1 2023-01-14 GDB Administrator <gdbadmin@sourceware.org>
3 Automatic date update in version.in
5 2023-01-13 Nick Clifton <nickc@redhat.com>
7 Updated Romainian translation for the bfd sub-directory
9 2023-01-13 GDB Administrator <gdbadmin@sourceware.org>
11 Automatic date update in version.in
13 2023-01-12 Hans-Peter Nilsson <hp@axis.com>
15 ARM: Fix ld bloat introduced between binutils-2.38 and 2.39
16 Since commit 9833b7757d24, "PR28824, relro security issues",
17 ELF_MAXPAGESIZE matters much more, with regards to layout of
18 the linked file. That commit fixed an actual bug, but also
19 exposes a problem for targets were that value is too high.
21 For example, for ARM(32, a.k.a. "Aarch32") specifically
22 bfd_arch_arm, it's set to 64 KiB, making all Linux(/GNU)
23 targets pay an extra amount of up to 60 KiB of bloat in
24 DSO:s and executables. This matters when there are many
25 such files, and where storage is expensive.
27 It's *mostly* bloat when using a Linux kernel, as ARM(32) is
28 a good example of an target where ELF_MAXPAGESIZE is set to
29 an extreme value for an obscure corner-case. The ARM
30 (32-bit) kernel has 4 KiB pages, has had that value forever,
31 and can't be configured to any other value. The use-case is
32 IIUC "Aarch32" emulation on an "Aarch64" (arm64) kernel, but
33 not just that, but a setup where the Linux page-size is
34 configured to something other than the *default* 4 KiB. Not
35 sure there actually any such systems in use, again with
36 both Aarch32 compatibility support and a non-4KiB pagesize,
37 with all the warnings in the kernel config and requiring the
38 "EXPERT" level set on.
40 So, let's do like x86-64 in a2267dbfc9e1 "x86-64: Use only
41 one default max-page-size" and set ELF_MAXPAGESIZE to 4096.
44 * elf32-arm.c (ELF_MAXPAGESIZE): Always set to 0x1000.
45 (cherry picked from commit 1a26a53a0dee39106ba58fcb15496c5f13074652)
47 2023-01-12 Hans-Peter Nilsson <hp@axis.com>
49 ld/testsuite: Adjust for ELF_MAXPAGESIZE 0x1000
50 Many tests reflect a setting of ELF_MAXPAGESIZE to 64 KiB.
51 With ELF_MAXPAGESIZE changed to 4 KiB, layout is sometimes
52 different and symbols end up in other places. Avoid churn
53 and regexpification of old test patterns by passing the
54 max-page-size setting active at the time.
58 * testsuite/ld-arm/arm-elf.exp,
59 testsuite/ld-arm/non-contiguous-arm2.d,
60 testsuite/ld-arm/non-contiguous-arm3.d,
61 testsuite/ld-arm/non-contiguous-arm5.d,
62 testsuite/ld-arm/non-contiguous-arm6.d,
63 testsuite/ld-arm/thumb-plt-got.d, testsuite/ld-arm/thumb-plt.d:
64 Pass -z max-page-size=0x10000 explicitly to test that rely on
65 that value in output-matching patterns.
66 (cherry picked from commit b305015577bb92d3041e55a72ca8cd43f7c05748)
68 2023-01-12 Nick Alcock <nick.alcock@oracle.com>
70 libctf: ctf-link outdated input check faulty
71 This check has a pair of faults which, combined, can lead to memory
72 corruption. Firstly, it assumes that the values of the ctf_link_inputs
73 hash are ctf_dict_t's: they are not, they are ctf_link_input_t's, a much
74 shorter structure. So the flags check which is the core of this is
75 faulty (but happens, by chance, to give the right output on most
76 architectures, since usually we happen to get a 0 here, so the test that
77 checks this usually passes). Worse, the warning that is emitted when
78 the test fails is added to the wrong dict -- it's added to the input
79 dict, whose warning list is never consumed, rendering the whole check
80 useless. But the dict it adds to is still the wrong type, so we end up
81 overwriting something deep in memory (or, much more likely,
82 dereferencing a garbage pointer and crashing).
84 Fixing both reveals another problem: the link input is an *archive*
85 consisting of multiple members, so we have to consider whether to check
86 all of them for the outdated-func-info thing we are checking here.
87 However, no compiler exists that emits a mixture of members with this
88 flag on and members with it off, and the linker always reserializes (and
89 upgrades) such things when it sees them: so all members in a given
90 archive must have the same value of the flag, so we only need to check
91 one member per input archive.
95 * ctf-link.c (ctf_link_warn_outdated_inputs): Get the types of
96 members of ctf_link_inputs right, fixing a possible spurious
97 tesst failure / wild pointer deref / overwrite. Emit the
98 warning message into the right dict.
100 2023-01-12 Nick Clifton <nickc@redhat.com>
102 Ensure that libbacktrace/allocfail.sh is not deleted when creating release tarballs.
103 * Makefile.am (CLEANFILES): Import patch from upstream to prevent
104 allocafail.sh from being removed when running 'make clean'.
106 2023-01-12 GDB Administrator <gdbadmin@sourceware.org>
108 Automatic date update in version.in
110 2023-01-11 Nick Clifton <nickc@redhat.com>
112 Fix a potential illegal memory access in the BFD library when parsing a corrupt DWARF file.
114 * dwarf2.c (read_indexed_address): Fix check for an out of range
117 2023-01-11 Jan Beulich <jbeulich@suse.com>
119 gas/RISC-V: adjust assembler for opcode table re-ordering
122 With the single-operand JAL entry now sitting ahead of the two-operand
123 one, the parsing of a two-operand insn would first try to parse an 'a'-
124 style operand, resulting in the insertion of bogus (and otherwise
125 unused) undefined symbols in the symbol table, having register names.
126 Since 'a' is used as 1st operand only with J and JAL, and since JAL is
127 the only insn _also_ allowing for a register as 1st operand (and then
128 there being a 2nd one), special case this parsing aspect right there.
130 Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
132 2023-01-11 GDB Administrator <gdbadmin@sourceware.org>
134 Automatic date update in version.in
136 2023-01-10 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
138 IBM zSystems: Fix offset relative to static TLS
139 For local exec TLS relocations of the form foo@NTPOFF+x the addend was
144 * elf32-s390.c (elf_s390_relocate_section): Honor addend for
146 * elf64-s390.c (elf_s390_relocate_section): Honor addend for
151 * testsuite/ld-s390/reloctlsle-1.d: New test.
152 * testsuite/ld-s390/reloctlsle-1.s: New test.
154 (cherry picked from commit aefebe82dc89711384b85329daa48d04c1d3a45b)
156 2023-01-10 GDB Administrator <gdbadmin@sourceware.org>
158 Automatic date update in version.in
160 2023-01-09 Indu Bhagat <indu.bhagat@oracle.com>
162 sframe: fix the defined SFRAME_FRE_TYPE_*_LIMIT constants
163 An earlier commit 3f107464 defined the SFRAME_FRE_TYPE_*_LIMIT
164 constants. These constants are used (by gas and libsframe) to pick an
165 SFrame FRE type based on the function size. Those constants, however,
166 were buggy, causing the generated SFrame sections to be bloated as
167 SFRAME_FRE_TYPE_ADDR2/SFRAME_FRE_TYPE_ADDR4 got chosen more often than
171 * sframe-opt.c (sframe_estimate_size_before_relax): Use
173 (sframe_convert_frag): Likewise.
176 * sframe.c (sframe_calc_fre_type): Use a more appropriate type
177 for argument. Adjust the check for SFRAME_FRE_TYPE_ADDR4_LIMIT
178 to keep it warning-free but meaningful.
181 * sframe-api.h (sframe_calc_fre_type): Use a more appropriate
182 type for the argument.
183 * sframe.h (SFRAME_FRE_TYPE_ADDR1_LIMIT): Correct the constant.
184 (SFRAME_FRE_TYPE_ADDR2_LIMIT): Likewise.
185 (SFRAME_FRE_TYPE_ADDR4_LIMIT): Likewise.
187 (cherry picked from commit 725a19bfd142c845bf76ae28f6289972fd1cf5db)
189 2023-01-09 Indu Bhagat <indu.bhagat@oracle.com>
191 libsframe: adjust an incorrect check in flip_sframe
192 When sframe_encoder_write needs to flip the buffer containing the SFrame
193 section before writing, it is not necessary that the SFrame FDES are in
194 the order of their sfde_func_start_fre_off. On the contrary, SFrame
195 FDEs will be sorted in the order of their start address. So, remove
196 this incorrect assumption which is basically assuming that the last
197 sfde_func_start_fre_off seen will help determine the end of the flipped
200 The function now keeps track of the bytes_flipped and then compares it with
201 the expected value. Also, added two more checks at appropriate places:
202 - check that the SFrame FDE read is within bounds
203 - check that the SFrame FRE read is within bounds
207 * sframe.c (flip_sframe): Adjust an incorrect check.
208 Add other checks to ensure reads are within the buffer size.
210 (cherry picked from commit cd9aea32cffd8089f6f63f4eb86d4dccfc0b3850)
212 2023-01-09 Christophe Lyon <christophe.lyon@arm.com>
214 Skip ld/pr23169 test on arm.
215 The test is already skipped on several targets (including AArch64)
216 because it's invalid.
218 * testsuite/ld-ifunc/ifunc.exp: Skip pr23169 on arm.
220 2023-01-09 Christophe Lyon <christophe.lyon@arm.com>
222 Fix PR18841 ifunc relocation ordering
223 In order to get the ifunc relocs properly sorted the correct class
224 needs to be returned. The code mimics what has been done for AArch64.
227 FAIL: Run pr18841 with libpr18841b.so
228 FAIL: Run pr18841 with libpr18841c.so
229 FAIL: Run pr18841 with libpr18841bn.so (-z now)
230 FAIL: Run pr18841 with libpr18841cn.so (-z now)
234 * elf32-arm.c (elf32_arm_reloc_type_class): Return
235 reloc_class_ifunc for ifunc symbols.
238 * ld-arm/ifunc-12.rd: Update relocations order.
239 * ld-arm/ifunc-3.rd: Likewise.
240 * ld-arm/ifunc-4.rd: Likewise.
242 2023-01-09 Nick Clifton <nickc@redhat.com>
244 Updated transaltions for the gprof and binutils sub-directories
246 2023-01-09 GDB Administrator <gdbadmin@sourceware.org>
248 Automatic date update in version.in
250 2023-01-08 Alan Modra <amodra@gmail.com>
252 PR29972, inconsistent format specification in singular form
254 * readelf.c (process_dynamic_section): Correct format string.
256 (cherry picked from commit 02da71ee20ec71f7b3be85cf2266e09c124983bf)
258 2023-01-08 GDB Administrator <gdbadmin@sourceware.org>
260 Automatic date update in version.in
262 2023-01-07 GDB Administrator <gdbadmin@sourceware.org>
264 Automatic date update in version.in
266 2023-01-06 Jan Beulich <jbeulich@suse.com>
268 ld: yet another PDB build fix (or workaround)
269 Older bash looks to improperly deal with backslashes in here-documents,
270 leaving them in place on the escaped double quotes inside the parameter
271 expansion. Convert to a model without using such a construct, by simply
272 splitting the here-documents into three ones.
274 2023-01-06 Nick Clifton <nickc@redhat.com>
276 Updated Bulgarian and Russian translations for LD and BFD respectively
278 2023-01-06 GDB Administrator <gdbadmin@sourceware.org>
280 Automatic date update in version.in
282 2023-01-05 Nick Clifton <nickc@redhat.com>
284 Avoid unaligned pointer reads in PEP idata section
286 Updated Bulgarian and Russian translations for the gprof subdirectory
288 2023-01-05 GDB Administrator <gdbadmin@sourceware.org>
290 Automatic date update in version.in
292 2023-01-04 Alan Modra <amodra@gmail.com>
294 Merge config/picflag.m4 from gcc
296 Update some more copyright year ranges
297 These files disappear in commit 3002e78a7d3d but are still on the branch.
299 Update year range in gprofng copyright notices
300 This adds 'Innovative Computing Labs' as an external author to
301 update-copyright.py, to cover the copyright notice in
302 gprofng/common/opteron_pcbe.c, and uses that plus another external
303 author 'Oracle and' to update gprofng copyright dates. I'm not going
304 to commit 'Oracle and' as an accepted author, but that covers the
305 string "Copyright (c) 2006, 2012, Oracle and/or its affiliates. All
306 rights reserved." found in gprofng/testsuite/gprofng.display/jsynprog
309 Update year range in copyright notice of binutils files
310 The newer update-copyright.py fixes file encoding too, removing cr/lf
311 on binutils/bfdtest2.c and ld/testsuite/ld-cygwin/exe-export.exp, and
312 embedded cr in binutils/testsuite/binutils-all/ar.exp string match.
314 2023-01-04 Alan Modra <amodra@gmail.com>
316 Update etc/update-copyright.py
317 This picks up some improvements from gcc/contrib. exceptions must
318 derive from BaseException, port to python3, retain original file mode,
319 fix name of script in examples.
321 Adds libsframe to list of default dirs. I would have added gprofng
322 too but there are some files claiming copyright by authors other than
323 the Free Software Foundation.
325 2023-01-04 Andreas K. Huettel <dilfridge@gentoo.org>
327 Fix AArch64 linker testsuite failures triggered by differences in build environments.
329 * testsuite/ld-aarch64/bti-plt-5.d: Relax regxps slightly to allow
330 for differences in build environments.
331 * testsuite/ld-aarch64/tls-relax-gdesc-le-now.d: Likewise.
333 2023-01-04 Mark Harmstone <mark@harmstone.com>
335 Avoid unaligned pointer reads in PEP .idata section
336 This is something I discovered when working on aarch64, though it's
337 relevant to x86_64 too.
339 The PE32+ imports are located in the .idata section, which starts off
340 with a 20-byte structure for each DLL, containing offsets into the rest
341 of the section. This is the Import Directory Table in
342 https://learn.microsoft.com/en-us/windows/win32/debug/pe-format, which
343 is a concatenation of the .idata$2 sections. This is then followed by an
344 20 zero bytes generated by the linker script, which calls this .idata$3.
346 After this comes the .idata$4 entries for each function, which the
347 loader overwrites with the function pointers. Because there's no padding
348 between .idata$3 and .idata$4, this means that if there's an even number
349 of DLLs, the function pointers won't be aligned on an 8-byte boundary.
351 Misaligned reads are slower on x86_64, but this is more important on
352 aarch64, as the e.g. `ldr x0, [x0, :lo12:__imp__func]` the compiler
353 might generate requires __imp__func (the .idata$4 entry) to be aligned
354 to 8 bytes. Without this you get IMAGE_REL_ARM64_PAGEOFFSET_12L overflow
357 2023-01-04 GDB Administrator <gdbadmin@sourceware.org>
359 Automatic date update in version.in
361 2023-01-03 Nick Clifton <nickc@redhat.com>
363 Updated translations for various languages and sub-directories
365 2023-01-03 GDB Administrator <gdbadmin@sourceware.org>
367 Automatic date update in version.in
369 2023-01-02 GDB Administrator <gdbadmin@sourceware.org>
371 Automatic date update in version.in
373 2022-12-31 Nick Clifton <nickc@redhat.com>
375 Update version number and regenerate files
377 Add markers for 2.40 branch
379 sync libiberty sources with gcc mainline
381 2022-12-31 Tom de Vries <tdevries@suse.de>
383 [gdb/cli] Add maintenance ignore-probes
384 There's a command "disable probes", but SystemTap probes, for instance
385 libc:longjmp cannot be disabled:
387 $ gdb -q -batch a.out -ex start -ex "disable probes libc ^longjmp$"
389 Probe libc:longjmp cannot be disabled.
390 Probe libc:longjmp cannot be disabled.
391 Probe libc:longjmp cannot be disabled.
394 Add a command "maintenance ignore-probes" that ignores probes during
395 get_probes, such that we can easily pretend to use a libc without the
398 (gdb) maint ignore-probes -verbose libc ^longjmp$
399 ignore-probes filter has been set to:
401 PROBE_NAME: '^longjmp$'
405 Ignoring SystemTap probe libc longjmp in /lib64/libc.so.6.^M
406 Ignoring SystemTap probe libc longjmp in /lib64/libc.so.6.^M
407 Ignoring SystemTap probe libc longjmp in /lib64/libc.so.6.^M
410 The "Ignoring ..." messages can be suppressed by not using -verbose.
412 Note that as with "disable probes", running simply "maint ignore-probes"
415 The ignore-probes filter can be reset by using:
417 (gdb) maint ignore-probes -reset
418 ignore-probes filter has been reset
421 For now, the command is only supported for SystemTap probes.
424 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=27159
426 2022-12-31 Mark Harmstone <mark@harmstone.com>
428 ld/testsuite: Don't add index to sizes in pdb.exp
430 ld: Handle LF_VFTABLE types in PDBs
432 2022-12-31 Mark Harmstone <mark@harmstone.com>
434 ld: Handle extended-length data structures in PDB types
435 A few fixes to minor issues I've discovered in my PDB patches.
437 * If sizes or offsets are greater than 0x8000, they get encoded as
438 extended values in the same way as for enum values - e.g. a LF_ULONG
439 .short followed by a .long.
441 * I've managed to coax MSVC to produce another type, LF_VFTABLE, which
442 is seen when dealing with COM. I don't think LLVM emits this. Note that
443 we can't just implement everything in Microsoft's header files, as most
446 * Fixes a stupid bug in the test program, where I was adding an index to
447 a size. The index was hard-coded to 0, so this didn't cause any actual
450 2022-12-31 Nick Clifton <nickc@redhat.com>
452 Updated Romanian translation for the binutils sub-directory
454 2022-12-31 Tom de Vries <tdevries@suse.de>
456 [gdb/python] Fix gdb.python/py-finish-breakpoint2.exp for -m32
457 [ Partial resubmission of an earlier submission by Andrew (
458 https://sourceware.org/pipermail/gdb-patches/2012-September/096347.html ), so
459 listing him as co-author. ]
461 With x86_64-linux and target board unix/-m32, we have:
467 Breakpoint 3, throw_exception_1 (e=10) at py-finish-breakpoint2.cc:23^M
468 23 throw new int (e);^M
469 (gdb) FAIL: gdb.python/py-finish-breakpoint2.exp: \
470 check FinishBreakpoint in catch()
473 The following scenario happens:
474 - set breakpoint in throw_exception_1, a function that throws an exception
476 - hit breakpoint, with call stack main.c:38 -> throw_exception_1
477 - set a finish breakpoint
479 - hit the breakpoint again, with call stack main.c:48 -> throw_exception
482 Due to the exception, the function call did not properly terminate, and the
483 finish breakpoint didn't trigger. This is expected behaviour.
485 However, the intention is that gdb detects this situation at the next stop
486 and calls the out_of_scope callback, which would result here in this test-case
487 in a rather confusing "exception did not finish" message. So the problem is
488 that this message doesn't show up, in other words, the out_of_scope callback
491 [ Note that the fact that the situation is detected only at the next stop
492 (wherever that happens to be) could be improved upon, and the earlier
493 submission did that by setting a longjmp breakpoint. But I'm considering this
494 problem out-of-scope for this patch. ]
496 Note that the message does show up later, at thread exit:
498 [Inferior 1 (process 20046) exited with code 0236]^M
499 exception did not finish ...^M
502 The decision on whether to call the out_of_scope call back is taken in
503 bpfinishpy_detect_out_scope_cb, and the interesting bit is here:
505 if (b->pspace == current_inferior ()->pspace
506 && (!target_has_registers ()
507 || frame_find_by_id (b->frame_id) == NULL))
508 bpfinishpy_out_of_scope (finish_bp);
511 In the case of the thread exit, the callback triggers because
512 target_has_registers () == 0.
514 So why doesn't the callback trigger in the case of the breakpoint?
516 Well, the b->frame_id is the frame_id of the frame of main (the frame
517 in which the finish breakpoint is supposed to trigger), so AFAIU
518 frame_find_by_id (b->frame_id) == NULL will only be true once we've
519 left main, at which point I guess we don't stop till thread exit.
521 Fix this by saving the frame in which the finish breakpoint was created, and
522 using frame_find_by_id () == NULL on that frame instead, such that we have:
528 Breakpoint 3, throw_exception_1 (e=10) at py-finish-breakpoint2.cc:23^M
529 23 throw new int (e);^M
530 exception did not finish ...^M
531 (gdb) FAIL: gdb.python/py-finish-breakpoint2.exp: \
532 check FinishBreakpoint in catch()
535 Still, the test-case is failing because it's setup to match the behaviour that
536 we get on x86_64-linux with target board unix/-m64:
541 stopped at ExceptionFinishBreakpoint^M
542 (gdb) PASS: gdb.python/py-finish-breakpoint2.exp: \
543 check FinishBreakpoint in catch()
546 So what happens here? Again, due to the exception, the function call did not
547 properly terminate, but the finish breakpoint still triggers. This is somewhat
548 unexpected. This happens because it just so happens to be that the frame
549 return address at which the breakpoint is set, is also the first instruction
550 after the exception has been handled. This is a know problem, filed as
551 PR29909, so KFAIL it, and modify the test-case to expect the out_of_scope
554 Also add a breakpoint after setting the finish breakpoint but before throwing
555 the exception, to check that we don't call the out_of_scope callback too early.
557 Tested on x86_64-linux, with target boards unix/-m32.
559 Co-Authored-By: Andrew Burgess <aburgess@redhat.com>
561 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=27247
563 2022-12-31 Tom de Vries <tdevries@suse.de>
565 [gdb/testsuite] Fix gdb.base/print-symbol-loading.exp on ubuntu 22.04.1
566 On ubuntu 22.04.1 x86_64, I run into:
568 (gdb) PASS: gdb.base/print-symbol-loading.exp: shlib off: \
569 set print symbol-loading off
571 Symbols already loaded for /lib/x86_64-linux-gnu/libc.so.6^M
572 Symbols already loaded for /lib/x86_64-linux-gnu/libpthread.so.0^M
573 (gdb) FAIL: gdb.base/print-symbol-loading.exp: shlib off: load shared-lib
576 The test-case expects the libc.so line, but not the libpthread.so line.
580 $ ldd /lib/x86_64-linux-gnu/libc.so.6
581 linux-vdso.so.1 (0x00007ffd7f7e7000)
582 libgtk3-nocsd.so.0 => /lib/x86_64-linux-gnu/libgtk3-nocsd.so.0 (0x00007f4468c00000)
583 /lib64/ld-linux-x86-64.so.2 (0x00007f4469193000)
584 libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007f4468f3e000)
585 libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007f4468f39000)
587 so it's not unexpected that libpthread.so is loaded if libc.so is loaded.
589 Fix this by accepting the libpthread.so line.
591 Tested on x86_64-linux.
594 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29919
596 2022-12-31 Tom de Vries <tdevries@suse.de>
598 [gdb/testsuite] Replace deprecated pthread_yield in gdb.threads/watchpoint-fork.exp
599 On Ubuntu 22.04.1 x86_64, with glibc 2.35 I run into:
601 watchpoint-fork-mt.c: In function 'start':^M
602 watchpoint-fork-mt.c:67:7: warning: 'pthread_yield' is deprecated: \
603 pthread_yield is deprecated, use sched_yield instead \
604 [-Wdeprecated-declarations]^M
605 67 | i = pthread_yield ();^M
609 Fix this as suggested, by using sched_yield instead.
611 Tested on x86_64-linux.
613 2022-12-31 Tom de Vries <tdevries@suse.de>
615 [gdb/testsuite] Fix gdb.base/corefile.exp with glibc 2.35
616 On Ubuntu 22.04.1 x86_64 (with glibc 2.35), I run into:
618 (gdb) PASS: gdb.base/corefile.exp: $_exitcode is void
620 #0 __pthread_kill_implementation (...) at ./nptl/pthread_kill.c:44^M
621 #1 __pthread_kill_internal (...) at ./nptl/pthread_kill.c:78^M
622 #2 __GI___pthread_kill (...) at ./nptl/pthread_kill.c:89^M
623 #3 0x00007f4985e1a476 in __GI_raise (...) at ../sysdeps/posix/raise.c:26^M
624 #4 0x00007f4985e007f3 in __GI_abort () at ./stdlib/abort.c:79^M
625 #5 0x0000556b4ea4b504 in func2 () at gdb.base/coremaker.c:153^M
626 #6 0x0000556b4ea4b516 in func1 () at gdb.base/coremaker.c:159^M
627 #7 0x0000556b4ea4b578 in main (...) at gdb.base/coremaker.c:171^M
628 (gdb) PASS: gdb.base/corefile.exp: backtrace
630 #1 __pthread_kill_internal (...) at ./nptl/pthread_kill.c:78^M
631 78 in ./nptl/pthread_kill.c^M
632 (gdb) FAIL: gdb.base/corefile.exp: up
635 The problem is that the regexp used here:
637 gdb_test "up" "#\[0-9\]* *\[0-9xa-fH'\]* in .* \\(.*\\).*" "up"
639 does not fit the __pthread_kill_internal line which lacks the instruction
640 address due to inlining.
642 Fix this by making the regexp less strict.
644 Tested on x86_64-linux.
646 2022-12-31 GDB Administrator <gdbadmin@sourceware.org>
648 Automatic date update in version.in
650 2022-12-30 Tom de Vries <tdevries@suse.de>
652 [gdb/testsuite] Fix gdb.threads/dlopen-libpthread.exp for upstream glibc
653 On ubuntu 22.04.1 x86_64, I run into:
655 (gdb) info probes all rtld rtld_map_complete^M
657 (gdb) XFAIL: gdb.threads/dlopen-libpthread.exp: info probes all rtld rtld_map_complete
658 UNTESTED: gdb.threads/dlopen-libpthread.exp: no matching probes
660 This has been filed as PR testsuite/17016.
662 The problem is that the name rtld_map_complete is used, which was only
663 available in Fedora 17, and upstream the name map_complete was used.
665 In the email thread discussing a proposed patch (
666 https://sourceware.org/legacy-ml/gdb-patches/2014-09/msg00712.html ) it was
667 suggested to make the test-case handle both names.
669 So, handle both names: map_complete and rtld_map_complete.
671 This exposes the following FAIL:
673 (gdb) info sharedlibrary^M
674 From To Syms Read Shared Object Library^M
675 $hex $hex Yes /lib64/ld-linux-x86-64.so.2^M
676 $hex $hex Yes (*) /lib/x86_64-linux-gnu/libgtk3-nocsd.so.0^M
677 $hex $hex Yes /lib/x86_64-linux-gnu/libc.so.6^M
678 $hex $hex Yes /lib/x86_64-linux-gnu/libdl.so.2^M
679 $hex $hex Yes /lib/x86_64-linux-gnu/libpthread.so.0^M
680 (*): Shared library is missing debugging information.^M
681 (gdb) FAIL: gdb.threads/dlopen-libpthread.exp: libpthread.so not found
683 due to using a glibc (v2.35) that has libpthread integrated into libc.
685 Fix this by changing the FAIL into UNSUPPORTED.
687 Tested on x86_64-linux.
689 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=17016
691 2022-12-30 Tom de Vries <tdevries@suse.de>
693 [gdb/testsuite] Fix gdb.reverse/step-indirect-call-thunk.exp with -fcf-protection
694 On Ubuntu 22.04.1 x86_64, I run into:
696 gdb.reverse/step-indirect-call-thunk.c: In function 'inc':^M
697 gdb.reverse/step-indirect-call-thunk.c:22:1: error: '-mindirect-branch' and \
698 '-fcf-protection' are not compatible^M
703 Fix this by forcing -fcf-protection=none, if supported.
705 Tested on x86_64-linux.
707 2022-12-30 Tom de Vries <tdevries@suse.de>
709 [gdb/testsuite] Fix gdb.cp/step-and-next-inline.exp with -fcf-protection
710 On Ubuntu 22.04.1 x86_64, I run into:
712 (gdb) PASS: gdb.cp/step-and-next-inline.exp: no_header: not in inline 1
715 (gdb) FAIL: gdb.cp/step-and-next-inline.exp: no_header: next step 1
718 This is due to -fcf-protection, which adds the endbr64 at the start of get_alias_set:
720 0000000000001180 <_Z13get_alias_setP4tree>:
721 1180: f3 0f 1e fa endbr64
722 1184: 48 85 ff test %rdi,%rdi
724 so the extra insn gets an is-stmt line number entry:
726 INDEX LINE ADDRESS IS-STMT PROLOGUE-END
728 11 50 0x0000000000001180 Y
729 12 50 0x0000000000001180
730 13 51 0x0000000000001184 Y
731 14 54 0x0000000000001184
733 and when stepping into get_alias_set we step to line 50:
735 (gdb) PASS: gdb.cp/step-and-next-inline.exp: no_header: in main
737 get_alias_set (t=t@entry=0x555555558018 <xx>) at step-and-next-inline.cc:50^M
741 In contrast, with -fcf-protection=none, we get:
743 0000000000001170 <_Z13get_alias_setP4tree>:
744 1170: 48 85 ff test %rdi,%rdi
748 INDEX LINE ADDRESS IS-STMT PROLOGUE-END
750 11 50 0x0000000000001170 Y
751 12 51 0x0000000000001170 Y
752 13 54 0x0000000000001170
754 so when stepping into get_alias_set we step to line 51:
756 (gdb) PASS: gdb.cp/step-and-next-inline.exp: no_header: in main
758 get_alias_set (t=t@entry=0x555555558018 <xx>) at step-and-next-inline.cc:51^M
762 Fix this by rewriting the gdb_test issuing the step command to check which
763 line the step lands on, and issuing an extra next if needed.
765 Tested on x86_64-linux, both with and without -fcf-protection=none.
768 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29920
770 2022-12-30 Tom de Vries <tdevries@suse.de>
772 [gdb/symtab] Make comp_unit_head.length private
773 Make comp_unit_head.length private, to enforce using accessor functions.
775 Replace accessor function get_length with get_length_with_initial and
776 get_length_without_initial, to make it explicit which variant we're using.
778 Tested on x86_64-linux.
781 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29343
783 2022-12-30 Alan Modra <amodra@gmail.com>
785 PR29948, heap-buffer-overflow in display_debug_lines_decoded
786 This fixes a couple of places in display_debug_lines_decoded that were
787 off by one in checking DWARF5 .debug_line directory indices. It also
788 displays the DWARF5 entry 0 for the program current directory rather
789 than "." as is done for pre-DWARF5. I decided against displaying
790 DW_AT_comp_dir for pre-DWARF5 since I figure it is better for readelf
791 to minimally interpret debug info.
795 * dwarf.c (display_debug_lines_decoded): Display the given
796 directory entry 0 for DWARF5. Properly check directory index
797 against number of entries in the table. Revert to using
798 unsigned int for n_directories and associated variables.
799 Correct warning messages.
801 * testsuite/gas/elf/dwarf-5-loc0.d: Update.
803 2022-12-30 GDB Administrator <gdbadmin@sourceware.org>
805 Automatic date update in version.in
807 2022-12-29 Tsukasa OI <research_trasio@irq.a4lg.com>
809 RISC-V: Simplify riscv_csr_address logic on state enable extensions
810 This commit makes CSR class handling for 'Smstateen' and 'Ssstateen'
811 extensions simpler using fall-throughs (as used in CSR_CLASS_I{,_32}).
815 * config/tc-riscv.c (riscv_csr_address): Simplify the logic for
816 'Smstateen' and 'Ssstateen' extensions.
818 2022-12-29 GDB Administrator <gdbadmin@sourceware.org>
820 Automatic date update in version.in
822 2022-12-28 Tom Tromey <tom@tromey.com>
824 Use $decimal in timestamp.exp
825 This patch fixes a review comment by Tom de Vries. He pointed out
826 that the new timestamp.exp should use the $decimal convenience regexp.
828 2022-12-28 Tom Tromey <tom@tromey.com>
830 Fix "set debug timestamp"
831 PR cli/29945 points out that "set debug timestamp 1" stopped working
832 -- this is a regression due to commit b8043d27 ("Remove a ui-related
835 This patch fixes the bug and adds a regression test.
837 I think this should probably be backported to the gdb 13 branch.
839 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29945
841 2022-12-28 GDB Administrator <gdbadmin@sourceware.org>
843 Automatic date update in version.in
845 2022-12-27 H.J. Lu <hjl.tools@gmail.com>
847 x86-64: Allocate input section memory if needed
848 When --no-keep-memory is used, the input section memory may not be cached.
849 Allocate input section memory for -z pack-relative-relocs if needed.
854 * elfxx-x86.c (elf_x86_size_or_finish_relative_reloc): Allocate
855 input section memory if needed.
860 * testsuite/ld-elf/dt-relr-2i.d: New test.
862 2022-12-27 Christoph Müllner <christoph.muellner@vrull.eu>
864 RISC-V: Fix T-Head Fmv vendor extension encoding
865 A recent change in the XTheadFmv spec fixed an encoding bug in the
866 document. This patch changes the code to follow this bugfix.
868 Spec patch can be found here:
869 https://github.com/T-head-Semi/thead-extension-spec/pull/11
871 2022-12-27 Tom Tromey <tromey@adacore.com>
873 Handle SIGSEGV in gdb selftests
874 The gdb.gdb self-tests were timing out for me, which turned out to be
875 PR testsuite/29325. Looking into it, the problem is that the version
876 of the Boehm GC that is used by Guile on my machine causes a SEGV
877 during stack probing. This unexpected stop confuses the tests and
878 causes repeated timeouts.
880 This patch adapts the two failing tests. This makes them work for me,
881 and reduces the running time of gdb.gdb from 20 minutes to about 11
884 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29325
886 2022-12-27 Mike Frysinger <vapier@gentoo.org>
888 sim: build: clean up unused codegen logic
889 Now that all igen ports are in the top-level makefile, we don't need
890 this logic in any subdirs anymore, so clean it up.
892 sim: mips: hoist "multi" igen rules up to common builds
893 Since these are the last mips igen rules, we can clean up a number of
894 bits in the local Makefile.in.
896 sim: mips: hoist "m16" igen rules up to common builds
898 sim: mips: hoist "single" igen rules up to common builds
900 sim: mips: rename "igen" generation mode to "single"
901 The naming in here has grown organically and is confusing to follow.
902 Originally there was only one set of rules for generating code from
903 the igen sources, so calling it "tmp-igen" and such made sense. But
904 when other multigen modes were added ("m16" & "multi") which also
905 used igen, it's not clear what's common igen and what's specific to
906 this generation mode. So rename the set of rules from "igen" to
907 "single" so it's easier to follow.
909 sim: mips: hoist itable igen rules up to common builds
910 Since this rule is pretty simple, hoist it up to the common build.
912 2022-12-27 Mike Frysinger <vapier@gentoo.org>
914 sim: mips: unify itable generation (a bit)
915 The m16 & multi targets generate itable once even when all the other
916 modules are generated multiple times. The default igen target will
917 generate itable with everything else out of convenience. This means
918 flags are passed which don't affect the generated itable there.
920 We can unify the itable generation by making sure the right -F/-M
921 filter variables are passed down. Since there's already a dedicated
922 rule & variable in the multi build mode, generalize that and switch
923 the m16 & igen builds over too.
925 I spent a lot of time staring at this code, building for diff mips
926 targets, and exploring all the shell code paths. I think this is
927 safe, but only time (and users) will really tell.
929 2022-12-27 Mike Frysinger <vapier@gentoo.org>
931 sim: mips: rename multi_flags to igen_itable_flags
932 This variable is only used to generate the itable files. In preparation
933 for merging the itable logic among all ports, rename "multi_flags" to a
934 more appropriate "igen_itable_flags" variable. There should be no real
935 chagnes here otherwise.
937 sim: mips: drop unused micromips igen logic
938 This code appears to be unused since it was first merged. When
939 micromips was enabled, it was via the "MULTI" config, not the
940 "MICROMIPS" config, and the multi configs have sep vars. Since
941 nothing sets SIM_MIPS_GEN=MICROMIPS in the config, all of this
942 should be unreachable, so punt it to simplify. Further, the
943 SIM_MIPS_MICROMIPS16_FLAGS & SIM_MIPS_MICROMIPS_FLAGS settings
944 rely on sim_mips_micromips{,16}_{filter,machine} variables that
945 are never set in the configure script.
947 2022-12-27 GDB Administrator <gdbadmin@sourceware.org>
949 Automatic date update in version.in
951 2022-12-26 Tom Tromey <tom@tromey.com>
953 Add initializers to comp_unit_head
954 PR symtab/29343 points out that it would be beneficial if
955 comp_unit_head had a constructor and used initializers. This patch
956 implements this. I'm unsure if this is sufficient to close the bug,
957 but at least it's a step.
959 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29343
961 2022-12-26 Alan Modra <amodra@gmail.com>
963 bfd/dwarf2.c: allow use of DWARF5 directory entry 0
964 I think the test for table->files[file].dir being non-zero is wrong
965 for DWARF5 where index zero is allowed and is the current directory of
966 the compilation. Most times this will be covered by the use of
967 table->comp_dir (from DW_AT_comp_dir) in concat_filename but the point
968 of putting the current dir in .debug_line was so the section could
969 stand alone without .debug_info.
971 Also, there is no need to check for table->dirs non-NULL, the
972 table->num_dirs test is sufficient.
974 * dwarf2.c (concat_filename): Correct and simplify tests of
977 2022-12-26 Flavio Cruz <flaviocruz@gmail.com>
979 Add support for x86_64-*-gnu-* targets to build x86_64 gnumach/hurd
981 2022-12-26 GDB Administrator <gdbadmin@sourceware.org>
983 Automatic date update in version.in
985 2022-12-25 Mike Frysinger <vapier@gentoo.org>
987 sim: build: drop support for subdir distclean
988 All ports that need to clean things up at distclean time have moved
989 to the top-level build, so we can drop support for this hook.
991 sim: mips: move distclean settings to common build
992 This was missed when mips/configure was merged into the top-level.
994 2022-12-25 Indu Bhagat <indu.bhagat@oracle.com>
996 libsframe: write out SFrame FRE start address correctly
997 The following test was failing on ppc64 and s390x:
998 "FAIL: encode-1: Encode buffer match"
1000 The offending stub was how we memcpy the FRE start address to the buffer
1001 (on-disk format). When the host is big-endian, the address of the
1002 source buffer for the memcpy needs to point to the uint8_t/uint16_t sized
1003 value of the FRE start addr, not uint32_t sized value; we intend to copy
1004 out only the fre_start_addr_sz number of bytes.
1008 * libsframe/sframe.c (sframe_encoder_write_fre_start_addr): New
1010 (sframe_encoder_write_fre): Use it instead of memcpy.
1012 2022-12-25 Mike Frysinger <vapier@gentoo.org>
1014 sim: smp: plumb igen flag down to all users
1015 While mips has respected sim_igen_smp at configure time (which was
1016 always empty since it defaulted smp to off), no other igen port did.
1017 Move this to a makefile variable and plumb it through the common
1018 IGEN_RUN variable instead so everyone gets it by default. We also
1019 clean up some redundant -N0 setting with multirun mips.
1021 sim: smp: make option available again
1022 At some point we want this to work, but it's not easy to test if
1023 the configure option isn't available. Restore it, but keep the
1026 sim: cpu: change default init to handle all cpus
1027 All the runtimes were only initializing a single CPU. When SMP is
1028 enabled, things quickly crash as none of the other CPU structs are
1029 setup. Change the default from 0 to the compile time value.
1031 sim: msp430: add basic SMP cpu init
1032 There's no need to assert there's only 1 CPU when setting them all
1035 sim: m32r: fix iterator typo when setting up cpus
1036 This code loops over available cpus with "c", but then looks up the
1037 cpu with "i". Fix the typo so the code works correctly with smp.
1039 sim: v850: fix SMP compile
1040 The igen tool sets up the SD & CPU defines for code fragments to use,
1041 but v850 was expecting "sd". Change all the igen related code to use
1042 SD so it actually compiles, and fix a few places to use "CPU" instead
1045 sim: or1k: fix iterator typo when setting up cpus
1046 This code loops over available cpus with "c", but then looks up the
1047 cpu with "i". Fix the typo so the code works correctly with smp.
1049 sim: mn10300: fix SMP compile
1050 The igen tool sets up the SD define for code fragments to use, but
1051 mn10300 was expecting "sd". Change all the igen related code to use
1052 SD so it actually compiles.
1054 sim: cpu: fix SMP msg prefix helper
1055 This code fails to compile when SMP is enabled due to some obvious
1056 errors. Fix those and change the logic to avoid CPP to prevent any
1057 future rot from creeping back in.
1059 sim: mips: clean up a bit after mips/configure removal
1060 Now that there is no subdir configure script, we can clean up some
1061 logic that was spread between the files.
1063 sim: mips: move igen settings to top-level configure
1064 This is the last bit of logic that exists in the mips configure
1065 script, so move it to the top-level configure to kill it off.
1066 We still have to move the Makefile.in igen logic to local.mk,
1067 but this is a required first step for that.
1069 sim: mips: namespace igen configure vars
1070 To prepare moving this logic to the top-level configure, the vars
1071 need to be namespaced. Do that here to make it easier to review.
1072 Basically sim_xxx -> SIM_MIPS_XXX when a var is exported from the
1073 configure script to the Makefile, and sim_xxx -> sim_mips_xxx when
1074 the var is internal in the configure script.
1076 sim: mips: add igen recursive dep
1077 Make sure the igen tool exists before trying to compile the mips
1078 subdir. This happens to work when mips has a subconfigure, but
1079 hits a race condition when that is removed.
1081 sim: mips: drop unused ENGINE_ISSUE_POSTFIX_HOOK
1082 Nothing defines this, and it isn't called in all the engine runtimes,
1083 so drop it entirely to avoid confusion.
1085 sim: igen: drop move-if-changed usage
1086 Now that igen itself has this logic, drop these custom build rules
1087 to greatly simplify.
1089 sim: igen: support in-place updates ourself
1090 Every file that igen outputs is then processed with the move-if-changed
1091 shell script. This creates a lot of boilerplate in the build and not an
1092 insignificant amount of build-time overhead. Move the simple "is the file
1093 changed" logic into igen itself.
1095 sim: igen: constify itable data structures
1096 These are const data arrays of strings and numbers. We don't want
1097 or need them to be writable, so mark them all const.
1099 2022-12-25 GDB Administrator <gdbadmin@sourceware.org>
1101 Automatic date update in version.in
1103 2022-12-24 Andrew Burgess <aburgess@redhat.com>
1105 gdb/testsuite: fix buffer overflow in gdb.base/signed-builtin-types.exp
1108 commit 9f50fe0835850645bd8ea9bb1efe1fe6c48dfb12
1109 Date: Wed Dec 7 15:55:25 2022 +0000
1111 gdb/testsuite: new test for recent dwarf reader issue
1113 A new test (gdb.base/signed-builtin-types.exp) was added that made use
1114 of 'info sources' to figure out if the debug information for a
1115 particular object file had been fully expanded or not. Unfortunately
1116 some lines of the 'info sources' output can be very long, this was
1117 observed on some systems where the debug information for the
1118 dynamic-linker was installed, in this case, the list of source files
1119 associated with the dynamic linker was so long it would cause expect's
1120 internal buffer to overflow.
1122 This commit switches from using 'info sources' to 'maint print
1123 objfile', the output from the latter command is more compact, but
1124 also, can be restricted to a single named object file.
1126 With this change in place I am no longer seeing buffer overflow errors
1127 from expect when running gdb.base/signed-builtin-types.exp.
1129 2022-12-24 Mike Frysinger <vapier@gentoo.org>
1131 sim: or1k: move arch-specific settings to internal header
1132 There's no need for these settings to be in sim-main.h which is shared
1133 with common/ sim code, so move it all out to the existing or1k-sim.h.
1134 Unfortunately, we can't yet drop the or1k-sim.h include from sim-main.h
1135 as many of the generated CGEN files refer only to sim-main.h. We'll
1136 have to improve the CGEN interface before we can make more progress,
1137 but this is at least a minor improvement.
1139 2022-12-24 GDB Administrator <gdbadmin@sourceware.org>
1141 Automatic date update in version.in
1143 2022-12-23 Tom Tromey <tom@tromey.com>
1145 Use bool for dwarf2_has_info
1146 This changes dwarf2_has_info to return bool.
1148 2022-12-23 Indu Bhagat <indu.bhagat@oracle.com>
1150 libsframe: testsuite: fix memory leaks in testcases
1153 * libsframe/testsuite/libsframe.decode/be-flipping.c: Free
1155 * libsframe/testsuite/libsframe.decode/frecnt-1.c: Likewise.
1156 * libsframe/testsuite/libsframe.decode/frecnt-2.c: Likewise.
1158 2022-12-23 Indu Bhagat <indu.bhagat@oracle.com>
1160 libsframe: fix a memory leak in sframe_decode
1161 sframe_decode () needs to malloc a temporary buffer of the same size as
1162 the input buffer (containing the SFrame section bytes) when endian
1163 flipping is needed. The decoder keeps the endian flipped contents in
1164 this buffer for its usage. This code is necessary when the target
1165 endianneess is not the same as host endianness.
1167 The malloc'd buffer needs to be kept track of, so that it can freed up in
1168 sframe_decoder_free () later.
1172 * libsframe/sframe-impl.h (struct sframe_decoder_ctx): Add new
1173 member to keep track of the internally malloc'd buffer.
1174 * libsframe/sframe.c (sframe_decoder_free): Free it up.
1175 (sframe_decode): Update the reference to the buffer.
1177 2022-12-23 Simon Marchi <simon.marchi@polymtl.ca>
1179 gdb/testsuite: remove MPFR detection in gdb.base/float128.exp
1180 I see this fail since commit 991180627851 ("Use toplevel configure for
1181 GMP and MPFR for gdb"):
1183 FAIL: gdb.base/float128.exp: show configuration
1185 The test fails to find --with-mpfr or --without-mpfr in the "show
1186 configuration" output. Since MPFR has become mandatory, we can just
1187 remove that check and simplify the test to assume MPFR support is there.
1189 Change-Id: I4f3458470db0029705b390dfefed3a66dfc0633a
1190 Approved-By: Tom de Vries <tdevries@suse.de>
1192 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1194 sim: m32r: move arch-specific settings to internal header
1195 There's no need for these settings to be in sim-main.h which is shared
1196 with common/ sim code, so move it all out to the existing m32r-sim.h.
1197 Unfortunately, we can't yet drop the m32r-sim.h include from sim-main.h
1198 as many of the generated CGEN files refer only to sim-main.h. We'll
1199 have to improve the CGEN interface before we can make more progress,
1200 but this is at least a minor improvement.
1202 sim: bfin: move arch-specific settings to internal header
1203 There's no need for these settings to be in sim-main.h which is shared
1204 with common/ sim code, so drop the bfin.h include and move the remaining
1205 bfin-specific settings into it.
1207 sim: m68hc11: move arch-specific settings to internal header
1208 There's no need for these settings to be in sim-main.h which is shared
1209 with common/ sim code, so move it all out to a new header which only
1210 this port will include.
1212 sim: sh: move arch-specific settings to internal header
1213 There's no need for these settings to be in sim-main.h which is shared
1214 with common/ sim code, so move it all out to a new header which only
1215 this port will include.
1217 sim: mcore: move arch-specific settings to internal header
1218 There's no need for these settings to be in sim-main.h which is shared
1219 with common/ sim code, so move it all out to a new header which only
1220 this port will include.
1222 sim: h8300: move arch-specific settings to internal header
1223 There's no need for these settings to be in sim-main.h which is shared
1224 with common/ sim code, so move it all out to a new header which only
1225 this port will include.
1227 sim: pru: move arch-specific settings to internal header
1228 There's no need for these settings to be in sim-main.h which is shared
1229 with common/ sim code, so drop the pru.h include and move the remaining
1230 pru-specific settings into it.
1232 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1234 sim: mn10300: standardize the arch-specific settings a little
1235 Rename mn10300_sim.h to mn10300-sim.h to match other ports, and move most
1236 of the arch-specific content out of sim-main.h to it. This isn't a big
1237 win though as we still have to include the header in sim-main.h due to the
1238 igen interface: it hardcodes including sim-main.h in its files. So until
1239 we can fix that, we have to keep bleeding these settings into the common
1242 Also take the opportunity to purge a lot of unused headers from these.
1243 The local modules should already include the right headers, so there's
1244 no need to force everyone to pull them in. A lot of this is a hold over
1245 from the pre-igen days of this port.
1247 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1249 sim: microblaze: move arch-specific settings to internal header
1250 There's no need for these settings to be in sim-main.h which is shared
1251 with common/ sim code, so move it all out to a new header which only
1252 this port will include.
1254 sim: example-synacor: move arch-specific settings to internal header
1255 There's no need for these settings to be in sim-main.h which is shared
1256 with common/ sim code, so move it all out to a new header which only
1257 this port will include.
1259 sim: moxie: move arch-specific settings to internal header
1260 There's no need for these settings to be in sim-main.h which is shared
1261 with common/ sim code, so move it all out to a new header which only
1262 this port will include.
1264 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1266 sim: riscv: move arch-specific settings to internal header
1267 There's no need for these settings to be in sim-main.h which is shared
1268 with common/ sim code, so move it all out to a new header which only
1269 this port will include.
1271 We can also move the machs.h include out since the model logic was all
1272 generalized from compile-time to runtime last year.
1274 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1276 sim: v850: standardize the arch-specific settings a little
1277 Rename v850_sim.h to v850-sim.h to match other ports, and move most
1278 of the arch-specific content out of sim-main.h to it. This isn't a
1279 big win though as we still have to include the header in sim-main.h
1280 due to the igen interface: it hardcodes including sim-main.h in its
1281 files. So until we can fix that, we have to keep bleeding these
1282 settings into the common codes.
1284 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1286 sim: msp430: move arch-specific settings to internal header
1287 There's no need for these settings to be in sim-main.h which is shared
1288 with common/ sim code, so drop the msp430-sim.h include and move it to
1289 the few files that actually need it.
1291 While we're here, drop redundant includes from sim-main.h:
1292 * sim-config.h & sim-types.h included by sim-basics.h already
1293 * sim-engine.h included by sim-base.h already
1294 And move sim-options.h to the one file that needs it.
1296 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1298 sim: ft32: move arch-specific settings to internal header
1299 There's no need for these settings to be in sim-main.h which is shared
1300 with common/ sim code, so drop the ft32-sim.h include and move it to
1301 the few files that actually need it.
1303 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1305 sim: d10v: move arch-specific settings to internal header
1306 There's no need for these settings to be in sim-main.h which is shared
1307 with common/ sim code, so drop the d10v_sim.h include and move it to
1308 the few files that actually need it.
1310 Also rename the file to standardize it a bit better with other ports.
1312 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1314 sim: cr16: move arch-specific settings to internal header
1315 There's no need for these settings to be in sim-main.h which is shared
1316 with common/ sim code, so drop the cr16_sim.h include and move it to
1317 the few files that actually need it.
1319 Also rename the file to standardize it a bit better with other ports.
1321 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1323 sim: arm: move arch-specific settings to internal header
1324 There's no need for these settings to be in sim-main.h which is shared
1325 with common/ sim code, so move it all out to a new header which only
1326 this port will include.
1328 The BIT override would be better in the place where it's redefined, so
1329 move it to armdefs.h instead.
1331 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1333 sim: aarch64: move arch-specific settings to internal header
1334 There's no need for these settings to be in sim-main.h which is shared
1335 with common/ sim code, so move it all out to a new header which only
1336 this port will include.
1338 While we're here, drop redundant includes from sim-main.h:
1339 * sim-types.h is included by sim-base.h already
1340 * sim-base.h is included twice
1341 * sim-io.h is included by sim-base.h already
1343 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1345 sim: avr: move arch-specific settings to internal header
1346 There's no need for these settings to be in sim-main.h which is shared
1347 with common/ sim code, so move it all out to a new header which only
1348 this port will include.
1350 2022-12-23 Nick Clifton <nickc@redhat.com>
1352 Fix illegal memory access parsing corrupt DWARF information.
1354 * dwarf2.c (concat_filename): Fix check for a directory index off
1355 the end of the directory table.
1357 2022-12-23 Eli Zaretskii <eliz@gnu.org>
1359 Fix MinGW build using mingw.org's MinGW
1360 This allows to build GDB even though the default value of
1361 _WIN32_WINNT is lower than the one needed to expose some
1362 new APIs used here, and leave the test for their actual
1363 support to run time.
1364 * gdb/nat/windows-nat.c (EXTENDED_STARTUPINFO_PRESENT): Define if
1366 (create_process_wrapper): Use 'gdb_lpproc_thread_attribute_list'
1367 instead of 'PPROC_THREAD_ATTRIBUTE_LIST' (which might not be defined
1368 at compile time). This fixes compilation error using mingw.org's
1371 2022-12-23 Mark Harmstone <mark@harmstone.com>
1373 ld: Write linker symbols in PDB
1375 ld: Copy other symbols into PDB file
1377 ld: Write globals stream in PDB
1379 ld: Parse LF_UDT_SRC_LINE records when creating PDB file
1381 ld: Write types into IPI stream of PDB
1383 ld: Write types into TPI stream of PDB
1385 ld: Write DEBUG_S_LINES entries in PDB file
1387 ld: Fix segfault in populate_publics_stream
1389 ld: Write DEBUG_S_FILECHKSMS entries in PDBs
1391 ld: Generate PDB string table
1393 2022-12-23 Alan Modra <amodra@gmail.com>
1396 Enable compilation of ld/pdb.c just for x86, as is done for bfd/pdb.c.
1397 This reduces the size of ld and is necessary with the following
1398 patches that call a COFF-only bfd function from ld/pdb.c. Without it
1399 we'd break every non-COFF target build.
1401 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1403 sim: lm32/m32r: drop redundant opcode/cgen.h include
1404 The xxx-desc.h header file already includes this, and it's how the
1405 other cgen ports are getting it, so drop it from these two.
1407 sim: ppc: drop unused types from sim-main.h
1408 The common sim headers should define these for us already, so there's
1409 no need for the ppc header to set them up.
1411 sim: cgen: move symcat.h include to where it's used
1412 Move this out of the global sim-main.h and to the few files that
1413 actually use functions from it. Only the cgen ports were pulling
1414 this, so this makes cgen & non-cgen behave more the same.
1416 sim: cgen: move cgen-types.h include to cgen-defs.h
1417 The cgen-types.h header sets up types that are needed by cgen-defs.h,
1418 so move the include out of sim-main.h and to that header. It might
1419 be needed in other specific modules, but for now let's kick it out of
1420 sim-main.h to make some progress. Things still build with just this.
1422 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1424 Revert "sim: mn10300: drop unused sim-main.c"
1425 This reverts commit 681a422b855e4b20086554b170dae051361f00c7.
1427 I missed that this was included via common/sim-inline.c. I thought
1428 I had grepped the top of the tree, but I must have only done mn10300.
1430 Add a comment to make it clear where/how this file is used.
1432 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1434 sim: mn10300: drop unused sim-main.c
1435 Nothing compiles or references this, so punt it.
1437 sim: endian: move bfd.h from header to source
1438 The bfd APIs are used only by sim-n-endian.h which is only included by
1439 sim-endian.c, so move the bfd.h include there and out of sim-endian.h
1440 which is included by many other modules.
1442 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1444 sim: move bfd.h include out of sim-main.h
1445 Not all arches include this in sim-main.h, and the ones that do don't
1446 actually use bfd defines in the sim-main.h header. Prune it to make
1447 sim-main.h simpler so we can kill it off entirely in the future.
1449 We add the include to the files that utilize e.g. bfd_vma though.
1451 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1453 sim: mcore: replace custom "word" type with int32_t
1454 This is a 32-bit architecture with 32-bit registers, so replace the
1455 custom "word" long int typedef with an explicit int32_t. This is
1456 a correctness fix since long will be 64-bits on most 64-bit hosts.
1458 sim: moxie: replace custom "word" type with int32_t
1459 This is a 32-bit architecture with 32-bit registers, so replace the
1460 custom "word" int typedef with an explicit int32_t. Practically
1461 speaking, this produces the same code, but it should hopefully make
1462 it easier to merge common code in the future.
1464 sim: cr16/d10v/mcore/moxie: clean up unused word & uword types
1465 Nothing actually uses these, so punt them. Some of the ports are
1466 using local "word" types, but we'll clean those up in a follow up.
1468 sim: mips: trim redundant igen settings
1469 These variables are setting the same value as the defaults. Trim
1470 this redundant logic to make it easier to see the real differences
1471 so we can try to keep unifying cases.
1473 sim: mips: merge mips64* with existing multi-run build
1474 Change the default (unhandled) mips64* targets to use the existing
1475 mips64 multi-run build. It already handles the formats, we just
1476 have to list the mips8000 bfd for it.
1478 sim: mips: merge mips64vr5000 with existing multi-run build
1479 The existing mips64vr-* multi-run build already handles mips5000
1480 targets, so reuse that for mips64vr5* targets too. This moves
1481 more logic from build-time to runtime so we can have a single
1482 binary that supports many targets.
1484 2022-12-23 Nelson Chu <nelson@rivosinc.com>
1486 RISC-V: Relax the order checking for the architecture string
1487 * riscv-toolchain-conventions,
1488 PR, https://github.com/riscv-non-isa/riscv-toolchain-conventions/pull/14
1489 Issue, https://github.com/riscv-non-isa/riscv-toolchain-conventions/issues/11
1491 * Refer to the commit afc41ffb,
1492 RISC-V: Reorder the prefixed extensions which are out of order.
1494 In the past we only allow to reorder the prefixed extensions. But according
1495 to the PR 14 in the riscv-toolchain-convention, we can also relax the order
1496 checking to allow the whole extensions be written out of orders, including
1497 the single standard extensions and the prefixed multi-letter extensions.
1498 Just that we still need to follow the following rules as usual,
1500 1. prefixed extensions need to be seperated with `_'.
1501 2. prefixed extensions need complete <major>.<minor> version if set.
1503 Please see the details in the march-ok-reorder gas testcase.
1505 Passed the riscv-gnu-toolchain regressions.
1508 * elfxx-riscv.c (enum riscv_prefix_ext_class): Changed RV_ISA_CLASS_UNKNOWN
1509 to RV_ISA_CLASS_SINGLE, since everything that does not belong to the
1510 multi-keyword will possible be a single extension for the current parser.
1511 (parse_config): Likewise.
1512 (riscv_get_prefix_class): Likewise.
1513 (riscv_compare_subsets): Likewise.
1514 (riscv_parse_std_ext): Removed, and merged with riscv_parse_prefixed_ext
1515 into riscv_parse_extensions.
1516 (riscv_parse_prefixed_ext): Likewise.
1517 (riscv_parse_subset): Only need to call riscv_parse_extensions to parse
1518 both single standard and prefixed extensions.
1520 * testsuite/gas/riscv/march-fail-order-std.d: Removed since the relaxed
1522 * testsuite/gas/riscv/march-fail-order-std.l: Likewise.
1523 * testsuite/gas/riscv/march-fail-order-x-std.d: Likewise.
1524 * testsuite/gas/riscv/march-fail-order-z-std.d: Likewise.
1525 * testsuite/gas/riscv/march-fail-order-zx-std.l: Likewise.
1526 * testsuite/gas/riscv/march-fail-unknown-std.l: Updated.
1527 * testsuite/gas/riscv/march-ok-reorder.d: New testcase.
1529 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1531 sim: drop unused SIM_ADDR type [PR sim/7504]
1532 Now that sim APIs either use 64-bit addresses all the time, or more
1533 appropriate target-specific types, drop this now-unused 32-bit-only
1536 Bug: https://sourceware.org/PR7504
1538 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1540 sim: mips: switch from SIM_ADDR to address_word
1541 The latter type matches the address size configured for this sim.
1543 Also take the opportunity to simplify printf logic by leveraging
1546 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1548 sim: v850: switch from SIM_ADDR to address_word
1549 The latter type matches the address size configured for this sim.
1551 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1553 sim: switch sim_{read,write} APIs to 64-bit all the time [PR sim/7504]
1554 We've been using SIM_ADDR which has always been 32-bit. This means
1555 the upper 32-bit address range in 64-bit sims is inaccessible. Use
1556 64-bit addresses all the time since we want the APIs to be stable
1557 regardless of the active arch backend (which can be 32 or 64-bit).
1559 The length is also 64-bit because it's completely feasible to have
1560 a program that is larger than 4 GiB in size/image/runtime. Forcing
1561 the caller to manually chunk those accesses up into 4 GiB at a time
1562 doesn't seem useful to anyone.
1564 Bug: https://sourceware.org/PR7504
1566 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1568 sim: use bfd_vma when reading start addr from bfd info
1569 Since SIM_ADDR is always 32-bit, it might truncate the address with
1570 64-bit ELFs. Since we load that addr from the bfd, use the bfd_vma
1571 type which matches the bfd_get_start_address API.
1573 sim: m32r: include sim-hw.h for sim_hw_parse
1575 2022-12-23 Alan Modra <amodra@gmail.com>
1577 COFF build-id writes uninitialised data to file
1578 1) The first write in write_build_id wrote rubbish past the struct
1579 external_IMAGE_DEBUG_DIRECTORY, which was later overwritten with
1580 correct data. No user visible problem there, except that tools like
1582 2) The size for the pdb name was incorrectly calculated.
1584 * emultempl/pe.em (write_build_id): Write the debug directory,
1585 not the entire section contents.
1586 (setup_build_id): Add size for the base name of pdb_name, not
1588 * emultempl/pep.em: Likewise.
1589 * testsuite/ld-pe/pdb2-section-contrib.d: Update.
1591 2022-12-23 Mike Frysinger <vapier@gentoo.org>
1593 sim: mips: merge mips64vr4300 with existing multi-run build
1594 The existing mips64vr-* multi-run build already handles mips4300
1595 targets, so reuse that for mips64vr43* targets too. This moves
1596 more logic from build-time to runtime so we can have a single
1597 binary that supports many targets.
1599 2022-12-23 GDB Administrator <gdbadmin@sourceware.org>
1601 Automatic date update in version.in
1603 2022-12-22 Indu Bhagat <indu.bhagat@oracle.com>
1605 sframe: doc: update documentation for pauth key in SFrame FDE
1608 * libsframe/doc/sframe-spec.texi
1610 2022-12-22 Indu Bhagat <indu.bhagat@oracle.com>
1612 gas: sframe: testsuite: add testcase for .cfi_b_key_frame
1613 This is actually a composite test that checks SFrame unwind information
1614 generation for both the .cfi_negate_ra_state and .cfi_b_key_frame
1615 directives on aarch64.
1619 * testsuite/gas/cfi-sframe/cfi-sframe-aarch64-pac-ab-key-1.d:
1621 * testsuite/gas/cfi-sframe/cfi-sframe-aarch64-pac-ab-key-1.s:
1623 * testsuite/gas/cfi-sframe/cfi-sframe.exp: Run new test.
1625 2022-12-22 Indu Bhagat <indu.bhagat@oracle.com>
1627 objdump/readelf: sframe: emit marker for SFrame FDE with B key
1630 * libsframe/sframe-dump.c (is_sframe_abi_arch_aarch64): New
1632 (dump_sframe_func_with_fres): Emit a string if B key is used.
1634 2022-12-22 Indu Bhagat <indu.bhagat@oracle.com>
1636 gas: sframe: add support for .cfi_b_key_frame
1637 Gather the information from the DWARF FDE on whether frame's return
1638 addresses are signed using the B key or A key. Reflect the information in
1639 the SFrame counterpart data structure, the SFrame FDE.
1643 * gas/gen-sframe.c (get_dw_fde_pauth_b_key_p): New definition.
1644 (sframe_v1_set_func_info): Add new argument for pauth_key.
1645 (sframe_set_func_info): Likewise.
1646 (output_sframe_funcdesc): Likewise.
1647 * gas/gen-sframe.h (struct sframe_version_ops): Add new argument
1648 to the function pointer declaration.
1649 * gas/sframe-opt.c (sframe_convert_frag): Handle pauth_key.
1651 2022-12-22 Indu Bhagat <indu.bhagat@oracle.com>
1653 sframe.h: add support for .cfi_b_key_frame
1654 ARM 8.3 provides five separate keys that can be used to authenticate
1655 pointers. There are two key for executable (instruction) pointers. The
1656 enum pointer_auth_key in gas/config/tc-aarch64.h currently holds two keys:
1657 enum pointer_auth_key {
1658 AARCH64_PAUTH_KEY_A,
1662 Analogous to the above, in SFrame format V1, a bit is reserved in the SFrame
1663 FDE to indicate which key is used for signing the frame's return addresses:
1664 - SFRAME_AARCH64_PAUTH_KEY_A has a value of 0
1665 - SFRAME_AARCH64_PAUTH_KEY_B has a value of 1
1667 Note that the information in this bit will always be used along with the
1668 mangled_ra_p bit, the latter indicates whether the return addresses are
1669 mangled/contain PAC auth bits.
1673 * sframe.h (SFRAME_AARCH64_PAUTH_KEY_A): New definition.
1674 (SFRAME_AARCH64_PAUTH_KEY_B): Likewise.
1675 (SFRAME_V1_FUNC_INFO): Adjust to accommodate pauth_key.
1676 (SFRAME_V1_FUNC_PAUTH_KEY): New macro.
1677 (SFRAME_V1_FUNC_INFO_UPDATE_PAUTH_KEY): Likewise.
1679 2022-12-22 Jan Beulich <jbeulich@suse.com>
1681 gas: re-arrange listing output for .irp and alike
1682 It is kind of odd to have the expansions of such constructs ahead of
1683 their definition in listings with macro expansion enabled. Adjust this
1684 by pulling ahead the output of the definition lines, taking care to
1685 avoid producing a listing line for (non-existing) line 0 when the source
1688 Note that with the code movement the conditional operator isn't
1689 necessary anymore - list->line now match up.
1691 2022-12-22 Jan Beulich <jbeulich@suse.com>
1693 x86: correct/improve TSX controls
1694 TSXLDTRK takes RTM as a prereq. Additionally introduce an umbrella "tsx"
1695 extension option covering both RTM and HLE, paralleling the "abm" one we
1698 2022-12-22 Jan Beulich <jbeulich@suse.com>
1700 x86: add dependencies on SVME
1701 SEV-ES is an extension to SVME. SNP in turn is an extension to SEV-ES,
1702 and yet in turn RMPQUERY is a SNP extension.
1704 Note that cpu_arch[] has no SNP entry, so CPU_ANY_SNP_FLAGS remains
1705 unused (just like CPU_SNP_FLAGS already is).
1707 2022-12-22 Jan Beulich <jbeulich@suse.com>
1709 x86: add dependencies on VMX
1710 Both EPT and VMFUNC are extensions to VMX.
1712 2022-12-22 Jan Beulich <jbeulich@suse.com>
1714 x86: correct XSAVE* dependencies
1715 Like various other features AMX-TILE takes XSAVE as a prereq.
1717 XSAVES, unconditionally using compacted format, in turn effectively
1718 takes XSAVEC as a prereq (an SDM clarification to this effect is in the
1721 2022-12-22 Jan Beulich <jbeulich@suse.com>
1723 x86: correct dependencies of a few AVX512 sub-features
1724 Like AVX512-FP16, several other extensions require wider than 16-bit
1725 mask registers. As a result they take AVX512BW as a prereq, not (just)
1726 AVX512F. Which in turn points out wrong expectations in the noavx512-1
1729 x86: rework noavx512-1 testcase
1730 So far the set of ".noavx512*" has been accumulating, which isn't ideal.
1731 In particular this hides issues with dependencies between features.
1732 Switch back to the default ISA before disabling a particular subset.
1733 Furthermore limit redundancy by wrapping the repeated block of insns in
1736 x86: add dependencies on AVX2
1737 Like AVX-VNNI both VAES and VPCLMUL take AVX2 as a prereq, for operating
1738 on up to 256-bit packed integer vectors.
1740 x86: correct SSE dependencies
1741 SSE itself takes FXSR as a prereq. Like AES, PCLMUL, and SHA both GFNI
1742 and KL take SSE2 as a prereq, for operating on packed integers. And
1743 while correcting KL also record it as a prereq to WIDEKL.
1745 2022-12-22 Jan Beulich <jbeulich@suse.com>
1747 x86: correct what gets disabled by certain ".arch .no*"
1748 Features with prereqs as well as features with dependents cannot re-use
1749 CPU_*_MASK for the 3rd argument of SUBARCH() - they need to use
1750 CPU_ANY_*_MASK in order to avoid disabling too many (when there are
1751 prereqs) and/or too few (when there are dependents) features.
1753 Generally any CPU_ANY_*_MASK which exist should not remain unused.
1755 - FISTTP which has no corresponding entry in cpu_arch[],
1756 - IAMCU which is a base architecture and hence uses ARCH(), not
1757 SUBARCH() (only extensions can be disabled, but unlike for Cpu*86 it
1758 would be a little more clumsy to suppress generating of the #define).
1760 2022-12-22 Jan Beulich <jbeulich@suse.com>
1762 x86: re-work ISA extension dependency handling
1763 Getting both forward and reverse ISA dependencies right / consistent has
1764 been a permanent source of mistakes. Reduce what needs specifying
1765 manually to just the direct forward dependencies. Transitive forward
1766 dependencies as well as reverse ones are now derived and hence cannot go
1767 out of sync anymore (at least in the vast majority of cases; there are a
1768 few special cases to still take care of manually). In the course of this
1769 several CPU_ANY_*_FLAGS disappear, requiring adjustment to the
1770 assembler's cpu_arch[].
1772 Note that to retain the correct reverse dependency of AVX512F wrt
1773 AVX512-VP2INTERSECT, the latter has the previously missing AVX512F
1776 Note further that to avoid adding the following undue prereqs:
1777 * ATHLON, K8, and AMDFAM10 gain CMOV and FXSR,
1779 auxiliary table entries (including a colon-separated modifier) are
1780 introduced in addition to the ones representing from converting the old
1783 To maintain forward-only dependencies between AVX (XOP) and SSE* (SSE4a)
1784 (i.e. "nosse" not disabling AVX), reverse dependency tracking is
1785 artifically suppressed.
1787 As a side effect disabling of SSE or SSE2 will now also disable AES,
1788 PCLMUL, and SHA (respective elements were missing from
1789 CPU_ANY_SSE2_FLAGS).
1791 2022-12-22 Mike Frysinger <vapier@gentoo.org>
1793 sim: mips: match target on cpu settings
1794 We don't need to enforce larger target settings when the only thing
1795 the sim should care about is the CPU target. So reduce most of the
1796 target matches to only check the CPU.
1798 sim: mips: move fpu bitsize defines to top-level configure
1799 This drops support for the --enable-sim-float configure option,
1800 but it's not clear anyone ever actually used that. Eventually
1801 we'll want this to be a runtime option anyways.
1803 sim: mips: move bitsize defines to top-level configure
1804 Since the msb value is always defined as the wordsize-1, stop
1805 hardcoding that value directly, and use a CPP value instead.
1807 sim: mips: move subtarget defines to top-level configure
1808 We want to kill off mips/configure entirely. Move this small part
1809 out now to get started.
1811 sim: mips: always resolve active bfd mach dynamically
1812 Don't assume that the default bfd that we configured for is the one
1813 that is always active when running a program. We already have access
1814 to the real runtime value, so use it directly. This simplifies the
1815 code quite a bit, and will make it easier to support multiple mach's
1818 sim: hw-config.h: move generation to top-level
1819 In order to compile arch objects from the top-level, we need to
1820 generate the hw-config.h header, so move that logic up to the top
1823 sim: build: hoist lists of hw devices up
1824 We need these in the top-level to generate libsim.a, but also in the
1825 subdirs to generate hw-config.h. Move it to the local.mk, and pass
1826 it down when running recursive make. This avoids duplication, and
1827 makes it available to both. We can simplify this once we move the
1828 various steps up to the top-level too.
1830 sim: build: hoist lists of common objects up
1831 In order to create libsim.a in the common dir, we need the list of
1832 objects for each target. To avoid duplicating the list with the
1833 recursive make in each port, pass it down as a variable. This is
1834 a temporary hack until the top-level creates libsim.a for ports.
1836 2022-12-22 GDB Administrator <gdbadmin@sourceware.org>
1838 Automatic date update in version.in
1840 2022-12-21 Alan Modra <amodra@gmail.com>
1842 PR29925, Memory leak in find_abstract_instance
1843 The testcase in the PR had a variable with both DW_AT_decl_file and
1844 DW_AT_specification, where the DW_AT_specification also specified
1845 DW_AT_decl_file. This leads to a memory leak as the file name is
1846 malloced and duplicates are not expected.
1848 I've also changed find_abstract_instance to not use a temp for "name",
1849 because that can result in a change in behaviour from the usual last
1850 of duplicate attributes wins.
1853 * dwarf2.c (find_abstract_instance): Delete "name" variable.
1854 Free *filename_ptr before assigning new file name.
1855 (scan_unit_for_symbols): Similarly free func->file and
1856 var->file before assigning.
1858 2022-12-21 Andrew Pinski <apinski@marvell.com>
1860 Fix compiling of top.c
1861 When I moved my last patch forward, somehow I missed removing
1862 the #endif for the HAVE_LIBMPFR case.
1864 Committed as obvious after a quick build.
1867 * top.c: Remove the extra #endif which was missed.
1869 2022-12-21 Andrew Pinski <apinski@marvell.com>
1871 Use toplevel configure for GMP and MPFR for gdb
1872 This patch uses the toplevel configure parts for GMP/MPFR for
1873 gdb. The only thing is that gdb now requires MPFR for building.
1874 Before it was a recommended but not required library.
1875 Also this allows building of GMP and MPFR with the toplevel
1876 directory just like how it is done for GCC.
1877 We now error out in the toplevel configure of the version
1878 of GMP and MPFR that is wrong.
1880 OK after GDB 13 branches? Build gdb 3 ways:
1881 with GMP and MPFR in the toplevel (static library used at that point for both)
1882 With only MPFR in the toplevel (GMP distro library used and MPFR built from source)
1883 With neither GMP and MPFR in the toplevel (distro libraries used)
1886 * Updated gdb/README and gdb/doc/gdb.texinfo.
1887 * Regenerated using unmodified autoconf-2.69
1893 * Makefile.def: Add configure-gdb dependencies
1894 on all-gmp and all-mpfr.
1895 * configure.ac: Split out MPC checking from MPFR.
1896 Require GMP and MPFR if the gdb directory exist.
1897 * Makefile.in: Regenerate.
1898 * configure: Regenerate.
1903 * configure.ac: Remove AC_LIB_HAVE_LINKFLAGS
1905 Use GMPLIBS and GMPINC which is provided by the
1907 * Makefile.in (LIBGMP, LIBMPFR): Remove.
1908 (GMPLIBS, GMPINC): Add definition.
1909 (INTERNAL_CFLAGS_BASE): Add GMPINC.
1910 (CLIBS): Exchange LIBMPFR and LIBGMP
1912 * target-float.c: Make the code conditional on
1913 HAVE_LIBMPFR unconditional.
1914 * top.c: Remove code checking HAVE_LIBMPFR.
1915 * configure: Regenerate.
1916 * config.in: Regenerate.
1917 * README: Update GMP/MPFR section of the config
1919 * doc/gdb.texinfo: Likewise.
1921 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28500
1923 2022-12-21 Bruno Larsen <blarsen@redhat.com>
1925 gdb/c++: validate 'using' directives based on the current line
1926 When asking GDB to print a variable from an imported namespace, we only
1927 want to see variables imported in lines that the inferior has already
1928 gone through, as is being tested last in gdb.cp/nsusing.exp. However
1929 with the proposed change to gdb.cp/nsusing.exp, we get the following
1932 (gdb) PASS: gdb.cp/nsusing.exp: continue to breakpoint: marker10 stop
1935 (gdb) FAIL: gdb.cp/nsusing.exp: print x, before using statement
1938 (gdb) PASS: gdb.cp/nsusing.exp: using namespace M
1941 (gdb) PASS: gdb.cp/nsusing.exp: print x, only using M
1943 Showing that the feature wasn't functioning properly, it just so
1944 happened that gcc ordered the namespaces in a convenient way.
1945 This happens because GDB doesn't take into account the line where the
1946 "using namespace" directive is written. So long as it shows up in the
1947 current scope, we assume it is valid.
1949 To fix this, add a new member to struct using_direct, that stores the
1950 line where the directive was written, and a new function that informs if
1951 the using directive is valid already.
1953 Unfortunately, due to a GCC bug, the failure still shows up. Compilers
1954 that set the declaration line of the using directive correctly (such as
1955 Clang) do not show such a bug, so the test includes an XFAIL for gcc
1958 Finally, because the final test of gdb.cp/nsusing.exp has turned into
1959 multiple that all would need XFAILs for older GCCs (<= 4.3), and that
1960 GCC is very old, if it is detected, the test just exits early.
1962 Approved-by: Tom Tromey <tom@tromey.com>
1964 2022-12-21 Nick Clifton <nickc@redhat.com>
1966 Updated Romanian translation for the BFD sub-directory.
1968 Fix an attempt to allocate an unreasonably large amount of memory when parsing a corrupt ELF file.
1970 * objdump.c (load_specific_debug_section): Check for excessively
1973 Keep the .drectve section when performing a relocateable link.
1975 * scripttempl/pe.sc: Keep the .drectve section when performing a
1977 * scripttempl/pep.sc: Likewise.
1979 2022-12-21 Jan Beulich <jbeulich@suse.com>
1981 x86: rename CheckRegSize to CheckOperandSize
1982 While originally indeed used for register size checking only, the
1983 attribute has been used for memory operand size checking as well already
1984 for quite a while, with more such uses recently having been added.
1986 gprofng/testsuite: restrict testing to native configurations
1987 The binaries involved in testing gprofng are native ones, and hence a
1988 cross build of binutils won't really test intended functionality. Since
1989 this testing takes quite a bit of time (typically more than running all
1990 of binutils, gas, and ld testsuites together), restrict the testing to
1991 native configurations only.
1993 2022-12-21 Alan Modra <amodra@gmail.com>
1995 enable-non-contiguous-regions warnings
1996 The warning about discarded sections in elf_link_input_bfd doesn't
1997 belong there since the code is dealing with symbols. Multiple symbols
1998 in a discarded section will result in multiple identical warnings
1999 about the section. Move the warning to a new function in ldlang.c.
2001 The patch also tidies the warning quoting of section and file names,
2002 consistently using `%pA' and `%pB'. I'm no stickler for one style of
2003 section and file name quoting, but they ought to be consistent within
2004 a warning, eg. see the first one fixed in ldlang.c, and when a warning
2005 is emitted for multiple targets they all ought to use exactly the same
2006 format string to reduce translation work. elf64-ppc.c loses the
2007 build_one_stub errors since we won't get there before hitting the
2008 fatal errors in size_one_stub.
2011 * elflink.c (elf_link_input_bfd): Don't warn here about
2013 * elf32-arm.c (arm_build_one_stub): Use consistent style in
2014 --enable-non-contiguous-regions error.
2015 * elf32-csky.c (csky_build_one_stub): Likewise.
2016 * elf32-hppa.c (hppa_build_one_stub): Likewise.
2017 * elf32-m68hc11.c (m68hc11_elf_build_one_stub): Likewise.
2018 * elf32-m68hc12.c (m68hc12_elf_build_one_stub): Likewise.
2019 * elf32-metag.c (metag_build_one_stub): Likewise.
2020 * elf32-nios2.c (nios2_build_one_stub): Likewise.
2021 * elfnn-aarch64.c (aarch64_build_one_stub): Likewise.
2022 * xcofflink.c (xcoff_build_one_stub): Likewise.
2023 * elf64-ppc.c (ppc_size_one_stub): Likewise.
2024 (ppc_build_one_stub): Delete dead code.
2026 * ldlang.c (lang_add_section): Use consistent style in
2027 --enable-non-contiguous-regions warnings.
2028 (size_input_section): Likewise.
2029 (warn_non_contiguous_discards): New function.
2030 (lang_process): Call it.
2031 * testsuite/ld-arm/non-contiguous-arm.d: Update.
2032 * testsuite/ld-arm/non-contiguous-arm4.d: Update.
2033 * testsuite/ld-arm/non-contiguous-arm7.d: Add
2034 --enable-non-contiguous-regions-warnings.
2035 * testsuite/ld-arm/non-contiguous-arm7.err: New.
2036 * testsuite/ld-powerpc/non-contiguous-powerpc.d: Update.
2037 * testsuite/ld-powerpc/non-contiguous-powerpc64.d: Update.
2039 2022-12-21 Alan Modra <amodra@gmail.com>
2041 PR29922, SHT_NOBITS section avoids section size sanity check
2043 * dwarf2.c (find_debug_info): Ignore sections without
2046 2022-12-21 Mike Frysinger <vapier@gentoo.org>
2048 sim: fully merge sim_cpu_base into sim_cpu
2049 Now that all ports have migrated to the new framework, drop support
2050 for the old sim_cpu_base layout. There's a lot of noise here, so
2051 it's been split into a dedicated commit.
2053 sim: enable common sim_cpu usage everywhere
2054 All ports should be migrated now. Drop the SIM_HAVE_COMMON_SIM_CPU
2055 knob and require it be used everywhere now.
2057 sim: or1k: invert sim_cpu storage
2058 The cpu.h change is in generated cgen code, but that has been sent
2059 upstream too, so the next regen should include it automatically.
2061 sim: m32r: invert sim_cpu storage
2062 The cpu*.h changes are in generated cgen code, but that has been sent
2063 upstream too, so the next regen should include it automatically.
2065 sim: lm32: invert sim_cpu storage
2066 The cpu.h change is in generated cgen code, but that has been sent
2067 upstream too, so the next regen should include it automatically.
2069 sim: iq2000: invert sim_cpu storage
2070 The cpu.h change is in generated cgen code, but that has been sent
2071 upstream too, so the next regen should include it automatically.
2073 sim: frv: invert sim_cpu storage
2074 The cpu.h change is in generated cgen code, but that has been sent
2075 upstream too, so the next regen should include it automatically.
2077 sim: cris: invert sim_cpu storage
2078 The cpu*.h changes are in generated cgen code, but that has been sent
2079 upstream too, so the next regen should include it automatically.
2081 sim: bpf: invert sim_cpu storage
2082 The cpu.h change is in generated cgen code, but that has been sent
2083 upstream too, so the next regen should include it automatically.
2085 sim: cgen: prep for inverting sim_cpu storage
2086 Some common cgen code changes to allow cgen ports to invert their
2087 sim_cpu storage one-by-one.
2089 sim: riscv: invert sim_cpu storage
2091 sim: pru: invert sim_cpu storage
2093 sim: example-synacor: invert sim_cpu storage
2095 sim: h8300: invert sim_cpu storage
2097 sim: m68hc11: invert sim_cpu storage
2099 sim: mips: invert sim_cpu storage
2101 sim: v850: invert sim_cpu storage
2103 sim: mcore: invert sim_cpu storage
2105 sim: aarch64: invert sim_cpu storage
2107 sim: microblaze: invert sim_cpu storage
2109 sim: avr: invert sim_cpu storage
2111 sim: moxie: invert sim_cpu storage
2113 sim: msp430: invert sim_cpu storage
2115 sim: ft32: invert sim_cpu storage
2117 sim: bfin: invert sim_cpu storage
2119 2022-12-21 Mike Frysinger <vapier@gentoo.org>
2121 sim: sim_cpu: invert sim_cpu storage
2122 Currently all ports have to declare sim_cpu themselves in their
2123 sim-main.h and then embed the common sim_cpu_base in it. This
2124 dynamic makes it impossible to share common object code among
2125 multiple ports because the core data structure is always different.
2127 Let's invert this relationship: common code declares sim_cpu, and
2128 the port uses the new arch_data field for its per-cpu state.
2130 This is the first in a series of changes: it adds a define to select
2131 between the old & new layouts, then converts all the ports that don't
2132 need custom state over to the new layout. This includes mn10300 that,
2133 while it defines custom fields in its cpu struct, never uses them.
2135 2022-12-21 Mike Frysinger <vapier@gentoo.org>
2137 sim: move register headers into sim/ namespace [PR sim/29869]
2138 These headers define the register numbers for each port to implement
2139 the sim_fetch_register & sim_store_register interfaces. While gdb
2140 uses these, the APIs are part of the sim, not gdb. Move the headers
2141 out of the gdb/ include namespace and into sim/ instead.
2143 sim: ppc: drop old dgen.c generator
2144 The spreg.[ch] files live in the source tree now and are created
2145 with the dgen.py script, so we don't need this old tool anymore.
2147 2022-12-21 Mike Frysinger <vapier@gentoo.org>
2149 sim: ppc: move spreg.[ch] files to the source tree
2150 Simplify the build by moving the generation of these files from
2151 build-time (via dgen.c that we have to compile & execute on the
2152 build system) to maintainer/release mode (via spreg-gen.py that
2153 we only ever execute when the spreg table actually changes). It
2154 speeds up the build process and makes it easier for us to reason
2155 about & review changes to the code generator.
2157 The tool is renamed from "dgen" because it's hardcoded to only
2158 generated spreg files. It isn't a generalized tool for creating
2161 2022-12-21 GDB Administrator <gdbadmin@sourceware.org>
2163 Automatic date update in version.in
2165 2022-12-20 Hannes Domani <ssbssa@yahoo.de>
2167 Fix install-strip target
2168 The libtool patch broke install-strip of gdb:
2170 /bin/sh ../../gdb/../mkinstalldirs /src/gdb/inst/share/gdb/python/gdb
2171 transformed_name=`t='s,y,y,'; \
2172 echo gdb | sed -e "$t"` ; \
2173 if test "x$transformed_name" = x; then \
2174 transformed_name=gdb ; \
2178 /bin/sh ../../gdb/../mkinstalldirs /src/gdb/inst/bin ; \
2179 /bin/sh ./libtool --mode=install STRIPPROG='strip' /bin/sh /src/gdb/gdb.git/install-sh -c -s \
2181 /src/gdb/inst/bin/$transformed_name ; \
2182 /bin/sh ../../gdb/../mkinstalldirs /src/gdb/inst/include/gdb ; \
2183 /usr/bin/install -c -m 644 jit-reader.h /src/gdb/inst/include/gdb/jit-reader.h
2184 libtool: install: `/src/gdb/inst/bin/gdb' is not a directory
2185 libtool: install: Try `libtool --help --mode=install' for more information.
2187 Since INSTALL_PROGRAM_ENV is no longer at the beginning of the command, the
2188 gdb executable is not installed with install-strip.
2190 2022-12-20 Torbjörn SVENSSON <torbjorn.svensson@foss.st.com>
2192 bfd: Discard symbol regardless of warning flag
2193 The discard of symbols should be performed whether the warning for
2194 the discard is enabled or not.
2195 Without this patch, ld would segfault in bfd_section_removed_from_list,
2196 called in the if-statement right after this block, as the argument
2197 isec->output_section can be NULL.
2199 Co-Authored-By: Yvan ROUX <yvan.roux@foss.st.com>
2201 2022-12-20 Alan Modra <amodra@gmail.com>
2203 PR29915, bfdio.c does not compile with mingw.org's MinGW
2205 * configure.ac: Add AC_CHECK_DECLS test ___lc_codepage_func.
2206 * configure: Regenerate.
2207 * config.in: Regenerate.
2208 * bfdio.c (___lc_codepage_func): Move declaration to..
2209 (_bfd_real_fopen): ..here, and use !HAVE_DECL____LC_CODEPAGE_FUNC.
2211 Re: x86: remove i386-opc.c
2212 Regen opcodes/po/POTFILES.in
2214 2022-12-20 Mike Frysinger <vapier@gentoo.org>
2216 sim: ppc: change spreg switch table generation to compile-time
2217 Simplify the generator by always outputting the switch tables, and
2218 leave the choice of whether to use them to the compiler via a -D
2221 2022-12-20 Mike Frysinger <vapier@gentoo.org>
2223 sim: dv-core: add hw_detach_address method [PR sim/25211]
2224 The core device has an attach address method as the root of the tree
2225 which calls out to the sim API. But it doesn't have a corresponding
2226 detach method which means we just crash if anything tries to detach
2227 itself from the core. In practice, the m68hc11 is the only model
2228 that actually tries to detach itself on the fly, so no one noticed
2231 With this in place, we can delete the existing detach code from the
2232 m68hc11 model since it defaults to "passthru" callback which will in
2233 turn call the dv-core detach, and they have the same behavior -- call
2234 the sim core API to detach from the address space.
2236 Bug: https://sourceware.org/PR25211
2238 2022-12-20 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2240 gprofng: PR29646 Various warnings
2242 2022-12-19 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2245 * common/core_pcbe.c: Fix missingReturn warning.
2246 * libcollector/iolib.c: Fix -Waddress warnings.
2247 * src/Settings.cc: Likewise.
2248 * src/checks.cc: Likewise.
2249 * libcollector/linetrace.c: Likewise.
2250 * libcollector/iotrace.c: Fix va_end_missing error.
2251 * libcollector/libcol_util.c: Fix uninitvar warning.
2252 * src/Command.cc: Fix arrayIndexOutOfBounds error.
2253 * src/Function.cc: Fix uninitStructMember warning.
2254 * src/ipc.cc: Fix -Wwrite-strings warnings.
2256 2022-12-20 GDB Administrator <gdbadmin@sourceware.org>
2258 Automatic date update in version.in
2260 2022-12-19 Tom Tromey <tromey@adacore.com>
2262 Avoid compiler warning in dwarf-do-refresh
2263 The Emacs 28 compiler warns about dwarf-mode.el:
2265 Warning (comp): dwarf-mode.el:180:32: Warning: Unused lexical argument `ignore'
2267 This is easily fixed by prepending "_" to the parameter's name.
2270 2022-12-19 Tom Tromey <tromey@adacore.com>
2272 * dwarf-mode.el (dwarf-do-refresh): Avoid compiler warning.
2274 2022-12-19 Tom Tromey <tom@tromey.com>
2277 This changes bpstat to use 'bool' rather than 'char', and updates the
2280 Use bool constants for value_print_options
2281 This changes the uses of value_print_options to use 'true' and 'false'
2282 rather than integers.
2284 Remove quick_symbol_functions::relocated
2285 quick_symbol_functions::relocated is only needed for psymtabs, and
2286 there it is only needed for Rust. However, because we've switched the
2287 DWARF reader away from psymtabs, this means there's no longer a need
2288 for this method at all.
2290 2022-12-19 Tom Tromey <tom@tromey.com>
2293 MI version 1 is long since obsolete. Several years ago, I filed
2294 PR mi/23170 for this. I think it's finally time to remove this.
2295 Any users of MI 1 can and should upgrade to a newer version.
2297 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=23170
2299 2022-12-19 Tom Tromey <tom@tromey.com>
2301 Remove vestiges of MI version 0
2302 I found a few vestiges of MI version 0 in the test suite. This patch
2305 2022-12-19 Alan Modra <amodra@gmail.com>
2307 Tidy PR29893 and PR29908 fix
2310 * dwarf.c (display_debug_addr): Combine dwarf5 unit_length checks.
2313 2022-12-19 Jan Vrany <jan.vrany@labware.com>
2315 gdb: fix command lookup in execute_command ()
2316 Commit b5661ff2 ("gdb: fix possible use-after-free when
2317 executing commands") used lookup_cmd_exact () to lookup
2318 command again after its execution to avoid possible
2319 use-after-free error.
2321 However this change broke test gdb.base/define.exp which
2322 defines a post-hook for subcommand ("target testsuite").
2323 In this case, lookup_cmd_exact () returned NULL because
2324 there's no command 'testsuite' in top-level commands.
2326 This commit fixes this case by looking up the command again
2327 using the original command line via lookup_cmd ().
2329 Approved-By: Simon Marchi <simon.marchi@efficios.com>
2331 2022-12-19 Nick Clifton <nickc@redhat.com>
2333 Fix potential illegal memory accesses when parsing corrupt DWARF data.
2335 * dwarf.c (fetch_indexed_value): Fail if the section is not big
2336 enough to contain a header size field.
2337 (display_debug_addr): Fail if the computed address size is too big
2340 New Romainian translation for the GOLD subdirectory.
2342 2022-12-19 Jan Beulich <jbeulich@suse.com>
2344 gprofng/testsuite: skip Java test without JDK
2345 There's no point in even trying the Java test when gprofng was built
2346 without Java support, and when the building of the constituents of the
2347 testcase also would fail. On such systems this converts the respective
2348 tests from "unresolved" to "unsupported", making the overall testsuite
2349 run no longer report failure just because of this.
2351 gprofng/testsuite: eliminate bogus casts
2352 Casting pointers to unsigned int is generally problematic and hence
2353 compilers tend to warn about such. While here they're used only in
2354 fprintf(), it still seems better to omit such casts, even if only to
2355 avoid setting bad precedents.
2357 gprofng/testsuite: correct line continuation in endcases.c
2358 A backslash used to indicate line continuation (in a macro definition
2359 here) is not supposed to be followed by blanks or other white space; the
2360 end-of-line indicator is to follow immediately.
2362 gprofng/testsuite: correct names for signal handling tests
2363 The signal handling tests spend most of their time in the signal
2364 handlers, and hence for profile output to match anything in program
2365 output, the respective name fields need to hold the handler function
2366 names. This converts both respective tests from "unresolved" to actually
2369 gprofng/testsuite: adjust linking of synprog
2370 In order for so_syn.so and so_syx.so to be able to access the main
2371 program's "testtime" variable, that variable needs exposing in the
2372 dynamic symbol table. Since this is a test program only, do it the brute
2373 force way and simply expose all global symbols.
2375 Arm: break gas dependency on libopcodes
2376 gas doesn't use anything from libopcodes (anymore?) - suppress linking
2379 x86: omit Cpu prefixes from opcode table
2380 These enumerators can be used in only one specific field, and hence the
2381 Cpu prefix isn't needed ther for disambiguation / name space separation.
2383 2022-12-19 GDB Administrator <gdbadmin@sourceware.org>
2385 Automatic date update in version.in
2387 2022-12-18 Alan Modra <amodra@gmail.com>
2389 Comment bfd_get_section_limit_octets and bfd_get_section_alloc_size
2390 * bfd.c (bfd_get_section_limit_octets): Add comment.
2391 (bfd_get_section_alloc_size): Likewise.
2392 * libbfd.c (_bfd_generic_get_section_contents): Use
2393 bfd_get_section_limit_octets.
2394 * section.c (bfd_get_section_contents): Likewise.
2395 * bfd-in2.h: Regenerate.
2397 2022-12-18 Alan Modra <amodra@gmail.com>
2399 ld bootstrap test in build dir with path containing symlinks
2400 This allows the bootstrap test to run if you have a symlink somewhere
2401 in the build path directory. $ld depends on $base_dir which is set
2402 via tcl [pwd], collapsing the symlink like /usr/bin/pwd, while $objdir
2403 contains the symlink.
2405 * testsuite/ld-bootstrap/bootstrap.exp: Normalize paths when
2406 checking for ld build directory.
2408 2022-12-18 Joel Brobecker <brobecker@adacore.com>
2410 Update gdb/NEWS after GDB 13 branch creation.
2411 This commit a new section for the next release branch, and renames
2412 the section of the current branch, now that it has been cut.
2414 2022-12-18 Joel Brobecker <brobecker@adacore.com>
2416 Bump version to 14.0.50.DATE-git.
2417 Now that the GDB 13 branch has been created,
2418 this commit bumps the version number in gdb/version.in to
2421 For the record, the GDB 13 branch was created
2422 from commit 71c90666e601c511a5f495827ca9ba545e4cb463.
2424 Also, as a result of the version bump, the following changes
2425 have been made in gdb/testsuite:
2427 * gdb.base/default.exp: Change $_gdb_major to 14.
2429 2022-12-18 GDB Administrator <gdbadmin@sourceware.org>
2431 Automatic date update in version.in
2433 2022-12-17 Alan Modra <amodra@gmail.com>
2435 bfd_get_relocated_section_contents allow NULL data buffer
2436 This patch removes the bfd_malloc in default_indirect_link_order and
2437 bfd_simple_get_relocated_section_contents, pushing the allocation down
2438 to bfd_get_relocated_section_contents. The idea is to make use of the
2439 allocation done with sanity checking in bfd_get_full_section_contents,
2440 which is called by bfd_generic_get_relocated_section_contents.
2442 Doing this exposed a bug in bfd_get_full_section_contents. With
2443 relaxation it is possible that an input section rawsize is different
2444 to the section size. In that case we want to use the larger of
2445 rawsize (the on-disk size for input sections) and size.
2447 * reloc.c (bfd_generic_get_relocated_section_contents),
2448 * reloc16.c (bfd_coff_reloc16_get_relocated_section_contents),
2449 * coff-alpha.c (alpha_ecoff_get_relocated_section_contents),
2450 * coff-sh.c (sh_coff_get_relocated_section_contents),
2451 * elf-m10200.c (mn10200_elf_get_relocated_section_contents),
2452 * elf-m10300.c (mn10300_elf_get_relocated_section_contents),
2453 * elf32-avr.c (elf32_avr_get_relocated_section_contents),
2454 * elf32-cr16.c (elf32_cr16_get_relocated_section_contents),
2455 * elf32-crx.c (elf32_crx_get_relocated_section_contents),
2456 * elf32-h8300.c (elf32_h8_get_relocated_section_contents),
2457 * elf32-nds32.c (nds32_elf_get_relocated_section_contents),
2458 * elf32-sh.c (sh_elf_get_relocated_section_contents),
2459 * elfxx-mips.c (_bfd_elf_mips_get_relocated_section_contents):
2460 Handle NULL data buffer.
2461 * bfd.c (bfd_get_section_alloc_size): New function.
2462 * bfd-in2.h: Regenerate.
2463 * compress.c (bfd_get_full_section_contents): Correct section
2465 * linker.c (default_indirect_link_order): Don't malloc memory
2466 here before calling bfd_get_relocated_section_contents.
2467 * simple.c (bfd_simple_get_relocated_section_contents): Likewise.
2469 2022-12-17 Alan Modra <amodra@gmail.com>
2471 asan: elf.c:12621:18: applying zero offset to null pointer
2472 That's this line in elf_parse_notes:
2473 while (p < buf + size)
2475 * elf.c (_bfd_elf_make_section_from_shdr): Don't call
2476 elf_parse_notes when sh_size is zero.
2478 2022-12-17 Alan Modra <amodra@gmail.com>
2480 Re: The problem with warning in elf_object_p
2481 Commit 5aa0f10c424e added a per_xvec_warn array to provide support for
2482 warnings from elf_object_p (and a later patch for warnings from
2483 pe_bfd_object_p) to be cached and then only printed if the target
2484 matches. It was quite limited in the style of message supported, only
2485 one message could be printed, and didn't really meet the stated aim of
2486 only warning when a target matches: There are many other errors and
2487 warnings that can be emitted by functions called from elf_object_p.
2489 So this patch extends the error handler functions to support printing
2490 to a string buffer, extends per_xvec_warn to support multiple errors/
2491 warnings, and hooks this all into bfd_check_format_matches. If
2492 bfd_check_format_matches succeeds then any errors/warnings are printed
2493 for the matching target. If bfd_check_format_matches fails either due
2494 to no match or to multiple matches and only one target vector produced
2495 errors, then those errors are printed.
2497 * bfd.c (MAX_ARGS): Define, use throughout.
2498 (print_func): New typedef.
2499 (_bfd_doprnt): Add new print param. Replace calls to fprintf
2501 (PRINT_TYPE): Similarly.
2502 (error_handler_fprintf): Renamed from error_handler_internal.
2503 Use _bfd_get_error_program_name. Add fprintf arg. Move code
2505 (_bfd_doprnt_scan): ..to here. Add ap param.
2506 (struct buf_stream): New.
2507 (err_sprintf): New function.
2508 (error_handler_bfd): New static variable.
2509 (error_handler_sprintf): New function.
2510 (_bfd_set_error_handler_caching): New function.
2511 (_bfd_get_error_program_name): New function.
2512 * elfcode.h (elf_swap_shdr_in): Use _bfd_error_handler in
2514 (elf_object_p): Likewise.
2515 * format.c (print_warnmsg): New function.
2516 (clear_warnmsg): Rewrite.
2517 (null_error_handler): New function.
2518 (bfd_check_format_matches): Ignore warnings from recursive calls
2519 checking first element of an archive. Use caching error handler
2520 otherwise. Print warnings on successful match, or when only one
2521 target has emitted warnings/errors.
2522 * peicode.h (pe_bfd_object_p): Use _bfd_error_handler in
2524 * targets.c (per_xvec_warn): Change type of array elements.
2525 (struct per_xvec_message): New.
2526 (_bfd_per_xvec_warn): Rewrite.
2527 * Makefile.am (LIBBFD_H_FILES): Add bfd.c.
2528 * Makefile.in: Regenerate.
2529 * bfd-in2.h: Regenerate.
2530 * libbfd.h: Regenerate.
2532 2022-12-17 Indu Bhagat <indu.bhagat@oracle.com>
2534 sframe: doc: update spec for the mangled-RA bit in FRE
2537 * libsframe/doc/sframe-spec.texi
2539 2022-12-17 Indu Bhagat <indu.bhagat@oracle.com>
2541 gas: sframe: testsuite: add testcase for .cfi_negate_ra_state
2542 Add a new test to check that .cfi_negate_ra_state on aarch64 is handled
2543 well (a non-empty SFrame section with valid SFrame FREs is generated).
2547 * testsuite/gas/cfi-sframe/cfi-sframe-aarch64-2.d: New test.
2548 * testsuite/gas/cfi-sframe/cfi-sframe-aarch64-2.s: Likewise.
2549 * testsuite/gas/cfi-sframe/cfi-sframe.exp: Adjust the list
2552 2022-12-17 Indu Bhagat <indu.bhagat@oracle.com>
2554 objdump/readelf: sframe: emit marker for FREs with mangled RA
2555 In the textual dump of the SFrame section, when an SFrame FRE recovers a
2556 mangled RA, use string "[s]" in the output to indicate that the return
2557 address is a signed (mangled) one.
2561 * libsframe/sframe-dump.c (dump_sframe_func_with_fres): Postfix
2562 with "[s]" if RA is signed with authorization code.
2564 2022-12-17 Indu Bhagat <indu.bhagat@oracle.com>
2566 libsframe: provide new access API for mangled RA bit
2569 * sframe-api.h (sframe_fre_get_ra_mangled_p): New declaration.
2573 * libsframe/sframe.c (sframe_get_fre_ra_mangled_p): New
2575 (sframe_fre_get_ra_mangled_p): New static function.
2577 2022-12-17 Indu Bhagat <indu.bhagat@oracle.com>
2579 gas: sframe: add support for .cfi_negate_ra_state
2580 DW_CFA_AARCH64_negate_ra_state in aarch64 is multiplexed with
2581 DW_CFA_GNU_window_save in the DWARF format.
2583 Remove the common-empty-4 testcase because the generated SFrame section
2584 will not be be empty anymore. A relevant test will be added in a later
2589 * gas/gen-sframe.c (sframe_v1_set_fre_info): Add new argument
2591 (sframe_set_fre_info): Likewise.
2592 (output_sframe_row_entry): Handle mangled_ra_p.
2593 (sframe_row_entry_new): Reset mangled_ra_p.
2594 (sframe_row_entry_initialize): Initialize mangled_ra_p.
2595 (sframe_xlate_do_gnu_window_save): New definition.
2596 (sframe_do_cfi_insn): Handle DW_CFA_GNU_window_save.
2597 * gas/gen-sframe.h (struct sframe_row_entry): New member.
2598 (struct sframe_version_ops): Add a new argument for
2600 * gas/testsuite/gas/cfi-sframe/cfi-sframe.exp: Remove test.
2601 * gas/testsuite/gas/cfi-sframe/common-empty-4.d: Removed.
2602 * gas/testsuite/gas/cfi-sframe/common-empty-4.s: Removed.
2604 2022-12-17 Indu Bhagat <indu.bhagat@oracle.com>
2606 sframe.h: add support for .cfi_negate_ra_state
2607 Use the last remaining bit in the 'SFrame FRE info' word to store whether
2608 the RA is signed/unsigned with PAC authorization code: this bit is named
2609 as the "mangled RA" bit. This bit is still unused for x86-64.
2611 The behaviour of the mangled-RA info bit in SFrame format closely
2612 follows the behaviour of DW_CFA_AARCH64_negate_ra_state in DWARF. During
2613 unwinding, whenever an SFrame FRE with non-zero "mangled RA" bit is
2614 encountered, it means the upper bits of the return address contain Pointer
2615 Authentication code. The unwinder, hence, must use appropriate means to
2616 restore LR correctly in such cases.
2620 * sframe.h (SFRAME_V1_FRE_INFO_UPDATE_MANGLED_RA_P): New macro.
2621 (SFRAME_V1_FRE_MANGLED_RA_P): Likewise.
2623 2022-12-17 GDB Administrator <gdbadmin@sourceware.org>
2625 Automatic date update in version.in
2627 2022-12-16 Pedro Alves <pedro@palves.net>
2629 Delay checking whether /proc/pid/mem is writable (PR gdb/29907)
2630 As of 1bcb0708f229 ("gdb/linux-nat: Check whether /proc/pid/mem is
2631 writable"), GDB checks if /proc/pid/mem is writable. This is done
2632 early at GDB startup, in order to get a consistent warning, instead of
2633 a warning that depends on whenever GDB writes to inferior memory.
2635 PR gdb/29907 points out that some build systems (like QEMU's,
2636 apparently) may call 'gdb --version' to check GDB's presence & its
2637 version on the system, and that Gentoo's build process has sandboxing
2638 which blocks the /proc/pid/mem access and thus GDB warns, which
2639 results in build fails.
2641 To help with that, this patch delays the /proc/pid/mem check until we
2642 start or attach to an inferior. Ends up potentially emiting a warning
2643 close where we already emit other ptrace- and /proc- related warnings,
2644 which just Feels Right.
2646 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29907
2647 Change-Id: I5537653ecfbbe76a04ab035e40e59d09b4980763
2649 2022-12-16 Nick Clifton <nickc@redhat.com>
2651 Fix previous delta to allow for compilation on 32-bit systems
2653 2022-12-16 Tom de Vries <tdevries@suse.de>
2655 [gdb/testsuite] Fix race in gdb.threads/detach-step-over.exp
2656 Once in a while I run into:
2658 FAIL: gdb.threads/detach-step-over.exp: \
2659 breakpoint-condition-evaluation=host: target-non-stop=off: non-stop=off: \
2660 displaced=off: iter 1: all threads running
2663 In can easily reproduce this by doing:
2665 # Wait a bit, to give time for the threads to hit the
2672 Fix this by counting the running threads in a loop, effectively allowing 10
2673 seconds (instead of 1) for the threads to start running, but only sleeping if
2676 Reduces total execution time from 1m27s to 56s.
2678 Tested on x86_64-linux.
2680 2022-12-16 Andrew Burgess <aburgess@redhat.com>
2682 gdb: fix crash when getting the value of a label symbol
2683 When the source program contains a goto label, it turns out it's
2684 actually pretty hard for a user to find out more about that label.
2688 No symbol "some_label" in current context.
2689 (gdb) disassemble some_label
2690 No symbol "some_label" in current context.
2691 (gdb) x/10i some_label
2692 No symbol "some_label" in current context.
2693 (gdb) break some_label
2694 Breakpoint 2 at 0x401135: file /tmp/py-label-symbol-value.c, line 35.
2696 In all cases, some_label is a goto label within the current frame.
2697 Only placing a breakpoint on the label worked.
2699 This all seems a little strange to me, it feels like asking about a
2700 goto label would not be an unreasonable thing for a user to do.
2702 This commit doesn't fix any of the above issues, I mention them just
2703 to provide a little context for why the following issue has probably
2704 not been seen before.
2706 It turns out there is one way a user can access the symbol for a goto
2707 label, through the Python API:
2709 python frame = gdb.selected_frame()
2710 python frame_pc = frame.pc()
2711 python block = gdb.current_progspace().block_for_pc(frame_pc)
2712 python symbol,_ = gdb.lookup_symbol('some_label', block, gdb.SYMBOL_LABEL_DOMAIN)
2713 python print(str(symbol.value()))
2714 ../../src/gdb/findvar.c:204: internal-error: store_typed_address: Assertion `type->is_pointer_or_reference ()' failed.
2716 The problem is that label symbols are created using the
2717 builtin_core_addr type, which is a pure integer type.
2719 When GDB tries to fetch the value of a label symbol then we end up in
2720 findvar.c, in the function language_defn::read_var_value, in the
2721 LOC_LABEL case. From here store_typed_address is called to store the
2722 address of the label into a value object with builtin_core_addr type.
2724 The problem is that store_typed_address requires that the destination
2725 type be a pointer or reference, which the builtin_core_addr type is
2728 Now it's not clear what type a goto label address should have, but
2729 GCC has an extension that allows users to take the address of a goto
2730 label (using &&), in that case the result is of type 'void *'.
2732 I propose that when we convert the CORE_ADDR value to a GDB value
2733 object, we use builtin_func_ptr type instead of builtin_core_addr,
2734 this means the result will be of type 'void (*) ()'. The benefit of
2735 this approach is that when gdbarch_address_to_pointer is called the
2736 target type will be correctly identified as a pointer to code, which
2737 should mean any architecture specific adjustments are done correctly.
2739 We can then cast the new value to 'void *' type with a call to
2740 value_cast_pointer, this should not change the values bit
2741 representation, but will just update the type.
2743 After this asking for the value of a label symbol works just fine:
2745 (gdb) python print(str(symbol.value()))
2748 And the type is maybe what we'd expect:
2750 (gdb) python print(str(symbol.value().type))
2753 2022-12-16 Simon Marchi <simon.marchi@polymtl.ca>
2755 gdb: convert linux-osdata.c from buffer to std::string
2756 Replace the use of struct buffer in linux-osdata.c with std::string.
2757 There is no change in the logic, so there should be no user-visible
2760 Change-Id: I27f53165d401650bbd0bebe8ed88221e25545b3f
2761 Approved-By: Pedro Alves <pedro@palves.net>
2763 2022-12-16 Simon Marchi <simon.marchi@polymtl.ca>
2765 gdbsupport: add string_xml_appendf
2766 Add a version of buffer_xml_printf (defined in gdbsupport/buffer.{c,h})
2767 that appends to an std::string, rather than a struct buffer. Call it
2768 "string" rather than "buffer" since it operates on an std::string rather
2769 than a buffer. And call it "appendf" rather than "printf", since it
2770 appends to and does not replace the string's content. This mirrors
2773 Place the new version in gdbsupport/xml-utils.h.
2775 The code is a direct copy of buffer_xml_printf. The old version is
2776 going to disappear at some point, which is why I didn't do any effort to
2779 Change-Id: I30e030627ab4970fd0b9eba3b7e8cec78fa561ba
2780 Approved-By: Pedro Alves <pedro@palves.net>
2782 2022-12-16 Andrew Burgess <aburgess@redhat.com>
2784 gdb: clean up some inefficient std::string usage
2787 commit 53cf95c3389a3ecd97276d322e4a60fe3396a201
2788 Date: Wed Dec 14 14:17:44 2022 +0000
2790 gdb: make more use of make_target_connection_string
2792 Introduced a couple of inefficient uses of std::string, both of which
2793 are fixed in this commit.
2795 There should be no user visible changes after this commit.
2797 Approved-By: Simon Marchi <simon.marchi@efficios.com>
2799 2022-12-16 Nick Clifton <nickc@redhat.com>
2801 Fix a potential illegal memory access when parsing corrupt DWARF information.
2803 * dwarf.c (display_debug_addr): Check for corrupt header lengths.
2805 2022-12-16 Jan Vrany <jan.vrany@labware.com>
2807 gdb/testsuite: add test for Python commands redefining itself
2808 This commit adds a test that creates a Python command that redefines
2809 itself during its execution. This is to test use-after-free in
2812 This test needs run with ASan enabled in order to fail when it
2815 Approved-By: Simon Marchi <simon.marchi@efficios.com>
2817 2022-12-16 Luis Machado <luis.machado@arm.com>
2819 [aarch64] Fix removal of non-address bits for PAuth
2822 The address_significant gdbarch setting was introduced as a way to remove
2823 non-address bits from pointers, and it is specified by a constant. This
2824 constant represents the number of address bits in a pointer.
2826 Right now AArch64 is the only architecture that uses it, and 56 was a
2827 correct option so far.
2829 But if we are using Pointer Authentication (PAuth), we might use up to 2 bytes
2830 from the address space to store the required information. We could also have
2831 cases where we're using both PAuth and MTE.
2833 We could adjust the constant to 48 to cover those cases, but this doesn't
2834 cover the case where GDB needs to sign-extend kernel addresses after removal
2835 of the non-address bits.
2837 This has worked so far because bit 55 is used to select between kernel-space
2838 and user-space addresses. But trying to clear a range of bits crossing the
2839 bit 55 boundary requires the hook to be smarter.
2841 The following patch renames the gdbarch hook from significant_addr_bit to
2842 remove_non_address_bits and passes a pointer as opposed to the number of
2843 bits. The hook is now responsible for removing the required non-address bits
2844 and sign-extending the address if needed.
2846 While at it, make GDB and GDBServer share some more code for aarch64 and add a
2847 new arch-specific testcase gdb.arch/aarch64-non-address-bits.exp.
2849 Bug-url: https://sourceware.org/bugzilla/show_bug.cgi?id=28947
2851 Approved-By: Simon Marchi <simon.marchi@efficios.com>
2853 2022-12-16 Jan Beulich <jbeulich@suse.com>
2855 gas: restore Dwarf info generation after macro diagnostic adjustments
2856 While 6fdb723799e2 ("gas: re-work line number tracking for macros and
2857 their expansions") was meant to leave generated Dwarf as is, it really
2858 didn't (and the testcase intended to catch that wasn't covering the case
2859 which broke). Its adjustment to buffer_and_nest() didn't go far enough,
2860 leading to the "linefile" directive inserted at the top to also be
2861 processed later in the PR gas/16908 workaround (which clearly isn't
2862 intended - it's being put there for processing during macro expansion
2863 only). That unnoticed flaw in turn led me to work around it by a
2864 (suspicious to me already at the time) conditional in as_where().
2866 x86: change representation of extension opcode
2867 Having a "None" field in the vast majority of entries is needlessly
2868 cluttering the overall table. Instead of this being a separate field,
2869 use a representation matching that of Intel SDM and AMD PM for the main
2870 use of the field: Append the value after a / as the separator.
2872 2022-12-16 Simon Marchi <simon.marchi@polymtl.ca>
2874 gdbsupport: change xml_escape_text_append's parameter from pointer to reference
2875 The passed in string can't be nullptr, it makes more sense to pass in a
2878 Change-Id: Idc8bd38abe1d6d9b44aa227d7856956848c233b3
2880 2022-12-16 Simon Marchi <simon.marchi@polymtl.ca>
2882 gdb: remove static buffer in command_line_input
2883 [I sent this earlier today, but I don't see it in the archives.
2884 Resending it through a different computer / SMTP.]
2886 The use of the static buffer in command_line_input is becoming
2887 problematic, as explained here [1]. In short, with this patch [2] that
2888 attempt to fix a post-hook bug, when running gdb.base/commands.exp, we
2889 hit a case where we read a "define" command line from a script file
2890 using command_command_line_input. The command line is stored in
2891 command_line_input's static buffer. Inside the define command's
2892 execution, we read the lines inside the define using command_line_input,
2893 which overwrites the define command, in command_line_input's static
2894 buffer. After the execution of the define command, execute_command does
2895 a command look up to see if a post-hook is registered. For that, it
2896 uses a now stale pointer that used to point to the define command, in
2897 the static buffer, causing a use-after-free. Note that the pointer in
2898 execute_command points to the dynamically-allocated buffer help by the
2899 static buffer in command_line_input, not to the static object itself,
2900 hence why we see a use-after-free.
2902 Fix that by removing the static buffer. I initially changed
2903 command_line_input and other related functions to return an std::string,
2904 which is the obvious but naive solution. The thing is that some callees
2905 don't need to return an allocated string, so this this an unnecessary
2906 pessimization. I changed it to passing in a reference to an std::string
2907 buffer, which the callee can use if it needs to return
2908 dynamically-allocated content. It fills the buffer and returns a
2909 pointers to the C string inside. The callees that don't need to return
2910 dynamically-allocated content simply don't use it.
2912 So, it started with modifying command_line_input as described above, all
2913 the other changes derive directly from that.
2915 One slightly shady thing is in handle_line_of_input, where we now pass a
2916 pointer to an std::string's internal buffer to readline's history_value
2917 function, which takes a `char *`. I'm pretty sure that this function
2918 does not modify the input string, because I was able to change it (with
2919 enough massaging) to take a `const char *`.
2921 A subtle change is that we now clear a UI's line buffer using a
2922 SCOPE_EXIT in command_line_handler, after executing the command.
2923 This was previously done by this line in handle_line_of_input:
2925 /* We have a complete command line now. Prepare for the next
2926 command, but leave ownership of memory to the buffer . */
2927 cmd_line_buffer->used_size = 0;
2929 I think the new way is clearer.
2931 [1] https://inbox.sourceware.org/gdb-patches/becb8438-81ef-8ad8-cc42-fcbfaea8cddd@simark.ca/
2932 [2] https://inbox.sourceware.org/gdb-patches/20221213112241.621889-1-jan.vrany@labware.com/
2934 Change-Id: I8fc89b1c69870c7fc7ad9c1705724bd493596300
2935 Reviewed-By: Tom Tromey <tom@tromey.com>
2937 2022-12-16 GDB Administrator <gdbadmin@sourceware.org>
2939 Automatic date update in version.in
2941 2022-12-15 Indu Bhagat <indu.bhagat@oracle.com>
2943 libsframe asan: avoid generating misaligned loads
2944 There are two places where unaligned loads were seen on aarch64:
2945 - #1. access to the SFrame FRE stack offsets in the in-memory
2946 representation/abstraction provided by libsframe.
2947 - #2. access to the SFrame FRE start address in the on-disk representation
2948 of the frame row entry.
2950 For #1, we can fix this by reordering the struct members of
2951 sframe_frame_row_entry in libsframe/sframe-api.h.
2953 For #2, we need to default to using memcpy instead, and copy out the bytes
2954 to a location for output.
2956 SFrame format is an unaligned on-disk format. As such, there are other blobs
2957 of memory in the on-disk SFrame FRE that are on not on their natural
2958 boundaries. But that does not pose further problems yet, because the users
2959 are provided access to the on-disk SFrame FRE data via libsframe's
2960 sframe_frame_row_entry, the latter has its' struct members aligned on their
2961 respective natural boundaries (and initialized using memcpy).
2963 PR 29856 libsframe asan: load misaligned at sframe.c:516
2968 * bfd/elf64-x86-64.c: Adjust as the struct members have been
2970 * libsframe/sframe.c (sframe_decode_fre_start_address): Use
2971 memcpy to perform 16-bit/32-bit reads.
2972 * libsframe/testsuite/libsframe.encode/encode-1.c: Adjust as the
2973 struct members have been reordered.
2978 * sframe-api.h: Reorder fre_offsets for natural alignment.
2980 2022-12-15 Simon Marchi <simon.marchi@polymtl.ca>
2982 gdb/testsuite: don't delete command files in gdb.base/commands.exp
2983 Don't delete the runtime-generated command files. This makes it easier
2984 to reproduce tests by hand.
2986 Change-Id: I4e53484eea216512f1c5d7dfcb5c464b36950946
2987 Approved-By: Tom Tromey <tom@tromey.com>
2989 2022-12-15 Tom Tromey <tromey@adacore.com>
2991 Move streq and compare_cstrings to gdbsupport
2992 It seems to me that streq and compare_cstrings belong near the other
2993 string utility functions in common-utils.h; and furthermore that streq
2994 ought to be inlined. This patch makes this change.
2996 Approved-By: Simon Marchi <simon.marchi@efficios.com>
2998 2022-12-15 Tom Tromey <tromey@adacore.com>
3000 Remove subset_compare
3001 I stumbled across subset_compare today, and after looking at the
3002 callers I realized it could be removed and replaced with calls to
3005 Approved-By: Simon Marchi <simon.marchi@efficios.com>
3007 2022-12-15 Andrew Burgess <aburgess@redhat.com>
3009 gdb: use gdb_assert not internal_error
3010 Spotted a couple of places in findvar.c where we use:
3013 internal_error ("...");
3015 this commit changes these to be:
3017 gdb_assert ( CONDITION );
3019 which I think is better.
3021 Unless we happen to hit the internal_error calls (which was bad) there
3022 should be no user visible changes after this commit.
3024 2022-12-15 Andrew Burgess <aburgess@redhat.com>
3026 gdb: some int to bool conversion in remote-sim.c
3027 Some obvious int to bool conversion in remote-sim.c, there should be
3028 no user visible changes after this commit.
3030 2022-12-15 Andrew Burgess <aburgess@redhat.com>
3032 gdb: make more use of make_target_connection_string
3033 I noticed that we have a function make_target_connection_string which
3034 wraps all the logic for creating a string that describes a target
3035 connection - but in some places we are not calling this function,
3036 instead we duplicate the function's logic.
3038 This commit cleans this up, and calls make_target_connection_string
3041 There should be no user visible changes after this commit.
3043 2022-12-15 Andrew Burgess <aburgess@redhat.com>
3045 gdb: int to bool conversion in tracefile.c
3046 Some obvious int to bool conversion in tracefile.c.
3048 Should be no user visible changes after this commit.
3050 2022-12-15 Tom de Vries <tdevries@suse.de>
3052 [gdb/testsuite] Fix gdb.base/condbreak-multi-context.exp with gcc 4.8.5
3053 With gcc 4.8.5, I run into:
3055 Running gdb.base/condbreak-multi-context.exp ...
3056 gdb compile failed, condbreak-multi-context.cc:21:11: warning: non-static \
3057 data member initializers only available with -std=c++11 or -std=gnu++11 \
3058 [enabled by default]
3063 Fix this by making it a static const.
3065 Tested on x86_64-linux, with gcc 4.8.5, 7.5.0 and clang 13.0.1.
3067 2022-12-15 GDB Administrator <gdbadmin@sourceware.org>
3069 Automatic date update in version.in
3071 2022-12-14 Andrew Burgess <aburgess@redhat.com>
3073 gdb/maint: add core file name to 'maint info program-spaces' output
3074 Each program space can have an associated core file. Include this
3075 information in the output of 'maint info program-spaces'.
3077 2022-12-14 Andrew Burgess <aburgess@redhat.com>
3079 gdb: ensure all targets are popped before an inferior is destructed
3080 Now that the inferiors target_stack automatically manages target
3081 reference counts, we might think that we don't need to unpush targets
3082 when an inferior is deleted...
3084 ...unfortunately that is not the case. The inferior::unpush function
3085 can do some work depending on the type of target, so it is important
3086 that we still pass through this function.
3088 To ensure that this is the case, in this commit I've added an assert
3089 to inferior::~inferior that ensures the inferior's target_stack is
3090 empty (except for the ever present dummy_target).
3092 I've then added a pop_all_targets call to delete_inferior, otherwise
3093 the new assert will fire in, e.g. the gdb.python/py-inferior.exp test.
3095 2022-12-14 Andrew Burgess <aburgess@redhat.com>
3097 gdb: remove the pop_all_targets (and friends) global functions
3098 This commit removes the global functions pop_all_targets,
3099 pop_all_targets_above, and pop_all_targets_at_and_above, and makes
3100 them methods on the inferior class.
3102 As the pop_all_targets functions will unpush each target, which
3103 decrements the targets reference count, it is possible that the target
3106 Right now, closing a target, in some cases, depends on the current
3107 inferior being set correctly, that is, to the inferior from which the
3110 To facilitate this I have used switch_to_inferior_no_thread within the
3111 new methods. Previously it was the responsibility of the caller to
3112 ensure that the correct inferior was selected.
3114 In a couple of places (event-top.c and top.c) I have been able to
3115 remove a previous switch_to_inferior_no_thread call.
3117 In remote_unpush_target (remote.c) I have left the
3118 switch_to_inferior_no_thread call as it is required for the
3119 generic_mourn_inferior call.
3121 2022-12-14 Andrew Burgess <aburgess@redhat.com>
3123 gdb: remove decref_target
3124 The decref_target function is not really needed. Calling
3125 target_ops::decref will just redirect to decref_target anyway, so why
3126 not just rename decref_target to target_ops::decref?
3128 That's what this commit does.
3130 It's not exactly renaming to target_ops::decref, because the decref
3131 functionality is handled by a policy class, so the new name is now
3132 target_ops_ref_policy::decref.
3134 There should be no user visible change after this commit.
3136 2022-12-14 Andrew Burgess <aburgess@redhat.com>
3138 gdb: have target_stack automate reference count handling
3139 This commit changes the target_stack class from using a C style array
3140 of 'target_ops *' to using a C++ std::array<target_ops_ref, ...>. The
3141 benefit of this change is that some of the reference counting of
3142 target_ops objects is now done automatically.
3144 This commit fixes a crash in gdb.python/py-inferior.exp where GDB
3145 crashes at exit, leaving a core file behind.
3147 The crash occurs in connpy_connection_dealloc, and is actually
3148 triggered by this assert:
3150 gdb_assert (conn_obj->target == nullptr);
3152 Now a little aside...
3154 ... the assert is never actually printed, instead GDB crashes due
3155 to calling a pure virtual function. The backtrace at the point of
3156 crash looks like this:
3158 #7 0x00007fef7e2cf747 in std::terminate() () from /lib64/libstdc++.so.6
3159 #8 0x00007fef7e2d0515 in __cxa_pure_virtual () from /lib64/libstdc++.so.6
3160 #9 0x0000000000de334d in target_stack::find_beneath (this=0x4934d78, t=0x2bda270 <the_dummy_target>) at ../../s>
3161 #10 0x0000000000df4380 in inferior::find_target_beneath (this=0x4934b50, t=0x2bda270 <the_dummy_target>) at ../.>
3162 #11 0x0000000000de2381 in target_ops::beneath (this=0x2bda270 <the_dummy_target>) at ../../src/gdb/target.c:3047
3163 #12 0x0000000000de68aa in target_ops::supports_terminal_ours (this=0x2bda270 <the_dummy_target>) at ../../src/gd>
3164 #13 0x0000000000dde6b9 in target_supports_terminal_ours () at ../../src/gdb/target.c:1112
3165 #14 0x0000000000ee55f1 in internal_vproblem(internal_problem *, const char *, int, const char *, typedef __va_li>
3167 Notice in frame #12 we called target_ops::supports_terminal_ours,
3168 however, this is the_dummy_target, which is of type dummy_target,
3169 and so we should have called dummy_target::supports_terminal_ours.
3170 I believe the reason we ended up in the wrong implementation of
3171 supports_terminal_ours (which is a virtual function) is because we
3172 made the call during GDB's shut-down, and, I suspect, the vtables
3173 were in a weird state.
3175 Anyway, the point of this patch is not to fix GDB's ability to
3176 print an assert during exit, but to address the root cause of the
3177 assert. With that aside out of the way, we can return to the main
3180 Connections are represented in Python with gdb.TargetConnection
3181 objects (or its sub-classes). The assert in question confirms that
3182 when a gdb.TargetConnection is deallocated, the underlying GDB
3183 connection has itself been removed from GDB. If this is not true then
3184 we risk creating multiple different gdb.TargetConnection objects for
3185 the same connection, which would be bad.
3187 To ensure that we have one gdb.TargetConnection object for each
3188 connection, the all_connection_objects map exists, this maps the
3189 process_stratum_target object (the connection) to the
3190 gdb.TargetConnection object that represents the connection.
3192 When a connection is removed in GDB the connection_removed observer
3193 fires, which we catch with connpy_connection_removed, this function
3194 then sets conn_obj->target to nullptr, and removes the corresponding
3195 entry from the all_connection_objects map.
3197 The first issue here is that connpy_connection_dealloc is being called
3198 as part of GDB's exit code, which is run after the Python interpreter
3199 has been shut down. The connpy_connection_dealloc function is used to
3200 deallocate the gdb.TargetConnection Python object. Surely it is
3201 wrong for us to be deallocating Python objects after the interpreter
3204 The reason why connpy_connection_dealloc is called during GDB's exit
3205 is that the global all_connection_objects map is still holding a
3206 reference to the gdb.TargetConnection object. When the map is
3207 destroyed during GDB's exit, the gdb.TargetConnection objects within
3208 the map can finally be deallocated.
3210 The reason why all_connection_objects has contents when GDB exits, and
3211 the reason the assert fires, is that, when GDB exits, there are still
3212 some connections that have not yet been removed from GDB, that is,
3213 they have a non-zero reference count.
3215 If we take a look at quit_force (top.c) you can see that, for each
3216 inferior, we call pop_all_targets before we (later in the function)
3217 call do_final_cleanups. It is the do_final_cleanups call that is
3218 responsible for shutting down the Python interpreter. The
3219 pop_all_targets calls should, in theory, cause all the connections to
3220 be removed from GDB.
3222 That this isn't working indicates that some targets have a non-zero
3223 reference count even after this final pop_all_targets call, and
3224 indeed, when I debug GDB, that is what I see.
3226 I tracked the problem down to delete_inferior where we do some house
3227 keeping, and then delete the inferior object, which calls
3228 inferior::~inferior.
3230 In neither delete_inferior or inferior::~inferior do we call
3231 pop_all_targets, and it is this missing call that means we leak some
3232 references to the target_ops objects on the inferior's target_stack.
3234 In this commit I will provide a partial fix for the problem. I say
3235 partial fix, but this will actually be enough to resolve the crash.
3236 In a later commit I will provide the final part of the fix.
3238 As mentioned at the start of the commit message, this commit changes
3239 the m_stack in target_stack to hold target_ops_ref objects. This
3240 means that when inferior::~inferior is called, and m_stack is
3241 released, we automatically decrement the target_ops reference count.
3242 With this change in place we no longer leak any references, and now,
3243 in quit_force the final pop_all_targets calls will release the final
3244 references. This means that the targets will be correctly closed at
3245 this point, which means the connections will be removed from GDB and
3246 the Python objects deallocated before the Python interpreter shuts
3249 There's a slight oddity in target_stack::unpush, where we std::move
3250 the reference out of m_stack like this:
3252 auto ref = std::move (m_stack[stratum]);
3254 the `ref' isn't used explicitly, but it serves to hold the
3255 target_ops_ref until the end of the scope while allowing the m_stack
3256 entry to be reset back to nullptr. The alternative would be to
3257 directly set the m_stack entry to nullptr, like this:
3259 m_stack[stratum] = nullptr;
3261 The problem here is that when we set the m_stack entry to nullptr we
3262 first decrement the target_ops reference count, and then set the array
3265 If the decrement means that the target_ops object reaches a zero
3266 reference count then the target_ops object will be closed by calling
3267 target_close. In target_close we ensure that the target being closed
3268 is not in any inferiors target_stack.
3270 As we decrement before clearing, then this check in target_close will
3271 fail, and an assert will trigger.
3273 By using std::move to move the reference out of m_stack, this clears
3274 the m_stack entry, meaning the inferior no longer contains the
3275 target_ops in its target_stack. Now when the REF object goes out of
3276 scope and the reference count is decremented, target_close can run
3279 I've made use of the Python connection_removed listener API to add a
3280 test for this issue. The test installs a listener and then causes
3281 delete_inferior to be called, we can then see that the connection is
3282 then correctly removed (because the listener triggers).
3284 2022-12-14 Andrew Burgess <aburgess@redhat.com>
3286 gdb/remote: remove some manual reference count handling
3287 While working on some other target_ops reference count related code, I
3288 spotted that in remote.c we do some manual reference count handling,
3289 i.e. we call target_ops::incref and decref_target (which wraps
3290 target_ops::decref).
3292 I think it would be better to make use of gdb::ref_ptr to automate the
3293 reference count management.
3295 So, this commit updates scoped_mark_target_starting in two ways,
3296 first, I use gdb::ref_ptr to handle the reference counts. Then,
3297 instead of using the scoped_mark_target_starting constructor and
3298 destructor to set and reset the starting_up flag, I now use a
3299 scoped_restore_tmpl object to set and restore the flag.
3301 The above changes mean that the scoped_mark_target_starting destructor
3302 can be completely removed, and the constructor body is now empty.
3304 I've also fixed a typo in the class comment.
3306 The only change in behaviour after this commit is that previously we
3307 didn't care what the value of starting_up was, we just set it to true
3308 in the constructor and false in the destructor.
3310 Now, I assert that the flag is initially false, then set the flag true
3311 when the scoped_mark_target_starting is created.
3313 As the starting_up flag is initialized to false then, for the assert
3314 to fire, we would need to recursively enter
3315 remote_target::start_remote_1, which I don't think is something we
3316 should be doing, so I think the new assert is an improvement.
3318 2022-12-14 Alan Modra <amodra@gmail.com>
3320 Re: ld, gold: remove support for -z bndplt (MPX prefix)
3321 Don't attempt to run gold tests with -z bndplt
3323 * testsuite/Makefile.am (exception_x86_64_bnd_test, bnd_plt_1.sh),
3324 (bnd_ifunc_1.sh, bnd_ifunc_2.sh): Delete rules.
3325 * testsuite/Makefile.in: Regenerate.
3326 * testsuite/bnd_ifunc_1.s: Delete.
3327 * testsuite/bnd_ifunc_1.sh: Delete.
3328 * testsuite/bnd_ifunc_2.s: Delete.
3329 * testsuite/bnd_ifunc_2.sh: Delete.
3330 * testsuite/bnd_plt_1.s: Delete.
3331 * testsuite/bnd_plt_1.sh: Delete.
3333 2022-12-14 Alan Modra <amodra@gmail.com>
3335 asan: buffer overflow in sh_reloc
3336 * coff-sh.c (sh_reloc): Use bfd_reloc_offset_in_range.
3338 2022-12-14 Alan Modra <amodra@gmail.com>
3340 Fix haiku ld dependencies
3341 I noticed after commit 8ad93045ed, "ld, gold: remove support for -z
3342 bndplt (MPX prefix)", that some of my builds were failing with
3344 eelf_x86_64_haiku.c:650:9: error: no member named 'bndplt' in 'struct elf_linker_x86_params'
3345 params.bndplt = true;
3348 * emulparams/aarch64haiku.sh: Use "source_sh" rather than ".".
3349 * emulparams/armelf_haiku.sh: Likewise.
3350 * emulparams/elf32ppchaiku.sh: Likewise.
3351 * emulparams/elf_mipsel_haiku.sh: Likewise.
3352 * emulparams/elf_x86_64_haiku.sh: Likewise.
3354 2022-12-14 Andrew Burgess <aburgess@redhat.com>
3356 gdb: add SYMBOL_LOOKUP_SCOPED_DEBUG_ENTER_EXIT
3357 After the previous commit converted symbol-lookup debug to use the new
3358 debug scheme, this commit adds SYMBOL_LOOKUP_SCOPED_DEBUG_ENTER_EXIT.
3360 The previous commit didn't add SYMBOL_LOOKUP_SCOPED_DEBUG_ENTER_EXIT
3361 because symbol-lookup debug is controlled by an 'unsigned int' rather
3362 than a 'bool' control variable, we use the numeric value to offer
3363 different levels of verbosity for symbol-lookup debug.
3365 The *_SCOPED_DEBUG_ENTER_EXIT mechanism currently relies on capturing
3366 a reference to the bool control variable, and evaluating the variable
3367 both on entry, and at exit, this is done in the scoped_debug_start_end
3368 class (see gdbsupport/common-debug.h).
3370 This commit templates scoped_debug_start_end so that the class can
3371 accept either a 'bool &' or an invokable object, e.g. a lambda
3372 function, or a function pointer.
3374 The existing scoped_debug_start_end and scoped_debug_enter_exit macros
3375 in common-debug.h are updated to support scoped_debug_enter_exit being
3376 templated, however, nothing outside of common-debug.h needs to change.
3378 I've then added SYMBOL_LOOKUP_SCOPED_DEBUG_ENTER_EXIT in symtab.h, and
3379 added a couple of token uses in symtab.c. I didn't want to add too
3380 much in this first commit, this is really about updating
3381 common-debug.h to support this new functionality.
3383 Within symtab.h I created a couple of global functions that can be
3384 used to query the status of the symbol_lookup_debug control variable,
3385 these functions are then used within the two existing macros:
3387 symbol_lookup_debug_printf
3388 symbol_lookup_debug_printf_v
3390 and also in the new SYMBOL_LOOKUP_SCOPED_DEBUG_ENTER_EXIT macro.
3392 2022-12-14 Andrew Burgess <aburgess@redhat.com>
3394 gdb: convert 'set debug symbol-lookup' to new debug printing scheme
3395 Convert the implementation of 'set debug symbol-lookup' to the new
3396 debug printing scheme.
3398 In a few places I've updated the debug output to remove places where
3399 the printed debug message included the function name, the new debug
3400 scheme already adds that, but I haven't done all the possible updates.
3402 2022-12-14 Andrew Burgess <aburgess@redhat.com>
3404 gdb/testsuite: new test for recent dwarf reader issue
3405 This commit provides a test for this commit:
3407 commit 55fc1623f942fba10362cb199f9356d75ca5835b
3408 Date: Thu Nov 3 13:49:17 2022 -0600
3410 Add name canonicalization for C
3412 Which resolves PR gdb/29105. My reason for writing this test was a
3413 desire to better understand the above commit, my process was to study
3414 the commit until I thought I understood it, then write a test to
3415 expose the issue. As the original commit didn't have a test, I
3416 thought it wouldn't hurt to commit this upstream.
3418 The problem tested for here is already described in the above commit,
3419 but I'll give a brief description here. This description describes
3420 GDB prior to the above commit:
3422 - Builtin types are added to GDB using their canonical name,
3423 e.g. "short", not "signed short",
3425 - When the user does something like 'p sizeof(short)', then this is
3426 handled in c-exp.y, and results in a call to lookup_signed_type
3427 for the name "int". The "int" here is actually being looked up as
3428 the type for the result of the 'sizeof' expression,
3430 - In lookup_signed_type GDB first adds a 'signed' and looks for that
3431 type, so in this case 'signed int', and, if that lookup fails, GDB
3432 then looks up 'int',
3434 - The problem is that 'signed int' is not the canonical name for a
3435 signed int, so no builtin type with that name will be found, GDB
3436 will then go to each object file in turn looking for a matching
3439 - When checking each object file, GDB will first check the partial
3440 symtab to see if the full symtab should be expanded or not.
3441 Remember, at this point GDB is looking for 'signed int', there
3442 will be no partial symbols with that name, so GDB will not expand
3445 - However, GDB checks each partial symbol using multiple languages,
3446 not just the current language (C in this case), so, when GDB
3447 checks using the C++ language, the symbol name is first
3448 canonicalized (the code that does this can be found
3449 lookup_name_info::language_lookup_name). As the canonical form of
3450 'signed int' is just 'int', GDB then looks for any symbols with
3451 the name 'int', most partial symtabs will contain such a symbol,
3452 so GDB ends up expanding pretty much every symtab.
3454 The above commit fixes this by avoiding the use of non-canonical names
3455 with C, now the initial builtin type lookup will succeed, and GDB
3456 never even considers whether to expand any additional symtabs.
3458 The test case creates a library that includes char, short, int, and
3459 long types, and a test program that links against the library.
3461 In the test script we start the inferior, but don't allow it to
3462 progress far enough that the debug information for the library has
3463 been fully expanded yet.
3465 Then we evaluate some 'sizeof(TYPE)' expressions.
3467 In the buggy version of GDB this would cause the debug information
3468 for the library to be fully expanded, while in the fixed version of
3469 GDB this will not be the case.
3471 We use 'info sources' to determine if the debug information has been
3472 fully expanded or not.
3474 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29105
3476 2022-12-14 Andrew Burgess <aburgess@redhat.com>
3478 gdb/testsuite: fix readnow detection
3479 The following commit broke the readnow detection in the testsuite:
3481 commit dfaa040b440084dd73ebd359326752d5f44fc02c
3482 Date: Mon Mar 29 18:31:31 2021 -0600
3484 Remove some "OBJF_READNOW" code from dwarf2_debug_names_index
3486 The testsuite checks if GDB was started with the -readnow flag by
3487 using the 'maintenance print objfiles' command, and looking for the
3488 string 'faked for "readnow"' in the output. This is implemented in
3489 two helper procs `readnow` (gdb.exp) and `mi_readnow` (mi-support.exp).
3491 The following tests all currently depend on this detection:
3495 gdb.dwarf2/debug-aranges-duplicate-offset-warning.exp
3496 gdb.dwarf2/dw2-stack-boundary.exp
3497 gdb.dwarf2/dw2-zero-range.exp
3498 gdb.dwarf2/gdb-index-nodebug.exp
3499 gdb.mi/mi-info-sources.exp
3500 gdb.python/py-symbol.exp
3503 The following test also includes detection of 'readnow', but does the
3504 detection itself by checking $::GDBFLAGS for the readnow flag:
3506 gdb.opt/break-on-_exit.exp
3508 The above commit removed from GDB the code that produced the 'faked
3509 for "readnow"' string, as a consequence the testsuite can no longer
3510 correctly spot when readnow is in use, and many of the above tests
3511 will fail (at least partially).
3513 When looking at the above tests, I noticed that gdb.rust/traits.exp
3514 does call `readnow`, but doesn't actually use the result, so I've
3515 removed the readnow call, this simplifies the next part of this patch
3516 as gdb.rust/traits.exp was the only place an extra regexp was passed
3517 to the readnow call.
3519 Next I have rewritten `readnow` to check the $GDBFLAGS for the
3520 -readnow flag, and removed the `maintenance print objfiles` check. At
3521 least for all the tests above, when using the readnow board, this is
3522 good enough to get everything passing again.
3524 For the `mi_readnow` proc, I changed this to just call `readnow` from
3525 gdb.exp, I left the mi_readnow name in place - in the future it might
3526 be the case that we want to do some different checks here.
3528 Finally, I updated gdb.opt/break-on-_exit.exp to call the `readnow`
3531 With these changes, all of the tests listed above now pass correctly
3532 when using the readnow board.
3534 2022-12-14 Li Xu <xuli1@eswincomputing.com>
3536 RISC-V: Add string length check for operands in AS
3537 The current AS accepts invalid operands due to miss of operands length check.
3538 For example, "e6" is an invalid operand in (vsetvli a0, a1, e6, mf8, tu, ma),
3539 but it's still accepted by assembler. In detail, the condition check "strncmp
3540 (array[i], *s, len) == 0" in arg_lookup function passes with "strncmp ("e64",
3541 "e6", 2)" in the case above. So the generated encoding is same as that of
3542 (vsetvli a0, a1, e64, mf8, tu, ma).
3544 This patch fixes issue above by prompting an error in such case and also adds
3549 * config/tc-riscv.c (arg_lookup): Add string length check for operands.
3550 * testsuite/gas/riscv/vector-insns-fail-vsew.d: New testcase for an illegal vsew.
3551 * testsuite/gas/riscv/vector-insns-fail-vsew.l: Likewise.
3552 * testsuite/gas/riscv/vector-insns-fail-vsew.s: Likewise.
3554 2022-12-14 Jan Beulich <jbeulich@suse.com>
3556 x86: adjust type checking constructs
3557 As Alan points out, ASAN takes issue with these constructs, for
3558 current_templates being NULL. Wrap them in sizeof(), so the expressions
3559 aren't actually evaluated.
3561 2022-12-14 Martin Liska <mliska@suse.cz>
3563 ld, gold: remove support for -z bndplt (MPX prefix)
3566 * elf-linker-x86.h (struct elf_linker_x86_params): Remove
3568 * elf64-x86-64.c (elf_x86_64_scan_relocs): Ignore
3570 (elf_x86_64_relocate_section): Similarly here.
3571 (elf_x86_64_link_setup_gnu_properties): Ignore bndplt.
3572 * elfxx-x86.c: Likewise.
3573 * elfxx-x86.h: Likewise.
3577 * NEWS: Document -z bndplt.
3578 * options.h (class General_options): Remove bndplt option.
3579 * x86_64.cc (class Output_data_plt_x86_64_bnd): Remove.
3580 (Target_x86_64::do_make_data_plt): Do not use
3581 Output_data_plt_x86_64_bnd.
3582 (Target_x86_64::Scan::get_reference_flags): Likewise.
3583 (Target_x86_64::Scan::check_non_pic): Likewise.
3584 (Target_x86_64::Scan::local): Likewise.
3585 (Target_x86_64::Scan::global): Likewise.
3589 * NEWS: Document -z bndplt.
3590 * emulparams/elf_x86_64.sh: Remove bndplt option.
3591 * ld.texi: Likewise.
3592 * testsuite/ld-x86-64/x86-64.exp:
3593 * testsuite/ld-x86-64/bnd-branch-1-now.d: Removed.
3594 * testsuite/ld-x86-64/bnd-branch-1.d: Removed.
3595 * testsuite/ld-x86-64/bnd-branch-1.s: Removed.
3596 * testsuite/ld-x86-64/bnd-ifunc-1-now.d: Removed.
3597 * testsuite/ld-x86-64/bnd-ifunc-1.d: Removed.
3598 * testsuite/ld-x86-64/bnd-ifunc-1.s: Removed.
3599 * testsuite/ld-x86-64/bnd-ifunc-2-now.d: Removed.
3600 * testsuite/ld-x86-64/bnd-ifunc-2.d: Removed.
3601 * testsuite/ld-x86-64/bnd-ifunc-2.s: Removed.
3602 * testsuite/ld-x86-64/bnd-plt-1-now.d: Removed.
3603 * testsuite/ld-x86-64/bnd-plt-1.d: Removed.
3604 * testsuite/ld-x86-64/mpx.exp: Removed.
3605 * testsuite/ld-x86-64/mpx1.out: Removed.
3606 * testsuite/ld-x86-64/mpx1a.c: Removed.
3607 * testsuite/ld-x86-64/mpx1a.rd: Removed.
3608 * testsuite/ld-x86-64/mpx1b.c: Removed.
3609 * testsuite/ld-x86-64/mpx1c.c: Removed.
3610 * testsuite/ld-x86-64/mpx1c.rd: Removed.
3611 * testsuite/ld-x86-64/mpx2.out: Removed.
3612 * testsuite/ld-x86-64/mpx2a.c: Removed.
3613 * testsuite/ld-x86-64/mpx2a.rd: Removed.
3614 * testsuite/ld-x86-64/mpx2b.c: Removed.
3615 * testsuite/ld-x86-64/mpx2c.c: Removed.
3616 * testsuite/ld-x86-64/mpx2c.rd: Removed.
3617 * testsuite/ld-x86-64/mpx3.dd: Removed.
3618 * testsuite/ld-x86-64/mpx3a.s: Removed.
3619 * testsuite/ld-x86-64/mpx3b.s: Removed.
3620 * testsuite/ld-x86-64/mpx3n.dd: Removed.
3621 * testsuite/ld-x86-64/mpx4.dd: Removed.
3622 * testsuite/ld-x86-64/mpx4a.s: Removed.
3623 * testsuite/ld-x86-64/mpx4b.s: Removed.
3624 * testsuite/ld-x86-64/mpx4n.dd: Removed.
3625 * testsuite/ld-x86-64/pr20800a.S: Removed.
3626 * testsuite/ld-x86-64/pr20800b.S: Removed.
3627 * testsuite/ld-x86-64/pr21038a-now.d: Removed.
3628 * testsuite/ld-x86-64/pr21038a.d: Removed.
3629 * testsuite/ld-x86-64/pr21038a.s: Removed.
3630 * testsuite/ld-x86-64/pr21038b-now.d: Removed.
3631 * testsuite/ld-x86-64/pr21038b.d: Removed.
3632 * testsuite/ld-x86-64/pr21038b.s: Removed.
3633 * testsuite/ld-x86-64/pr21038c-now.d: Removed.
3634 * testsuite/ld-x86-64/pr21038c.d: Removed.
3635 * testsuite/ld-x86-64/pr21038c.s: Removed.
3637 2022-12-14 Alan Modra <amodra@gmail.com>
3639 asan: signed integer overflow in display_debug_frames
3640 * dwarf.c (struct Frame_Chunk): Make col_offset an int64_t.
3641 Adjust all places allocating col_offset and col_type to use
3642 the size of the array element rather than the size of a type.
3643 (frame_display_row): Adjust printing of col_offset.
3644 (display_debug_frames): Factor out multiplication by
3645 code_factor and data_factor. Avoid signed overflow. Use
3648 2022-12-14 Alan Modra <amodra@gmail.com>
3650 Don't access freed memory printing objcopy warning
3651 abfd->filename will be freed if bfd_close gets far enough to delete
3652 the bfd. It's possible to have an error from fclose at this point.
3654 * objcopy.c (copy_archive): Dup filename before closing bfd for
3655 potential use in bfd_nonfatal_message.
3657 2022-12-14 GDB Administrator <gdbadmin@sourceware.org>
3659 Automatic date update in version.in
3661 2022-12-13 Tom Tromey <tromey@adacore.com>
3663 Fix control-c handling on Windows
3664 As Hannes pointed out, the Windows target-async patches broke C-c
3665 handling there. Looking into this, I found a few oddities, fixed
3668 First, windows_nat_target::interrupt calls GenerateConsoleCtrlEvent.
3669 I think this event can be ignored by the inferior, so it's not a great
3670 way to interrupt. Instead, using DebugBreakProcess (or a more
3671 complicated thing for Wow64) seems better.
3673 Second, windows_nat_target did not implement the pass_ctrlc method.
3674 Implementing this lets us remove the special code to call
3675 SetConsoleCtrlHandler and instead integrate into gdb's approach to C-c
3676 handling. I believe that this should also fix the race that's
3677 described in the comment that's being removed.
3679 Initially, I thought a simpler version of this patch would work.
3680 However, I think what happens is that some other library (I'm not sure
3681 what) calls SetConsoleCtrlHandler while gdb is running, and this
3682 intercepts and handles C-c -- so that the gdb SIGINT handler is not
3683 called. C-break continues to work, presumably because whatever
3684 handler is installed ignores it.
3686 This patch works around this issue by ensuring that the gdb handler
3689 2022-12-13 Tom Tromey <tromey@adacore.com>
3691 Refactor code to check for terminal sharing
3692 This refactors the code to check for terminal sharing.
3693 is_gdb_terminal is exported, and sharing_input_terminal_1 is renamed,
3694 slightly refactored, and moved to posix-hdep.c. A new
3695 Windows-specific implementation of this function is added to
3698 MSDN has a warning about GetConsoleProcessList
3700 This API is not recommended and does not have a virtual terminal
3701 equivalent. [...] Applications remoting via cross-platform
3702 utilities and transports like SSH may not work as expected if
3705 However, we believe this isn't likely to be an issue for gdb.
3707 2022-12-13 Tom Tromey <tromey@adacore.com>
3709 Use gdb::optional for sigint_ours
3710 sigint_ours (and sigquit_ours) can be used without being set. Avoid
3711 this problem by changing them to gdb::optional and checking that they
3712 are in fact set before using the value.
3714 Rename install_sigint_handler
3715 A subsequent patch will introduce a global 'install_sigint_handler'
3716 function, so first rename the static one in extension.c.
3718 2022-12-13 Tom de Vries <tdevries@suse.de>
3720 [gdb/tdep] Fix s390_linux_nat_target::stopped_by_watchpoint
3721 On s390x-linux, I run into:
3725 breakpoint.c:5784: internal-error: bpstat_stop_status_nowatch: \
3726 Assertion `!target_stopped_by_watchpoint ()' failed.^M
3727 A problem internal to GDB has been detected,^M
3728 further debugging may prove unreliable.^M
3729 FAIL: gdb.threads/watchpoint-fork.exp: parent: singlethreaded: \
3730 breakpoint after the first fork (GDB internal error)
3733 What happens is the follow:
3734 - a watchpoint event triggers
3735 - the event is processed, s390_linux_nat_target::stopped_by_watchpoint is called and
3736 it returns true, as expected
3737 - the watchpoint event is reported by gdb, and gdb stops
3738 - we issue a continue command
3739 - a fork event triggers
3740 - the event is processed, and during processing that event
3741 s390_linux_nat_target::stopped_by_watchpoint is called again, and returns
3743 - the assertion fails, because the function is expected to return false
3745 The function s390_linux_nat_target::stopped_by_watchpoint returns true the
3746 second time, because it looks at the exact same data that was looked at when
3747 it was called the first time, and that data hasn't changed.
3749 There's code in the same function that intends to prevent that from happening:
3751 /* Do not report this watchpoint again. */
3752 memset (&per_lowcore, 0, sizeof (per_lowcore));
3753 if (ptrace (PTRACE_POKEUSR_AREA, s390_inferior_tid (), &parea, 0) < 0)
3754 perror_with_name (_("Couldn't clear watchpoint status"));
3756 and that probably used to work for older kernels, but no longer does since
3757 linux kernel commit 5e9a26928f55 ("[S390] ptrace cleanup").
3759 Fix this by copying this:
3762 if (!linux_nat_get_siginfo (inferior_ptid, &siginfo))
3764 if (siginfo.si_signo != SIGTRAP
3765 || (siginfo.si_code & 0xffff) != TRAP_HWBKPT)
3768 from aarch64_linux_nat_target::stopped_data_address and remove the
3769 obsolete watchpoint status clearing code.
3771 Tested on s390x-linux.
3773 Approved-By: Ulrich Weigand <uweigand@de.ibm.com>
3775 2022-12-13 H.J. Lu <hjl.tools@gmail.com>
3777 gold: Remove BND from 64-bit x86-64 IBT PLT
3778 Since MPX support has been removed from x86-64 psABI, remove BND from
3779 64-bit IBT PLT by using 32-bit IBT PLT.
3782 * x86_64.cc (Output_data_plt_x86_64_ibt<32>::first_plt_entry):
3784 (Output_data_plt_x86_64_ibt<size>::first_plt_entry): This.
3785 (Output_data_plt_x86_64_ibt<64>::first_plt_entry): Removed.
3786 (Output_data_plt_x86_64_ibt<size>::do_fill_first_plt_entry):
3787 Drop the size == 32 check.
3788 (Output_data_plt_x86_64_ibt<32>::plt_entry): Renamed to ...
3789 (Output_data_plt_x86_64_ibt<size>::plt_entry): This.
3790 (Output_data_plt_x86_64_ibt<64>::plt_entry): Removed.
3791 (Output_data_plt_x86_64_ibt<32>::aplt_entry): Renamed to ...
3792 (Output_data_plt_x86_64_ibt<size>::aplt_entry): This.
3793 (Output_data_plt_x86_64_ibt<64>::aplt_entry): Removed.
3794 (Output_data_plt_x86_64_ibt<size>::do_fill_plt_entry): Drop the
3796 (Output_data_plt_x86_64_ibt<size>::fill_aplt_entry): Likewise.
3798 2022-12-13 Tom Tromey <tromey@adacore.com>
3800 Remove two unnecessary casts
3801 A couple of calls to parse_probe_linespec had an unnecessary cast. I
3802 suspect this cast was never needed, but once commands were changed to
3803 take a 'const' argument, they became completely obsolete. Tested by
3806 2022-12-13 Andrew Burgess <aburgess@redhat.com>
3808 gdb/testsuite: avoid creating temp file in gdb/testsuite/ directory
3811 commit 33c1395cf5e9deec7733691ba32c450e5c27f757
3812 Date: Fri Nov 11 15:26:46 2022 +0000
3814 gdb/testsuite: fix gdb.trace/unavailable-dwarf-piece.exp with Clang
3816 The gdb.trace/unavailable-dwarf-piece.exp test script was creating a
3817 temporary file in the build/gdb/testsuite/ directory, instead of in
3818 the expected place in the outputs directory.
3820 Fix this by adding a call to standard_output_file.
3822 2022-12-13 Tom de Vries <tdevries@suse.de>
3824 [gdb/testsuite] Fix gdb.python/py-disasm.exp on s390x
3825 On s390x-linux, I run into:
3827 (gdb) disassemble test^M
3828 Dump of assembler code for function test:^M
3829 0x0000000001000638 <+0>: stg %r11,88(%r15)^M
3830 0x000000000100063e <+6>: lgr %r11,%r15^M
3831 0x0000000001000642 <+10>: nop 0^M
3832 => 0x0000000001000646 <+14>: nop 0^M
3833 0x000000000100064a <+18>: nop 0^M
3834 0x000000000100064e <+22>: lhi %r1,0^M
3835 0x0000000001000652 <+26>: lgfr %r1,%r1^M
3836 0x0000000001000656 <+30>: lgr %r2,%r1^M
3837 0x000000000100065a <+34>: lg %r11,88(%r11)^M
3838 0x0000000001000660 <+40>: br %r14^M
3839 End of assembler dump.^M
3840 (gdb) FAIL: gdb.python/py-disasm.exp: global_disassembler=: disassemble test
3843 The problem is that the test-case expects "nop" but on s390x we have instead
3846 Fix this by allowing the insn.
3848 Tested on s390x-linux and x86_64-linux.
3850 2022-12-13 Jan Beulich <jbeulich@suse.com>
3852 gas: re-work line number tracking for macros and their expansions
3853 The PR gas/16908 workaround aimed at uniformly reporting line numbers
3854 to reference macro invocation sites. As mentioned in a comment this may
3855 be desirable for small macros, but often isn't for larger ones. As a
3856 first step improve diagnostics to report both locations, while aiming at
3857 leaving generated debug info unaltered.
3859 Note that macro invocation context is lost for any diagnostics issued
3860 only after all input was processed (or more generally for any use of
3861 as_*_where(), as the functions can't know whether the passed in location
3862 is related to [part of] the present stack of locations). To maintain the
3863 intended workaround behavior for PR gas/16908, a new as_where() is
3864 introduced to "look through" macro invocations, while the existing
3865 as_where() is renamed (and used in only very few places for now). Down
3866 the road as_where() will likely want to return a list of (file,line)
3869 2022-12-13 Jan Beulich <jbeulich@suse.com>
3871 Arm: avoid unhelpful uses of .macro in testsuite
3872 Macros with just a single use site are a little pointless to have, and
3873 even in further cases .irp is more suitable for the purpose. Expand such
3874 inline, avoiding the need to touch the testcases when diagnostics are
3875 changed for code resulting from macro expansion.
3877 While there also make what was "iter_mla" in sp-usage-thumb2-relax cover
3878 smlatt as well, rather than testing smlabt twice.
3880 2022-12-13 Tom Tromey <tom@tromey.com>
3882 Fix crash in is_nocall_function
3883 is_nocall_function anticipates only being called for a function or a
3884 method. However, PR gdb/29871 points out a situation where an unusual
3885 expression -- but one that parses to a valid, if extremely weird,
3886 function call -- breaks this assumption.
3888 This patch changes is_nocall_function to remove this assert and
3889 instead simply return 'false' in this case.
3891 Approved-By: Simon Marchi <simon.marchi@efficios.com>
3892 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29871
3894 2022-12-13 Johnson Sun <j3.soon777@gmail.com>
3896 Replace gdbpy_should_stop with gdbpy_breakpoint_cond_says_stop
3897 In 2014, the function `gdbpy_should_stop' has been replaced with
3898 `gdbpy_breakpoint_cond_says_stop'
3900 This replaces `gdbpy_should_stop' with `gdbpy_breakpoint_cond_says_stop' in the
3903 Since `gdbpy_should_stop' has been renamed as noted in `gdb/ChangeLog-2014':
3905 * python/py-breakpoint.c (gdbpy_breakpoint_cond_says_stop): Renamed
3906 from gdbpy_should_stop. Change result type to enum scr_bp_stop.
3908 Change-Id: I0ef3491ce5e057c5e75ef8b569803b30a5838575
3909 Approved-By: Simon Marchi <simon.marchi@efficios.com>
3911 2022-12-13 Alan Modra <amodra@gmail.com>
3913 asan: mips_hi16_list segfault in bfd_get_section_limit_octets
3914 static variables like mips_hi16_list are nasty for applications using
3915 bfd. It is possible when opening and closing bfds with mis-matched
3916 hi/lo relocs to leave a stale section pointer on the list. That can
3917 cause a segfault if multiple bfds are being processed.
3919 Tidying the list when closing is sufficient to stop this happening
3920 (and fixes small memory leaks). This patch goes further and moves
3921 mips_hi16_list to where it belongs in the bfd tdata.
3923 * elf32-mips.c (bfd_elf32_close_and_cleanup(: Define.
3924 * elf64-mips.c (bfd_elf64_close_and_cleanup): Define.
3925 * elfn32-mips.c (bfd_elf32_close_and_cleanup(: Define.
3926 * elfxx-mips.c (struct mips_hi16): Move earlier.
3927 (mips_hi16_list): Move to..
3928 (struct mips_elf_obj_tdata): ..here.
3929 (_bfd_mips_elf_close_and_cleanup): New function.
3930 (_bfd_mips_elf_hi16_reloc, _bfd_mips_elf_lo16_reloc),
3931 (_bfd_elf_mips_get_relocated_section_contents): Adjust uses of
3933 * elfxx-mips.h (_bfd_mips_elf_close_and_cleanup): Declare.
3935 2022-12-13 GDB Administrator <gdbadmin@sourceware.org>
3937 Automatic date update in version.in
3939 2022-12-12 Indu Bhagat <indu.bhagat@oracle.com>
3941 libctf: remove unnecessary zstd constructs
3942 This patch is essentially a revert of
3943 commit-id: 8818c80cbd4116ef5af171ec47c61167179e225c
3944 (libctf: Add ZSTD_LIBS to LIBS so that ac_cv_libctf_bfd_elf can be true)
3946 As the specific configure check now uses libtool, this explicit mention of the
3947 dependency $ZSTD_LIBS is not needed anymore.
3951 * libctf/Makefile.in: Regenerated.
3952 * libctf/aclocal.m4: Likewise.
3953 * libctf/config.h.in: Likewise.
3954 * libctf/configure: Likewise.
3955 * libctf/configure.ac: Remove ZSTD_LIBS from LIBS. Cleanup
3958 2022-12-12 Indu Bhagat <indu.bhagat@oracle.com>
3960 libctf: remove AC_CONFIG_MACRO_DIR
3961 ACLOCAL_AMFLAGS is being set already. So using AC_CONFIG_MACRO_DIR is
3966 * libctf/configure: Regenerated.
3967 * libctf/configure.ac: remove AC_CONFIG_MACRO_DIR usage.
3969 2022-12-12 Indu Bhagat <indu.bhagat@oracle.com>
3971 libctf: remove unnecessary zlib constructs
3972 This dependency is managed via libtool. So explicit addition to LDFLAGS
3973 and LIBS is not necessary anymore.
3977 * libctf/configure: Regenerated.
3978 * libctf/configure.ac: remove zlib from LDFLAGS and LIBS.
3980 2022-12-12 Tom de Vries <tdevries@suse.de>
3982 [gdb/testsuite] Fix PR20630 regression test in gdb.base/printcmds.exp
3983 On s390x-linux, I run into:
3985 (gdb) print {unsigned char}{65}^M
3987 (gdb) FAIL: gdb.base/printcmds.exp: print {unsigned char}{65}
3990 In contrast, on x86_64-linux, we have:
3992 (gdb) print {unsigned char}{65}^M
3994 (gdb) PASS: gdb.base/printcmds.exp: print {unsigned char}{65}
3997 The first problem here is that the test is supposed to be a regression test
3998 for PR20630, which can be reproduced (for an unfixed gdb) like this:
4000 (gdb) p {unsigned char[]}{0x17}
4001 gdbtypes.c:4641: internal-error: copy_type: \
4002 Assertion `TYPE_OBJFILE_OWNED (type)' failed.
4004 but it's not due to insufficient quoting (note the dropped '[]').
4006 That's easy to fix, but after that we have on s390 (big endian):
4008 (gdb) print {unsigned char[]}{65}^M
4011 and on x86_64 (little endian):
4013 (gdb) print {unsigned char[]}{65}^M
4017 Fix this by using 0xffffffff, such that in both cases we have:
4019 (gdb) print {unsigned char[]}{0xffffffff}^M
4020 $749 = "\377\377\377\377"^M
4023 Tested on x86_64-linux and s390x-linux.
4025 2022-12-12 Alan Modra <amodra@gmail.com>
4027 PR29893, buffer overflow in display_debug_addr
4029 * dwarf.c (display_debug_addr): Sanity check dwarf5 unit_length
4030 field. Don't read past end.
4032 2022-12-12 Tom Tromey <tom@tromey.com>
4034 Another Rust operator precedence bug
4035 My earlier patch to fix PR rust/29859 introduced a new operator
4036 precedence bug in the Rust parser. Assignment operators are
4037 right-associative in Rust. And, while this doesn't often matter, as
4038 Rust assignments always have the value (), still as a matter of
4039 principle we should get this correct.
4041 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29859
4043 2022-12-12 Tom de Vries <tdevries@suse.de>
4045 [gdb/testsuite] Fix gdb.base/write_mem.exp for big endian
4046 On s390x-linux (big endian), I run into:
4049 0x1000638 <main>: 0x0000^M
4050 (gdb) FAIL: gdb.base/write_mem.exp: x /xh main
4053 In contrast, on x86_64-linux (little endian), we have the expected:
4056 0x4004a7 <main>: 0x4242^M
4057 (gdb) PASS: gdb.base/write_mem.exp: x /xh main
4060 The problem is that the test-case hard-codes expectations about endiannes by
4061 writing an int-sized value (4 bytes in this case) and then printing only a
4062 halfword by using "/h" (so, two bytes).
4064 If we print 4 bytes, we have for s390x:
4066 0x1000638 <main>: 0x00004242^M
4070 0x4004a7 <main>: 0x00004242^M
4073 Fix this by removing the "/h".
4075 Tested on x86_64-linux and s390x-linux.
4077 2022-12-12 Jan Vrany <jan.vrany@labware.com>
4079 gdb: fix possible use-after-free when executing commands
4080 In principle, `execute_command()` does following:
4082 struct cmd_list_element *c;
4083 c = lookup_cmd ( ... );
4085 /* If this command has been pre-hooked, run the hook first. */
4086 execute_cmd_pre_hook (c);
4088 /* ...execute the command `c` ...*/
4090 execute_cmd_post_hook (c);
4092 This may lead into use-after-free error. Imagine the command
4093 being executed is a user-defined Python command that redefines
4094 itself. In that case, struct `cmd_list_element` pointed to by
4095 `c` is deallocated during its execution so it is no longer valid
4096 when post hook is executed.
4098 To fix this case, this commit looks up the command once again
4099 after it is executed to get pointer to (possibly newly allocated)
4102 2022-12-12 Jan Beulich <jbeulich@suse.com>
4104 x86: further re-work insn/suffix recognition to also cover MOVSX
4107 Having templates with a suffix explicitly present has always been
4108 quirky. After prior adjustment all that's left to also eliminate the
4109 anomaly from move-with-sign-extend is to consolidate the insn templates
4110 and to make may_need_pass2() cope (plus extend testsuite coverage).
4112 2022-12-12 Jan Beulich <jbeulich@suse.com>
4114 x86: drop (now) stray IsString
4115 The need for them on the operand-less string insns has gone away with
4116 the removal of maybe_adjust_templates() and associated logic. Since
4117 i386_index_check() needs adjustment then anyway, take the opportunity
4118 and also simplify it, possible again as a result of said removal (plus
4119 the opcode template adjustments done here).
4121 2022-12-12 Jan Beulich <jbeulich@suse.com>
4123 x86: move bad-use-of-TLS-reloc check
4124 Having it in match_template() is unhelpful. Neither does looking for the
4125 next template to possibly match make any sense in that case, nor is the
4126 resulting diagnostic making clear what the problem is.
4128 While moving the check, also generalize it to include all SIMD and VEX-
4129 encoded insns. This way an existing conditional can be re-used in
4130 md_assemble(). Note though that this still leaves a lof of insns which
4131 are also wrong to use with these relocations.
4133 Further fold the remaining check (BFD_RELOC_386_GOT32) with the XRELEASE
4134 related one a few lines down. This again allows re-using an existing
4137 2022-12-12 Jan Beulich <jbeulich@suse.com>
4139 x86-64: allow HLE store of accumulator to absolute 32-bit address
4140 In commit 1212781b35c9 ("ix86: allow HLE store of accumulator to
4141 absolute address") I was wrong to exclude 64-bit code. Dropping the
4142 check also leads to better diagnostics in 64-bit code ("MOV", after
4143 all, isn't invalid with "XRELEASE").
4145 While there also limit the amount of further checks done: The operand
4146 type checks that were there were effectively redundant with other ones
4147 anyway, plus it's quite fine to also have "xrelease mov <disp>, %eax"
4148 look for the next MOV template (in fact again also improving
4151 2022-12-12 Jan Beulich <jbeulich@suse.com>
4153 ix86: don't recognize/derive Q suffix in the common case
4154 Have its use, except where actually legitimate, result in the same "only
4155 supported in 64-bit mode" diagnostic as emitted for other 64-bit only
4156 insns. Also suppress deriving of the suffix in Intel mode except in the
4157 legitimate cases. This in exchange allows dropping the respective code
4158 from match_template().
4160 To maintain reasonable diagnostics (in particular to avoid "`mov' is
4161 only supported in 64-bit mode" on the SIMD forms of MOVQ) we need to
4162 defer parse_insn()'s emitting of errors unrelated to prefix parsing.
4163 Utilize i.error just like match_template() does.
4165 Oddly enough despite gcc's preference towards FILDQ and FIST{,T}Q we
4166 had no testcase whatsoever for these. Therefore such tests are being
4167 added. Note that the removed line in the x86-64-lfence-load testcase
4168 was redundant with the exact same one a few lines up.
4170 2022-12-12 Jan Beulich <jbeulich@suse.com>
4172 x86: re-work insn/suffix recognition
4173 Having templates with a suffix explicitly present has always been
4174 quirky. Introduce a 2nd matching pass in case the 1st one couldn't find
4175 a suitable template _and_ didn't itself already need to trim off a
4176 suffix to find a match at all. This requires error reporting adjustments
4177 (albeit luckily fewer than I was afraid might be necessary), as errors
4178 previously reported during matching now need deferring until after the
4179 2nd pass (because, obviously, we must not emit any error if the 2nd pass
4180 succeeds). While also related to PR gas/29524, it was requested that
4181 move-with-sign-extend be left as broken as it always was.
4184 Note that with the dropped CMPSD and MOVSD Intel Syntax string insn
4185 templates taking operands, mixed IsString/non-IsString template groups
4186 (with memory operands) cannot occur anymore. With that
4187 maybe_adjust_templates() becomes unnecessary (and is hence being
4191 Note further that while the additions to the intel16 testcase aren't
4192 really proper Intel syntax, we've been permitting all of those except
4193 for the MOVD variant. The test therefore is to avoid re-introducing such
4196 2022-12-12 Jan Beulich <jbeulich@suse.com>
4198 x86: constify parse_insn()'s input
4199 The function doesn't alter its input buffer: Reflect this in its
4200 prototype. To avoid using any kind of cast, simply calculate the update
4201 of "line" from the function's input and output.
4203 x86: revert disassembler parts of "x86: Allow 16-bit register source for LAR and LSL"
4204 This reverts the disassembler parts of 859aa2c86dc9 ("x86: Allow 16-bit
4205 register source for LAR and LSL"), adjusting testcases as necessary.
4206 That change was itself a partial revert of c9f5b96bdab0 ("x86: correct
4207 handling of LAR and LSL"), without actually saying so. While the earlier
4208 commit was properly agreed upon, the partial revert was not, and hence
4209 should not have been committed. This is even more so that the revert
4210 part of that change wasn't even necessary to address PR gas/29844.
4212 2022-12-12 Alan Modra <amodra@gmail.com>
4214 PR29892, Field file_table of struct module is uninitialized
4216 * vms-alphs.c (new_module): Use bfd_zmalloc to alloc file_table.
4217 (parse_module): Rewrite file_table reallocation code and clear.
4219 Lack of bounds checking in vms-alpha.c parse_module
4239 * vms-alpha.c (parse_module): Make length param bfd_size_type.
4240 Delete length == -1 checks. Sanity check record_length.
4241 Sanity check DST__K_MODBEG, DST__K_RTNBEG, DST__K_RTNEND lengths.
4242 Sanity check DST__K_SOURCE and DST__K_LINE_NUM elements
4244 (build_module_list): Pass dst_section size to parse_module.
4246 2022-12-12 Alan Modra <amodra@gmail.com>
4248 PR29872, uninitialised value in display_debug_lines_decoded dwarf.c:5413
4249 Plus segvs if the C-library doesn't handle printf %s of NULL.
4252 * dwarf.c (null_name): New function.
4253 (process_debug_info): Use it here..
4254 (display_debug_lines_raw): ..and here..
4255 (display_debug_lines_decoded): ..and here. xcalloc directory_table.
4256 Simplify xcalloc of file_table.
4258 2022-12-12 Jan Beulich <jbeulich@suse.com>
4260 gas/codeview: avoid "shadowing" of glibc function name
4261 While not "index" this time, old enough glibc also has an (unguarded)
4262 declaration of fileno() in stdio.h, which triggers a "shadows a global
4263 declaration" warning with our choice of warning level and with at least
4266 x86: generate template sets data at build time
4267 Speed up gas startup by avoiding runtime allocation of the instances of
4268 type "templates". At the same time cut the memory requirement to just
4269 very little over half (not even accounting for any overhead
4270 notes_alloc() may incur) by reusing the "end" slot of a preceding entry
4271 for the "start" slot of the subsequent one.
4273 x86: drop sentinel from i386_optab[]
4274 Now that the table is local to gas, ARRAY_SIZE() can be used to
4275 determine the end of the table. Re-arrange the processing loop in
4276 md_begin() accordingly, at the same time folding the two calls to
4277 notes_alloc() into just one.
4279 x86: add generated tables dependency check to gas
4280 As requested by H.J., just for the sake of people potentially building
4281 in gas/ alone, add a check that the generated files in opcodes/ are
4282 actually up-to-date. Personally I think this should at best be a
4283 warning, but I can see how this may not be easily noticable among other
4284 make output (depending in particular on the verbosity level).
4286 x86: break gas dependency on libopcodes
4287 gas doesn't use anything from libopcodes anymore - suppress linking in
4290 x86: remove i386-opc.c
4291 Remove the now empty i386-opc.c. To compensate, tie table generation in
4292 opcodes/ to the building of i386-dis.o, despite the file not really
4293 depending on the generated data.
4295 2022-12-12 Jan Beulich <jbeulich@suse.com>
4297 x86: instantiate i386_{op,reg}tab[] in gas instead of in libopcodes
4298 Unlike many other architectures, x86 does not share an opcode table
4299 between assembly and disassembly. Any consumer of libopcodes would only
4300 ever access one of the two. Since gas is the only consumer of the
4301 assembly data, move it there. While doing so mark respective entities
4302 "static" in i386-gen (we may want to do away with i386_regtab_size
4305 This also shrinks the number of relocations to be processed for
4306 libopcodes.so by about 30%.
4308 2022-12-12 GDB Administrator <gdbadmin@sourceware.org>
4310 Automatic date update in version.in
4312 2022-12-11 Alan Modra <amodra@gmail.com>
4314 PR29870, objdump SEGV in display_debug_lines_decoded dwarf.c:5524
4315 DWARF5 directory and file table allow more opportunity for fuzzers
4316 to break things. There are likely other places in dwarf.c that should
4320 * dwarf.c (display_debug_lines_decoded): Handle NULL file_table
4323 2022-12-11 GDB Administrator <gdbadmin@sourceware.org>
4325 Automatic date update in version.in
4327 2022-12-10 Tom de Vries <tdevries@suse.de>
4329 [gdb/tdep] Fix larl handling in s390_displaced_step_fixup
4330 On s390x-linux with target board unix/-m31, I run into:
4332 (gdb) PASS: gdb.guile/scm-lazy-string.exp: bad length
4334 $1 = 0x804006b0 <error: Cannot access memory at address 0x804006b0>^M
4335 (gdb) FAIL: gdb.guile/scm-lazy-string.exp: ptr: print ptr
4338 A minimal example is:
4340 $ gdb -q -batch -ex "set trace-commands on" -x gdb.in
4341 +file scm-lazy-string
4343 Breakpoint 1 at 0x4005d2: file scm-lazy-string.c, line 23.
4346 Breakpoint 1, main () at scm-lazy-string.c:23
4347 23 const char *ptr = "pointer";
4349 24 const char array[] = "array";
4351 $1 = 0x804006b0 <error: Cannot access memory at address 0x804006b0>
4354 If we delete the breakpoint after running to it, we have instead the expected:
4358 24 const char array[] = "array";
4360 $1 = 0x4006b0 "pointer"
4363 The problem is in displaced stepping, forced by the presence of the breakpoint,
4364 when stepping over this insn:
4366 0x4005d2 <main+10> larl %r1,0x4006b0
4369 With normal stepping we have:
4374 but with displaced stepping we have instead (note the 0x80000000 difference):
4381 The difference comes from this code in s390_displaced_step_fixup:
4383 /* Handle LOAD ADDRESS RELATIVE LONG. */
4384 else if (is_ril (insn, op1_larl, op2_larl, &r1, &i2))
4387 regcache_write_pc (regs, from + insnlen);
4388 /* Recompute output address in R1. */
4389 regcache_cooked_write_unsigned (regs, S390_R0_REGNUM + r1,
4390 amode | (from + i2 * 2));
4393 where the "amode |" adds the 0x80000000.
4395 Fix this by removing the "amode |".
4397 Tested on s390-linux, with native and target board unix/-m31.
4399 Approved-By: Ulrich Weigand <uweigand@de.ibm.com>
4401 2022-12-10 GDB Administrator <gdbadmin@sourceware.org>
4403 Automatic date update in version.in
4405 2022-12-09 Indu Bhagat <indu.bhagat@oracle.com>
4407 objdump: sframe: fix memory leaks
4410 * binutils/objdump.c (dump_section_sframe): free up contents and
4411 SFrame decoder context on exit.
4413 2022-12-09 Indu Bhagat <indu.bhagat@oracle.com>
4415 libsframe: rename API sframe_fde_func_info to sframe_fde_create_func_info
4416 The new name better reflects the purpose of the function.
4420 * bfd/elfxx-x86.c (_bfd_x86_elf_create_sframe_plt): Use new
4422 * libsframe/sframe.c (sframe_fde_create_func_info): Rename
4423 sframe_fde_func_info to this.
4424 * libsframe/testsuite/libsframe.encode/encode-1.c: Use new name.
4428 * sframe-api.h (sframe_fde_create_func_info): Rename
4429 sframe_fde_func_info to this.
4431 2022-12-09 Indu Bhagat <indu.bhagat@oracle.com>
4433 gas: sframe: fine tune the fragment fixup for SFrame func info
4434 SFrame function info is an unsigned 8-bit field comprising of the following
4437 - 1-bit: FRE start address encoding
4440 At the moment, the most-significat 4-bits are zero (The FRE start
4441 address encoding of SFRAME_FDE_TYPE_PCINC has a value of zero, and the upper
4442 3-bits are unused). So the current implementation works without this patch.
4444 To be precise, however, the fragment fixup logic is meant to fixup only the
4445 least-significant 4-bits (i.e., only the FRE type needs to be updated
4446 according to the function size).
4448 This patch makes the gas implementation a bit more resilient: In the
4449 future, when the format does evolve to make use of the currently unused
4450 3-bits in various ways, the values in those 3-bits can be propagated
4451 unchanged while the fragment fixup continues to update the lowermost
4452 4-bits to indicate the selected FRE type.
4456 * gas/gen-sframe.c (create_func_info_exp): New definition.
4457 (output_sframe_funcdesc): Call create_func_info_exp.
4458 * gas/sframe-opt.c (sframe_estimate_size_before_relax): The
4459 associated fragment uses O_modulus now.
4460 (sframe_convert_frag): Adjust the fragment fixup code according
4461 to the new composite exp.
4463 2022-12-09 Indu Bhagat <indu.bhagat@oracle.com>
4465 sframe: gas: libsframe: define constants and remove magic numbers
4466 Define constants in sframe.h for the various limits associated with the
4467 range of offsets that can be encoded in the start address of an SFrame
4468 FRE. E.g., sframe_frame_row_entry_addr1 is used when start address
4469 offset can be encoded as 1-byte unsigned value.
4471 Update the code in gas to use these defined constants as it checks for
4472 these limits, and remove the usage of magic numbers.
4476 * gas/sframe-opt.c (sframe_estimate_size_before_relax):
4477 (sframe_convert_frag): Do not use magic numbers.
4478 * libsframe/sframe.c (sframe_calc_fre_type): Likewise.
4482 * sframe.h (SFRAME_FRE_TYPE_ADDR1_LIMIT): New constant.
4483 (SFRAME_FRE_TYPE_ADDR2_LIMIT): Likewise.
4484 (SFRAME_FRE_TYPE_ADDR4_LIMIT): Likewise.
4486 2022-12-09 Indu Bhagat <indu.bhagat@oracle.com>
4488 sframe.h: make some macros more precise
4491 * sframe.h (SFRAME_V1_FUNC_INFO): Use specific bits only.
4492 (SFRAME_V1_FRE_INFO): Likewise.
4494 2022-12-09 Indu Bhagat <indu.bhagat@oracle.com>
4496 libsframe: minor formatting nits
4499 * libsframe/sframe.c: Fix formatting nits.
4501 2022-12-09 Luis Machado <luis.machado@arm.com>
4503 [aarch64] Add TPIDR2 register support for Linux
4504 With the AArch64 Scalable Matrix Extension we have a new TPIDR2 register, and
4505 it will be added to the existing NT_ARM_TLS register set. Kernel patches are
4506 being reviewed here:
4508 https://lore.kernel.org/linux-arm-kernel/20220818170111.351889-1-broonie@kernel.org/
4510 From GDB's perspective, we handle it in a similar way to the existing TPIDR
4511 register. But we need to consider cases of systems that only have TPIDR and
4512 systems that have both TPIDR and TPIDR2.
4514 With that in mind, the following patch adds the required code to support
4515 TPIDR2 and turns the org.gnu.gdb.aarch64.tls feature into a
4516 dynamically-generated target description as opposed to a static target
4517 description containing only TPIDR.
4519 That means we can remove the gdb/features/aarch64-tls.xml file and replace the
4520 existing gdb/features/aarch64-tls.c auto-generated file with a new file that
4521 dynamically generates the target description containing either TPIDR alone or
4524 In the future, when *BSD's start to support this register, they can just
4525 enable it as is being done for the AArch64 Linux target.
4527 The core file read/write code has been updated to support TPIDR2 as well.
4529 On GDBserver's side, there is a small change to the find_regno function to
4530 expose a non-throwing version of it.
4532 It always seemed strange to me how find_regno causes the whole operation to
4533 abort if it doesn't find a particular register name. The patch moves code
4534 from find_regno into find_regno_no_throw and makes find_regno call
4535 find_regno_no_throw instead.
4537 This allows us to do register name lookups to find a particular register
4538 number without risking erroring out if nothing is found.
4540 The patch also adjusts the feature detection code for aarch64-fbsd, since
4541 the infrastructure is shared amongst all aarch64 targets. I haven't added
4542 code to support TPIDR2 in aarch64-fbsd though, as I'm not sure when/if
4545 2022-12-09 Alan Modra <amodra@gmail.com>
4547 PR28306, segfault in _bfd_mips_elf_reloc_unshuffle
4548 Access to section data during relocation processing should be bounds
4549 checked, as it is in bfd_perform_relocation. bfd_perform_relocation
4550 does these checks after any special_function is called. So a reloc
4551 special_function needs to do its own bounds checking before accessing
4552 section data. This patch adds many such checks to the mips backend.
4554 Checking mips relocs is not without some difficulty. See the comment
4555 in _bfd_mips_reloc_offset_in_range. In a multitple reloc sequence
4556 applied to the same location, relocs that may appear somewhere other
4557 than the last one of the sequence need to be treated specially since
4558 they apply to the addend for the next relocation rather than the
4559 section contents. If the addend is in the section then it needs to be
4560 checked but not when the addend is in the reloc. check_inplace
4561 handles this situation. _bfd_mips_reloc_offset_in_range with
4562 check_shuffle handles the case where contents are shuffled before
4563 applying the relocation.
4566 * elf32-mips.c (_bfd_mips_elf32_gprel16_reloc): Check reloc
4567 address using _bfd_mips_reloc_offset_in_range.
4568 (gprel32_with_gp, mips16_gprel_reloc): Likewise.
4569 * elf64-mips.c (mips_elf64_gprel32_reloc): Likewise.
4570 (mips16_gprel_reloc): Likewise.
4571 * elfn32-mips.c (mips16_gprel_reloc): Likewise.
4572 (gprel32_with_gp): Check reloc address using
4573 bfd_reloc_offset_in_range.
4574 * elfxx-mips.h (enum reloc_check): Define.
4575 (_bfd_mips_reloc_offset_in_range): Declare.
4576 * elfxx-mips.c (needs_shuffle): New function.
4577 (_bfd_mips_elf_reloc_unshuffle, _bfd_mips_elf_reloc_shuffle): Use it.
4578 (_bfd_mips_reloc_offset_in_range): New function.
4579 (_bfd_mips_elf_gprel16_with_gp): Move reloc address checks to
4580 partial_inplace handling. Use bfd_reloc_offset_in_range.
4581 (_bfd_mips_elf_lo16_reloc): Check reloc address using
4582 bfd_reloc_offset_in_range.
4583 (_bfd_mips_elf_generic_reloc): Check reloc address using
4584 _bfd_mips_reloc_offset_in_range.
4585 (mips_elf_calculate_relocation): Check reloc address before calling
4586 mips_elf_nullify_got_load.
4587 (_bfd_mips_elf_check_relocs): Likewise.
4588 (mips_elf_read_rel_addend): Add sec param, check reloc address
4589 before reading. Adjust callers.
4590 (mips_elf_add_lo16_rel_addend): Add sec param, adjust callers.
4592 2022-12-09 Tom de Vries <tdevries@suse.de>
4594 [gdb/testsuite] Fix gdb.guile/scm-symtab.exp for ppc64le
4595 On powerpc64le-linux, I run into:
4597 (gdb) PASS: gdb.guile/scm-symtab.exp: step out of func2
4598 guile (print (> (sal-line (find-pc-line (frame-pc (selected-frame)))) line))^M
4600 (gdb) FAIL: gdb.guile/scm-symtab.exp: test find-pc-line with resume address
4603 The problem is as follows: the instructions for the call to func2 are:
4605 1000070c: 39 00 00 48 bl 10000744 <func1>
4606 10000710: 00 00 00 60 nop
4607 10000714: 59 00 00 48 bl 1000076c <func2>
4608 10000718: 00 00 00 60 nop
4609 1000071c: 00 00 20 39 li r9,0
4611 and the corresponding line number info is:
4614 File name Line number Starting address View Stmt
4615 scm-symtab.c 42 0x1000070c x
4616 scm-symtab.c 43 0x10000714 x
4617 scm-symtab.c 44 0x1000071c x
4620 The test-case looks at the line numbers for two insns:
4621 - the insn of the call to func2 (0x10000714), and
4622 - the insn after that (0x10000718),
4623 and expects the line number of the latter to be greater than the line number
4626 However, both insns have the same line number: 43.
4628 Fix this by replacing ">" with ">=".
4630 Tested on x86_64-linux and powerpc64le-linux.
4632 2022-12-09 GDB Administrator <gdbadmin@sourceware.org>
4634 Automatic date update in version.in
4636 2022-12-08 H.J. Lu <hjl.tools@gmail.com>
4638 x86-64: Remove BND from 64-bit IBT PLT
4639 Since MPX support has been removed from x86-64 psABI, remove BND from
4640 64-bit IBT PLT by using x32 IBT PLT.
4645 * elf64-x86-64.c (elf_x86_64_get_synthetic_symtab): Also check
4646 x32 IBT PLT for 64-bit.
4647 (elf_x86_64_link_setup_gnu_properties): Always use x32 IBT PLT.
4652 * testsuite/ld-x86-64/ibt-plt-1.d: Updated.
4653 * testsuite/ld-x86-64/ibt-plt-2a.d: Likewise.
4654 * testsuite/ld-x86-64/ibt-plt-2b.d: Likewise.
4655 * testsuite/ld-x86-64/ibt-plt-2c.d: Likewise.
4656 * testsuite/ld-x86-64/ibt-plt-2d.d: Likewise.
4657 * testsuite/ld-x86-64/ibt-plt-3a.d: Likewise.
4658 * testsuite/ld-x86-64/ibt-plt-3b.d: Likewise.
4659 * testsuite/ld-x86-64/ibt-plt-3c.d: Likewise.
4660 * testsuite/ld-x86-64/ibt-plt-3d.d: Likewise.
4661 * testsuite/ld-x86-64/plt-main-ibt-x32.dd: Moved to ...
4662 * testsuite/ld-x86-64/plt-main-ibt.dd: This.
4663 * testsuite/ld-x86-64/x86-64.exp: Don't use plt-main-ibt-x32.dd.
4665 2022-12-08 Tom de Vries <tdevries@suse.de>
4667 [gdb/testsuite] Require debug info for gdb.tui/tui-layout-asm-short-prog.exp
4668 When running test-case gdb.tui/tui-layout-asm-short-prog.exp on SLE-12-SP3
4669 aarch64, I run into:
4671 FAIL: gdb.tui/tui-layout-asm-short-prog.exp: check asm box contents
4672 FAIL: gdb.tui/tui-layout-asm-short-prog.exp: check asm box contents again
4676 (gdb) file tui-layout-asm-short-prog^M
4677 Reading symbols from tui-layout-asm-short-prog...^M
4678 (No debugging symbols found in tui-layout-asm-short-prog)^M
4681 I managed to reproduce the same behaviour on openSUSE Leap 15.4 x86_64, by
4682 removing the debug option.
4684 Fix this by making the test-case unsupported if no debug info is found.
4686 Tested on x86_64-linux.
4688 2022-12-08 Enze Li <enze.li@hotmail.com>
4690 gdb/testsuite: update a pattern in gdb_file_cmd
4691 When building GDB with the following CFLAGS and CXXFLAGS as part of
4694 CFLAGS=-std=gnu11 CXXFLAGS=-std=gnu++11
4696 Then run the selftest.exp, I see:
4699 Running /home/lee/dev/binutils-gdb/gdb/testsuite/gdb.gdb/selftest.exp
4701 FAIL: gdb.gdb/selftest.exp: run until breakpoint at captured_main
4702 WARNING: Couldn't test self
4706 # of unexpected failures 1
4707 /home/lee/dev/binutils-gdb/gdb/gdb version 13.0.50.20221206-git -nw -nx
4708 -iex "set height 0" -iex "set width 0" -data-directory
4709 /home/lee/dev/binutils-gdb/gdb/testsuite/../data-directory
4712 It is the fact that when I use the previously mentioned CFLAGS and
4713 CXXFLAGS as part of the configuration line, the default value (-O2 -g)
4714 is overridden, then GDB has no debug information. When there's no debug
4715 information, GDB should not run the testcase in selftest.exp.
4717 The root cause of this FAIL is that the $gdb_file_cmd_debug_info didn't
4718 get the right value ("nodebug") during the gdb_file_cmd procedure.
4720 That's because in this commit,
4722 commit 3453e7e409f44a79ac6695589836edb8a49bfb08
4723 Date: Sat May 19 11:25:20 2018 -0600
4725 Clean up "Reading symbols" output
4727 It changed "no debugging..." to "No debugging..." which causes the above
4728 problem. This patch only updates the corresponding pattern to fix this
4731 With this patch applied, I see:
4734 Running /home/lee/dev/binutils-gdb/gdb/testsuite/gdb.gdb/selftest.exp
4739 # of untested testcases 1
4740 /home/lee/dev/binutils-gdb/gdb/gdb version 13.0.50.20221206-git -nw -nx
4741 -iex "set height 0" -iex "set width 0" -data-directory
4742 /home/lee/dev/binutils-gdb/gdb/testsuite/../data-directory
4745 Tested on x86_64-linux.
4747 Approved-By: Simon Marchi <simon.marchi@efficios.com>
4749 2022-12-08 Nick Clifton <nickc@redhat.com>
4751 Update the description of the linker script's TYPE directive.
4753 * ld.texi (Output Section Type): Note that setting the output
4754 section type only works if the section contains untyped data.
4756 2022-12-08 Jan Vrany <jan.vrany@labware.com>
4758 gdb: skip objfiles with no BFD in DWARF unwinder
4759 While playing with JIT reader I experienced GDB to crash on null-pointer
4760 dereference when stepping through non-jitted code.
4762 The problem was that dwarf2_frame_find_fde () assumed that all objfiles
4763 have BFD but that's not always true. To address this problem, this
4764 commit skips such objfiles.
4766 To test the fix we put breakpoint in jit_function_add (). The JIT reader
4767 does not know how unwind this function so unwinding eventually falls
4768 back to DWARF unwinder which in turn iterates over objfiles. Since the
4769 the code is jitted, it is guaranteed it would eventually process JIT
4772 Approved-By: Simon Marchi <simon.marchi@efficios.com>
4774 2022-12-08 Alan Modra <amodra@gmail.com>
4776 libctf: avoid potential double free
4777 * ctf-link.c (ctf_link_add_cu_mapping): Set t NULL after free.
4779 2022-12-08 GDB Administrator <gdbadmin@sourceware.org>
4781 Automatic date update in version.in
4783 2022-12-07 Peter Bergner <bergner@linux.ibm.com>
4785 PowerPC: Add support for RFC02655 - Saturating Subtract Instruction
4787 * ppc-opc.c (XOL): New define.
4788 (XOL_MASK): Likewise.
4789 (powerpc_opcodes): Add subfus, subfus., subwus, subwus., subdus, subdus.
4792 * testsuite/gas/ppc/rfc02655.s: New test.
4793 * testsuite/gas/ppc/rfc02655.d: Likewise
4794 * testsuite/gas/ppc/future-raw.s: Likewise.
4795 * testsuite/gas/ppc/future-raw.d: Likewise.
4796 * testsuite/gas/ppc/ppc.exp: Run them.
4798 2022-12-07 Peter Bergner <bergner@linux.ibm.com>
4800 PowerPC: Add support for RFC02656 - Enhanced Load Store with Length Instructions
4802 * ppc-opc.c (PPCVSXF): New define.
4803 (powerpc_opcodes): Add lxvrl, lxvrll, lxvprl, lxvprll, stxvrl,
4804 stxvrll, stxvprl, stxvprl.
4807 * testsuite/gas/ppc/rfc02656.s: New test.
4808 * testsuite/gas/ppc/rfc02656.d: Likewise.
4809 * testsuite/gas/ppc/ppc.exp: Run it.
4811 2022-12-07 Simon Marchi <simon.marchi@polymtl.ca>
4813 gdb: add invalidate_selected_frame function
4814 Instead of using `select_frame (nullptr)` to invalidate the selected
4815 frame, introduce a function to do that. There is no change in behavior,
4816 but it makes the intent a bit clearer. It also allows adding an assert
4817 in select_frame that fi is not nullptr, so it avoids passing nullptr by
4820 Change-Id: I61643f46bc8eca428334513ebdaadab63997bdd0
4821 Reviewed-By: Bruno Larsen <blarsen@redhat.com>
4823 2022-12-07 Tom de Vries <tdevries@suse.de>
4825 [gdb/testsuite] Add KFAILs in gdb.base/longjmp.exp
4826 Add KFAILs in test-case gdb.base/longjmp.exp for PR gdb/26967, covering
4827 various ways that gdb is unable to recover the longjmp target if the libc
4828 probe is not supported.
4830 Tested on x86_64-linux.
4832 Approved-By: Simon Marchi <simon.marchi@efficios.com>
4834 2022-12-07 Tom Tromey <tromey@adacore.com>
4836 Remove unnecessary xstrdup from bppy_init
4837 I saw that bppy_init used a non-const "char *". Fixing this revealed
4838 that the xstrdup here was also unnecessary, so this patch removes it.
4840 2022-12-07 Alan Modra <amodra@gmail.com>
4842 coff make_a_section_from_file tidy
4843 Also support compressing a few more sections.
4845 * coffgen.c (make_a_section_from_file): Rename return_section
4846 to newsect. Don't try to be clever matching section name.
4847 Compress .gnu.debuglto_.debug_ and .gnu.linkonce.wi. too.
4848 Only rename debug sections when decompressing for linker.
4850 2022-12-07 Alan Modra <amodra@gmail.com>
4852 gas compress_debug tidy
4853 * write.c (compress_debug): Don't set up "ob" until after
4854 seginfo NULL check. Simplify SEC_CONTENTS test. Localise
4855 variables. Use bfd_debug_name_to_zdebug.
4857 _bfd_elf_slurp_secondary_reloc_section sanity check
4858 * elf.c (_bfd_elf_slurp_secondary_reloc_section): Sanity check
4859 section header against file size. Avoid overflow in
4862 bfd_compress_section_contents access to elf_section_data
4863 * compress.c (bfd_compress_section_contents): Don't access
4864 elf_section_data for non-ELF.
4866 2022-12-07 Alan Modra <amodra@gmail.com>
4868 Compression tidy and fixes
4870 - Move stuff from bfd-in.h and libbfd.c to compress.c
4871 - Delete COMPRESS_DEBUG from enum compressed_debug_section_type
4872 - Move compress_debug field out of link_info to ld_config.
4874 - Correct test in bfd_convert_section_setup to use obfd flags,
4876 - Apply bfd_applicable_file_flags to compression bfd flags added
4877 by gas and ld to the output bfd.
4880 * bfd-in.h (enum compressed_debug_section_type),
4881 (struct compressed_type_tuple),
4882 (bfd_get_compression_algorithm),
4883 (bfd_get_compression_algorithm_name),
4884 * libbfd.c (compressed_debug_section_names),
4885 (bfd_get_compression_algorithm),
4886 (bfd_get_compression_algorithm_name): Move..
4887 * compress.c: ..to here, deleting COMPRESS_DEBUG from
4888 enum compressed_debug_section_type.
4889 (bfd_convert_section_setup): Test obfd flags not ibfd for
4891 * elf.c (elf_fake_sections): Replace link_info->compress_debug
4892 test with abfd->flags test.
4893 * bfd-in2.h: Regenerate.
4895 * objcopy.c (copy_file): Tidy setting of bfd compress flags.
4898 * write.c (compress_debug): Test bfd compress flags rather than
4899 flag_compress_debug.
4900 (write_object_file): Apply bfd_applicable_file_flags to compress
4901 debug flags added to output bfd.
4903 * bfdlink.h (struct bfd_link_info): Delete compress_debug.
4905 * ld.h (ld_config_type): Add compress_debug.
4906 * emultempl/elf.em: Replace references to link_info.compress_debug
4907 with config.compress_debug.
4908 * lexsup.c (elf_static_list_options): Likewise.
4909 * ldmain.c (main): Likewise. Apply bfd_applicable_file_flags
4910 to compress debug flags added to output bfd.
4912 2022-12-07 GDB Administrator <gdbadmin@sourceware.org>
4914 Automatic date update in version.in
4916 2022-12-06 H.J. Lu <hjl.tools@gmail.com>
4918 bfd: Avoid signed overflow for new_size adjustment
4919 When bfd_size_type is unsigned 64-bit integer and sizeof is unsigned
4920 32-bit integer, subtraction in
4922 *new_size += sizeof (Elf32_External_Chdr) - sizeof (Elf64_External_Chdr);
4926 *new_size -= sizeof (Elf64_External_Chdr) - sizeof (Elf32_External_Chdr);
4931 * compress.c (bfd_convert_section_setup): Avoid signed overflow
4932 for new_size adjustment.
4934 2022-12-06 Tom Tromey <tromey@adacore.com>
4936 Cosmetic fix in ppc-sysv-tdep.c
4937 This is just a couple of cosmetic fixes in ppc-sysv-tdep.c: fixing
4938 some formatting and correcting a typo.
4940 Fix operator precedence bug in Rust parser
4941 PR rust/29859 points out an operator precedence bug in the Rust
4942 parser. This patch fixes it and adds a regression test.
4944 2022-12-06 Nick Clifton <nickc@redhat.com>
4946 Fix a dereference of NULL when scanning the symbol hashes array in the ARM linker.
4948 * elf32-arm.c (cmse_scan): Check for NULL entries in the
4950 (elf32_arm_gc_mark_extra_sections): Likewise.
4952 2022-12-06 Tom de Vries <tdevries@suse.de>
4954 [gdb/testsuite] Fix test names in gdb.base/longjmp.exp
4955 When running test-case gdb.base/longjmp.exp, we have:
4957 PASS: gdb.base/longjmp.exp: next over setjmp (1)
4959 PASS: gdb.base/longjmp.exp: next over setjmp (2)
4962 The trailing " (1)" and " (2)" are interpreted as comments rather than parts
4963 of the test name, and therefore this is a duplicate, which is currently not
4964 detected by our duplicate detection mechanism (PR testsuite/29772).
4966 Fix the duplicate by using with_test_prefix.
4968 Tested on x86_64-linux.
4970 2022-12-06 Tom de Vries <tdevries@suse.de>
4972 [gdb/testsuite] Make gdb.base/longjmp.exp FAIL more stable across archs
4973 When running test-case gdb.base/longjmp.exp on x86_64-linux, the master
4974 longjmp breakpoint is set using probes and the test-case passes:
4976 (gdb) PASS: gdb.base/longjmp.exp: next to longjmp (1)
4978 0x00000000004005cc 49 if (setjmp (env) == 0) /* patt1 */^M
4979 (gdb) PASS: gdb.base/longjmp.exp: next over longjmp(1)
4982 (gdb) PASS: gdb.base/longjmp.exp: next into else block (1)
4985 However, if I disable
4986 create_longjmp_master_breakpoint_probe, we have instead:
4988 (gdb) PASS: gdb.base/longjmp.exp: next to longjmp (1)
4991 (gdb) FAIL: gdb.base/longjmp.exp: next over longjmp(1)
4994 At first glance, the failure mode doesn't look too bad: we stop
4995 a few insns later than the passing scenario.
4997 For contrast, if we do the same on powerpc64le, the failure mode is:
4999 (gdb) PASS: gdb.base/longjmp.exp: next to longjmp (1)
5002 Breakpoint 3, main () at longjmp.c:59^M
5003 59 i = 1; /* miss_step_1 */^M
5004 (gdb) FAIL: gdb.base/longjmp.exp: next over longjmp(1)
5006 Here we only stop because of running into the safety net breakpoint at
5009 So, how does this happen on x86_64? Let's look at the code:
5011 4005c7: e8 94 fe ff ff call 400460 <_setjmp@plt>
5012 4005cc: 85 c0 test %eax,%eax
5013 4005ce: 75 1e jne 4005ee <main+0x3b>
5014 4005d0: 8b 05 8e 0a 20 00 mov 0x200a8e(%rip),%eax # 601064 <longjmps>
5015 4005d6: 83 c0 01 add $0x1,%eax
5016 4005d9: 89 05 85 0a 20 00 mov %eax,0x200a85(%rip) # 601064 <longjmps>
5017 4005df: be 01 00 00 00 mov $0x1,%esi
5018 4005e4: bf 80 10 60 00 mov $0x601080,%edi
5019 4005e9: e8 82 fe ff ff call 400470 <longjmp@plt>
5020 4005ee: 8b 05 74 0a 20 00 mov 0x200a74(%rip),%eax # 601068 <resumes>
5022 The next over the longjmp call at 4005e9 is supposed to stop at the longjmp
5023 target at 4005cc, but instead we stop at 4005ee, where we have the step-resume
5024 breakpoint inserted by the next. In other words, we accidentally "return"
5025 from the longjmp call to the insn immediately after it (even though
5026 a longjmp is a noreturn function).
5028 Try to avoid this accident and make the failure mode on x86_64 the same as on
5029 powerpc64le, by switching the then and else branch.
5031 Tested on x86_64-linux.
5033 2022-12-06 Xiao Zeng <zengxiao@eswincomputing.com>
5035 gdb/riscv: correct dwarf to gdb register number mapping
5036 According to the riscv psabi, the mapping relationship between the
5037 DWARF registers and the machine registers is as follows:
5039 DWARF Number | Register Name | Description
5040 0 - 31 | x0 - x31 | Integer Registers
5041 32 - 63 | f0 - f31 | Floating-point Registers
5043 This is not modelled quite right in riscv_dwarf_reg_to_regnum, the
5044 DWARF register numbers 31 and 63 are not handled correctly due to a
5045 use of '<' instead of '<='. This commit fixes this issue.
5047 2022-12-06 Haochen Jiang <haochen.jiang@intel.com>
5049 x86: Remove unnecessary vex.w check for xh_mode in disassembler
5050 For all the xh_mode usage in table, they are all using %XH, which will
5051 print "{bad}" while EVEX.W=1. This makes this vex.w check unnecessary.
5055 * i386-dis.c (OP_E_memory): Remove vex.w check for xh_mode.
5057 2022-12-06 Alan Modra <amodra@gmail.com>
5059 Get rid of SEC_ELF_COMPRESS
5060 This flag also isn't needed, except for some sanity checks which we
5063 * elf.c (elf_fake_sections): Don't set SEC_ELF_COMPRESS for
5064 compressed debug sections, just leave sh_name as -1.
5065 (assign_file_positions_for_non_load_sections),
5066 (assign_file_positions_except_relocs): Decide whether a section
5067 needs compressing and thus should not have its file offset set
5068 by looking at sh_name.
5069 (_bfd_elf_assign_file_positions_for_non_load): Similarly decide
5070 which sections need compressing.
5071 * elflink.c (bfd_elf_final_link): Don't test SEC_ELF_COMPRESS.
5072 * merge.c (_bfd_write_merged_section): Likewise.
5073 * section.c (SEC_ELF_COMPRESS): Don't define.
5074 (SEC_ELF_PURECODE): Renumber.
5075 * bfd-in2.h: Regenerate.
5077 2022-12-06 Alan Modra <amodra@gmail.com>
5079 Get rid of SEC_ELF_RENAME
5080 SEC_ELF_RENAME is a flag used to effect section name changes when
5081 compressing/decompressing zlib-gnu debug sections. This can be
5082 accomplished more directly in one of the objcopy specific bfd
5083 functions. Renaming for ld input is simplified too. Ld input object
5084 files always have BFD_DECOMPRESS set.
5087 * compress.c (bfd_convert_section_size): Rename to..
5088 (bfd_convert_section_setup): ..this. Handle objcopy renaming
5089 of compressed/decompressed debug sections.
5090 * elf.c (_bfd_elf_make_section_from_shdr): Only rename zdebug
5092 (elf_fake_sections): Don't handle renaming of debug sections for
5094 * section.c (SEC_ELF_RENAME): Delete.
5095 * bfd-in2.h: Regenerate.
5097 * objcopy.c (setup_section): Call bfd_convert_section_setup.
5098 Don't call bfd_convert_section_size.
5100 2022-12-06 Alan Modra <amodra@gmail.com>
5102 Compression header enum
5103 Define an enum instead of using ELFCOMPRESS_ZLIB and ELFCOMPRESS_ZSTD
5104 in bfd and binutils, and move some functions from bfd.c to compress.c.
5105 When looking at the COFF/PE debug compression support, I wondered
5106 about extending it to support zstd. I likely won't do that, but
5107 the compression header ch_type field isn't just ELF specific if these
5108 headers are to be used in COFF/PE too.
5111 * bfd.c (bfd_update_compression_header),
5112 (bfd_check_compression_header, bfd_get_compression_header_size),
5113 (bfd_convert_section_size, bfd_convert_section_contents): Move to..
5114 * compress.c: ..here.
5115 (enum compression_type): New. Use it throughout file.
5116 * elf.c (_bfd_elf_make_section_from_shdr): Replace uses of
5117 ELFCOMPRESS_ZLIB and ELFCOMPRESS_ZSTD with ch_compress_zlib and
5119 * bfd-in2.h: Regenerate.
5121 * readelf.c (process_section_headers, dump_section_as_strings),
5122 (dump_section_as_bytes, load_specific_debug_section): Replace
5123 uses of ELFCOMPRESS_ZLIB and ELFCOMPRESS_ZSTD with
5124 ch_compress_zlib and ch_compress_zstd.
5126 2022-12-06 mengqinggang <mengqinggang@loongson.cn>
5128 LoongArch: Fix dynamic reloc not generated bug in some cases.
5131 * elfnn-loongarch.c (loongarch_elf_relocate_section): Likewise.
5133 2022-12-06 Alan Modra <amodra@gmail.com>
5135 PR29855, ch_type in bfd_init_section_decompress_status can be uninitialized
5137 * compress.c (bfd_init_section_decompress_status): Set ch_type
5138 to zero for zlib-gnu case.
5140 2022-12-06 GDB Administrator <gdbadmin@sourceware.org>
5142 Automatic date update in version.in
5144 2022-12-05 Simon Marchi <simon.marchi@polymtl.ca>
5146 gdb/linux-nat: add ptid parameter to linux_xfer_siginfo
5147 Make the inferior_ptid bubble up to linux_nat_target::xfer_partial.
5149 Change-Id: I62dbc5734c26648bb465f449c2003c73751cd812
5151 2022-12-05 Simon Marchi <simon.marchi@polymtl.ca>
5153 gdb/linux-nat: use l linux_nat_get_siginfo in linux_xfer_siginfo
5154 I noticed we could reduce duplication a bit here.
5156 Change-Id: If24e54d1dac71b46f7c1f68a18a073d4c084b644
5158 2022-12-05 Simon Marchi <simon.marchi@polymtl.ca>
5160 gdb/linux-nat: check ptrace return value in linux_nat_get_siginfo
5161 Not a big deal, but it seems strange to check errno instead of the
5162 ptrace return value to know whether it succeeded.
5164 Change-Id: If0a6d0280ab0e5ecb077e546af0d6fe489c5b9fd
5166 2022-12-05 Simon Marchi <simon.marchi@polymtl.ca>
5168 gdb/linux-nat: don't memset siginfo on failure in linux_nat_get_siginfo
5169 No caller cares about the value of *SIGINFO on failure. It's also
5170 documented in the function doc that *SIGINFO is uninitialized (I
5171 understand "untouched") on failure.
5173 Change-Id: I5ef38a5f58e3635e109b919ddf6f827f38f1225a
5175 2022-12-05 Simon Marchi <simon.marchi@polymtl.ca>
5177 gdb/linux-nat: bool-ify linux_nat_get_siginfo
5178 Change return type to bool.
5180 Change-Id: I1bf0360bfdd1b5994cd0f96c268d806f96fe51a4
5182 2022-12-05 Simon Marchi <simon.marchi@polymtl.ca>
5184 gdb/linux-nat: use get_ptrace_pid in two spots
5185 No behavior change expected.
5187 Change-Id: Ifaa64ecd619483646b024fd7c62e571e92a8eedb
5189 2022-12-05 Simon Marchi <simon.marchi@efficios.com>
5191 gdb/testsuite: remove perror calls when failing to run
5192 I noticed that when running these two tests in sequence:
5194 Running /home/smarchi/src/binutils-gdb/gdb/testsuite/gdb.ada/arrayptr.exp ...
5195 ERROR: GDB process no longer exists
5196 ERROR: Couldn't run foo-all
5197 Running /home/smarchi/src/binutils-gdb/gdb/testsuite/gdb.ada/assign_1.exp ...
5199 The results in gdb.sum are:
5201 Running /home/smarchi/src/binutils-gdb/gdb/testsuite/gdb.ada/arrayptr.exp ...
5202 PASS: gdb.ada/arrayptr.exp: scenario=all: compilation foo.adb
5203 ERROR: GDB process no longer exists
5204 UNRESOLVED: gdb.ada/arrayptr.exp: scenario=all: gdb_breakpoint: set breakpoint at foo.adb:40 (eof)
5205 ERROR: Couldn't run foo-all
5206 Running /home/smarchi/src/binutils-gdb/gdb/testsuite/gdb.ada/assign_1.exp ...
5207 UNRESOLVED: gdb.ada/assign_1.exp: changing the language to ada
5208 PASS: gdb.ada/assign_1.exp: set convenience variable $xxx to 1
5210 The UNRESOLVED for arrayptr.exp is fine, as GDB crashes in that test,
5211 while trying to run to main. However, the UNRESOLVED in assign_1.exp
5212 doesn't make sense, GDB behaves as expected in that test:
5214 (gdb) set lang ada^M
5215 (gdb) UNRESOLVED: gdb.ada/assign_1.exp: changing the language to ada
5218 (gdb) PASS: gdb.ada/assign_1.exp: set convenience variable $xxx to 1
5220 The problem is that arrayptr.exp calls perror when failing to run to
5221 main, then returns. perror makes it so that the next test (as in
5222 pass/fail) will be recorded as UNRESOLVED. However, here, the next test
5223 (as in pass/fail) is in the next test (as in .exp). Hence the spurious
5224 UNRESOLVED in assign_1.exp.
5226 These perror when failing to run to X are not really useful, especially
5227 since runto records a FAIL on error, by default. Remove all the
5228 perrors on runto failure I could find.
5230 When there wasn't one already, add a return statement when failing to
5231 run, to avoid running the test of the test unnecessarily.
5233 I thought of adding a check ran between test (in gdb_finish
5234 probably) where we would emit a warning if errcnt > 0, meaning a test
5235 quit and left a perror "active". However, reading that variable would
5236 poke into the DejaGNU internals, not sure it's a good idea.
5238 Change-Id: I2203df6d06e199540b36f56470d1c5f1dc988f7b
5240 2022-12-05 Luis Machado <luis.machado@arm.com>
5242 Add missing newline to gdbarch_tdep debugging output
5243 The missing newline causes testsuite issues because the gdb prompt gets output
5244 to an unexpected location.
5246 2022-12-05 Nick Clifton <nickc@redhat.com>
5248 Prevent an illegal memory access when comparing the prefix of a section name regexp.
5250 * ldlang.c (spec_match): Check that there is sufficient length in
5251 the target name to match the spec's prefix.
5253 2022-12-05 Martin Liska <mliska@suse.cz>
5255 testsuite: support mold linker
5256 Mold linker demotes symbols like main to be local and the patch
5257 adjusts expected output from nm.
5259 Moreover, simplify the regex patterns.
5261 2022-12-05 Thiago Jung Bauermann <thiago.bauermann@linaro.org>
5263 gdbarch.py: Fix indentation in the generated set_gdbarch_* definitions
5264 Use as many tabs as possible for indentation and pad with spaces to keep
5265 the argument aligned to the opening parenthesis in the line above.
5267 Co-developed-by: Simon Marchi <simon.marchi@efficios.com>
5268 Approved-By: Simon Marchi <simon.marchi@efficios.com>
5270 2022-12-05 Thiago Jung Bauermann <thiago.bauermann@linaro.org>
5272 gdbarch.py: Fix indentation in the generated gdbarch_dump function
5273 Use tab for the first eight spaces of indentation, and align the gdb_printf
5274 arguments to the open parenthesis of the function call.
5276 Approved-By: Simon Marchi <simon.marchi@efficios.com>
5278 2022-12-05 Thiago Jung Bauermann <thiago.bauermann@linaro.org>
5280 gdb: Update my email address in MAINTAINERS
5282 2022-12-05 Jan Beulich <jbeulich@suse.com>
5284 gas: add Dwarf line number test for .macro expansions
5285 Before fiddling with the code let's put in place a test covering what
5286 PR/gas 16908 aimed at.
5288 gas: squash (some) .linefile from listings
5289 Not so long ago we started to insert these artificially when expanding
5290 certain macro-like constructs; zap them as cluttering what actually
5291 results from user input.
5293 2022-12-05 Jan Beulich <jbeulich@suse.com>
5295 gas: avoid inserting extra newline in buffer_and_nest()
5296 In "-alm" listings I've noticed an odd blank line following the inserted
5297 .linefile one. This results from the explicit NL inserted being
5298 redundant with the one left in place from the original input line by all
5299 respective callers. Note that we need to compensate for the removed line
5300 by bumping the directive argument (which in turn is decremented again in
5301 s_linefile() before calling new_logical_line_flags(), and I have to
5302 confess that when putting together the original change I was a little
5303 puzzled by the imbalance of increments/decrements, but then I forgot to
5304 actually go look for the cause).
5306 While there also switch to sb_add_string() instead of effectively open-
5307 coding it to some degree.
5309 2022-12-05 Jan Beulich <jbeulich@suse.com>
5311 Arm: .noinit and .persistent are not supported for Linux targets
5312 Respective tests being run means guaranteed failures.
5314 2022-12-05 Nick Clifton <nickc@redhat.com>
5316 Fix an illegal memory access when parsing a corrupt VMS Alpha file.
5318 * vms-alpha.c (parse_module): Fix potential out of bounds memory
5321 2022-12-05 Andrew Burgess <aburgess@redhat.com>
5323 libopcodes/mips: add support for disassembler styling
5324 This commit adds disassembler styling support for MIPS. After this
5325 commit objdump and GDB will style disassembler output.
5327 This is a pretty straight forward change, we switch to use the
5328 disassemble_info::fprintf_styled_func callback, and pass an
5329 appropriate style through as needed. No additional tricks were
5330 needed (compared to say i386, or ARM).
5332 Tested by running all of the objdump commands used by the gas
5333 testsuite and manually inspecting the styled output, everything looks
5334 reasonable, though I'm not a MIPS expert, so it is possible that I've
5335 missed some corner cases. Worst case though is that something will be
5336 styled incorrectly, the actual content should be unchanged.
5338 All the gas, ld, and binutils tests still pass for me.
5340 2022-12-05 Andrew Burgess <aburgess@redhat.com>
5342 opcodes/mips: use .word/.short for undefined instructions
5343 While working on disassembler styling for MIPS, I noticed that
5344 undefined instructions are printed by the disassembler as raw number
5345 with no assembler directive prefix (e.g. without .word or .short).
5347 I think adding something like .word, or .short, helps to make it
5348 clearer the size of the value that is being displayed, and is inline
5349 with what many of the other libopcode disassemblers do.
5351 In this commit I've added the .word and .short directives, and updated
5352 all the tests that I spotted that failed as a result.
5354 2022-12-05 Alan Modra <amodra@gmail.com>
5356 Re: Renaming .debug to .zdebug and vice versa
5357 * compress.c (bfd_debug_name_to_zdebug): Fix C++ compile error.
5358 (bfd_zdebug_name_to_debug): Likewise.
5359 * bfd-in2.h: Regenerate.
5361 2022-12-05 GDB Administrator <gdbadmin@sourceware.org>
5363 Automatic date update in version.in
5365 2022-12-04 Alan Modra <amodra@gmail.com>
5367 PR29846, segmentation fault in objdump.c compare_symbols
5368 Fixes a fuzzed object file problem where plt relocs were manipulated
5369 in such a way that two synthetic symbols were generated at the same
5370 plt location. Won't occur in real object files.
5374 * objdump.c (compare_symbols): Test symbol flags to exclude
5375 section and synthetic symbols before attempting to check flavour.
5377 2022-12-04 Alan Modra <amodra@gmail.com>
5379 COFF compressed debug support
5380 Since commit 4bea06d73c04 COFF support for compressed debug sections
5381 has been broken due to the "flags" variable not getting SEC_HAS_CONTENTS.
5383 * coffgen.c (make_a_section_from_file): Correct section flags
5384 handling. Delete extraneous condition. Update error messages
5385 to be the same as in elf.c.
5387 2022-12-04 Alan Modra <amodra@gmail.com>
5389 Renaming .debug to .zdebug and vice versa
5390 Move a couple of elf.c functions to compress.c.
5392 * compress.c (bfd_debug_name_to_zdebug): New inline function.
5393 (bfd_zdebug_name_to_debug): Likewise.
5394 * elf.c (convert_debug_to_zdebug, convert_zdebug_to_debug): Delete.
5395 (_bfd_elf_make_section_from_shdr, elf_fake_sections),
5396 (_bfd_elf_assign_file_positions_for_non_load): Adjust to suit.
5397 * coffgen.c (make_a_section_from_file): Use new inlines here.
5399 2022-12-04 GDB Administrator <gdbadmin@sourceware.org>
5401 Automatic date update in version.in
5403 2022-12-03 H.J. Lu <hjl.tools@gmail.com>
5405 Revert "ld: Add .note.GNU-stack to ld-plugin/dummy.s"
5406 This reverts commit 44e59b5a7d8a874f6546a1471b8a003911853aa0.
5408 It works only for ELF targets.
5410 2022-12-03 H.J. Lu <hjl.tools@gmail.com>
5412 ld: Add .note.GNU-stack to ld-plugin/dummy.s
5413 * testsuite/ld-plugin/dummy.s: Add .note.GNU-stack.
5415 2022-12-03 H.J. Lu <hjl.tools@gmail.com>
5417 x86: Allow 16-bit register source for LAR and LSL
5418 Since LAR and LSL only access 16 bits of the source operand, regardless
5419 of operand size, allow 16-bit register source for LAR and LSL, and always
5420 disassemble LAR and LSL with 16-bit source operand.
5425 * testsuite/gas/i386/i386.s: Add tests for LAR and LSL.
5426 * testsuite/gas/i386/x86_64.s: Likewise.
5427 * testsuite/gas/i386/intelbad.s: Remove "lar/lsl eax, ax".
5428 * testsuite/gas/i386/i386-intel.d: Updated.
5429 * testsuite/gas/i386/i386.d: Likewise.
5430 * testsuite/gas/i386/intel-intel.d: Likewise.
5431 * testsuite/gas/i386/intel.d: Likewise.
5432 * testsuite/gas/i386/intelbad.l: Likewise.
5433 * testsuite/gas/i386/x86_64-intel.d: Likewise.
5434 * testsuite/gas/i386/x86_64.d: Likewise.
5439 * i386-dis.c (MOD_0F02): Removed.
5440 (MOD_0F03): Likewise.
5441 (dis386_twobyte): Restore larS and lslS.
5442 (mod_table): Remove MOD_0F02 and MOD_0F03.
5443 * i386-opc.tbl: Allow 16-bit register source for LAR and LSL.
5444 * i386-tbl.h: Regenerated.
5446 2022-12-03 GDB Administrator <gdbadmin@sourceware.org>
5448 Automatic date update in version.in
5450 2022-12-02 Simon Marchi <simon.marchi@polymtl.ca>
5452 gdb/linux-nat: add pid parameter to linux_proc_xfer_memory_partial
5453 Add a pid parameter to linux_proc_xfer_memory_partial, making the
5454 inferior_ptid reference bubble up close to the target_ops::xfer_partial
5455 boundary. No behavior change expected.
5457 Change-Id: I58171b00ee1bba1ea22efdbb5dcab8b1ab3aac4c
5459 2022-12-02 Simon Marchi <simon.marchi@efficios.com>
5461 gdb: add some debug statements to solib-svr4.c
5462 Add a few debug statements that were useful to me when debugging why the
5463 glibc probes interface wasn't getting used.
5465 Change-Id: Ic20744f9fc80a90f196896b0829949411620c540
5467 2022-12-02 Simon Marchi <simon.marchi@polymtl.ca>
5469 gdb: merge solib-frv aix-solib debug options into "set/show debug solib"
5470 solib implementations are typically used one at a time. So it will be
5471 rare that you will want to enable debug for one solib kind, and
5472 absolutely want to keep the others disabled. To make things simpler,
5473 instead of adding separate variables / macros / commands for each solib
5474 implementation, merge the existing ones (frv and aix) into a unified
5475 "set/show debug solib", with the solib_debug_printf macro.
5477 Change-Id: I6e18bbc7401724f37ae66681badb079d75ecf7fa
5479 2022-12-02 Nick Clifton <nickc@redhat.com>
5481 Add Jan Beulich as an x86_64 maintainer.
5483 2022-12-02 Jan Beulich <jbeulich@suse.com>
5485 x86: drop most OPERAND_TYPE_* (and rework the rest)
5486 With the general use of C99 there's no need anymore to have i386-gen
5487 produce these. For more frequently used ones introduce local #define-s,
5488 while others are simply spelled out directly. While doing this move
5489 some static constants into more narrow scopes.
5491 Note that as a "side effect" this corrects type_names[]'es imm8s entry.
5493 2022-12-02 Jan Beulich <jbeulich@suse.com>
5495 x86: simplify and slightly correct XCHG vs NOP checking
5496 For one, because of CheckRegSize, there's no need to check the size of
5497 both (register) operands. And then in process_suffix() check opcode
5498 space rather than the (potentially ambiguous) extension opcode.
5500 x86: also use D for XCHG and TEST
5501 Leverage the C (commutative) attribute to also reduce the number of XCHG
5502 and TEST templates we have. This way the reg <-> r/m (and reg <-> reg for
5503 XCHG) forms can also be folded into a single template each, utilizing D.
5505 2022-12-02 Tom de Vries <tdevries@suse.de>
5507 [gdb/testsuite] Prevent timeout in gdb.ada/float-bits.exp
5508 Recent commit 32a5aa26256 ("[gdb/testsuite] Fix gdb.ada/float-bits.exp
5509 for powerpc64le") started using command "maint print architecture", which
5510 produces ~275 lines.
5512 Rewrite the corresponding gdb_test_multiple to read line-by-line, to prevent
5513 timeouts on slower test setups.
5515 Note that this doesn't fix a timeout in the test-case on aarch64 due to:
5517 gdbarch_dump: read_core_file_mappings = <0x817438>
5518 (gdb) aarch64_dump_tdep: Lowest pc = 0x0x8000
5521 Tested on x86_64-linux.
5523 2022-12-02 GDB Administrator <gdbadmin@sourceware.org>
5525 Automatic date update in version.in
5527 2022-12-01 Carl Love <cel@us.ibm.com>
5529 PowerPC, fix gdb.reverse/finish-reverse-bkpt.exp and gdb.reverse/next-reverse-bkpt-over-sr.exp
5530 The tests set a break point with the command break *func. This sets a
5531 breakpoint on the first instruction of the function. PowerPC uses
5532 Global Entry Points (GEP) and Local Entry Points (LEP). The first
5533 instruction in the function is the GEP. The GEP sets up register
5534 r2 before reaching the LEP. When the function is called with func() the
5535 function is entered via the LEP and the test fails because GDB does not
5536 see the breakpoint on the GEP. However, if the function is called via a
5537 function pointer, execution begins at the GEP as the test expects.
5539 Currently finish-reverse-bkpt.exp uses source file finish-reverse.c and
5540 next-reverse-bpkt-over-sr.exp uses source file step-reverse.c A new
5541 source file was created for tests finish-reverse-bkpt.exp and
5542 next-reverse-bkpt-over-sr.exp. The new files use the new function
5543 pointer method to call the functions so the tests will work correctly on
5544 both PowerPC with a GEP and LEP as well as on other systems. The GEP is
5545 the same as the LEP on non PowerPC systems.
5547 The expect files were changed to use the new source files and to set the
5548 initial break point for the rest of the test on the function pointer call
5551 This patch fixes two PowerPC test failures in each of the tests
5552 gdb.reverse/finish-reverse-bkpt.exp and
5553 gdb.reverse/next-reverse-bkpt-over-sr.exp.
5555 Patch tested on PowerPC and Intel X86-64 with no regressions.
5557 Reviewed-By: Bruno Larsen <blarsen@redhat.com>
5559 2022-12-01 Tom Tromey <tromey@adacore.com>
5561 Remove call to registers_changed from windows-nat.c
5562 I noticed that windows_nat_target::interrupt calls registers_changed.
5563 However, I don't think there's any reason to do this, because this
5564 will happen automatically when the inferior stop is processed.
5566 Approved-By: Simon Marchi <simon.marchi@efficios.com>
5568 2022-12-01 Tom Tromey <tromey@adacore.com>
5570 Remove the_windows_nat_target global
5571 I belatedly realized that the "the_windows_nat_target" global isn't
5572 really necessary. It's only used in one place, where 'this' would be
5573 simpler and clearer. This patch removes the global entirely.
5575 Approved-By: Simon Marchi <simon.marchi@efficios.com>
5577 2022-12-01 Simon Marchi <simon.marchi@polymtl.ca>
5579 gdb: make frame_register static
5580 It is only used inside frame.c.
5582 Change-Id: I44eb46a5992412f8f8b4954b2284b0ef3b549504
5584 2022-12-01 Tom Tromey <tromey@adacore.com>
5586 Add name canonicalization for C
5587 PR symtab/29105 shows a number of situations where symbol lookup can
5588 result in the expansion of too many CUs.
5590 What happens is that lookup_signed_typename will try to look up a type
5591 like "signed int". In cooked_index_functions::expand_symtabs_matching,
5592 when looping over languages, the C++ case will canonicalize this type
5593 name to be "int" instead. Then this method will proceed to expand
5594 every CU that has an entry for "int" -- i.e., nearly all of them. A
5595 crucial component of this is that the caller, objfile::lookup_symbol,
5596 does not do this canonicalization, so when it tries to find the symbol
5597 for "signed int", it fails -- causing the loop to continue.
5599 This patch fixes the problem by introducing name canonicalization for
5600 C. The idea here is that, by making C and C++ agree on the canonical
5601 name when a symbol name can have multiple spellings, we avoid the bad
5602 behavior in objfile::lookup_symbol (and any other such code -- I don't
5603 know if there is any).
5605 Unlike C++, C only has a few situations where canonicalization is
5606 needed. And, in particular, due to the lack of overloading (thus
5607 avoiding any issues in linespec) and due to the way c-exp.y works, I
5608 think that no canonicalization is needed during symbol lookup -- only
5609 during symtab construction. This explains why lookup_name_info is not
5612 The stabs reader is modified on a "best effort" basis.
5614 The DWARF reader needed one small tweak in dwarf2_name to avoid a
5615 regression in dw2-unusual-field-names.exp. I think this is adequately
5616 explained by the comment, but basically this is a scenario that should
5617 not occur in real code, only the gdb test suite.
5619 lookup_signed_typename is simplified. It used to search for two
5620 different type names, but now gdb can search just for the canonical
5623 gdb.dwarf2/enum-type.exp needed a small tweak, because the
5624 canonicalizer turns "unsigned integer" into "unsigned int integer".
5625 It seems better here to use the correct C type name.
5627 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29105
5628 Tested-by: Simon Marchi <simark@simark.ca>
5629 Reviewed-by: Andrew Burgess <aburgess@redhat.com>
5631 2022-12-01 Tom Tromey <tromey@adacore.com>
5633 Refactor cooked_index::do_finalize
5634 This refactors cooked_index::do_finalize, reordering an 'if' to make
5635 it a little less redundant. This change makes a subsequent patch
5638 Reviewed-by: Andrew Burgess <aburgess@redhat.com>
5640 2022-12-01 Tom Tromey <tromey@adacore.com>
5642 Remove language check from dwarf2_compute_name
5643 dwarf2_compute_name has a redundant check of the CU's language -- this
5644 is also checked in dwarf2_canonicalize_name. Removing this slightly
5645 simplifies a future patch.
5647 Reviewed-by: Andrew Burgess <aburgess@redhat.com>
5649 2022-12-01 Simon Marchi <simon.marchi@efficios.com>
5651 gdb/dwarf: add some QUIT macros
5652 While testing the fix for PR 29105, I noticed I couldn't ctrl-C my way
5653 out of GDB expanding many symtabs. GDB was busy in a loop in
5654 cooked_index_functions::expand_symtabs_matching. Add a QUIT there. I
5655 also happened to see a spot in
5656 cooked_index_functions::expand_matching_symbols where a QUIT would be
5657 useful too, since we iterate over a potentially big number of index
5658 entries and expand CUs in the loop. Add one there too.
5660 Change-Id: Ie1d650381df7f944c16d841b3e592d2dce7306c3
5661 Approved-By: Kevin Buettner <kevinb@redhat.com>
5663 2022-12-01 Simon Marchi <simon.marchi@efficios.com>
5665 gdb: remove prune_threads in thread_db_target::update_thread_list
5666 Pedro mentioned that this prune_threads call in
5667 thread_db_target::update_thread_list was not needed, and it was probably
5668 an oversight to leave it there in the work following commit e8032dde10b
5669 ("Push pruning old threads down to the target"). That commit changed
5670 the "find new threads" target operation to "update thread list", making
5671 the target responsible of adding new threads and removing exited
5672 threads, rather than just adding new threads. Commit e8032dde10b moved
5673 the prune_threads calls previously done in common code into each
5674 target's update_thread_list method, in order to keep the existing
5675 behavior, which is why this prune_threads call ended up there.
5677 In the mean time, the linux-nat target was taught to update_thread_list,
5678 and thread_db_target::update_thread_list defers to that for any live
5679 inferior, so the prune_threads call is not needed there. Otherwise, the
5680 thread_db_target::update_thread_list implementation based on
5681 td_ta_thr_iter_p only knows how to add new threads, not how to delete
5682 exited threads, but that is only used for non-live inferiors, where
5683 threads can't exit anyway. So the prune_threads call is not needed for
5686 Change-Id: I127fd4f84c25086f97853dadf34c5cec6816840d
5687 Approved-By: Pedro Alves <pedro@palves.net>
5689 2022-12-01 H.J. Lu <hjl.tools@gmail.com>
5691 opcodes: Remove i386-init.h and i386-tbl.h from HFILES
5692 i386-init.h and i386-tbl.h are generated files. There is nothing to
5693 translate. Remove them from HFILES (POTFILES).
5695 * Makefile.am (HFILES): Remove i386-init.h and i386-tbl.h.
5696 * Makefile.in: Regenerated.
5697 * po/POTFILES.in: Likewise.
5699 2022-12-01 Tom Tromey <tromey@adacore.com>
5701 Avoid timeouts in gdb.compile
5702 PR compile/29541 points out that some of the C++ tests in gdb.compile
5703 will time out when the glibc debuginfo is installed. This was
5704 interfering with my hacking on gdb by making test runs extremely long,
5705 so I looked into it.
5707 Internally the bug seems to be that gdb tries to convert multiple
5708 symbols named "var" via the compiler interface; one such symbol (I
5709 didn't track it down too far) causes the C++ compiler plugin to crash.
5711 Unfortunately, the crash is reported as a timeout, as the gdb side of
5712 the plugin simply hangs. This seems like a bug in the plugin RPC
5713 mechanism and, worse, apparently when I wrote this stuff I didn't
5714 really consider error reporting very much at all, so gdb can't really
5715 detect failures in the first place.
5717 Anyway... this patch works around the timeout by compiling a simple
5718 test that should provoke this bug, and then using "untested" if it
5719 notices a GCC crash.
5721 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29541
5723 2022-12-01 Tom Tromey <tromey@adacore.com>
5725 Remove obsolete check from skip_compile_feature_tests
5726 skip_compile_feature_tests checks for "Command not supported on this
5727 host", but this error was removed by commit e8d8cce6 ("Import mkdtemp
5728 gnulib module, fix mingw build"). This patch removes the obsolete
5731 Remove one copy of skip_compile_feature_tests
5732 I noticed that there are two identical copies of
5733 skip_compile_feature_tests in the test suite. This removes one from
5734 gdb.exp, in favor of the one in compile-support.exp.
5736 2022-12-01 Clément Chigot <chigot@adacore.com>
5738 binutils: improve holes detection in .debug_loclists.
5739 The previous warnings about holes in .debug_loclists sections don't
5740 take into account the headers of each CU and could include the locviews
5741 if they precede the loclist.
5743 The following warning can be triggered between two CU.
5744 ... <previous CU views> ...
5745 0000001d <End of list>
5747 0000002a v000000000000000 v000000000000000 location view pair
5748 0000002c v000000000000000 v000000000000000 location view pair
5750 readelf: Warning: There is a hole [0x1e - 0x2e] in .debug_loclists section.
5751 0000002e v000000000000000 v000000000000000 views at 0000002a for:
5754 But [0x1e - 0x2a] corresponds to the CU header and [0x2a - 0x2e] are
5755 the locviews. Thus there is no hole here.
5759 * dwarf.c (display_debug_loc): Adjust holes detections for
5760 headers and locviews.
5762 2022-12-01 Nick Clifton <nickc@redhat.com>
5764 Fix verilog output when the width is > 1.
5766 bfd * bfd.c (VerilogDataEndianness): New variable.
5767 (verilog_write_record): Use VerilogDataEndianness, if set, to
5768 choose the endianness of the output.
5769 (verilog_write_section): Adjust the address by the data width.
5771 binutils* objcopy.c (copy_object): Set VerilogDataEndianness to the
5772 endianness of the input file.
5773 (copy_main): Verifiy the value set by the --verilog-data-width
5775 * testsuite/binutils-all/objcopy.exp: Add tests of the new behaviour.
5776 * testsuite/binutils-all/verilog-I4.hex: New file.
5778 2022-12-01 Jan Beulich <jbeulich@suse.com>
5780 x86: rework of match_template()'s suffix checking
5781 (Ab)using i386_opcode_modifier for this has been overkill, as the logic
5782 doesn't really require the full structure. With the removal of
5783 LONG_DOUBLE_MNEM_SUFFIX and No_ldSuf there's no good reason at all
5784 anymore to pull out such a loop invariant: We're dealing a check of a
5785 bit in the loop for a simple comparison. Do the original compares inside
5786 the loop, thus also making it easier to understand what is actually
5790 With LONG_DOUBLE_MNEM_SUFFIX gone there'salso no use for No_ldSuf
5793 x86/Intel: drop LONG_DOUBLE_MNEM_SUFFIX
5794 With the removal of its use for FPU insns the suffix is now finally
5795 properly misnamed. Drop its use altogether, replacing it by a separate
5798 x86/Intel: restrict use of LONG_DOUBLE_MNEM_SUFFIX
5799 As a comment near the top of match_template() already says: We really
5800 only need this pseudo-suffix for far branch handling. Stop "deriving" it
5801 for floating point insns. (Don't bother renaming the now properly
5802 misnamed LONG_DOUBLE_MNEM_SUFFIX, to e.g. FAR_BRANCH_SUFFIX - it's going
5803 to disappear anyway.)
5805 2022-12-01 Tom de Vries <tdevries@suse.de>
5807 [gdb/testsuite] Wait longer for core generation
5808 When I run the gdb testsuite on a powerpc64le-linux system with (slow) nfs
5809 file system, I run into timeouts due to core generation, like for instance:
5811 (gdb) gcore $outputs/gdb.ada/task_switch_in_core/crash.gcore^M
5812 FAIL: gdb.ada/task_switch_in_core.exp: save a corefile (timeout)
5815 Fix this by using with_timeout_factor 3 in gdb_gcore_cmd.
5817 Tested on powerpc64le-linux.
5818 Approved-By: Tom Tromey <tom@tromey.com>
5820 2022-12-01 Tom de Vries <tdevries@suse.de>
5822 [gdb/testsuite] Fix gdb.ada/float-bits.exp for powerpc64le
5823 On powerpc64le-linux, I run into:
5825 (gdb) print 16llf#4000921fb54442d18469898cc51701b8#^M
5826 $9 = <invalid float value>^M
5827 (gdb) FAIL: gdb.ada/float-bits.exp: print \
5828 16llf#4000921fb54442d18469898cc51701b8#
5831 The problem is that we're using a hex string for the 128-bit IEEE quad long
5832 double format, but the actual long double float format is:
5834 gdbarch_dump: long_double_format = floatformat_ibm_long_double_little^M
5837 Fix this by using the hex string obtained by compiling test.c:
5839 long double a = 5.0e+25L;
5843 $ gcc -mlittle test.c -c -g
5847 $ gdb -q -batch test.o -ex "p /x a"
5848 $1 = 0xc1e1c000000000004544adf4b7320335
5850 and likewise for -mbig:
5852 $ gdb -q -batch test.o -ex "p /x a"
5853 $1 = 0x4544adf4b7320335c1e1c00000000000
5856 Tested on powerpc64le-linux.
5858 I excercised the case of floatformat_ibm_long_double_big by
5859 using "set endian big" in the test-case.
5861 Note that for this patch to work correctly, recent commit aaa79cd62b8 ("[gdb]
5862 Improve printing of float formats") is required.
5865 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29816
5866 Approved-By: Tom Tromey <tom@tromey.com>
5868 2022-12-01 GDB Administrator <gdbadmin@sourceware.org>
5870 Automatic date update in version.in
5872 2022-11-30 Tom de Vries <tdevries@suse.de>
5874 [gdb/testsuite] Fix DUPLICATEs in s390-multiarch.exp
5875 On s390x-linux, I run into:
5877 DUPLICATE: gdb.arch/s390-multiarch.exp: Linux v2
5878 DUPLICATE: gdb.arch/s390-multiarch.exp: Linux v2
5879 DUPLICATE: gdb.arch/s390-multiarch.exp: Linux v2
5882 Fix this by using with_test_prefix.
5884 Tested on s390x-linux.
5886 2022-11-30 Tom de Vries <tdevries@suse.de>
5888 [gdb/testsuite] Enable gdb.arch/s390-disassembler-options.exp for --enable-targets=all
5889 On s390x-linux, I run into:
5891 DUPLICATE: gdb.arch/s390-disassembler-options.exp: \
5892 show disassembler-options esa
5895 First, reproduce this on x86_64-linux with --enable-targets=all, by replacing
5896 the test for 'istarget "s390*-*-*"' with a test for 'get_set_option_choices
5897 "set architecture" "s390"'.
5899 Fix the DUPLICATE by using with_test_prefix.
5901 Also modernize the test-case by using clean_restart instead of gdb_exit/gdb_start.
5903 Tested on x86_64-linux.
5905 2022-11-30 Michael Matz <matz@suse.de>
5907 section-select: Fix exclude-file-3
5908 this testcase wasn't correctly testing everything, it passed, even
5909 though sections from an excluded file were included. Fixing this
5910 reveals a problem in the new section selector. This fixes that as
5913 section-select: Remove unused code
5914 walk_wild_file, hence walk_wild_section and walk_wild_section_handler
5915 aren't called with the prefix tree. Hence initialization of the latter
5916 and all potential special cases for it aren't used anymore. That also
5917 removes the need to handler_data[] and some associated helper functions.
5918 So, remove all of that.
5920 2022-11-30 Michael Matz <matz@suse.de>
5922 section-select: Implement a prefix-tree
5923 Now that we have a list of potentially matching sections per wild
5924 statement we can actually pre-fill that one by going once over all input
5925 sections and match their names against a prefix-tree that points to the
5926 potentially matching wild statements.
5928 So instead of looking at all sections names for each glob for each wild
5929 statement we now look at the sections only once and then only check
5930 against those globs that have a possibility of a match at all (usually
5933 This pushes the whole section selection off the profiles.
5935 2022-11-30 Michael Matz <matz@suse.de>
5937 section-select: Completely rebuild matches
5938 The check_relocs callback (and others) might have created new
5939 section behind our back and some of them (e.g. on powerpc the
5940 "linker stubs" .got) need to come in front of all others, despite
5941 being created late (a symptom would be "TOC opt*" failing on powerpc).
5943 This resets all section matches before updating for newly created
5944 sections (i.e. completely rebuilds the matches).
5946 2022-11-30 Michael Matz <matz@suse.de>
5948 section-select: Lazily resolve section matches
5949 and remember the results. Before this the order of section matching
5952 foreach script-wild-stmt S
5953 foreach pattern P of S
5955 foreach section S of I
5957 if match: do action for S
5959 And this process is done three or four times: for each top-level call to
5960 walk_wild() or wild(), that is: check_input_sections, lang_gc_sections,
5961 lang_find_relro_sections and of course map_input_to_output_sections.
5963 So we iterate over all sections of all files many many times (for each
5964 glob). Reality is a bit more complicated (some special glob types don't
5965 need the full iteration over all sections, only over all files), but
5966 that's the gist of it.
5968 For future work this shuffles the whole ordering a bit by lazily doing
5969 the matching process and memoizing results, trading a little memory for
5970 a 75% speedup of the overall section selection process.
5972 This lazy resolution introduces a problem with sections added late
5973 that's corrected in the next patch.
5975 2022-11-30 Tom Tromey <tromey@adacore.com>
5977 Bounds check access to Ada task state names
5978 While looking into Ada tasking a little, I noticed that no bounds
5979 checking is done on accesses to the Ada task state names arrays. This
5980 isn't a problem currently, but if the runtime ever added numbers -- or
5981 if there was some kind of runtime corruption -- it could cause a gdb
5984 This patch adds range checking. It also adds a missing _() call when
5985 printing from the 'task_states' array.
5987 2022-11-30 Tom Tromey <tromey@adacore.com>
5989 Use ui_file_up in mi_interp
5990 This changes mi_interp to use ui_file_up rather than explicit
5993 Approved-By: Simon Marchi <simon.marchi@efficios.com>
5995 2022-11-30 Tom Tromey <tromey@adacore.com>
5997 Rename fields of cli_interp_base::saved_output_files
5998 This renames the fields of cli_interp_base::saved_output_files, as
5999 requested by Simon. I tried to choose names that more obviously
6000 reflect what the field is used for. I also added a couple of
6003 Approved-By: Simon Marchi <simon.marchi@efficios.com>
6005 2022-11-30 Tom de Vries <tdevries@suse.de>
6007 [gdb] Improve printing of float formats
6008 Currently, on x86_64, a little endian target, I get:
6010 $ gdb -q -batch -ex "maint print architecture" | grep " = floatformat"
6011 gdbarch_dump: bfloat16_format = floatformat_bfloat16_big
6012 gdbarch_dump: double_format = floatformat_ieee_double_big
6013 gdbarch_dump: float_format = floatformat_ieee_single_big
6014 gdbarch_dump: half_format = floatformat_ieee_half_big
6015 gdbarch_dump: long_double_format = floatformat_i387_ext
6017 which suggests big endian.
6019 This is due to this bit of code in pformat:
6021 /* Just print out one of them - this is only for diagnostics. */
6022 return format[0]->name;
6025 Fix this by using gdbarch_byte_order to pick the appropriate index, such that
6026 we have the more accurate:
6028 gdbarch_dump: bfloat16_format = floatformat_bfloat16_little
6029 gdbarch_dump: half_format = floatformat_ieee_half_little
6030 gdbarch_dump: float_format = floatformat_ieee_single_little
6031 gdbarch_dump: double_format = floatformat_ieee_double_little
6032 gdbarch_dump: long_double_format = floatformat_i387_ext
6035 Tested on x86_64-linux.
6037 2022-11-30 Alan Modra <amodra@gmail.com>
6039 Correct ordering problem in comm-data.exp
6040 * testsuite/ld-elf/comm-data.exp: Build libcomm-data.so before
6041 attempting to read it to set ELF64.
6043 regen SRC-POTFILES.in
6045 2022-11-30 Jan Beulich <jbeulich@suse.com>
6047 x86/Intel: adjustment to restricted suffix derivation
6048 In "x86/Intel: restrict suffix derivation" I think I screwed up
6049 slightly, bringing a piece of code out of sync with its comment, and
6050 resulting in a suffix potentially being derived when one isn't needed.
6052 2022-11-30 Jan Beulich <jbeulich@suse.com>
6054 x86: clean up after removal of support for gcc <= 2.8.1
6055 At the very least a comment in process_operands() is stale. Beyond that
6056 there are effectively two options:
6057 1) It is possible that FADDP and FMULP were mistakenly not marked as
6058 being in need of dealing with the compiler anomaly, and hence the
6059 respective templates weren't removed at the time when they should
6061 2) It is also possible that there are indeed uses known beyond compiler
6062 generated output for these two commutative opcodes, and hence the
6063 templates need to stay.
6064 To be on the safe side assume 2: Update the comment and fold the
6065 templates into their "normal" ones (utilizing D), adjusting consuming
6068 For FMULP also add a comment paralleling a similar one FADDP has.
6070 2022-11-30 Jan Beulich <jbeulich@suse.com>
6073 There are just 4 templates using it, which can be easily identified by
6074 other means, as D is set only on a very limited number of FPU templates.
6075 Also move the respective conditional out of the code path taken by all
6076 "reverse match" insns (it probably should have been this way already
6077 before, to avoid the one conditional in the common case).
6079 With this the templates which had FloatR dropped no longer differ from
6080 their AT&T syntax + mnemonic counterparts - the only difference is now
6081 which of the two would be recognized. For this, however, we don't need
6082 two templates - we can simply arrange the condition for setting
6083 Opcode_FloatR accordingly.
6085 2022-11-30 Jan Beulich <jbeulich@suse.com>
6087 x86: extend FPU test coverage for AT&T / Intel mnemonic differences
6088 Before touching the templates, let's ensure we actually cover things:
6089 For one FSUB{,R} and FDIV{,R} would better be tested with operands in
6090 both possible orders. And then -mmnemonic=intel wasn't tested at all.
6092 2022-11-30 Joel Brobecker <brobecker@adacore.com>
6094 src-release.sh: Fix gdb source tarball build failure due to libsframe
6095 This script was recently changed as follow:
6097 | commit e619dddb3a45780ae66d762756882a3b896b617d
6098 | Date: Tue Nov 15 15:07:13 2022 -0800
6099 | Subject: src-release.sh: Add libsframe
6101 | Add libsframe to the list of top level directories that will be included
6104 Since then, the gdb source tarball has been failing with the error
6105 below during the "make configure-host configure-target" phase:
6107 | make[3]: *** No rule to make target '../libsframe/libsframe.la',
6108 | needed by 'libbfd.la'. Stop.
6109 | make[3]: Leaving directory '/tmp/gdb-public/bfd'
6111 This patch fixes the issue by adding libsframe to the list of
6112 GDB_SUPPORT_DIRS, similar to what was done for BINUTILS.
6116 * src-release.sh (GDB_SUPPORT_DIRS): Add libsframe.
6118 2022-11-30 GDB Administrator <gdbadmin@sourceware.org>
6120 Automatic date update in version.in
6122 2022-11-29 Tom de Vries <tdevries@suse.de>
6124 [gdb/testsuite] Fix gdb.base/vla-optimized-out.exp for ppc64le
6125 On powerpc64le-linux, I run into:
6127 (gdb) PASS: gdb.base/vla-optimized-out.exp: o1: printed optimized out vla
6129 $2 = <optimized out>^M
6130 (gdb) FAIL: gdb.base/vla-optimized-out.exp: o1: \
6131 printed size of optimized out vla
6134 The problem happens as follows.
6136 In order to find the size of the optimized out vla, gdb needs to evaluate:
6138 <155> DW_AT_upper_bound : 13 byte block: f3 1 53 23 1 8 20 24 8 20 26 31 1c \
6139 (DW_OP_GNU_entry_value: (DW_OP_reg3 (r3)); DW_OP_plus_uconst: 1;
6140 DW_OP_const1u: 32; DW_OP_shl; DW_OP_const1u: 32; DW_OP_shra; DW_OP_lit1;
6144 When trying to evaluate DW_OP_GNU_entry_value, it looks for a call site
6145 matching the pc, but doesn't find it:
6147 $ gdb -q -batch outputs/gdb.base/vla-optimized-out/vla-optimized-out-o1 \
6148 -ex "break f1" -ex run -ex "set debug entry-values 1" -ex "print sizeof (a)"
6149 Breakpoint 1 at 0x1000067c: file vla-optimized-out.c, line 34.
6151 Breakpoint 1, f1 (i=5) at vla-optimized-out.c:34
6153 DW_OP_entry_value resolving cannot find DW_TAG_call_site 0x100006b0 in main
6154 $1 = <optimized out>
6157 The call site lookup fails because the call site label .LVL4:
6159 bl f1 # 11 *call_value_nonlocal_aixdi [length = 8]
6163 is not placed directly after the bl insn. This is gcc PR target/107909.
6165 However, after manually fixing the .s file we have instead:
6167 Cannot find matching parameter at DW_TAG_call_site 0x10000690 at main
6168 $1 = <optimized out>
6170 due to the fact that the call site has no call site parameters.
6172 The call site does have a reference to the corresponding function f1, with
6173 parameter i, for which we find location list entries:
6175 0037 1000067c 10000680 (DW_OP_reg3 (r3))
6176 004a 10000680 10000690 (DW_OP_GNU_entry_value: (DW_OP_reg3 (r3));
6179 and we could use the fact that the current pc is in the 1000067c-10000680
6180 range, and that that the range starts at the start of the function, to deduce
6181 that DW_OP_GNU_entry_value: (DW_OP_reg3 (r3)) == DW_OP_reg3 (r3).
6182 But that's a non-trivial enhancement, filed as enhancement PR symtab/29836.
6184 Fix this by allowing <optimized out> for target powerpc and the gcc compiler.
6186 Reviewed-By: Carl Love <cel@us.ibm.com>
6187 Tested-By: Carl Love <cel@us.ibm.com>
6189 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29813
6191 2022-11-29 Simon Marchi <simon.marchi@polymtl.ca>
6193 gdb/testsuite: make gdb_unload use gdb_test_multiple
6194 In the failure seen by Philippe here:
6196 https://inbox.sourceware.org/gdb-patches/20221120173024.3647464-1-philippe.waroquiers@skynet.be/
6198 gdb_unload crashed GDB, leaving no trace in the test results. Change it
6199 to use gdb_test_multiple, so that it leaves an UNRESOLVED result. I
6200 think it is good practice anyway.
6202 Make it return the result of gdb_test_multiple directly, change
6203 gdb.python/py-objfile.exp accordingly.
6205 Change gdb.base/endian.exp as well to avoid duplicate test names.
6207 Change gdb.base/gnu-debugdata.exp to avoid recording a test result,
6208 since gdb_unload does it already now.
6210 Change-Id: I59a1e4947691330797e6ce23277942547c437a48
6211 Approved-By: Tom de Vries <tdevries@suse.de>
6213 2022-11-29 Simon Marchi <simon.marchi@polymtl.ca>
6215 gdb/testsuite: make gdb_test_multiple return immediately if send_gdb fails
6216 In the failure seen by Philippe here:
6218 https://inbox.sourceware.org/gdb-patches/20221120173024.3647464-1-philippe.waroquiers@skynet.be/
6220 ... the testsuite only outputs PASSes, and an ERROR, resulting from an
6221 uncaught exception. This is a bit sneaky, because ERRORs are not
6222 reported in the test summary. In certain circumstances, it can be easy
6225 Normally, gdb_test_multiple outputs an UNRESOLVED when GDB crashes. But
6226 this is only if it manages to send the command, and it's that command
6227 that crashes GDB. Here, the ERROR is due to the fact that GDB had
6228 already crashed by the time we entered gdb_test_multiple and tried to
6229 send a command. GDB was crashed by the previous "file" command, sent by
6230 gdb_unload. Because gdb_unload uses bare expect, it didn't record a
6231 test failure when crashing GDB (this will be addressed separately).
6233 In this patch, I propose to make gdb_test_multiple call unresolved
6234 directly and return -1 send_gdb fails. This way, if GDB is already
6235 crashed by the time we enter gdb_test_multiple, it will leave a trace in
6236 the test results in the form of an UNRESOLVED. It will also spare us
6237 the not-so-useful-in-my-opinion TCL backtrace.
6239 Before, it looks like:
6241 ERROR: Couldn't send python print(objfile.filename) to GDB.
6242 ERROR: : spawn id exp9 not open
6246 -re ".*A problem internal to GDB has been detected" {
6247 fail "$message (GDB internal error)"
6248 gdb_internal_error..."
6249 ("uplevel" body line 1)
6251 "uplevel $body" NONE : spawn id exp9 not open
6255 Couldn't send python print(objfile.filename) to GDB.
6256 UNRESOLVED: gdb.python/py-objfile.exp: objfile.filename after objfile is unloaded
6258 Change-Id: I72af8dc0d687826fc3f76911c27a9e5f91b677ba
6259 Approved-By: Tom de Vries <tdevries@suse.de>
6261 2022-11-29 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6263 gprofng: remove unused gprofng/src/DbeSession.cc.1
6265 2022-11-29 Max Filippov <jcmvbkbc@gmail.com>
6267 xtensa: allow dynamic configuration
6268 Import include/xtensa-dynconfig.h that defines XCHAL_* macros as fields
6269 of a structure returned from the xtensa_get_config_v<x> function call.
6270 Define that structure and fill it with default parameter values
6271 specified in the include/xtensa-config.h.
6272 Define reusable function xtensa_load_config that tries to load
6273 configuration and return an address of an exported object from it.
6274 Define functions xtensa_get_config_v{1,2} that use xtensa_load_config
6275 to get structures xtensa_config_v{1,2}, either dynamically configured
6279 * Makefile.am (BFD32_BACKENDS, BFD32_BACKENDS_CFILES): Append
6281 * Makefile.in: Regenerate.
6282 * configure: Regenerate.
6283 * configure.ac (xtensa_elf32_be_vec, xtensa_elf32_le_vec): Add
6284 xtensa-dynconfig.lo to the tb.
6285 * elf32-xtensa.c (xtensa-config.h): Replace #include with
6287 (XSHAL_ABI, XTHAL_ABI_WINDOWED, XTHAL_ABI_CALL0): Remove
6289 * xtensa-dynconfig.c: New file.
6290 * xtensa-isa.c (xtensa-dynconfig.h): New #include.
6291 (xtensa_get_modules): New function.
6292 (xtensa_isa_init): Call xtensa_get_modules instead of taking
6293 address of global xtensa_modules.
6296 * config/tc-xtensa.c (xtensa-config.h): Replace #include with
6298 (XTHAL_ABI_WINDOWED, XTHAL_ABI_CALL0, XTENSA_MARCH_EARLIEST):
6300 * config/tc-xtensa.h (xtensa-config.h): Replace #include with
6302 * config/xtensa-relax.c (xtensa-config.h): Replace #include with
6304 (XCHAL_HAVE_WIDE_BRANCHES): Remove definition.
6307 * xtensa-dynconfig.h: New file.
6310 * emultempl/xtensaelf.em (xtensa-config.h): Replace #include
6311 with xtensa-dynconfig.h.
6312 (XTHAL_ABI_WINDOWED, XTHAL_ABI_CALL0): Remove definitions.
6314 2022-11-29 GDB Administrator <gdbadmin@sourceware.org>
6316 Automatic date update in version.in
6318 2022-11-28 Andrew Burgess <aburgess@redhat.com>
6320 gdb/testsuite: remove use of then keyword from library files
6321 The canonical form of 'if' in modern TCL is 'if {} {}'. But there's
6322 still a bunch of places in the testsuite where we make use of the
6323 'then' keyword, and sometimes these get copies into new tests, which
6324 just spreads poor practice.
6326 This commit removes all use of the 'then' keyword from the testsuite
6327 library files (in boards/, config/, and lib/). Previous commits have
6328 removed all uses of the 'then' keyword from the test script files,
6329 this commit just cleans up the library files.
6331 There should be no changes in what is tested after this commit.
6333 2022-11-28 Andrew Burgess <aburgess@redhat.com>
6335 gdb/testsuite: remove use of then keyword from gdb.*/*.exp scripts
6336 The canonical form of 'if' in modern TCL is 'if {} {}'. But there's
6337 still a bunch of places in the testsuite where we make use of the
6338 'then' keyword, and sometimes these get copies into new tests, which
6339 just spreads poor practice.
6341 This commit removes all use of the 'then' keyword from the remaining
6342 gdb.*/*.exp scripts. Previous commits have done the bulk of this
6343 removal, this commit just handles the remaining directories that each
6344 contain a low number of instances.
6346 There should be no changes in what is tested after this commit.
6348 2022-11-28 Andrew Burgess <aburgess@redhat.com>
6350 gdb/testsuite: remove use of then keyword from gdb.multi/*.exp
6351 The canonical form of 'if' in modern TCL is 'if {} {}'. But there's
6352 still a bunch of places in the testsuite where we make use of the
6353 'then' keyword, and sometimes these get copies into new tests, which
6354 just spreads poor practice.
6356 This commit removes all use of the 'then' keyword from the gdb.multi/
6357 test script directory.
6359 There should be no changes in what is tested after this commit.
6361 2022-11-28 Andrew Burgess <aburgess@redhat.com>
6363 gdb/testsuite: remove use of then keyword from gdb.fortran/*.exp
6364 The canonical form of 'if' in modern TCL is 'if {} {}'. But there's
6365 still a bunch of places in the testsuite where we make use of the
6366 'then' keyword, and sometimes these get copies into new tests, which
6367 just spreads poor practice.
6369 This commit removes all use of the 'then' keyword from the gdb.fortran/
6370 test script directory.
6372 There should be no changes in what is tested after this commit.
6374 2022-11-28 Andrew Burgess <aburgess@redhat.com>
6376 gdb/testsuite: remove use of then keyword from gdb.disasm/*.exp
6377 The canonical form of 'if' in modern TCL is 'if {} {}'. But there's
6378 still a bunch of places in the testsuite where we make use of the
6379 'then' keyword, and sometimes these get copies into new tests, which
6380 just spreads poor practice.
6382 This commit removes all use of the 'then' keyword from the gdb.disasm/
6383 test script directory.
6385 There should be no changes in what is tested after this commit.
6387 2022-11-28 Andrew Burgess <aburgess@redhat.com>
6389 gdb/testsuite: remove use of then keyword from gdb.reverse/*.exp
6390 The canonical form of 'if' in modern TCL is 'if {} {}'. But there's
6391 still a bunch of places in the testsuite where we make use of the
6392 'then' keyword, and sometimes these get copies into new tests, which
6393 just spreads poor practice.
6395 This commit removes all use of the 'then' keyword from the gdb.reverse/
6396 test script directory.
6398 There should be no changes in what is tested after this commit.
6400 2022-11-28 Andrew Burgess <aburgess@redhat.com>
6402 gdb/testsuite: remove use of then keyword from gdb.trace/*.exp
6403 The canonical form of 'if' in modern TCL is 'if {} {}'. But there's
6404 still a bunch of places in the testsuite where we make use of the
6405 'then' keyword, and sometimes these get copies into new tests, which
6406 just spreads poor practice.
6408 This commit removes all use of the 'then' keyword from the gdb.trace/
6409 test script directory.
6411 There should be no changes in what is tested after this commit.
6413 2022-11-28 Andrew Burgess <aburgess@redhat.com>
6415 gdb/testsuite: remove use of then keyword from gdb.threads/*.exp
6416 The canonical form of 'if' in modern TCL is 'if {} {}'. But there's
6417 still a bunch of places in the testsuite where we make use of the
6418 'then' keyword, and sometimes these get copies into new tests, which
6419 just spreads poor practice.
6421 This commit removes all use of the 'then' keyword from the gdb.threads/
6422 test script directory.
6424 There should be no changes in what is tested after this commit.
6426 2022-11-28 Andrew Burgess <aburgess@redhat.com>
6428 gdb/testsuite: remove use of then keyword from gdb.python/*.exp
6429 The canonical form of 'if' in modern TCL is 'if {} {}'. But there's
6430 still a bunch of places in the testsuite where we make use of the
6431 'then' keyword, and sometimes these get copies into new tests, which
6432 just spreads poor practice.
6434 This commit removes all use of the 'then' keyword from the gdb.python/
6435 test script directory.
6437 There should be no changes in what is tested after this commit.
6439 2022-11-28 Andrew Burgess <aburgess@redhat.com>
6441 gdb/testsuite: remove use of then keyword from gdb.cp/*.exp
6442 The canonical form of 'if' in modern TCL is 'if {} {}'. But there's
6443 still a bunch of places in the testsuite where we make use of the
6444 'then' keyword, and sometimes these get copies into new tests, which
6445 just spreads poor practice.
6447 This commit removes all use of the 'then' keyword from the gdb.cp/
6448 test script directory.
6450 There should be no changes in what is tested after this commit.
6452 2022-11-28 Andrew Burgess <aburgess@redhat.com>
6454 gdb/testsuite: remove use of then keyword from gdb.arch/*.exp
6455 The canonical form of 'if' in modern TCL is 'if {} {}'. But there's
6456 still a bunch of places in the testsuite where we make use of the
6457 'then' keyword, and sometimes these get copies into new tests, which
6458 just spreads poor practice.
6460 This commit removes all use of the 'then' keyword from the gdb.arch/
6461 test script directory.
6463 There should be no changes in what is tested after this commit.
6465 2022-11-28 Andrew Burgess <aburgess@redhat.com>
6467 gdb/testsuite: remove use of then keyword from gdb.base/*.exp
6468 The canonical form of 'if' in modern TCL is 'if {} {}'. But there's
6469 still a bunch of places in the testsuite where we make use of the
6470 'then' keyword, and sometimes these get copies into new tests, which
6471 just spreads poor practice.
6473 This commit removes all use of the 'then' keyword from the gdb.base/
6474 test script directory.
6476 There should be no changes in what is tested after this commit.
6478 2022-11-28 Andrew Burgess <aburgess@redhat.com>
6480 gdb/testsuite: remove use of then keyword from gdb.ada/*.exp
6481 The canonical form of 'if' in modern TCL is 'if {} {}'. But there's
6482 still a bunch of places in the testsuite where we make use of the
6483 'then' keyword, and sometimes these get copies into new tests, which
6484 just spreads poor practice.
6486 This commit removes all use of the 'then' keyword from the gdb.ada/
6487 test script directory.
6489 There should be no changes in what is tested after this commit.
6491 2022-11-28 Andrew Burgess <aburgess@redhat.com>
6493 gdb/testsuite: remove DOS line endings from a test script
6494 The gdb.fortran/nested-funcs.exp test script has DOS line endings. I
6495 can see no reason why this script needs DOS line endings.
6497 Convert to UNIX line endings.
6499 There should be no change in what is tested after this commit.
6501 2022-11-28 Tom Tromey <tromey@adacore.com>
6503 Don't let gdb_stdlog use pager
6504 When using the "set logging" commands, cli_interp_base::set_logging
6505 will send gdb_stdlog output (among others) to the tee it makes for
6506 gdb_stdout. However, this has the side effect of also causing logging
6507 to use the pager. This is PR gdb/29787.
6509 This patch fixes the problem by keeping stderr and stdlog separate
6510 from stdout, preserving the rule that only gdb_stdout should page.
6512 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29787
6514 2022-11-28 Tom Tromey <tromey@adacore.com>
6516 Don't let tee_file own a stream
6517 Right now, tee_file owns the second stream it writes to. This is done
6518 for the convenience of the users. In a subsequent patch, this will no
6519 longer be convenient, so this patch moves the responsibility for
6520 ownership to the users of tee_file.
6522 Remove 'saved_output' global
6523 CLI redirect uses a global variable, 'saved_output'. However, globals
6524 are generally bad, and there is no need for this one -- it can be a
6525 member of cli_interp_base. This patch makes this change.
6527 2022-11-28 Hannes Domani <ssbssa@yahoo.de>
6529 Remove no longer used jump label
6530 The out label is unused since wait_for_debug_event is in a different thread.
6532 Actually set m_is_async to current async mode
6533 Looks like this was missed in the async mode implementation.
6535 2022-11-28 Hannes Domani <ssbssa@yahoo.de>
6537 Don't use auto for lambda parameter
6538 Older gcc versions (here 4.9.2) can't handle auto for a lambda parameter:
6540 ../../gdb/windows-nat.c: In member function 'void windows_nat_target::delete_thread(ptid_t, DWORD, bool)':
6541 ../../gdb/windows-nat.c:629:12: error: use of 'auto' in lambda parameter declaration only available with -std=c++1y or -std=gnu++1y [-Werror]
6545 2022-11-28 Hannes Domani <ssbssa@yahoo.de>
6547 Fix calling convention of thread entry point
6548 For i686 the CreateThread entry point function needs the WINAPI (stdcall)
6551 ../../gdb/windows-nat.c: In constructor 'windows_nat_target::windows_nat_target()':
6552 ../../gdb/windows-nat.c:450:56: error: invalid user-defined conversion from 'windows_nat_target::windows_nat_target()::<lambda(LPVOID)>' to 'LPTHREAD_START_ROUTINE' {aka 'long unsigned int (__attribute__((stdcall)) *)(void*)'} [-fpermissive]
6553 450 | HANDLE bg_thread = CreateThread (nullptr, 64 * 1024, fn, this, 0, nullptr);
6555 ../../gdb/windows-nat.c:444:13: note: candidate is: 'constexpr windows_nat_target::windows_nat_target()::<lambda(LPVOID)>::operator DWORD (*)(LPVOID)() const' (near match)
6556 444 | auto fn = [] (LPVOID self) -> DWORD
6558 ../../gdb/windows-nat.c:444:13: note: no known conversion from 'DWORD (*)(LPVOID)' {aka 'long unsigned int (*)(void*)'} to 'LPTHREAD_START_ROUTINE' {aka 'long unsigned int (__attribute__((stdcall)) *)(void*)'}
6560 Since it's not possible to change the calling convention of a lambda, I've
6561 moved it to a separate function.
6563 2022-11-28 Andrew Burgess <aburgess@redhat.com>
6565 gdb: mark disassembler function callback types as noexcept
6566 In disasm.h we define a set of types that are used by the various
6567 disassembler classes to hold callback functions before passing the
6568 callbacks into libopcodes.
6570 Because libopcodes is C code, and on some (many?) targets, C code is
6571 compiled without exception support, it is important that GDB not try
6572 to throw an exception over libopcode code.
6574 In the previous commit all the existing callbacks were marked as
6575 noexcept, however, this doesn't protect us from a future change to GDB
6576 either adding a new callback that is not noexcept, or removing the
6577 noexcept keyword from an existing callback.
6579 In this commit I mark all the callback types as noexcept. This means
6580 that GDB's disassembler classes will no longer compile if we try to
6581 pass a callback that is not marked as noexcept.
6583 At least, that's the idea. Unfortunately, it's not that easy.
6585 Before C++17, the noexcept keyword on a function typedef would be
6588 using func_type = void (*) (void) noexcept;
6597 some_func (func_type f)
6609 Will compile just fine for C++11 and C++14 with GCC. Clang on the
6610 other hand complains that 'noexcept' should not appear on function
6611 types, but then does appear to correctly complain that passing a_func
6612 is a mismatch in the set of exceptions that could be thrown.
6614 Switching to C++17 and both GCC and Clang correctly point out that
6615 passing a_func is an invalid conversion relating to the noexcept
6616 keyword. Changing a_func to:
6622 And for C++17 both GCC and Clang compile this just fine.
6624 My conclusion then is that adding the noexcept keyword to the function
6625 types is pointless while GDB is not compiled as C++17, and silencing
6626 the warnings would require us to jump through a bunch of hoops.
6628 And so, in this commit, I define a macro LIBOPCODE_CALLBACK_NOEXCEPT,
6629 this macro expands to noexcept when compiling for C++17, but otherwise
6630 expands to nothing. I then add this macro to the function types.
6632 I've compiled GDB as the default C++11 and also forced the compile to
6633 C++17. When compiling as C++17 I spotted a few additional places
6634 where callbacks needed to be marked noexcept (these fixes were merged
6635 into the previous commit, but this confirmed to be that the macro is
6636 working as expected).
6638 2022-11-28 Andrew Burgess <aburgess@redhat.com>
6640 gdb/disasm: mark functions passed to the disassembler noexcept
6641 While working on another patch, Simon pointed out that GDB could be
6642 improved by marking the functions passed to the disassembler as
6645 https://sourceware.org/pipermail/gdb-patches/2022-October/193084.html
6647 The reason this is important is the on some hosts, libopcodes, being C
6648 code, will not be compiled with support for handling exceptions. As
6649 such, an attempt to throw an exception over libopcodes code will cause
6652 See bug gdb/29712 for an example of when this happened.
6654 In this commit all the functions that are passed to the disassembler,
6655 and which might be used as callbacks by libopcodes are marked
6658 Ideally, I would have liked to change these typedefs:
6660 using read_memory_ftype = decltype (disassemble_info::read_memory_func);
6661 using memory_error_ftype = decltype (disassemble_info::memory_error_func);
6662 using print_address_ftype = decltype (disassemble_info::print_address_func);
6663 using fprintf_ftype = decltype (disassemble_info::fprintf_func);
6664 using fprintf_styled_ftype = decltype (disassemble_info::fprintf_styled_func);
6666 which are declared in disasm.h, as including the noexcept keyword.
6667 However, when I tried this, I ran into this warning/error:
6669 In file included from ../../src/gdb/disasm.c:25:
6670 ../../src/gdb/disasm.h: In constructor ‘gdb_printing_disassembler::gdb_printing_disassembler(gdbarch*, ui_file*, gdb_disassemble_info::read_memory_ftype, gdb_disassemble_info::memory_error_ftype, gdb_disassemble_info::print_address_ftype)’:
6671 ../../src/gdb/disasm.h:116:3: error: mangled name for ‘gdb_printing_disassembler::gdb_printing_disassembler(gdbarch*, ui_file*, gdb_disassemble_info::read_memory_ftype, gdb_disassemble_info::memory_error_ftype, gdb_disassemble_info::print_address_ftype)’ will change in C++17 because the exception specification is part of a function type [-Werror=noexcept-type]
6672 116 | gdb_printing_disassembler (struct gdbarch *gdbarch,
6673 | ^~~~~~~~~~~~~~~~~~~~~~~~~
6675 So I've left that change out. This does mean that if somebody adds a
6676 new use of the disassembler classes in the future, and forgets to mark
6677 the callbacks as noexcept, this will compile fine. We'll just have to
6678 manually check for that during review.
6680 2022-11-28 Andrew Burgess <aburgess@redhat.com>
6682 gdb/python: avoid throwing an exception over libopcodes code
6683 Bug gdb/29712 identifies a problem with the Python disassembler API.
6684 In some cases GDB will try to throw an exception through the
6685 libopcodes disassembler code, however, not all targets include
6686 exception unwind information when compiling C code, for targets that
6687 don't include this information GDB will terminate when trying to pass
6688 the exception through libopcodes.
6690 To explain what GDB is trying to do, consider the following trivial
6691 use of the Python disassembler API:
6693 class ExampleDisassembler(gdb.disassembler.Disassembler):
6695 class MyInfo(gdb.disassembler.DisassembleInfo):
6696 def __init__(self, info):
6697 super().__init__(info)
6699 def read_memory(self, length, offset):
6700 return super().read_memory(length, offset)
6703 super().__init__("ExampleDisassembler")
6705 def __call__(self, info):
6706 info = self.MyInfo(info)
6707 return gdb.disassembler.builtin_disassemble(info)
6709 This disassembler doesn't add any value, it defers back to GDB to do
6710 all the actual work, but it serves to allow us to discuss the problem.
6712 The problem occurs when a Python exception is raised by the
6713 MyInfo.read_memory method. The MyInfo.read_memory method is called
6714 from the C++ function gdbpy_disassembler::read_memory_func. The C++
6715 stack at the point this function is called looks like this:
6717 #0 gdbpy_disassembler::read_memory_func (memaddr=4198805, buff=0x7fff9ab9d2a8 "\220ӹ\232\377\177", len=1, info=0x7fff9ab9d558) at ../../src/gdb/python/py-disasm.c:510
6718 #1 0x000000000104ba06 in fetch_data (info=0x7fff9ab9d558, addr=0x7fff9ab9d2a9 "ӹ\232\377\177") at ../../src/opcodes/i386-dis.c:305
6719 #2 0x000000000104badb in ckprefix (ins=0x7fff9ab9d100) at ../../src/opcodes/i386-dis.c:8571
6720 #3 0x000000000104e28e in print_insn (pc=4198805, info=0x7fff9ab9d558, intel_syntax=-1) at ../../src/opcodes/i386-dis.c:9548
6721 #4 0x000000000104f4d4 in print_insn_i386 (pc=4198805, info=0x7fff9ab9d558) at ../../src/opcodes/i386-dis.c:9949
6722 #5 0x00000000004fa7ea in default_print_insn (memaddr=4198805, info=0x7fff9ab9d558) at ../../src/gdb/arch-utils.c:1033
6723 #6 0x000000000094fe5e in i386_print_insn (pc=4198805, info=0x7fff9ab9d558) at ../../src/gdb/i386-tdep.c:4072
6724 #7 0x0000000000503d49 in gdbarch_print_insn (gdbarch=0x5335560, vma=4198805, info=0x7fff9ab9d558) at ../../src/gdb/gdbarch.c:3351
6725 #8 0x0000000000bcc8c6 in disasmpy_builtin_disassemble (self=0x7f2ab07f54d0, args=0x7f2ab0789790, kw=0x0) at ../../src/gdb/python/py-disasm.c:324
6727 ### ... snip lots of frames as we pass through Python itself ...
6729 #22 0x0000000000bcd860 in gdbpy_print_insn (gdbarch=0x5335560, memaddr=0x401195, info=0x7fff9ab9e3c8) at ../../src/gdb/python/py-disasm.c:783
6730 #23 0x00000000008995a5 in ext_lang_print_insn (gdbarch=0x5335560, address=0x401195, info=0x7fff9ab9e3c8) at ../../src/gdb/extension.c:939
6731 #24 0x0000000000741aaa in gdb_print_insn_1 (gdbarch=0x5335560, vma=0x401195, info=0x7fff9ab9e3c8) at ../../src/gdb/disasm.c:1078
6732 #25 0x0000000000741bab in gdb_disassembler::print_insn (this=0x7fff9ab9e3c0, memaddr=0x401195, branch_delay_insns=0x0) at ../../src/gdb/disasm.c:1101
6734 So gdbpy_disassembler::read_memory_func is called from the libopcodes
6735 disassembler to read memory, this C++ function then calls into user
6736 supplied Python code to do the work.
6738 If the user supplied Python code raises an gdb.MemoryError exception
6739 indicating the memory read failed, this is fine. The C++ code
6740 converts this exception back into a return value that libopcodes can
6741 understand, and returns to libopcodes.
6743 However, if the user supplied Python code raises some other exception,
6744 what we want is for this exception to propagate through GDB and appear
6745 as if raised by the call to gdb.disassembler.builtin_disassemble. To
6746 achieve this, when gdbpy_disassembler::read_memory_func spots an
6747 unknown Python exception, we must pass the information about this
6748 exception from frame #0 to frame #8 in the above backtrace. Frame #8
6749 is the C++ implementation of gdb.disassembler.builtin_disassemble, and
6750 so it is this function that we want to re-raise the unknown Python
6751 exception, so the user can, if they want, catch the exception in their
6754 The previous mechanism by which the exception was passed was to pack
6755 the details of the Python exception into a C++ exception, then throw
6756 the exception from frame #0, and catch the exception in frame #8,
6757 unpack the details of the Python exception, and re-raise it.
6759 However, this relies on the exception passing through frames #1 to #7,
6760 some of which are in libopcodes, which is C code, and so, might not be
6761 compiled with exception support.
6763 This commit proposes an alternative solution that does not rely on
6764 throwing a C++ exception.
6766 When we spot an unhandled Python exception in frame #0, we will store
6767 the details of this exception within the gdbpy_disassembler object
6768 currently in use. Then we return to libopcodes a value indicating
6769 that the memory_read failed.
6771 libopcodes will now continue to disassemble as though that memory read
6772 failed (with one special case described below), then, when we
6773 eventually return to disasmpy_builtin_disassemble we check to see if
6774 there is an exception stored in the gdbpy_disassembler object. If
6775 there is then this exception can immediately be installed, and then we
6776 return back to Python, when the user will be able to catch the
6779 There is one extra change in gdbpy_disassembler::read_memory_func.
6780 After the first call that results in an exception being stored on the
6781 gdbpy_disassembler object, any future calls to the ::read_memory_func
6782 function will immediately return as if the read failed. This avoids
6783 any additional calls into user supplied Python code.
6785 My thinking here is that should the first call fail with some unknown
6786 error, GDB should not keep trying with any additional calls. This
6787 maintains the illusion that the exception raised from
6788 MyInfo.read_memory is immediately raised by
6789 gdb.disassembler.builtin_disassemble. I have no tests for this change
6790 though - to trigger this issue would rely on a libopcodes disassembler
6791 that will try to read further memory even after the first failed
6792 read. I'm not aware of any such disassembler that currently does
6793 this, but that doesn't mean such a disassembler couldn't exist in the
6796 With this change in place the gdb.python/py-disasm.exp test should now
6799 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29712
6801 Approved-By: Simon Marchi <simon.marchi@efficios.com>
6803 2022-11-28 Tom Tromey <tom@tromey.com>
6805 Remove reset_ecs and execution_control_state::reset
6806 I noticed that execution_control_state has a 'reset' method, and
6807 there's also a 'reset_ecs' function that calls it. This patch cleans
6808 this area up a little by adding a parameter to the constructor and (a
6809 change Simon suggested) removing the reset method. Some extraneous
6810 variables are also removed, like:
6812 - struct execution_control_state ecss;
6813 - struct execution_control_state *ecs = &ecss;
6815 Here 'ecs' is never changed, so this patch removes it entirely in
6816 favor of just using the object everywhere.
6818 Regression tested on x86-64 Fedora 34.
6820 Approved-By: Simon Marchi <simon.marchi@efficios.com>
6822 2022-11-28 Andrew Burgess <aburgess@redhat.com>
6824 gdb: relax requirement for the map_failed stap probe to be present
6825 From glibc 2.35 and later, the "map_failed" stap probe is no longer
6826 included in glibc. The removal of the probe looks like an accident,
6827 but it was caused by a glibc commit which meant that the "map_failed"
6828 probe could no longer be reached; the compiler then helpfully
6829 optimised out the probe.
6831 In GDB, in solib-svr4.c, we have a list of probes that we look for
6832 related to the shared library loading detection. If any of these
6833 probes are missing then GDB will fall back to the non-probe based
6834 mechanism for detecting shared library loading. The "map_failed"
6835 probe is include in the list of required probes.
6837 This means that on glibc 2.35 (or later) systems, GDB is going to
6838 always fall back to the non-probes based mechanism for detecting
6839 shared library loading.
6841 I raised a glibc bug to discuss this issue:
6843 https://sourceware.org/bugzilla/show_bug.cgi?id=29818
6845 But, whatever the ultimate decision from the glibc team, given there
6846 are version of glibc in the wild without the "map_failed" probe, we
6847 probably should update GDB to handle this situation.
6849 The "map_failed" probe is already a little strange, very early
6850 versions of glibc didn't include this probe, so, in some cases, if
6851 this probe is missing GDB is happy to ignore it. This is fine, the
6852 action associated with this probe inside GDB is DO_NOTHING, this means
6853 the probe isn't actually required in order for GDB to correctly detect
6854 the loading of shared libraries.
6856 In this commit I propose changing the rules so that any probe whose
6857 action is DO_NOTHING, is optional.
6859 There is one possible downside to this change, and that concerns 'set
6860 stop-on-solib-events on'. If a probe is removed from glibc, but the
6861 old style breakpoint based mechanism is still in place within glibc
6862 for that same event, then GDB will stop when using the old style
6863 non-probe based mechanism, but not when using the probes based
6866 For the map_failed case this is not a problem, both the map_failed
6867 probe, and the call to the old style breakpoint location were
6868 optimised out, and so neither event (probes based, or breakpoint
6869 based) will trigger. This would only become an issue if glibc removed
6870 a probe, but left the breakpoint in place (this would almost certainly
6873 For now, I'm proposing that we just don't worry about this. Because
6874 some probes have actions that are not DO_NOTHING, then we know the
6875 user will always seem _some_ stops when a shared library is
6876 loaded/unloaded, and (I'm guessing), in most cases, that's all they
6877 care about. I figure when someone complains then we can figure out
6878 what the right solution is then.
6880 With this commit in place, then, when using a glibc 2.35 or later
6881 system, GDB will once again use the stap probes for shared library
6884 Reviewed-By: Lancelot SIX <lancelot.six@amd.com>
6886 2022-11-28 Tom de Vries <tdevries@suse.de>
6888 [gdb/testsuite] Require hw watchpoint in gdb.ada/task_watch.exp
6889 On powerpc64le-linux I run into:
6891 (gdb) PASS: gdb.ada/task_watch.exp: info tasks before inserting breakpoint
6892 watch -location value task 3^M
6893 Watchpoint 2: -location value^M
6894 (gdb) PASS: gdb.ada/task_watch.exp: watch -location value task 3
6897 [Thread 0x7ffff7ccf170 (LWP 65550) exited]^M
6898 [Thread 0x7ffff7abf170 (LWP 65551) exited]^M
6899 FAIL: gdb.ada/task_watch.exp: continue to watchpoint (timeout)
6902 On x86_64-linux (where the test-case passes), a hardware watchpoint is used:
6904 (gdb) PASS: gdb.ada/task_watch.exp: info tasks before inserting breakpoint
6905 watch -location value task 3^M
6906 Hardware watchpoint 2: -location value^M
6908 and after forcing "set can-use-hw-watchpoints 0" we can intermittently
6909 reproduce the same failure.
6911 In the gdb documentation related to watchpoints in multi-threaded programs, we
6914 Warning: In multi-threaded programs, software watchpoints have only limited
6915 usefulness. If GDB creates a software watchpoint, it can only watch the value
6916 of an expression in a single thread. If you are confident that the expression
6917 can only change due to the current thread’s activity (and if you are also
6918 confident that no other thread can become current), then you can use software
6919 watchpoints as usual. However, GDB may not notice when a non-current thread’s
6920 activity changes the expression. (Hardware watchpoints, in contrast, watch an
6921 expression in all threads.)
6924 Since the ada task construct is mapped onto threads, it seems that the
6925 same limitation holds for tasks.
6927 Fix this by using skip_hw_watchpoint_tests.
6929 Tested on powerpc64-linux.
6930 Tested-By: Carl Love <cel@us.ibm.com>
6932 2022-11-28 Tom de Vries <tdevries@suse.de>
6934 [gdb/testsuite] Fix gdb.ada/out_of_line_in_inlined.exp for ppc64le
6935 On powerpc64le-linux, with test-case gdb.ada/out_of_line_in_inlined.exp I run
6939 Starting program: foo_o224_021-all ^M
6941 Breakpoint 1, 0x0000000010002f48 in foo_o224_021.child1.child2 (s=...) at \
6942 foo_o224_021.adb:24^M
6943 24 function Child2 (S : String) return Boolean is -- STOP^M
6944 (gdb) FAIL: gdb.ada/out_of_line_in_inlined.exp: scenario=all: \
6945 run to foo_o224_021.child1.child2
6948 The breakpoint is correctly set at the local entry point, and given that the
6949 local entry point doesn't correspond to a line number entry, the instruction
6950 address of the breakpoint is shown.
6952 The problem is that test-case doesn't expect the breakpoint address.
6954 Fix this by allowing the breakpoint address to occur.
6956 Tested on powerpc64le-linux.
6958 2022-11-28 Michael Matz <matz@suse.de>
6960 Only use wild_sort_fast
6961 there's no reason why the tree-based variant can't always be used
6962 when sorting is required, it merely needs to also support filename
6963 sorting and have a fast path for insertion at end (aka rightmost tree
6966 The filename sorting isn't tested anywhere and the only scripttempl
6967 that uses it is avr (for 'SORT(*)(.ctors)'), and I believe even there it
6968 was a mistake. Either way, this adds a testcase for filename sorting as
6971 Then the non-BST based sorting can be simplified to only support
6972 the fast case of no sorting required at all (at the same time renaming
6973 the two variants to _sort and _nosort).
6975 2022-11-28 Michael Matz <matz@suse.de>
6977 Special case more simple patterns
6978 fnmatch is slow, so avoiding it in more cases is good. This implements
6979 a more generic version of match_simple_wild which needs some
6980 pre-processing of patterns. In particular it supports patterns of the
6981 form PREFIX*SUFFIX (where all parts are optional), i.e. a super set of
6982 what's handled now. Most section matchers of this form and hence don't
6983 need any calls to fnmatch anymore.
6985 We retain the implementation of match_simple_wild for the filename
6986 matchers (they aren't called often enough to matter).
6988 2022-11-28 Simon Marchi <simon.marchi@polymtl.ca>
6990 gdb/testsuite: fail if gdb_start_cmd fails
6991 I broke gdb.ada/start.exp, and did not notice it, because it outputs an
6992 UNTESTED if gdb_start_cmd fails. I don't really see when start would
6993 fail and it's not a problem that should be looked at. Change all spots
6994 that call untested after a gdb_start_cmd failure, use fail instead.
6996 Doing so caused some failures with the native-gdbserver board. Some
6997 tests that use "start" were relying on the fact that start would fail
6998 with that board to just return with "untested". Change them to add an
6999 early return if use_gdb_stub returns true.
7001 Some gdb.pascal tests also failed with native-gdbserver, because they
7002 did use gdb_start_cmd to start the inferior, for no good reason.
7003 Convert them to use runto_main instead, which does the right thing if
7004 the target is a stub.
7006 A further refactoring could be to make gdb_start_cmd match the expected
7007 breakpoint hit and the prompt, which it doesn't do currently (it leaves
7008 that to the callers, but not all of them do).
7010 Change-Id: I097370851213e798ff29fb6cf8ba25ef7d2be007
7011 Reviewed-By: Bruno Larsen <blarsen@redhat.com>
7012 Approved-By: Andrew Burgess <aburgess@redhat.com>
7014 2022-11-28 Tom Tromey <tromey@adacore.com>
7016 Fix range type signed-ness heuristic
7017 The code to create a range type has a heuristic to decide whether the
7018 range is unsigned. However, this heuristic can fail if the upper
7019 bound of the range has its high bit set, because the test is done
7022 With this patch, if the underlying type of a range is unsigned, then
7023 the range will always be unsigned. A new test is included.
7025 Regression tested on x86-64 Fedora 34. We've also been using this
7026 internally at AdaCore for a while.
7028 2022-11-28 Simon Marchi <simon.marchi@efficios.com>
7030 gdb: disable commit resumed in target_kill
7031 New in this version:
7033 - Better comment in target_kill
7034 - Uncomment line in test to avoid hanging when exiting, when testing on
7035 native-extended-gdbserver
7037 PR 28275 shows that doing a sequence of:
7039 - Run inferior in background (run &)
7040 - kill that inferior
7043 We get into this assertion:
7045 /home/smarchi/src/binutils-gdb/gdb/target.c:2590: internal-error: target_wait: Assertion `!proc_target->commit_resumed_state' failed.
7047 #0 internal_error_loc (file=0x5606b344e740 "/home/smarchi/src/binutils-gdb/gdb/target.c", line=2590, fmt=0x5606b344d6a0 "%s: Assertion `%s' failed.") at /home/smarchi/src/binutils-gdb/gdbsupport/errors.cc:54
7048 #1 0x00005606b6296475 in target_wait (ptid=..., status=0x7fffb9390630, options=...) at /home/smarchi/src/binutils-gdb/gdb/target.c:2590
7049 #2 0x00005606b5767a98 in startup_inferior (proc_target=0x5606bfccb2a0 <the_amd64_linux_nat_target>, pid=3884857, ntraps=1, last_waitstatus=0x0, last_ptid=0x0) at /home/smarchi/src/binutils-gdb/gdb/nat/fork-inferior.c:482
7050 #3 0x00005606b4e6c9c5 in gdb_startup_inferior (pid=3884857, num_traps=1) at /home/smarchi/src/binutils-gdb/gdb/fork-child.c:132
7051 #4 0x00005606b50f14a5 in inf_ptrace_target::create_inferior (this=0x5606bfccb2a0 <the_amd64_linux_nat_target>, exec_file=0x604000039f50 "/home/smarchi/build/binutils-gdb/gdb/test", allargs="", env=0x61500000a580, from_tty=1)
7052 at /home/smarchi/src/binutils-gdb/gdb/inf-ptrace.c:105
7053 #5 0x00005606b53b6d23 in linux_nat_target::create_inferior (this=0x5606bfccb2a0 <the_amd64_linux_nat_target>, exec_file=0x604000039f50 "/home/smarchi/build/binutils-gdb/gdb/test", allargs="", env=0x61500000a580, from_tty=1)
7054 at /home/smarchi/src/binutils-gdb/gdb/linux-nat.c:978
7055 #6 0x00005606b512b79b in run_command_1 (args=0x0, from_tty=1, run_how=RUN_NORMAL) at /home/smarchi/src/binutils-gdb/gdb/infcmd.c:468
7056 #7 0x00005606b512c236 in run_command (args=0x0, from_tty=1) at /home/smarchi/src/binutils-gdb/gdb/infcmd.c:526
7058 When running the kill command, commit_resumed_state for the
7059 process_stratum_target (linux-nat, here) is true. After the kill, when
7060 there are no more threads, commit_resumed_state is still true, as
7061 nothing touches this flag during the kill operation. During the
7062 subsequent run command, run_command_1 does:
7064 scoped_disable_commit_resumed disable_commit_resumed ("running");
7066 We would think that this would clear the commit_resumed_state flag of
7067 our native target, but that's not the case, because
7068 scoped_disable_commit_resumed iterates on non-exited inferiors in order
7069 to find active process targets. And after the kill, the inferior is
7070 exited, and the native target was unpushed from it anyway. So
7071 scoped_disable_commit_resumed doesn't touch the commit_resumed_state
7072 flag of the native target, it stays true. When reaching target_wait, in
7073 startup_inferior (to consume the initial expect stop events while the
7074 inferior is starting up and working its way through the shell),
7075 commit_resumed_state is true, breaking the contract saying that
7076 commit_resumed_state is always false when calling the targets' wait
7079 (note: to be correct, I think that startup_inferior should toggle
7080 commit_resumed between the target_wait and target_resume calls, but I'll
7081 ignore that for now)
7083 I can see multiple ways to fix this. In the end, we need
7084 commit_resumed_state to be cleared by the time we get to that
7085 target_wait. It could be done at the end of the kill command, or at the
7086 beginning of the run command.
7088 To keep things in a coherent state, I'd like to make it so that after
7089 the kill command, when the target is left with no threads, its
7090 commit_resumed_state flag is left to false. This way, we can keep
7091 working with the assumption that a target with no threads (and therefore
7092 no running threads) has commit_resumed_state == false.
7094 Do this by adding a scoped_disable_commit_resumed in target_kill. It
7095 clears the target's commit_resumed_state on entry, and leaves it false
7096 if the target does not have any resumed thread on exit. That means,
7097 even if the target has another inferior with stopped threads,
7098 commit_resumed_state will be left to false, which makes sense.
7100 Add a test that tries to cover various combinations of actions done
7101 while an inferior is running (and therefore while commit_resumed_state
7102 is true on the process target).
7104 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28275
7105 Change-Id: I8e6fe6dc1f475055921520e58cab68024039a1e9
7106 Approved-By: Andrew Burgess <aburgess@redhat.com>
7108 2022-11-28 Simon Marchi <simon.marchi@efficios.com>
7110 gdbserver: switch to right process in find_one_thread
7111 New in this version: add a dedicated test.
7115 $ ./gdb -nx --data-directory=data-directory -q \
7117 -ex "maint set target-non-stop on" \
7118 -ex "tar ext :1234" \
7119 -ex "set remote exec-file /bin/sleep" \
7123 Reading symbols from /bin/sleep...
7124 (No debugging symbols found in /bin/sleep)
7125 Remote debugging using :1234
7126 Starting program: /bin/sleep 1231
7127 Reading /lib64/ld-linux-x86-64.so.2 from remote target...
7128 warning: File transfers from remote targets can be slow. Use "set sysroot" to access files locally instead.
7129 Reading /lib64/ld-linux-x86-64.so.2 from remote target...
7130 Reading /usr/lib/debug/.build-id/a6/7a1408f18db3576757eea210d07ba3fc560dff.debug from remote target...
7132 Added inferior 2 on connection 1 (extended-remote :1234)
7133 [Switching to inferior 2 [<null>] (<noexec>)]
7134 (gdb) Reading /lib/x86_64-linux-gnu/libc.so.6 from remote target...
7136 Attaching to process 3659848
7137 /home/smarchi/src/binutils-gdb/gdb/thread.c:85: internal-error: inferior_thread: Assertion `current_thread_ != nullptr' failed.
7139 Note the "attach" command just above. When doing it on the command-line
7140 with a -ex switch, the bug doesn't trigger.
7142 The internal error of GDB is actually caused by GDBserver crashing, and
7143 the error recovery of GDB is not on point. This patch aims to fix just
7144 the GDBserver crash, not the GDB problem.
7146 GDBserver crashes with a segfault here:
7149 #0 0x00005555557fb3f4 in find_one_thread (ptid=...) at /home/smarchi/src/binutils-gdb/gdbserver/thread-db.cc:177
7150 #1 0x00005555557fd5cf in thread_db_thread_handle (ptid=<error reading variable: Cannot access memory at address 0xffffffffffffffa0>, handle=0x7fffffffc400, handle_len=0x7fffffffc3f0)
7151 at /home/smarchi/src/binutils-gdb/gdbserver/thread-db.cc:461
7152 #2 0x000055555578a0b6 in linux_process_target::thread_handle (this=0x5555558a64c0 <the_x86_target>, ptid=<error reading variable: Cannot access memory at address 0xffffffffffffffa0>, handle=0x7fffffffc400,
7153 handle_len=0x7fffffffc3f0) at /home/smarchi/src/binutils-gdb/gdbserver/linux-low.cc:6905
7154 #3 0x00005555556dfcc6 in handle_qxfer_threads_worker (thread=0x60b000000510, buffer=0x7fffffffc8a0) at /home/smarchi/src/binutils-gdb/gdbserver/server.cc:1645
7155 #4 0x00005555556e00e6 in operator() (__closure=0x7fffffffc5e0, thread=0x60b000000510) at /home/smarchi/src/binutils-gdb/gdbserver/server.cc:1696
7156 #5 0x00005555556f54be in for_each_thread<handle_qxfer_threads_proper(buffer*)::<lambda(thread_info*)> >(struct {...}) (func=...) at /home/smarchi/src/binutils-gdb/gdbserver/gdbthread.h:159
7157 #6 0x00005555556e0242 in handle_qxfer_threads_proper (buffer=0x7fffffffc8a0) at /home/smarchi/src/binutils-gdb/gdbserver/server.cc:1694
7158 #7 0x00005555556e04ba in handle_qxfer_threads (annex=0x629000000213 "", readbuf=0x621000019100 '\276' <repeats 200 times>..., writebuf=0x0, offset=0, len=4097)
7159 at /home/smarchi/src/binutils-gdb/gdbserver/server.cc:1732
7160 #8 0x00005555556e1989 in handle_qxfer (own_buf=0x629000000200 "qXfer:threads", packet_len=26, new_packet_len_p=0x7fffffffd630) at /home/smarchi/src/binutils-gdb/gdbserver/server.cc:2045
7161 #9 0x00005555556e720a in handle_query (own_buf=0x629000000200 "qXfer:threads", packet_len=26, new_packet_len_p=0x7fffffffd630) at /home/smarchi/src/binutils-gdb/gdbserver/server.cc:2685
7162 #10 0x00005555556f1a01 in process_serial_event () at /home/smarchi/src/binutils-gdb/gdbserver/server.cc:4176
7163 #11 0x00005555556f4457 in handle_serial_event (err=0, client_data=0x0) at /home/smarchi/src/binutils-gdb/gdbserver/server.cc:4514
7164 #12 0x0000555555820f56 in handle_file_event (file_ptr=0x607000000250, ready_mask=1) at /home/smarchi/src/binutils-gdb/gdbsupport/event-loop.cc:573
7165 #13 0x0000555555821895 in gdb_wait_for_event (block=1) at /home/smarchi/src/binutils-gdb/gdbsupport/event-loop.cc:694
7166 #14 0x000055555581f533 in gdb_do_one_event (mstimeout=-1) at /home/smarchi/src/binutils-gdb/gdbsupport/event-loop.cc:264
7167 #15 0x00005555556ec9fb in start_event_loop () at /home/smarchi/src/binutils-gdb/gdbserver/server.cc:3512
7168 #16 0x00005555556f0769 in captured_main (argc=4, argv=0x7fffffffe0d8) at /home/smarchi/src/binutils-gdb/gdbserver/server.cc:3992
7169 #17 0x00005555556f0e3f in main (argc=4, argv=0x7fffffffe0d8) at /home/smarchi/src/binutils-gdb/gdbserver/server.cc:4078
7171 The reason is a wrong current process when find_one_thread is called.
7172 The current process is the 2nd one, which was just attached. It does
7173 not yet have thread_db data (proc->priv->thread_db is nullptr). As we
7174 iterate on all threads of all process to fulfull the qxfer:threads:read
7175 request, we get to a thread of process 1 for which we haven't read
7176 thread_db information yet (lwp_info::thread_known is false), so we get
7177 into find_one_thread. find_one_thread uses
7178 `current_process ()->priv->thread_db`, assuming the current process
7179 matches the ptid passed as a parameter, which is wrong. A segfault
7180 happens when trying to dereference that thread_db pointer.
7182 Fix this by making find_one_thread not assume what the current process /
7183 current thread is. If it needs to call into libthread_db, which we know
7184 will try to read memory from the current process, then temporarily set
7185 the current process.
7187 In the case where the thread is already know and we return early, we
7188 don't need to switch process.
7190 Add a test to reproduce this specific situation.
7192 Change-Id: I09b00883e8b73b7e5f89d0f47cb4e9c0f3d6caaa
7193 Approved-By: Andrew Burgess <aburgess@redhat.com>
7195 2022-11-28 Tom Tromey <tromey@adacore.com>
7197 Fix crash in "document" command
7198 PR cli/29800 points out that "document" will now crash when the
7199 argument is an undefined command. This is a regression due to the
7200 "document user-defined aliases" patch.
7202 Approved-By: Joel Brobecker <brobecker@adacore.com>
7203 Reviewed-By: Philippe Waroquiers <philippe.waroquiers@skynet.be>
7204 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29800
7206 2022-11-28 Tom de Vries <tdevries@suse.de>
7208 [gdb/testsuite] Fix gdb.opt/solib-intra-step.exp for powerpc64le
7209 On powerpc64le-linux, I run into:
7211 (gdb) PASS: gdb.opt/solib-intra-step.exp: first-hit
7213 28 { /* first-retry */^M
7214 (gdb) FAIL: gdb.opt/solib-intra-step.exp: second-hit
7217 It's a bit easier to understand what happens if we do a full stepping session:
7219 Temporary breakpoint 1, main ()
7220 at solib-intra-step-main.c:23
7223 shlib_first () at solib-intra-step-lib.c:29
7224 29 shlib_second (0); /* first-hit */
7226 28 { /* first-retry */
7228 29 shlib_second (0); /* first-hit */
7230 shlib_second (dummy=0)
7231 at solib-intra-step-lib.c:23
7232 23 abort (); /* second-hit */
7234 and compare that to the line info:
7236 CU: solib-intra-step-lib.c:
7237 File name Line number Starting address View Stmt
7238 solib-intra-step-lib.c 22 0x710 x
7239 solib-intra-step-lib.c 23 0x724 x
7240 solib-intra-step-lib.c 28 0x740 x
7241 solib-intra-step-lib.c 29 0x74c x
7242 solib-intra-step-lib.c 28 0x750 x
7243 solib-intra-step-lib.c 29 0x758 x
7244 solib-intra-step-lib.c 30 0x760 x
7245 solib-intra-step-lib.c - 0x77c
7248 So we step from line 29 to line 28, and back to line 29, which is behaviour
7249 that matches the line table. The peculiar order is due to using optimization.
7250 The problem is that the test-case doesn't expect this order.
7252 Fix this by allowing this order in the test-case.
7254 Tested on powerpc64le-linux.
7257 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29792
7259 2022-11-28 Andrew Burgess <aburgess@redhat.com>
7261 gdb: fix assert when quitting GDB while a thread is stepping
7262 This commit addresses one of the issues identified in PR gdb/28275.
7264 Bug gdb/28275 identifies a number of situations in which this assert:
7266 Assertion `!proc_target->commit_resumed_state' failed.
7268 could be triggered. There's actually a number of similar places where
7269 this assert is found in GDB, the two of interest in gdb/28275 are in
7270 target_wait and target_stop.
7272 In one of the comments:
7274 https://sourceware.org/bugzilla/show_bug.cgi?id=28275#c1
7276 steps to trigger the assertion within target_stop were identified when
7277 using a modified version of the gdb.threads/detach-step-over.exp test
7280 In the gdb.threads/detach-step-over.exp test, we attach to a
7281 multi-threaded inferior, and continue the inferior in asynchronous
7282 (background) mode. Each thread is continuously hitting a conditional
7283 breakpoint where the condition is always false. While the inferior is
7284 running we detach. The goal is that we detach while GDB is performing a
7285 step-over for the conditional breakpoint in at least one thread.
7287 While detaching, if a step-over is in progress, then GDB has to
7288 complete the step over before we can detach. This involves calling
7289 target_stop and target_wait (see prepare_for_detach).
7291 As far as gdb/28275 is concerned, the interesting part here, is the
7292 the process_stratum_target::commit_resumed_state variable must be
7293 false when target_stop and target_wait are called.
7295 This is currently ensured because, in detach_command (infrun.c), we
7296 create an instance of scoped_disable_commit_resumed, this ensures that
7297 when target_detach is called, ::commit_resumed_state will be false.
7299 The modification to the test that I propose here, and which exposed
7300 the bug, is that, instead of using "detach" to detach from the
7301 inferior, we instead use "quit". Quitting GDB after attaching to an
7302 inferior will cause GDB to first detach, and then exit.
7304 When we quit GDB we end up calling target_detach via a different code
7305 path, the stack looks like:
7312 Along this path there is no scoped_disable_commit_resumed created.
7313 ::commit_resumed_state can be true when we reach prepare_for_detach,
7314 which calls target_wait and target_stop, so the assertion will trigger.
7316 In this commit, I propose fixing this by adding the creation of a
7317 scoped_disable_commit_resumed into target_detach. This will ensure
7318 that ::commit_resumed_state is false when calling prepare_for_detach
7319 from within target_detach.
7321 I did consider placing the scoped_disable_commit_resumed in
7322 prepare_for_detach, however, placing it in target_detach ensures that
7323 the target's commit_resumed_state flag is left to false if the detached
7324 inferior was the last one for that target. It's the same rationale as
7325 for patch "gdb: disable commit resumed in target_kill" that comes later
7326 in this series, but for detach instead of kill.
7328 detach_command still includes a scoped_disable_commit_resumed too, but I
7329 think it is still relevant to cover the resumption at the end of the
7332 Co-Authored-By: Simon Marchi <simon.marchi@efficios.com>
7333 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28275
7334 Change-Id: Ie128f7aba6ef0e018859275eca372e6ea738e96f
7336 2022-11-28 Andrew Burgess <aburgess@redhat.com>
7338 gdb/testsuite: refactor gdb.threads/detach-step-over.exp
7339 Factor out some bits of gdb.threads/detach-step-over.exp to procs in
7340 preparation to adding some new variations of the test. Rename the
7341 existing "test" proc and make it use proc_with_prefix.
7343 Co-Authored-By: Simon Marchi <simon.marchi@efficios.com>
7344 Change-Id: Ib4412545c81c8556029e0f7bfa9dd48d7a9f3189
7346 2022-11-28 Simon Marchi <simon.marchi@efficios.com>
7348 gdb/testsuite: remove global declarations in gdb.threads/detach-step-over.exp
7349 Before doing further changes to this file, change to use the :: notation
7350 instead of declaring global variables with the `global` keyword.
7352 Change-Id: I72301fd8f4693fea61aac054ba17245a1f4442fb
7353 Approved-By: Andrew Burgess <aburgess@redhat.com>
7355 2022-11-28 Tom de Vries <tdevries@suse.de>
7357 [gdb/testsuite] Fix gdb.arch/altivec-regs.exp with gcc 4.8.5
7358 On powerpc64le-linux, using gcc 4.8.5, I run into:
7360 (gdb) PASS: gdb.arch/altivec-regs.exp: next (1)
7362 11 c = vec_add (a, b);^M
7363 (gdb) PASS: gdb.arch/altivec-regs.exp: next (2)
7365 $67 = {0xfefefefe, 0xfefefefe, 0xfefefefe, 0xfefefefe}^M
7366 (gdb) FAIL: gdb.arch/altivec-regs.exp: print vector parameter a
7369 Looking at the disassembly and the debug info, it's clear why there's
7372 The debug info says that the variable can be found at some stack location, but
7373 the instructions don't seem to be writing there.
7375 We can work around this by marking variable a volatile. Likewise for b.
7377 Note that marking the variables as volatile doesn't change the location
7380 Tested on power64le-linux.
7382 2022-11-28 Tom de Vries <tdevries@suse.de>
7384 [gdb/tdep] Fix gdb.base/msym-bp-shl.exp for ppc64le
7385 With test-case gdb.base/msym-bp-shl.exp on powerpc64le-linux, I run into:
7387 (gdb) PASS: gdb.base/msym-bp-shl.exp: debug=0: before run: break foo
7389 Num Type Disp Enb Address What^M
7390 1 breakpoint keep y <MULTIPLE> ^M
7391 1.1 y 0x00000000000008d4 <foo+12>^M
7392 1.2 y 0x0000000000000a34 crti.S:88^M
7393 (gdb) FAIL: gdb.base/msym-bp-shl.exp: debug=0: before run: info breakpoint
7396 The problem is that the prologue skipper walks from foo@plt at 0xa28 to 0xa34:
7398 0000000000000a28 <foo@plt>:
7399 a28: c0 ff ff 4b b 9e8 <__glink_PLTresolve>
7401 Disassembly of section .fini:
7403 0000000000000a2c <_fini>:
7404 a2c: 02 00 4c 3c addis r2,r12,2
7405 a30: d4 74 42 38 addi r2,r2,29908
7406 a34: a6 02 08 7c mflr r0
7409 This is caused by ppc_elfv2_elf_make_msymbol_special which marks foo@plt as
7410 having a local entry point, due to incorrectly accessing an asymbol struct
7411 using a (larger) elf_symbol_type.
7413 Fix this by simply ignoring artificial symbols in
7414 ppc_elfv2_elf_make_msymbol_special.
7416 Tested on powerpc64le.
7418 Approved-By: Ulrich Weigand <uweigand@de.ibm.com>
7419 Reviewed-By: Carl Love <cel@us.ibm.com>
7420 Tested-By: Carl Love <cel@us.ibm.com>
7422 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29814
7424 2022-11-28 Alan Modra <amodra@gmail.com>
7426 PR10368, ISO 8859 mentioned as 7bit encoding in strings documentation
7428 * doc/binutils.texi (strings): Delete example of 7-bit encoding.
7430 Use bfd_rename_section in msp430.em
7431 * emultempl/msp430.em (add_region_prefix <REGION_EITHER>): Use
7433 * testsuite/ld-msp430-elf/msp430-tiny-rom.ld: Handle varian data
7434 and bss input sections.
7436 asan: pef: buffer overflow
7437 * pef.c (bfd_pef_parse_traceback_table): Correct size moved when
7438 stripping leading dot.
7440 regen gas/Makefile.in
7442 2022-11-28 Tsukasa OI <research_trasio@irq.a4lg.com>
7444 RISC-V: Allow merging 'H' extension
7445 Because riscv_merge_std_ext function did not merge the 'H' extension, linked
7446 executables lacked 'H' extension when multiple objects are merged.
7448 This issue is found while building OpenSBI with 'H' extension (resulting
7449 ELF files did not contain "h1p0" in "Tag_RISCV_arch" even if *all* linked
7450 object files contained it).
7452 This commit adds 'h' to standard_exts variable to merge 'H' extension.
7456 * elfnn-riscv.c (riscv_merge_std_ext): Add 'H' extension merging.
7458 2022-11-28 Tsukasa OI <research_trasio@irq.a4lg.com>
7460 RISC-V: Better support for long instructions (tests)
7461 This commit tests both (assembler and disassembler) fixes of "Better support
7462 for long instructions".
7466 * testsuite/gas/riscv/insn.s: Add testcases such that big number
7467 handling is required and should be disassembled as long ".byte"
7468 sequence with correct instruction bits.
7469 * testsuite/gas/riscv/insn.d: Likewise.
7470 * testsuite/gas/riscv/insn-na.d: Likewise.
7471 * testsuite/gas/riscv/insn-dwarf.d: Likewise.
7473 2022-11-28 Tsukasa OI <research_trasio@irq.a4lg.com>
7475 RISC-V: Better support for long instructions (assembler)
7476 Commit bb996692bd96 ("RISC-V/gas: allow generating up to 176-bit
7477 instructions with .insn") tried to start supporting long instructions but
7478 it was insufficient.
7480 1. It heavily depended on the bignum internals (radix of 2^16),
7481 2. It generates "value conflicts with instruction length" even if a big
7482 number instruction encoding does not exceed its expected length and
7483 3. Because long opcode was handled separately (from struct riscv_cl_insn),
7484 some information like DWARF line number correspondence was missing.
7486 To resolve these problems, this commit:
7488 1. Handles bignum (and its encodings) precisely and
7489 2. Incorporates long opcode handling into regular instruction handling.
7491 This commit will be tested on the separate commit.
7495 * config/tc-riscv.c (struct riscv_cl_insn): Add long opcode field.
7496 (create_insn) Clear long opcode marker.
7497 (install_insn) Install longer opcode as well.
7498 (s_riscv_insn) Likewise.
7499 (riscv_ip_hardcode): Make big number handling stricter. Length and
7500 the value conflicts only if the bignum size exceeds the expected
7503 2022-11-28 Tsukasa OI <research_trasio@irq.a4lg.com>
7505 RISC-V: Better support for long instructions (disassembler)
7506 Commit bb996692bd96 ("RISC-V/gas: allow generating up to 176-bit
7507 instructions with .insn") tried to start supporting long instructions but
7508 it was insufficient.
7510 On the disassembler, correct ".byte" output was limited to the first 64-bits
7511 of an instruction. After that, zeroes are incorrectly printed.
7513 Note that, it only happens on ".byte" output (instruction part) and not on
7514 hexdump (data) part. For example, before this commit, hexdump and ".byte"
7515 produces different values:
7518 .insn 22, 0xfedcba98765432100123456789abcdef55aa33cc607f
7519 objdump output example (before the fix):
7520 10: 607f 33cc 55aa cdef .byte 0x7f, 0x60, 0xcc, 0x33, 0xaa, 0x55, 0xef, 0xcd, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00
7521 18: 89ab 4567 0123 3210
7524 Note that, after 0xcd (after first 64-bits of the target instruction), all
7525 ".byte" values are incorrectly printed as zero while hexdump prints correct
7528 To resolve this, this commit adds "packet" argument to support dumping
7529 instructions longer than 64-bits (to print correct instruction bits on
7530 ".byte"). This commit will be tested on the separate commit.
7533 .insn 22, 0xfedcba98765432100123456789abcdef55aa33cc607f
7534 objdump output example (after the fix):
7535 10: 607f 33cc 55aa cdef .byte 0x7f, 0x60, 0xcc, 0x33, 0xaa, 0x55, 0xef, 0xcd, 0xab, 0x89, 0x67, 0x45, 0x23, 0x01, 0x10, 0x32, 0x54, 0x76, 0x98, 0xba, 0xdc, 0xfe
7536 18: 89ab 4567 0123 3210
7541 * riscv-dis.c (riscv_disassemble_insn): Print unknown instruction
7542 using the new argument packet.
7543 (riscv_disassemble_data): Add unused argument packet.
7544 (print_insn_riscv): Pass packet to the disassemble function.
7546 2022-11-28 GDB Administrator <gdbadmin@sourceware.org>
7548 Automatic date update in version.in
7550 2022-11-27 Philippe Waroquiers <philippe.waroquiers@skynet.be>
7552 Fix leak in the dwarf reader
7553 Valgrind reports a leak in the dwarf reader (see details below).
7554 The function dw2_get_file_names_reader is interning in the per_objfile
7555 all the file names it finds, except the name of 'fnd file name and directory'.
7556 Instead, it was xstrdup-ing the name.
7557 Fix the leaks by also interning the name.
7559 This was validated running the tests natively, and under valgrind.
7560 Leaks have decreased as mentionned below.
7561 Valgrind detected no error such as double free or use after free.
7563 Stack trace of the leak:
7564 ==4113266== 490,735 bytes in 17,500 blocks are definitely lost in loss record 7,061 of 7,074
7565 ==4113266== at 0x483979B: malloc (vg_replace_malloc.c:393)
7566 ==4113266== by 0x25A454: xmalloc (alloc.c:57)
7567 ==4113266== by 0x7D1E1E: xstrdup (xstrdup.c:34)
7568 ==4113266== by 0x39D141: dw2_get_file_names_reader (read.c:2825)
7569 ==4113266== by 0x39D141: dw2_get_file_names(dwarf2_per_cu_data*, dwarf2_per_objfile*) (read.c:2851)
7570 ==4113266== by 0x39DD6C: dw_expand_symtabs_matching_file_matcher(dwarf2_per_objfile*, gdb::function_view<bool (char const*, bool)>) (read.c:4149)
7571 ==4113266== by 0x3BC8B5: cooked_index_functions::expand_symtabs_matching(objfile*, gdb::function_view<bool (char const*, bool)>, lookup_name_info const*, gdb::function_view<bool (char const*)>, gdb::function_view<bool (compunit_symtab*)>, enum_flags<block_search_flag_values>, domain_enum, search_domain) (read.c:18688)
7572 ==4113266== by 0x5DD1EA: objfile::map_symtabs_matching_filename(char const*, char const*, gdb::function_view<bool (symtab*)>) (symfile-debug.c:207)
7573 ==4113266== by 0x5F04CC: iterate_over_symtabs(char const*, gdb::function_view<bool (symtab*)>) (symtab.c:633)
7574 ==4113266== by 0x477EE3: collect_symtabs_from_filename(char const*, program_space*) (linespec.c:3712)
7575 ==4113266== by 0x477FC1: symtabs_from_filename(char const*, program_space*) (linespec.c:3726)
7576 ==4113266== by 0x47A9B8: convert_explicit_location_spec_to_linespec(linespec_state*, linespec*, char const*, char const*, symbol_name_match_type, char const*, line_offset) (linespec.c:2329)
7577 ==4113266== by 0x47E86E: convert_explicit_location_spec_to_sals (linespec.c:2388)
7578 ==4113266== by 0x47E86E: location_spec_to_sals(linespec_parser*, location_spec const*) (linespec.c:3104)
7579 ==4113266== by 0x47EDAC: decode_line_full(location_spec*, int, program_space*, symtab*, int, linespec_result*, char const*, char const*) (linespec.c:3149)
7582 Without the fix, the top 10 leaks are:
7583 ./gdb/testsuite/outputs/gdb.base/condbreak-bad/gdb.log:345:==3213924== definitely lost: 130,937 bytes in 5,409 blocks
7584 ./gdb/testsuite/outputs/gdb.base/hbreak2/gdb.log:619:==3758919== definitely lost: 173,323 bytes in 7,204 blocks
7585 ./gdb/testsuite/outputs/gdb.mi/mi-var-cp/gdb.log:1320:==4152873== definitely lost: 172,826 bytes in 7,207 blocks
7586 ./gdb/testsuite/outputs/gdb.base/advance-until-multiple-locations/gdb.log:398:==2992643== definitely lost: 172,965 bytes in 7,211 blocks
7587 ./gdb/testsuite/outputs/gdb.mi/mi-var-cmd/gdb.log:2302:==4159476== definitely lost: 173,129 bytes in 7,211 blocks
7588 ./gdb/testsuite/outputs/gdb.cp/gdb2384/gdb.log:222:==3811851== definitely lost: 218,106 bytes in 7,761 blocks
7589 ./gdb/testsuite/outputs/gdb.cp/mb-templates/gdb.log:310:==3787344== definitely lost: 290,311 bytes in 10,340 blocks
7590 ./gdb/testsuite/outputs/gdb.mi/mi-var-rtti/gdb.log:2568:==4158350== definitely lost: 435,427 bytes in 15,507 blocks
7591 ./gdb/testsuite/outputs/gdb.mi/mi-catch-cpp-exceptions/gdb.log:1704:==4119722== definitely lost: 435,405 bytes in 15,510 blocks
7592 ./gdb/testsuite/outputs/gdb.mi/mi-vla-fortran/gdb.log:768:==4113266== definitely lost: 508,585 bytes in 18,109 blocks
7595 ./gdb/testsuite/outputs/gdb.base/fork-running-state/gdb.log:1536:==2924193== indirectly lost: 13,848 bytes in 98 blocks
7596 ./gdb/testsuite/outputs/gdb.base/fork-running-state/gdb.log:1675:==2928777== indirectly lost: 13,848 bytes in 98 blocks
7597 ./gdb/testsuite/outputs/gdb.python/py-inferior-leak/gdb.log:4729:==3353335== definitely lost: 3,360 bytes in 140 blocks
7598 ./gdb/testsuite/outputs/gdb.base/kill-detach-inferiors-cmd/gdb.log:210:==2746927== indirectly lost: 13,246 bytes in 154 blocks
7599 ./gdb/testsuite/outputs/gdb.base/inferior-clone/gdb.log:179:==3034984== indirectly lost: 12,921 bytes in 161 blocks
7600 ./gdb/testsuite/outputs/gdb.base/interrupt-daemon/gdb.log:209:==3006248== indirectly lost: 20,683 bytes in 174 blocks
7601 ./gdb/testsuite/outputs/gdb.threads/watchpoint-fork/gdb.log:714:==3512403== indirectly lost: 20,707 bytes in 175 blocks
7602 ./gdb/testsuite/outputs/gdb.threads/watchpoint-fork/gdb.log:962:==3514498== indirectly lost: 20,851 bytes in 178 blocks
7603 ./gdb/testsuite/outputs/gdb.base/multi-forks/gdb.log:336:==2585839== indirectly lost: 53,630 bytes in 386 blocks
7604 ./gdb/testsuite/outputs/gdb.base/multi-forks/gdb.log:1338:==2592417== indirectly lost: 100,008 bytes in 1,154 blocks
7606 Approved-By: Simon Marchi <simon.marchi@efficios.com>
7608 2022-11-27 Philippe Waroquiers <philippe.waroquiers@skynet.be>
7610 fix leak in gdb_environ
7611 valgrind reports a leak when assigning a gdb_environ to another gdb_environ.
7612 The memory allocated for the target gdb_environ env variables is not released.
7613 The gdb_environ selftest reproduces the leak (see below).
7614 Fix the leak by clearing the target gdb_environ before std::move-ing the
7617 Tested natively and re-running all tests under valgrind.
7619 ==3261873== 4,842 bytes in 69 blocks are definitely lost in loss record 6,772 of 6,839
7620 ==3261873== at 0x483979B: malloc (vg_replace_malloc.c:393)
7621 ==3261873== by 0x25A454: xmalloc (alloc.c:57)
7622 ==3261873== by 0x7D1E4E: xstrdup (xstrdup.c:34)
7623 ==3261873== by 0x7E2A51: gdb_environ::from_host_environ() (environ.cc:56)
7624 ==3261873== by 0x66F1C8: test_reinit_from_host_environ (environ-selftests.c:78)
7625 ==3261873== by 0x66F1C8: selftests::gdb_environ_tests::run_tests() (environ-selftests.c:285)
7626 ==3261873== by 0x7EFC43: operator() (std_function.h:622)
7629 Approved-By: Simon Marchi <simon.marchi@efficios.com>
7631 2022-11-27 Philippe Waroquiers <philippe.waroquiers@skynet.be>
7633 Use false/true for some inferior class members instead of 0/1
7634 Some class members were changed to bool, but there was
7635 still some assignments or comparisons using 0/1.
7637 Approved-By: Simon Marchi <simon.marchi@efficios.com>
7639 2022-11-27 Tom de Vries <tdevries@suse.de>
7641 [gdb/server] Emit warning for SIGINT failure
7642 Consider the executable from test-case gdb.base/interrupt-daemon.exp.
7644 When starting it using gdbserver:
7646 $ ./build/gdbserver/gdbserver localhost:2345 \
7647 ./outputs/gdb.base/interrupt-daemon/interrupt-daemon
7649 and connecting to it using gdb:
7651 $ gdb -q -ex "target remote localhost:2345" \
7652 -ex "set follow-fork-mode child" \
7653 -ex "break daemon_main" -ex cont
7655 we are setup to do the same as in the test-case: interrupt a running inferior
7664 After pressing ^C, nothing happens. This a known problem, filed as
7667 The problem is that in linux_process_target::request_interrupt, a kill is used
7668 to send a SIGINT, but it fails. And it fails silently.
7670 Make the failure verbose by adding a warning, such that the gdbserver output
7671 becomes more helpful:
7673 Process interrupt-daemon created; pid = 15068
7674 Listening on port 2345
7675 Remote debugging from host ::1, port 35148
7676 Detaching from process 15068
7677 Detaching from process 15085
7678 gdbserver: Sending SIGINT to process group of pid 15068 failed: \
7682 Note that the failure can easily be reproduced using the test-case and target
7683 board native-gdbserver:
7687 PASS: gdb.base/interrupt-daemon.exp: fg: continue
7688 ^CFAIL: gdb.base/interrupt-daemon.exp: fg: ctrl-c stops process (timeout)
7690 as reported in PR server/23382.
7692 Tested on x86_64-linux.
7693 Approved-By: Simon Marchi <simon.marchi@efficios.com>
7695 2022-11-27 GDB Administrator <gdbadmin@sourceware.org>
7697 Automatic date update in version.in
7699 2022-11-26 Tom de Vries <tdevries@suse.de>
7701 [gdb/testsuite] Don't generate core in gdb.base/bt-on-fatal-signal.exp
7702 When running test-case gdb.base/bt-on-fatal-signal.exp on powerpc64le-linux I
7705 FAIL: gdb.base/bt-on-fatal-signal.exp: SEGV: scan for backtrace (timeout)
7708 The timeout is 10 seconds, but generating the core file takes more than a
7709 minute, probably due to slow NFS.
7711 I managed to reproduce this behaviour independently of gdb, by compiling
7712 "int main (void) { __builtin_abort (); }" and running it, which took 1.5
7713 seconds for a core file 50 times smaller than the one for gdb.
7715 Fix this by preventing the core file from being generated, using a wrapper
7716 around gdb that does "ulimit -c 0".
7718 Tested on x86_64-linux.
7720 2022-11-26 Tom de Vries <tdevries@suse.de>
7722 [gdb/symtab] Handle failure to open .gnu_debugaltlink file
7723 If we instrument cc-with-tweaks.sh to remove the .gnu_debugaltlink file after
7724 dwz has created it, with test-case
7725 gdb.threads/access-mem-running-thread-exit.exp and target board cc-with-dwz-m
7728 (gdb) file access-mem-running-thread-exit^M
7729 Reading symbols from access-mem-running-thread-exit...^M
7730 could not find '.gnu_debugaltlink' file for access-mem-running-thread-exit^M
7732 followed a bit later by:
7734 (gdb) file access-mem-running-thread-exit^M
7735 Reading symbols from access-mem-running-thread-exit...^M
7736 gdb/dwarf2/read.c:7284: internal-error: create_all_units: \
7737 Assertion `per_objfile->per_bfd->all_units.empty ()' failed.^M
7740 The problem is that create_units does not catch the error thrown by
7741 dwarf2_get_dwz_file.
7743 Fix this by catching the error and performing the necessary cleanup, getting
7744 the same result for the first and second file command.
7747 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29805
7749 2022-11-26 Philippe Waroquiers <philippe.waroquiers@skynet.be>
7751 Fix jump on uninit producer_is_clang bit of cu.h dwarf2_cu struct.
7752 Valgrind reports a "jump on unitialised bit error" when running
7753 e.g. gdb.base/macro-source-path.exp (see details below).
7755 Fix this by initializing producer_is_clang member variable of dwarf2_cu.
7757 Tested on amd64/debian11 and re-running gdb.base/macro-source-path.exp
7760 ==2140965== Conditional jump or move depends on uninitialised value(s)
7761 ==2140965== at 0x5211F7: dwarf_decode_macro_bytes(dwarf2_per_objfile*, buildsym_compunit*, bfd*, unsigned char const*, unsigned char const*, macro_source_file*, line_header const*, dwarf2_section_info const*, int, int, unsigned int, dwarf2_section_info*, dwarf2_section_info*, gdb::optional<unsigned long>, htab*, dwarf2_cu*) (macro.c:676)
7762 ==2140965== by 0x52158A: dwarf_decode_macros(dwarf2_per_objfile*, buildsym_compunit*, dwarf2_section_info const*, line_header const*, unsigned int, unsigned int, dwarf2_section_info*, dwarf2_section_info*, gdb::optional<unsigned long>, int, dwarf2_cu*) (macro.c:967)
7763 ==2140965== by 0x523BC4: dwarf_decode_macros(dwarf2_cu*, unsigned int, int) (read.c:23379)
7764 ==2140965== by 0x552AB5: read_file_scope(die_info*, dwarf2_cu*) (read.c:9687)
7765 ==2140965== by 0x54F7B2: process_die(die_info*, dwarf2_cu*) (read.c:8660)
7766 ==2140965== by 0x5569C7: process_full_comp_unit (read.c:8429)
7767 ==2140965== by 0x5569C7: process_queue (read.c:7675)
7768 ==2140965== by 0x5569C7: dw2_do_instantiate_symtab (read.c:2063)
7769 ==2140965== by 0x5569C7: dw2_instantiate_symtab(dwarf2_per_cu_data*, dwarf2_per_objfile*, bool) (read.c:2085)
7770 ==2140965== by 0x55700B: dw2_expand_symtabs_matching_one(dwarf2_per_cu_data*, dwarf2_per_objfile*, gdb::function_view<bool (char const*, bool)>, gdb::function_view<bool (compunit_symtab*)>) (read.c:3984)
7771 ==2140965== by 0x557EA3: cooked_index_functions::expand_symtabs_matching(objfile*, gdb::function_view<bool (char const*, bool)>, lookup_name_info const*, gdb::function_view<bool (char const*)>, gdb::function_view<bool (compunit_symtab*)>, enum_flags<block_search_flag_values>, domain_enum, search_domain) (read.c:18781)
7772 ==2140965== by 0x778977: objfile::lookup_symbol(block_enum, char const*, domain_enum) (symfile-debug.c:276)
7774 ==2140965== Uninitialised value was created by a heap allocation
7775 ==2140965== at 0x4839F01: operator new(unsigned long) (vg_replace_malloc.c:434)
7776 ==2140965== by 0x533A64: cutu_reader::cutu_reader(dwarf2_per_cu_data*, dwarf2_per_objfile*, abbrev_table*, dwarf2_cu*, bool, abbrev_cache*) (read.c:6264)
7777 ==2140965== by 0x5340C2: load_full_comp_unit(dwarf2_per_cu_data*, dwarf2_per_objfile*, dwarf2_cu*, bool, language) (read.c:7729)
7778 ==2140965== by 0x548338: load_cu(dwarf2_per_cu_data*, dwarf2_per_objfile*, bool) (read.c:2021)
7779 ==2140965== by 0x55634C: dw2_do_instantiate_symtab (read.c:2048)
7780 ==2140965== by 0x55634C: dw2_instantiate_symtab(dwarf2_per_cu_data*, dwarf2_per_objfile*, bool) (read.c:2085)
7781 ==2140965== by 0x55700B: dw2_expand_symtabs_matching_one(dwarf2_per_cu_data*, dwarf2_per_objfile*, gdb::function_view<bool (char const*, bool)>, gdb::function_view<bool (compunit_symtab*)>) (read.c:3984)
7782 ==2140965== by 0x557EA3: cooked_index_functions::expand_symtabs_matching(objfile*, gdb::function_view<bool (char const*, bool)>, lookup_name_info const*, gdb::function_view<bool (char const*)>, gdb::function_view<bool (compunit_symtab*)>, enum_flags<block_search_flag_values>, domain_enum, search_domain) (read.c:18781)
7783 ==2140965== by 0x778977: objfile::lookup_symbol(block_enum, char const*, domain_enum) (symfile-debug.c:276)
7786 2022-11-26 Philippe Waroquiers <philippe.waroquiers@skynet.be>
7788 remove the declared but undefined/unused method find_partial_die
7790 struct partial_die_info *find_partial_die (sect_offset sect_off);
7791 in cu.h is defined, but is used nowhere and not implemented.
7793 2022-11-26 GDB Administrator <gdbadmin@sourceware.org>
7795 Automatic date update in version.in
7797 2022-11-25 Martin Liska <mliska@suse.cz>
7799 Revert "readelf: Do not require EI_OSABI for IFUNC."
7800 This reverts commit ffbbab0b3a1000f862b6d4ce3d9a76ed14f08801.
7802 2022-11-25 Christoph Müllner <christoph.muellner@vrull.eu>
7804 riscv: Add AIA extension support (Smaia, Ssaia)
7805 This commit adds the AIA extensions (Smaia and Ssaia) CSRs.
7809 * elfxx-riscv.c: Add 'smaia' and 'ssaia' to the list
7810 of known standard extensions.
7814 * config/tc-riscv.c (enum riscv_csr_class):
7815 (riscv_csr_address): Add CSR classes for Smaia/Ssaia.
7816 * testsuite/gas/riscv/csr-dw-regnums.d: Add new CSRs.
7817 * testsuite/gas/riscv/csr-dw-regnums.s: Likewise.
7818 * testsuite/gas/riscv/csr-version-1p10.d: Likewise.
7819 * testsuite/gas/riscv/csr-version-1p10.l: Likewise.
7820 * testsuite/gas/riscv/csr-version-1p11.d: Likewise.
7821 * testsuite/gas/riscv/csr-version-1p11.l: Likewise.
7822 * testsuite/gas/riscv/csr-version-1p12.d: Likewise.
7823 * testsuite/gas/riscv/csr-version-1p12.l: Likewise.
7824 * testsuite/gas/riscv/csr-version-1p9p1.d: Likewise.
7825 * testsuite/gas/riscv/csr-version-1p9p1.l: Likewise.
7826 * testsuite/gas/riscv/csr.s: Likewise.
7830 * opcode/riscv-opc.h (CSR_MISELECT): New CSR macro.
7831 (CSR_MIREG): Likewise.
7832 (CSR_MTOPEI): Likewise.
7833 (CSR_MTOPI): Likewise.
7834 (CSR_MVIEN): Likewise.
7835 (CSR_MVIP): Likewise.
7836 (CSR_MIDELEGH): Likewise.
7837 (CSR_MIEH): Likewise.
7838 (CSR_MVIENH): Likewise.
7839 (CSR_MVIPH): Likewise.
7840 (CSR_MIPH): Likewise.
7841 (CSR_SISELECT): Likewise.
7842 (CSR_SIREG): Likewise.
7843 (CSR_STOPEI): Likewise.
7844 (CSR_STOPI): Likewise.
7845 (CSR_SIEH): Likewise.
7846 (CSR_SIPH): Likewise.
7847 (CSR_HVIEN): Likewise.
7848 (CSR_HVICTL): Likewise.
7849 (CSR_HVIPRIO1): Likewise.
7850 (CSR_HVIPRIO2): Likewise.
7851 (CSR_VSISELECT): Likewise.
7852 (CSR_VSIREG): Likewise.
7853 (CSR_VSTOPEI): Likewise.
7854 (CSR_VSTOPI): Likewise.
7855 (CSR_HIDELEGH): Likewise.
7856 (CSR_HVIENH): Likewise.
7857 (CSR_HVIPH): Likewise.
7858 (CSR_HVIPRIO1H): Likewise.
7859 (CSR_HVIPRIO2H): Likewise.
7860 (CSR_VSIEH): Likewise.
7861 (CSR_VSIPH): Likewise.
7862 (DECLARE_CSR): Add CSRs for Smaia and Ssaia.
7865 - Imply ssaia for smaia
7866 - Imply zicsr for ssaia (and transitively smaia)
7867 - Move hypervisor CSRs to Ssaia+H
7868 - Rebase on upstream/master
7871 - Add hypervisor and VS CSRs
7872 - Fix whitespace issue
7874 2022-11-25 GDB Administrator <gdbadmin@sourceware.org>
7876 Automatic date update in version.in
7878 2022-11-24 Indu Bhagat <indu.bhagat@oracle.com>
7880 sframe/doc: remove usage of xrefautomaticsectiontitle
7881 xrefautomaticsectiontitle appears to be available from texinfo 5.0 or
7882 greater. As such, it is not worthwhile to add requirement for a minimum
7883 necessary makeinfo version. So remove the usage of it.
7885 Also align node name with section title where possible.
7889 * libsframe/doc/sframe-spec.texi: Remove usage of
7890 xrefautomaticsectiontitle.
7892 2022-11-24 Andrew Burgess <aburgess@redhat.com>
7894 gdb: fix typo in debug output message
7895 Spotted a minor type, a missing ')', in a debug message.
7897 2022-11-24 Simon Marchi <simon.marchi@polymtl.ca>
7899 gdb/testsuite/gdb.base/break.exp: split test_break
7900 Move all the remaining tests to a single test_break proc. It's a bit
7901 big, but all of these are kind of tied together. The procs starts by
7902 setting breakpoints, checks that we can see them in "info breakpoints",
7903 and tries stopping on them.
7905 Move all the "set bp_locationX" calls together at the top.
7907 Change-Id: Id05f98957e1a3462532d2dbd577cd0a7c7263900
7908 Approved-By: Kevin Buettner <kevinb@redhat.com>
7910 2022-11-24 Simon Marchi <simon.marchi@polymtl.ca>
7912 gdb/testsuite/gdb.base/break.exp: split test_tbreak
7913 Leave setting bp_location11 in the global scope, so that it's accessible
7916 Change-Id: I8928f01640d3a1e993649b2168b9eda0724ee1d9
7917 Approved-By: Kevin Buettner <kevinb@redhat.com>
7919 2022-11-24 Simon Marchi <simon.marchi@polymtl.ca>
7921 gdb/testsuite/gdb.base/break.exp: split test_no_break_on_catchpoint
7922 Change-Id: Ifa7070943f1de22c2839fedf5f346d6591bb5a76
7923 Approved-By: Kevin Buettner <kevinb@redhat.com>
7925 gdb/testsuite/gdb.base/break.exp: split test_break_nonexistent_line
7926 Change-Id: I4390dd5da23bae83ccc513ad0de0169ddff7df12
7927 Approved-By: Kevin Buettner <kevinb@redhat.com>
7929 2022-11-24 Simon Marchi <simon.marchi@polymtl.ca>
7931 gdb/testsuite/gdb.base/break.exp: split test_break_default
7932 One special thing here is that the part just above this one, that sets
7933 catchpoints and verifies they are not hit, requires that we resume
7934 execution to verify that the catchpoints are indeed not hit. I guess
7935 it was previously achieved by the until command, but it doesn't happen
7936 now that the until is moved into test_break_default. Add a
7937 gdb_continue_to_end after setting the catchpoints. If any catchpoint
7938 were to be hit, it would catch the problem.
7940 Change-Id: I5d4b43da91886b1beda9f6e56b05aa04331a9c05
7941 Approved-By: Kevin Buettner <kevinb@redhat.com>
7943 2022-11-24 Simon Marchi <simon.marchi@polymtl.ca>
7945 gdb/testsuite/gdb.base/break.exp: split test_break_silent_and_more
7946 This one is a bit tricky. The clear tests seem to depend on the various
7947 breakpoints that have been set before, starting with the "silent"
7948 breakpoints. So, move all this in a single chunk, it can always be
7949 split later if needed.
7951 Change-Id: I7ba61a5b130ade63eda0c4790534840339f8a72f
7952 Approved-By: Kevin Buettner <kevinb@redhat.com>
7954 2022-11-24 Simon Marchi <simon.marchi@polymtl.ca>
7956 gdb/testsuite/gdb.base/break.exp: split test_break_line_convenience_var
7957 Change-Id: I593002373da971a0a4d6b5355d3fe321873479ab
7958 Approved-By: Kevin Buettner <kevinb@redhat.com>
7960 gdb/testsuite/gdb.base/break.exp: split test_break_user_call
7961 Change-Id: I9151ce9db9435722b758f41c6606b461bf15f320
7962 Approved-By: Kevin Buettner <kevinb@redhat.com>
7964 gdb/testsuite/gdb.base/break.exp: split test_finish_arguments
7965 Change-Id: Id84babed1eeb3ce7d14b94ff332795964e8ead51
7966 Approved-By: Kevin Buettner <kevinb@redhat.com>
7968 2022-11-24 Simon Marchi <simon.marchi@polymtl.ca>
7970 gdb/testsuite/gdb.base/break.exp: use proc_with_prefix for test_next_with_recursion
7971 This one is already in a proc, just make the proc use proc_with_prefix,
7974 Change-Id: I313ecf5097ff04526c29396529baeba84e37df5a
7975 Approved-By: Kevin Buettner <kevinb@redhat.com>
7977 2022-11-24 Simon Marchi <simon.marchi@polymtl.ca>
7979 gdb/testsuite/gdb.base/break.exp: split test_break_optimized_prologue
7980 Change-Id: Ibf17033c8ce72aa5cfe1b739be2902e84a5e945d
7981 Approved-By: Kevin Buettner <kevinb@redhat.com>
7983 gdb/testsuite/gdb.base/break.exp: split test_rbreak_shlib
7984 Change-Id: I130e8914c2713095aab03e84aba1481b4c7af978
7985 Approved-By: Kevin Buettner <kevinb@redhat.com>
7987 gdb/testsuite/gdb.base/break.exp: split test_break_file_line_convenience_var
7988 Change-Id: I0c31b037669b2917e062bf431372fb6531f8f53c
7989 Approved-By: Kevin Buettner <kevinb@redhat.com>
7991 gdb/testsuite/gdb.base/break.exp: split test_break_commands_clear
7992 Change-Id: Ia58f90117d52fc419fc494836d9b4ed5d902fe9b
7993 Approved-By: Kevin Buettner <kevinb@redhat.com>
7995 2022-11-24 Nick Clifton <nickc@redhat.com>
7997 Impport libiberty commit: 885b6660c17f from gcc mainline. Fix gas's acinclude.m4 to stop a potwntial configure time warning message.
7999 2022-11-24 Martin Liska <mliska@suse.cz>
8001 readelf: Do not require EI_OSABI for IFUNC.
8006 * readelf.c (get_symbol_type): Consider STT_GNU_IFUNC as
8009 2022-11-24 Jan Beulich <jbeulich@suse.com>
8011 x86: widen applicability and use of CheckRegSize
8012 First of all make operand_type_register_match() apply to all sized
8013 operands, i.e. in Intel Syntax also to respective memory ones. This
8014 addresses gas wrongly accepting certain SIMD insns where register and
8015 memory operand sizes should match but don't. This apparently has
8016 affected all templates with one memory-only operand and one or more
8017 register ones, both permitting at least two sizes, due to CheckRegSize
8020 Then also add CheckRegSize to a couple of non-SIMD templates matching
8021 that same pattern of memory-only vs register operands. This replaces
8022 bogus (for Intel Syntax) diagnostics referring to a wrong suffix (when
8023 none was used at all) by "type mismatch" ones, just like already emitted
8024 for insns where the template allows a register operand alongside a
8025 memory one at any particular position.
8027 This also is a prereq to limiting (ideally eliminating in the long run)
8028 suffix "derivation" in Intel Syntax mode.
8030 While making the code adjustment also flip order of checks to do the
8031 cheaper one first in both cases.
8033 2022-11-24 Jan Beulich <jbeulich@suse.com>
8035 x86: add missing CheckRegSize
8036 To properly and predictably determine operand size encoding (operand
8037 size or REX.W prefixes), consistent operand sizes need to be specified.
8038 Add CheckRegSize where this was previously missing.
8040 x86: correct handling of LAR and LSL
8041 Both uniformly only ever take 16-bit memory operands while at the same
8042 time requiring matching (in size) register operands, which then also
8043 should disassemble that way. This in particular requires splitting each
8044 of the templates for the assembler and separating decode of the
8045 register and memory forms in the disassembler.
8047 2022-11-24 Alan Modra <amodra@gmail.com>
8049 Tidy objdump printing of section size
8050 * objdump.c (load_specific_debug_section): Use PRIx64 format.
8052 Constify nm format array
8053 * nm.c (formats, format): Make const.
8055 2022-11-24 Alan Modra <amodra@gmail.com>
8057 PR16995, m68k coldfire emac immediate to macsr incorrect disassembly
8058 Mode/reg bits for these insns are 000 Dy, 001 Ay, and 111 100 for the
8061 * m68k-opc.c (m68k_opcodes): Only accept 000 and 001 as mode
8062 for move reg to macsr/mask insns.
8064 2022-11-24 Mark Harmstone <mark@harmstone.com>
8066 gas: Disable --gcodeview on PE targets with no O_secrel
8068 2022-11-24 GDB Administrator <gdbadmin@sourceware.org>
8070 Automatic date update in version.in
8072 2022-11-23 Torbjörn SVENSSON <torbjorn.svensson@foss.st.com>
8074 gdb/arm: Include FType bit in EXC_RETURN pattern on v8m
8075 For v8m, the EXC_RETURN pattern, without security extension, consists of
8076 FType, Mode and SPSEL. These are the same bits that are used in v7m.
8077 This patch extends the list of patterns to include also the FType bit
8078 and not just Mode and SPSEL bits for v8m targets without security
8081 2022-11-23 Alan Modra <amodra@gmail.com>
8085 2022-11-23 Alan Modra <amodra@gmail.com>
8087 PR22509 - Null pointer dereference on coff_slurp_reloc_table
8088 This extends the commit 4581a1c7d304 fix to more targets, which
8089 hardens BFD a little. I think the real underlying problem was the
8090 bfd_canonicalize_reloc call in load_specific_debug_section which
8091 passed a NULL for "symbols". Fix that too.
8095 * aoutx.h (swap_ext_reloc_out): Gracefully handle NULL symbols.
8096 * i386lynx.c (swap_ext_reloc_out): Likewise.
8097 * pdp11.c (pdp11_aout_swap_reloc_out): Likewise.
8098 * coff-tic30.c (reloc_processing): Likewise.
8099 * coff-tic4x.c (tic4x_reloc_processing): Likewise.
8100 * coff-tic54x.c (tic54x_reloc_processing): Likewise.
8101 * coff-z80.c (reloc_processing): Likewise.
8102 * coff-z8k.c (reloc_processing): Likewise.
8103 * ecoff.c (ecoff_slurp_reloc_table): Likewise.
8104 * som.c (som_set_reloc_info): Likewise.
8106 * objdump.c (load_specific_debug_section): Pass syms to
8107 bfd_canonicalize_reloc.
8109 2022-11-23 Alan Modra <amodra@gmail.com>
8111 asan: NULL deref in filter_symbols
8112 If tdata->symbols is NULL, make tdata->symcount zero too. This makes
8113 wasm_get_symtab_upper_bound return the proper result and stops
8116 * wasm-module.c (wasm_scan_name_function_section): Clear
8117 tdata->symcount on error.
8119 2022-11-23 Luis Machado <luis.machado@arm.com>
8121 Document the memory_tagged argument for memory region callbacks
8122 There were no comments in some instances (gdb/defs.h, gdb/core.c and
8123 gdb/linux-tdep.c), so address that by adding comments where those are missing.
8125 2022-11-23 Tom de Vries <tdevries@loganberry-1.arch.suse.de>
8127 Fix gdb.cp/gdb2495.exp on powerpc64le
8128 On powerpc64le-linux I ran into this FAIL:
8130 (gdb) p exceptions.throw_function()^M
8131 terminate called after throwing an instance of 'int'^M
8133 Program received signal SIGABRT, Aborted.^M
8134 0x00007ffff7979838 in raise () from /lib64/libc.so.6^M
8135 The program being debugged was signaled while in a function called from GDB.^M
8136 GDB remains in the frame where the signal was received.^M
8137 To change this behavior use "set unwindonsignal on".^M
8138 Evaluation of the expression containing the function^M
8139 (SimpleException::throw_function()) will be abandoned.^M
8140 When the function is done executing, GDB will silently stop.^M
8141 (gdb) FAIL: gdb.cp/gdb2495.exp: call a function that raises an exception \
8145 The following happens:
8146 - we start an inferior call
8147 - an internal breakpoint is set on the global entry point of std::terminate
8148 - the inferior call uses the local entry point
8149 - the breakpoint is not triggered
8150 - we run into std::terminate
8152 We can fix this by simply adding the missing gdbarch_skip_entrypoint call in
8153 create_std_terminate_master_breakpoint, but we try to do this a bit more
8155 - adding a variant of function create_internal_breakpoint which takes a
8156 minimal symbol instead of an address as argument
8157 - in the new function:
8158 - using both gdbarch_convert_from_func_ptr_addr and gdbarch_skip_entrypoint
8159 - documenting why we don't need to use gdbarch_addr_bits_remove
8160 - adding a note about possibly
8161 needing gdbarch_deprecated_function_start_offset.
8162 - using the new function in:
8163 - create_std_terminate_master_breakpoint
8164 - create_exception_master_breakpoint_hook, which currently uses only
8165 gdbarch_convert_from_func_ptr_addr.
8167 Note: we could use the new function in more locations in breakpoint.c, but
8168 as we're not aware of any related failures, we declare this out of scope for
8171 Tested on x86_64-linux, powerpc64le-linux.
8173 Co-Authored-By: Ulrich Weigand <uweigand@de.ibm.com>
8174 Tested-by: Carl Love <cel@us.ibm.com>
8176 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29793
8178 2022-11-23 Xiao Zeng <zengxiao@eswincomputing.com>
8180 RISC-V: Make R_RISCV_SUB6 conforms to riscv ABI standard
8181 According to the riscv psabi, R_RISCV_SUB6 only allows 6 least significant
8182 bits are valid, but since binutils implementation, we usually get 8 bits
8183 field for it. That means, the high 2 bits could be other field and have
8184 different purpose. Therefore, we should filter the 8 bits to 6 bits before
8185 calculate, and then only encode the valid 6 bits back. By the way, we also
8186 need the out-of-range check for R_RISCV_SUB6, and the overflow checks for
8187 all R_RISCV_ADD/SUB/SET relocations, but we can add them in the future patches.
8189 Passing riscv-gnu-toolchain regressions.
8193 * elfnn-riscv.c (riscv_elf_relocate_section): Take the R_RISCV_SUB6
8194 lower 6 bits as the significant bit.
8195 * elfxx-riscv.c (riscv_elf_add_sub_reloc): Likewise.
8197 2022-11-23 Mark Harmstone <mark@harmstone.com>
8199 gas: Add --gcodeview option
8201 ld: Add section contributions substream to PDB files
8203 2022-11-23 GDB Administrator <gdbadmin@sourceware.org>
8205 Automatic date update in version.in
8207 2022-11-22 John Baldwin <jhb@FreeBSD.org>
8209 aarch64-fbsd: Use a static regset for the TLS register set.
8210 This uses custom collect/supply regset handlers which pass the TLS
8211 register number from the gdbarch_tdep as the base register number.
8213 Approved-By: Simon Marchi <simon.marchi@efficios.com>
8215 2022-11-22 John Baldwin <jhb@FreeBSD.org>
8217 arm-fbsd: Use a static regset for the TLS register set.
8218 This uses custom collect/supply regset handlers which pass the TLS
8219 register number from the gdbarch_tdep as the base register number.
8221 Approved-By: Simon Marchi <simon.marchi@efficios.com>
8223 2022-11-22 John Baldwin <jhb@FreeBSD.org>
8225 fbsd-nat: Pass an optional register base to the register set helpers.
8226 This is needed to permit using the helpers for register sets with a
8227 variable base. In particular regnum needs to be converted into a
8228 relative register number before passed to regcache_map_supplies.
8230 Approved-By: Simon Marchi <simon.marchi@efficios.com>
8232 2022-11-22 John Baldwin <jhb@FreeBSD.org>
8234 fbsd-nat: Use regset supply/collect methods.
8235 fbsd-nat includes various helper routines for fetching and storing
8236 register sets via ptrace where the register set is described by a
8237 regset. These helper routines directly invoke the
8238 supply/collect_regset regcache methods which doesn't permit a regset
8239 to provide custom logic when fetching or storing a register set.
8240 Instead, just use the function pointers from the struct regset
8243 Approved-By: Simon Marchi <simon.marchi@efficios.com>
8245 2022-11-22 John Baldwin <jhb@FreeBSD.org>
8247 regcache: Add collect/supply_regset variants that accept a register base.
8248 Some register sets described by an array of regcache_map_entry
8249 structures do not have fixed register numbers in their associated
8250 architecture but do describe a block of registers whose numbers are at
8251 fixed offsets relative to some base register value. An example of
8252 this are the TLS register sets for the ARM and AArch64 architectures.
8254 Currently OS-specific architectures create register maps and register
8255 sets dynamically using the register base number. However, this
8256 requires duplicating the code to create the register map and register
8257 set. To reduce duplication, add variants of the collect_regset and
8258 supply_regset regcache methods which accept a base register number.
8259 For valid register map entries (i.e. not REGCACHE_MAP_SKIP), add this
8260 base register number to the value from the map entry to determine the
8261 final register number.
8263 Approved-By: Simon Marchi <simon.marchi@efficios.com>
8265 2022-11-22 H.J. Lu <hjl.tools@gmail.com>
8267 x86: Don't define _TLS_MODULE_BASE_ for ld -r
8271 * elfxx-x86.c (_bfd_x86_elf_always_size_sections): Don't define
8272 _TLS_MODULE_BASE_ for ld -r.
8277 * testsuite/ld-x86-64/pr29820.d: New file.
8278 * testsuite/ld-x86-64/pr29820.s: Likewise.
8279 * testsuite/ld-x86-64/x86-64.ex: Run pr29820.
8281 2022-11-22 Alan Modra <amodra@gmail.com>
8283 Don't use "long" in readelf for file offsets
8284 The aim here is to improve readelf handling of large 64-bit object
8285 files on LLP64 hosts (Windows) where long is only 32 bits. The patch
8286 changes more than just file offsets. Addresses and sizes are also
8287 changed to avoid "long". Most places get to use uint64_t even where
8288 size_t may be more appropriate, because that allows some overflow
8289 checks to be implemented easily (*alloc changes).
8291 * dwarf.c (cmalloc, xcmalloc, xcrealloc, xcalloc2): Make nmemb
8293 * dwarf.h: Update prototypes.
8294 (struct dwarf_section): Make num_relocs uint64_t.
8295 * elfcomm.c (setup_archive): Update error format.
8296 * elfcomm.h (struct archive_info): Make sym_size, longnames_size,
8297 nested_member_origin, next_arhdr_offset uint64_t.
8298 * readelf.c (struct filedata): Make archive_file_offset,
8299 archive_file_size, string_table_length, dynamic_addr,
8300 dynamic_nent, dynamic_strings_length, num_dynamic_syms,
8301 dynamic_syminfo_offset uint64_t.
8302 (many functions): Replace uses of "unsigned long" with
8303 "uint64_t" or "size_t".
8305 2022-11-22 Alan Modra <amodra@gmail.com>
8307 Re: readelf: use fseeko64 or fseeko if possible
8308 Replace the macros with a small wrapper function that verifies the fseek
8309 offset arg isn't overlarge.
8311 * readelf.c (FSEEK_FUNC): Delete, replace uses with..
8312 (fseek64): ..this new function.
8313 (process_program_headers): Don't cast p_offset to long.
8315 2022-11-22 Torbjörn SVENSSON <torbjorn.svensson@foss.st.com>
8317 gdb/arm: Fix obvious typo in b0b23e06c3a
8318 As part of the rebase of the patch, I managed to loose the local
8319 changes I had for the comments from Tomas in
8320 https://sourceware.org/pipermail/gdb-patches/2022-November/193413.html
8321 This patch corrects the obvious two typos.
8323 2022-11-22 Michael Matz <matz@suse.de>
8325 binutils/configure.ac: integrate last change
8326 Integrate back checks for fseeko{,64} into configure.ac, so
8327 that regeneration works.
8330 * configure.ac: Add fseeko, fseeko64 checks.
8331 * configure: Regenerate.
8333 2022-11-22 Shahab Vahedi <shahab@synopsys.com>
8335 opcodes: Correct address for ARC's "isa_config" aux reg
8336 This patch changes the address for "isa_config" auxiliary register
8337 from 0xC2 to the correct value 0xC1. Moreover, it only exists in
8338 arc700+ and not all ARCs.
8342 * arc-regs.h: Change isa_config address to 0xc1.
8343 isa_config exists for ARC700 and ARCV2 and not ARCALL.
8345 2022-11-22 Bruno Larsen <blarsen@redhat.com>
8347 gdb/testsuite: remove gcc restriction from gdb.dwarf2/clang-cli-macro.exp
8348 With the recent changes to the dwarf assembler, there is no longer a
8349 need to test for gcc in gdb.dwarf2/clang-cli-macro.exp and mark it as
8350 untested. This commit removes that logic.
8352 2022-11-22 Jan Beulich <jbeulich@suse.com>
8354 gas/sframe: avoid "shadowing" of glibc function name
8355 Once again: Old enough glibc has an (unguarded) declaration of index()
8356 in string.h, which triggers a "shadows a global declaration" warning
8357 with our choice of wanring level and with at least some gcc versions.
8359 2022-11-22 GDB Administrator <gdbadmin@sourceware.org>
8361 Automatic date update in version.in
8363 2022-11-21 Brett Werling <bwerl.dev@gmail.com>
8365 readelf: use fseeko64 or fseeko if possible
8366 Changes readelf to make use first of fseeko64 and then fseeko,
8367 depending on which of those is available. If neither is available,
8368 reverts to the previous behavior of using fseek.
8370 This is necessary when building readelf for LLP64 systems, where a
8371 long will only be 32 bits wide. If the elf file in question is >= 2 GiB,
8372 that is greater than the max long value and therefore fseek will fail
8373 indicating that the offset is negative. On such systems, making use of
8374 fseeko64 or fseeko will result in the ability so seek past the 2 GiB
8377 Note that large archive handling in readelf remains to be fixed.
8379 2022-11-21 Alan Modra <amodra@gmail.com>
8381 PR29807, SIGSEGV when linking fuzzed PE object
8383 * cofflink.c (_bfd_coff_generic_relocate_section): Skip relocs
8384 against symbols with a NULL section.
8386 2022-11-21 Alan Modra <amodra@gmail.com>
8388 Re: ld: Always output local symbol for relocatable link
8389 Remove this code dating back to commit 98790d3a95fc entirely, what it
8390 was trying to do is done elsewhere.
8393 * elflink.c (elf_link_output_symstrtab): Don't handle symbols
8394 in SEC_EXCLUDE sections here.
8396 2022-11-21 Philippe Waroquiers <philippe.waroquiers@skynet.be>
8398 When getting the locno of a bpstat, handle the case of bp with null locations.
8399 The test py-objfile.exp unloads the current file while debugging the process.
8400 This results in bpstat bs->b->loc to become nullptr.
8401 Handle this case in breakpoint.c:bpstat_locno.
8403 Note: GDB crashes on this problem with an internal error,
8404 but the end of gdb summary shows:
8408 # of expected passes 36
8410 The output also does not contain a 'FAIL:'.
8411 After the fix, the nr of expected passes increased.
8413 In the gdb.log output, one can see:
8415 Fatal signal: Segmentation fault
8416 ----- Backtrace -----
8417 0x55698905c5b9 gdb_internal_backtrace_1
8418 ../../binutils-gdb/gdb/bt-utils.c:122
8419 0x55698905c5b9 _Z22gdb_internal_backtracev
8422 ERROR: Couldn't send python print(objfile.filename) to GDB.
8423 ERROR: : spawn id exp9 not open
8427 -re ".*A problem internal to GDB has been detected" {
8428 fail "$message (GDB internal error)"
8429 gdb_internal_error..."
8430 ("uplevel" body line 1)
8434 Wondering if it might be possible to improve gdb_test to have
8435 gdb_test "python print(objfile.filename)" "None" \
8436 "objfile.filename after objfile is unloaded"
8437 reporting a failed result instead of just producing the internal error.
8439 2022-11-21 Philippe Waroquiers <philippe.waroquiers@skynet.be>
8441 Fix use after free introduced by $_hit_bpnum/$_hit_locno variables.
8442 If the commands of the bpstat bs contain commands such as step or next or
8443 continue, the BS and its commands are freed by execute_control_command.
8445 So, we cannot remember the BS that was printed. Instead, remember
8446 the bpnum and locno.
8448 Regtested on debian/amd64 and re-run a few tests under valgrind.
8450 2022-11-21 Philippe Waroquiers <philippe.waroquiers@skynet.be>
8452 Fix step-over-syscall.exp matching regexp for $bpnum.$locno matching
8453 step-over-syscall.exp has some specific tests for gdbserver.
8454 The regexp matching breakpoint hit must take the added locno into account.
8456 Test re-run in 3 modes (normal, native-gdbserver and native-extended-gdbserver).
8458 2022-11-21 Nick Clifton <nickc@redhat.com>
8460 Fix ARM and AArch64 assembler tests to work in a multi-arch environment.
8462 gas * testsuite/gas/arm/cpu-cortex-a76ae.d: Add arm prefix to the -m
8463 option passed to objdump.
8464 * testsuite/gas/arm/cpu-cortex-a77.d: Likewise.
8465 * testsuite/gas/aarch64/cpu-cortex-a76ae.d: Add aarch64 prefix to
8466 the -m option passed to objdump.
8467 * testsuite/gas/aarch64/cpu-cortex-a77.d: Likewise.
8469 bfd * cpu-arm.c (scan): Accept machine names prefixed with "arm:".
8470 * cpu-aarch64.c (scan): Accept machine names prefixed with "aarch64:".
8472 bin * doc/binutils.texi (objdump): Note that the -m option supports
8473 the <architecture>:<machine> syntax.
8475 2022-11-21 Torbjörn SVENSSON <torbjorn.svensson@foss.st.com>
8477 gdb/arm: Ensure that stack pointers are in sync
8478 For targets with secext, msp and psp can be seen as an alias for one
8479 of msp_s, msp_ns, psp_s or psp_ns.
8480 Without this patch, sp might be secure, but msp or psp is non-secure
8481 (this state can not happen in the hardware).
8483 gdb/arm: Update active msp/psp when switching stack
8484 For targets with secext, msp and psp can be seen as an alias for one
8485 of msp_s, msp_ns, psp_s or psp_ns. When switching active sp, the
8486 corresponding msp/psp needs to be switched too.
8488 2022-11-21 Jiangshuai Li <jiangshuai_li@linux.alibaba.com>
8490 gdb/csky just return type from csky_vector_type() for vector resgisters
8491 Some gdb stubs may not describe the type for vector registers in the
8492 tdesc-xml and only send bitsize="128", gdb can't deal with a reg
8493 with default type int with bitsize==128. So Just return csky_vector_type()
8494 for vector resgisters.
8496 gdb/csky return type int32 for float and vector pseudo regs
8497 When reg_nr is one of the float and vector pseudo registers,
8498 return builtin_type (gdbarch)->builtin_int32 for it.
8500 2022-11-21 GDB Administrator <gdbadmin@sourceware.org>
8502 Automatic date update in version.in
8504 2022-11-20 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8506 [PR build/29791] gnulib: Disable _GL_ATTRIBUTE_DEALLOC on Solaris
8507 gdbsupport compilation badly fails with GCC 12 on Solaris, with errors
8510 ../gnulib/config.h:1693:72: error: ‘malloc’ attribute argument 1 is ambiguous
8511 1693 | # define _GL_ATTRIBUTE_DEALLOC(f, i) __attribute__ ((__malloc__ (f, i)))
8513 ../gnulib/config.h:1693:72: note: use a cast to the expected type to disambiguate
8515 We've not yet been able to determine where the ambiguity actually lies,
8516 so this patch works around the issue by disabling _GL_ATTRIBUTE_DEALLOC
8517 on Solaris, at least as a workaround for GDB 13.
8519 As Tom suggested in the PR, this is done using our infrastructure for
8520 local gnulib patches.
8522 Tested on sparcv9-sun-solaris2.11, amd64-pc-solaris2.11, and
8523 x86_64-pc-linux-gnu.
8525 Approved-By: Simon Marchi <simon.marchi@efficios.com>
8527 2022-11-20 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8529 Fix sol-thread.c compilation on 32-bit Solaris
8530 sol-thread.c fails to compile on 32-bit Solaris: there are several
8533 In file included from /vol/src/gnu/gdb/hg/master/local/gdb/../gdbsupport/common-defs.h:203,
8534 from /vol/src/gnu/gdb/hg/master/local/gdb/defs.h:28,
8535 from /vol/src/gnu/gdb/hg/master/local/gdb/sol-thread.c:51:
8536 /vol/src/gnu/gdb/hg/master/local/gdb/sol-thread.c: In member function ‘virtual void sol_thread_target::resume(ptid_t, int, gdb_signal)’:
8537 /vol/src/gnu/gdb/hg/master/local/gdb/sol-thread.c:416:20: error: format ‘%ld’ expects argument of type ‘long int’, but argument 2 has type ‘ULONGEST’ {aka ‘long long unsigned int’} [-Werror=format=]
8538 416 | warning (_("Specified thread %ld seems to have terminated"),
8539 | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
8540 /vol/src/gnu/gdb/hg/master/local/gdb/../gdbsupport/gdb_locale.h:28:29:
8541 note: in definition of macro ‘_’
8542 28 | # define _(String) gettext (String)
8544 /vol/src/gnu/gdb/hg/master/local/gdb/sol-thread.c:416:40: note: format
8545 string is defined here
8546 416 | warning (_("Specified thread %ld seems to have terminated"),
8552 Fixed by using pulongest () instead.
8554 Tested on i386-pc-solaris2.11, amd64-pc-solaris2.11,
8555 sparc-sun-solaris2.11, and sparcv9-sun-solaris2.11 (together with
8556 Simon's patch for PR build/29798).
8558 2022-11-20 GDB Administrator <gdbadmin@sourceware.org>
8560 Automatic date update in version.in
8562 2022-11-19 Philippe Waroquiers <philippe.waroquiers@skynet.be>
8564 Add missing gdb_prompt in ctxobj.exp to avoid random failure, fix typo.
8565 ctxobj.exp fails randomly when computer is loaded.
8566 With the addition of $gdb_prompt in the regexp testing for breakpoint hit,
8567 I could not make it fail anymore.
8569 Also fixed a typo in a comment.
8571 2022-11-19 Philippe Waroquiers <philippe.waroquiers@skynet.be>
8573 Show locno for 'multi location' breakpoint hit msg+conv var $_hit_bbnum $_hit_locno PR breakpoints/12464
8574 This implements the request given in PR breakpoints/12464.
8576 Before this patch, when a breakpoint that has multiple locations is reached,
8578 Thread 1 "zeoes" hit Breakpoint 1, some_func () at somefunc1.c:5
8580 This patch changes the message so that bkpt_print_id prints the precise
8581 encountered breakpoint:
8582 Thread 1 "zeoes" hit Breakpoint 1.2, some_func () at somefunc1.c:5
8584 In mi mode, bkpt_print_id also (optionally) prints a new table field "locno":
8585 locno is printed when the breakpoint hit has more than one location.
8586 Note that according to the GDB user manual node 'GDB/MI Development and Front
8587 Ends', it is ok to add new fields without changing the MI version.
8589 Also, when a breakpoint is reached, the convenience variables
8590 $_hit_bpnum and $_hit_locno are set to the encountered breakpoint number
8591 and location number.
8593 $_hit_bpnum and $_hit_locno can a.o. be used in the command list of a
8594 breakpoint, to disable the specific encountered breakpoint, e.g.
8595 disable $_hit_bpnum.$_hit_locno
8597 In case the breakpoint has only one location, $_hit_locno is set to
8598 the value 1, so as to allow a command such as:
8599 disable $_hit_bpnum.$_hit_locno
8600 to disable the breakpoint even when the breakpoint has only one location.
8602 This also fixes a strange behaviour: when a breakpoint X has only
8605 is accepted but transforms the breakpoint in a multiple locations
8606 breakpoint having only one location.
8608 The changes in RFA v4 handle the comments of Tom Tromey:
8609 - Changed convenience var names from $bkptno/$locno to
8610 $_hit_bpnum/$_hit_locno.
8611 - updated the tests and user manual accordingly.
8612 User manual also explictly describes that $_hit_locno is set to 1
8613 for a breakpoint with a single location.
8614 - The variable values are now set in bpstat_do_actions_1 so that
8615 they are set for silent breakpoints, and when several breakpoints
8616 are hit at the same time, that the variables are set to the printed
8619 The changes in RFA v3 handle the additional comments of Eli:
8622 - Use 'code location' instead of 'location'.
8624 - Ensure that disable $bkptno and disable $bkptno.$locno have
8625 each their explanation inthe example
8626 - Reworded the 'breakpoint-hit' paragraph.
8628 - Use 'code location' instead of 'location'.
8629 - Add a note to clarify the distinction between $bkptno and $bpnum.
8630 - Use @kbd instead of examples with only one command.
8632 Compared to RFA v1, the changes in v2 handle the comments given by
8633 Keith Seitz and Eli Zaretskii:
8634 - Use %s for the result of paddress
8635 - Use bkptno_numopt_re instead of 2 different -re cases
8637 - Add index entries for $bkptno and $locno
8638 - Added an example for "locno" in the mi interface
8639 - Added examples in the Break command manual.
8641 2022-11-19 Tsukasa OI <research_trasio@irq.a4lg.com>
8643 RISC-V: Add 'Ssstateen' extension and its CSRs
8644 This commit adds 'Ssstateen' extension, which is a supervisor-visible view
8645 of the 'Smstateen' extension. It means, this extension implements sstateen*
8646 and hstateen* CSRs of the 'Smstateen' extension.
8648 Note that 'Smstateen' extension itself is unchanged but due to
8649 implementation simplicity, it is implemented so that 'Smstateen' implies
8650 'Ssstateen' (just like 'M' implies 'Zmmul').
8652 This is based on the latest version of RISC-V Profiles
8653 (version 0.9-draft, Frozen):
8654 <https://github.com/riscv/riscv-profiles/commit/226b7f643067b29abc6723fac60d5f6d3f9eb901>
8658 * elfxx-riscv.c (riscv_implicit_subsets): Update implication rules.
8659 (riscv_supported_std_s_ext) Add 'Ssstateen' extension.
8663 * config/tc-riscv.c (enum riscv_csr_class): Rename
8664 CSR_CLASS_SMSTATEEN_AND_H{,_32} to CSR_CLASS_SSSTATEEN_...
8665 Add CSR_CLASS_SSSTATEEN.
8666 (riscv_csr_address): Support new/renamed CSR classes.
8667 * testsuite/gas/riscv/csr.s: Add 'Ssstateen' extension to comment.
8668 * testsuite/gas/riscv/csr-version-1p9p1.l: Reflect changes to
8670 * testsuite/gas/riscv/csr-version-1p10.l: Likewise.
8671 * testsuite/gas/riscv/csr-version-1p11.l: Likewise.
8672 * testsuite/gas/riscv/csr-version-1p12.l: Likewise.
8673 * testsuite/gas/riscv/ssstateen-csr.s: Test for 'Ssstateen' CSRs.
8674 * testsuite/gas/riscv/ssstateen-csr.d: Likewise.
8675 * testsuite/gas/riscv/smstateen-csr-s.d: Test to make sure that
8676 supervisor/hypervisor part of 'Smstateen' CSRs are accessible from
8677 'RV32IH_Smstateen', not just from 'RV32IH_Ssstateen' that is tested
8682 * opcode/riscv-opc.h: Update DECLARE_CSR declarations with
8685 2022-11-19 GDB Administrator <gdbadmin@sourceware.org>
8687 Automatic date update in version.in
8689 2022-11-18 Simon Marchi <simon.marchi@efficios.com>
8691 gdbserver/linux-x86: move lwp declaration out of __x86_64__ region
8692 Commit 4855cbdc3d8f ("gdbserver/linux-x86: make is_64bit_tdesc accept
8693 thread as a parameter") caused this when building in 32 bits / i386
8697 In file included from /home/smarchi/src/binutils-gdb/gdbserver/linux-x86-low.cc:24:
8698 /home/smarchi/src/binutils-gdb/gdbserver/linux-x86-low.cc: In member function ‘virtual int x86_target::low_get_thread_area(int, CORE_ADDR*)’:
8699 /home/smarchi/src/binutils-gdb/gdbserver/linux-x86-low.cc:357:47: error: ‘lwp’ was not declared in this scope
8700 357 | struct thread_info *thr = get_lwp_thread (lwp);
8702 /home/smarchi/src/binutils-gdb/gdbserver/linux-low.h:709:31: note: in definition of macro ‘get_lwp_thread’
8703 709 | #define get_lwp_thread(lwp) ((lwp)->thread)
8706 This is because it moved the lwp variable declaration inside the
8707 __x86_64__ guard, making it unavailable when building in 32 bits mode.
8708 Move the lwp variable outside of the __x86_64__ region.
8710 Change-Id: I7fa3938c6b44b345c27a52c8b8d3ea12aba53e05
8712 2022-11-18 Simon Marchi <simon.marchi@efficios.com>
8714 gdbserver: use current_process in ps_getpid
8715 The following patch ("gdbserver: switch to right process in
8716 find_one_thread") makes it so find_one_thread calls into libthread_db
8717 with a current process but no current thread. This tripped on ps_getpid
8718 using current_thread in order to get the process' pid. Get the pid from
8719 `current_process ()` instead, which removes the need to have a current
8720 thread. Eventually, it would be good to get it from the
8721 gdb_ps_prochandle_t structure, to avoid the need for a current process
8724 Reviewed-By: Andrew Burgess <aburgess@redhat.com>
8725 Change-Id: I9d2fae266419199a2fbc2fde0a5104c6e0dbd2d5
8727 2022-11-18 Simon Marchi <simon.marchi@efficios.com>
8729 gdbserver/linux-x86: make is_64bit_tdesc accept thread as a parameter
8730 ps_get_thread_area receives as a parameter the lwpid it must work on.
8731 It then calls is_64bit_tdesc, which uses the current_thread as the
8732 thread to work on. However, it is not said that both are the same.
8734 This became a problem when working in a following patch that makes
8735 find_one_thread switch to a process but to no thread (current_thread ==
8736 nullptr). When libthread_db needed to get the thread area,
8737 is_64bit_tdesc would try to get the regcache of a nullptr thread.
8739 Fix that by making is_64bit_tdesc accept the thread to work on as a
8740 parameter. Find the right thread from the context, when possible (when
8741 we know the lwpid to work on). Otherwise, pass "current_thread", to
8742 retain the existing behavior.
8744 Reviewed-By: Andrew Burgess <aburgess@redhat.com>
8745 Change-Id: I44394d6be92392fa28de71982fd04517ce8a3007
8747 2022-11-18 Simon Marchi <simon.marchi@polymtl.ca>
8749 gdbserver/linux: take condition out of callback in find_lwp_pid
8750 Just a small optimization, it's not necessary to recompute lwp at each
8753 While at it, change the variable type to long, as ptid_t::lwp returns a
8756 Reviewed-By: Andrew Burgess <aburgess@redhat.com>
8757 Change-Id: I181670ce1f90b59cb09ea4899367750be2ad9105
8759 2022-11-18 Johnson Sun <j3.soon777@gmail.com>
8761 Fix deletion of FinishBreakpoints
8762 Currently, FinishBreakpoints are set at the return address of a frame based on
8763 the `finish' command, and are meant to be temporary breakpoints. However, they
8764 are not being cleaned up after use, as reported in PR python/18655. This was
8765 happening because the disposition of the breakpoint was not being set
8768 This commit fixes this issue by correctly setting the disposition in the
8769 post-stop hook of the breakpoint. It also adds a test to ensure this feature
8770 isn't regressed in the future.
8772 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=18655
8774 2022-11-18 Simon Marchi <simon.marchi@polymtl.ca>
8776 gdb: fix symtab.c build on 32 bit targets
8777 When building on Ubuntu 22.04, gcc 12, x86-64 with -m32 and -O2, I get:
8780 /home/smarchi/src/binutils-gdb/gdb/symtab.c: In member function ‘std::vector<symbol_search> global_symbol_searcher::search() const’:
8781 /home/smarchi/src/binutils-gdb/gdb/symtab.c:4961:44: error: ‘__builtin___sprintf_chk’ may write a terminating nul past the end of the destination [-Werror=format-overflow=]
8782 4961 | sprintf (tmp, "operator%.*s%s", fix, " ", opname);
8784 In file included from /usr/include/stdio.h:894,
8785 from ../gnulib/import/stdio.h:43,
8786 from /home/smarchi/src/binutils-gdb/gdb/../gdbsupport/common-defs.h:86,
8787 from /home/smarchi/src/binutils-gdb/gdb/defs.h:28,
8788 from /home/smarchi/src/binutils-gdb/gdb/symtab.c:20:
8789 In function ‘int sprintf(char*, const char*, ...)’,
8790 inlined from ‘std::vector<symbol_search> global_symbol_searcher::search() const’ at /home/smarchi/src/binutils-gdb/gdb/symtab.c:4961:16:
8791 /usr/include/i386-linux-gnu/bits/stdio2.h:38:34: note: ‘__builtin___sprintf_chk’ output between 9 and 2147483648 bytes into a destination of size 2147483647
8792 38 | return __builtin___sprintf_chk (__s, __USE_FORTIFY_LEVEL - 1,
8793 | ~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
8794 39 | __glibc_objsize (__s), __fmt,
8795 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
8796 40 | __va_arg_pack ());
8799 PR build/29798 shows a similar error message but on Solaris.
8801 Work around that by using string_printf. It is a good thing to get rid
8802 of the alloca anyway.
8804 Change-Id: Ifbac11fee3062ad7f134d596b4e2229dc5d166f9
8805 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29798
8807 2022-11-18 Andrew Burgess <aburgess@redhat.com>
8809 gdb/testsuite: rewrite gdb.cp/call-method-register.exp with dwarf assembler
8810 Convert the gdb.cp/call-method-register.exp test to make use of the
8813 The existing gdb.cp/call-method-register.exp test relies on a GCC
8814 extension - forcing a local variable into a particular named register.
8816 This means that the test will only work with Clang, and, as we have to
8817 name the register into which the variable will be placed, will only
8818 work for those targets where we've selected a suitable register,
8819 currently this is x86-64, i386, and ppc64.
8821 By switching to the DWARF assembler, the test will work with gcc and
8822 clang, and should work on most, if not all, architectures.
8824 The test creates a small structure, something that can fit within a
8825 register, and then tries to call a method on the structure from within
8826 GDB. This should fail because GDB can't take the address of the in
8827 register structure (for the `this` pointer).
8829 As the test is for a failure case, then we don't really care _which_
8830 register the structure is in, and I take advantage of this for the
8831 DWARF assembler test, I just declare that the variable is in
8832 DW_OP_reg0, whatever that might be. I've tested the new test on
8833 x86-64, ppc, aarch64, and risc-v, and the test runs, and passes on all
8834 these architectures, which is already more than we used to cover.
8836 Additionally, on x86-64, I've tested with Clang and gcc, and the test
8837 runs and passed with both compilers.
8839 Reviewed-By: Lancelot SIX <lancelot.six@amd.com>
8841 2022-11-18 Andrew Burgess <aburgess@redhat.com>
8843 gdb/testsuite: fix gdb.debuginfod/fetch_src_and_symbols.exp with Clang
8844 The gdb.debuginfod/fetch_src_and_symbols.exp test is showing a single
8845 failure when run with some older versions of Clang, e.g. 9.0.1.
8847 The problem appears to be with Clang's generated line table. The test
8848 source looks like this:
8853 asm ("main_label: .globl main_label");
8857 In GDB, when we 'start', we expect to stop at the 'return 0;' line.
8858 This is the behaviour when the compiler is gcc, or later versions of
8861 However, with Clang 9.0.2, I see GDB stop on the 'asm' line.
8863 In this commit I'll fix this issue by placing a breakpoint on the
8864 return line, and then using gdb_continue_to_breakpoint to ensure we
8865 have stopped in the correct place.
8867 Of course, using gdb_continue_to_breakpoint will only work if we are
8868 not already stopped at the breakpoint location, so I've added some
8869 filler work before the 'return 0;' line. With this done we can use
8870 gdb_continue_to_breakpoint in all cases.
8872 As a result of adding the new filler work, one of the later tests,
8873 that used the 'list' command, no longer see the correct expected
8874 output (the top line of the source file is no longer included in the
8875 output). I've fixed this by listing a known specific line, the test
8876 is checking that GDB managed to find the source file, it doesn't
8877 matter which source line we list, as long as we can list something.
8879 2022-11-18 Andrew Burgess <aburgess@redhat.com>
8881 gdb/testsuite: rename source file gdb.debuginfod/main.c
8882 The test gdb.debuginfod/fetch_src_and_symbols.exp uses a source file
8883 named main.c. I can't see any particular reason why the file is named
8886 Usually test source files are named after the test script.
8888 This commit just renames the source file inline with the test script,
8889 and updates the call to standard_testfile (removing the reference to
8892 There's no particular reason for this change other than seeing the
8893 file named main.c made me thing that the source file must be shared
8894 with some other test (it isn't).
8896 There should be no change in what is tested after this commit.
8898 2022-11-18 Andrew Burgess <aburgess@redhat.com>
8900 gdb/testsuite: add (and use) a new build-id compile option
8901 I noticed that the gdb.debuginfod/fetch_src_and_symbols.exp test was
8902 failing when run with Clang as the compiler.
8904 This test relies on the compiled binaries having a build-id within
8905 them. For GCC, really GNU ld, the default is to always include a
8908 When compiling with Clang though, the default is for no build-id.
8910 I did consider *always* turning on the build-id feature when the
8911 compiler is Clang, but that felt a little weird.
8913 Instead, I propose that we add a new 'build-id' compiler option to
8914 gdb_compile, this flag indicates that the test _requires_ a build-id.
8915 In gcc_compile we can then add the required flags if the compiler is
8916 Clang so that we do get a build-id.
8918 With this change the gdb.debuginfod/fetch_src_and_symbols.exp test
8919 now (mostly) passes with Clang 9.0.1 and 15.0.2, and still passes with
8920 gcc. The 'mostly' part is an unrelated issue, and will be addressed
8921 in a later commit in this series.
8923 Reviewed-By: Lancelot SIX <lancelot.six@amd.com>
8925 2022-11-18 Andrew Burgess <aburgess@redhat.com>
8927 gdb/testsuite: fix gdb.compile/compile-ops.exp with clang
8928 I noticed that the gdb.compile/compile-ops.exp test was failing when
8929 run with Clang as the compiler.
8931 This test makes use of the DWARF assembler, and, it turns out, uses
8932 a technique which is not portable to Clang. This problem is
8933 described in the comment on the function_range proc in lib/dwarf.exp,
8936 # If the compiler is gcc, we can do the following to get function start
8937 # and end address too:
8939 # asm ("func_start: .globl func_start");
8940 # static void func (void) {}
8941 # asm ("func_end: .globl func_end");
8943 # however, this isn't portable, because other compilers, such as clang,
8944 # may not guarantee the order of global asms and function. The code
8947 # asm ("func_start: .globl func_start");
8948 # asm ("func_end: .globl func_end");
8949 # static void func (void) {}
8951 These start/end labels are used for computing the function start, end,
8952 and length. The portable solution is to place a label within the
8953 function, like this:
8957 # asm ("main_label: .globl main_label");
8961 And make use of 'proc function_range' (from lib/dwarf.exp).
8963 So, that's what I do in this commit.
8965 One consequence of this change is that we need to compile the source
8966 file, and have it loaded into a GDB session, before calling
8967 function_range, so I've added an early call to prepare_for_testing.
8969 Additionally, this test script was generating the DWARF assembler into
8970 a file called gdbjit-ops.S, I suspect a copy and paste issue there, so
8971 I've switched this to use compile-ops-dbg.S instead, which is more
8972 inline with what other DWARF assembler tests do.
8974 The only other change, which might be a problem, is that I also
8975 deleted these two lines from the source file:
8977 asm (".section \".text\"");
8980 These lines were setting the alignment of the .text section. What I
8981 don't know is whether this was significant or not. If it is
8982 significant, then I can't see why.
8984 On x86-64, the test still passes fine without these lines, but that
8985 doesn't mean the test wont start failing on some other architecture.
8987 Still, I figure, lets remove them, then, if/when we find a test that
8988 starts failing, we can add the lines back, along with an explanation
8989 for why the extra alignment is required.
8991 But, if people would prefer to be more conservative, then I'm happy to
8992 just add the lines back.
8994 Reviewed-By: Lancelot SIX <lancelot.six@amd.com>
8996 2022-11-18 Andrew Burgess <aburgess@redhat.com>
8998 gdb/testsuite: fix gdb.trace/unavailable-dwarf-piece.exp with Clang
8999 I noticed that the test gdb.trace/unavailable-dwarf-piece.exp was
9000 failing when run with Clang. Or rather, the test was not running as
9001 the test executable failed to compile.
9003 The problem is that Clang was emitting this warning:
9005 warning: argument unused during compilation: '-fdiagnostics-color=never' [-Wunused-command-line-argument]
9007 This warning is emitted when compiling the assembler file generated
9008 by the DWARF assembler.
9010 Most DWARF assembler tests generate the assembler file into a file
9011 with the '.S' extension. However, this particular test uses a '.s'
9014 Now a .S file will be passed through the preprocessor, while a .s will
9015 be sent straight to the assembler. My guess is that Clang doesn't
9016 support the -fdiagnostics-color=never option for the assembler, but
9017 does for the preprocessor.
9019 That's a little annoying, but easily worked around. We don't care if
9020 our assembler file is passed through the preprocessor, so, in this
9021 commit, I just change the file extension from .s to .S, and the
9024 Currently, the unavailable-dwarf-piece.exp script names the assembler
9025 file using standard_output_file, in this commit I've switched to make
9026 use of standard_testfile, as that seems to be the more common way of
9027 doing this sort of thing.
9029 With these changes the test now passes with Clang 9.0.1 and 15.0.2,
9030 and also still passes with gcc.
9032 Reviewed-By: Lancelot SIX <lancelot.six@amd.com>
9034 2022-11-18 Andrew Burgess <aburgess@redhat.com>
9036 gdb/testsuite: don't avoid DWARF assembler tests with Clang
9037 Two tests make the claim that the DWARF assembler requires gcc,
9038 however, this isn't true. I think at one point, when the DWARF
9039 assembler was first added, we did use some techniques that were not
9040 portable (see the comments in lib/dwarf.exp on function_range for
9041 details), however, I think most DWARF assembler tests will now work
9044 The two tests that I modify in this commit both work fine with Clang,
9045 at least, I've tested with Clang 9.0.1 and 15.0.2, and don't see any
9046 problems, so I'm removing the early return logic that stops these
9047 tests from running with Clang.
9049 Reviewed-By: Lancelot SIX <lancelot.six@amd.com>
9051 2022-11-18 Zac Walker <zac.walker@linaro.org>
9053 GAS fix alignment for aarch64-pe
9054 Fixes issue where various values of '.align' causes writing of COFF files to fail.
9055 Specific to the aarch64-pe target.
9057 2022-11-18 Alan Modra <amodra@gmail.com>
9059 PR29799 heap buffer overflow in display_gdb_index dwarf.c:10548
9061 * dwarf.c (display_gdb_index): Typo fix.
9064 * coff-stgo32 (go32exe_check_format): Sanity check stubsize against
9065 filesize before malloc.
9067 Regen potfiles for sframe
9069 2022-11-18 GDB Administrator <gdbadmin@sourceware.org>
9071 Automatic date update in version.in
9073 2022-11-17 Indu Bhagat <indu.bhagat@oracle.com>
9075 [gas, aarch64]: fix build breakage for aarch64-pe
9076 SFrame is supported for ELF only. Keep the definitions and declarations
9077 guarded with OBJ_ELF consistently.
9081 * gas/config/tc-aarch64.h: Guard SFrame related definitions
9084 2022-11-17 Tom Tromey <tromey@adacore.com>
9086 Fix static initialization order problem in windows-nat.c
9087 This patch fixes a static initialization order problem in
9088 windows-nat.c that was pointed out by Jon Turney. The underlying
9089 problem is that the windows_nat_target constructor relies on
9090 serial_logfile already being constructed, but this is not enforced by
9091 C++ rules. This patch fixes the problem by initializing the global
9092 windows_nat_target later.
9094 2022-11-17 H.J. Lu <hjl.tools@gmail.com>
9096 opcodes: Define NoSuf in i386-opc.tbl
9097 Use NoSuf to replace No_bSuf|No_wSuf|No_lSuf|No_sSuf|No_qSuf|No_ldSuf
9098 and add the explicit NoSuf to AddrPrefixOpReg in templates.
9100 * i386-opc.tbl (NoSuf): New macro.
9101 (AddrPrefixOpReg): Remove No_?Suf.
9102 Replace No_bSuf|No_wSuf|No_lSuf|No_sSuf|No_qSuf|No_ldSuf with
9104 Add NoSuf to AddrPrefixOpReg in templates.
9106 2022-11-17 H.J. Lu <hjl.tools@gmail.com>
9108 i386: Move i386_seg_prefixes to gas
9111 * config/tc-i386.c (i386_seg_prefixes): New. Moved from opcodes.
9115 * i386-opc.c (i386_seg_prefixes): Removed.
9116 * i386-opc.h (i386_seg_prefixes): Likewise.
9118 2022-11-17 Carl Love <cel@us.ibm.com>
9120 PowerPC, fix gdb.base/retval-large-struct.exp
9121 Support for printining non-trivial return values was recently added in
9124 commit a0eda3df5b750ae32576a9be092b361281a41787
9125 Author: Carl Love <cel@us.ibm.com>
9126 Date: Mon Nov 14 16:22:37 2022 -0500
9128 PowerPC, fix support for printing the function return value for non-trivial values.
9130 The functionality can now be used to fix gdb.base/retval-large-struct.exp.
9131 The test just needs to be compiled with -fvar-tracking to enable GDB to
9132 determine the address off the return buffer when the function is called.
9134 The current output from the test:
9136 34 return big_struct;
9137 (gdb) PASS: gdb.base/retval-large-struct.exp: continue to breakpoint: Break in print_large_struct
9139 warning: Cannot determine the function return value.
9140 Try compiling with -fvar-tracking.
9141 Run till exit from #0 return_large_struct () at binutils-gdb-current/gdb/testsuite/gdb.base/retval-large-struct.c:34
9142 main (argc=1, argv=0x7fffffffcd58) at binutils-gdb-current/gdb/testsuite/gdb.base/retval-large-struct.c:44
9144 Value returned has type: struct big_struct_t. Cannot determine contents
9145 (gdb) FAIL: gdb.base/retval-large-struct.exp: finish from return_large_struct
9146 testcase binutils-gdb-current/gdb/testsuite/gdb.base/retval-large-struct.exp completed in 1 seconds
9148 This patch adds the command line argument -fvar-tracking to enable gdb to
9149 determine the return vaule and thus fixing the test.
9151 Patch tested on Power 10 with no regressions.
9153 2022-11-17 Tom Tromey <tromey@adacore.com>
9155 Use boolean literals for pagination_enabled
9156 I noticed a couple of spots that used '0' rather than 'false' when
9157 modifying pagination_enabled. This patch cleans these up.
9159 2022-11-17 Carl Love <cel@us.ibm.com>
9161 Change NULL to nullptr in gdb/infcmd.c and gdb/infrun.c
9162 The GDB coding standard specifies that nullptr should be used instead of
9163 NULL. There are numerous uses of NULL and nullptr in files infcmd.c and
9164 infrun.c. This patch replaces the various uses of NULL with nullptr in
9165 the source files. The use of NULL in the comments was not changed.
9167 The patch does not introduce any functional changes.
9169 The patch has been tested on PowerPC and Intel X86_64 with no new unexpected
9170 test failures, unresolved tests, new core files etc.
9172 2022-11-17 H.J. Lu <hjl.tools@gmail.com>
9174 ld: Always call elf_backend_output_arch_local_syms
9175 Always call elf_backend_output_arch_local_syms since only the backend
9176 knows if elf_backend_output_arch_local_syms is needed when all symbols
9177 are striped. elf_backend_output_arch_local_syms is defined only for
9178 x86, ARM and AARCH64. On x86, elf_backend_output_arch_local_syms must
9179 be called to handle local IFUNC symbols even if all symbols are striped.
9180 Update ARM and AARCH64 to skip elf_backend_output_arch_local_syms when
9181 symbols aren't needed.
9186 * elf32-arm.c (elf32_arm_output_arch_local_syms): Skip if symbols
9188 * elfnn-aarch64.c (elfNN_aarch64_output_arch_local_syms):
9190 * elflink.c (bfd_elf_final_link): Always call
9191 elf_backend_output_arch_local_syms if available.
9196 * testsuite/ld-elf/linux-x86.exp: Run PR ld/29797 test.
9197 * testsuite/ld-elf/pr29797.c: New file.
9199 2022-11-17 Andrew Burgess <aburgess@redhat.com>
9201 gdb: new $_inferior_thread_count convenience variable
9202 Add a new convenience variable $_inferior_thread_count that contains
9203 the number of live (non-exited) threads in the current inferior. This
9204 can be used in command scripts, or breakpoint conditions, etc to
9205 adjust the behaviour for multi-threaded inferiors.
9207 This value is only stable in all-stop mode. In non-stop mode, where
9208 new threads can be started, and existing threads exit, at any time,
9209 this convenience variable can give a different value each time it is
9212 2022-11-17 Tom Tromey <tromey@adacore.com>
9214 Remove two obsolete declarations
9215 I happened to find a couple of obsolete declarations in cli-interp.h.
9216 This patch removes them. Tested by rebuilding.
9218 2022-11-17 Andrew Burgess <aburgess@redhat.com>
9220 gdb/testsuite: fix failure in gdb.python/py-send-packet.exp
9221 While working on another patch I noticed that, when run on an AArch64
9222 target, the test gdb.python/py-send-packet.exp was failing:
9224 Traceback (most recent call last):
9225 File "<string>", line 1, in <module>
9226 File "/tmp/build/gdb/testsuite/outputs/gdb.python/py-send-packet/py-send-packet.py",
9227 line 106, in run_auxv_send_packet_test
9228 assert string == expected_result
9230 Error while executing Python code.
9231 (gdb) FAIL: gdb.python/py-send-packet.exp: call python run_auxv_send_packet_test function
9233 The test uses 'maint packet ...' to send a packet to gdbserver, and
9234 then captures the output in TCL. This output is then passed through
9235 to a Python function, which performs some actions using the Python
9236 API, and compares the results from the Python API to the results
9237 captured in TCL from 'maint packet ...'.
9239 The problem is that the output captured in TCL contains lots of things
9240 like '\x000', when this is passed through to Python the '\x' causes
9241 this to be treated as an escape code, which isn't what we want - we
9242 want the actual string "\x000".
9244 So, in the TCL part of the test we were expanding '\x' to '\\x', this
9245 seemed to work fine for my testing on x86-64.
9247 However, on AArch64 what I see is that the results from 'maint packet
9248 ...' contain a literal '\' character followed by a literal 'x'
9249 character. When GDB prints this in the 'maint packet' output, GDB
9250 escapes the '\' for us, thus we get '\\x' printed by 'maint packet'.
9252 However, now our TCL test script kicks in and tries to "fix" the '\x',
9253 this means we now have '\\\x', which isn't correct.
9255 The problem is that in the TCL script we are too restrictive, we
9256 expand '\x' to '\\x', but really, we should be expanding all '\'
9257 characters, regardless of what follows them. This is what this patch
9260 After this the gdb.python/py-send-packet.exp test passes on AArch64
9263 2022-11-17 Aditya Vidyadhar Kamath <Aditya.Kamath1@ibm.com>
9265 Fix call functions command bug in 64 bits programs for AIX
9266 In AIX for 64 bit programs we need to zero extend variables
9267 of integer or enum or char data type.
9269 Otherwise a zero will get dumped in the register as we memset
9270 our word to 0 and we copy non zero extended contents to the cache.
9272 2022-11-17 Andrew Burgess <aburgess@redhat.com>
9274 gdb/fortran/testsuite: print values and types of string variables
9275 While looking through the Fortran tests, I couldn't find a test of GDB
9276 printing the value and type of a Fortran string defined using the
9277 'character*SIZE' notation.
9279 This works fine in GDB right now, but I thought it wouldn't hurt to
9280 have a test for this, so this commit adds such a test.
9282 The test also includes printing a string that includes some embedded
9283 special characters: \n \r \t \000 - that's right, as Fortran strings
9284 are stored as an address and length, it is fine to include an embedded
9285 null, so this test includes an example of that.
9287 Standard Fortran doesn't support backslash escape sequences within
9288 strings, the special characters must be generated using the `achar`
9289 function. However, when GDB prints the strings we currently print
9290 using the standard C like backslash sequences.
9292 I'm not currently proposing to change that behaviour, the backslash
9293 sequences are more compact than the standard Fortran way of doing
9294 things, and are so widely used that I suspect most Fortran programmers
9295 will understand them.
9297 2022-11-17 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9299 Fix various procfs.c compilation errors
9300 procfs.c has accumulated several compilation errors lately (some of them
9301 new with GCC 12), which are fixed by this patch:
9305 /vol/src/gnu/gdb/hg/master/dist/gdb/procfs.c:144:7: error:
\91int
9306 procfs_target::auxv_parse(gdb_byte**, gdb_byte*, CORE_ADDR*, CORE_ADDR*)
\92
9307 marked
\91override
\92, but does not override
9308 144 | int auxv_parse (gdb_byte **readptr,
9311 Obviouly, procfs.c was missed in the auxv_parse constification.
9313 * dead_procinfo has:
9315 /vol/src/gnu/gdb/hg/master/dist/gdb/procfs.c: In function
\91void
9316 dead_procinfo(procinfo*, const char*, int)
\92:
9317 /vol/src/gnu/gdb/hg/master/dist/gdb/procfs.c:563:11: warning: the address
9318 of
\91procinfo::pathname
\92 will never be NULL [-Waddress]
9319 563 | if (pi->pathname)
9321 /vol/src/gnu/gdb/hg/master/dist/gdb/procfs.c:238:8: note:
9322 \91procinfo::pathname
\92 declared here
9323 238 | char pathname[MAX_PROC_NAME_SIZE]; /* Pathname to /proc entry */
9326 The warning is correct, so the code can lose support for the NULL
9329 * create_inferior has this ugly warning:
9331 /vol/src/gnu/gdb/hg/master/dist/gdb/procfs.c: In member function
\91virtual void procfs_target::create_inferior(const char*, const std::string&, char**, int)
\92:
9332 /vol/src/gnu/gdb/hg/master/dist/gdb/procfs.c:2815:19: warning:
\91char* std::strncpy(char*, const char*, size_t)
\92 output truncated before terminating nul copying as many bytes from a string as its length [-Wstringop-truncation]
9333 2815 | strncpy (tryname, p, len);
9334 | ~~~~~~~~^~~~~~~~~~~~~~~~~
9335 /vol/src/gnu/gdb/hg/master/dist/gdb/procfs.c:2814:26: note: length computed here
9336 2814 | len = strlen (p);
9339 It seems that this is another case of GCC PR middle-end/88059, which
9340 Martin Sebor refuses to fix. So I'm using the hack suggested in the
9341 PR to use memcpy instead of strncpy.
9343 * find_memory_regions_callback fails with
9345 /vol/src/gnu/gdb/hg/master/dist/gdb/procfs.c: In function
\91int find_memory_regions_callback(prmap*, find_memory_region_ftype, void*)
\92:
9346 /vol/src/gnu/gdb/hg/master/dist/gdb/procfs.c:3167:18: error: too few arguments to function
9347 3167 | return (*func) ((CORE_ADDR) map->pr_vaddr,
9348 | ~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~
9349 3168 | map->pr_size,
9351 3169 | (map->pr_mflags & MA_READ) != 0,
9352 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
9353 3170 | (map->pr_mflags & MA_WRITE) != 0,
9354 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
9355 3171 | (map->pr_mflags & MA_EXEC) != 0,
9356 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
9357 3172 | 1, /* MODIFIED is unknown, pass it as true. */
9358 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
9362 Again, procfs.c was overlooked when adding the new memory_tagged arg.
9363 Unfortunately, it wasn't even documented in gdb/defs.h when it was
9366 commit 68cffbbd4406b4efe1aa6e18460b1d7ca02549f1
9367 Author: Luis Machado <luis.machado@arm.com>
9368 Date: Thu Mar 31 11:42:35 2022 +0100
9370 [AArch64] MTE corefile support
9372 With those changes, procfs.c compiles again. Together with the hack
9373 from the Solaris gdbsupport breakage reported in PR build/29791, I was
9374 able to build and test gdb on both amd64-pc-solaris2.11 and
9375 sparcv9-sun-solaris2.11.
9377 Approved-By: Simon Marchi <simon.marchi@efficios.com>
9379 2022-11-17 Christoph Müllner <christoph.muellner@vrull.eu>
9381 RISC-V: Add T-Head Int vendor extension
9382 This patch adds the XTheadInt extension, which provides interrupt
9383 stack management instructions.
9385 The XTheadFmv extension is documented in the RISC-V toolchain
9387 https://github.com/riscv-non-isa/riscv-toolchain-conventions
9389 Co-developed-by: Lifang Xia <lifang_xia@linux.alibaba.com>
9391 2022-11-17 Christoph Müllner <christoph.muellner@vrull.eu>
9393 RISC-V: Add T-Head Fmv vendor extension
9394 This patch adds the XTheadFmv extension, which allows to access the
9395 upper 32 bits of a double-precision floating-point register in RV32.
9397 The XTheadFmv extension is documented in the RISC-V toolchain
9399 https://github.com/riscv-non-isa/riscv-toolchain-conventions
9401 Co-developed-by: Lifang Xia <lifang_xia@linux.alibaba.com>
9403 2022-11-17 Tom de Vries <tdevries@jostaberry-8.arch.suse.de>
9405 [gdb/testsuite] Fix DUPLICATE in gdb.arch/ppc-fp.exp
9408 DUPLICATE: gdb.arch/ppc-fp.exp: next
9411 Fix this by adding unique test names.
9413 Tested on powerpc64le-linux.
9415 2022-11-17 GDB Administrator <gdbadmin@sourceware.org>
9417 Automatic date update in version.in
9419 2022-11-16 Kévin Le Gouguec <legouguec@adacore.com>
9421 Add myself to the gdb/MAINTAINERS write-after-approval list
9423 2022-11-16 Kévin Le Gouguec <legouguec@adacore.com>
9425 Guard against frame.c destructors running before frame-info.c's
9426 On x86_64-windows, since 04e2ac7b2a7, we observe this internal error:
9428 [...]/gdbsupport/intrusive_list.h:458: internal-error: erase_element:
9429 Assertion `elem_node->prev != INTRUSIVE_LIST_UNLINKED_VALUE' failed.
9431 Breaking in the destructors for intrusive_list and frame_info_ptr shows that
9432 in this configuration, the destructors for frame.c's statically-stored
9433 objects are run before frame-info.c's:
9435 Thread 1 hit Breakpoint 7, intrusive_list<frame_info_ptr,
9436 intrusive_base_node<frame_info_ptr> >::~intrusive_list (this=0x7ff69c418c90
9437 <frame_info_ptr::frame_list>, __in_chrg=<optimized out>)
9438 [...]/../gdbsupport/intrusive_list.h:250
9441 #0 intrusive_list<frame_info_ptr, intrusive_base_node<frame_info_ptr> >
9442 ::~intrusive_list (this=0x7ff69c418c90 <frame_info_ptr::frame_list>,
9443 __in_chrg=<optimized out>) [...]/../gdbsupport/intrusive_list.h:250
9444 #1 0x00007ff69b78edba in __tcf_1 () [...]/frame-info.c:27
9445 #2 0x00007ff9c457aa9f in msvcrt!_initterm_e ()
9446 from C:\Windows\System32\msvcrt.dll
9447 #3 0x00007ff69b8246a6 in captured_main_1 (context=0x5ffe00)
9449 #4 0x00007ff69b825149 in captured_main (data=0x5ffe00) [...]/main.c:1320
9450 #5 0x00007ff69b8251b1 in gdb_main (args=0x5ffe00) [...]/main.c:1345
9451 #6 0x00007ff69b5d1730 in main (argc=2, argv=0x751730) [...]/gdb.c:32
9455 Thread 1 hit Breakpoint 8, frame_info_ptr::~frame_info_ptr
9456 (this=0x7ff69c418e20 <selected_frame>, __in_chrg=<optimized out>)
9457 [...]/frame-info.h:74
9458 74 if (is_linked ())
9460 #0 frame_info_ptr::~frame_info_ptr (this=0x7ff69c418e20 <selected_frame>,
9461 __in_chrg=<optimized out>) [...]/frame-info.h:74
9462 #1 0x00007ff69b79a643 in __tcf_1 () [...]/frame.c:1675
9463 #2 0x00007ff9c457aa9f in msvcrt!_initterm_e () from
9464 C:\Windows\System32\msvcrt.dll
9465 #3 0x00007ff69b8246a6 in captured_main_1 (context=0x5ffe00)
9467 #4 0x00007ff69b825149 in captured_main (data=0x5ffe00) [...]/main.c:1320
9468 #5 0x00007ff69b8251b1 in gdb_main (args=0x5ffe00) [...]/main.c:1345
9469 #6 0x00007ff69b5d1730 in main (argc=2, argv=0x751730) [...]/gdb.c:32
9471 Approved-By: Simon Marchi <simon.marchi@efficios.com>
9473 2022-11-16 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9475 PR29788, gprofng cannot display Java's generated assembly code
9477 2022-11-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9480 * src/Experiment.h: Declare dyntext_name.
9481 * src/Experiment.cc: Use dyntext_name to initialize img_fname.
9483 2022-11-16 Tom de Vries <tdevries@suse.de>
9485 [gdb/testsuite] Use gdb_gcore_cmd in gdb.threads/gcore-thread.exp
9486 I noticed a plain gcore command in test-case gdb.threads/gcore-thread.exp:
9488 gdb_test "gcore $core0file" "Saved corefile .*" \
9489 "save a zeroed-threads corefile"
9492 Use gdb_gcore_cmd instead.
9494 Tested on x86_64-linux.
9496 2022-11-16 Carl Love <cel@us.ibm.com>
9498 Bug fix in commit for printing the function return value for non-trivial values
9501 commit a0eda3df5b750ae32576a9be092b361281a41787
9502 Author: Carl Love <cel@us.ibm.com>
9503 Date: Mon Nov 14 16:22:37 2022 -0500
9505 PowerPC, fix support for printing the function return value for non-trivial values.
9507 Is generating a segmentation fault on x86_64-linux.
9511 PASS: gdb.asm/asm-source.exp: info source asmsrc1.s
9512 ERROR: GDB process no longer exists
9513 UNRESOLVED: gdb.asm/asm-source.exp: finish from foo3
9516 Reproduced on command line:
9518 $ gdb -q -batch -x outputs/gdb.asm/asm-source/gdb.in.1
9521 The problem seems to be that:
9523 Thread 1 "gdb" received signal SIGSEGV, Segmentation fault.
9524 0x000000000043de7a in symbol::type (this=0x0) at
9525 .../gdb_versions/devel/src/gdb/symtab.h:1287
9531 #1 0x0000000000852d94 in finish_command (arg=0x0, from_tty=0)
9532 at .../gdb_versions/devel/src/gdb/infcmd.c:1887
9533 1887 = check_typedef (sm->function->type ()->target_type ());
9534 (gdb) p sm->function
9537 The code is not checking if sm->function is NULL. If sm->function is NULL
9538 the check for the return buffer should be skipped.
9540 2022-11-16 Tom Tromey <tromey@adacore.com>
9542 Update Ada tasks documentation
9543 My co-worker Kévin noticed that the Ada tasks documentation is
9544 slightly out of date -- it does not document all the states that can
9545 be reported by ada-tasks.c.
9547 This patch adds the missing states to the appropriate node, and
9548 updates one state to reflect a change made some time ago.
9550 2022-11-16 Pedro Alves <palves@redhat.com>
9552 gdb: add "set style tui-current-position on|off", default to off
9555 https://sourceware.org/pipermail/gdb-patches/2020-June/169519.html
9557 this patch disables source and assembly code highlighting for the
9558 text highlighted by the TUI's current position indicator, and adds a
9559 command to enable it back.
9561 2022-11-16 Tom de Vries <tdevries@suse.de>
9563 [gdb/testsuite] Modernize gdb.arch/i386-biarch-core.exp
9564 I noticed in test-case gdb.arch/i386-biarch-core.exp that we run into the
9565 completion limit for "complete set gnutarget":
9567 set gnutarget vms-libtxt^M
9568 set gnutarget *** List may be truncated, max-completions reached. ***^M
9569 (gdb) PASS: gdb.arch/i386-biarch-core.exp: complete set gnutarget
9572 Fix this by using get_set_option_choices.
9574 Also use get_set_option_choices for "complete set architecture i386", which
9575 required extending get_set_option_choices to accept a second argument, such
9578 set archs [get_set_option_choices "set architecture" "i386"]
9580 because this returns an empty list:
9582 set archs [get_set_option_choices "set architecture i386"]
9584 because it does "complete set architecture i386 ".
9586 Also clean up the explicit gdb_exit/gdb_start and use clean_restart instead.
9588 Tested on x86_64-linux.
9590 2022-11-16 Tom de Vries <tdevries@suse.de>
9592 [gdb/testsuite] Fix gdb.arch/ppc64-symtab-cordic.exp without bzip2
9593 After de-installing bzip2, I run into:
9595 Running ppc64-symtab-cordic.exp ...
9596 sh: bzip2: command not found
9597 PATH: gdb.arch/ppc64-symtab-cordic.exp: failed bzip2 for \
9598 src/gdb/testsuite/gdb.arch/cordic.ko.bz2
9602 - using remote_exec instead of catch system, and
9603 - using file tail in the untested message.
9605 I've tried making output redirection work with remote_exec, but that seems to
9606 be broken, so we now:
9607 - copy the file $f.bz2 into the desired location $dir/$f.bz2, and
9608 - decompress the bz2 file using "bzip2 -df $dir/$f.bz2", resulting in a file
9611 Factor out new function decompress_bz2 to make the test-case less verbose, and
9612 also use it in gdb.arch/i386-biarch-core.exp.
9614 Tested on x86_64-linux, without and with bzip2 installed.
9616 2022-11-16 GDB Administrator <gdbadmin@sourceware.org>
9618 Automatic date update in version.in
9620 2022-11-15 Indu Bhagat <indu.bhagat@oracle.com>
9622 doc: add SFrame spec file
9625 * libsframe/Makefile.am: Add info-in-builddir to
9626 AUTOMAKE_OPTIONS. Include doc/local.mk.
9627 * libsframe/Makefile.in: Regenerated.
9628 * libsframe/configure: Likewise.
9629 * libsframe/configure.ac: Check for makeinfo and set BUILD_INFO.
9630 * libsframe/doc/local.mk: New file.
9631 * libsframe/doc/sframe-spec.texi: Likewise.
9633 2022-11-15 Indu Bhagat <indu.bhagat@oracle.com>
9635 gas/NEWS: add text about new command line option and SFrame support
9638 * gas/NEWS: Add SFrame related news.
9640 2022-11-15 Indu Bhagat <indu.bhagat@oracle.com>
9642 binutils/NEWS: add text for SFrame support
9645 * binutils/NEWS: Add item for SFrame support.
9647 2022-11-15 Indu Bhagat <indu.bhagat@oracle.com>
9649 src-release.sh: Add libsframe
9650 Add libsframe to the list of top level directories that will be included
9655 * src-release.sh: Add libsframe
9657 2022-11-15 Indu Bhagat <indu.bhagat@oracle.com>
9659 readelf/objdump: support for SFrame section
9660 This patch adds support for SFrame in readelf and objdump. The arguments
9661 of --sframe are optional for both readelf and objdump.
9665 * sframe-api.h (dump_sframe): New function declaration.
9669 * binutils/Makefile.am: Add dependency on libsframe for
9670 readelf and objdump.
9671 * binutils/Makefile.in: Regenerate.
9672 * binutils/doc/binutils.texi: Document --sframe=[section].
9673 * binutils/doc/sframe.options.texi: New file.
9674 * binutils/objdump.c: Add support for SFrame format.
9675 * binutils/readelf.c: Likewise.
9676 * include/sframe-api.h: Add new API for dumping .sframe
9678 * libsframe/Makefile.am: Add sframe-dump.c.
9679 * libsframe/Makefile.in: Regenerate.
9680 * libsframe/sframe-dump.c: New file.
9682 2022-11-15 Indu Bhagat <indu.bhagat@oracle.com>
9684 bfd: linker: merge .sframe sections
9685 The linker merges all the input .sframe sections. When merging, the
9686 linker verifies that all the input .sframe sections have the same
9689 The linker uses libsframe library to perform key actions on the
9690 .sframe sections - decode, read, and create output data. This
9691 implies buildsystem changes to make and install libsframe before
9694 The linker places the output .sframe section in a new segment of its
9695 own: PT_GNU_SFRAME. A new segment is not added, however, if the
9696 generated .sframe section is empty.
9698 When a section is discarded from the final link, the corresponding
9699 entries in the .sframe section for those functions are also deleted.
9701 The linker sorts the SFrame FDEs on start address by default and sets
9702 the SFRAME_F_FDE_SORTED flag in the .sframe section.
9704 This patch also adds support for generation of SFrame unwind
9705 information for the .plt* sections on x86_64. SFrame unwind info is
9706 generated for IBT enabled PLT, lazy/non-lazy PLT.
9708 The existing linker option --no-ld-generated-unwind-info has been
9709 adapted to include the control of whether .sframe unwind information
9710 will be generated for the linker generated sections like PLT.
9712 Changes to the linker script have been made as necessary.
9716 * Makefile.def: Add install dependency on libsframe for libbfd.
9717 * Makefile.in: Regenerated.
9718 * bfd/Makefile.am: Add elf-sframe.c
9719 * bfd/Makefile.in: Regenerated.
9720 * bfd/bfd-in2.h (SEC_INFO_TYPE_SFRAME): Regenerated.
9721 * bfd/configure: Regenerate.
9722 * bfd/configure.ac: Add elf-sframe.lo.
9723 * bfd/elf-bfd.h (struct sframe_func_bfdinfo): New struct.
9724 (struct sframe_dec_info): Likewise.
9725 (struct sframe_enc_info): Likewise.
9726 (struct elf_link_hash_table): New member for encoded .sframe
9728 (struct output_elf_obj_tdata): New member.
9729 (elf_sframe): New access macro.
9730 (_bfd_elf_set_section_sframe): New declaration.
9731 * bfd/elf.c (get_segment_type): Handle new segment
9733 (bfd_section_from_phdr): Likewise.
9734 (get_program_header_size): Likewise.
9735 (_bfd_elf_map_sections_to_segments): Likewise.
9736 * bfd/elf64-x86-64.c (elf_x86_64_link_setup_gnu_properties): Add
9737 contents to the .sframe sections or .plt* entries.
9738 * bfd/elflink.c (elf_section_ignore_discarded_relocs): Handle
9739 SEC_INFO_TYPE_SFRAME.
9740 (_bfd_elf_default_action_discarded): Handle .sframe section.
9741 (elf_link_input_bfd): Merge .sframe section.
9742 (bfd_elf_final_link): Write the output .sframe section.
9743 (bfd_elf_discard_info): Handle discarding .sframe section.
9744 * bfd/elfxx-x86.c (_bfd_x86_elf_size_dynamic_sections): Create
9745 .sframe section for .plt and .plt.sec.
9746 (_bfd_x86_elf_finish_dynamic_sections): Handle .sframe from
9748 * bfd/elfxx-x86.h (PLT_SFRAME_FDE_START_OFFSET): New
9750 (SFRAME_PLT0_MAX_NUM_FRES): Likewise.
9751 (SFRAME_PLTN_MAX_NUM_FRES): Likewise.
9752 (struct elf_x86_sframe_plt): New structure.
9753 (struct elf_x86_link_hash_table): New member.
9754 (struct elf_x86_init_table): New members for .sframe
9756 * bfd/section.c: Add new definition SEC_INFO_TYPE_SFRAME.
9757 * binutils/readelf.c (get_segment_type): Handle new segment
9759 * ld/ld.texi: Update documentation for
9760 --no-ld-generated-unwind-info.
9761 * ld/scripttempl/elf.sc: Support .sframe sections.
9762 * ld/Makefile.am (TESTSFRAMELIB): Use it.
9763 (check-DEJAGNU): Likewise.
9764 * ld/Makefile.in: Regenerated.
9765 * ld/configure.ac (TESTSFRAMELIB): Set to the .so or .a like TESTBFDLIB.
9766 * ld/configure: Regenerated.
9767 * bfd/elf-sframe.c: New file.
9771 * elf/common.h (PT_GNU_SFRAME): New definition.
9772 * elf/internal.h (struct elf_segment_map): Handle new segment
9775 ld/testsuite/ChangeLog:
9777 * ld/testsuite/ld-bootstrap/bootstrap.exp: Add SFRAMELIB.
9778 * ld/testsuite/ld-aarch64/aarch64-elf.exp: Add new test
9780 * ld/testsuite/ld-aarch64/sframe-bar.s: New file.
9781 * ld/testsuite/ld-aarch64/sframe-foo.s: Likewise.
9782 * ld/testsuite/ld-aarch64/sframe-simple-1.d: Likewise.
9783 * ld/testsuite/ld-sframe/sframe-empty.d: New test.
9784 * ld/testsuite/ld-sframe/sframe-empty.s: New file.
9785 * ld/testsuite/ld-sframe/sframe.exp: New testsuite.
9786 * ld/testsuite/ld-x86-64/sframe-bar.s: New file.
9787 * ld/testsuite/ld-x86-64/sframe-foo.s: Likewise.
9788 * ld/testsuite/ld-x86-64/sframe-simple-1.d: Likewise.
9789 * ld/testsuite/ld-x86-64/sframe-plt-1.d: Likewise.
9790 * ld/testsuite/ld-x86-64/sframe-simple-1.d: Likewise.
9791 * ld/testsuite/ld-x86-64/x86-64.exp: Add new tests -
9792 sframe-simple-1, sframe-plt-1.
9793 * ld/testsuite/lib/ld-lib.exp: Add new proc to check if
9794 assembler supports SFrame section.
9795 * ld/testsuite/ld-sframe/discard.d: New file.
9796 * ld/testsuite/ld-sframe/discard.ld: Likewise.
9797 * ld/testsuite/ld-sframe/discard.s: Likewise.
9799 2022-11-15 Weimin Pan <weimin.pan@oracle.com>
9801 libsframe: add the SFrame library
9802 libsframe is a library that allows you to:
9803 - decode a .sframe section
9804 - probe and inspect a .sframe section
9805 - encode (and eventually write) a .sframe section.
9807 This library is currently being used by the linker, readelf, objdump.
9808 This library will also be used by the SFrame unwinder which is still
9811 The file include/sframe-api.h defines the user-facing APIs for decoding,
9812 encoding and probing .sframe sections. A set of error codes together
9813 with their error message strings are also defined.
9815 Endian flipping is performed automatically at read and write time, if
9816 cross-endianness is detected.
9820 * Makefile.def: Add libsframe as new module with its
9822 * Makefile.in: Regenerated.
9823 * binutils/Makefile.am: Add libsframe.
9824 * binutils/Makefile.in: Regenerated.
9825 * configure: Regenerated
9826 * configure.ac: Add libsframe to host_libs.
9827 * libsframe/Makefile.am: New file.
9828 * libsframe/Makefile.in: New file.
9829 * libsframe/aclocal.m4: New file.
9830 * libsframe/config.h.in: New file.
9831 * libsframe/configure: New file.
9832 * libsframe/configure.ac: New file.
9833 * libsframe/sframe-error.c: New file.
9834 * libsframe/sframe-impl.h: New file.
9835 * libsframe/sframe.c: New file.
9839 * sframe-api.h: New file.
9841 testsuite/ChangeLog:
9843 * libsframe/testsuite/Makefile.am: New file.
9844 * libsframe/testsuite/Makefile.in: Regenerated.
9845 * libsframe/testsuite/libsframe.decode/Makefile.am: New
9847 * libsframe/testsuite/libsframe.decode/Makefile.in:
9849 * libsframe/testsuite/libsframe.decode/decode.exp: New file.
9850 * libsframe/testsuite/libsframe.encode/Makefile.am:
9852 * libsframe/testsuite/libsframe.encode/Makefile.in:
9854 * libsframe/testsuite/libsframe.encode/encode.exp: New file.
9855 * libsframe/testsuite/libsframe.encode/encode-1.c: Likewise.
9856 * libsframe/testsuite/libsframe.decode/be-flipping.c: Likewise.
9857 * libsframe/testsuite/libsframe.decode/frecnt-1.c: Likewise.
9858 * libsframe/testsuite/libsframe.decode/frecnt-2.c: Likewise.
9859 * libsframe/testsuite/libsframe.decode/DATA-BE: New file.
9860 * libsframe/testsuite/libsframe.decode/DATA1: Likewise.
9861 * libsframe/testsuite/libsframe.decode/DATA2: Likewise.
9863 2022-11-15 Indu Bhagat <indu.bhagat@oracle.com>
9865 gas: testsuite: add new tests for SFrame unwind info
9866 Earlier these tests were in the same commit as previous which adds the
9867 support in GNU assembler to generate .sframe section from CFI
9868 directives. Splitting this out here for ease of applying and testing.
9872 * gas/testsuite/gas/cfi-sframe/cfi-sframe-aarch64-1.d: New file.
9873 * gas/testsuite/gas/cfi-sframe/cfi-sframe-aarch64-1.s: Likewise.
9874 * gas/testsuite/gas/cfi-sframe/cfi-sframe-common-1.d: Likewise.
9875 * gas/testsuite/gas/cfi-sframe/cfi-sframe-common-1.s: Likewise.
9876 * gas/testsuite/gas/cfi-sframe/cfi-sframe-common-2.d: Likewise.
9877 * gas/testsuite/gas/cfi-sframe/cfi-sframe-common-2.s: Likewise.
9878 * gas/testsuite/gas/cfi-sframe/cfi-sframe-common-3.d: Likewise.
9879 * gas/testsuite/gas/cfi-sframe/cfi-sframe-common-3.s: Likewise.
9880 * gas/testsuite/gas/cfi-sframe/cfi-sframe-common-4.d: Likewise.
9881 * gas/testsuite/gas/cfi-sframe/cfi-sframe-common-4.s: Likewise.
9882 * gas/testsuite/gas/cfi-sframe/cfi-sframe-common-5.d: Likewise.
9883 * gas/testsuite/gas/cfi-sframe/cfi-sframe-common-5.s: Likewise.
9884 * gas/testsuite/gas/cfi-sframe/cfi-sframe-common-6.d: Likewise.
9885 * gas/testsuite/gas/cfi-sframe/cfi-sframe-common-6.s: Likewise.
9886 * gas/testsuite/gas/cfi-sframe/cfi-sframe-common-7.d: Likewise.
9887 * gas/testsuite/gas/cfi-sframe/cfi-sframe-common-7.s: Likewise.
9888 * gas/testsuite/gas/cfi-sframe/cfi-sframe-common-8.d: Likewise.
9889 * gas/testsuite/gas/cfi-sframe/cfi-sframe-common-8.s: Likewise.
9890 * gas/testsuite/gas/cfi-sframe/cfi-sframe-x86_64-1.d: Likewise.
9891 * gas/testsuite/gas/cfi-sframe/cfi-sframe-x86_64-1.s: Likewise.
9892 * gas/testsuite/gas/cfi-sframe/cfi-sframe.exp: Likewise.
9893 * gas/testsuite/gas/cfi-sframe/common-empty-1.d: Likewise.
9894 * gas/testsuite/gas/cfi-sframe/common-empty-1.s: Likewise.
9895 * gas/testsuite/gas/cfi-sframe/common-empty-2.d: Likewise.
9896 * gas/testsuite/gas/cfi-sframe/common-empty-2.s: Likewise.
9897 * gas/testsuite/gas/cfi-sframe/common-empty-3.d: Likewise.
9898 * gas/testsuite/gas/cfi-sframe/common-empty-3.s: Likewise.
9899 * gas/testsuite/gas/cfi-sframe/common-empty-4.d: Likewise.
9900 * gas/testsuite/gas/cfi-sframe/common-empty-4.s: Likewise.
9902 2022-11-15 Indu Bhagat <indu.bhagat@oracle.com>
9904 gas: generate .sframe from CFI directives
9905 Currently supported for x86_64 and aarch64 only.
9907 [PS: Currently, the compiler has not been adapted to generate
9908 ".cfi_sections" with ".sframe" in it. The newly added command line
9909 option of --gsframe provides an easy way to try out .sframe support
9912 gas interprets the CFI directives to generate DWARF-based .eh_frame
9913 info. These internal DWARF structures are now consumed by
9914 gen-sframe.[ch] sub-system to, in turn, create the SFrame unwind
9915 information. These internal DWARF structures are read-only for the
9916 purpose of SFrame unwind info generation.
9918 SFrame unwind info generation does not impact .eh_frame unwind info
9919 generation. Both .eh_frame and .sframe can co-exist in an ELF file,
9920 if so desired by the user.
9922 Recall that SFrame unwind information only contains the minimal
9923 necessary information to generate backtraces and does not provide
9924 information to recover all callee-saved registers. The reason being
9925 that callee-saved registers other than FP are not needed for stack
9926 unwinding, and hence are not included in the .sframe section.
9928 Consequently, gen-sframe.[ch] only needs to interpret a subset of
9929 DWARF opcodes in gas. More details follow.
9931 [Set 1, Interpreted] The following opcodes are interpreted:
9932 - DW_CFA_advance_loc
9934 - DW_CFA_def_cfa_register
9935 - DW_CFA_def_cfa_offset
9937 - DW_CFA_remember_state
9938 - DW_CFA_restore_state
9941 [Set 2, Bypassed] The following opcodes are acknowledged but are not
9942 necessary for generating SFrame unwind info:
9946 Anything else apart from the two above-mentioned sets is skipped
9947 altogether. This means that any function containing a CFI directive not
9948 in Set 1 or Set 2 above, will not have any SFrame unwind information
9949 generated for them. Holes in instructions covered by FREs of a single
9950 FDE are not representable in the SFrame unwind format.
9952 As few examples, following opcodes are not processed for .sframe
9953 generation, and are skipped:
9957 - .cfi_negate_ra_state
9960 Not processing .cfi_escape, .cfi_negate_ra_state will cause SFrame
9961 unwind information to be absent for SFrame FDEs that contain these CFI
9962 directives, hence affecting the asynchronicity.
9964 x86-64 and aarch64 backends need to have a few new definitions and
9965 functions for .sframe generation. These provide gas with architecture
9966 specific information like the SP/FP/RA register numbers and an
9967 SFrame-specific ABI marker.
9969 Lastly, the patch also implements an optimization for size, where
9970 specific fragments containing SFrame FRE start address and SFrame FDE
9971 function are fixed up. This is similar to other similar optimizations
9972 in gas, where fragments are sized and fixed up when the associated
9973 symbols can be resolved. This optimization is controlled by a #define
9974 SFRAME_FRE_TYPE_SELECTION_OPT and should be easy to turn off if needed.
9975 The optimization is on by default for both x86_64 and aarch64.
9979 * gas/Makefile.am: Include gen-sframe.c and sframe-opt.c.
9980 * gas/Makefile.in: Regenerated.
9981 * gas/as.h (enum _relax_state): Add new state rs_sframe.
9982 (sframe_estimate_size_before_relax): New function.
9983 (sframe_relax_frag): Likewise.
9984 (sframe_convert_frag): Likewise.
9985 * gas/config/tc-aarch64.c (aarch64_support_sframe_p): New
9987 (aarch64_sframe_ra_tracking_p): Likewise.
9988 (aarch64_sframe_cfa_ra_offset): Likewise.
9989 (aarch64_sframe_get_abi_arch): Likewise.
9990 (md_begin): Set values of sp/fp/ra registers.
9991 * gas/config/tc-aarch64.h (aarch64_support_sframe_p): New
9993 (support_sframe_p): Likewise.
9994 (SFRAME_CFA_SP_REG): Likewise.
9995 (SFRAME_CFA_FP_REG): Likewise.
9996 (SFRAME_CFA_RA_REG): Likewise.
9997 (aarch64_sframe_ra_tracking_p): Likewise.
9998 (sframe_ra_tracking_p): Likewise.
9999 (aarch64_sframe_cfa_ra_offset): Likewise.
10000 (sframe_cfa_ra_offset): Likewise.
10001 (aarch64_sframe_get_abi_arch): Likewise.
10002 (sframe_get_abi_arch): Likewise.
10003 * gas/config/tc-i386.c (x86_support_sframe_p): New definition.
10004 (x86_sframe_ra_tracking_p): Likewise.
10005 (x86_sframe_cfa_ra_offset): Likewise.
10006 (x86_sframe_get_abi_arch): Likewise.
10007 * gas/config/tc-i386.h (x86_support_sframe_p): New declaration.
10008 (support_sframe_p): Likewise.
10009 (SFRAME_CFA_SP_REG): Likewise.
10010 (SFRAME_CFA_FP_REG): Likewise.
10011 (x86_sframe_ra_tracking_p): Likewise.
10012 (sframe_ra_tracking_p): Likewise.
10013 (x86_sframe_cfa_ra_offset): Likewise.
10014 (sframe_cfa_ra_offset): Likewise.
10015 (x86_sframe_get_abi_arch): Likewise.
10016 (sframe_get_abi_arch): Likewise.
10017 * gas/config/tc-xtensa.c (unrelaxed_frag_max_size): Add case for
10019 * gas/doc/as.texi: Add .sframe to the documentation for
10021 * gas/dw2gencfi.c (cfi_finish): Create a .sframe section.
10022 * gas/dw2gencfi.h (CFI_EMIT_sframe): New definition.
10023 * gas/write.c (cvt_frag_to_fill): Handle rs_sframe.
10024 (relax_segment): Likewise.
10025 * gas/gen-sframe.c: New file.
10026 * gas/gen-sframe.h: New file.
10027 * gas/sframe-opt.c: New file.
10029 2022-11-15 Indu Bhagat <indu.bhagat@oracle.com>
10031 gas: add new command line option --gsframe
10032 When --gsframe is specified, the assembler will generate a .sframe
10033 section from the CFI directives in the assembly.
10037 * gas/as.c (parse_args): Parse args and set flag_gen_sframe.
10038 * gas/as.h: Introduce skeleton for --gsframe.
10039 * gas/doc/as.texi: document --gsframe.
10041 2022-11-15 Indu Bhagat <indu.bhagat@oracle.com>
10043 sframe.h: Add SFrame format definition
10044 The header sframe.h defines the SFrame format.
10046 The SFrame format is the Simple Frame format. It can be used to
10047 represent the minimal necessary unwind information required for
10048 backtracing. The current version supports AMD64 and AARCH64.
10050 More details of the SFrame format are included in the documentation
10051 of the header file in this patch.
10054 * sframe.h: New file.
10056 2022-11-15 Alan Modra <amodra@gmail.com>
10058 Re: [gas] arm: Add support for new unwinder directive ".pacspval".
10059 * testsuite/gas/arm/ehabi-pacbti-m.d: Limit test to ELF.
10061 2022-11-15 Alan Modra <amodra@gmail.com>
10063 aarch64-pe can't fill 16 bytes in section .text
10064 Without commit b66e671854, this:
10069 results in an error when coff_frob_section attempts to pad out the
10070 section to a 16-byte boundary. Due to miscalculating the pad pattern
10071 repeat count, write.c:write_contents attempts to shove 16 bytes of
10072 padding into the remaining 4 bytes of the .text section.
10074 * config/obj-coff.c (coff_frob_section): Correct fill count.
10075 Don't pad after errors.
10077 2022-11-15 Jose E. Marchesi <jose.marchesi@oracle.com>
10079 gdb/configure: regenerate
10080 The commit bbaabc767a4293492817a0840819aef2768cce90 introduced an
10081 incorrect thunk for the `configure' script. This patch regenerates
10082 configure by calling autoreconf.
10084 2022-11-15 Jose E. Marchesi <jose.marchesi@oracle.com>
10086 gdb: use libtool in GDB_AC_CHECK_BFD
10087 The GDB_AC_CHECK_BFD macro defined in gdb/acinclude.m4 uses the
10088 AC_LINK_IFELSE autoconf macro in order to link a simple program to
10089 check features of libbfd.
10091 If libbfd's link dependencies change, it was necessary to reflect them
10092 either in the definition of the macro, or as a consequence of checking
10093 for them with an autoconf macro resulting in an addition to LIBS.
10095 This patch modifies the definition of the GDB_CHECK_BFD macro in order
10096 to use libtool to perform the test link. This makes it possible to
10097 not have to list dependencies of libbfd (which are indirect to GDB) at
10102 configure:28553: checking for ELF support in BFD
10103 configure:28573: ./libtool --quiet --mode=link gcc -o conftest \
10104 -I../../gdb/../include -I../bfd \
10105 -I../../gdb/../bfd -g -O2 \
10106 -L../bfd -L../libiberty conftest.c -lbfd -liberty \
10107 -lncursesw -lm -ldl >&5
10108 configure:28573: $? = 0
10109 configure:28583: result: yes
10113 - Configure --with-system-zlib and --without-system-zlib.
10114 - Check link dependencies of installed GDB with both --enable-shared
10115 and --disable-shared.
10116 - Run installed GDB in both cases.
10118 Approved-By: Simon Marchi <simon.marchi@efficios.com>
10120 2022-11-15 Tom Tromey <tromey@adacore.com>
10122 Fix crash in ada_print_type
10123 The "varstring" paramter to ada_print_type can be null, but one spot
10124 failed to check this. This could cause a crash in some situations.
10126 As this is Ada-specific, and we've been using it internally at AdaCore
10127 for a while, I am going to push it.
10129 2022-11-15 Tejas Joshi <TejasSanjay.Joshi@amd.com>
10131 Add AMD znver4 processor support
10132 2022-09-28 Tejas Joshi <TejasSanjay.Joshi@amd.com>
10136 * config/tc-i386.c (cpu_arch): Add znver4 ARCH and rmpquery SUBARCH.
10137 (md_assemble): Expand comment before swap_operands() with rmpquery.
10138 * doc/c-i386.texi: Add znver4.
10139 * testsuite/gas/i386/arch-14-1.d: New.
10140 * testsuite/gas/i386/arch-14-1.s: New.
10141 * testsuite/gas/i386/arch-14-znver4.d: New.
10142 * testsuite/gas/i386/i386.exp: Add new znver4 test cases.
10143 * testsuite/gas/i386/rmpquery.d: New.
10144 * testsuite/gas/i386/rmpquery.s: New.
10145 * testsuite/gas/i386/x86-64-arch-4-1.d: New.
10146 * testsuite/gas/i386/x86-64-arch-4-1.s: New.
10147 * testsuite/gas/i386/x86-64-arch-4-znver4.d: New.
10151 * i386-dis.c (x86_64_table): Add rmpquery.
10152 * i386-gen.c (cpu_flag_init): Add CPU_ZNVER4_FLAGS and
10153 CPU_RMPQUERY_FLAGS.
10154 (cpu_flags): Add CpuRMPQUERY.
10155 * i386-opc.h (enum): Add CpuRMPQUERY.
10156 (i386_cpu_flags): Add cpurmpquery.
10157 * i386-opc.tbl: Add rmpquery insn.
10158 * i386-init.h: Re-generated.
10159 * i386-tbl.h: Re-generated.
10161 2022-11-15 Simon Marchi <simon.marchi@polymtl.ca>
10163 gdb/testsuite: get_set_option_choices: expect \r\n after each item
10164 I get some random failures since commit 8d45c3a82a0e ("[gdb/testsuite]
10165 Set completions to unlimited in get_set_option_choices"), which can be
10168 $ make check-read1 TESTS="gdb.base/parse_number.exp"
10172 set architecture A^M
10173 Ambiguous item "A".^M
10174 (gdb) FAIL: gdb.base/parse_number.exp: arch=A: set architecture A
10176 The problem is the regexp in get_set_option_choices, it is possible that
10177 is only matches part of a completion result. With check-read1, that is
10180 Fix this by expecting the \r\n at the end of the line, so we only match
10181 entire results. Use ^ in match patterns to ensure we don't miss any
10184 Approved-By: Tom de Vries <tdevries@suse.de>
10185 Change-Id: Ib1733737feab7dde0f7095866e089081a891054e
10187 2022-11-15 Andre Vieira <andre.simoesdiasvieira@arm.com>
10189 aarch64, testsuite: Fixed recently added cssc.d
10190 Fixed wrong paste in cssc.d.
10194 * testsuite/gas/aarch64/cssc.d: Removed duplicate head.
10196 2022-11-15 Tom de Vries <tdevries@suse.de>
10198 [gdb/testsuite] Normalize gdbserver path name
10199 Currently for the target board remote-gdbserver-on-localhost we use the
10200 gdbserver file on build, using a file name which includes "/../".
10202 Fix this by using a normalized file name instead.
10204 This allows us to be more restrictive about which files REMOTE_TARGET_USERNAME
10207 - remote_exec build "chmod go-rx $objdir/outputs"
10208 + remote_exec build "chmod go-rx $objdir"
10211 Tested on x86_64-linux.
10213 2022-11-15 Tom de Vries <tdevries@suse.de>
10215 [gdb/testsuite] Fix gdb.base/jit-elf-so.exp for remote target
10216 With test-case gdb.base/jit-elf-so.exp and target board
10217 remote-gdbserver-on-localhost (using REMOTE_TARGET_USERNAME) we run into some
10221 - setting jit_libname with the name as returned by gdb_load_shlib
10222 - allowing the libraries to be prefixed with the remote target directory.
10224 Tested on x86_64-linux.
10226 Co-Authored-by: Ivan Tetyushkin <ivan.tetyushkin@syntacore.com>
10228 2022-11-15 Tom de Vries <tdevries@suse.de>
10230 [gdb/testsuite] Fix gdb.base/jit-reader-exec.exp for remote target
10231 With test-case gdb.base/jit-reader-exec.exp and target board
10232 remote-gdbserver-on-localhost (using REMOTE_TARGET_USERNAME) we run into some
10235 Fix this by adding the missing gdb_remote_download.
10237 Tested on x86_64-linux.
10239 Co-Authored-by: Ivan Tetyushkin <ivan.tetyushkin@syntacore.com>
10241 2022-11-15 Tom de Vries <tdevries@suse.de>
10243 [gdb/testsuite] Fix gdb.base/info-shared.exp for remote target
10244 With test-case gdb.base/info-shared.exp and target board
10245 remote-gdbserver-on-localhost (using REMOTE_TARGET_USERNAME) we run into some
10248 Fix these by adding the missing gdb_load_shlib.
10250 Tested on x86_64-linux.
10252 Co-Authored-by: Ivan Tetyushkin <ivan.tetyushkin@syntacore.com>
10254 2022-11-15 Tom de Vries <tdevries@suse.de>
10256 [gdb/testsuite] Fix gdb.base/solib-vanish.exp for remote target
10257 When running test-case gdb.base/solib-vanish.exp with target board
10258 remote-gdbserver-on-localhost (using REMOTE_TARGET_USERNAME) we run into some
10261 Fix these by adding the missing gdb_load_shlib.
10263 Tested on x86_64-linux.
10265 Co-Authored-by: Ivan Tetyushkin <ivan.tetyushkin@syntacore.com>
10267 2022-11-15 Tom de Vries <tdevries@suse.de>
10269 [gdb/testsuite] Fix gdb.base/infcall-exec.exp for remote target
10270 When running test-case gdb.base/infcall-exec.exp with target board
10271 remote-gdbserver-on-localhost (using REMOTE_TARGET_USERNAME) we run into:
10273 (gdb) call (int) execlp ("$outputs/gdb.base/infcall-exec/infcall-exec2", \
10274 "$outputs/gdb.base/infcall-exec/infcall-exec2", (char *)0)^M
10276 (gdb) FAIL: gdb.base/infcall-exec.exp: call execlp
10279 Fix this by using just:
10281 (gdb) call (int) execlp ("infcall-exec2", "infcall-exec2", (char *)0)^M
10283 and using putenv ("PATH=...") to allow infcall-exec to exec infcall-exec2
10284 if it's available alongside.
10286 Also fix the exec name in the test-case, such that we can successfully
10289 $ ./outputs/gdb.base/infcall-exec/infcall-exec
10290 PATH SETTING: 'PATH=./outputs/gdb.base/infcall-exec'
10294 Tested on x86_64-linux.
10296 Co-Authored-by: Ivan Tetyushkin <ivan.tetyushkin@syntacore.com>
10298 2022-11-15 Tom de Vries <tdevries@suse.de>
10300 [gdb/testsuite] Fix gdb.base/print-file-var.exp for remote target
10301 When running test-case gdb.base/print-file-var.exp with target board
10302 remote-gdbserver-on-localhost (using REMOTE_TARGET_USERNAME) we run into some
10305 Fix these by using the name of a shared lib as returned by gdb_load_shlib.
10307 This required splitting up the gdb_load_shlib functionality, which is now
10310 proc gdb_load_shlib { file } {
10311 set dest [gdb_download_shlib $file]
10312 gdb_locate_shlib $file
10316 such that we can do gdb_download_shlib before gdb is started.
10318 Tested on x86_64-linux.
10320 Co-Authored-by: Ivan Tetyushkin <ivan.tetyushkin@syntacore.com>
10322 2022-11-15 Tom de Vries <tdevries@suse.de>
10324 [gdb/testsuite] Add REMOTE_TARGET_USERNAME in remote-gdbserver-on-localhost.exp
10326 ( https://sourceware.org/pipermail/gdb-patches/2022-October/193147.html ) a
10327 number of test-cases fails with a remote target setup, for instance test-case
10328 gdb.base/print-file-var.exp.
10330 So, why don't we see these fails with our remote target boards in
10331 gdb/testsuite/boards, say remote-gdbserver-on-localhost.exp?
10333 The problem is that the target board uses the same machine and user for
10334 both (by-definition-local) build and remote target, and when using absolute
10335 pathnames to refer to files on build, we can access those files on target,
10336 which in a real remote target setup wouldn't be the case: we'd have to
10337 download them to target first, and then the filename would also be different.
10339 For aforementioned test-case, this happens when the name of a shared library is
10340 passed as absolute file name to gcc:
10342 gcc ... -DSHLIB_NAME="$outputs/gdb.base/print-file-var/\
10343 print-file-var-lib2-hidden0-dlopen1-version_id_main0_c.so"
10346 Make these problems visible with remote-gdbserver-on-localhost.exp by
10347 adding an option to specify a test account (still on the same machine)
10348 using REMOTE_TARGET_USERNAME.
10350 We make sure by restricting file permissions, that the test account cannot see
10351 the build files on the $USER account, and that the $USER account cannot see
10352 the target files on the test account.
10354 And so we can reproduce the reported fails:
10357 $ tc="gdb.base/print-file-var.exp"
10358 $ tb="--target_board remote-gdbserver-on-localhost"
10359 $ tbu="REMOTE_TARGET_USERNAME=remote-target"
10360 $ make check RUNTESTFLAGS="$tb $tbu $tc"
10362 FAIL: gdb.base/print-file-var.exp: lang=c: hidden=0: dlopen=1: \
10363 version_id_main=0: continue to STOP marker
10366 Tested on x86_64-linux.
10368 Reported-by: Ivan Tetyushkin <ivan.tetyushkin@syntacore.com>
10370 2022-11-15 Tom de Vries <tdevries@suse.de>
10372 [gdb/testsuite] Fix gdb.base/info_sources_2.exp for remote target
10373 With test-case gdb.base/info_sources_2.exp and target board
10374 remote-gdbserver-on-localhost (using REMOTE_TARGET_USERNAME) we run into some
10377 Fix these by adding the missing gdb_load_shlib.
10379 Tested on x86_64-linux.
10381 2022-11-15 Tom de Vries <tdevries@suse.de>
10383 [gdb/testsuite] Fix gdb.base/foll-exec.exp for remote target
10384 When running test-case gdb.base/foll-exec.exp with target board
10385 remote-gdbserver-on-localhost.exp, I run into:
10387 (gdb) PASS: gdb.base/foll-exec.exp: insert first exec catchpoint
10390 [Inferior 1 (process 4476) exited normally]^M
10391 (gdb) FAIL: gdb.base/foll-exec.exp: continue to first exec catchpoint (the program e\
10395 The problem is that the foll-exec executable expects the exec-ed executable
10396 execd-prog alongside it, but it's missing.
10398 Fix this by adding the missing gdb_remote_download.
10400 Likewise in a few other test-cases.
10402 Tested on x86_64-linux.
10404 2022-11-15 Tom de Vries <tdevries@suse.de>
10406 [gdb/testssuite] Skip aarch64 in skip_gdbserver_test if no xml support
10407 On aarch64-linux, with a gdb build without libexpat, so without xml support, I
10410 (gdb) builtin_spawn attach-no-multi-process^M
10412 Attaching to Remote target^M
10413 warning: Can not parse XML target description; XML support was disabled at \
10415 Reading symbols from attach-no-multi-process...^M
10416 Remote 'g' packet reply is too long (expected 788 bytes, got 796 bytes): ... ^M
10419 The test-case checks for skip_gdbserver_tests and that one contains a check
10422 # If GDB is lack of XML support, and targets, like arm, have
10423 # multiple target descriptions, GDB doesn't know which target
10424 # description GDBserver uses, and may fail to parse 'g' packet
10425 # after connection.
10426 if { [gdb_skip_xml_test]
10427 && ([istarget "arm*-*-linux*"]
10428 || [istarget "mips*-*-linux*"]
10429 || [istarget "powerpc*-*-linux*"]
10430 || [istarget "s390*-*-linux*"]
10431 || [istarget "x86_64-*-linux*"]
10432 || [istarget "i\[34567\]86-*-linux*"]) } {
10436 but it doesn't trigger because aarch64 is missing.
10438 Fix this by adding istarget "aarch64*-*-linux*".
10440 Tested on aarch64-linux.
10442 Approved-By: Luis Machado <luis.machado@arm.com>
10444 2022-11-15 Aditya Vidyadhar Kamath <Aditya.Kamath1@ibm.com>
10446 Enable multi-process debugging for AIX
10447 This patch adds multi-process debugging feature in AIX.
10449 Till now AIX supported debugging only one inferior at a time,
10450 now we can be able to debug multi process. Users can use set
10451 follow fork mode in child or parent and set detach on fork on
10452 or off to enable/disable simultaneous debugging of parent/child.
10454 2022-11-15 GDB Administrator <gdbadmin@sourceware.org>
10456 Automatic date update in version.in
10458 2022-11-14 Carl Love <cel@us.ibm.com>
10460 PowerPC, fix support for printing the function return value for non-trivial values.
10461 Currently, a non-trivial return value from a function cannot currently be
10462 reliably determined on PowerPC. This is due to the fact that the PowerPC
10463 ABI uses register r3 to store the address of the buffer containing the
10464 non-trivial return value when the function is called. The PowerPC ABI
10465 does not guarantee the value in register r3 is not modified in the
10466 function. Thus the value in r3 cannot be reliably used to obtain the
10467 return addreses on exit from the function.
10469 This patch adds a new gdbarch method to allow PowerPC to access the value
10470 of r3 on entry to a function. On PowerPC, the new gdbarch method attempts
10471 to use the DW_OP_entry_value for the DWARF entries, when exiting the
10472 function, to determine the value of r3 on entry to the function. This
10473 requires the use of the -fvar-tracking compiler option to compile the
10474 user application thus generating the DW_OP_entry_value in the binary. The
10475 DW_OP_entry_value entries in the binary file allows GDB to resolve the
10476 DW_TAG_call_site entries. This new gdbarch method is used to get the
10477 return buffer address, in the case of a function returning a nontrivial
10478 data type, on exit from the function. The GDB function should_stop checks
10479 to see if RETURN_BUF is non-zero. By default, RETURN_BUF will be set to
10480 zero by the new gdbarch method call for all architectures except PowerPC.
10481 The get_return_value function will be used to obtain the return value on
10482 all other architectures as is currently being done if RETURN_BUF is zero.
10483 On PowerPC, the new gdbarch method will return a nonzero address in
10484 RETURN_BUF if the value can be determined. The value_at function uses the
10485 return buffer address to get the return value.
10487 This patch fixes five testcase failures in gdb.cp/non-trivial-retval.exp.
10488 The correct function return values are now reported.
10490 Note this patch is dependent on patch: "PowerPC, function
10491 ppc64_sysv_abi_return_value add missing return value convention".
10493 This patch has been tested on Power 10 and x86-64 with no regressions.
10495 2022-11-14 Carl Love <cel@us.ibm.com>
10497 PowerPC, function ppc64_sysv_abi_return_value add missing return value convention
10498 This patch address five testcase failures in gdb.cp/non-trivial-retval.exp.
10499 The following commit resulted in the five testcases failures on PowerPC.
10500 The value returned by the function is being reported incorrectly.
10502 commit b1718fcdd1d2a5c514f8ee504ba07fb3f42b8608
10503 Author: Andrew Burgess <aburgess@redhat.com>
10504 Date: Mon Dec 13 16:56:16 2021 +0000
10506 gdb: on x86-64 non-trivial C++ objects are returned in memory
10508 Fixes PR gdb/28681. It was observed that after using the `finish`
10509 command an incorrect value was displayed in some cases. Specifically,
10510 this behaviour was observed on an x86-64 target.
10514 enum return_value_convention
10515 ppc64_sysv_abi_return_value (struct gdbarch *gdbarch, struct value *function,
10516 struct type *valtype, struct regcache *regcache,
10517 gdb_byte *readbuf, const gdb_byte *writebuf)
10519 should return RETURN_VALUE_STRUCT_CONVENTION if the valtype->code() is
10520 TYPE_CODE_STRUCT and if the language_pass_by_reference is not
10521 trivially_copyable.
10523 This patch adds the needed code to return the value
10524 RETURN_VALUE_STRUCT_CONVENTION in this case.
10526 With this patch, the five test cases still fail but with the message "Value
10527 returned has type: A. Cannot determine contents". The PowerPC ABI stores
10528 the address of the buffer containing the function return value in register
10529 r3 on entry to the function. However, the PowerPC ABI does not guarentee
10530 that r3 will not be modified in the function. So when the function returns,
10531 the return buffer address cannot be reliably obtained from register r3.
10532 Thus the message "Cannot determine contents" is appropriate in this case.
10534 2022-11-14 Tom Tromey <tromey@adacore.com>
10536 Remove dump_prefix_expression
10537 Since the expression rewrite, dump_prefix_expression has been
10538 misnamed. This patch cleans this up by removing the function, turning
10539 it into a method on struct expression.
10541 2022-11-14 Andre Vieira <andre.simoesdiasvieira@arm.com>
10543 aarch64: Add support for Common Short Sequence Compression extension
10544 This patch adds support for the CSSC extension and its corresponding
10545 instructions: ABS, CNT, CTZ, SMAX, UMAX, SMIN, UMIN.
10549 * config/tc-aarch64.c (parse_operands): Handle new operand types.
10550 * doc/c-aarch64.texi: Document new extension.
10551 * testsuite/gas/aarch64/cssc.d: New test.
10552 * testsuite/gas/aarch64/cssc.s: New test.
10556 * opcode/aarch64.h (AARCH64_FEATURE_CSSC): New feature Macro.
10557 (enum aarch64_opnd): New operand types.
10558 (enum aarch64_insn_class): New instruction class.
10562 * aarch64-asm-2.c: Regenerate.
10563 * aarch64-dis-2.c: Regenerate.
10564 * aarch64-opc-2.c: Regenerate.
10565 * aarch64-opc.c (operand_general_constraint_met_p): Update for new
10567 (aarch64_print_operand): Likewise.
10568 * aarch64-opc.h (enum aarch64_field_kind): Declare FLD_CSSC_imm8 field.
10569 * aarch64-tbl.h (aarch64_feature_cssc): Define new feature set.
10570 (CSSC): Define new feature set Macro.
10571 (CSSC_INSN): Define new instruction type.
10572 (aarch64_opcode_table): Add new instructions.
10574 2022-11-14 Jan Beulich <jbeulich@suse.com>
10576 x86: fold special-operand insn attributes into a single enum
10577 Attributes which aren't used together in any single insn template can be
10578 converted from individual booleans to a single enum, as was done for a few
10579 other attributes before. This is more space efficient. Collect together
10580 all attributes which express special operand constraints (and which fit
10581 the criteria for folding).
10583 2022-11-14 Dimitar Dimitrov <dimitar@dinux.eu>
10585 pru: bfd: Correct default to no execstack
10586 Data and instruction memories are strictly separated, so it is not
10587 possible to execute instructions from the stack memory on PRU.
10589 I don't see any difference in testsuite results with or without this
10594 * elf32-pru.c (elf_backend_default_execstack): Define as 0.
10598 * testsuite/ld-elf/elf.exp (target_defaults_to_execstack):
10601 2022-11-14 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
10603 [gas] arm: Add support for new unwinder directive ".pacspval".
10604 This patch adds the assembler support for the new unwinder
10605 directive ".pacspval" and encodes this directives with opcode
10606 "0xb5". This opcode indicates the unwinder to use effective
10607 vsp as modifier for PAC validation.
10611 2022-11-07 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
10613 * doc/c-arm.texi: Document directive.
10614 * config/tc-arm.c (s_arm_unwind_pacspval): Define function.
10615 (md_pseudo_table): Add entry for pacspval directive.
10616 * testsuite/gas/arm/ehabi-pacbti-m.d: New test.
10617 * testsuite/gas/arm/ehabi-pacbti-m.s: Likewise.
10619 2022-11-14 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
10621 [readelf] arm: Support for new pacbti unwind opcode 0xb5.
10622 This patch adds readelf support for decoding the exception
10623 table opcode "0xb5", which indicates to use effective vsp
10624 as modifier for PAC validation as defined by EHABI
10625 (https://github.com/ARM-software/abi-aa/releases/download/2022Q3/ehabi32.pdf
10628 binutils/ChangeLog:
10630 2022-11-07 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
10632 * readelf.c (decode_arm_unwind_bytecode): Add entry to decode opcode 0xb5.
10634 2022-11-14 Tsukasa OI <research_trasio@irq.a4lg.com>
10636 gdb/unittests: PR28413, suppress warnings generated by Gnulib
10637 Gnulib generates a warning if the system version of certain functions
10638 are used (to redirect the developer to use Gnulib version). It caused a
10639 compiler error when...
10641 - Compiled with Clang
10642 - -Werror is specified (by default)
10643 - C++ standard used by Clang is before C++17 (by default as of 15.0.0)
10644 when this unit test is activated.
10646 This issue is raised as PR28413.
10648 However, previous proposal to fix this issue (a "fix" to Gnulib):
10649 <https://lists.gnu.org/archive/html/bug-gnulib/2021-10/msg00003.html>
10650 was rejected because it ruins the intent of Gnulib warnings.
10652 So, we need a Binutils/GDB-side solution.
10654 This commit tries to address this issue on the GDB side. We have
10655 "include/diagnostics.h" to disable certain warnings only when necessary.
10657 This commit suppresses the Gnulib warnings by surrounding entire #include
10658 block with DIAGNOSTIC_IGNORE_USER_DEFINED_WARNINGS to disable Gnulib-
10659 generated warnings on all standard C++ header files.
10661 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28413
10662 Approved-By: Simon Marchi <simon.marchi@efficios.com>
10663 Change-Id: Ieeb5a31a6902808d4c7263a2868ae19a35e0ccaa
10665 2022-11-14 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
10667 arm: Add support for Cortex-X1C CPU.
10668 This patch adds support for Cortex-X1C CPU in Arm.
10672 2022-11-09 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
10674 * cpu-arm.c (processors): Add Cortex-X1C CPU entry.
10678 2022-11-09 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
10680 * NEWS: Update docs.
10681 * config/tc-arm.c (arm_cpus): Add cortex-x1c to -mcpu.
10682 * doc/c-arm.texi: Update docs.
10683 * testsuite/gas/arm/cpu-cortex-x1c.d: New test.
10685 2022-11-14 Tom de Vries <tdevries@suse.de>
10687 [gdb/testsuite] Run gdb.arch/ppc64-symtab-cordic.exp for --enable-targets=all
10688 While looking at test-case gdb.arch/ppc64-symtab-cordic.exp I realized that
10689 the test-case is too restrictive here:
10691 if {![istarget "powerpc*"] || ![is_lp64_target]} {
10692 verbose "Skipping powerpc64 separate debug file symtab test."
10696 and can also be run on x86_64-linux, if "set arch powerpc:common64" is
10697 supported, which is the case if we've build gdb with --enable-targets=all.
10699 Fix this by instead checking if powerpc:common64 is in the completion list for
10702 This allows us to remove the 'untested "powerpc:common64 is not supported"'.
10704 While we're at it, clean up the test-case by using clean_restart.
10706 Tested on x86_64-linux.
10708 2022-11-14 Tom de Vries <tdevries@suse.de>
10710 [gdb/testsuite] Handle with_set arch
10711 I realized that the more irregular output of show arch:
10714 The target architecture is set to "auto" (currently "i386").^M
10716 would be a problem for something like:
10718 with_set arch powerpc:common64 {}
10722 (gdb) set arch powerpc:common64^M
10723 The target architecture is set to "powerpc:common64".^M
10724 (gdb) FAIL: gdb.base/foo.exp: set arch powerpc:common64
10728 (gdb) set arch set to "auto" (currently "i386")^M
10729 Undefined item: "set".^M
10732 Fix this in with_set by handling this type of output.
10734 Tested on x86_64-linux.
10736 2022-11-14 Tom de Vries <tdevries@suse.de>
10738 [gdb/testsuite] Set completions to unlimited in get_set_option_choices
10739 In some test-case I tried to use get_set_option_choices "set architecture" and
10740 ran into max-completions:
10742 set architecture simple^M
10743 set architecture tomcat^M
10744 set architecture xscale^M
10745 set architecture *** List may be truncated, max-completions reached. ***^M
10746 (gdb) PASS: gdb.base/foo.exp: complete set architecture
10749 There's only one test-case using this currently: gdb.base/parse_number.exp,
10750 and it locally sets max-completions to unlimited.
10753 - factoring out a new proc with_set out of proc with_complaints, and
10754 - using it to temporarily set max-completions to unlimited in
10755 get_set_option_choice.
10757 Tested on x86_64-linux, by running test-cases that excercise
10758 get_set_option_choice and with_complaints.
10760 2022-11-14 Alan Modra <amodra@gmail.com>
10762 Re: objcopy renaming section with explicit flags
10763 For now, xfail the new test. Some header/aux-header rewriting is
10764 required at the very least.
10766 * testsuite/binutils-all/rename-section-01.d: xfail xcoff.
10768 2022-11-14 Alan Modra <amodra@gmail.com>
10770 objcopy renaming section with explicit flags
10771 This tidies SEC_RELOC handling in bfd, in the process fixing a bug
10772 with objcopy when renaming sections.
10775 * reloc.c (_bfd_generic_set_reloc): Set/clear SEC_RELOC depending
10777 * elf64-sparc.c (elf64_sparc_set_reloc): Likewise.
10779 * objcopy.c (copy_relocations_in_section): Remove now unnecessary
10780 clearing of SEC_RELOC.
10781 * testsuite/binutils-all/rename-section-01.d: New test.
10782 * testsuite/binutils-all/objcopy.exp: Run it.
10784 * write.c (size_seg): Remove unneccesary twiddle of SEC_RELOC.
10785 (write_relocs): Likewise. Always call bfd_set_reloc.
10787 2022-11-14 GDB Administrator <gdbadmin@sourceware.org>
10789 Automatic date update in version.in
10791 2022-11-13 Jon Turney <jon.turney@dronecode.org.uk>
10793 Fix Cygwin build after 02d04eac
10794 Commit 02d04eac "Use strwinerror in gdb/windows-nat.c" also moves
10795 strwinerror() under the USE_WIN32API conditional, which is not defined
10796 for Cygwin (and looks like it shouldn't be, as appears to imply
10797 non-POSIX and MiNGW and WinSock...)
10799 Also enable the declaration and definition of strwinerror() when
10800 __CYGWIN__ is defined.
10802 2022-11-13 Jon Turney <jon.turney@dronecode.org.uk>
10804 Drop apparently unneeded include of winsock2.h
10805 Commit d08bae3d ("Implement target async for Windows") unconditionally
10806 includes winsock2.h. We don't want to do that on Cygwin, since
10807 including both winsock2.h and sys/select.h causes incompatible
10808 redefinition problems.
10810 Since that include is apparently unneeded, just drop it.
10814 2022-11-13 GDB Administrator <gdbadmin@sourceware.org>
10816 Automatic date update in version.in
10818 2022-11-12 Dimitar Dimitrov <dimitar@dinux.eu>
10820 sim: pru: Fix behaviour when loop count is zero
10821 If the counter for LOOP instruction is provided by a register with
10822 value zero, then the instruction must cause a PC jump directly to the
10823 loop end. But in that particular case simulator must not initialize
10824 its internal loop variables, because loop body will not be executed.
10825 Instead, simulator must obtain the loop's end address directly from
10826 the LOOP instruction.
10828 2022-11-12 Alan Modra <amodra@gmail.com>
10830 PowerPC64 paddi -Mraw
10833 disassembled with -Mpower10 results in
10834 0: 00 00 10 06 pla r8,0 # 0
10836 0: R_PPC64_PCREL34 foo
10837 but with -Mpower10 -Mraw
10838 0: 00 00 10 06 .long 0x6100000
10839 0: R_PPC64_PCREL34 foo
10840 4: 00 00 00 39 addi r8,0,0
10842 The instruction is unrecognised due to the hack we have in
10843 extract_pcrel0 in order to disassemble paddi with RA0=0 and R=1 as
10844 pla. I could have just added "&& !(dialect & PPC_OPCODE_RAW)" to the
10845 condition in extract_pcrel0 under which *invalid is set, but went for
10846 this larger patch that reorders the extended insn pla to the more
10847 usual place before its underlying machine insn. (la is after addi
10848 because we never disassemble to la.)
10851 * testsuite/gas/ppc/raw.d,
10852 * testsuite/gas/ppc/raw.s: Add pla.
10854 * ppc-opc.c (extract_pcrel1): Rename from extract_pcrel0 and
10855 invert *invalid logic.
10856 (PCREL1): Rename from PCREL0.
10857 (prefix_opcodes): Sort pla before paddi, adjusting R operand
10858 for pla, paddi and psubi.
10860 2022-11-12 Indu Bhagat <indu.bhagat@oracle.com>
10862 libctf: use libtool for link test in configure
10863 The configure check for ELF support in BFD uses the AC_TRY_LINK. If
10864 libbfd's dependencies change, this macro will need to be updated
10865 manually with explicit additions to LDFLAGS and LIBS.
10867 This patch updates the check to use libtool instead.
10871 * libctf/configure.ac: Use libtool instead.
10872 * libctf/configure: Regenerated.
10874 2022-11-12 GDB Administrator <gdbadmin@sourceware.org>
10876 Automatic date update in version.in
10878 2022-11-11 Simon Marchi <simon.marchi@polymtl.ca>
10880 gdb: fix start breakpoint expression not working in some languages
10881 Commit 0be837be9fb4 ("gdb: make "start" breakpoint inferior-specific")
10882 regresses gdb.ada/start.exp:
10885 Error in expression, near `1'.
10886 (gdb) UNTESTED: gdb.ada/start.exp: start failed to land inside the right procedure
10888 This is because in Ada, the equality operator is =, not ==.
10890 I checked the other languages supported by GDB, these other languages
10891 use = for equality:
10893 - Pascal: tests like gdb.pascal/hello.exp are affected too
10894 - Modula-2: I tried building a Modula-2 hello world using gm2, but it
10895 seems like the generated DWARF doesn't specify the Modula-2 language
10896 in the CUs, it's C++ and C, so the selected language isn't
10897 "modula-2". But if I manually do "set language modula-2" on a dummy
10898 program and then "start", I get the same error.
10900 Other languages all use ==.
10902 So, a short term fix would be to use = or == in the expression, based on
10903 the current language. If this was meant to be permanent, I would
10904 suggest adding something like an "equality_operator" method to
10905 language_defn, that returns the right equality operator for the
10906 language. But the goal is to replace all this with proper
10907 inferior-specific breakpoints, so I hope all this is temporary.
10909 Approved-By: Tom de Vries <tdevries@suse.de>
10910 Change-Id: Id4d38e14a80e6bbbb1ad2b2277f974dd55192969
10912 2022-11-11 Mike Frysinger <vapier@gentoo.org>
10914 sim: igen: cleanup archaic pointer-to-long printf casts
10915 Use proper %p to printf a pointer instead of casting it to long and
10916 using 0x%lx. It's cleaner, more correct, and doesn't break on LLP64.
10918 2022-11-11 Tom de Vries <tdevries@suse.de>
10920 [gdb/testsuite] Don't timeout on prompt in gdb_start_cmd
10921 We're currently running into a timeout at:
10924 Error in expression, near `1'.^M
10925 (gdb) UNTESTED: gdb.ada/start.exp: start failed to land inside the right \
10928 due to the fact that gdb_start_cmd doesn't handle a prompt as reaction to
10931 Fix this by handling the prompt. Reduces execution time of the test-case from
10934 Tested on x86_64-linux.
10936 2022-11-11 Tom de Vries <tdevries@suse.de>
10938 [gdb/testsuite] Better error checking in has_hw_wp_support
10939 With gdb 12.1, on powerpc64le I ran into ERRORs related to has_hw_wp_support
10940 usage, which was already fixed on trunk by commits:
10941 - 13f72372413 ("gdb/testsuite: fix gdb.base/break-idempotent.exp on ppc"), and
10942 - 01a32ee0b8c ("PowerPC, fix gdb.base/watchpoint.exp on Power 9")
10944 While looking into these ERRORs and the commits that fix them, it occurred to
10945 me that while the commits fix the root cause, the failure mode is not great.
10947 The test-cases expect a running instance of gdb upon return, which is not
10948 there, so there's an long stream of ERRORs generated as a result.
10950 Fix this at the start of has_hw_wp_support, by (instead of accomodating a
10951 running gdb instance by calling gdb_exit), checking whether it's called
10952 without a running gdb instance, and erroring out otherwise. This way, there's
10955 I also noticed that in case we do an early exit due to !runto_main, we don't
10956 clean up, so copy the missing cleanups (gdb_exit and $obj file deletion) from
10959 Tested on x86_64-linux, using has_hw_wp_support for x86_64 in
10960 skip_hw_watchpoint_tests.
10962 2022-11-11 Lancelot SIX <lancelot.six@amd.com>
10964 gdb/py-inferior: Keep inferior threads in a map
10965 The python code maintains a list of threads for each inferior. This
10966 list is implemented as a linked list. When the number of threads grows
10967 high, this implementation can begin to be a performance bottleneck as
10968 finding a particular thread_object in the list has a complexity of O(N).
10970 We see this in ROCgdb[1], a downstream port of GDB for AMDGUP. On
10971 AMDGPU devices, the number of threads can get significantly higher than
10972 on usual GDB workloads.
10974 In some situations, we can reach the end of the inferior process with
10975 GDB still having a substantial list of known threads. While running
10976 target_mourn_inferior, we end up in inferior::clear_thread_list which
10977 iterates over all remaining threads and marks each thread exited. This
10978 fires the gdb::observers::thread_exit observer and eventually
10979 py-inferior.c:set_thread_exited gets called. This function searches in
10980 the linked list with poor performances.
10982 This patch proposes to change the linked list that keeps the per
10983 inferior_object list of thread_objects into a std::unordered_map. This
10984 allows to have the search operation complexity be O(1) on average
10987 With this patch, we can complete clear_thread_list in about 2.5 seconds
10988 compared to 10 minutes without it.
10990 Except for the performance change, no user visible change is expected.
10992 Regression tested on Ubuntu-22.04 x86_64.
10994 [1] https://github.com/ROCm-Developer-Tools/ROCgdb
10996 2022-11-11 Luis Machado <luis.machado@arm.com>
10998 Make sure a copy_insn_closure is available when we have a match in copy_insn_closure_by_addr
11001 Investigating PR29272, it was mentioned a particular test used to work on
11002 GDB 10, but it started failing with GDB 11 onwards. I tracked it down to
11003 some displaced stepping improvements on commit
11004 187b041e2514827b9d86190ed2471c4c7a352874.
11006 In particular, one of the corner cases using copy_insn_closure_by_addr got
11007 silently broken. It is hard to spot because it doesn't have any good tests
11008 for it, and the situation is quite specific to the Arm target.
11010 Essentially, the change from the displaced stepping improvements made it so
11011 we could still invoke copy_insn_closure_by_addr correctly to return the
11012 pointer to a copy_insn_closure, but it always returned nullptr due to
11013 the order of the statements in displaced_step_buffer::prepare.
11015 The way it is now, we first write the address of the displaced step buffer
11016 to PC and then save the copy_insn_closure pointer.
11018 The problem is that writing to PC for the Arm target requires figuring
11019 out if the new PC is thumb mode or not.
11021 With no copy_insn_closure data, the logic to determine the thumb mode
11022 during displaced stepping doesn't work, and gives random results that
11023 are difficult to track (SIGILL, SIGSEGV etc).
11025 Fix this by reordering the PC write in displaced_step_buffer::prepare
11026 and, for safety, add an assertion to
11027 displaced_step_buffer::copy_insn_closure_by_addr so GDB stops right
11028 when it sees this invalid situation. If this gets broken again in the
11029 future, it will be easier to spot.
11031 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29272
11033 Approved-By: Simon Marchi <simon.marchi@efficios.com>
11035 2022-11-11 Felix Willgerodt <felix.willgerodt@intel.com>
11037 gdb, btrace: Fix rn-dl-bind.exp for new icx remark.
11038 When running the test with the latest Intel compiler:
11040 Running gdb/gdb/testsuite/gdb.btrace/rn-dl-bind.exp ...
11041 gdb compile failed, icpx: warning: treating 'c' input as 'c++' when
11042 in C++ mode, this behavior is deprecated [-Wdeprecated]
11044 The test doesn't seem to test something specifically for C++,
11045 so I removed the C++ compilation option. Alternatively we could rename
11046 rn-dl-bind.exp.c to rn-dl-bind.exp.cc.
11048 2022-11-11 Bruno Larsen <blarsen@redhat.com>
11050 gdb/testsuite: disable gdb.cp/call-method-register.exp when not using gcc
11051 The test gdb.cp/call-method-register.exp assumes that the class will be
11052 placed on a register. However, this keyword has been deprecated since
11053 C++11, and Clang, for instance, does not feel the need to follow it.
11054 Since this test is not usable without this working, this commit marks
11055 this test as untested.
11057 Approved-by: Tom Tromey <tom@tromey.com>
11059 2022-11-11 Bruno Larsen <blarsen@redhat.com>
11061 gdb/testsuite: remove XFAIL on gdb.cp/temargs.exp
11062 gdb.cp/temargs.exp last 2 tests always setup an XFAILs, despite checking
11063 for old gcc versions. However, Clang does not fail in this test,
11064 turning into XPASSes and slighty annoying when comparing between
11065 compilers. To change this, make the xfails only happen if we using gcc.
11067 Approved-by: Tom Tromey <tom@tromey.com>
11069 2022-11-11 Bruno Larsen <blarsen@redhat.com>
11071 gdb/testsuite: disable some tests of gdb.cp/typeid.exp when using Clang
11072 Since Clang chooses to not add any debug information for base types,
11073 expecting it to be included with libraries' informations, gdb.cp/typeid.exp
11074 will always fail if the program hasn't started. This commit fixes that by
11075 making it so when using Clang, the base type variables aren't tested.
11077 Approved-by: Tom Tromey <tom@tromey.com>
11079 2022-11-11 Bruno Larsen <blarsen@redhat.com>
11081 gdb/testsuite: skip gdb.cp/anon-struct.exp when using Clang
11082 When Clang compiles anonymous structures, it does not add linkage names in
11083 their dwarf representations. This is compounded by Clang not adding linkage
11084 names to subprograms of those anonymous structs (for instance, the
11085 constructor). With these 2 things together, GDB is unable to refer to
11086 any of them, so there is no way to pass any of the tests of
11087 gdb.cp/anon-struct.exp
11089 Since this isn't a bug on Clang or GDB according to the DWARF
11090 specifications as DW_AT_name is optional for all DIEs, the test was marked
11093 Since I was already touching the file, I also added a comment at the top
11094 of the file explaining what it is testing for.
11096 Approved-by: Tom Tromey <tom@tromey.com>
11098 2022-11-11 Bruno Larsen <blarsen@redhat.com>
11100 gdb/testsuite: allow for Clang style destructors on gdb.cp/m-static.exp
11101 when running gdb.cp/m-static.exp using Clang, we get the following
11104 print test1.~gnu_obj_1^M
11105 $6 = {void (gnu_obj_1 * const)} 0x555555555470 <gnu_obj_1::~gnu_obj_1()>^M
11106 (gdb) FAIL: gdb.cp/m-static.exp: simple object instance, print destructor
11107 ptype test1.~gnu_obj_1^M
11108 type = void (gnu_obj_1 * const)^M
11109 (gdb) FAIL: gdb.cp/m-static.exp: simple object instance, ptype destructor
11110 print test1.'~gnu_obj_1'^M
11111 $7 = {void (gnu_obj_1 * const)} 0x555555555470 <gnu_obj_1::~gnu_obj_1()>^M
11112 (gdb) FAIL: gdb.cp/m-static.exp: simple object instance, print quoted destructor
11114 This is because the test is expecting an extra integer parameter on the
11115 destructor. Looking at the debuginfo, it seems that there is nothing
11116 actually wrong with this output, so these tests were changed to test
11117 multiple possible regexps.
11119 Approved-by: Tom Tromey <tom@tromey.com>
11121 2022-11-11 Bruno Larsen <blarsen@redhat.com>
11123 gdb/testsuite: add XFAIL to gdb.cp/derivation.exp when using Clang
11124 When running gdb.cp/derivation.exp using Clang, we get an unexpected
11125 failure when printing the type of a class with an internal typedef. This
11126 happens because Clang doesn't add accessibility information for typedefs
11127 inside classes (see https://github.com/llvm/llvm-project/issues/57608
11128 for more info). To help with Clang testing, an XFAIL was added to this
11131 Approved-by: Tom Tromey <tom@tromey.com>
11133 2022-11-11 Bruno Larsen <blarsen@redhat.com>
11135 gdb/testsuite: account for clang's nested destructor calls on gdb.cp/mb-ctor.exp
11136 When compiling virtual classes's destructors, two versions are compiled,
11137 one with a single parameter (this) and the other with 2 parameters (this
11140 GCC's compilation makes it so either the version with 1
11141 parameter or the one with 2 parameters is called, depending on whether
11142 the destructor is being called by the class itself or by an inherited
11143 class. On the test gdb.cp/mb-ctor.exp, this means that the breakpoint
11144 set at the destructor will be hit 4 times.
11146 Clang, on the other hand, makes the single-parameter version call the 2
11147 parameter version, probably in an attempt to reduce the size of the
11148 resulting executable. This means that the gdb.cp/mb-ctor.exp will hit 6
11149 breakpoints before finishing, and is the reason why this test was
11150 failing. To make this test stop failing, a compiler check is added and
11151 another "continue" instruction is issued to account for this difference.
11153 Approved-by: Tom Tromey <tom@tromey.com>
11155 2022-11-11 Bruno Larsen <blarsen@redhat.com>
11157 gdb/testsuite: enable running gdb.cp/classes.exp with clang
11158 When attempting to run the gdb.cp/classes.exp test using Clang++, the
11159 test fails to prepare with -Wnon-c-typedef-for-linkage like the
11160 previously fixed gdb.cp/class2.exp. Upon fixing this, the test shows 5
11161 unexpected failures. One such failures is:
11163 ptype/r class class_with_public_typedef
11164 type = class class_with_public_typedef {
11168 class_with_public_typedef::INT b;
11173 (gdb) FAIL: gdb.cp/classes.exp: ptype class class_with_public_typedef // wrong access specifier for typedef: private
11175 While g++ provided the following output:
11177 ptype/r class class_with_public_typedef
11178 type = class class_with_public_typedef {
11182 class_with_public_typedef::INT b;
11186 (gdb) PASS: gdb.cp/classes.exp: ptype class class_with_public_typedef
11188 This error happens because Clang does not add DW_AT_accessibility to
11189 typedefs inside classes, and without this information GDB defaults to
11190 assuming the typedef is private. Since there is nothing that GDB can do
11191 about this, these tests have been set as xfails, and Clang bug 57608 has
11194 Bug: https://github.com/llvm/llvm-project/issues/57608
11195 Approved-by: Tom Tromey <tom@tromey.com>
11197 2022-11-11 Bruno Larsen <blarsen@redhat.com>
11199 gdb/testsuite: ignore Non-C-typedefs for gdb.cp/class2.exp
11200 When attempting to test gdb.cp/class2.exp using Clang, it fails to
11201 prepare with the following error:
11203 Executing on host: clang++ -fdiagnostics-color=never -Wno-unknown-warning-option -c -g -o /home/blarsen/Documents/fsf_build/gdb/testsuite/outputs/gdb.cp/class2/class20.o /home/blarsen/Documents/fsf_build/gdb/testsuite/../../../binutils-gdb/gdb/testsuite/gdb.cp/class2.cc (timeout = 300)
11204 builtin_spawn -ignore SIGHUP clang++ -fdiagnostics-color=never -Wno-unknown-warning-option -c -g -o /home/blarsen/Documents/fsf_build/gdb/testsuite/outputs/gdb.cp/class2/class20.o /home/blarsen/Documents/fsf_build/gdb/testsuite/../../../binutils-gdb/gdb/testsuite/gdb.cp/class2.cc
11205 /home/blarsen/Documents/fsf_build/gdb/testsuite/../../../binutils-gdb/gdb/testsuite/gdb.cp/class2.cc:53:14: warning: anonymous non-C-compatible type given name for linkage purposes by typedef declaration; add a tag name here [-Wnon-c-typedef-for-linkage]
11209 /home/blarsen/Documents/fsf_build/gdb/testsuite/../../../binutils-gdb/gdb/testsuite/gdb.cp/class2.cc:54:2: note: type is not C-compatible due to this member declaration
11212 /home/blarsen/Documents/fsf_build/gdb/testsuite/../../../binutils-gdb/gdb/testsuite/gdb.cp/class2.cc:58:3: note: type is given name 'Dbase' for linkage purposes by this typedef declaration
11215 1 warning generated.
11216 gdb compile failed, /home/blarsen/Documents/fsf_build/gdb/testsuite/../../../binutils-gdb/gdb/testsuite/gdb.cp/class2.cc:53:14: warning: anonymous non-C-compatible type given name for linkage purposes by typedef declaration; add a tag name here [-Wnon-c-typedef-for-linkage]
11220 /home/blarsen/Documents/fsf_build/gdb/testsuite/../../../binutils-gdb/gdb/testsuite/gdb.cp/class2.cc:54:2: note: type is not C-compatible due to this member declaration
11223 /home/blarsen/Documents/fsf_build/gdb/testsuite/../../../binutils-gdb/gdb/testsuite/gdb.cp/class2.cc:58:3: note: type is given name 'Dbase' for linkage purposes by this typedef declaration
11226 1 warning generated.
11227 UNTESTED: gdb.cp/class2.exp: failed to prepare
11229 This can be silenced by adding -Wno-non-c-typedef-for-linkage for Clang
11230 11 or later. The test shows no failures with this change.
11232 Approved-by: Tom Tromey <tom@tromey.com>
11234 2022-11-11 Jan Beulich <jbeulich@suse.com>
11236 gas: accept custom ".linefile <n> ."
11237 While .linefile is generally intended for gas internal use only, its use
11238 in a source file would better not result in an internal error. Give use
11239 of it outside of any macro(-like) construct the meaning of restoring the
11240 original (physical) input file name.
11242 x86: drop stray IsString from PadLock insns
11243 The need for IsString on the PadLock insns went away with the
11244 introduction of RepPrefixOk. Drop these leftovers.
11246 x86: drop duplicate sse4a entry from cpu_arch[]
11247 Of the two instances the first is correct in using ANY_SSE4A as 3rd
11248 argument to SUBARCH(), so drop the wrong/redundant/dead 2nd one.
11250 2022-11-11 Alan Modra <amodra@gmail.com>
11252 PR28834, PR26946 sanity checking section size
11253 This patch provides a new function to sanity check section sizes.
11254 It's mostly extracted from what we had in bfd_get_full_section_contents
11255 but also handles compressed debug sections.
11257 - section file offset is taken into account,
11258 - added checks that a compressed section can be read from file.
11260 The function is then used when handling multiple .debug_* sections
11261 that need to be read into a single buffer, to sanity check sizes
11262 before allocating the buffer.
11265 * Makefile.am (LIBBFD_H_FILES): Add section.c.
11266 * compress.c (bfd_get_full_section_contents): Move section size
11268 * section.c (_bfd_section_size_insane): ..to here. New function.
11269 * dwarf2.c (read_section): Use _bfd_section_size_insane.
11270 (_bfd_dwarf2_slurp_debug_info): Likewise.
11271 * Makefile.in: Regenerate.
11272 * libbfd.h: Regenerate.
11274 2022-11-11 Alan Modra <amodra@gmail.com>
11276 Sanity check SHT_MIPS_OPTIONS size
11277 * elfxx-mips.c (_bfd_mips_elf_section_from_shdr): Use
11278 bfd_malloc_and_get_section to read contents of .MIPS.options.
11280 Re: gold: add --compress-debug-sections=zstd [PR 29641]
11282 compressed_output.cc:86:8: error: assignment of read-only variable ‘size’
11283 86 | size = ZSTD_compress(*compressed_data + header_size, size, uncompressed_data,
11285 2022-11-11 Fangrui Song <maskray@google.com>
11287 gold: add --compress-debug-sections=zstd [PR 29641]
11288 This option compresses output debug sections with zstd and sets ch_type
11289 to ELFCOMPRESS_ZSTD. Latest gdb and lldb support ELFCOMPRESS_ZSTD.
11291 There will be an error if zstd is not enabled at configure time.
11293 error: --compress-debug-sections=zstd: gold is not built with zstd support
11295 2022-11-11 Fangrui Song <maskray@google.com>
11297 gold, dwp: support zstd compressed input debug sections [PR 29641]
11298 This feature is enabled if config/zstd.m4 uses zstd.
11300 2022-11-11 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
11302 gprofng: fix typo in configure.ac
11304 2022-11-10 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
11306 * configure.ac: Fix typo in redirect operator.
11307 * configure: Rebuild.
11309 2022-11-11 Vladislav Khmelevsky <och95@yandex.ru>
11311 Fix adrp distance check
11313 * aarch64.cc (aarch64_valid_for_adrp_p): Shift offset
11314 as a signed number.
11316 2022-11-11 GDB Administrator <gdbadmin@sourceware.org>
11318 Automatic date update in version.in
11320 2022-11-10 Mike Frysinger <vapier@gentoo.org>
11322 sim: v850: rename v850.dc to align with other ports
11323 Other arches use the .dc extension for the instruction decode table.
11325 sim: igen: fix hang when decoding boolean rule constants
11326 The parser for boolean rules fails to skip over the , separator in
11327 the options which makes it hang forever. No dc files in the tree
11328 use boolean rules atm which is why no one noticed.
11330 sim: igen: mark error func as noreturn since it exits
11332 sim: igen: mark output funcs with printf attribute
11333 ... and fix the legitimate bug that it catches.
11335 sim: igen: constify various func arguments
11337 sim: ppc: rename ppc-instructions to powerpc.igen
11338 To make it clear this is an input to the igen tool, rename it with an
11339 igen extension. This matches the other files in the ppc dir (altivec
11340 & e500 igen files), and the other igen ports (mips, mn10300, v850).
11342 2022-11-10 H.J. Lu <hjl.tools@gmail.com>
11344 i386: Check invalid (%dx) usage
11345 (%dx) isn't a valid memory address in any modes. It is used as a special
11346 memory operand for input/output port address in AT&T syntax and should
11347 only be used with input/output instructions. Update i386_att_operand to
11348 set i.input_output_operand to true for (%dx) and issue an error if (%dx)
11349 is used with non-input/output instructions.
11352 * config/tc-i386.c (_i386_insn): Add input_output_operand.
11353 (md_assemble): Issue an error if input/output memory operand is
11354 used with non-input/output instructions.
11355 (i386_att_operand): Set i.input_output_operand to true for
11357 * testsuite/gas/i386/inval.l: Updated.
11358 * testsuite/gas/i386/x86-64-inval.l: Likewise.
11359 * testsuite/gas/i386/inval.s: Add tests for invalid (%dx) usage.
11360 * testsuite/gas/i386/x86-64-inval.s: Likewise.
11362 2022-11-10 Simon Marchi <simon.marchi@efficios.com>
11364 gdb: make "start" breakpoint inferior-specific
11365 I saw this failure on a CI:
11370 (gdb) PASS: gdb.threads/vfork-multi-inferior.exp: method=non-stop: add-inferior
11372 [Switching to inferior 2 [<null>] (<noexec>)]
11373 (gdb) PASS: gdb.threads/vfork-multi-inferior.exp: method=non-stop: inferior 2
11375 The program is not being run.
11376 (gdb) file /home/jenkins/workspace/binutils-gdb_master_linuxbuild/platform/jammy-amd64/target_board/unix/tmp/tmp.GYATAXR8Ku/gdb/testsuite/outputs/gdb.threads/vfork-multi-inferior/vfork-multi-inferior-sleep
11377 Reading symbols from /home/jenkins/workspace/binutils-gdb_master_linuxbuild/platform/jammy-amd64/target_board/unix/tmp/tmp.GYATAXR8Ku/gdb/testsuite/outputs/gdb.threads/vfork-multi-inferior/vfork-multi-inferior-sleep...
11379 Starting program: /home/jenkins/workspace/binutils-gdb_master_linuxbuild/platform/jammy-amd64/target_board/unix/tmp/tmp.GYATAXR8Ku/gdb/testsuite/outputs/gdb.threads/vfork-multi-inferior/vfork-multi-inferior-sleep
11380 (gdb) PASS: gdb.threads/vfork-multi-inferior.exp: method=non-stop: run inferior 2
11382 [Switching to inferior 1 [<null>] (<noexec>)]
11383 (gdb) PASS: gdb.threads/vfork-multi-inferior.exp: method=non-stop: inferior 1
11385 The program is not being run.
11386 (gdb) file /home/jenkins/workspace/binutils-gdb_master_linuxbuild/platform/jammy-amd64/target_board/unix/tmp/tmp.GYATAXR8Ku/gdb/testsuite/outputs/gdb.threads/vfork-multi-inferior/vfork-multi-inferior
11387 Reading symbols from /home/jenkins/workspace/binutils-gdb_master_linuxbuild/platform/jammy-amd64/target_board/unix/tmp/tmp.GYATAXR8Ku/gdb/testsuite/outputs/gdb.threads/vfork-multi-inferior/vfork-multi-inferior...
11388 (gdb) break should_break_here
11389 Breakpoint 1 at 0x11b1: file /home/jenkins/workspace/binutils-gdb_master_linuxbuild/platform/jammy-amd64/target_board/unix/src/binutils-gdb/gdb/testsuite/gdb.threads/vfork-multi-inferior.c, line 25.
11390 (gdb) PASS: gdb.threads/vfork-multi-inferior.exp: method=non-stop: break should_break_here
11391 [Thread debugging using libthread_db enabled]
11392 Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
11394 Temporary breakpoint 2 at 0x11c0: -qualified main. (2 locations)
11395 Starting program: /home/jenkins/workspace/binutils-gdb_master_linuxbuild/platform/jammy-amd64/target_board/unix/tmp/tmp.GYATAXR8Ku/gdb/testsuite/outputs/gdb.threads/vfork-multi-inferior/vfork-multi-inferior
11396 [Thread debugging using libthread_db enabled]
11397 Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
11399 Thread 2.1 "vfork-multi-inf" hit Temporary breakpoint 2, main () at /home/jenkins/workspace/binutils-gdb_master_linuxbuild/platform/jammy-amd64/target_board/unix/src/binutils-gdb/gdb/testsuite/gdb.threads/vfork-multi-inferior-sleep.c:23
11401 (gdb) FAIL: gdb.threads/vfork-multi-inferior.exp: method=non-stop: start inferior 1
11405 1. We start inferior 2 with "run&", it runs very slowly, takes time to
11407 2. We switch to inferior 1, and run "start"
11408 3. The temporary breakpoint inserted by "start" applies to all inferiors
11409 4. Inferior 2 hits that breakpoint and GDB reports that hit
11411 To avoid this, breakpoints inserted by "start" should be
11412 inferior-specific. However, we don't have a nice way to make
11413 inferior-specific breakpoints yet. It's possible to make
11414 pspace-specific breakpoints (for example how the internal_breakpoint
11415 constructor does) by creating a symtab_and_line manually. However,
11416 inferiors can share program spaces (usually on particular embedded
11417 targets), so we could have a situation where two inferiors run the same
11418 code in the same program space. In that case, it would just not be
11419 possible to insert a breakpoint in one inferior but not the other.
11421 A simple solution that should work all the time is to add a condition to
11422 the breakpoint inserted by "start", to check the inferior reporting the
11423 hit is the expected one. This is what this patch implements.
11425 Add a test that does:
11427 - start in background inferior 1 that sleeps before reaching its main
11428 function (using a sleep in a global C++ object's constructor)
11429 - start inferior 2 with the "start" command, which also sleeps before
11430 reaching its main function
11431 - validate that we hit the breakpoint in inferior 2
11433 Without the fix, we hit the breakpoint in inferior 1 pretty much all the
11434 time. There could be some unfortunate scheduling causing the test not
11435 to catch the bug, for instance if the scheduler decides not to schedule
11436 inferior 1 for a long time, but it would be really rare. If the bug is
11437 re-introduced, the test will catch it much more often than not, so it
11440 Reviewed-By: Bruno Larsen <blarsen@redhat.com>
11441 Approved-By: Pedro Alves <pedro@palves.net>
11442 Change-Id: Ib0148498a476bfa634ed62353c95f163623c686a
11444 2022-11-10 Bruno Larsen <blarsen@redhat.com>
11446 gdb: Fix regressions caused by 041de3d73aa121f2ff0c077213598963bfb34b79
11447 Commit 041de3d73aa changed the output format of all error messages when
11448 GDB couldn't determine a compatible overload for a given function, but
11449 it was only supposed to change if the failure happened due to incomplete
11450 types. This commit removes the stray . that was added
11452 2022-11-10 Aaron Merey <amerey@redhat.com>
11454 gdb/debuginfod: Improve progress updates
11455 If the download size is known, a progress bar is displayed along with
11456 the percentage of completion and the total download size.
11458 Downloading separate debug info for /lib/libxyz.so
11459 [############ ] 25% (10.01 M)
11461 If the download size is not known, a progress indicator is displayed
11462 with a ticker ("###") that moves across the screen at a rate of 1 tick
11465 Downloading separate debug info for /lib/libxyz.so
11468 If the output stream is not a tty, batch mode is enabled, the screen is
11469 too narrow or width has been set to 'unlimited', then only a static
11470 description of the download is printed. No bar or ticker is displayed.
11472 Downloading separate debug info for /lib/libxyz.so...
11474 In any case, if the size of the download is known at the time the
11475 description is printed then it will be included in the description.
11477 Downloading 10.01 MB separate debug info for /lib/libxyz.so...
11479 2022-11-10 Simon Marchi <simon.marchi@efficios.com>
11481 gdb: add special handling for frame level 0 in frame_info_ptr
11482 I noticed this problem while preparing the initial submission for the
11483 ROCm GDB port. One particularity of this patch set is that it does not
11484 support unwinding frames, that requires support of some DWARF extensions
11485 that will come later. It was still possible to run to a breakpoint and
11486 print frame #0, though.
11488 When rebasing on top of the frame_info_ptr work, GDB started tripping on
11489 a prepare_reinflate call, making it not possible anymore to event print
11490 the frame when stopping on a breakpoint. One thing to know about frame
11491 0 is that its id is lazily computed when something requests it through
11494 https://gitlab.com/gnutools/binutils-gdb/-/blob/23912acd402f5af9caf91b257e5209ec4c58a09c/gdb/frame.c#L2070-2080
11496 So, up to that prepare_reinflate call, frame 0's id was not computed,
11497 and prepare_reinflate, calling get_frame_id, forces it to be computed.
11498 Computing the frame id generally requires unwinding the previous frame,
11499 which with my ROCm GDB patch fails. An exception is thrown and the
11500 printing of the frame is simply abandonned.
11502 Regardless of this ROCm GDB problem (which is admittedly temporary, it
11503 will be possible to unwind with subsequent patches), we want to avoid
11504 prepare_reinflate to force the computing of the frame id, for the same
11505 reasons we lazily compute it in the first place.
11507 In addition, frame 0's id is subject to change across a frame cache
11508 reset. This is why save_selected_frame and restore_selected_frame have
11509 special handling for frame 0:
11511 https://gitlab.com/gnutools/binutils-gdb/-/blob/23912acd402f5af9caf91b257e5209ec4c58a09c/gdb/frame.c#L1841-1863
11513 For this last reason, we also need to handle frame 0 specially in
11514 prepare_reinflate / reinflate. Because the frame id of frame 0 can
11515 change across a frame cache reset, we must not rely on the frame id from
11516 that frame to reinflate it. We should instead just re-fetch the current
11517 frame at that point.
11519 This patch adds a frame_info_ptr::m_cached_level field, set in
11520 frame_info_ptr::prepare_reinflate, so we can tell if a frame is frame 0.
11521 There are cases where a frame_info_ptr object wraps a sentinel frame,
11522 for which frame_relative_level returns -1, so I have chosen the value -2
11523 to represent "invalid frame level", for when the frame_info_ptr object
11526 In frame_info_ptr::prepare_reinflate, only cache the frame id if the
11527 frame level is not 0. It's fine to cache the frame id for the sentinel
11528 frame, it will be properly handled by frame_find_by_id later.
11530 In frame_info_ptr::reinflate, if the frame level is 0, call
11531 get_current_frame to get the target's current frame. Otherwise, use
11532 frame_find_by_id just as before.
11534 This patch should not have user-visible changes with upstream GDB. But
11535 it will avoid forcing the computation of frame 0's when calling
11536 prepare_reinflate. And, well, it fixes the upcoming ROCm GDB patch
11539 Change-Id: I176ed7ee9317ddbb190acee8366e087e08e4d266
11540 Reviewed-By: Bruno Larsen <blarsen@redhat.com>
11542 2022-11-10 Simon Marchi <simon.marchi@polymtl.ca>
11544 gdb: add missing prepare_reinflate call in print_frame_info
11545 print_frame_info calls frame_info_ptr::reinflate, but not
11546 frame_info_ptr::prepare_reinflate, add the call to prepare_reinflate.
11547 It works right now, because all callers of print_frame_info that could
11548 possibly lead to the pretty printers being called, and the frame_info
11549 objects being invalidated, do call prepare_reinflate themselves. And
11550 since the cached frame id is copied when passing a frame_info_ptr by
11551 value, print_frame_info does have a cached frame id on entry. So
11552 technically, this change isn't needed. But I don't think it's good for
11553 a function to rely on its callers to have called prepare_reinflate, if
11554 it intends to call reinflate.
11556 Change-Id: Ie332b2d5479aef46f83fdc1120c7c83f4e84d1b0
11557 Reviewed-By: Bruno Larsen <blarsen@redhat.com>
11559 2022-11-10 Simon Marchi <simon.marchi@efficios.com>
11561 gdb: use frame_id_p instead of comparing to null_frame_id in frame_info_ptr::reinflate
11564 gdb_assert (m_cached_id != null_frame_id);
11566 is always true, as comparing equal to null_frame_id is always false
11567 (it's the first case in frame_id::operator==, not sure why it's not this
11568 way, but that's what it is).
11570 Replace the comparison with a call to frame_id_p.
11572 Approved-By: Tom Tromey <tom@tromey.com>
11573 Change-Id: I93986e6a85ac56353690792552e5b3b4cedec7fb
11575 2022-11-10 Simon Marchi <simon.marchi@efficios.com>
11577 gdb: remove manual frame_info reinflation code in backtrace_command_1
11578 With the following patch applied (gdb: use frame_id_p instead of
11579 comparing to null_frame_id in frame_info_ptr::reinflate), I would get:
11581 $ ./gdb -q -nx --data-directory=data-directory testsuite/outputs/gdb.base/bt-selected-frame/bt-selected-frame -ex "b breakpt" -ex r -ex "bt full"
11582 Reading symbols from testsuite/outputs/gdb.base/bt-selected-frame/bt-selected-frame...
11583 Breakpoint 1 at 0x1131: file /home/smarchi/src/binutils-gdb/gdb/testsuite/gdb.base/bt-selected-frame.c, line 22.
11584 Starting program: /home/smarchi/build/binutils-gdb/gdb/testsuite/outputs/gdb.base/bt-selected-frame/bt-selected-frame
11585 [Thread debugging using libthread_db enabled]
11586 Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
11588 Breakpoint 1, breakpt () at /home/smarchi/src/binutils-gdb/gdb/testsuite/gdb.base/bt-selected-frame.c:22
11590 #0 breakpt () at /home/smarchi/src/binutils-gdb/gdb/testsuite/gdb.base/bt-selected-frame.c:22
11592 /home/smarchi/src/binutils-gdb/gdb/frame-info.c:42: internal-error: reinflate: Assertion `frame_id_p (m_cached_id)' failed.
11594 This is because the code in backtrace_command_1 to manually reinflate
11595 `fi` steps overs frame_info_ptr's toes.
11599 fi.prepare_reinflate ();
11601 `fi` gets properly filled with the cached frame id. But when this
11604 fi = frame_find_by_id (frame_id);
11606 `fi` gets replaced by a brand new frame_info_ptr that doesn't have a
11607 cached frame id. Then this is called without a cached frame id:
11611 That doesn't cause any problem currently, since
11613 - the gdb_assert in the reinflate method doesn't actually do anything
11614 (the following patch fixes that)
11615 - `fi.m_ptr` will always be non-nullptr, since we just got it from
11616 frame_find_by_id, so reinflate will not do anything, it won't try to
11619 Fix that by removing the code to manually re-fetch the frame. That
11620 should be taken care of by frame_info_ptr::reinflate.
11622 Note that the old code checked if we successfully re-inflated the frame
11623 or not, and if not it did emit a warning. The equivalent in
11624 frame_info_ptr::reinflate asserts that the frame has been successfully
11625 re-inflated. It's not clear if / when this can happen, but if it can
11626 happen, we'll need to find a solution to this problem globally
11627 (everywhere a frame_info_ptr can be re-inflated), not just here. So I
11628 propose to leave it like this, until it does become a problem.
11630 Reviewed-By: Bruno Larsen <blarsen@redhat.com>
11631 Change-Id: I07b783d94e2853e0a2d058fe7deaf04eddf24835
11633 2022-11-10 Simon Marchi <simon.marchi@polymtl.ca>
11635 gdb: move frame_info_ptr method implementations to frame-info.c
11636 I don't see any particular reason why the implementations of the
11637 frame_info_ptr object are in the header file. It only seems to add some
11638 complexity. Since we can't include frame.h in frame-info.h, we have to
11639 add declarations of functions defined in frame.c, in frame-info.h. By
11640 moving the implementations to a new frame-info.c, we can avoid that.
11642 Change-Id: I435c828f81b8a3392c43ef018af31effddf6be9c
11643 Reviewed-By: Bruno Larsen <blarsen@redhat.com>
11644 Reviewed-By: Tom Tromey <tom@tromey.com>
11646 2022-11-10 Simon Marchi <simon.marchi@polymtl.ca>
11648 gdb: add prepare_reinflate/reinflate around print_frame_args in info_frame_command_core
11649 I noticed this crash:
11651 $ ./gdb --data-directory=data-directory -nx -q \
11652 testsuite/outputs/gdb.python/pretty-print-call-by-hand/pretty-print-call-by-hand \
11653 -x testsuite/outputs/gdb.python/pretty-print-call-by-hand/pretty-print-call-by-hand.py \
11656 Stack level 0, frame at 0x7fffffffdd80:
11657 rip = 0x555555555160 in g
11658 (/home/simark/src/binutils-gdb/gdb/testsuite/gdb.python/pretty-print-call-by-hand.c:41); saved rip = 0x5555555551a3
11659 called by frame at 0x7fffffffdda0
11661 Arglist at 0x7fffffffdd70, args: mt=mytype is 0x555555556004 "hello world",
11664 Fatal signal: Segmentation fault
11666 This is another case of frame_info being invalidated under a function's
11667 feet. The stack trace when the frame_info get invalidated looks like:
11669 ... many frames to pretty print the arg, that eventually invalidate the frame_infos ...
11670 #35 0x00005568d0a8ab24 in print_frame_arg (fp_opts=..., arg=0x7ffc3216bcb0) at /home/simark/src/binutils-gdb/gdb/stack.c:489
11671 #36 0x00005568d0a8cc75 in print_frame_args (fp_opts=..., func=0x621000233210, frame=..., num=-1, stream=0x60b000000300)
11672 at /home/simark/src/binutils-gdb/gdb/stack.c:898
11673 #37 0x00005568d0a9536d in info_frame_command_core (fi=..., selected_frame_p=true) at /home/simark/src/binutils-gdb/gdb/stack.c:1682
11675 print_frame_args knows that print_frame_arg can invalidate frame_info
11676 objects, and therefore calls prepare_reinflate/reinflate. However,
11677 info_frame_command_core has a separate frame_info_ptr instance (it is
11678 passed by value / copy). So info_frame_command_core needs to know that
11679 print_frame_args can invalidate frame_info objects, and therefore needs
11680 to prepare_reinflate/reinflate as well. Add those calls, and enhance
11681 the gdb.python/pretty-print-call-by-hand.exp test to test that command.
11683 Reviewed-By: Bruno Larsen <blarsen@redhat.com>
11684 Change-Id: I9edaae06d62e97ffdb30938d364437737238a960
11686 2022-11-10 Simon Marchi <simon.marchi@polymtl.ca>
11688 gdb: clear other.m_cached_id in frame_info_ptr's move ctor
11689 We do it in the move assignment operator, so I think it makes sense to
11690 do it here too for consistency. I don't think it's absolutely necessary
11691 to clear the other object's fields (in other words, copy constructor and
11692 move constructor could be the same), as there is no exclusive resource
11693 being transfered. The important thing is to leave the moved-from object
11694 in an unknown, but valid state. But still, I think that clearing the
11695 fields of the moved-from object is not a bad idea, it helps ensure we
11696 don't rely on the moved-from object after.
11698 Change-Id: Iee900ff9d25dad51d62765d694f2e01524351340
11699 Reviewed-By: Bruno Larsen <blarsen@redhat.com>
11701 2022-11-10 Bruno Larsen <blarsen@redhat.com>
11703 gdb/c++: Improve error messages in overload resolution
11704 When resolving overloaded functions, GDB relies on knowing relationships
11705 between types, i.e. if a type inherits from another. However, some
11706 compilers may not add complete information for given types as a way to
11707 reduce unnecessary debug information. In these cases, GDB would just say
11708 that it couldn't resolve the method or function, with no extra
11711 The problem is that sometimes the user may not know that the type
11712 information is incomplete, and may just assume that there is a bug in
11713 GDB. To improve the user experience, we attempt to detect if the
11714 overload match failed because of an incomplete type, and warn the user
11717 This commit also adds a testcase confirming that the message is only
11718 triggered in the correct scenario. This test was not developed as an
11719 expansion of gdb.cp/overload.cc because it needed the dwarf assembler,
11720 and porting all of overload.cc seemed unnecessary.
11722 Approved-By: Tom Tromey <tom@tromey.com>
11724 2022-11-10 Bruno Larsen <blarsen@redhat.com>
11726 gdb/testsuite: allowed for function_range to deal with mangled functions
11727 When calling get_func_info inside a test case, it would cause failures
11728 if the function was printed using a C++ style mangled name. The current
11729 patch fixes this by allowing for mangled names along with the current
11732 Approved-By: Tom Tromey <tom@tromey.com>
11734 2022-11-10 Clément Chigot <chigot@adacore.com>
11736 ld/testsuite: skip ld-size when -shared is not supported
11739 * testsuite/ld-size/size.exp: Skip when -shared is not
11742 2022-11-10 Alan Modra <amodra@gmail.com>
11744 mach-o reloc size overflow
11745 * mach-o.c (bfd_mach_o_canonicalize_reloc): Set bfd_error on
11748 2022-11-10 Alan Modra <amodra@gmail.com>
11750 Sanity check reloc count in get_reloc_upper_bound
11751 The idea here is the stop tools from allocating up to 32G per section
11752 for the arelent pointer array, only to find a little later that the
11753 section reloc count was fuzzed. This usually doesn't hurt much (on
11754 systems that allow malloc overcommit) except when compiled with asan.
11756 We already do this for ELF targets, and while fixing the logic
11757 recently I decided other targets ought to do the same.
11759 * elf64-sparc.c (elf64_sparc_get_reloc_upper_bound): Sanity check
11760 section reloc count against file size.
11761 * mach-o.c (bfd_mach_o_get_reloc_upper_bound): Likewise.
11762 * aoutx.h (get_reloc_upper_bound): Likewise, and don't duplicate
11763 check done in bfd_get_reloc_upper_bound.
11764 * pdp11.c (get_reloc_upper_bound): Likewise.
11765 * coffgen.c (coff_get_reloc_upper_bound): Likewise.
11767 2022-11-10 Lancelot SIX <lancelot.six@amd.com>
11769 gdb/testsuite: Fix rtld-step-nodebugsym.exp
11770 The test case introduced in bafcc335266 (Fix stepping in rtld without
11771 debug symbol) fails on some systems as reported by PR/29768. This can
11772 be seen if the system does not have debug info for the libc:
11775 Single stepping until exit from function main,^M
11776 which has no line number information.^M
11777 hello world[Inferior 1 (process 48203) exited normally]^M
11778 (gdb) PASS: gdb.base/rtld-step-nodebugsym.exp: step
11780 The program is not being run.^M
11781 (gdb) FAIL: gdb.base/rtld-step-nodebugsym.exp: continue until exit (the program is no longer running)
11783 Without glibc debug info, GDB steps until the program finishes, and
11784 then "gdb_continue_to_end" fails.
11786 As this test was designed to check that GDB does not crash in the "step"
11787 command, the continue does not carry real meaning to the test.
11789 Replace it by "print 0" so we still check that after the step command
11790 GDB is still alive, which is what we care about.
11792 Tested on Ubuntu-22.04 x86_64, with and without libc6-dbg.
11794 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29768
11795 Approved-By: Simon Marchi <simon.marchi@efficios.com>
11797 2022-11-10 Mike Frysinger <vapier@gentoo.org>
11799 sim: ppc: drop old makefile fragment
11800 Support for these files was dropped almost 30 years ago, but the ppc
11801 arch was missed. Clean that up now too.
11803 sim: ppc: drop support for dgen -L option
11804 Nothing passes this to dgen, and even if it did, nothing would happen
11805 because the generated spreg.[ch] files don't include any references
11806 back to the original data table. So drop it to simplify.
11808 sim: ppc: collapse is_readonly & length switch tables heavily
11809 Since we know we'll return 0 by default, we don't have to output case
11810 statements for readonly or length fields whose values are also zero.
11811 This is the most common case by far and thus generates a much smaller
11812 switch table in the end.
11814 2022-11-10 Mike Frysinger <vapier@gentoo.org>
11816 sim: ppc: collapse is_valid switch table more
11817 Instead of writing:
11824 Output a single return so we get:
11830 This saves ~100 lines of code. Hopefully the compiler was already
11831 smart enough to optimize to the same code, but if not, this probably
11832 helps there too :).
11834 2022-11-10 Mike Frysinger <vapier@gentoo.org>
11836 sim: ppc: pull default switch return out
11837 This saves a single line for the same result. By itself, it's not
11838 interesting, but we can further optimize the generated output and
11839 completely omit the switch table in some cases. Which we'll do in
11842 sim: ppc: constify spreg table
11843 This internal table is only ever read, so constify it.
11845 2022-11-10 Mark Harmstone <mark@harmstone.com>
11847 ld: Add module information substream to PDB files
11849 2022-11-10 Luis Machado <luis.machado@arm.com>
11851 [opcodes/arm] Fix potential null pointer dereferences
11854 As pointed out in the bug ticket, we have a couple potential null pointer
11855 dereferencing situations. Harden those.
11857 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29598
11859 2022-11-10 Luis Machado <luis.machado@arm.com>
11861 [gdb/aarch64] Use safer memory read routines
11864 As reported, we are using some memory read routines that don't handle read
11865 errors gracefully. Convert those to use the safe_* versions if available.
11867 This allows the code to handle those read errors in a more sensible way.
11869 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28796
11871 2022-11-10 GDB Administrator <gdbadmin@sourceware.org>
11873 Automatic date update in version.in
11875 2022-11-09 Lancelot SIX <lancelot.six@amd.com>
11877 Fix stepping in rtld without debug symbol
11878 Commit be6276e0aed "Allow debugging of runtime loader / dynamic linker"
11879 introduced a small regression when stepping into the runtime loader /
11880 dynamic linker from function we do not have debug information for. This
11881 is reported in PR/29747.
11883 This can be shown by the following example (given by Simon Marchi in
11884 buzilla bug report):
11894 $ gcc test.c -O0 -o test
11895 $ ./gdb -q -nx --data-directory=data-directory test -ex start -ex s
11896 Reading symbols from test...
11897 (No debugging symbols found in test)
11898 Temporary breakpoint 1 at 0x1151
11899 Starting program: .../binutils-gdb/gdb/test
11900 [Thread debugging using libthread_db enabled]
11901 Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
11903 Temporary breakpoint 1, 0x0000555555555151 in main ()
11904 Single stepping until exit from function main,
11905 which has no line number information.
11906 /home/smarchi/src/binutils-gdb/gdb/infrun.c:6960:64: runtime error: member call on null pointer of type 'struct symbol'
11908 The crash happens here:
11910 #0 __sanitizer::Die () at ../../../../src/libsanitizer/sanitizer_common/sanitizer_termination.cpp:50
11911 #1 0x00007ffff5dd7128 in __ubsan::__ubsan_handle_type_mismatch_v1_abort (Data=<optimized out>, Pointer=<optimized out>) at ../../../../src/libsanitizer/ubsan/ubsan_handlers.cpp:148
11912 #2 0x000055556183e1a7 in process_event_stop_test (ecs=0x7fffffffccd0) at .../binutils-gdb/gdb/infrun.c:6960
11913 #3 0x0000555561838ea4 in handle_signal_stop (ecs=0x7fffffffccd0) at .../binutils-gdb/gdb/infrun.c:6615
11914 #4 0x000055556182f77b in handle_inferior_event (ecs=0x7fffffffccd0) at .../binutils-gdb/gdb/infrun.c:5866
11918 6956 if (execution_direction != EXEC_REVERSE
11919 6957 && ecs->event_thread->control.step_over_calls == STEP_OVER_UNDEBUGGABLE
11920 6958 && in_solib_dynsym_resolve_code (ecs->event_thread->stop_pc ())
11921 6959 && !in_solib_dynsym_resolve_code (
11922 6961 ecs->event_thread->control.step_start_function->value_block ()
11923 6962 ->entry_pc ()))
11925 we dereference, ecs->event_thread->control.step_start_function which is
11928 This patch changes this condition so it evaluates to true if
11929 ecs->event_thread->control.step_start_function is nullptr since this
11930 matches the behaviour before be6276e0aed.
11932 Tested on ubuntu-22.04 x86_64.
11934 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29747
11935 Reviewed-By: Bruno Larsen <blarsen@redhat.com>
11936 Approved-By: Kevin Buettner <kevinb@redhat.com>
11938 2022-11-09 Mike Frysinger <vapier@gentoo.org>
11940 sim: igen: add missing newline to various error messages
11941 The error() function expects a trailing newline in its message.
11942 Most callers do this already, so adding it to the few that don't.
11944 sim: restore lstat & mkdir func checks
11945 When merging ppc configure checks into the top-level, these 2 funcs
11946 were accidentally dropped (probably due to incorrect resolution of
11947 conflicts). Restore them since the ppc code utilizes them both.
11949 sim: ppc: drop obsolete USE_WIN32API check
11950 This controls only one thing: how to call mkdir(). The gnulib code
11951 already has a mkdir module that provides this exact logic for us, so
11952 punt the code entirely.
11954 2022-11-09 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
11956 gdbserver: do not report btrace support if target does not announce it
11957 Gdbserver unconditionally reports support for btrace packets. Do not
11958 report the support, if the underlying target does not say it supports
11959 it. Otherwise GDB would query the server with btrace-related packets
11962 2022-11-09 Tom Tromey <tromey@adacore.com>
11964 Allow 'ptype/o' for assembly
11965 PR exp/28359 points out that 'ptype/o' does not work when the current
11968 I tracked this down to a hard-coded list of languages in typeprint.c.
11969 This patch replaces this list with a method on 'language_defn'
11970 instead. If all languages are ever updated to have this feature, the
11971 method could be removed; but in the meantime this lets each language
11972 control what happens.
11974 I looked at having each print_type method simply modify the flags
11975 itself, but this doesn't work very well with the feature that disables
11976 method-printing by default (but allows it via a flag).
11978 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28359
11979 Approved-By: Andrew Burgess <aburgess@redhat.com>
11980 Approved-By: Keith Seitz <keiths@redhat.com>
11982 2022-11-09 Mike Frysinger <vapier@gentoo.org>
11984 sim: ppc: add missing parens with e500 macro
11985 This macro expansion was missing a set of outer-most parenthesis which
11986 some compilers would complain about depending on how the macro is used.
11987 This is just standard good macro hygiene too.
11989 sim: ppc: drop useless linking of helper tools
11990 We've never run these helper programs directly. The igen program
11991 includes the relevant source files directly and runs the code that
11992 way. So stop wasting developer CPU time linking programs that are
11993 never run. We leave the rules in place for people who need to test
11994 and debug the specific bits of code every now & then.
11996 2022-11-09 Jan Beulich <jbeulich@suse.com>
11998 x86/Intel: don't accept malformed EXTRQ / INSERTQ
11999 Operand swapping was mistakenly suppressed when the first two operands
12000 were immediate ones, not taking into account overall operand count. This
12001 way EXTRQ / INSERTQ would have been accepted also with kind-of-AT&T
12004 For the testcase being extended, in order to not move around "GAS
12005 LISTING" expectations, suppress pagination.
12007 2022-11-09 Alan Modra <amodra@gmail.com>
12009 Re: Fuzzed files in archives
12010 Like commit ffbe89531c2e this avoids more silliness writing output
12011 that is going to be deleted. bfd_close and bfd_close_all_done differ
12012 in that only the former calls _bfd_write_contents.
12014 * objcopy.c (copy_archive): Don't call bfd_close for elements
12015 that are going to be deleted, call bfd_close_all_done instead.
12016 Do the same for the archive itself.
12018 2022-11-09 Christoph Müllner <christoph.muellner@vrull.eu>
12020 RISC-V: xtheadfmemidx: Use fp register in mnemonics
12021 Although the encoding for scalar and fp registers is identical,
12022 we should follow common pratice and use fp register names
12023 when referencing fp registers.
12025 The xtheadmemidx extension consists of indirect load/store instructions
12026 which all load to or store from fp registers.
12027 Let's use fp register names in this case and adjust the test cases
12031 * testsuite/gas/riscv/x-thead-fmemidx-fail.l: Updated since rd need to
12033 * testsuite/gas/riscv/x-thead-fmemidx-fail.s: Likewise.
12034 * testsuite/gas/riscv/x-thead-fmemidx.d: Likewise.
12035 * testsuite/gas/riscv/x-thead-fmemidx.s: Likewise.
12037 * riscv-opc.c (riscv_opcodes): Updated since rd need to be float register.
12039 2022-11-09 H.J. Lu <hjl.tools@gmail.com>
12041 ld: Always output local symbol for relocatable link
12043 * elflink.c (elf_link_output_symstrtab): Don't skip local symbol
12044 in SEC_EXCLUDE section for relocatable link.
12046 2022-11-09 GDB Administrator <gdbadmin@sourceware.org>
12048 Automatic date update in version.in
12050 2022-11-08 Simon Marchi <simon.marchi@efficios.com>
12052 gdb/linux-nat: get core count using /sys/devices/system/cpu/possible
12053 I get this test failure on my CI;
12055 FAIL: gdb.base/info-os.exp: get process list
12057 The particularity of this setup is that builds are done in containers
12058 who are allocated 4 CPUs on a machine that has 40. The code in
12059 nat/linux-osdata.c fails to properly fetch the core number for each
12062 linux_xfer_osdata_processes uses `sysconf (_SC_NPROCESSORS_ONLN)`, which
12063 returns 4, so it allocates an array of 4 integers. However, the core
12064 numbers read from /proc/pid/task/tid/stat, by function
12065 linux_common_core_of_thread, returns a value anywhere between 0 and 39.
12066 The core numbers above 3 are therefore ignored, many processes end up
12067 with no core value, and the regexp in the test doesn't match (it
12068 requires an integer as the core field).
12070 The way this the CPUs are exposed to the container is that the container
12071 sees 40 CPUs "present" and "possible", but only 4 arbitrary CPUs
12074 root@ci-node-jammy-amd64-04-08:~# cat /sys/devices/system/cpu/present
12076 root@ci-node-jammy-amd64-04-08:~# cat /sys/devices/system/cpu/online
12078 root@ci-node-jammy-amd64-04-08:~# cat /sys/devices/system/cpu/possible
12081 The solution proposed in this patch is to find out the number of
12082 possible CPUs using /sys/devices/system/cpu/possible. In practice, this
12083 will probably always contain `0-N`, where N is the number of CPUs, minus
12084 one. But the documentation [1] doesn't such guarantee, so I'll assume
12085 that it can contain a more complex range list such as `2,4-31,32-63`,
12086 like the other files in that directory can have. The solution is to
12087 iterate over these numbers to find the highest possible CPU id, and
12088 use that that value plus one as the size of the array to allocate.
12090 [1] https://www.kernel.org/doc/Documentation/admin-guide/cputopology.rst
12092 Change-Id: I7abce2e43b000c1327fa94cd7b99d46e49d7ccf3
12094 2022-11-08 Simon Marchi <simon.marchi@efficios.com>
12096 gdbsupport, gdb: add read_text_file_to_string, use it in linux_common_core_of_thread
12097 I would like to add more code to nat/linux-osdata.c that reads an entire
12098 file from /proc or /sys and processes it as a string afterwards. I
12099 would like to avoid duplicating the somewhat error-prone code that reads
12100 an entire file to a buffer. I think we should have a utility function
12103 Add read_file_to_string to gdbsupport/filestuff.{c,h}, and make
12104 linux_common_core_of_thread use it. I want to make the new function
12105 return an std::string, and because strtok doesn't play well with
12106 std::string (it requires a `char *`, std::string::c_str returns a `const
12107 char *`), change linux_common_core_of_thread to use std::string methods
12110 Approved-By: Tom Tromey <tom@tromey.com>
12111 Change-Id: I1793fda72a82969c28b944a84acb953f74c9230a
12113 2022-11-08 Peter Bergner <bergner@linux.ibm.com>
12115 PowerPC: Add XSP operand define
12117 * ppc-opc.c (XSP): New define.
12118 (powerpc_opcodes) <stxvp, stxvpx, pstxvp>: Use it.
12120 2022-11-08 Tom de Vries <tdevries@suse.de>
12122 [gdb/cli] Make quit really quit after remote connection closed
12123 Consider a hello world a.out, started using gdbserver:
12125 $ gdbserver --once 127.0.0.1:2345 ./a.out
12126 Process ./a.out created; pid = 15743
12127 Listening on port 2345
12129 that we can connect to using gdb:
12131 $ gdb -ex "target remote 127.0.0.1:2345"
12132 Remote debugging using 127.0.0.1:2345
12133 Reading /home/vries/a.out from remote target...
12135 0x00007ffff7dd4550 in _start () from target:/lib64/ld-linux-x86-64.so.2
12139 After that, we can for instance quit with confirmation:
12142 A debugging session is active.
12144 Inferior 1 [process 16691] will be killed.
12146 Quit anyway? (y or n) y
12150 Or, kill with confirmation and quit:
12153 Kill the program being debugged? (y or n) y
12154 [Inferior 1 (process 16829) killed]
12159 Or, monitor exit, kill with confirmation, and quit:
12163 Kill the program being debugged? (y or n) y
12164 Remote connection closed
12169 But when doing monitor exit followed by quit with confirmation, we get the gdb
12170 prompt back, requiring us to do quit once more:
12174 A debugging session is active.
12176 Inferior 1 [process 16944] will be killed.
12178 Quit anyway? (y or n) y
12179 Remote connection closed
12184 So, the first quit didn't quit. This happens as follows:
12185 - quit_command calls query_if_trace_running
12186 - a TARGET_CLOSE_ERROR is thrown
12187 - it's caught in remote_target::get_trace_status, but then
12188 rethrown because it's TARGET_CLOSE_ERROR
12189 - catch_command_errors catches the error, at which point the quit command
12192 The TARGET_CLOSE_ERROR is defined as:
12194 /* Target throwing an error has been closed. Current command should be
12195 aborted as the inferior state is no longer valid. */
12196 TARGET_CLOSE_ERROR,
12198 so in a way this is expected behaviour. But aborting quit because the inferior
12199 state (which we've already confirmed we're not interested in) is no longer
12200 valid, and having to type quit again seems pointless.
12202 Furthermore, the purpose of not catching errors thrown by
12203 query_if_trace_running as per commit 2f9d54cfcef ("make -gdb-exit call
12204 disconnect_tracing too, and don't lose history if the target errors on
12205 "quit""), was to make sure that error (_("Not confirmed.") had effect.
12207 Fix this in quit_command by catching only the TARGET_CLOSE_ERROR exception
12208 during query_if_trace_running and reporting it:
12212 A debugging session is active.
12214 Inferior 1 [process 19219] will be killed.
12216 Quit anyway? (y or n) y
12217 Remote connection closed
12221 Tested on x86_64-linux.
12224 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=15746
12225 Approved-By: Tom Tromey <tom@tromey.com>
12227 2022-11-08 Tom de Vries <tdevries@suse.de>
12229 [gdb/testsuite] Remove test-case from test name
12230 Remove test-cases from test-names, such that we don't have the redundant:
12232 PASS: gdb.base/corefile.exp: backtrace in corefile.exp
12236 PASS: gdb.base/corefile.exp: backtrace
12239 Fixed all instances found using:
12241 $ grep ":.*:.*\.exp" gdb.sum
12244 Tested on x86_64-linux.
12246 2022-11-08 Tom de Vries <tdevries@suse.de>
12248 [gdb/testsuite] Fix find_core_file for core named core
12249 With test-case gdb.base/bigcore.exp I run into:
12251 (gdb) PASS: gdb.base/bigcore.exp: get inferior pid
12253 Continuing with signal SIGABRT.^M
12255 Program terminated with signal SIGABRT, Aborted.^M
12256 The program no longer exists.^M
12257 (gdb) PASS: gdb.base/bigcore.exp: signal SIGABRT
12258 UNTESTED: gdb.base/bigcore.exp: can't generate a core file
12260 due to find_core_file returning "".
12262 There is a core file name core:
12264 $ ls ./outputs/gdb.base/bigcore
12265 bigcore bigcore.corefile core gdb.cmd.1 gdb.in.1 gdbserver.cmd.1
12267 but it's not found.
12269 The problem is this statement:
12271 lappend files [list ${::testfile}.core core]
12273 which adds a single list item "${::testfile}.core core".
12275 Fix this in the most readable way:
12277 lappend files ${::testfile}.core
12281 Tested on x86_64-linux.
12283 2022-11-08 Mike Frysinger <vapier@gentoo.org>
12285 sim: mips: call Unpredictable instead of setting bogus values [PR sim/29276]
12286 The intention of this code seems to be to indicate that this insn
12287 should not be used and produces undefined behavior, so instead of
12288 setting registers to bogus values, call Unpredictable. This fixes
12289 build warnings due to 32-bit/64-bit type conversions, and outputs
12290 a log message for users at runtime instead of silent corruption.
12292 Bug: https://sourceware.org/PR29276
12294 2022-11-08 Mike Frysinger <vapier@gentoo.org>
12296 sim: drop unused CORE_ADDR_TYPE
12297 This hasn't been used by gdb in decades, and doesn't make sense with
12298 a standalone sim program/library where the ABI is fixed. So punt it
12299 to simplify the code.
12301 2022-11-08 Haochen Jiang <haochen.jiang@intel.com>
12303 x86: Correct wrong comments in vex_w_table
12306 This wrong comment was introduced by previous AVX-VNNI-INT8 commit.
12308 Committed as obvious fix.
12315 * i386-dis.c (VEX_W_0F3851): Corrected from
12318 2022-11-08 Kong Lingling <lingling.kong@intel.com>
12320 Support Intel RAO-INT
12323 * NEWS: Support Intel RAO-INT.
12324 * config/tc-i386.c: Add raoint.
12325 * doc/c-i386.texi: Document .raoint.
12326 * testsuite/gas/i386/i386.exp: Run RAO_INT tests.
12327 * testsuite/gas/i386/raoint-intel.d: New test.
12328 * testsuite/gas/i386/raoint.d: Ditto.
12329 * testsuite/gas/i386/raoint.s: Ditto.
12330 * testsuite/gas/i386/x86-64-raoint-intel.d: Ditto.
12331 * testsuite/gas/i386/x86-64-raoint.d: Ditto.
12332 * testsuite/gas/i386/x86-64-raoint.s: Ditto.
12336 * i386-dis.c (PREFIX_0F38FC): New.
12337 (prefix_table): Add PREFIX_0F38FC.
12338 * i386-gen.c: (cpu_flag_init): Add CPU_RAO_INT_FLAGS and
12339 CPU_ANY_RAO_INT_FLAGS.
12340 * i386-init.h: Regenerated.
12341 * i386-opc.h: (CpuRAO_INT): New.
12342 (i386_cpu_flags): Add cpuraoint.
12343 * i386-opc.tbl: Add RAO_INT instructions.
12344 * i386-tbl.h: Regenerated.
12346 2022-11-08 GDB Administrator <gdbadmin@sourceware.org>
12348 Automatic date update in version.in
12350 2022-11-07 Tom Tromey <tromey@adacore.com>
12352 Silence libtool during link
12353 The switch to linking with libtool now shows a very long link line
12354 even when V=0. This patch arranges to silence libtool in this
12357 Approved-By: Simon Marchi <simon.marchi@efficios.com>
12359 2022-11-07 Simon Marchi <simon.marchi@efficios.com>
12361 gdb: make lookup_selected_frame static
12362 Change-Id: Ide2749a34333110c7f0112b25852c78cace0d2b4
12364 2022-11-07 Mike Frysinger <vapier@gentoo.org>
12366 sim: riscv: add missing AC_MSG_RESULT call
12367 Previous commit in here forgot to include this.
12369 sim: v850: drop subdir configure logic
12370 We've been using this only to set the default word size to 32. We
12371 can easily move this into the makefile via a -D compiler flag and
12372 clean up the build logic quite a bit.
12374 sim: mn10300: drop subdir configure logic
12375 We've been using this only to set the default word size to 32. We
12376 can easily move this into the makefile via a -D compiler flag and
12377 clean up the build logic quite a bit.
12379 sim: or1k: drop subdir configure logic
12380 We've been using this only to set the default word size to 32. We
12381 can easily move this into the makefile via a -D compiler flag and
12382 clean up the build logic quite a bit.
12384 sim: bpf: drop subdir configure logic
12385 We've been using this only to set the default word size to 64. We
12386 can easily move this into the makefile via a -D compiler flag and
12387 clean up the build logic quite a bit.
12389 sim: riscv: drop subdir configure logic
12390 We've been using this only to set the default word size to 32-vs-64
12391 based on the $target. We can easily merge this with the top-level
12392 configure script to clean things up a bit.
12394 2022-11-07 Jose E. Marchesi <jose.marchesi@oracle.com>
12396 gdb: link executables with libtool
12397 This patch changes the GDB build system in order to use libtool to
12398 link the several built executables. This makes it possible to refer
12399 to libtool libraries (.la files) in CLIBS.
12401 As an application of the above,
12403 BFD now refers to ../libbfd/libbfd.la
12404 OPCODES now refers to ../opcodes/libopcodes.la
12405 LIBBACKTRACE_LIB now refers to ../libbacktrace/libbacktrace.la
12406 LIBCTF now refers to ../libctf/libctf.la
12408 NOTE1: The addition of libtool adds a few new configure-time options
12409 to GDB. Among these, --enable-shared and --disable-shared, which were
12410 previously ignored. Now GDB shall honor these options when linking,
12411 picking up the right version of the referred libtool libraries
12414 NOTE2: I have not tested the insight build.
12416 NOTE3: For regenerating configure I used an environment with Autoconf
12417 2.69 and Automake 1.15.1. This should match the previously
12418 used version as announced in the configure script.
12420 NOTE4: Now the installed shared objects libbfd.so, libopcodes.so and
12421 libctf.so are used by gdb if binutils is installed with
12426 - --enable-shared and --disable-shared (the default in binutils) work
12427 as expected: the linked executables link with the archive or shared
12428 libraries transparently.
12430 - Makefile.in modified for EXEEXT = .exe. It installs the binaries
12431 just fine. The installed gdb.exe runs fine.
12433 - Native build regtested in x86_64. No regressions found.
12435 - Cross build for aarch64-linux-gnu built to exercise
12436 program_transform_name and friends. The installed
12437 aarch64-linux-gnu-gdb runs fine.
12439 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29372
12440 Approved-By: Simon Marchi <simon.marchi@efficios.com>
12442 2022-11-07 Simon Marchi <simon.marchi@polymtl.ca>
12444 gdb/testsuite: use a more unique name in gdb.mi/mi-breakpoint-multiple-locations.exp
12445 I see failures in this test, due to the function name "add" being too
12446 generic, and unexpected breakpoint locations being found in my
12447 libstdc++, such as (wrapped for readability):
12450 number="2.4",enabled="y",addr="0x00007ffff7d67e68",
12451 func="(anonymous namespace)::fast_float::bigint::add",
12452 file="/usr/src/debug/gcc/libstdc++-v3/src/c++17/fast_float/fast_float.h",
12453 fullname="/usr/src/debug/gcc/libstdc++-v3/src/c++17/fast_float/fast_float.h",
12454 line="1815", thread-groups=["i1"]
12457 Change the test to use a more unique name.
12459 Change-Id: I91de781be62d246eb41c73eaa410ebdd12633d1d
12461 2022-11-07 Pedro Alves <pedro@palves.net>
12463 Don't explicitly set clone child ptrace options
12464 linux_handle_extended_wait calls target_post_attach if we're handling
12465 a PTRACE_EVENT_CLONE, and libthread_db.so isn't active.
12466 target_post_attach just calls linux_init_ptrace_procfs to set the
12467 lwp's ptrace options. However, this is completely unnecessary,
12468 because, as man ptrace [1] says, options are inherited:
12470 "Flags are inherited by new tracees created and "auto-attached" via
12471 active PTRACE_O_TRACEFORK, PTRACE_O_TRACEVFORK, or PTRACE_O_TRACECLONE
12474 This removes the unnecessary call.
12476 [1] - https://man7.org/linux/man-pages/man2/ptrace.2.html
12478 Approved-By: Simon Marchi <simon.marchi@efficios.com>
12479 Change-Id: I533eaa60b700f7e40760311fc0d344d0b3f19a78
12481 2022-11-07 Mike Frysinger <vapier@gentoo.org>
12483 sim: .gdbinit: generate for all arch subdirs
12484 This was being skipped for ports that had a recursive configure,
12485 but we want it for them too.
12487 sim: build: add a proper var for enabled arches
12488 The install code was using $SUBDIRS to track all enabled arches. This
12489 works, but isn't great if we want to add a subdir that isn't an arch
12490 port, or as we merge the subdirs into the top-level. Create a new var
12491 explicitly to track the list of enabled arches instead.
12493 2022-11-07 Christophe Lyon <christophe.lyon@arm.com>
12495 configure: require libzstd >= 1.4.0
12496 gas uses ZSTD_compressStream2 which is only available with libzstd >=
12497 1.4.0, leading to build errors when an older version is installed.
12499 This patch updates the check libzstd presence to check its version is
12500 >= 1.4.0. However, since gas seems to be the only component requiring
12501 such a recent version this may imply that we disable ZSTD support for
12502 all components although some would still benefit from an older
12505 I ran 'autoreconf -f' in all directories containing a configure.ac
12506 file, using vanilla autoconf-2.69 and automake-1.15.1. I noticed
12507 several errors from autoheader in readline, as well as warnings in
12508 intl, but they are unrelated to this patch.
12510 This should fix some of the buildbots.
12518 2022-11-07 Clément Chigot <chigot@adacore.com>
12520 ld/testsuite: skip tests related to -shared when disabled
12521 Call the helper function "check_shared_lib_support" to ensure -shared
12522 is enabled before launching ld-shared, ld-elfweak and ld-elfvers.
12523 This allows to catch custom targets explicitly disabling it.
12527 * testsuite/ld-elfvers/vers.exp: Call check_shared_lib_support.
12528 * testsuite/ld-elfweak/elfweak.exp: Likewise.
12529 * testsuite/ld-shared/shared.exp: Likewise.
12531 2022-11-07 Tsukasa OI <research_trasio@irq.a4lg.com>
12533 RISC-V: Remove RV32EF conflict
12534 Despite that the RISC-V ISA Manual version 2.2 prohibited "RV32EF", later
12535 versions beginning with the version 20190608-Base-Ratified removed this
12536 restriction. Because the 'E' extension is still a draft, the author chose
12537 to *just* remove the conflict (not checking the ISA version).
12539 Note that, because RV32E is only used with a soft-float calling convention,
12540 there's no valid official ABI for RV32EF. It means, even if we can assemble
12541 a program with -march=rv32ef -mabi=ilp32e, floating-point registers are kept
12542 in an unmanaged state (outside ABI management).
12544 The purpose of this commit is to suppress unnecessary errors while parsing
12545 an ISA string and/or disassembling, not to allow hard-float with RVE.
12549 * elfxx-riscv.c (riscv_parse_check_conflicts): Accept RV32EF
12550 because only older specifications disallowed it.
12554 * testsuite/gas/riscv/march-fail-rv32ef.d: Remove as not directly
12556 * testsuite/gas/riscv/march-fail-rv32ef.l: Likewise.
12558 2022-11-07 GDB Administrator <gdbadmin@sourceware.org>
12560 Automatic date update in version.in
12562 2022-11-06 Mike Frysinger <vapier@gentoo.org>
12564 sim: build: respect AM_MAKEFLAGS when entering subdirs
12565 This doesn't matter right now, but it will as we add more flags to
12566 the recursive make step to pass state down.
12568 sim: build: stop passing down SIM_PRIMARY_TARGET
12569 This was needed when the install step was run in subdirs, but now
12570 that we process that entirely in the top-level, we don't need to
12571 pass this down, so drop it.
12573 2022-11-06 GDB Administrator <gdbadmin@sourceware.org>
12575 Automatic date update in version.in
12577 2022-11-05 Tom Tromey <tom@tromey.com>
12579 Deprecate MI version 1
12580 MI version 1 is long since obsolete. Rather than remove it
12581 immediately (though I did send a patch for that), instead let's
12582 deprecate it in GDB 13 and then remove it for GDB 14.
12584 This version of the patch incorporates Simon's warning change, and
12585 Luis' recommendation to mention the gdb versions here.
12587 2022-11-05 Mike Frysinger <vapier@gentoo.org>
12589 sim: fix readline linkage
12590 Now that we link programs in the top dir instead of the arch subdir,
12591 update the readline library path to be relative to the top dir.
12593 sim: use libtool to install programs
12594 Now that we use libtool to link, we have to use it to install instead
12595 of keeping the manual logic so we don't install wrapper shell scripts.
12597 sim: bfin: move linux-fixed-code.h to top-level
12599 2022-11-05 Mike Frysinger <vapier@gentoo.org>
12601 sim: run: move linking into top-level
12602 Automake will run each subdir individually before moving on to the next
12603 one. This means that the linking phase, a single threaded process, will
12604 not run in parallel with anything else. When we have to link ~32 ports,
12605 that's 32 link steps that don't take advantage of parallel systems. On
12606 my really old 4-core system, this cuts a multi-target build from ~60 sec
12607 to ~30 sec. We eventually want to move all compile+link steps to this
12608 common dir anyways, so might as well move linking now for a nice speedup.
12610 We use noinst_PROGRAMS instead of bin_PROGRAMS because we're taking care
12611 of the install ourselves rather than letting automake process it.
12613 2022-11-05 Mike Frysinger <vapier@gentoo.org>
12615 sim: build: add uninstall support
12616 This never worked before, but adding it to the common top-level dir
12617 is pretty easy to do now that we're unified.
12619 sim: build: move install steps to the top-level
12620 We still have to maintain custom install rules due to how we rename
12621 arch-specific files with an arch prefix in their name, but we can at
12622 least unify the logic in the common dir.
12624 sim: cris: move rvdummy linking to top-level
12625 This is only used by `make check`, so we can move it out of the
12628 sim: build: add SIM_HW_CFLAGS to top-level build too
12629 This matches what we do with targets already.
12631 sim: drop unused SIM_HARDWARE variable
12632 This hasn't been used since the refactor way back in commit
12633 f872d0d643968c1101bb8c07b252edd54f626da2 ("Only enable H/W
12634 on some mips targets."), so punt it.
12636 sim: adjust sim_hw options style
12637 We use uppercase for other variables, and are already turning it to
12638 uppercase in the arch-subdir.mk, so convert it in the configure step.
12640 sim: ppc: drop unused /dev/zero logic
12641 Nothing in the tree checks this option, or has checked for decades.
12642 The pre-cvs-import ChangeLog suggests this was added & removed back
12643 then, but can't be sure as that history doesn't exist in the VCS.
12645 sim: ppc: delete unused host bitsize settings
12646 Nothing checks this define anywhere, so drop all the logic. We don't
12647 want this to be a configure option in the first place as all such usage
12648 should be automatic & following proper types.
12650 sim: ppc: inline the sim-packages option
12651 This has only ever had a single option that's enabled by default.
12652 The objects it adds are pretty small and don't add overhead at
12653 runtime if it isn't used, so just enable it all the time to make
12654 the build code simpler.
12656 2022-11-05 GDB Administrator <gdbadmin@sourceware.org>
12658 Automatic date update in version.in
12660 2022-11-04 H.J. Lu <hjl.tools@gmail.com>
12662 binutils: Run PR binutils/26160 test
12663 Update expected PR binutils/26160 test output for readelf out change
12664 and run PR binutils/26160 test.
12667 * testsuite/binutils-all/pr26160.r: Updated.
12668 * testsuite/binutils-all/readelf.exp: Run PR binutils/26160 test.
12670 2022-11-04 Lancelot SIX <lancelot.six@amd.com>
12672 [testsuite] gdb.base/dlmopen: Fix test name and use gdb_attach
12673 One test name in gdb.base/dlmopen.exp changes from run to run
12674 since it includes a process id:
12676 PASS: gdb.base/dlmopen.exp: attach 3442682
12678 This is not convenient do diff gdb.sum files to compare test runs.
12680 Fix by using gdb_attach helper function to handle attaching to the
12681 process as it produce a constant test name.
12683 While at it also check gdb_attach's return value to only run the
12684 rest of the test if the attach was successful.
12686 Approved-By: Simon Marchi <simon.marchi@efficios.com>
12688 2022-11-04 Carl Love <cel@us.ibm.com>
12690 PowerPC update comments for the MMA instruction name changes.
12691 The mnemonics for the pmxvf16ger*, pmxvf32ger*,pmxvf64ger*, pmxvi4ger8*,
12692 pmxvi8ger4*, and pmxvi16ger2* instructions were officially changed to
12693 pmdmxbf16ger*, pmdmxvf32ger*, pmdmxvf64ger*, pmdmxvi4ger8*, pmdmxvi8ger4*,
12694 pmdmxvi16ger* respectively. The old mnemonics are still supported by the
12695 assembler as extended mnemonics. The disassembler generates the new
12696 mnemonics. The name changes occurred in commit:
12698 commit bb98553cad4e017f1851153fa5de91f2cee98fb2
12699 Author: Peter Bergner <bergner@linux.ibm.com>
12700 Date: Sat Oct 8 16:19:51 2022 -0500
12702 PowerPC: Add support for RFC02658 - MMA+ Outer-Product Instructions
12705 * config/tc-ppc.c (md_assemble): Only check for prefix opcodes.
12706 * testsuite/gas/ppc/rfc02658.s: New test.
12707 * testsuite/gas/ppc/rfc02658.d: Likewise.
12708 * testsuite/gas/ppc/ppc.exp: Run it.
12711 * ppc-opc.c (XMSK8, P_GERX4_MASK, P_GERX2_MASK, XX3GERX_MASK): New.
12712 (powerpc_opcodes): Add dmxvi8gerx4pp, dmxvi8gerx4, dmxvf16gerx2pp,
12713 dmxvf16gerx2, dmxvbf16gerx2pp, dmxvf16gerx2np, dmxvbf16gerx2,
12714 dmxvi8gerx4spp, dmxvbf16gerx2np, dmxvf16gerx2pn, dmxvbf16gerx2pn,
12715 dmxvf16gerx2nn, dmxvbf16gerx2nn, pmdmxvi8gerx4pp, pmdmxvi8gerx4,
12716 pmdmxvf16gerx2pp, pmdmxvf16gerx2, pmdmxvbf16gerx2pp, pmdmxvf16gerx2np,
12717 pmdmxvbf16gerx2, pmdmxvi8gerx4spp, pmdmxvbf16gerx2np, pmdmxvf16gerx2pn,
12718 pmdmxvbf16gerx2pn, pmdmxvf16gerx2nn, pmdmxvbf16gerx2nn.
12720 This patch updates the comments in the various gdb files to reflect the
12721 name changes. There are no functional changes made by this patch.
12723 The older instruction names are still used in the test
12724 gdb.reverse/ppc_record_test_isa_3_1.exp for backwards compatibility.
12726 Patch has been tested on Power 10 with no regressions.
12728 2022-11-04 Carl Love <cel@us.ibm.com>
12730 PowerPC fix for the gdb.arch/powerpc-power10.exp test.
12731 The mnemonics for the pmxvf16ger*, pmxvf32ger*,pmxvf64ger*, pmxvi4ger8*,
12732 pmxvi8ger4*, pmxvi16ger2* instructions were officially changed to
12733 pmdmxvf16ger*, pmdmxvf32ger*, pmdmxvf64ger*, pmdmxvi4ger8*, pmdmxvi8ger4*,
12734 pmdmxvi16ger* respectively. The old mnemonics are still supported by the
12735 assembler as extended mnemonics. The disassembler generates the new
12736 mnemonics. The name changes occurred in commit:
12738 commit bb98553cad4e017f1851153fa5de91f2cee98fb2
12739 Author: Peter Bergner <bergner@linux.ibm.com>
12740 Date: Sat Oct 8 16:19:51 2022 -0500
12742 PowerPC: Add support for RFC02658 - MMA+ Outer-Product Instructions
12745 * config/tc-ppc.c (md_assemble): Only check for prefix opcodes.
12746 * testsuite/gas/ppc/rfc02658.s: New test.
12747 * testsuite/gas/ppc/rfc02658.d: Likewise.
12748 * testsuite/gas/ppc/ppc.exp: Run it.
12751 * ppc-opc.c (XMSK8, P_GERX4_MASK, P_GERX2_MASK, XX3GERX_MASK): New.
12752 (powerpc_opcodes): Add dmxvi8gerx4pp, dmxvi8gerx4, dmxvf16gerx2pp,
12753 dmxvf16gerx2, dmxvbf16gerx2pp, dmxvf16gerx2np, dmxvbf16gerx2,
12754 dmxvi8gerx4spp, dmxvbf16gerx2np, dmxvf16gerx2pn, dmxvbf16gerx2pn,
12755 dmxvf16gerx2nn, dmxvbf16gerx2nn, pmdmxvi8gerx4pp, pmdmxvi8gerx4,
12756 pmdmxvf16gerx2pp, pmdmxvf16gerx2, pmdmxvbf16gerx2pp, pmdmxvf16gerx2np,
12757 pmdmxvbf16gerx2, pmdmxvi8gerx4spp, pmdmxvbf16gerx2np, pmdmxvf16gerx2pn,
12758 pmdmxvbf16gerx2pn, pmdmxvf16gerx2nn, pmdmxvbf16gerx2nn.
12760 The above commit results in about 224 failures on Power 10 since the
12761 disassembled names do not match the expected names in the test. This
12762 patch updates the expected names in the test to match the values produced
12763 by the disassembler.
12765 This patch updates file gdb.arch/powerpc-power10.exp with the new expected
12766 values to the instructions. The comment giving the name of the instruction
12767 for each binary value in the file gdb.arch/powerpc-power10.c is updated
12768 with the new name. There are no functional changes in file
12769 gdb.arch/powerpc-power10.c.
12771 2022-11-04 Carl Love <cel@us.ibm.com>
12773 Powerpc fix for gdb.base/unwind-on-each-insn.exp
12774 The test disassembles function foo and searches for the line
12775 "End of assembler dump" to determing the last address in the function. The
12776 assumption is the last instruction will be given right before the line
12777 "End of assembler dump". This assumption fails on PowerPC.
12779 The PowerPC disassembly of the function foo looks like:
12780 Dump of assembler code for function foo:
12781 # => 0x00000000100006dc <+0>: std r31,-8(r1)
12782 # 0x00000000100006e0 <+4>: stdu r1,-48(r1)
12783 # 0x00000000100006e4 <+8>: mr r31,r1
12784 # 0x00000000100006e8 <+12>: nop
12785 # 0x00000000100006ec <+16>: addi r1,r31,48
12786 # 0x00000000100006f0 <+20>: ld r31,-8(r1)
12787 # 0x00000000100006f4 <+24>: blr
12788 # 0x00000000100006f8 <+28>: .long 0x0
12789 # 0x00000000100006fc <+32>: .long 0x0
12790 # 0x0000000010000700 <+36>: .long 0x1000180
12791 # End of assembler dump.
12793 The blr instruction is the last instruction in function foo. The lines
12794 with .long following the blr instruction need to be ignored.
12796 This patch adds a new condition to the gdb_test_multiple "disassemble foo"
12797 test to ignore the lines with the .long.
12799 The patch has been tested on PowerPC and Intel X86-64.
12801 2022-11-04 Jan Beulich <jbeulich@suse.com>
12803 x86: adjust recently introduced testcases
12804 The issue addressed by 2c02c72c62d2 ("re: Support Intel AMX-FP16") has
12805 been introduced once again in a number of new tests.
12807 2022-11-04 Nick Clifton <nickc@redhat.com>
12809 Update release documentation with regard to uploading gprofng docs
12811 2022-11-04 Bruno Larsen <blarsen@redhat.com>
12813 gdb/testsuite: add KFAILs to gdb.reverse/step-reverse.exp
12814 Recent changes to gdb.reverse/step-reverse.exp revealed the latent bug
12815 PR record/29745, where we can't skip one funcion forward if we're using
12816 native-gdbserver. This commit just adds kfails to the test.
12818 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29745
12819 Approved-By: Simon Marchi <simon.marchi@efficios.com>
12821 2022-11-04 Andrew Burgess <aburgess@redhat.com>
12823 opcodes/arm: silence compiler warning about uninitialized variable use
12826 commit 6576bffe6cbbb53c5756b2fccd2593ba69b74cdf
12827 Date: Thu Jul 7 13:43:45 2022 +0100
12829 opcodes/arm: add disassembler styling for arm
12831 Some people were seeing their builds failing with complaints about a
12832 possible uninitialized variable usage. I previously fixed an instance
12833 of this issue in this commit:
12835 commit 2df82cd4b459fbc32120e0ad1ce19e26349506fe
12836 Date: Tue Nov 1 10:36:59 2022 +0000
12838 opcodes/arm: silence compiler warning about uninitialized variable use
12840 which did fix the build problems that the sourceware buildbot was
12841 hitting, however, an additional instance of the same problem was
12842 brought to my attention, and that is fixed in this commit.
12844 Where commit 2df82cd4b4 fixed the uninitialized variable problem in
12845 print_mve_unpredictable, this commit fixes the same problem in
12846 print_mve_undefined.
12848 As with the previous commit, I don't believe we could really ever get
12849 an uninitialized variable usage, based on the current usage of the
12850 function, so I have just initialized the reason variable to "??".
12852 2022-11-04 Mike Frysinger <vapier@gentoo.org>
12854 sim: rx: drop unused $arch setting
12855 This is only needed for CGEN ports which RX isn't, so drop it.
12857 2022-11-04 Mike Frysinger <vapier@gentoo.org>
12859 sim: build: remove various obsolete generation dep variables
12860 These manual settings were necessary when we weren't doing automatic
12861 header dependency tracking. That was changed a while ago, and we use
12862 automake now to do it all for us. As a result, many of these vars
12863 aren't even referenced anymore.
12865 Further, some of the source file generation (e.g. .c files, or igen,
12866 or cgen outputs) were moved to the common automake build, and it takes
12867 care of dependency tracking for us with the object files.
12869 2022-11-04 Mike Frysinger <vapier@gentoo.org>
12871 sim: don't hardcode -ldl for SDL support
12872 Since we use AC_SEARCH_LIBS to find dlopen, we don't need to hardcode
12873 -ldl when using SDL ourselves.
12875 2022-11-04 konglin1 <lingling.kong@intel.com>
12877 Support Intel AVX-NE-CONVERT
12880 * NEWS: Support Intel AVX-NE-CONVERT.
12881 * config/tc-i386.c: Add avx_ne_convert.
12882 * doc/c-i386.texi: Document .avx_ne_convert.
12883 * testsuite/gas/i386/i386.exp: Run AVX NE CONVERT tests.
12884 * testsuite/gas/i386/avx-ne-convert-intel.d: New test.
12885 * testsuite/gas/i386/avx-ne-convert.d: Ditto.
12886 * testsuite/gas/i386/avx-ne-convert.s: Ditto.
12887 * testsuite/gas/i386/x86-64-avx-ne-convert-intel.d: Ditto.
12888 * testsuite/gas/i386/x86-64-avx-ne-convert.d: Ditto.
12889 * testsuite/gas/i386/x86-64-avx-ne-convert.s: Ditto.
12893 * i386-dis.c (Mw): New.
12894 (PREFIX_VEX_0F3872): Ditto.
12895 (PREFIX_VEX_0F38B0_W_0): Ditto.
12896 (PREFIX_VEX_0F38B1_W_0): Ditto.
12897 (VEX_W_0F3872_P_1): Ditto.
12898 (VEX_W_0F38B0): Ditto.
12899 (VEX_W_0F38B1): Ditto.
12900 (prefix_table): Add PREFIX_VEX_0F3872, PREFIX_VEX_0F38B0_W_0,
12901 PREFIX_VEX_0F38B1_W_0.
12902 (vex_w_table): Add VEX_W_0F3872_P_1, VEX_W_0F38B0, VEX_W_0F38B1.
12903 * i386-gen.c (cpu_flag_init): Add CPU_AVX_NE_CONVERT_FLGAS and
12904 CPU_ANY_AVX_NE_CONVERT_FLAGS.
12905 (cpu_flags): Add CpuAVX_NE_CONVERT.
12906 * i386-init.h: Regenerated.
12907 * i386-opc.h (CpuAVX_NE CONVERT): New.
12908 (i386_cpu_flags): Add cpuavx_ne_convert.
12909 * i386-opc.tbl: Add Intel AVX-NE-CONVERT instructions.
12910 * i386-tbl.h: Regenerated.
12912 2022-11-04 konglin1 <lingling.kong@intel.com>
12914 i386: Rename <xy> template.
12916 * i386-opc.tbl: Rename <xy> template for VEX insn with x/y suffix to <Vxy>.
12917 Rename <xy> for EVEX insn with x/y suffix to <Exy>.
12919 2022-11-04 Jojo R <rjiejie@linux.alibaba.com>
12921 Support multiple .eh_frame sections
12922 This patch is based on MULTIPLE_FRAME_SECTIONS and EH_FRAME_LINKONCE,
12923 it allows backend to enable this feature and use '--gc-sections' simply.
12925 * gas/dw2gencfi.h (TARGET_MULTIPLE_EH_FRAME_SECTIONS): New.
12926 (MULTIPLE_FRAME_SECTIONS): Add TARGET_MULTIPLE_EH_FRAME_SECTIONS.
12927 * gas/dw2gencfi.c (EH_FRAME_LINKONCE): Add TARGET_MULTIPLE_EH_FRAME_SECTIONS.
12928 (is_now_linkonce_segment): Likewise.
12929 (get_cfi_seg): Create relocation info between .eh_frame.* and .text.* section.
12931 * bfd/elf-bfd.h (elf_backend_can_make_multiple_eh_frame): New.
12932 * bfd/elfxx-target.h (elf_backend_can_make_multiple_eh_frame): Likewise.
12933 * bfd/elflink.c (_bfd_elf_default_action_discarded): Add checking for
12934 elf_backend_can_make_multiple_eh_frame.
12936 2022-11-04 Jojo R <rjiejie@linux.alibaba.com>
12938 gas/doc/internals.texi: fix typo
12939 * gas/doc/internals.texi (md_emit_single_noop_insn):
12940 fix '@var missing closing brace'
12941 * gas/doc/internals.texi (Hash tables):
12942 fix '@menu reference to nonexistent node `Hash tables''
12944 2022-11-04 Mike Frysinger <vapier@gentoo.org>
12946 sim: drop -lm from SIM_EXTRA_LIBS
12947 We have configure tests for this in the top-level configure script
12948 to link this when necessary, so we don't need to explicitly list it
12949 for specific ports.
12951 sim: build: change AC_CHECK_LIB to AC_SEARCH_LIBS
12952 With more C libraries moving functions entirely into the main -lc,
12953 change the AC_CHECK_LIB calls to AC_SEARCH_LIBS so we look in there
12954 first and avoid extra linkage when possible.
12956 sim: build: drop duplicate $(LIBS) usage
12957 COMMON_LIBS is set to $(LIBS), and CONFIG_LIBS is set to that plus
12958 @LIBS@. This leds to the values being used twice. Inline the
12959 CONFIG_LIBS variable without @LIBS@ since it's used only once.
12961 sim: build: switch to bfd & opcodes libtool linker scripts
12962 Now that we use libtool to link, we don't need to duplicate all the
12963 libs that bfd itself uses. This simplifies the configure & Makefile.
12965 sim: build: switch to libtool for linking
12966 The top-level already sets up a libtool script for the host, so use
12967 that when linking rather than invoking CC directly. This will also
12968 happen when we (someday) move the building to pure automake.
12970 2022-11-04 GDB Administrator <gdbadmin@sourceware.org>
12972 Automatic date update in version.in
12974 2022-11-03 Mike Frysinger <vapier@gentoo.org>
12976 sim: testsuite: fix cris stat3 in diff setups
12977 This test uses the test itself as an input to stating regular files.
12978 This gets funky though: when we run check in parallel, the output
12979 object dir is the subdir that matches the .exp file. When we run
12980 with -j1, the output object dir is the sim builddir itself.
12982 The old test would append argv[0] to find the file, while the new
12983 test uses basename on it. Each method works in only one of the
12984 aforementioned build scenarios. Rather than complicate this any
12985 more, switch to a different file that we know will always exist:
12988 2022-11-03 Mike Frysinger <vapier@gentoo.org>
12990 sim: testsuite: fix cris badarch in multi-target builds
12991 This test assumes that /bin/sh will never be a CRIS ELF by way of
12992 assuming that the current bfd cannot load it (since a basic cris
12993 cross-compiler only understands CRIS ELFs). In a multi-target
12994 build though, bfd understands just about every ELF out there, so
12995 we're able to read the /bin/sh format before failing at a diff
12996 point in the cris code.
12998 Let's switch to using / instead since it'll fail for a similar
12999 reason (at least similar enough for what this test is testing).
13001 2022-11-03 Mike Frysinger <vapier@gentoo.org>
13003 sim: cleanup unused SIM_EXTRA_CFLAGS
13004 We want to eventually delete this, so at least drop the empty ones.
13006 sim: m32c/rx: drop useless $(ENDLIST)
13007 This is used to allow for dangling \ in object lists, but these are the
13008 only ports that do it, and it isn't really necessary. Punt it to keep
13009 the various makefiles harmonized.
13011 sim: mips: simplify fpu configure logic
13012 The configure code always defaults to HARD_FLOATING_POINT, so inline
13013 that value and drop redundant target checks as a result.
13015 2022-11-03 Mike Frysinger <vapier@gentoo.org>
13017 sim: erc32: link sis to run program
13018 The erc32 sim does a lot itself, including handling of the CLI. It
13019 used to provide a run-compatible interface in the pre-nrun days, but
13020 it was dropped when the old run interface was punted. Since the old
13021 commit 465fb143c87076b6416a8d0d5dd79bb016060fe3 ("sim: make nrun the
13022 default run program"), the erc32 run & sis programs have been the
13023 same, and erc32 hasn't provide a real run-compatible interface.
13025 Simplify this by linking the two programs via ln/cp instead of running
13026 the linking phase twice to produce the same result. If/when we fix up
13027 the erc32 port to have a proper run interface, it should be easy to
13028 split these back apart into real programs.
13030 Note: the interf.o reference in here is a bit of a misdirect. Since
13031 that object is placed into libsim.a, it's never been linked into the
13032 programs since the linker ignores objects that aren't referenced, and
13033 only gdb uses those symbols.
13035 2022-11-03 Nick Clifton <nickc@redhat.com>
13037 V850 Linker: do not complain about RWX segments.
13039 * configure.tgt (ac_default_ld_warn_rwx_segments): Set to 0 for
13042 2022-11-03 Mike Frysinger <vapier@gentoo.org>
13044 sim: v850: switch to standard (high-level) trace defines
13045 The v850 port uses -DDEBUG to control whether to enable internal tracing.
13046 We already have such options via the common trace framework, and those
13047 can be controlled at build time via configure flags (which the v850 code
13048 currently cannot). So switch it over to WITH_TRACE_ANY_P to simplify the
13049 v850 build code even if it doesn't (yet) respect any other trace options.
13051 sim: ppc: include copyright & license in --version
13052 This makes it match the other sim run programs and GNU tools.
13054 sim: update --version copyright year
13055 Probably should have done this 11 months ago ...
13057 sim: ppc: drop use of DATE & TIME
13058 No other tool does this, sim or otherwise, and it makes the ppc build
13059 non-reproducible. Drop it to simplify & make reproducible.
13061 2022-11-03 Bruno Larsen <blarsen@redhat.com>
13063 gdb: Fix issue with Clang CLI macros
13064 Clang up to version 15 (current) adds macros that were defined in the
13065 command line or by "other means", according to the Dwarf specification,
13066 after the last DW_MACRO_end_file, instead of before the first
13067 DW_MACRO_start_file, as the specification dictates. When GDB reads the
13068 macros after the last file is closed, the macros never end up "in scope"
13069 and so we can't print them. This has been submitted as a bug to Clang
13070 developers (https://github.com/llvm/llvm-project/issues/54506), and PR
13071 macros/29034 was opened for GDB to keep track of this.
13073 Seeing as there is no expected date for it to be fixed, add a workaround
13074 for all current versions of Clang. The workaround detects when
13075 the main file would be closed and if the producer is Clang, and turns
13076 that operation into a noop, so we keep a reference to the current_file
13077 as those macros are read.
13079 A test case was added to confirm the functionality, and the KFAIL for
13080 running gdb.base/macro-source-path when using clang.
13082 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29034
13083 Approved-By: Simon Marchi <simon.marchi@efficios.com>
13085 2022-11-03 Nick Clifton <nickc@redhat.com>
13087 AVR Linker: Allow the start of the data region to be specified on the linker command line. [Fix PR number in ChangeLog entry]
13089 * scripttempl/avr.sc (__DATA_REGION_ORIGIN__): Define. If a value
13090 has not been provided on the command line then use DATA_ORIGIN.
13091 (MEMORY): Use __DATA_REGION_ORIGIN__ as the start of the data region.
13093 AVR Linker: Allow the start of the data region to specified on the command line.
13095 * scripttempl/avr.sc (__DATA_REGION_ORIGIN__): Define. If a value
13096 has not been provided on the command line then use DATA_ORIGIN.
13097 (MEMORY): Use __DATA_REGION_ORIGIN__ as the start of the data region.
13099 2022-11-03 Mike Frysinger <vapier@gentoo.org>
13101 sim: move common flags to default AM_CPPFLAGS
13102 Since all host files we compile use these settings, move them out of
13103 libcommon.a and into the default AM_CPPFLAGS. This has the effect of
13104 dropping the custom per-target automake rules. Currently it saves us
13105 ~150 lines, but since it's about ~8 lines per object, the overhead
13106 will increase quite a bit as we merge more files into a single build.
13108 This also changes the object output names, so we have to tweak the
13109 rules that were pulling in the common objects when linking.
13111 2022-11-03 Mike Frysinger <vapier@gentoo.org>
13113 sim: merge gnulib logic into the top-level
13114 We aren't using this just yet, but we will, so make it available to
13115 building of common sim files.
13117 sim: common: remove unused include paths
13118 A bunch of these paths don't include any headers, and most likely
13119 never will, so there's no real need to keep them. This will let
13120 us harmonize paths with the top-level Makefile more easily, which
13121 will in turn make it easier to move more compile steps there.
13123 2022-11-03 GDB Administrator <gdbadmin@sourceware.org>
13125 Automatic date update in version.in
13127 2022-11-02 Christophe Lyon <christophe.lyon@arm.com>
13129 arm: PR 29739 Fix typo where ';' should not have been replaced with '@'
13130 ';' does not always indicate the start of a comment, and commit
13131 8cb6e17571f3fb66ccd4fa19f881602542cd06fc incorrectly replaced 3
13132 instances of ';' with '@' in expected diagnostics, leading to tests
13135 This patch restores the original ';' as needed in these testcases.
13137 Fixes bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29739
13139 2022-11-02 Mike Frysinger <vapier@gentoo.org>
13141 sim: split CPPFLAGS between build & host
13142 In order to merge more common/ files into the top-level, we need to
13143 add more host flags to CPPFLAGS, and that conflicts with our current
13144 use with build-time tools. So split them apart like we do with all
13145 other build flags to avoid the issue.
13147 sim: h8300: switch to cpu for state
13148 Rather than rely on pulling out the first cpu from the sim state
13149 for cpu state, pass down the active cpu that's already available.
13151 sim: common: change sim_{fetch,store}_register helpers to use void* buffers
13152 When reading/writing arbitrary data to the system's memory, the unsigned
13153 char pointer type doesn't make that much sense. Switch it to void so we
13154 align a bit with standard C library read/write functions, and to avoid
13155 having to sprinkle casts everywhere.
13157 2022-11-02 Jon Turney <jon.turney@dronecode.org.uk>
13159 Fix Cygwin build after 20489cca
13160 Update code under __CYGWIN__ which accesses inferior process information
13161 which is now stored in windows_process_info rather than globals.
13163 2022-11-02 Jon Turney <jon.turney@dronecode.org.uk>
13165 Fix Cygwin build after bcb9251f
13166 Absent _UNICODE being defined (which gdb's Makefile doesn't do),
13167 windows.h will always define STARTUPINFO is as STARTUPINFOA, so this
13168 cast isn't correct when create_process expects a STARTUPINFOW
13169 parameter (i.e. in a Cygwin build).
13171 Instead write this as &info_ex.StartupInfo (which is always of the
13174 2022-11-02 Jan Beulich <jbeulich@suse.com>
13176 x86: drop bogus Tbyte
13177 Prior to commit 1cb0ab18ad24 ("x86/Intel: restrict suffix derivation")
13178 the Tbyte modifier on the FLDT and FSTPT templates was pointless, as
13179 No_ldSuf would have prevented it being accepted. Due to the special
13180 nature of LONG_DOUBLE_MNEM_SUFFIX said commit, however, has led to these
13181 insns being accepted in Intel syntax mode even when "tbyte ptr" was
13182 present. Restore original behavior by dropping Tbyte there. (Note that
13183 these insns in principle should by marked AT&T syntax only, but since
13184 they haven't been so far we probably shouldn't change that.)
13186 x86: simplify expressions in update_imm()
13187 Comparing the sum of the relevant .imm<N> fields against a constant imo
13188 makes more obvious what is actually meant. It allows dropping of two
13189 static variables, with a 3rd drop requiring two more minor adjustments
13190 elsewhere, utilizing that "i" is zeroed first thing in md_assemble().
13191 This also increases the chances of the compiler doing the calculations
13194 2022-11-02 Nelson Chu <nelson@rivosinc.com>
13196 RISC-V: Fixed the missing $x+arch when adding odd paddings for alignment.
13199 .option arch, rv32i
13206 Assembler adds $d for the odd .byte, and then adds $x+arch for the
13207 alignment. Since norelax, riscv_add_odd_padding_symbol will add the
13208 $d and $x for the odd alignment, but accidently remove the $x+arch because
13209 it has the same address as $d. Therefore, we will get the unexpected result
13210 before applying this patch,
13213 .align 2 # odd alignment, $xrv32ic replaced by $d + $x
13215 After this patch, the expected result should be,
13218 .align 2 # odd alignment, $xrv32ic replaced by $d + $xrv32ic
13221 * config/tc-riscv.c (make_mapping_symbol): If we are adding mapping symbol
13222 for odd alignment, then we probably will remove the $x+arch by accidently
13223 when it has the same address of $d. Try to add the removed $x+arch back
13224 after the $d rather than just $x.
13225 (riscv_mapping_state): Updated since parameters of make_mapping_symbol are
13227 (riscv_add_odd_padding_symbol): Likewise.
13228 (riscv_remove_mapping_symbol): Removed and moved the code into the
13229 riscv_check_mapping_symbols.
13230 (riscv_check_mapping_symbols): Updated.
13231 * testsuite/gas/riscv/mapping-dis.d: Updated and added new testcase.
13232 * testsuite/gas/riscv/mapping-symbols.d: Likewise.
13233 * testsuite/gas/riscv/mapping.s: Likewise.
13235 2022-11-02 Hu, Lin1 <lin1.hu@intel.com>
13237 Support Intel MSRLIST
13240 * NEWS: Support Intel MSRLIST.
13241 * config/tc-i386.c: Add msrlist.
13242 * doc/c-i386.texi: Document .msrlist.
13243 * testsuite/gas/i386/i386.exp: Add MSRLIST tests.
13244 * testsuite/gas/i386/msrlist-inval.l: New test.
13245 * testsuite/gas/i386/msrlist-inval.s: Ditto.
13246 * testsuite/gas/i386/x86-64-msrlist-intel.d: Ditto.
13247 * testsuite/gas/i386/x86-64-msrlist.d: Ditto.
13248 * testsuite/gas/i386/x86-64-msrlist.s: Ditto.
13252 * i386-dis.c (X86_64_0F01_REG_0_MOD_3_RM_6_P_1): New.
13253 (X86_64_0F01_REG_0_MOD_3_RM_6_P_3): Ditto.
13254 (prefix_table): New entry for msrlist.
13255 (x86_64_table): Add X86_64_0F01_REG_0_MOD_3_RM_6_P_1
13256 and X86_64_0F01_REG_0_MOD_3_RM_6_P_3.
13257 * i386-gen.c (cpu_flag_init): Add CPU_MSRLIST_FLAGS
13258 and CPU_ANY_MSRLIST_FLAGS.
13259 * i386-init.h: Regenerated.
13260 * i386-opc.h (CpuMSRLIST): New.
13261 (i386_cpu_flags): Add cpumsrlist.
13262 * i386-opc.tbl: Add MSRLIST instructions.
13263 * i386-tbl.h: Regenerated.
13265 2022-11-02 Hu, Lin1 <lin1.hu@intel.com>
13267 Support Intel WRMSRNS
13270 * NEWS: Support Intel WRMSRNS.
13271 * config/tc-i386.c: Add wrmsrns.
13272 * doc/c-i386.texi: Document .wrmsrns.
13273 * testsuite/gas/i386/i386.exp: Add WRMSRNS tests.
13274 * testsuite/gas/i386/wrmsrns-intel.d: New test.
13275 * testsuite/gas/i386/wrmsrns.d: Ditto.
13276 * testsuite/gas/i386/wrmsrns.s: Ditto.
13277 * testsuite/gas/i386/x86-64-wrmsrns-intel.d: Ditto.
13278 * testsuite/gas/i386/x86-64-wrmsrns.d: Ditto.
13282 * i386-dis.c (PREFIX_0F01_REG_0_MOD_3_RM_6): New.
13283 (prefix_table): Add PREFIX_0F01_REG_0_MOD_3_RM_6.
13284 (rm_table): New entry for wrmsrns.
13285 * i386-gen.c (cpu_flag_init): Add CPU_WRMSRNS_FLAGS
13286 and CPU_ANY_WRMSRNS_FLAGS.
13287 (cpu_flags): Add CpuWRMSRNS.
13288 * i386-init.h: Regenerated.
13289 * i386-opc.h (CpuWRMSRNS): New.
13290 (i386_cpu_flags): Add cpuwrmsrns.
13291 * i386-opc.tbl: Add WRMSRNS instructions.
13292 * i386-tbl.h: Regenerated.
13294 2022-11-02 Kong Lingling <lingling.kong@intel.com>
13296 Add handler for more i386_cpu_flags
13299 * config/tc-i386.c (cpu_flags_all_zero): Add new ARRAY_SIZE handle.
13300 (cpu_flags_equal): Ditto.
13301 (cpu_flags_and): Ditto.
13302 (cpu_flags_or): Ditto.
13303 (cpu_flags_and_not): Ditto.
13305 2022-11-02 Haochen Jiang <haochen.jiang@intel.com>
13307 Support Intel CMPccXADD
13310 * NEWS: Support Intel CMPccXADD.
13311 * config/tc-i386.c: Add cmpccxadd.
13312 (build_modrm_byte): Add operations for Vex.VVVV reg
13313 on operand 0 while have memory operand.
13314 * doc/c-i386.texi: Document .cmpccxadd.
13315 * testsuite/gas/i386/i386.exp: Run CMPccXADD tests.
13316 * testsuite/gas/i386/cmpccxadd-inval.s: New test.
13317 * testsuite/gas/i386/cmpccxadd-inval.l: Ditto.
13318 * testsuite/gas/i386/x86-64-cmpccxadd-intel.d: Ditto.
13319 * testsuite/gas/i386/x86-64-cmpccxadd.s: Ditto.
13320 * testsuite/gas/i386/x86-64-cmpccxadd.d: Ditto.
13324 * i386-dis.c (Mdq): New.
13325 (X86_64_VEX_0F38E0): Ditto.
13326 (X86_64_VEX_0F38E1): Ditto.
13327 (X86_64_VEX_0F38E2): Ditto.
13328 (X86_64_VEX_0F38E3): Ditto.
13329 (X86_64_VEX_0F38E4): Ditto.
13330 (X86_64_VEX_0F38E5): Ditto.
13331 (X86_64_VEX_0F38E6): Ditto.
13332 (X86_64_VEX_0F38E7): Ditto.
13333 (X86_64_VEX_0F38E8): Ditto.
13334 (X86_64_VEX_0F38E9): Ditto.
13335 (X86_64_VEX_0F38EA): Ditto.
13336 (X86_64_VEX_0F38EB): Ditto.
13337 (X86_64_VEX_0F38EC): Ditto.
13338 (X86_64_VEX_0F38ED): Ditto.
13339 (X86_64_VEX_0F38EE): Ditto.
13340 (X86_64_VEX_0F38EF): Ditto.
13341 (x86_64_table): Add X86_64_VEX_0F38E0, X86_64_VEX_0F38E1,
13342 X86_64_VEX_0F38E2, X86_64_VEX_0F38E3, X86_64_VEX_0F38E4,
13343 X86_64_VEX_0F38E5, X86_64_VEX_0F38E6, X86_64_VEX_0F38E7,
13344 X86_64_VEX_0F38E8, X86_64_VEX_0F38E9, X86_64_VEX_0F38EA,
13345 X86_64_VEX_0F38EB, X86_64_VEX_0F38EC, X86_64_VEX_0F38ED,
13346 X86_64_VEX_0F38EE, X86_64_VEX_0F38EF.
13347 * i386-gen.c (cpu_flag_init): Add CPU_CMPCCXADD_FLAGS and
13348 CPU_ANY_CMPCCXADD_FLAGS.
13349 (cpu_flags): Add CpuCMPCCXADD.
13350 * i386-init.h: Regenerated.
13351 * i386-opc.h (CpuCMPCCXADD): New.
13352 (i386_cpu_flags): Add cpucmpccxadd. Comment unused for it is actually 0.
13353 * i386-opc.tbl: Add Intel CMPccXADD instructions.
13354 * i386-tbl.h: Regenerated.
13356 2022-11-02 Cui,Lili <lili.cui@intel.com>
13358 Support Intel AVX-VNNI-INT8
13360 * NEWS: Support Intel AVX-VNNI-INT8.
13361 * config/tc-i386.c: Add avx_vnni_int8.
13362 * doc/c-i386.texi: Document avx_vnni_int8.
13363 * testsuite/gas/i386/avx-vnni-int8-intel.d: New file.
13364 * testsuite/gas/i386/avx-vnni-int8.d: Likewise.
13365 * testsuite/gas/i386/avx-vnni-int8.s: Likewise.
13366 * testsuite/gas/i386/x86-64-avx-vnni-int8-intel.d: Likewise.
13367 * testsuite/gas/i386/x86-64-avx-vnni-int8.d: Likewise.
13368 * testsuite/gas/i386/x86-64-avx-vnni-int8.s: Likewise.
13369 * testsuite/gas/i386/i386.exp: Run AVX VNNI INT8 tests.
13372 * i386-dis.c: (PREFIX_VEX_0F3850) New.
13373 (PREFIX_VEX_0F3851): Likewise.
13374 (VEX_W_0F3850_P_0): Likewise.
13375 (VEX_W_0F3850_P_1): Likewise.
13376 (VEX_W_0F3850_P_2): Likewise.
13377 (VEX_W_0F3850_P_3): Likewise.
13378 (VEX_W_0F3851_P_0): Likewise.
13379 (VEX_W_0F3851_P_1): Likewise.
13380 (VEX_W_0F3851_P_2): Likewise.
13381 (VEX_W_0F3851_P_3): Likewise.
13382 (VEX_W_0F3850): Delete.
13383 (VEX_W_0F3851): Likewise.
13384 (prefix_table): Add PREFIX_VEX_0F3850 and PREFIX_VEX_0F3851.
13385 (vex_table): Add PREFIX_VEX_0F3850 and PREFIX_VEX_0F3851,
13386 delete VEX_W_0F3850 and VEX_W_0F3851.
13387 (vex_w_table): Add VEX_W_0F3850_P_0, VEX_W_0F3850_P_1, VEX_W_0F3850_P_2
13388 VEX_W_0F3850_P_3, VEX_W_0F3851_P_0, VEX_W_0F3851_P_1, VEX_W_0F3851_P_2
13389 and VEX_W_0F3851_P_3, delete VEX_W_0F3850 and VEX_W_0F3851.
13390 * i386-gen.c: (cpu_flag_init): Add CPU_AVX_VNNI_INT8_FLAGS
13391 and CPU_ANY_AVX_VNNI_INT8_FLAGS.
13392 (cpu_flags): Add CpuAVX_VNNI_INT8.
13393 * i386-opc.h (CpuAVX_VNNI_INT8): New.
13394 * i386-opc.tbl: Add Intel AVX_VNNI_INT8 instructions.
13395 * i386-init.h: Regenerated.
13396 * i386-tbl.h: Likewise.
13398 2022-11-02 Hongyu Wang <hongyu.wang@intel.com>
13399 Haochen Jiang <haochen.jiang@intel.com>
13401 Support Intel AVX-IFMA
13402 x86: Support Intel AVX-IFMA
13404 Intel AVX IFMA instructions are marked with CpuVEX_PREFIX, which is
13405 cleared by default. Without {vex} pseudo prefix, Intel IFMA instructions
13406 are encoded with EVEX prefix. {vex} pseudo prefix will turn on VEX
13407 encoding for Intel IFMA instructions.
13411 * NEWS: Support Intel AVX-IFMA.
13412 * config/tc-i386.c (cpu_arch): Add avx_ifma.
13413 * doc/c-i386.texi: Document .avx_ifma.
13414 * testsuite/gas/i386/avx-ifma.d: New file.
13415 * testsuite/gas/i386/avx-ifma-intel.d: Likewise.
13416 * testsuite/gas/i386/avx-ifma.s: Likewise.
13417 * testsuite/gas/i386/x86-64-avx-ifma.d: Likewise.
13418 * testsuite/gas/i386/x86-64-avx-ifma-intel.d: Likewise.
13419 * testsuite/gas/i386/x86-64-avx-ifma.s: Likewise.
13420 * testsuite/gas/i386/i386.exp: Run AVX IFMA tests.
13424 * i386-dis.c (PREFIX_VEX_0F38B4): New.
13425 (PREFIX_VEX_0F38B5): Likewise.
13426 (VEX_W_0F38B4_P_2): Likewise.
13427 (VEX_W_0F38B5_P_2): Likewise.
13428 (prefix_table): Add PREFIX_VEX_0F38B4 and PREFIX_VEX_0F38B5.
13429 (vex_table): Add VEX_W_0F38B4_P_2 and VEX_W_0F38B5_P_2.
13430 * i386-dis-evex.h: Fold AVX512IFMA entries to AVX-IFMA.
13431 * i386-gen.c (cpu_flag_init): Clear the CpuAVX_IFMA bit in
13432 CPU_UNKNOWN_FLAGS. Add CPU_AVX_IFMA_FLGAS and
13433 CPU_ANY_AVX_IFMA_FLAGS. Add CpuAVX_IFMA to CPU_AVX2_FLAGS.
13434 (cpu_flags): Add CpuAVX_IFMA.
13435 * i386-opc.h (CpuAVX_IFMA): New.
13436 (i386_cpu_flags): Add cpuavx_ifma.
13437 * i386-opc.tbl: Add Intel AVX IFMA instructions.
13438 * i386-init.h: Regenerated.
13439 * i386-tbl.h: Likewise.
13441 2022-11-02 GDB Administrator <gdbadmin@sourceware.org>
13443 Automatic date update in version.in
13445 2022-11-01 Andrew Burgess <aburgess@redhat.com>
13447 opcodes/arm: don't pass non-string literal to printf like function
13448 The earlier commit:
13450 commit 6576bffe6cbbb53c5756b2fccd2593ba69b74cdf
13451 Date: Thu Jul 7 13:43:45 2022 +0100
13453 opcodes/arm: add disassembler styling for arm
13455 introduced two places where a register name was passed as the format
13456 string to the disassembler's fprintf_styled_func callback. This will
13457 cause a warning from some compilers, like this:
13459 ../../binutils-gdb/opcodes/arm-dis.c: In function ‘print_mve_vld_str_addr’:
13460 ../../binutils-gdb/opcodes/arm-dis.c:6005:3: error: format not a string literal and no format arguments [-Werror=format-security]
13461 6005 | func (stream, dis_style_register, arm_regnames[gpr]);
13464 This commit fixes these by using "%s" as the format string.
13466 2022-11-01 Andrew Burgess <aburgess@redhat.com>
13468 opcodes/arm: silence compiler warning about uninitialized variable use
13469 The earlier commit:
13471 commit 6576bffe6cbbb53c5756b2fccd2593ba69b74cdf
13472 Date: Thu Jul 7 13:43:45 2022 +0100
13474 opcodes/arm: add disassembler styling for arm
13476 was causing a compiler warning about a possible uninitialized variable
13477 usage within opcodes/arm-dis.c.
13479 The problem is in print_mve_unpredictable, and relates to the reason
13480 variable, which is set by a switch table.
13482 Currently the switch table does cover every valid value, though there
13483 is no default case. The variable switched on is passed in as an
13484 argument to the print_mve_unpredictable function.
13486 Looking at how print_mve_unpredictable is used, there is only one use,
13487 the second argument is the one that is used for the switch table,
13488 looking at how this argument is set, I don't believe it is possible
13489 for this argument to take an invalid value.
13491 So, I think the compiler warning is a false positive. As such, my
13492 proposed solution is to initialize the reason variable to the string
13493 "??", this will silence the warning, and the "??" string should never
13494 end up being printed.
13496 2022-11-01 Andrew Burgess <aburgess@redhat.com>
13498 opcodes/arm: add disassembler styling for arm
13499 This commit adds disassembler styling for the ARM architecture.
13501 The ARM disassembler is driven by several instruction tables,
13502 e.g. cde_opcodes, coprocessor_opcodes, neon_opcodes, etc
13504 The type for elements in each table can vary, but they all have one
13505 thing in common, a 'const char *assembler' field. This field
13506 contains a string that describes the assembler syntax of the
13509 Embedded within that assembler syntax are various escape characters,
13510 prefixed with a '%'. Here's an example of a very simple instruction
13511 from the arm_opcodes table:
13515 The '%a' indicates a particular type of operand, the function
13516 print_insn_arm processes the arm_opcodes table, and includes a switch
13517 statement that handles the '%a' operand, and takes care of printing
13518 the correct value for that instruction operand.
13520 It is worth noting that there are many print_* functions, each
13521 function handles a single *_opcodes table, and includes its own switch
13522 statement for operand handling. As a result, every *_opcodes table
13523 uses a different mapping for the operand escape sequences. This means
13524 that '%a' might print an address for one *_opcodes table, but in a
13525 different *_opcodes table '%a' might print a register operand.
13527 Notice as well that in our example above, the instruction mnemonic
13528 'pld' is embedded within the assembler string. Some instructions also
13529 include comments within the assembler string, for example, also from
13530 the arm_opcodes table:
13532 "nop\t\t\t@ (mov r0, r0)"
13534 here, everything after the '@' is a comment that is displayed at the
13535 end of the instruction disassembly.
13537 The next complexity is that the meaning of some escape sequences is
13538 not necessarily fixed. Consider these two examples from arm_opcodes:
13540 "ldrex%c\tr%12-15d, [%16-19R]"
13543 Here, the '%d' escape is used with a bitfield modifier, '%12-15d' in
13544 the first instruction, and '%9-9d' in the second instruction, but,
13545 both of these are the '%d' escape.
13547 However, in the first instruction, the '%d' is used to print a
13548 register number, notice the 'r' immediately before the '%d'. In the
13549 second instruction the '%d' is used to print an immediate, notice the
13550 '#' just before the '%d'.
13552 We have two problems here, first, the '%d' needs to know if it should
13553 use register style or immediate style, and secondly, the 'r' and '#'
13554 characters also need to be styled appropriately.
13556 The final thing we must consider is that some escape codes result in
13557 more than just a single operand being printed, for example, the '%q'
13558 operand as used in arm_opcodes ends up calling arm_decode_shift, which
13559 can print a register name, a shift type, and a shift amount, this
13560 could end up using register, sub-mnemonic, and immediate styles, as
13561 well as the text style for things like ',' between the different
13564 I propose a three layer approach to adding styling:
13566 (1) Basic state machine:
13568 When we start printing an instruction we should maintain the idea
13569 of a 'base_style'. Every character from the assembler string will
13570 be printed using the base_style.
13572 The base_style will start as mnemonic, as each instruction starts
13573 with an instruction mnemonic. When we encounter the first '\t'
13574 character, the base_style will change to text. When we encounter
13575 the first '@' the base_style will change to comment_start.
13577 This simple state machine ensures that for simple instructions the
13578 basic parts, except for the operands themselves, will be printed in
13581 (2) Simple operand styling:
13583 For operands that only have a single meaning, or which expand to
13584 multiple parts, all of which have a consistent meaning, then I
13585 will simply update the operand printing code to print the operand
13586 with the correct style. This will cover a large number of the
13587 operands, and is the most consistent with how styling has been
13588 added to previous architectures.
13590 (3) New styling syntax in assembler strings:
13592 For cases like the '%d' that I describe above, I propose adding a
13593 new extension to the assembler syntax. This extension will allow
13594 me to temporarily change the base_style. Operands like '%d', will
13595 then print using the base_style rather than using a fixed style.
13597 Here are the two examples from above that use '%d', updated with
13598 the new syntax extension:
13600 "ldrex%c\t%{R:r%12-15d%}, [%16-19R]"
13601 "setpan\t%{I:#%9-9d%}"
13603 The syntax has the general form '%{X:....%}' where the 'X'
13604 character changes to indicate a different style. In the first
13605 instruction I use '%{R:...%}' to change base_style to the register
13606 style, and in the second '%{I:...%}' changes base_style to
13609 Notice that the 'r' and '#' characters are included within the new
13610 style group, this ensures that these characters are printed with
13611 the correct style rather than as text.
13613 The function decode_base_style maps from character to style. I've
13614 included a character for each style for completeness, though only
13615 a small number of styles are currently used.
13617 I have updated arm-dis.c to the above scheme, and checked all of the
13618 tests in gas/testsuite/gas/arm/, and the styling looks reasonable.
13620 There are no regressions on the ARM gas/binutils/ld tests that I can
13621 see, so I don't believe I've changed the output layout at all. There
13622 were two binutils tests for which I needed to force the disassembler
13625 I can't guarantee that I've not missed some untested corners of the
13626 disassembler, or that I might have just missed some incorrectly styled
13627 output when reviewing the test results, but I don't believe I've
13628 introduced any changes that could break the disassembler - the worst
13629 should be some aspect is not styled correctly.
13631 2022-11-01 Andrew Burgess <aburgess@redhat.com>
13633 opcodes/arm: use '@' consistently for the comment character
13634 Looking at the ARM disassembler output, every comment seems to start
13635 with a ';' character, so I assumed this was the correct character to
13636 start an assembler comment.
13638 I then spotted a couple of places where there was no ';', but instead,
13639 just a '@' character. I thought that this was a case of a missing
13640 ';', and proposed a patch to add the missing ';' characters.
13642 Turns out I was wrong, '@' is actually the ARM assembler comment
13643 character, while ';' is the statement separator. Thus this:
13647 is two statements, the first is the 'nop' instruction, while the
13648 second contains no instructions, just the '@ comment' comment text.
13654 is a single 'nop' instruction followed by a comment. And finally,
13659 is two statements, the first contains the 'nop' instruction, while the
13660 second contains the instruction 'comment', which obviously isn't
13661 actually an instruction at all.
13663 Why this matters is that, in the next commit, I would like to add
13664 libopcodes syntax styling support for ARM.
13666 The question then is how should the disassembler style the three cases
13669 As '@' is the actual comment start character then clearly the '@' and
13670 anything after it can be styled as a comment. But what about ';' in
13671 the second example? Style as text? Style as a comment?
13673 And the third example is even harder, what about the 'comment' text?
13674 Style as an instruction mnemonic? Style as text? Style as a comment?
13676 I think the only sensible answer is to move the disassembler to use
13677 '@' consistently as its comment character, and remove all the uses of
13680 Then, in the next commit, it's obvious what to do.
13682 There's obviously a *lot* of tests that get updated by this commit,
13683 the only actual code changes are in opcodes/arm-dis.c.
13685 2022-11-01 GDB Administrator <gdbadmin@sourceware.org>
13687 Automatic date update in version.in
13689 2022-10-31 Tom Tromey <tromey@adacore.com>
13691 Add missing TYPE_CODE_* constants to Python
13692 A user noticed that TYPE_CODE_FIXED_POINT was not exported by the gdb
13693 Python layer. This patch fixes the bug, and prevents future
13694 occurences of this type of bug.
13696 2022-10-31 Carl Love <cel@us.ibm.com>
13698 Remove REPARSE condition to force hardware resource checking when updating watchpoints
13699 Currently the resource checking is done if REPARSE is true. The hardware
13700 watchpoint resource checking in update_watchpoint needs to be redone on
13701 each call to function update_watchpoints as the value chain may have
13702 changed. The number of hardware registers needed for a watchpoint can
13703 change if the variable being watched changes. This situation occurs in
13704 this test when watching variable **global_ptr_ptr. Initially when the
13705 watch command is issued, only two addresses need to be watched as
13706 **global_ptr_ptr has not yet been initialized. Once the value of
13707 **global_ptr_ptr is initialized the locations to be tracked increase to
13708 three addresses. However, update_watchpoints is not called again with
13709 REPARSE set to 1 to force the resource checking to be redone. When the
13710 test is run on Power 10, an internal gdb error occurs when the PowerPC
13711 routine tries to setup the three hardware watchpoint address since the hw
13712 only has two hardware watchpoint registers. The error occurs because the
13713 resource checking was not redone in update_watchpoints after
13714 **global_ptr_ptr changed.
13716 The following descibes the situation in detail that occurs on Power 10 with
13717 gdb running on the binary for gdb.base/watchpoint.c.
13721 3 watch *global_ptr
13722 4 next execute source code: buf[0] = 3;
13723 5 next execute source code: global_ptr = buf;
13724 6 next execute source code: buf[0] = 7;
13725 7 delete 2 (delete watch *global_ptr)
13726 8 watch **global_ptr_ptr
13727 9 next execute source code: buf[1] = 5;
13728 10 next global_ptr_ptr = &global_ptr;
13729 11 next buf[0] = 9;
13731 In step 8, the the watch **global_ptr_prt command calls update_watchpoint
13732 in breakpoint.c with REPARSE set to 1. The function update_watchpoint
13733 calls can_use_hardware_watchpoint to see if there are enough
13734 resources available to add the watchpoint since REPARSE is set to 1. At
13735 this point, **global_ptr_ptr has not been initialized so only two addresses
13736 are watched. The val_chain contains the address for **global_ptr_ptr and 0
13737 since **global_ptr_ptr has not been initialized. The update_watchpoint
13738 updates the breakpoint list as follows:
13741 loc 0: b->address = 0x100009c0
13743 loc 1: b->address = 0x7ffff7f838a0
13745 loc 2: b->address = 0x7ffff7b7fc54
13747 loc 3: b->address = 0x7ffff7a5788c
13749 loc 4: b->address = 0x0 <-- location pointed to by global_ptr_ptr
13750 loc 5: b->address = 0x100200b8 <-- global_ptr_ptr watchpoint
13752 loc 6: b->address = 0x7ffff7b7fc54
13754 In step 10, the next command executes the source code
13755 global_ptr_ptr = &global_ptr. This changes the set of locations to be
13756 watched for the watchpoint **global_ptr_prt. The list of addresses for the
13757 breakpoint consist of the address for global_ptr_prt, global_ptr and buf.
13758 The breakpoint list gets updated by update_watchpoint as follows:
13761 loc 0: b->address = 0x100009c0
13763 loc 1: b->address = 0x7ffff7f838a0
13765 loc 2: b->address = 0x7ffff7b7fc54
13767 loc 3: b->address = 0x7ffff7a5788c
13769 loc 4: b->address = 0x10020050 buf
13770 loc 5: b->address = 0x100200b0 watch *global_ptr
13771 loc 6: b->address = 0x100200b8 watch **global_ptr_ptr
13773 loc 7: b->address = 0x7ffff7b7fc54
13776 However, the hardware resource checking was not redone because
13777 update_breakpoint was called with REPARSE equal to 0.
13779 Step 11, execute the third next command. The function
13780 ppc_linux_nat_target::low_prepare_to_resume() attempts a ptrace
13781 call to setup each of the three address for breakpoint 4. The slot
13782 value returned for the third ptrace call is -1 indicating an error
13783 because there are only two hardware watchpoint registers available on
13786 This patch removes just the statement "if (reparse)" in function
13787 update_watchpoint to force the resources to be rechecked on every call to
13788 the function. This ensures that any changes to the val_chain resulting
13789 in needing more resources then available will be caught.
13791 The patch has been tested on Power 8, Power 10 and X86-64. Note the patch
13792 has no effect on Power 9 since hardware watchpoint support is disabled on
13795 2022-10-31 Carl Love <cel@us.ibm.com>
13797 PowerPC, add support for recording pipe2 system call.
13798 Test gdb.reverse/pipe-reverse.exp fails on Power 10 running the fedora 36
13799 distro. The gdb record error message is:
13801 Process record and replay target doesn't support syscall number 317.
13803 System call 317 on PowerPC maps to the pipe2 system call.
13805 This patch adds support for the missing pipe2 system call for PowerPC.
13807 Patch fixes the test failure in gdb.reverse/pipe-reverse.exp.
13809 The patch has been tested on Power 10 with no regression failures.
13811 2022-10-31 Jan Beulich <jbeulich@suse.com>
13813 x86: minor improvements to optimize_imm() (part III)
13814 Earlier tidying still missed an opportunity: There's no need for the
13815 "anyimm" static variable. Instead of using it in the loop to mask
13816 "allowed" (which is necessary to satisfy operand_type_or()'s assertions)
13817 simply use "mask", requiring it to be calculated first. That way the
13818 post-loop masking by "mask" ahead of the operand_type_all_zero() can be
13821 2022-10-31 Mike Frysinger <vapier@gentoo.org>
13823 sim: reg: constify store helper
13824 These functions only read from memory, so mark the pointer as const.
13826 sim: constify various integer readers
13827 These functions only read from memory, so mark the pointer as const.
13829 sim: cgen: constify GETT helpers
13830 These functions only read from memory, so mark the pointer as const.
13832 sim: common: change sim_read & sim_write to use void* buffers
13833 When reading/writing arbitrary data to the system's memory, the unsigned
13834 char pointer type doesn't make that much sense. Switch it to void so we
13835 align a bit with standard C library read/write functions, and to avoid
13836 having to sprinkle casts everywhere.
13838 2022-10-31 H.J. Lu <hjl.tools@gmail.com>
13840 x86: Silence GCC 12 warning on tc-i386.c
13841 Silence GCC 12 warning on tc-i386.c:
13843 gas/config/tc-i386.c: In function ‘md_assemble’:
13844 gas/config/tc-i386.c:5039:16: error: too many arguments for format [-Werror=format-extra-args]
13845 5039 | as_warn (_("only support RIP-relative address"), i.tm.name);
13847 * config/tc-i386.c (md_assemble): Print mnemonic in RIP-relative
13849 * estsuite/gas/i386/x86-64-prefetchi-warn.l: Updated.
13851 2022-10-31 Tom Tromey <tromey@adacore.com>
13853 Use enum for gdbarch's call_dummy_location
13854 This changes gdbarch to use an enum for call_dummy_location, providing
13855 a little more type safety.
13857 Inline initialization of gdbarch members
13858 This changes gdbarch to use the "predefault" to initialize its members
13859 inline. This required changing a couple of the Value instantiations
13860 to avoid a use of "gdbarch" during initialization, but on the whole I
13861 think this is better -- it removes a hidden ordering dependency.
13863 2022-10-31 Tom Tromey <tromey@adacore.com>
13865 Fix regression in pointer-to-member printing
13866 PR c++/29243 points out that "info func" on a certain C++ executable
13867 will cause an infinite loop in gdb.
13869 I tracked this down to a bug introduced by commit 6b5a7bc76 ("Handle
13870 member pointers directly in generic_value_print"). Before this
13871 commit, the C++ code to print a member pointer would wind up calling
13872 value_print_scalar_formatted; but afterward it simply calls
13873 generic_value_print and gets into a loop.
13875 This patch restores the previous behavior and adds a regression test.
13877 2022-10-31 Nick Clifton <nickc@redhat.com>
13879 Updated Romainain translation for the binutils sub-directory and Swedish translations for the ld and opcodes sub-directories.
13881 2022-10-31 Cui, Lili <lili.cui@intel.com>
13883 Support Intel PREFETCHI
13886 * NEWS: Add support for Intel PREFETCHI instruction.
13887 * config/tc-i386.c (load_insn_p): Use prefetch* to fold all prefetches.
13888 (md_assemble): Add warning for illegal input of PREFETCHI.
13889 * doc/c-i386.texi: Document .prefetchi.
13890 * testsuite/gas/i386/i386.exp: Run PREFETCHI tests.
13891 * testsuite/gas/i386/x86-64-lfence-load.d: Add PREFETCHI.
13892 * testsuite/gas/i386/x86-64-lfence-load.s: Likewise.
13893 * testsuite/gas/i386/x86-64-prefetch.d: New test.
13894 * testsuite/gas/i386/x86-64-prefetchi-intel.d: Likewise.
13895 * testsuite/gas/i386/x86-64-prefetchi-inval-register.d: Likewise..
13896 * testsuite/gas/i386/x86-64-prefetchi-inval-register.s: Likewise.
13897 * testsuite/gas/i386/x86-64-prefetchi-warn.l: Likewise.
13898 * testsuite/gas/i386/x86-64-prefetchi-warn.s: Likewise.
13899 * testsuite/gas/i386/x86-64-prefetchi.d: Likewise.
13900 * testsuite/gas/i386/x86-64-prefetchi.s: Likewise.
13904 * i386-dis.c (reg_table): Add MOD_0F18_REG_6 and MOD_0F18_REG_7
13905 (x86_64_table): Add X86_64_0F18_REG_6_MOD_0 and X86_64_0F18_REG_7_MOD_0.
13906 (mod_table): Add MOD_0F18_REG_6 and MOD_0F18_REG_7.
13907 (prefix_table): Add PREFIX_0F18_REG_6_MOD_0_X86_64 and
13908 PREFIX_0F18_REG_7_MOD_0_X86_64.
13909 (PREFETCHI_Fixup): New.
13910 * i386-gen.c (cpu_flag_init): Add CPU_PREFETCHI_FLAGS.
13911 (cpu_flags): Add CpuPREFETCHI.
13912 * i386-opc.h (CpuPREFETCHI): New.
13913 (i386_cpu_flags): Add cpuprefetchi.
13914 * i386-opc.tbl: Add Intel PREFETCHI instructions.
13915 * i386-init.h: Regenerated.
13916 * i386-tbl.h: Likewise.
13918 2022-10-31 Bruno Larsen <blarsen@redhat.com>
13920 gdb/testsuite: fix gdb.cp/converts.exp to run with clang
13921 Clang attempts to minimize the size of the debug-info by not adding
13922 complete information about types that aren't constructed in a given
13923 file. Specifically, this meant that there was minimal information about
13924 class B in the test gdb.cp/converts.exp. To fix this, we just need to
13925 construct any object of type B in that file.
13927 Approved-By: Andrew Burgess <aburgess@redhat.com>
13929 2022-10-31 Bruno Larsen <blarsen@redhat.com>
13931 gdb/testsuite: add XFAIL to gdb.cp/ptype-flags.exp when using clang
13932 When running gdb.cp/ptype-flags.exp using Clang, we get an unexpected
13933 failure when printing the type of a class with an internal typedef. This
13934 happens because Clang doesn't add accessibility information for typedefs
13935 inside classes (see https://github.com/llvm/llvm-project/issues/57608
13936 for more info). To help with Clang testing, an XFAIL was added to this
13939 2022-10-31 Yoshinori Sato <ysato@users.sourceforge.jp>
13941 RX assembler: switch arguments of thw MVTACGU insn.
13943 2022-10-31 Nick Clifton <nickc@redhat.com>
13945 objdump: Add configure time option to enable colored disassembly output by default.
13947 * configure.ac: Add --enable-colored-disassembly.
13948 * objdump.c: Add --disassembler-color=terminal.
13949 * doc/binutils.texi (objdump): Document the new option.
13950 * NEWS: Mention new feature.
13951 * config.in: Regenerate in.
13952 * configure: Regenerate.
13954 2022-10-31 Mark Harmstone <mark@harmstone.com>
13956 ld: Add publics stream to PDB files
13958 ld: Add section header stream to PDB files
13960 ld: Use %E in einfo in pdb.c
13961 Resubmission, taking into account
13962 https://sourceware.org/pipermail/binutils/2022-October/123948.html.
13964 2022-10-31 GDB Administrator <gdbadmin@sourceware.org>
13966 Automatic date update in version.in
13968 2022-10-30 Alan Modra <amodra@gmail.com>
13970 Pool section entries for DWP version 1
13971 Ref: https://gcc.gnu.org/wiki/DebugFissionDWP?action=recall&rev=3
13973 Fuzzers have found a weakness in the code stashing pool section
13974 entries. With random nonsensical values in the index entries (rather
13975 than each index pointing to its own set distinct from other sets),
13976 it's possible to overflow the space allocated, losing the NULL
13977 terminator. Without a terminator, find_section_in_set can run off the
13978 end of the shndx_pool buffer. Fix this by scanning the pool directly.
13981 * dwarf.c (add_shndx_to_cu_tu_entry): Delete range check.
13982 (end_cu_tu_entry): Likewise.
13983 (process_cu_tu_index): Fill shndx_pool by directly scanning
13984 pool, rather than indirectly from index entries.
13986 2022-10-30 GDB Administrator <gdbadmin@sourceware.org>
13988 Automatic date update in version.in
13990 2022-10-29 Maciej W. Rozycki <macro@embecosm.com>
13992 gdb/testsuite: Wrap `param_integer_error' in gdb.guile/scm-parameter.exp
13993 Wrap an overlong line in the definition of `param_integer_error' in
13994 gdb.guile/scm-parameter.exp. No functional change.
13996 2022-10-29 Tsukasa OI <research_trasio@irq.a4lg.com>
13998 sim/sh: Remove redundant function declaration
13999 Clang generates a warning if there is a function declaration/definition
14000 with zero arguments. Such declarations/definitions without a prototype (an
14001 argument list) are deprecated forms of indefinite arguments
14002 ("-Wdeprecated-non-prototype"). On the default configuration, it causes a
14003 build failure (unless "--disable-werror" is specified).
14005 But there is another issue. This function declaration in sim/sh/interp.c
14006 is completely redundant. This commit just removes that declaration.
14008 2022-10-29 Tsukasa OI <research_trasio@irq.a4lg.com>
14010 sim/m32r: Initialize "list" variable
14011 The variable "list" is only initialized when arg1 > 0 and when arg1 == 0,
14012 an uninitialized value is passed to translate_endian_h2t function.
14014 Although this behavior is harmless, this commit adds initialization to avoid
14015 a GCC warning ("-Wmaybe-uninitialized").
14017 2022-10-29 Tsukasa OI <research_trasio@irq.a4lg.com>
14019 sim/erc32: Use int32_t as IRQ callback argument
14020 Clang generates a warning if an argument is passed to a function without
14021 prototype (zero arguments, even without (void)). Such calls are deprecated
14022 forms of indefinite arguments passing ("-Wdeprecated-non-prototype").
14023 On the default configuration, it (somehow) doesn't cause a build failure but
14024 a warning is generated.
14026 But because the cause is the same as the issue the author fixed in
14027 "sim/erc32: Use int32_t as event callback argument", it would be better to
14028 fix it now to prevent problems in the future.
14030 To fix the issue, this commit makes struct irqcall to use int32_t as a
14031 callback (callback) argument of an IRQ.
14033 2022-10-29 Tsukasa OI <research_trasio@irq.a4lg.com>
14035 sim/erc32: Use int32_t as event callback argument
14036 Clang generates a warning if an argument is passed to a function without
14037 prototype (zero arguments, even without (void)). Such calls are deprecated
14038 forms of indefinite arguments passing ("-Wdeprecated-non-prototype").
14039 On the default configuration, it causes a build failure (unless
14040 "--disable-werror" is specified).
14042 To fix that, this commit makes struct evcell to use int32_t as a callback
14043 (cfunc) argument of an event. int32_t is chosen because "event" function
14044 accepts "int32_t arg".
14046 2022-10-29 Tsukasa OI <research_trasio@irq.a4lg.com>
14048 sim/erc32: Insert void parameter
14049 Clang generates a warning if there is a function declaration/definition
14050 with zero arguments. Such declarations/definitions without a prototype (an
14051 argument list) are deprecated forms of indefinite arguments
14052 ("-Wdeprecated-non-prototype"). On the default configuration, it causes a
14053 build failure (unless "--disable-werror" is specified).
14055 This commit replaces () with (void) to avoid this warning.
14057 2022-10-29 Tom de Vries <tdevries@suse.de>
14059 [gdb/testsuite] Use ssh -t in remote-*.exp
14060 When running test-case gdb.server/multi-ui-errors.exp on target board
14061 remote-gdbserver-on-localhost.exp, I run into:
14063 (gdb) PASS: gdb.server/multi-ui-errors.exp: connect to gdbserver
14066 PASS: gdb.server/multi-ui-errors.exp: continue - extra UI
14067 Remote debugging from host ::1, port 35466^M
14068 FAIL: gdb.server/multi-ui-errors.exp: ensure inferior is running
14071 The problem is that the target board uses ssh -T, which fails to guarantee
14072 that output from the inferior will be available.
14074 Fix this by copying proc ${board}_spawn from local-remote-host.exp, which
14075 ensures using ssh -t. [ It would be nice to define an ssh base board to
14076 get rid of the copies, but I'm not addressing that in this commit. ]
14078 Likewise for target board remote-stdio-gdbserver.exp.
14080 Tested on x86_64-linux.
14082 2022-10-29 Tom de Vries <tdevries@suse.de>
14084 [gdb/testsuite] Fix gdb.server/multi-ui-errors.exp with local-remote-host-notty
14085 With test-case gdb.server/multi-ui-errors.exp and host board
14086 local-remote-host-notty, I run into:
14088 (gdb) PASS: gdb.server/multi-ui-errors.exp: interact with GDB's main UI
14089 Executing on target: kill -9 29666 (timeout = 300)
14090 builtin_spawn -ignore SIGHUP kill -9 29666^M
14092 Remote connection closed^M
14093 (gdb) (gdb) FAIL: gdb.server/multi-ui-errors.exp: \
14094 main UI, prompt after gdbserver dies (timeout)
14097 In contrast, with local-remote-host (so, everything the same but editing off):
14099 (gdb) PASS: gdb.server/multi-ui-errors.exp: interact with GDB's main UI
14100 Executing on target: kill -9 31245 (timeout = 300)
14101 builtin_spawn -ignore SIGHUP kill -9 31245^M
14102 Remote connection closed^M
14104 (gdb) PASS: gdb.server/multi-ui-errors.exp: main UI, prompt after gdbserver dies
14107 The test-case issues a kill, which results in a "Remote connection closed"
14108 message and a prompt.
14110 The problem is that the prompt is not consumed, so the subsequent echo may be
14111 issued before that prompt, which causes a mismatch when matching the result
14114 Fix this by consuming the "Remote connection closed" message and prompt.
14116 Tested on x86_64-linux.
14118 2022-10-29 Tom de Vries <tdevries@suse.de>
14120 [gdb/testsuite] Consume output asap in gdb.server/multi-ui-errors.exp
14121 With test-case gdb.server/multi-ui-errors.exp we see:
14123 (gdb) PASS: multi-ui-errors.exp: main UI, prompt after gdbserver dies
14127 (gdb) PASS: multi-ui-errors.exp: extra UI, prompt after gdbserver dies
14130 The continue is issued earlier in the test-case, but the output has not been
14131 consumed, which makes it show up much later.
14133 Consume the continue output asap, to make it clear when the continue is issued:
14135 (gdb) PASS: gdb.server/multi-ui-errors.exp: connect to gdbserver
14138 PASS: gdb.server/multi-ui-errors.exp: continue - extra UI
14141 Tested on x86_64-linux.
14143 2022-10-29 Tom de Vries <tdevries@suse.de>
14145 [gdb/testsuite] Remove REMOTE_PORTNUM in remote-stdio-gdbserver.exp
14146 The usage for board remote-stdio-gdbserver.exp is advertised as:
14148 # bash$ make check RUNTESTFLAGS="--target_board=remote-stdio-gdbserver \
14149 # REMOTE_USERNAME=... REMOTE_HOSTNAME=... REMOTE_PORTNUM=... \
14150 # [REMOTE_TMPDIR=${remote_dir}] [GDBSERVER=${remote_gdbserver}]"
14152 but when adding REMOTE_PORTNUM=22, I run into:
14154 Running stop-reply-no-thread-multi.exp ...
14155 ERROR: tcl error sourcing stop-reply-no-thread-multi.exp.
14156 ERROR: couldn't execute "/usr/bin/ssh -p22": no such file or directory
14158 "builtin_spawn {/usr/bin/ssh -p22} -l vries localhost {/usr/bin/gdbserver \
14159 --once localhost:2346 \
14160 /home/vries/gdb_versions/devel/build/gdb/testsuite/outp..."
14163 Fix this by simply removing REMOTE_PORTNUM.
14165 Tested on x86_64-linux.
14167 2022-10-29 Tsukasa OI <research_trasio@irq.a4lg.com>
14169 sim, sim/{m32c,ppc,rl78}: Use getopt_long
14170 Because of a Libiberty hack, getopt on GNU libc (2.25 or earlier) is
14171 currently unusable on sim, causing a regression on CentOS 7.
14173 This is caused as follows:
14175 1. If HAVE_DECL_GETOPT is defined (getopt declaration with known prototype
14176 is detected while configuration), a declaration of getopt in
14177 "include/getopt.h" is suppressed.
14178 The author started to define HAVE_DECL_GETOPT in sim with the commit
14179 340aa4f6872c ("sim: Check known getopt definition existence").
14180 2. GNU libc (2.25 or earlier)'s <unistd.h> includes <getopt.h> with a
14181 special purpose macro defined to declare only getopt function but due
14182 to include path (not tested while configuration), it causes <unistd.h>
14183 to include Libiberty's "include/getopt.h".
14184 3. If both 1. and 2. are satisfied, despite that <unistd.h> tries to
14185 declare getopt by including <getopt.h>, "include/getopt.h" does not do
14186 so, causing getopt function undeclared.
14188 Getting rid of "include/getopt.h" (e.g. renaming this header file) is the
14189 best solution to avoid hacking but as a short-term solution, this commit
14190 replaces getopt with getopt_long under sim/.
14192 2022-10-29 Alan Modra <amodra@gmail.com>
14194 pef: sanity check before malloc
14195 And do the sanity check in a way that can't overflow.
14197 * pef.c (bfd_pef_parse_function_stubs): Sanity check header
14198 imported_library_count and total_imported_symbol_count before
14201 2022-10-29 Alan Modra <amodra@gmail.com>
14203 Fix small objcopy memory leak
14204 * objcopy.c (copy_archive): Free l->name.
14206 2022-10-29 Alan Modra <amodra@gmail.com>
14208 NULL dereference read in som_write_object_contents
14209 objcopy copy_object may omit the call to bfd_copy_private_bfd_data for
14210 various conditions deemed non-fatal, in which case obj_som_exec_data
14211 will be NULL for the output file.
14213 * som.c (som_finish_writing): Don't dereference NULL
14216 2022-10-29 Nelson Chu <nelson@rivosinc.com>
14218 RISC-V: Always generate mapping symbols at the start of the sections.
14219 Before figuring out the suppress rule of mapping symbol with architecture
14220 (changed back to $x), always generate them at the start of the sections.
14223 * config/tc-riscv.c (need_arch_map_symbol): Removed.
14224 (riscv_mapping_state): Updated.
14225 (riscv_check_mapping_symbols): Updated.
14226 * testsuite/gas/riscv/mapping-non-arch.d: Removed.
14227 * testsuite/gas/riscv/mapping-non-arch.s: Likewise.
14229 2022-10-29 GDB Administrator <gdbadmin@sourceware.org>
14231 Automatic date update in version.in
14233 2022-10-28 Palmer Dabbelt <palmer@rivosinc.com>
14235 gas: NEWS: Note support for RISC-V Zawrs
14236 This has been supported since eb668e50036 ("RISC-V: Add Zawrs ISA
14237 extension support").
14239 gas: NEWS: Add a missing newline
14241 2022-10-28 Tom Tromey <tom@tromey.com>
14243 Convert compunit_language to a method
14244 This changes compunit_language to be a method on compunit_symtab.
14246 Approved-By: Simon Marchi <simon.marchi@efficios.com>
14248 2022-10-28 Tsukasa OI <research_trasio@irq.a4lg.com>
14250 RISC-V: Improve "bits undefined" diagnostics
14251 This commit improves internal error message
14252 "internal: bad RISC-V opcode (bits 0x%lx undefined): %s %s"
14253 to display actual unused bits (excluding non-instruction bits).
14257 * config/tc-riscv.c (validate_riscv_insn): Exclude non-
14258 instruction bits from displaying internal diagnostics.
14259 Change error message slightly.
14261 2022-10-28 Tsukasa OI <research_trasio@irq.a4lg.com>
14263 RISC-V: Fallback for instructions longer than 64b
14264 We don't support instructions longer than 64-bits yet. Still, we can
14265 modify validate_riscv_insn function to prevent unexpected behavior by
14266 limiting the "length" of an instruction to 64-bit (or less).
14270 * config/tc-riscv.c (validate_riscv_insn): Fix function
14271 description comment based on current spec. Limit instruction
14272 length up to 64-bit for now. Make sure that required_bits does
14273 not corrupt even if unsigned long long is longer than 64-bit.
14275 2022-10-28 Jan Beulich <jbeulich@suse.com>
14277 RISC-V/gas: fix build with certain gcc versions
14278 Some versions of gcc warn by default about shadowed outer-scope
14279 declarations. This affects frag_align_code, which is declared in
14280 frags.h. Rename the offending function parameter. While there also
14281 switch to using true/false at the function call sites.
14283 2022-10-28 Tsukasa OI <research_trasio@irq.a4lg.com>
14285 RISC-V: Fix build failure for -Werror=maybe-uninitialized
14286 Commit 40f1a1a4564b ("RISC-V: Output mapping symbols with ISA string.")
14287 caused a build failure on GCC 12 as follows:
14289 make[3]: Entering directory '$(builddir)/gas'
14290 CC config/tc-riscv.o
14291 In file included from $(srcdir)/gas/config/tc-riscv.c:23:
14292 $(srcdir)/gas/as.h: In function ‘make_mapping_symbol’:
14293 $(srcdir)/gas/as.h:123:15: error: ‘buff’ may be used uninitialized [-Werror=maybe-uninitialized]
14294 123 | #define xfree free
14296 $(srcdir)/gas/config/tc-riscv.c:487:9: note: ‘buff’ was declared here
14299 cc1: all warnings being treated as errors
14300 make[3]: *** [Makefile:1425: config/tc-riscv.o] Error 1
14302 This is caused by a false positive of "maybe uninitialized" variable
14303 detection (-Wmaybe-uninitialized). To avoid this error, this commit
14304 initializes the local variable buff to NULL first in all cases.
14308 * config/tc-riscv.c (make_mapping_symbol): Initialize variable
14309 buff with NULL to avoid build failure caused by a GCC's false
14310 positive of maybe uninitialized variable detection.
14312 2022-10-28 Markus Metzger <markus.t.metzger@intel.com>
14314 gdb, btrace: fix family and model computation
14315 In gdb/nat/linux-btrace.c:btrace_this_cpu() we initialize the cpu
14316 structure given to the libipt btrace decoder.
14318 We only consider the extended model field for family 0x6 and forget about
14319 family 0xf and we don't consider the extended family field. Fix it.
14321 2022-10-28 Tsukasa OI <research_trasio@irq.a4lg.com>
14323 include: Define macro to ignore -Wdeprecated-declarations on GCC
14324 "-Wdeprecated-declarations" warning option can be helpful to track
14325 deprecated function delarations but sometimes we need to disable this
14326 warning for a good reason.
14328 DIAGNOSTIC_IGNORE_DEPRECATED_DECLARATIONS is an existing macro but only
14329 defined on Clang. Since "-Wdeprecated-declarations" is also available on
14330 GCC (>= 3.4.0), this commit adds equivalent definition as Clang.
14332 __GNUC__ and __GNUC_MINOR__ are not checked because this header file seems
14333 to assume GCC >= 4.6 (with "GCC diagnostic push/pop").
14337 * diagnostics.h (DIAGNOSTIC_IGNORE_DEPRECATED_DECLARATIONS):
14338 Define also on GCC.
14340 2022-10-28 Nelson Chu <nelson@rivosinc.com>
14342 RISC-V: Output mapping symbols with ISA string.
14343 RISC-V Psabi pr196,
14344 https://github.com/riscv-non-isa/riscv-elf-psabi-doc/pull/196
14347 * elfxx-riscv.c (riscv_release_subset_list): Free arch_str if needed.
14348 (riscv_copy_subset_list): Copy arch_str as well.
14349 * elfxx-riscv.h (riscv_subset_list_t): Store arch_str for each subset list.
14351 * config/tc-riscv.c (riscv_reset_subsets_list_arch_str): Update the
14352 architecture string in the subset_list.
14353 (riscv_set_arch): Call riscv_reset_subsets_list_arch_str after parsing new
14354 architecture string.
14355 (s_riscv_option): Likewise.
14356 (need_arch_map_symbol): New boolean, used to indicate if .option
14357 directives do affect instructions.
14358 (make_mapping_symbol): New boolean parameter reset_seg_arch_str. Need to
14359 generate $x+arch for MAP_INSN, and then store it into tc_segment_info_data
14360 if reset_seg_arch_str is true.
14361 (riscv_mapping_state): Decide if we need to add $x+arch for MAP_INSN. For
14362 now, only add $x+arch if the architecture strings in subset list and segment
14363 are different. Besides, always add $x+arch at the start of section, and do
14364 not add $x+arch for code alignment, since rvc for alignment can be judged
14365 from addend of R_RISCV_ALIGN.
14366 (riscv_remove_mapping_symbol): If current and previous mapping symbol have
14367 same value, then remove the current $x only if the previous is $x+arch;
14368 Otherwise, always remove previous.
14369 (riscv_add_odd_padding_symbol): Updated.
14370 (riscv_check_mapping_symbols): Don't need to add any $x+arch if
14371 need_arch_map_symbol is false, so changed them to $x.
14372 (riscv_frag_align_code): Updated since riscv_mapping_state is changed.
14373 (riscv_init_frag): Likewise.
14374 (s_riscv_insn): Likewise.
14375 (riscv_elf_final_processing): Call riscv_release_subset_list to release
14376 subset_list of riscv_rps_as, rather than only release arch_str in the
14377 riscv_write_out_attrs.
14378 (riscv_write_out_attrs): No need to call riscv_arch_str, just get arch_str
14379 from subset_list of riscv_rps_as.
14380 * config/tc-riscv.h (riscv_segment_info_type): Record current $x+arch mapping
14381 symbol of each segment.
14382 * testsuite/gas/riscv/mapping-0*: Merged and replaced by mapping.s.
14383 * testsuite/gas/riscv/mapping.s: New testcase, to test most of the cases in
14385 * testsuite/gas/riscv/mapping-symbols.d: Likewise.
14386 * testsuite/gas/riscv/mapping-dis.d: Likewise.
14387 * testsuite/gas/riscv/mapping-non-arch.s: New testcase for the case that
14388 does need any $x+arch.
14389 * testsuite/gas/riscv/mapping-non-arch.d: Likewise.
14390 * testsuite/gas/riscv/option-arch-01a.d: Updated.
14392 * riscv-dis.c (riscv_disassemble_insn): Set riscv_fpr_names back to
14393 riscv_fpr_names_abi or riscv_fpr_names_numeric when zfinx is disabled
14394 for some specfic code region.
14395 (riscv_get_map_state): Recognized mapping symbols $x+arch, and then reset
14396 the architecture string once the ISA is different.
14398 2022-10-28 Lifang Xia <lifang_xia@linux.alibaba.com>
14400 binutils: Update my e-mail and Yunhai's e-mail
14402 * MAINTAINERS(C-SKY): update e-mails of Lifang & Yunhai.
14404 2022-10-28 Peter Bergner <bergner@linux.ibm.com>
14406 PowerPC: Add support for RFC02658 - MMA+ Outer-Product Instructions
14408 * config/tc-ppc.c (md_assemble): Only check for prefix opcodes.
14409 * testsuite/gas/ppc/rfc02658.s: New test.
14410 * testsuite/gas/ppc/rfc02658.d: Likewise.
14411 * testsuite/gas/ppc/ppc.exp: Run it.
14414 * ppc-opc.c (XMSK8, P_GERX4_MASK, P_GERX2_MASK, XX3GERX_MASK): New.
14415 (powerpc_opcodes): Add dmxvi8gerx4pp, dmxvi8gerx4, dmxvf16gerx2pp,
14416 dmxvf16gerx2, dmxvbf16gerx2pp, dmxvf16gerx2np, dmxvbf16gerx2,
14417 dmxvi8gerx4spp, dmxvbf16gerx2np, dmxvf16gerx2pn, dmxvbf16gerx2pn,
14418 dmxvf16gerx2nn, dmxvbf16gerx2nn, pmdmxvi8gerx4pp, pmdmxvi8gerx4,
14419 pmdmxvf16gerx2pp, pmdmxvf16gerx2, pmdmxvbf16gerx2pp, pmdmxvf16gerx2np,
14420 pmdmxvbf16gerx2, pmdmxvi8gerx4spp, pmdmxvbf16gerx2np, pmdmxvf16gerx2pn,
14421 pmdmxvbf16gerx2pn, pmdmxvf16gerx2nn, pmdmxvbf16gerx2nn.
14423 2022-10-28 Peter Bergner <bergner@linux.ibm.com>
14425 PowerPC: Add support for RFC02653 - Dense Math Facility
14427 * config/tc-ppc.c (pre_defined_registers): Add dense math registers.
14428 (md_assemble): Check dmr specified in correct operand.
14429 * testsuite/gas/ppc/outerprod.s <dmsetaccz, dmxvbf16ger2,
14430 dmxvbf16ger2nn, dmxvbf16ger2np, dmxvbf16ger2pn, dmxvbf16ger2pp,
14431 dmxvf16ger2, dmxvf16ger2nn, dmxvf16ger2np, dmxvf16ger2pn, dmxvf16ger2pp,
14432 dmxvf32ger, dmxvf32gernn, dmxvf32gernp, dmxvf32gerpn, dmxvf32gerpp,
14433 dmxvf64ger, dmxvf64gernn, dmxvf64gernp, dmxvf64gerpn, dmxvf64gerpp,
14434 dmxvi16ger2, dmxvi16ger2pp, dmxvi16ger2s, dmxvi16ger2spp, dmxvi4ger8,
14435 dmxvi4ger8pp, dmxvi8ger4, dmxvi8ger4pp, dmxvi8ger4spp, dmxxmfacc,
14436 dmxxmtacc, pmdmxvbf16ger2, pmdmxvbf16ger2nn, pmdmxvbf16ger2np,
14437 pmdmxvbf16ger2pn, pmdmxvbf16ger2pp, pmdmxvf16ger2, pmdmxvf16ger2nn,
14438 pmdmxvf16ger2np, pmdmxvf16ger2pn, pmdmxvf16ger2pp, pmdmxvf32ger,
14439 pmdmxvf32gernn, pmdmxvf32gernp, pmdmxvf32gerpn, pmdmxvf32gerpp,
14440 pmdmxvf64ger, pmdmxvf64gernn, pmdmxvf64gernp, pmdmxvf64gerpn,
14441 pmdmxvf64gerpp, pmdmxvi16ger2, pmdmxvi16ger2pp, pmdmxvi16ger2s,
14442 pmdmxvi16ger2spp, pmdmxvi4ger8, pmdmxvi4ger8pp, pmdmxvi8ger4,
14443 pmdmxvi8ger4pp, pmdmxvi8ger4spp>: Add new tests.
14444 * testsuite/gas/ppc/outerprod.d: Likewise.
14445 * testsuite/gas/ppc/rfc02653.s: New test.
14446 * testsuite/gas/ppc/rfc02653.d: Likewise.
14447 * testsuite/gas/ppc/ppc.exp: Run it.
14450 * opcode/ppc.h (PPC_OPERAND_DMR): Define. Renumber following
14451 PPC_OPERAND defines.
14454 * ppc-dis.c (print_insn_powerpc): Prepend 'dm' when printing DMR regs.
14455 * ppc-opc.c (insert_p2, (extract_p2, (insert_xa5, (extract_xa5,
14456 insert_xb5, (extract_xb5): New functions.
14457 (insert_xa6a, extract_xa6a, insert_xb6a, extract_xb6a): Disallow
14458 operand overlap only on Power10.
14459 (DMR, DMRAB, P1, P2, XA5p, XB5p, XDMR_MASK, XDMRDMR_MASK, XX2ACC_MASK,
14460 XX2DMR_MASK, XX3DMR_MASK): New defines.
14461 (powerpc_opcodes): Add dmmr, dmsetaccz, dmsetdmrz, dmxor, dmxvbf16ger2,
14462 dmxvbf16ger2nn, dmxvbf16ger2np, dmxvbf16ger2pn, dmxvbf16ger2pp,
14463 dmxvf16ger2, dmxvf16ger2nn, dmxvf16ger2np, dmxvf16ger2pn, dmxvf16ger2pp,
14464 dmxvf32ger, dmxvf32gernn, dmxvf32gernp, dmxvf32gerpn, dmxvf32gerpp,
14465 dmxvf64ger, dmxvf64gernn, dmxvf64gernp, dmxvf64gerpn, dmxvf64gerpp,
14466 dmxvi16ger2, dmxvi16ger2pp, dmxvi16ger2s, dmxvi16ger2spp, dmxvi4ger8,
14467 dmxvi4ger8pp, dmxvi8ger4, dmxvi8ger4pp, dmxvi8ger4spp, dmxxextfdmr256,
14468 dmxxextfdmr512, dmxxinstdmr256, dmxxinstdmr512, dmxxmfacc, dmxxmtacc,
14469 pmdmxvbf16ger2, pmdmxvbf16ger2nn, pmdmxvbf16ger2np, pmdmxvbf16ger2pn,
14470 pmdmxvbf16ger2pp, pmdmxvf16ger2, pmdmxvf16ger2nn, pmdmxvf16ger2np,
14471 pmdmxvf16ger2pn, pmdmxvf16ger2pp, pmdmxvf32ger, pmdmxvf32gernn,
14472 pmdmxvf32gernp, pmdmxvf32gerpn, pmdmxvf32gerpp, pmdmxvf64ger,
14473 pmdmxvf64gernn, pmdmxvf64gernp, pmdmxvf64gerpn, pmdmxvf64gerpp,
14474 pmdmxvi16ger2, pmdmxvi16ger2pp, pmdmxvi16ger2s, pmdmxvi16ger2spp,
14475 pmdmxvi4ger8, pmdmxvi4ger8pp, pmdmxvi8ger4, pmdmxvi8ger4pp,
14478 2022-10-28 GDB Administrator <gdbadmin@sourceware.org>
14480 Automatic date update in version.in
14482 2022-10-27 Andrew Burgess <aburgess@redhat.com>
14484 sim/cgen: initialize variable at creation in engine_run_n
14485 Zero initialize engine_fns entirely at creation, then override those
14486 fields we intend to use, rather than zero just initializing the unused
14489 There should be no user visible changes after this commit.
14491 2022-10-27 Tom de Vries <tdevries@suse.de>
14493 [gdb/testsuite] Remove address from test names
14494 I noticed an address in a test name:
14496 PASS: gdb.base/eh_return.exp: gdb_breakpoint: \
14497 set breakpoint at *0x000000000040071b
14500 Stabilize the test name by using "set breakpoint on address" instead.
14502 Likewise in two other test-cases.
14504 Tested on x86_64-linux.
14506 2022-10-27 Tom de Vries <tdevries@suse.de>
14508 [gdb/testsuite] Disable styling in host board local-remote-host-native.exp
14509 Propagate fix from commit 17c68d98f74 ("[gdb/testsuite] Disable styling in host
14510 board local-remote-host.exp") to local-remote-host-native.exp.
14512 Tested on x86_64-linux.
14514 2022-10-27 Tom de Vries <tdevries@suse.de>
14516 [gdb/testsuite] Fix silent timeouts in gdb.mi/mi-exec-run.exp with remote host
14517 I noticed that running test-case gdb.mi/mi-exec-run.exp with host board
14518 local-remote-host.exp takes about 44 seconds.
14520 I found two silent timeouts responsible for this.
14522 The first is in mi_gdb_exit, where we have:
14524 if { [is_remote host] && [board_info host exists fileid] } {
14525 send_gdb "999-gdb-exit\n"
14531 -re "Undefined command.*$gdb_prompt $" {
14535 -re "DOSEXIT code" { }
14539 so in gdb.log we see:
14543 =thread-exited,id="1",group-id="i1"^M
14544 =thread-group-exited,id="i1"^M
14546 after which expect just waits for the timeout.
14548 Fix this by adding a gdb_expect clause to parse the exit:
14550 -re "\r\n999\\^exit\r\n" { }
14553 Note that we're not parsing the thread-exited/thread-group-exited messages, because
14554 they may not be present:
14557 =thread-group-added,id="i1"
14564 After fixing that, we have:
14568 PASS: gdb.mi/mi-exec-run.exp: inferior-tty=separate: mi=separate: \
14569 force-fail=1: run failure detected
14574 What seems to be happening is that default_gdb_exit sends a cli interpreter
14575 quit command to an mi interpreter, after which again expect just waits for the
14578 Fix this by adding mi_gdb_exit to the end of the test-case, as in many other
14579 gdb.mi/*.exp test-cases.
14581 After these two fixes, the test-case takes about 4 seconds.
14583 Tested on x86_64-linux.
14585 2022-10-27 Tom de Vries <tdevries@suse.de>
14587 [gdb/testsuite] Use remote_exec chmod instead of remote_spawn
14588 I build gdb using -O2, and ran the testsuite using taskset -c 0, and ran into:
14590 (gdb) PASS: gdb.server/connect-with-no-symbol-file.exp: sysroot=: \
14591 action=delete: setup: adjust sysroot
14592 builtin_spawn gdbserver --once localhost:2385 /connect-with-no-symbol-file^M
14593 /bin/bash: connect-with-no-symbol-file: Permission denied^M
14594 /bin/bash: line 0: exec: connect-with-no-symbol-file: cannot execute: \
14595 Permission denied^M
14596 During startup program exited with code 126.^M
14598 target remote localhost:2385^M
14599 `connect-with-no-symbol-file' has disappeared; keeping its symbols.^M
14600 localhost:2385: Connection timed out.^M
14601 (gdb) FAIL: gdb.server/connect-with-no-symbol-file.exp: sysroot=: \
14602 action=delete: connection to GDBserver succeeded
14605 The expected series of events is (skipping disconnect and detach as I don't
14606 think they're relevant to the problem):
14607 - enter scenario "permission"
14608 - cp $exec.bak $exec
14609 - gdbserver start with $exec
14611 - connect to gdbserver
14612 - enter scenario "delete"
14613 - cp $exec.bak $exec
14614 - gdbserver start with $exec
14616 - connect to gdbserver
14618 The problem is that the chmod is executed using remote_spawn:
14620 } elseif { $action == "permission" } {
14621 remote_spawn target "chmod 000 $target_exec"
14624 without waiting on the resulting spawn id, so we're not sure when the
14625 chmod will have effect.
14627 The FAIL we're seeing above is explained by the chmod having effect during the
14628 delete scenario, after the "cp $exec.bak $exec" and before the "gdbserver
14631 Fix this by using remote_exec instead.
14633 Likewise, fix a similar case in gdb.mi/mi-exec-run.exp.
14635 Tested on x86_64-linux.
14637 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29726
14639 2022-10-27 Nelson Chu <nelson@rivosinc.com>
14641 RISC-V: Fix build failures for -Werror=sign-compare.
14642 elfnn-riscv.c: In function ‘riscv_relax_resolve_delete_relocs’:
14643 elfnn-riscv.c:4256:30: error: operand of ‘?:’ changes signedness from ‘int’ to ‘unsigned int’ due to unsignedness of other operand [-Werror=sign-compare]
14645 So make the operands unsigned could resolve problem.
14648 * elfnn-riscv.c (riscv_relax_resolve_delete_relocs): Fixed build
14649 failures for -Werror=sign-compare.
14651 2022-10-27 Alan Modra <amodra@gmail.com>
14653 Fuzzed files in archives
14654 Given a fuzzed object file in an archive with section size exceeding
14655 file size, objcopy will report an error like "section size (0xfeffffff
14656 bytes) is larger than file size (0x17a bytes)" but will create a copy
14657 of the object laid out for the large section. That means a large
14658 temporary file on disk that is read back and written to the output
14659 archive, which can take a while. The output archive is then deleted
14660 due to the error. Avoid some of this silliness.
14662 * objcopy.c (copy_section): If section contents cannot be read
14663 set output section size to zero.
14665 2022-10-27 Martin Liska <mliska@suse.cz>
14667 tests: use canonical option name
14670 * testsuite/ld-size/size.exp: Use canonical option name.
14672 2022-10-27 Alan Modra <amodra@gmail.com>
14674 re: Support Intel AMX-FP16
14675 Fix these fails due to the target padding out sections with nops.
14676 x86_64-w64-mingw32 +FAIL: x86_64 AMX-FP16 insns
14677 x86_64-w64-mingw32 +FAIL: x86_64 AMX-FP16 insns (Intel disassembly)
14679 * testsuite/gas/i386/x86-64-amx-fp16-intel.d: Accept trailing nops.
14680 * testsuite/gas/i386/x86-64-amx-fp16.d: Likewise.
14682 2022-10-27 Alan Modra <amodra@gmail.com>
14684 Re: ld/testsuite: adjust ld-arm to run shared tests only when supported
14685 commit 67527cffcd enabled previously disabled tests unresolved-1-dyn,
14686 thumb-plt and thumb-plt-got for nacl. The first fails due to trying
14687 to link against mixed-lib.so which isn't compiled for nacl. The last
14689 objdump: tmpdir/dump(.rel.plt): relocation 0 has invalid symbol index 14885104
14691 readelf: Error: bad symbol index: 00e320f0 in reloc
14693 Relocation section '.rel.plt' at offset 0x128 contains 1 entry:
14694 Offset Info Type Sym. Value Symbol's Name
14695 e320f000 e320f000 R_ARM_NONE
14697 * testsuite/ld-arm/arm-elf.exp: Disable unresolved-1-dyn,
14698 thumb-plt and thumb-plt-got for nacl.
14700 2022-10-27 GDB Administrator <gdbadmin@sourceware.org>
14702 Automatic date update in version.in
14704 2022-10-26 Simon Marchi <simon.marchi@efficios.com>
14706 gdb/testsuite: fix gdb.guile/scm-parameter.exp "wrong type argument" test pattern for Guile >= 2.2
14707 Since commit 90319cefe3 ("GDB/Guile: Don't assert that an integer value
14708 is boolean"), I see:
14710 FAIL: gdb.guile/scm-parameter.exp: kind=PARAM_ZINTEGER: test-PARAM_ZINTEGER-param: guile (set-parameter-value! test-PARAM_ZINTEGER-param #:unlimited)
14711 FAIL: gdb.guile/scm-parameter.exp: kind=PARAM_ZUINTEGER: test-PARAM_ZUINTEGER-param: guile (set-parameter-value! test-PARAM_ZUINTEGER-param #:unlimited)
14713 This comes from the fact that GDB outputs this:
14715 ERROR: In procedure set-parameter-value!:
14716 In procedure gdbscm_set_parameter_value_x: Wrong type argument in position 2 (expecting integer): #:unlimited
14717 Error while executing Scheme code.
14719 while the test expects an additional "ERROR:" on the second line,
14720 something like this:
14722 ERROR: In procedure set-parameter-value!:
14723 ERROR: In procedure gdbscm_set_parameter_value_x: Wrong type argument in position 2 (expecting integer): #:unlimited
14724 Error while executing Scheme code.
14726 Guile 2.0 outputs the `ERROR:` on the second line, while later versions
14727 do not. Change the pattern to accept both outputs. This is similar to
14728 commit 6bbe1a929c6 ("[gdb/testsuite] Fix gdb.guile/scm-breakpoint.exp
14731 Change-Id: I9dc45e7492a4f08340cad974610242ed689de959
14733 2022-10-26 Luis Machado <luis.machado@arm.com>
14735 gdb/arm: Fix M-profile EXC_RETURN
14736 Arm v8-M Architecture Reference Manual,
14737 D1.2.95 EXC_RETURN, Exception Return Payload
14741 Exception Secure. The security domain the exception was taken to.
14742 The possible values of this bit are:
14746 arm-tdep.c:3443, arm_m_exception_cache () function tests this bit:
14748 exception_domain_is_secure = (bit (lr, 0) == 0);
14750 The test is negated!
14752 Later on line 3553, the condition evaluates if an additional state
14753 context is stacked:
14755 /* With the Security extension, the hardware saves R4..R11 too. */
14756 if (tdep->have_sec_ext && secure_stack_used
14757 && (!default_callee_register_stacking || exception_domain_is_secure))
14759 RM, B3.19 Exception entry, context stacking
14761 RPLHM "In a PE with the Security Extension, on taking an exception,
14764 2. If exception entry requires a transition from Secure state to
14765 Non-secure state, the PE hardware extends the stack frame and also
14766 saves additional state context."
14768 So we should test for !exception_domain_is_secure instead of non-negated
14770 These two bugs compensate each other so unstacking works correctly.
14772 But another test of exception_domain_is_secure (negated due to the
14773 first bug) prevents arm_unwind_secure_frames to work as expected:
14775 /* Unwinding from non-secure to secure can trip security
14776 measures. In order to avoid the debugger being
14777 intrusive, rely on the user to configure the requested
14779 if (secure_stack_used && !exception_domain_is_secure
14780 && !arm_unwind_secure_frames)
14782 Test with GNU gdb (GDB) 13.0.50.20221016-git.
14783 Stopped in a non-secure handler:
14785 (gdb) set arm unwind-secure-frames 0
14787 #0 HAL_SYSTICK_Callback () at C:/dvl/stm32l5trustzone/GPIO_IOToggle_TrustZone/NonSecure/Src/nsmain.c:490
14788 #1 0x0804081c in SysTick_Handler ()
14789 at C:/dvl/stm32l5trustzone/GPIO_IOToggle_TrustZone/NonSecure/Src/nsstm32l5xx_it.c:134
14790 #2 <signal handler called>
14791 #3 HAL_GPIO_ReadPin (GPIOx=0x52020800, GPIO_Pin=8192)
14792 at C:/dvl/stm32l5trustzone/GPIO_IOToggle_TrustZone/Drivers/STM32L5xx_HAL_Driver/Src/stm32l5xx_hal_gpio.c:386
14793 #4 0x0c000338 in SECURE_Mode () at C:/dvl/stm32l5trustzone/GPIO_IOToggle_TrustZone/Secure/Src/main.c:86
14794 #5 0x080403f2 in main () at C:/dvl/stm32l5trustzone/GPIO_IOToggle_TrustZone/NonSecure/Src/nsmain.c:278
14795 Backtrace stopped: previous frame inner to this frame (corrupt stack?)
14797 The frames #3 and #4 are secure. backtrace should stop before #3.
14799 Stopped in a secure handler:
14802 #0 HAL_SYSTICK_Callback () at C:/dvl/stm32l5trustzone/GPIO_IOToggle_TrustZone/Secure/Src/main.c:425
14803 #1 0x0c000b6a in SysTick_Handler ()
14804 at C:/dvl/stm32l5trustzone/GPIO_IOToggle_TrustZone/Secure/Src/stm32l5xx_it.c:234
14805 warning: Non-secure to secure stack unwinding disabled.
14806 #2 <signal handler called>
14808 The exception from secure to secure erroneously stops unwinding. It should
14809 continue as far as the security unlimited backtrace:
14811 (gdb) set arm unwind-secure-frames 1
14812 (gdb) si <-- used to rebuild frame cache after change of unwind-secure-frames
14813 0x0c0008e6 425 if (SecureTimingDelay != 0U)
14815 #0 0x0c0008e6 in HAL_SYSTICK_Callback () at C:/dvl/stm32l5trustzone/GPIO_IOToggle_TrustZone/Secure/Src/main.c:425
14816 #1 0x0c000b6a in SysTick_Handler ()
14817 at C:/dvl/stm32l5trustzone/GPIO_IOToggle_TrustZone/Secure/Src/stm32l5xx_it.c:234
14818 #2 <signal handler called>
14819 #3 0x0c000328 in SECURE_Mode () at C:/dvl/stm32l5trustzone/GPIO_IOToggle_TrustZone/Secure/Src/main.c:88
14820 #4 0x080403f2 in main () at C:/dvl/stm32l5trustzone/GPIO_IOToggle_TrustZone/NonSecure/Src/nsmain.c:278
14822 Backtrace stopped: previous frame inner to this frame (corrupt stack?)
14824 Set exception_domain_is_secure to the value expected by its name.
14825 Fix exception_domain_is_secure usage in the additional state context
14826 stacking condition.
14828 2022-10-26 Luis Machado <luis.machado@arm.com>
14830 gdb/arm: fix IPSR field test in arm_m_exception_cache ()
14831 Arm v8-M Architecture Reference Manual,
14832 D1.2.141 IPSR, Interrupt Program Status Register reads
14833 "Exception, bits [8:0]"
14835 9 bits, not 8! It is uncommon but true!
14837 2022-10-26 Luis Machado <luis.machado@arm.com>
14839 gdb/arm: Terminate frame unwinding in M-profile lockup
14840 In the lockup state the PC value of the the outer frame is irreversibly
14841 lost. The other registers are intact so LR likely contains
14842 PC of some frame next to the outer one, but we cannot analyze
14843 the nearest outer frame without knowing its PC
14844 therefore we do not know SP fixup for this frame.
14846 The frame unwinder possibly gets mad due to the wrong SP value.
14847 To prevent problems terminate unwinding if PC contains the magic
14848 value of the lockup state.
14850 Example session wihtout this change,
14851 Cortex-M33 CPU in lockup, gdb 13.0.50.20221016-git:
14856 Program received signal SIGINT, Interrupt.
14857 0xeffffffe in ?? ()
14859 #0 0xeffffffe in ?? ()
14860 #1 0x0c000a9c in HardFault_Handler ()
14861 at C:/dvl/stm32l5trustzone/GPIO_IOToggle_TrustZone/Secure/Src/stm32l5xx_it.c:99
14862 #2 0x2002ffd8 in ?? ()
14863 Backtrace stopped: previous frame identical to this frame (corrupt stack?)
14866 The frame #1 is at correct PC taken from LR, #2 is a total nonsense.
14873 Program received signal SIGINT, Interrupt.
14874 warning: ARM M in lockup state, stack unwinding terminated.
14875 <signal handler called>
14877 #0 <signal handler called>
14881 There is a visible drawback of emitting a warning in a cache buildnig routine
14882 as introduced in Torbjörn SVENSSON's
14883 [PATCH v4] gdb/arm: Stop unwinding on error, but do not assert
14884 The warning is printed just once and not repeated on each backtrace command.
14886 2022-10-26 Mike Frysinger <vapier@gentoo.org>
14888 gdb: copyright: make file header scan a bit more pythonic
14889 Should be functionally the same, but uses more pythonic idioms to get
14890 fewer lines of code, and to make sure to not leak open file handles.
14892 Approved-By: Simon Marchi <simon.marchi@efficios.com>
14894 2022-10-26 Mike Frysinger <vapier@gentoo.org>
14896 gdb: make copyright.py interface a bit nicer
14897 This way people can run `./copyright.py --help` and get some info as
14898 to what this does without it going and modifying the tree.
14900 sim: testsuite: improve parallel test processing
14901 The current logic limits itself to a maxdepth of 4 when looking for
14902 results. This wouldn't be a problem if cris didn't have a testsuite
14903 at a depth of 5 which we end up ignoring when summarizing. Rather
14904 than bump the number from 4 to 5, rework the code so that we gather
14905 the exact set of tests that we tried to run.
14907 2022-10-26 Alan Modra <amodra@gmail.com>
14909 buffer overflow in _bfd_XX_print_ce_compressed_pdata
14912 * peXXigen.c (_bfd_XX_print_ce_compressed_pdata): Use smaller of
14913 virt_size and bfd section size as limit of function table.
14915 2022-10-26 Alan Modra <amodra@gmail.com>
14917 Correct ELF reloc size sanity check
14918 The external reloc size check was wrong. Here asect is the code/data
14919 section, not the reloc section. So using this_hdr gave the size of
14920 the code/data section.
14922 * elf.c (_bfd_elf_get_reloc_upper_bound): Properly get
14923 external size from reloc headers.
14925 2022-10-26 Alan Modra <amodra@gmail.com>
14927 segfault in objdump.c reloc_at
14928 bfd_canonicalize_reloc returns -1L on errors.
14930 * objdump.c (load_specific_debug_section): Properly handle
14931 error return from bfd_canonicalize_reloc.
14933 2022-10-26 Alan Modra <amodra@gmail.com>
14935 som.c reloc sanity checking
14936 This patch checks that relocations emitted in som_write_fixups have
14937 offsets that are monotonic and within a section. To do that properly
14938 using bfd_reloc_offset_in_range it is necessary to set the reloc howto
14939 size field, which isn't used otherwise by the som backend. Note that
14940 the sizes used are not exactly those in the old sizing switch
14941 statement deleted from som_write_fixups, but all relocs handled by the
14942 main switch statement there get the same size. Most unhandled relocs
14943 get a zero size (exceptions being R_RELOCATION, R_SPACE_REF,
14944 R_MILLI_REL, R_BREAKPOINT which all involve writing one word according
14945 to my SOM reference). I figure it doesn't matter since any unhandled
14946 reloc is converted to 0xff R_RESERVED, and a default of zero is better
14947 for a "don't know" reloc.
14949 Besides tidying the code, stringizing name from type in SOM_HOWTO
14950 fixes R_REPEATED_INIT name.
14952 * som.c (SOM_HOWTO): Add SIZE arg, delete NAME. Stringize type
14954 (som_hppa_howto_table): Update with sizes.
14955 (som_write_fixups): Delete sizing switch statement. Sanity check
14956 bfd_reloc address against subsection size.
14958 2022-10-26 Alan Modra <amodra@gmail.com>
14960 som.c buffer overflow
14961 Fuzzed object files can put random values in bfd_reloc->address,
14962 leading to large som_reloc_skip output.
14964 * som.c (som_write_fixups): Allow for maximal som_reloc_skip.
14966 2022-10-26 Alan Modra <amodra@gmail.com>
14968 PR29720, objdump -S crashes if build-id is missing
14970 * objdump.c (slurp_file): Don't call debuginfod_find_source
14971 when build_id is NULL.
14973 2022-10-26 GDB Administrator <gdbadmin@sourceware.org>
14975 Automatic date update in version.in
14977 2022-10-25 Simon Marchi <simon.marchi@efficios.com>
14979 gdb: remove spurious spaces after frame_info_ptr
14980 Fix some whitespace issues introduced with the frame_info_ptr patch.
14982 Change-Id: I158d30d8108c97564276c647fc98283ff7b12163
14984 2022-10-25 Michael Matz <matz@suse.de>
14986 x86-64: Use only one default max-page-size
14987 On x86-64 the default ELF_MAXPAGESIZE depends on a configure
14988 option (--disable-separate-code). Since 9833b775
14989 ("PR28824, relro security issues") we use max-page-size for relro
14990 alignment (with a short interval, from 31b4d3a ("PR28824, relro
14991 security issues, x86 keep COMMONPAGESIZE relro") to its revert
14992 a1faa5ea, where x86-64 only used COMMONPAGESIZE as relro alignment
14995 But that means that a linker configured with --disable-separate-code
14996 behaves different from one configured with --enable-separate-code
14997 (the default), _even if using "-z {no,}separate-code" option to use
14998 the non-configured behaviour_ . In particular it means that when
14999 configuring with --disable-separate-code the linker will produce
15000 binaries aligned to 2MB pages on disk, and hence generate 2MB
15001 executables for a hello world (and even 6MB when linked with
15002 "-z separate-code").
15004 Generally we can't have constants that ultimately land in static
15005 variables be depending on configure options if those only influence
15006 behaviour that is overridable by command line options.
15008 So, do away with that, make the default MAXPAGESIZE be 4k (as is default
15009 for most x86-64 configs anyway, as most people won't configure with
15010 --disable-separate-code). If people need more they can use the
15011 "-z max-page-size" (with would have been required right now for a
15012 default configure binutils).
15015 * elf64-x86-64.c (ELF_MAXPAGESIZE): Don't depend on
15016 DEFAULT_LD_Z_SEPARATE_CODE.
15018 2022-10-25 Simon Marchi <simon.marchi@efficios.com>
15020 gdb/testsuite: make sure to consume the prompt in gdb.base/unwind-on-each-insn.exp
15021 This test fails quite reliably for me when ran as:
15023 $ taskset -c 1 make check TESTS="gdb.base/unwind-on-each-insn.exp" RUNTESTFLAGS="--target_board=native-gdbserver"
15027 $ make check-read1 TESTS="gdb.base/unwind-on-each-insn.exp"
15029 The problem is that the gdb_test_multiple call that grabs the frame id
15030 from "maint print frame-id" does not consume the prompt. Well, it does
15031 sometimes due to the trailing .*, but not always. If the prompt is not
15032 consumed, the tests that follow get confused:
15034 FAIL: gdb.base/unwind-on-each-insn.exp: gdb_breakpoint: set breakpoint at *foo
15035 FAIL: gdb.base/unwind-on-each-insn.exp: disassemble foo
15036 FAIL: gdb.base/unwind-on-each-insn.exp: get $sp and frame base in foo: get hexadecimal valueof "$sp"
15039 Use -wrap to make gdb_test_multiple consume the prompt.
15041 While at it, remove the bit that consumes the command name and do
15042 exp_continue, it's not really necessary. And for consistency, do the
15043 same changes to the gdb_test_multiple that consumes the stack address,
15044 although that one was fine, it did consume the prompt explicitly.
15046 Change-Id: I2b7328c8844c7e98921ea494c4c05107162619fc
15047 Reviewed-By: Bruno Larsen <blarsen@redhat.com>
15049 2022-10-25 Tom de Vries <tdevries@suse.de>
15051 [gdb/testsuite] Handle missing .note.GNU-stack
15052 On openSUSE Tumbleweed I run into this for the dwarf assembly test-cases, and
15053 some hardcoded assembly test-cases:
15055 Running gdb.dwarf2/fission-absolute-dwo.exp ...
15056 gdb compile failed, ld: warning: fission-absolute-dwo.o: \
15057 missing .note.GNU-stack section implies executable stack
15058 ld: NOTE: This behaviour is deprecated and will be removed in a future \
15059 version of the linker
15061 === gdb Summary ===
15063 # of untested testcases 1
15066 Fix the dwarf assembly test-cases by adding the missing .note.GNU-stack in
15067 proc Dwarf::assemble.
15069 Fix the hard-coded test-cases using this command:
15071 $ for f in $(find gdb/testsuite/gdb.* -name *.S); do
15072 if ! grep -q note.GNU-stack $f; then
15073 echo -e "\t.section\t.note.GNU-stack,\"\",@progbits" >> $f;
15078 Likewise for .s files, and gdb/testsuite/lib/my-syscalls.S.
15080 The idiom for arm seems to be to use %progbits instead, see commit 9a5911c08be
15081 ("gdb/testsuite/gdb.dwarf2: Replace @ with % for ARM compatability"), so
15082 hand-edit gdb/testsuite/gdb.arch/arm-disp-step.S to use %progbits instead.
15084 Note that dwarf assembly testcases use %progbits as decided by proc _section.
15086 Tested on x86_64-linux.
15088 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29674
15090 2022-10-25 Tom de Vries <tdevries@suse.de>
15092 [gdb/testsuite] Add missing skip_gdbserver_tests in gdb.multi/attach-no-multi-process.exp
15093 I build gdb without gdbserver, and ran into:
15095 (gdb) PASS: gdb.multi/attach-no-multi-process.exp: target_non_stop=off: \
15096 switch to inferior 2
15097 spawn of --once --multi localhost:2346 failed
15098 ERROR: tcl error sourcing attach-no-multi-process.exp.
15099 ERROR: tcl error code NONE
15100 ERROR: Timeout waiting for gdbserver response.
15103 Add the missing skip_gdbserver_tests.
15105 Tested on x86_64-linux.
15107 2022-10-25 Tom de Vries <tdevries@suse.de>
15109 [gdb] Rewrite RETHROW_ON_TARGET_CLOSE_ERROR into function
15110 Recent commit b2829fcf9b5 ("[gdb] Fix rethrow exception slicing in
15111 insert_bp_location") introduced macro RETHROW_ON_TARGET_CLOSE_ERROR.
15113 I wrote this as a macro in order to have the rethrowing throw be part of the
15114 same function as the catch, but as it turns out that's not necessary.
15116 Rewrite into a function.
15118 Tested on x86_64-linux.
15120 2022-10-25 Simon Marchi <simon.marchi@polymtl.ca>
15122 gdb: internal_error -> internal_error_loc in gdb-gdb.gdb.in
15123 Commit f34652de0b ("internal_error: remove need to pass
15124 __FILE__/__LINE__") renamed the internal_error function to
15125 internal_error_loc. Change gdb-gdb.gdb.in accordingly.
15127 Change-Id: I876e1623607b6becf74ade53d102ead53a74ed86
15129 2022-10-25 Nelson Chu <nelson@rivosinc.com>
15131 RISC-V: Should reset `again' flag for _bfd_riscv_relax_pc.
15132 The R_RISCV_DELETE relocations are no longer deleted at another relax pass,
15133 so we should reset 'again' flag to true for _bfd_riscv_relax_pc, while the
15134 deleted bytes are marked as R_RISCV_DELETE.
15137 * elfnn-riscv.c (_bfd_riscv_relax_pc): Set `again' to true while the
15138 deleted bytes are marked as R_RISCV_DELETE.
15140 2022-10-25 Patrick O'Neill <patrick@rivosinc.com>
15142 RISC-V: Improve link time complexity.
15143 The riscv port does deletion and symbol table update for each relocation
15144 while relaxing, so we are moving section bytes and scanning symbol table once
15145 for each relocation. Compared to microblaze port, they record the relaxation
15146 changes into a table, then do the deletion and symbol table update once per
15147 section, rather than per relocation. Therefore, they should have better link
15148 time complexity than us.
15150 To improve the link time complexity, this patch try to make the deletion in
15151 linear time. Compared to record the relaxation changes into a table, we
15152 replace the unused relocation with R_RISCV_DELETE for the deleted bytes, and
15153 then resolve them at the end of the section. Assuming the number of
15154 R_RISCV_DELETE is m, and the number of symbols is n, the total link complexity
15155 should be O(m) for moving section bytes, and O(m*n^2) for symbol table update.
15156 If we record the relaxation changes into the table, and then sort the symbol
15157 table by values, then probably can reduce the time complexity to O(m*n*log(n))
15158 for updating symbol table, but it doesn't seem worth it for now.
15161 * elfnn-riscv.c (_riscv_relax_delete_bytes): Renamed from
15162 riscv_relax_delete_bytes, updated to reduce the tiem complexity to O(m)
15164 (typedef relax_delete_t): Function pointer declaration of delete functions.
15165 (riscv_relax_delete_bytes): Can choose to use _riscv_relax_delete_piecewise
15166 or _riscv_relax_delete_immediate for deletion.
15167 (_riscv_relax_delete_piecewise): Just mark the deleted bytes as R_RISCV_DELETE.
15168 (_riscv_relax_delete_immediate): Delete some bytes from a section while
15170 (riscv_relax_resolve_delete_relocs): Delete the bytes for R_RISCV_DELETE
15171 relocations from a section, at the end of _bfd_riscv_relax_section.
15172 (_bfd_riscv_relax_call): Mark deleted bytes as R_RISCV_DELETE by reusing
15174 (_bfd_riscv_relax_lui): Likewise, but reuse R_RISCV_HI20 for lui, and reuse
15175 R_RISCV_RELAX for c.lui.
15176 (_bfd_riscv_relax_tls_le): Likewise, but resue R_RISCV_TPREL_HI20 and
15178 (_bfd_riscv_relax_pc): Likewise, but resue R_RISCV_PCREL_HI20 for auipc.
15179 (_bfd_riscv_relax_align): Updated, don't need to resue relocation since
15180 calling _riscv_relax_delete_immediate.
15181 (_bfd_riscv_relax_delete): Removed.
15182 (_bfd_riscv_relax_section): Set riscv_relax_delete_bytes for each relax_func,
15183 to delete bytes immediately or later. Call riscv_relax_resolve_delete_relocs
15184 to delete bytes for DELETE relocations from a section.
15186 2022-10-25 GDB Administrator <gdbadmin@sourceware.org>
15188 Automatic date update in version.in
15190 2022-10-24 Andrew Burgess <aburgess@redhat.com>
15192 gdb/doc: reword description of DisassembleInfo.read_memory
15193 While reading the documentation of DisassembleInfo.read_memory I
15194 spotted the word 'available' in one sentence where it didn't make
15197 2022-10-24 Andrew Burgess <aburgess@redhat.com>
15199 sim/lm32: fix some missing function declaration warnings
15200 In the lm32 simulator, I was seeing some warnings about missing
15201 function declarations.
15203 The lm32 simulator has a weird header structure, in order to pull in
15204 the full cpu.h header we need to define WANT_CPU_LM32BF. This is done
15205 in some files, but not in others. Critically, it's not done in some
15206 files that then use functions declared in cpu.h
15208 In this commit I added the missing #define so that the full cpu.h can
15211 After doing this there are still a few functions that are used
15212 undeclared, these functions appear to be missing any declarations at
15213 all, so I've added some to cpu.h.
15215 With this done all the warnings when compiling lm32 are resolved for
15216 both gcc and clang, so I've removed the SIM_WERROR_CFLAGS line from
15217 Makefile.in, this allows lm32 to build with -Werror.
15219 2022-10-24 Andrew Burgess <aburgess@redhat.com>
15221 sim/h8300: avoid self assignment
15222 There are two places in the h8300 simulator where we assign a variable
15223 to itself. Clang gives a warning for this, which is converted into an
15226 Silence the warning by removing the self assignments. As these
15227 assignments were in a complex if/then/else tree, rather than try to
15228 adjust all the conditions, I've just replaced the self assignments
15229 with a comment and an empty statement.
15231 2022-10-24 Andrew Burgess <aburgess@redhat.com>
15233 sim/aarch64: remove two unused functions
15234 These functions are not used. Clang warns about the unused functions,
15235 which is then converted into an error by -Werror.
15237 Delete the unused functions.
15239 2022-10-24 Andrew Burgess <aburgess@redhat.com>
15241 sim/ppc: fix for operator precedence warning from clang
15242 In the ppc simulator, clang was warning about some code like this:
15244 busy_ptr->nr_writebacks = 1 + (PPC_ONE_BIT_SET_P(out_vmask)) ? 1 : 2;
15248 operator '?:' has lower precedence than '+'; '+' will be evaluated first
15250 I suspect that this is not the original authors intention.
15251 PPC_ONE_BIT_SET_P is going to be 0 or 1, so if we evaluate the '+'
15252 first, the condition will always be non-zero, so true. The whole
15253 expression could then be simplified to just '1', which doesn't make
15256 I suspect the answer the author was expecting was either 2 or 3. Why
15257 they didn't just write:
15259 busy_ptr->nr_writebacks = (PPC_ONE_BIT_SET_P(out_vmask)) ? 2 : 3;
15261 I have no clue, however, to keep the structure of the code unchanged,
15262 I've updated things to:
15264 busy_ptr->nr_writebacks = 1 + (PPC_ONE_BIT_SET_P (out_vmask) ? 1 : 2);
15266 which silences the warning from clang, and is, I am guessing, what the
15267 original author intended.
15269 2022-10-24 Andrew Burgess <aburgess@redhat.com>
15271 sim/ppc: initialize a memory buffer in all cases
15272 In the ppc simulator's do_fstat function, which provides the fstat
15273 call for the simulator, if the fstat is going to fail then we
15274 currently write an uninitialized buffer into the simulated target.
15276 In theory, I think this is fine, we also write the error status into
15277 the simulated target, so, given that the fstat has failed, the target
15278 shouldn't be relying on the buffer contents.
15280 However, writing an uninitialized buffer means we might leak simulator
15281 private data into the simulated target, which is probably a bad thing.
15282 Plus it probably makes life easier if something consistent, like all
15283 zeros, is written rather than random junk, which might look like a
15284 successful call (except for the error code).
15286 So, in this commit, I initialize the stat buffer to zero before
15287 it is potentially used. If the stat call is not made then the buffer
15288 will be left initialized as all zeros.
15290 2022-10-24 Andrew Burgess <aburgess@redhat.com>
15292 sim/ppc: don't try to print an uninitialized variable
15293 The ppc simulator, in sim_create_inferior, tries to print the function
15294 local entry_point variable before the variable is initialized.
15296 In this commit, I defer the debug print line until the variable has
15299 2022-10-24 Andrew Burgess <aburgess@redhat.com>
15301 sim/sh: use fabs instead of abs
15302 The sh simulator incorrectly uses integer abs instead of the floating
15303 point fabs on some floating point values, fixed in this commit.
15305 2022-10-24 Tom de Vries <tdevries@suse.de>
15307 [gdb] Fix rethrow exception slicing in insert_bp_location
15308 The preferred way of rethrowing an exception is by using throw without
15309 expression, because it avoids object slicing of the exception [1].
15311 Fix this in insert_bp_location.
15313 Tested on x86_64-linux.
15315 [1] https://en.cppreference.com/w/cpp/language/throw
15317 Approved-By: Andrew Burgess <aburgess@redhat.com>
15319 2022-10-24 Tom de Vries <tdevries@suse.de>
15321 [gdb] Fix rethrow exception slicing in pretty_print_insn
15322 The preferred way of rethrowing an exception is by using throw without
15323 expression, because it avoids object slicing of the exception [1].
15325 Fix this in gdb_pretty_print_disassembler::pretty_print_insn.
15327 Tested on x86_64-linux.
15329 [1] https://en.cppreference.com/w/cpp/language/throw
15331 Approved-By: Andrew Burgess <aburgess@redhat.com>
15333 2022-10-24 Clément Chigot <chigot@adacore.com>
15335 ld/testsuite: adjust ld-arm to run shared tests only when supported
15336 If a custom arm-elf target is disabling the shared support, a lot of
15337 failures are reported by the testsuite.
15338 Moreover, some tests try to access libraries which have been explicitly
15339 skipped earlier (eg mixed-lib.so).
15343 * testsuite/ld-arm/arm-elf.exp: Separate tests needing shared
15346 2022-10-24 Clément Chigot <chigot@adacore.com>
15348 ld/testsuite: skip ld-elf/exclude when -shared is not supported
15351 * testsuite/ld-elf/exclude.exp: Call check_shared_lib_support.
15352 to skip for all targets without shared lib support.
15354 2022-10-24 Jan Beulich <jbeulich@suse.com>
15356 x86: consolidate VPCLMUL tests
15357 There's little point in having Intel syntax disassembler tests when the
15358 purpose of a test is assembler functionality: Drop all
15359 *avx512*_vpclmulqdq-wig1-intel.
15361 For *avx512*_vpclmulqdq-wig1 share source with *avx512*_vpclmulqdq.
15363 Finally put in place similar tests for -mvexwig=1.
15365 2022-10-24 Jan Beulich <jbeulich@suse.com>
15367 x86: consolidate VAES tests
15368 There's little point in having Intel syntax disassembler tests when the
15369 purpose of a test is assembler functionality: Drop all
15370 *avx512*_vaes-wig1-intel.
15372 For *avx512*_vaes-wig1 share source with *avx512*_vaes. This in
15373 particular makes sure that the 32-bit VL test actually tests any EVEX
15374 encodings in the first place.
15376 Finally put in place similar tests for -mvexwig=1.
15378 2022-10-24 Jan Beulich <jbeulich@suse.com>
15380 x86: emit {evex} prefix when disassembling ambiguous AVX512VL insns
15381 When no AVX512-specific functionality is in use, the disassembly of
15382 AVX512VL insns is indistinguishable from their AVX counterparts (if such
15383 exist). Emit the {evex} pseudo-prefix in such cases.
15385 Where applicable drop stray uses of PREFIX_OPCODE from table entries.
15387 2022-10-24 Tom de Vries <tdevries@suse.de>
15389 [gdb/testsuite] Add skip_python_tests in gdb.python/tui-window-names.exp
15390 I did a gdb build without python support, and during testing ran into FAILs in
15391 test-case gdb.python/tui-window-names.exp.
15393 Fix this by adding the missing skip_python_test.
15395 Tested on x86_64-linux.
15397 2022-10-24 GDB Administrator <gdbadmin@sourceware.org>
15399 Automatic date update in version.in
15401 2022-10-23 Mike Frysinger <vapier@gentoo.org>
15403 sim: testsuite: update ignored .exp files [PR sim/29596]
15404 Now that we run `check/foo.exp` instead of `check/./foo.exp`,
15405 update the config/ & lib/ exceptions to cover both paths.
15407 Bug: https://sourceware.org/PR29596
15409 2022-10-23 Mike Frysinger <vapier@gentoo.org>
15411 sim: testsuite: tweak parallel find invocation [PR sim/29596]
15412 Make sure we invoke runtest with the same exp filenames when running in
15413 parallel as it will find when run single threaded. When `runtest` finds
15414 files itself, it will use paths like "aarch64/allinsn.exp". When we run
15415 `find .` with the %p option, it produces "./aarch64/allinsn.exp". Switch
15416 to %P to get "aarch64/allinsn.exp".
15418 Bug: https://sourceware.org/PR29596
15420 2022-10-23 Mike Frysinger <vapier@gentoo.org>
15422 sim: mips/ppc/riscv: re-add AC_CANONICAL_SYSTEM [PR sim/29439]
15423 These configure scripts check $target and change behavior. They
15424 shouldn't be doing that, but until we can rework the sim to change
15425 behavior based on the input ELF, restore AC_CANONICAL_SYSTEM to
15426 these so that $target is correctly populated.
15428 This was lost in the d3562f83a7b8a1ae6e333cd5561419d3da18fcb4
15429 ("sim: unify toolchain probing logic") refactor as the logic was
15430 hoisted up to the common code. But the fact the vars weren't
15431 passed down to the sub-configure scripts was missed.
15433 Bug: https://sourceware.org/PR29439
15435 2022-10-23 GDB Administrator <gdbadmin@sourceware.org>
15437 Automatic date update in version.in
15439 2022-10-22 Simon Marchi <simon.marchi@polymtl.ca>
15441 gdb/testsuite: add max number of instructions check in gdb.base/unwind-on-each-insn.exp
15442 This test sends my CI in an infinite loop of failures. We expect to
15443 have a handful of iterations (5 on my development machine, where the
15444 test passes fine)but the log shows that it went up to 104340 iterations:
15446 FAIL: gdb.base/unwind-on-each-insn.exp - instruction 104340: maint print frame-id
15447 DUPLICATE: gdb.base/unwind-on-each-insn.exp - instruction 104340: maint print frame-id
15448 FAIL: gdb.base/unwind-on-each-insn.exp - instruction 104340: [string equal $fid $main_fid]
15449 FAIL: gdb.base/unwind-on-each-insn.exp - instruction 104340: get hexadecimal valueof "$pc"
15451 Add a max instruction check, exit the loop if we reach 100 iterations.
15452 This should allow the test to fail fast if there's a problem, but 100
15453 iterations should be more than enough for when things are working.
15455 Change-Id: I77978d593aca046068f9209272d82e1675ba17c2
15457 2022-10-22 GDB Administrator <gdbadmin@sourceware.org>
15459 Automatic date update in version.in
15461 2022-10-21 Pedro Alves <pedro@palves.net>
15463 Improve Python Unwinders documentation
15464 - avoid "GDB proper" to refer to global locus, as object files and
15465 program spaces are also GDB proper.
15467 - gdb.register_unwinder does not accept locus=gdb.
15469 - "a unwinder" -> "an unwinder"
15471 Approved-by: Eli Zaretskii <eliz@gnu.org>
15472 Change-Id: I98c1b1000e1063815238e945ca71ec6f37b5702e
15474 2022-10-21 Simon Marchi <simon.marchi@polymtl.ca>
15476 gdb: make inherit_abstract_dies use vector iterators
15477 Small cleanup to use std::vector iterators rather than raw pointers.
15479 Approved-By: Tom Tromey <tom@tromey.com>
15480 Change-Id: I8d50dbb3f2d8dad7ff94066a578d523f1f31b590
15482 2022-10-21 Simon Marchi <simon.marchi@polymtl.ca>
15484 gdb: check for empty offsets vector in inherit_abstract_dies
15485 When building GDB with clang and --enable-ubsan, I get:
15487 UNRESOLVED: gdb.dwarf2/frame-inlined-in-outer-frame.exp: starti prompt
15491 $ ./gdb --data-directory=data-directory -nx -q -readnow testsuite/outputs/gdb.dwarf2/frame-inlined-in-outer-frame/frame-inlined-in-outer-frame
15492 Reading symbols from testsuite/outputs/gdb.dwarf2/frame-inlined-in-outer-frame/frame-inlined-in-outer-frame...
15493 Expanding full symbols from testsuite/outputs/gdb.dwarf2/frame-inlined-in-outer-frame/frame-inlined-in-outer-frame...
15494 /home/simark/src/binutils-gdb/gdb/dwarf2/read.c:11954:47: runtime error: applying non-zero offset 8 to null pointer
15496 I found this to happen with ld-linux on at least Arch Linux and Ubuntu
15499 $ ./gdb --data-directory=data-directory -nx -q -readnow -iex "set debuginfod enabled on" /lib64/ld-linux-x86-64.so.2
15500 Reading symbols from /lib64/ld-linux-x86-64.so.2...
15501 Reading symbols from /home/simark/.cache/debuginfod_client/22bd7a2c03d8cfc05ef7092bfae5932223189bc1/debuginfo...
15502 Expanding full symbols from /home/simark/.cache/debuginfod_client/22bd7a2c03d8cfc05ef7092bfae5932223189bc1/debuginfo...
15503 /home/simark/src/binutils-gdb/gdb/dwarf2/read.c:11954:47: runtime error: applying non-zero offset 8 to null pointer
15505 The problem happens when doing this:
15507 sect_offset *offsetp = offsets.data () + 1
15509 When `offsets` is an empty vector, `offsets.data ()` returns nullptr.
15510 Fix it by wrapping that in a `!offsets.empty ()` check.
15512 Change-Id: I6d29ba2fe80ba4308f68effd9c57d4ee8d67c29f
15513 Approved-By: Tom Tromey <tom@tromey.com>
15515 2022-10-21 Fangrui Song <maskray@google.com>
15517 readelf: support zstd compressed debug sections [PR 29640]
15519 2022-10-21 Tom Tromey <tromey@adacore.com>
15521 Fix incorrect .gdb_index with new DWARF scanner
15522 PR symtab/29694 points out a regression caused by the new DWARF
15523 scanner when the cc-with-gdb-index target board is used.
15525 What happens here is that an older version of gdb will make an index
15526 describing the "A" type as:
15528 [737] A: 1 [global, type]
15530 whereas the new gdb says:
15532 [1008] A: 0 [global, type]
15534 Here the old one is correct because the A in CU 0 is just a
15535 declaration without a size:
15537 <1><45>: Abbrev Number: 10 (DW_TAG_structure_type)
15538 <46> DW_AT_name : A
15539 <48> DW_AT_declaration : 1
15540 <48> DW_AT_sibling : <0x6d>
15542 This patch fixes the problem by introducing the idea of a "type
15543 declaration". I think gdb still needs to recurse into these types,
15544 searching for methods, but by marking the type itself as a
15545 declaration, gdb can skip this type during lookups and when writing
15548 Regression tested on x86-64 using the cc-with-gdb-index board.
15550 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29694
15552 2022-10-21 Tom Tromey <tromey@adacore.com>
15554 Fix crash in value_print_array_elements
15555 A user noticed that gdb would crash when printing a packed array after
15556 doing "set lang c". Packed arrays don't exist in C, but it's
15557 occasionally useful to print things in C mode when working in a non-C
15558 language -- this lets you see under the hood a little bit.
15560 The bug here is that generic value printing does not handle packed
15561 arrays at all. This patch fixes the bug by introducing a new function
15562 to extract a value from a bit offset and width.
15564 The new function includes a hack to avoid problems with some existing
15565 test cases when using -fgnat-encodings=all. Cleaning up this code
15566 looked difficult, and since "all" is effectively deprecated, I thought
15567 it made sense to simply work around the problems.
15569 2022-10-21 Tom Tromey <tromey@adacore.com>
15571 Fix bug in Ada packed array handling
15572 A user found a bug where an array of packed arrays was printed
15573 incorrectly. The bug here is that the packed array has a bit stride,
15574 but the outer array does not -- and should not. However,
15575 update_static_array_size does not distinguish between an array of
15576 packed arrays and a multi-dimensional packed array, and for the
15577 latter, only the innermost array will end up with a stride.
15579 This patch fixes the problem by adding a flag to indicate whether a
15580 given array type is a constituent of a multi-dimensional array.
15582 2022-10-21 Simon Marchi <simon.marchi@polymtl.ca>
15584 gdb: declare variables on first use in inherit_abstract_dies
15585 Move variable declarations to where they are first use, plus some random
15588 Change-Id: Idf40d60f9034996fa6a234165cd989a721eb4148
15590 2022-10-21 Nick Clifton <nickc@redhat.com>
15592 Add a -w option to the linker to suppress warning and error messages.
15594 * ld.h (struct ld_config_type): Add no_warnings field.
15595 * ldlex.h (enum option_values): Add OPTION_NO_WARNINGS.
15596 * lexsup.c (ld_options): Add --no-warnings.
15597 (parse_args): Add support for -w and --no-warnings.
15598 * ldmisc.c (vfinfo): Return early if the message is a warning and
15599 -w has been enabled.
15600 * ld.texi (options): Document new command line option.
15601 * NEWS: Mention the new feature.
15603 Add a note to the binutils/NEWS file about DCO signed contributions.
15605 2022-10-21 Bruno Larsen <blarsen@redhat.com>
15607 gdb/reverse: Fix stepping over recursive functions
15608 Currently, when using GDB to do reverse debugging, if we try to use the
15609 command "reverse next" to skip a recursive function, instead of skipping
15610 all of the recursive calls and stopping in the previous line, we stop at
15611 the second to last recursive call, and need to manually step backwards
15612 until we leave the first call. This is well documented in PR gdb/16678.
15614 This bug happens because when GDB notices that a reverse step has
15615 entered into a function, GDB will add a step_resume_breakpoint at the
15616 start of the function, then single step out of the prologue once that
15617 breakpoint is hit. The problem was happening because GDB wouldn't give
15618 that step_resume_breakpoint a frame-id, so the first time the breakpoint
15619 was hit, the inferior would be stopped. This is fixed by giving the
15620 current frame-id to the breakpoint.
15622 This commit also changes gdb.reverse/step-reverse.c to contain a
15623 recursive function and attempt to both, skip it altogether, and to skip
15624 the second call from inside the first call, as this setup broke a
15625 previous version of the patch.
15627 2022-10-21 Bruno Larsen <blarsen@redhat.com>
15629 Change calculation of frame_id by amd64 epilogue unwinder
15630 When GDB is stopped at a ret instruction and no debug information is
15631 available for unwinding, GDB defaults to the amd64 epilogue unwinder, to
15632 be able to generate a decent backtrace. However, when calculating the
15633 frame id, the epilogue unwinder generates information as if the return
15634 instruction was the whole frame.
15636 This was an issue especially when attempting to reverse debug, as GDB
15637 would place a step_resume_breakpoint from the epilogue of a function if
15638 we were to attempt to skip that function, and this breakpoint should
15639 ideally have the current function's frame_id to avoid other problems
15640 such as PR record/16678.
15642 This commit changes the frame_id calculation for the amd64 epilogue,
15643 so that it is always the same as the dwarf2 unwinder's frame_id.
15645 It also adds a test to confirm that the frame_id will be the same,
15646 regardless of using the epilogue unwinder or not, thanks to Andrew
15649 Co-Authored-By: Andrew Burgess <aburgess@redhat.com>
15651 2022-10-21 Nick Clifton <nickc@redhat.com>
15653 Updated Hungarian translation for the gprof sub-directory.
15654 * po/hu.po: Updated Hungarian translation.
15656 2022-10-21 Maciej W. Rozycki <macro@embecosm.com>
15658 GDB/Python: Make `None' stand for `unlimited' in setting integer parameters
15659 Similarly to booleans and following the fix for PR python/29217 make
15660 `gdb.parameter' accept `None' for `unlimited' with parameters of the
15661 PARAM_UINTEGER, PARAM_INTEGER, and PARAM_ZUINTEGER_UNLIMITED types, as
15662 `None' is already returned by parameters of the two former types, so
15663 one might expect to be able to feed it back. It also makes it possible
15664 to avoid the need to know what the internal integer representation is
15665 for the special setting of `unlimited'.
15667 Expand the testsuite accordingly.
15669 Approved-By: Simon Marchi <simon.marchi@polymtl.ca>
15671 2022-10-21 Maciej W. Rozycki <macro@embecosm.com>
15673 GDB/testsuite: Expand Python integer parameter coverage across all types
15674 Also verify PARAM_UINTEGER, PARAM_INTEGER, and PARAM_ZINTEGER parameter
15675 types, in addition to PARAM_ZUINTEGER and PARAM_ZUINTEGER_UNLIMITED
15676 already covered, and verify a choice of existing GDB parameters. Add
15677 verification for reading parameters via `<parameter>.value' in addition
15678 to `gdb.parameter('<parameter>')' as this covers different code paths.
15680 Approved-By: Simon Marchi <simon.marchi@polymtl.ca>
15682 2022-10-21 Maciej W. Rozycki <macro@embecosm.com>
15684 GDB/Guile: Don't assert that an integer value is boolean
15685 Do not assert that a value intended for an integer parameter, of either
15686 the PARAM_UINTEGER or the PARAM_ZUINTEGER_UNLIMITED type, is boolean,
15687 causing error messages such as:
15689 ERROR: In procedure make-parameter:
15690 ERROR: In procedure gdbscm_make_parameter: Wrong type argument in position 15 (expecting integer or #:unlimited): 3
15691 Error while executing Scheme code.
15693 when initialization with a number is attempted. Instead assert that it
15694 is integer. Keep matching `#:unlimited' keyword as an alternative. Add
15695 suitable test cases.
15697 Approved-By: Simon Marchi <simon.marchi@polymtl.ca>
15699 2022-10-21 Tom de Vries <tdevries@suse.de>
15701 [gdb/testsuite] Silence compilation fail in gdb.base/rtld-step.exp
15702 With gcc 7.5.0 and test-case gdb.base/rtld-step.exp, I run into:
15704 gdb compile failed, gcc: error: unrecognized command line option \
15705 '-static-pie'; did you mean '-static'?
15708 Silence this by checking in the test-case that -static-pie is supported, and
15711 UNTESTED: gdb.base/rtld-step.exp: \
15712 failed to compile (-static-pie not supported or static libc missing)
15715 Tested on x86_64-linux, with:
15716 - gcc 7.5.0: UNTESTED
15717 - gcc 12.2.1 with static glibc not installed: UNTESTED
15718 - gcc 12.2.1 with static glibc installed: PASS
15720 2022-10-21 Cui,Lili <lili.cui@intel.com>
15722 Support Intel AMX-FP16
15725 * NEWS: Add support for Intel AMX-FP16 instruction.
15726 * config/tc-i386.c: Add amx_fp16.
15727 * doc/c-i386.texi: Document .amx_fp16.
15728 * testsuite/gas/i386/i386.exp: Add AMX-FP16 tests.
15729 * testsuite/gas/i386/x86-64-amx-fp16-intel.d: New test.
15730 * testsuite/gas/i386/x86-64-amx-fp16.d: Likewise.
15731 * testsuite/gas/i386/x86-64-amx-fp16.s: Likewise.
15732 * testsuite/gas/i386/x86-64-amx-fp16-bad.d: Likewise.
15733 * testsuite/gas/i386/x86-64-amx-fp16-bad.s: Likewise.
15737 * i386-dis.c (MOD_VEX_0F385C_X86_64_P_3_W_0): New.
15738 (VEX_LEN_0F385C_X86_64_P_3_W_0_M_0): Likewise.
15739 (VEX_W_0F385C_X86_64_P_3): Likewise.
15740 (prefix_table): Add VEX_W_0F385C_X86_64_P_3.
15741 (vex_len_table): Add VEX_LEN_0F385C_X86_64_P_3_W_0_M_0.
15742 (vex_w_table): Add VEX_W_0F385C_X86_64_P_3.
15743 (mod_table): Add MOD_VEX_0F385C_X86_64_P_3_W_0.
15744 * i386-gen.c (cpu_flag_init): Add AMX-FP16_FLAGS.
15745 (CPU_ANY_AMX_TILE_FLAGS): Add CpuAMX_FP16.
15746 (cpu_flags): Add CpuAMX-FP16.
15747 * i386-opc.h (enum): Add CpuAMX-FP16.
15748 (i386_cpu_flags): Add cpuamx_fp16.
15749 * i386-opc.tbl: Add Intel AMX-FP16 instruction.
15750 * i386-init.h: Regenerate.
15751 * i386-tbl.h: Likewise.
15753 2022-10-21 Tsukasa OI <research_trasio@irq.a4lg.com>
15755 sim: Remove unused CXXFLAGS substitution
15756 Not only that sim/configure.ac does not AC_SUBST CXXFLAGS,
15757 unless we need C++ compiler like CXX, substitution @CXXFLAGS@ is useless.
15758 Because of this, this commit removes this substitution.
15760 2022-10-21 GDB Administrator <gdbadmin@sourceware.org>
15762 Automatic date update in version.in
15764 2022-10-20 H.J. Lu <hjl.tools@gmail.com>
15766 x86: Check VEX/EVEX encoding before checking vector operands
15769 commit 837e225ba1992f9745e5bbbd5e8443243a7f475f
15770 Author: Jan Beulich <jbeulich@suse.com>
15771 Date: Thu Oct 20 10:01:12 2022 +0200
15773 x86: re-work AVX-VNNI support
15775 moved AVX-VNNI after AVX512-VNNI, vector Disp8 is applied even when VEX
15776 encoding is selected. Check VEX/EVEX encoding before checking vector
15777 operands to avoid vector Disp8 with VEX encoding.
15780 * config/tc-i386.c (match_template): Check VEX/EVEX encoding
15781 before checking vector operands.
15782 * testsuite/gas/i386/avx-vnni.d: Updated.
15783 * testsuite/gas/i386/x86-64-avx-vnni.d: Likewise.
15784 * testsuite/gas/i386/avx-vnni.s: Add a Disp32 test.
15785 * testsuite/gas/i386/x86-64-avx-vnni.s: Likewise.
15787 2022-10-20 Andrew Burgess <aburgess@redhat.com>
15789 gdb/python: break more dependencies between gdbpy_initialize_* functions
15790 In a later commit in this series I will propose removing all of the
15791 explicit gdbpy_initialize_* calls from python.c and replace these
15792 calls with a more generic mechanism.
15794 One of the side effects of this generic mechanism is that the order in
15795 which the various Python sub-systems within GDB are initialized is no
15798 On the whole I don't think this matters, most of the sub-systems are
15799 independent of each other, though testing did reveal a few places
15800 where we did have dependencies, though I don't think those
15801 dependencies were explicitly documented in comment anywhere.
15803 This commit is similar to the previous one, and fixes the second
15804 dependency issue that I found.
15806 In this case the finish_breakpoint_object_type uses the
15807 breakpoint_object_type as its tp_base, this means that
15808 breakpoint_object_type must have been initialized with a call to
15809 PyType_Ready before finish_breakpoint_object_type can be initialized.
15811 Previously we depended on the ordering of calls to
15812 gdbpy_initialize_breakpoints and gdbpy_initialize_finishbreakpoints in
15815 After this commit a new function gdbpy_breakpoint_init_breakpoint_type
15816 exists, this function ensures that breakpoint_object_type has been
15817 initialized, and can be called from any gdbpy_initialize_* function.
15819 I feel that this change makes the dependency explicit, which I think
15822 There should be no user visible changes after this commit.
15824 2022-10-20 Andrew Burgess <aburgess@redhat.com>
15826 gdb/python: break dependencies between gdbpy_initialize_* functions
15827 In a later commit in this series I will propose removing all of the
15828 explicit gdbpy_initialize_* calls from python.c and replace these
15829 calls with a more generic mechanism.
15831 One of the side effects of this generic mechanism is that the order in
15832 which the various Python sub-systems within GDB are initialized is no
15835 On the whole I don't think this matters, most of the sub-systems are
15836 independent of each other, though testing did reveal a few places
15837 where we did have dependencies, though I don't think those
15838 dependencies were explicitly documented in a comment anywhere.
15840 This commit removes the first dependency issue, with this and the next
15841 commit, all of the implicit inter-sub-system dependencies will be
15842 replaced by explicit dependencies, which will allow me to, I think,
15843 clean up how the sub-systems are initialized.
15845 The dependency is around the py_insn_type. This type is setup in
15846 gdbpy_initialize_instruction and used in gdbpy_initialize_record.
15847 Rather than depend on the calls to these two functions being in a
15848 particular order, in this commit I propose adding a new function
15849 py_insn_get_insn_type. This function will take care of setting up the
15850 py_insn_type type and calling PyType_Ready. This helper function can
15851 be called from gdbpy_initialize_record and
15852 gdbpy_initialize_instruction, and the py_insn_type will be initialized
15855 To me this is better, the dependency is now really obvious, but also,
15856 we no longer care in which order gdbpy_initialize_record and
15857 gdbpy_initialize_instruction are called.
15859 There should be no user visible changes after this commit.
15861 2022-10-20 Andrew Burgess <aburgess@redhat.com>
15863 gdb: some int to bool conversion in breakpoint.c
15864 Some int to bool conversion in breakpoint.c. I've only updated the
15865 function signatures of static functions, but I've updated some
15866 function local variables throughout the file.
15868 There should be no user visible changes after this commit.
15870 Approved-By: Simon Marchi <simon.marchi@efficios.com>
15872 2022-10-20 Andrew Burgess <aburgess@redhat.com>
15874 gdb: make use of scoped_restore in unduplicated_should_be_inserted
15875 I noticed that we could make use of a scoped_restore in the function
15876 unduplicated_should_be_inserted. I've also converted the function
15877 return type from int to bool.
15879 This change shouldn't make any difference, as I don't think anything
15880 within should_be_inserted could throw an exception, but the change
15881 doesn't hurt, and will help keep us safe if anything ever changes in
15884 Approved-By: Simon Marchi <simon.marchi@efficios.com>
15886 2022-10-20 Andrew Burgess <aburgess@redhat.com>
15888 gdb: used scoped_restore_frame in update_watchpoint
15889 I was doing some int to bool cleanup in update_watchpoint, and I
15890 noticed a manual version of scoped_restore_selected_frame. As always
15891 when these things are done manually, there is the chance that, in an
15892 error case, we might leave the wrong frame selected.
15894 This commit updates things to use scoped_restore_selected_frame, and
15895 also converts a local variable from int to bool.
15897 The only user visible change after this commit is in the case where
15898 update_watchpoint throws an error - we should now correctly restore
15899 the previously selected frame. Otherwise, this commit should be
15900 invisible to the user.
15902 Approved-By: Simon Marchi <simon.marchi@efficios.com>
15904 2022-10-20 Andrew Burgess <aburgess@redhat.com>
15906 gdb: make some bp_location arguments const in breakpoint.c
15907 I spotted a few places where I could make some 'bp_location *'
15908 arguments constant in breakpoint.c.
15910 There should be no user visible changes after this commit.
15912 Approved-By: Simon Marchi <simon.marchi@efficios.com>
15914 2022-10-20 Дилян Палаузов <dilyan.palauzov@aegee.org>
15916 Reapply "Don't build readline/libreadline.a, when --with-system-readline is supplied"
15917 Commit 228cf97dd3c8 ("Merge configure.ac from gcc project") undid the
15918 change originally done in commit 69961a84c9b ("Don't build
15919 readline/libreadline.a, when --with-system-readline is supplied").
15922 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=18632
15924 2022-10-20 Jan Beulich <jbeulich@suse.com>
15926 x86: re-work AVX-VNNI support
15927 By putting the templates after their AVX512 counterparts, the AVX512
15928 flavors will be picked by default. That way the need to always use {vex}
15929 ceases to exist once respective CPU features (AVX512-VNNI or AVX512VL as
15930 a whole) have been disabled. This way the need for the PseudoVexPrefix
15931 attribute also disappears.
15933 2022-10-20 Tom de Vries <tdevries@suse.de>
15935 [gdb/testsuite] Fix gdb.debuginfod/fetch_src_and_symbols.exp with check-read1
15936 With test-case gdb.debuginfod/fetch_src_and_symbols.exp and check-read1, I run
15939 (gdb) FAIL: gdb.debuginfod/fetch_src_and_symbols.exp: local_url: \
15940 file fetch_src_and_symbols (got interactive prompt)
15943 The problem is that this output:
15945 Enable debuginfod for this session? (y or [n]) y^M
15947 is matched using regexp "Enable debuginfod?.*" with matches only the first two
15948 words of the output, after which an implicit clause in gdb_test_multiple triggers
15949 on the second part containing the interactive prompt.
15951 Fix this by included the interactive prompt in the regexp.
15953 Tested on x86_64-linux.
15955 2022-10-20 Tom de Vries <tdevries@suse.de>
15957 [gdb/testsuite] Fix gdb.mi/mi-disassemble.exp with check-read1
15958 With test-case gdb.mi/mi-disassemble.exp and check-read1 I run into:
15960 FAIL: gdb.mi/mi-disassemble.exp: disassemble /b main
15961 FAIL: gdb.mi/mi-disassemble.exp: get valueof "*((unsigned char *) 0x400549)"
15964 The problem for both FAILs is that the output is parsed using
15965 gdb_test_multiple, which has implicit clauses using $gdb_prompt, which can
15966 match before the explicit clauses using $mi_gdb_prompt.
15968 Fix this by passing -prompt "$mi_gdb_prompt$" to gdb_test_multiple.
15970 Tested on x86-64-linux.
15972 2022-10-20 Alan Modra <amodra@gmail.com>
15974 Re: aarch64-pe support for LD, GAS and BFD
15975 Fix dependencies for eaarch64pe.c. Generated files aren't handled
15976 fully automatically.
15978 2022-10-20 Mark Harmstone <mark@harmstone.com>
15980 ld: Add minimal pdb generation
15982 ld: Add --pdb option
15983 Second patch incorporates fixes for endian and UB issues in calc_hash, as per
15984 https://sourceware.org/pipermail/binutils/2022-October/123514.html.
15986 2022-10-20 Kevin Buettner <kevinb@redhat.com>
15988 Test stepping within a runtime loader / dynamic linker
15989 See the remarks in rtld-step.exp for a description of what this
15992 This test case has been tested using gcc on the following x86-64 Linux
15993 distributions/releases:
16006 It's also been tested (and found to be working) with
16007 RUNTESTFLAGS="CC_FOR_TARGET=clang" on all of the above expect for
16008 Fedora 28. The (old) version of clang available on F28 did not
16009 accept the -static-pie option.
16011 I also tried to make this test work on FreeBSD 13.1. While I think I
16012 made significant progress, I was ultimately stymied by this message
16013 which occurs when attempting to run the main program which has been
16014 set to use the fake/pretend RTLD as the ELF interpreter:
16016 ELF interpreter /path/to/rtld-step-rtld not found, error 22
16018 I have left one of the flags (-static) in place which I believe
16019 to be needed for FreeBSD (though since I never got it to work, I
16020 don't know for sure.) I've also left some declarations needed
16021 for FreeBSD in rtld-step-rtld.c. They're currently disabled via
16022 a #if 0; you'll need to enable them if you want to try to make
16023 it work on FreeBSD.
16025 2022-10-20 Kevin Buettner <kevinb@redhat.com>
16027 Allow debugging of runtime loader / dynamic linker
16028 At present, GDB does not allow for the debugging of the runtime loader
16029 and/or dynamic linker. Much of the time, this makes sense. An
16030 application programmer doesn't normally want to see symbol resolution
16031 code when stepping into a function that hasn't been resolved yet.
16033 But someone who wishes to debug the runtime loader / dynamic linker
16034 might place a breakpoint in that code and then wish to debug it
16035 as normal. At the moment, this is not possible. Attempting to step
16036 will cause GDB to internally step (and not stop) until code
16037 unrelated to the dynamic linker is reached.
16039 This commit makes a minor change to infrun.c which allows the dynamic
16040 loader / linker to be debugged in the case where a step, next, etc.
16041 is initiated from within that code.
16043 While developing this fix, I tried some approaches which weren't quite
16044 right. The GDB testusite definitely contains tests which FAIL when
16045 it's done incorrectly. (At one point, I saw 17 regressions!) This
16046 commit has been tested on x86-64 linux with no regressions.
16048 2022-10-20 Tsukasa OI <research_trasio@irq.a4lg.com>
16050 binutils: Remove unused substitution PROGRAM
16051 Unlike other substitution, this substitution of @PROGRAM@ was done in
16052 binutils/Makefile and it was intended for binutils/cxxfilt.man. @PROGRAM@
16053 in binutils/cxxfilt.man is removed in the commit 0285c67df190 ("Automate
16054 generate on man pages") in 2001 and @PROGRAM@ is ineffective since then.
16056 Because PROGRAM substitution does nothing, removing this manual
16057 substitution should be completely safe.
16059 binutils/ChangeLog:
16061 * doc/local.mk: Remove unused substitution PROGRAM.
16062 * Makefile.in: Regenerate.
16064 2022-10-20 GDB Administrator <gdbadmin@sourceware.org>
16066 Automatic date update in version.in
16068 2022-10-20 Alan Modra <amodra@gmail.com>
16071 * config.bfd: Obsolete *-*-beos*. Simplify x86 beos match.
16073 Regen ld/po/BLD-POTFILES.in
16075 2022-10-19 Tom de Vries <tdevries@suse.de>
16077 [gdb] Fix assert in handle_jit_event
16078 With the cc-with-tweaks.sh patch submitted here (
16079 https://sourceware.org/pipermail/gdb-patches/2022-October/192586.html ) we run
16082 $ export STRIP_ARGS_STRIP_DEBUG=--strip-all
16083 $ make check RUNTESTFLAGS="gdb.base/jit-reader.exp \
16084 --target_board cc-with-gnu-debuglink"
16086 into the following assert:
16089 Starting program: jit-reader ^M
16090 gdb/jit.c:1247: internal-error: jit_event_handler: \
16091 Assertion `jiter->jiter_data != nullptr' failed.^M
16094 Fix this by handling the
16095 jit_bp_sym.objfile->separate_debug_objfile_backlink != nullptr case in
16098 Tested on x86_64-linux.
16100 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29277
16102 2022-10-19 Pedro Alves <pedro@palves.net>
16104 internal_error: remove need to pass __FILE__/__LINE__
16105 Currently, every internal_error call must be passed __FILE__/__LINE__
16108 internal_error (__FILE__, __LINE__, "foo %d", var);
16110 The need to pass in explicit __FILE__/__LINE__ is there probably
16111 because the function predates widespread and portable variadic macros
16112 availability. We can use variadic macros nowadays, and in fact, we
16113 already use them in several places, including the related
16114 gdb_assert_not_reached.
16116 So this patch renames the internal_error function to something else,
16117 and then reimplements internal_error as a variadic macro that expands
16118 __FILE__/__LINE__ itself.
16120 The result is that we now should call internal_error like so:
16122 internal_error ("foo %d", var);
16124 Likewise for internal_warning.
16126 The patch adjusts all calls sites. 99% of the adjustments were done
16127 with a perl/sed script.
16129 The non-mechanical changes are in gdbsupport/errors.h,
16130 gdbsupport/gdb_assert.h, and gdb/gdbarch.py.
16132 Approved-By: Simon Marchi <simon.marchi@efficios.com>
16133 Change-Id: Ia6f372c11550ca876829e8fd85048f4502bdcf06
16135 2022-10-19 Nick Clifton <nickc@redhat.com>
16137 Fix an illegal memory access when parsing an ELF file containing corrupt symbol version information.
16139 * elf.c (_bfd_elf_slurp_version_tables): Fail if the sh_info field
16140 of the section header is zero.
16142 2022-10-19 Andrew Burgess <aburgess@redhat.com>
16144 sim/iq2000: silence pointer-sign warnings
16145 When building the iq2000 simulator I see a few warnings like this:
16147 /tmp/build/sim/../../src/sim/iq2000/iq2000.c: In function ‘fetch_str’:
16148 /tmp/build/sim/../../src/sim/iq2000/iq2000.c:50:54: error: pointer targets in passing argument 3 of ‘sim_read’ differ in signedness [-Werror=pointer-sign]
16149 50 | sim_read (CPU_STATE (current_cpu), CPU2DATA(addr), buf, nr);
16154 I've silenced these warnings by casting buf to 'unsigned char *'.
16155 With this change I now see no warnings when compiling iq2000.c, so
16156 I've removed the line from Makefile.in that disables -Werror.
16158 Makefile.in was also disabling -Werror when compiling mloop.c,
16159 however, I'm not seeing any warnings when compiling that file, so I've
16160 removed the -Werror disable in that case too.
16162 2022-10-19 Andrew Burgess <aburgess@redhat.com>
16164 sim/erc32: avoid dereferencing type-punned pointer warnings
16165 When building the erc32 simulator I get a few warnings like this:
16167 /tmp/build/sim/../../src/sim/erc32/exec.c:1377:21: warning: dereferencing type-punned pointer will break strict-aliasing rules [-Wstrict-aliasing]
16168 1377 | sregs->fs[rd] = *((float32 *) & ddata[0]);
16169 | ~^~~~~~~~~~~~~~~~~~~~~~~
16171 The type of '& ddata[0]' will be 'uint32_t *', which is what triggers
16174 This commit makes use of memcpy when performing the type-punning,
16175 which resolves the above warnings.
16177 With this change, I now see no warnings when compiling exec.c, which
16178 means that the line in Makefile.in that disables -Werror can be
16181 There should be no change in behaviour after this commit.
16183 2022-10-19 Andrew Burgess <aburgess@redhat.com>
16185 sim/ppc: mark device_error function as ATTRIBUTE_NORETURN
16186 The device_error function always ends up calling the error function,
16187 which is itself marked as ATTRIBUTE_NORETURN, so it makes sense that
16188 device_error should also be marked ATTRIBUTE_NORETURN.
16190 Doing this resolves a few warnings from hw_ide.c about possibly
16191 uninitialized variables - the variables are only uninitialized after
16192 passing through a call to device_error, which obviously means the
16193 variables are never really used uninitialized, the simulation will
16194 terminate with the device_error call.
16196 2022-10-19 Andrew Burgess <aburgess@redhat.com>
16198 sim/ppc: fix warnings related to printf format strings
16199 This commit is a follow on to:
16201 commit 182421c9d2eea8c4877d983a2124e591f0aca710
16202 Date: Tue Oct 11 15:02:08 2022 +0100
16204 sim/ppc: fixes for arguments to printf style functions
16206 where commit 182421c9d2ee addressed issues with printf format
16207 arguments that were causing the compiler to give an error, this commit
16208 addresses issues that caused the compiler to emit a warning.
16210 This commit is mostly either changing the format string to match the
16211 argument, or in some cases, excess, unused arguments are removed.
16213 2022-10-19 Andrew Burgess <aburgess@redhat.com>
16215 sim/cgen: mask uninitialized variable warning in cgen-run.c
16216 I see an uninitialized variable warning (with gcc 9.3.1) from
16217 cgen-run.c, like this:
16219 /tmp/build/sim/../../src/sim/cris/../common/cgen-run.c: In function ‘sim_resume’:
16220 /tmp/build/sim/../../src/sim/cris/../common/cgen-run.c:259:5: warning: ‘engine_fns$’ may be used uninitialized in this function [-Wmaybe-uninitialized]
16221 259 | (* engine_fns[next_cpu_nr]) (cpu);
16222 | ~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
16223 /tmp/build/sim/../../src/sim/cris/../common/cgen-run.c:232:14: note: ‘engine_fns$’ was declared here
16224 232 | ENGINE_FN *engine_fns[MAX_NR_PROCESSORS];
16227 This is a false positive - we over allocate engine_fn, and then only
16228 initialize the nr_cpus entries which we will later go on to use.
16230 However, we can easily silence this warning by initializing the unused
16231 entries in engine_fns to NULL, this might also help if anyone ever
16232 looks at engine_fns in a debugger, it should now be obvious which
16233 entries are in use, and which are not.
16235 With this change the warning is gone.
16237 There should be no change in behaviour with this commit.
16239 2022-10-19 Alan Modra <amodra@gmail.com>
16241 Fix addr2line test for ppc64 elfv1 and mingw
16242 * testsuite/binutils-all/addr2line.exp: Tidy. For powerpc64
16243 arrange to pass --synthetic to nm, and extract .main and .fn
16244 symbol address for addr2line test. Handle default executable
16245 extension on cygwin/mingw compilers.
16247 2022-10-19 Nick Clifton <nickc@redhat.com>
16249 Update MAINTAINERS file with details about accepting DCO signed contributions.
16250 * MAINTAINERS: Add section on patches, copyright and DCO.
16252 2022-10-19 Andrew Burgess <aburgess@redhat.com>
16254 gdb/testsuite: avoid temporary file in gdb/testsuite (unittest.exp)
16255 I spotted that the gdb.gdb/unittest.exp script causes a temporary file
16256 inserters_extractors-2.txt to be created in build/gdb/testsuite/
16257 instead of in build/gdb/testsuite/output/gdb.gdb/unittest/.
16259 This is because some of the 'maint selftest' tests create temporary
16260 files in GDB's current directory, specifically, the two source files:
16262 gdb/unittests/basic_string_view/inserters/wchar_t/2.cc
16263 gdb/unittests/basic_string_view/inserters/char/2.cc
16265 both create a temporary file called inserters_extractors-2.txt, though
16266 we only run the second of these as part of GDB's selftests.
16268 I initially proposed just using GDB's 'cd' command in unittest.exp to
16269 switch to the test output directory before running the selftests,
16270 however, Pedro pointed out that there was a risk here that, if GDB
16271 crashed during shutdown, the generated core file would be left in the
16272 test output directory rather than in the testsuite directory. As a
16273 result, our clever core file spotting logic would fail to spot the
16274 core file and alert the user.
16276 Instead, I propose this slightly more involved solution. I've added a
16277 new with_gdb_cwd directory proc, used like this:
16279 with_gdb_cwd $directory {
16283 The new proc temporarily switches to $directory and then runs the
16284 tests within the block. After running the tests the previous current
16285 working directory is restored.
16287 Additionally, after switching back to the previous cwd, we check that
16288 GDB is still responsive. This means that if GDB crashed immediately
16289 prior to restoring the previous directory, and left the core file in
16290 the wrong place, then the responsiveness check will fail, and a FAIL
16291 will be emitted, this should be enough to alert the user that
16292 something has gone wrong.
16294 With this commit in place the unittest.exp script now leaves its
16295 temporary file in the test output directory.
16297 2022-10-19 Andrew Burgess <aburgess@redhat.com>
16299 gdb/testsuite: avoid creating files in gdb/testsuite directory
16300 I spotted that the test gdb.dwarf2/dw2-using-debug-str.exp was
16301 creating an output file called debug_str_section in the root
16302 build/gdb/testsuite directory instead of using the
16303 build/gdb/testsuite/output/gdb.dwarf2/dw2-using-debug-str/ directory.
16305 This appears to be caused by a missing '$' character. We setup a
16306 variable debug_str_section which contains a path within the output
16307 directory, but then when we build the objcopy command we use
16308 'debug_str_section' without a '$' prefix, as a result, we create the
16309 debug_str_section file.
16311 This commit adds the missing '$', the file is now created in the
16314 2022-10-19 Andrew Burgess <aburgess@redhat.com>
16316 bfd: fix undefined references to aarch64_pe_le_vec
16319 commit c60b3806799abf1d7f6cf5108a1b0e733a950b13
16320 Date: Wed Oct 19 10:57:12 2022 +0200
16322 aarch64-pe support for LD, GAS and BFD
16324 It appears that bfd/Makefile.in and bfd/configure were not regenerated
16325 correctly. The differences in the configure file are only whitespace,
16326 but in Makefile.in a critical reference to pe-aarch64.lo was missing.
16328 2022-10-19 Jedidiah Thompson <wej22007@outlook.com>
16329 Jedidiah Thompson <wej22007@outlook.com>
16330 Zac Walker <zac.walker@linaro.org>
16332 aarch64-pe support for LD, GAS and BFD
16333 Allows aarch64-pe to be targeted natively, not having to use objcopy to convert it from ELF to PE.
16334 Based on initial work by Jedidiah Thompson
16336 2022-10-19 rupesh potharla <rupesh.potharla@amd.com>
16338 Binutils: Adding new testcase for addr2line.
16339 * binutils/testsuite/config/default.exp: Set ADDR2LINE and ADDR2LINEFLAGS.
16340 * binutils/testsuite/binutils-all/addr2line.exp: New file.
16342 2022-10-19 Tom de Vries <tdevries@suse.de>
16344 [gdb/testsuite] Fix ERROR in gdb.python/py-breakpoint.exp
16345 With test-case gdb.python/py-breakpoint.exp I run into:
16347 (gdb) ERROR: tcl error sourcing gdb.python/py-breakpoint.exp.
16348 ERROR: can't read "skip_hw_watchpoint_tests_p": no such variable
16350 "if {$skip_hw_watchpoint_tests_p} {
16351 gdb_test_no_output "set can-use-hw-watchpoints 0" ""
16355 Fix this by adding the missing "global skip_hw_watchpoint_tests_p" in two
16358 Tested on x86_64-linux.
16360 2022-10-19 Andreas Krebbel <krebbel@linux.ibm.com>
16362 IBM zSystems: Issue error for *DBL relocs on misaligned symbols
16363 Relocs like PC32DBL require a right shift of the symbol value. There
16364 is no situation where dropping symbol value bits with the right shift
16365 is a good thing. Hence we now issue an error to detect such problems.
16367 2022-10-19 Simon Marchi <simon.marchi@efficios.com>
16369 gdb: check for groups with duplicate names in reggroups:add
16370 In the downstream ROCm GDB port, we would create multiple register
16371 groups with duplicate names. While it did not really hurt, it certainly
16372 wasn't the intent. And I don't think it ever makes sense to do so.
16374 To catch these, change the assert in reggroups::add to check for
16375 duplicate names. It's no longer necessary to check for duplicate
16376 reggroup pointers, because adding the same group twice would be caught
16377 by the duplicate name check.
16379 Change-Id: Id216a58acf91f1b314d3cba2d02de73656f8851d
16380 Approved-By: Tom Tromey <tom@tromey.com>
16382 2022-10-19 GDB Administrator <gdbadmin@sourceware.org>
16384 Automatic date update in version.in
16386 2022-10-18 H.J. Lu <hjl.tools@gmail.com>
16388 x86: Disable AVX-VNNI when disabling AVX2
16389 Since AVX-VNNI requires AVX2, disable AVX-VNNI when disabling AVX2.
16391 * i386-gen.c (cpu_flag_init): Add CpuAVX_VNNI to
16392 CPU_ANY_AVX2_FLAGS.
16393 * i386-init.h: Regenerate.
16395 2022-10-18 Tom Tromey <tom@tromey.com>
16397 Remove dead code from py-finishbreakpoint.c
16398 PR python/16324 points out that comparing a frame id to null_frame_id
16399 can never succeed, and proposes simply removing the dead code. That
16400 is what this patch does.
16402 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=16324
16404 2022-10-18 Carl Love <cel@us.ibm.com>
16406 Update tests to use skip_hw_watchpoint_tests to test for HW watchpoint support.
16407 The hardware watchpoint check has been updated in a couple of recent
16408 patches. This patch updates the hardware watchpoint test in the remaining
16411 The issue is the PowerPC processors support hardware watchpoints with the
16412 exception of Power 9. The hardware watchpoint support is disabled on
16413 Power 9. The test skip_hw_watchpoint_tests must be used to correctly
16414 determine if the PowerPC processor supports hardware watchpoints.
16416 This patch fixes 6 test failures in test gdb.threads/watchpoint-fork.exp.
16418 Test gdb.base/watch-vfork.exp runs with can-use-hw-watchpoints set to
16419 true and false. When the test is run with can-use-hw-watchpoints set to
16420 true, gdb just falls back to using software watchpoints. The
16421 patch reduces the number of expected passes by 2 since because it now
16422 only runs once with can-use-hw-watchpoints set to false.
16424 Test gdb.mi/mi-watch.exp runs the test with argument hw and sw. If the
16425 argument is hw and hardware watchpoints are not supported the test exits.
16426 The number of expected passes is cut in half with the patch as it now only
16427 runs the test using software breakpoints. Previously the pass to use
16428 hardware watchpoints was not skipped and the test actually ran using
16429 software watchpoints.
16431 The following tests run the same with and without the patch. The tests
16432 are supposed to execute the gdb command "set can-use-hw-watchpoints 0" if
16433 the processor does not support hardware bwatchpoints. However the command
16434 was not being executed and gdb was falling back to using software
16435 watchpoints since the Power 9 watchpoint resource check fails. With the
16436 patch, the tests now execute the command and the test runs using software
16437 watchpoints as it did previously. The tests are:
16439 gdb.base/commands.exp
16440 gdb.base/cond-eval-mode.exp
16441 gdb.base/display.exp
16442 gdb.base/gdb11531.exp
16443 gdb.base/recurse.exp
16444 gdb.base/value-double-free.exp
16445 gdb.base/watch-bitfields.exp
16446 gdb.base/watch-cond-infcall.exp
16447 gdb.base/watch-cond.exp
16448 gdb.base/watchpoint-solib.exp
16449 gdb.base/watchpoints.exp
16451 The following two tests are not supported on the Power 9 system used to
16452 test the changes. The patch does not change the tests results for these
16455 gdb.python/py-breakpoint.exp
16456 gdb.mi/mi-watch-nonstop.exp
16458 2022-10-18 Tom de Vries <tdevries@suse.de>
16460 [gdb/testsuite] Handle header files with local-remote-host.exp
16461 With test-case gdb.base/included.exp and host board local-remote-host.exp with
16462 tentative fix for PR29697 I run into:
16464 included.c:18:10: fatal error: included.h: No such file or directory
16465 #include "included.h"
16467 compilation terminated.
16470 Fix this by adding the missing gdb_remote_download calls.
16472 Likewise in a few other test-cases.
16474 Tested on x86_64-linux.
16476 2022-10-18 Tom de Vries <tdevries@suse.de>
16478 [gdb/testsuite] Fix gdb.server/no-thread-db.exp with local-remote-host.exp
16479 With test-case gdb.server/no-thread-db.exp and host board local-remote-host.exp
16480 with a tentative fix for PR29697 I run into:
16483 Cannot find thread-local storage for Thread 29613.29613, executable file \
16484 $HOME/no-thread-db:^M
16485 Remote target failed to process qGetTLSAddr request^M
16486 (gdb) FAIL: gdb.server/no-thread-db.exp: print foo
16489 The regexp in the test-case expects the full $binfile pathname, but we have
16490 instead $HOME/no-thread-db.
16492 Fix this by making the regexp less strict.
16494 Tested on x86_64-linux.
16496 2022-10-18 Tom de Vries <tdevries@suse.de>
16498 [gdb/testsuite] Fix gdb.base/return-nodebug.exp with local-remote-host.exp
16499 With host board local-remote-host.exp and test-case
16500 gdb.base/return-nodebug.exp, I run into:
16502 Executing on host: gcc -fno-stack-protector -fdiagnostics-color=never \
16503 -DTYPE=signed\ char -c -g -o return-nodebug-signed-char0.o \
16504 /home/vries/gdb_versions/devel/src/gdb/testsuite/gdb.base/return-nodebug.c \
16506 builtin_spawn [open ...]^M
16507 gcc: error: char: No such file or directory
16510 Avoid the quoting problem by not using spaces in the define.
16512 Tested on x86_64-linux.
16514 2022-10-18 Tom de Vries <tdevries@suse.de>
16516 [gdb/testsuite] Fix gdb.server/file-transfer.exp with local-remote-host.exp
16517 When running test-case gdb.server/file-transfer.exp with host
16518 board local-remote-host.exp, I get:
16520 Executing on host: cmp -s $outputs/gdb.server/file-transfer/file-transfer \
16521 down-server (timeout = 300)
16522 builtin_spawn [open ...]^M
16524 FAIL: gdb.server/file-transfer.exp: compare intermediate binary file
16527 The remote host and remote target cases are handled here together here in proc
16528 test_file_transfer:
16530 if {![is_remote host] && ![is_remote target]} {
16531 set up_server [standard_output_file $up_server]
16532 set down_server [standard_output_file $down_server]
16536 Fix this by handling them separately.
16538 Tested on x86_64-linux.
16540 2022-10-18 Tom de Vries <tdevries@suse.de>
16542 [gdb/testsuite] Update boards/README
16543 Update gdb/testsuite/boards/README to reflect recent commit c4c8c27263d
16544 ("[gdb/testsuite] Fix host board local-remote-host-notty.exp timeouts"), which
16545 means the board now uses a pseudo-tty, but with editing disabled.
16547 2022-10-18 Markus Metzger <markus.t.metzger@intel.com>
16549 gdb, solib-svr4: support namespaces in DSO iteration
16550 When looking up names, GDB needs to stay within one linker namespace to
16551 find the correct instance in case the same name is provided in more than
16554 Modify svr4_iterate_over_objfiles_in_search_order() to stay within the
16555 namespace of the current_objfile argument. If no current_objfile is
16556 provided (i.e. it is nullptr), iterate over objfiles in the initial
16559 For objfiles that do not have a corresponding so_list to provide the
16560 namespace, assume that the objfile was loaded into the initial namespace.
16561 This would cover the main executable objfile (which is indeed loaded into
16562 the initial namespace) as well as manually added symbol files.
16566 - gdb.base/non-lazy-array-index.exp: the expression parser may lookup
16567 global symbols, which may result in xfers to read auxv for determining
16568 the debug base as part of svr4_iterate_over_objfiles_in_search_order().
16570 - gdb.server/non-lazy-array-index.exp: symbol lookup may access the
16571 target to read AUXV in order to determine the debug base for SVR4
16576 - get_symbol_address() and get_msymbol_address() search objfiles for a
16577 'better' match. This was introduced by
16579 4b610737f02 Handle copy relocations
16581 to handle copy relocations but it now causes a wrong address to be
16582 read after symbol lookup actually cound the correct symbol. This can
16583 be seen, for example, with gdb.base/dlmopen.exp when compiled with
16586 - gnu ifuncs are only looked up in the initial namespace.
16588 - lookup_minimal_symbol() and lookup_minimal_symbol_text() directly
16589 iterate over objfiles and are not aware of linker namespaces.
16591 2022-10-18 Markus Metzger <markus.t.metzger@intel.com>
16593 gdb: update gnu ifunc resolve
16594 Update elf_gnu_ifunc_resolve_by_cache() and elf_gnu_ifunc_resolve_by_got()
16595 to use gdbarch_iterate_over_objfiles_in_search_order() in order to
16596 restrict the objfile traversal to the initial namespace.
16598 In order to extend this to other namespaces, we'd need to provide context,
16599 e.g. via an objfile inside that namespace.
16601 2022-10-18 Markus Metzger <markus.t.metzger@intel.com>
16603 gdb, symtab: inline find_quick_global_symbol_language
16604 There is only one use of find_quick_global_symbol_language that calls it
16605 for the special symbol "main".
16607 Inline the function as it is probably not correct in the general case
16608 where we may have multiple instances of global symbols with the same name
16609 but different languages in different libraries in different linker
16612 Further, change the objfiles iteration into a call to
16613 gdbarch_iterate_over_objfiles_in_search_order, which would only search the
16614 initial linker namespace, where we expect "main" to be located.
16616 2022-10-18 Markus Metzger <markus.t.metzger@intel.com>
16618 gdb, hppa: remove unused hppa_lookup_stub_minimal_symbol
16619 I stumbled over this while reviewing all objfiles traversals with regards
16620 to impact of linker namespaces.
16622 Recursive grep only finds two occurrences of hppa_lookup_stub_minimal_symbol:
16623 - the declaration in hppa-tdep.h.
16624 - the definition in hppa-tdep.c.
16626 There appear to be no calls to this function. Remove it.
16628 2022-10-18 Markus Metzger <markus.t.metzger@intel.com>
16630 gdb, cp: update add_symbol_overload_list_qualified
16631 Iterate over objfiles in search order using the objfile of the selected
16632 block as current_objfile so the iteration can stay inside the block's
16635 fixup! gdb, ada: update ada_lookup_simple_minsym
16636 remove get_selected_block()
16638 gdb, ada: update ada_lookup_simple_minsym
16639 Iterate over objfile in search order using the objfile of the context
16640 block as current_objfile so the iteration can stay inside the block's
16643 gdb, ada: collect standard exceptions in all objfiles
16644 When searching for standard exceptions for Ada, we lookup the minimal
16645 symbol of each exception. With linker namespaces there can be multiple
16646 instances in different namespaces. Collect them all.
16648 2022-10-18 Markus Metzger <markus.t.metzger@intel.com>
16650 gdb, python: use gdbarch_iterate_over_objfiles_in_search_order
16651 The implementation of gdb.lookup_objfile() iterates over all objfiles and
16652 compares their name or build id to the user-provided search string.
16654 This will cause problems when supporting linker namespaces as the first
16655 objfile in any namespace will be found. Instead, use
16656 gdbarch_iterate_over_objfiles_in_search_order to only consider the
16657 namespace of gdb.current_objfile() for the search, which defaults to the
16658 initial namespace when gdb.current_objfile() is None.
16660 2022-10-18 Markus Metzger <markus.t.metzger@intel.com>
16662 gdb, compile: unlink objfile stored in module
16663 When cleaning up after a compile command, we iterate over all objfiles and
16664 unlink the first objfile with the same name as the one we compiled.
16666 Since we already store a pointer to that objfile in the module and use it
16667 to get the name we're comparing against, there's no reason to iterate, at
16668 all. We can simply use that objfile.
16670 This further avoids potential issues when an objfile with the same name is
16671 loaded into a different linker namespace.
16673 2022-10-18 Markus Metzger <markus.t.metzger@intel.com>
16675 gdb, gdbserver: extend RSP to support namespaces
16676 Introduce a new qXfer:libraries-svr4:read annex key/value pair
16678 lmid=<namespace identifier>
16680 to be used together with start and prev to provide the namespace of start
16681 and prev to gdbserver.
16683 Unknown key/value pairs are ignored by gdbserver so no new supports check
16686 Introduce a new library-list-svr4 library attribute
16690 to provide the namespace of a library entry to GDB.
16692 This implementation uses the address of a namespace's r_debug object as
16693 namespace identifier.
16695 This should have incremented the minor version but since unknown XML
16696 attributes are ignored, anyway, and since changing the version results in
16697 a warning from GDB, the version is left at 1.0.
16699 2022-10-18 Markus Metzger <markus.t.metzger@intel.com>
16701 gdbserver: move main_lm handling into caller
16702 When listing SVR4 shared libraries, special care has to be taken about the
16703 first library in the default namespace as that refers to the main
16704 executable. The load map address of this main executable is provided in
16705 an attribute of the library-list-svr4 element.
16707 Move that code from where we enumerate libraries inside a single namespace
16708 to where we generate the rest of the library-list-svr4 element. This
16709 allows us to complete the library-list-svr4 element inside one function.
16711 There should be no functional change.
16713 2022-10-18 Markus Metzger <markus.t.metzger@intel.com>
16714 Lu, Hongjiu <hongjiu.lu@intel.com>
16716 gdb, gdbserver: support dlmopen()
16717 In glibc, the r_debug structure contains (amongst others) the following
16721 Version number for this protocol. It should be greater than 0.
16723 If r_version is 2, struct r_debug is extended to struct r_debug_extended
16724 with one additional field:
16726 struct r_debug_extended *r_next;
16727 Link to the next r_debug_extended structure. Each r_debug_extended
16728 structure represents a different namespace. The first r_debug_extended
16729 structure is for the default namespace.
16731 1. Change solib_svr4_r_map argument to take the debug base.
16732 2. Add solib_svr4_r_next to find the link map in the next namespace from
16734 3. Update svr4_current_sos_direct to get the link map in the next namespace
16735 from the r_next field.
16736 4. Don't check shared libraries in other namespaces when updating shared
16737 libraries in a new namespace.
16738 5. Update svr4_same to check the load offset in addition to the name
16739 6. Update svr4_default_sos to also set l_addr_inferior
16740 7. Change the flat solib_list into a per-namespace list using the
16741 namespace's r_debug address to identify the namespace.
16743 Add gdb.base/dlmopen.exp to test this.
16745 To remain backwards compatible with older gdbserver, we reserve the
16746 namespace zero for a flat list of solibs from all namespaces. Subsequent
16747 patches will extend RSP to allow listing libraries grouped by namespace.
16749 This fixes PR 11839.
16751 2022-10-18 Markus Metzger <markus.t.metzger@intel.com>
16753 gdb, solib-svr4: remove locate_base()
16754 Whenever we call locate_base(), we clear info->debug_base directly before
16755 the call. Thus, we never cache the base location as locate_base() had
16758 Move the svr4_have_link_map_offsets() check into elf_locate_base(), inline
16759 locate_base() at all call sites, and remove it.
16761 2022-10-18 Markus Metzger <markus.t.metzger@intel.com>
16763 gdb, testsuite: extend gdb_test_multiple checks
16766 warning: Corrupted shared library list
16771 warning: Probes-based dynamic linker interface failed.
16772 Reverting to original interface.
16774 in gdb_test_multiple.
16776 2022-10-18 Jan Beulich <jbeulich@suse.com>
16778 x86: generalize gas documentation for disabling of ISA extensions
16779 As of commit ae89daecb132 ("x86: generalize disabling of sub-
16780 architectures") there's no arbitrary subset of ISAs which can also be
16781 disabled. This should have been reflected in documentation right away.
16782 Since I failed to do so, correct this now.
16784 x86: correct CPU_AMX_{BF16,INT8}_FLAGS
16785 AMX-TILE is a prereq to these, as already correctly expressed by
16786 CPU_ANY_AMX_TILE_FLAGS. Express the dependency also in the reverse
16787 ("positive") direction.
16789 2022-10-18 GDB Administrator <gdbadmin@sourceware.org>
16791 Automatic date update in version.in
16793 2022-10-17 Tom Tromey <tromey@adacore.com>
16795 kfail an Ada test for GCC < 12
16796 I noticed one particular Ada test was failing on Fedora 34, but works
16797 when I switch to GCC 12. This patch arranges to kfail the test when
16798 an older compiler is used.
16800 I tested this with GCC 11, 12, and 13. I'm going to check it in.
16802 2022-10-17 Tom Tromey <tromey@adacore.com>
16804 Remove a nullptr check in DWARF scanner
16805 In scan_attributes, The DWARF scanner checks whether maybe_defer is
16806 nullptr, but this can never happen. This patch removes the check.
16808 2022-10-17 Pedro Alves <pedro@palves.net>
16810 gdbarch-components.py: Remove spurious space from "frame_info_ptr " params
16811 If you run gdbarch.py today, you'll get local modifications compared
16812 to what's in the tree, like:
16814 --- c/gdb/gdbarch-gen.h
16815 +++ w/gdb/gdbarch-gen.h
16816 @@ -315,8 +315,8 @@ extern void set_gdbarch_register_type (struct gdbarch *gdbarch, gdbarch_register
16817 should match the address at which the breakpoint was set in the dummy
16820 -typedef struct frame_id (gdbarch_dummy_id_ftype) (struct gdbarch *gdbarch, frame_info_ptr this_frame);
16821 -extern struct frame_id gdbarch_dummy_id (struct gdbarch *gdbarch, frame_info_ptr this_frame);
16822 +typedef struct frame_id (gdbarch_dummy_id_ftype) (struct gdbarch *gdbarch, frame_info_ptr this_frame);
16823 +extern struct frame_id gdbarch_dummy_id (struct gdbarch *gdbarch, frame_info_ptr this_frame);
16824 extern void set_gdbarch_dummy_id (struct gdbarch *gdbarch, gdbarch_dummy_id_ftype *dummy_id);
16828 The extra space comes from the "frame_info_ptr " param that appears in
16829 a number of gdbarch methods in gdbarch-components.py. With the extra
16830 space removed, running ./gdbarch.py generates the exact code that's in
16833 Change-Id: If7d20b8c6b2fd9ff466142a01bd2611c9ef9f53e
16835 2022-10-17 Tom Tromey <tromey@adacore.com>
16837 Change .gdb_index de-duplication implementation
16838 While investigating PR symtab/29179, I found that one Ada test failed
16839 because, although a certain symbol was present in the index, with the
16840 new DWARF reader it pointed to a different CU than was chosen by
16841 earlier versions of gdb.
16843 This patch changes how symbol de-duplication is done, deferring the
16844 process until the entire symbol table has been constructed. This way,
16845 it's possible to always choose the lower-numbered CU among duplicates,
16846 which is how gdb (implicitly) previously worked.
16848 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29179
16850 2022-10-17 Tom Tromey <tromey@adacore.com>
16852 Improve Ada support in .gdb_index
16853 The cooked index work changed how .gdb_index is constructed, and in
16854 the process broke .gdb_index support. This is PR symtab/29179.
16856 This patch partially fixes the problem. It arranges for Ada names to
16857 be encoded in the form expected by the index code. In particular,
16858 linkage names for Ada are emitted, including the "main" name; names
16859 are Ada-encoded; and names are no longer case-folded, something that
16860 prevented operator names from round-tripping correctly.
16862 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29179
16864 2022-10-17 Tom Tromey <tromey@adacore.com>
16866 Don't add type linkage names to cooked index
16867 The compiler will sometimes emit a linkage name for a type, like:
16869 <1d3> DW_AT_linkage_name: (indirect string, offset: 0x106f): 11__mbstate_t
16871 These names aren't very useful, and this patch changes the DWARF
16872 reader so that they are ignored by the cooked index.
16874 2022-10-17 Tom Tromey <tromey@adacore.com>
16876 Fix regression in c-linkage-name.exp with gdb index
16877 c-linkage-name.exp started failing with the gdb-index target board due
16878 to an earlier patch. The problem here is that some linkage names must
16879 be in the index -- but, based on inspection, not C++ linkage names.
16880 This patch updates the code to exclude only these.
16882 2022-10-17 TaiseiIto <taisei1212@outlook.jp>
16884 Fix null pointer representations
16885 Since "NULL" and "0" are used to represent invalid address in function
16886 "gdbarch_find_by_info" in "binutils-gdb/gdb/arch-utils.c", I modified
16889 2022-10-17 Simon Marchi <simon.marchi@efficios.com>
16891 gdb: silence unused-but-set-variable warning about yynerrs in cp-name-parser.y
16892 When building with clang 15 on Ubuntu 20.04, I get:
16894 CXX cp-name-parser.o
16895 cp-name-parser.c.tmp:1777:9: error: variable 'cpnameyynerrs' set but not used [-Werror,-Wunused-but-set-variable]
16898 /home/smarchi/src/binutils-gdb/gdb/yy-remap.h:58:18: note: expanded from macro 'yynerrs'
16899 #define yynerrs GDB_YY_REMAP (yynerrs)
16901 /home/smarchi/src/binutils-gdb/gdb/yy-remap.h:40:29: note: expanded from macro 'GDB_YY_REMAP'
16902 #define GDB_YY_REMAP(YYSYM) GDB_YY_REMAP_1 (GDB_YY_REMAP_PREFIX, YYSYM)
16904 /home/smarchi/src/binutils-gdb/gdb/yy-remap.h:39:39: note: expanded from macro 'GDB_YY_REMAP_1'
16905 #define GDB_YY_REMAP_1(PREFIX, YYSYM) GDB_YY_REMAP_2 (PREFIX, YYSYM)
16907 /home/smarchi/src/binutils-gdb/gdb/yy-remap.h:38:39: note: expanded from macro 'GDB_YY_REMAP_2'
16908 #define GDB_YY_REMAP_2(PREFIX, YYSYM) PREFIX ## YYSYM
16910 <scratch space>:45:1: note: expanded from here
16914 This is because clang 15 warns for something like this:
16920 whereas previous versions do not.
16922 yynerrs is defined in yyparse and is there for actions to use. Since
16923 the actions in cp-name-parser.y don't use it, we get a warning. We see
16924 this problem on this particular .y file because it uses `%pure-parser`
16925 [1], which makes yynerrs a local rather than a global.
16927 I initially fixed this by using
16928 DIAGNOSTIC_IGNORE_UNUSED_BUT_SET_VARIABLE (like in commit f7aa1a5acc5
16929 ("gold: Suppress "unused" variable warning on Clang")), but then I
16930 realized we could suppress the warning in a more fine-grained way using
16935 [1] https://www.gnu.org/software/bison/manual/html_node/Error-Reporting-Function.html
16937 Change-Id: I6cae7a4207c19fe1b719e2ac19be69122ebe3af1
16939 2022-10-17 Clément Chigot <chigot@adacore.com>
16941 ld/testsuite: consistently add board_ldflags when linking with GCC
16942 Currently, the functions checking if the compiler is available or if a
16943 feature is available add both board_cflags and board_ldflags.
16944 However, functions running the tests only retrieve board_cflags. This
16945 can lead to unexpected errors when mandaratory flags are defined in
16946 board_ldflags and not board_cflags.
16950 * testsuite/ld-unique/unique.exp: Add board_ldflags when
16952 * testsuite/lib/ld-lib.exp: Likewise.
16954 2022-10-17 CaiJingtao <caijingtao@huawei.com>
16956 Allow explicit size specifier for predicate operand of {sq, uq, }{incp, decp}
16957 Omitting predicate size specifier in vector form of {sq, uq, }{decp, incp} is deprecated and will be prohibited in a future release of the aarch64,
16958 see https://developer.arm.com/documentation/ddi0602/2021-09/SVE-Instructions/DECP--vector---Decrement-vector-by-count-of-true-predicate-elements-.
16960 This allows explicit size specifier, e.g. `decp z0.h, p0.h`, for predicate operand of these SVE instructions.
16961 The existing behaviour of not requiring the specifier is preserved.
16962 And the disasembly is with the specifier with this patch.
16964 The GAS tests passed under our local tests.
16967 * aarch64-asm.c: Modify `sve_size_hsd` encoding.
16968 * aarch64-tbl.h (aarch64_opcode_table): Add QUALS's type OP_SVE_Vv_HSD
16969 for decp, incp, sqdecp, sqincp, uqdecp and uqincp.
16972 * testsuite/gas/aarch64/sve-movprfx_23.s: Update movprfx_23 testcase's
16973 test_sametwo macro, where take the predicate size specifier.
16974 * testsuite/gas/aarch64/sve-movprfx_23.d: Update movprfx_23 testcase's
16975 expected disassembly.
16976 * testsuite/gas/aarch64/sve-movprfx_23.l: Update movprfx_23 testcase's
16977 expected assembler messages.
16978 * testsuite/gas/aarch64/sve.s: Add sve testcase's instructions for
16979 decp, incp, sqdecp, sqincp, uqdecp and uqincp, which take the
16980 predicate size specifier.
16981 * testsuite/gas/aarch64/sve.d: Update sve testcase's expected
16984 2022-10-17 Richard Sandiford <richard.sandiford@arm.com>
16986 aarch64: Tweak handling of F_STRICT
16987 Current F_STRICT qualifier checking is enforced after the fact
16988 rather than as part of the match. This makes it impossible to
16997 * aarch64-opc.c (aarch64_find_best_match): Handle F_STRICT here
16999 (match_operands_qualifier): ...here.
17001 2022-10-17 Jan Beulich <jbeulich@suse.com>
17003 x86: properly decode EVEX.W for AVX512_4{FMAPS,VNNIW} insns
17004 These require EVEX.W=0. Use %XS to facilitate the checking, even if for
17005 the AVX512_4VNNIW ones this is kind of an abuse (as 's' there stands for
17006 "signed", not "single").
17008 While there also correct the 3rd operand for the AVX512_4VNNIW entries:
17009 Only the memory form is allowed (just like for AVX512_4FMAPS, where the
17010 correct type is already in use).
17012 2022-10-17 Jan Beulich <jbeulich@suse.com>
17014 x86: fold AVX512-VNNI disassembler entries with AVX-VNNI ones
17015 Make %XV also print the separating blank in the VEX case, while making
17016 it do nothing for EVEX-encoded insns. This way the AVX-VNNI entries
17017 can be re-used for AVX512-VNNI, at the same time fixing the lack of
17020 For the AVX-VNNI ones further make sure only VEX.66 forms are actually
17023 2022-10-17 GDB Administrator <gdbadmin@sourceware.org>
17025 Automatic date update in version.in
17027 2022-10-16 Tom Tromey <tom@tromey.com>
17029 More uses of checked_static_cast
17030 This patch changes a few more uses of static_cast to use
17031 checked_static_cast. In this patch, cast-to-references are converted
17032 by moving the dereference outside of the cast, as checked_static_cast
17033 only handles pointers.
17035 2022-10-16 Tom Tromey <tom@tromey.com>
17037 Use checked_static_cast in more places
17038 I looked through all the uses of static_cast<... *> in gdb and
17039 converted many of them to checked_static_cast.
17041 I couldn't test a few of these changes.
17043 2022-10-16 Alan Modra <amodra@gmail.com>
17045 PowerPC se_rfmci and VLE, SPE2 and LSP insns with -many
17046 I noticed recently that se_rfmci, a VLE mode instruction, was being
17047 accepted by non-VLE cpus, and also that se_rfmci by itself in a
17048 section did not cause SHF_PPC_VLE to be set. ie. both testcases added
17049 by this patch fail without the changes to tc-ppc.c here.
17051 Also, VLE, SPE2 and LSP insns were not accepted by the assembler with
17052 -many nor were SPE2 and LSP being disassembled with -Many.
17055 * config/tc-ppc.c (ppc_setup_opcodes): Wrap long lines. Add
17056 vle_opcodes when PPC_OPCODE_VLE or PPC_OPCODE_ANY. Simplify
17057 disassembler index segment checks. Add LSP and SPE2 opcodes
17058 when PPC_OPCODE_ANY too.
17059 (md_assemble): Correct logic adding PPC_APUINFO_VLE and
17061 * testsuite/gas/ppc/se_rfmci.s
17062 * testsuite/gas/ppc/se_rfmci.d,
17063 * testsuite/gas/ppc/se_rfmci_bad.d: New tests.
17064 * testsuite/gas/ppc/ppc.exp: Run them.
17066 * ppc-dis.c (print_insn_powerpc): Disassemble SPE2 and LSP insn
17068 * ppc-opc.c (vle_opcodes <se_rfmci>): Comment.
17070 2022-10-16 Alan Modra <amodra@gmail.com>
17072 zlib-gabi to zstd woes
17073 So we had a zlib-gabi .debug_info section that increased in size with
17074 zstd, so much so that it was better to leave the section
17075 uncompressed. Things went horribly wrong when the section was read
17076 again later. The section was read again off disk using the
17077 uncompressed size. So you get the zlib section again with some
17078 garbage at the end. Fix that particular problem by setting the
17079 section flag SEC_IN_MEMORY. Any future read will get sec->contents.
17081 Also, if the section is to be left uncompressed, the input
17082 SHF_COMPRESSED flag needs to be reset otherwise objcopy will copy it
17085 Finally, bfd_convert_section_contents needed a small update to handle
17086 zstd compressed sections, and I've deleted bfd_cache_section_contents.
17088 * bfd.c (bfd_convert_section_contents): Handle zstd.
17089 * compress.c (bfd_compress_section_contents): When section
17090 contents are uncompressed set SEC_IN_MEMORY flag,
17091 compress_status to COMRESS_SECTION_NONE, and clear
17092 SHF_COMPRESSED. Set SEC_IN_MEMORY for compressed contents.
17093 (bfd_get_full_section_contents): Don't check section size
17094 against file size when SEC_IN_MEMORY.
17095 (bfd_cache_section_contents): Delete function.
17096 * elf32-arm.c (elf32_arm_get_synthetic_symtab): Expand
17097 bfd_cache_section_contents here.
17098 * bfd-in2.h: Regenerate.
17100 2022-10-16 GDB Administrator <gdbadmin@sourceware.org>
17102 Automatic date update in version.in
17104 2022-10-15 Torbjörn SVENSSON <torbjorn.svensson@foss.st.com>
17106 gdb/arm: Don't rely on loop detection to stop unwinding
17107 Setting SP of the next frame to the same address as the current frame
17108 is an ugly way to stop the unwinding. A cleaner way is to rely on
17109 the frame_unwind_stop_reason function to return UNWIND_OUTERMOST.
17111 2022-10-15 GDB Administrator <gdbadmin@sourceware.org>
17113 Automatic date update in version.in
17115 2022-10-14 Tom de Vries <tdevries@suse.de>
17117 [gdb/testsuite] Add boards/README
17118 Add a file gdb/testsuite/boards/README, to make it easier to get a high-level
17119 overview of the various boards.
17121 2022-10-14 Tom de Vries <tdevries@suse.de>
17123 [gdb/contrib] Handle STRIP_ARGS_{STRIP,KEEP}_DEBUG in cc-with-tweaks.sh
17124 Handle new environment variable STRIP_ARGS_STRIP_DEBUG, defaulting to
17125 --strip-debug in gdb/contrib/cc-with-tweaks.sh, such that we can easily
17126 reproduce the PR29277 assert using:
17128 $ export STRIP_ARGS_STRIP_DEBUG=--strip-all
17129 $ make check RUNTESTFLAGS="gdb.base/jit-reader.exp \
17130 --target_board cc-with-gnu-debuglink"
17133 For completeness sake and to avoid confusion about which of the two used strip
17134 invocations the passed args apply to, likewise add STRIP_ARGS_KEEP_DEBUG,
17135 defaulting to --only-keep-debug.
17137 Script checked with shellcheck, no new warnings added.
17139 Tested on x86_64-linux.
17141 2022-10-14 Tom de Vries <tdevries@suse.de>
17143 [gdb] Fix heap-buffer-overflow in find_program_interpreter
17144 With the test-case included in this patch, we run into:
17146 (gdb) target remote localhost:2347^M
17147 `target:twice-connect' has disappeared; keeping its symbols.^M
17148 Remote debugging using localhost:2347^M
17149 warning: Unable to find dynamic linker breakpoint function.^M
17150 GDB will be unable to debug shared library initializers^M
17151 and track explicitly loaded dynamic code.^M
17152 Reading /usr/lib/debug/.build-id/$hex/$hex.debug from remote target...^M
17153 0x00007ffff7dd4550 in ?? ()^M
17154 (gdb) PASS: gdb.server/twice-connect.exp: session=second: gdbserver started
17155 FAIL: gdb.server/twice-connect.exp: found interpreter
17158 The problem originates in find_program_interpreter, where
17159 bfd_get_section_contents is called to read .interp, but fails. The function
17160 returns false but the result is ignored, so find_program_interpreter returns
17161 some random string.
17163 Fix this by checking the result of the call to bfd_get_section_contents.
17165 Tested on x86_64-linux.
17167 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29652
17169 2022-10-14 Tom de Vries <tdevries@suse.de>
17171 [gdb/testsuite] Fix gdb.server/unittest.exp with host board local-remote-host.exp
17172 With test-case gdb.server/unittest.exp and host board local-remote-host.exp I
17175 builtin_spawn build/gdbserver/gdbserver --selftest^M
17176 ERROR: : spawn id exp7 not open
17179 -i exp7 -timeout 10
17180 -i $server_spawn_id
17181 -re "Ran ($decimal) unit tests, 0 failed" {
17182 set num_ran $expect_out(1,string)
17184 ("uplevel" body line 1)
17185 invoked from within
17186 "uplevel $body" NONE : spawn id exp7 not open
17187 UNRESOLVED: gdb.server/unittest.exp: unit tests
17190 The problem is (as fixed for avr in commit df5b8876083 ("gdb/testsuite: better
17191 handle failures in simavr board, reap simavr process")), that gdb_expect through
17192 remote_expect adds a "-i <gdb spawn id> -timeout 10", which is the one causing
17195 As in aforementioned commit, fix this by using expect instead.
17197 Tested on x86_64-linux.
17199 2022-10-14 Tom de Vries <tdevries@suse.de>
17201 [gdb/testsuite] Fix host board local-remote-host-notty.exp timeouts
17202 With test-case gdb.server/stop-reply-no-thread-multi.exp and host board
17203 local-remote-host-notty.exp we occasionally run into a silent out, due to
17207 (gdb) The program is not being run.^M
17209 instead of the expected:
17212 The program is not being run.^M
17216 Likewise, we occasionally run into a nonsilent timeout:
17219 (gdb) You can't do that when your target is `exec'^M
17220 FAIL: gdb.server/stop-reply-no-thread.exp: to_disable=Tthread: t_nonstop=on: \
17221 disconnect (timeout)
17224 Typically, this results in the test-case taking more than two minutes to run.
17226 The problem can be reproduced using just:
17228 $ ssh -l $USER 127.0.0.1 gdb -q -ex kill
17231 Note that ssh by default uses -T which disables pseudo-tty allocation (as
17232 opposed to -t which forces pseudo-tty allocation):
17234 $ ssh -l $USER 127.0.0.1 -T tty
17236 $ ssh -l $USER 127.0.0.1 -t tty
17238 Connection to 127.0.0.1 closed.
17240 and according to https://stackoverflow.com/a/63241102 the behaviour we're
17241 seeing is specific to using '-T'.
17243 The related host board local-remote-host.exp does use '-t', and the only
17244 difference between the two boards mentioned is whether editing is on or off.
17247 - moving the content of local-remote-host-notty.exp into
17248 local-remote-host.exp
17249 - consequently, extending the copyright years in local-remote-host.exp
17250 - including local-remote-host.exp in local-remote-host-notty.exp
17251 (making local-remote-host-notty.exp use '-t')
17252 - adding -iex "set editing off" to GDBFLAGS in local-remote-host-notty.exp
17254 This results in the test-case taking just 6 seconds to run.
17256 Tested on x86_64-linux.
17258 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29669
17260 2022-10-14 Tom de Vries <tdevries@suse.de>
17262 [gdb/testsuite] Disable styling in host board local-remote-host.exp
17263 With test-case gdb.server/stop-reply-no-thread.exp and host board
17264 local-remote-host.exp, I run into:
17266 Breakpoint 1, ^[[33mmain^[[m () at ^[[32mstop-reply-no-thread.c^[[m:21^M
17267 21 ^[[01;34mreturn^[[m ^[[35m0^[[m^[[31m;^[[m^M
17268 (gdb) FAIL: gdb.server/stop-reply-no-thread.exp: to_disable=: t_nonstop=off: \
17272 The problem is that styling is enabled, and that is causing a regexp mismatch.
17274 With native, styling is disabled in default_gdb_init by doing
17275 'setenv TERM "dumb"', but that only has effect because the build (where we
17276 execute runtest, and consequently the setenv) and the host (where we execute
17277 gdb) are the same. For this host board however, gdb executes on a remote
17278 host, and the setenv has no effect.
17280 We could try to make some generic way to set TERM on the host, but for the
17281 purposes of this test-case it seems sufficient to just add:
17283 set GDBFLAGS "${GDBFLAGS} -iex \"set style enabled off\""
17285 so let's go with that for now.
17287 Tested on x86_64-linux.
17289 2022-10-14 Tom Tromey <tromey@adacore.com>
17291 Use scoped_value_mark in more places
17292 I looked at all the spots using value_mark, and converted all the
17293 straightforward ones to use scoped_value_mark instead.
17295 Regression tested on x86-64 Fedora 34.
17297 2022-10-14 Torbjörn SVENSSON <torbjorn.svensson@foss.st.com>
17299 gdb/arm: Stop unwinding on error, but do not assert
17300 When it's impossible to read the FPCCR and XPSR, the unwinding is
17301 unpredictable as the it's not possible to determine the correct
17302 frame size or padding.
17303 The only sane thing to do in this condition is to stop the unwinding.
17305 Example session without this patch:
17308 #0 SVC_Handler () at .../GPIO/GPIO_EXTI/Src/stm32f4xx_it.c:112
17309 .../gdb/arm-tdep.c:3594: internal-error: arm_m_exception_cache: Assertion `safe_read_memory_unsigned_integer (FPCCR, ARM_INT_REGISTER_SIZE, byte_order, &fpccr)' failed.
17310 A problem internal to GDB has been detected,
17311 further debugging may prove unreliable.
17312 ----- Backtrace -----
17313 0x5583bfb2a157 gdb_internal_backtrace_1
17315 ---------------------
17317 This is a bug, please report it. For instructions, see:
17318 <https://www.gnu.org/software/gdb/bugs/>.
17320 Aborted (core dumped)
17322 Example session with this patch:
17325 #0 SVC_Handler () at .../GPIO/GPIO_EXTI/Src/stm32f4xx_it.c:112
17326 warning: Could not fetch required FPCCR content. Further unwind is impossible.
17327 #1 <signal handler called>
17330 Reviewed-by: Pedro Alves <pedro@palves.net>
17332 2022-10-14 Alan Modra <amodra@gmail.com>
17334 PowerPC SPE disassembly and tests
17335 Where sub and subf forms of an instruction exist we generally
17336 disassemble to the extended insn sub form rather than the underlying
17337 machine subf instruction. Do so for SPE evsubw and evsubiw too.
17339 spe_ambiguous.d always was a bit too optimistic. There is no sensible
17340 way to disassemble identical bytes back to different and original
17341 source. Instead change the test to check -Mraw results.
17344 * testsuite/gas/ppc/ppc.exp: Run spe_ambiguous test.
17345 * testsuite/gas/ppc/spe.d: Expect evsubw and evsubiw rather than
17346 evsubfw and evsubifw.
17347 * testsuite/gas/ppc/spe_ambiguous.s: Test evnor form equivalent
17349 * testsuite/gas/ppc/spe_ambiguous.d: Test Mraw.
17351 * ppc-opc.c (powerpc_opcodes): Move evsubw before evsubfw and
17352 evsubiw before evsubifw and mark EXT.
17354 2022-10-14 Alan Modra <amodra@gmail.com>
17357 It has bothered me for a long time that we have disabled LSP (and SPE)
17358 tests. Also the LSP test comment indicating there is something wrong
17359 with get_powerpc_dialect. I don't think there is. Decoding of a VLE
17360 instruction depends on whether the processor is in VLE mode (some
17361 processors support both VLE and standard PPC) which we flag per
17362 section with SHF_PPC_VLE for decoding when disassembling.
17364 Background: Some versions of powerpc e200 have "Lightweight Signal
17365 Processing" support, examples being e200z215 and e200z425. As far as
17366 I can tell, LSP and SPE are mutually exclusive. This seems to be
17367 borne out by insn encoding, for example LSP "zvaddih" and SPE "evaddw"
17368 have the same encoding. So none of the processor descriptions in
17369 ppc_opts ought to have both PPC_OPCODE_LSP and PPC_OPCODE_SPE/2, if we
17370 want disassembly to work. I also could not find anything to suggest
17371 that the LSP insns are enabled only in VLE mode, which means the LSP
17372 insns should not be in vle_opcodes.
17374 Fix all this by moving the LSP insns to their own table, and add a new
17375 e200z2 cpu entry with LSP support, removing LSP from -me200z4 and from
17376 -mvle. (Yes, I know, as I said above some of the e200z4 processors
17377 have LSP. Others have SPE. It's hard to choose good options. Think
17378 of z2 as meaning earlier, z4 as later.) Also add -mlsp to allow
17379 adding the LSP insn set.
17382 * opcode/ppc.h (lsp_opcodes, lsp_num_opcodes): Declare.
17383 (LSP_OP_TO_SEG): Define.
17385 * doc/binutils.texi: Update ppc docs.
17387 * config/tc-ppc.c (ppc_setup_opcodes): Add lsp opcodes to ppc_hash.
17388 * doc/c-ppc.texi: Document e200 and lsp.
17389 * testsuite/gas/ppc/lsp-checks.d: Assemble with -me200z2.
17390 * testsuite/gas/ppc/lsp.d: Likewise, disassembly too.
17391 * testsuite/gas/ppc/ppc.exp: Don't xfail lsp test.
17393 * ppc-dis.c (ppc_opts): Add e200z2 and lsp. Don't set
17394 PPC_OPCODE_LSP for e200z4 or vle.
17395 (ppc_parse_cpu): Mutually exclude LSP and SPE.
17396 (LSP_OPCD_SEGS): Define.
17397 (lsp_opcd_indices): New array.
17398 (disassemble_init_powerpc): Init lsp_opcd_indices.
17399 (lookup_lsp): New function.
17400 (print_insn_powerpc): Call it.
17401 * ppc-opc.c: Include libiberty.h for ARRAY_SIZE and use throughout.
17402 (vle_opcodes): Move LSP opcodes to..
17403 (lsp_opcodes): ..here, and sort.
17404 (lsp_num_opcodes): New.
17406 2022-10-14 Alan Modra <amodra@gmail.com>
17408 PR29677, Field `the_bfd` of `asymbol` is uninitialised
17409 Besides not initialising the_bfd of synthetic symbols, counting
17410 symbols when sizing didn't match symbols created if there were any
17411 dynsyms named "". We don't want synthetic symbols without names
17412 anyway, so get rid of them. Also, simplify and correct sanity checks.
17415 * mach-o.c (bfd_mach_o_get_synthetic_symtab): Rewrite.
17417 2022-10-14 Tom de Vries <tdevries@suse.de>
17419 [gdb/testsuite] Drop unnecessary -Wl,-soname in gdb.base/skip-solib.exp
17420 I noticed in gdb.base/skip-solib.exp:
17422 if {[gdb_compile_shlib ${srcdir}/${subdir}/${srcfile_lib} ${binfile_lib} \
17423 [list debug -Wl,-soname,${libname}.so]] != ""} {
17427 that the -Wl,-soname argument is missing an ldflags= prefix, but adding it
17428 gives us a duplicate:
17430 Executing on host: gcc -fno-stack-protector \
17431 outputs/gdb.base/skip-solib/skip-solib-lib.c.o -fdiagnostics-color=never \
17432 -shared -g -Wl,-soname,libskip-solib.so -Wl,-soname,libskip-solib.so -lm \
17433 -o outputs/gdb.base/skip-solib/libskip-solib.so (timeout = 300)
17435 so apparently it's taken care of by gdb_compile_shlib.
17437 Drop the inactive and also unnecessary -Wl,-soname,${libname}.so from the
17438 flags list for the gdb_compile_shlib call.
17440 Tested on x86_64-linux.
17442 2022-10-14 Tom de Vries <tdevries@suse.de>
17444 [gdb/testsuite] Fix gdb.base/infoline-reloc-main-from-zero.exp with PIE
17445 With test-case gdb.base/infoline-reloc-main-from-zero.exp and target board
17446 unix/-fPIE/-pie I run into:
17448 gdb compile failed, ld: infoline-reloc-main-from-zero: error: \
17449 PHDR segment not covered by LOAD segment
17450 collect2: error: ld returned 1 exit status
17453 When running with native, I find that the executable is static:
17455 $ file infoline-reloc-main-from-zero
17456 infoline-reloc-main-from-zero: ELF 64-bit LSB executable, x86-64, \
17457 version 1 (SYSV), statically linked, BuildID[sha1]=$hex, with debug_info, \
17460 despite not having been compiled with -static.
17462 Fix the compilation by adding -static to the compilation flags.
17464 Tested on x86_64-linux.
17466 2022-10-14 Tom de Vries <tdevries@suse.de>
17468 [gdb/testsuite] Fix gdb.base/infoline-reloc-main-from-zero.exp with clang
17469 With test-case gdb.base/infoline-reloc-main-from-zero.exp and clang I run into:
17471 gdb compile failed, clang-13.0: warning: -e main: 'linker' input unused \
17472 [-Wunused-command-line-argument]
17473 clang-13.0: warning: -Wl,-Ttext=0x00: 'linker' input unused \
17474 [-Wunused-command-line-argument]
17475 clang-13.0: warning: -Wl,-N: 'linker' input unused \
17476 [-Wunused-command-line-argument]
17477 UNTESTED: gdb.base/infoline-reloc-main-from-zero.exp: \
17478 infoline-reloc-main-from-zero.exp
17479 UNTESTED: gdb.base/infoline-reloc-main-from-zero.exp: failed to compile
17482 Fix this by using ldflags instead of additional_flags.
17484 Likewise, fix all occurrences of:
17486 $ find gdb/testsuite -name *.exp | xargs grep additional_flags.*Wl
17489 Tested on x86_64-linux.
17491 2022-10-14 Tom de Vries <tdevries@suse.de>
17493 [gdb/testsuite] Fix nopie test-cases with target board unix/-fPIE/-pie
17494 Compilers default to either PIE or no-PIE executables.
17496 In order to test PIE executables with a compiler that produces non-PIE by
17497 default, we can use target board unix/-fPIE/-pie, which set the multilib_flags
17498 of the target board to "-fPIE -pie".
17500 Likewise, we can use target board unix/-fno-PIE/-no-pie with a compiler that
17501 produces PIE by default.
17503 The target board unix/-fno-PIE/-no-pie has a potential problem when compiling
17504 shared libs, because the multilib_flags will override the attempts of
17505 gdb_compile_shlib to compile with -fPIC. This is taken care of by running the
17506 body of gdb_compile_shlib wrapped in with_PIE_multilib_flags_filtered.
17508 The target board unix/-fPIE/-pie has a problem with nopie compilations. The
17509 current approach is to do the compilation hoping for the best, and if we find
17510 out that the resulting executable is PIE despite specifying nopie, we error
17511 out with the standard error message "nopie failed to prevent PIE executable".
17513 That however does not work for hard-coded assembly nopie test-cases, which will
17514 just noisily refuse to compile:
17516 ld: amd64-disp-step0.o: relocation R_X86_64_32S against `.text' can not be \
17517 used when making a PIE object; recompile with -fPIE^M
17520 Fix this in gdb_compile by filtering out the PIE settings in the target board
17521 multilib_flags when pie or nopie is specified.
17523 Tested on x86_64-linux.
17525 2022-10-14 Tom de Vries <tdevries@suse.de>
17527 [gdb/testsuite] Factor out with_PIE_multilib_flags_filtered
17528 Factor out new procs with_PIE_multilib_flags_filtered and
17529 with_multilib_flags_filtered from proc gdb_compile_shlib.
17531 Tested on x86_64-linux.
17533 2022-10-14 Tom de Vries <tdevries@suse.de>
17535 [gdb/testsuite] Add cond_wrap proc
17536 Add a new proc cond_wrap, that can be used to replace the repetitive:
17548 cond_wrap $cond wrap {
17553 Tested on x86_64-linux.
17555 2022-10-14 Torbjörn SVENSSON <torbjorn.svensson@foss.st.com>
17557 gdb: add Torbjörn Svensson to gdb/MAINTAINERS
17559 2022-10-14 Jan Beulich <jbeulich@suse.com>
17561 RISC-V: Zicbo{m,p,z} adjustments to riscv_multi_subset_supports_ext()
17562 The lack thereof did caused gas to issue "internal: unreachable
17563 INSN_CLASS_*" errors when trying to assemble respective insns without
17564 the feature(s) enabled via e.g. ".option arch, ...". Of course a proper
17565 hint towards the missing extension then wasn't given either.
17567 2022-10-14 Tsukasa OI <research_trasio@irq.a4lg.com>
17569 RISC-V: Imply 'Zicsr' from privileged extensions with CSRs
17570 'H', 'Smstateen', 'Sscofpmf' and 'Sstc' are four privileged extensions with
17571 their CSR definitions and 'Smepmp' is a privileged extension with additional
17574 Volume II: Privileged Architecture of the RISC-V ISA Manual states that the
17575 privileged architecture requires the 'Zicsr' extension. However, current
17576 GNU Binutils has no direct way whether the program has dependency to the
17577 privileged architecture itself.
17579 As a workaround, we should add implications from privileged extensions that
17580 either add new CSRs, extend existing CSRs or depends on using CSRs.
17582 This commit adds such implications for existing privileged extensions that
17583 satisfy this condition.
17587 * testsuite/gas/riscv/march-imply-h.d: New test, at least for 'H'.
17591 * elfxx-riscv.c (riscv_implicit_subsets): Add 'Zicsr'
17592 implicications for privileged extensions 'H', 'Smstateen',
17593 'Sscofpmf', 'Sstc' and 'Smepmp'.
17595 2022-10-14 Tsukasa OI <research_trasio@irq.a4lg.com>
17597 opcodes/riscv-dis.c: Remove last_map_state
17598 Before changing the core disassembler, we take care of minor code clarity
17599 issues and improve readability.
17601 This commit removes unused variable last_map_state (set by the
17602 print_insn_riscv function but not read anywhere else).
17606 * riscv-dis.c (last_map_state): Remove.
17607 (print_insn_riscv): Remove setting last_map_state.
17609 2022-10-14 Tsukasa OI <research_trasio@irq.a4lg.com>
17611 opcodes/riscv-dis.c: Make XLEN variable static
17612 Before changing the core disassembler, we take care of minor code clarity
17613 issues and improve readability.
17615 Since xlen variable is not (and should not) used outside riscv-dis.c,
17616 this commit makes this variable static.
17620 * riscv-dis.c (xlen): Make this variable static.
17622 2022-10-14 Tsukasa OI <research_trasio@irq.a4lg.com>
17624 opcodes/riscv-dis.c: Use bool type whenever possible
17625 Before changing the core disassembler, we take care of minor code clarity
17626 issues and improve readability.
17628 This commit replaces uses of int with bool whenever possible.
17632 * riscv-dis.c (no_aliases) Change type to bool.
17633 (set_default_riscv_dis_options): Use boolean.
17634 (parse_riscv_dis_option_without_args): Likewise.
17635 (riscv_disassemble_insn): Use boolean keywords.
17637 2022-10-14 Tsukasa OI <research_trasio@irq.a4lg.com>
17639 opcodes/riscv-dis.c: Tidying with spacing
17640 Before changing the core disassembler, we take care of minor code clarity
17641 issues and improve readability.
17643 This commit takes care of improper spacing for code clarity.
17647 * riscv-dis.c (riscv_disassemble_insn): Tidying with spacing.
17649 2022-10-14 Tsukasa OI <research_trasio@irq.a4lg.com>
17651 opcodes/riscv-dis.c: Tidying with comments/clarity
17652 Before changing the core disassembler, we take care of minor code clarity
17653 issues and improve readability.
17655 First, we need to clarify the roles of variables and code portions.
17659 * riscv-dis.c (xlen): Move before default_isa_spec. Add comment.
17660 (default_isa_spec, default_priv_spec): Add comment.
17661 (riscv_gpr_names, riscv_fpr_names): Likewise.
17662 (parse_riscv_dis_option_without_args): Likewise.
17663 (parse_riscv_dis_option, parse_riscv_dis_options): Likewise.
17664 (maybe_print_address): Likewise.
17665 (riscv_disassemble_insn): Fix comment about the Zfinx "extension".
17666 Add comment about the riscv_multi_subset_supports call.
17668 2022-10-14 Tsukasa OI <research_trasio@irq.a4lg.com>
17670 RISC-V: Test DWARF register number for "fp"
17671 This commit adds "fp" (x8 or s0) to dw-regnums.{s,d}.
17675 * testsuite/gas/riscv/dw-regnums.s: Add "fp".
17676 * testsuite/gas/riscv/dw-regnums.d: Likewise.
17678 2022-10-14 Tsukasa OI <research_trasio@irq.a4lg.com>
17680 RISC-V: Move standard hints before all instructions
17681 Because all standard hints must be placed before corresponding instruction
17682 for the disassembler, they may taint basic RVI instruction section.
17684 This commit moves all standard hints before all basic RVI instructions
17685 to improve maintainability.
17689 * riscv-opc.c (riscv_opcodes): Move all standard hints before all
17690 standard instructions.
17692 2022-10-14 Tsukasa OI <research_trasio@irq.a4lg.com>
17694 RISC-V: Move certain arrays to riscv-opc.c
17695 This is a part of small tidying (declare tables in riscv-opc.c).
17699 * opcode/riscv.h (riscv_rm, riscv_pred_succ): Move declarations to
17700 opcodes/riscv-opc.c. New non-static definitions.
17704 * riscv-opc.c (riscv_rm, riscv_pred_succ): Move from
17705 include/opcode/riscv.h. Add description.
17707 2022-10-14 Fangrui Song <i@maskray.me>
17709 ld: Add --undefined-version
17710 This cancels a previous --no-undefined-version.
17711 gold has had --undefined-version for a long time.
17713 2022-10-14 GDB Administrator <gdbadmin@sourceware.org>
17715 Automatic date update in version.in
17717 2022-10-13 Carl Love <cel@us.ibm.com>
17719 PowerPC, fix gdb.base/watchpoint.exp on Power 9
17720 Test gdb.base/watchpoint.exp generates 4 test errors on Power 9. The
17721 test uses the test [target_info exists gdb,no_hardware_watchpoints] to
17722 determine if the processor supports hardware watchpoints. The check
17723 only examines the processor type to determine if it supports hardware
17726 The PowerPC processors support hardware watchpoints with the
17727 exception of Power 9. The hardware watchpoint support is disabled on
17728 Power 9. The test skip_hw_watchpoint_tests must be used to correctly
17729 determine if the PowerPC processor supports hardware watchpoints.
17731 This patch replaces the [target_info exists gdb,no_hardware_watchpoints]
17732 with the skip_hw_watchpoint_tests_p check. With the patch, the test runs
17733 on Power 9 with hardware watchpoint force-disabled. The test runs on
17734 all other PowerPC processors with and without hardware watchpoints
17737 The patch has been tested on Power 9 to verify the test only runs with
17738 hardware breakpoints disabled. The patch has been tested on X86-64 with
17739 no regression failures. The test fails on Power 10 due to an internal GDB
17740 error due to resource management. The resource management issue will be
17741 addressed in another patch.
17743 2022-10-13 Tom de Vries <tdevries@suse.de>
17745 [gdb/testsuite] Fix gdb.dwarf2/macro-source-path.exp with -m32
17746 With test-case gdb.dwarf2/macro-source-path.exp and target board unix/-m32, I
17749 as: macro-source-path-gcc11-ld238-dw5-filename-641.o: \
17750 unsupported relocation type: 0x1^M
17753 The problem is that we have 64-bit dwarf so the debug_line offset in the
17754 .debug_macro section is an 8-byte entity, emitted using ".8byte":
17756 .section .debug_macro
17758 .2byte 5 /* version */
17759 .byte 3 /* flags */
17760 .8byte .LLlines3 /* debug_line offset */
17762 but the linker doesn't support 8-byte relocation types on a 32-bit architecture.
17764 This is similar to what was fixed in commit a5ac8e7fa3b
17765 ("[gdb/testsuite] Fix 64-bit dwarf test-cases with -m32") for for instance
17768 Fix this in the same way, by using _op_offset to emit the debug_line offset.
17770 Tested on x86_64-linux with native and target board unix/-m32.
17772 2022-10-13 Tom de Vries <tdevries@suse.de>
17774 [gdb/testsuite] Fix gdb.dwarf2/entry-value-typedef.exp with -m32
17775 With test-case gdb.dwarf2/entry-value-typedef.exp and target board unix/-m32,
17778 builtin_spawn -ignore SIGHUP g++ -fno-stack-protector \
17779 gdb/testsuite/gdb.dwarf2/entry-value-typedef-amd64.S \
17780 -fdiagnostics-color=never -Lbuild/libiberty -lm -m32 \
17781 -o outputs/gdb.dwarf2/entry-value-typedef/entry-value-typedef^M
17782 entry-value-typedef.cpp: Assembler messages:^M
17783 entry-value-typedef.cpp:38: Error: bad register name `%rbp'^M
17786 The problem is that the test-cases selects an amd64 .S file based on the check:
17788 if { [istarget "x86_64-*-linux*"] } {
17790 which is also true for target board unix/-m32 on x86_64-linux.
17792 Fix this by adding the missing is_lp64_target check.
17794 Tested on x86_64-linux, using native and target board unix/-m32.
17796 2022-10-13 Tom de Vries <tdevries@suse.de>
17798 [gdb/testsuite] Fix gdb.mi/mi-disassemble.exp with -m32
17799 With target board unix/-m32 and test-case gdb.mi/mi-disassemble.exp we have:
17802 print/x *((unsigned char *) 0x8048485)^M
17803 &"print/x *((unsigned char *) 0x8048485)\n"^M
17807 PASS: gdb.mi/mi-disassemble.exp: get valueof "*((unsigned char *) 0x8048485)"
17808 FAIL: gdb.mi/mi-disassemble.exp: byte at 0x8048485 matches
17810 The test-case passes with native.
17812 With native we see in gdb.log that variable longest_insn_bytes is:
17814 Longest instruction at 0x0000000000400549 with bytes '48 8b 05 20 01 00 00'
17816 and variable split_bytes (added debug puts) ends up as:
17818 SPLIT_BYTES: 48 8b 05 20 01 00 00
17821 But with unix/-m32 we have longest_insn_byte:
17823 Longest instruction at 0x08048481 with bytes '8d 4c 24 04 '
17825 and split_bytes ends up as:
17827 SPLIT_BYTES: 8d 4c 24 04 {} {} {} {} {} {} {} {}
17829 so the trailing whitespace is translated by split to empty bytes, and the
17830 mismatch FAILs are generated for those.
17832 Fix this by stripping the whitespace, which makes us end up with a different
17833 and indeed longer insn:
17835 Longest instruction at 0x08048492 with bytes 'dd 05 98 85 04 08'
17838 Tested on x86_64-linux, with native and target board unix/-m32.
17840 2022-10-13 GDB Administrator <gdbadmin@sourceware.org>
17842 Automatic date update in version.in
17844 2022-10-12 Carl Love <cel@us.ibm.com>
17846 PowerPC, fix test gdb.base/watchpoint-stops-at-right-insn.exp
17847 Test gdb.base/watchpoint-stops-at-right-insn.exp generates 4 test errors
17848 on Power 9. The test uses the test [target_info exists gdb,
17849 no_hardware_watchpoints] to determine if the processor supports hardware
17850 watchpoints. The check only examines the processor type to determine if
17851 it supports hardware watchpoints. Note, the test works fine on Power 10.
17853 The PowerPC processors support hardware watchpoints with the
17854 exception of Power 9. The hardware watchpoint support is disabled on
17855 Power 9. The test skip_hw_watchpoint_tests must be used to correctly
17856 determine if the PowerPC processor supports hardware watchpoints.
17858 This patch replaces the [target_info exists gdb,no_hardware_watchpoints]
17859 with the skip_hw_watchpoint_tests_p check. With the patch, the test is
17860 disabled on Power 9 but runs on all other PowerPC processors.
17862 The patch has been tested on Power 9, Power 10 and X86-64 with no
17863 regression failures.
17865 2022-10-12 Jan Beulich <jbeulich@suse.com>
17867 x86: drop "regmask" static variable
17868 Replace its two uses by more direct checks, paralleling what's already
17869 there for SIMD registers.
17871 2022-10-12 Tom de Vries <tdevries@suse.de>
17873 [gdb/symtab] Factor out elf_symfile_read_dwarf2
17874 Factor out elf_symfile_read_dwarf2 from elf_symfile_read. NFC.
17876 Tested on x86_64-linux.
17878 2022-10-12 Tom de Vries <tdevries@suse.de>
17880 [gdb/testsuite] Fix ctf test-cases on openSUSE Tumbleweed
17881 When running test-case gdb.base/ctf-constvars.exp on openSUSE Tumbleweed (with
17882 system gcc version 12, providing gcc -gctf support, enabling the ctf test-cases
17883 in the gdb testsuite), I run into:
17886 'vox' has unknown type; cast it to its declared type^M
17887 (gdb) FAIL: gdb.base/ctf-constvars.exp: print vox
17890 There are two causes for this:
17891 - the linker flags are missing --ctf-variables, so the information for variable
17892 vox is missing (reported in PR29468), and
17893 - the executable contains some dwarf2 due to some linked-in glibc objects,
17894 so the ctf info is ignored (reported in PR29160).
17897 - -Wl,--ctf-variable,
17898 - -Wl,--strip-debug, and
17899 we can make the test-case and some similar test-cases pass.
17901 Tested on x86_64-linux.
17903 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29160
17904 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29468
17906 2022-10-12 Tom de Vries <tdevries@suse.de>
17908 [gdb/testsuite] Silence warnings about obsolete -gstabs
17909 When running test-case gdb.base/gdbindex-stabs.exp on openSUSE Tumbleweed (with
17912 gdb compile failed, gdb/testsuite/gdb.base/gdbindex-stabs.c: warning: \
17913 STABS debugging information is obsolete and not supported anymore
17916 Silence the warning by passing quiet to gdb_compile. Likewise in two other
17919 2022-10-12 Tom de Vries <tdevries@suse.de>
17921 [gdb/testsuite] Replace remaining -gt with -gctf
17922 With test-cases gdb.base/cvexpr.exp and gdb.base/whatis.exp I run into:
17924 gdb compile failed, gcc: error: unrecognized debug output level 't'
17927 This is due to using additional_flags=-gt.
17929 Commit ffb3f587933 ("CTF: multi-CU and archive support") replaced
17930 additional_flags=-gt with additional_flags=-gctf in gdb.ctf/*.exp and
17931 gdb.base/ctf-*.exp.
17933 Do the same in these two test-cases.
17935 Tested on x86_64-linux.
17937 2022-10-12 Tom de Vries <tdevries@suse.de>
17939 [gdb/testsuite] Fix gdb.base/infoline-reloc-main-from-zero.exp with recent ld
17940 On openSUSE Tumbleweed (with ld 2.39) and test-case
17941 gdb.base/infoline-reloc-main-from-zero.exp, I get:
17943 gdb compile failed, ld: warning: infoline-reloc-main-from-zero has a LOAD \
17944 segment with RWX permissions
17945 UNTESTED: gdb.base/infoline-reloc-main-from-zero.exp: \
17946 infoline-reloc-main-from-zero.exp
17949 Fix this by compiling with -Wl,--no-warn-rwx-segments.
17951 Tested on x86_64-linux.
17953 2022-10-12 Tom de Vries <tdevries@suse.de>
17955 [gdb/testsuite] Fix gdb.base/nested-subp{2,3}.exp with recent ld
17956 On openSUSE Tumbleweed (with ld 2.39) I get for test-case
17957 gdb.base/nested-subp2.exp:
17959 gdb compile failed, ld: warning: tmp.o: requires executable stack \
17960 (because the .note.GNU-stack section is executable)
17963 Fix this by compiling with -Wl,--no-warn-execstack.
17965 Likewise in gdb.base/nested-subp3.exp
17967 Tested on x86_64-linux.
17969 2022-10-12 Tom de Vries <tdevries@suse.de>
17971 [gdb/testsuite] Remove unnecessary perror in some test-cases
17972 On openSUSE Tumbleweed I noticed:
17974 UNTESTED: gdb.dwarf2/fission-absolute-dwo.exp: fission-absolute-dwo.exp
17975 ERROR: failed to compile fission-absolute-dwo
17978 The ERROR is unnecessary, given that an UNTESTED is already emitted.
17980 Furthermore, it could be argued that it is incorrect because it's not a
17981 testsuite error to not be able to compile something, and UNTESTED or
17982 UNSUPPORTED is more appropriate.
17984 Remove the perror call, likewise in fission-relative-dwo.exp.
17986 Tested on x86_64-linux.
17988 2022-10-12 Nick Clifton <nickc@redhat.com>
17990 Fix objcopy's error message when it cannot add a .gnu_debuglink section because the section already exists.
17992 * objcopy.c (copy_object): Use the input filename when
17993 reporting that a .gnu_debuglink section already exists.
17995 2022-10-12 Tsukasa OI <research_trasio@irq.a4lg.com>
17997 sim/ppc: Fix core_find_mapping diagnostics
17998 Because "%p" is the pointer conversion specifier to print a pointer in an
17999 implementation-defined manner, the result with format string containing
18000 "0x%p" can be strange. For instance, core_map_find_mapping prints error
18001 containing "0x0x...." (processor is not NULL) or "0x(null)" (processor is
18004 This commit replaces "0x%p" with "%p" to prevent unpredictable behavior.
18006 2022-10-12 Andrew Burgess <aburgess@redhat.com>
18008 sim/ppc: fixes for arguments to printf style functions
18009 After the recent series of fixes to mark more functions in the
18010 simulator with ATTRIBUTE_PRINTF, there were some build failures in the
18011 ppc sim due, in some cases, to bugs with the arguments being passed,
18012 and in other cases, the issues were (maybe) less serious, with
18013 arguments being the wrong size, or type, for the printf format being
18016 This commit fixes all of the issues that I ran into.
18018 In each case I selected the easiest solution to the problem, which is
18019 usually just casting the argument to the correct type. If anyone
18020 later on thinks the print format should change, please feel free to do
18021 that. What we have here should keep the simulator basically working
18022 as it does currently, which is my goal with this commit.
18024 2022-10-12 Tom de Vries <tdevries@suse.de>
18026 [gdb/contrib] Use OBJCOPY everywhere in cc-with-tweaks.sh
18027 I noticed that the $want_gnu_debuglink code in gdb/contrib/cc-with-tweaks.sh
18028 uses objcopy instead of $OBJCOPY. Fix this.
18030 Script checked with shellcheck, no new warnings added.
18032 Tested on x86_64-linux.
18034 2022-10-12 Simon Marchi <simon.marchi@efficios.com>
18036 Re-apply "Pass PKG_CONFIG_PATH down from top-level Makefile"
18037 Commit 228cf97dd3c8 ("Merge configure.ac from gcc project") undid the
18038 change originally done in de83289ef32e ("Pass PKG_CONFIG_PATH down from
18039 top-level Makefile"). Re-apply it.
18041 Change-Id: I91138dfca41c43b05e53e445f62e4b27882536bf
18043 2022-10-12 Simon Marchi <simon.marchi@polymtl.ca>
18045 gdb: rename target_read_auxv(target_ops *) to target_read_auxv_raw
18046 Having two overloads of target_read_auxv that don't have the same goals
18047 is confusing. Rename the one that reads from an explicit target_ops to
18048 target_read_auxv_raw. Also, it occured to me that the non-raw version
18049 could use the raw version, that reduces duplication a bit.
18051 Change-Id: I28e5f7cecbfcacd0174d4686efb3e4a23b4ad491
18053 2022-10-12 GDB Administrator <gdbadmin@sourceware.org>
18055 Automatic date update in version.in
18057 2022-10-12 Alan Modra <amodra@gmail.com>
18059 Re: Merge configure.ac from gcc project
18060 Also copy over config.acx.m4, and regenerate.
18062 2022-10-11 Simon Marchi <simon.marchi@polymtl.ca>
18064 gdb: fix auxv caching
18065 There's a flaw in the interaction of the auxv caching and the fact that
18066 target_auxv_search allows reading auxv from an arbitrary target_ops
18067 (passed in as a parameter). This has consequences as explained in this
18070 https://inbox.sourceware.org/gdb-patches/20220719144542.1478037-1-luis.machado@arm.com/
18072 In summary, when loading an AArch64 core file with MTE support by
18073 passing the executable and core file names directly to GDB, we see the
18076 $ ./gdb -nx --data-directory=data-directory -q aarch64-mte-gcore aarch64-mte-gcore.core
18078 Program terminated with signal SIGSEGV, Segmentation fault
18079 Memory tag violation while accessing address 0x0000ffff8ef5e000
18082 #0 0x0000aaaade3d0b4c in ?? ()
18085 But if we do it as two separate commands (file and core) we don't:
18087 $ ./gdb -nx --data-directory=data-directory -q -ex "file aarch64-mte-gcore" -ex "core aarch64-mte-gcore.core"
18089 Program terminated with signal SIGSEGV, Segmentation fault.
18090 #0 0x0000aaaade3d0b4c in ?? ()
18093 The problem with the latter is that auxv data gets improperly cached
18094 between the two commands. When executing the file command, auxv gets
18095 first queried here, when loading the executable:
18097 #0 target_auxv_search (ops=0x55555b842400 <exec_ops>, match=0x9, valp=0x7fffffffc5d0) at /home/simark/src/binutils-gdb/gdb/auxv.c:383
18098 #1 0x0000555557e576f2 in svr4_exec_displacement (displacementp=0x7fffffffc8c0) at /home/simark/src/binutils-gdb/gdb/solib-svr4.c:2482
18099 #2 0x0000555557e594d1 in svr4_relocate_main_executable () at /home/simark/src/binutils-gdb/gdb/solib-svr4.c:2878
18100 #3 0x0000555557e5989e in svr4_solib_create_inferior_hook (from_tty=1) at /home/simark/src/binutils-gdb/gdb/solib-svr4.c:2933
18101 #4 0x0000555557e6e49f in solib_create_inferior_hook (from_tty=1) at /home/simark/src/binutils-gdb/gdb/solib.c:1253
18102 #5 0x0000555557f33e29 in symbol_file_command (args=0x7fffffffe01c "aarch64-mte-gcore", from_tty=1) at /home/simark/src/binutils-gdb/gdb/symfile.c:1655
18103 #6 0x00005555573319c3 in file_command (arg=0x7fffffffe01c "aarch64-mte-gcore", from_tty=1) at /home/simark/src/binutils-gdb/gdb/exec.c:555
18104 #7 0x0000555556e47185 in do_simple_func (args=0x7fffffffe01c "aarch64-mte-gcore", from_tty=1, c=0x612000047740) at /home/simark/src/binutils-gdb/gdb/cli/cli-decode.c:95
18105 #8 0x0000555556e551c9 in cmd_func (cmd=0x612000047740, args=0x7fffffffe01c "aarch64-mte-gcore", from_tty=1) at /home/simark/src/binutils-gdb/gdb/cli/cli-decode.c:2543
18106 #9 0x00005555580e63fd in execute_command (p=0x7fffffffe02c "e", from_tty=1) at /home/simark/src/binutils-gdb/gdb/top.c:692
18107 #10 0x0000555557771913 in catch_command_errors (command=0x5555580e55ad <execute_command(char const*, int)>, arg=0x7fffffffe017 "file aarch64-mte-gcore", from_tty=1, do_bp_actions=true) at /home/simark/src/binutils-gdb/gdb/main.c:513
18108 #11 0x0000555557771fba in execute_cmdargs (cmdarg_vec=0x7fffffffd570, file_type=CMDARG_FILE, cmd_type=CMDARG_COMMAND, ret=0x7fffffffd230) at /home/simark/src/binutils-gdb/gdb/main.c:608
18109 #12 0x00005555577755ac in captured_main_1 (context=0x7fffffffda10) at /home/simark/src/binutils-gdb/gdb/main.c:1299
18110 #13 0x0000555557775c2d in captured_main (data=0x7fffffffda10) at /home/simark/src/binutils-gdb/gdb/main.c:1320
18111 #14 0x0000555557775cc2 in gdb_main (args=0x7fffffffda10) at /home/simark/src/binutils-gdb/gdb/main.c:1345
18112 #15 0x00005555568bdcbe in main (argc=10, argv=0x7fffffffdba8) at /home/simark/src/binutils-gdb/gdb/gdb.c:32
18114 Here, target_auxv_search is called on the inferior's target stack. The
18115 target stack only contains the exec target, so the query returns empty
18116 auxv data. This gets cached for that inferior in `auxv_inferior_data`.
18118 In its constructor (before it is pushed to the inferior's target stack),
18119 the core_target needs to identify the right target description from the
18120 core, and for that asks the gdbarch to read a target description from
18121 the core file. Because some implementations of
18122 gdbarch_core_read_description (such as AArch64's) need to read auxv data
18123 from the core in order to determine the right target description, the
18124 core_target passes a pointer to itself, allowing implementations to call
18125 target_auxv_search it. However, because we have previously cached
18126 (empty) auxv data for that inferior, target_auxv_search searched that
18127 cached (empty) auxv data, not auxv data read from the core. Remember
18128 that this data was obtained by reading auxv on the inferior's target
18129 stack, which only contained an exec target.
18131 The problem I see is that while target_auxv_search offers the
18132 flexibility of reading from an arbitrary (passed as an argument) target,
18133 the caching doesn't do the distinction of which target is being queried,
18134 and where the cached data came from. So, you could read auxv from a
18135 target A, it gets cached, then you try to read auxv from a target B, and
18136 it returns the cached data from target A. That sounds wrong. In our
18137 case, we expect to read different auxv data from the core target than
18138 what we have read from the target stack earlier, so it doesn't make
18139 sense to hit the cache in this case.
18141 To fix this, I propose splitting the code paths that read auxv data from
18142 an inferior's target stack and those that read from a passed-in target.
18143 The code path that reads from the target stack will keep caching,
18144 whereas the one that reads from a passed-in target won't. And since,
18145 searching in auxv data is independent from where this data came from,
18146 split the "read" part from the "search" part.
18148 From what I understand, auxv caching was introduced mostly to reduce
18149 latency on remote connections, when doing many queries. With the change
18150 I propose, only the queries done while constructing the core_target
18151 end up not using cached auxv data. This is fine, because there are just
18152 a handful of queries max, done at this point, and reading core files is
18155 The changes to auxv functions are:
18157 - Introduce 2 target_read_auxv functions. One reads from an explicit
18158 target_ops and doesn't do caching (to be used in
18159 gdbarch_core_read_description context). The other takes no argument,
18160 reads from the current inferior's target stack (it looks just like a
18161 standard target function wrapper) and does caching.
18163 The first target_read_auxv actually replaces get_auxv_inferior_data,
18164 since it became a trivial wrapper around it.
18166 - Change the existing target_auxv_search to not read auxv data from the
18167 target, but to accept it as a parameter (a gdb::byte_vector). This
18168 function doesn't care where the data came from, it just searches in
18169 it. It still needs to take a target_ops and gdbarch to know how to
18170 parse auxv entries.
18172 - Add a convenience target_auxv_search overload that reads auxv
18173 data from the inferior's target stack and searches in it. This
18174 overload is useful to replace the exist target_auxv_search calls that
18175 passed the `current_inferior ()->top_target ()` target and keep the
18178 - Modify parse_auxv to accept a target_ops and gdbarch to use for
18179 parsing entries. Not strictly related to the rest of this change,
18180 but it seems like a good change in the context.
18182 Changes in architecture-specific files (tdep and nat):
18184 - In linux-tdep, linux_get_hwcap and linux_get_hwcap2 get split in two,
18185 similar to target_auxv_search. One version receives auxv data,
18186 target and arch as parameters. The other gets everything from the
18187 current inferior. The latter is for convenience, to avoid making
18188 call sites too ugly.
18190 - Call sites of linux_get_hwcap and linux_get_hwcap2 are adjusted to
18191 use either of the new versions. The call sites in
18192 gdbarch_core_read_description context explicitly read auxv data from
18193 the passed-in target and call the linux_get_hwcap{,2} function with
18194 parameters. Other call sites use the versions without parameters.
18196 - Same idea for arm_fbsd_read_description_auxv.
18198 - Call sites of target_auxv_search that passed
18199 `current_inferior ()->top_target ()` are changed to use the
18200 target_auxv_search overload that works in the current inferior.
18202 Reviewed-By: John Baldwin <jhb@FreeBSD.org>
18203 Reviewed-By: Luis Machado <luis.machado@arm.com>
18204 Change-Id: Ib775a220cf1e76443fb7da2fdff8fc631128fe66
18206 2022-10-11 Tom de Vries <tdevries@suse.de>
18208 [gdb/testsuite] Fix gdb.debuginfod/fetch_src_and_symbols.exp with native-gdbserver
18209 When running test-case gdb.debuginfod/fetch_src_and_symbols.exp with target
18210 board native-gdbserver, I get:
18212 Running gdb.debuginfod/fetch_src_and_symbols.exp ...
18213 ERROR: tcl error sourcing gdb.debuginfod/fetch_src_and_symbols.exp.
18214 ERROR: gdbserver does not support start without extended-remote
18216 "error "gdbserver does not support $command without extended-remote""
18217 (procedure "gdb_test_multiple" line 51)
18218 invoked from within
18219 "gdb_test_multiple $command $message {*}$opts $user_code"
18220 (procedure "gdb_test" line 56)
18221 invoked from within
18222 "gdb_test "start" "Temporary breakpoint.*""
18225 Fix this by replacing gdb_test "start" with runto_main.
18227 Tested on x86_64-linux.
18229 2022-10-11 Nick Clifton <nickc@redhat.com>
18231 Re: Error: attempt to get value of unresolved symbol `L0'
18232 * symbols.c (S_GET_VALUE): If the unresolved symbol is the fake
18233 label provide a more helpful error message to the user.
18234 (S_GET_VALUE_WHERE): Like S_GET_VALUE, but includes a file/line
18235 number for error reporting purposes.
18236 * symbols.h (S_GET_VALUE_WHERE): Prototype.
18237 * write.c (fixup_segment): Use S_GET_VALUE_WHERE.
18239 2022-10-11 Tsukasa OI <research_trasio@irq.a4lg.com>
18241 sim: Initialize pbb_br_* by default
18242 On the files generated by sim/common/genmloop.sh, variables pbb_br_type and
18243 pbb_br_npc are declared uninitialized and passed to other functions in some
18244 cases. Despite that those are harmless, they will generate GCC warnings
18245 ("-Wmaybe-uninitialized").
18247 This commit ensures that pbb_br_type and pbb_br_npc variables are
18248 initialized to a harmless value.
18250 2022-10-11 Tsukasa OI <research_trasio@irq.a4lg.com>
18252 sim: Check known getopt definition existence
18253 Clang generates a warning if there is a function declaration/definition
18254 with zero arguments. Such declarations/definitions without a prototype (an
18255 argument list) are deprecated forms of indefinite arguments
18256 ("-Wdeprecated-non-prototype"). On the default configuration, it causes a
18257 build failure (unless "--disable-werror" is specified).
18259 include/getopt.h defines some getopt function definitions but one of them
18260 has a form "extern int getopt ();". If this form is selected in
18261 include/getopt.h, Clang generates a warning and the build fails by default.
18263 In really old environments, this getopt definition with no arguments is
18264 necessary (because the definition may change between environments).
18265 However, this definition is now a cause of problems on modern environments.
18267 A good news is, this definition is not always selected (e.g. if used by
18268 binutils/*.c). This is because configuration scripts of binutils, gas,
18269 gprof and ld tries to find known definition of getopt function is used and
18270 defines HAVE_DECL_GETOPT macro. If this macro is defined when getopt.h is
18271 included, a good form of getopt is used and Clang won't generate warnings.
18273 This commit adds a modified portion of ld/configure.ac to find the known
18274 getopt definition. If we could find one (and we *will* in most modern
18275 environments), we don't need to rely on the deprecated definition.
18277 2022-10-11 Tsukasa OI <research_trasio@irq.a4lg.com>
18278 Andrew Burgess <aburgess@redhat.com>
18280 sim: Suppress non-literal printf warning
18281 Clang generates a warning if the format string of a printf-like function is
18282 not a literal ("-Wformat-nonliteral"). On the default configuration, it
18283 causes a build failure (unless "--disable-werror" is specified).
18285 To avoid this warning, this commit now uses vsnprintf to format error
18286 message and pass the message to sim_engine_abort function with another
18287 printf-style formatting.
18289 This patch is mostly authored by Andrew Burgess and slightly modified by
18292 2022-10-11 Tsukasa OI <research_trasio@irq.a4lg.com>
18294 sim: Make WITH_{TRACE,PROFILE}-based macros bool
18295 Clang generates a warning if there is an ambiguous expression (possibly a
18296 bitwise operation (& or |), but a logical operator (&& or ||) is used;
18297 "-Wconstant-logical-operand"). On the default configuration, it causes a
18298 build failure (unless "--disable-werror" is specified).
18300 This is caused by predicate macros that use the form (base_variable & flag).
18301 Clang considers them as regular integer values (not boolean) and
18302 generates that warning.
18304 This commit makes Clang think those predicate macros to be boolean.
18306 2022-10-11 Tsukasa OI <research_trasio@irq.a4lg.com>
18308 sim: Remove self-assignments
18309 Clang generates a warning if there is a redundant self-assignment
18310 ("-Wself-assign"). On the default configuration, it causes a build failure
18311 (unless "--disable-werror" is specified).
18313 This commit removes redundant self-assignments from two files.
18315 2022-10-11 Tsukasa OI <research_trasio@irq.a4lg.com>
18317 sim/rl78: Add ATTRIBUTE_PRINTF
18318 Clang generates a warning if the format string of a printf-like function is
18319 not a literal ("-Wformat-nonliteral"). On the default configuration, it
18320 causes a build failure (unless "--disable-werror" is specified).
18322 To avoid warnings on the printf-like wrapper, it requires proper
18323 __attribute__((format)) and we have ATTRIBUTE_PRINTF macro for this reason.
18325 This commit adds ATTRIBUTE_PRINTF to the printf-like functions.
18327 2022-10-11 Tsukasa OI <research_trasio@irq.a4lg.com>
18329 sim/ppc: Add ATTRIBUTE_PRINTF
18330 Clang generates a warning if the format string of a printf-like function is
18331 not a literal ("-Wformat-nonliteral"). On the default configuration, it
18332 causes a build failure (unless "--disable-werror" is specified).
18334 To avoid warnings on the printf-like wrapper, it requires proper
18335 __attribute__((format)) and we have ATTRIBUTE_PRINTF macro for this reason.
18337 This commit adds ATTRIBUTE_PRINTF to the printf-like functions.
18339 For the error function defined in sim_calls.c, the ATTRIBUTE_NORETURN
18340 has been moved to the function declaration.
18342 2022-10-11 Tsukasa OI <research_trasio@irq.a4lg.com>
18344 sim/m68hc11: Add ATTRIBUTE_PRINTF
18345 Clang generates a warning if the format string of a printf-like function is
18346 not a literal ("-Wformat-nonliteral"). On the default configuration, it
18347 causes a build failure (unless "--disable-werror" is specified).
18349 To avoid warnings on the printf-like wrapper, it requires proper
18350 __attribute__((format)) and we have ATTRIBUTE_PRINTF macro for this reason.
18352 This commit adds ATTRIBUTE_PRINTF to a printf-like function.
18354 2022-10-11 Tsukasa OI <research_trasio@irq.a4lg.com>
18356 sim/m32c: Add ATTRIBUTE_PRINTF
18357 Clang generates a warning if the format string of a printf-like function is
18358 not a literal ("-Wformat-nonliteral"). On the default configuration, it
18359 causes a build failure (unless "--disable-werror" is specified).
18361 To avoid warnings on the printf-like wrapper, it requires proper
18362 __attribute__((format)) and we have ATTRIBUTE_PRINTF macro for this reason.
18364 This commit adds ATTRIBUTE_PRINTF to the printf-like functions.
18366 2022-10-11 Tsukasa OI <research_trasio@irq.a4lg.com>
18368 sim/erc32: Add ATTRIBUTE_PRINTF
18369 Clang generates a warning if the format string of a printf-like function is
18370 not a literal ("-Wformat-nonliteral"). On the default configuration, it
18371 causes a build failure (unless "--disable-werror" is specified).
18373 To avoid warnings on the printf-like wrapper, it requires proper
18374 __attribute__((format)) and we have ATTRIBUTE_PRINTF macro for this reason.
18376 This commit adds ATTRIBUTE_PRINTF to the printf-like functions.
18378 2022-10-11 Tsukasa OI <research_trasio@irq.a4lg.com>
18380 sim/cris: Add ATTRIBUTE_PRINTF
18381 Clang generates a warning if the format string of a printf-like function is
18382 not a literal ("-Wformat-nonliteral"). On the default configuration, it
18383 causes a build failure (unless "--disable-werror" is specified).
18385 To avoid warnings on the printf-like wrapper, it requires proper
18386 __attribute__((format)) and we have ATTRIBUTE_PRINTF macro for this reason.
18388 This commit adds ATTRIBUTE_PRINTF to a printf-like function.
18390 2022-10-11 Tsukasa OI <research_trasio@irq.a4lg.com>
18392 sim/common: Add ATTRIBUTE_PRINTF
18393 Clang generates a warning if the format string of a printf-like function is
18394 not a literal ("-Wformat-nonliteral"). On the default configuration, it
18395 causes a build failure (unless "--disable-werror" is specified).
18397 To avoid warnings on the printf-like wrapper, it requires proper
18398 __attribute__((format)) and we have ATTRIBUTE_PRINTF macro for this reason.
18400 This commit adds ATTRIBUTE_PRINTF to a printf-like function.
18402 2022-10-11 Martin Liska <mliska@suse.cz>
18404 fix compressed_debug_section_names definition for "zlib"
18407 * libbfd.c: Set COMPRESS_DEBUG_GABI_ZLIB for "zlib" value.
18409 2022-10-11 Martin Liska <mliska@suse.cz>
18411 add --enable-default-compressed-debug-sections-algorithm configure option
18414 * configure.ac: Add --enable-default-compressed-debug-sections-algorithm.
18415 * configure: Regenerate.
18419 * NEWS: Document the new option.
18420 * as.c (flag_compress_debug): Set default algorithm based
18421 on the configure option.
18422 * configure.ac: Add --enable-default-compressed-debug-sections-algorithm.
18423 * configure: Regenerate.
18424 * config.in: Likewise.
18428 * NEWS: Document the new option.
18429 * configure.ac: Add --enable-default-compressed-debug-sections-algorithm.
18430 * configure: Regenerate.
18431 * config.in: Likewise.
18432 * ldmain.c: Set default algorithm based
18433 on the configure option.
18435 2022-10-11 Martin Liska <mliska@suse.cz>
18437 refactor usage of compressed_debug_section_type
18440 * bfd-in.h (bfd_hash_set_default_size): Add COMPRESS_UNKNOWN
18442 (struct compressed_type_tuple): New.
18443 * bfd-in2.h (bfd_hash_set_default_size): Regenerate.
18444 (struct compressed_type_tuple): Likewise.
18445 * libbfd.c (ARRAY_SIZE): New macro.
18446 (bfd_get_compression_algorithm): New function.
18447 (bfd_get_compression_algorithm_name): Likewise.
18451 * as.c: Do not special-case, use the new functions.
18455 * emultempl/elf.em: Do not special-case, use the new functions.
18456 * lexsup.c (elf_static_list_options): Likewise.
18458 2022-10-11 Tsukasa OI <research_trasio@irq.a4lg.com>
18460 sim/riscv: fix multiply instructions on simulator
18463 commit 0938b032daa52129b4215d8e0eedb6c9804f5280
18464 Date: Wed Feb 2 10:06:15 2022 +0900
18466 RISC-V: Add 'Zmmul' extension in assembler.
18468 some instructions in the RISC-V simulator stopped working as a new
18469 instruction class 'INSN_CLASS_ZMMUL' was added, and some existing
18470 instructions were moved into this class.
18472 The simulator doesn't currently handle this instruction class, and so
18473 the instructions will now cause an illegal instruction trap.
18475 This commit adds support for INSN_CLASS_ZMMUL, and adds a test that
18476 ensures the affected instructions can be executed by the simulator.
18478 Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
18479 Reviewed-by: Andrew Burgess <aburgess@redhat.com>
18481 2022-10-11 Nick Clifton <nickc@redhat.com>
18483 Error: attempt to get value of unresolved symbol `L0'
18484 * symbols.c (S_GET_VALUE): If the unresolved symbol is the fake
18485 label provide a more helpful error message to the user.
18487 2022-10-11 Tsukasa OI <research_trasio@irq.a4lg.com>
18489 sim/moxie: add custom directory stamp rule
18490 Because sim/moxie/moxie-gdb.dtb is neither a program nor a library, automake
18491 does not generate dirstamp file ($builddir/sim/moxie/.dirstamp) for it.
18493 When maintainer mode is enabled, it tries to rebuild sim/moxie/moxie-gdb.dtb
18494 but fails because there's no rules for automake-generated dirstamp file
18495 which moxie-gdb.dtb depends.
18497 This commit adds its own rule for the directory stamp (modified copy of the
18498 automake output) and adds the directory stamp file to DISTCLEANFILES to
18499 mimic automake-generated behavior (although "make distclean" does not work
18500 when maintainer mode is enabled).
18502 2022-10-11 Bruno Larsen <blarsen@redhat.com>
18504 gdb/testsuite: Fix formatting of python script
18505 The python black formatter was complaining about formatting on the
18506 script gdb.python/pretty-print-call-by-hand.py. This commit changed
18507 the offending lines to make the formatter happy.
18509 2022-10-11 Tom de Vries <tdevries@suse.de>
18511 [gdb/testsuite] Fix prompt parsing in capture_command_output
18512 I noticed in capture_command_output that the output of a single command is
18513 matched using two gdb_test_multiples:
18514 - the first one matching the echoed command and skipping an optional prefix,
18515 - the second one matching the output and the prompt.
18517 This is error-prone, because the first gdb_test_multiple has implicit
18518 clauses which may consume the prompt.
18520 The problem is easy to spot with an example. First consider:
18522 set output [capture_command_output "print 1" "\\\$1 = "]
18523 gdb_assert { [string equal $output "1"] }
18527 PASS: [string equal $output "1"]
18530 If we change the prefix string to a no-match, say "1 = ", and update the
18531 output string match accordingly, we get instead:
18533 FAIL: capture_command_output for print 1
18534 FAIL: [string equal $output "\$1 = 1"]
18537 The first FAIL is produced by the first gdb_test_multiple, consuming the prompt.
18539 The second gdb_test_multiple then silently times out waiting for another prompt,
18540 after which the second FAIL is produced. Note that the timeout is silent
18541 because the gdb_test_multiple is called with an empty message argument.
18543 The second FAIL is because capture_command_output returns "", given that all
18544 the command output was consumed by the first gdb_test_multiple.
18546 Fix this by rewriting capture_command_output to use only a single
18549 Tested on x86_64-linux.
18551 2022-10-11 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
18553 gprofng: no need to build version.texi
18555 2022-10-10 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
18559 * doc/Makefile.am: No need to build version.texi.
18560 * doc/Makefile.in: Rebuild.
18562 2022-10-11 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
18564 gprofng: use the --libdir path to find libraries
18566 2022-10-10 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
18569 * src/Makefile.am: Add -DLIBDIR to CPPFLAGS.
18570 * src/Makefile.in: Rebuild.
18571 * src/envsets.cc (putenv_libcollector_ld_misc): Use LIBDIR to find
18572 the gprofng libraries.
18574 2022-10-11 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
18576 gprofng: run tests without installation
18578 2022-10-10 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
18581 * testsuite/config/default.exp: Set up environment to run gprofng tests
18582 without installation.
18583 * testsuite/lib/Makefile.skel: Likewise.
18584 * testsuite/lib/display-lib.exp: Likewise.
18586 2022-10-11 Simon Marchi <simon.marchi@polymtl.ca>
18588 gdb/testsuite: fix race in gdb.base/async-shell.exp
18589 I see some random failures in this test:
18591 FAIL: gdb.base/async-shell.exp: run & (timeout)
18593 It can be reliably reproduced on a recent enough GNU/Linux with this
18596 diff --git a/gdb/testsuite/lib/gdb.exp b/gdb/testsuite/lib/gdb.exp
18597 index 44cc28b30051..2a3c8253ba5a 100644
18598 --- a/gdb/testsuite/lib/gdb.exp
18599 +++ b/gdb/testsuite/lib/gdb.exp
18600 @@ -1301,6 +1301,7 @@ proc gdb_test_multiple { command message args } {
18602 set gdb_test_name "$message"
18606 set code [catch {gdb_expect $code} string]
18608 "recent enough" means a system where libpthread.so was merged with
18609 libc.so, so at least glibc 2.34.
18611 The problem is that the `run &` command prints some things after the
18614 (gdb) [Thread debugging using libthread_db enabled]
18615 Using host libthread_db library "/usr/lib/../lib/libthread_db.so.1".
18617 If expect is quick enough, it will consume only up to the prompt. But
18618 if it is slow enough, it will consume those messages at the same time as
18619 the prompt, in which case the gdb_test used for "run &" won't match. By
18620 default, the prompt used by gdb_test uses a `$` to anchor the match at
18621 the end of the buffer. If there's anything following the prompt, it
18624 The diff above adds a delay between sending the command and consuming
18625 the output, giving GDB more time to output the messages, giving a good
18626 chance that expect consumes them at the same time as the prompt.
18628 This is normally handled by using gdb_test_multiple and specifying a
18629 pattern that ends with "$gdb_prompt", but not a trailing $. I think
18630 this is common enough that it deserves its own gdb_test option.
18631 Therefore, add the -no-anchor-prompt option to gdb_test, and
18632 gdb_test_no_output for completeness. Use it in
18633 gdb.base/async-shell.exp.
18635 Change-Id: I9051d8800d1c10a2e95db1a575991f7723492f1b
18636 Approved-By: Tom de Vries <tdevries@suse.de>
18638 2022-10-11 GDB Administrator <gdbadmin@sourceware.org>
18640 Automatic date update in version.in
18642 2022-10-10 Tom Tromey <tom@tromey.com>
18644 Fix a latent bug in print_wchar
18645 print_wchar keeps track of when escape sequences are emitted, to force
18646 an escape sequence if needed by a subsequent character. For example
18647 for the string concatenation "\0" "1", gdb will print "\000\061" --
18648 because printing "\0001" might be confusing.
18650 However, this code has two errors. First, this logic is not needed
18651 for octal escapes, because there is a length limit of 3 for octal
18652 escapes, and gdb always prints these with "%.3o". Second, though,
18653 this *is* needed for hex escapes, because those do not have a length
18656 This patch fixes these problems and adds the appropriate tests.
18658 2022-10-10 Tom Tromey <tom@tromey.com>
18660 Don't use wchar_printable in print_wchar
18661 print_wchar uses wchar_printable, but this isn't needed -- all the
18662 relevant cases are already handled by the 'switch'. This changes the
18663 code to use gdb_iswprint, and removes a somewhat confusing comment
18664 related to this code.
18667 This renames c_printstr, removing a layer of indirection.
18670 This renames c_emit_char, removing a layer of indirection.
18672 Boolify need_escape in generic_emit_char
18673 This changes 'need_escape' in generic_emit_char to be of type bool,
18676 2022-10-10 Tom Tromey <tom@tromey.com>
18677 Andrew Burgess <aburgess@redhat.com>
18679 Fix latent quote char bug in generic_printstr
18680 generic_printstr prints an empty string like:
18682 fputs_filtered ("\"\"", stream);
18684 However, this seems wrong to me if the quote character is something
18685 other than double quote. This patch fixes this latent bug. Thanks to
18686 Andrew for the test case.
18688 2022-10-10 Tom Tromey <tromey@adacore.com>
18690 Fix the guile build
18691 The frame_info_ptr patches broke the build with Guile. This patch
18692 fixes the problem. In mos cases I chose to preserve the use of
18693 frame_info_ptr, at least where I could be sure that the object
18694 lifetime did not interact with Guile's longjmp-based exception scheme.
18696 Tested on x86-64 Fedora 34.
18698 2022-10-10 Tom de Vries <tdevries@suse.de>
18700 [gdb/testsuite] Detect trailing ^C/^D in command
18701 Detect a trailing ^C/^D in the command argument of gdb_test_multiple, and
18704 Tested on x86_64-linux.
18706 2022-10-10 Tom de Vries <tdevries@suse.de>
18708 [gdb/testsuite] Fix error message for cmd with trailing newline
18709 I noticed that the error message in gdb_test_multiple about trailing newline
18710 in a command does not mention the offending command, nor the word command:
18712 if [string match "*\[\r\n\]" $command] {
18713 error "Invalid trailing newline in \"$message\" test"
18717 Fix this by using instead:
18719 error "Invalid trailing newline in \"$command\" command"
18722 Also add a test-case to trigger this: gdb.testsuite/gdb-test.exp.
18724 Tested on x86_64-linux.
18726 2022-10-10 Andrew Burgess <aburgess@redhat.com>
18728 gdb: include the base address in in-memory bfd filenames
18729 The struct target_buffer (in gdb_bfd.c) is used to hold information
18730 about an in-memory BFD object created by GDB. For now this mechanism
18731 is used by GDB when loading information about JIT symfiles.
18733 This commit updates target_buffer (in gdb_bfd.c) to be more C++ like,
18734 and, at the same time, adds the base address of the symfile into the
18737 Right now, every in-memory BFD is given the filename "<in-memory>".
18738 This filename is visible in things like 'maint info symtabs' and
18739 'maint info line-table'. If there are multiple in-memory BFD objects
18740 then it can be hard to match keep track if which BFD is which. This
18741 commit changes the name to be "<in-memory@ADDRESS>" where ADDRESS is
18742 replaced with the base address for where the in-memory symbol file was
18745 As an example of how this is useful, here's the output of 'maint info
18746 jit' showing a single loaded JIT symfile:
18748 (gdb) maintenance info jit
18749 jit_code_entry address symfile address symfile size
18750 0x00000000004056b0 0x0000000007000000 17320
18752 And here's part of the output from 'maint info symtabs':
18754 (gdb) maintenance info symtabs
18756 { objfile <in-memory@0x7000000> ((struct objfile *) 0x5258250)
18757 { ((struct compunit_symtab *) 0x4f0afb0)
18758 debugformat DWARF 4
18759 producer GNU C17 9.3.1 20200408 (Red Hat 9.3.1-2) -mtune=generic -march=x86-64 -g -fno-stack-protector -fpic
18760 name jit-elf-solib.c
18761 dirname /tmp/binutils-gdb/build/gdb/testsuite
18762 blockvector ((struct blockvector *) 0x5477850)
18763 user ((struct compunit_symtab *) (null))
18764 { symtab /tmp/binutils-gdb/build/gdb/testsuite/../../../src/gdb/testsuite/gdb.base/jit-elf-solib.c ((struct symtab *) 0x4f0b030)
18766 linetable ((struct linetable *) 0x5477880)
18771 I've added a new test that checks the new in-memory file names are
18772 generated correctly, and also checks that the in-memory JIT files can
18773 be dumped back out using 'dump binary memory'.
18775 2022-10-10 Andrew Burgess <aburgess@redhat.com>
18777 gdb: remove filename arg from gdb_bfd_open_from_target_memory
18778 The filename argument to gdb_bfd_open_from_target_memory was never
18779 used; this argument had a default value of nullptr, and the only call
18780 to this function, in jit.c, relied on the default value.
18782 In the next commit I'm going to make some changes to the
18783 gdb_bfd_open_from_target_memory function, and, though I could take
18784 account of a filename parameter, it seems pointless to maintain an
18787 This commit removes the filename argument.
18789 There should be no user visible changes after this commit.
18791 2022-10-10 Andrew Burgess <aburgess@redhat.com>
18793 gdb: add infcall specific debugging
18794 Add two new commands:
18796 set debug infcall on|off
18799 These enable some new debugging related to when GDB makes inferior
18800 function calls. I've added some basic debugging for what I think are
18801 the major steps in the inferior function call process, but I'm sure we
18802 might want to add more later.
18804 2022-10-10 Andrew Burgess <aburgess@redhat.com>
18806 gdb: extra debug output in thread.c
18807 Add some extra 'threads' debug in a couple of places in thread.c.
18808 I've also added an additional gdb_assert in one case.
18810 gdb: improve infrun_debug_show_threads output
18811 This commit switches to use INFRUN_SCOPED_DEBUG_START_END in the
18812 infrun_debug_show_threads function, which means the output will get an
18813 extra level of indentation, this looks a little nicer I think.
18815 2022-10-10 Nick Clifton <nickc@redhat.com>
18817 Add ability to create reproducible source tarballs.
18818 * src-release.sh: Add "-r <date>" option to create reproducible
18819 tarballs based upon a fixed timestamp of <date>.
18820 * binutils/README-how-to-make-a-release: Add a line showing how to
18821 use -r <date> when creating a binutils release.
18823 2022-10-10 Bruno Larsen <blarsen@redhat.com>
18825 gdb/frame: Add reinflation method for frame_info_ptr
18826 Currently, despite having a smart pointer for frame_infos, GDB may
18827 attempt to use an invalidated frame_info_ptr, which would cause internal
18828 errors to happen. One such example has been documented as PR
18829 python/28856, that happened when printing frame arguments calls an
18832 To avoid failures, the smart wrapper was changed to also cache the frame
18833 id, so the pointer can be reinflated later. For this to work, the
18834 frame-id stuff had to be moved to their own .h file, which is included
18837 Frame_id caching is done explicitly using the prepare_reinflate method.
18838 Caching is done manually so that only the pointers that need to be saved
18839 will be, and reinflating has to be done manually using the reinflate
18840 method because the get method and the -> operator must not change
18841 the internals of the class. Finally, attempting to reinflate when the
18842 pointer is being invalidated causes the following assertion errors:
18844 check_ptrace_stopped_lwp_gone: assertion `lp->stopped` failed.
18845 get_frame_pc: Assertion `frame->next != NULL` failed.
18847 As for performance concerns, my personal testing with `time make
18848 chec-perf GDB_PERFTEST_MODE=run` showed an actual reduction of around
18849 10% of time running.
18851 This commit also adds a testcase that exercises the python/28856 bug with
18852 7 different triggers, run, continue, step, backtrace, finish, up and down.
18853 Some of them can seem to be testing the same thing twice, but since this
18854 test relies on stale pointers, there is always a chance that GDB got lucky
18855 when testing, so better to test extra.
18857 Regression tested on x86_64, using both gcc and clang.
18859 Approved-by: Tom Tomey <tom@tromey.com>
18861 2022-10-10 Tom Tromey <tom@tromey.com>
18863 Change GDB to use frame_info_ptr
18864 This changes GDB to use frame_info_ptr instead of frame_info *
18865 The substitution was done with multiple sequential `sed` commands:
18867 sed 's/^struct frame_info;/class frame_info_ptr;/'
18868 sed 's/struct frame_info \*/frame_info_ptr /g' - which left some
18869 issues in a few files, that were manually fixed.
18870 sed 's/\<frame_info \*/frame_info_ptr /g'
18871 sed 's/frame_info_ptr $/frame_info_ptr/g' - used to remove whitespace
18874 The changed files were then manually checked and some 'sed' changes
18875 undone, some constructors and some gets were added, according to what
18876 made sense, and what Tromey originally did
18878 Co-Authored-By: Bruno Larsen <blarsen@redhat.com>
18879 Approved-by: Tom Tomey <tom@tromey.com>
18881 2022-10-10 Tom Tromey <tom@tromey.com>
18883 Introduce frame_info_ptr smart pointer class
18884 This adds frame_info_ptr, a smart pointer class. Every instance of
18885 the class is kept on an intrusive list. When reinit_frame_cache is
18886 called, the list is traversed and all the pointers are invalidated.
18887 This should help catch the typical GDB bug of keeping a frame_info
18888 pointer alive where a frame ID was needed instead.
18890 Co-Authored-By: Bruno Larsen <blarsen@redhat.com>
18891 Approved-by: Tom Tomey <tom@tromey.com>
18893 2022-10-10 Tom Tromey <tom@tromey.com>
18896 This replaces frame_id_eq with operator== and operator!=. I wrote
18897 this for a version of this series that I later abandoned; but since it
18898 simplifies the code, I left this patch in.
18900 Approved-by: Tom Tomey <tom@tromey.com>
18902 2022-10-10 Andrew Burgess <aburgess@redhat.com>
18904 gdb/testsuite: use 'end' at the end of python blocks
18905 Within the testsuite, use the keyword 'end' to terminate blocks of
18906 Python code being sent to GDB, rather than sending \004. I could only
18907 find three instances of this, all in tests that I originally wrote. I
18908 have no memory of there being any special reason why I used \004
18909 instead of 'end' - I assume I copied this from somewhere else that has
18912 Non of the tests being changed here are specifically about whether
18913 \004 can be used to terminate a Python block, so I think switching to
18914 the more standard 'end' keyword is the right choice.
18916 2022-10-10 Simon Marchi <simon.marchi@polymtl.ca>
18918 gdbsupport: re-generate configure
18919 I get this diff when re-generating configure, probably leftover from
18920 67d1991b785 ("egrep in binutils").
18922 Change-Id: I759c88c2bad648736d33ff98089db45c9b686356
18924 2022-10-10 Alan Modra <amodra@gmail.com>
18926 Merge configure.ac from gcc project
18927 To merge with gcc's copy of configure.ac we need to revert changes to
18928 configure.ac in the following gcc commits:
18929 dc832fb39fc0 2022-08-25
18930 fc259b522c0f 2022-06-25
18931 Then reapply configure.ac changes in binutils from these binutils
18933 50ad1254d503 2021-01-09
18934 bb368aad297f 2022-03-11
18935 e5f2f7d901ee 2022-07-26
18936 2cac01e3ffff 2022-09-26
18937 Plus copy over gcc's config/ax_cxx_compile_stdcxx.m4, then regenerate
18940 2022-10-10 GDB Administrator <gdbadmin@sourceware.org>
18942 Automatic date update in version.in
18944 2022-10-09 GDB Administrator <gdbadmin@sourceware.org>
18946 Automatic date update in version.in
18948 2022-10-08 Tom Tromey <tom@tromey.com>
18950 Merge both implementations of debug_names::insert
18951 The class debug_names has two 'insert' overloads, but only one of them
18952 is ever called externally, and it simply forwards to the other
18953 implementation. It seems cleaner to me to have a single method, so
18954 this patch merges the two.
18956 2022-10-08 Tom de Vries <tdevries@suse.de>
18958 [gdb/testsuite] Fix silent fail in gdb.server/connect-with-no-symbol-file.exp
18959 With native and target boards native-gdbserver, remote-gdbserver-on-localhost and
18960 remote-stdio-gdbserver I have for gdb.server/connect-with-no-symbol-file.exp:
18962 # of expected passes 8
18964 but with native-extended-gdbserver I have instead:
18966 # of expected passes 8
18967 # of unexpected failures 4
18970 The extra FAILs are of the form:
18973 Detaching from pid process 28985^M
18974 [Inferior 1 (process 28985) detached]^M
18975 (gdb) FAIL: gdb.server/connect-with-no-symbol-file.exp: sysroot=: \
18976 action=permission: connection to GDBserver succeeded
18978 and are due to the fact that the actual gdb output doesn't match the regexp:
18980 gdb_test "detach" \
18981 ".*Detaching from program: , process.*Ending remote debugging.*" \
18982 "connection to GDBserver succeeded"
18985 With native, the actual gdb output is:
18988 Detaching from pid process 29657^M
18989 Ending remote debugging.^M
18990 [Inferior 1 (process 29657) detached]^M
18991 (gdb) Remote debugging from host ::1, port 51028^M
18993 and because the regexp doesn't match, it triggers an implicit clause for
18994 "Ending remote debugging" in gdb_test_multiple, which has the consequence
18995 that the FAIL is silent.
18998 - the regexp by making it less strict
18999 - the silent fail by rewriting into a gdb_test_multiple, and adding an
19000 explicit fail clause.
19002 Tested on x86_64-linux, using native and aforementioned target boards.
19004 2022-10-08 GDB Administrator <gdbadmin@sourceware.org>
19006 Automatic date update in version.in
19008 2022-10-07 Lancelot SIX <lancelot.six@amd.com>
19010 gdb/testsuite: fix gdb.threads/linux-dp.exp regex
19011 On ubuntu 22.04 with the libc6-dbg package installed, I have the
19015 #0 print_philosopher (n=3, left=33 '!', right=33 '!') at .../gdb/testsuite/gdb.threads/linux-dp.c:105
19016 #1 0x000055555555576a in philosopher (data=0x55555555937c) at .../gdb/testsuite/gdb.threads/linux-dp.c:148
19017 #2 0x00007ffff7e11b43 in start_thread (arg=<optimized out>) at ./nptl/pthread_create.c:442
19018 #3 0x00007ffff7ea3a00 in clone3 () at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:81
19019 (gdb) FAIL: gdb.threads/linux-dp.exp: first thread-specific breakpoint hit
19021 The regex for this test accounts for different situations (with /
19022 without debug symbol) but assumes that if debug info is present the
19023 backtrace shows execution under pthread_create. However, for the
19024 implementation under test, we are under start_thread.
19026 Update the regex to accept start_thread.
19028 Tested on Ubuntu-22.04 x86_64 with and without libc6-dbg debug symbols
19031 Change-Id: I1e1536279890bca2cd07f038e026b41e46af44e0
19033 2022-10-07 Tom de Vries <tdevries@suse.de>
19035 [gdb/testsuite] Handle host cleanfiles
19036 When running test-case gdb.server/abspath.exp with host board
19037 local-remote-host-notty, I get:
19041 deleted: gdb/testsuite/gdb.xml/trivial.xml
19044 This happens as follows. The test-case calls skip_gdbserver_test, which calls
19045 gdb_skip_xml_test, which does:
19047 set xml_file [gdb_remote_download host "${srcdir}/gdb.xml/trivial.xml"]
19050 Then proc gdb_remote_download appends $xml_file (which for this particular
19051 host board happens to be ${srcdir}/gdb.xml/trivial.xml) to cleanfiles, which
19052 ends up being handled in gdb_finish by:
19054 eval remote_file target delete $cleanfiles
19057 The problem is that a host file is deleted using target delete.
19059 Fix this by splitting cleanfiles up in cleanfiles_target and cleanfiles_host.
19061 Tested on x86_64-linux.
19063 2022-10-07 Tom de Vries <tdevries@suse.de>
19065 [gdb/testsuite] Remove unnecessary warning in gdb.base/default.exp
19066 When running test-case gdb.base/default.exp with target board
19067 native-gdbserver, we get:
19069 WARNING: Skipping backtrace and break tests because of GDB stub.
19072 There's no need for such a warning, so remove it.
19074 Tested on x86_64-linux with native and target board native-gdbserver.
19076 2022-10-07 Tom de Vries <tdevries@suse.de>
19078 [gdb/testsuite] Fix have_mpx with remote-gdbserver-on-localhost
19079 With target board remote-gdbserver-on-localhost and gdb.arch/i386-mpx-call.exp
19082 FAIL: gdb.arch/i386-mpx-call.exp: upper_bnd0: continue to a bnd violation
19085 This is due to the have_mpx test which should return 0, but instead returns 1
19086 because the captured output:
19091 does not match the used regexp:
19093 set status [expr ($status == 0) \
19094 && ![regexp "^No MPX support\r\n" $output]]
19096 which does match the captured output with native:
19102 Fix this by making the \r in the regexp optional.
19104 Tested on x86_64-linux, with native and target board
19105 remote-gdbserver-on-localhost.
19107 2022-10-07 Tom de Vries <tdevries@suse.de>
19109 [gdb/testsuite] Fix DUPLICATEs with remote-gdbserver-on-localhost
19110 Fix some DUPLICATEs that we run into with target board
19111 remote-gdbserver-on-localhost, by using test_with_prefix.
19113 Tested on x86_64-linux, with native and target board
19114 remote-gdbserver-on-localhost.
19116 2022-10-07 Tom de Vries <tdevries@suse.de>
19118 [gdb/testsuite] Fix path in test name in gdb_load_shlib
19119 When running test-case gdb.server/solib-list.exp with target board
19120 remote-gdbserver-on-localhost, I run into:
19122 (gdb) set solib-search-path $outputs/gdb.server/solib-list^M
19123 (gdb) PASS: gdb.server/solib-list.exp: non-stop 0: \
19124 set solib-search-path $outputs/gdb.server/solib-list
19125 PATH: gdb.server/solib-list.exp: non-stop 0: \
19126 set solib-search-path $outputs/gdb.server/solib-list
19129 This is due to this code in gdb_load_shlib:
19131 gdb_test "set solib-search-path [file dirname $file]" "" ""
19134 Fix this by setting an explicit test name.
19136 Tested on x86_64-linux, with native and target boards
19137 remote-gdbserver-on-localhost, native-gdbserver and native-extended-gdbserver.
19139 2022-10-07 Alan Modra <amodra@gmail.com>
19141 PR29653, objcopy/strip: fuzzed small input file induces large output file
19142 _bfd_check_format functions should not print errors or warnings if
19143 they return NULL. A NULL return means the particular target under
19144 test does not match, so there isn't any reason to make a complaint
19145 about the target. In fact there isn't a good reason to warn even if
19146 the target matches, except via the _bfd_per_xvec_warn mechanism; Some
19147 other target might be a better match.
19149 This patch tidies pe_bfd_object_p with the above in mind, and
19150 restricts the PE optional header SectionAlignment and FileAlignment
19151 fields somewhat. I chose to warn on nonsense values rather than
19152 refusing to match. Refusing to match would be OK too.
19155 * peXXigen.c (_bfd_XXi_swap_aouthdr_in): Don't emit error about
19156 invalid NumberOfRvaAndSizes here. Limit loop copying data
19157 directory to IMAGE_NUMBEROF_DIRECTORY_ENTRIES.
19158 * peicode.h (pe_bfd_object_p): Don't clear and test bfd_error
19159 around bfd_coff_swap_aouthdr_in. Warn on invalid SectionAlignment,
19160 FileAlignment and NumberOfRvaAndSizes. Don't return NULL on
19161 invalid NumberOfRvaAndSizes.
19163 2022-10-07 GDB Administrator <gdbadmin@sourceware.org>
19165 Automatic date update in version.in
19167 2022-10-06 Tom Tromey <tromey@adacore.com>
19169 Fix indentation in riscv-tdep.c
19170 This just fixes some indentation in riscv-tdep.c.
19172 2022-10-06 Torbjörn SVENSSON <torbjorn.svensson@foss.st.com>
19174 gdb/arm: Handle lazy FPU state preservation
19175 Read LSPEN, ASPEN and LSPACT bits from FPCCR and use them together
19176 with FPCAR to identify if lazy FPU state preservation is active for
19177 the current frame. See "Lazy context save of FP state", in B1.5.7,
19178 also ARM AN298, supported by Cortex-M4F architecture for details on
19179 lazy FPU register stacking. The same conditions are valid for other
19180 Cortex-M cores with FPU.
19182 This patch has been verified on a STM32F4-Discovery board by:
19183 a) writing a non-zero value (lets use 0x1122334455667788 as an
19184 example) to all the D-registers in the main function
19185 b) configured the SysTick to fire
19186 c) in the SysTick_Handler, write some other value (lets use
19187 0x0022446688aaccee as an example) to one of the D-registers (D0 as
19188 an example) and then do "SVC #0"
19189 d) in the SVC_Handler, write some other value (lets use
19190 0x0099aabbccddeeff) to one of the D-registers (D0 as an example)
19192 In GDB, suspend the execution in the SVC_Handler function and compare
19193 the value of the D-registers for the SVC_handler frame and the
19194 SysTick_Handler frame. With the patch, the value of the modified
19195 D-register (D0) should be the new value (0x009..eff) on the
19196 SVC_Handler frame, and the intermediate value (0x002..cee) for the
19197 SysTick_Handler frame. Now compare the D-register value for the
19198 SysTick_Handler frame and the main frame. The main frame should
19199 have the initial value (0x112..788).
19201 2022-10-06 Tom de Vries <tdevries@suse.de>
19203 [gdb/symtab] Factor out have_complaint
19204 After committing 8ba677d3560 ("[gdb/symtab] Don't complain about function
19205 decls") I noticed that quite a bit of code in read_func_scope is used to decide
19206 whether to issue the "cannot get low and high bounds for subprogram DIE at
19207 $hex" complaint, which executes unnecessarily if we have the default
19208 "set complaints 0".
19211 - factoring out new static function have_complaint from macro complaint, and
19212 - using it to wrap the relevant code in read_func_scope.
19214 Tested on x86_64-linux.
19216 2022-10-06 Andrew Burgess <aburgess@redhat.com>
19218 gdb: add missing nullptr checks in bpstat_check_breakpoint_conditions
19219 Add a couple of missing nullptr checks in the function
19220 bpstat_check_breakpoint_conditions.
19222 No user visible change after this commit.
19224 2022-10-06 Andrew Burgess <aburgess@redhat.com>
19226 gdb: more infrun debug from breakpoint.c
19227 This commit adds additional infrun debug from the breakpoint.c file.
19228 The new debug output all relates to breakpoint condition evaluation.
19230 There is already some infrun debug emitted from the breakpoint.c file,
19231 so hopefully, adding more will not be contentious. I think the
19232 functions being instrumented make sense as part of the infrun process,
19233 the inferior stops, evaluates the condition, and then either stops or
19234 continues. This new debug gives more insight into that process.
19236 I had to make the bp_location* argument to find_loc_num_by_location
19237 const, and add a declaration for find_loc_num_by_location.
19239 There should be no user visible changes unless they turn on debug
19242 2022-10-06 Andrew Burgess <aburgess@redhat.com>
19244 gdb: add some additional debug in mark_async_event_handler
19245 Extend the existing debug printf call to include the previous state of
19246 the async_event_handler object.
19248 2022-10-06 Tsukasa OI <research_trasio@irq.a4lg.com>
19250 RISC-V: Print XTheadMemPair literal as "immediate"
19251 The operand type "Xl(...)" denotes that (...) is a literal. Specifically,
19252 they are intended to be a constant immediate value.
19254 This commit prints "Xl(...)" operand with dis_style_immediate style,
19255 not dis_style_text.
19259 * riscv-dis.c (print_insn_args): Use dis_style_immediate on
19260 the constant literal of the "Xl..." operand.
19262 2022-10-06 Tsukasa OI <research_trasio@irq.a4lg.com>
19264 RISC-V: Fix T-Head immediate types on printing
19265 This commit fixes two minor typing-related issues for
19266 T-Head immediate operands.
19268 1. A signed type must be specified when printing with %i.
19269 2. unsigned/signed int is not portable enough for max 32-bit immediates.
19270 Instead, we should use unsigned/signed long.
19271 The format string is changed accordingly.
19275 * riscv-dis.c (print_insn_args): Fix T-Head immediate types on
19278 2022-10-06 Tsukasa OI <research_trasio@irq.a4lg.com>
19280 RISC-V: Print comma and tabs as the "text" style
19281 On the RISC-V disassembler, some separators have non-text style when
19282 printed with another word with another style.
19284 This commit splits those, making sure that those comma and tabs are printed
19285 with the "text" style.
19289 * riscv-dis.c (print_insn_args): Split and print the comma as
19290 text. (riscv_disassemble_insn): Split and print tabs as text.
19291 (riscv_disassemble_data): Likewise.
19293 2022-10-06 Tsukasa OI <research_trasio@irq.a4lg.com>
19295 RISC-V: Optimize riscv_disassemble_data printf
19296 This commit makes types of printf arguments on riscv_disassemble_data
19297 as small as possible (as long as we can preserve the portability) to reduce
19298 the cost of printf (especially on 32-bit host).
19302 * riscv-dis.c (riscv_disassemble_data): Use smallest possible type
19305 2022-10-06 Tsukasa OI <research_trasio@irq.a4lg.com>
19307 RISC-V: Fix printf argument types corresponding %x
19308 "%x" format specifier requires unsigned type, not int. This commit
19309 fixes this issue on the RISC-V disassembler.
19313 * riscv-dis.c (print_insn_args): Fix printf argument types where
19314 the format specifier is "%x".
19316 2022-10-06 Tsukasa OI <research_trasio@irq.a4lg.com>
19318 RISC-V: Fix immediates to have "immediate" style
19319 This commit fixes certain print calls on immediate operands to have
19320 dis_style_immediate.
19324 * riscv-dis.c (print_insn_args): Fix immediates to have
19325 "immediate" style. (riscv_disassemble_data): Likewise.
19327 2022-10-06 GDB Administrator <gdbadmin@sourceware.org>
19329 Automatic date update in version.in
19331 2022-10-06 Alan Modra <amodra@gmail.com>
19333 Re: bfd BLD-POTFILES.in dependencies
19334 Removing $BLD_POTFILES from BFD-POTFILES.in was correct, but left a
19335 hole in dependencies.
19336 make[4]: Entering directory '/home/alan/build/gas/all/bfd/po'
19337 make[4]: *** No rule to make target '../elf32-aarch64.c', needed by '/home/alan/src/binutils-gdb/bfd/po/bfd.pot'. Stop.
19339 * Makefile.am (BUILT_SOURCES): Add BUILD_CFILES.
19340 * Makefile.in: Regenerate.
19342 2022-10-05 Jan Beulich <jbeulich@suse.com>
19344 x86/gas: support quoted address scale factor in AT&T syntax
19345 An earlier attempt (e68c3d59acd0 ["x86: better respect quotes in
19346 parse_operands()"]) needed undoing (cc0f96357e0b ["x86: permit
19347 parenthesized expressions again as addressing scale factor"]) as far its
19348 effect here went. As indicated back then, the issue is the backwards
19349 scanning of the operand string to find the matching opening parenthesis.
19350 Switch to forward scanning, finding the last outermost unquoted opening
19351 parenthesis (which is the one matching the trailing closing one).
19353 Arm64: support CLEARBHB alias
19354 While the Arm v8 ARM (rev I-a) still doesn't mention this alias, it is
19355 (typically via a macro) already in use in kernels and alike.
19357 2022-10-05 Alan Modra <amodra@gmail.com>
19359 PR29647, objdump -S looping
19360 Fuzzed input with this in .debug_line
19361 [0x0000003b] Special opcode 115: advance Address by 8 to 0x401180 and Line by -2 to -1
19364 * objdump.c (print_line): Don't decrement line number here..
19365 (dump_lines): ..do so here instead, ensuring loop terminates.
19367 2022-10-05 Alan Modra <amodra@gmail.com>
19369 Re: stab nearest_line bfd_malloc_and_get_section
19370 It didn't take long for the fuzzers to avoid size checks in
19371 bfd_malloc_and_get_section. Plug this hole.
19373 * syms.c (_bfd_stab_section_find_nearest_line): Ignore fuzzed
19374 sections with no contents.
19376 2022-10-05 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
19378 gprofng: fix build with --enable-pgo-build=lto
19380 2022-10-04 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
19383 * libcollector/dispatcher.c: Fix the symbol version in SYMVER_ATTRIBUTE.
19384 * libcollector/iotrace.c: Likewise.
19385 * libcollector/linetrace.c: Likewise.
19386 * libcollector/mmaptrace.c: Likewise.
19387 * libcollector/synctrace.c: Likewise.
19389 2022-10-05 GDB Administrator <gdbadmin@sourceware.org>
19391 Automatic date update in version.in
19393 2022-10-04 Tom Tromey <tom@tromey.com>
19395 Remove decode_location_spec_default
19396 This removes decode_location_spec_default, inlining it into its sole
19399 Regression tested on x86-64 Fedora 34.
19401 2022-10-04 Palmer Dabbelt <palmer@rivosinc.com>
19403 gas: NEWS: Mention the T-Head extensions that were recently added
19405 2022-10-04 Tom de Vries <tdevries@suse.de>
19407 [gdb/symtab] Don't complain about function decls
19408 [ Requires "[gdb/symtab] Don't complain about inlined functions" as
19410 https://sourceware.org/pipermail/gdb-patches/2022-September/191762.html ). ]
19412 With the test-case included in this patch, we get:
19415 During symbol reading: cannot get low and high bounds for subprogram DIE \
19417 type = int (void)^M
19418 (gdb) FAIL: gdb.dwarf2/anon-ns-fn.exp: ptype main without complaints
19421 The DIE causing the complaint is a function declaration:
19423 <2><c1>: Abbrev Number: 3 (DW_TAG_subprogram)
19424 <c2> DW_AT_name : foo
19425 <c8> DW_AT_declaration : 1
19427 which is referred to from the DIE representing the function definition:
19429 <1><f4>: Abbrev Number: 7 (DW_TAG_subprogram)
19430 <f5> DW_AT_specification: <0xc1>
19431 <f9> DW_AT_low_pc : 0x4004c7
19432 <101> DW_AT_high_pc : 0x7
19434 which does contain the low and high bounds.
19436 Fix this by not complaining about function declarations.
19438 Tested on x86_64-linux.
19440 2022-10-04 Tom de Vries <tdevries@suse.de>
19442 [gdb/symtab] Don't complain about inlined functions
19443 With the test-case included in this patch, we get:
19446 During symbol reading: cannot get low and high bounds for subprogram DIE \
19448 During symbol reading: cannot get low and high bounds for subprogram DIE \
19450 type = int (void)^M
19451 (gdb) FAIL: gdb.dwarf2/inline.exp: ptype main
19454 The complaints are about foo, with DW_AT_inline == DW_INL_inlined:
19456 <1><11f>: Abbrev Number: 6 (DW_TAG_subprogram)
19457 <120> DW_AT_name : foo
19458 <126> DW_AT_prototyped : 1
19459 <126> DW_AT_type : <0x10c>
19460 <12a> DW_AT_inline : 1 (inlined)
19462 and foo2, with DW_AT_inline == DW_INL_declared_inlined:
19464 <1><113>: Abbrev Number: 5 (DW_TAG_subprogram)
19465 <114> DW_AT_name : foo2
19466 <11a> DW_AT_prototyped : 1
19467 <11a> DW_AT_type : <0x10c>
19468 <11e> DW_AT_inline : 3 (declared as inline and inlined)
19471 Fix this by not complaining about inlined functions.
19473 Tested on x86_64-linux.
19475 2022-10-04 Tsukasa OI <research_trasio@irq.a4lg.com>
19477 gdb/riscv: Partial support for instructions up to 176-bit
19478 Because riscv_insn_length started to support instructions up to 176-bit,
19479 we need to increase buf size to 176-bit in size.
19481 Also, that would break an assumption in riscv_insn::decode so this commit
19482 fixes it, noting that instructions longer than 64-bit are not fully
19485 2022-10-04 Tsukasa OI <research_trasio@irq.a4lg.com>
19487 RISC-V: Fix buffer overflow on print_insn_riscv
19488 Because riscv_insn_length started to support instructions up to 176-bit,
19489 we need to increase packet buffer size to 176-bit in size.
19493 * opcode/riscv.h (RISCV_MAX_INSN_LEN): Max instruction length for
19494 use in buffer size.
19498 * riscv-dis.c (print_insn_riscv): Increase buffer size for max
19499 176-bit length instructions.
19501 2022-10-04 Nelson Chu <nelson@rivosinc.com>
19503 RISC-V: Renamed INSN_CLASS for floating point in integer extensions.
19504 Just added suffix _INX for those INSN_CLASS should be enough to represent
19505 their fpr can be replaced by gpr.
19507 2022-10-04 Nick Clifton <nickc@redhat.com>
19509 Note that at least dejagnu version 1.5.3 is required in order to be ale to run the testsuites.
19510 * README-maintainer-mode: Add a minimum version of dejagnu
19513 2022-10-04 Andrew Burgess <aburgess@redhat.com>
19515 opcodes/riscv: style csr names as registers
19516 While reviewing another patch I noticed that RISC-V CSR names are
19517 given the text style, not the register style. This patch fixes this
19520 2022-10-04 Luis Machado <luis.machado@arm.com>
19522 [AArch64] Update FPSR/FPCR fields for FPU and SVE
19523 I noticed some missing flags/fields from FPSR and FPCR registers in
19524 both the FPU and SVE target descriptions.
19526 This patch adds those and makes the SVE versions of FPSR and FPCR
19527 use the proper flags/bitfields types.
19529 2022-10-04 Alan Modra <amodra@gmail.com>
19531 Support objcopy changing compression to or from zstd
19532 Commit 2cac01e3ffff lacked support for objcopy changing compression
19533 style. Add that support, which meant a rewrite of
19534 bfd_compress_section_contents. In the process I've fixed some memory
19537 * compress.c (bfd_is_section_compressed_info): Rename from
19538 bfd_is_section_compressed_with_header and add ch_type param
19539 to return compression header ch_type field.
19540 Update all callers.
19541 (decompress_section_contents): Remove buffer and size params.
19542 Rewrite. Update callers.
19543 (bfd_init_section_compress_status): Free contents on failure.
19544 (bfd_compress_section): Likewise.
19545 * elf.c (_bfd_elf_make_section_from_shdr): Support objcopy
19546 changing between any of the three compression schemes. Report
19547 "unable to compress/decompress" rather than "unable to
19548 initialize compress/decompress status" on compress/decompress
19550 * bfd-in2.h: Regenerate.
19552 2022-10-04 Alan Modra <amodra@gmail.com>
19554 Re: compress .gnu.debuglto_.debug_* sections if requested
19555 Enable zlib-gnu compression for .gnu.debuglto_.debug_*. This differs
19556 from zlib-gnu for .debug_* where the name is changed to .zdebug_*.
19557 The name change isn't really needed.
19560 * elf.c (elf_fake_sections): Replace "." with ".z" in debug
19561 section names only when name was ".d*", ie. ".debug_*".
19562 (_bfd_elf_assign_file_positions_for_non_load): Likewise.
19564 * write.c (compress_debug): Compress .gnu.debuglto_.debug_*
19565 for zlib-gnu too. Compress .gnu.linkonce.wi.*.
19567 2022-10-04 Martin Liska <mliska@suse.cz>
19569 compress .gnu.debuglto_.debug_* sections if requested
19570 Right now, when using LTO, the intermediate object files do contain
19571 debug info in sections starting with .gnu.debuglto_ prefix and are
19572 not compressed when --compress-debug-sections is used.
19574 It's a mistake and we can save quite some disk space. The following
19575 example comes from tramp3d when the corresponding LTO sections
19576 are compressed with zlib:
19578 $ bloaty tramp3d-v4-v2.o -- tramp3d-v4.o
19580 -------------- --------------
19581 +83% +10 [ = ] 0 [Unmapped]
19582 -68.0% -441 [ = ] 0 .gnu.debuglto_.debug_line
19583 -52.3% -759 [ = ] 0 .gnu.debuglto_.debug_line_str
19584 -62.4% -3.24Ki [ = ] 0 .gnu.debuglto_.debug_abbrev
19585 -64.8% -1.12Mi [ = ] 0 .gnu.debuglto_.debug_info
19586 -88.8% -4.58Mi [ = ] 0 .gnu.debuglto_.debug_str
19587 -27.7% -5.70Mi [ = ] 0 TOTAL
19591 * elf.c (_bfd_elf_make_section_from_shdr): Compress all debug
19596 * write.c (compress_debug): Compress also ".gnu.debuglto_.debug_"
19597 if the compression algorithm is different from zlib-gnu.
19599 2022-10-04 Jan Beulich <jbeulich@suse.com>
19601 RISC-V/gas: allow generating up to 176-bit instructions with .insn
19602 For the time being simply utilize O_big to avoid widening other fields,
19603 bypassing append_insn() etc.
19605 RISC-V/gas: don't open-code insn_length()
19606 Use the helper when it can be used.
19608 RISC-V/gas: drop stray call to install_insn()
19609 add_fixed_insn(), by calling move_insn(), already invokes install_insn().
19611 RISC-V/gas: drop riscv_subsets static variable
19612 It's fully redundant with the subset_list member of riscv_rps_as.
19614 RISC-V: don't cast expressions' X_add_number to long in diagnostics
19615 There's no need for such workarounds anymore now that we use C99
19616 uniformly. This addresses several testsuite failures encountered when
19617 (cross-)building on a 32-bit host.
19619 2022-10-04 Potharla, Rupesh <Rupesh.Potharla@amd.com>
19621 ignore DWARF debug information for -gsplit-dwarf with dwarf-5
19622 Skip dwo_id for split dwarf.
19624 * dwarf2.c (parse_comp_unit): Skip DWO_id for DW_UT_skeleton.
19626 2022-10-04 GDB Administrator <gdbadmin@sourceware.org>
19628 Automatic date update in version.in
19630 2022-10-03 Jan-Benedict Glaw <jbglaw@lug-owl.de>
19632 Fix self-move warning check for GCC 13+
19633 GCC 13 got the self-move warning (0abb78dda084a14b3d955757c6431fff71c263f3),
19634 but that warning is only checked for clang, resulting in:
19636 /usr/lib/gcc-snapshot/bin/g++ -x c++ -I. -I. -I./config -DLOCALEDIR="\"/tmp/gdb-m68k-linux/share/locale\"" -DHAVE_CONFIG_H -I./../include/opcode -I./../readline/readline/.. -I./../zlib -I../bfd -I./../bfd -I./../include -I../libdecnumber -I./../libdecnumber -I./../gnulib/import -I../gnulib/import -I./.. -I.. -I./../libbacktrace/ -I../libbacktrace/ -DTUI=1 -I./.. -pthread -Wall -Wpointer-arith -Wno-unused -Wunused-value -Wunused-variable -Wunused-function -Wno-switch -Wno-char-subscripts -Wempty-body -Wunused-but-set-parameter -Wunused-but-set-variable -Wno-sign-compare -Wno-error=maybe-uninitialized -Wno-mismatched-tags -Wsuggest-override -Wimplicit-fallthrough=3 -Wduplicated-cond -Wshadow=local -Wdeprecated-copy -Wdeprecated-copy-dtor -Wredundant-move -Wmissing-declarations -Wstrict-null-sentinel -Wformat -Wformat-nonliteral -Werror -g -O2 -c -o unittests/environ-selftests.o -MT unittests/environ-selftests.o -MMD -MP -MF unittests/.deps/environ-selftests.Tpo unittests/environ-selftests.c
19637 unittests/environ-selftests.c: In function 'void selftests::gdb_environ_tests::test_self_move()':
19638 unittests/environ-selftests.c:228:7: error: moving 'env' of type 'gdb_environ' to itself [-Werror=self-move]
19639 228 | env = std::move (env);
19640 | ~~~~^~~~~~~~~~~~~~~~~
19641 unittests/environ-selftests.c:228:7: note: remove 'std::move' call
19642 cc1plus: all warnings being treated as errors
19643 make[1]: *** [Makefile:1896: unittests/environ-selftests.o] Error 1
19644 make[1]: Leaving directory '/var/lib/laminar/run/gdb-m68k-linux/3/binutils-gdb/gdb'
19645 make: *** [Makefile:13193: all-gdb] Error 2
19647 2022-10-03 Simon Marchi <simon.marchi@polymtl.ca>
19649 gdb: constify inferior::target_is_pushed
19650 Change-Id: Ia4143b9c63cb76e2c824ba773c66f5c5cd94b2aa
19652 2022-10-03 Luis Machado <luis.machado@arm.com>
19654 [AArch64] Handle W registers as pseudo-registers instead of aliases of X registers
19655 The aarch64 port handles W registers as aliases of X registers. This is
19656 incorrect because X registers are 64-bit and W registers are 32-bit.
19658 This patch teaches GDB how to handle W registers as pseudo-registers of
19659 32-bit, the bottom half of the X registers.
19663 2022-10-03 Luis Machado <luis.machado@arm.com>
19665 [AArch64] Fix pseudo-register numbering in the presence of unexpected additional registers
19666 When using AArch64 GDB with the QEMU debugging stub (in user mode), we get
19667 additional system registers that GDB doesn't particularly care about, so
19668 it doesn't number those explicitly.
19670 But given the pseudo-register numbers are above the number of real registers,
19671 we need to setup/account for the real registers first before going ahead and
19672 numbering the pseudo-registers. This has to happen at the end of
19673 aarch64_gdbarch_init, after the call to tdesc_use_registers, as that
19674 updates the total number of real registers.
19676 This is in preparation to supporting pointer authentication for bare metal
19679 2022-10-03 Nick Clifton <nickc@redhat.com>
19681 readelf: DO not load section headers from file offset zero
19682 * readelf.c (get_32bit_section_headers): Return false if the
19683 e_shoff field is zero.
19684 (get_64bit_section_headers): Likewise.
19686 2022-10-03 Tsukasa OI <research_trasio@irq.a4lg.com>
19688 RISC-V: Move supervisor instructions after all unprivileged ones
19689 This location of supervisor instructions is out of place (because many other
19690 privileged instructions are located at the tail but after the supervisor
19691 instructions, we have many unprivileged instructions including bit
19692 manipulation / crypto / vector instructions).
19694 Not only that, this is harmful to implement pseudoinstructions in the latest
19695 'P'-extension proposal (CLROV and RDOV). This commit moves supervisor
19696 instructions after all unprivileged instructions.
19700 * riscv-opc.c (riscv_opcodes): Adjust indents. Move supervisor
19701 instructions after all unprivileged instructions.
19703 2022-10-03 Bruno Larsen <blarsen@redhat.com>
19705 Improve GDB's baseclass detection with typedefs
19706 When a class inherits from a typedef'd baseclass, GDB may be unable to
19707 find the baseclass if the user is not using the typedef'd name, as is
19708 tested on gdb.cp/virtbase2.exp; the reason that test case is working
19709 under gcc is that the dwarf generated by gcc links the class to the
19710 original definition of the baseclass, not to the typedef. If the
19711 inheritance is linked to the typedef, such as how clang does it,
19712 gdb.cp/virtbase2.exp starts failing.
19714 This can also be seen in gdb.cp/impl-this.exp, when attempting to print
19715 D::Bint::i, and GDB not being able to find the baseclass Bint.
19717 This happens because searching for baseclasses only uses the macro
19718 TYPE_BASECLASS_NAME, which returns the typedef'd name. However, we can't
19719 switch that macro to checking for typedefs, otherwise we wouldn't be
19720 able to find the typedef'd name anymore. This is fixed by searching for
19721 members or baseclasses by name, we check both the saved name and the
19722 name after checking for typedefs.
19724 This also fixes said long-standing bug in gdb.cp/impl-this.exp when the
19725 compiler adds information about typedefs in the debuginfo.
19727 2022-10-03 Tsukasa OI <research_trasio@irq.a4lg.com>
19729 RISC-V: Assign DWARF numbers to vector registers
19730 This commit assigns DWARF register numbers to vector registers (v0-v31:
19731 96..127) to implement RISC-V DWARF Specification version 1.0-rc4
19732 (now in the frozen state):
19734 https://github.com/riscv-non-isa/riscv-elf-psabi-doc/releases/tag/v1.0-rc4
19736 binutils/ChangeLog:
19738 * dwarf.c (dwarf_regnames_riscv): Assign DWARF register numbers
19739 96..127 to vector registers v0-v31.
19743 * config/tc-riscv.c (tc_riscv_regname_to_dw2regnum): Support
19745 * testsuite/gas/riscv/dw-regnums.s: Add vector registers to the
19746 DWARF register number test.
19747 * testsuite/gas/riscv/dw-regnums.d: Likewise.
19749 2022-10-03 Tsukasa OI <research_trasio@irq.a4lg.com>
19751 RISC-V: Add testcase for DWARF register numbers
19752 Although it had csr-dw-regnums.d (for CSRs), it didn't have DWARF register
19753 number test for GPRs/FPRs.
19755 This commit adds dw-regnums.{s,d} to test such registers.
19759 * testsuite/gas/riscv/dw-regnums.s: New DWARF register number test
19761 * testsuite/gas/riscv/dw-regnums.d: Likewise.
19763 2022-10-03 GDB Administrator <gdbadmin@sourceware.org>
19765 Automatic date update in version.in
19767 2022-10-02 Tom de Vries <tdevries@suse.de>
19769 [gdb/testsuite] Fix waitpid testing in next-fork-other-thread.c
19770 In next-fork-other-thread.c, there's this loop:
19774 ret = waitpid (pid, &stat, 0);
19775 } while (ret == EINTR);
19778 The loop condition tests for "ret == EINTR" but waitpid signals EINTR by
19779 returning -1 and setting errno to EINTR.
19781 Fix this by changing the loop condition to "ret == -1 && errno == EINTR".
19783 Tested on x86_64-linux.
19785 2022-10-02 Andrew Burgess <aburgess@redhat.com>
19787 gdb/testsuite: handle invalid .exp names passed in TESTS
19788 I ran some tests like:
19790 $ make check-gdb TESTS="gdb.base/break.exp"
19792 then, then I went to rerun the tests later, I managed to corrupt the
19793 command line, like this:
19795 $ make check-gdb TESTS="gdb.base/breakff.exp"
19797 the make command did exit with an error, but DejaGnu appeared to
19798 report that every test passed! The tail end of the output looks like
19801 Illegal Argument "no-matching-tests-found"
19802 try "runtest --help" for option list
19803 === gdb Summary ===
19805 # of expected passes 115
19806 /tmp/build/gdb/gdb version 13.0.50.20220831-git -nw -nx -iex "set height 0" -iex "set width 0" -data-directory /tmp/build/gdb/testsuite/../data-directory
19808 make[3]: *** [Makefile:212: check-single] Error 1
19809 make[3]: Leaving directory '/tmp/build/gdb/testsuite'
19810 make[2]: *** [Makefile:161: check] Error 2
19811 make[2]: Leaving directory '/tmp/build/gdb/testsuite'
19812 make[1]: *** [Makefile:1916: check] Error 2
19813 make[1]: Leaving directory '/tmp/build/gdb'
19814 make: *** [Makefile:13565: check-gdb] Error 2
19816 For a while, I didn't spot that DejaGnu had failed at all, I saw the
19817 115 passes, and thought everything had run correctly - though I was
19818 puzzled that make was reporting an error.
19820 What happens is that in gdb/testsuite/Makefile, in the check-single
19821 rule, we first run DejaGnu, then run the dg-add-core-file-count.sh
19822 script, and finally, we use sed to extract the results from the
19825 In my case, with the invalid test name, DejaGnu fails, but the
19826 following steps are still run, the final result, the 115 passes, is
19827 then extracted from the pre-existing gdb.sum file.
19829 If I use 'make -jN' then the 'check-parallel' rule, rather than the
19830 'check-single' rule is used. In this case the behaviour is slightly
19831 different, the tail end of the output now looks like this:
19833 No matching tests found.
19835 make[4]: Leaving directory '/tmp/build/gdb/testsuite'
19836 find: ‘outputs’: No such file or directory
19837 Usage: ../../../src/gdb/testsuite/../../contrib/dg-extract-results.py [-t tool] [-l variant-list] [-L] log-or-sum-file ...
19839 tool The tool (e.g. g++, libffi) for which to create a
19840 new test summary file. If not specified then output
19841 is created for all tools.
19842 variant-list One or more test variant names. If the list is
19843 not specified then one is constructed from all
19844 variants in the files for <tool>.
19845 sum-file A test summary file with the format of those
19846 created by runtest from DejaGnu.
19847 If -L is used, merge *.log files instead of *.sum. In this
19848 mode the exact order of lines may not be preserved, just different
19849 Running *.exp chunks should be in correct order.
19850 find: ‘outputs’: No such file or directory
19851 Usage: ../../../src/gdb/testsuite/../../contrib/dg-extract-results.py [-t tool] [-l variant-list] [-L] log-or-sum-file ...
19853 tool The tool (e.g. g++, libffi) for which to create a
19854 new test summary file. If not specified then output
19855 is created for all tools.
19856 variant-list One or more test variant names. If the list is
19857 not specified then one is constructed from all
19858 variants in the files for <tool>.
19859 sum-file A test summary file with the format of those
19860 created by runtest from DejaGnu.
19861 If -L is used, merge *.log files instead of *.sum. In this
19862 mode the exact order of lines may not be preserved, just different
19863 Running *.exp chunks should be in correct order.
19864 make[3]: Leaving directory '/tmp/build/gdb/testsuite'
19865 make[2]: Leaving directory '/tmp/build/gdb/testsuite'
19866 make[1]: Leaving directory '/tmp/build/gdb'
19868 Rather than DejaGnu failing, we now get a nice 'No matching tests
19869 found' message, followed by some other noise. This other noise is
19870 first `find` failing, followed by the dg-extract-results.py script
19873 What happens here is that, in the check-parallel rule, the outputs
19874 directory is deleted before DejaGnu is invoked. Then we try to run
19875 all the tests, and finally we use find and dg-extract-results.py to
19876 combine all the separate .sun and .log files together. However, if
19877 there are no tests run then the outputs/ directory is never created,
19878 so the find command and consequently the dg-extract-results.py script,
19881 This commit aims to fix the following issues:
19883 (1) For check-single and check-parallel rules, don't run any of the
19884 post-processing steps if DejaGnu failed to run. This will avoid all
19885 the noise after the initial failure of DejaGnu,
19887 (2) For check-single ensure that we don't accidentally report
19888 previous results, this is related to the above, but is worth calling
19889 out as a separate point, and
19891 (3) For check-single, print the 'No matching tests found' message
19892 just like we do for a parallel test run. This makes the parallel and
19893 non-parallel testing behaviour more similar, and I think is clearer
19894 than the current 'Illegal Argument' error message.
19896 Points (1) and (2) will be handled by moving the post processing steps
19897 inside an if block within the recipe. For check-single I propose
19898 deleting the gdb.sum and gdb.log files before running DejaGnu, this is
19899 similar (I think) to how we delete the outputs/ directory in the
19900 check-parallel rule.
19902 For point (3) I plan to split the check-single rule in two, the
19903 existing check-single will be renamed do-check-single, then a new
19904 check-single rule will be added. The new check-single rule can either
19905 depend on the new do-check-single, or will ensure the 'No matching
19906 tests found' message is printed when appropriate.
19908 2022-10-02 Andrew Burgess <aburgess@redhat.com>
19910 gdb/disasm: better intel flavour disassembly styling with Pygments
19911 This commit was inspired by this stackoverflow post:
19913 https://stackoverflow.com/questions/73491793/why-is-there-a-%C2%B1-in-lea-rax-rip-%C2%B1-0xeb3
19915 One of the comments helpfully links to this Python test case:
19917 from pygments import formatters, lexers, highlight
19919 def colorize_disasm(content, gdbarch):
19921 lexer = lexers.get_lexer_by_name("asm")
19922 formatter = formatters.TerminalFormatter()
19923 return highlight(content, lexer, formatter).rstrip().encode()
19927 print(colorize_disasm("lea [rip+0x211] # COMMENT", None).decode())
19929 Run the test case and you should see that the '+' character is
19930 underlined, and could be confused with a combined +/- symbol.
19932 What's happening is that Pygments is failing to parse the input text,
19933 and the '+' is actually being marked in the error style. The error
19934 style is red and underlined.
19936 It is worth noting that the assembly instruction being disassembled
19937 here is an x86-64 instruction in the 'intel' disassembly style, rather
19938 than the default att style. Clearly the Pygments module expects the
19939 att syntax by default.
19941 If we change the test case to this:
19943 from pygments import formatters, lexers, highlight
19945 def colorize_disasm(content, gdbarch):
19947 lexer = lexers.get_lexer_by_name("asm")
19948 lexer.add_filter('raiseonerror')
19949 formatter = formatters.TerminalFormatter()
19950 return highlight(content, lexer, formatter).rstrip().encode()
19954 res = colorize_disasm("lea rax,[rip+0xeb3] # COMMENT", None)
19956 print(res.decode())
19958 print("No result!")
19960 Here I've added the call: lexer.add_filter('raiseonerror'), and I am
19961 now checking to see if the result is None or not. Running this and
19962 the test now print 'No result!' - instead of styling the '+' in the
19963 error style, we instead give up on the styling attempt.
19965 There are two things we need to fix relating to this disassembly
19966 text. First, Pygments is expecting att style disassembly, not the
19967 intel style that this example uses. Fortunately, Pygments also
19968 supports the intel style, all we need to do is use the 'nasm' lexer
19969 instead of the 'asm' lexer.
19971 However, this leads to the second problem; in our disassembler line we
19972 have '# COMMENT'. The "official" Intel disassembler style uses ';'
19973 for its comment character, however, gas and libopcodes use '#' as the
19974 comment character, as gas uses ';' for an instruction separator.
19976 Unfortunately, Pygments expects ';' as the comment character, and
19977 treats '#' as an error, which means, with the addition of the
19978 'raiseonerror' filter, that any line containing a '#' comment, will
19979 not get styled correctly.
19981 However, as the i386 disassembler never produces a '#' character other
19982 than for comments, we can easily "fix" Pygments parsing of the
19983 disassembly line. This is done by creating a filter. This filter
19984 looks for an Error token with the value '#', we then change this into
19985 a comment token. Every token after this (until the end of the line)
19986 is also converted into a comment.
19988 In this commit I do the following:
19990 1. Check the 'disassembly-flavor' setting and select between the
19991 'asm' and 'nasm' lexers based on the setting. If the setting is not
19992 available then the 'asm' lexer is used by default,
19994 2. Use "add_filter('raiseonerror')" to ensure that the formatted
19995 output will not include any error text, which would be underlined,
19996 and might be confusing,
19998 3. If the 'nasm' lexer is selected, then add an additional filter
19999 that will format '#' and all other text on the line, as a comment,
20002 4. If Pygments throws an exception, instead of returning None,
20003 return the original, unmodified content. This will mean that this
20004 one instruction is printed without styling, but GDB will continue to
20005 call into the Python code to style later instructions.
20007 I haven't included a test specifically for the above error case,
20008 though I have manually check that the above case now styles
20009 correctly (with no underline). The existing style tests check that
20010 the disassembler styling still works though, so I know I've not
20011 generally broken things.
20013 One final thought I have after looking at this issue is that I wonder
20014 now if using Pygments for styling disassembly from every architecture
20015 is actually a good idea?
20017 Clearly, the 'asm' lexer is OK with att style x86-64, but not OK with
20018 intel style x86-64, so who knows how well it will handle other random
20021 When I first added this feature I tested it against some random
20022 RISC-V, ARM, and X86-64 (att style) code, and it seemed fine, but I
20023 never tried to make an exhaustive check of all instructions, so its
20024 quite possible that there are corner cases where things are styled
20027 With the above changes I think that things should be a bit better
20028 now. If a particular instruction doesn't parse correctly then our
20029 Pygments based styling code will just not style that one instruction.
20030 This is combined with the fact that many architectures are now moving
20031 to libopcodes based styling, which is much more reliable.
20033 So, I think it is fine to keep using Pygments as a fallback mechanism
20034 for styling all architectures, even if we know it might not be perfect
20037 2022-10-02 Andrew Burgess <aburgess@redhat.com>
20039 gdb: improve disassembler styling when Pygments raises an exception
20040 While working on another issue relating to GDB's use of the Python
20041 Pygments package for disassembly styling I noticed an issue in the
20042 case where the Pygments package raises an exception.
20044 The intention of the current code is that, should the Pygments package
20045 raise an exception, GDB will disable future attempts to call into the
20046 Pygments code. This was intended to prevent repeated errors during
20047 disassembly if, for some reason, the Pygments code isn't working.
20049 Since the Pygments based styling was added, GDB now supports
20050 disassembly styling using libopcodes, but this is only available for
20051 some architectures. For architectures not covered by libopcodes
20052 Pygments is still the only option.
20054 What I observed is that, if I disable the libopcodes styling, then
20055 setup GDB so that the Pygments based styling code will indicate an
20056 error (by returning None), GDB does, as expected, stop using the
20057 Pygments based styling. However, the libopcodes based styling will
20058 instead be used, despite this feature having been disabled.
20060 The problem is that the disassembler output is produced into a
20061 string_file buffer. When we are using Pygments, this buffer is
20062 created without styling support. However, when Pygments fails, we
20063 recreate the buffer with styling support.
20065 The problem is that we should only recreate the buffer with styling
20066 support only if libopcodes styling is enabled. This was an oversight
20069 commit 4cbe4ca5da5cd7e1e6331ce11f024bf3c07b9744
20070 Date: Mon Feb 14 14:40:52 2022 +0000
20072 gdb: add support for disassembler styling using libopcodes
20076 1. Factors out some of the condition checking logic into two new
20077 helper functions use_ext_lang_for_styling and
20078 use_libopcodes_for_styling,
20080 2. Reorders gdb_disassembler::m_buffer and gdb_disassembler::m_dest,
20081 this allows these fields to be initialised m_dest first, which means
20082 that the new condition checking functions can rely on m_dest being
20083 set, even when called from the gdb_disassembler constructor,
20085 3. Make use of the new condition checking functions each time
20086 m_buffer is initialised,
20088 4. Add a new test that forces the Python disassembler styling
20089 function to return None, this will cause GDB to disable use of
20090 Pygments for styling, and
20092 5. When we reinitialise m_buffer, and re-disassemble the
20093 instruction, call reset the in-comment flag. If the instruction
20094 being disassembler ends in a comment then the first disassembly pass
20095 will have set the in-comment flag to true. This shouldn't be a
20096 problem as we will only be using Pygments, and thus performing a
20097 re-disassembly pass, if libopcodes is disabled, so the in-comment
20098 flag will never be checked, even if it is set incorrectly. However,
20099 I think that having the flag set correctly is a good thing, even if
20100 we don't check it (you never know what future uses might come up).
20102 2022-10-02 Andrew Burgess <aburgess@redhat.com>
20104 gdb/testsuite: extend styling test for libopcodes styling
20105 This commit extends the gdb.base/style.exp test to cover disassembler
20106 styling using libopcodes (where available).
20108 The test will try to enable libopcode based styling, if this
20109 works (because such styling is available) then some tests are run to
20110 check that the output is styled, and that the styling can be disabled
20111 using 'set style disassembler enabled off'. If libopcodes styling is
20112 not available on the current architecture then these new tests will be
20115 I've moved the Python Pygments module check inside the
20116 test_disable_disassembler_styling function now, so that the test will
20117 be run even when Python Pygments is not available, this allows the new
20118 tests discussed above.
20120 If the Pygments module is not available then the Pygments based tests
20121 will be skipped just as they were before.
20123 2022-10-02 Andrew Burgess <aburgess@redhat.com>
20125 gdb: update now gdbarch_register_name doesn't return nullptr
20126 After the previous few commit, gdbarch_register_name no longer returns
20127 nullptr. This commit audits all the calls to gdbarch_register_name
20128 and removes any code that checks the result against nullptr.
20130 There should be no visible change after this commit.
20132 2022-10-02 Andrew Burgess <aburgess@redhat.com>
20134 gdb: final cleanup of various gdbarch_register_name methods
20135 Building on the previous commits, this commit goes through the various
20136 gdbarch_register_name methods and removes all the remaining 'return
20137 NULL' cases, I claim that these either couldn't be hit, or should be
20138 returning the empty string.
20140 In all cases the return of NULL was used if the register number being
20141 passed to gdbarch_register_name was "invalid", i.e. negative, or
20142 greater than the total number of declared registers. I don't believe
20143 either of these cases can occur, and the previous commit asserts that
20144 this is the case. As a result we can simplify the code by removing
20145 these checks. In many cases, where the register names are held in an
20146 array, I was able to add a static assert that the array contains the
20147 correct number of strings, after that, a direct access into the array
20150 I don't have any means of testing these changes.
20152 2022-10-02 Andrew Burgess <aburgess@redhat.com>
20154 gdb/csky: remove nullptr return from csky_pseudo_register_name
20155 Building on the previous commits, in this commit I remove two
20156 instances of 'return NULL' from csky_pseudo_register_name, and replace
20157 them with a return of the empty string.
20159 These two are particularly interesting, and worth pulling into their
20160 own commit, because these returns of NULL appear to be depended on
20161 within other parts of the csky code.
20163 In csky-linux-tdep.c in the register collect/supply code, GDB checks
20164 for the register name being nullptr in order to decide if a target
20165 supports a particular feature or not. I've updated the code to check
20166 for the empty string.
20168 I have no way of testing this change.
20170 2022-10-02 Andrew Burgess <aburgess@redhat.com>
20172 gdb: add asserts to gdbarch_register_name
20173 This commit adds asserts to gdbarch_register_name that validate the
20174 parameters, and the return value.
20176 The interesting thing here is that gdbarch_register_name is generated
20177 by gdbarch.py, and so, to add these asserts, I need to update the
20180 I've added two new arguments for Functions and Methods (as declared in
20181 gdbarch-components.py), these arguments are 'param_checks' and
20182 'result_checks'. Each of these new arguments can be used to list some
20183 expressions that are then used within gdb_assert calls in the
20186 The asserts that validate the API as described in the comment I added
20187 to gdbarch_register_name a few commits back; the register number
20188 passed in needs to be a valid cooked register number, and the result
20189 being returned should not be nullptr.
20191 2022-10-02 Andrew Burgess <aburgess@redhat.com>
20193 gdb: check for duplicate register names in selftest
20194 Building on the previous commit, this commit extends the register_name
20195 selftest to check for duplicate register names.
20197 If two registers in the cooked register set (real + pseudo registers)
20198 have the same name, then this will show up as duplicate registers in
20199 the 'info all-registers' output, but the user will only be able to
20200 interact with one copy of the register.
20202 In this commit I extend the selftest that I added in the previous
20203 commit to check for duplicate register names, I didn't include this
20204 functionality in the previous commit because one architecture needed
20205 fixing, and I wanted to keep those fixes separate from the fixes in
20206 the previous commit.
20208 The problematic architecture(s) are powerpc:750 and powerpc:604. In
20209 both of these cases the 'dabr' register appears twice, there's a
20210 definition of dabr in power-oea.xml which is included into both
20211 powerpc-604.xml and powerpc-750.xml. Both of these later two xml
20212 files also define the dabr register.
20214 I'm hopeful that this change shouldn't break anything, but I don't
20215 have the ability to actually test this change, however:
20217 On the gdbserver side, neither powerpc-604.xml nor powerpc-750.xml are
20218 mentioned in gdbserver/configure.srv, which I think means that
20219 gdbserver will never use these descriptions, and,
20221 Within GDB the problematic descriptions are held in the variables
20222 tdesc_powerpc_604 and tdesc_powerpc_750, which are only mentioned in
20223 the variants array in rs6000-tdep.c, this is used when looking up a
20224 description based on the architecture.
20226 For a native Linux target however, this will not be used as
20227 ppc_linux_nat_target::read_description exists, which calls
20228 ppc_linux_match_description, which I don't believe can return either
20229 of the problematic descriptions.
20231 This leaves the other native targets, FreeBSD, AIX, etc. These don't
20232 appear to override the ::read_description method, so will potentially
20233 return the problematic descriptions, but, in each case I think the
20234 ::fetch_registers and ::store_registers methods will ignore the dabr
20235 register, which will leave the register as <unavailable>.
20237 So, my proposed solution is to just remove the duplicate register from
20238 each of powerpc-604.xml and powerpc-750.xml, then regenerate the
20239 corresponding C++ source file. With this change made, the selftest
20240 now passes for all architectures.
20242 2022-10-02 Andrew Burgess <aburgess@redhat.com>
20244 gdb: add a gdbarch_register_name self test, and fix some architectures
20245 This commit adds a self-test that checks that gdbarch_register_name
20246 never returns nullptr for any valid register number.
20248 Most architectures already met this requirement, there were just 6
20249 that failed the new selftest, and are updated in this commit.
20251 Beyond the self-tests I don't have any facilities to test that the
20252 architectures I've adjusted still work correctly.
20254 If you review all the various gdbarch_register_name implementations
20255 then you will see that there are far more architectures that seem like
20256 they might return nullptr in some situations, e.g. alpha, avr, bpf,
20257 etc. This commit doesn't attempt to address these cases as non of
20258 them are hit during the selftest. Many of these cases can never be
20259 hit, for example, in alpha_register_name GDB checks for a register
20260 number less than zero, this case can't happen and could be changed
20263 A later commit in this series will have a general cleanup of all the
20264 various register_name methods, and remove all references to NULL from
20265 their code, however, as that commit will be mostly adjusting code that
20266 is never hit, I want to keep those changes separate.
20268 The selftest has been tested on x86-64, but I don't have access to
20269 suitable systems to fully test any of the *-tdep.c code I've changed
20272 2022-10-02 Andrew Burgess <aburgess@redhat.com>
20274 gdb/gdbarch: add a comment to gdbarch_register_name
20275 After the previous commit, this commit sets out to formalise the API
20276 for gdbarch_register_name. Not every architecture is actually in
20277 compliance with the API I set out here, but I believe that most are.
20279 I think architectures that don't comply with the API laid out here
20280 will fail the gdb.base/completion.exp test.
20282 The claims in the comment are I feel, best demonstrated with the
20283 asserts in this code:
20286 gdbarch_register_name (struct gdbarch *gdbarch, int regnr)
20288 gdb_assert (regnr >= 0);
20289 gdb_assert (regnr < gdbarch_num_cooked_regs (gdbarch));
20291 const char *name = gdbarch->register_name (gdbarch, regnr);
20293 gdb_assert (name != nullptr);
20298 Like I said, I don't believe every architecture follows these rules
20299 right now, which is why I'm not actually adding any asserts. Instead,
20300 this commit adds a comment to gdbarch_register_name, this comment is
20301 where I'd like to get to, rather than where we are right now.
20303 Subsequent commits will fix all targets to be in compliance with this
20304 comment, and will even add the asserts shown above to
20305 gdbarch_register_name.
20307 2022-10-02 Andrew Burgess <aburgess@redhat.com>
20309 gdb/riscv: fix failure in gdb.base/completion.exp
20310 I noticed a test failure in gdb.base/completion.exp for RISC-V on
20311 a native Linux target, this is the failure:
20313 (gdb) FAIL: gdb.base/completion.exp: complete 'info registers '
20315 The problem is caused by a mismatch in the output of 'maint print
20316 registers' and the completion list for 'info registers'. The 'info
20317 registers' completion list contains less registers than
20318 expected. Additionally, the list of registers extracted from the
20319 'maint print registers' list was wrong too, in some cases the test was
20320 grabbing the register number, rather than a register name,
20322 Both of these problems have the same root cause, riscv_register_name
20323 returns nullptr for some registers when it should return an empty
20326 The gdbarch_register_name API is not clearly documented anywhere, and
20327 at first glance it would appear that the function can return either
20328 nullptr, or an empty string to indicate that a register is not
20329 available on the current target. Indeed, there are plenty of places
20330 in GDB where we compare the output of gdbarch_register_name to both
20331 nullptr and '\0' in order to see if a register is supported or not,
20332 and there are plenty of targets that return empty string in some
20333 cases, and nullptr in others.
20335 However, the 'info registers' completion code (reg_or_group_completer)
20336 clearly depends on user_reg_map_regnum_to_name only returning nullptr
20337 when the passed in regnum is greater than the maximum possible
20338 register number (i.e. after all physical registers, pseudo-registers,
20339 and user-registers), this means that gdbarch_register_name should not
20340 be returning nullptr.
20342 I did consider "fixing" user_reg_map_regnum_to_name, if
20343 gdbarch_register_name returns nullptr, I could convert to an empty
20344 string at this point, but that felt like a real hack, so I discarded
20347 The next possibility I considered was "fixing" reg_or_group_completer
20348 to not rely on nullptr to indicate the end marker. Or rather, I could
20349 have reg_or_group_completer use gdbarch_num_cooked_regs, we know that
20350 we should check at least that many register numbers. Then, once we're
20351 passed that limit, we keep checking until we hit a nullptr. This
20352 would absolutely work, and didn't actually feel that bad, but, it
20353 still felt a little weird that gdbarch_register_name could return
20354 nullptr OR the empty string to mean the same thing, so I wondered if
20355 the "right" solution was to have gdbarch_register_name not return
20356 nullptr. With this in mind I tried an experiment:
20358 I added a self-test that, for each architecture, calls
20359 gdbarch_register_name for every register number up to the
20360 gdbarch_num_cooked_regs limit, and checks that the name is not
20363 Only a handful of architectures failed this test, RISC-V being one of
20366 This seems to suggest that most architectures agree that the correct
20367 API for gdbarch_register_name is to return an empty string for
20368 registers that are not supported on the current target, and that
20369 returning nullptr is really a mistake.
20371 In this commit I will update the RISC-V target so that GDB no longer
20372 returns nullptr from riscv_register_name, instead we return the empty
20375 In subsequent commits I will add the selftest that I mention above,
20376 and will fix the targets that fail the selftest.
20378 With this change the gdb.base/completion.exp test now passes.
20380 2022-10-02 Andrew Burgess <aburgess@redhat.com>
20382 gdb/testsuite: rewrite capture_command_output proc
20383 I noticed a test failure in gdb.base/completion.exp for RISC-V on a
20384 native Linux target. Upon investigation I discovered a couple of
20385 reasons for the failure, this commit addresses one of them. A later
20386 commit will address the other issue.
20388 The completion.exp test makes use of the capture_command_output proc
20389 to collect the output of the 'maint print registers' command. For
20390 RISC-V this command produces a lot of output.
20392 Currently the capture_command_output proc tries to collect the
20393 complete command output in a single expect buffer, and what I see is
20394 an error caused by the expect buffer becoming full.
20396 This commit rewrites capture_command_output to make use of
20397 gdb_test_multiple to collect the command output line at a time, in
20398 this way we avoid overflowing the expect buffer.
20400 The capture_command_output proc has some logic for skipping a prefix
20401 pattern, which is passed in to the proc as an argument. In order to
20402 handle this correctly (only matching the prefix at the start of the
20403 command output), I use two gdb_test_multiple calls, the first spots
20404 and discards the echoed command and the (optional) prefix pattern, the
20405 second gdb_test_multiple call then collects the rest of the command
20406 output line at a time until a prompt is seen.
20408 There is one slight oddity with the current implementation, which I
20409 have changed in my rewrite, this does, slightly, change the behaviour
20412 The current implementation uses this pattern:
20414 -re "[string_to_regexp ${command}]\[\r\n\]+${prefix}(.*)\[\r\n\]+$gdb_prompt $"
20416 Now a typical command output will look like this:
20421 As the TCL regexp matching is greedy, TCL will try to match as much as
20422 possible in one part of the pattern before moving on to the next.
20423 Thus, when this matches against (.*)[\r\n]+, the (.*) will end up
20424 matching against 'output here\r' and the [\r\n]+ will match '\n' only.
20426 In short the previous implementation would leave the '\r' on the end
20427 of the returned text, but not the final trailing '\n'.
20429 Now clearly I could make a new version of capture_command_output that
20430 maintained this behaviour, but I couldn't see any reason to do this.
20431 So, my new implementation drops the final '\r\n' completely, in our
20432 example above we now return 'output here' with no '\r'.
20434 This change doesn't seem to affect any of the existing tests, but I
20435 thought it was worth mentioning.
20437 2022-10-02 Andrew Burgess <aburgess@redhat.com>
20439 gdb/mi: new options for -data-disassemble command
20440 Now that the disassembler has two different strategies for laying out
20441 the opcode bytes of an instruction (see /r vs /b for the disassemble
20442 command), I wanted to add support for this to the MI disassemble
20445 Currently the -data-disassemble command takes a single 'mode' value,
20446 which currently has 6 different values (0 -> 5), 3 of these modes
20447 relate to opcode display.
20449 So, clearly I should just add an additional 3 modes to handle the new
20450 opcode format, right?
20452 No, I didn't think that was a great idea either.
20454 So, I wonder, could I adjust the -data-disassemble command in a
20455 backward compatible way, that would allow GDB to move away from using
20456 the mode value altogether?
20460 In this commit, I propose adding two new options to -data-disassemble,
20463 --opcodes none|bytes|display
20466 Additionally, I will make the mode optional, and default to mode 0 if
20467 no mode value is given. Mode 0 is the simplest, no source code, no
20468 opcodes disassembly mode.
20470 The two new options are only valid for mode 0, if they are used with
20471 any other mode then an error is thrown.
20473 The --opcodes option can add opcodes to the result, with 'bytes' being
20474 equivalent to 'disassemble /b' and 'display' being 'disassemble /r'.
20476 The --source option will enable the /s style source code display, this
20477 is equivalent to modes 4 and 5. There is no way, using the new
20478 command options to get the now deprecated /m style source code
20479 display, that is mode 1 and 3.
20481 My hope is that new users of the MI will not use the mode at all, and
20482 instead will just use the new options to achieve the output they need.
20483 Existing MI users can continue to use the mode, and will not need to
20484 be updated to use the new options.
20486 2022-10-02 Andrew Burgess <aburgess@redhat.com>
20488 gdb/mi: some int to bool conversion
20489 Just some simple int to bool conversion in mi_cmd_disassemble. There
20490 should be no user visible changes after this commit.
20492 2022-10-02 Andrew Burgess <aburgess@redhat.com>
20494 gdb/doc: update syntax of -data-disassemble command arguments
20495 The argument documentation for -data-disassemble looks like this:
20498 [ -s @var{start-addr} -e @var{end-addr} ]
20499 | [ -a @var{addr} ]
20500 | [ -f @var{filename} -l @var{linenum} [ -n @var{lines} ] ]
20503 However, I believe, according to the 'Notation and Terminology'
20504 section, this means that the there are 3 optional location
20505 specification argument groups for the command, followed by a
20506 non-optional '-- mode'.
20508 However, this is not true, one of the location specifications must be
20509 given, i.e. we can't choose to give NO location specification, which
20510 is what the above implies.
20512 I propose that we change this to instead be:
20515 ( -s @var{start-addr} -e @var{end-addr}
20517 | -f @var{filename} -l @var{linenum} [ -n @var{lines} ] )
20520 By placing all the location specifications within '( ... )' we
20521 indication that these are a group, from which one of the options,
20522 separated by '|', must be selected.
20524 However, the 'Notation and Terminology' section only describes two
20525 uses for parenthesis: '( GROUP )*' and '( GROUP )+', in the first case
20526 GROUP is repeated zero or more times, and in the second GROUP is
20527 repeated 1 or more times.
20529 Neither of those exactly describe what I want, which is GROUP must
20530 appear exactly once. I propose to extend 'Notation and Terminology'
20531 to include '( GROUP )' which means that GROUP should appear exactly
20534 This change is important because, in a later commit, I want to add
20535 additional optional arguments to the -data-disassemble command, and
20536 things start to get confusing with the original syntax.
20538 2022-10-02 Andrew Burgess <aburgess@redhat.com>
20540 gdb: make gdb_disassembly_flag unsigned
20541 In a later commit I want to use operator~ on a gdb_disassembly_flag
20542 flag value. This is currently not possible as gdb_disassembly_flag
20543 is, by default, signed.
20545 This commit just makes this enum unsigned.
20547 There should be no user visible changes after this commit.
20549 2022-10-02 Andrew Burgess <aburgess@redhat.com>
20551 gdb: disassembler opcode display formatting
20552 This commit changes the format of 'disassemble /r' to match GNU
20553 objdump. Specifically, GDB will now display the instruction bytes in
20554 as 'objdump --wide --disassemble' does.
20556 Here is an example for RISC-V before this patch:
20558 (gdb) disassemble /r 0x0001018e,0x0001019e
20559 Dump of assembler code from 0x1018e to 0x1019e:
20560 0x0001018e <call_me+66>: 03 26 84 fe lw a2,-24(s0)
20561 0x00010192 <call_me+70>: 83 25 c4 fe lw a1,-20(s0)
20562 0x00010196 <call_me+74>: 61 65 lui a0,0x18
20563 0x00010198 <call_me+76>: 13 05 85 6a addi a0,a0,1704
20564 0x0001019c <call_me+80>: f1 22 jal 0x10368 <printf>
20565 End of assembler dump.
20567 And here's an example after this patch:
20569 (gdb) disassemble /r 0x0001018e,0x0001019e
20570 Dump of assembler code from 0x1018e to 0x1019e:
20571 0x0001018e <call_me+66>: fe842603 lw a2,-24(s0)
20572 0x00010192 <call_me+70>: fec42583 lw a1,-20(s0)
20573 0x00010196 <call_me+74>: 6561 lui a0,0x18
20574 0x00010198 <call_me+76>: 6a850513 addi a0,a0,1704
20575 0x0001019c <call_me+80>: 22f1 jal 0x10368 <printf>
20576 End of assembler dump.
20578 There are two differences here. First, the instruction bytes after
20579 the patch are grouped based on the size of the instruction, and are
20580 byte-swapped to little-endian order.
20582 Second, after the patch, GDB now uses the bytes-per-line hint from
20583 libopcodes to add whitespace padding after the opcode bytes, this
20584 means that in most cases the instructions are nicely aligned.
20586 It is still possible for a very long instruction to intrude into the
20587 disassembled text space. The next example is x86-64, before the
20590 (gdb) disassemble /r main
20591 Dump of assembler code for function main:
20592 0x0000000000401106 <+0>: 55 push %rbp
20593 0x0000000000401107 <+1>: 48 89 e5 mov %rsp,%rbp
20594 0x000000000040110a <+4>: c7 87 d8 00 00 00 01 00 00 00 movl $0x1,0xd8(%rdi)
20595 0x0000000000401114 <+14>: b8 00 00 00 00 mov $0x0,%eax
20596 0x0000000000401119 <+19>: 5d pop %rbp
20597 0x000000000040111a <+20>: c3 ret
20598 End of assembler dump.
20600 And after the patch:
20602 (gdb) disassemble /r main
20603 Dump of assembler code for function main:
20604 0x0000000000401106 <+0>: 55 push %rbp
20605 0x0000000000401107 <+1>: 48 89 e5 mov %rsp,%rbp
20606 0x000000000040110a <+4>: c7 87 d8 00 00 00 01 00 00 00 movl $0x1,0xd8(%rdi)
20607 0x0000000000401114 <+14>: b8 00 00 00 00 mov $0x0,%eax
20608 0x0000000000401119 <+19>: 5d pop %rbp
20609 0x000000000040111a <+20>: c3 ret
20610 End of assembler dump.
20612 Most instructions are aligned, except for the very long instruction.
20613 Notice too that for x86-64 libopcodes doesn't request that GDB group
20614 the instruction bytes. This matches the behaviour of objdump.
20616 In case the user really wants the old behaviour, I have added a new
20617 modifier 'disassemble /b', this displays the instruction byte at a
20618 time. For x86-64, which never groups instruction bytes, /b and /r are
20619 equivalent, but for RISC-V, using /b gets the old layout back (except
20620 that the whitespace for alignment is still present). Consider our
20621 original RISC-V example, this time using /b:
20623 (gdb) disassemble /b 0x0001018e,0x0001019e
20624 Dump of assembler code from 0x1018e to 0x1019e:
20625 0x0001018e <call_me+66>: 03 26 84 fe lw a2,-24(s0)
20626 0x00010192 <call_me+70>: 83 25 c4 fe lw a1,-20(s0)
20627 0x00010196 <call_me+74>: 61 65 lui a0,0x18
20628 0x00010198 <call_me+76>: 13 05 85 6a addi a0,a0,1704
20629 0x0001019c <call_me+80>: f1 22 jal 0x10368 <printf>
20630 End of assembler dump.
20632 Obviously, this patch is a potentially significant change to the
20633 behaviour or /r. I could have added /b with the new behaviour and
20634 left /r alone. However, personally, I feel the new behaviour is
20635 significantly better than the old, hence, I made /r be what I consider
20636 the "better" behaviour.
20638 The reason I prefer the new behaviour is that, when I use /r, I almost
20639 always want to manually decode the instruction for some reason, and
20640 having the bytes displayed in "instruction order" rather than memory
20641 order, just makes this easier.
20643 The 'record instruction-history' command also takes a /r modifier, and
20644 has been modified in the same way as disassemble; /r gets the new
20645 behaviour, and /b has been added to retain the old behaviour.
20647 Finally, the MI command -data-disassemble, is unchanged in behaviour,
20648 this command now requests the raw bytes of the instruction, which is
20649 equivalent to the /b modifier. This means that the MI output will
20650 remain backward compatible.
20652 2022-10-02 Andrew Burgess <aburgess@redhat.com>
20654 gdb/disasm: read opcodes bytes with a single read_code call
20655 This commit reduces the number of times we call read_code when
20656 printing the instruction opcode bytes during disassembly.
20658 I've added a new gdb::byte_vector within the
20659 gdb_pretty_print_disassembler class, in line with all the other
20660 buffers that gdb_pretty_print_disassembler needs. This byte_vector is
20661 then resized as needed, and filled with a single read_code call for
20664 There should be no user visible changes after this commit.
20666 2022-10-02 Andrew Burgess <aburgess@redhat.com>
20668 gdb/testsuite: new test for -data-disassemble opcodes format
20669 Add another test for the output of MI command -data-disassemble. The
20670 new check validates the format of the 'opcodes' field, specifically,
20671 this test checks that the field contains a series of bytes, separated
20674 We also check that the bytes are in the correct order, that is, the
20675 first byte is from the lowest address, and subsequent bytes are from
20676 increasing addresses.
20678 The motivation for this test (besides more tests being generally good)
20679 is that I plan to make changes to how opcode bytes are displayed in
20680 the disassembler output, and I want to ensure that I don't break any
20681 existing MI behaviour.
20683 There should be no user visible changes to GDB after this commit.
20685 2022-10-02 GDB Administrator <gdbadmin@sourceware.org>
20687 Automatic date update in version.in
20689 2022-10-01 GDB Administrator <gdbadmin@sourceware.org>
20691 Automatic date update in version.in
20693 2022-09-30 Tsukasa OI <research_trasio@irq.a4lg.com>
20695 RISC-V: Relax "fmv.[sdq]" requirements
20696 This commit relaxes requirements to "fmv.s" instructions from 'F' to ('F'
20697 or 'Zfinx'). The same applies to "fmv.d" and "fmv.q". Note that 'Zhinx'
20698 extension already contains "fmv.h" instruction (as well as 'Zfh').
20702 * testsuite/gas/riscv/zfinx.s: Add "fmv.s" instruction.
20703 * testsuite/gas/riscv/zfinx.d: Likewise.
20704 * testsuite/gas/riscv/zdinx.s: Add "fmv.d" instruction.
20705 * testsuite/gas/riscv/zdinx.d: Likewise.
20706 * testsuite/gas/riscv/zqinx.d: Add "fmv.q" instruction.
20707 * testsuite/gas/riscv/zqinx.s: Likewise.
20711 * riscv-opc.c (riscv_opcodes): Relax requirements to "fmv.[sdq]"
20712 instructions to support those in 'Zfinx'/'Zdinx'/'Zqinx'.
20714 2022-09-30 Tsukasa OI <research_trasio@irq.a4lg.com>
20716 RISC-V: Reorganize and enhance 'Zfinx' tests
20717 This commit adds certain test cases for 'Zfinx'/'Zdinx'/'Zqinx' extensions
20718 and reorganizes them, fixing coding style while improving coverage.
20719 This is partially based on jiawei's 'Zhinx' testcases.
20723 * testsuite/gas/riscv/zfinx.s: Use different registers for
20724 better encode space testing. Make indentation consistent.
20725 Add tests for instruction with rounding mode. Change march
20726 to minimum required extensions. Remove source line.
20727 * testsuite/gas/riscv/zfinx.d: Likewise.
20728 * testsuite/gas/riscv/zdinx.s: Likewise.
20729 * testsuite/gas/riscv/zdinx.d: Likewise.
20730 * testsuite/gas/riscv/zqinx.s: Likewise.
20731 Also use even-numbered registers to use valid register pairs.
20732 * testsuite/gas/riscv/zqinx.d: Likewise.
20734 2022-09-30 Christoph Müllner <christoph.muellner@vrull.eu>
20736 RISC-V: Eliminate long-casts of X_add_number in diagnostics
20737 There is no need for casts to (signed/unsigned) long, as we can use
20738 C99's PRId64/PRIu64 format specifiers.
20740 2022-09-30 Jan Beulich <jbeulich@suse.com>
20742 RISC-V: fallout from "re-arrange opcode table for consistent alias handling"
20743 Several new testcasee have appeared since the submission of said change,
20744 some of which now also need adjustment.
20746 RISC-V: fix build after "Add support for arbitrary immediate encoding formats"
20747 Pre- and post-increment/decrement are side effects, the behavior of
20748 which is undefined when combined with passing an address of the accessed
20749 variable in the same function invocation. There's no need for the
20750 increments here - simply adding 1 achieves the intended effect without
20751 triggering compiler diagnostics (which are fatal with -Werror).
20753 objcopy: avoid "shadowing" of remove() function name
20754 remove() is a standard library function (declared in stdio.h), which
20755 triggers a "shadows a global declaration" warning with some gcc versions.
20757 RISC-V: drop stray INSN_ALIAS flags
20758 FENCE.TSO isn't an alias. ZIP and UNZIP in the long run likely are, but
20759 presently they aren't. This fixes disassembly of these insns with
20762 2022-09-30 Jan Beulich <jbeulich@suse.com>
20764 RISC-V: re-arrange opcode table for consistent alias handling
20765 For disassembly to pick up aliases in favor of underlying insns (helping
20766 readability in the common case), the aliases need to come ahead of the
20767 "base" insns. Slightly more code movement is needed because of insns
20768 with the same name needing to stay next to each other.
20770 Note that the "rorw" alias entry also has the missing INSN_ALIAS added
20773 Clone a few testcases to exercise -Mno-aliases some more, better
20774 covering the differences between the default and that disassembly mode.
20776 2022-09-30 Jan Beulich <jbeulich@suse.com>
20778 x86: correct build dependencies in opcodes/
20779 With the command in the rule merely being "echo", i386-tbl.h won't be
20780 rebuilt if missing, when at the same time i386-init.h is present and
20781 up-to-date. Use a pattern rule instead to express the multiple targets
20782 correctly (the &: rule separator is supported only by GNU make 4.3 and
20783 newer). Note that now, for the opposite case to work (i386-tbl.h is
20784 up-to-date but i386-init.h is missing), i386-init.h also needs
20785 mentioning as a dependency somewhere: Add a fake dependency for
20786 i386-opc.lo ("fake" because i386-opc.c doesn't include that header).
20788 At the same time use $(AM_V_GEN) in the actual rule, replacing the
20789 earlier (open-coded) "echo". And while there also drop a duplicate
20790 dependency of i386-gen.o on i386-opc.h.
20792 2022-09-30 Jan Beulich <jbeulich@suse.com>
20794 x86: improve match_template()'s diagnostics
20797 extractps $0, %xmm0, %xmm0
20798 insertps $0, %xmm0, %eax
20800 (both having respectively the same mistake of using the wrong kind of
20801 destination register) it is easy to see that current behavior is far
20802 from ideal: The former results in "unsupported instruction" for 32-bit
20803 code simply because the 2nd template we have is a Cpu64 one. Instead we
20804 should aim at emitting the "best" possible error, which will typically
20805 be the one where we passed the largest number of checks. Generalize the
20806 original "specific_error" approach by making it apply to the entire
20807 matching loop, utilizing that line numbers increase as we pass further
20810 2022-09-30 Jan Beulich <jbeulich@suse.com>
20812 x86/Intel: restrict suffix derivation
20813 While in some cases deriving an AT&T-style suffix from an Intel syntax
20814 memory operand size specifier is necessary, in many cases this is not
20815 only pointless, but has led to the introduction of various workarounds:
20816 Excessive use of IgnoreSize and NoRex64 as well as the ToDword and
20817 ToQword attributes. Suppress suffix derivation when we can clearly tell
20818 that the memory operand's size isn't going to be needed to infer the
20819 possible need for the low byte/word opcode bit or an operand size prefix
20822 As a result ToDword and ToQword can be dropped entirely, plus a fair
20823 number of IgnoreSize and NoRex64 can also be got rid of. Note that
20824 IgnoreSize needs to remain on legacy encoded SIMD insns with GPR
20825 operand, to avoid emitting an operand size prefix in 16-bit mode. (Since
20826 16-bit code using SIMD insns isn't well tested, clone an existing
20827 testcase just enough to cover a few insns which are potentially
20828 problematic but are being touched here.)
20830 Note that while folding the VCVT{,T}S{S,D}2SI templates, VCVT{,T}SH2SI
20831 isn't included there. This is to fulfill the request of not allowing L
20832 and Q suffixes there, despite the inconsistency with VCVT{,T}S{S,D}2SI.
20834 2022-09-30 liuzhensong <liuzhensong@loongson.cn>
20836 LoongArch: Update ELF e_flags handling according to specification.
20837 Update handling of e_flags according to the documentation
20838 update [1] (discussions [2][3]).
20840 Object file bitness is now represented in the EI_CLASS byte.
20841 The e_flags field is now interpreted as follows:
20843 e_flags[2:0]: Base ABI modifier
20846 - 0x2: single-precision hard-float
20847 - 0x3: double-precision hard-float
20849 e_flags[7:6]: ELF object ABI version
20854 [1]: https://github.com/loongson/LoongArch-Documentation/blob/main/docs/LoongArch-ELF-ABI-EN.adoc#e_flags-identifies-abi-type-and-version
20855 [2]: https://github.com/loongson/LoongArch-Documentation/pull/61
20856 [3]: https://github.com/loongson/LoongArch-Documentation/pull/47
20858 2022-09-30 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
20860 gprofng: fix cppcheck warnings
20862 2022-09-29 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
20864 * common/hwcdrv.c: Fix cppcheck warning.
20865 * src/ABS.h: Likewise.
20866 * src/CompCom.cc: Likewise.
20868 2022-09-30 Tom de Vries <tdevries@suse.de>
20870 [gdb/testsuite] Fix gdb.mi/mi-sym-info.exp on openSUSE Tumbleweed
20871 On openSUSE Tumbleweed, I run into:
20873 FAIL: gdb.mi/mi-sym-info.exp: List all functions from debug information only
20876 The problem is in matching this string:
20878 {name="_start",type="void (void)",description="void _start(void);"}
20880 using regexp fun_re, which requires a line field:
20883 "\{line=\"$decimal\",name=${qstr},type=${qstr},description=${qstr}\}"
20886 Fix this by making the line field optional in fun_re.
20888 Tested on x86_64-linux.
20890 2022-09-30 Tsukasa OI <research_trasio@irq.a4lg.com>
20892 RISC-V: Add privileged extensions without instructions/CSRs
20893 Currently, GNU Binutils does not support following privileged extensions:
20899 as they do not provide new CSRs or new instructions ('Smepmp' extends the
20900 privileged architecture CSRs but does not define the CSR itself). However,
20901 adding them might be useful as we no longer have to "filter" ISA strings
20902 just for toolchains (if full ISA string is given by a vendor, we can
20903 straightly use it).
20905 And there's a fact that supports this theory: there's already an
20906 (unprivileged) extension which does not provide CSRs or instructions (but
20907 only an architectural guarantee): 'Zkt' (constant timing guarantee for
20908 certain subset of RISC-V instructions).
20910 This simple commit simply adds three privileged extensions listed above.
20914 * elfxx-riscv.c (riscv_supported_std_s_ext): Add 'Smepmp',
20915 'Svnapot' and 'Svpbmt' extensions.
20917 2022-09-30 Tsukasa OI <research_trasio@irq.a4lg.com>
20919 gdb: Remove unused extra_lines variable
20920 Clang generates a warning if there is a variable that is set but not used
20921 otherwise ("-Wunused-but-set-variable"). On the default configuration, it
20922 causes a build failure (unless "--disable-werror" is specified).
20924 The only extra_lines use in arrange_linetable function is removed on the
20925 commit 558802e4d1c5dcbd0df7d2c6ef62a6deac247a2f
20926 ("gdb: change subfile::line_vector to an std::vector"). So, this variable
20927 should be removed to prevent a build failure.
20929 2022-09-30 Tom de Vries <tdevries@suse.de>
20931 [gdb/testsuite] Add aranges to gdb.dwarf2/dw2-dir-file-name.exp
20932 Since commit 52b920c5d20 ("[gdb/testsuite] Fix gdb.dwarf2/dw2-dir-file-name.exp
20933 for ppc64le"), the test-case fails with target board cc-with-debug-names, due
20934 to missing .debug_aranges info.
20936 Add the missing .debug_aranges info.
20938 Also add a file_id option to Dwarf::assemble, to make it possible to contribute
20939 to an already open file.
20941 Tested on x86_64-linux.
20943 2022-09-30 Tom de Vries <tdevries@suse.de>
20945 [gdb/c++] Print destructor the same for gcc and clang
20946 Consider the test-case contained in this patch.
20948 With g++ (7.5.0) we have for "ptype A":
20958 and with clang++ (13.0.1):
20968 and we observe that the destructor is printed differently.
20970 There's a difference in debug info between the two cases: in the clang case,
20971 there's one artificial parameter, but in the g++ case, there are two, and
20972 these similar cases are handled differently in cp_type_print_method_args.
20974 This is due to this slightly convoluted bit of code:
20976 i = staticp ? 0 : 1;
20983 gdb_printf (stream, "...");
20984 else if (language == language_cplus)
20985 gdb_printf (stream, "void");
20988 The purpose of "i = staticp ? 0 : 1" is to skip the printing of the implicit
20991 In commit 5f4d1085085 ("c++/8218: Destructors w/arguments"), skipping of other
20992 artificial parameters was added, but using a different method: rather than
20993 adjusting the potential loop start, it skips the parameter in the loop.
20995 The observed difference in printing is explained by whether we enter the loop:
20996 - in the clang case, the loop is not entered and we print "void".
20997 - in the gcc case, the loop is entered, and nothing is printed.
20999 Fix this by rewriting the code to:
21000 - always enter the loop
21001 - handle whether arguments need printing in the loop
21002 - keep track of how many arguments are printed, and
21003 use that after the loop to print void etc.
21004 such that we have the same for both gcc and clang:
21010 Note that I consider the discussion of whether we want to print:
21011 - A(void) / ~A(void), or
21013 out-of-scope for this patch.
21015 Tested on x86_64-linux.
21017 2022-09-30 Alan Modra <amodra@gmail.com>
21019 PR29626, Segfault when disassembling ARM code
21021 * arm-dis.c (mapping_symbol_for_insn): Return false on zero
21022 symtab_size. Delete later symtab_size test.
21024 2022-09-30 GDB Administrator <gdbadmin@sourceware.org>
21026 Automatic date update in version.in
21028 2022-09-29 Simon Marchi <simon.marchi@polymtl.ca>
21030 gdb: make target_auxv_parse static and rename
21031 It is only used in auxv.c. Also, given it is not strictly a wrapper
21032 around target_ops::auxv (since 27a48a9223d0 "Add auxv parsing to the
21033 architecture vector."), I think that the name prefixed with target is a
21034 bit misleading. Rename to just parse_auxv.
21036 Change-Id: I41cca055b92c8ede37c258ba6583746a07d8f77e
21038 2022-09-29 Simon Marchi <simon.marchi@polymtl.ca>
21040 gdb: make fprint_target_auxv static
21041 It's only used in auxv.c.
21043 Change-Id: I4992d9aae37b6631a074ab99bbab2f619725b642
21045 2022-09-29 Simon Marchi <simon.marchi@polymtl.ca>
21047 gdb: constify auxv parse functions
21048 Constify the input parameters of the various auxv parse functions, they
21049 don't need to modify the raw auxv data.
21051 Change-Id: I13eacd5ab8e925ec2b5c1f7722cbab39c41516ec
21053 2022-09-29 Simon Marchi <simon.marchi@polymtl.ca>
21055 gdb: constify target_stack::is_pushed
21056 The target_ops parameters here can be made const.
21058 Change-Id: Ibc18b17d6b21d06145251a03e68aca90538117d6
21060 2022-09-29 Keith Seitz <keiths@redhat.com>
21062 Constify target_desc declarations
21063 This patch changes various global target_desc declarations to const, thereby
21064 correcting a prominent source of ODR violations in PowerPC-related target code.
21065 The majority of files/changes are mechanical const-ifications accomplished by
21066 regenerating the C files in features/.
21068 This also required manually updating mips-linux-tdep.h, s390-linux-tdep.h,
21069 nios2-tdep.h, s390-tdep.h, arch/ppc-linux-tdesc.h, arch/ppc-linux-common.c,
21072 Patch tested against the sourceware trybot, and fully regression tested against
21073 our (Red Hat's) internal test infrastructure on Rawhide aarch64, s390x, x86_64,
21076 With this patch, I can finally enable LTO in our GDB package builds. [Tested
21077 with a rawhide scratch build containing this patch.]
21079 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=22395
21080 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=24835
21082 2022-09-29 Keith Seitz <keiths@redhat.com>
21084 cleanup: Add missing feature/ XML files to Makefile
21085 This patch adds some missing .xml files to features/Makefile so that when the
21086 directory's C files are regenerated, all files are appropriately remade.
21088 This has demonstrated that there have been several "misses" in regenerating
21089 files in this directory. Namely, arm-secext.c and sparc{32,64}-solaris.c. For
21090 the former case, there was what essentially amounts to a typo regarding the
21091 create feature function's name. In the later case, this file has missed at least
21092 one important update in July, 2020, when allocate_target_description was
21093 changed to return a unique pointer.
21095 Those corrections are included.
21097 2022-09-29 Nick Clifton <nickc@redhat.com>
21099 Add -B to the help output from gprof, and add suitable documentation.
21101 * gprof.c (usage): Add -B.
21102 * gprof.texi (synopsis): Add -B.
21103 (Output Options): Add entry for -B.
21105 2022-09-29 GDB Administrator <gdbadmin@sourceware.org>
21107 Automatic date update in version.in
21109 2022-09-28 Pedro Alves <pedro@palves.net>
21111 Fix GDB build: ELF support check & -lzstd
21112 GDB fails to build for me, on Ubuntu 20.04. I get:
21116 /usr/bin/ld: linux-tdep.o: in function `linux_corefile_thread(thread_info*, linux_corefile_thread_data*)':
21117 /home/pedro/gdb/binutils-gdb/src/gdb/linux-tdep.c:1831: undefined reference to `gcore_elf_build_thread_register_notes(gdbarch*, thread_info*, gdb_signal, bfd*, std::unique_ptr<char, gdb::xfree_deleter<char> >*, int*)'
21118 /usr/bin/ld: linux-tdep.o: in function `linux_make_corefile_notes(gdbarch*, bfd*, int*)':
21119 /home/pedro/gdb/binutils-gdb/src/gdb/linux-tdep.c:2117: undefined reference to `gcore_elf_make_tdesc_note(bfd*, std::unique_ptr<char, gdb::xfree_deleter<char> >*, int*)'
21120 collect2: error: ld returned 1 exit status
21121 make[2]: *** [Makefile:2149: gdb] Error 1
21122 make[2]: Leaving directory '/home/pedro/gdb/binutils-gdb/build/gdb'
21123 make[1]: *** [Makefile:11847: all-gdb] Error 2
21124 make[1]: Leaving directory '/home/pedro/gdb/binutils-gdb/build'
21125 make: *** [Makefile:1004: all] Error 2
21127 Those undefined functions exist in gdb/gcore-elf.c, which is only
21128 included in the build if GDB's configure thinks that the target you're
21129 configuring for is an ELF target. GDB's configure thinks my system
21130 isn't ELF, which is incorrect.
21132 For the ELF support check, gdb/config.log shows:
21134 configure:17387: checking for ELF support in BFD
21135 configure:17407: gcc -o conftest -I/home/pedro/gdb/binutils-gdb/src/gdb/../include -I../bfd -I/home/pedro/gdb/binutils-gdb/src/gdb/../bfd -g3 -O0 -L../bfd -L../libiberty -lzstd conftest.c -lbfd -liberty -lz -lncursesw -lm -ldl >&5
21136 /usr/bin/ld: ../bfd/libbfd.a(compress.o): in function `decompress_contents':
21137 /home/pedro/gdb/binutils-gdb/src/bfd/compress.c:42: undefined reference to `ZSTD_decompress'
21138 /usr/bin/ld: /home/pedro/gdb/binutils-gdb/src/bfd/compress.c:44: undefined reference to `ZSTD_isError'
21139 /usr/bin/ld: ../bfd/libbfd.a(compress.o): in function `bfd_compress_section_contents':
21140 /home/pedro/gdb/binutils-gdb/src/bfd/compress.c:195: undefined reference to `ZSTD_compress'
21141 /usr/bin/ld: /home/pedro/gdb/binutils-gdb/src/bfd/compress.c:198: undefined reference to `ZSTD_isError'
21142 collect2: error: ld returned 1 exit status
21143 configure:17407: $? = 1
21145 configure:17417: result: no
21147 Note how above, in the gcc command line, "-lzstd" appears before
21148 "-lbfd". That explain the link failure. It should appear after, like
21151 This commit fixes it, by moving ZSTD_LIBS from LDFLAGS to LIBS, next
21152 to -lz, in GDB_AC_CHECK_BFD, and regenerating gdb/configure.
21154 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29630
21155 Change-Id: I1f4128dde634e8ea04c9002904f1005a8b3a6863
21157 2022-09-28 Simon Marchi <simon.marchi@efficios.com>
21159 gdb: remove trailing spaces in README
21160 Change-Id: Ic7f8e415acd1bff6194cf08ed646bff45571f165
21162 2022-09-28 Tom Tromey <tromey@adacore.com>
21164 Treat Character as a discrete type in Ada
21165 A user noticed that gdb would assert when printing a certain array
21166 with array-indexes enabled. This turned out to be caused by the array
21167 having an index type of Character, which is completely valid in Ada.
21168 This patch changes the Ada support to recognize Character as a
21169 discrete type, and adds some tests.
21171 Because this is Ada-specific and was also reviewed internally, I am
21174 2022-09-28 Nick Clifton <nickc@redhat.com>
21176 The help document of size misses an option.
21178 * size.c (usage): Add -f.
21179 * doc/binutils.texi (size): Add -f.
21181 2022-09-28 Clément Chigot <chigot@adacore.com>
21183 ld/testsuite: force warnings when dealing with execstack tests
21184 Binutils can be configured to avoid printing the execstack or RWD
21185 segment warnings. In this case, the first test of PR ld/29072 will fail.
21186 Fix that by always manually forcing the warnings for it.
21190 * testsuite/ld-elf/elf.exp (PR ld/29072): Force execstack and
21191 RWD segment warnings.
21193 2022-09-28 Alan Modra <amodra@gmail.com>
21195 Re: egrep in binutils
21196 Multi-line patterns for grep are not supported on some old versions
21200 * embedspu.sh: Replace multi-line grep with sed.
21202 * testsuite/ld-elfvers/vers.exp: Replace multi-line grep with sed.
21204 2022-09-28 Pedro Alves <pedro@palves.net>
21206 Renenerate {gdb,gdbserver}/configure
21207 Pick up config/lib-ld.m4 changes from:
21209 commit 67d1991b785bdfef1d70cddfa0202b99b43ccce9
21210 Author: Alan Modra <amodra@gmail.com>
21211 AuthorDate: Wed Sep 28 13:37:31 2022 +0930
21212 Commit: Alan Modra <amodra@gmail.com>
21213 CommitDate: Wed Sep 28 13:37:31 2022 +0930
21217 Change-Id: Ifc84d30f1fca015e80bafa80f9a35616b0077220
21219 2022-09-28 Nick Clifton <nickc@redhat.com>
21221 The help document of as misses some many options
21223 * as.c (show_usage): Document the --dump-config,
21224 --gdwarf-cie-version, --hash-size, --multibyte-handling,
21225 and --reduce-memory-overheads options.
21226 * config/tc-i386.c (md_show_usage): Document the -O option.
21227 * doc/as.texi: Document the --dump-config, --emulation,
21228 --hash-size, and --reduce-memory-overheads options.
21230 2022-09-28 Alan Modra <amodra@gmail.com>
21233 Apparently some distros have a nagging egrep that helpfully tells you
21234 egrep is deprecated and to use "grep -E". The nag message causes a ld
21235 testsuite failure. What's more the advice isn't that good. The "-E"
21236 flag may not be available with older versions of grep.
21238 This patch fixes bare invocation of egrep within binutils, replacing
21239 it with the autoconf $EGREP or with grep.
21242 * lib-ld.m4 (AC_LIB_PROG_LD_GNU): Require AC_PROG_EGREP and
21244 (AC_LIB_PROG_LD): Likewise.
21246 * configure: Regenerate.
21247 * embedspu.sh: Replace egrep with grep.
21249 * testsuite/Makefile.am (flagstest_compress_debug_sections.check):
21250 Replace egrep with grep.
21251 * testsuite/Makefile.in: Regenerate.
21252 * testsuite/bnd_ifunc_1.sh: Replace egrep with $EGREP.
21253 * testsuite/bnd_ifunc_2.sh: Likewise.
21254 * testsuite/bnd_plt_1.sh: Likewise.
21255 * testsuite/discard_locals_test.sh: Likewise.
21256 * testsuite/gnu_property_test.sh: Likewise.
21257 * testsuite/no_version_test.sh: Likewise.
21258 * testsuite/pr18689.sh: Likewise.
21259 * testsuite/pr26936.sh: Likewise.
21260 * testsuite/retain.sh: Likewise.
21261 * testsuite/split_i386.sh: Likewise.
21262 * testsuite/split_s390.sh: Likewise.
21263 * testsuite/split_x32.sh: Likewise.
21264 * testsuite/split_x86_64.sh: Likewise.
21265 * testsuite/ver_test_pr16504.sh: Likewise.
21267 * configure: Regenerate.
21269 * testsuite/ld-elfvers/vers.exp (test_ar): Replace egrep with grep.
21271 2022-09-28 Alan Modra <amodra@gmail.com>
21273 regen bfd/configure
21275 2022-09-28 Alan Modra <amodra@gmail.com>
21277 asan: _bfd_stab_section_find_nearest_line segv
21278 The segv was on "info->strs[strsize - 1] = 0;" with strsize zero. OK,
21279 if strsize is zero we don't have any filenames in stabs so no useful
21282 * syms.c (_bfd_stab_section_find_nearest_line): Exit if either
21283 stabsize or strsize is zero.
21285 2022-09-28 Alan Modra <amodra@gmail.com>
21287 asan: segv in _bfd_archive_close_and_cleanup
21288 Uninitialised arelt_data->parent_cache led to this segv.
21290 * pdb.c (pdb_get_elt_at_index): Clear arelt_data.
21292 2022-09-28 GDB Administrator <gdbadmin@sourceware.org>
21294 Automatic date update in version.in
21296 2022-09-27 Fangrui Song <maskray@google.com>
21298 sim: Link ZSTD_LIBS
21299 This fixes linker errors in a `../../configure --enable-targets
21300 --enable-sim; make all-gdb` build.
21302 2022-09-27 Tsukasa OI <research_trasio@irq.a4lg.com>
21304 gold: Suppress "unused" variable warning on Clang
21305 Clang generates a warning if there is a variable that is set but not used
21306 otherwise ("-Wunused-but-set-variable"). On the default configuration, it
21307 causes a build failure (unless "--disable-werror" is specified).
21309 Because the cause of this error is in the Bison-generated code
21310 ($(srcdir)/gold/yyscript.y -> $(builddir)/gold/yyscript.c),
21311 this commit suppresses this warning ("-Wunused-but-set-variable") by placing
21312 DIAGNOSTIC_IGNORE_UNUSED_BUT_SET_VARIABLE macro at the end of user
21313 prologue on yyscript.y.
21315 * yyscript.y: Suppress -Wunused-but-set-variable warning on
21316 the Bison-generated code.
21318 2022-09-27 Fangrui Song <i@maskray.me>
21320 libctf: Add ZSTD_LIBS to LIBS so that ac_cv_libctf_bfd_elf can be true
21322 2022-09-27 Fangrui Song <maskray@google.com>
21324 binutils, gdb: support zstd compressed debug sections
21325 PR29397 PR29563: Add new configure option --with-zstd which defaults to
21326 auto. If pkgconfig/libzstd.pc is found, define HAVE_ZSTD and support
21327 zstd compressed debug sections for most tools.
21329 * bfd: for addr2line, objdump --dwarf, gdb, etc
21330 * gas: support --compress-debug-sections=zstd
21331 * ld: support ELFCOMPRESS_ZSTD input and --compress-debug-sections=zstd
21332 * objcopy: support ELFCOMPRESS_ZSTD input for
21333 --decompress-debug-sections and --compress-debug-sections=zstd
21334 * gdb: support ELFCOMPRESS_ZSTD input. The bfd change references zstd
21335 symbols, so gdb has to link against -lzstd in this patch.
21337 If zstd is not supported, ELFCOMPRESS_ZSTD input triggers an error. We
21338 can avoid HAVE_ZSTD if binutils-gdb imports zstd/ like zlib/, but this
21339 is too heavyweight, so don't do it for now.
21343 ld/ld-new: a.o: section .debug_abbrev is compressed with zstd, but BFD is not built with zstd support
21346 % ld/ld-new a.o --compress-debug-sections=zstd
21347 ld/ld-new: --compress-debug-sections=zstd: ld is not built with zstd support
21349 % binutils/objcopy --compress-debug-sections=zstd a.o b.o
21350 binutils/objcopy: --compress-debug-sections=zstd: binutils is not built with zstd support
21352 % binutils/objcopy b.o --decompress-debug-sections
21353 binutils/objcopy: zstd.o: section .debug_abbrev is compressed with zstd, but BFD is not built with zstd support
21357 2022-09-27 Alan Modra <amodra@gmail.com>
21359 PR29617, ld segfaults when bfd_close fails
21361 * ldmain.c (main): Don't access output_bfd after bfd_close.
21363 2022-09-27 GDB Administrator <gdbadmin@sourceware.org>
21365 Automatic date update in version.in
21367 2022-09-26 Simon Marchi <simon.marchi@polymtl.ca>
21369 gdb/testsuite: update field names in gdb-gdb.py.in
21370 Patches that renamed the type::length and type::target_type fields
21371 didn't update gdb-gdb.py.in accordingly, do that.
21373 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29599
21374 Change-Id: I0f3f37a94d43497789156b0ded4d2f2dd5b89496
21376 2022-09-26 Simon Marchi <simon.marchi@polymtl.ca>
21378 gdb/testsuite: use gdb_test in gdb.gdb/python-helper.exp
21379 If some command in there gives the wrong answer, we currently have to
21380 wait for a timeout for the test to continue. For instance, I currently
21384 $1 = Python Exception <class 'gdb.error'>: Cannot take address of method length.
21386 (outer-gdb) FAIL: gdb.gdb/python-helper.exp: pretty print type (timeout)
21388 We can avoid this and modernize the test at the same time by using the
21389 -prompt option of gdb_test.
21391 gdb_test_no_output currently accepts a -prompt_re option (the variable
21392 name passed to parse_args defines the option name), but I think
21393 it's a typo. It's supposed to be -prompt, like gdb_test. I can't find
21394 anything using -prompt_re using grep. Change it to just "prompt".
21396 Change-Id: Icc0a9a0ef482e62460c708bccdd544c11d711eca
21398 2022-09-26 Simon Marchi <simon.marchi@polymtl.ca>
21400 gdb/testsuite: bump duration for the whole test in do_self_tests
21401 When running gdb.gdb/python-helper.exp, I get some timeouts:
21407 FAIL: gdb.gdb/python-helper.exp: hit breakpoint in outer gdb (timeout)
21409 At this time, GDB is actually processing the stop and reading in some
21410 CUs. selftest_setup does bump the timeout, but it's not for the whole
21413 Since debugging GDB with GDB is (unfortunately) a bit slow, bump the
21414 timeout for the whole duration of the setup and body. On my optimized
21415 build, the command takes just a bit more than the current timeout of 10
21416 seconds. But it's much slower if running the test on an unoptimized
21417 build, so I think it's necessary to bump the timeout for that in any
21420 Change-Id: I4d38285870e76c94f9d0bfdb60648a2e7f2cfa5d
21422 2022-09-26 Clément Chigot <chigot@adacore.com>
21424 binutils/testsuite: handle the different install names of c++filt
21425 c++filt is always named cxxfilt in a build directory, but in a install
21426 directory it would be named either cxxfilt or c++filt (depending on
21427 the host). Handle this last case in testsuite.
21429 binutils/ChangeLog:
21430 * testsuite/config/default.exp (CXXFILE): if cxxfilt not found,
21433 2022-09-26 Clément Chigot <chigot@adacore.com>
21435 binutils/testsuite: skip gentestdlls related tests if missing
21436 When launching the testsuite through runtest outside the build tree,
21437 gentestdlls might not be available, this binary being created by make
21439 Simply untested the related tests instead of crashing.
21441 binutils/ChangeLog:
21443 * testsuite/binutils-all/objdump.exp: Skip dotnet tests if
21444 gentestdlls is not available.
21446 2022-09-26 Tom de Vries <tdevries@suse.de>
21448 [gdb/testsuite] Fix gdb.dwarf2/dw2-unspecified-type-foo.c with -m32
21449 When running test-case gdb.dwarf2/dw2-unspecified-type-foo.c with target board
21450 unix/-m32, I run into:
21452 (gdb) PASS: gdb.dwarf2/dw2-unspecified-type.exp: ptype foo
21453 p ((int (*) ()) foo) ()^M
21457 Add the missing "return 0" in foo, which fixes this.
21459 Tested on x86_64-linux.
21461 2022-09-26 Alan Modra <amodra@gmail.com>
21463 PR29613, use of uninitialized value in objcopy
21465 * elf.c (_bfd_elf_write_secondary_reloc_section): Trim sh_size
21466 back to relocs written. Use better types for vars.
21468 2022-09-26 Alan Modra <amodra@gmail.com>
21470 PR29542, PowerPC gold internal error in get_output_view,
21471 We were attempting to set a BSS style section contents.
21474 * powerpc.cc (Output_data_plt_powerpc::do_write): Don't set .plt,
21475 .iplt or .lplt section contents when position independent.
21477 2022-09-26 Alan Modra <amodra@gmail.com>
21479 stab nearest_line bfd_malloc_and_get_section
21480 bfd_malloc_and_get_section performs some sanity checks on the section
21481 size before allocating memory. This patch avails the stab
21482 nearest_line code of that sanity checking, and tidies up memory
21485 * coffgen.c (_bfd_coff_close_and_cleanup): Call _bfd_stab_cleanup.
21486 * elf.c (_bfd_elf_close_and_cleanup): Likewise.
21487 * syms.c (_bfd_stab_section_find_nearest_line): Set *pinfo earlier.
21488 Use bfd_malloc_and_get_section. Free malloc'd buffers on failure.
21490 (_bfd_stab_cleanup): New function.
21491 * libbfd-in.h (_bfd_stab_cleanup): Declare.
21492 * libbfd.h: Regnerate.
21494 2022-09-26 Alan Modra <amodra@gmail.com>
21496 PKG_CHECK_MODULES for msgpack and jansson
21497 Using AS_IF rather than shell "if" is recommended for conditionals
21498 that contain non-trivial autoconf macros, because autoconf will emit
21499 any AC_REQUIREd autoconf macro expansions outside of the conditional.
21500 This makes them available elsewhere in the configure script.
21503 * configure.ac (msgpack): Use "AS_IF" rather than "if".
21504 * configure: Regenerate.
21506 * configure.ac (jansson): Use "AS_IF" rather than "if".
21507 * configure: Regenerate.
21509 2022-09-26 GDB Administrator <gdbadmin@sourceware.org>
21511 Automatic date update in version.in
21513 2022-09-25 GDB Administrator <gdbadmin@sourceware.org>
21515 Automatic date update in version.in
21517 2022-09-24 Magne Hov <mhov@undo.io>
21519 gdb/source.c: Fix undefined behaviour dereferencing empty string
21520 When a source file's dirname is solely made up of directory separators
21521 we end up trying to dereference the last character of an empty string
21522 with std::string::back, which results in undefined behaviour. A typical
21523 use case where this can happen is when the root directory "/" is used as
21524 a compilation directory.
21526 With libstdc++.so.6.0.28 we get no out-of-bounds checks and the byte
21527 preceding the storage of the empty string is returned. The character
21528 value of this byte depends on heap implementation and usage, but when
21529 this byte happens to hold the value of the directory separator character
21530 we go on to call std::string::pop_back on the empty string which results
21531 in an out_of_range exception which terminates GDB.
21533 Fix this by using path_join. prepare_path_for_appending ensures that the
21534 filename component is relative.
21536 The testsuite has been run before and after the change and no
21537 regressions were found.
21539 2022-09-24 Enze Li <enze.li@hotmail.com>
21541 gdbserver: remove unused for loop
21544 commit cf6c1e710ee162a5adb0ae47acb731f2bfecc956
21545 Date: Mon Jul 11 20:53:48 2022 +0800
21547 gdbserver: remove unused variable
21549 I removed an unused variable in handle_v_run. Pedro then pointed out
21550 that the for loop after it was also unused. After a period of smoke
21551 testing, no exceptions were found.
21553 Tested on x86_64-linux.
21555 2022-09-24 Alan Modra <amodra@gmail.com>
21557 The problem with warning in elf_object_p
21558 elfcode.h can emit three warnings in elf_object_p for various things,
21559 "section extending past end of file", "corrupt string table index",
21560 and "program header with invalid alignment". The problem with doing
21561 this is that the warning can be emitted for multiple possible targets
21562 as each one is tried. I was looking at a fuzzer testcase that had an
21563 object file with 6144 program headers, 5316 of which had invalid
21564 alignment. It would be bad enough to get 5316 messages all the same,
21565 but this object was contained in an archive and resulted in 4975776
21568 Some trimming can be done by not warning if the bfd is already marked
21569 read_only, as is done for the "section extending past end of file"
21570 warning, but that still results in an unacceptable number of
21571 warnings for object files in archives. Besides that, it is just wrong
21572 to warn about a problem detected by a target elf_object_p other than
21573 the one that actually matches. At some point we might have more
21574 target specific warnings.
21576 So what to do? One obvious solution is to remove the warnings.
21577 Another is to poke any warning strings into the target xvec, emitting
21578 them if that xvec is the final one chosen. This also has the benefit
21579 of solving the archive problem. A warning when recursing into
21580 _bfd_check_format for the first element of the archive (to find the
21581 correct target for the archive) will still be on the xvec at the point
21582 that target is chosen for the archive. However, target xvecs are
21583 read-only. Thus the need for per_xvec_warn to logically extend
21584 bfd_target with a writable field. I've made per_xvec_warn one larger
21585 than bfd_target_vector to provide one place for user code that makes
21586 private copies of target xvecs.
21588 * elfcode.h (elf_swap_shdr_in, elf_object_p): Stash potential
21589 warnings in _bfd_per_xvec_warn location.
21590 * format.c (clear_warnmsg): New function.
21591 (bfd_check_format_matches): Call clear_warnmsg before trying
21592 a new xvec. Print warnings for the successful non-archive
21594 * targets.c: Include libiberty.h.
21595 (_bfd_target_vector_entries): Use ARRAY_SIZE.
21596 (per_xvec_warn): New.
21597 (_bfd_per_xvec_warn): New function.
21598 * Makefile.am (LIBBFD_H_FILES): Add targets.c.
21599 * Makefile.in: Regenerate.
21600 * libbfd.h: Regenerate.
21602 2022-09-24 Alan Modra <amodra@gmail.com>
21604 Re: bfd_cleanup for object_p
21605 Bits still missing from commit cb001c0d283d.
21607 * aoutx.h (aout_@var{size}_some_aout_object_p): Correct synopsis.
21608 * i386lynx.c (lynx_core_file_p): Correct return type.
21609 * ptrace-core.c (ptrace_unix_core_file_p): Likewise.
21611 2022-09-24 GDB Administrator <gdbadmin@sourceware.org>
21613 Automatic date update in version.in
21615 2022-09-23 John Baldwin <jhb@FreeBSD.org>
21617 Support AT_USRSTACKBASE and AT_USRSTACKLIM.
21618 FreeBSD's kernel has recently added two new ELF auxiliary vector
21619 entries to describe the location of the user stack for the initial
21620 thread in a process.
21622 This change displays the proper name and description of these entries
21625 2022-09-23 Christoph Müllner <christoph.muellner@vrull.eu>
21627 RISC-V: Add Zawrs ISA extension support
21628 This patch adds support for the Zawrs ISA extension
21629 ("wrs.nto" and "wrs.sto" instructions).
21631 The specification can be found here:
21632 https://github.com/riscv/riscv-zawrs/blob/main/zawrs.adoc
21634 2022-09-23 Simon Marchi <simon.marchi@polymtl.ca>
21636 gdb/testsuite/tui: start GDB with "set filename-display basename"
21637 The test gdb.tui/tui-missing-src.exp fails on my CI machine, and I
21638 concluded that it is caused by the long source directory name:
21640 /home/jenkins/workspace/binutils-gdb_master_linuxbuild/platform/jammy-amd64/target_board/unix/src/binutils-gdb
21642 The long name causes some particular redrawing that doesn't happen for
21643 shorter directories, and causes a Term::command call to return too
21646 This can be reproduced by cloning the binutils-gdb repo in a directory
21647 with a name similar to the one shown above.
21650 /home/simark/workspace/binutils-gdb_master_linuxbuild/platform/jammy-amd64/target_board/unix/src/binutils-gdb/build/gdb
21651 $ make check-read1 TESTS="gdb.tui/tui-missing-src.exp"
21652 FAIL: gdb.tui/tui-missing-src.exp: checking if inside f2 ()
21653 FAIL: gdb.tui/tui-missing-src.exp: f2.c must be displayed in source window
21654 FAIL: gdb.tui/tui-missing-src.exp: check source box is empty after return
21655 FAIL: gdb.tui/tui-missing-src.exp: Back in main
21657 Note that using "make check" instead of "make check-read1" only shows
21658 the last 2 failures for me.
21660 When running gdb.tui/tui-missing-src.exp in a directory with a shorter
21661 name, the terminal looks like this by the time the "checking if inside
21664 Screen Dump (size 80 columns x 24 rows, cursor at column 6, row 23):
21665 0 +-...ld/binutils-gdb-noasan/gdb/testsuite/outputs/gdb.tui/tui-missing-src/f2.c-+
21671 6 | 6 return x+5; |
21679 14 +------------------------------------------------------------------------------+
21680 15 multi-thre Thread 0x7ffff7cc07 In: f2 L5 PC: 0x555555555143
21681 16 at /home/simark/build/binutils-gdb-noasan/gdb/testsuite/outputs/gdb.tui/tui-
21682 17 missing-src/main.c:6
21686 21 at /home/simark/build/binutils-gdb-noasan/gdb/testsuite/outputs/gdb.tui/tui-
21687 22 missing-src/f2.c:5
21689 PASS: gdb.tui/tui-missing-src.exp: checking if inside f2 ()
21691 When running the `Term::command "step"` just before, GDB writes the
21692 "step", which makes the `wait_for` proc go in the "looking for the
21693 prompt" mode, to know when the command's execution is complete. As some
21694 new output appears, lines that must disappear are deleted using the
21695 "Delete Line" operation [1] and some new ones are drawn. The source
21696 window gets redrawn with the contents of the f2.c file. Then, GDB
21697 writes the prompt (at line 23 above), which satisfies `wait_for`, which
21698 then returns. The state of the terminal is therefore correct for the
21699 "check if inside f2" and "f2.c must be displayed in the source window"
21702 In the non-working case, the terminal looks like this by the time the
21703 "check if inside f2" test runs:
21705 Screen Dump (size 80 columns x 24 rows, cursor at column 6, row 17):
21706 0 +------------------------------------------------------------------------------+
21713 7 | [ No Source Available ] |
21720 14 +------------------------------------------------------------------------------+
21721 15 multi-thre Thread 0x7ffff7cc1b In: main L7 PC: 0x555555555128
21722 16 sing-src/main.c:6
21723 17 (gdb) ary breakpoint 1, main ()
21724 18 at /home/simark/workspace/binutils-gdb_master_linuxbuild/platform/jammy-amd6
21725 19 4/target_board/unix/src/binutils-gdb/build/gdb/testsuite/outputs/gdb.tui/tui-mis
21726 20 sing-src/main.c:6
21730 FAIL: gdb.tui/tui-missing-src.exp: checking if inside f2 ()
21732 What happened is: GDB wrote the "step" command, which make the
21733 `wait_for` proc go in its "looking for the prompt" mode. However,
21734 curses decided to redraw whatever scrolled up to line 17 using some
21735 standard character insertion operations:
21737 +++ Cursor Down (1), cursor: (16, 0) -> (17, 0)
21738 +++ Inserting string '('
21739 +++ Inserted char '(', cursor: (17, 0) -> (17, 1)
21740 +++ Inserted string '(', cursor: (17, 0) -> (17, 1)
21741 +++ Inserting string 'g'
21742 +++ Inserted char 'g', cursor: (17, 1) -> (17, 2)
21743 +++ Inserted string 'g', cursor: (17, 1) -> (17, 2)
21744 +++ Inserting string 'd'
21745 +++ Inserted char 'd', cursor: (17, 2) -> (17, 3)
21746 +++ Inserted string 'd', cursor: (17, 2) -> (17, 3)
21747 +++ Inserting string 'b'
21748 +++ Inserted char 'b', cursor: (17, 3) -> (17, 4)
21749 +++ Inserted string 'b', cursor: (17, 3) -> (17, 4)
21750 +++ Inserting string ')'
21751 +++ Inserted char ')', cursor: (17, 4) -> (17, 5)
21752 +++ Inserted string ')', cursor: (17, 4) -> (17, 5)
21753 +++ Inserting string ' '
21754 +++ Inserted char ' ', cursor: (17, 5) -> (17, 6)
21755 +++ Inserted string ' ', cursor: (17, 5) -> (17, 6)
21757 And that causes `wait_for` to think the "step" command is complete.
21758 This is wrong, as the prompt at line 17 isn't the prompt drawn after the
21759 completion of the "step" command. The subsequent tests now run with a
21760 partially updated screen (what is shown above) and obviously fail.
21762 The ideal way to fix this would be for `wait_for` to be smarter, to
21763 avoid it confusing the different prompts drawn.
21765 However, I would also like to reduce the variations in TUI test results
21766 due to the directories (source and build) in which tests are ran. TUI
21767 tests are more prone to differences in test results due to variations in
21768 directory names than other tests, as it makes curses take different
21769 redrawing decisions. So in this patch, I propose to make TUI tests use
21770 "set filename-display basename", which makes GDB omit directory names
21771 when it prints file names. This way, regardless of where you run the
21772 tests, you should get the same results (all other things being equal).
21774 Doing this happens to fix my failures and makes my CI happy (which in
21775 turns makes me happy). To be clear, I understand that this does not fix
21776 the root issue of `proc wait_for` being confused. However, it makes TUI
21777 test runs be more similar for everyone, such that there's less chance of
21778 TUI tests randomly failing for somebody. If some other change triggers
21779 the `wait_for` problem again in the future, hopefully everybody will see
21780 the problem and we can work on getting it fixed more easily than if just
21781 one unlucky person sees the problem.
21783 Note that there are other reasons why TUI tests could vary, like
21784 different curses library versions taking different re-drawing decisions.
21785 However, I think my change is a good step towards more stable test
21788 [1] https://vt100.net/docs/vt510-rm/DL.html
21790 Change-Id: Ib18da83317e7b78a46f77892af0d2e39bd261bf5
21792 2022-09-23 Jiangshuai Li <jiangshuai_li@linux.alibaba.com>
21794 gdb/csky add cskyv2-linux.xml for cskyv2-linux.c
21795 Add cskyv2-linux.xml for re-generating cskyv2-linux.c if needed.
21796 Also update cskyv2-linux.c.
21798 2022-09-23 Alan Modra <amodra@gmail.com>
21800 pdb: _bfd_generic_close_and_cleanup
21801 Every format that might appear inside a generic archive needs to call
21802 _bfd_generic_close_and_cleanup, so that the archive element lookup
21803 htab can be tidied on closing an element. Otherwise you get stale
21804 entries in the htab pointing at freed and perhaps reused memory,
21805 resulting in segfaults when the archive is closed.
21807 Calling _bfd_generic_close_and_cleanup on close means tdata needs to
21808 be set up too, since pdb claims to be of format bfd_archive.
21810 * pdb.c (pdb_close_and_cleanup): Define as
21811 _bfd_generic_close_and_cleanup.
21812 (pdb_archive_p): Set up tdata for bfd_archive format.
21814 2022-09-23 Alan Modra <amodra@gmail.com>
21816 Don't attempt to compress bss sections
21817 It doesn't make sense to try to compress a section without contents
21818 since those sections take no space on disk. Compression can only
21819 increase the disk image size.
21821 * coffgen.c (make_a_section_from_file): Exclude !SEC_HAS_CONTENTS
21822 sections from compression and decompression.
21823 * elf.c (_bfd_elf_make_section_from_shdr): Likewise.
21825 2022-09-23 GDB Administrator <gdbadmin@sourceware.org>
21827 Automatic date update in version.in
21829 2022-09-22 Lancelot SIX <lancelot.six@amd.com>
21831 gdb/testsuite/lib/future.exp: follow dejagnu default_target_compile
21832 GDB's testsuite can override dejagnu's default_target_compile if the
21833 system provided dejagnu installation does not provide support to compile
21834 languages GDB needs.
21836 Recent version of dejagnu (1.6.3, installed on RHEL-9) includes ba60272
21837 "Establish a default C compiler by evaluating [find_gcc] if no other
21838 compiler is given."[1]. This commit removed calls such as
21839 `set_board_info compiler "[find_gcc]"` from the various baseboards
21840 and has default_target_compile call `find_gcc` itself to find a compiler
21841 if none was specified by the board description.
21843 On systems with dejagnu-1.6.3, if GDB's overrides is needed to support
21844 languages still unknown to dejagnu, we end up in the following
21846 - The system board files do not set the C compiler anymore,
21847 - GDB's replacement for default_target_compile assumes that the
21848 compiler should have been set up by the board file.
21850 In this situation, no one sets the C compiler for the board and as a
21851 result many test are not compiled and not executed:
21854 Running .../gdb/testsuite/gdb.base/bt-on-error-and-warning.exp ...
21855 gdb compile failed, default_target_compile: No compiler to compile with
21856 Running .../gdb/testsuite/gdb.base/dprintf-non-stop.exp ...
21857 gdb compile failed, default_target_compile: No compiler to compile with
21858 Running .../gdb/testsuite/gdb.base/structs3.exp ...
21859 gdb compile failed, default_target_compile: No compiler to compile with
21862 We are observing this error with ROCgdb[2], a downstream port of GDB
21863 supporting AMD GPUs. This port needs to use GDB's override of
21864 default_target_compile to compile HIP programs since dejagnu does not
21865 provide support for this language yet.
21867 This patch changes gdb_default_target_compile_1 in a similar way
21868 default_target_compile has been updated so both implementations remain
21869 compatible. Even if this is not strictly required by GDB just yet,
21870 I believe keeping both implementations in sync desirable.
21872 Using board files provided with dejagnu <=1.6.2 is still supported: if
21873 the compiler is set by the board file, gdb_default_target_compile_1 uses
21874 it and does not need `find_gcc`.
21876 Patch tested on x86_64 RHEL-9 and ubuntu-20.04 on top of GDB and ROCgdb.
21878 [1] http://git.savannah.gnu.org/gitweb/?p=dejagnu.git;a=commit;h=ba60272a5ac6f6a7012acca03f596a6ed003f044
21879 [2] https://github.com/ROCm-Developer-Tools/ROCgdb
21881 Change-Id: Ibff52684d9cab8243a7c6748ecbd29f50c37e669
21883 2022-09-22 Christoph Müllner <christoph.muellner@vrull.eu>
21885 RISC-V: Add T-Head MemPair vendor extension
21886 T-Head has a range of vendor-specific instructions.
21887 Therefore it makes sense to group them into smaller chunks
21888 in form of vendor extensions.
21890 This patch adds the XTheadMemPair extension, a collection of T-Head specific
21891 two-GP-register memory operations.
21892 The 'th' prefix and the "XTheadMemPair" extension are documented in a PR
21893 for the RISC-V toolchain conventions ([1]).
21895 [1] https://github.com/riscv-non-isa/riscv-toolchain-conventions/pull/19
21897 Co-developed-by: Lifang Xia <lifang_xia@linux.alibaba.com>
21899 2022-09-22 Christoph Müllner <christoph.muellner@vrull.eu>
21901 RISC-V: Add support for literal instruction arguments
21902 This patch introduces support for arbitrary literal instruction
21903 arguments, that are not encoded in the opcode.
21905 A typical use case for this feature would be an instruction that
21906 applies an implicit shift by a constant value on an immediate
21907 (that is a real operand). With this patch it is possible to make
21908 this shift visible in the dissasembly and support such artificial
21909 parameter as part of the asssembly code.
21911 Co-developed-by: Lifang Xia <lifang_xia@linux.alibaba.com>
21913 2022-09-22 Christoph Müllner <christoph.muellner@vrull.eu>
21915 RISC-V: Add T-Head MemIdx vendor extension
21916 T-Head has a range of vendor-specific instructions.
21917 Therefore it makes sense to group them into smaller chunks
21918 in form of vendor extensions.
21920 This patch adds the XTheadMemIdx extension, a collection of T-Head specific
21921 GPR memory access instructions.
21922 The 'th' prefix and the "XTheadMemIdx" extension are documented in a PR
21923 for the RISC-V toolchain conventions ([1]).
21925 In total XTheadCmo introduces the following 44 instructions
21926 (BU,HU,WU only for loads (zero-extend instead of sign-extend)):
21928 * {L,S}{D,W,WU,H,HU,B,BU}{IA,IB} rd, rs1, imm5, imm2
21929 * {L,S}R{D,W,WU,H,HU,B,BU} rd, rs1, rs2, imm2
21930 * {L,S}UR{D,W,WU,H,HU,B,BU} rd, rs1, rs2, imm2
21932 [1] https://github.com/riscv-non-isa/riscv-toolchain-conventions/pull/19
21934 Co-developed-by: Lifang Xia <lifang_xia@linux.alibaba.com>
21936 2022-09-22 Christoph Müllner <christoph.muellner@vrull.eu>
21938 RISC-V: Add T-Head FMemIdx vendor extension
21939 T-Head has a range of vendor-specific instructions.
21940 Therefore it makes sense to group them into smaller chunks
21941 in form of vendor extensions.
21943 This patch adds the XTheadFMemIdx extension, a collection of
21944 T-Head-specific floating-point memory access instructions.
21945 The 'th' prefix and the "XTheadFMemIdx" extension are documented
21946 in a PR for the RISC-V toolchain conventions ([1]).
21948 [1] https://github.com/riscv-non-isa/riscv-toolchain-conventions/pull/19
21950 Co-developed-by: Lifang Xia <lifang_xia@linux.alibaba.com>
21952 2022-09-22 Christoph Müllner <christoph.muellner@vrull.eu>
21954 RISC-V: Add T-Head MAC vendor extension
21955 T-Head has a range of vendor-specific instructions.
21956 Therefore it makes sense to group them into smaller chunks
21957 in form of vendor extensions.
21959 This patch adds the XTheadMac extension, a collection of
21960 T-Head-specific multiply-accumulate instructions.
21961 The 'th' prefix and the "XTheadMac" extension are documented
21962 in a PR for the RISC-V toolchain conventions ([1]).
21964 [1] https://github.com/riscv-non-isa/riscv-toolchain-conventions/pull/19
21966 Co-developed-by: Lifang Xia <lifang_xia@linux.alibaba.com>
21968 2022-09-22 Christoph Müllner <christoph.muellner@vrull.eu>
21970 RISC-V: Add T-Head CondMov vendor extension
21971 T-Head has a range of vendor-specific instructions.
21972 Therefore it makes sense to group them into smaller chunks
21973 in form of vendor extensions.
21975 This patch adds the XTheadCondMov extension, a collection of
21976 T-Head-specific conditional move instructions.
21977 The 'th' prefix and the "XTheadCondMov" extension are documented
21978 in a PR for the RISC-V toolchain conventions ([1]).
21980 [1] https://github.com/riscv-non-isa/riscv-toolchain-conventions/pull/19
21982 Co-developed-by: Lifang Xia <lifang_xia@linux.alibaba.com>
21984 2022-09-22 Christoph Müllner <christoph.muellner@vrull.eu>
21986 RISC-V: Add T-Head Bitmanip vendor extension
21987 T-Head has a range of vendor-specific instructions.
21988 Therefore it makes sense to group them into smaller chunks
21989 in form of vendor extensions.
21991 This patch adds the XThead{Ba,Bb,Bs} extensions, a collection of
21992 T-Head-specific bitmanipulation instructions.
21993 The 'th' prefix and the "XThead{Ba,Bb,Bs}" extension are documented
21994 in a PR for the RISC-V toolchain conventions ([1]).
21996 [1] https://github.com/riscv-non-isa/riscv-toolchain-conventions/pull/19
21998 Co-developed-by: Lifang Xia <lifang_xia@linux.alibaba.com>
22000 2022-09-22 Christoph Müllner <christoph.muellner@vrull.eu>
22002 RISC-V: Add support for arbitrary immediate encoding formats
22003 This patch introduces support for arbitrary signed or unsigned immediate
22004 encoding formats. The formats have the form "XsN@S" and "XuN@S" with N
22005 being the number of bits and S the LSB position.
22007 For example an immediate field of 5 bytes that encodes a signed value
22008 and is stored in the bits 24-20 of the instruction word can use the
22009 format specifier "Xs5@20".
22011 Co-developed-by: Lifang Xia <lifang_xia@linux.alibaba.com>
22013 2022-09-22 Christoph Müllner <christoph.muellner@vrull.eu>
22015 RISC-V: Add T-Head SYNC vendor extension
22016 T-Head has a range of vendor-specific instructions.
22017 Therefore it makes sense to group them into smaller chunks
22018 in form of vendor extensions.
22020 This patch adds the XTheadSync extension, a collection of
22021 T-Head-specific multi-processor synchronization instructions.
22022 The 'th' prefix and the "XTheadSync" extension are documented in a PR
22023 for the RISC-V toolchain conventions ([1]).
22025 [1] https://github.com/riscv-non-isa/riscv-toolchain-conventions/pull/19
22027 Co-developed-by: Lifang Xia <lifang_xia@linux.alibaba.com>
22029 2022-09-22 Christoph Müllner <christoph.muellner@vrull.eu>
22031 RISC-V: Add T-Head CMO vendor extension
22032 T-Head has a range of vendor-specific instructions.
22033 Therefore it makes sense to group them into smaller chunks
22034 in form of vendor extensions.
22036 This patch adds the XTheadCmo extension, a collection of T-Head specific
22037 cache management operations.
22038 The 'th' prefix and the "XTheadCmo" extension are documented in a PR
22039 for the RISC-V toolchain conventions ([1]).
22041 In total XTheadCmo introduces the following 21 instructions:
22043 * DCACHE.{C,CI,I}ALL
22044 * DCACHE.{C,CI,I}{PA,VA,SW} rs1
22045 * DCACHE.C{PAL1,VAL1} rs1
22046 * ICACHE.I{ALL,ALLS}
22047 * ICACHE.I{PA,VA} rs1
22048 * L2CACHE.{C,CI,I}ALL
22050 Contrary to Zicbom, the XTheadCmo instructions don't have a constant
22051 displacement, therefore we have a different syntax for the arguments.
22052 To clarify this is intended behaviour, there is a set of negative test
22053 for Zicbom-style arguments in x-thead-cmo-fail.s.
22055 [1] https://github.com/riscv-non-isa/riscv-toolchain-conventions/pull/19
22058 - Add missing DECLARE_INSN() list
22061 Co-developed-by: Lifang Xia <lifang_xia@linux.alibaba.com>
22063 2022-09-22 Christoph Müllner <christoph.muellner@vrull.eu>
22065 RISC-V: Add generic support for vendor extensions
22066 This patch introduces changes that allow the integration of vendor ISA
22068 * Define a list of vendor extensions (riscv_supported_vendor_x_ext)
22069 where vendor extensions can be added
22070 * Introduce a section with a table in the documentation where vendor
22071 extensions can be added
22073 To add a vendor extension that consists of instructions only,
22074 the following things need to be done:
22075 * Add the extension to the riscv_supported_vendor_x_ext list
22076 * Add lookup entry in riscv_multi_subset_supports
22077 * Documenting the extension in c-riscv.texti
22078 * Add test cases for all instructions
22079 * Add MATCH*/MASK* constants and DECLARE_INSN() for all instructions
22080 * Add new instruction class to enum riscv_insn_class
22081 * Define the instructions in riscv_opcodes
22082 * Additional changes if necessary (depending on the instructions)
22084 Co-developed-by: Lifang Xia <lifang_xia@linux.alibaba.com>
22086 2022-09-22 Tom de Vries <tdevries@suse.de>
22088 [gdb/symtab] Add all_comp_units/all_type_units views on all_units
22089 Add all_comp_units/all_type_units views on all_units.
22091 Having the views allows us to:
22092 - easily get the number of CUs or TUs in all_units, and
22093 - easily access the nth CU or TU.
22095 This minimizes the use of tu_stats.nr_tus.
22097 Tested on x86_64-linux.
22099 2022-09-22 Tom de Vries <tdevries@suse.de>
22101 [gdb/symtab] Rename all_comp_units to all_units
22102 Mechanically rename all_comp_units to all_units:
22104 $ sed -i 's/all_comp_units/all_units/' gdb/dwarf2/*
22107 Tested on x86_64-linux.
22109 2022-09-22 Yoshinori Sato <ysato@users.sourceforge.jp>
22111 opcodes: SH fix bank register disassemble.
22112 * sh-dis.c (print_insn_sh): Enforce bit7 of LDC Rm,Rn_BANK and STC
22113 Rm_BANK,Rn is always 1.
22115 2022-09-22 Tsukasa OI <research_trasio@irq.a4lg.com>
22117 include: Add macro to ignore -Wunused-but-set-variable
22118 "-Wunused-but-set-variable" warning option can be helpful to track variables
22119 that are written but not read thereafter. But it can be harmful if some of
22120 the code is auto-generated and we have no ways to deal with it.
22122 The particular example is Bison-generated code.
22124 The new DIAGNOSTIC_IGNORE_UNUSED_BUT_SET_VARIABLE macro can be helpful on
22125 such cases. A typical use of this macro is to place this macro before the
22126 end of user prologues on Bison (.y) files.
22130 * diagnostics.h (DIAGNOSTIC_IGNORE_UNUSED_BUT_SET_VARIABLE): New.
22132 2022-09-22 Tsukasa OI <research_trasio@irq.a4lg.com>
22134 include: Add macro to ignore -Wuser-defined-warnings
22135 User-defined warnings (on Clang, "-Wuser-defined-warnings") can be harmful
22136 if we have specified "-Werror" and we have no control to disable the warning
22137 ourself. The particular example is Gnulib.
22139 Gnulib generates a warning if the system version of certain functions
22140 are used (to redirect the developer to use Gnulib version). However,
22141 it can be harmful if we cannot easily replace them (e.g. the target is in
22142 the standard C++ library).
22144 The new DIAGNOSTIC_IGNORE_USER_DEFINED_WARNINGS macro can be helpful on such
22145 cases. A typical use of this macro is to place this macro before including
22146 certain system headers.
22150 * diagnostics.h (DIAGNOSTIC_IGNORE_USER_DEFINED_WARNINGS): New.
22152 2022-09-22 Andrew Burgess <aburgess@redhat.com>
22154 gdb/python: restrict the names accepted by gdb.register_window_type
22155 I noticed that, from Python, I could register a new TUI window that
22156 had whitespace in its name, like this:
22158 gdb.register_window_type('my window', MyWindowType)
22160 however, it is not possible to then use this window in a new TUI
22163 (gdb) tui new-layout foo my window 1 cmd 1
22164 Unknown window "my"
22165 (gdb) tui new-layout foo "my window" 1 cmd 1
22166 Unknown window ""my"
22167 (gdb) tui new-layout foo my\ window 1 cmd 1
22168 Unknown window "my\"
22170 GDB clearly uses the whitespace to split the incoming command line.
22172 I could fix this by trying to add a mechanism by which we can use
22173 whitespace within a window name, but it seems like an easier solution
22174 if we just forbid whitespace within a window name. Not only is this
22175 easier, but I think this is probably the better solution, identifier
22176 names with spaces in would mean we'd need to audit all the places a
22177 window name could be printed and ensure that the use of a space didn't
22178 make the output ambiguous.
22180 So, having decided to disallow whitespace, I then thought about other
22181 special characters. We currently accept anything as a window name,
22182 and I wondered if this was a good idea.
22184 My concerns were about how special characters used in a window name
22185 might cause confusion, for example, we allow '$' in window names,
22186 which is maybe fine now, but what if one day we wanted to allow
22187 variable expansion when creating new layouts? Or what about starting
22188 a window name with '-'? We already support a '-horizontal' option,
22189 what if we want to add more in the future? Or use of the special
22190 character '{' which has special meaning within a new layout?
22192 In the end I figured it might make sense to place some restrictive
22193 rules in place, and then relax the rules later if/when users complain,
22194 we can consider each relaxation as its requested.
22196 So, I propose that window names should match this regular expression:
22198 [a-zA-Z][-_.a-zA-Z0-9]*
22200 There is a chance that there is user code in the wild which will break
22201 with the addition of this change, but hopefully adapting to the new
22202 restrictions shouldn't be too difficult.
22204 2022-09-22 Bruno Larsen <blarsen@redhat.com>
22206 gdb/testsuite: Add test to step through function epilogue
22207 The testsuite implicitly tests GDB's ability to step through epilogues
22208 in multiple tests, without doing it explicitly anywhere. This is
22209 unfortunate, as clang does not emit epilogue information, so using clang
22210 on our testsuite makes many tests fail. This patch adds a central,
22211 explicit test for walking through the epilogue so we can safely remove
22212 this from other tests and have them working with clang.
22214 The test created attempts to step through a simple epilogue, an
22215 epilogue that ends on another epilogue, and epilogues leading to other
22218 2022-09-22 Bruno Larsen <blarsen@redhat.com>
22220 gdb.base/skip.exp: Use finish to exit functions
22221 gdb.base/skip.exp was making use of a fixed number of step commands to
22222 exit some functions. This caused some problems when using clang to test
22223 GDB, as GDB would need fewer steps to reach the desired spots. For
22224 instance, when testing in the section "step after disabling 3", the log
22227 Breakpoint 4, main () at binutils-gdb/gdb/testsuite/gdb.base/skip.c:32
22228 32 x = baz ((bar (), foo ()));
22230 bar () at binutils-gdb/gdb/testsuite/gdb.base/skip1.c:21
22232 (gdb) PASS: gdb.base/skip.exp: step after disabling 3: step 1
22234 foo () at binutils-gdb/gdb/testsuite/gdb.base/skip.c:42
22236 (gdb) PASS: gdb.base/skip.exp: step after disabling 3: step 2
22238 main () at binutils-gdb/gdb/testsuite/gdb.base/skip.c:34
22239 34 test_skip_file_and_function ();
22241 test_skip_file_and_function () at binutils-gdb/gdb/testsuite/gdb.base/skip.c:59
22243 (gdb) FAIL: gdb.base/skip.exp: step after disabling 3: step 3
22245 test_skip () at binutils-gdb/gdb/testsuite/gdb.base/skip.c:48
22247 (gdb) PASS: gdb.base/skip.exp: step after disabling 3: step 4
22249 test_skip_file_and_function () at binutils-gdb/gdb/testsuite/gdb.base/skip.c:60
22250 60 skip1_test_skip_file_and_function ();
22251 (gdb) FAIL: gdb.base/skip.exp: step after disabling 3: step 5
22253 This shows that the feature is working but because the inferior lands in
22254 a different location, it registers as a failure. Seeing as along with
22255 this difference, there are also some differences that depend on gcc
22256 versions (where gdb might stop back at line 32 before entering foo), it
22257 would not be easy to test for this behavior using steps and analzing
22258 where the inferior stops at each point. On the other hand, using
22259 gdb_step_until is not feasible because we'd possibly gloss over stepping
22260 into baz and rendering the whole test useless. Instead, skip.exp now
22261 uses finish to leave functions, synchronizing through compilers and
22262 compiler versions. Some test names were also changed to be a bit more
22263 descriptive. The new log looks like this, independently of compiler used:
22265 Breakpoint 4, main () at binutils-gdb/gdb/testsuite/gdb.base/skip.c:32
22266 32 x = baz ((bar (), foo ()));
22268 bar () at binutils-gdb/gdb/testsuite/gdb.base/skip1.c:21
22270 (gdb) PASS: gdb.base/skip.exp: step after disabling 3: step into bar
22272 Run till exit from #0 bar () at binutils-gdb/gdb/testsuite/gdb.base/skip1.c:21
22273 main () at binutils-gdb/gdb/testsuite/gdb.base/skip.c:32
22274 32 x = baz ((bar (), foo ()));
22275 Value returned is $2 = 1
22276 (gdb) PASS: gdb.base/skip.exp: step after disabling 3: return from bar
22278 foo () at binutils-gdb/gdb/testsuite/gdb.base/skip.c:42
22280 (gdb) PASS: gdb.base/skip.exp: step after disabling 3: step into foo
22282 Run till exit from #0 foo () at binutils-gdb/gdb/testsuite/gdb.base/skip.c:42
22283 main () at binutils-gdb/gdb/testsuite/gdb.base/skip.c:32
22284 32 x = baz ((bar (), foo ()));
22285 Value returned is $3 = 0
22286 (gdb) PASS: gdb.base/skip.exp: step after disabling 3: Return from foo
22288 34 test_skip_file_and_function ();
22289 (gdb) PASS: gdb.base/skip.exp: step after disabling 3: step and skip baz
22291 2022-09-22 Bruno Larsen <blarsen@redhat.com>
22293 fix gdb.base/jit-elf.exp when testing with clang
22294 When using clang as the compiler for the target, gdb.base/jit-elf.exp
22295 was failing because the filename displayed when GDB attached to the
22296 inferior was only showing up as with a relative path, like so:
22298 (gdb) attach 3674146
22299 Attaching to program: /home/blarsen/Documents/gdb-build/gdb/testsuite/outputs/gdb.base/jit-elf/jit-elf-main, process 3674146
22300 Reading symbols from /lib64/libm.so.6...
22301 Reading symbols from .gnu_debugdata for /lib64/libm.so.6...
22302 (No debugging symbols found in .gnu_debugdata for /lib64/libm.so.6)
22303 Reading symbols from /lib64/libc.so.6...
22304 (No debugging symbols found in /lib64/libc.so.6)
22305 Reading symbols from /lib64/ld-linux-x86-64.so.2...
22306 [Thread debugging using libthread_db enabled]
22307 Using host libthread_db library "/lib64/libthread_db.so.1".
22308 0x00000000004013ff in main (argc=3, argv=0x7fffffffd820) at ../../../common/git-repos/binutils-gdb/gdb/testsuite/gdb.base/jit-elf-main.c:118
22309 118| WAIT_FOR_GDB; i = 0; /* gdb break here 1 */
22310 (gdb) FAIL: gdb.base/jit-elf.exp: attach: one_jit_test-2: break here 1: attach
22312 While gcc's output is as follows:
22314 (gdb) attach 3592961
22315 Attaching to program: /home/blarsen/Documents/gdb-build/gdb/testsuite/outputs/gdb.base/jit-elf/jit-elf-main, process 3592961
22316 Reading symbols from /lib64/libm.so.6...
22317 Reading symbols from .gnu_debugdata for /lib64/libm.so.6...
22318 (No debugging symbols found in .gnu_debugdata for /lib64/libm.so.6)
22319 Reading symbols from /lib64/libc.so.6...
22320 (No debugging symbols found in /lib64/libc.so.6)
22321 Reading symbols from /lib64/ld-linux-x86-64.so.2...
22322 [Thread debugging using libthread_db enabled]
22323 Using host libthread_db library "/lib64/libthread_db.so.1".
22324 main (argc=3, argv=0x7fffffffd860) at /home/blarsen/Documents/gdb-build/gdb/testsuite/../../../common/git-repos/binutils-gdb/gdb/testsuite/gdb.base/jit-elf-main.c:118
22325 118| WAIT_FOR_GDB; i = 0; /* gdb break here 1 */
22326 (gdb) PASS: gdb.base/jit-elf.exp: attach: one_jit_test-2: break here 1: attach
22328 This difference only happens when GDB's configure is ran using a
22329 relative path, but seeing as testing the full path is not important for
22330 this specific test, it feels worth fixing anyway. To fix the false
22331 positive, the regexp for checking where gdb has stopped was relaxed a
22332 little to allow the relative path.
22334 2022-09-22 Bruno Larsen <blarsen@redhat.com>
22336 gdb/testsuite: fix gdb.base/msym-bp-shl when running with Clang
22337 When trying to test gdb.base/msym-bp-shl.exp using clang, it would have
22338 many failures because one of the version of the foo function was being
22339 optimized away. Adding __attribute__ ((used)) to it fixed this.
22341 2022-09-22 Bruno Larsen <blarsen@redhat.com>
22343 gdb/testsuite: fix testing gdb.base/skip-inline.exp with clang
22344 When testing gdb.base/skip-inline.exp using clang, we get failures
22345 when trying to step out of functions, since clang requires one fewer
22346 step when compared to gcc. The inferior gets increasingly out of sync
22347 as the test continues because of this difference, which generates those
22350 This commit fixes this by switching those hardcoded steps to
22351 gdb_step_until, to guarantee that the inferior is always synced to what
22352 the test expects. This approach does not work for the parts that use
22353 step 2 or step 3, so when we identify that clang is being used, those
22356 2022-09-22 Bruno Larsen <blarsen@redhat.com>
22358 Change gdb.base/skip-solib.exp deal with lack of epilogue information
22359 When running gdb.base/skip-solib.exp, the backtrace tests could fail with
22360 compilers that associated epilogue instructions with the last statement
22361 line of the function, instead of associating it with the closing brace,
22362 despite the feature being fully functional. As an example, when testing
22363 skipping the function square, the testsuite would show
22365 Breakpoint 1, main () at (...)/binutils-gdb/gdb/testsuite/gdb.base/skip-solib-main.c:5
22366 5 return square(0);
22368 0x00007ffff7cef560 in __libc_start_call_main () from /lib64/libc.so.6
22369 (gdb) PASS: gdb.base/skip-solib.exp: ignoring solib file: step
22371 #0 0x00007ffff7cef560 in __libc_start_call_main () from /lib64/libc.so.6
22372 #1 0x00007ffff7cef60c in __libc_start_main_impl () from /lib64/libc.so.6
22373 #2 0x0000000000401065 in _start ()
22374 (gdb) FAIL: gdb.base/skip-solib.exp: ignoring solib file: bt
22376 Which means that the feature is working, the testsuite is just
22377 mis-identifying it. To avoid this problem, the skipped function calls
22378 have been sent to a line before `return`, so epilogues won't factor in.
22380 2022-09-22 Bruno Larsen <blarsen@redhat.com>
22382 gdb/testsuite: Add a proc to test where compiler links the epilogue
22383 Different compilers link the epilogue of functions to different lines.
22384 As an example, gcc links it to the closing brace of the function,
22385 whereas clang links it to the last statement of the function. This
22386 difference is important for the testsuite, since the where GDB will land
22387 after a step can be wildly different. Where possible, this dependency
22388 should be side-stepped in the testsuite, but it isn't always possible,
22389 so this commit adds a gdb_caching_proc that is able to detect where the
22390 epilogue is linked, so tests can react accordingly.
22392 2022-09-22 Clément Chigot <chigot@adacore.com>
22394 ld/testsuite: allow to force another directory for gcc linker
22395 Add a new variable "ld_testsuite_tmpdir" to enable manual configuration
22396 of the -B flag added to gcc calls. This flag ensure that gcc is invoking
22397 the linker and the assembler we want to test.
22399 When launching the testsuite outside of the build tree, the links made
22400 by the testsuite in tmpdir/ld will point to nothing. Thus, even with the
22401 PATH correctly setup towards the linker directory, gcc might end up
22402 falling back to its default linker. Hence this variable to ensure that
22403 gcc, whatever happens, is using the linker we want.
22407 * testsuite/config/default.exp: Allow to change -B flag with
22408 ld_testsuite_bindir variable.
22410 2022-09-22 Clément Chigot <chigot@adacore.com>
22412 ld/testsuite: skip bootstrap.exp when OFILES are missing
22413 OFILES are normally provided through an environment variable set by
22414 Makefiles. However, when launching the testsuite directly through
22415 runtest outside the build tree, it can be hard to retrieve them.
22416 Thus, they can be missing.
22417 Instead of letting tcl raise an error when trying to access this
22418 OFILES variable, skip bootstrap.exp if it doesn't exist.
22422 * testsuite/ld-bootstrap/bootstrap.exp: Skip if OFILES is
22425 2022-09-22 Tsukasa OI <research_trasio@irq.a4lg.com>
22427 RISC-V: Remove "b" operand type from disassembler
22428 There are a few operand types not used by any RISC-V instructions.
22437 But most of them has a reasoning to keep them:
22439 - Cx : Same as "Ct" except it has a constraint to have rd == rs2
22440 (similar to "Cw"). Although it hasn't used, its role is clear
22441 enough to implement a new instruction with this operand type.
22442 - Vf, Ve : Used by vector AMO instructions (not ratified and real
22443 instructions are not upstreamed yet).
22444 - [, ] : Unused tokenization symbols. Reserving them is not harmful
22445 and a vendor may use this symbol for special purposes.
22447 ... except "b". I could not have found any reference to this operand type
22448 except it works like the "s" operand type. Historically, it seems... it's
22449 just unused from the beginning. Its role is not clear either.
22451 On such cases, we should vacate this room for the new operand type with
22452 much clearer roles.
22456 * riscv-dis.c (print_insn_args): Remove 'b' operand type.
22458 2022-09-22 Tsukasa OI <research_trasio@irq.a4lg.com>
22460 RISC-V: Add macro-only operands to validate_riscv_insn
22461 Although they are not (and should not be) reachable, following macro-only
22462 operands are parsed in the `validate_riscv_insn' function and ignored.
22463 That function also notes that they are macro-only.
22469 Following this convention, this commit adds three remaining macro-only
22470 operands to this function. By doing this, we could instead choose to reject
22471 those operands from appearing in regular instructions later.
22473 - "c" (used by call, tail and jump macros)
22474 - "VM" (used by vmsge.vx and vmsgeu.vx macros)
22479 * config/tc-riscv.c (validate_riscv_insn): Add "c", "VM" and "VT"
22480 macro-only operand types.
22482 2022-09-22 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
22484 gprofng: fix -Wduplicated-cond warning
22486 2022-09-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
22488 * src/collctrl.cc: Fix -Wduplicated-cond warning.
22490 2022-09-22 GDB Administrator <gdbadmin@sourceware.org>
22492 Automatic date update in version.in
22494 2022-09-21 Alan Modra <amodra@gmail.com>
22496 bfd BLD-POTFILES.in dependencies
22497 A file that consists of a list of files doesn't depend on those files
22498 being built. This patch came from trying to avoid a maintainer-mode
22499 make -j bug, where the recipe for targmatch.h was being run twice in
22500 parallel. Typical output shown below.
22502 make[2]: Entering directory '/build/gas/all/bfd'
22508 make[3]: Entering directory '/build/gas/all/bfd/po'
22509 cd .. && make po/SRC-POTFILES.in
22510 cd .. && make po/BLD-POTFILES.in
22511 make[4]: Entering directory '/build/gas/all/bfd'
22512 GEN elf32-aarch64.c
22513 GEN elf64-aarch64.c
22516 GEN elf32-loongarch.c
22517 GEN elf64-loongarch.c
22523 GEN pe-aarch64igen.c
22525 make[4]: Entering directory '/build/gas/all/bfd'
22526 CCLD doc/chew.stamp
22527 mv: cannot stat 'targmatch.new': No such file or directory
22528 make[4]: *** [Makefile:2325: targmatch.h] Error 1
22530 * Makefile.am (po/BLD-POTFILES.in): Don't depend on $(BLD_POTFILES).
22531 (po/SRC-POTFILES.in): Don't depend on $(SRC_POTFILES).
22533 2022-09-21 Simon Marchi <simon.marchi@polymtl.ca>
22535 gdbsupport: move fileio_errno_to_host to fileio.{h,cc} and rename
22536 gdb_bfd.c and remote.c contain identical implementations of a
22537 fileio_error -> errno function. Factor that out to
22538 gdbsupport/fileio.{h,cc}.
22540 Rename it fileio_error_to_host, for symmetry with host_to_fileio_error.
22542 Change-Id: Ib9b8807683de2f809c94a5303e708acc2251a0df
22544 2022-09-21 Simon Marchi <simon.marchi@efficios.com>
22546 gdbsupport: convert FILEIO_* macros to an enum
22547 Converting from free-form macros to an enum gives a bit of type-safety.
22548 This caught places where we would assign host error numbers to what
22549 should contain a target fileio error number, for instance in
22550 target_fileio_pread.
22552 I added the FILEIO_SUCCESS enumerator, because
22553 remote.c:remote_hostio_parse_result initializes the remote_errno output
22554 variable to 0. It seems better to have an explicit enumerator than to
22555 assign a value for which there is no enumerator. I considered
22556 initializing this variable to FILEIO_EUNKNOWN instead, such that if the
22557 remote side replies with an error and omits the errno value, we'll get
22558 an errno that represents an error instead of 0 (which reprensents no
22559 error). But it's not clear what the consequences of that change would
22560 be, so I prefer to err on the side of caution and just keep the existing
22561 behavior (there is no intended change in behavior with this patch).
22563 Note that remote_hostio_parse_resul still reads blindly what the remote
22564 side sends as a target errno into this variable, so we can still end up
22565 with a nonsensical value here. It's not good, but out of the scope of
22568 Convert host_to_fileio_error and fileio_errno_to_host to return / accept
22569 a fileio_error instead of an int, and cascade the change in the whole
22570 chain that uses that.
22572 Change-Id: I454b0e3fcf0732447bc872252fa8e57d138b0e03
22574 2022-09-21 Simon Marchi <simon.marchi@efficios.com>
22576 gdbsupport: move include/gdb/fileio.h contents to fileio.h
22577 I don't see why include/gdb/fileio.h is placed there. It's not
22578 installed by "make install", and it's not included by anything outside
22579 of gdb/gdbserver/gdbsupport.
22581 Move its content back to gdbsupport/fileio.h. I have omitted the bits
22582 inside an `#if 0`, since it's obviously not used, as well as the
22583 "limits" constants, which are also unused.
22585 Change-Id: I6fbc2ea10fbe4cfcf15f9f76006b31b99c20e5a9
22587 2022-09-21 Simon Marchi <simon.marchi@efficios.com>
22589 gdbsupport: change path_join parameter to array_view<const char *>
22590 When a GDB built with -D_GLIBCXX_DEBUG=1 reads a binary with a single
22591 character name, we hit this assertion failure:
22593 $ ./gdb -q --data-directory=data-directory -nx ./x
22594 /usr/include/c++/12.1.0/string_view:239: constexpr const std::basic_string_view<_CharT, _Traits>::value_type& std::basic_string_view<_CharT, _Traits>::operator[](size_type) const [with _CharT = char; _Traits = std::char_traits<char>; const_reference = const char&; size_type = long unsigned int]: Assertion '__pos < this->_M_len' failed.
22598 #3 0x00007ffff6c0f002 in std::__glibcxx_assert_fail (file=<optimized out>, line=<optimized out>, function=<optimized out>, condition=<optimized out>) at /usr/src/debug/gcc/libstdc++-v3/src/c++11/debug.cc:60
22599 #4 0x000055555da8a864 in std::basic_string_view<char, std::char_traits<char> >::operator[] (this=0x7fffffffcc30, __pos=1) at /usr/include/c++/12.1.0/string_view:239
22600 #5 0x00005555609dcb88 in path_join[abi:cxx11](gdb::array_view<std::basic_string_view<char, std::char_traits<char> > const>) (paths=...) at /home/simark/src/binutils-gdb/gdbsupport/pathstuff.cc:203
22601 #6 0x000055555e0443f4 in path_join<char const*, char const*> () at /home/simark/src/binutils-gdb/gdb/../gdbsupport/pathstuff.h:84
22602 #7 0x00005555609dc336 in gdb_realpath_keepfile[abi:cxx11](char const*) (filename=0x6060000a8d40 "/home/simark/build/binutils-gdb-one-target/gdb/./x") at /home/simark/src/binutils-gdb/gdbsupport/pathstuff.cc:122
22603 #8 0x000055555ebd2794 in exec_file_attach (filename=0x7fffffffe0f9 "./x", from_tty=1) at /home/simark/src/binutils-gdb/gdb/exec.c:471
22604 #9 0x000055555f2b3fb0 in catch_command_errors (command=0x55555ebd1ab6 <exec_file_attach(char const*, int)>, arg=0x7fffffffe0f9 "./x", from_tty=1, do_bp_actions=false) at /home/simark/src/binutils-gdb/gdb/main.c:513
22605 #10 0x000055555f2b7e11 in captured_main_1 (context=0x7fffffffdb60) at /home/simark/src/binutils-gdb/gdb/main.c:1209
22606 #11 0x000055555f2b9144 in captured_main (data=0x7fffffffdb60) at /home/simark/src/binutils-gdb/gdb/main.c:1319
22607 #12 0x000055555f2b9226 in gdb_main (args=0x7fffffffdb60) at /home/simark/src/binutils-gdb/gdb/main.c:1344
22608 #13 0x000055555d938c5e in main (argc=5, argv=0x7fffffffdcf8) at /home/simark/src/binutils-gdb/gdb/gdb.c:32
22610 The problem is this line in path_join:
22612 gdb_assert (strlen (path) == 0 || !IS_ABSOLUTE_PATH (path));
22614 ... where `path` is "x". IS_ABSOLUTE_PATH eventually calls
22617 #define HAS_DRIVE_SPEC_1(dos_based, f) \
22618 ((f)[0] && ((f)[1] == ':') && (dos_based))
22620 This macro accesses indices 0 and 1 of the input string. However, `f`
22621 is a string_view of length 1, so it's incorrect to try to access index
22622 1. We know that the string_view's underlying object is a null-terminated
22623 string, so in practice there's no harm. But as far as the string_view
22624 is concerned, index 1 is considered out of bounds.
22626 This patch makes the easy fix, that is to change the path_join parameter
22627 from a vector of to a vector of `const char *`. Another solution would
22628 be to introduce a non-standard gdb::cstring_view class, which would be a
22629 view over a null-terminated string. With that class, it would be
22630 correct to access index 1, it would yield the NUL character. If there
22631 is interest in having this class (it has been mentioned a few times in
22632 the past) I can do it and use it here.
22634 This was found by running tests such as gdb.ada/arrayidx.exp, which
22635 produce 1-char long filenames, so adding a new test is not necessary.
22637 Change-Id: Ia41a16c7243614636b18754fd98a41860756f7af
22639 2022-09-21 Simon Marchi <simon.marchi@polymtl.ca>
22641 gdb: remove TYPE_LENGTH
22642 Remove the macro, replace all uses with calls to type::length.
22644 Change-Id: Ib9bdc954576860b21190886534c99103d6a47afb
22646 2022-09-21 Simon Marchi <simon.marchi@polymtl.ca>
22648 gdb: add type::length / type::set_length
22649 Add the `length` and `set_length` methods on `struct type`, in order to remove
22650 the `TYPE_LENGTH` macro. In this patch, the macro is changed to use the
22651 getter, so all the call sites of the macro that are used as a setter are
22652 changed to use the setter method directly. The next patch will remove the
22655 Change-Id: Id1090244f15c9856969b9be5006aefe8d8897ca4
22657 2022-09-21 Simon Marchi <simon.marchi@polymtl.ca>
22659 gdb: remove TYPE_TARGET_TYPE
22660 Remove the macro, replace all uses by calls to type::target_type.
22662 Change-Id: Ie51d3e1e22f94130176d6abd723255282bb6d1ed
22664 2022-09-21 Simon Marchi <simon.marchi@polymtl.ca>
22666 gdb: add type::target_type / type::set_target_type
22667 Add the `target_type` and `set_target_type` methods on `struct type`, in order
22668 to remove the `TYPE_TARGET_TYPE` macro. In this patch, the macro is changed to
22669 use the getter, so all the call sites of the macro that are used as a setter
22670 are changed to use the setter method directly. The next patch will remove the
22673 Change-Id: I85ce24d847763badd34fdee3e14b8c8c14cb3161
22675 2022-09-21 Tsukasa OI <research_trasio@irq.a4lg.com>
22677 RISC-V: Fix riscv_set_tso declaration
22678 To avoid -Werror=strict-prototypes, this commit changes () to (void).
22679 This is because "()" possibly means a function prototype with indeterminate
22680 arguments on old C standards.
22684 * config/tc-riscv.c (riscv_set_tso): Fix declaration.
22686 2022-09-21 Alan Modra <amodra@gmail.com>
22688 PR29566, objdump -p considers an empty .gnu.version_r invalid
22689 Allow and ignore an empty section.
22692 * elf.c (bfd_section_from_shdr): Don't set elf_dynverdef or
22693 elf_dynverref for empty sections.
22694 (_bfd_elf_slurp_version_tables): Remove now redundant tests.
22696 2022-09-21 Tsukasa OI <research_trasio@irq.a4lg.com>
22698 RISC-V: Set EF_RISCV_TSO also on .option arch
22699 This is a minor fix to commit 96462b012988d35ebb1137a2ad9fd0a96547d79a
22700 ("RISC-V: Implement Ztso extension"). Currently, it sets EF_RISCV_TSO ELF
22701 flag when initial ISA string contains the 'Ztso' extension. However, GAS
22702 has a way to update the ISA string: ".option arch".
22704 When the architecture is updated by ".option arch", EF_RISCV_RVC ELF flag
22705 is set when the 'C' extension is detected. Analogously, this commit sets
22706 the EF_RISCV_TSO when the 'Ztso' extension is detected.
22710 * config/tc-riscv.c (s_riscv_option): Set TSO ELF flag if the
22711 'Ztso' extension is specified via ".option arch" directive.
22713 2022-09-21 Alan Modra <amodra@gmail.com>
22715 PR29573, addr2line doesn't display file/line for local symbols
22716 The DWARF standard is clear that DW_AT_linkage_name is optional.
22717 Compilers may not provide the attribute on functions and variables,
22718 even though the language mangles names. g++ does not for local
22719 variables and functions. Without DW_AT_linkage_name, mangled object
22720 file symbols can't be directly matched against the source-level
22721 DW_AT_name in DWARF info. One possibility is demangling the object
22722 file symbols, but that comes with its own set of problems:
22723 1) A demangler might not be available for the compiler/language.
22724 2) Demangling doesn't give the source function name as stored in
22725 DW_AT_name. Class and template parameters must be stripped at
22728 So this patch takes a simpler approach. A symbol matches DWARF info
22729 if the DWARF address matches the symbol address, and if the symbol
22730 name contains the DWARF name as a sub-string. Very likely the name
22731 matching is entirely superfluous.
22734 * dwarf.c (lookup_symbol_in_function_table): Match a symbol
22735 containing the DWARF source name as a substring.
22736 (lookup_symbol_in_variable_table): Likewise.
22737 (_bfd_dwarf2_find_nearest_line_with_alt): If stash_find_line_fast
22738 returns false, fall back to comp_unit_find_line.
22740 2022-09-21 Alan Modra <amodra@gmail.com>
22742 dwarf2.c: simplify best_fit_len tests
22743 * dwarf2.c (lookup_address_in_function_table): Simplify
22745 (info_hash_lookup_funcinfo): Likewise.
22746 (lookup_symbol_in_function_table): Likewise, also reorder tests
22747 and check "file" is set.
22748 (lookup_symbol_in_variable_table): Reorder tests.
22750 2022-09-21 Alan Modra <amodra@gmail.com>
22752 dwarf2.c: mangle_style
22753 non_mangled incorrectly returned "true" for Ada. Correct that, and
22754 add a few more non-mangled entries. Return a value suitable for
22755 passing to cplus_demangle to control demangling.
22757 * dwarf2.c: Include demangle.h.
22758 (mangle_style): Rename from non_mangled. Return DMGL_* value
22759 to suit lang. Adjust all callers.
22761 2022-09-21 Alan Modra <amodra@gmail.com>
22763 dwarf2.c remove varinfo and funcinfo sec field
22764 The "sec" field in these structures is only set and used in lookup
22765 functions. It always starts off as NULL. So the only possible effect
22766 of the field is to modify the return of the lookup, which was its
22767 purpose back in 2005 when HJ fixed PR990. Since then we solved the
22768 problem of relocatable object files with the fix for PR2338, so this
22769 field is now redundant.
22771 * dwarf.c (struct funcinfo, struct varinfo): Remove "sec" field.
22772 (lookup_symbol_in_function_table): Don't set or test "sec".
22773 (lookup_symbol_in_variable_table): Likewise.
22774 (info_hash_lookup_funcinfo, info_hash_lookup_varinfo): Likewise.
22776 2022-09-21 Tsukasa OI <research_trasio@irq.a4lg.com>
22778 configure: Pass CPPFLAGS_FOR_BUILD to subdirs
22779 Because CPPFLAGS_FOR_BUILD is used in some subdirectories (through
22780 bfd/warning.m4), not AC_SUBSTing the variable causes minor issues.
22782 Fortunately, it didn't cause severe errors but error messages related to
22783 @CPPFLAGS_FOR_BUILD@ (not AC_SUBSTed CPPFLAGS_FOR_BUILD variable passed
22784 to subdirectories through Makefile) remain in config.log.
22786 To avoid invalid invocation of preprocessor for build environment, we
22787 need to set proper CPPFLAGS_FOR_BUILD (may be empty) and pass it to
22788 subdirectories that need it. This is what this commit does.
22792 * configure.ac: Pass CPPFLAGS_FOR_BUILD to subdirectories.
22793 * configure: Regenerate.
22795 2022-09-21 Shihua <shihua@iscas.ac.cn>
22797 RISC-V: Implement Ztso extension
22798 This patch support ZTSO extension. It will turn on the tso flag for elf_flags
22799 once we have enabled Ztso extension. This is intended to implement v0.1 of
22800 the proposed specification which can be found in Chapter 25 of,
22801 https://github.com/riscv/riscv-isa-manual/releases/download/draft-20220723-10eea63/riscv-spec.pdf.
22805 * elfnn-riscv.c (_bfd_riscv_elf_merge_private_bfd_data): Set TSO flag.
22806 * elfxx-riscv.c: Add Ztso's arch.
22808 binutils\ChangeLog:
22810 * readelf.c (get_machine_flags): Set TSO flag.
22814 * config/tc-riscv.c (riscv_set_tso): Ditto.
22815 (riscv_set_arch): Ditto.
22816 * testsuite/gas/riscv/ztso.d: New test.
22820 * elf/riscv.h (EF_RISCV_TSO): Ditto.
22822 2022-09-21 Nelson Chu <nelson@rivosinc.com>
22824 RISC-V: Always generate R_RISCV_CALL_PLT reloc for call in assembler.
22825 Since we have the same behaviors of CALL and CALL_PLT relocs in linker for now,
22826 https://github.com/bminor/binutils-gdb/commit/3b1450b38c644f99aa2e211747b428b9f8d15cca
22828 And the psabi already deprecate the CALL reloc,
22829 https://github.com/riscv-non-isa/riscv-elf-psabi-doc/commit/a0dced85018d7a0ec17023c9389cbd70b1dbc1b0
22831 Therefore, we should always generate R_RISCV_CALL_PLT reloc for call, even if
22832 it has @plt postfix. I believe LLVM (https://reviews.llvm.org/D132530) already
22833 support this, so GNU as should do the same thing.
22836 * config/tc-riscv.c (riscv_ip): Always generate CALL_PLT reloc for
22837 call, even if it has @plt postfix.
22838 * testsuite/gas/riscv/no-relax-reloc.d: Updated CALL to CALL_PLT.
22839 * testsuite/gas/riscv/relax-reloc.d: Likewise.
22841 * testsuite/ld-riscv-elf/variant_cc-r.d: Updated CALL to CALL_PLT.
22843 2022-09-21 GDB Administrator <gdbadmin@sourceware.org>
22845 Automatic date update in version.in
22847 2022-09-21 Alan Modra <amodra@gmail.com>
22849 Re: PowerPC64 pcrel got relocs against local symbols
22850 The last patch wasn't all that shiny. There are rather a lot more
22851 relocations that can hit the assertion in md_apply_fix if the symbol
22852 is local or absolute. Fix them all.
22854 * config/tc-ppc.c (ppc_force_relocation): Add all relocs that
22855 expect a symbol in md_apply_fix. Remove tls pcrel relocs
22856 already covered in general tls match range.
22858 2022-09-21 Alan Modra <amodra@gmail.com>
22860 looping in alpha_vms_slurp_relocs
22861 The direct cause for the looping was failing to test for error return
22862 from _bfd_vms_get_object_record inside a while(1) loop. Fix that.
22863 Also record status of first alpha_vms_slurp_relocs call and return
22864 that for all subsequent calls. (The object format has one set of
22865 relocation records for all sections.) If the first call fails, all
22868 * vms-alpha.c (struct vms_private_data_struct): Make reloc_done
22870 (alpha_vms_slurp_relocs): Set reloc_done to 1 on success, -1 on
22871 failure. Return that status on subsequent calls. Check
22872 _bfd_vms_get_object_record return status.
22873 (alpha_vms_get_reloc_upper_bound): Return status from
22874 alpha_vms_slurp_relocs.
22875 (alpha_vms_write_exec): Exclude sections with contents NULL due
22876 to previous errors from layout, and don't try to write them.
22878 2022-09-20 Dmitry Selyutin <ghostmansd@gmail.com>
22880 ppc/svp64: test setvl ms operand
22882 2022-09-20 Tom Tromey <tom@tromey.com>
22884 Make stdin_event_handler static
22885 I noticed that stdin_event_handler is only used in event-top.c, so
22886 this patch changes it to be 'static'.
22888 2022-09-20 Tom Tromey <tromey@adacore.com>
22890 Constify some target_so_ops instances
22891 This changes some target_so_ops instances to be const. This makes
22892 their use a little more obvious (they can't be mutated) and also
22893 allows for the removal of some initialization code.
22895 Move solib_ops into gdbarch
22896 This changs solib_ops to be an ordinary gdbarch value and updates all
22897 the uses. This removes a longstanding FIXME and makes the code
22898 somewhat cleaner as well.
22900 Remove current_target_so_ops
22901 current_target_so_ops is only set in a single place. It seems better
22902 to simply remove it.
22904 2022-09-20 Andrew Burgess <aburgess@redhat.com>
22906 gdb/testsuite: add a debuginfod-support.exp helper library
22907 We currently have a single test for GDB's debuginfod support, this is
22908 gdb.debuginfod/fetch_src_and_symbols.exp, this script does all the
22909 setup, starts debuginfod, and then does the testing.
22911 This commit tries to split the existing script in two, there is a new
22912 library lib/debuginfod-support.exp, which contains a helper functions
22913 related to running debuginfod tests. All the code in the new library
22914 is basically copied from the existing test case (which is why I
22915 retained the copyright date range on the new library), with some minor
22916 adjustments to try and make the code a little more generic.
22918 One change I made, for example, is the library offers functions to
22919 shut down debuginfod, previously we just relied on expect shutting
22920 down debuginfod when dejagnu completed.
22922 The existing test script is updated to make use of the new library
22923 code, and this test is still passing for me. The only change in the
22924 test results is a single test where I changed the name to remove the
22925 port number from the test name - the port number can change from run
22926 to run, so could make it hard to compare test results.
22928 I have also done a little light house keeping on the original test
22929 script, updating and adding new comments, and making use of
22930 proc_with_prefix in a couple of places.
22932 2022-09-20 liuzhensong <liuzhensong@loongson.cn>
22934 LoongArch: Set macro SUB_SEGMENT_ALIGN to 0.
22936 2022-09-20 Nick Clifton <nickc@redhat.com>
22938 Stop strip from complaining about empty note sections when stripping a binary for a second time.
22939 * objcopy.c (copy_object): Do not issue a warning message when
22940 encountering empty .gnu.build.attribute sections.
22942 New Serbian translations for various binutils sub-directories.
22944 2022-09-20 Zeke Lu <lvzecai@gmail.com>
22946 Bug 29580 - typo in warning message: .note.gnu.build-id data size is too bug
22948 2022-09-20 Xi Ruoyao <xry111@xry111.site>
22950 LoongArch: Fix R_LARCH_IRELATIVE insertion after elf_link_sort_relocs
22951 loongarch_elf_finish_dynamic_symbol is called after elf_link_sort_relocs
22952 if -z combreloc. elf_link_sort_relocs redistributes the contents of
22953 .rela.* sections those would be merged into .rela.dyn, so the slot for
22954 R_LARCH_IRELATIVE may be out of relplt->contents now.
22956 To make things worse, the boundary check
22958 dyn < dyn + relplt->size / sizeof (*dyn)
22960 is obviously wrong ("x + 10 < x"? :), causing the issue undetected
22961 during the linking process and the resulted executable suddenly crashes
22964 The issue was found during an attempt to add static-pie support to the
22967 Fix it by iterating through the inputs of .rela.dyn to find the slot.
22969 2022-09-20 Xi Ruoyao <xry111@xry111.site>
22971 LoongArch: Don't write into GOT for local ifunc
22972 Local ifuncs are always resolved at runtime via R_LARCH_IRELATIVE, so
22973 there is no need to write anything into GOT. And when we write the GOT
22974 we actually trigger a heap-buffer-overflow: If a and b are different
22975 sections, we cannot access something in b with "a->contents + (offset
22976 from a)" because "a->contents" and "b->contents" are heap buffers
22977 allocated separately, not slices of a large buffer.
22979 So stop writing into GOT for local ifunc now.
22981 2022-09-20 GDB Administrator <gdbadmin@sourceware.org>
22983 Automatic date update in version.in
22985 2022-09-19 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
22987 gprofng: build documentation only if BUILD_MAN is true
22989 2022-09-16 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
22992 * gprofng/Makefile.am: Build documentation only if BUILD_MAN is true
22993 * gprofng/Makefile.in: Rebuild.
22994 * gprofng/configure: Rebuild.
22996 2022-09-19 Enze Li <enze.li@hotmail.com>
22998 gdb: add ATTRIBUTE_PRINTF to gdb_bfd_error_handler
22999 I see this error when building with clang,
23002 gdb_bfd.c:1180:43: error: format string is not a string literal [-Werror,-Wformat-nonliteral]
23003 const std::string str = string_vprintf (fmt, ap_copy);
23007 This patch adds missing ATTRIBUTE_PRINTF to fix the error.
23009 Tested on x86_64-linux with gcc 12 and clang 14.
23011 2022-09-19 GDB Administrator <gdbadmin@sourceware.org>
23013 Automatic date update in version.in
23015 2022-09-18 GDB Administrator <gdbadmin@sourceware.org>
23017 Automatic date update in version.in
23019 2022-09-17 Tom de Vries <tdevries@suse.de>
23021 [gdb/symtab] Fix "file index out of range" complaint
23022 With the test-case included in this commit, we run into this FAIL:
23025 During symbol reading: file index out of range^M
23027 (gdb) FAIL: gdb.dwarf2/dw2-no-code-cu.exp: p var with no complaints
23030 This is a regression since commit 6d263fe46e0 ("Avoid bad breakpoints with
23031 --gc-sections"), which contains this change in read_file_scope:
23033 - handle_DW_AT_stmt_list (die, cu, fnd, lowpc);
23034 + if (lowpc != highpc)
23035 + handle_DW_AT_stmt_list (die, cu, fnd, lowpc);
23038 The change intends to avoid a problem with a check in
23039 lnp_state_machine::check_line_address, but also prevents the file and dir
23040 tables from being read, which causes the complaint.
23042 Fix the FAIL by reducing the scope of the "lowpc != highpc" condition to the
23043 call to dwarf_decode_lines in handle_DW_AT_stmt_list.
23045 Tested on x86_64-linux.
23047 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29561
23049 2022-09-17 GDB Administrator <gdbadmin@sourceware.org>
23051 Automatic date update in version.in
23053 2022-09-17 Kevin Buettner <kevinb@redhat.com>
23055 BFD error message suppression test case
23056 This commit adds a GDB test case which tests GDB's BFD error handler
23057 hook for suppressing output of all but the first identical messages.
23059 See the comment at the beginning of bfd-errors.exp for details about
23062 I've tested this test for both 32- and 64-bit ELF files and also
23063 on both little endian and big endian machines. It also works for
23064 both native and remote targets. The only major restriction is that
23065 it only works for ELF targets.
23067 2022-09-17 Kevin Buettner <kevinb@redhat.com>
23069 Suppress printing of superfluous BFD error messages
23070 This commit adds a hook to the BFD error handler for suppressing
23071 identical messages which have been output once already.
23073 It's motivated by this Fedora bug...
23075 https://bugzilla.redhat.com/show_bug.cgi?id=2083315
23077 ...in which over 900,000 BFD error messages are output when attaching
23078 to firefox. From the bug report, the messages all say:
23080 BFD: /usr/lib/debug/usr/lib64/firefox/libxul.so-100.0-2.fc35.x86_64.debug: attempt to load strings from a non-string section (number 38)
23082 Since there's no (additional) context which might assist the user in
23083 determining what's wrong, there's really no point in outputting more
23084 than one message. Of course, if BFD should output some
23085 other/different message, it should be output too, but all future
23086 messages identical to those already output should be suppressed.
23088 For the firefox problem, it turned out that there were only 37
23089 sections, but something was referring to section #38. I haven't
23090 investigated further to find out how this came to be.
23092 Despite this problem, useful debugging might still be done, especially
23093 if the user doesn't care about debugging the problematic library.
23095 If it turns out that knowing the quantity of messages might be useful,
23096 I've implemented the suppression mechanism by keeping a count of each
23097 identical message. A new GDB command, perhaps a 'maintenance'
23098 command, could be added to print out each message along with the
23099 count. I haven't implemented this though because I'm not convinced of
23100 its utility. Also, the BFD message printer has support for BFD-
23101 specific format specifiers. The BFD message strings that GDB stores
23102 in its map are sufficient for distinguishing messages from each
23103 other, but are not identical to those output by BFD's default error
23104 handler. So, that problem would need to be solved too.
23106 2022-09-16 Tom de Vries <tdevries@suse.de>
23108 [gdb/symtab] Handle named DW_TAG_unspecified_type DIE
23109 With the test-case included in the patch, we run into:
23111 (gdb) info types -q std::nullptr_t^M
23112 During symbol reading: unsupported tag: 'DW_TAG_unspecified_type'^M
23114 File /usr/include/c++/7/x86_64-suse-linux/bits/c++config.h:^M
23115 2198: typedef decltype(nullptr) std::nullptr_t;^M
23116 (gdb) FAIL: gdb.dwarf2/nullptr_t.exp: info types -q std::nullptr_t \
23120 Fix the complaint by handling DW_TAG_unspecified_type in new_symbol, and verify
23121 in the test-case using "maint print symbols" that the symbol exists.
23123 Tested on x86_64-linux, with gcc 7.5.0 and clang 13.0.
23125 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=17271
23127 2022-09-16 Tom de Vries <tdevries@suse.de>
23129 [gdb/tdep] Fix PowerPC IEEE 128-bit format arg passing
23130 On a powerpc system with gcc 12 built to default to 128-bit IEEE long double,
23133 (gdb) print find_max_long_double_real(4, ldc1, ldc2, ldc3, ldc4)^M
23135 (gdb) FAIL: gdb.base/varargs.exp: print \
23136 find_max_long_double_real(4, ldc1, ldc2, ldc3, ldc4)
23139 This is due to incorrect handling of the argument in ppc64_sysv_abi_push_param.
23141 Fix this and similar cases, and expand the test-case to test handling of
23142 homogeneous aggregates.
23144 Tested on ppc64le-linux, power 10.
23146 Co-Authored-By: Ulrich Weigand <uweigand@de.ibm.com>
23147 Tested-by: Carl Love <cel@us.ibm.com>
23148 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29543
23150 2022-09-16 Tom de Vries <tdevries@suse.de>
23152 [gdb/testsuite] Fix gdb.dwarf2/dw2-dir-file-name.exp for aarch64
23153 [ Another attempt at fixing the problem described in commit cd919f5533c
23154 ("[gdb/testsuite] Fix gdb.dwarf2/dw2-dir-file-name.exp"). ]
23156 When running the test-case gdb.dwarf2/dw2-dir-file-name.exp with
23157 aarch64-linux, we run into:
23162 Breakpoint 2, compdir_missing__ldir_missing__file_basename () at \
23163 tmp-dw2-dir-file-name.c:999^M
23164 (gdb) FAIL: gdb.dwarf2/dw2-dir-file-name.exp: \
23165 compdir_missing__ldir_missing__file_basename: continue to breakpoint: \
23166 compdir_missing__ldir_missing__file_basename
23169 The breakpoint set at compdir_missing__ldir_missing__file_basename_label,
23170 address 0x400608 starts at a line entry:
23172 CU: tmp-dw2-dir-file-name.c:
23173 File name Line number Starting address View Stmt
23174 tmp-dw2-dir-file-name.c 999 0x400608 x
23175 tmp-dw2-dir-file-name.c 1000 0x40062c x
23176 tmp-dw2-dir-file-name.c - 0x40062c
23178 and therefore the breakpoint is printed without instruction address.
23180 In contrast, for x86_64-linux, we have the breakpoint printed with instruction
23186 Breakpoint 2, 0x004004c1 in compdir_missing__ldir_missing__file_basename () \
23187 at tmp-dw2-dir-file-name.c:999^M
23188 (gdb) PASS: gdb.dwarf2/dw2-dir-file-name.exp: \
23189 compdir_missing__ldir_missing__file_basename: continue to breakpoint: \
23190 compdir_missing__ldir_missing__file_basename
23193 The breakpoint set at compdir_missing__ldir_missing__file_basename_label,
23194 address 0x004004c1 doesn't start at a line entry:
23196 CU: tmp-dw2-dir-file-name.c:
23197 File name Line number Starting address View Stmt
23198 tmp-dw2-dir-file-name.c 999 0x4004bd x
23199 tmp-dw2-dir-file-name.c 1000 0x4004d3 x
23200 tmp-dw2-dir-file-name.c - 0x4004d3
23204 - unifying behaviour between the archs by adding an explicit line number entry
23205 for the address compdir_missing__ldir_missing__file_basename_label, making
23206 the FAIL reproducible on x86_64-linux.
23207 - expecting the breakpoint to be printed without instruction address.
23209 Tested on x86_64-linux and aarch64-linux.
23211 2022-09-16 Tom de Vries <tdevries@suse.de>
23213 [gdb] Handle pending ^C after rl_callback_read_char
23214 In completion tests in various test-cases, we've been running into these
23215 "clearing input line" timeouts:
23217 (gdb) $cmd^GPASS: gdb.gdb/unittest.exp: tab complete "$cmd"
23218 FAIL: gdb.gdb/unittest.exp: tab complete "$cmd" (clearing input line) (timeout)
23220 where $cmd == "maintenance selftest name_that_does_not_exist".
23222 AFAIU, the following scenario happens:
23223 - expect sends "$cmd\t"
23224 - gdb detects the stdin event, and calls rl_callback_read_char until it
23226 - readline interprets the \t as completion, tries to complete, fails to do so,
23227 outputs a bell (^G)
23228 - expect sees the bell, and proceeds to send ^C
23229 - readline is still in the call to rl_callback_read_char, and stores the
23230 signal in _rl_caught_signal
23231 - readline returns from the call to rl_callback_read_char, without having
23232 handled _rl_caught_signal
23233 - gdb goes to wait for the next event
23234 - expect times out waiting for "Quit", the expected reaction for ^C
23236 Fix this by handling pending signals after each call to rl_callback_read_char.
23238 The fix is only available for readline 8.x, if --with-system-readline provides
23239 an older version, then the fix is disabled due to missing function
23242 Tested on x86_64-linux.
23244 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=27813
23246 2022-09-16 Alan Modra <amodra@gmail.com>
23248 PowerPC64 pcrel got relocs against local symbols
23249 Not that anyone would want to indirect via the GOT when an address can
23250 be loaded directly with pla, the following:
23255 leads to "Internal error in md_apply_fix", because the generic parts
23256 of assembler fixup handling convert the fx_pcrel fixup to one without
23257 a symbol. Stop that happening.
23259 * config/tc-ppc.c (ppc_force_relocation): Add PLT_PCREL34 and
23260 assorted GOT_PCREL34 relocs.
23262 2022-09-16 Alan Modra <amodra@gmail.com>
23264 pdb sanity check block_size
23265 * pdb.c (pdb_get_elt_at_index): Only allow block_size to be
23266 512, 1024, 2048, or 4096.
23268 2022-09-16 Nelson Chu <nelson@rivosinc.com>
23270 RISC-V: Make g imply zmmul extension.
23272 * elfxx-riscv.c (riscv_implicit_subset): Moved entry of m after g,
23273 so that g can imply zmmul.
23275 * testsuite/gas/riscv/attribute-01.d: Updated.
23276 * testsuite/gas/riscv/attribute-02.d: Likewise.
23277 * testsuite/gas/riscv/attribute-03.d: Likewise.
23278 * testsuite/gas/riscv/attribute-04.d: Likewise.
23279 * testsuite/gas/riscv/attribute-05.d: Likewise.
23280 * testsuite/gas/riscv/attribute-10.d: Likewise.
23281 * testsuite/gas/riscv/march-imply-g.d: Likewise.
23282 * testsuite/gas/riscv/march-imply-unsupported.d: Likewise.
23284 2022-09-16 GDB Administrator <gdbadmin@sourceware.org>
23286 Automatic date update in version.in
23288 2022-09-15 Tsukasa OI <research_trasio@irq.a4lg.com>
23290 bfd, binutils, gas: Remove/mark unused variables
23291 Clang generates a warning on unused (technically, written but not read
23292 thereafter) variables. By the default configuration (with "-Werror"), it
23293 causes a build failure (unless "--disable-werror" is specified).
23295 This commit adds ATTRIBUTE_UNUSED attribute to some of them, which means
23296 they are *possibly* unused (can be used but no warnings occur when
23297 unused) and removes others.
23301 * elf32-lm32.c (lm32_elf_size_dynamic_sections): Mark unused
23302 rgot_count variable.
23303 * elf32-nds32.c (elf32_nds32_unify_relax_group): Remove unused
23305 * mmo.c (mmo_scan): Mark unused lineno variable.
23307 binutils/ChangeLog:
23309 * windmc.c (write_rc): Remove unused i variable.
23313 * config/tc-riscv.c (riscv_ip): Remove unused argnum variable.
23317 * pe-dll.c (generate_reloc): Remove unused bi and page_count
23320 2022-09-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
23322 gprofng: fix build issues on musl
23324 2022-09-14 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
23327 * configure.ac: Set __MUSL_LIBC.
23328 * configure: Rebuild.
23329 * common/config.h.in: Rebuild.
23330 * src/collector_module.h: Fix compiler errors because mmap64, open64,
23331 pwrite64 are macros and getcontext() is absent on musl.
23332 * libcollector/collector.c: Likewise.
23333 * libcollector/hwprofile.c: Likewise.
23334 * libcollector/iolib.c: Likewise.
23335 * libcollector/libcol_util.c: Likewise.
23336 * libcollector/linetrace.c: Likewise.
23337 * libcollector/memmgr.c: Likewise.
23338 * libcollector/profile.c: Likewise.
23339 * libcollector/unwind.c: Likewise.
23340 * libcollector/dispatcher.c: Likewise.
23341 * src/Experiment.cc: Likewise.
23342 * libcollector/collector.h: Use dlsym() because dlvsym() is not defined
23344 * libcollector/iotrace.c: Remove interposition of versioned functions.
23345 * libcollector/mmaptrace.c: Likewise.
23346 * libcollector/libcol_util.h: Fix -Wint-to-pointer-cast warnings.
23347 * libcollector/jprofile.c: Likewise.
23348 * libcollector/synctrace.c: Include "collector.h".
23349 * src/Print.cc: Use get_basename() because basename() is not defined
23351 * common/hwcdrv.c: Fix -Wformat= warnings.
23353 2022-09-15 GDB Administrator <gdbadmin@sourceware.org>
23355 Automatic date update in version.in
23357 2022-09-14 Rupesh Potharla <Rupesh.Potharla@amd.com>
23359 Binutils: Readelf testcase failing with clang
23360 * testsuite/binutils-all/readelf.exp (readelf_wi_test): Extend
23361 regexps to allow for output genreated by the Clang compiler.
23363 2022-09-14 Alan Modra <amodra@gmail.com>
23365 looping in bfd_mach_o_fat_openr_next_archived_file
23366 mach-o.c doesn't sanity check mach-o-fat archives, making it easy for
23367 fuzzers to create an archive with mach_o_fat_archentry headers that
23368 point to the same offset. bfd_mach_o_fat_openr_next_archived_file
23369 uses the previous element offset to find its header, and thus the next
23370 element. If two offsets are the same, any tool reading the archive
23371 will get stuck. This patch rejects such archives, and any with
23372 overlapping elements.
23374 * mach-o.c (overlap_previous): New function.
23375 (bfd_mach_o_fat_archive_p): Sanity check that elements do not
23376 overlap each other or the file and archive headers.
23378 2022-09-14 Alan Modra <amodra@gmail.com>
23382 2022-09-14 Tsukasa OI <research_trasio@irq.a4lg.com>
23384 bfd: Stop using -Wstack-usage=262144 when built with Clang
23385 Some components of GNU Binutils will pass "-Wstack-usage=262144" when
23386 "GCC >= 5.0" is detected. However, Clang does not support "-Wstack-usage",
23387 despite that related configuration part in bfd/warning.m4 handles the latest
23388 Clang (15.0.0 as of this writing) as "GCC >= 5.0".
23390 The option "-Wstack-usage" was ignored when the first version of Clang is
23391 released but even this "ignoring" behavior is removed before Clang 4.0.0.
23392 So, if we give Clang "-Wstack-usage=262144", it generates a warning, making
23395 This commit checks "__clang__" macro to prevent adding the option if the
23396 compiler is identified as Clang.
23400 * warning.m4: Stop appending "-Wstack-usage=262144" option when
23401 compiled with Clang.
23402 * configure: Regenerate.
23404 binutils/ChangeLog:
23406 * configure: Regenerate.
23410 * configure: Regenerate.
23414 * configure: Regenerate.
23418 * configure: Regenerate.
23422 * configure: Regenerate.
23426 * configure: Regenerate.
23428 2022-09-14 Alan Modra <amodra@gmail.com>
23430 Modify ld-ctf test files to suit ARM
23431 The "@" char starts a comment on ARM.
23433 * testsuite/ld-ctf/diag-ctf-version-0.s: Replace @progbits with
23435 * testsuite/ld-ctf/diag-ctf-version-2-unsupported-feature.s: Likewise.
23436 * testsuite/ld-ctf/diag-ctf-version-f.s: Likewise.
23437 * testsuite/ld-ctf/diag-cttname-invalid.s: Likewise.
23438 * testsuite/ld-ctf/diag-cttname-null.s: Likewise.
23439 * testsuite/ld-ctf/diag-cuname.s: Likewise.
23440 * testsuite/ld-ctf/diag-decompression-failure.s: Likewise.
23441 * testsuite/ld-ctf/diag-parlabel.s: Likewise.
23442 * testsuite/ld-ctf/diag-parname.s: Likewise.
23443 * testsuite/ld-ctf/diag-strlen-invalid.s: Likewise.
23444 * testsuite/ld-ctf/diag-unsupported-flag.s: Likewise.
23445 * testsuite/ld-ctf/diag-wrong-magic-number.s: Likewise.
23447 2022-09-14 Alan Modra <amodra@gmail.com>
23449 asan: som_set_reloc_info heap buffer overflow
23450 Also a bugfix. The first time the section was read, the contents
23451 didn't supply an addend.
23453 * som.c (som_set_reloc_info): Sanity check offset. Do process
23454 contents after reading. Tidy section->contents after freeing.
23456 2022-09-14 Alan Modra <amodra@gmail.com>
23458 ubsan: som_is_space null dereference
23459 On objcopy of fuzzed file.
23461 * som.c (som_write_fixups): Exit loop if space sections all
23464 2022-09-14 Alan Modra <amodra@gmail.com>
23466 msan: vms-alpha use-of-uninitialized-value in dst_retrieve_location
23467 * vms-alpha.c (dst_define_location): Init any unused entries.
23469 2022-09-14 Alan Modra <amodra@gmail.com>
23471 ubsan: arm-dis.c index out of bounds
23472 We are way off in the weeds with this one, and will be printing
23473 <UNPREDICTABLE> for S > 10.
23475 * arm-dis.c (print_insn_cde): Wrap 'T' value.
23477 2022-09-14 Alan Modra <amodra@gmail.com>
23479 PR29540, R_PPC64_NONE in .rela.dyn when linking Linux vdso
23481 * elf64-ppc.c (allocate_dynrelocs): Don't alloc space for relocs
23482 against discarded sections.
23483 (ppc64_elf_size_dynamic_sections): Use standard test for discarded
23485 * elf32-ppc.c (allocate_dynrelocs): Don't alloc space for relocs
23486 against discarded sections.
23487 (ppc_elf_size_dynamic_sections): Use standard test for discarded
23490 2022-09-14 GDB Administrator <gdbadmin@sourceware.org>
23492 Automatic date update in version.in
23494 2022-09-13 Aaron Merey <amerey@redhat.com>
23496 objdump: '-S' should trigger search for separate debuginfo.
23497 Add with_source_code to the command line options that trigger
23498 might_need_separate_debug_info and dump_any_debugging. This helps
23499 'objdump -S' download missing files via debuginfod without the need for
23500 specifying extra command line options like '-L'.
23502 2022-09-13 Bruno Larsen <blarsen@redhat.com>
23504 gdb/testsuite: Update gdb.base/so-impl-ld.exp
23505 gdb.base/so-impl-ld.exp was setup assuming that the compiler would add
23506 epilogue information and that GDB would stop in the } line. This would
23507 make clang tests fail like so:
23510 solib_main (arg=10000) at ../../../common/git-repos/binutils-gdb/gdb/testsuite/gdb.base/solib1.c:7^M
23511 7|__ return arg*arg;|__|___/* HERE */^M
23512 (gdb) PASS: gdb.base/so-impl-ld.exp: step into solib call
23514 main () at ../../../common/git-repos/binutils-gdb/gdb/testsuite/gdb.base/so-impl-ld.c:22^M
23516 (gdb) FAIL: gdb.base/so-impl-ld.exp: step in solib call
23518 0x00007ffff7cef560 in __libc_start_call_main () from /lib64/libc.so.6^M
23519 (gdb) FAIL: gdb.base/so-impl-ld.exp: step out of solib call
23521 This patch changes it so solib_main has 2 lines where GDB can stop
23522 regardless of compiler choices, and updates the exp file to
23523 generically deal with unknown number of steps until leaving that
23526 2022-09-13 Bruno Larsen <blarsen@redhat.com>
23528 gdb/testsuite: introduce gdb_step_until
23529 Currently, GDB's testsuite uses a set amount of step commands to exit
23530 functions. This is a problem if a compiler emits different epilogue
23531 information from gcc, or emits no epilogue information at all. It was
23532 most noticeable if Clang was used to test GDB.
23534 To fix this unreliability, this commit introduces a new proc that will
23535 step the inferior until it is stopped at a line that matches the given
23536 regexp, or until it steps too many times - defined as an optional
23537 argument. If the line is found, it shows up as a single PASS in the
23538 test, and if the line is not found, a single FAIL is emitted.
23540 This patch only introduces this proc, but does not add it to any
23541 existing tests, these will be introduced in the following commit.
23543 2022-09-13 Bruno Larsen <blarsen@redhat.com>
23545 explicitly test for stderr in gdb.base/dprintf.exp
23546 Not all compilers add stderr debug information when compiling a
23547 program. Clang, for instance, prefers to add nothing from standard
23548 libraries and let an external debug package have this information.
23549 Because of this, gdb.base/dprintf.exp was failing when GDB attempted to
23550 use dprintf as a call to fprintf(stderrr, ...), like this:
23552 (gdb) PASS: gdb.base/dprintf.exp: call: fprintf: set dprintf style to call
23556 also to stderr 1234
23557 'stderr' has unknown type; cast it to its declared type
23558 (gdb) FAIL: gdb.base/dprintf.exp: call: fprintf: 1st dprintf (timeout)
23560 To avoid this false positive, we explicitly test to see if
23561 the compiler has added information about stderr at all, and abort
23562 testing dprintf as an fprintf call if it is unavailable.
23564 2022-09-13 Mark Harmstone <mark@harmstone.com>
23566 Add pdb archive format
23567 Resubmitted with changes in
23568 https://sourceware.org/pipermail/binutils/2022-September/122791.html
23571 2022-09-13 Jiangshuai Li <jiangshuai_li@linux.alibaba.com>
23573 gdb/csky rm csky_print_registers_info
23574 The reason for implementing this interface is that we want to print
23575 GPR, PC, EPC, PSR and EPSR when the "info register" command
23578 A prev patch has added PC, EPC, PSR and EPSR to reggroup
23579 general_group, the purpose has been achieved, so this function is
23580 no longer required.
23582 2022-09-13 Jiangshuai Li <jiangshuai_li@linux.alibaba.com>
23584 gdb/csky rm csky_memory_insert/remove_breakpoint
23585 Software breakpoints are inserted or removed by the gdb stub via
23586 remote protocol, these two functions are no longer needed.
23588 2022-09-13 Jiangshuai Li <jiangshuai_li@linux.alibaba.com>
23590 gdb/csky add unwinder for long branch cases
23591 There are two sequences of instructions for long branch:
23592 1. jmpi [pc+4] //insn code: 0xeac00001
23595 2. lrw t1, [pc+8] //insn code: 0xea8d0002
23596 jmp t1 //insn code: 0x7834
23597 nop //insn code: 0x6c03
23600 2022-09-13 Jiangshuai Li <jiangshuai_li@linux.alibaba.com>
23602 gdbserver/csky add csky gdbserver support
23606 gdb/features/cskyv2-linux.c
23607 gdbserver/linux-csky-low.cc
23609 1. In gdb/arch/csky.c file, add function "csky_create_target_description()"
23610 for csky_target::low_arch_setup(). later, it can be used for csky native gdb.
23612 2. In gdb/features/cskyv2-linux.c file, create target_tdesc for csky, include
23613 gprs, pc, hi, lo, float, vector and float control registers.
23615 3. In gdbserver/linux-csky-low.cc file, using PTRACE_GET/SET_RGESET to
23616 get/set registers. The main data structures in asm/ptrace.h are:
23628 unsigned long orig_a0;
23637 unsigned long regs[10];
23640 unsigned long exregs[15];
23644 unsigned long dcsr;
23648 unsigned long vr[96];
23650 unsigned long fesr;
23652 unsigned long reserved;
23655 2022-09-13 GDB Administrator <gdbadmin@sourceware.org>
23657 Automatic date update in version.in
23659 2022-09-12 Tom Tromey <tromey@adacore.com>
23661 Use checked_static_cast in more places
23662 I went through all the uses of dynamic_cast<> in gdb, looking for ones
23663 that could be replaced with checked_static_cast. This patch is the
23664 result. Regression tested on x86-64 Fedora 34.
23666 2022-09-12 Peter Bergner <bergner@linux.ibm.com>
23668 ppc: Document the -mfuture and -Mfuture options and make them usable
23669 The -mfuture and -Mfuture options which are used for adding potential
23670 new ISA instructions were not documented. They also lacked a bitmask
23671 so new instructions could not be enabled by those options. Fixed.
23674 * doc/binutils.texi: Document -Mfuture.
23677 * config/tc-ppc.c: Document -mfuture
23678 * doc/c-ppc.texi: Likewise.
23681 * opcode/ppc.h (PPC_OPCODE_FUTURE): Define.
23684 * ppc-dis.c (ppc_opts) <future>: Use it.
23685 * ppc-opc.c (FUTURE): Define.
23687 2022-09-12 Bruno Larsen <blarsen@redhat.com>
23689 add xfails to gdb.base/complex-parts.exp when testing with clang
23690 clang doesn't add encoding to the name of complex variables, only says
23691 that the type name is complex, making the relevant tests fail.
23692 This patch adds the xfails to the tests that expect the variable name to
23695 2022-09-12 Bruno Larsen <blarsen@redhat.com>
23697 Fix gdb.base/call-ar-st to work with Clang
23698 When running gdb.base/call-ar-st.exp against Clang, we see one FAIL,
23701 print_all_arrays (array_i=<main.integer_array>, array_c=<main.char_array> "ZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZa
23702 ZaZaZaZaZaZaZaZaZaZaZaZa", array_f=<main.float_array>, array_d=<main.double_array>) at ../../../src/gdb/testsuite/gdb.base/call-ar-st.c:274
23703 274 print_int_array(array_i); /* -step1- */
23704 (gdb) FAIL: gdb.base/call-ar-st.exp: step inside print_all_arrays
23706 With GCC we instead see:
23708 print_all_arrays (array_i=<integer_array>, array_c=<char_array> "ZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZaZa", array_f=<float_array>, array_d=<double_array>) at /home/pedro/gdb/build/gdb/testsuite/../../../src/gdb/testsuite/gdb.base/call-ar-st.c:274
23709 274 print_int_array(array_i); /* -step1- */
23710 (gdb) PASS: gdb.base/call-ar-st.exp: step inside print_all_arrays
23712 The difference is that with Clang we get:
23714 array_i=<main.integer_array>, ...
23718 array_i = <integer_array>, ...
23720 These symbols are local static variables, and "main" is the name of
23721 the function they are defined in. GCC instead appends a sequence
23722 number to the linkage name:
23724 $ nm -A call-ar-st.gcc | grep integer_
23725 call-ar-st/call-ar-st:00000000000061a0 b integer_array.3968
23727 $ nm -A call-ar-st.clang | grep integer_
23728 call-ar-st:00000000004061a0 b main.integer_array
23730 This commit changes the testcase to accept both outputs, as they are
23731 functionally identical.
23733 Co-Authored-By: Pedro Alves <pedro@palves.net>
23734 Change-Id: Iaf2ccdb9d5996e0268ed12f595a6e04b368bfcb4
23736 2022-09-12 Bruno Larsen <blarsen@redhat.com>
23738 fix gdb.base/access-mem-running.exp for clang testing
23739 Clang was optimizing global_var away because it was not being used
23740 anywhere. this commit fixes that by adding the attribute used it.
23742 update gdb.base/info-program.exp to not fail with clang
23743 The test specifically mentions that it doesn't care where the program
23744 stops, however it was still testing for a specific location. The clang
23745 compiler emits different line information for epilogue, so GDB reports a
23746 different stopping location, depending on the used compiler. With this
23747 patch the test works even with clang.
23749 2022-09-12 Bruno Larsen <blarsen@redhat.com>
23751 gdb/testsuite: change gdb.base/nodebug.exp to not fail with clang
23752 Clang organizes the variables differently to gcc in the original version
23753 of this code, leading to the following differences when testing
23754 p (int*) &dataglobal + 1
23757 $16 = (int *) 0x404034 <datalocal>
23760 $16 = (int *) 0x404034 <dataglobal8>
23762 However, since the important part of this test doesn't seem to be which
23763 symbol is linked, but rather if GDB is correctly increasing the
23764 address. This test was changed to actually measure address changes,
23765 instead of assuming the ordering and naming of symbols.
23767 Co-Authored-By: Andrew Burgess <aburgess@redhat.com>
23769 2022-09-12 Martin Storsjö <martin@martin.st>
23771 ld: pe: Apply review suggestions on the existing exports/imports arrays
23772 Use a separate explicit max_exports/imports field, instead of
23773 deducing it from the number of allocated elements. Use a named
23774 constant for the incremental growth of the array.
23776 Use bool instead of int for boolean values.
23778 Remove an unnecessary if statement/scope in the def_file_free
23781 Add more verbose comments about parameters, and about insertion
23782 into an array of structs.
23784 Generally use unsigned integers for all array indices and sizes.
23785 The num_exports/imports fields are kept as is as signed integers,
23786 since changing them to unsigned would require a disproportionate
23787 amount of changes ti pe-dll.c to avoid comparisons between signed
23790 Simply use xrealloc instead of a check and xmalloc/xrealloc;
23791 xrealloc can take NULL as the first parameter (and does a similar
23792 check internally). (This wasn't requested in review though,
23793 but noticed while working on the code.)
23795 2022-09-12 Martin Storsjö <martin@martin.st>
23797 ld: pe: Improve performance of object file exclude symbol directives
23798 Store the list of excluded symbols in a sorted list, speeding up
23799 checking for duplicates when inserting new entries.
23801 This is done in the same way as is done for exports and imports
23802 (while the previous implementation was done with a linked list,
23803 based on the implementation for aligncomm).
23805 When linking object files with excluded symbols, there can potentially
23806 be very large numbers of excluded symbols (just like builds with
23807 exports can have a large number of exported symbols).
23809 This improves the link performance somewhat, when linking with large
23810 numbers of excluded symbols.
23812 The later actual use of the excluded symbols within pe-dll.c
23813 handles them via an unordered linked list still, though.
23815 2022-09-12 Tom de Vries <tdevries@suse.de>
23817 [gdb] Fix abort in selftest run_on_main_thread with ^C
23818 When running selftest run_on_main_thread and pressing ^C, we can run into:
23820 Running selftest run_on_main_thread.
23821 terminate called without an active exception
23823 Fatal signal: Aborted
23826 The selftest function looks like this:
23831 std::thread thread;
23836 gdb::block_signals blocker;
23838 thread = std::thread (set_done);
23841 while (!done && gdb_do_one_event () >= 0)
23844 /* Actually the test will just hang, but we want to test
23852 The error message we see is due to the destructor of thread being called while
23853 thread is joinable.
23855 This is supposed to be taken care of by thread.join (), but the ^C prevents
23856 that one from being called, while the destructor is still called.
23858 Fix this by ensuring thread.join () is called (if indeed required) before the
23859 destructor using SCOPE_EXIT.
23861 Tested on x86_64-linux.
23863 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29549
23865 2022-09-12 Tom de Vries <tdevries@suse.de>
23867 [gdb/symtab] Support .gdb_index section with TUs in .debug_info
23868 The .gdb_index variant of commit d878bb39e41 ("[gdb/symtab] Support
23869 .debug_names section with TUs in .debug_info").
23871 Tested on x86_64-linux.
23873 2022-09-12 Tom de Vries <tdevries@suse.de>
23875 [gdb/testsuite] Fix gdb.dwarf2/dw2-dir-file-name.exp for ppc64le
23876 In commit cd919f5533c ("[gdb/testsuite] Fix
23877 gdb.dwarf2/dw2-dir-file-name.exp"), I made gdb.dwarf2/dw2-dir-file-name.exp
23878 independent of prologue analyzers, using this change:
23880 - gdb_breakpoint $func
23881 + gdb_breakpoint *$func
23884 That however caused a regression on ppc64le. For PowerPC, as described in the
23885 ELFv2 ABI, a function can have a global and local entry point.
23887 Setting a breakpoint on *$func effectively creates a breakpoint for the global
23888 entry point, so if the function is entered through the local entry point, the
23889 breakpoint doesn't trigger.
23891 Fix this by reverting commit cd919f5533c, and setting the breakpoint on
23892 ${func}_label instead.
23894 Tested on x86_64-linux and ppc64le-linux.
23896 2022-09-12 Tom de Vries <tdevries@suse.de>
23898 [gdb/testsuite] Fix gdb.dwarf2/dw2-dir-file-name.exp with clang
23899 When running test-case gdb.dwarf2/dw2-dir-file-name.exp with clang, we run
23902 (gdb) break *compdir_missing__ldir_missing__file_basename^M
23903 Breakpoint 2 at 0x400580^M
23907 Breakpoint 2, 0x0000000000400580 in \
23908 compdir_missing.ldir_missing.file_basename ()^M
23909 (gdb) FAIL: gdb.dwarf2/dw2-dir-file-name.exp: \
23910 compdir_missing__ldir_missing__file_basename: continue to breakpoint: \
23911 compdir_missing__ldir_missing__file_basename
23914 The problem is that the test-case uses labels outside functions, which is know
23915 to cause problem with clang, as documented in the comment for proc
23918 Fix this by using get_func_info instead.
23920 Tested on x86_64-linux, with both gcc 7.5.0 and clang 13.0.0.
23922 2022-09-12 Jan Beulich <jbeulich@suse.com>
23924 x86: avoid i386_dis_printf()'s staging area for a fair part of output
23925 While PR binutils/29483 has now been addressed differently, this
23926 originally proposed change still has its merits: Avoiding vsnprintf()
23927 for typically far more than half of the overall output results in a 2-3%
23928 performance gain in my testing (with debug builds of objdump, libbfd,
23931 With that part of output no longer using staging_area[], the array also
23932 doesn't need to be quite as large anymore (the largest presently used
23933 size is 27, from "64-bit address is disabled").
23935 While limiting the scope of "res" it became apparent that
23936 - no caller cares about the function's return value,
23937 - the comment about the return value was wrong,
23938 - a particular positive return value would have been meaningless to the
23940 Therefore convert the function to return "void" at the same time.
23942 2022-09-12 Nelson Chu <nelson@rivosinc.com>
23944 RISC-V: PR28509, the default visibility symbol cannot be referenced by R_RISCV_JAL.
23945 When generating the shared object, the default visibility symbols may bind
23946 externally, which means they will be exported to the dynamic symbol table,
23947 and are preemptible by default. These symbols cannot be referenced by the
23948 non-pic R_RISCV_JAL and R_RISCV_RVC_JUMP. However, consider that linker
23949 may relax the R_RISCV_CALL relocations to R_RISCV_JAL or R_RISCV_RVC_JUMP,
23950 if these relocations are relocated to the plt entries, then we won't report
23951 error for them. Perhaps we also need the similar checks for the
23952 R_RISCV_BRANCH and R_RISCV_RVC_BRANCH relocations.
23954 After applying this patch, and revert the following glibc patch,
23955 riscv: Fix incorrect jal with HIDDEN_JUMPTARGET
23956 https://sourceware.org/git/?p=glibc.git;a=commit;h=68389203832ab39dd0dbaabbc4059e7fff51c29b
23958 I get the expected errors as follows,
23959 ld: relocation R_RISCV_RVC_JUMP against `__sigsetjmp' which may bind externally can not be used when making a shared object; recompile with -fPIC
23960 ld: relocation R_RISCV_JAL against `exit' which may bind externally can not be used when making a shared object; recompile with -fPIC
23962 Besides, we also have similar changes for libgcc,
23963 RISC-V: jal cannot refer to a default visibility symbol for shared object
23964 https://github.com/gcc-mirror/gcc/commit/45116f342057b7facecd3d05c2091ce3a77eda59
23968 * elfnn-riscv.c (riscv_elf_relocate_section): Report errors when
23969 makeing a shard object, and the referenced symbols of R_RISCV_JAL
23970 relocations are default visibility. Besides, we should handle most
23971 of the cases here, so don't need the unresolvable check later for
23972 R_RISCV_JAL and R_RISCV_RVC_JUMP.
23975 * testsuite/ld-riscv-elf/ld-riscv-elf.exp: Updated.
23976 * testsuite/ld-riscv-elf/lib-nopic-01a.s: Removed.
23977 * testsuite/ld-riscv-elf/lib-nopic-01b.d: Likewise.
23978 * testsuite/ld-riscv-elf/lib-nopic-01b.s: Likewise.
23979 * testsuite/ld-riscv-elf/shared-lib-nopic-01.d: New testcase.
23980 * testsuite/ld-riscv-elf/shared-lib-nopic-01.s: Likewise.
23981 * testsuite/ld-riscv-elf/shared-lib-nopic-02.d: Likewise.
23982 * testsuite/ld-riscv-elf/shared-lib-nopic-02.s: Likewise.
23983 * testsuite/ld-riscv-elf/shared-lib-nopic-03.d: Likewise.
23984 * testsuite/ld-riscv-elf/shared-lib-nopic-03.s: Likewise.
23985 * testsuite/ld-riscv-elf/shared-lib-nopic-04.d: Likewise.
23986 * testsuite/ld-riscv-elf/shared-lib-nopic-04.s: Likewise.
23988 2022-09-12 GDB Administrator <gdbadmin@sourceware.org>
23990 Automatic date update in version.in
23992 2022-09-11 Tom de Vries <tdevries@suse.de>
23994 [gdb/symtab] Fix handling of DW_TAG_unspecified_type
23995 Currently, the test-case contained in this patch fails:
23997 (gdb) p (int) foo ()^M
23999 (gdb) FAIL: gdb.dwarf2/dw2-unspecified-type.exp: p (int) foo ()
24001 because DW_TAG_unspecified_type is translated as void.
24003 There's some code in read_unspecified_type that marks the type as stub, but
24004 that's only active for ada:
24006 if (cu->lang () == language_ada)
24007 type->set_is_stub (true);
24011 - marking the type as a stub for all languages, and
24012 - handling the stub return type case in call_function_by_hand_dummy.
24014 Tested on x86_64-linux.
24016 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29558
24018 2022-09-11 GDB Administrator <gdbadmin@sourceware.org>
24020 Automatic date update in version.in
24022 2022-09-10 Alan Modra <amodra@gmail.com>
24024 Re: PR29466, APP/NO_APP with linefile
24025 It looks like I copied the SIZE init across from
24026 binutils/testsuite/config/default.exp without some necessary editing.
24028 * testsuite/config/default.exp (SIZE): Adjust relative path.
24030 2022-09-10 GDB Administrator <gdbadmin@sourceware.org>
24032 Automatic date update in version.in
24034 2022-09-09 Nick Clifton <nickc@redhat.com>
24036 Support debuginfo files with empty group sections.
24038 bfd * elf.c (setup_group): Do not return false if there is no group
24039 information available.
24041 bionutils* objcopy.c (setup_section): Leave group sections intact when
24042 creating separate debuginfo files.
24044 2022-09-09 Tsukasa OI <research_trasio@irq.a4lg.com>
24046 RISC-V: Fix vector CSR requirements
24047 Vector CSRs are also required on smaller vector subsets.
24049 Not only that the most of vector CSRs are general purpose (and must be
24050 accessible for every vector subsets), current minimum vector subset 'Zve32x'
24051 requires fixed point arithmetic, making remaining non-general purpose
24052 (fixed point arithmetic only) CSRs mandatory for such subsets.
24054 So, those CSRs must be accessible from 'Zve32x', not just from 'V'.
24055 This commit fixes this issue which caused CSR accessibility warnings.
24059 * config/tc-riscv.c (riscv_csr_address): Change vector CSR
24060 requirement from 'V' to 'Zve32x'.
24061 * testsuite/gas/riscv/csr-version-1p9p1.l: Change vector CSR
24062 requirement from 'V' to 'Zve32x'.
24063 * testsuite/gas/riscv/csr-version-1p10.l: Likewise.
24064 * testsuite/gas/riscv/csr-version-1p11.l: Likewise.
24065 * testsuite/gas/riscv/csr-version-1p12.l: Likewise.
24067 2022-09-09 GDB Administrator <gdbadmin@sourceware.org>
24069 Automatic date update in version.in
24071 2022-09-08 Carl Love <cel@us.ibm.com>
24073 Fix hardware watchpoint check in test gdb.base/watchpoint-reuse-slot.exp
24074 This test generates 48 failures on Power 9 when testing with HW watchpoints
24075 enabled. Note HW watchpoint support is disabled on Power 9 due to a HW bug.
24076 The skip_hw_watchpoint_tests proc must be used to correctly determine
24077 if the processor supports HW watchpoints.
24079 This patch replaces the [target_info exists gdb,no_hardware_watchpoints]
24080 with the skip_hw_watchpoint_tests check.
24082 This patch was tested on Power 9, Power 10 and X86-64 with no regressions.
24084 2022-09-08 Nick Clifton <nickc@redhat.com>
24086 Gas generated incorrect debug info (top-level DW_TAG_unspecified_type DIE)
24088 * dwarf2dbg.c (out_debug_info): Place DW_TAG_unspecified_type at
24089 the end of the list of children, not at the start of the CU
24091 * testsuite/gas/elf/dwarf-3-func.d: Update expected output.
24092 * testsuite/gas/elf/dwarf-5-func-global.d: Likewise.
24093 * testsuite/gas/elf/dwarf-5-func-local.d: Likewise.
24094 * testsuite/gas/elf/dwarf-5-func.d: Likewise.
24096 2022-09-08 Tsukasa OI <research_trasio@irq.a4lg.com>
24098 gdbsupport: Fix config.status dependency
24099 Commit 171fba11ab27 ("Make GDBserver abort on internal error in development mode")
24100 created a new substitution CONFIG_STATUS_DEPENDENCIES but this is used by
24101 Makefile.in (which is not regenerated by that commit). After regenerating
24102 it, it is found that CONFIG_STATUS_DEPENDENCIES value is not valid, making
24103 gdbsupport fail to build.
24105 Since the CONFIG_STATUS_DEPENDENCIES value is used in the Makefile, macro
24106 substitution must have a Makefile format but commit 171fba11ab27 used shell
24107 format "$srcdir/../bfd/development.sh".
24109 This commit fixes this issue by substituting "$srcdir" (shell format) to
24110 "$(srcdir)" (Makefile format). It preserves the dependency as Pedro
24111 intended and fixes the build problem.
24113 It also regenerates corresponding files with the maintainer mode.
24115 gdbsupport/ChangeLog:
24117 * configure.ac: Fix config.status dependency.
24118 * Makefile.in: Regenerate.
24119 * configure: Regenerate.
24121 2022-09-08 Nick Clifton <nickc@redhat.com>
24123 Maintainer mode: wrong gettext version?
24124 * README-maintainer-mode: Update minimum version of gettext
24127 i686-w64-mingw32-objdump -WL returns incorrect file paths
24129 * dwarf.c (display_debug_lines_decoded): Correctly handle DWARF-5
24130 directory and filename tables.
24132 2022-09-08 GDB Administrator <gdbadmin@sourceware.org>
24134 Automatic date update in version.in
24136 2022-09-07 Tom de Vries <tdevries@suse.de>
24138 [gdb/testsuite] xfail gdb.ada/O2_float_param.exp for aarch64 and gcc 7.5.0
24139 On aarch64-linux, with gcc 7.5.0, we run into:
24142 #0 callee.increment (val=99.0, val@entry=9.18340949e-41, msg=...) at \
24144 21 if Val > 200.0 then^M
24145 (gdb) FAIL: gdb.ada/O2_float_param.exp: scenario=all: frame
24148 The problem is a GCC bug, filed as "PR98148 - [AArch64] Wrong location
24149 expression for function entry values" (
24150 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98148 ).
24152 Xfail the test for aarch64 and gcc 7.
24154 Tested on x86_64-linux and aarch64-linux.
24156 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29418
24158 2022-09-07 Tom de Vries <tdevries@suse.de>
24160 [gdb/testsuite] Fix gdb.ada/access_tagged_param.exp for aarch64
24161 On aarch64-linux, I run into:
24163 Breakpoint 2, pck.inspect (obj=0x430eb0 \
24164 <system.pool_global.global_pool_object>, <objL>=0) at pck.adb:17^M
24165 17 procedure Inspect (Obj: access Top_T'Class) is^M
24166 (gdb) FAIL: gdb.ada/access_tagged_param.exp: continue
24168 while on x86_64-linux, I see:
24170 Breakpoint 2, pck.inspect (obj=0x62b2a0, <objL>=2) at pck.adb:19^M
24172 (gdb) PASS: gdb.ada/access_tagged_param.exp: continue
24174 Note the different line numbers, 17 vs 19.
24176 The difference comes from the gdbarch_skip_prologue implementation.
24178 The amd64_skip_prologue implementation doesn't use gcc line numbers, and falls
24179 back to the architecture-specific prologue analyzer, which correctly skips
24180 past the prologue, to address 0x4022f7:
24182 00000000004022ec <pck__inspect>:
24183 4022ec: 55 push %rbp
24184 4022ed: 48 89 e5 mov %rsp,%rbp
24185 4022f0: 48 89 7d f8 mov %rdi,-0x8(%rbp)
24186 4022f4: 89 75 f4 mov %esi,-0xc(%rbp)
24189 4022f9: 5d pop %rbp
24193 The aarch64_skip_prologue implementation does use gcc line numbers, which are:
24195 File name Line number Starting address View Stmt
24196 pck.adb 17 0x402580 x
24197 pck.adb 17 0x402580 1 x
24198 pck.adb 19 0x40258c x
24199 pck.adb 20 0x402590 x
24201 and which are represented like this internally in gdb:
24203 INDEX LINE ADDRESS IS-STMT PROLOGUE-END
24204 0 17 0x0000000000402580 Y
24205 1 17 0x0000000000402580 Y
24206 2 19 0x000000000040258c Y
24207 3 20 0x0000000000402590 Y
24208 4 END 0x00000000004025a0 Y
24211 The second entry is interpreted as end-of-prologue, so 0x402580 is used, while
24212 the actual end of the prologue is at 0x40258c:
24214 0000000000402580 <pck__inspect>:
24215 402580: d10043ff sub sp, sp, #0x10
24216 402584: f90007e0 str x0, [sp, #8]
24217 402588: b90007e1 str w1, [sp, #4]
24218 40258c: d503201f nop
24219 402590: d503201f nop
24220 402594: 910043ff add sp, sp, #0x10
24221 402598: d65f03c0 ret
24222 40259c: d503201f nop
24225 Note that the architecture-specific prologue analyzer would have gotten this
24228 (gdb) p /x aarch64_analyze_prologue (gdbarch, pc, pc + 128, 0)
24232 Fix the FAIL by making the test-case more robust against problems in prologue
24233 skipping, by setting the breakpoint on line 19 instead.
24235 Likewise in a few similar test-cases.
24237 Tested on x86_64-linux and aarch64-linux.
24239 2022-09-07 Luis Machado <luis.machado@arm.com>
24240 Tom de Vries <tdevries@suse.de>
24242 Fix endianness handling for arm record self tests
24245 - Add 32-bit Arm instruction selftest
24246 - Refactored abstract memory reader into abstract instruction reader
24247 - Adjusted code to use templated type and to use host endianness as
24248 opposed to target endianness.
24250 The arm record tests handle 16-bit and 32-bit thumb instructions, but the
24251 code is laid out in a way that handles the 32-bit thumb instructions as
24254 This is fine, but it is prone to host-endianness issues given how the two
24255 16-bit parts are stored and how they are accessed later on. Arm is
24256 little-endian by default, so running this test with a GDB built with
24257 --enable-targets=all and on a big endian host will run into the following:
24259 Running selftest arm-record.
24260 Process record and replay target doesn't support syscall number -2036195
24261 Process record does not support instruction 0x7f70ee1d at address 0x0.
24262 Self test failed: self-test failed at ../../binutils-gdb/gdb/arm-tdep.c:14482
24264 It turns out the abstract memory reader class is more generic than it needs to
24265 be, and we can simplify the code a bit by assuming we have a simple instruction
24266 reader that only reads up to 4 bytes, which is the length of a 32-bit
24269 Instead of returning a bool, we return instead the instruction that has been
24270 read. This way we avoid having to deal with the endianness conversion, and use
24271 the host endianness instead. The Arm selftests can be executed on non-Arm
24274 While at it, Tom suggested adding a 32-bit Arm instruction selftest to increase
24275 the coverage of the selftests.
24277 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29432
24279 2022-09-07 Tom de Vries <tdevries@suse.de>
24281 [gdb/testsuite] Use prototype to call libc functions
24282 On openSUSE Tumbleweed (using glibc 2.36), I run into:
24284 (gdb) print /d (int) munmap (4198400, 4096)^M
24286 (gdb) FAIL: gdb.base/break-main-file-remove-fail.exp: cmdline: \
24287 get integer valueof "(int) munmap (4198400, 4096)"
24290 The problem is that after starting the executable, the symbol has type
24294 $1 = {<text variable, no debug info>} 0x401030 <munmap@plt>
24298 $2 = {void (void)} 0x7ffff7feb9a0 <__GI_munmap>
24300 which causes the "Invalid cast" error.
24302 Looking at the debug info for glibc for symbol __GI_munmap:
24304 <0><189683>: Abbrev Number: 1 (DW_TAG_compile_unit)
24305 <189691> DW_AT_name : ../sysdeps/unix/syscall-template.S
24306 <189699> DW_AT_producer : GNU AS 2.39.0
24307 <1><1896ae>: Abbrev Number: 2 (DW_TAG_subprogram)
24308 <1896af> DW_AT_name : __GI___munmap
24309 <1896b3> DW_AT_external : 1
24310 <1896b4> DW_AT_low_pc : 0x10cad0
24311 <1896bc> DW_AT_high_pc : 37
24313 that's probably caused by this bit (or similar bits for other munmap aliases).
24315 This is fixed in gas on trunk by commit 5578fbf672e ("GAS: Add a return type
24316 tag to DWARF DIEs generated for function symbols").
24318 Work around this (for say gas 2.39) by explicitly specifying the prototype for
24321 Likewise for getpid in a couple of other test-cases.
24323 Tested on x86_64-linux.
24325 2022-09-07 mengqinggang <mengqinggang@loongson.cn>
24327 LoongArch: fix gas BFD_RELOC_8/16/24 bug
24328 If fixP->fx_subsy is NULL, BFD_RELOC_8/16/24 can't convert to
24329 BFD_RELOC_LARCH_xxx.
24331 gas/config/tc-loongarch.c
24333 2022-09-07 GDB Administrator <gdbadmin@sourceware.org>
24335 Automatic date update in version.in
24337 2022-09-06 Aaron Merey <amerey@redhat.com>
24338 Nick Clifton <nickc@redhat.com>
24340 Add debuginfod support for objdump -S
24341 Currently objdump -S is not able to make use files downloaded from debuginfod.
24342 This is due to bfd_find_nearest_line_discriminator being unable to locate any
24343 separate debuginfo files in the debuginfod cache. Additionally objdump lacked
24344 a call to debuginfod_find_source in order to download missing source files.
24346 Fix this by using bfd_find_nearest_line_with_alt instead of
24347 bfd_find_nearest_line_discriminator. Also add a call to
24348 debuginfod_find_source in order to download missing source files.
24350 2022-09-06 Aaron Merey <amerey@redhat.com>
24352 bfd: Add bfd_find_nearest_line_with_alt
24353 bfd_find_nearest_line_with_alt functions like bfd_find_nearest_line with
24354 the addition of a parameter for specifying the filename of a supplementary
24355 debug file such as one referenced by .gnu_debugaltlink or .debug_sup.
24357 This patch focuses on implementing bfd_find_nearest_line_with_alt
24358 support for ELF/DWARF2 .gnu_debugaltlink. For other targets this
24359 function simply sets the invalid_operation bfd_error.
24361 2022-09-06 Tsukasa OI <research_trasio@irq.a4lg.com>
24363 gdb: add Tsukasa Oi to gdb/MAINTAINERS
24365 2022-09-06 Andrew Burgess <aburgess@redhat.com>
24367 gdb: move a write after approval entry into the correct place
24368 Noticed in passing that an entry in the MAINTAINERS write after
24369 approval list was in the wrong place.
24371 2022-09-06 Tsukasa OI <research_trasio@irq.a4lg.com>
24373 gdb: Add non-enum disassembler options
24374 This is paired with "opcodes: Add non-enum disassembler options".
24376 There is a portable mechanism for disassembler options and used on some
24386 However, it only supports following forms:
24389 - [NAME]=[ENUM_VALUE]
24391 Valid values for [ENUM_VALUE] must be predefined in
24392 disasm_option_arg_t.values. For instance, for -M cpu=[CPU] in ARC
24393 architecture, opcodes/arc-dis.c builds valid CPU model list from
24394 include/elf/arc-cpu.def.
24396 In this commit, it adds following format:
24398 - [NAME]=[ARBITRARY_VALUE] (cannot contain "," though)
24400 This is identified by NULL value of disasm_option_arg_t.values
24401 (normally, this is a non-NULL pointer to a NULL-terminated list).
24405 * gdb/disasm.c (set_disassembler_options): Add support for
24406 non-enum disassembler options.
24407 (show_disassembler_options_sfunc): Likewise.
24409 2022-09-06 Tom de Vries <tdevries@suse.de>
24411 [gdb/symtab] Support .debug_names section with TUs in .debug_info
24412 When running test-case gdb.cp/cpexprs-debug-types.exp on target board
24413 cc-with-debug-names/gdb:debug_flags=-gdwarf-5, we get an executable with
24414 a .debug_names section, but no .debug_types section. For dwarf-5, the TUs
24415 are no longer put in a separate unit, but instead they're put in the
24416 .debug_info section.
24418 When loading the executable, the .debug_names section is silently ignored
24419 because of this check in dwarf2_read_debug_names:
24421 if (map->tu_count != 0)
24423 /* We can only handle a single .debug_types when we have an
24425 if (per_bfd->types.size () != 1)
24428 which triggers because per_bfd->types.size () == 0.
24430 The intention of the check is to make sure we don't have more that one
24431 .debug_types section, as can happen in a object file (see PR12984):
24433 $ grep "\.debug_types" 11.s
24434 .section .debug_types,"G",@progbits,wt.75c042c23a9a07ee,comdat
24435 .section .debug_types,"G",@progbits,wt.c59c413bf50a4607,comdat
24439 - changing the check condition to "per_bfd->types.size () > 1", and
24440 - handling per_bfd->types.size () == 0.
24442 Tested on x86_64-linux.
24444 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29385
24446 2022-09-06 Tom de Vries <tdevries@suse.de>
24448 [gdb/testsuite] Add gdb.dwarf2/debug-names-bad-cu-index.exp
24449 Add test-case gdb.dwarf2/debug-names-bad-cu-index.exp, a regression test for
24450 commit 2fe9a3c41fa ("[gdb/symtab] Fix bad compile unit index complaint").
24452 Tested on x86_64-linux.
24454 2022-09-06 Tom de Vries <tdevries@suse.de>
24456 [gdb/testsuite] Add gdb.dwarf2/debug-names-tu.exp
24457 Add a test-case gdb.dwarf2/debug-names-tu.exp, that uses the dwarf assembler
24458 to specify a .debug_names index with the TU list referring to a TU from the
24459 .debug_types section.
24461 This is intended to produce something similar to:
24463 $ gcc -g -fdebug-types-section ~/hello.c -gdwarf-4
24464 $ gdb-add-index -dwarf-5 a.out
24467 Tested on x86_64-linux.
24469 2022-09-06 Tsukasa OI <research_trasio@irq.a4lg.com>
24471 opcodes: Add non-enum disassembler options
24472 This is paired with "gdb: Add non-enum disassembler options".
24474 There is a portable mechanism for disassembler options and used on some
24484 However, it only supports following forms:
24487 - [NAME]=[ENUM_VALUE]
24489 Valid values for [ENUM_VALUE] must be predefined in
24490 disasm_option_arg_t.values. For instance, for -M cpu=[CPU] in ARC
24491 architecture, opcodes/arc-dis.c builds valid CPU model list from
24492 include/elf/arc-cpu.def.
24494 In this commit, it adds following format:
24496 - [NAME]=[ARBITRARY_VALUE] (cannot contain "," though)
24498 This is identified by NULL value of disasm_option_arg_t.values
24499 (normally, this is a non-NULL pointer to a NULL-terminated list).
24503 * dis-asm.h (disasm_option_arg_t): Update comment of values
24504 to allow non-enum disassembler options.
24508 * riscv-dis.c (print_riscv_disassembler_options): Support
24509 non-enum disassembler options on printing disassembler help.
24510 * arc-dis.c (print_arc_disassembler_options): Likewise.
24511 * mips-dis.c (print_mips_disassembler_options): Likewise.
24513 2022-09-06 GDB Administrator <gdbadmin@sourceware.org>
24515 Automatic date update in version.in
24517 2022-09-05 Tsukasa OI <research_trasio@irq.a4lg.com>
24519 sim/riscv: Complete tidying up with SBREAK
24520 This commit removes SBREAK-related references on the simulator as it's
24521 renamed to EBREAK in 2016 (the RISC-V ISA, version 2.1).
24525 * riscv/sim-main.c (execute_i): Use "ebreak" instead of "sbreak".
24527 2022-09-05 GDB Administrator <gdbadmin@sourceware.org>
24529 Automatic date update in version.in
24531 2022-09-04 Andrew Burgess <aburgess@redhat.com>
24533 sim/erc32: fix gdb with simulator build
24536 commit 7b01c1cc1d111ba0afa51e60fa9842d3b971e2d1
24537 Date: Mon Apr 4 22:38:04 2022 +0100
24539 sim: fixes for libopcodes styled disassembler
24541 changes were made to the simulator source to handle the new libopcodes
24542 disassembler styling API.
24544 Unfortunately, these changes broke building GDB with the erc32 (sparc)
24545 simulator, like this:
24547 ../src/configure --target=sparc-linux
24550 /usr/bin/ld: ../sim/erc32/libsim.a(interf.o): in function `sim_open':
24551 /tmp/build/sim/../../src/sim/erc32/interf.c:247: undefined reference to `fprintf_styled'
24552 collect2: error: ld returned 1 exit status
24554 The problem is that in commit 7b01c1cc1d11 the fprintf_styled function
24555 was added into sis.c. This file is only used when building the 'run'
24556 binary, that is, the standalone simulator, and is not included in the
24559 Now, the obvious fix would be to move fprintf_styled into libsim.a,
24560 however, that turns out to be tricky.
24562 The erc32 simulator currently has two copies of the function run_sim,
24563 one in sis.c, and one in interf.c, both of these copies are global.
24565 Currently, the 'run' binary links fine, though I suspect this might be
24566 pure luck. When I tried moving fprintf_styled into interf.c, I ran
24567 into multiple-definition (of run_sim) errors. I suspect that by
24568 requiring the linker to pull in fprintf_styled from libsim.a I was
24569 changing the order in which symbols were loaded, and the linker was
24570 now seeing both copies of run_sim, while currently we only see one
24573 The ideal solution of course, would be to merge the two similar, but
24574 slightly different copies of run_sim, and just use the one copy. Then
24575 we could safely move fprintf_styled into interf.c too, and all would
24578 But I don't have time right now to start debugging the erc32
24579 simulator, so I wanted a solution that fixes the build without
24580 introducing multiple definition errors.
24582 The easiest solution I think is to just have two copies of
24583 fprintf_styled, one in sis.c, and one in interf.c. Unlike run_sim,
24584 these two copies are both static, so we will not run into multiple
24585 definition issues with this function. The functions themselves are
24586 not very big, so it's not a huge amount of duplicate code.
24588 I am very aware that this is not an ideal solution, and I would
24589 welcome anyone who wants to take on fixing the run_sim problem
24590 properly, and then cleanup the fprintf_styled duplication.
24592 2022-09-04 GDB Administrator <gdbadmin@sourceware.org>
24594 Automatic date update in version.in
24596 2022-09-03 GDB Administrator <gdbadmin@sourceware.org>
24598 Automatic date update in version.in
24600 2022-09-02 Max Filippov <jcmvbkbc@gmail.com>
24602 xtensa: bfd: fix TLS relocations generated for PIE
24603 When generating TLS dynamic relocations the existing xtensa BFD code
24604 treats linking to a PIE exactly as linking to a shared object, resulting
24605 in generation of wrong relocations for TLS entries. Fix that and add
24609 * elf32-xtensa.c (elf_xtensa_check_relocs): Use bfd_link_dll
24610 instead of bfd_link_pic. Add elf_xtensa_dynamic_symbol_p test
24611 when generating GOT entries.
24612 (elf_xtensa_relocate_section): Use bfd_link_dll instead of
24615 * testsuite/ld-xtensa/tlspie.dd: New file.
24616 * testsuite/ld-xtensa/tlspie.rd: New file.
24617 * testsuite/ld-xtensa/tlspie.sd: New file.
24618 * testsuite/ld-xtensa/tlspie.td: New file.
24619 * testsuite/ld-xtensa/xtensa-linux.exp (TLS PIE transitions):
24622 2022-09-02 Max Filippov <jcmvbkbc@gmail.com>
24624 xtensa: adjust expected output in ld TLS tests
24625 objdump output for l32r opcode was changed in commit b3ea76397a07
24626 ("opcodes: xtensa: display loaded literal value"), but xtensa linker TLS
24627 relaxation tests weren't adjusted accordingly.
24628 readelf output was changed in commit 23356397449a ("Adjust readelf's
24629 output so that section symbols without a name as shown with their
24630 section name."), but xtensa linker TLS relaxation tests weren't adjusted
24632 Fix expected output changes in xtensa ld TLS relaxation tests.
24635 * testsuite/ld-xtensa/tlsbin.dd: Adjust expected output for l32r
24637 * testsuite/ld-xtensa/tlsbin.rd: Adjust expected output to allow
24638 for named section symbols.
24639 * testsuite/ld-xtensa/tlspic.dd: Adjust expected output for l32r
24641 * testsuite/ld-xtensa/tlspic.rd: Adjust expected output to allow
24642 for named section symbols.
24644 2022-09-02 Frederic Cambus <fred@statdns.com>
24646 Add OpenBSD ARM Little Endian BFD support.
24647 * config.bfd (arm-*-openbsd*): Restore target.
24649 2022-09-02 Tsukasa OI <research_trasio@irq.a4lg.com>
24651 RISC-V: Print highest address (-1) on the disassembler
24652 This patch makes possible to print the highest address (-1) and the addresses
24653 related to gp which value is -1. This is particularly useful if the highest
24654 address space is used for I/O registers and corresponding symbols are defined.
24655 Besides, despite that it is very rare to have GP the highest address, it would
24656 be nice because we enabled highest address printing on regular cases.
24660 * testsuite/gas/riscv/dis-addr-topaddr.s: New test for the top
24661 address (-1) printing.
24662 * testsuite/gas/riscv/dis-addr-topaddr-32.d: Likewise.
24663 * testsuite/gas/riscv/dis-addr-topaddr-64.d: Likewise.
24664 * testsuite/gas/riscv/dis-addr-topaddr-gp.s: New test for
24665 GP-relative addressing when GP is the highest address (-1).
24666 * testsuite/gas/riscv/dis-addr-topaddr-gp-32.d: Likewise.
24667 * testsuite/gas/riscv/dis-addr-topaddr-gp-64.d: Likewise.
24671 * riscv-dis.c (struct riscv_private_data): Add `to_print_addr' to
24672 enable printing the highest address.
24673 (maybe_print_address): Utilize `to_print_addr'.
24674 (riscv_disassemble_insn): Likewise.
24676 2022-09-02 Tsukasa OI <research_trasio@irq.a4lg.com>
24678 RISC-V: PR29342, Fix RV32 disassembler address computation
24679 If either the base register is `zero', `tp' or `gp' and XLEN is 32, an
24680 incorrectly sign-extended address is produced when printing. This commit
24681 fixes this by fitting an address into a 32-bit value on RV32.
24683 Besides, H. Peter Anvin discovered that we have wrong address computation
24684 for JALR instruction (the initial bug is back in 2018). This commit also
24685 fixes that based on the idea of Palmer Dabbelt.
24689 * testsuite/gas/riscv/lla32.d: Reflect RV32 address computation fix.
24690 * testsuite/gas/riscv/dis-addr-overflow.s: New testcase.
24691 * testsuite/gas/riscv/dis-addr-overflow-32.d: Likewise.
24692 * testsuite/gas/riscv/dis-addr-overflow-64.d: Likewise.
24695 * riscv-dis.c (maybe_print_address): Fit address into 32-bit on RV32.
24696 (print_insn_args): Fix JALR address by adding EXTRACT_ITYPE_IMM.
24698 2022-09-02 Tsukasa OI <research_trasio@irq.a4lg.com>
24700 RISC-V: Add address printer tests with ADDIW
24701 Address sequences involving ADDIW/C.ADDIW instructions require special
24702 handling to sign-extend lower 32-bits of the original result.
24704 This commit tests whether this sign-extension works.
24708 * testsuite/gas/riscv/dis-addr-addiw.s: New to test the address
24709 computation with sign extension as used in ADDIW/C.ADDIW.
24710 * testsuite/gas/riscv/dis-addr-addiw-a.d: Test PC sign bit 0.
24711 * testsuite/gas/riscv/dis-addr-addiw-b.d: Test PC sign bit 1.
24715 * testsuite/gas/riscv/dis-addr-addiw-a.d: New test.
24716 * testsuite/gas/riscv/dis-addr-addiw-b.d: New test.
24717 * testsuite/gas/riscv/dis-addr-addiw.s: New test.
24719 2022-09-02 GDB Administrator <gdbadmin@sourceware.org>
24721 Automatic date update in version.in
24723 2022-09-01 Tsukasa OI <research_trasio@irq.a4lg.com>
24725 sim: Update mailing list address
24726 The commit bf1102165389 "* MAINTAINERS: Perform some obvious fixups."
24727 back in 2009 changed the mailing list address gdb-patches@sources.redhat.com
24728 to gdb-patches@sourceware.org.
24730 This commit does the same to sim/MAINTAINERS.
24734 * MAINTAINERS: Update mailing list address.
24736 Change-Id: I56c6bf21a4bddfb35ffc3336ffcba7ff9b39926e
24738 2022-09-01 Nick Clifton <nickc@redhat.com>
24740 dllwrap, windres and dlltools use mktemp, which should be avoided
24742 * dllwrap.c: Replace uses of choose_temp_base() with
24744 * dlltool.c: Likewise.
24745 * resrc.c: Likewise.
24747 2022-09-01 Maciej W. Rozycki <macro@embecosm.com>
24749 GDB/doc: Document the Guile `#:unlimited' keyword
24750 Document the Guile `#:unlimited' keyword and deprecate the internal
24751 integer representation it corresponds to for integer parameters.
24753 2022-09-01 Lancelot SIX <lancelot.six@amd.com>
24755 gdb/python-config: replace deprecated distutils.sysconfig
24756 When running the gdb/configure script on ubuntu 22.04 with
24757 python-3.10.4, I see:
24759 checking for python... no
24760 checking for python3... /usr/bin/python3
24761 [...]/gdb/python/python-config.py:7: DeprecationWarning: The distutils package is deprecated and slated for removal in Python 3.12. Use setuptools or check PEP 632 for potential alternatives
24762 from distutils import sysconfig
24763 [...]/gdb/python/python-config.py:7: DeprecationWarning: The distutils.sysconfig module is deprecated, use sysconfig instead
24764 from distutils import sysconfig
24765 [...]/gdb/python/python-config.py:7: DeprecationWarning: The distutils package is deprecated and slated for removal in Python 3.12. Use setuptools or check PEP 632 for potential alternatives
24766 from distutils import sysconfig
24767 [...]/gdb/python/python-config.py:7: DeprecationWarning: The distutils.sysconfig module is deprecated, use sysconfig instead
24768 from distutils import sysconfig
24769 [...]/gdb/python/python-config.py:7: DeprecationWarning: The distutils package is deprecated and slated for removal in Python 3.12. Use setuptools or check PEP 632 for potential alternatives
24770 from distutils import sysconfig
24771 [...]/gdb/python/python-config.py:7: DeprecationWarning: The distutils.sysconfig module is deprecated, use sysconfig instead
24772 from distutils import sysconfig
24773 checking for python... yes
24775 The distutils module is deprecated as per the PEP 632[1] and will be
24776 removed in python-3.12.
24778 This patch migrates gdb/python/python-config.py from distutils.sysconfig
24779 to the sysconfig module[2].
24781 The sysconfig module has has been introduced in the standard library in
24782 python 3.2. Given that support for python < 3.2 has been removed by
24783 edae3fd6600f: "gdb/python: remove Python 2 support", this patch does not
24784 need to support both implementations for backward compatibility.
24786 Tested on ubuntu-22.04 and ubuntu 20.04.
24788 [1] https://peps.python.org/pep-0632/
24789 [2] https://docs.python.org/3/library/sysconfig.html
24791 Change-Id: Id0df2baf3ee6ce68bd01c236b829ab4c0a4526f6
24793 2022-09-01 GDB Administrator <gdbadmin@sourceware.org>
24795 Automatic date update in version.in
24797 2022-08-31 Tom Tromey <tromey@adacore.com>
24799 Fix interpreter-exec crash
24800 PR mi/10347 points out that using interpreter-exec inside of a
24801 "define" command will crash gdb. The bug here is that
24802 gdb_setup_readline doesn't check for the case where instream==nullptr.
24804 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=10347
24806 2022-08-31 Tom Tromey <tromey@adacore.com>
24808 Fix "source" with interpreter-exec
24809 PR mi/15811 points out that "source"ing a file that uses
24810 interpreter-exec will put gdb in a weird state, where the CLI stops
24811 working. The bug is that tui_interp::suspend does not unregister the
24812 event file descriptor.
24814 The test case is from Andrew Burgess.
24816 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=15811
24818 2022-08-31 Tom Tromey <tromey@adacore.com>
24820 Remove a call to clear_interpreter_hooks
24821 mi_interp::resume does not need to call clear_interpreter_hooks,
24822 because this is already done by interp_set.
24824 TUI stdout buffering cleanup
24825 The TUI checks against gdb_stdout to decide when to buffer. It seems
24826 much cleaner to me to simply record this as an attribute of the stream
24829 Remove a ui-related memory leak
24830 gdb_setup_readline makes new streams and assigns to the various stream
24831 members of struct ui. However, these assignments cause the previous
24832 values to leak. As far as I can, this code is simply unnecessary and
24833 can be removed -- with the exception of the assignment to gdb_stdtarg,
24834 which is not initialized anywhere else.
24837 tui_out_new is just a simple wrapper for 'new' and can be removed,
24838 simplifying gdb a tiny bit.
24840 Use scoped_restore in safe_parse_type
24841 This changes safe_parse_type to use scoped_restore rather than
24842 explicit assignments.
24844 Use member initialization in 'struct ui'
24845 This changes 'struct ui' to use member initialization. This is
24846 simpler to understand.
24848 Remove two unused members from mi_interp
24849 These members of mi_interp aren't used and can be removed.
24851 Remove obsolete filtering comment
24852 top.h has an obsolete comment about the use of _unfiltered.
24854 Remove the "for moment" comments
24855 A few spots setting some gdb output stream variables have a "for
24856 moment" comment. These comments aren't useful and I think the moment
24857 has passed -- these are permanent now.
24859 Use ui_out_redirect_pop in more places
24860 This changes ui_out_redirect_pop to also perform the redirection, and
24861 then updates several sites to use this, rather than explicit
24864 Free ui::line_buffer
24865 A ui initializes its line_buffer, but never calls buffer_free on it.
24866 This patch fixes the oversight. I found this by inspection.
24868 Remove some dead code
24869 This patch removes some dead code and an old FIXME. These no longer
24870 seem useful, even for documentation purposes.
24872 Let ui::input_fd be -1
24873 This changes gdb so that, if ui::input_fd is set to -1, then it will
24874 not be registered with the event loop. This is useful for the DAP
24875 support code I wrote, but as it turns out to also be useful to
24876 Insight, it seems best to check it in separately.
24878 2022-08-31 Andrew Burgess <aburgess@redhat.com>
24880 gdb/riscv: better support for fflags and frm registers
24881 First, some background on the RISC-V registers fflags, frm, and fcsr.
24883 These three registers all relate to the floating-point status and
24884 control mechanism on RISC-V. The fcsr is the floatint-point control
24885 status register, and consists of two parts, the flags (bits 0 to 4)
24886 and the rounding-mode (bits 5 to 7).
24888 The fcsr register is just one of many control/status registers (or
24889 CSRs) available on RISC-V. The fflags and frm registers are also
24890 CSRs. These CSRs are aliases for the relevant parts of the fcsr
24891 register. So fflags is an alias for bits 0 to 4 of fcsr, and frm is
24892 an alias for bits 5 to 7 of fcsr.
24894 This means that a user can change the floating-point rounding mode
24895 either, by writing a complete new value into fcsr, or by writing just
24896 the rounding mode into frm.
24898 How this impacts on GDB is like this: a target description could,
24899 legitimately include all three registers, fcsr, fflags, and frm. The
24900 QEMU target currently does this, and this makes sense. The target is
24901 emulating the complete system, and has all three CSRs available, so
24902 why not tell GDB about this.
24904 In contrast, the RISC-V native Linux target only has access to the
24905 fcsr. This is because the ptrace data structure that the kernel uses
24906 for reading and writing floating point state only contains a copy of
24907 the fcsr, after all, this one field really contains both the fflags
24908 and frm fields, so why carry around duplicate data.
24910 So, we might expect that the target description for the RISC-V native
24911 Linux GDB would only contain the fcsr register. Unfortunately, this
24912 is not the case. The RISC-V native Linux target uses GDB's builtin
24913 target descriptions by calling riscv_lookup_target_description, this
24914 will then add an fpu feature from gdb/features/riscv, either
24915 32bit-fpu.xml or 64bit-fpu.xml. The problem, is that these features
24916 include an entry for fcsr, fflags, and frm. This means that GDB
24917 expects the target to handle reading and writing these registers. And
24918 the RISC-V native Linux target currently doesn't.
24920 In riscv_linux_nat_target::store_registers and
24921 riscv_linux_nat_target::fetch_registers only the fcsr register is
24922 handled, this means that, for RISC-V native Linux, the fflags and frm
24923 registers always show up as <unavailable> - they are present in the
24924 target description, but the target doesn't know how to access the
24927 A final complication relating to these floating pointer CSRs is which
24928 target description feature the registers appear in.
24930 These registers are CSRs, so it would seem sensible that these
24931 registers should appear in the CSR target description feature.
24933 However, when I first added RISC-V target description support, I was
24934 using a RISC-V simulator that didn't support any CSRs other than the
24935 floating point related ones. This simulator bundled all the float
24936 related CSRs into the fpu target feature. This didn't feel completely
24937 unreasonable to me, and so I had GDB check for these registers in
24938 either target feature.
24940 In this commit I make some changes relating to how GDB handles the
24941 three floating point CSR:
24943 1. Remove fflags and frm from 32bit-fpu.xml and 64bit-fpu.xml. This
24944 means that the default RISC-V target description (which RISC-V native
24945 FreeBSD), and the target descriptions created for RISC-V native Linux,
24946 will not include these registers. There's nothing stopping some other
24947 target (e.g. QEMU) from continuing to include all three of these CSRs,
24948 the code in riscv-tdep.c continues to check for all three of these
24949 registers, and will handle them correctly if they are present.
24951 2. If a target supplied fcsr, but does not supply fflags and/or frm,
24952 then RISC-V GDB will now create two pseudo registers in order to
24953 emulate the two missing CSRs. These new pseudo-registers do the
24954 obvious thing of just reading and writing the fcsr register.
24956 3. With the new pseudo-registers we can no longer make use of the GDB
24957 register numbers RISCV_CSR_FFLAGS_REGNUM and RISCV_CSR_FRM_REGNUM.
24958 These will be the numbers used if the target supplies the registers in
24959 its target description, but, if GDB falls back to using
24960 pseudo-registers, then new, unique numbers will be used. To handle
24961 this I've added riscv_gdbarch_tdep::fflags_regnum and
24962 riscv_gdbarch_tdep::frm_regnum, I've then updated the RISC-V code to
24963 compare against these fields.
24965 When adding the pseudo-register support, it is important that the
24966 pseudo-register numbers are calculated after the call to
24967 tdesc_use_registers. This is because we don't know the total number
24968 of physical registers until after this call, and the psuedo-register
24969 numbers must follow on from the real (target supplied) registers.
24971 I've updated some tests to include more testing of the fflags and frm
24972 registers, as well as adding a new test.
24974 2022-08-31 Andrew Burgess <aburgess@redhat.com>
24976 gdb: Add tdesc_found_register function to tdesc API
24977 This commit adds a new function to the target description API within
24978 GDB. This new function is not used in this commit, but will be used
24979 in the next commit, I'm splitting it out into a separate patch for
24982 What I want to do in the next commit is check to see if a target
24983 description supplied a particular register, however, the register in
24984 question could appear in one of two possible features.
24986 The new function allows me to ask the tdesc_arch_data whether a
24987 register was found and assigned a particular GDB register number once
24988 all of the features have been checked. I think this is a much simpler
24989 solution than adding code such that, while checking each feature, I
24990 spot if the register I'm processing is the one I care about.
24992 No tests here as the new code is not used, but this code will be
24993 exercised in the next commit.
24995 2022-08-31 Andrew Burgess <aburgess@redhat.com>
24997 gdb/riscv: improve (and fix) display of frm field in 'info registers'
24998 On RISC-V the FCSR (float control/status register) is split into two
24999 parts, FFLAGS (the flags) and FRM (the rounding mode). Both of these
25000 two fields are part of the FCSR register, but can also be accessed as
25001 separate registers in their own right. And so, we have three separate
25002 registers, $fflags, $frm, and $fcsr, with the last of these being the
25003 combination of the first two.
25005 Here's how the bits of FCSR are split between FRM and FFLAGS:
25009 76543210 <----- FCSR
25013 Here's how GDB currently displays these registers:
25015 (gdb) info registers $fflags $frm $fcsr
25016 fflags 0x0 RD:0 NV:0 DZ:0 OF:0 UF:0 NX:0
25017 frm 0x0 FRM:0 [RNE (round to nearest; ties to even)]
25018 fcsr 0x0 RD:0 NV:0 DZ:0 OF:0 UF:0 NX:0 FRM:0 [RNE (round to nearest; ties to even)]
25020 Notice the 'RD' field which is present in both $fflags and $fcsr.
25021 This field contains the value of the FRM field, which makes sense when
25022 displaying the $fcsr, but makes no sense when displaying $fflags, as
25023 the $fflags doesn't include the FRM field.
25025 Additionally, the $fcsr already includes an FRM field, so the
25026 information in 'RD' is duplicated. Consider this:
25028 (gdb) set $frm = 0x3
25029 (gdb) info registers $fflags $frm $fcsr │
25030 fflags 0x0 RD:0 NV:0 DZ:0 OF:0 UF:0 NX:0
25031 frm 0x3 FRM:3 [RUP (Round up towards +INF)]
25032 fcsr 0x60 RD:3 NV:0 DZ:0 OF:0 UF:0 NX:0 FRM:3 [RUP (Round up towards +INF)]
25034 See how the 'RD' field in $fflags still displays 0, while the 'RD' and
25035 'FRM' fields in $fcsr show the same information.
25037 The first change I propose in this commit is to remove the 'RD'
25038 field. After this change the output now looks like this:
25040 (gdb) info registers $fflags $frm $fcsr
25041 fflags 0x0 NV:0 DZ:0 OF:0 UF:0 NX:0
25042 frm 0x0 FRM:0 [RNE (round to nearest; ties to even)]
25043 fcsr 0x0 NV:0 DZ:0 OF:0 UF:0 NX:0 FRM:0 [RNE (round to nearest; ties to even)]
25045 Next, I spotted that the text that goes along with the 'FRM' field was
25046 not wrapped in the i18n markers for internationalisation, so I added
25049 Next, I spotted that:
25052 (gdb) info registers $fflags $frm $fcsr
25053 fflags 0x0 RD:0 NV:0 DZ:0 OF:0 UF:0 NX:0
25054 frm 0x7 FRM:3 [RUP (Round up towards +INF)]
25055 fcsr 0xe0 RD:7 NV:0 DZ:0 OF:0 UF:0 NX:0 FRM:3 [RUP (Round up towards +INF)]
25057 Notice that despite being a 3-bit field, FRM masks to 2-bits.
25058 Checking the manual I can see that the FRM field is 3-bits, and is
25059 defined for all 8 values. That GDB masks to 2-bits is just a bug I
25060 think, so I've fixed this.
25062 Finally, the 'FRM' text for value 0x7 is wrong. Currently we use the
25063 text 'dynamic rounding mode' for value 0x7. However, this is not
25066 A RISC-V instruction can either encode the rounding mode within the
25067 instruction, or a RISC-V instruction can choose to use a global,
25068 dynamic rounding mode.
25070 So, for the rounding-mode field of an _instruction_ the value 0x7
25071 indicates "dynamic round mode", the instruction should defer to the
25072 rounding mode held in the FRM field of the $fcsr.
25074 But it makes no sense for the FRM of $fcsr to itself be set to
25075 0x7 (dynamic rounding mode), and indeed, section 11.2, "Floating-Point
25076 Control and Status Register" of the RISC-V manual, says that a value
25077 of 0x7 in the $fcsr FRM field is invalid, and if an instruction has
25078 _its_ round-mode set to dynamic, and the FRM field is also set to 0x7,
25079 then an illegal instruction exception is raised.
25081 And so, I propose changing the text for value 0x7 of the FRM field to
25082 be "INVALID[7] (Dynamic rounding mode)". We already use the text
25083 "INVALID[5]" and "INVALID[6]" for the two other invalid fields,
25084 however, I think adding the extra "Dynamic round mode" hint might be
25087 I've added a new test that uses 'info registers' to check what GDB
25088 prints for the three registers related to this patch. There is one
25089 slight oddity with this test - for the fflags and frm registers, the
25090 test accepts both the "normal" output (as described above), but also
25091 allows these registers to be reported as '<unavailable>'.
25093 The reason why I accept <unavailable> is that currently, the RISC-V,
25094 native Linux target advertises these registers in its target
25095 description, but then doesn't support reading or writing of these
25096 registers, this results in the registers being reported as
25099 A later patch in this series will address this issue, and will remove
25100 this check for <unavailable>.
25102 2022-08-31 Frederic Cambus <fred@statdns.com>
25104 Add OpenBSD AArch64 GAS support.
25105 * configure.tgt (aarch64*-*-openbsd*): Add target.
25107 2022-08-31 Nils-Christian Kempke <nils-christian.kempke@intel.com>
25109 gdb, dwarf: create symbols for template tags without names
25110 The following GDB behavior was also reported as a GDB bug in
25112 https://sourceware.org/bugzilla/show_bug.cgi?id=28396
25114 I will reiterate the problem a bit and give some more information here.
25115 This patch closes the above mentioned bug.
25117 The DWARF 5 standard 2.23 'Template Parameters' reads:
25119 A template type parameter is represented by a debugging information
25120 entry with the tag DW_TAG_template_type_parameter. A template value
25121 parameter is represented by a debugging information entry with the tag
25122 DW_TAG_template_value_parameter. The actual template parameter entries
25123 appear in the same order as the corresponding template formal
25124 parameter declarations in the source progam.
25126 A type or value parameter entry may have a DW_AT_name attribute, whose
25127 value is a null-terminated string containing the name of the
25128 corresponding formal parameter.
25130 So the DW_AT_name attribute for DW_TAG_template_type_parameter and
25131 DW_TAG_template_value_parameter is optional.
25133 Within GDB, creating a new symbol from some read DIE usually requires the
25134 presence of a DW_AT_name for the DIE (an exception here is the case of
25135 unnamed namespaces or the existence of a linkage name).
25137 This patch makes the presence of the DW_AT_name for template value/type
25138 tags optional, similar to the unnamed namespaces.
25140 For unnamed namespaces dwarf2_name simply returns the constant string
25141 CP_ANONYMOUS_NAMESPACE_STR '(anonymous namespace)'. For template tags a
25142 case was added to the switch statement calling the
25143 unnamed_template_tag_name helper. Within the scope of parent which
25144 the template parameter is a child of, the helper counts the position
25145 of the template tag within the unnamed template tags and returns
25146 '<unnamedNUMBER>' where NUMBER is its position. This way we end up with
25147 unique names within the respective scope of the function/class/struct
25148 (these are the only currenltly supported template kinds within GDB and
25149 usually the compilers) where we discovered the template tags in.
25151 While I do not know of a way to bring GCC to emit template tags without
25152 names there is one for clang/icpx. Consider the following example
25154 template<typename A, typename B, typename C>
25157 template<typename, typename B, typename>
25161 Foo<double, int, float> f;
25165 The forward declaration for 'Foo' with the missing template type names
25166 'A' and 'C' makes clang emit a bunch of template tags without names:
25169 <2><43>: Abbrev Number: 3 (DW_TAG_variable)
25170 <44> DW_AT_location : 2 byte block: 91 78 (DW_OP_fbreg: -8)
25171 <47> DW_AT_name : (indirect string, offset: 0x63): f
25172 <4b> DW_AT_decl_file : 1
25173 <4c> DW_AT_decl_line : 8
25174 <4d> DW_AT_type : <0x59>
25176 <1><59>: Abbrev Number: 5 (DW_TAG_class_type)
25177 <5a> DW_AT_calling_convention: 5 (pass by value)
25178 <5b> DW_AT_name : (indirect string, offset: 0x74): Foo<double, int, float>
25179 <5f> DW_AT_byte_size : 1
25180 <60> DW_AT_decl_file : 1
25181 <61> DW_AT_decl_line : 2
25182 <2><62>: Abbrev Number: 6 (DW_TAG_template_type_param)
25183 <63> DW_AT_type : <0x76>
25184 <2><67>: Abbrev Number: 7 (DW_TAG_template_type_param)
25185 <68> DW_AT_type : <0x52>
25186 <6c> DW_AT_name : (indirect string, offset: 0x6c): B
25187 <2><70>: Abbrev Number: 6 (DW_TAG_template_type_param)
25188 <71> DW_AT_type : <0x7d>
25191 Befor this patch, GDB would not create any symbols for the read template
25192 tag DIEs and thus lose knowledge about them. Breaking at the return
25193 statement and printing f's type would read
25196 type = class Foo<double, int, float> [with B = int] {
25200 After this patch GDB does generate symbols from the DWARF (with their
25204 type = class Foo<double, int, float> [with <unnamed0> = double, B = int,
25205 <unnamed1> = float] {
25209 The same principle theoretically applies to template functions. Also
25210 here, GDB would not record unnamed template TAGs but I know of no visual
25211 way to trigger and test this changed behavior. Template functions do
25212 not emit a '[with...]' list and their name generation also does not
25213 suffer from template tags without names. GDB does not check whether or
25214 not a template tag has a name in 'dwarf2_compute_name' and thus, the
25215 names of the template functions are created independently of whether or
25216 not the template TAGs have a DW_TAT_name attribute. A testcase has
25217 been added in the gdb.dwarf2 for template classes and structs.
25219 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28396
25221 2022-08-31 Nils-Christian Kempke <nils-christian.kempke@intel.com>
25223 gdb, testsuite: adapt function_range expected name
25224 When writing a dwarf testcase for some C++ code I wanted to use the
25225 MACRO_AT_range which in turn uses the function_range proc in dwarf.exp
25226 to extract the bounds of 'main'.
25228 However, the macro failed as GDB prints the C++ 'main' with its
25229 arguments as 'main(int, char**)' or 'main()'.
25231 The reason for this is that in read.c::dwarf2_compute_name we call
25232 c_type_print_args on C++ functions and append their arguments to the
25233 function name. This happens to all C++ functions, but is only visible
25234 when the function doesn't have a linkage name.
25236 An example might make this more clear. Given the following code
25239 int foo (int a, float b)
25244 int main (int argc, char **argv)
25249 which is legal in both languages, C and C++, and compiling it with
25250 e.g. clang or gcc will make the disassemble command look like:
25253 clang version 10.0.0-4ubuntu1
25255 >> clang -O0 -g ./c.cpp
25256 >> gdb -q ./a.out -ex "start"
25258 (gdb) disassemble main
25259 Dump of assembler code for function main(int, char**):
25260 0x0000000000401120 <+0>: push %rbp
25261 0x0000000000401121 <+1>: mov %rsp,%rbp
25263 0x0000000000401135 <+21>: ret
25264 End of assembler dump.
25265 (gdb) disassemble foo
25266 Dump of assembler code for function _Z3fooif:
25267 0x0000000000401110 <+0>: push %rbp
25268 0x0000000000401111 <+1>: mov %rsp,%rbp
25270 0x000000000040111f <+15>: ret
25271 End of assembler dump.
25273 Note, that main is emitted with its arguments while for foo the linkage
25274 name is being printed, as also visible in its DWARF:
25276 >> objdump ./a.out --dwarf=info | grep "foo" -A3 -B3
25277 <2b> DW_AT_low_pc : 0x401110
25278 <33> DW_AT_high_pc : 0x10
25279 <37> DW_AT_frame_base : 1 byte block: 56 (DW_OP_reg6 (rbp))
25280 <39> DW_AT_linkage_name: (indirect string, offset: 0x39): _Z3fooif
25281 <3d> DW_AT_name : (indirect string, offset: 0x42): foo
25282 <41> DW_AT_decl_file : 1
25283 <42> DW_AT_decl_line : 1
25284 <43> DW_AT_type : <0x9a>
25286 Now, let's rename the C++ file and compile it as C:
25289 >> clang -O0 -g ./c.c
25290 >> gdb -q ./a.out -ex "start'
25292 (gdb) disassemble main
25293 Dump of assembler code for function main:
25294 0x0000000000401120 <+0>: push %rbp
25295 0x0000000000401121 <+1>: mov %rsp,%rbp
25297 0x0000000000401135 <+21>: ret
25298 End of assembler dump.
25299 (gdb) disassemble foo
25300 Dump of assembler code for function foo:
25301 0x0000000000401110 <+0>: push %rbp
25302 0x0000000000401111 <+1>: mov %rsp,%rbp
25304 0x000000000040111f <+15>: ret
25305 End of assembler dump.
25307 Note, for foo we did not get a linkage name emitted in DWARF, so
25308 it is printed by its name:
25310 >> objdump --dwarf=info ./a.out | grep foo -A3 -B3
25311 <2b> DW_AT_low_pc : 0x401110
25312 <33> DW_AT_high_pc : 0x10
25313 <37> DW_AT_frame_base : 1 byte block: 56 (DW_OP_reg6 (rbp))
25314 <39> DW_AT_name : (indirect string, offset: 0x37): foo
25315 <3d> DW_AT_decl_file : 1
25316 <3e> DW_AT_decl_line : 1
25317 <3f> DW_AT_prototyped : 1
25319 To make the macro and proc work with C++ as well, an optional argument
25320 list was added to the regex matching the function name in the
25321 disassemble command in function_range. This does not change any used
25322 behavior as currently, there exists no C++ test using the proc
25325 2022-08-31 Aaron Merey <amerey@redhat.com>
25327 gdb/elfread.c: Use bfd filename instead of objfile->original_name
25328 The call to debuginfod_debuginfo_query in elf_symfile_read is given
25329 objfile->original_name as the filename to print when downloading the
25330 objfile's debuginfo.
25332 In some cases original_name is prefixed with gdb's working directory
25333 even though the objfile is not located in the working directory. This
25334 causes debuginfod to display the wrong path of the objfile during a download.
25336 Fix this by using the objfile's bfd filename instead.
25338 2022-08-31 GDB Administrator <gdbadmin@sourceware.org>
25340 Automatic date update in version.in
25342 2022-08-30 Martin Storsjö <martin@martin.st>
25344 ld: pe: Fix linking against Microsoft import libraries with multiple DLLs
25345 Initially, since c6c37250e98f113755e0d787f7070e2ac80ce77e (in 1999),
25346 in order to fix linking against Microsoft import libraries, ld did
25347 internally rename members of such libraries. At that point, the
25348 criteria for being considered a Microsoft import library was that
25349 every archive member had the same name (no regard for exactly what
25352 This was later amended in 44dbf3639f127af46d569ad96b6242dfbc4c0a89
25353 (in 2003) to allow for Microsoft import libraries with intermixed
25354 static object files. At this point, the criteria were extended, so
25355 that all members following the first member named *.dll either had
25356 the exact same member name, or be named *.obj. (Curiously, this would
25357 allow members with any name if it precedes the first one named *.dll.)
25359 In practice, Microsoft style import libraries can contain
25360 members for linking against more than one DLL (built by merging
25361 multiple regular import libraries into one).
25363 Instead of trying to do validation of the whole archive before
25364 considering it a Microsoft style import library, relax the criteria
25365 for doing the member renaming: If an archive member is named *.dll
25366 and it contains .idata sections, assume that that member is a
25367 Microsoft import file, and apply the renaming scheme.
25369 This works for imports for any number of DLLs in the same library,
25370 intermixed with other static object files (regardless of their
25371 names), and vastly simplifies the code.
25373 LLVM generates Microsoft style import libraries, and Rust builds
25374 seem to bundle up multiple import libraries together with some
25375 Rust specific static objects. This fixes linking directly against
25378 2022-08-30 Simon Marchi <simon.marchi@polymtl.ca>
25380 gdbsupport: add wrapper around result_of and invoke_result
25381 When building with Clang 14 (using gcc 12 libstdc++ headers), I get:
25384 In file included from /home/simark/src/binutils-gdb/gdb/dwarf2/read.c:94:
25385 /home/simark/src/binutils-gdb/gdb/../gdbsupport/parallel-for.h:142:21: error: 'result_of<(lambda at /home/simark/src/binutils-gdb/gdb/dwarf2/read.c:7124:5) (__gnu_debug::_Safe_iterator<__gnu_cxx::__normal_iterator<std::unique_ptr<dwarf2_per_cu_data, dwarf2_per_cu_data_deleter> *, std::__cxx1998::vector<std::unique_ptr<dwarf2_per_cu_data, dwarf2_per_cu_data_deleter>>>, std::vector<std::unique_ptr<dwarf2_per_cu_data, dwarf2_per_cu_data_deleter>>, std::random_access_iterator_tag>, __gnu_debug::_Safe_iterator<__gnu_cxx::__normal_iterator<std::unique_ptr<dwarf2_per_cu_data, dwarf2_per_cu_data_deleter> *, std::__cxx1998::vector<std::unique_ptr<dwarf2_per_cu_data, dwarf2_per_cu_data_deleter>>>, std::vector<std::unique_ptr<dwarf2_per_cu_data, dwarf2_per_cu_data_deleter>>, std::random_access_iterator_tag>)>' is deprecated: use 'std::invoke_result' instead [-Werror,-Wdeprecated-declarations]
25386 = typename std::result_of<RangeFunction (RandomIt, RandomIt)>::type;
25388 /home/simark/src/binutils-gdb/gdb/dwarf2/read.c:7122:14: note: in instantiation of function template specialization 'gdb::parallel_for_each<__gnu_debug::_Safe_iterator<__gnu_cxx::__normal_iterator<std::unique_ptr<dwarf2_per_cu_data, dwarf2_per_cu_data_deleter> *, std::__cxx1998::vector<std::unique_ptr<dwarf2_per_cu_data, dwarf2_per_cu_data_deleter>>>, std::vector<std::unique_ptr<dwarf2_per_cu_data, dwarf2_per_cu_data_deleter>>, std::random_access_iterator_tag>, (lambda at /home/simark/src/binutils-gdb/gdb/dwarf2/read.c:7124:5)>' requested here
25389 = gdb::parallel_for_each (1, per_bfd->all_comp_units.begin (),
25391 /usr/bin/../lib64/gcc/x86_64-pc-linux-gnu/12.1.1/../../../../include/c++/12.1.1/type_traits:2597:9: note: 'result_of<(lambda at /home/simark/src/binutils-gdb/gdb/dwarf2/read.c:7124:5) (__gnu_debug::_Safe_iterator<__gnu_cxx::__normal_iterator<std::unique_ptr<dwarf2_per_cu_data, dwarf2_per_cu_data_deleter> *, std::__cxx1998::vector<std::unique_ptr<dwarf2_per_cu_data, dwarf2_per_cu_data_deleter>>>, std::vector<std::unique_ptr<dwarf2_per_cu_data, dwarf2_per_cu_data_deleter>>, std::random_access_iterator_tag>, __gnu_debug::_Safe_iterator<__gnu_cxx::__normal_iterator<std::unique_ptr<dwarf2_per_cu_data, dwarf2_per_cu_data_deleter> *, std::__cxx1998::vector<std::unique_ptr<dwarf2_per_cu_data, dwarf2_per_cu_data_deleter>>>, std::vector<std::unique_ptr<dwarf2_per_cu_data, dwarf2_per_cu_data_deleter>>, std::random_access_iterator_tag>)>' has been explicitly marked deprecated here
25392 { } _GLIBCXX17_DEPRECATED_SUGGEST("std::invoke_result");
25394 /usr/bin/../lib64/gcc/x86_64-pc-linux-gnu/12.1.1/../../../../include/c++/12.1.1/x86_64-pc-linux-gnu/bits/c++config.h:120:45: note: expanded from macro '_GLIBCXX17_DEPRECATED_SUGGEST'
25395 # define _GLIBCXX17_DEPRECATED_SUGGEST(ALT) _GLIBCXX_DEPRECATED_SUGGEST(ALT)
25397 /usr/bin/../lib64/gcc/x86_64-pc-linux-gnu/12.1.1/../../../../include/c++/12.1.1/x86_64-pc-linux-gnu/bits/c++config.h:96:19: note: expanded from macro '_GLIBCXX_DEPRECATED_SUGGEST'
25398 __attribute__ ((__deprecated__ ("use '" ALT "' instead")))
25401 It complains about the use of std::result_of, which is deprecated in
25402 C++17 and removed in C++20:
25404 https://en.cppreference.com/w/cpp/types/result_of
25406 Given we'll have to transition to std::invoke_result eventually, make a
25407 GDB wrapper to mimimc std::invoke_result, which uses std::invoke_result
25408 for C++ >= 17 and std::result_of otherwise. This way, it will be easy
25409 to remove the wrapper in the future, just replace gdb:: with std::.
25411 Tested by building with gcc 12 in -std=c++11 and -std=c++17 mode, and
25412 clang in -std=c++17 mode (I did not test fully with clang in -std=c++11
25413 mode because there are other unrelated issues).
25415 Change-Id: I50debde0a3307a7bc67fcf8fceefda51860efc1d
25417 2022-08-30 Tom Tromey <tromey@adacore.com>
25419 Fix flush for sys.stderr
25420 GDB overwrites Python's sys.stdout and sys.stderr, but does not
25421 properly implement the 'flush' method -- it only ever will flush
25422 stdout. This patch fixes the bug. I couldn't find a straightforward
25423 way to write a test for this.
25425 Fix gdb.flush documentation
25426 The gdb.flush documentation does not mention the 'stream' argument in
25427 the function signature, only in the description. This patch fixes the
25430 2022-08-30 Nick Clifton <nickc@redhat.com>
25432 BFD library: Use entry 0 in directory and filename tables of DWARF-5 debug info.
25434 * dwarf2.c (struct line_info_table): Add new field:
25435 use_dir_and_file_0.
25436 (concat_filename): Use new field to help select the correct table
25438 (read_formatted_entries): Do not skip entry 0.
25439 (decode_line_info): Set new field depending upon the version of
25440 DWARF being parsed. Initialise filename based upon the setting of
25443 2022-08-30 Enze Li <enze.li@hotmail.com>
25445 gdb: update ranged_breakpoint::print_one_detail in comments
25446 The print_one_detail_ranged_breakpoint has been renamed to
25447 ranged_breakpoint::print_one_detail in this commit:
25449 commit ec45bb676c9c69c30783bcf35ffdac8280f3b8bc
25450 Date: Sat Jan 15 16:34:51 2022 -0700
25452 Convert ranged breakpoints to vtable ops
25454 So their comments should be updated as well.
25456 2022-08-30 Nick Clifton <nickc@redhat.com>
25458 Add a testcase for PR 29494.
25460 * testsuite/gas/arm/pr29494.s: New test source file.
25461 * testsuite/gas/arm/pr29494.d: New test driver.
25463 2022-08-30 liuzhensong <liuzhensong@loongson.cn>
25465 LoongArch: Fix redefinition of "PACKAGE".
25466 Running configure and make in binutils-gdb.
25470 In file included from ./as.h:37,
25471 from ./config/loongarch-lex.l:21,
25472 from config/loongarch-lex-wrapper.c:20:
25473 ./config.h:206: error: “PACKAGE” redefined [-Werror]
25474 #define PACKAGE "gas"
25478 * loongarch-lex-wrapper.c
25480 2022-08-30 Tsukasa OI <research_trasio@irq.a4lg.com>
25482 RISC-V: Add 'Zmmul' extension in assembler.
25483 Three-part patch set from Tsukasa OI to support zmmul in assembler.
25485 The 'Zmmul' is a RISC-V extension consisting of only multiply instructions
25486 (a subset of 'M' which has multiply and divide instructions).
25489 * elfxx-riscv.c (riscv_implicit_subsets): Add 'Zmmul' implied by 'M'.
25490 (riscv_supported_std_z_ext): Add 'Zmmul' extension.
25491 (riscv_multi_subset_supports): Add handling for new instruction class.
25493 * testsuite/gas/riscv/attribute-09.d: Updated implicit 'Zmmul' by 'M'.
25494 * testsuite/gas/riscv/option-arch-02.d: Likewise.
25495 * testsuite/gas/riscv/m-ext.s: New test.
25496 * testsuite/gas/riscv/m-ext-32.d: New test (RV32).
25497 * testsuite/gas/riscv/m-ext-64.d: New test (RV64).
25498 * testsuite/gas/riscv/zmmul-32.d: New expected output.
25499 * testsuite/gas/riscv/zmmul-64.d: Likewise.
25500 * testsuite/gas/riscv/m-ext-fail-xlen-32.d: New test (failure
25501 by using RV64-only instructions in RV32).
25502 * testsuite/gas/riscv/m-ext-fail-xlen-32.l: Likewise.
25503 * testsuite/gas/riscv/m-ext-fail-zmmul-32.d: New failure test
25504 (RV32 + Zmmul but with no M).
25505 * testsuite/gas/riscv/m-ext-fail-zmmul-32.l: Likewise.
25506 * testsuite/gas/riscv/m-ext-fail-zmmul-64.d: New failure test
25507 (RV64 + Zmmul but with no M).
25508 * testsuite/gas/riscv/m-ext-fail-zmmul-64.l: Likewise.
25509 * testsuite/gas/riscv/m-ext-fail-noarch-64.d: New failure test
25511 * testsuite/gas/riscv/m-ext-fail-noarch-64.l: Likewise.
25513 * opcode/riscv.h (enum riscv_insn_class): Added INSN_CLASS_ZMMUL.
25515 * testsuite/ld-riscv-elf/attr-merge-arch-01.d: We don't care zmmul in
25516 these testcases, so just replaced m by a.
25517 * testsuite/ld-riscv-elf/attr-merge-arch-01a.s: Likewise.
25518 * testsuite/ld-riscv-elf/attr-merge-arch-01b.s: Likewise.
25519 * testsuite/ld-riscv-elf/attr-merge-arch-02.d: Likewise.
25520 * testsuite/ld-riscv-elf/attr-merge-arch-02a.s: Likewise.
25521 * testsuite/ld-riscv-elf/attr-merge-arch-03.d: Likewise.
25522 * testsuite/ld-riscv-elf/attr-merge-arch-03a.s: Likewise.
25523 * testsuite/ld-riscv-elf/attr-merge-user-ext-01.d: Likewise.
25524 * testsuite/ld-riscv-elf/attr-merge-user-ext-rv32i2p1_a2p0.s: Renamed.
25525 * testsuite/ld-riscv-elf/attr-merge-user-ext-rv32i2p1_a2p1.s: Renamed.
25527 * riscv-opc.c (riscv_opcodes): Updated multiply instructions to zmmul.
25529 2022-08-30 Tom de Vries <tdevries@suse.de>
25531 [gdb/symtab] Fix assert in set_length
25532 When running the included test-case, we run into:
25534 (gdb) break _start^M
25535 read.h:309: internal-error: set_length: \
25536 Assertion `m_length == length' failed.^M
25539 The problem is that while there are two CUs:
25541 $ readelf -wi debug-names-missing-cu | grep @
25542 Compilation Unit @ offset 0x0:
25543 Compilation Unit @ offset 0x2d:
25545 the CU table in the .debug_names section only contains the first one:
25551 The incomplete CU table makes create_cus_from_debug_names_list set the size of
25552 the CU at 0x0 to the actual size of both CUs combined.
25554 This eventually leads to the assert, when we read the actual size from the CU
25557 While having an incomplete CU table in a .debug_names section is incorrect,
25558 we need a better failure mode than asserting.
25560 The easiest way to fix this is to set the length to 0 (meaning: unkown) in
25561 create_cus_from_debug_names_list.
25563 This makes the failure mode to accept the incomplete CU table, but to ignore
25566 It would be nice to instead reject the .debug_names index, and build a
25567 complete CU list, but the point where we find this out is well after
25568 dwarf2_initialize_objfile, so it looks rather intrusive to restart at that
25571 Tested on x86_64-linux.
25573 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29453
25575 2022-08-30 Tom de Vries <tdevries@suse.de>
25577 [gdb/tdep] Declare score-*-* target obsolete
25578 I tried out the script gdb/gdb_mbuild.sh, and ran into:
25581 ... configure --target=score-elf
25584 score-elf: gdb dumped core
25588 Gdb runs into this internal error in initialize_current_architecture:
25590 if (! gdbarch_update_p (info))
25591 internal_error (__FILE__, __LINE__,
25592 _("initialize_current_architecture: Selection of "
25593 "initial architecture failed"));
25596 The call to gdbarch_update_p fails because commit 575b4c298a6 ("gdb: Remove
25597 support for S+core") removed support for the architecture.
25599 Fix this by adding score-*-* to the list of obsolete targets in
25600 gdb/configure.tgt, such that we're no longer able to build the configuration:
25602 *** Configuration score-unknown-elf is obsolete.
25603 *** Support has been REMOVED.
25604 make: *** [Makefile:12806: configure-gdb] Error 1
25607 Also remove the related line from the "Target Instruction Set Architectures"
25608 list in gdb/MAINTAINERS, such that gdb/gdb_mbuild.sh no longer tries to build
25611 2022-08-30 GDB Administrator <gdbadmin@sourceware.org>
25613 Automatic date update in version.in
25615 2022-08-29 GDB Administrator <gdbadmin@sourceware.org>
25617 Automatic date update in version.in
25619 2022-08-28 Alan Modra <amodra@gmail.com>
25621 PR29494 Trailing jump table on ARM
25622 out_inc_line_addr and relax_inc_line_addr are passed INT_MAX as
25623 line_delta to flag end of section. This filters its way down to
25624 size_inc_line_addr and emit_inc_line_addr. Pass line_delta on to
25625 scale_addr_delta where it can be used to omit an unaligned opcode
25629 * dwarf2dbg.c (scale_addr_delta): Delete unnecessary forward decl.
25630 Add line_delta param. Don't print error at end of section, just
25631 round the address down.
25632 (size_inc_line_addr, emit_inc_line_addr): Adjust calls.
25634 2022-08-28 GDB Administrator <gdbadmin@sourceware.org>
25636 Automatic date update in version.in
25638 2022-08-27 rupothar <rupesh.potharla@amd.com>
25640 bfd: Fix minor bug in read_indexed_address function.
25641 read_indexed_address function is using offset_size instead of
25642 addr_size while reading addrx forms.
25644 2022-08-27 GDB Administrator <gdbadmin@sourceware.org>
25646 Automatic date update in version.in
25648 2022-08-26 Simon Marchi <simon.marchi@polymtl.ca>
25650 gdbsupport: fix gdb::optional compilation with C++11 && _GLIBCXX_DEBUG
25651 Similar to 911438f9f4 ("gdbsupport: fix array-view compilation with
25652 c++11 && _GLIBCXX_DEBUG"), but for gdb::optional.
25654 I get this error when building with Clang 14 and -std=c++11:
25657 In file included from /home/simark/src/binutils-gdb/gdbsupport/agent.cc:20:
25658 In file included from /home/simark/src/binutils-gdb/gdbsupport/common-defs.h:210:
25659 In file included from /home/simark/src/binutils-gdb/gdbsupport/common-debug.h:23:
25660 /home/simark/src/binutils-gdb/gdbsupport/../gdbsupport/gdb_optional.h:213:5: error: use of this statement in a constexpr function is a C++14 extension [-Werror,-Wc++14-extensions]
25661 gdb_assert (this->has_value ());
25663 /home/simark/src/binutils-gdb/gdbsupport/gdb_assert.h:35:3: note: expanded from macro 'gdb_assert'
25664 ((void) ((expr) ? 0 : \
25667 Change-Id: If0cf55607fc9dbd1925ccb97cd9abbf8993ff264
25669 2022-08-26 Simon Marchi <simon.marchi@polymtl.ca>
25671 gdb: change bpstat_print's kind parameter to target_waitkind
25672 Change from int to target_waitkind, which is really what is is. While
25673 at it, remove some outdated doc. The return value is described by a
25674 relatively self-describing enum, not a numerical value like the doc
25677 Change-Id: Id899c853a857c7891c45e5b1639024067d5b59cd
25679 2022-08-26 Simon Marchi <simon.marchi@efficios.com>
25681 gdb, gdbsupport: configure: factor out yes/no/auto value checking
25682 Factor out the code that checks that a value is yes/no or yes/no/auto.
25683 Add two macros to gdbsupport/common.m4 and use them in gdb/configure.ac
25685 I inspected the changes to configure. Other than whitespace changes, we
25686 have some benign changes to the error messages (one of them had an error
25687 actually). There are changes to the --enable-source-highlight and
25688 --enable-libbacktrace handling, but setting enable_source_highlight /
25689 enable_libbacktrace was not really useful anyway, they already had the
25692 Change-Id: I92587aec36874309e1605e2d60244649f09a757a
25694 2022-08-26 Alan Modra <amodra@gmail.com>
25696 PR12265, Compiling ld/ fails on Solaris 8
25697 The fail was due to -Werror and headers included by dlfcn.h and
25698 elf-bfd.h disagreeing about AT_DCACHEBSIZE and other AT_*. Not a
25699 serious problem obviously, since release versions of binutils don't
25700 enable -Werror and the defines are not used. Anyway, reduce the
25701 number of files that might hit this problem by only including dlfcn.h
25702 where it is needed.
25705 * sysdep.h: Don't include dlfcn.h here.
25706 * plugin.c: Include it here.
25708 2022-08-26 GDB Administrator <gdbadmin@sourceware.org>
25710 Automatic date update in version.in
25712 2022-08-25 Philippe Waroquiers <philippe.waroquiers@skynet.be>
25714 Allow to document user-defined aliases.
25715 Compared to the previous version, this version fixes the comments reported by
25716 Tom Tromey and ensures that the 'help some-user-documented-alias'
25717 shows the alias definition to ensure the user understands this is an
25718 alias even if specifically documented.
25720 When using 'help ALIASNAME', GDB shows the help of the aliased command.
25721 This is a good default behaviour.
25723 However, GDB alias command allows to define aliases with arguments
25724 possibly changing or tuning significantly the behaviour of
25725 the aliased command. In such a case, showing the help of the aliased
25726 command might not be ideal.
25728 This is particularly true when defining an alias as a set of
25729 nested 'with' followed by a last command to launch, such as:
25730 (gdb) alias pp10 = with print pretty -- with print elements 10 -- print
25731 Asking 'help pp10' shows the help of the 'with' command, which is
25732 not particularly useful:
25735 alias pp10 = with print pretty -- with print elements 10 -- print
25736 Temporarily set SETTING to VALUE, run COMMAND, and restore SETTING.
25737 Usage: with SETTING [VALUE] [-- COMMAND]
25740 Such an alias can now be documented by the user:
25741 (gdb) document pp10
25742 >Pretty printing an expressiong, printing 10 elements.
25743 >Usage: pp10 [PRINT-COMMAND-OPTIONS] EXP
25744 >See 'help print' for more information.
25747 alias pp10 = with print pretty -- with print elements 10 -- print
25748 Pretty printing an expressiong, printing 10 elements.
25749 Usage: pp10 [PRINT-COMMAND-OPTIONS] EXP
25750 See 'help print' for more information.
25753 When a user-defined alias is documented specifically, help and apropos
25754 use the provided alias documentation instead of the documentation of
25755 the aliased command.
25757 Such a documented alias is also not shown anymore in the help of the
25758 aliased command, and the alias is not listed anymore in the help
25759 of the aliased command. In particular for cases such as pp10 example above,
25760 indicating that pp10 is an alias of the 'with' command is confusing.
25762 2022-08-25 Jan-Benedict Glaw <jbglaw@lug-owl.de>
25764 sim/aarch64: Fix aarch64_get_CPSR_bits() declaration
25765 Noticed while doing mass builds with a very recent GCC:
25767 /usr/lib/gcc-snapshot/bin/gcc -DHAVE_CONFIG_H -DWITH_HW=1 -DHAVE_DV_SOCKSER -DDEFAULT_INLINE=0 -Wall -Wdeclaration-after-statement -Wpointer-arith -Wno-unused -Wunused-value -Wunused-function -Wno-switch -Wno-char-subscripts -Wempty-body -Wunused-but-set-parameter -Wno-error=maybe-uninitialized -Wmissing-declarations -Wmissing-prototypes -Wdeclaration-after-statement -Wmissing-parameter-type -Wpointer-sign -Wold-style-declaration -Werror -I. -I/var/lib/laminar/run/gdb-aarch64-elf/1/binutils-gdb/sim/aarch64 -I../common -I/var/lib/laminar/run/gdb-aarch64-elf/1/binutils-gdb/sim/aarch64/../common -I../../include -I/var/lib/laminar/run/gdb-aarch64-elf/1/binutils-gdb/sim/aarch64/../../include -I../../bfd -I/var/lib/laminar/run/gdb-aarch64-elf/1/binutils-gdb/sim/aarch64/../../bfd -I../../opcodes -I/var/lib/laminar/run/gdb-aarch64-elf/1/binutils-gdb/sim/aarch64/../../opcodes -I../.. -I/var/lib/laminar/run/gdb-aarch64-elf/1/binutils-gdb/sim/aarch64/../../gnulib/import -I../../gnulib/import -g -O2 -c -o cpustate.o -MT cpustate.o -MMD -MP -MF .deps/cpustate.Tpo cpustate.c
25768 cpustate.c:270:1: error: conflicting types for 'aarch64_get_CPSR_bits' due to enum/integer mismatch; have 'uint32_t(sim_cpu *, FlagMask)' {aka 'unsigned int(struct _sim_cpu *, FlagMask)'} [-Werror=enum-int-mismatch]
25769 270 | aarch64_get_CPSR_bits (sim_cpu *cpu, FlagMask mask)
25770 | ^~~~~~~~~~~~~~~~~~~~~
25771 In file included from sim-main.h:30,
25772 from cpustate.c:28:
25773 cpustate.h:310:20: note: previous declaration of 'aarch64_get_CPSR_bits' with type 'uint32_t(sim_cpu *, uint32_t)' {aka 'unsigned int(struct _sim_cpu *, unsigned int)'}
25774 310 | extern uint32_t aarch64_get_CPSR_bits (sim_cpu *, uint32_t);
25775 | ^~~~~~~~~~~~~~~~~~~~~
25776 cc1: all warnings being treated as errors
25778 2022-08-25 H.J. Lu <hjl.tools@gmail.com>
25780 x86: Ignore protected visibility in shared libraries on Solaris
25781 On x86, the PLT entry in executable may be used as function address for
25782 functions in shared libraries. If functions are protected, the function
25783 address used in executable can be different from the function address
25784 used in shared library. This will lead to incorrect run-time behavior
25785 if function pointer equality is needed. By default, x86 linker issues
25786 an error in this case.
25788 On Solaris, linker issued an error for
25790 struct tm *tb = (kind == CPP_time_kind::FIXED ? gmtime : localtime) (&tt);
25792 where gmtime is a protected function in libc.so. Use gmtime's PLT entry
25793 in executable as function address is safe since function pointer equality
25794 isn't needed. Ignore protected visibility in shared libraries on Solaris
25795 to disable linker error. If function pointer equality is needed, linker
25796 will silently generate executable with incorrect run-time behavior on
25800 * elf32-i386.c (elf_i386_scan_relocs): Ignore protected
25801 visibility in shared libraries on Solaris.
25802 * elf64-x86-64.c (elf_x86_64_scan_relocs): Likewise.
25804 2022-08-25 Nick Clifton <nickc@redhat.com>
25806 GAS: Add a return type tag to DWARF DIEs generated for function symbols.
25808 * dwarf2dbg.c (GAS_ABBREV_COMP_UNIT): New defined constant.
25809 (GAS_ABBREV_SUBPROG): New defined constant.
25810 (GAS_ABBREV_NO_TYPE): New defined constant.
25811 (out_debug_abbrev): Use the new defined constants when emitting
25812 abbreviation numbers. Generate an abbreviation for an unspecified
25814 (out_debug_info): Use the new defined constants when referring to
25815 abbreviations. Generate a use of the no_type abbreviation.
25816 Reference the use when generating DIEs for functions.
25817 * testsuite/gas/elf/dwarf-3-func.d: Update to allow for newly
25818 extended output from the assembler.
25819 * testsuite/gas/elf/dwarf-5-func-global.d: Likewise.
25820 * testsuite/gas/elf/dwarf-5-func-local.d: Likewise.
25821 * testsuite/gas/elf/dwarf-5-func.d: Likewise.
25823 GAS: Allow AArch64 pseudo-ops to accept the command line separator character.
25825 * config/tc-aarch64.c (s_unreq): Use find_end_of_line().
25826 (s_aarch64_cpu): Likewise.
25827 (s_aarch64_arch): Likewise.
25828 (s_aarch64_arch_extension): Likewise.
25829 * testsuite/gas/aarch64/pr29519.d: New test driver file.
25830 * testsuite/gas/aarch64/pr29519.s: New test source file.
25832 2022-08-25 Palmer Dabbelt <palmer@rivosinc.com>
25834 gas: NEWS: Add the RISC-V features for 2.39
25836 gas: NEWS: Add the RISC-V features for 2.38
25838 gas: NEWS: Add the RISC-V features for 2.37
25840 gas: NEWS: Add the RISC-V features for 2.36
25842 gas: NEWS: Add the RISC-V features for 2.35
25844 gas: NEWS: Add the RISC-V features for 2.31
25846 2022-08-25 Alan Modra <amodra@gmail.com>
25848 PR11290, avr-ld "out of range error" is confusing
25849 Don't overload bfd_reloc_outofrange with what is really a domain error
25850 (target at odd address), or an overflow.
25853 * reloc.c (bfd_reloc_other): Correct comment.
25854 * elf32-avr.c (avr_final_link_relocate): Return bfd_reloc_other
25855 for unaligned reloc target values. Return bfd_reloc_overflow
25856 when stubs are too far away and when R_AVR_LDS_STS_16,
25857 R_AVR_PORT6, or R_AVR_PORT5 overflow.
25858 (elf32_avr_relocate_section): Report more descriptive relocation
25860 * bfd-in2.h: Regenerate.
25862 2022-08-25 Martin Storsjö <martin@martin.st>
25864 ld: pe: Move the return type to a separate line from the function name
25865 This fixes the coding style of an old, preexisting function.
25867 2022-08-25 Alan Modra <amodra@gmail.com>
25869 PR10372, SH: ld test with sim/sh/run fails always
25871 * testsuite/ld-sh/start.s: Add _start sym. Use trapa 34. Create
25872 an alloc .stack section.
25874 2022-08-25 Alan Modra <amodra@gmail.com>
25876 Re: LoongArch: ld: Fix bug not generate plt when link a dso
25877 Fixes loongarch32-elf +FAIL: medium jirl plt
25879 * testsuite/ld-loongarch-elf/cmodel.exp: Don't run test when
25880 no shared library support.
25882 2022-08-25 GDB Administrator <gdbadmin@sourceware.org>
25884 Automatic date update in version.in
25886 2022-08-24 Martin Storsjö <martin@martin.st>
25888 ld: pe: Make archive member file extension comparisons case insensitive when cross compiling too
25889 On Windows, filename_cmp is case insensitive, but when cross compiling
25890 with libraries that may contain members with uppercase file names, we
25891 should keep those comparisons case insensitive when running the build
25892 tools on other OSes too.
25894 Also make the check for .def consistent with the other ones, fixing
25895 out of bounds reads if file names are shorter than 4 characters.
25897 2022-08-24 Richard Earnshaw <rearnsha@arm.com>
25899 gas: arm: handle multiple .directives on a single line (PR29519)
25900 There's been a long-standing bug in the arm backend where
25901 target-specific directives did not correctly handle lines with
25902 multiple statements. This patch fixes the issue for all the cases
25903 I've been able to find.
25905 It does result in a slight change in behaviour when errors are
25906 encountered: where, previously,
25910 would result in the error "junk at end of line, first unrecognized
25911 character is `b'", we now get "unknown cpu `arm6 bar'", which I think
25912 is slightly more helpful anyway. Similar errors are generated for
25915 2022-08-24 Andrew Burgess <aburgess@redhat.com>
25917 gdb: new 'maint print frame-id' command
25918 When debugging a certain class of GDB bug, I often end up wanting to
25919 know what GDB thinks the frame-id is in a particular frame. It's
25920 not too hard to pull this from some debug output, but I thought it
25921 might be nice if there was a maintenance command that could tell us.
25923 This commit adds 'maint print frame-id' which prints the frame-id of
25924 the currently selected frame. You can also pass a frame level number
25925 to find the frame-id for a specific frame.
25927 There's a new test too.
25929 2022-08-24 liuzhensong <liuzhensong@loongson.cn>
25931 LoongArch: ld: Fix bug not generate plt when link a dso
25932 Fix the bug that can not generate func@plt
25933 when linking a undefined function with cmodel=medium.
25937 * elfnn-loongarch.c
25938 ld/testsuite/ld-loongarch-elf/
25939 * cmodel-libjirl.dd
25943 2022-08-24 GDB Administrator <gdbadmin@sourceware.org>
25945 Automatic date update in version.in
25947 2022-08-23 Alan Modra <amodra@gmail.com>
25949 SHT_RELR sh_link and sh_info
25950 I don't think it makes any sense for a SHT_RELR section to specify a
25951 symbol table with sh_link. SHT_RELR relocations don't use symbols.
25952 There is no real need to specify sh_info either, SHT_RELR is not for
25953 relocatable object files. Anyway, fuzzers of course don't restrict
25954 themselves to even half-sensible objects. So they found a hole in
25955 objcopy using a non-alloc SHT_RELR in an ET_EXEC. In that case BFD
25956 set up the SHT_RELR section as if it were a SHT_REL against the
25957 sh_info target section. When it came to reading in the target section
25958 relocs, the count was horribly wrong which caused a buffer overflow.
25960 * elf.c (bfd_section_from_shdr <SHT_RELR>): Always just make a
25961 normal section, don't treat it as a reloc section.
25963 2022-08-23 Alan Modra <amodra@gmail.com>
25965 Re: bfd_elf_set_group_contents assertion
25966 Further to commit 7744e3278b9f.
25968 * elf.c (bfd_elf_set_group_contents): Restrict loc in loop writing
25969 contents, and add another assertion.
25971 2022-08-23 Nick Clifton <nickc@redhat.com>
25973 Add an option to dlltool to allow the creation of deterministic libraries.
25975 * dlltool.c (deterministic): New variable.
25976 (gen_lib_file): If deterministic is true set the
25977 BFD_DETERMINISTIC_OUTPUT flag.
25978 (usage): Mention --deterministic-libraries and
25979 --non-deterministic-libraries.
25980 (long_options): Add new options.
25981 (main): Parse new options.
25982 * doc/binutils.texi: Document the new options.
25983 * NEWS: Mention the new feature.
25985 2022-08-23 Nelson Chu <nelson@rivosinc.com>
25987 binutils: Updated my email address.
25989 * MAINTAINERS (RISC-V): Updated my email address.
25991 2022-08-23 GDB Administrator <gdbadmin@sourceware.org>
25993 Automatic date update in version.in
25995 2022-08-22 Tom Tromey <tromey@adacore.com>
25997 Implement target async for Windows
25998 This implements target async for Windows. The basic idea is to have
25999 the worker thread block in WaitForDebugEvent, then notify the event
26000 loop when an event is seen. In a few situations, this blocking
26001 behavior is undesirable, so the functions passed to do_synchronously
26002 are changed to return a boolean indicating which behavior is needed.
26004 2022-08-22 Tom Tromey <tromey@adacore.com>
26006 Move some Windows operations to worker thread
26007 On Windows, certain debugging APIs can only be called from the thread
26008 that started (or attached) to the inferior. Also, there is no way on
26009 Windows to wait for a debug event in addition to other events.
26010 Therefore, in order to implement target async for Windows, gdb will
26011 have to call some functions in a worker thread.
26013 This patch implements the worker thread and moves the necessary
26014 operations there. Target async isn't yet implemented, so this patch
26015 does not cause any visible changes.
26017 2022-08-22 Tom Tromey <tromey@adacore.com>
26019 Avoid crash with Ravenscar tasks
26020 When using Ravenscar, gdb can crash if the user sets a breakpoint very
26021 early in task startup. This happens because gdb thinks the runtime is
26022 initialized, but in practice the particular task isn't sufficiently
26023 initialized. This patch avoids the issue by turning an assertion into
26026 I tested this using the AdaCore internal test suite. I don't know how
26027 to test Ravenscar using the FSF test suite.
26029 2022-08-22 Nick Clifton <nickc@redhat.com>
26031 Fix compile time warning from Clang about error messages not being printed safely.
26033 Have readelf warn users if it is asked to decode a LLVM bitcode file or a golang object file.
26034 * readelf.c (check_magic_number): New function. Checks the magic
26035 bytes at the start of a file. If they are not the ELF format
26036 magic values, then attempts to generate a helpful error message.
26037 (process_file_header): Call check_magic_number.
26039 2022-08-22 Frederic Cambus <fred@statdns.com>
26041 Add OpenBSD AArch64 Little Endian BFD support.
26042 * config.bfd (aarch64-*-openbsd*): Add target.
26044 2022-08-22 tangxiaolin <tangxiaolin@loongson.cn>
26046 LoongArch: gas: add support using constant variable in instructions.
26047 Instructions that can load immediate support using constant
26048 variable like ".equ var, 123 li.w/d resgister, var".
26051 * config/loongarch-parse.y
26052 * config/tc-loongarch.c
26054 Add four testcases.One is a program using constant variable,
26055 one test using label is unsupported, and another two test
26056 almost instructions that can load immediate.
26059 * testsuite/gas/loongarch/li.d
26060 * testsuite/gas/loongarch/li.s
26061 * testsuite/gas/loongarch/imm_ins_label-fail.d
26062 * testsuite/gas/loongarch/imm_ins_label-fail.l
26063 * testsuite/gas/loongarch/imm_ins_label-fail.s
26064 * testsuite/gas/loongarch/imm_ins.d
26065 * testsuite/gas/loongarch/imm_ins.s
26066 * testsuite/gas/loongarch/imm_ins_32.d
26067 * testsuite/gas/loongarch/imm_ins_32.s
26069 2022-08-22 GDB Administrator <gdbadmin@sourceware.org>
26071 Automatic date update in version.in
26073 2022-08-21 Tom Tromey <tom@tromey.com>
26075 Fix crash in gdbpy_parse_register_id
26076 I noticed that gdbpy_parse_register_id would assert if passed a Python
26077 object of a type it was not expecting. The included test case shows
26078 this crash. This patch fixes the problem and also changes
26079 gdbpy_parse_register_id to be more "Python-like" -- it always ensures
26080 the Python error is set when it fails, and the callers now simply
26081 propagate the existing exception.
26083 2022-08-21 GDB Administrator <gdbadmin@sourceware.org>
26085 Automatic date update in version.in
26087 2022-08-20 Alan Modra <amodra@gmail.com>
26089 symbols for bfd_simple_get_relocated_section_contents
26090 If symbols are provided by the caller of this function they are
26091 passed on to bfd_get_relocated_section_contents. No surprises there.
26092 It gets a little weird if they are not provided. In that case they
26093 are read from the bfd by _bfd_generic_link_add_symbols, and global
26094 symbols are added to the generic linker hash table. Global symbols
26095 are not added to the linker hash table if symbols *are* provided. Now
26096 the linker hash table symbols are not used by the generic
26097 bfd_get_relocated_section_conents, and also not by most target
26098 versions when called from bfd_simple_get_relocated_section_contents
26099 except for symbols like "_gp". So it mostly doesn't matter whether
26100 symbols are in the linker hash table, but it's odd that there is a
26101 difference. We could always add them, but I'm inclined to think that
26102 is unnecessary work so this patch always leaves them out.
26104 Also, symbols are canonicalized and written into a malloc'd buffer.
26105 The buffer isn't freed, see commit 8e16317ca5eb. I don't know whether
26106 that matters any more, but in any case I can't see why we need another
26107 copy of the symbols when _bfd_generic_link_read_symbols has already
26110 * simple.c (bfd_simple_get_relocated_section_contents): If not
26111 provided, read symbols via bfd_generic_link_read_symbols. Do
26112 not create another copy of symbols. Tidy failure exits.
26113 Minor tidy of bfd_get_relocated_section_contents and
26114 bfd_get_full_section_contents arguments.
26116 2022-08-20 Alan Modra <amodra@gmail.com>
26118 Re: Missing linking test case for pe dll using a def file
26119 Fixes this when cross-compiling from x86_64-linux
26120 x86_64-w64-mingw32 +FAIL: compiling shared lib fastcall/stdcall
26122 * testsuite/ld-pe/pe-run2-def.exp (test_direct2_link_dll_def):
26123 Use CC_FOR_TARGET and CFLAGS_FOR_TARGET rather than CC and CFLAGS.
26125 2022-08-20 GDB Administrator <gdbadmin@sourceware.org>
26127 Automatic date update in version.in
26129 2022-08-19 Patrick Monnerat <patrick@monnerat.net>
26131 gdb_do_one_event: use integer test syntax
26132 Timeout is an int, not a bool.
26134 2022-08-19 Tom Tromey <tom@tromey.com>
26136 Remove two initialization functions
26137 I noticed a couple of initialization functions that aren't really
26138 needed, and that currently require explicit calls in gdb_init. This
26139 patch removes these functions, simplifying gdb a little.
26141 Regression tested on x86-64 Fedora 34.
26143 2022-08-19 Simon Marchi <simon.marchi@efficios.com>
26145 gdb/testsuite: re-compile entry-value-typedef .S files with -fPIE
26146 As Luis pointed out here [1], the AArch64 variant of the test doesn't
26147 work on systems that use PIE by default. For example, on this Debian
26150 $ make check TESTS="gdb.dwarf2/entry-value-typedef.exp"
26151 gdb compile failed, /usr/bin/ld: /tmp/ccJE8ZSr.o: relocation R_AARCH64_ADR_PREL_PG_HI21 against symbol `_ZNSsD1Ev@@GLIBCXX_3.4' which may bind externally can not be used when making a shared object; recompile with -fPIC
26152 /usr/bin/ld: /tmp/ccJE8ZSr.o(.text+0x38): unresolvable R_AARCH64_ADR_PREL_PG_HI21 relocation against symbol `_ZNSsD1Ev@@GLIBCXX_3.4'
26154 This is because entry-value-typedef-aarch64.S was generated on an old
26155 system that does not generate position-independent code by default, but
26156 the system the test runs on tries to link the test executable as
26157 position-independent. Fix this by regenerating the same binary on the
26158 same system as the original one, but with -fPIE this time. Do the same
26159 for the amd64 binary, although this one was already position-independent
26160 so the generated code doesn't change.
26162 With this patch applied, the test passes on the Debian 11 AArch64
26165 [1] https://sourceware.org/pipermail/gdb-patches/2022-August/191462.html
26167 Change-Id: I68d55adaa56a7a3eddb0c13980b1a98b791f8144
26169 2022-08-19 Felix Willgerodt <felix.willgerodt@intel.com>
26171 gdb, testsuite: Adapt gdb.base/callfuncs.exp for new clang warning.
26172 Clang 15.0.0 enabled the warning for deprecated non-prototype functions
26173 by default: https://reviews.llvm.org/D122895
26174 Callfuncs.exp is impacted and won't run due to new warnings:
26176 callfuncs.c:339:5: warning: a function declaration without a prototype is
26177 deprecated in all versions of C and is not supported in C2x
26178 [-Wdeprecated-non-prototype]
26179 int t_float_values (float_arg1, float_arg2)
26181 This patch disables those warnings with -Wno-deprecated-non-prototype.
26182 Removing the test for deprecated syntax would also be an option. But I will
26183 leave that up for others to decide/implement.
26185 2022-08-19 Felix Willgerodt <felix.willgerodt@intel.com>
26187 gdb, testsuite: Enable testcases that suppress specific warnings, for icc/icx.
26189 Some C/C++ testcases unconditionally pass -Wno-foo as additional
26190 options to disable some warning. That is OK with GCC, because
26191 by design, GCC accepts any -Wno-foo option, even if it doesn't
26192 support -Wfoo. Clang however warns about unknown -Wno-foo by
26193 default, unless you pass -Wno-unknown-warning-option as well.
26194 We do that here, so that individual testcases don't have to
26197 This patch adds the same option that already exists for clang for icx and
26198 adds the equivalent icc option.
26200 2022-08-19 Tiezhu Yang <yangtiezhu@loongson.cn>
26202 gdb: LoongArch: Handle variadic arguments
26203 According to LoongArch ELF ABI specification [1], variadic arguments
26204 are passed in GARs in the same manner as named arguments. And after
26205 a variadic argument has been passed on the stack, all future arguments
26206 will also be passed on the stack, i.e., the last argument register may
26207 be left unused due to the aligned register pair rule. long double data
26208 tpye is passed in an aligned GAR pair, the first register in the pair
26211 [1] https://loongson.github.io/LoongArch-Documentation/LoongArch-ELF-ABI-EN.html
26213 2022-08-19 Alan Modra <amodra@gmail.com>
26215 loongarch64_pei_vec garbage in objcopy'd relocs
26216 Like commit a9c09a3667cc, but for loongarch64.
26218 * coff-loongarch64.c (SWAP_IN_RELOC_OFFSET): Define.
26219 (SWAP_OUT_RELOC_OFFSET): Define.
26221 2022-08-19 GDB Administrator <gdbadmin@sourceware.org>
26223 Automatic date update in version.in
26225 2022-08-18 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
26227 gprofng: fix bug 29479 Collection fails when built without java support
26229 2022-08-17 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
26232 * libcollector/collector.c: Add #if defined(GPROFNG_JAVA_PROFILING) for
26233 java specific code.
26234 * libcollector/unwind.c: Likewise.
26236 2022-08-18 Simon Marchi <simon.marchi@polymtl.ca>
26238 gdb: call check_typedef at beginning of dwarf_expr_context::fetch_result
26239 Bug 29374 shows this crash:
26241 $ ./gdb -nx --data-directory=data-directory -q -batch -ex "catch throw" -ex r -ex bt a.out
26243 /home/simark/src/binutils-gdb/gdb/../gdbsupport/array-view.h:217: internal-error: copy: Assertion `dest.size () == src.size ()' failed.
26247 #0 internal_error (file=0x5555606504c0 "/home/simark/src/binutils-gdb/gdb/../gdbsupport/array-view.h", line=217, fmt=0x55556064b700 "%s: Assertion `%s' failed.") at /home/simark/src/binutils-gdb/gdbsupport/errors.cc:51
26248 #1 0x000055555d41c0bb in gdb::copy<unsigned char const, unsigned char> (src=..., dest=...) at /home/simark/src/binutils-gdb/gdb/../gdbsupport/array-view.h:217
26249 #2 0x000055555deef28c in dwarf_expr_context::fetch_result (this=0x7fffffffb830, type=0x621007a86830, subobj_type=0x621007a86830, subobj_offset=0, as_lval=false) at /home/simark/src/binutils-gdb/gdb/dwarf2/expr.c:1040
26250 #3 0x000055555def0015 in dwarf_expr_context::evaluate (this=0x7fffffffb830, addr=0x62f00004313e "0", len=1, as_lval=false, per_cu=0x60b000069550, frame=0x621007c9e910, addr_info=0x0, type=0x621007a86830, subobj_type=0x621007a86830, subobj_offset=0) at /home/simark/src/binutils-gdb/gdb/dwarf2/expr.c:1091
26251 #4 0x000055555e084327 in dwarf2_evaluate_loc_desc_full (type=0x621007a86830, frame=0x621007c9e910, data=0x62f00004313e "0", size=1, per_cu=0x60b000069550, per_objfile=0x613000006080, subobj_type=0x621007a86830, subobj_byte_offset=0, as_lval=false) at /home/simark/src/binutils-gdb/gdb/dwarf2/loc.c:1485
26252 #5 0x000055555e0849e2 in dwarf2_evaluate_loc_desc (type=0x621007a86830, frame=0x621007c9e910, data=0x62f00004313e "0", size=1, per_cu=0x60b000069550, per_objfile=0x613000006080, as_lval=false) at /home/simark/src/binutils-gdb/gdb/dwarf2/loc.c:1529
26253 #6 0x000055555e0828c6 in dwarf_entry_parameter_to_value (parameter=0x621007a96e58, deref_size=0x0, type=0x621007a86830, caller_frame=0x621007c9e910, per_cu=0x60b000069550, per_objfile=0x613000006080) at /home/simark/src/binutils-gdb/gdb/dwarf2/loc.c:1235
26254 #7 0x000055555e082f55 in value_of_dwarf_reg_entry (type=0x621007a86890, frame=0x621007acc510, kind=CALL_SITE_PARAMETER_DWARF_REG, kind_u=...) at /home/simark/src/binutils-gdb/gdb/dwarf2/loc.c:1332
26255 #8 0x000055555e083449 in value_of_dwarf_block_entry (type=0x621007a86890, frame=0x621007acc510, block=0x61e000033568 "T\004\205\001\240\004\004\243\001T\237\004\240\004\261\004\001T\004\261\004\304\005\004\243\001T\237\004\304\005\310\005\001T\004\310\005\311\005\004\243\001T\237", block_len=1) at /home/simark/src/binutils-gdb/gdb/dwarf2/loc.c:1365
26256 #9 0x000055555e094d40 in loclist_read_variable_at_entry (symbol=0x621007a99bd0, frame=0x621007acc510) at /home/simark/src/binutils-gdb/gdb/dwarf2/loc.c:3889
26257 #10 0x000055555f5192e0 in read_frame_arg (fp_opts=..., sym=0x621007a99bd0, frame=0x621007acc510, argp=0x7fffffffbf20, entryargp=0x7fffffffbf60) at /home/simark/src/binutils-gdb/gdb/stack.c:559
26258 #11 0x000055555f51c352 in print_frame_args (fp_opts=..., func=0x621007a99ad0, frame=0x621007acc510, num=-1, stream=0x6030000bad90) at /home/simark/src/binutils-gdb/gdb/stack.c:887
26259 #12 0x000055555f521919 in print_frame (fp_opts=..., frame=0x621007acc510, print_level=1, print_what=LOCATION, print_args=1, sal=...) at /home/simark/src/binutils-gdb/gdb/stack.c:1390
26260 #13 0x000055555f51f22e in print_frame_info (fp_opts=..., frame=0x621007acc510, print_level=1, print_what=LOCATION, print_args=1, set_current_sal=0) at /home/simark/src/binutils-gdb/gdb/stack.c:1116
26261 #14 0x000055555f526c6d in backtrace_command_1 (fp_opts=..., bt_opts=..., count_exp=0x0, from_tty=0) at /home/simark/src/binutils-gdb/gdb/stack.c:2079
26262 #15 0x000055555f527ae5 in backtrace_command (arg=0x0, from_tty=0) at /home/simark/src/binutils-gdb/gdb/stack.c:2198
26264 The problem is that the type that gets passed down to
26265 dwarf_expr_context::fetch_result (the type of a variable of which we're
26266 trying to read the entry value) is a typedef whose size has never been
26267 computed yet (check_typedef has never been called on it). As we get in
26268 the DWARF_VALUE_STACK case (line 1028 of dwarf2/expr.c), the `len`
26269 variable is therefore set to 0, instead of the actual type length. We
26270 then call allocate_value on subobj_type, which does call check_typedef,
26271 so the length of the typedef gets filled in at that point. We end up
26272 passing to the copy function a source array view of length 0 and a
26273 target array view of length 4, and the assertion fails.
26275 Fix this by calling check_typedef on both type and subobj_type at the
26276 beginning of fetch_result.
26278 I tried writing a test for this using the DWARF assembler, but I haven't
26279 succeeded. It's possible that we need to get into this specific code
26280 path (value_of_dwarf_reg_entry and all) to manage to get to
26281 dwarf_expr_context::fetch_result with a typedef type that has never been
26282 resolved. In all my attempts, the typedef would always be resolved
26283 already, so the bug wouldn't show up.
26285 As a fallback, I made a gdb.dwarf2 test with compiler-generated .S
26286 files. I don't particularly like those, but I think it's better than no
26287 test. The .cpp source code is the smallest reproducer I am able to make
26288 from the reproducer given in the bug (thanks to Pedro for suggestions on
26289 how to minimize it further than I had). Since I tested on both amd64
26290 and aarch64, I added versions of the test for these two architectures.
26292 Change-Id: I182733ad08e34df40d8bcc47af72c482fabf4900
26293 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29374
26295 2022-08-18 Luis Machado <luis.machado@arm.com>
26297 [aarch64] Remove handling of ADR/ADRP from prologue analyzer
26298 As reported by Tom in https://sourceware.org/pipermail/gdb-patches/2022-August/191357.html,
26299 the aarch64 prologue analyzer considers the adrp instruction in the
26300 gdb.dwarf2/dw2-dir-file-name.exp testcase to be part of a prologue.
26302 The function has no prologue though, and it only loads the volatile variable
26303 from memory. GDB should not skip any instructions in this case.
26305 Doing some archaeology, it seems handling for adr/adrp in prologues was
26306 included with the original aarch64 port. It might've been an oversight.
26308 In the particular case of gdb.dwarf2/dw2-dir-file-name.exp, the analyzer skips
26309 a couple instructions and leaves us in a nice spot where the address to the
26310 variable "v" is already in w0. But no prologues exists.
26312 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29481
26314 2022-08-18 Tom Tromey <tom@tromey.com>
26316 Change bookmark allocation
26317 This changes how bookmarks are allocated and stored, replacing a
26318 linked list with a vector and removing some ALL_* iterator macros.
26319 Regression tested on x86-64 Fedora 34.
26321 2022-08-18 Thiago Jung Bauermann <thiago.bauermann@linaro.org>
26323 Add test for AArch64 Scalable Vector Extension
26324 It exercises a bug that GDB previously had where it would lose track of
26325 some registers when the inferior changed its vector length.
26327 It also checks that the vg register and the size of the z0-z31 registers
26328 correctly reflect the new vector length.
26330 2022-08-18 Thiago Jung Bauermann <thiago.bauermann@linaro.org>
26332 Fix thread's gdbarch when SVE vector length changes
26333 When the inferior program changes the SVE length, GDB can stop tracking
26334 some registers as it obtains the new gdbarch that corresponds to the
26337 Breakpoint 1, do_sve_ioctl_test () at sve-ioctls.c:44
26338 44 res = prctl(PR_SVE_SET_VL, i, 0, 0, 0, 0);
26341 (gdb) info registers
26343 [ snip registers x0 to x30 ]
26345 sp 0xffffffffeff0 0xffffffffeff0
26346 pc 0xaaaaaaaaa8ac 0xaaaaaaaaa8ac <do_sve_ioctl_test+112>
26347 cpsr 0x60000000 [ EL=0 BTYPE=0 C Z ]
26351 tpidr 0xfffff7fcb320 0xfffff7fcb320
26354 (gdb) info registers
26356 [ snip registers x0 to x30 ]
26358 sp 0xffffffffeff0 0xffffffffeff0
26359 pc 0xaaaaaaaaa8cc 0xaaaaaaaaa8cc <do_sve_ioctl_test+144>
26360 cpsr 0x200000 [ EL=0 BTYPE=0 SS ]
26366 Notice that register tpidr disappeared when vg (which holds the vector
26367 length) changed from 8 to 4. The tpidr register is provided by the
26368 org.gnu.gdb.aarch64.tls feature.
26370 This happens because the code that searches for a new gdbarch to match the
26371 new vector length in aarch64_linux_nat_target::thread_architecture doesn't
26372 take into account the features present in the target description associated
26373 with the previous gdbarch. This patch makes it do that.
26375 Since the id member of struct gdbarch_info is now unused, it's removed.
26377 2022-08-18 Ralf Habacker <ralf.habacker@freenet.de>
26379 Missing linking test case for pe dll using a def file.
26381 * testsuite/ld-pe/pe-run2-def.exp: New file.
26383 2022-08-18 Patrick Monnerat <patrick@monnerat.net>
26385 gdbsupport/event-loop: add a timeout parameter to gdb_do_one_event
26386 Since commit b2d8657, having a per-interpreter event/command loop is not
26389 As Insight uses a GUI that has its own event loop, gdb and GUI event
26390 loops have then to be "merged" (i.e.: work together). But this is
26391 problematic as gdb_do_one_event is not aware of this alternate event
26392 loop and thus may wait forever.
26394 A solution is to delegate GUI events handling to the gdb events handler.
26395 Insight uses Tck/Tk as GUI and the latter offers a "notifier" feature to
26396 implement such a delegation. The Tcl notifier spec requires the event wait
26397 function to support a timeout parameter. Unfortunately gdb_do_one_event
26398 does not feature such a parameter.
26399 This timeout cannot be implemented externally with a gdb timer, because
26400 it would become an event by itself and thus can cause a legitimate event to
26401 be missed if the timeout is 0.
26402 Tcl implements "idle events" that are (internally) triggered only when no
26403 other event is pending. For this reason, it can call the event wait function
26404 with a 0 timeout quite often.
26406 This patch implements a wait timeout to gdb_do_one_event. The initial
26407 pending events monitoring is performed as before without the possibility
26408 to enter a wait state. If no pending event has been found during this
26409 phase, a timer is then created for the given timeout in order to re-use
26410 the implemented timeout logic and the event wait is then performed.
26411 This "internal" timer only limits the wait time and should never be triggered.
26412 It is deleted upon gdb_do_one_event exit.
26414 The new parameter defaults to "no timeout" (-1): as it is used by Insight
26415 only, there is no need to update calls from the gdb source tree.
26417 2022-08-18 Patrick Monnerat <patrick@monnerat.net>
26419 gdb: add Patrick Monnerat to gdb/MAINTAINERS
26421 2022-08-18 Jan Beulich <jbeulich@suse.com>
26423 x86: move / quiesce pre-386 non-16-bit warning
26424 Emitting this warning for every insn, including ones having actual
26425 errors, is annoying. Introduce a boolean variable to emit the warning
26426 just once on the first insn after .arch may have changed the things, and
26427 move the warning to output_insn(). (I didn't want to go as far as
26428 checking whether the .arch actually turned off the i386 bit, but doing
26429 so would be an option.)
26431 x86: insert "no error" enumerator in i386_error enumeration
26432 The value of zero would better not indicate any error, but rather hit
26433 the abort() at the top of the consuming switch().
26435 2022-08-18 GDB Administrator <gdbadmin@sourceware.org>
26437 Automatic date update in version.in
26439 2022-08-17 Maciej W. Rozycki <macro@embecosm.com>
26441 GDB/testsuite: Fix PARAM_ZUINTEGER reported for PARAM_ZUINTEGER_UNLIMITED
26442 Correctly report PARAM_ZUINTEGER_UNLIMITED rather than PARAM_ZUINTEGER
26443 in testing a Python parameter of the PARAM_ZUINTEGER_UNLIMITED type.
26445 2022-08-17 Alan Modra <amodra@gmail.com>
26447 bfd_elf_set_group_contents assertion
26448 objcopy of broken SHT_GROUP sections shouldn't write garbage.
26450 * elf.c (bfd_elf_set_group_contents): If number of entries is
26451 unexpected, fill out section with zeros.
26453 2022-08-17 Alan Modra <amodra@gmail.com>
26455 timeout in mmo_get_symbols
26456 Fix mmo_get_byte to return a fail-safe value, not just on the first
26457 call with a read error but on subsequent calls too.
26459 * mmo.c (mmo_get_byte): Return the fail-safe value on every
26460 call after a read error.
26462 2022-08-17 Alan Modra <amodra@gmail.com>
26464 mmo.c leak in mmo_make_section
26465 * mmo.c (mmo_make_section): Alloc name using bfd_alloc. Use
26466 bfd_error_no_memory.
26467 (mmo_decide_section): Check for NULL return from mmo_make_section.
26469 2022-08-17 Alan Modra <amodra@gmail.com>
26471 asan: heap buffer overflow in mmo_scan
26472 mmo_get_loc needs to handle arbitrary vma and size chunks. Fuzzers
26473 found that it wasn't working so well when the end of chunks were
26474 getting close to address wrap-around.
26476 * mmo.c (mmo_get_loc): Make "size" unsigned. Avoid arithmetic
26477 overflow when calculating whether range hits an existing chunk.
26479 2022-08-17 Alan Modra <amodra@gmail.com>
26482 Swap params of is_note, so they are section, segment like others used
26483 in rewrite_elf_program_header. Whitespace fixes, plus wrapping of
26486 2022-08-17 GDB Administrator <gdbadmin@sourceware.org>
26488 Automatic date update in version.in
26490 2022-08-16 Torbjörn SVENSSON <torbjorn.svensson@foss.st.com>
26492 bfd: Define ___lc_codepage_func prototype for older MinGW-w64
26493 In commit 68e80d96a84282d547f3b3c1234c99009521630c, the usage of
26494 ___lc_codepage_func was introduced to determine the current encoding.
26496 Prior to version 9.0 of MinGW-w64, the function prototype for
26497 ___lc_codepage_func was missing and trying to build BFD caused the
26500 error: implicit declaration of function ‘___lc_codepage_func’
26502 This changeset adds a conditonal definition of
26503 ___lc_codepage_func to allow a sucessful build with MinGW-w64.
26505 2022-08-16 H.J. Lu <hjl.tools@gmail.com>
26507 i386: Add MAX_OPERAND_BUFFER_SIZE
26508 When displaying operands, invalid opcodes may overflow operand buffer
26509 due to additional styling characters. Each style is encoded with 3
26510 bytes. Define MAX_OPERAND_BUFFER_SIZE for operand buffer size and
26511 increase it from 100 bytes to 128 bytes to accommodate 9 sets of styles
26517 * testsuite/gas/i386/i386.exp: Run pr29483.
26518 * testsuite/gas/i386/pr29483.d: New file.
26519 * testsuite/gas/i386/pr29483.s: Likewise.
26524 * i386-dis.c (MAX_OPERAND_BUFFER_SIZE): New.
26525 (obuf): Replace 100 with MAX_OPERAND_BUFFER_SIZE.
26526 (staging_area): Likewise.
26527 (op_out): Likewise.
26529 2022-08-16 Andrew Burgess <aburgess@redhat.com>
26531 gdb/riscv: fix gdb.arch/riscv-unwind-long-insn.exp on RV64
26532 I noticed that the gdb.arch/riscv-unwind-long-insn.exp test was
26533 failing when run on a 64-bit RISC-V target.
26535 The problem was that GDB was failing to stop after a finish command,
26536 and was then running to an unexpected location.
26538 The reason GDB failed to stop at the finish breakpoint was that the
26539 frame-id of the inferior, when we reached the finish breakpoint,
26540 didn't match the expected frame-id that was stored on the breakpoint.
26542 The reason for this mismatch was that the assembler code that is
26543 included in this test, was written only taking 32-bit RISC-V into
26544 account, as a result, the $fp register was being corrupted, and this
26545 was causing the frame-id mismatch.
26547 Specifically, the $fp register would end up being sign-extended from
26548 32 to 64 bits. If the expected $fp value has some significant bits
26549 above bit 31 then the computed and expected frame-ids would not match.
26551 To fix this I propose merging the two .s files into a single .S file,
26552 and making use of preprocessor macros to specialise the file for the
26553 correct size of $fp. There are plenty of existing tests that already
26554 make use of preprocessor macros in assembler files, so I assume this
26557 Once I'd decided to make use of preprocessor macros to solve the 32/64
26558 bit issue, then I figured I might as well merge the two test assembler
26559 files, they only differed by a single instruction.
26561 With this change in place I now see this test fully passing on 32 and
26562 64 bit RISC-V targets.
26564 2022-08-16 Simon Marchi <simon.marchi@polymtl.ca>
26566 gdb/testsuite: fix breakpoint script output in gdb.mi/mi-break.exp
26567 Commit 9db0d8536dbc ("gdb/mi: fix breakpoint script field output") fixed
26568 the output of the script key in the MI breakpoint output, from
26570 script={"print 10","continue"}
26574 script=["print 10","continue"]
26576 However, it missed updating this test case, which still tests for the
26577 old (broken) form, causing:
26579 FAIL: gdb.mi/mi-break.exp: mi-mode=main: test_breakpoint_commands: breakpoint commands: check that commands are set (unexpected output)
26580 FAIL: gdb.mi/mi-break.exp: mi-mode=separate: test_breakpoint_commands: breakpoint commands: check that commands are set (unexpected output)
26582 Update the test to expect the new form.
26584 Change-Id: I174919d4eea53e96d914ca9bd1cf6f01c8de30b8
26586 2022-08-16 Tom Tromey <tromey@adacore.com>
26588 Use strwinerror in gdb/windows-nat.c
26589 When working on windows-nat.c, it's useful to see an error message in
26590 addition to the error number given by GetLastError. This patch moves
26591 strwinerror from gdbserver to gdbsupport, and then updates
26592 windows-nat.c to use it. A couple of minor changes to strwinerror
26593 (constify the return type and use the ARRAY_SIZE macro) are also
26596 2022-08-16 Tom Tromey <tom@tromey.com>
26598 Remove register_gdbarch_init
26599 This removes the deprecated register_gdbarch_init in favor a default
26600 argument to gdbarch_register. Regression tested on x86-64 Fedora 34.
26602 2022-08-16 Alan Modra <amodra@gmail.com>
26604 PR29495, rewrite_elf_program_header looping
26605 This patch, in order of significance:
26606 1) Replaces some macros with inline functions.
26607 2) Those inline functions catch and avoid arithmetic overflows when
26608 comparing addresses.
26609 3) When assigning sections to segments (IS_SECTION_IN_INPUT_SEGMENT)
26610 use bed->want_p_paddr_set_to_zero to decide whether lma vs p_paddr
26611 or vma vs p_vaddr should be tested. When remapping, use the same
26612 test, and use is_note rather than the more restrictive
26615 It's important that the later tests not be more restrictive. If they
26616 are it can lead to the situation triggered by the testcases, where a
26617 section seemingly didn't fit and thus needed a new mapping. It didn't
26618 fit the new mapping either, and this repeated until memory exhausted.
26621 * elf.c (SEGMENT_END, SECTION_SIZE, IS_CONTAINED_BY_VMA): Delete.
26622 (IS_CONTAINED_BY_LMA, IS_NOTE, IS_COREFILE_NOTE): Delete.
26623 (segment_size, segment_end, section_size): New inline function.
26624 (is_contained_by, is_note): Likewise.
26625 (rewrite_elf_program_header): Use new functions.
26627 2022-08-16 Jan Beulich <jbeulich@suse.com>
26629 x86: shorten certain template names
26630 Now that we can purge templates, let's use this to improve readability a
26631 little by shortening a few of their names, making functionally similar
26632 ones also have identical names in their multiple incarnations.
26634 2022-08-16 Jan Beulich <jbeulich@suse.com>
26636 x86: template-ize certain vector conversion insns
26637 Many of the vector conversion insns come with X/Y/Z suffixed forms, for
26638 disambiguation purposes in AT&T syntax. All of these gorups follow
26639 certain patterns. Introduce "xy" and "xyz" templates to reduce
26642 To facilitate using a uniform name for both AVX and AVX512, further
26643 introduce a means to purge a previously defined template: A standalone
26644 <name> will be recognized to have this effect.
26646 Note that in the course of the conversion VFPCLASSPH is properly split
26647 to separate AT&T and Intel syntax forms, matching VFPCLASSP{S,D} and
26648 yielding the intended "ambiguous operand size" diagnostic in Intel mode.
26650 2022-08-16 Jan Beulich <jbeulich@suse.com>
26652 x86: template-ize vector packed byte/word integer insns
26653 Many of the vector integer insns come in byte/word element pairs. Most
26654 of these pairs follow certain encoding patterns. Introduce a "bw"
26655 template to reduce redundancy.
26657 Note that in the course of the conversion
26658 - the AVX VPEXTRW template which is not being touched needs to remain
26659 ahead of the new "combined" ones, as (a) this should be tried first
26660 when matching insns against templates and (b) its Load attributes
26661 requires it to be first,
26662 - this add a benign/meaningless IgnoreSize attribute to the memory form
26663 of PEXTRB; it didn't seem worth avoiding this.
26665 2022-08-16 Jan Beulich <jbeulich@suse.com>
26667 x86: re-order AVX512 S/G templates
26668 The AVX2 gather ones are nicely grouped - do the same for the various
26669 AVX512 scatter/gather ones. On the moved lines also convert EVex=<n> to
26672 2022-08-16 Jan Beulich <jbeulich@suse.com>
26674 x86: template-ize vector packed dword/qword integer insns
26675 Many of the vector integer insns come in dword/qword element pairs. Most
26676 of these pairs follow certain encoding patterns. Introduce a "dq"
26677 template to reduce redundancy.
26679 Note that in the course of the conversion
26680 - a few otherwise untouched templates are moved, so they end up next to
26682 - drop an unhelpful Cpu64 from the GPR form of VPBROADCASTQ, matching
26683 what we already have for KMOVQ - the diagnostic is better this way for
26684 insns with multiple forms (i.e. the same Cpu64 attributes on {,V}MOVQ,
26685 {,V}PEXTRQ, and {,V}PINSRQ are useful to keep),
26686 - this adds benign/meaningless IgnoreSize attributes to the GPR forms of
26687 KMOVD and VPBROADCASTD; it didn't seem worth avoiding this.
26689 2022-08-16 Jan Beulich <jbeulich@suse.com>
26691 x86: template-ize packed/scalar vector floating point insns
26692 The vast majority of vector FP insns comes in single/double pairs. Many
26693 pairs follow certain encoding patterns. Introduce an "sd" template to
26694 reduce redundancy. Similarly, to further cover similarities between
26695 AVX512F and AVX512-FP16, introduce an "sdh" template.
26697 For element-size Disp8 shift generalize i386-gen's broadcast size
26698 determination, allowing Disp8MemShift to be specified without an operand
26699 in the affected templated templates. While doing the adjustment also
26700 eliminate an unhelpful (lost information) diagnostic combined with a use
26701 after free in what is now get_element_size().
26703 Note that in the course of the conversion
26704 - the AVX512F form of VMOVUPD has a stray (leftover) Load attribute
26706 - VMOVSH has a benign IgnoreSize added (the attribute is still strictly
26707 necessary for VMOVSD, and necessary for VMOVSS as long as we permit
26708 strange combinations like "-march=i286+avx"),
26709 - VFPCLASSPH is properly split to separate AT&T and Intel syntax forms,
26710 matching VFPCLASSP{S,D}.
26712 2022-08-16 Jan Beulich <jbeulich@suse.com>
26714 revert "x86: Also pass -P to $(CPP) when processing i386-opc.tbl"
26715 This reverts commit 384f368958f2a5bb083660e58e5f8a010e6ad429, which
26716 broke i386-gen's emitting of diagnostics. As a replacement to address
26717 the original issue of newer gcc no longer splicing lines when dropping
26718 the line continuation backslashes, switch to using + as the line
26719 continuation character, doing the line splicing in i386-gen.
26721 2022-08-16 GDB Administrator <gdbadmin@sourceware.org>
26723 Automatic date update in version.in
26725 2022-08-15 Alan Modra <amodra@gmail.com>
26727 PR29362, some binutils memory leaks
26728 2022-08-16 Alan Modra <amodra@gmail.com>
26729 Cunlong Li <shenxiaogll@163.com>
26732 * dwarf.c (free_debug_information): New function, extracted..
26733 (free_debug_memory): ..from here.
26734 (process_debug_info): Use it when before clearing out unit
26735 debug_information. Clear all fields.
26736 * objcopy.c (delete_symbol_htabs): New function.
26737 (main): Call it via xatexit.
26738 (copy_archive): Free "dir".
26739 * objdump.c (free_debug_section): Free reloc_info.
26741 2022-08-15 Jiangshuai Li <jiangshuai_li@linux.alibaba.com>
26743 gdb/csky add unwinder for sigtramp frame when kernel 4.x and later
26744 When kernel veriosn >= V4.x, the characteristic values used to
26745 determine whether it is a signal function call are:
26749 Registers are saved at (sp + CSKY_SIGINFO_OFFSET + CSKY_SIGINFO_SIZE
26750 + CSKY_UCONTEXT_SIGCONTEXT + CSKY_SIGCONTEXT_PT_REGS_TLS). The order
26751 is described in csky_linux_rt_sigreturn_init_pt_regs.
26753 2022-08-15 Alan Modra <amodra@gmail.com>
26756 I know this target is just a skeleton, but let's not write out relocs
26757 with uninitialised garbage.
26759 * coff-aarch64.c (SWAP_IN_RELOC_OFFSET): Define.
26760 (SWAP_OUT_RELOC_OFFSET): Define.
26762 2022-08-15 GDB Administrator <gdbadmin@sourceware.org>
26764 Automatic date update in version.in
26766 2022-08-14 Andrew Burgess <aburgess@redhat.com>
26768 gdb/riscv: improve a comment about fcsr, fflags, and frm registers
26769 There's a comment in riscv-tdep.c that explains some of the background
26770 about how we check for the fcsr, fflags, and frm registers within a
26771 riscv target description.
26773 This comment (and the functionality it describes) relates to how QEMU
26774 advertises these registers within its target description.
26776 Unfortunately, QEMU includes these three registers in both the fpu and
26777 crs target description features. To work around this GDB uses one of
26778 the register declarations, and ignores the other, this means the GDB
26779 user sees a single copy of each register, and things just work.
26781 When I originally wrote the comment I thought it didn't matter which
26782 copy of the register GDB selected, the fpu copy or the csr copy, so
26783 long as we just used one of them. The comment reflected this belief.
26785 Upon further investigation, it turns out I was wrong. GDB has to use
26786 the csr copy of the register. If GDB tries to use the register from
26787 the fpu feature then QEMU will return an error when GDB tries to read
26788 or write the register.
26790 Luckily, the code within GDB (currently) will always select the csr
26791 copy of the register, so nothing is broken, but the comment is wrong.
26792 This commit updates the comment to better describe what is actually
26795 Of course, I should probably also send a patch to QEMU to fix up the
26796 target description that is sent to GDB.
26798 2022-08-14 Andrew Burgess <aburgess@redhat.com>
26800 gdb/nds32: update features/nds32.c
26803 commit 7b7c365c5c663ffdfb2b3f696db35c23cdccd921
26804 Date: Wed Sep 15 10:10:46 2021 +0200
26806 [bfd] Ensure unique printable names for bfd archs
26808 The printable name field of the default nds32 bfd_arch_info changed
26809 from 'n1h' to 'n1'. As a consequence the generated feature file
26810 within GDB should have been recreated. Recreate it now.
26812 2022-08-14 Tom Tromey <tom@tromey.com>
26814 Move decode_location_spec to code_breakpoint
26815 breakpoint::decode_location_spec just asserts if called. It turned
26816 out to be relatively easy to remove this method from breakpoint and
26817 instead move the base implementation to code_breakpoint.
26819 Change location_spec_to_sals to a method
26820 location_spec_to_sals is only ever called for code breakpoints, so
26821 make it a protected method there.
26823 Change breakpoint_re_set_default to a method
26824 breakpoint_re_set_default is only ever called from breakpoint re_set
26825 methods, so make it a protected method on code_breakpoint.
26827 2022-08-14 GDB Administrator <gdbadmin@sourceware.org>
26829 Automatic date update in version.in
26831 2022-08-13 Alan Modra <amodra@gmail.com>
26833 PR29482 - strip: heap-buffer-overflow
26835 * coffcode.h (coff_set_section_contents): Sanity check _LIB.
26837 asan: NULL dereference in spu_elf_object_p
26838 * elf32-spu.c (spu_elf_object_p): Don't dereference NULL
26841 ubsan: undefined shift in sign_extend
26842 * libhppa.h (sign_extend): Avoid undefined behaviour.
26844 asan: NULL dereference in som_set_reloc_info
26845 * som.c (som_set_reloc_info): Ignore non-existent previous
26848 readelf: print 0x0 as 0, and remove trailing spaces
26849 This changes readelf output a little, removing the 0x prefix on hex
26850 output when the value is 0, except in cases where a fixed field
26851 width is shown. %#010x is not a good replacement for 0x%08x.
26853 Make dwarf_vma uint64_t
26854 This replaces dwarf_vma, dwarf_size_type and dwarf_signed_vma with
26855 uint64_t and int64_t everywhere. The patch also gets rid of
26856 DWARF_VMA_FMT since we can't use that with uint64_t, and all of the
26857 configure support for deciding the flavour of HOST_WIDEST_INT.
26858 dwarf_vmatoa also disappears, replacing most uses with one of
26859 PRIx64, PRId64 or PRIu64. Printing of size_t and ptrdiff_t values
26860 now use %z and %t rather than by casting to unsigned long. Also,
26861 most warning messages that used 0x%lx or similar now use %#lx and a
26862 few that didn't print the 0x hex prefix now also use %#. The patch
26863 doesn't change normal readelf output, except in odd cases where values
26864 previously might have been truncated.
26866 Don't use bfd_vma in readelf.c
26867 This replaces bfd_vma with uint64_t in readelf, defines BFD64
26868 unconditionally, removes tests of BFD64 and sizeof (bfd_vma), and
26869 removes quite a few now unnecessary casts.
26871 Don't use bfd_size_type in readelf.c and dwarf.c
26872 Replacing bfd_size_type with dwarf_size_type or uint64_t is mostly
26873 cosmetic. The point of the change is to avoid use of a BFD type
26874 in readelf, where we'd like to keep as independent of BFD as
26875 possible. Also, the patch is a step towards using standard types.
26877 Replace elf_vma with uint64_t
26878 This patch replaces all uses of elf_vma with uint64_t, removes
26879 tests of sizeof (elf_vma), and does a little tidying of
26880 byte_get_little_endian and byte_get_big_endian.
26882 2022-08-13 GDB Administrator <gdbadmin@sourceware.org>
26884 Automatic date update in version.in
26886 2022-08-12 Tom de Vries <tdevries@suse.de>
26888 [gdb/testsuite] Fix gdb.dwarf2/dw2-dir-file-name.exp
26889 When running test-case gdb.dwarf2/dw2-dir-file-name.exp on x86_64-linux, we
26892 (gdb) break compdir_missing__ldir_missing__file_basename^M
26893 Breakpoint 2 at 0x4004c4: file tmp-dw2-dir-file-name.c, line 999.^M
26897 Breakpoint 2, 0x00000000004004c4 in \
26898 compdir_missing__ldir_missing__file_basename () \
26899 at tmp-dw2-dir-file-name.c:999^M
26900 (gdb) PASS: gdb.dwarf2/dw2-dir-file-name.exp: \
26901 compdir_missing__ldir_missing__file_basename: continue to breakpoint: \
26902 compdir_missing__ldir_missing__file_basename
26905 When trying to set a breakpoint on
26906 compdir_missing__ldir_missing__file_basename, the architecture-specific
26907 prologue skipper starts at 0x4004c0 and skips past two insns, to 0x4004c4:
26909 00000000004004c0 <compdir_missing__ldir_missing__file_basename>:
26910 4004c0: 55 push %rbp
26911 4004c1: 48 89 e5 mov %rsp,%rbp
26912 4004c4: 8b 05 72 1b 20 00 mov 0x201b72(%rip),%eax # 60203c <v>
26913 4004ca: 83 c0 01 add $0x1,%eax
26914 4004cd: 89 05 69 1b 20 00 mov %eax,0x201b69(%rip) # 60203c <v>
26916 4004d4: 5d pop %rbp
26920 And because the line table info is rudamentary:
26922 CU: tmp-dw2-dir-file-name.c:
26923 File name Line number Starting address View Stmt
26924 tmp-dw2-dir-file-name.c 999 0x4004c0 x
26925 tmp-dw2-dir-file-name.c 1000 0x4004d6 x
26926 tmp-dw2-dir-file-name.c - 0x4004d6
26928 the address does not fall at an actual line, so the breakpoint is shown with
26929 address, both when setting it and hitting it.
26931 when running the test-case with aarch64-linux, we have similarly:
26933 (gdb) break compdir_missing__ldir_missing__file_basename^M
26934 Breakpoint 2 at 0x400618: file tmp-dw2-dir-file-name.c, line 999.^M
26936 due to the architecture-specific prologue skipper starting at 0x400610 and
26937 skipping past two insns, to 0x400618:
26939 0000000000400610 <compdir_missing__ldir_missing__file_basename>:
26940 400610: 90000100 adrp x0, 420000 <__libc_start_main@GLIBC_2.17>
26941 400614: 9100b000 add x0, x0, #0x2c
26942 400618: b9400000 ldr w0, [x0]
26943 40061c: 11000401 add w1, w0, #0x1
26944 400620: 90000100 adrp x0, 420000 <__libc_start_main@GLIBC_2.17>
26945 400624: 9100b000 add x0, x0, #0x2c
26946 400628: b9000001 str w1, [x0]
26947 40062c: d503201f nop
26948 400630: d65f03c0 ret
26951 But interestingly, the aarch64 architecture-specific prologue skipper is
26952 wrong. There is no prologue, and the breakpoint should be set at 0x400610.
26954 By using "break *compdir_missing__ldir_missing__file_basename"
26955 we can get the breakpoint set at 0x400610:
26957 (gdb) break *compdir_missing__ldir_missing__file_basename^M
26958 Breakpoint 2 at 0x400610: file tmp-dw2-dir-file-name.c, line 999.^M
26960 and make the test-case independent of prologue analysis.
26962 This requires us to update the expected patterns.
26964 The fix ensures that once the aarch64 architecture-specific prologue skipper
26965 will be fixed, this test-case won't start failing.
26967 Tested on x86_64-linux.
26969 2022-08-12 GDB Administrator <gdbadmin@sourceware.org>
26971 Automatic date update in version.in
26973 2022-08-11 Lancelot SIX <lancelot.six@amd.com>
26975 gdb/varobj: Only re-evaluate invalid globals during re_set
26976 When doing varobj_re_set, we currently try to recreate floating varobj.
26977 This was introduced by 4e969b4f0128 "Re-evaluate floating varobj as part
26978 of varobj_invalidate" to deal with use a after free issue. However
26979 since bc20e562ec0 "Fix use after free in varobj" we now ensure that we
26980 never have dangling pointers so this all recreation is not strictly
26981 needed anymore for floating varobjs.
26983 This commit proposes to remove this recreation process for floating
26986 Tested on x86_64-linux.
26988 2022-08-11 Tom de Vries <tdevries@suse.de>
26989 Lancelot SIX <lancelot.six@amd.com>
26991 gdb/varobj: Reset varobj after relocations have been computed
26992 [This patch is a followup to the discussion in
26993 https://sourceware.org/pipermail/gdb-patches/2022-August/191188.html]
26995 PR/29426 shows failures when running the gdb.mi/mi-var-invalidate-shlib
26996 test when using a compiler which does not produce a PIE executable by
26999 In the testcase, a varobj is created to track a global variable, and
27000 then the main binary is reloaded in GDB (using the file command).
27002 During the load of the new binary, GDB tries to recreate the varobj to
27003 track the global in the new binary (varobj_invalidate_iter). At this
27004 point, the old process is still in flight. So when we try to access to
27005 the value of the global, in a PIE executable we only have access to the
27006 unrelocated address (the objfile's text_section_offset () is 0). As a
27007 consequence down the line read_value_memory fails to read the unrelated
27008 address, so cannot evaluate the value of the global. Note that the
27009 expression used to access to the global’s value is valid, so the varobj
27010 can be created. When using a non PIE executable, the address of the
27011 global GDB knows about at this point does not need relocation, so
27012 read_value_memory can access the (old binary’s) value.
27014 So at this point, in the case of a non-PIE executable the value field is
27015 set, while it is cleared in the case of PIE executable. Later when the
27016 test issues a "-var-update global_var", the command sees no change in
27017 the case of the non-PIE executable, while in the case of the PIE
27018 executable install_new_value sees that value changes, leading to a
27021 This patch makes sure that, as we do for breakpoints, we wait until
27022 relocation has happened before we try to recreate varobjs. This way we
27023 have a consistent behavior between PIE and non-PIE binaries.
27025 Tested on x86_64-linux.
27027 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29426
27029 2022-08-11 Lancelot SIX <lancelot.six@amd.com>
27031 gdb/varobj: Do not invalidate locals in varobj_invalidate_iter
27032 The varobj_invalidate_iter function has logic to invalidate any local
27033 varobj it can find. However since bc20e562ec0 "gdb/varobj: Fix use
27034 after free in varobj" all varobj containing references to an objfile are
27035 cleared when the objfile goes out of scope. This means that at this
27036 point any local varobj seen by varobj_invalidate_iter either has
27037 already been invalidated by varobj_invalidate_if_uses_objfile or only
27038 contains valid references and there is no reason to invalidate it.
27040 This patch proposes to remove this unnecessary invalidation and adds a
27041 testcase which exercises a scenario where a local varobj can legitimately
27042 survive a call to varobj_invalidate_iter.
27044 At this point the varobj_invalidate and varobj_invalidate_iter seem
27045 misnamed since they deal with re-creating invalid objects and do not do
27046 invalidation, but this will be fixed in a following patch.
27048 Tested on x86_64-linux.
27050 2022-08-11 Dmitry Selyutin <ghostmansd@gmail.com>
27052 ppc/svp64: support svindex instruction
27053 https://libre-soc.org/openpower/sv/
27054 https://libre-soc.org/openpower/sv/remap/#svindex
27055 https://libre-soc.org/openpower/isa/simplev/
27057 ppc/svp64: support svremap instruction
27058 https://libre-soc.org/openpower/sv/
27059 https://libre-soc.org/openpower/sv/remap/#svremap
27060 https://libre-soc.org/openpower/isa/simplev/
27062 ppc/svp64: support svshape instruction
27063 https://libre-soc.org/openpower/sv/
27064 https://libre-soc.org/openpower/sv/remap/#svshape
27065 https://libre-soc.org/openpower/isa/simplev/
27067 ppc/svp64: support svstep instructions
27068 https://libre-soc.org/openpower/sv/
27069 https://libre-soc.org/openpower/sv/svstep/
27070 https://libre-soc.org/openpower/isa/simplev/
27072 ppc/svp64: support setvl instructions
27073 https://libre-soc.org/openpower/sv/
27074 https://libre-soc.org/openpower/sv/setvl/
27075 https://libre-soc.org/openpower/isa/simplev/
27077 ppc/svp64: introduce non-zero operand flag
27078 svstep and svshape instructions subtract 1 before encoding some of the
27079 operands. Obviously zero is not supported for these operands. Whilst
27080 PPC_OPERAND_PLUS1 fits perfectly to mark that maximal value should be
27081 incremented, there is no flag which marks the fact that zero values are
27082 not allowed. This patch adds a new flag, PPC_OPERAND_NONZERO, for this
27085 2022-08-11 Dmitry Selyutin <ghostmansd@gmail.com>
27087 ppc/svp64: support LibreSOC architecture
27088 This patch adds support for LibreSOC machine and SVP64 extension flag
27089 for PowerPC architecture. SV (Simple-V) is a strict RISC-paradigm
27090 Scalable Vector Extension for the Power ISA. SVP64 is the 64-bit
27091 Prefixed instruction format implementing SV. Funded by NLnet through EU
27092 Grants No: 825310 and 825322, SV is in DRAFT form and is to be publicly
27093 submitted via the OpenPOWER Foundation ISA Working Group via the
27094 newly-created External RFC Process.
27096 For more details, visit https://libre-soc.org.
27098 2022-08-11 Torbjörn SVENSSON <torbjorn.svensson@foss.st.com>
27100 [Arm] Cleanup arm_m_exception_cache
27101 With this change, only valid contents of LR are accepted when unwinding
27102 exception frames for m-profile targets.
27104 If the contents of LR are anything but EXC_RETURN or FNC_RETURN, it
27105 will cause GDB to print an error and/or abort unwinding of the frame as
27106 it's an invalid state for the unwinder.
27108 The FNC_RETURN pattern requires Security Extensions to be enabled.
27110 2022-08-11 Fangrui Song <maskray@google.com>
27112 RISC-V: Remove R_RISCV_GNU_VTINHERIT/R_RISCV_GNU_VTENTRY
27113 They were legacy relocation types copied from other ports. The related
27114 -fvtable-gc was removed from GCC in 2003.
27116 The associated assembler directives (.vtable_inherit and .vtable_entry)
27117 have never been supported by the RISC-V port. Remove related ld code.
27119 Link: https://github.com/riscv-non-isa/riscv-elf-psabi-doc/pull/323
27121 2022-08-11 Alan Modra <amodra@gmail.com>
27123 PR29466, APP/NO_APP with .linefile
27124 Commit 53f2b36a54b9 exposed a bug in sb_scrub_and_add_sb that could
27125 result in losing input. If scrubbing results in expansion past the
27126 holding capacity of do_scrub_chars output buffer, then do_scrub_chars
27127 stashes the extra input for the next call. That call never came
27128 because sb_scrub_and_add_sb wrongly decided it was done. Fix that by
27129 allowing sb_scrub_and_add_sb to see whether there is pending input.
27130 Also allow a little extra space so that in most cases we won't need
27131 to resize the output buffer.
27133 sb_scrub_and_add_sb also limited output to the size of the input,
27134 rather than the actual output buffer size. Fixing that resulted in a
27135 fail of gas/testsuite/macros/dot with an extra warning: "end of file
27136 not at end of a line; newline inserted". OK, so the macro in dot.s
27137 really does finish without end-of-line. Apparently the macro
27138 expansion code relied on do_scrub_chars returning early. So fix that
27139 too by adding a newline if needed in macro_expand_body.
27142 * app.c (do_scrub_pending): New function.
27143 * as.h: Declare it.
27144 * input-scrub.c (input_scrub_include_sb): Add extra space for
27145 two .linefile directives.
27146 * sb.c (sb_scrub_and_add_sb): Take into account pending input.
27147 Allow output to max.
27148 * macro.c (macro_expand_body): Add terminating newline.
27149 * testsuite/config/default.exp (SIZE, SIZEFLAGS): Define.
27150 * testsuite/gas/macros/app5.d,
27151 * testsuite/gas/macros/app5.s: New test.
27152 * testsuite/gas/macros/macros.exp: Run it.
27154 2022-08-11 Alan Modra <amodra@gmail.com>
27158 2022-08-11 GDB Administrator <gdbadmin@sourceware.org>
27160 Automatic date update in version.in
27162 2022-08-10 Simon Marchi <simon.marchi@polymtl.ca>
27164 gdb/mi: fix breakpoint script field output
27165 The "script" field, output whenever information about a breakpoint with
27166 commands is output, uses wrong MI syntax.
27168 $ ./gdb -nx -q --data-directory=data-directory -x script -i mi
27169 =thread-group-added,id="i1"
27170 =breakpoint-created,bkpt={number="1",type="breakpoint",disp="keep",enabled="y",addr="0x000000000000111d",func="main",file="test.c",fullname="/home/simark/build/binutils-gdb-one-target/gdb/test.c",line="3",thread-groups=["i1"],times="0",original-location="main"}
27171 =breakpoint-modified,bkpt={number="1",type="breakpoint",disp="keep",enabled="y",addr="0x000000000000111d",func="main",file="test.c",fullname="/home/simark/build/binutils-gdb-one-target/gdb/test.c",line="3",thread-groups=["i1"],times="0",script={"aaa","bbb","ccc"},original-location="main"}
27174 ^done,BreakpointTable={nr_rows="1",nr_cols="6",hdr=[{width="7",alignment="-1",col_name="number",colhdr="Num"},{width="14",alignment="-1",col_name="type",colhdr="Type"},{width="4",alignment="-1",col_name="disp",colhdr="Disp"},{width="3",alignment="-1",col_name="enabled",colhdr="Enb"},{width="18",alignment="-1",col_name="addr",colhdr="Address"},{width="40",alignment="2",col_name="what",colhdr="What"}],body=[bkpt={number="1",type="breakpoint",disp="keep",enabled="y",addr="0x000000000000111d",func="main",file="test.c",fullname="/home/simark/build/binutils-gdb-one-target/gdb/test.c",line="3",thread-groups=["i1"],times="0",script={"aaa","bbb","ccc"},original-location="main"}]}
27177 In both the =breakpoint-modified and -break-info output, we have:
27179 script={"aaa","bbb","ccc"}
27181 According to the output syntax [1], curly braces means tuple, and a
27182 tuple contains key=value pairs. This looks like it should be a list,
27183 but uses curly braces by mistake. This would make more sense:
27185 script=["aaa","bbb","ccc"]
27187 Fix it, keeping the backwards compatibility by introducing a new MI
27188 version (MI4), in exactly the same way as was done when fixing
27189 multi-locations breakpoint output in [2].
27191 - Add a fix_breakpoint_script_output uiout flag. MI uiouts will use
27192 this flag if the version is >= 4.
27193 - Add a fix_breakpoint_script_output_globally variable and the
27194 -fix-breakpoint-script-output MI command to set it, if frontends want
27195 to use the fixed output for this without using the newer MI version.
27196 - When emitting the script field, use list instead of tuple, if we want
27197 the fixed output (depending on the two criteria above)
27200 [1] https://sourceware.org/gdb/onlinedocs/gdb/GDB_002fMI-Output-Syntax.html#GDB_002fMI-Output-Syntax
27201 [2] https://gitlab.com/gnutools/binutils-gdb/-/commit/b4be1b0648608a2578bbed39841c8ee411773edd
27203 Change-Id: I7113c6892832c8d6805badb06ce42496677e2242
27204 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=24285
27206 2022-08-10 Andrew Burgess <aburgess@redhat.com>
27208 objdump: fix extended (256) disassembler colors
27211 commit a88c79b77036e4778e70d62081c3cfd1044bb8e3
27212 Date: Tue Aug 9 14:57:48 2022 +0100
27214 Default to enabling colored disassembly if output is to a terminal.
27216 The 256 extended-color support for --disassembler-color was broken.
27217 This is fixed in this commit.
27220 * objdump (objdump_styled_sprintf): Check disassembler_color
27221 against an enum value, don't treat it as a bool.
27223 2022-08-10 mga-sc <mark.goncharov@syntacore.com>
27225 gdb/riscv: implement cannot_store_register gdbarch method
27226 The x0 (zero) register is read-only on RISC-V. Implement the
27227 cannot_store_register gdbarch method to tell GDB this.
27229 Without this method GDB will try to write to x0, and relies on the
27230 target to ignore such writes. If you are using a target that
27231 complains (or throws an error) when writing to x0, this change will
27232 prevent this from happening.
27234 The gdb.arch/riscv-reg-aliases.exp test exercises writing to x0, and
27235 will show the errors when using a suitable target.
27237 2022-08-10 Luis Machado <luis.machado@arm.com>
27239 Disable year 2038 support on 32-bit hosts by default
27240 With a recent import of gnulib, code has been pulled that tests and enables
27241 64-bit time_t by default on 32-bit hosts that support it.
27243 Although gdb can use the gnulib support, bfd doesn't use gnulib and currently
27244 doesn't do these checks.
27246 As a consequence, if we have a 32-bit host that supports 64-bit time_t, we'll
27247 have a mismatch between gdb's notion of time_t and bfd's notion of time_t.
27249 This will lead to mismatches in the struct stat size, leading to memory
27250 corruption and crashes.
27252 This patch disables the year 2038 check for now, which makes things work
27255 I'd consider this a temporary fix until we have proper bfd checks for the year
27256 2038, if it makes sense. 64-bit hosts seems to be more common these days, so
27257 I'm not sure how important it is to have this support enabled and how soon
27258 we want to enable it.
27262 2022-08-10 Jan Beulich <jbeulich@suse.com>
27264 gas/Dwarf: properly skip zero-size functions
27267 While out_debug_abbrev() properly skips such functions, out_debug_info()
27268 mistakenly didn't. It needs to calculate the high_pc expression ahead of
27269 time, in order to skip emitting any data for the function if the value
27272 The one case which would still leave a zero-size entry is when
27273 symbol_get_obj(symp)->size ends up evaluating to zero. I hope we can
27274 expect that to not be the case, otherwise we'd need to have a way to
27275 post-process .debug_info contents between resolving expressions and
27276 actually writing the data out to the file. Even then it wouldn't be
27277 entirely obvious in which way to alter the data.
27279 2022-08-10 Alan Modra <amodra@gmail.com>
27281 PR29462, internal error in relocate, at powerpc.cc:10796
27282 Prior to the inline plt call support (commit 08be322439), the only
27283 local syms with plt entries were local ifunc symbols. There shouldn't
27284 be stubs for other local symbols so don't look for them. The patch
27285 also fixes minor bugs in get_reference_flags; Many relocs are valid
27286 only for ppc64 and a couple only for ppc32.
27289 * powerpc.cc (Target_powerpc::Relocate::relocate): Rename
27290 use_plt_offset to pltcal_to_direct, invert logic. For relocs
27291 not used with inline plt sequences against local symbols, only
27292 look for stubs when the symbol is an ifunc.
27293 (Target_powerpc::Scan::get_reference_flags): Correct reloc
27294 handling for relocs not valid for both 32-bit and 64-bit.
27296 2022-08-10 Youling Tang <tangyouling@loongson.cn>
27298 bfd: Add support for LoongArch64 EFI (efi-*-loongarch64).
27299 This adds support for efi-loongarch64 by virtue of adding a new PEI target
27300 pei-loongarch64. This is not a full target and only exists to support EFI at
27303 This means that this target does not support relocation processing and is mostly
27304 a container format. This format has been added to elf based loongarch64 targets
27305 such that efi images can be made natively on Linux.
27307 However this target is not valid for use with gas but only with objcopy.
27309 We should't limit addresses to 32-bits for 64-bit vma, otherwise there will be
27310 "RVA truncated" error when using objcopy on loongarch64.
27312 With these changes the resulting file is recognized as an efi image.
27314 Any magic number is based on the Microsoft PE specification [1].
27316 The test results are as follows:
27317 $ make check-binutils RUNTESTFLAGS='loongarch64.exp'
27318 PASS: Check if efi app format is recognized
27320 $ objdump -h -f tmpdir/loongarch64copy.o
27321 tmpdir/loongarch64copy.o: file format pei-loongarch64
27322 architecture: Loongarch64, flags 0x00000132:
27323 EXEC_P, HAS_SYMS, HAS_LOCALS, D_PAGED
27324 start address 0x0000000000000000
27327 Idx Name Size VMA LMA File off Algn
27328 0 .text 0000003c 00000000200000b0 00000000200000b0 00000200 2**2
27329 CONTENTS, ALLOC, LOAD, READONLY, CODE
27331 [1] https://docs.microsoft.com/en-us/windows/win32/debug/pe-format
27334 * .gitignore (pe-loongarch64igen.c): New.
27335 * Makefile.am (pei-loongarch64.lo, pe-loongarch64igen.lo, pei-loongarch64.c,
27336 pe-loongarch64igen.c): Add support.
27337 * Makefile.in: Likewise.
27338 * bfd.c (bfd_get_sign_extend_vma): Add pei-loongarch64.
27339 * coff-loongarch64.c: New file.
27340 * coffcode.h (coff_set_arch_mach_hook, coff_set_flags,
27341 coff_write_object_contents) Add loongarch64 (loongarch64_pei_vec) support.
27342 * config.bfd: Likewise.
27343 * configure: Likewise.
27344 * configure.ac: Likewise.
27345 * libpei.h (GET_OPTHDR_IMAGE_BASE, PUT_OPTHDR_IMAGE_BASE,
27346 GET_OPTHDR_SIZE_OF_STACK_RESERVE, PUT_OPTHDR_SIZE_OF_STACK_RESERVE,
27347 GET_OPTHDR_SIZE_OF_STACK_COMMIT, PUT_OPTHDR_SIZE_OF_STACK_COMMIT,
27348 GET_OPTHDR_SIZE_OF_HEAP_RESERVE, PUT_OPTHDR_SIZE_OF_HEAP_RESERVE,
27349 GET_OPTHDR_SIZE_OF_HEAP_COMMIT, PUT_OPTHDR_SIZE_OF_HEAP_COMMIT,
27350 GET_PDATA_ENTRY, _bfd_peLoongArch64_bfd_copy_private_bfd_data_common,
27351 _bfd_peLoongArch64_bfd_copy_private_section_data,
27352 _bfd_peLoongArch64_get_symbol_info, _bfd_peLoongArch64_only_swap_filehdr_out,
27353 _bfd_peLoongArch64_print_private_bfd_data_common,
27354 _bfd_peLoongArch64i_final_link_postscript,
27355 _bfd_peLoongArch64i_only_swap_filehdr_out, _bfd_peLoongArch64i_swap_aouthdr_in,
27356 _bfd_peLoongArch64i_swap_aouthdr_out, _bfd_peLoongArch64i_swap_aux_in,
27357 _bfd_peLoongArch64i_swap_aux_out, _bfd_peLoongArch64i_swap_lineno_in,
27358 _bfd_peLoongArch64i_swap_lineno_out, _bfd_peLoongArch64i_swap_scnhdr_out,
27359 _bfd_peLoongArch64i_swap_sym_in, _bfd_peLoongArch64i_swap_sym_out,
27360 _bfd_peLoongArch64i_swap_debugdir_in, _bfd_peLoongArch64i_swap_debugdir_out,
27361 _bfd_peLoongArch64i_write_codeview_record,
27362 _bfd_peLoongArch64i_slurp_codeview_record,
27363 _bfd_peLoongArch64_print_ce_compressed_pdata): New.
27364 * peXXigen.c (_bfd_XXi_swap_aouthdr_in, _bfd_XXi_swap_aouthdr_out,
27365 _bfd_XXi_swap_scnhdr_out, pe_print_pdata, _bfd_XX_print_private_bfd_data_common,
27366 _bfd_XX_bfd_copy_private_section_data, _bfd_XXi_final_link_postscript):
27367 Support COFF_WITH_peLoongArch64,
27368 * pei-loongarch64.c: New file.
27369 * peicode.h (coff_swap_scnhdr_in, pe_ILF_build_a_bfd, pe_ILF_object_p):
27370 Support COFF_WITH_peLoongArch64.
27371 (jtab): Add dummy entry that traps.
27372 * targets.c (loongarch64_pei_vec): New.
27375 * testsuite/binutils-all/loongarch64/loongarch64.exp: New file.
27376 * testsuite/binutils-all/loongarch64/pei-loongarch64.d: New test.
27377 * testsuite/binutils-all/loongarch64/pei-loongarch64.s: New test.
27380 * coff/loongarch64.h: New file.
27381 * coff/pe.h (IMAGE_FILE_MACHINE_LOONGARCH64): New.
27383 2022-08-10 GDB Administrator <gdbadmin@sourceware.org>
27385 Automatic date update in version.in
27387 2022-08-09 Andrew Burgess <aburgess@redhat.com>
27389 gdb/riscv/testsuite: fix failures in gdb.arch/riscv-reg-aliases.exp
27390 When running on a native RISC-V Linux target I currently see failures
27391 in the gdb.arch/riscv-reg-aliases.exp test like this:
27393 set $ft0.float = 501
27394 (gdb) PASS: gdb.arch/riscv-reg-aliases.exp: write non-zero value to ft0
27397 (gdb) FAIL: gdb.arch/riscv-reg-aliases.exp: read ft0 after non-zero write to ft0
27399 This test started failing after this commit:
27401 commit 56262a931b7ca8ee3ec9104bc7e9e0b40cf3d64e
27402 Date: Thu Feb 17 13:43:59 2022 -0700
27404 Change how "print/x" displays floating-point value
27406 The problem is that when 501 is written to $ft0.float the value is
27407 converted to floating point format and stored in the register. Prior
27408 to the above commit printing with /x and /d would first extract the
27409 value as a float, and then convert the value to an integer for
27410 display. After the above commit GDB now uses the raw register value
27411 when displaying /x and /d, and so we see this behaviour:
27413 (gdb) info registers $ft0
27414 ft0 {float = 501, double = 5.6347704700123827e-315} (raw 0x0000000043fa8000)
27415 (gdb) p/f $ft0.float
27417 (gdb) p/d $ft0.float
27419 (gdb) p/x $ft0.float
27422 To fix this test I now print the float registers using the /f format
27423 rather than /d. With this change the test now passes.
27425 2022-08-09 Stepan Nemec <stepnem@gmail.com>
27427 Another gas manual typo correction.
27429 Fix typos in assembler documentation.
27431 2022-08-09 Feiyang Chen <chenfeiyang@loongson.cn>
27433 gdb/gdbserver: LoongArch: Improve implementation of fcc registers
27434 The current implementation of the fcc register is referenced to the
27435 user_fp_state structure of the kernel uapi [1].
27437 struct user_fp_state {
27443 But it is mistakenly defined as a 64-bit fputype register, resulting
27444 in a confusing output of "info register".
27446 (gdb) info register
27448 fcc {f = 0x0, d = 0x0} {f = 0, d = 0}
27451 According to "Condition Flag Register" in "LoongArch Reference Manual"
27452 [2], there are 8 condition flag registers of size 1. Use 8 registers of
27453 uint8 to make it easier for users to view the fcc register groups.
27455 (gdb) info register
27467 [1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/loongarch/include/uapi/asm/ptrace.h
27468 [2] https://loongson.github.io/LoongArch-Documentation/LoongArch-Vol1-EN.html#_condition_flag_register
27470 2022-08-09 Nick Clifton <nickc@redhat.com>
27472 Default to enabling colored disassembly if output is to a terminal.
27474 * objdump.c (disassembler_color): Change type to an enum.
27475 (disassembler_extended_color): Remove.
27477 (objdump_color_for_assembler_style): Update.
27478 (main): Update initialisation of disassembler_color. If not
27479 initialised via a command line option, set based upon terminal
27481 * doc/binutils.texi: Update description of disassmbler-color
27483 * testsuite/binutils-all/arc/objdump.exp: Add
27484 --disassembler-color=off option when disassembling.
27485 * testsuite/binutils-all/arm/objdump.exp: Likewise.
27487 2022-08-09 Aditya Vidyadhar Kamath <Aditya.Kamath1@ibm.com>
27489 Fix-for-multiple-thread-detection-in-AIX.
27490 In AIX multiple threads were not added. This patch is a fix for the same
27492 When we create a pthread debug session we have callbacks to read
27493 symbols and memory. One of those call backs is pdc_read_data.
27495 Before we come into aix-thread wait() we switch to no thread and
27496 therefore the current thread is null.
27498 When we get into pdc_read_data we have a dependency that we need to
27499 be in the correct current thread that has caused an event of new
27500 thread, inorder to read memory.
27502 Hence we switch to the correct thread.
27504 This is done by passing the pid in the pthdb_user_t user_current_pid
27505 parameter in every call back.
27507 2022-08-09 Tom de Vries <tdevries@suse.de>
27509 [gdb/testsuite] Fix gdb.dwarf2/debug-names.exp
27510 When running test-case gdb.dwarf2/debug-names.exp on openSUSE Tumbleweed, I
27513 (gdb) maint info symtabs^M
27515 ERROR: internal buffer is full.
27516 UNRESOLVED: gdb.dwarf2/debug-names.exp: break _start expanded symtab
27519 Fix this by simplifying the test-case to print _start rather running to it.
27521 Tested on x86_64-linux.
27523 2022-08-09 Andrew Burgess <aburgess@redhat.com>
27525 gdb/riscv: use register name enum values in riscv-linux-nat.c
27526 There were a few places where we were using integer values rather than
27527 the RISCV_*_REGNUM constants defined in riscv-tdep.h. This commit
27528 replaces 0 with RISCV_ZERO_REGNUM and 32 with RISCV_PC_REGNUM in a few
27531 There should be no user visible changes after this commit.
27533 2022-08-09 Jan Beulich <jbeulich@suse.com>
27535 x86-64: adjust MOVQ to/from SReg attributes
27536 It is unclear to me why the corresponding MOV (no Q suffix) can be
27537 issued without REX.W, but MOVQ has to have that prefix (bit). Add
27538 NoRex64 and in exchange drop Size64.
27540 x86: adjust MOVSD attributes
27541 The non-SSE2AVX form of the SIMD variant of the instruction needlessly
27542 has the (still multi-purpose) IgnoreSize attribute. All other similar
27543 SSE2 insns use NoRex64 instead. Make this consistent, noting that the
27544 SSE2AVX form can't have the same change made - there the memory operand
27545 doesn't at the same time permit RegXMM (which logic uses when deciding
27546 whether a Q suffix is okay outside of 64-bit mode).
27548 x86: fold AVX VGATHERDPD / VPGATHERDQ
27549 While the other three variants each differ in attributes and hence can't
27550 be folded, these two pairs actually can be (and were previously
27551 overlooked). This effectively matches their AVX512VL counterparts, which
27552 are also expressed as a single template.
27554 x86: allow use of broadcast with X/Y/Z-suffixed AVX512-FP16 insns
27555 While the x/y/z suffix isn't necessary to use in this case, it is still
27556 odd that these forms don't support broadcast (unlike their AVX512F /
27557 AVX512DQ counterparts). The lack thereof can e.g. make macro-ized
27558 programming more difficult.
27560 x86/Intel: split certain AVX512-FP16 VCVT*2PH templates
27561 One more place where pre-existing templates should have been taken as a
27562 basis: In Intel syntax we want to consistently issue an "ambiguous
27563 operand size" error when a size-less memory operand is specified for an
27564 insn where register use alone isn't sufficient for disambiguation.
27566 2022-08-09 Jiangshuai Li <jiangshuai_li@linux.alibaba.com>
27568 gdb/csky fix build error in ubuntu20_04
27569 build error in: https://builder.sourceware.org/buildbot/#/builders/170/builds/246
27571 ../../binutils-gdb/gdb/csky-linux-tdep.c: In function ‘void
27572 csky_supply_fregset(const regset*, regcache*, int, const void*, size_t)’:
27573 ../../binutils-gdb/gdb/csky-linux-tdep.c:194:18: error: format ‘%ld’
27574 expects argument of type ‘long int’, but argument 2 has type ‘size_t’
27575 {aka ‘unsigned int’} [-Werror=format=]
27576 194 | warning (_("Unknow size %ld of section .reg2, can not get
27579 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
27580 195 | " of float registers."), len);
27583 Fix it via using %s vs pulongest suggested by Tom.
27585 2022-08-09 GDB Administrator <gdbadmin@sourceware.org>
27587 Automatic date update in version.in
27589 2022-08-08 Tom Tromey <tromey@adacore.com>
27591 Fix regression from gdbarch registry change
27592 The gdbarch registry patch introduced a regression that could cause a
27593 crash when opening files in gdb. The bug is that, previously, the
27594 solib ops would default to current_target_so_ops; but the patch
27595 changed this code to default to nullptr. This patch fixes the bug by
27596 reintroducing the earlier behavior. This is PR gdb/29449.
27598 I managed to reproduce the bug with a riscv-elf build and then
27599 verified that this fixes the problem.
27601 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29449
27603 2022-08-08 Martin Liska <mliska@suse.cz>
27605 add splay tree for info_ptr -> CU mapping
27606 While using perf top for MozillaThunderbird I noticed quite some slow
27607 dissably call with source code involved. E.g.
27609 time ./objdump --start-address=0x0000000004e0dcd0 --stop-address=0x0000000004e0df8b -l -d --no-show-raw-insn -S -C /usr/lib64/thunderbird/libxul.so
27611 took 2.071s and I noticed quite some time is spent in
27612 find_abstract_instance:
27614 33.46% objdump objdump [.] find_abstract_instance
27615 18.22% objdump objdump [.] arange_add
27616 13.77% objdump objdump [.] read_attribute_value
27617 4.82% objdump objdump [.] comp_unit_maybe_decode_line_info
27618 3.10% objdump libc.so.6 [.] __memset_avx2_unaligned_erms
27620 where linked list of CU is iterated when searing for where info_ptr
27623 : 3452 for (u = unit->prev_unit; u != NULL; u = u->prev_unit)
27624 0.00 : 4c61f7: mov 0x10(%rbx),%rax
27625 0.00 : 4c61fb: test %rax,%rax
27626 0.00 : 4c61fe: je 4c6215 <find_abstract_instance+0x365>
27627 : 3453 if (info_ptr >= u->info_ptr_unit && info_ptr < u->end_ptr)
27628 0.00 : 4c6200: cmp 0x60(%rax),%rdx
27629 83.20 : 4c6204: jb 4c620c <find_abstract_instance+0x35c>
27630 0.00 : 4c6206: cmp 0x78(%rax),%rdx
27631 6.89 : 4c620a: jb 4c6270 <find_abstract_instance+0x3c0>
27632 : 3452 for (u = unit->prev_unit; u != NULL; u = u->prev_unit)
27633 0.00 : 4c620c: mov 0x10(%rax),%rax
27634 7.90 : 4c6210: test %rax,%rax
27635 0.00 : 4c6213: jne 4c6200 <find_abstract_instance+0x350>
27637 The following scan can be replaced with search in a splay tree and with
27638 that I can get to 1.5s and there are other symbols where the difference
27644 * dwarf2.c (struct addr_range): New.
27645 (addr_range_intersects): Likewise.
27646 (splay_tree_compare_addr_range): Likewise.
27647 (splay_tree_free_addr_range): Likewise.
27648 (struct dwarf2_debug_file): Add comp_unit_tree.
27649 (find_abstract_instance): Use the splay tree when searching
27651 (stash_comp_unit): Insert to the splay tree.
27652 (_bfd_dwarf2_cleanup_debug_info): Clean up the splay tree.
27654 2022-08-08 Martin Liska <mliska@suse.cz>
27656 dwarf: use find_abstract_instance for vars and DW_AT_specification
27657 The following simple test case fails when dwz is used:
27661 enum { _S_fixed, _S_floatfield = _S_fixed };
27672 $ g++ demo.C -g && cp a.out b.out && dwz -m xxx.so a.out b.out && objdump -S a.out >/dev/null
27673 objdump: DWARF error: could not find variable specification at offset 0x3d3
27675 As seen the reference is defined in xxx.so shared part:
27677 $ eu-readelf -w -N a.out | grep -A3 -B3 3d3
27678 decl_column (data1) 11
27679 sibling (ref_udata) [ 387]
27680 [ 387] variable abbrev: 30
27681 specification (GNU_ref_alt) [ 3d3]
27684 [ 396] subprogram abbrev: 32
27686 $ eu-readelf -w -N a.out | less
27690 Compilation unit at offset 920:
27691 Version: 5, Abbreviation section offset: 0, Address size: 8, Offset size: 4
27692 Unit type: partial (3)
27694 [ 3d3] variable abbrev: 31
27695 name (strp) "__ioinit"
27696 decl_file (data1) demo.C (10)
27697 decl_line (data1) 6
27698 decl_column (data1) 3
27699 type (ref_udata) [ 3c4]
27700 declaration (flag_present) yes
27702 With the patch the same output is emitted as before usage of dwz.
27707 * dwarf2.c (struct varinfo): Use const char * type.
27708 (scan_unit_for_symbols): Call find_abstract_instance for
27709 DW_AT_specification for variables that can be in a different CU
27712 2022-08-08 Tsukasa OI <research_trasio@irq.a4lg.com>
27714 Mach-O: i18n enablement on some error messages.
27715 * config/obj-macho.c (obj_mach_o_get_section_names): Wrap two
27716 string literals within with gettext macro.
27718 2022-08-08 Martin Liska <mliska@suse.cz>
27723 * NEWS: Fix 2 typos.
27725 2022-08-08 Nick Clifton <nickc@redhat.com>
27727 Add a link to the NEWS files in the release announcement email.
27729 2022-08-08 Jiangshuai Li <jiangshuai_li@linux.alibaba.com>
27731 gdb/csky support .reg2 for kernel 4.x and later
27732 When kernel's version >= 4.x, the size of .reg2 section will be 400.
27733 Contents of .reg2 are {
27734 unsigned long vr[96];
27736 unsigned long fesr;
27738 unsigned long reserved;
27741 VR[96] means: (vr0~vr15) + (fr16~fr31), each Vector register is
27742 128-bits, each Float register is 64 bits, the total size is
27745 In addition, for fr0~fr15, each FRx is the lower 64 bits of the
27746 corresponding VRx. So fr0~fr15 and vr0~vr15 regisetrs use the same
27749 2022-08-08 GDB Administrator <gdbadmin@sourceware.org>
27751 Automatic date update in version.in
27753 2022-08-07 Tom de Vries <tdevries@suse.de>
27755 [gdb/build] Fix build with gcc 4.8.5
27756 When building with gcc 4.8.5, I run into:
27758 user-regs.c:85:1: error: could not convert \
27759 ‘{0l, (& builtin_user_regs.gdb_user_regs::first)}’ from \
27760 ‘<brace-enclosed initializer list>’ to ‘gdb_user_regs’
27765 Fix this by removing the initialization and handling regs.last == nullptr in
27768 Tested on x86_64-linux.
27770 2022-08-07 Tom de Vries <tdevries@suse.de>
27772 [gdb/symtab] Fix assert in read_addrmap_from_aranges
27773 When loading the debug-names-duplicate-cu executable included in this
27774 test-case, we run into:
27776 (gdb) file debug-names-duplicate-cu^M
27777 Reading symbols from debug-names-duplicate-cu...^M
27778 src/gdb/dwarf2/read.c:2353: internal-error: read_addrmap_from_aranges: \
27779 Assertion `insertpair.second' failed.^M
27782 This assert was added in recent commit 75337cbc147 ("[gdb/symtab] Fix
27783 .debug_aranges duplicate offset warning").
27785 The assert triggers because the CU table in the .debug_names section contains
27789 Augmentation string: 47 44 42 00 ("GDB")
27795 Fix this by rejecting the .debug_names index:
27797 (gdb) file debug-names-duplicate-cu^M
27798 Reading symbols from debug-names-duplicate-cu...^M
27799 warning: Section .debug_names has duplicate entry in CU table, \
27800 ignoring .debug_names.^M
27803 Likewise for the case where the CU table is not sorted by increasing offset.
27805 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29436
27807 2022-08-07 Tom de Vries <tdevries@suse.de>
27809 [gdb/testsuite] Add support for .debug_names in dwarf assembler
27811 - support for a per-module .debug_names section in the dwarf assembler, and
27812 - a test-case excercising this new functionality.
27814 A per-module .debug_names section needs to have an entry in the CU list for
27815 each CU in the module, which is made more difficult by two things:
27816 - linking in other objects, which may contain additional CUs
27817 (typically the case on openSUSE), and
27818 - adding dummy CUs in the dwarf assembler.
27820 - compiling with -nostartfiles (so the test-case contains _start rather than
27822 - disabling the dummy CU generation for the test-case.
27824 I've kept things simple by having the test-case specify the hash value, rather
27825 than adding that functionality in the dwarf assembler.
27827 Also I've kept the bucket count to 1, which makes it trivial to satisfy the
27828 requirement that "the symbol is entered into a bucket whose index is the hash
27829 value modulo bucket_count".
27831 The readelf dump of the .debug_names section from the test-case looks like:
27834 Augmentation string: 47 44 42 00 ("GDB")
27842 Used 1 of 1 bucket.
27843 Out of 2 items there are 1 bucket clashes (longest of 1 entries).
27846 [ 0] #eddb6232 _start: <1> DW_TAG_subprogram DW_IDX_compile_unit=0
27847 [ 1] #0b888030 int: <2> DW_TAG_base_type DW_IDX_compile_unit=0
27850 Tested on x86_64-linux.
27852 2022-08-07 GDB Administrator <gdbadmin@sourceware.org>
27854 Automatic date update in version.in
27856 2022-08-06 Alan Modra <amodra@gmail.com>
27858 asan: heap buffer overflow in _bfd_error_handler
27859 On coff_slurp_symbol_table printing "unrecognized storage class"
27860 for a symbol error. If the symbol name is the last string in its
27861 section and not terminated, we run off the end of the buffer.
27863 * coffgen.c (build_debug_section): Terminate the section with
27866 2022-08-06 Alan Modra <amodra@gmail.com>
27868 asan: segfault in coff_write_auxent_fname
27869 More fuzzed input file nonsense.
27871 * coffgen.c (coff_write_symbol): Don't call coff_write_auxent_fname
27872 when extrap is NULL.
27874 2022-08-06 Alan Modra <amodra@gmail.com>
27876 msan: bfd_mach_o_layout_commands use of uninitialised value
27877 Catches fuzzed input with unterminated strings that later run off the
27878 end of their buffers when calling strlen.
27880 * mach-o.c: Use size_t vars where approprite.
27881 (bfd_mach_o_alloc_and_read): Add "extra" param. Allocate that
27882 much extra and clear. Update all callers, those that set up
27883 strings with one extra byte.
27885 2022-08-06 Alan Modra <amodra@gmail.com>
27887 objcopy section alignment
27888 bfd_set_section_alignment currently always returns true. This patch
27889 changes it to return false on silly alignment values, avoiding yet
27890 another way to trigger ubsan errors like coffcode.h:3192:12: runtime
27891 error: shift exponent 299 is too large for 32-bit type 'int'. We'll
27892 catch that one in objcopy.c:setup_sections. However, setup_sections
27893 gives up on other setup operations that are necessary even after an
27894 error of some sort. Change that to keep going, which might change the
27895 error message but that shouldn't matter in the least.
27898 * section.c (bfd_set_section_alignment): Return false and
27899 don't set alignment_power for stupidly large alignments.
27900 * bfd-in2.h: Regenerate.
27901 * coffcode.h (coff_compute_section_file_positions): Don't use
27902 an int constant when calculating alignment.
27904 * objcopy.c (setup_section): Keep on going after hitting
27907 2022-08-06 Alan Modra <amodra@gmail.com>
27909 ubsan: som.c undefined shift in som_set_reloc_info
27910 Do the shift using unsigned variables to avoid UB on << 8.
27912 * som.c (som_set_reloc_info): Make v unsigned. Localise some
27913 variables to their blocks.
27915 2022-08-06 GDB Administrator <gdbadmin@sourceware.org>
27917 Automatic date update in version.in
27919 2022-08-05 Alan Modra <amodra@gmail.com>
27921 Get rid of BFD_VMA_FMT
27922 Remove the BFD_VMA_FMT defines in bfd.h and configure support.
27924 * bfd-in.h (BFD_VMA_FMT): Don't define.
27925 * configure.ac (BFD_INT64_FMT): Remove configure test.
27926 * configure.com: Likewise.
27927 * Makefile.in: Regenerate.
27928 * bfd-in2.h: Regenerate.
27929 * configure: Regenerate.
27931 2022-08-05 Alan Modra <amodra@gmail.com>
27933 Don't use BFD_VMA_FMT in gdb and sim
27934 Like commit b82817674f, this replaces BFD_VMA_FMT "x" in sim/ with
27935 PRIx64 and casts to promote bfd_vma to uint64_t. The one file using
27936 BFD_VMA_FMT in gdb/ instead now uses hex_string, and a typo in the
27937 warning message is fixed.
27939 2022-08-05 Tom de Vries <tdevries@suse.de>
27941 [gdb/build] Fix build breaker in language.c with gcc 7.5.0
27942 When building gdb on openSUSE Leap 15.3, using gcc 7.5.0, I run into:
27944 gdb/language.c: In constructor ‘constexpr language_gdbarch::language_gdbarch()’:
27945 gdb/language.c:921:8: error: use of deleted function \
27946 ‘language_arch_info::language_arch_info(const language_arch_info&)’
27947 struct language_gdbarch
27949 In file included from gdbsupport/common-defs.h:104:0,
27950 from gdb/defs.h:28,
27951 from gdb/language.c:31:
27952 gdb/language.h:95:28: note: declared here
27953 DISABLE_COPY_AND_ASSIGN (language_arch_info);
27955 include/ansidecl.h:342:3: note: in definition of macro \
27956 ‘DISABLE_COPY_AND_ASSIGN’
27957 TYPE (const TYPE&) = delete; \
27959 gdb/language.c: In function ‘language_gdbarch* get_language_gdbarch(gdbarch*)’:
27960 gdb/language.c:936:22: note: synthesized method ‘constexpr \
27961 language_gdbarch::language_gdbarch()’ first required here
27962 l = new struct language_gdbarch;
27966 This seems to be fixed by this change in the struct language_gdbarch
27969 - struct language_arch_info arch_info[nr_languages] {};
27970 + struct language_arch_info arch_info[nr_languages];
27973 Tested on x86_64-linux.
27975 2022-08-05 Tom de Vries <tdevries@suse.de>
27977 [gdb] Add unit test for gdb::sequential_for_each
27978 With commit 18a5766d09c ("[gdbsupport] Add sequential_for_each") I added a
27979 drop-in replacement for gdb::parallel_for_each, but there's nothing making
27980 sure that the two remain in sync.
27982 Extend the unit test for gdb::parallel_for_each to test both.
27984 Do this using a slightly unusual file-self-inclusion. Doing so keep things
27985 readable and maintainable, and avoids macrofying functions.
27987 Tested on x86_64-linux.
27989 2022-08-05 Tom de Vries <tdevries@suse.de>
27991 [gdb/symtab] Use task size in parallel_for_each in dwarf2_build_psymtabs_hard
27992 In dwarf2_build_psymtabs_hard, we use a parallel_for_each to distribute CUs
27995 Ensuring a fair distribution over the worker threads and main thread in terms
27996 of number of CUs might not be the most efficient way, given that CUs can vary
27999 Fix this by using per_cu->get_length () as the task size.
28001 I've used this experiment to verify the performance impact:
28003 $ for n in $(seq 1 10); do \
28004 time gdb -q -batch ~/firefox/libxul.so-93.0-1.1.x86_64.debug \
28009 and without the patch got:
28022 and with the patch:
28035 so that seems a reasonable improvement.
28037 With parallel_for_each_debug set to true, we get some more detail about
28038 the difference in behaviour. Without the patch we have:
28040 Parallel for: n_elements: 2818
28041 Parallel for: minimum elements per thread: 1
28042 Parallel for: elts_per_thread: 704
28043 Parallel for: elements on worker thread 0 : 705
28044 Parallel for: elements on worker thread 1 : 705
28045 Parallel for: elements on worker thread 2 : 704
28046 Parallel for: elements on worker thread 3 : 0
28047 Parallel for: elements on main thread : 704
28049 and with the patch:
28051 Parallel for: n_elements: 2818
28052 Parallel for: total_size: 1483674865
28053 Parallel for: size_per_thread: 370918716
28054 Parallel for: elements on worker thread 0 : 752 (size: 371811790)
28055 Parallel for: elements on worker thread 1 : 360 (size: 371509370)
28056 Parallel for: elements on worker thread 2 : 1130 (size: 372681710)
28057 Parallel for: elements on worker thread 3 : 0 (size: 0)
28058 Parallel for: elements on main thread : 576 (size: 367671995)
28061 Tested on x86_64-linux.
28063 2022-08-05 Tom de Vries <tdevries@suse.de>
28065 [gdbsupport] Add task size parameter in parallel_for_each
28066 Add a task_size parameter to parallel_for_each, defaulting to nullptr, and use
28067 the task size to distribute similarly-sized chunks to the threads.
28069 Tested on x86_64-linux.
28071 2022-08-05 Pedro Alves <pedro@palves.net>
28073 Introduce gdb::make_function_view
28074 This adds gdb::make_function_view, which lets you create a function
28075 view from a callable without specifying the function_view's template
28076 parameter. For example, this:
28078 auto lambda = [&] (int) { ... };
28079 auto fv = gdb::make_function_view (lambda);
28083 auto lambda = [&] (int) { ... };
28084 gdb::function_view<void (int)> fv = lambda;
28086 It is particularly useful if you have a template function with an
28087 optional function_view parameter, whose type depends on the function's
28088 template parameters. Like:
28090 template<typename T>
28091 void my_function (T v, gdb::function_view<void(T)> callback = nullptr);
28093 For such a function, the type of the callback argument you pass must
28094 already be a function_view. I.e., this wouldn't compile:
28096 auto lambda = [&] (int) { ... };
28097 my_function (1, lambda);
28099 With gdb::make_function_view, you can write the call like so:
28101 auto lambda = [&] (int) { ... };
28102 my_function (1, gdb::make_function_view (lambda));
28104 Unit tests included.
28106 Tested by building with GCC 9.4, Clang 10, and GCC 4.8.5, on x86_64
28107 GNU/Linux, and running the unit tests.
28109 Change-Id: I5c4b3b4455ed6f0d8878cf1be189bea3ee63f626
28111 2022-08-05 Nick Clifton <nickc@redhat.com>
28113 Update following 2.39 release
28115 2022-08-05 Alan Modra <amodra@gmail.com>
28117 asan: ppc64_elf_get_synthetic_symtab heap buffer overflow
28118 Fuzzed input files with sizes of .dynamic not a multiple of dynamic
28119 tag size can result in reading past the end of the buffer with the
28120 current simple checks. Fix that, and use the same check in other
28121 files that process input object .dynamic section. (There is no need
28122 for buffer overflow checks in the linker's generated .dynamic
28125 * elf32-ppc.c (ppc_elf_get_synthetic_symtab): Sanity check
28126 .dynamic content buffer reads.
28127 * elf64-ppc.c (ppc64_elf_get_synthetic_symtab): Likewise.
28128 * elf64-ia64-vms.c (elf64_vms_link_add_object_symbols): Likewise.
28129 * elf.c (_bfd_elf_print_private_bfd_data): Simplify .dynamic
28130 buffer sanity checks.
28131 * elflink.c (elf_link_add_object_symbols): Avoid possible UB
28132 subtracting sizeof_dyn from pointer.
28134 2022-08-05 Alan Modra <amodra@gmail.com>
28136 Sanity check loc_offsets index
28137 Fixes a segfault found by the fuzzers.
28139 * dwarf.c (fetch_indexed_value): Return -1 on error.
28140 (read_and_display_attr_value): Don't display string when
28141 fetch_indexed_value returns an error. Sanity check loc_offsets
28144 2022-08-05 Jan Beulich <jbeulich@suse.com>
28146 binutils/Dwarf: avoid "shadowing" of glibc function name
28147 As before: Old enough glibc has an (unguarded) declaration of index()
28148 in string.h, which triggers a "shadows a global declaration" warning.
28150 2022-08-05 Tsukasa OI <research_trasio@irq.a4lg.com>
28152 gas: fix a testcase broken by new ZSTD support
28153 The commit 1369522f36eece1b37139a81f7f2139ba3915172 ("Recognize the new ELF
28154 compression type for ZSTD.") added the new ELF compression type but it
28155 accidentally broke a GAS testcase. Since testing for the section type
28156 "2048" (SHF_COMPRESSED) is not going to be portable in the long term, it
28157 now tests SHF_LINK_ORDER ("128") instead.
28159 Using SHF_LINK_ORDER (with possibly sh_link == 0) is an idea by Jan Beulich.
28163 * testsuite/gas/elf/section10.s: Use SHF_LINK_ORDER to test
28164 mixed numeric and alpha values.
28165 * testsuite/gas/elf/section10.d: Reflect the change above.
28167 2022-08-05 Nick Clifton <nickc@redhat.com>
28169 When gas/read.c calls mbstowcs with a NULL destination, it should set size to 0
28171 * read.c (read_symbol_name): Pass 0 as the length parameter when
28172 invoking mbstowc in order to check the validity of a wide string.
28174 2022-08-05 Tom de Vries <tdevries@suse.de>
28176 [gdb] Add debug_{exp,val}
28177 When debugging cc1 I heavily rely on simple one-parameter debug functions
28178 that allow me to inspect a variable of a common type, like:
28179 - debug_generic_expr
28180 - debug_gimple_stmt
28182 and I miss similar functions in gdb.
28184 Add functions to dump variables of types 'value' and 'expression':
28188 Tested on x86_64-linux, by breaking on varobj_create, and doing:
28190 (gdb) call debug_exp (var->root->exp.get ())
28191 &"Operation: OP_VAR_VALUE\n"
28192 &" Block symbol:\n"
28194 &" Block: 0x2d064f0\n"
28199 (gdb) call debug_val (value)
28204 2022-08-05 Luca Boccassi <bluca@debian.org>
28206 Add gold support for --package-metadata option.
28207 Following the same format as the implementation in ld:
28208 9e2bb0cb5e74aed4158f08495534922d7108f928
28210 Generate a .note.package FDO package metadata ELF note, following
28211 the spec: https://systemd.io/ELF_PACKAGE_METADATA/
28213 If the jansson library is available at build time (and it is explicitly
28214 enabled), link ld to it, and use it to validate that the input is
28215 correct JSON, to avoid writing garbage to the file. The
28216 configure option --enable-jansson has to be used to explicitly enable
28217 it (error out when not found). This allows bootstrappers (or others who
28218 are not interested) to seamlessly skip it without issues.
28221 * elfcpp.h: Add FDO_PACKAGING_METADATA note type.
28224 * Makefile.am: Add jansson flags and libraries.
28225 * configure.ac: Check for jansson library.
28226 * layout.cc (Layout::create_notes): Call create_package_metadata().
28227 (Layout::create_package_metadata): New function.
28228 * layout.h (Layout::create_package_metadata): New function.
28229 (Layout::package_metadata_note_): New data member.
28230 * options.h (class General_options): Add --package-metadata option.
28231 * testsuite/Makefile.am (object_unittest): Add jansson libraries.
28232 (binary_unittest): Likewise.
28233 (leb128_unittest): Likewise.
28234 (overflow_unittest): Likewise.
28235 (package_metadata_test): New test.
28236 * testsuite/package_metadata_main.c: New test source.
28238 2022-08-05 Cary Coutant <ccoutant@gmail.com>
28240 Recognize the new ELF compression type for ZSTD.
28241 There is more work to be done to actually support compression and
28242 decompression using the zstd library, but I will leave that to the
28243 champions of the new compression option.
28246 * binutils/readelf.c (process_section_headers): Add support for
28249 2022-08-05 GDB Administrator <gdbadmin@sourceware.org>
28251 Automatic date update in version.in
28253 2022-08-04 Tom Tromey <tom@tromey.com>
28255 Use registry in gdbarch
28256 gdbarch implements its own registry-like approach. This patch changes
28257 it to instead use registry.h. It's a rather large patch but largely
28258 uninteresting -- it's mostly a straightforward conversion from the old
28259 approach to the new one.
28261 The main benefit of this change is that it introduces type safety to
28262 the gdbarch registry. It also removes a bunch of code.
28264 One possible drawback is that, previously, the gdbarch registry
28265 differentiated between pre- and post-initialization setup. This
28266 doesn't seem very important to me, though.
28268 2022-08-04 Tom Tromey <tom@tromey.com>
28270 Allow registry to refer to const types
28271 So far, the registry hasn't been used to refer to a 'const' type, but
28272 this changes with the gdbarch change. This patch arranges to let the
28273 registry store a pointer-to-const, by removing const in the 'set'
28276 Use new and delete for gdbarch
28277 This changes gdbarch to use new and delete.
28279 Use bool in gdbarch
28280 This changes gdbarch to use bool for initialized_p.
28282 2022-08-04 Tom de Vries <tdevries@suse.de>
28284 [gdb/testsuite] Fix .debug_aranges in gdb.dwarf2/fission-loclists.S
28285 When running test-case gdb.dwarf2/fission-loclists.exp, I noticed:
28287 warning: Section .debug_aranges in fission-loclists has duplicate \
28288 debug_info_offset 0x8f, ignoring .debug_aranges.^M
28291 Fix this by removing the duplicate .debug_aranges entry.
28293 Tested on x86_64-linux.
28295 2022-08-04 Tom de Vries <tdevries@suse.de>
28297 [gdb/testsuite] Fix ERROR in gdb.base/watchpoint-unaligned.exp
28298 In PR23888 an error is reported:
28300 ERROR: tcl error sourcing watchpoint-unaligned.exp.
28301 ERROR: expected boolean value but got ""
28306 This presumably happens when:
28307 - skip_hw_watchpoint_tests returns 0 meaning hw watchpoints are supported
28308 - gdb fails to set a hw watchpoint and instead sets a sw watchpoint
28310 That particular situation is handled for arm:
28312 -re "Watchpoint (\[0-9\]+): .*\r\n$gdb_prompt $" {
28313 if {[istarget "arm*-*-*"]} {
28319 but not for any other targets so wpnum remains "", triggering the ERROR.
28321 Possibly this has been fixed for powerpc by commit 8d4e4d13afb ("gdb Power 9
28322 add test for HW watchpoint support."), but it's still possible for other
28326 - initializing wpnum to 0 instead of ""
28327 - signalling the failure to set a hw watchpoint by a fail
28329 Tested on x86_64-linux, also by adding:
28331 gdb_test_no_output "set can-use-hw-watchpoints 0"
28333 and verifying that it triggers the fail.
28335 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=23888
28337 2022-08-04 Tom de Vries <tdevries@suse.de>
28339 [gdb/tdep] Fix gdb.base/large-frame.exp for aarch64
28340 On aarch64, I run into:
28342 FAIL: gdb.base/large-frame.exp: optimize=-O0: backtrace
28345 The problem is that the architecture-specific prologue analyzer fails to
28346 handle the first two insns in the prologue properly:
28348 0000000000400610 <func>:
28349 400610: d2880210 mov x16, #0x4010
28350 400614: cb3063ff sub sp, sp, x16
28351 400618: a9007bfd stp x29, x30, [sp]
28352 40061c: 910003fd mov x29, sp
28353 400620: 910043a0 add x0, x29, #0x10
28354 400624: 97fffff0 bl 4005e4 <blah>
28358 $ gdb -q -batch ./outputs/gdb.base/large-frame/large-frame-O0 -ex "b func"
28359 Breakpoint 1 at 0x400614
28363 - fixing the support for the first insn to extract the immediate operand, and
28364 - adding support for the second insn,
28367 Breakpoint 1 at 0x400624
28369 Note that we're overshooting by one insn (0x400620 is the first insn after the
28370 prologue), but that's a pre-existing problem.
28372 Tested on aarch64-linux.
28374 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29408
28376 2022-08-04 Alan Modra <amodra@gmail.com>
28378 Don't use BFD_VMA_FMT in binutils
28379 BFD_VMA_FMT can't be used in format strings that need to be
28380 translated, because the translation won't work when the type of
28381 bfd_vma differs from the machine used to compile .pot files. We've
28382 known about this for a long time, but patches slip through review.
28384 So just get rid of BFD_VMA_FMT, instead using the appropriate PRId64,
28385 PRIu64, PRIx64 or PRIo64 and SCN variants for scanf. The patch is
28386 mostly mechanical, the only thing requiring any thought is casts
28387 needed to preserve PRId64 output from bfd_vma values, or to preserve
28388 one of the unsigned output formats from bfd_signed_vma values.
28390 2022-08-04 Alan Modra <amodra@gmail.com>
28392 Re: Get rid of fprintf_vma and sprintf_vma
28393 Commit f493c2174e messed the formatting in linker map files,
28394 particularly for 32-bit builds where a number of tests using map files
28395 regressed. I should have noticed the BFD64 conditional printing of
28396 spaces to line up output due to the original %V printing hex vmas with
28397 16 digits when BFD64 and 8 digits when not. Besides that, it is nicer
28398 to print 32-bit vmas for 32-bit targets. So change %V back to be
28399 target dependent, now using bfd_sprintf_vma. Since minfo doesn't
28400 return the number of chars printed, that means some places that
28401 currently use %V must instead sprintf to a buffer in order to find the
28404 * ldmisc.h (print_spaces): Declare.
28405 (print_space): Change to a macro.
28406 * ldmisc.c (vfinfo): Use bfd_sprintf_vma for %V. Tidy %W case.
28407 (print_space): Delete.
28408 (print_spaces): New function.
28409 * emultempl/aix.em (print_symbol): Use print_spaces.
28410 * ldctor.c (ldctor_build_sets): Likewise.
28411 * ldmain.c (add_archive_element): Likewise.
28412 * ldlang.c (print_one_symbol, lang_print_asneeded): Likewise.
28413 (print_output_section_statement, print_data_statement): Likewise.
28414 (print_reloc_statement, print_padding_statement): Likewise.
28415 (print_assignment): Likewise. Also replace %V printing of vmas
28416 with printing to a buffer in order to properly format output.
28417 (print_input_section, lang_one_common): Likewise.
28419 2022-08-04 Alan Modra <amodra@gmail.com>
28421 MIPS: Use R_MIPS_REL16 for BFD_RELOC_16
28422 R_MIPS_REL16 isn't a pc-relative reloc as the name might indicate.
28424 * elf64-mips.c (mips_reloc_map): Map BFD_RELOC_16 to R_MIPS_REL16.
28425 * elfn32-mips.c (mips_reloc_map): Likewise.
28427 2022-08-04 GDB Administrator <gdbadmin@sourceware.org>
28429 Automatic date update in version.in
28431 2022-08-03 H.J. Lu <hjl.tools@gmail.com>
28433 elf: Reset alignment for each PT_LOAD segment
28434 Reset alignment for each PT_LOAD segment to avoid using alignment from
28435 the previous PT_LOAD segment.
28440 * elf.c (assign_file_positions_for_load_sections): Reset
28441 alignment for each PT_LOAD segment.
28446 * testsuite/ld-elf/pr29435.d: New file.
28447 * testsuite/ld-elf/pr29435.s: Likewise.
28449 2022-08-03 Tom Tromey <tom@tromey.com>
28451 Use unique_ptr to destroy per-bfd object
28452 In some cases, the objfile owns the per-bfd object. This is yet
28453 another object that can sometimes be destroyed before the registry is
28454 destroyed, possibly reslting in a use-after-free. Also, I noticed
28455 that the condition for deleting the object is not the same as the
28456 condition used to create it -- so it could possibly result in a memory
28457 leak in some situations. This patch fixes the problem by introducing
28458 a new unique_ptr that holds this object when necessary.
28460 Use auto_obstack in objfile
28461 This changes objfile to use an auto_obstack. This helps prevent
28462 use-after-free bugs, because it ensures that anything allocated on the
28463 objfile obstack will live past the point at which the registry object
28466 Use gdb_bfd_ref_ptr in objfile
28467 This changes struct objfile to use a gdb_bfd_ref_ptr. In addition to
28468 removing some manual memory management, this fixes a use-after-free
28469 that was introduced by the registry rewrite series. The issue there
28470 was that, in some cases, registry shutdown could refer to memory that
28471 had already been freed. This help fix the bug by delaying the
28472 destruction of the BFD reference (and thus the per-bfd object) until
28473 after the registry has been shut down.
28475 2022-08-03 Ruud van der Pas <ruud.vanderpas@oracle.com>
28477 gprofng: fix bug 29410 - Argument " 0." isn't numeric in numeric gt (>)
28479 2022-08-02 Ruud van der Pas <ruud.vanderpas@oracle.com>
28482 * gp-display-html/gp-display-html.in: Remove non-breaking spaces.
28484 2022-08-03 Alan Modra <amodra@gmail.com>
28486 Fix a conflict between the linker's need to rename some PE format input libraries and the BFD library's file caching mechanism.
28488 bfd * bfd.c (BFD_CLOSED_BY_CACHE): New bfd flag.
28489 * cache.c (bfd_cache_delete): Set BFD_CLOSED_BY_DELETE on the
28491 (bfd_cache_lookup_worker): Clear BFD_CLOSED_BY_DELETE on the newly
28493 * opncls.c (bfd_set_filename): Refuse to change the name of a bfd
28494 that has been closed by bfd_cache_delete. Mark changed bfds as
28496 * bfd-in2.h: Regenerate.
28498 ld * ldlang.h (lang_input_statement_struct): Add sort_key field.
28499 * emultempl/pe.em (after_open): If multiple import libraries refer
28500 to the same bfd, store their names in the sort_key field.
28501 * emultempl/pep.em (after_open): Likewise.
28502 * ldlang.c (sort_filename): New function. Returns the filename to
28503 be used when sorting input files.
28504 (wild_sort): Use the sort_filename function.
28506 2022-08-03 Enze Li <enze.li@hotmail.com>
28508 gdb/amd64: clean up unused variable
28509 When building with clang 15, I got this,
28512 amd64-tdep.c:1410:13: error: variable 'insn' set but not used[-Werror,-Wunused-but-set-variable]
28513 gdb_byte *insn = insn_details->raw_insn + modrm_offset;
28517 The function that uses this variable has been removed in this commit,
28519 commit 870f88f7551b0f2d6aaaa36fb684b5ff8f468107
28520 Date: Mon Apr 18 13:16:27 2016 -0400
28522 remove trivialy unused variables
28524 Fix this by removing unused variable.
28526 Tested by rebuilding on x86_64-linux with clang 15 and gcc 12.
28528 2022-08-03 Lancelot SIX <lancelot.six@amd.com>
28530 gdb: Fix regression in varobj recreation
28531 Commit bc20e562ec0 "Fix use after free in varobj" introduced a
28532 regression. This commit makes sure that the varobj object does not
28533 keeps stale references to object being freed when we unload an objfile.
28534 This includes the "valid_block" field which is reset to nullptr if the
28535 pointed to block is tied to an objfile being freed.
28537 However, at some point varobj_invalidate_iter might try to recreate
28538 varobjs tracking either floating or globals. Varobj tracking globals
28539 are identified as having the "valid_block" field set nullptr, but as
28540 bc20e562ec0 might clear this field, we have lost the ability to
28541 distinguish between varobj referring to globals and non globals.
28543 Fix this by introducing a "global" flag which tracks if a given varobj
28544 was initially created as tracking a global.
28546 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29426
28548 2022-08-03 Alan Modra <amodra@gmail.com>
28551 All of these buffer overrun tests are better written as a comparison
28552 against size remaining, due to ISO C 9899 standard 6.5.2 para 8
28553 regarding adding a constant to a pointer:
28555 "If both the pointer operand and the result point to elements of the
28556 same array object, or one past the last element of the array object,
28557 the evaluation shall not produce an overflow; otherwise, the behavior
28560 So "ex_dta + 4" might be undefined behaviour, if you interpret "the
28561 array object" in this case to be the malloc'd section contents!
28563 * pei-x86_64.c (pex64_get_unwind_info): Tidy sanity checks.
28564 (pex64_xdata_print_uwd_codes): Likewise.
28566 2022-08-03 Jan Beulich <jbeulich@suse.com>
28568 x86: improve/shorten vector zeroing-idiom optimization conditional
28569 - Drop the rounding type check: We're past template matching, and none
28570 of the involved insns support embedded rounding.
28571 - Drop the extension opcode check: None of the involved opcodes have
28572 variants with it being other than None.
28573 - Instead check opcode space, even if just to be on the safe side going
28575 - Reduce the number of comparisons by folding two groups.
28577 x86: properly mark i386-only insns
28578 Just like all Size64 insns are marked Cpu64, all Size32 insns ought to
28581 x86: also use D for MOVBE
28582 First of all rename the meanwhile misleading Opcode_SIMD_FloatD, as it
28583 has also been used for KMOV* and BNDMOV. Then simplify the condition
28584 selecting which form if "reversing" to use - except for the MOV to/from
28585 control/debug/test registers all extended opcode space insns use bit 0
28586 (rather than bit 1) to indicate the direction (from/to memory) of an
28587 operation. With that, D can simply be set on the first of the two
28588 templates, while the other can be dropped.
28590 2022-08-03 GDB Administrator <gdbadmin@sourceware.org>
28592 Automatic date update in version.in
28594 2022-08-03 Cary Coutant <ccoutant@gmail.com>
28596 Add ELFCOMPRESS_ZSTD.
28598 * common.h: Add ELFCOMPRESS_ZSTD.
28600 2022-08-02 John Baldwin <jhb@FreeBSD.org>
28602 fbsd-nat: Correct the return type of the have_regset method.
28603 During the development of 40c23d880386d6e8202567eaa2a6b041feb1a652,
28604 the return value of fbsd_nat_target::have_regset was changed from a
28605 simple boolean to returning the size of the register set. The
28606 comments and callers were all updated for this change, but the actual
28607 return type was accidentally left as a bool. This change fixes the
28608 return type to be a size_t.
28610 Current callers of this only checked the value against 0 and thus
28611 still worked correctly.
28613 2022-08-02 Jan Beulich <jbeulich@suse.com>
28615 ELF: emit symbol table when there are relocations
28616 Even when there are no symbols (e.g. all relocations being against
28617 absolute values), a symbol table (with just the first placeholder entry)
28618 needs to be emitted. Otherwise tools like objdump won't properly process
28619 the relocations. The respective checks in assign_section_numbers() and
28620 _bfd_elf_compute_section_file_positions() support also this view. Oddly
28621 enough so far HAS_RELOC was only set when reading in an object file, but
28622 not when generating one anew; the flag would only have been cleared when
28623 no relocations were found (anymore).
28625 While there also amend the affected function's leading comment to also
28628 2022-08-02 Matthew Malcomson <hardenedapple@gmail.com>
28630 ld: aarch64: Adjust TLS relaxation condition
28631 In aarch64_tls_transition_without_check and elfNN_aarch64_tls_relax we
28632 choose whether to perform a relaxation to an IE access model or an LE
28633 access model based on whether the symbol itself is marked as local (i.e.
28636 This is problematic in two ways. The first is that sometimes a global
28637 dynamic access can be relaxed to an initial exec access when creating a
28638 shared library, and if that happens on a local symbol then we currently
28639 relax it to a local exec access instead. This usually does not happen
28640 since we only relax an access if aarch64_can_relax_tls returns true and
28641 aarch64_can_relax_tls does not have the same problem. However, it can
28642 happen when we have seen both an IE and GD access on the same symbol.
28643 This case is exercised in the newly added testcase tls-relax-gd-ie-2.
28645 The second problem is that deciding based on whether the symbol is local
28646 misses the case when the symbol is global but is still non-interposable
28647 and known to be located in the executable. This happens on all global
28648 symbols in executables.
28649 This case is exercised in the newly added testcase tls-relax-ie-le-4.
28651 Here we adjust the condition we base our relaxation on so that we relax
28652 to local-exec if we are creating an executable and the relevant symbol
28653 we're accessing is stored inside that executable.
28655 -- Updating tests for new relaxation criteria
28657 Many of the tests added to check our relaxation to IE were implemented
28658 by taking advantage of the fact that we did not relax a global symbol
28659 defined in an executable.
28661 Since a global symbol defined in an executable is still not
28662 interposable, we know that a TLS version of such a symbol will be in the
28663 main TLS block. This means that we can perform a stronger relaxation on
28664 such symbols and relax their accesses to a local-exec access.
28666 Hence we have to update all tests that relied on the older suboptimal
28669 The two cases when we still would want to relax a general dynamic access
28670 to an initial exec one are:
28671 1) When in a shared library and accessing a symbol which we have already
28672 seen accessed with an initial exec access sequence.
28673 2) When in an executable and accessing a symbol defined in a shared
28676 Both of these require shared library support, which means that these
28677 tests are now only available on targets with that.
28679 I have chosen to switch the existing testcases from a plain executable
28680 to one dynamically linked to a shared object as that doesn't require
28681 changing the testcases quite so much (just requires accessing a
28682 different variable rather than requiring adding another code sequence).
28684 The tls-relax-all testcase was an outlier to the above approach, since
28685 it included a general dynamic access to both a local and global symbol
28686 and inspected for the difference accordingly.
28688 2022-08-02 Matthew Malcomson <hardenedapple@gmail.com>
28690 ld: aarch64: Update test linker scripts relocs.ld and relocs-ilp32.ld
28691 The updates are to ensure that the .data section exists. This means
28692 that we always have a data section. That means that we don't create a
28693 RWX segment and avoid the corresponding warning.
28695 We get this warning when testing aarch64-none-elf with -mcmodel=tiny.
28696 N.b. this changes quite a few testcases from fail to pass.
28698 2022-08-02 Victor Do Nascimento <Victor.DoNascimento@arm.com>
28700 arm: Add cfi expression support for ra_auth_code
28701 This patch extends assembler support for the use of register names to
28702 allow for pseudo-registers, e.g. ra_auth_code register.
28703 This is done particularly with CFI directives in mind, allowing for
28704 expressions of the type:
28706 .cfi_register ra_auth_code, 12
28710 * config/tc-arm.c (tc_arm_regname_to_dw2regnum): Add
28711 REG_TYPE_PSEUDO handling.
28712 * testsuite/gas/arm/cfi-pacbti-m-readelf.d: New.
28713 * testsuite/gas/arm/cfi-pacbti-m.s: New.
28715 2022-08-02 Victor Do Nascimento <Victor.DoNascimento@arm.com>
28717 arm: Use DWARF numbering convention for pseudo-register representation
28718 This patch modifies the internal `struct reg_entry' numbering of DWARF
28719 pseudo-registers to match values assigned in DWARF standards (see "4.1
28720 DWARF register names" in [1])so ra_auth_code goes from 12 to 143 and
28721 amends the unwinder .save directive-processing code to correctly handle
28722 mixed register-type save directives.
28724 The mechanism for splitting the register list is also re-written to
28725 comply with register ordering on push statements, being that registers
28726 are stored on the stack in numerical order, with the lowest numbered
28727 register at the lowest address [2].
28729 Consequently, the parsing of the hypothetical directive
28731 .save{r4-r7, r10, ra_auth_core, lr}
28733 has been changed such as rather than producing
28736 .save{ra_auth_code}
28739 as was the case with previous implementation, now produces:
28742 .save{ra_auth_code}
28745 [1] <https://github.com/ARM-software/abi-aa/blob/main/aadwarf32/aadwarf32.rst>
28746 [2] <https://developer.arm.com/documentation/dui0473/j/arm-and-thumb-instructions/push>
28750 * config/tc-arm.c (REG_RA_AUTH_CODE): New.
28751 (parse_dot_save): Likewise.
28752 (parse_reg_list): Remove obsolete code.
28753 (reg_names): Set ra_auth_code to 143.
28754 (s_arm_unwind_save): Handle core and pseudo-register lists via
28756 (s_arm_unwind_save_mixed): Deleted.
28757 (s_arm_unwind_save_pseudo): Handle one register at a time.
28758 * testsuite/gas/arm/unwind-pacbti-m-readelf.d: Fix test.
28759 * testsuite/gas/arm/unwind-pacbti-m.d: Likewise.
28761 2022-08-02 Alan Modra <amodra@gmail.com>
28764 objdump -x on PE executables produces lots of "xdata section corrupt"
28765 and "corrupt unwind data" warnings, and refuses to dump that info. It
28766 turns out that the sanity checks were bad, not the data. Fix them.
28768 * pei-x86_64.c (pex64_get_unwind_info): Correct buffer overrun
28770 (pex64_xdata_print_uwd_codes): Similarly.
28772 2022-08-02 Jan Beulich <jbeulich@suse.com>
28774 x86: XOP shift insns don't really allow B suffix
28775 By mistake it was permitted to be used from the very introduction of XOP
28778 2022-08-02 GDB Administrator <gdbadmin@sourceware.org>
28780 Automatic date update in version.in
28782 2022-08-01 Martin Storsjö <martin@martin.st>
28784 ld: Support the -exclude-symbols option via COFF def files, with the EXCLUDE_SYMBOLS keyword
28785 This was requested in review.
28787 ld: Add support for a new option, -exclude-symbols, in COFF object file directives
28788 This maps to the same as ld's --exclude-symbols command line option,
28789 but allowing specifying the option via directives embedded in the
28790 object files instead of passed manually on the command line.
28792 2022-08-01 Tom de Vries <tdevries@suse.de>
28794 [gdb/symtab] Fix .debug_aranges duplicate offset warning
28795 The function read_addrmap_from_aranges contains code to issue a warning:
28797 if (!insertpair.second)
28799 warning (_("Section .debug_aranges in %s has duplicate "
28800 "debug_info_offset %s, ignoring .debug_aranges."),
28801 objfile_name (objfile), sect_offset_str (per_cu->sect_off));
28805 but the warning is in fact activated when all_comp_units has duplicate
28806 entries, which is very misleading.
28809 - adding a test-case that should trigger the warning,
28810 - replacing the current implementation of the warning with an
28811 assert that all_comp_units should not contain duplicates, and
28812 - properly re-implementing the warning, such that it is triggered
28815 Tested on x86_64-linux.
28817 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29381
28819 2022-08-01 Jan Beulich <jbeulich@suse.com>
28821 x86: SKINIT with operand needs IgnoreSize
28822 Without it in 16-bit mode a pointless operand size prefix would be
28825 2022-08-01 WANG Xuerui <git@xen0n.name>
28827 opcodes: LoongArch: add "ret" instruction to reduce typing
28828 This syntactic sugar is present in both classical and emerging
28829 architectures, like Alpha, SPARC and RISC-V, and assembler macros
28830 doing the same thing can already be found in the wild e.g. [1], proving
28831 the feature's popularity. It's better to provide support directly in the
28832 assembler so downstream users wouldn't have to re-invent this over and
28835 [1]: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/unix/sysv/linux/loongarch/sysdep.h;h=c586df819cd90;hb=HEAD#l28
28837 2022-08-01 WANG Xuerui <git@xen0n.name>
28839 opcodes: LoongArch: make all non-native jumps desugar to canonical b{lt/ge}[u] forms
28840 Also re-order the jump/branch opcodes while at it, so that insns are
28841 sorted in ascending order according to opcodes, and the label form
28842 preceding the real definition.
28844 2022-08-01 Alan Modra <amodra@gmail.com>
28846 Get rid of fprintf_vma and sprintf_vma
28847 These two macros print either a 16 digit hex number or an 8 digit
28848 hex number. Unfortunately they depend on both target and host, which
28849 means that the output for 32-bit targets may be either 8 or 16 hex
28852 Replace them in most cases with code that prints a bfd_vma using
28853 PRIx64. In some cases, deliberately lose the leading zeros.
28854 This change some output, notably in base/offset fields of m68k
28855 disassembly which I think looks better that way, and in error
28856 messages. I've kept leading zeros in symbol dumps (objdump -t)
28857 and in PE header dumps.
28860 * bfd-in.h (fprintf_vma, sprintf_vma, printf_vma): Delete.
28861 * bfd-in2.h: Regenerate.
28862 * bfd.c (bfd_sprintf_vma): Don't use sprintf_vma.
28863 (bfd_fprintf_vma): Don't use fprintf_vma.
28864 * coff-rs6000.c (xcoff_reloc_type_tls): Don't use sprintf_vma.
28865 Instead use PRIx64 to print bfd_vma values.
28866 (xcoff_ppc_relocate_section): Likewise.
28867 * cofflink.c (_bfd_coff_write_global_sym): Likewise.
28868 * mmo.c (mmo_write_symbols_and_terminator): Likewise.
28869 * srec.c (srec_write_symbols): Likewise.
28870 * elf32-xtensa.c (print_r_reloc): Similarly for fprintf_vma.
28871 * pei-x86_64.c (pex64_dump_xdata): Likewise.
28872 (pex64_bfd_print_pdata_section): Likewise.
28873 * som.c (som_print_symbol): Likewise.
28874 * ecoff.c (_bfd_ecoff_print_symbol): Use bfd_fprintf_vma.
28876 * dis-buf.c (perror_memory, generic_print_address): Don't use
28877 sprintf_vma. Instead use PRIx64 to print bfd_vma values.
28878 * i386-dis.c (print_operand_value, print_displacement): Likewise.
28879 * m68k-dis.c (print_base, print_indexed): Likewise.
28880 * ns32k-dis.c (print_insn_arg): Likewise.
28881 * ia64-gen.c (_opcode_int64_low, _opcode_int64_high): Delete.
28882 (opcode_fprintf_vma): Delete.
28883 (print_main_table): Use PRIx64 to print opcode.
28885 * od-macho.c: Replace all uses of printf_vma with bfd_printf_vma.
28886 * objcopy.c (copy_object): Don't use sprintf_vma. Instead use
28887 PRIx64 to print bfd_vma values.
28888 (copy_main): Likewise.
28889 * readelf.c (CHECK_ENTSIZE_VALUES): Likewise.
28890 (dynamic_section_mips_val): Likewise.
28891 (print_vma): Don't use printf_vma. Instead use PRIx64 to print
28893 (dump_ia64_vms_dynamic_fixups): Likewise.
28894 (process_version_sections): Likewise.
28895 * rddbg.c (stab_context): Likewise.
28897 * config/tc-i386.c (offset_in_range): Don't use sprintf_vma.
28898 Instead use PRIx64 to print bfd_vma values.
28899 (md_assemble): Likewise.
28900 * config/tc-mips.c (load_register, macro): Likewise.
28901 * messages.c (as_internal_value_out_of_range): Likewise.
28902 * read.c (emit_expr_with_reloc): Likewise.
28903 * config/tc-ia64.c (note_register_values): Don't use fprintf_vma.
28904 Instead use PRIx64 to print bfd_vma values.
28905 (print_dependency): Likewise.
28906 * listing.c (list_symbol_table): Use bfd_sprintf_vma.
28907 * symbols.c (print_symbol_value_1): Use %p to print pointers.
28908 (print_binary): Likewise.
28909 (print_expr_1): Use PRIx64 to print bfd_vma values.
28910 * write.c (print_fixup): Use %p to print pointers. Don't use
28912 * testsuite/gas/all/overflow.l: Update expected output.
28913 * testsuite/gas/m68k/mcf-mov3q.d: Likewise.
28914 * testsuite/gas/m68k/operands.d: Likewise.
28915 * testsuite/gas/s12z/truncated.d: Likewise.
28917 * deffilep.y (def_file_print): Don't use fprintf_vma. Instead
28918 use PRIx64 to print bfd_vma values.
28919 * emultempl/armelf.em (gld${EMULATION_NAME}_finish): Don't use
28920 sprintf_vma. Instead use PRIx64 to print bfd_vma values.
28921 * emultempl/pe.em (gld${EMULATION_NAME}_finish): Likewise.
28922 * ldlang.c (lang_map): Use %V to print region origin.
28923 (lang_one_common): Don't use sprintf_vma.
28924 * ldmisc.c (vfinfo): Don't use fprintf_vma or sprintf_vma.
28925 * pe-dll.c (pe_dll_generate_def_file): Likewise.
28927 * remote.c (remote_target::trace_set_readonly_regions): Replace
28928 uses of sprintf_vma with bfd_sprintf_vma.
28930 2022-08-01 liuzhensong <liuzhensong@loongson.cn>
28932 LoongArch: Set defaults to exec stack 0.
28934 2022-08-01 Alan Modra <amodra@gmail.com>
28936 libctf: Avoid use of uninitialised variables
28937 * ctf-link.c (ctf_link_add_ctf_internal): Don't free uninitialised
28940 2022-08-01 Alan Modra <amodra@gmail.com>
28942 PR29348, BFD_VMA_FMT wrong
28943 There is a problem with my commit 0e3c1eebb2, which replaced
28944 bfd_uint64_t with uint64_t: Some hosts typedef int64_t to long long
28945 even when long is the same size as long long. That confuses the code
28946 choosing one of "l", "ll", or "I64" for BFD_VMA_FMT, and results in
28949 Write a direct configure test for the printf int64_t style instead.
28950 This removes the last use of BFD_HOST_64BIT_LONG, so delete that.
28951 Note that the changes to configure.com are pure guesswork.
28954 * bfd-in.h (BFD_HOST_64BIT_LONG): Don't define.
28955 (BFD_VMA_FMT): Define using BFD_INT64_FMT when 64-bit.
28956 (bfd_vma, bfd_signed_vma): Move comments to 64-bit typedefs.
28957 * configure.ac (BFD_HOST_64BIT_LONG): Delete.
28958 (BFD_INT64_FMT): New config test.
28959 * configure.com: Update similarly.
28960 * Makefile.in: Regenerate.
28961 * bfd-in2.h: Regenerate.
28962 * configure: Regenerate.
28964 2022-08-01 GDB Administrator <gdbadmin@sourceware.org>
28966 Automatic date update in version.in
28968 2022-07-31 GDB Administrator <gdbadmin@sourceware.org>
28970 Automatic date update in version.in
28972 2022-07-30 Tom de Vries <tdevries@suse.de>
28974 [gdb/testsuite] Fix gdb.ada/literals.exp with aarch64
28975 On aarch64-linux, I run into:
28977 (gdb) print 16#ffffffffffffffff#^M
28978 $7 = 18446744073709551615^M
28979 (gdb) FAIL: gdb.ada/literals.exp: print 16#ffffffffffffffff#
28981 while on x86_64-linux instead, I get:
28983 (gdb) print 16#ffffffffffffffff#^M
28985 (gdb) PASS: gdb.ada/literals.exp: print 16#ffffffffffffffff#
28988 We can easily reproduce this on x86_64-linux using:
28990 $ gdb -q -batch -ex "set lang ada" -ex "set arch i386" \
28991 -ex "print 16#ffffffffffffffff#"
28993 $ gdb -q -batch -ex "set lang ada" -ex "set arch aarch64" \
28994 -ex "print 16#ffffffffffffffff#"
28995 $1 = 18446744073709551615
28998 With i386, we have:
29004 (gdb) p long_long_bits
29007 and so in processInt we hit the fits-in-unsigned-long-long case where we use
29010 /* Note: Interprets ULLONG_MAX as -1. */
29011 yylval.typed_val.type = type_long_long (par_state);
29014 With aarch64, we have instead:
29020 (gdb) p long_long_bits
29023 and so in processInt we hit the fits-in-unsigned-long case where we use
29024 as type unsigned long:
29026 yylval.typed_val.type
29027 = builtin_type (par_state->gdbarch ())->builtin_unsigned_long;
29030 It's not clear why for ada we're using long long for the
29031 fits-in-unsigned-long-long case.
29033 Fix this by using unsigned long long for the fits-in-unsigned-long-long case,
29034 meaning the new reference output is 18446744073709551615 instead of -1.
29036 Tested on x86_64-linux.
29038 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29416
29040 2022-07-30 Simon Marchi <simon.marchi@efficios.com>
29042 gdb/testsuite: add macros test for source files compiled in various ways
29043 Using different ways of passing source file paths to compilers results n
29044 different file and directory paths in the line header. For example:
29051 Because of this, GDB sometimes failed to look up macros. The previous
29052 patch fixed that as much as possible. This patch adds the corresponding
29055 Add both a DWARF assembler-based test and a regular test. The DWARF
29056 assembled-based one tests some hard-coded debug info based on what I
29057 have observed some specific versions of gcc and clang generate. We want
29058 to make sure that GDB keeps handling all these cases correctly, even if
29059 it's not always clear whether they are really valid DWARF. Also, they
29060 will be tested no matter what the current target compiler is for a given
29063 The regular test is compiled using the target compiler, so it may help
29064 find bugs when testing against some other toolchains than what was used
29065 to generate the DWARF assembler-based test.
29067 For the DWARF assembler-based test, add to testsuite/lib/dwarf.exp the
29068 necessary code to generate a DWARF5 .debug_macro section. The design of
29069 the new procs is based on what was done for rnglists and loclists.
29071 To test against a specific compiler one can use this command, for
29074 $ make check TESTS="gdb.base/macro-source-path.exp" RUNTESTFLAGS="CC_FOR_TARGET=clang --target_board unix/gdb:debug_flags=-gdwarf-5"
29076 Change-Id: Iab8da498e57d10cc2a3d09ea136685d9278cfcf6
29078 2022-07-30 Simon Marchi <simon.marchi@polymtl.ca>
29080 gdb: remove code to prepend comp dir in buildsym_compunit::start_subfile
29081 The bit of code removed by this patch was introduced to fix the same
29082 kind of problem that the previous patch fixes. That is, to try to match
29083 existing subfiles when different name forms are used to refer to a same
29086 The thread for the patch that introduced this code is:
29088 https://pi.simark.ca/gdb-patches/45F8CBDF.9090501@hq.tensilica.com/
29090 The important bits are that the compiler produced a compilation unit
29093 DW_AT_name : test.c
29094 DW_AT_comp_dir : /home/maxim/W/BadgerPass/PR_14999
29096 and DWARF v2 line table with:
29098 The Directory Table:
29099 /home/maxim/W/BadgerPass/PR_14999
29101 The File Name Table:
29102 Entry Dir Time Size Name
29103 1 1 1173897037 152 test.c
29105 Because the main symtab was created with only DW_AT_name, it was named
29106 "test.c". And because the path built from the line header contained the
29107 "directory" part, it was "/home/maxim/W/BadgerPass/PR_14999/test.c".
29108 Because of this mismatch, thing didn't work, so they added this code to
29109 prepend the compilation directory to the existing subfile names, so that
29110 this specific case would work.
29112 With the changes done earlier in this series, where subfiles are
29113 identified using the "most complete path possible", this case would be
29114 handled. The main subfile's would be
29115 "/home/maxim/W/BadgerPass/PR_14999/test.c" from the start
29116 (DW_AT_comp_dir + DW_AT_name). It's not so different from some DWARF 5
29117 cases actually, which make the compilation directory explicit in the
29120 I therefore think that this code is no longer needed. It does feel like
29121 a quick hack to make one specific case work, and we have a more general
29122 solution now. Also, this code was introduced to work around a problem
29123 in the DWARF debug info or the DWARF debug info reader. In general, I
29124 think it's preferable for these hacks to be located in the specific
29125 debug info reader code, rather than in the common code.
29127 Even though this code was added to work around a DWARF reader problem,
29128 it's possible that some other debug info reader has started taking
29129 advantage of this code in the mean time. It's very difficult to
29130 know or verify, but I think the likelyhood is quite small, so I'm
29131 proposing to get rid of it to simplify things a little bit.
29133 Change-Id: I710b8ec0d449d1b110d67ddf9fcbdb2b37108306
29135 2022-07-30 Simon Marchi <simon.marchi@efficios.com>
29137 gdb: add "id" fields to identify symtabs and subfiles
29138 Printing macros defined in the main source file doesn't work reliably
29139 using various toolchains, especially when DWARF 5 is used. For example,
29140 using the binaries produced by either of these commands:
29145 GNU ld (GNU Binutils) 2.38
29146 $ gcc test.c -g3 -gdwarf-5
29149 clang version 13.0.1
29150 $ clang test.c -gdwarf-5 -fdebug-macro
29154 $ ./gdb -nx -q --data-directory=data-directory a.out
29156 Temporary breakpoint 1 at 0x111d: file test.c, line 6.
29157 Starting program: /home/simark/build/binutils-gdb-one-target/gdb/a.out
29159 Temporary breakpoint 1, main () at test.c:6
29162 No symbol "ZERO" in current context.
29164 When starting to investigate this (taking the gcc-compiled binary as an
29165 example), we see that GDB fails to look up the appropriate macro scope
29166 when evaluating the expression. While stopped in
29167 macro_lookup_inclusion:
29170 $1 = 0x62100011a980 "test.c"
29171 (top-gdb) p source.filename
29172 $2 = 0x62100011a9a0 "/home/simark/build/binutils-gdb-one-target/gdb/test.c"
29174 `source` is the macro_source_file that we would expect GDB to find.
29175 `name` comes from the symtab::filename field of the symtab we are
29176 stopped in. GDB doesn't find the appropriate macro_source_file because
29177 the name of the macro_source_file doesn't match exactly the name of the
29180 The name of the main symtab comes from the compilation unit's
29181 DW_AT_name, passed to the buildsym_compunit's constructor:
29183 https://gitlab.com/gnutools/binutils-gdb/-/blob/4815d6125ec580cc02a1094d61b8c9d1cc83c0a1/gdb/dwarf2/read.c#L10627-10630
29185 The contents of DW_AT_name, in this case, is "test.c". It is typically
29186 (what I witnessed all compilers do) the same string that was passed to
29187 the compiler on the command-line.
29189 The name of the macro_source_file comes from the line number program
29190 header's file table, from the call to the line_header::file_file_name
29193 https://gitlab.com/gnutools/binutils-gdb/-/blob/4815d6125ec580cc02a1094d61b8c9d1cc83c0a1/gdb/dwarf2/macro.c#L54-65
29195 line_header::file_file_name prepends the directory path that the file
29196 entry refers to, in the file table (if the file name is not already
29197 absolute). In this case, the file name is "test.c", appended to the
29198 directory "/home/simark/build/binutils-gdb-one-target/gdb".
29200 Because the symtab's name is not created the same way as the
29201 macro_source_file's name is created, we get this mismatch. GDB fails to
29202 find the appropriate macro scope for the symtab, and we can't print
29203 macros when stopped in that symtab.
29205 To make this work, we must ensure that paths produced in these two ways
29206 end up identical. This can be tricky because of the different ways a
29207 path can be passed to the compiler by the user.
29209 Another thing to consider is that while the main symtab's name (or
29210 subfile, before it becomes a symtab) is created using DW_AT_name, the
29211 main symtab is also referred to using its entry in the line table
29212 header's file table, when processing the line table. We must therefore
29213 ensure that the same name is produced in both cases, so that a call to
29214 "start_subfile" for the main subfile will correctly find the
29215 already-created subfile, created by buildsym_compunit's constructor. If
29216 we fail to do that, things still often work, because of a fallback: the
29217 watch_main_source_file_lossage method. This method determines that if
29218 the main subfile has no symbols but there exists another subfile with
29219 the same basename (e.g. "test.c") that does have symbols, it's probably
29220 because there was some filename mismatch. So it replaces the main
29221 subfile with that other subfile. I think that heuristic is useful as a
29222 last effort to work around any bug or bad debug info, but I don't think
29223 we should design things such as to rely on it. It's a heuristic, it can
29224 get things wrong. So in my search for a fix, it is important that given
29225 some good debug info, we don't end up relying on that for things to
29228 A first attempt at fixing this was to try to prepend the compilation
29229 directory here or not prepend it there. In practice, because of all the
29230 possible combinations of debug info the compilers produce, it was not
29231 possible to get something that would produce reliable, consistent paths.
29233 Another attempt at fixing this was to make both macro_source_file
29234 objects and symtab objects use the most complete form of path possible.
29235 That means to prepend directories at least until we get an absolute
29236 path. In theory, we should end up with the same path in all cases.
29237 This generally worked, but because it changed the symtab names, it
29238 resulted in user-visible changes (for example, paths to source files in
29239 Breakpoint hit messages becoming always absolute). I didn't find this
29240 very good, first because there is a "set filename-display" setting that
29241 lets the user control how they want the paths to be displayed, and that
29242 would suddenly make this setting completely ineffective (although even
29243 today, it is a bit dependent on the debug info). Second, it would
29244 require a good amount of testsuite tweaks to make tests accept these
29245 suddenly absolute paths.
29247 This new patch is a slight variation of that: it adds a new field called
29248 "filename_for_id" in struct symtab and struct subfile, next to the
29249 existing filename field. The goal is to separate the internal ids used
29250 for finding objects from the names used for presentation. This field is
29251 used for identifying subfiles, symtabs and macro_source_files
29252 internally. For DWARF symtabs, this new field is meant to contain the
29253 "most complete possible" path, as discussed above. So for a given file,
29254 it must always be in the same form, everywhere. The existing
29255 symtab::filename field remains the one used for printing to the user, so
29256 there shouldn't be any change in how paths are printed.
29258 Changes in the core symtab files are:
29260 - Add "name_for_id" and "filename_for_id" fields to "struct subfile"
29261 and "struct symtab", next to existing "name" and "filename" fields.
29262 - Make buildsym_compunit::buildsym_compunit and
29263 buildsym_compunit::start_subfile accept a "name_for_id" parameter
29264 next to the existing "name" ones.
29265 - Make buildsym_compunit::start_subfile use "name_for_id" for looking
29266 up existing subfiles. This is the key thing for making calls
29267 to start_subfile for the main source file look up the existing
29268 subfile successfully, and avoid relying on
29269 watch_main_source_file_lossage.
29270 - Make sal_macro_scope pass "filename_for_id", rather than "filename",
29271 to macro_lookup_inclusion. This is the key thing to making the
29272 lookup work and macro printing work.
29274 Changes in the DWARF files are:
29276 - Make line_header::file_file_name return the "most complete possible"
29277 name. The only pre-existing user of this method is the macro code,
29278 to give the macro_source_file objects their name. And we now want
29279 them to have this "most complete possible" name, which will match the
29280 corresponding symtab's "filename_for_id".
29281 - Make dwarf2_cu::start_compunit_symtab pass the "most complete
29282 possible" name for the main symtab's "filename_for_id". In this
29283 context, where the info comes from the compilation unit's DW_AT_name
29284 / DW_AT_comp_dir, it means prepending DW_AT_comp_dir to DW_AT_name if
29285 DW_AT_name is not already absolute.
29286 - Change dwarf2_start_subfile to build a name_for_id for the subfile
29287 being started. The simplest way is to re-use
29288 line_header::file_file_name, since the callers always have a
29289 file_entry handy. This ensures that it will get the exact same path
29290 representation as the macro code does, for the same file (since it
29291 also uses line_header::file_file_name).
29292 - Update calls to allocate_symtab to pass the "name_for_id" from the
29295 Tests exercising all this are added by the following patch.
29297 Of all the cases I tried, the only one I found that ends up relying on
29298 watch_main_source_file_lossage is the following one:
29301 clang version 13.0.1
29302 Target: x86_64-pc-linux-gnu
29303 Thread model: posix
29304 InstalledDir: /usr/bin
29305 $ clang ./test.c -g3 -O0 -gdwarf-4
29306 $ ./gdb -nx --data-directory=data-directory -q -readnow -iex "set debug symtab-create 1" a.out
29308 [symtab-create] start_subfile: name = test.c, name_for_id = /home/simark/build/binutils-gdb-one-target/gdb/test.c
29309 [symtab-create] start_subfile: name = ./test.c, name_for_id = /home/simark/build/binutils-gdb-one-target/gdb/./test.c
29310 [symtab-create] start_subfile: name = ./test.c, name_for_id = /home/simark/build/binutils-gdb-one-target/gdb/./test.c
29311 [symtab-create] start_subfile: found existing symtab with name_for_id /home/simark/build/binutils-gdb-one-target/gdb/./test.c (/home/simark/build/binutils-gdb-one-target/gdb/./test.c)
29312 [symtab-create] watch_main_source_file_lossage: using subfile ./test.c as the main subfile
29314 As we can see, there are two forms used for "test.c", one with a "." and
29315 one without. This comes from the fact that the compilation unit DIE
29318 DW_AT_name ("test.c")
29319 DW_AT_comp_dir ("/home/simark/build/binutils-gdb-one-target/gdb")
29321 without a ".", and the line table for that file contains:
29323 include_directories[ 1] = "."
29328 When assembling the filename from that entry, we get a ".".
29330 It is a bit unexpected that the main filename resulting from the line
29331 table header does not match exactly the name in the compilation unit.
29332 For instance, gcc uses "./test.c" for the DW_AT_name, which gives
29333 identical paths in the compilation unit and in the line table header.
29335 Similarly, with DWARF 5:
29337 $ clang ./test.c -g3 -O0 -gdwarf-5
29339 clang create two entries that refer to the same file but are of in a different
29342 include_directories[ 0] = "/home/simark/build/binutils-gdb-one-target/gdb"
29343 include_directories[ 1] = "."
29351 The first file name produces a path without a "." while the second does.
29352 This is not caught by watch_main_source_file_lossage, because of
29353 dwarf_decode_lines that creates a symtab for each file entry in the line
29354 table. It therefore appears as "non-empty" to
29355 watch_main_source_file_lossage. This results in two symtabs:
29357 (gdb) maintenance info symtabs
29358 { objfile /home/simark/build/binutils-gdb-one-target/gdb/a.out ((struct objfile *) 0x613000005d00)
29359 { ((struct compunit_symtab *) 0x62100011aca0)
29360 debugformat DWARF 5
29361 producer clang version 13.0.1
29363 dirname /home/simark/build/binutils-gdb-one-target/gdb
29364 blockvector ((struct blockvector *) 0x621000129ec0)
29365 user ((struct compunit_symtab *) (null))
29366 { symtab test.c ((struct symtab *) 0x62100011ad20)
29368 linetable ((struct linetable *) 0x0)
29370 { symtab ./test.c ((struct symtab *) 0x62100011ad60)
29372 linetable ((struct linetable *) 0x621000129ef0)
29377 I am not sure what is the consequence of this, but this is also what
29378 happens before my patch, so I think its acceptable to leave it as-is.
29380 To handle these two cases nicely, I think we will need a function that
29381 removes the unnecessary "." from path names, something that can be done
29384 Finally, I made a change in find_file_and_directory is necessary to
29385 avoid breaking test
29387 gdb.dwarf2/dw2-compdir-oldgcc.exp: info source gcc42
29389 Without that change, we would get:
29392 Current source file is /dir/d/dw2-compdir-oldgcc42.S
29393 Compilation directory is /dir/d
29395 whereas the expected result is:
29398 Current source file is dw2-compdir-oldgcc42.S
29399 Compilation directory is /dir/d
29401 This test was added here:
29403 https://sourceware.org/pipermail/gdb-patches/2012-November/098144.html
29405 Long story short, GCC <= 4.2 apparently had a bug where it would
29406 generate a DW_AT_name with a full path ("/dir/d/dw2-compdir-oldgcc42.S")
29407 and no DW_AT_comp_dir. The line table has one entry with filename
29408 "dw2-compdir-oldgcc42.S", which refers to directory 0. Directory 0
29409 normally refers to the compilation unit's comp dir, but it is
29410 non-existent in this case.
29412 This caused some symtab lookup problems, and to work around them, some
29413 workaround was added, which today reads as:
29415 if (res.get_comp_dir () == nullptr
29416 && producer_is_gcc_lt_4_3 (cu)
29417 && res.get_name () != nullptr
29418 && IS_ABSOLUTE_PATH (res.get_name ()))
29419 res.set_comp_dir (ldirname (res.get_name ()));
29421 Source: https://gitlab.com/gnutools/binutils-gdb/-/blob/6577f365ebdee7dda71cb996efa29d3714cbccd0/gdb/dwarf2/read.c#L9428-9432
29423 It extracts an artificial DW_AT_comp_dir from DW_AT_name, if there is no
29424 DW_AT_comp_dir and DW_AT_name is absolute.
29426 Prior to my patch, a subfile would get created with filename
29427 "/dir/d/dw2-compdir-oldgcc42.S", from DW_AT_name, and another would get
29428 created with filename "dw2-compdir-oldgcc42.S" from the line table's
29429 file table. Then watch_main_source_file_lossage would kick in and merge
29430 them, keeping only the "dw2-compdir-oldgcc42.S" one:
29432 [symtab-create] start_subfile: name = /dir/d/dw2-compdir-oldgcc42.S
29433 [symtab-create] start_subfile: name = dw2-compdir-oldgcc42.S
29434 [symtab-create] start_subfile: name = dw2-compdir-oldgcc42.S
29435 [symtab-create] start_subfile: found existing symtab with name dw2-compdir-oldgcc42.S (dw2-compdir-oldgcc42.S)
29436 [symtab-create] watch_main_source_file_lossage: using subfile dw2-compdir-oldgcc42.S as the main subfile
29438 And so "info source" would show "dw2-compdir-oldgcc42.S" as the
29441 With my patch applied, but without the change in
29442 find_file_and_directory, both DW_AT_name and the line table would try to
29443 start a subfile with the same filename_for_id, and there was no need for
29444 watch_main_source_file_lossage - which is what we want:
29446 [symtab-create] start_subfile: name = /dir/d/dw2-compdir-oldgcc42.S, name_for_id = /dir/d/dw2-compdir-oldgcc42.S
29447 [symtab-create] start_subfile: name = dw2-compdir-oldgcc42.S, name_for_id = /dir/d/dw2-compdir-oldgcc42.S
29448 [symtab-create] start_subfile: found existing symtab with name_for_id /dir/d/dw2-compdir-oldgcc42.S (/dir/d/dw2-compdir-oldgcc42.S)
29449 [symtab-create] start_subfile: name = dw2-compdir-oldgcc42.S, name_for_id = /dir/d/dw2-compdir-oldgcc42.S
29450 [symtab-create] start_subfile: found existing symtab with name_for_id /dir/d/dw2-compdir-oldgcc42.S (/dir/d/dw2-compdir-oldgcc42.S)
29452 But since the one with name == "/dir/d/dw2-compdir-oldgcc42.S", coming
29453 from DW_AT_name, gets created first, it wins, and the symtab ends up
29454 with "/dir/d/dw2-compdir-oldgcc42.S" as the name, "info source" shows
29455 "/dir/d/dw2-compdir-oldgcc42.S" and the test breaks.
29457 This is not wrong per-se, after all DW_AT_name is
29458 "/dir/d/dw2-compdir-oldgcc42.S", so it wouldn't be wrong to report the
29459 current source file as "/dir/d/dw2-compdir-oldgcc42.S". If you compile
29460 a file passing "/an/absolute/path.c", DW_AT_name typically contains (at
29461 least with GCC) "/an/absolute/path.c" and GDB tells you that the source
29462 file is "/an/absolute/path.c". But we can also keep the existing
29463 behavior fairly easily with a little change in find_file_and_directory.
29464 When extracting an artificial DW_AT_comp_dir from DW_AT_name, we now
29465 modify the name to just keep the file part. The result is coherent with
29466 what compilers do when you compile a file by just passing its filename
29469 DW_AT_name ("path.c")
29470 DW_AT_comp_dir ("/home/simark/build/binutils-gdb-one-target/gdb")
29472 With this change, filename_for_id is still the full name,
29473 "/dir/d/dw2-compdir-oldgcc42.S", but the filename of the subfile /
29474 symtab (what ends up shown by "info source") is just
29475 "dw2-compdir-oldgcc42.S", and that makes the test happy.
29477 Change-Id: I8b5cc4bb3052afdb172ee815c051187290566307
29479 2022-07-30 Simon Marchi <simon.marchi@polymtl.ca>
29481 gdb/dwarf: pass a file_entry to line_header::file_file_name
29482 In the following patch, there will be some callers of file_file_name
29483 that will already have access to the file_entry object for which they
29484 want the file name. It would be inefficient to have them pass an index,
29485 only for line_header::file_file_name to re-lookup the same file_entry
29486 object. Change line_header::file_file_name to accept a file_entry
29487 object reference, instead of an index to look up.
29489 I think this change makes sense in any case. Callers that have an index
29490 can first obtain a file_entry using line_header::file_name_at or
29491 line_header::file_names.
29493 When passing a file_entry object, we can assume that the file_entry's
29494 index is valid, unlike when passing an index. So, push the special case
29495 about an invalid index to the sole current caller of file_file_name,
29496 macro_start_file. I think that error belongs there anyway, since it
29497 specifically talks about "bad file number in macro information".
29499 This requires recording the file index in the file_entry structure, so
29502 Change-Id: Ic6e44c407539d92b7863d7ba82405ade17f384ad
29504 2022-07-30 Simon Marchi <simon.marchi@polymtl.ca>
29506 gdb/dwarf: pass compilation directory to line header
29507 The following patch changes line_header::file_file_name to prepend the
29508 compilation directory to the file name, if needed. For that, the line
29509 header needs to know about the compilation directory. Prepare for that
29510 by adding a constructor that takes it as a parameter, and passing the
29511 value down everywhere needed. Add a second constructor for the special
29512 case of building a line_header for doing a hash table lookup, since that
29513 case doesn't require a compilation directory value.
29515 Change-Id: Iba3ba0293e4e2d13a64b257cf9a3094684d54330
29517 2022-07-30 Simon Marchi <simon.marchi@polymtl.ca>
29519 gdb: add debug prints in buildsym.c
29520 Add a few debug prints in buildsym.c that were helpful to me in writing
29523 Change-Id: If10a818feaee3ce1b78a2a254013b62dd578002b
29525 2022-07-30 Simon Marchi <simon.marchi@polymtl.ca>
29527 gdb: introduce symtab_create_debug_printf
29528 Introduce symtab_create_debug_printf and symtab_create_debug_printf_v,
29529 to print the debug messages enabled by "set debug symtab-create".
29531 Change-Id: I442500903f72d4635c2dd9eaef770111f317dc04
29533 2022-07-30 GDB Administrator <gdbadmin@sourceware.org>
29535 Automatic date update in version.in
29537 2022-07-29 Tom de Vries <tdevries@suse.de>
29539 [gdb/testsuite] Fix gdb.ada/convvar_comp.exp with broken debug info
29540 On aarch64-linux I run into this failure with gcc 7.5.0:
29542 (gdb) print $item.started^M
29543 $1 = (-5312, 65535, 4202476)^M
29544 (gdb) FAIL: gdb.ada/convvar_comp.exp: print $item.started
29547 The test-case expects (0, 0, 0), but we're getting another value due to
29548 incorrect location information.
29550 Work around this by:
29551 - first printing the value, and then
29552 - verifying that the convenience variable matches the printed value.
29554 I've verified that the test-case still checks what it should by disabling
29555 the fix from commit cc0e770c0d0 ("memory error printing component of record
29556 from convenience variable") and observing the test-case fail.
29558 Tested on x86_64-linux and aarch64-linux.
29560 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29420
29562 2022-07-29 Alan Modra <amodra@gmail.com>
29564 Re: PR16005, avr linker crash on a particular instruction sequence with --relax
29565 The last patch wasn't so clever. The contents in fact have already
29566 been read, just not cached where relax_delete_bytes expects them.
29567 relax_delete_bytes also modifies relocs and syms, so they should be
29571 * elf32-avr.c (elf32_avr_relax_delete_bytes): Revert last change.
29572 (elf32_avr_relax_section): Cache contents, relocs and syms
29573 before calling relax_delete_bytes.
29575 2022-07-29 Andrew Burgess <aburgess@redhat.com>
29577 libopcodes/aarch64: add support for disassembler styling
29578 This commit enables disassembler styling for AArch64. After this
29579 commit it is possible to have objdump style AArch64 disassembler
29580 output (using --disassembler-color option). Once the required GDB
29581 patches are merged, GDB will also style the disassembler output.
29583 The changes to support styling are mostly split between two files
29584 opcodes/aarch64-dis.c and opcodes/aarch64-opc.c.
29586 The entry point for the AArch64 disassembler can be found in
29587 aarch64-dis.c, this file handles printing the instruction mnemonics,
29588 and assembler directives (e.g. '.byte', '.word', etc). Some operands,
29589 mostly relating to assembler directives are also printed from this
29590 file. This commit changes all of this to pass through suitable
29591 styling information.
29593 However, for most "normal" instructions, the instruction operands are
29594 printed using a two step process. From aarch64-dis.c, in the
29595 print_operands function, the function aarch64_print_operand is called,
29596 this function is in aarch64-opc.c, and converts an instruction operand
29597 into a string. Then, back in print_operands (aarch64-dis.c), the
29598 operand string is printed.
29600 Unfortunately, the string returned by aarch64_print_operand can be
29601 quite complex, it will include syntax elements, like '[' and ']', in
29602 addition to register names and immediate values. In some cases, a
29603 single operand will expand into what will appear (to the user) as
29604 multiple operands separated with a ','.
29606 This makes the task of styling more complex, all these different
29607 components need to by styled differently, so we need to get the
29608 styling information out of aarch64_print_operand in some way.
29610 The solution that I propose here is similar to the solution that I
29611 used for the i386 disassembler.
29613 Currently, aarch64_print_operand uses snprintf to write the operand
29614 text into a buffer provided by the caller.
29616 What I propose is that we pass an extra argument to the
29617 aarch64_print_operand function, this argument will be a structure, the
29618 structure contains a callback function and some state.
29620 When aarch64_print_operand needs to format part of its output this can
29621 be done by using the callback function within the new structure, this
29622 callback returns a string with special embedded markers that indicate
29623 which mode should be used for each piece of text. Back in
29624 aarch64-dis.c we can spot these special style markers and use this to
29625 split the disassembler output up and apply the correct style to each
29628 To make aarch64-opc.c clearer a series of new static functions have
29629 been added, e.g. 'style_reg', 'style_imm', etc. Each of these
29630 functions formats a piece of text in a different style, 'register' and
29631 'immediate' in this case.
29633 Here's an example taken from aarch64-opc.c of the new functions in
29636 snprintf (buf, size, "[%s, %s]!",
29637 style_reg (styler, base),
29638 style_imm (styler, "#%d", opnd->addr.offset.imm));
29640 The aarch64_print_operand function is also called from the assembler
29641 to aid in printing diagnostic messages. Right now I have no plans to
29642 add styling to the assembler output, and so, the callback function
29643 used in the assembler ignores the styling information and just returns
29646 I've used the source files in gas/testsuite/gas/aarch64/ for testing,
29647 and have manually gone through and checked that the styling looks
29648 reasonable, however, I'm not an AArch64 expert, so it is possible that
29649 the odd piece is styled incorrectly. Please point out any mistakes
29652 With objdump disassembler color turned off, there should be no change
29653 in the output after this commit.
29655 2022-07-29 Nick Clifton <nickc@redhat.com>
29657 Stop the linker from complaining about unrecognised DW_FORM-rnglistx and DW_FORM_loclistx format attributes.
29659 * dwarf2.c (read_attribute_value): Handle DW_FORM_rnglistx and
29662 2022-07-29 Alan Modra <amodra@gmail.com>
29664 PR16005, avr linker crash on a particular instruction sequence with --relax
29665 It's possible for relax_delete_bytes to be called with section
29666 contents NULL, as demonstrated by the testcase in this PR.
29669 * elf32-avr.c (elf32_avr_relax_delete_bytes): Get section contents
29670 if not already available.
29672 2022-07-29 Jan Beulich <jbeulich@suse.com>
29674 x86: drop stray NoRex64 from KeyLocker insns
29675 It's entirely unclear why some of the KeyLocker insns had NoRex64 on
29676 them - there's nothing here which could cause emission of REX.W (except
29677 of course a user-specified "rex.w", which we ought to honor anyway).
29679 2022-07-29 Jan Beulich <jbeulich@suse.com>
29681 Arm64: re-work PR gas/27217 fix
29682 The original approach has resulted in anomalies when . is involved in an
29683 operand of one of the affected insns. We cannot leave . unresolved, or
29684 else it'll be resolved at the end of assembly, then pointing to the
29685 address of a section rather than at the insn of interest. Undo part of
29686 the original change and instead check whether a relocation cannot be
29687 omitted in md_apply_fix().
29689 By resolving the expressions again, equates (see the adjustment of the
29690 respective testcase) will now be evaluated, and hence relocations
29691 against absolute addresses be emitted. This ought to be okay as long as
29692 the equates aren't global (and hence can't be overridden). If a need
29693 for such arises, quite likely the only way to address this would be to
29694 invent yet another expression evaluation mode, leaving everything
29695 _except_ . un-evaluated.
29697 There's a further anomaly in how transitive equates are handled. In
29703 add x0, x0, :lo12:x
29706 add x0, x0, :lo12:bar
29708 the first two relocations are now against *ABS*:0x12345678 (as said
29709 above), whereas the latter two relocations would be against x. (Before
29710 the change here, the first two relocations are against x and the latter
29711 two against bar.) But this is an issue seen elsewhere as well, and would
29712 likely require adjustments in the target-independent parts of the
29713 assembler instead of trying to hack around this for every target.
29715 2022-07-29 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
29717 ld: Extend ac_default_ld_warn_rwx_segments to all SPARC targets [PR29411]
29718 As discussed in PR ld/29411, the ld warning
29720 [...] has a LOAD segment with RWX permissions
29722 needs to be disabled on all SPARC targets, not just Solaris/SPARC: the
29723 .plt section is required to be RWX by the 32-bit SPARC ELF psABI and the
29724 64-bit SPARC Compliance Definition 2.4.1. Given that ld only supports
29725 SPARC ELF targets, this patch implements this.
29727 Tested on sparc64-unknown-linux-gnu and sparc-sun-solaris2.11.
29729 2022-07-28 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
29733 * configure.tgt (ac_default_ld_warn_rwx_segments): Extend to all
29734 sparc targets. Expand comment.
29736 2022-07-29 Tom de Vries <tdevries@suse.de>
29738 [gdb/testsuite] Fix gdb.threads/killed-outside.exp on aarch64
29739 On aarch64 (and likewise on arm), I run into:
29741 (gdb) PASS: gdb.threads/killed-outside.exp: get pid of inferior
29742 Executing on target: kill -9 11516 (timeout = 300)
29743 builtin_spawn -ignore SIGHUP kill -9 11516^M
29746 Unable to fetch general registers: No such process.^M
29747 (gdb) [Thread 0xfffff7d511e0 (LWP 11518) exited]^M
29749 Program terminated with signal SIGKILL, Killed.^M
29750 The program no longer exists.^M
29751 FAIL: gdb.threads/killed-outside.exp: prompt after first continue (timeout)
29753 due to a mismatch between the actual "No such process" line and the expected
29756 set no_such_process_msg "Couldn't get registers: No such process\."
29759 Fix this by updating the regexp.
29761 Tested on aarch64-linux, and x86_64-linux.
29763 2022-07-29 Tsukasa OI <research_trasio@irq.a4lg.com>
29765 RISC-V: Add `OP_V' to .insn named opcodes
29766 This commit adds `OP_V' (OP-V: vector instruction opcode for now
29767 ratified `V' extension) to .insn opcode name list. Although vector
29768 instruction encoding is not implemented in `.insn' directive, it will
29769 help future implementation of custom vector `.insn'.
29773 * config/tc-riscv.c (opcode_name_list): Add `OP_V'.
29774 * testsuite/gas/riscv/insn.s: Add testcase.
29775 * testsuite/gas/riscv/insn.d: Likewise.
29776 * testsuite/gas/riscv/insn-dwarf.d: Reflect insn.s update.
29778 2022-07-29 GDB Administrator <gdbadmin@sourceware.org>
29780 Automatic date update in version.in
29782 2022-07-28 Tom Tromey <tom@tromey.com>
29784 Remove some unneeded checks in Guile code
29785 The Guile code generally checks to see if an htab is non-null before
29786 destroying it. However, the registry code already ensures this, so we
29787 can change these checks to asserts and simplify the code a little.
29789 Change registry to use less memory
29790 The registry code creates "registry_data" objects that hold the free
29791 function and the index; then the registry keys refer to this object.
29792 However, only the index is really useful, and now that registries have
29793 a private implementation, just the index can be stored and we can
29794 reduce the memory use of registries a little bit. This also
29795 simplifies the code somewhat.
29797 2022-07-28 Tom Tromey <tom@tromey.com>
29800 This rewrites registry.h, removing all the macros and replacing it
29801 with relatively ordinary template classes. The result is less code
29802 than the previous setup. It replaces large macros with a relatively
29803 straightforward C++ class, and now manages its own cleanup.
29805 The existing type-safe "key" class is replaced with the equivalent
29806 template class. This approach ended up requiring relatively few
29807 changes to the users of the registry code in gdb -- code using the key
29808 system just required a small change to the key's declaration.
29810 All existing users of the old C-like API are now converted to use the
29811 type-safe API. This mostly involved changing explicit deletion
29812 functions to be an operator() in a deleter class.
29814 The old "save/free" two-phase process is removed, and replaced with a
29815 single "free" phase. No existing code used both phases.
29817 The old "free" callbacks took a parameter for the enclosing container
29818 object. However, this wasn't truly needed and is removed here as
29821 2022-07-28 Tom Tromey <tom@tromey.com>
29823 Remove some unused functions from guile code
29824 The guile code has a couple of unused functions that touch on the
29825 registry API. This patch removes them.
29827 2022-07-28 Tom Tromey <tom@tromey.com>
29829 Change allocation of type-copying hash table
29830 When an objfile is destroyed, types that are still in use and
29831 allocated on that objfile are copied. A temporary hash map is created
29832 during this process, and it is allocated on the destroyed objfile's
29833 obstack -- which normally is fine, as that is going to be destroyed
29836 However, this approach requires that the objfile be passed to registry
29837 destruction, and this won't be possible in the rewritten registry.
29838 This patch changes the copied type hash table to simply use the heap
29839 instead. It also removes the 'objfile' parameter from
29840 copy_type_recursive, to make this all more clear.
29842 This patch also fixes an apparent bug in copy_type_recursive.
29843 Previously it was copying the dynamic property list to the dying
29846 - = copy_dynamic_prop_list (&objfile->objfile_obstack,
29848 However I think this is incorrect -- that obstack is about to be
29851 2022-07-28 Tom Tromey <tom@tromey.com>
29853 Change address_space to use new and delete
29854 This changes address_space to use new and delete, and makes some other
29855 small C++-ification changes as well, like changing address_space_num
29858 This patch was needed for the subsequent patch to rewrite the registry
29861 2022-07-28 Simon Farre <simon.farre.cx@gmail.com>
29863 gdb/python: Add BreakpointLocation type
29867 This version addresses the issues pointed out by Tom.
29869 Added nullchecks for Python object creations.
29871 Changed from using PyLong_FromLong to the gdb_py-versions.
29873 Re-factored some code to make it look more cohesive.
29875 Also added the more safe Python reference count decrement PY_XDECREF,
29876 even though the BreakpointLocation type is never instantiated by the
29877 user (explicitly documented in the docs) decrementing < 0 is made
29878 impossible with the safe call.
29880 Tom pointed out that using the policy class explicitly to decrement a
29881 reference counted object was not the way to go, so this has instead been
29882 wrapped in a ref_ptr that handles that for us in blocpy_dealloc.
29884 Moved macro from py-internal to py-breakpoint.c.
29886 Renamed section at the bottom of commit message "Patch Description".
29889 This version addresses the points Pedro gave in review to this patch.
29891 Added the attributes `function`, `fullname` and `thread_groups`
29892 as per request by Pedro with the argument that it more resembles the
29893 output of the MI-command "-break-list". Added documentation for these attributes.
29895 Cleaned up left overs from copy+paste in test suite, removed hard coding
29896 of line numbers where possible.
29898 Refactored some code to use more c++-y style range for loops
29899 wrt to breakpoint locations.
29901 Changed terminology, naming was very inconsistent. Used a variety of "parent",
29902 "owner". Now "owner" is the only term used, and the field in the
29903 gdb_breakpoint_location_object now also called "owner".
29907 Changes in response to review by Tom Tromey:
29908 - Replaced manual INCREF/DECREF calls with
29909 gdbpy_ref ptrs in places where possible.
29910 - Fixed non-gdb style conforming formatting
29911 - Get parent of bploc increases ref count of parent.
29912 - moved bploc Python definition to py-breakpoint.c
29914 The INCREF of self in bppy_get_locations is due
29915 to the individual locations holding a reference to
29916 it's owner. This is decremented at de-alloc time.
29918 The reason why this needs to be here is, if the user writes
29921 py loc = gdb.breakpoints()[X].locations[Y]
29923 The breakpoint owner object is immediately going
29924 out of scope (GC'd/dealloced), and the location
29925 object requires it to be alive for as long as it is alive.
29927 Thanks for your review, Tom!
29930 Fixed remaining doc issues as per request
29934 Rewritten commit message, shortened + reworded,
29939 Currently, the Python API lacks the ability to
29940 query breakpoints for their installed locations,
29941 and subsequently, can't query any information about them, or
29942 enable/disable individual locations.
29944 This patch solves this by adding Python type gdb.BreakpointLocation.
29945 The type is never instantiated by the user of the Python API directly,
29946 but is produced by the gdb.Breakpoint.locations attribute returning
29947 a list of gdb.BreakpointLocation.
29949 gdb.Breakpoint.locations:
29950 The attribute for retrieving the currently installed breakpoint
29951 locations for gdb.Breakpoint. Matches behavior of
29952 the "info breakpoints" command in that it only
29953 returns the last known or currently inserted breakpoint locations.
29955 BreakpointLocation contains 7 attributes
29957 6 read-only attributes:
29958 owner: location owner's Python companion object
29959 source: file path and line number tuple: (string, long) / None
29960 address: installed address of the location
29961 function: function name where location was set
29962 fullname: fullname where location was set
29963 thread_groups: thread groups (inferiors) where location was set.
29965 1 writeable attribute:
29966 enabled: get/set enable/disable this location (bool)
29968 Access/calls to these, can all throw Python exceptions (documented in
29969 the online documentation), and that's due to the nature
29970 of how breakpoint locations can be invalidated
29971 "behind the scenes", either by them being removed
29972 from the original breakpoint or changed,
29973 like for instance when a new symbol file is loaded, at
29974 which point all breakpoint locations are re-created by GDB.
29975 Therefore this patch has chosen to be non-intrusive:
29976 it's up to the Python user to re-request the locations if
29977 they become invalid.
29979 Also there's event handlers that handle new object files etc, if a Python
29980 user is storing breakpoint locations in some larger state they've
29981 built up, refreshing the locations is easy and it only comes
29982 with runtime overhead when the Python user wants to use them.
29984 gdb.BreakpointLocation Python type
29985 struct "gdbpy_breakpoint_location_object" is found in python-internal.h
29987 Its definition, layout, methods and functions
29988 are found in the same file as gdb.Breakpoint (py-breakpoint.c)
29990 1 change was also made to breakpoint.h/c to make it possible
29991 to enable and disable a bp_location* specifically,
29992 without having its LOC_NUM, as this number
29993 also can change arbitrarily behind the scenes.
29995 Updated docs & news file as per request.
29997 Testsuite: tests the .source attribute and the disabling of
29998 individual locations.
30000 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=18385
30003 Change-Id: I302c1c50a557ad59d5d18c88ca19014731d736b0
30005 2022-07-28 yaowenbin <yaowenbin1@huawei.com>
30007 gdb/gdb_mbuild.sh: use return instead of continue to avoid shellcheck error
30010 In gdb_mbuild.sh line 174:
30012 ^------^ SC2104 (error): In functions, use return instead of continue.
30014 Change-Id: I5ce95b01359c5cfbb1612f2f48b80bfeea66c96c
30016 2022-07-28 GDB Administrator <gdbadmin@sourceware.org>
30018 Automatic date update in version.in
30020 2022-07-27 GDB Administrator <gdbadmin@sourceware.org>
30022 Automatic date update in version.in
30024 2022-07-27 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
30026 gprofng: check for the makeinfo version
30028 2022-07-25 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
30031 * configure.ac: Check for the makeinfo version.
30032 * configure: Rebuild.
30034 2022-07-26 Keith Seitz <keiths@redhat.com>
30036 gdb/linux_nat: Write memory using ptrace if /proc/pid/mem is not writable
30037 Commit 05c06f318fd9a112529dfc313e6512b399a645e4 enabled GDB to access
30038 memory while threads are running. It did this by accessing
30039 /proc/PID/task/LWP/mem.
30041 Unfortunately, this interface is not implemented for writing in older
30042 kernels (such as RHEL6). This means that GDB is unable to insert
30043 breakpoints on these hosts:
30045 $ ./gdb -q gdb -ex start
30046 Reading symbols from gdb...
30047 Temporary breakpoint 1 at 0x40fdd5: file ../../src/gdb/gdb.c, line 28.
30048 Starting program: /home/rhel6/fsf/linux/gdb/gdb
30050 Cannot insert breakpoint 1.
30051 Cannot access memory at address 0x40fdd5
30055 Before this patch, linux_proc_xfer_memory_partial (previously called
30056 linux_proc_xfer_partial) would return TARGET_XFER_EOF if the write to
30057 /proc/PID/mem failed. [More specifically, linux_proc_xfer_partial
30058 would not "bother for one word," but the effect is the essentially
30061 This status was checked by linux_nat_target::xfer_partial, which would
30062 then fallback to using ptrace to perform the operation.
30064 This is the specific hunk that removed the fallback:
30066 - xfer = linux_proc_xfer_partial (object, annex, readbuf, writebuf,
30067 - offset, len, xfered_len);
30068 - if (xfer != TARGET_XFER_EOF)
30070 + return linux_proc_xfer_memory_partial (readbuf, writebuf,
30071 + offset, len, xfered_len);
30074 return inf_ptrace_target::xfer_partial (object, annex, readbuf, writebuf,
30075 offset, len, xfered_len);
30077 This patch makes linux_nat_target::xfer_partial go straight to writing
30078 memory via ptrace if writing via /proc/pid/mem is not possible in the
30079 running kernel, enabling GDB to insert breakpoints on these older
30080 kernels. Note that a recent patch changed the return status from
30081 TARGET_XFER_EOF to TARGET_XFER_E_IO.
30083 Tested on {unix,native-gdbserver,native-extended-gdbserver}/-m{32,64}
30084 on x86_64, s390x, aarch64, and ppc64le.
30086 Change-Id: If1d884278e8c4ea71d8836bedd56e6a6c242a415
30088 2022-07-26 Pedro Alves <pedro@palves.net>
30090 gdb/linux-nat: Check whether /proc/pid/mem is writable
30091 Probe whether /proc/pid/mem is writable, by using it to write to a GDB
30092 variable. This will be used in the following patch to avoid falling
30093 back to writing to inferior memory with ptrace if /proc/pid/mem _is_
30096 Change-Id: If87eff0b46cbe5e32a583e2977a9e17d29d0ed3e
30098 2022-07-26 Tiezhu Yang <yangtiezhu@loongson.cn>
30100 gdb: LoongArch: Handle the function return value
30101 According to LoongArch ELF ABI specification [1], handle the function
30102 return value of various types.
30104 [1] https://loongson.github.io/LoongArch-Documentation/LoongArch-ELF-ABI-EN.html#_return_values
30106 2022-07-26 Tiezhu Yang <yangtiezhu@loongson.cn>
30108 gdb: LoongArch: Fix code style issues
30109 Fix some code style issues suggested by Tom Tromey and Andrew Burgess,
30112 (1) Put an introductory comment to explain the purpose for some functions.
30114 (2) Modify the the attribute code to make it portable.
30116 (3) Remove globals and pass pointers to locals.
30118 (4) Remove "*" in the subsequent comment lines.
30120 (5) Put two spaces before "{" and "}".
30122 2022-07-26 Nick Clifton <nickc@redhat.com>
30124 Stop the linker from complaining about RWX segments in sparc-solaris targets.
30126 * configure.tgt (ac_default_ld_warn_rwx_segments): Disable for
30127 sparc-solaris configurations.
30129 2022-07-26 Tom de Vries <tdevries@suse.de>
30131 [gdb/testsuite] Fix gdb.opt/inline-small-func.exp with clang
30132 When running test-case gdb.opt/inline-small-func.exp with clang 12.0.1, I run
30135 gdb compile failed, /usr/bin/ld: inline-small-func0.o: in function `main':
30136 inline-small-func.c:21: undefined reference to `callee'
30137 clang-12.0: error: linker command failed with exit code 1 \
30138 (use -v to see invocation)
30139 UNTESTED: gdb.opt/inline-small-func.exp: failed to prepare
30142 Fix this by using __attribute__((always_inline)).
30144 Tested on x86_64-linux.
30146 2022-07-26 Alan Modra <amodra@gmail.com>
30148 PowerPC32 ld test fails with --enable-targets=all
30149 Three pppc32 ld tests fail when spe support is included in the linker
30150 due to this snippet in ld/emulparams/elf32ppc.sh.
30152 if grep -q 'ld_elf32_spu_emulation' ldemul-list.h; then
30153 DATA_START_SYMBOLS="${RELOCATING+*crt1.o(.data .data.* .gnu.linkonce.d.*)
30154 PROVIDE (__spe_handle = .);
30156 . += 4 * (DEFINED (__spe_handle) || . != 0);}"
30159 * testsuite/ld-powerpc/tlsexe32.r: Pass with .data section present.
30160 * testsuite/ld-powerpc/tlsexe32no.r: Likewise.
30161 * testsuite/ld-powerpc/tlsso32.r: Likewise.
30163 2022-07-26 Enze Li <enze.li@hotmail.com>
30165 gdb/hurd: pass memory_tagged as false to find_memory_region_ftype
30166 I tried building GDB on GNU/Hurd, and ran into this error:
30169 gnu-nat.c: In member function ‘virtual int gnu_nat_target::find_memory_regions(find_memory_region_ftype, void*)’:
30170 gnu-nat.c:2620:21: error: too few arguments to function
30171 2620 | (*func) (last_region_address,
30172 | ~~~~~~~~^~~~~~~~~~~~~~~~~~~~~
30173 2621 | last_region_end - last_region_address,
30174 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
30175 2622 | last_protection & VM_PROT_READ,
30176 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
30177 2623 | last_protection & VM_PROT_WRITE,
30178 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
30179 2624 | last_protection & VM_PROT_EXECUTE,
30180 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
30181 2625 | 1, /* MODIFIED is unknown, pass it as true. */
30182 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
30185 gnu-nat.c:2635:13: error: too few arguments to function
30186 2635 | (*func) (last_region_address, last_region_end - last_region_address,
30187 | ~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
30188 2636 | last_protection & VM_PROT_READ,
30189 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
30190 2637 | last_protection & VM_PROT_WRITE,
30191 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
30192 2638 | last_protection & VM_PROT_EXECUTE,
30193 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
30194 2639 | 1, /* MODIFIED is unknown, pass it as true. */
30195 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
30198 make[2]: *** [Makefile:1926: gnu-nat.o] Error 1
30200 This is because in this commit:
30202 commit 68cffbbd4406b4efe1aa6e18460b1d7ca02549f1
30203 Date: Thu Mar 31 11:42:35 2022 +0100
30205 [AArch64] MTE corefile support
30207 Added a new argument to find_memory_region_ftype, but did not pass it to
30208 the function in gnu-nat.c. Fix this by passing memory_tagged as false.
30210 As Luis pointed out, similar bugs may also appear on FreeBSD and NetBSD,
30211 and I have reproduced them on both systems. This patch fixes them
30214 Tested by rebuilding on GNU/Hurd, FreeBSD/amd64 and NetBSD/amd64.
30216 2022-07-26 Enze Li <enze.li@hotmail.com>
30218 gdb/netbsd: add missing header file
30219 I ran into this error when building GDB on NetBSD:
30222 netbsd-nat.c: In member function 'virtual bool nbsd_nat_target::info_proc(const char*, info_proc_what)':
30223 netbsd-nat.c:314:3: error: 'gdb_argv' was not declared in this scope
30224 gdb_argv built_argv (args);
30226 netbsd-nat.c:314:3: note: suggested alternative: 'gdbarch'
30227 gdb_argv built_argv (args);
30230 netbsd-nat.c:315:7: error: 'built_argv' was not declared in this scope
30231 if (built_argv.count () == 0)
30233 netbsd-nat.c:315:7: note: suggested alternative: 'buildargv'
30234 if (built_argv.count () == 0)
30237 gmake[2]: *** [Makefile:1893: netbsd-nat.o] Error 1
30239 Fix this by adding the missing header file, as it is obvious.
30241 Tested by rebuilding on NetBSD/amd64.
30243 2022-07-26 Nick Clifton <nickc@redhat.com>
30245 Updated translations for various sub-directories
30247 2022-07-26 Andrew Burgess <aburgess@redhat.com>
30249 gdb: rename gdbarch_tdep struct to fix g++ 4.8 build
30252 commit 08106042d9f5fdff60c129bf33190639f1a98b2a
30253 Date: Thu May 19 13:20:17 2022 +0100
30255 gdb: move the type cast into gdbarch_tdep
30257 GDB would no longer build using g++ 4.8. The issue appears to be some
30258 confusion caused by GDB having 'struct gdbarch_tdep', but also a
30259 templated function called 'gdbarch_tdep'. Prior to the above commit
30260 the gdbarch_tdep function was not templated, and this compiled just
30261 fine. Note that the above commit compiles just fine with later
30262 versions of g++, so this issue was clearly fixed at some point, though
30263 I've not tried to track down exactly when.
30265 In this commit I propose to fix the g++ 4.8 build problem by renaming
30266 'struct gdbarch_tdep' to 'struct gdbarch_tdep_base'. This rename
30267 better represents that the struct is only ever used as a base class,
30268 and removes the overloading of the name, which allows GDB to build
30271 I've also updated the comment on 'struct gdbarch_tdep_base' to fix a
30272 typo, and the comment on the 'gdbarch_tdep' function, to mention that
30273 in maintainer mode a run-time type check is performed.
30275 2022-07-26 Nick Clifton <nickc@redhat.com>
30277 Fix indentation in loongarch code, preventing a compile time warning.
30279 2022-07-26 Lancelot SIX <lancelot.six@amd.com>
30281 gdb/varobj: Fix varobj_invalidate_iter
30282 The varobj_invalidate function is meant to be called when restarting a
30283 process, and check at this point if some of the previously existing
30284 varobj can be recreated in the context of the new process.
30286 Two kind of varobj are subject to re-creation: global varobj (i.e.
30287 varobj which reference a global variable), and floating varobj (i.e.
30288 varobj which are always re-evaluated in the context of whatever is
30289 the currently selected frame at the time of evaluation).
30291 However, in the re-creation process, the varobj_invalidate_iter
30292 recreates floating varobj as non-floating, due to an invalid parameter.
30293 This patches fixes this and adds an assertion to check that if a varobj
30294 is indeed recreated, it matches the original varobj "floating" property.
30296 Another issue is that if at this recreation time the expression watched
30297 by the floating varobj is not in scope, then the varobj is marked as
30298 invalid. If later the user selects a frame where the expression becomes
30299 valid, the varobj remains invalid and this is wrong. This patch also
30300 make sure that floating varobj are not invalidated if they cannot be
30303 The last important thing to note is that due to the previous patch, when
30304 varobj_invalidate is executed (in the context of a new process), any
30305 global var have already been invalidated (this has been done when the
30306 objfile it referred to got invalidated). As a consequence,
30307 varobj_invalidate tries to recreate vars which are already marked as
30308 invalid. This does not entirely feels right, but I keep this behavior
30309 for backward compatibility.
30311 Tested on x86_64-linux
30313 2022-07-26 Lancelot SIX <lancelot.six@amd.com>
30315 gdb/varobj: Fix use after free in varobj
30316 Varobj object contains references to types, variables (i.e. struct
30317 variable) and expression. All of those can reference data on an
30318 objfile's obstack. It is possible for this objfile to be deleted (and
30319 the obstack to be feed), while the varobj remains valid. Later, if the
30320 user uses the varobj, this will result in a use-after-free error. With
30321 address sanitizer build, this leads to a plain error. For non address
30322 sanitizer build we might see undefined behaviour, which manifest
30323 themself as assertion failures when accessing data backed by feed
30326 This can be observed if we create a varobj that refers to ta symbol in a
30327 shared library, after either the objfile gets reloaded (using the `file`
30328 command) or after the shared library is unloaded (with a call to dlclose
30331 This patch fixes those issues by:
30333 - Adding cleanup procedure to the free_objfile observable. When
30334 activated this observer clears expressions referencing the objfile
30335 being freed, and removes references to blocks belonging to this
30337 - Adding varobj support in the `preserve_values` (gdb.value.c). This
30338 ensures that before the objfile is unloaded, any type owned by the
30339 objfile referenced by the varobj is replaced by an equivalent type
30340 not owned by the objfile. This process is done here instead of in the
30341 free_objfile observer in order to reuse the type hash table already
30342 used for similar purpose when replacing types of values kept in the
30345 This patch also makes sure to keep a reference to the expression's
30346 gdbarch and language_defn members when the varobj->root->exp is
30347 initialized. Those structures outlive the objfile, so this is safe.
30348 This is done because those references might be used initialize a python
30349 context even after exp is invalidated. Another approach could have been
30350 to initialize the python context with default gdbarch and language_defn
30351 (i.e. nullptr) if expr is NULL, but since we might still try to display
30352 the value which was obtained by evaluating exp when it was still valid,
30353 keeping track of the context which was used at this time seems
30356 Tested on x86_64-Linux.
30358 Co-Authored-By: Pedro Alves <pedro@palves.net>
30360 2022-07-26 Pedro Alves <pedro@palves.net>
30362 MI: mi_runto -pending
30363 With the CLI testsuite's runto proc, we can pass "allow-pending" as an
30366 runto func allow-pending
30368 That is currently not possible with MI's mi_runto, however. This
30369 patch makes it possible, by adding a new "-pending" option to
30372 A pending breakpoint shows different MI attributes compared to a
30373 breakpoint with a location, so the regexp returned by
30374 mi_make_breakpoint isn't suitable. Thus, add a new
30375 mi_make_breakpoint_pending proc for pending breakpoints.
30377 Tweak mi_runto to let it take and pass down arguments.
30379 Change-Id: I185fef00ab545a1df2ce12b4dbc3da908783a37c
30381 2022-07-26 GDB Administrator <gdbadmin@sourceware.org>
30383 Automatic date update in version.in
30385 2022-07-25 Ruud van der Pas <ruud.vanderpas@oracle.com>
30387 gprofng: fix bug 29356 - Execution fails if gprofng is not included in PATH
30389 2022-07-22 Ruud van der Pas <ruud.vanderpas@oracle.com>
30392 * gp-display-html/gp-display-html.in: fixed a problem to execute
30393 gp-display-text in case gprofng is not included in the search
30396 2022-07-25 Ruud van der Pas <ruud.vanderpas@oracle.com>
30398 gprofng: fix bug 29392 - Unexpected line format in summary file
30400 2022-07-22 Ruud van der Pas <ruud.vanderpas@oracle.com>
30403 * gp-display-html/gp-display-html.in: modified a regex, plus the
30404 code to handle the results; renamed a variable to improve the
30405 consistency in naming.
30407 2022-07-25 Ruud van der Pas <ruud.vanderpas@oracle.com>
30409 gprofng: fix bug 29353 - Fix a lay-out issue in the html disassembly files
30411 2022-07-22 Ruud van der Pas <ruud.vanderpas@oracle.com>
30414 * gp-display-html/gp-display-html.in: fixed a problem in the
30415 generation of html for the disassembly where instructions
30416 without arguments were not handled correctly.
30418 2022-07-25 Ruud van der Pas <ruud.vanderpas@oracle.com>
30420 gprofng: fix bug 29352 - Fix the message Hexadecimal number > 0xffffffff non-portable
30422 2022-07-22 Ruud van der Pas <ruud.vanderpas@oracle.com>
30425 * gp-display-html/gp-display-html.in: the hex subroutine from
30426 the bigint module is now used.
30428 2022-07-25 Ruud van der Pas <ruud.vanderpas@oracle.com>
30430 gprofng: fix bug 29351 - Move dynamic loading of modules to a later stage
30432 2022-07-22 Ruud van der Pas <ruud.vanderpas@oracle.com>
30435 * gp-display-html/gp-display-html.in: the dynamic loading of
30436 modules occurred too early, resulting in the generation of the
30437 man page to fail in case a module is missing; the loading part is
30438 now done somewhat later in the execution to avoid this problem.
30440 2022-07-25 Kevin Buettner <kevinb@redhat.com>
30442 set/show python dont-write-bytecode fixes
30443 GDB uses the environment variable PYTHONDONTWRITEBYTECODE to
30444 determine whether or not to write the result of byte-compiling
30445 python modules when the "python dont-write-bytecode" setting
30446 is "auto". Simon noticed that GDB's implementation doesn't
30447 follow the Python documentation.
30449 At present, GDB only checks for the existence of this environment
30450 variable. That is not sufficient though. Regarding
30451 PYTHONDONTWRITEBYTECODE, this document...
30453 https://docs.python.org/3/using/cmdline.html
30457 If this is set to a non-empty string, Python won't try to write
30458 .pyc files on the import of source modules.
30460 This commit fixes GDB's handling of PYTHONDONTWRITEBYTECODE by adding
30461 an empty string check.
30463 This commit also corrects the set/show command documentation for
30464 "python dont-write-bytecode". The current doc was just a copy
30465 of that for set/show python ignore-environment.
30467 During his review of an earlier version of this patch, Eli Zaretskii
30468 asked that the help text that I proposed for "set/show python
30469 dont-write-bytecode" be expanded. I've done that in addition to
30470 clarifying the documentation of this option in the GDB manual.
30472 2022-07-25 Andrew Burgess <aburgess@redhat.com>
30474 gdb/python: fix invalid use disassemble_info::stream
30477 commit 81384924cdcc9eb2676dd9084b76845d7d0e0759
30478 Date: Tue Apr 5 11:06:16 2022 +0100
30480 gdb: have gdb_disassemble_info carry 'this' in its stream pointer
30482 The disassemble_info::stream field will no longer be a ui_file*. That
30483 commit failed to update one location in py-disasm.c though.
30485 While running some tests using the Python disassembler API, I
30486 triggered a call to gdbpy_disassembler::print_address_func, and, as I
30487 had compiled GDB with the undefined behaviour sanitizer, GDB crashed
30488 as the code currently (incorrectly) casts the stream field to be a
30491 In this commit I fix this error.
30493 In order to test this case I had to tweak the existing test case a
30494 little. I also spotted some debug printf statements in py-disasm.py,
30495 which I have removed.
30497 2022-07-25 Andrew Burgess <aburgess@redhat.com>
30499 gdb: fix use of uninitialised gdb_printing_disassembler::m_in_comment
30500 Simon pointed out that gdb_printing_disassembler::m_in_comment can be
30501 used uninitialised by the Python disassembler API code. This issue
30502 was spotted when GDB was built with the undefined behaviour sanitizer,
30503 and causes the gdb.python/py-disasm.exp test to fail like this:
30505 (gdb) PASS: gdb.python/py-disasm.exp: global_disassembler=GlobalPreInfoDisassembler: python add_global_disassembler(GlobalPreInfoDisassembler)
30507 Dump of assembler code for function main:
30508 0x0000555555555119 <+0>: push %rbp
30509 0x000055555555511a <+1>: mov %rsp,%rbp
30510 0x000055555555511d <+4>: nop
30511 /home/user/src/binutils-gdb/gdb/disasm.h:144:12: runtime error: load of value 118, which is not a valid value for type 'bool'
30513 The problem is that in disasmpy_builtin_disassemble we create a new
30514 instance of gdbpy_disassembler, which is a sub-class of
30515 gdb_printing_disassembler, however, the m_in_comment field is never
30518 This commit fixes the issue by providing a default initialisation
30519 value for m_in_comment in disasm.h. As we only ever disassemble a
30520 single instruction in disasmpy_builtin_disassemble then we don't need
30521 to worry about reseting m_in_comment back to false after the single
30522 instruction has been disassembled.
30524 With this commit the above issue is resolved and
30525 gdb.python/py-disasm.exp now passes.
30527 2022-07-25 H.J. Lu <hjl.tools@gmail.com>
30529 ld: Compile 2 CTF tests with -O2
30530 When GCC 12 is used to build binutils with -O0, the following 2 tests
30533 FAIL: Conflicted data syms, partially indexed, stripped, with variables
30534 FAIL: Conflicted data syms, partially indexed, stripped
30536 Compile 2 tests with -O2 to avoid test failures.
30539 * testsuite/ld-ctf/data-func-conflicted-vars.d: Compile with -O2.
30540 * testsuite/ld-ctf/data-func-conflicted.d: Likewise.
30542 2022-07-25 Pedro Alves <pedro@palves.net>
30544 struct packed: Add fallback byte array implementation
30545 Attribute gcc_struct is not implemented in Clang targeting Windows, so
30546 add a fallback standard-conforming implementation based on arrays.
30548 I ran the testsuite on x86_64 GNU/Linux with this implementation
30549 forced, and saw no regressions.
30551 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29373
30553 Change-Id: I023315ee03622c59c397bf4affc0b68179c32374
30555 2022-07-25 Pedro Alves <pedro@palves.net>
30557 struct packed: Unit tests and more operators
30558 For PR gdb/29373, I wrote an alternative implementation of struct
30559 packed that uses a gdb_byte array for internal representation, needed
30560 for mingw+clang. While adding that, I wrote some unit tests to make
30561 sure both implementations behave the same. While at it, I implemented
30562 all relational operators. This commit adds said unit tests and
30563 relational operators. The alternative gdb_byte array implementation
30566 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29373
30568 Change-Id: I023315ee03622c59c397bf4affc0b68179c32374
30570 2022-07-25 Pedro Alves <pedro@palves.net>
30572 struct packed: Use gcc_struct on Windows
30573 Building GDB on mingw/gcc hosts is currently broken, due to a static
30574 assertion failure in gdbsupport/packed.h:
30576 In file included from ../../../../../binutils-gdb/gdb/../gdbsupport/common-defs.h:201,
30577 from ../../../../../binutils-gdb/gdb/defs.h:28,
30578 from ../../../../../binutils-gdb/gdb/dwarf2/read.c:31:
30579 ../../../../../binutils-gdb/gdb/../gdbsupport/packed.h: In instantiation of 'packed<T, Bytes>::packed(T) [with T = dwarf_unit_type; long long unsigned int Bytes = 1]':
30580 ../../../../../binutils-gdb/gdb/dwarf2/read.h:181:74: required from here
30581 ../../../../../binutils-gdb/gdb/../gdbsupport/packed.h:41:40: error: static assertion failed
30582 41 | gdb_static_assert (sizeof (packed) == Bytes);
30583 | ~~~~~~~~~~~~~~~~^~~~~~~~
30584 ../../../../../binutils-gdb/gdb/../gdbsupport/gdb_assert.h:27:48: note: in definition of macro 'gdb_static_assert'
30585 27 | #define gdb_static_assert(expr) static_assert (expr, "")
30587 ../../../../../binutils-gdb/gdb/../gdbsupport/packed.h:41:40: note: the comparison reduces to '(4 == 1)'
30588 41 | gdb_static_assert (sizeof (packed) == Bytes);
30589 | ~~~~~~~~~~~~~~~~^~~~~~~~
30592 The issue is that mingw gcc defaults to "-mms-bitfields", which
30593 affects how bitfields are laid out. We can however tell GCC that we
30594 want the regular GCC layout instead using attribute gcc_struct.
30596 Attribute gcc_struct is not implemented in "clang -target
30597 x86_64-pc-windows-gnu", so that will need a different fix.
30599 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29373
30601 Change-Id: I023315ee03622c59c397bf4affc0b68179c32374
30603 2022-07-25 Andrew Burgess <aburgess@redhat.com>
30605 binutils-gdb/git: highlight whitespace errors in source files
30606 For a long time I've had this in my ~/.gitconfig:
30609 whitespace = space-before-tab,indent-with-non-tab,trailing-space
30611 which causes git to show me if I muck up and use spaces instead of
30612 tabs, or leave in trailing whitespace. I find this really useful.
30614 I recently proposed adding something like this to the .gitattributes
30615 files for the GDB sub-directories (gdb, gdbsupport, and gdbserver)[1],
30616 however, the question was asked - couldn't this be done at the top
30619 So, in this commit, I propose to update the top-level .gitattributes
30620 file, after this commit, any git diff on a C, C++, Expect, or TCL
30621 source file, will highlight the following whitespace errors:
30623 (a) Use a space before a tab at the start of a line,
30625 (b) Use of spaces where a tab could be used at the start of a line,
30628 (c) Any trailing whitespace.
30630 Errors are only highlighted in the diff on new or modified lines, so
30631 you don't get spammed for errors on context lines that you haven't
30634 The only downside I see to adding this at the top level is if there
30635 are any sub-directories that don't follow the tabs/spaces indentation
30636 rules very well already, in those directories you'll end up hitting
30637 issues any time you edit a line. For GDB we're usually pretty good,
30638 so having this highlighting isn't an issue.
30640 [1] https://sourceware.org/pipermail/gdb-patches/2022-July/190843.html
30642 2022-07-25 Yvan Roux <yvan.roux@foss.st.com>
30644 gdb/arm: Sync sp with other *sp registers
30645 For Arm Cortex-M33 with security extensions, there are 4 different
30646 stack pointers (msp_s, msp_ns, psp_s, psp_ns), without security
30647 extensions and for other Cortex-M targets, there are 2 different
30648 stack pointers (msp and psp).
30650 With this patch, sp will always be in sync with one of the real stack
30651 pointers on Arm targets that contain more than one stack pointer.
30653 2022-07-25 Torbjörn SVENSSON <torbjorn.svensson@foss.st.com>
30655 gdb/arm: Use if-else if instead of switch
30656 As the register numbers for the alternative Arm SP registers are not
30657 constant, it's not possible to use switch statement to define the
30658 rules. In order to not have a mix, replace the few existing
30659 switch statements with regular if-else if statements
30661 2022-07-25 Tom Tromey <tromey@adacore.com>
30663 Remove dead code from windows_nat_target::detach
30664 windows_nat_target::detach has a variable 'detached' that is only set
30665 after a call to 'error'. However, this can't happen because 'error'
30666 throws an exception.
30668 This patch removes the dead code.
30670 2022-07-25 Andrew Burgess <aburgess@redhat.com>
30672 gdb: handle dis_style_sub_mnemonic disassembler style
30675 commit 4f46c0bc36471b725de0253bfec1a42a36e2c5c5
30676 Date: Mon Jul 4 17:45:25 2022 +0100
30678 opcodes: add new sub-mnemonic disassembler style
30680 I added a new disassembler style dis_style_sub_mnemonic, but forgot to
30681 add GDB support for this style. Fix this oversight in this commit.
30683 2022-07-25 Andrew Burgess <aburgess@redhat.com>
30685 libopcodes/ppc: add support for disassembler styling
30686 This commit adds disassembler styling to the libopcodes ppc
30687 disassembler. This conversion was pretty straight forward, I just
30688 converted the fprintf_func calls to fprintf_styled_func calls and
30689 added an appropriate style.
30691 For testing the new styling I just assembled then disassembled the
30692 source files in gas/testsuite/gas/ppc and manually checked that the
30693 styling looked reasonable.
30695 I think the only slightly weird case was how things like '4*cr1+eq'
30696 are styled. As best I can tell, this construct, used for example in
30699 crand 4*cr1+lt,4*cr1+gt,4*cr1+eq
30701 is used to access a field of a control register. I initially tried
30702 styling this whole construct as a register[1], but during review it
30703 was suggested that instead different parts of the text should have
30704 different styles. In this commit I propose styling '4*cr1+lt' like
30713 If the user does not request styled output from objdump, then there
30714 should be no change in the disassembler output after this commit.
30716 [1] https://sourceware.org/pipermail/binutils/2022-July/121771.html
30718 2022-07-25 Andrew Burgess <aburgess@redhat.com>
30720 opcodes: add new sub-mnemonic disassembler style
30721 When adding libopcodes disassembler styling support for AArch64, it
30722 feels like the results would be improved by having a new sub-mnemonic
30723 style. This will be used in cases like:
30725 add w16, w7, w1, uxtb #2
30733 This commit just adds the new style, and prepares objdump to handle
30734 the style. A later commit will add AArch64 styling, and will actually
30735 make use of the style.
30737 As this style is currently unused, there should be no user visible
30738 changes after this commit.
30740 2022-07-25 liuzhensong <liuzhensong@loongson.cn>
30742 LoongArch: Add testcases for new relocate types.
30743 gas/testsuite/gas/all/
30745 gas/testsuite/gas/loongarch/
30752 macro_op_large_abs.d
30753 macro_op_large_abs.s
30754 macro_op_large_pc.d
30755 macro_op_large_pc.s
30759 ld/testsuite/ld-elf/
30762 ld/testsuite/ld-loongarch-elf/
30785 nopic-weak-global-so.rd
30786 nopic-weak-global-so.sd
30787 nopic-weak-global.out
30788 nopic-weak-global.s
30789 nopic-weak-global.sd
30790 nopic-weak-global.xd
30791 nopic-weak-local.out
30792 nopic-weak-local.rd
30794 nopic-weak-local.sd
30795 nopic-weak-local.xd
30799 2022-07-25 liuzhensong <liuzhensong@loongson.cn>
30801 bfd: Delete R_LARCH_NONE from dyn info of LoongArch.
30802 Some R_LARCH_64 in section .eh_frame will to generate
30803 R_LARCH_NONE, we change relocation to R_LARCH_32_PCREL
30804 from R_LARCH_64 in setction .eh_frame and not generate
30805 dynamic relocation for R_LARCH_32_PCREL.
30807 Add New relocate type R_LARCH_32_PCREL for .eh_frame.
30825 ld/testsuite/ld-elf/
30828 2022-07-25 liuzhensong <liuzhensong@loongson.cn>
30830 LoongArch: Move ifunc info to rela.dyn from rela.plt.
30831 Delete R_LARCH_IRELATIVE from dynamic loader (glibc ld.so) when
30832 loading lazy function (rela.plt section).
30834 In dynamic programes, move ifunc dynamic relocate info to section
30835 srelgot from srelplt.
30840 2022-07-25 liuzhensong <liuzhensong@loongson.cn>
30842 LoongArch: gas: Add new reloc types.
30843 Generate new relocate types while use new macro insns.
30851 2022-07-25 liuzhensong <liuzhensong@loongson.cn>
30853 LoongArch:opcodes: Add new reloc types.
30854 opcodes: Replace old insns with news and generate new relocate types
30855 while macro insns expanding.
30860 2022-07-25 liuzhensong <liuzhensong@loongson.cn>
30862 bfd: Add supported for LoongArch new relocations.
30863 Define new reloc types according to linker needs.
30876 2022-07-25 Alan Modra <amodra@gmail.com>
30878 Re: PowerPC64 .branch_lt address
30879 On seeing PR29369 my suspicion was naturally on a recent powerpc64
30880 change, commit 0ab80031430e. Without a reproducer, I spent time
30881 wondering what could have gone wrong, and while I doubt this patch
30882 would have fixed the PR, there are some improvements that can be made
30883 to cater for user silliness.
30885 I also noticed that when -z relro -z now sections are created out of
30886 order, with .got before .plt in the section headers but .got is laid
30887 out at a higher address. That's due to the address expression for
30888 .branch_lt referencing SIZEOF(.got) and so calling init_os (which
30889 creates a bfd section) for .got before the .plt section is created.
30890 Fix that by ignoring SIZEOF in exp_init_os. Unlike ADDR and LOADADDR
30891 which need to reference section vma and lma respectively, SIZEOF can
30892 and does cope with a missing bfd section by returning zero for its
30893 size, which of course is correct.
30896 * ldlang.c (exp_init_os): Don't create a bfd section for SIZEOF.
30897 * emulparams/elf64ppc.sh (OTHER_RELRO_SECTIONS_2): Revise
30898 .branch_lt address to take into account possible user sections
30899 with alignment larger than 8 bytes.
30901 2022-07-25 GDB Administrator <gdbadmin@sourceware.org>
30903 Automatic date update in version.in
30905 2022-07-24 Enze Li <enze.li@hotmail.com>
30907 gdb/testsuite: add a clear test to py-breakpoint.exp
30908 This patch adds a test case to try to clear an internal python
30909 breakpoint using the clear command.
30911 This was suggested by Pedro during a code review of the following
30914 commit a5c69b1e49bae4d0dcb20f324cebb310c63495c6
30915 Date: Sun Apr 17 15:09:46 2022 +0800
30917 gdb: fix using clear command to delete non-user breakpoints(PR cli/7161)
30919 Tested on x86_64 openSUSE Tumbleweed.
30921 2022-07-24 Enze Li <enze.li@hotmail.com>
30923 gdb/testsuite: rename get_maint_bp_addr and move it to gdb-utils.exp
30924 The get_maint_bp_addr procedure will be shared by other test suite, so
30925 move it to gdb-utils.exp.
30927 Following Andrew's suggestion, I renamed get_maint_bp_addr to
30928 gdb_get_bp_addr, since it would have handled normal breakpoints in
30929 addition to the internal ones. Note that there is still room for
30930 improvement in this procedure, which I indicated in comments nearby.
30932 2022-07-24 GDB Administrator <gdbadmin@sourceware.org>
30934 Automatic date update in version.in
30936 2022-07-23 GDB Administrator <gdbadmin@sourceware.org>
30938 Automatic date update in version.in
30940 2022-07-22 Tom de Vries <tdevries@suse.de>
30942 [gdb/symtab] Fix duplicate CUs in all_comp_units
30943 When running test-case gdb.cp/cpexprs-debug-types.exp with target board
30944 cc-with-debug-names on a system with gcc 12.1.1 (defaulting to dwarf 5), I
30947 (gdb) file cpexprs-debug-types^M
30948 Reading symbols from cpexprs-debug-types...^M
30949 warning: Section .debug_aranges in cpexprs-debug-types has duplicate \
30950 debug_info_offset 0x0, ignoring .debug_aranges.^M
30951 gdb/dwarf2/read.h:309: internal-error: set_length: \
30952 Assertion `m_length == length' failed.^M
30955 The exec contains a .debug_names section, which gdb rejects due to
30956 .debug_names containing a list of TUs, while the exec doesn't contain a
30957 .debug_types section (which is what you'd expect for dwarf 4).
30959 Gdb then falls back onto the cooked index, which calls create_all_comp_units
30960 to create all_comp_units. However, the failed index reading left some
30961 elements in all_comp_units, so we end up with duplicates in all_comp_units,
30962 which causes the misleading complaint and the assert.
30965 - asserting at the start of create_all_comp_units that all_comp_units is empty,
30966 as we do in create_cus_from_index and create_cus_from_debug_names, and
30967 - cleaning up all_comp_units when failing in dwarf2_read_debug_names.
30969 Add a similar cleanup in dwarf2_read_gdb_index.
30971 Tested on x86_64-linux.
30973 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29381
30975 2022-07-22 Simon Marchi <simon.marchi@polymtl.ca>
30977 gdb/testsuite: give binaries distinct names in Ada tests
30978 Some Ada tests repeat their test sequence with different gnat-encodings,
30979 typically "all" and "minimal". However, they give the same name to both
30980 binaries, meaning the second run overwrites the binary of the first run.
30981 This makes it difficult and confusing when trying to reproduce problems
30982 manually with the test artifacts. Change those tests to use unique
30983 names for each pass.
30985 Change-Id: Iaa3c9f041241249a7d67392e785c31aa189dcc88
30987 2022-07-22 Tom Tromey <tromey@adacore.com>
30989 Change target_ops::async to accept bool
30990 This changes the parameter of target_ops::async from int to bool.
30991 Regression tested on x86-64 Fedora 34.
30993 Fix typo in windows-nat.c
30994 I noticed a typo in a printf in windows-nat.c. This fixes it.
30996 2022-07-22 Tom de Vries <tdevries@suse.de>
30998 [gdb] Add empty range unit test for gdb::parallel_for_each
30999 Add a unit test that verifies that we can call gdb::parallel_for_each with an
31002 Tested on x86_64-linux.
31004 2022-07-22 Alan Modra <amodra@gmail.com>
31006 PR17122, OSX 10.9 build failure
31007 sbrk hasn't been used in binutils/ or ld/ for quite some time (so the
31008 PR was fixed a while ago). Tidy up configury.
31012 * configure.ac: Don't check for sbrk.
31013 * sysdep.h (sbrk): Don't supply fallback declaration.
31014 * config.in: Regenerate.
31015 * configure: Regenerate.
31017 * configure.ac: Don't check for sbrk.
31018 * config.in: Regenerate.
31019 * configure: Regenerate.
31021 2022-07-22 Jiangshuai Li <jiangshuai_li@linux.alibaba-inc.com>
31023 gdb/csky modify registers list for general_reggroup
31024 There are two modification points here:
31025 1. For the debugging of csky architecture, after executing "info register",
31026 we hope to print out GPRs, PC and the registers related to exceptions.
31027 2. With tdesc-xml, users can view the register groups described in XML.
31029 2022-07-22 Alan Modra <amodra@gmail.com>
31031 PR15951, binutils testsuite builds status wrapper unconditionally
31033 * testsuite/binutils-all/objcopy.exp: Build testglue.o when
31034 needs_status_wrapper.
31036 2022-07-22 GDB Administrator <gdbadmin@sourceware.org>
31038 Automatic date update in version.in
31040 2022-07-21 Peter Bergner <bergner@linux.ibm.com>
31042 Add ChangeLog entry from previous commit
31044 2022-07-21 Peter Bergner <bergner@linux.ibm.com>
31046 PowerPC: Create new MMA instruction masks and use them
31047 The MMA instructions use XX3_MASK|3<<21 as an instruction mask, but that
31048 misses the RC bit/bit 31, so if we disassemble a .long that represents an
31049 MMA instruction except that it also has bit 31 set, we will erroneously
31050 disassemble it to that MMA instruction. We create new masks defines that
31051 contain bit 31 so that doesn't happen anymore.
31054 * ppc-opc.c (XACC_MASK, XX3ACC_MASK): New defines.
31055 (P_GER_MASK, xxmfacc, xxmtacc, xxsetaccz, xvi8ger4pp, xvi8ger4,
31056 xvf16ger2pp, xvf16ger2, xvf32gerpp, xvf32ger, xvi4ger8pp, xvi4ger8,
31057 xvi16ger2spp, xvi16ger2s, xvbf16ger2pp, xvbf16ger2, xvf64gerpp,
31058 xvf64ger, xvi16ger2, xvf16ger2np, xvf32gernp, xvi8ger4spp, xvi16ger2pp,
31059 xvbf16ger2np, xvf64gernp, xvf16ger2pn, xvf32gerpn, xvbf16ger2pn,
31060 xvf64gerpn, xvf16ger2nn, xvf32gernn, xvbf16ger2nn, xvf64gernn: Use them.
31062 2022-07-21 H.J. Lu <hjl.tools@gmail.com>
31064 i386: Don't allow GOTOFF relocation against IFUNC symbol for PIC
31065 We can't use the PLT entry as the function address for PIC since the PIC
31066 register may not be set up properly for indirect call.
31071 * elf32-i386.c (elf_i386_relocate_section): Don't allow GOTOFF
31072 relocation against IFUNC symbol for PIC.
31077 * testsuite/ld-i386/pr27998a.d: Replace -shared with -e bar.
31078 * testsuite/ld-i386/pr27998b.d: Expect a linker error.
31079 * testsuite/ld-ifunc/ifunc-2-i386-now.d: Updated.
31080 * testsuite/ld-ifunc/ifunc-2-local-i386-now.d: Likewise.
31081 * testsuite/ld-ifunc/ifunc-2-i386.s: Replace @GOTOFF with @GOT.
31082 * testsuite/ld-ifunc/ifunc-2-local-i386.s: Likewise.
31084 2022-07-21 Andrew Burgess <aburgess@redhat.com>
31086 gdb: ensure the cast in gdbarch_tdep is valid
31087 This commit makes use of gdb::checked_static_cast when casting the
31088 generic gdbarch_tdep pointer to a specific sub-class type. This means
31089 that, when compiled in developer mode, GDB will validate that the cast
31092 In order to use gdb::checked_static_cast the types involved must have
31093 RTTI, which is why the gdbarch_tdep base class now has a virtual
31096 Assuming there are no bugs in GDB where we cast a gdbarch_tdep pointer
31097 to the wrong type, then there should be no changes after this commit.
31099 If any bugs do exist, then GDB will now assert (in a developer build).
31101 2022-07-21 Andrew Burgess <aburgess@redhat.com>
31103 gdbsupport: add checked_static_cast
31104 This commit was inspired by these mailing list posts:
31106 https://sourceware.org/pipermail/gdb-patches/2022-June/190323.html
31107 https://sourceware.org/pipermail/gdb-patches/2022-April/188098.html
31109 The idea is to add a new function gdb::checked_static_cast, which can,
31110 in some cases, be used as a drop-in replacement for static_cast. And
31111 so, if I previously wrote this:
31113 BaseClass *base = get_base_class_pointer ();
31114 DerivedClass *derived = static_cast<DerivedClass *> (base);
31118 BaseClass *base = get_base_class_pointer ();
31119 DerivedClass *derived = gdb::checked_static_cast<DerivedClass *> (base);
31121 The requirement is that BaseClass and DerivedClass must be
31124 The benefit of making this change is that, when GDB is built in
31125 developer mode, a run-time check will be made to ensure that `base`
31126 really is of type DerivedClass before the cast is performed. If
31127 `base` is not of type DerivedClass then GDB will assert.
31129 In a non-developer build gdb::checked_static_cast is equivalent to a
31130 static_cast, and there should be no performance difference.
31132 This commit adds the support function, but does not make use of this
31133 function, a use will be added in the next commit.
31135 Co-Authored-By: Pedro Alves <pedro@palves.net>
31136 Co-Authored-By: Tom Tromey <tom@tromey.com>
31138 2022-07-21 Andrew Burgess <aburgess@redhat.com>
31140 gdb: move the type cast into gdbarch_tdep
31141 I built GDB for all targets on a x86-64/GNU-Linux system, and
31142 then (accidentally) passed GDB a RISC-V binary, and asked GDB to "run"
31143 the binary on the native target. I got this error:
31145 (gdb) show architecture
31146 The target architecture is set to "auto" (currently "i386").
31147 (gdb) file /tmp/hello.rv32.exe
31148 Reading symbols from /tmp/hello.rv32.exe...
31149 (gdb) show architecture
31150 The target architecture is set to "auto" (currently "riscv:rv32").
31152 Starting program: /tmp/hello.rv32.exe
31153 ../../src/gdb/i387-tdep.c:596: internal-error: i387_supply_fxsave: Assertion `tdep->st0_regnum >= I386_ST0_REGNUM' failed.
31155 What's going on here is this; initially the architecture is i386, this
31156 is based on the default architecture, which is set based on the native
31157 target. After loading the RISC-V executable the architecture of the
31158 current inferior is updated based on the architecture of the
31161 When we "run", GDB does a fork & exec, with the inferior being
31162 controlled through ptrace. GDB sees an initial stop from the inferior
31163 as soon as the inferior comes to life. In response to this stop GDB
31164 ends up calling save_stop_reason (linux-nat.c), which ends up trying
31165 to read register from the inferior, to do this we end up calling
31166 target_ops::fetch_registers, which, for the x86-64 native target,
31167 calls amd64_linux_nat_target::fetch_registers.
31169 After this I eventually end up in i387_supply_fxsave, different x86
31170 based targets will end in different functions to fetch registers, but
31171 it doesn't really matter which function we end up in, the problem is
31172 this line, which is repeated in many places:
31174 i386_gdbarch_tdep *tdep = (i386_gdbarch_tdep *) gdbarch_tdep (arch);
31176 The problem here is that the ARCH in this line comes from the current
31177 inferior, which, as we discussed above, will be a RISC-V gdbarch, the
31178 tdep field will actually be of type riscv_gdbarch_tdep, not
31179 i386_gdbarch_tdep. After this cast we are relying on undefined
31180 behaviour, in my case I happen to trigger an assert, but this might
31181 not always be the case.
31183 The thing I tried that exposed this problem was of course, trying to
31184 start an executable of the wrong architecture on a native target. I
31185 don't think that the correct solution for this problem is to detect,
31186 at the point of cast, that the gdbarch_tdep object is of the wrong
31187 type, but, I did wonder, is there a way that we could protect
31188 ourselves from incorrectly casting the gdbarch_tdep object?
31190 I think that there is something we can do here, and this commit is the
31191 first step in that direction, though no actual check is added by this
31194 This commit can be split into two parts:
31196 (1) In gdbarch.h and arch-utils.c. In these files I have modified
31197 gdbarch_tdep (the function) so that it now takes a template argument,
31200 template<typename TDepType>
31201 static inline TDepType *
31202 gdbarch_tdep (struct gdbarch *gdbarch)
31204 struct gdbarch_tdep *tdep = gdbarch_tdep_1 (gdbarch);
31205 return static_cast<TDepType *> (tdep);
31208 After this change we are no better protected, but the cast is now
31209 done within the gdbarch_tdep function rather than at the call sites,
31210 this leads to the second, much larger change in this commit,
31212 (2) Everywhere gdbarch_tdep is called, we make changes like this:
31214 - i386_gdbarch_tdep *tdep = (i386_gdbarch_tdep *) gdbarch_tdep (arch);
31215 + i386_gdbarch_tdep *tdep = gdbarch_tdep<i386_gdbarch_tdep> (arch);
31217 There should be no functional change after this commit.
31219 In the next commit I will build on this change to add an assertion in
31220 gdbarch_tdep that checks we are casting to the correct type.
31222 2022-07-21 Andrew Burgess <aburgess@redhat.com>
31224 gdb: select suitable thread for gdbarch_adjust_breakpoint_address
31225 The three targets that implement gdbarch_adjust_breakpoint_address are
31226 arm, frv, and mips. In each of these targets the adjust breakpoint
31227 address function does some combination of reading the symbol table, or
31228 reading memory at the location the breakpoint could be placed.
31230 The problem is that performing these actions requires that the current
31231 inferior and program space be the one in which the breakpoint will be
31232 placed, and this is not currently always the case.
31234 Consider a GDB session with multiple inferiors. One inferior might be
31235 a native target while another could be a remote target of a completely
31236 different architecture. Alternatively, if we consider ARM and
31237 AArch64, one native inferior might be AArch64, while a second native
31238 inferior could be ARM.
31240 In these cases it is possible, and valid, for a user to have one
31241 inferior selected, and place a breakpoint in the other inferior by
31242 placing a breakpoint on a particular symbol.
31244 If this happens, then currently, when
31245 gdbarch_adjust_breakpoint_address is called, the wrong inferior (and
31246 program space) will be selected, and memory reads, and symbol look
31247 ups, will not return the expected results, this could lead to
31248 breakpoints being placed in the wrong location.
31250 There are currently two places where gdbarch_adjust_breakpoint_address
31253 1. In infrun.c, in the function handle_step_into_function. In this
31254 case, I believe that the correct inferior and program space will
31255 already be selected as this is called as part of the stop event
31256 handling, so I don't think we need to worry about this case, and
31258 2. In breakpoint.c, in the function adjust_breakpoint_address, which
31259 is itself called from code_breakpoint::add_location and
31262 The watch_command_1 case I don't think we need to worry about, this
31263 is for when a local watch expression is created, which can only be
31264 in the currently selected inferior, so this case should be fine.
31266 The code_breakpoint::add_location case is the one that needs fixing,
31267 this is what allows a breakpoint to be created between inferiors.
31269 To fix the code_breakpoint::add_location case, I propose that we pass
31270 the "correct" program_space (i.e. the program space in which the
31271 breakpoint will be created) to the adjust_breakpoint_address function.
31272 Then in adjust_breakpoint_address we can make use of
31273 switch_to_program_space_and_thread to switch program_space and
31274 inferior before calling gdbarch_adjust_breakpoint_address.
31276 I discovered this issue while working on a later patch in this
31277 series. This later patch will detect when we cast the result of
31278 gdbarch_tdep to the wrong type.
31280 With this later patch in place I ran gdb.multi/multi-arch.exp on an
31281 AArch64 target. In this situation, two inferiors are created, an
31282 AArch64 inferior, and an ARM inferior. The test selected the AArch64
31283 inferior and tries to create a breakpoint in the ARM inferior.
31285 As a result of this we end up in arm_adjust_breakpoint_address, which
31286 calls arm_pc_is_thumb. Before this commit the AArch64 inferior would
31287 be current. As a result, all of the checks in arm_pc_is_thumb would
31288 fail (they rely on reading symbols from the current program space),
31289 and so, at the end of arm_pc_is_thumb we would call
31290 arm_frame_is_thumb. However, remember, at this point the current
31291 inferior is the AArch64 inferior, so the current frame is an AArch64
31294 In arm_frame_is_thumb we call arm_psr_thumb_bit, which calls
31295 gdbarch_tdep and casts the result to arm_gdbarch_tdep. This is wrong,
31296 the tdep field is of type aarch64_gdbarch_tdep. After this we have
31297 undefined behaviour.
31299 With this patch in place, we will have switched to a thread in the ARM
31300 program space before calling arm_adjust_breakpoint_address. As a
31301 result, we now succeed in looking up the required symbols in
31302 arm_pc_is_thumb, and so we never call arm_frame_is_thumb.
31304 However, in the worst case scenario, if we did end up calling
31305 arm_frame_is_thumb, as the current inferior should now be the ARM
31306 inferior, the current frame should be an ARM frame, so we still should
31307 not hit undefined behaviour.
31309 I have added an assert to arm_frame_is_thumb.
31311 2022-07-21 Andrew Burgess <aburgess@redhat.com>
31313 gdb/mips: rewrite show_mask_address
31314 This commit is similar to the previous commit, but in this case GDB is
31315 actually relying on undefined behaviour.
31317 Consider building GDB for all targets on x86-64/GNU-Linux, then doing
31320 (gdb) show mips mask-address
31321 Zeroing of upper 32 bits of 64-bit addresses is auto.
31322 The 32 bit address mask is set automatically. Currently disabled
31325 The 'show mips mask-address' command ends up in show_mask_address in
31326 mips-tdep.c, and this function does this:
31328 mips_gdbarch_tdep *tdep
31329 = (mips_gdbarch_tdep *) gdbarch_tdep (target_gdbarch ());
31331 Later we might pass TDEP to mips_mask_address_p. However, in my
31332 example above, on an x86-64 native target, the current target
31333 architecture will be an x86-64 gdbarch, and the tdep field within the
31334 gdbarch will be of type i386_gdbarch_tdep, not of type
31335 mips_gdbarch_tdep, as a result the cast above was incorrect, and TDEP
31336 is not pointing at what it thinks it is.
31338 I also think the current output is a little confusing, we appear to
31339 have two lines that show the same information, but using different
31342 The first line comes from calling deprecated_show_value_hack, while
31343 the second line is printed directly from show_mask_address. However,
31344 both of these lines are printing the same mask_address_var value. I
31345 don't think the two lines actually adds any value here.
31347 Finally, none of the text in this function is passed through the
31348 internationalisation mechanism.
31350 It would be nice to remove another use of deprecated_show_value_hack
31351 if possible, so this commit does a complete rewrite of
31354 After this commit the output of the above example command, still on my
31355 x86-64 native target is:
31357 (gdb) show mips mask-address
31358 Zeroing of upper 32 bits of 64-bit addresses is "auto" (current architecture is not MIPS).
31360 The 'current architecture is not MIPS' text is only displayed when the
31361 current architecture is not MIPS. If the architecture is mips then we
31362 get the more commonly seen 'currently "on"' or 'currently "off"', like
31365 (gdb) set architecture mips
31366 The target architecture is set to "mips".
31367 (gdb) show mips mask-address
31368 Zeroing of upper 32 bits of 64-bit addresses is "auto" (currently "off").
31371 All the text is passed through the internationalisation mechanism, and
31372 we only call gdbarch_tdep when we know the gdbarch architecture is
31375 2022-07-21 Andrew Burgess <aburgess@redhat.com>
31377 gdb/arm: move fetch of arm_gdbarch_tdep to a more inner scope
31378 This is a small refactor to resolve an issue before it becomes a
31379 problem in a later commit.
31381 Move the fetching of an arm_gdbarch_tdep into a more inner scope
31382 within two functions in arm-tdep.c.
31384 The problem with the current code is that the functions in question
31385 are used as the callbacks for two set/show parameters. These set/show
31386 parameters are available no matter the current architecture, but are
31387 really about controlling an ARM architecture specific setting. And
31388 so, if I build GDB for all targets on an x86-64/GNU-Linux system, I
31394 After these calls we end up in show_fp_model and arm_show_abi
31395 respectively, where we unconditionally do this:
31397 arm_gdbarch_tdep *tdep
31398 = (arm_gdbarch_tdep *) gdbarch_tdep (target_gdbarch ());
31400 However, the gdbarch_tdep() result will only be a arm_gdbarch_tdep if
31401 the current architecture is ARM, otherwise the result will actually be
31402 of some other type.
31404 This isn't actually a problem, as in both cases the use of tdep is
31405 guarded by a later check that the gdbarch architecture is
31408 This commit just moves the call to gdbarch_tdep() after the
31409 architecture check.
31411 In a later commit gdbarch_tdep() will be able to spot when we are
31412 casting the result to the wrong type, and this function will trigger
31413 assertion failures if things are not fixed.
31415 There should be not user visible changes after this commit.
31417 2022-07-21 Torbjörn SVENSSON <torbjorn.svensson@foss.st.com>
31419 [arm] Rename arm_cache_is_sp_register to arm_is_alternative_sp_register
31420 All usages of this helper are really made to check if the register is
31421 one of the alternative SP registers (MSP/MSP_S/MSP_NS/PSP/PSP_S/PSP_NS)
31422 with the ARM_SP_REGNUM case being handled separately.
31424 2022-07-21 Tom de Vries <tdevries@suse.de>
31426 [gdb/python] Fix typo in test_python
31427 Fix typo in ref_output_0 variable in test_python.
31429 Tested by running the selftest on x86_64-linux with python 3.11.
31431 2022-07-21 Tom de Vries <tdevries@suse.de>
31433 [gdb/python] Fix python selftest with python 3.11
31434 With python 3.11 I noticed:
31436 $ gdb -q -batch -ex "maint selftest python"
31437 Running selftest python.
31438 Self test failed: self-test failed at gdb/python/python.c:2246
31439 Ran 1 unit tests, 1 failed
31445 $5 = "Traceback (most recent call last):\n File \"<string>\", line 0, \
31446 in <module>\nKeyboardInterrupt\n"
31448 $6 = "Traceback (most recent call last):\n File \"<string>\", line 1, \
31449 in <module>\nKeyboardInterrupt\n"
31452 Fix this by also allowing line number 0.
31454 Tested on x86_64-linux.
31456 This should hopefully fix buildbot builder gdb-rawhide-x86_64.
31458 2022-07-21 Tom de Vries <tdevries@suse.de>
31460 [gdbsupport] Fix type of parallel_for_each_debug
31461 When I changed the initialization of parallel_for_each_debug from 0 to false,
31462 I forgot to change the type from int to bool. Fix this.
31464 Tested by rebuilding on x86_64-linux.
31466 2022-07-21 Tom de Vries <tdevries@suse.de>
31468 [gdb/symtab] Fix bad compile unit index complaint
31469 I noticed this code in dw2_debug_names_iterator::next:
31471 case DW_IDX_compile_unit:
31472 /* Don't crash on bad data. */
31473 if (ull >= per_bfd->all_comp_units.size ())
31475 complaint (_(".debug_names entry has bad CU index %s"
31476 " [in module %s]"),
31478 objfile_name (objfile));
31481 per_cu = per_bfd->get_cu (ull);
31485 This code used to DTRT, before we started keeping both CUs and TUs in
31488 Fix by using "per_bfd->all_comp_units.size () - per_bfd->tu_stats.nr_tus"
31491 It's hard to produce a test-case for this, but let's try at least to trigger
31492 the complaint somehow. We start out by creating an exec with .debug_types and
31495 $ gcc -g ~/hello.c -fdebug-types-section
31496 $ gdb-add-index -dwarf-5 a.out
31498 and verify that we don't see any complaints:
31500 $ gdb -q -batch -iex "set complaints 100" ./a.out
31503 We look at the CU and TU table using readelf -w and conclude that we have
31504 nr_cus == 6 and nr_tus == 1.
31506 Now override ull in dw2_debug_names_iterator::next for the DW_IDX_compile_unit
31507 case to 6, and we have:
31509 $ gdb -q -batch -iex "set complaints 100" ./a.out
31510 During symbol reading: .debug_names entry has bad CU index 6 [in module a.out]
31513 After this, it still crashes because this code in
31514 dw2_debug_names_iterator::next:
31516 /* Skip if already read in. */
31517 if (m_per_objfile->symtab_set_p (per_cu))
31520 is called with per_cu == nullptr.
31522 Fix this by skipping the entry if per_cu == nullptr.
31524 Now revert the fix and observe that the complaint disappears, so we've
31525 confirmed that the fix is required.
31527 A somewhat similar issue for .gdb_index in dw2_symtab_iter_next has been filed
31530 Tested on x86_64-linux, with native and target board cc-with-debug-names.
31532 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29336
31534 2022-07-21 Jan Beulich <jbeulich@suse.com>
31536 x86: replace wrong attributes on VCVTDQ2PH{X,Y}
31537 A standalone (without SAE) StaticRounding attribute is meaningless, and
31538 indeed all other similar insns have ATTSyntax there instead. I can only
31539 assume this was some strange copy-and-paste mistake.
31541 x86/Intel: correct AVX512F scatter insn element sizes
31542 I clearly screwed up in 6ff00b5e12e7 ("x86/Intel: correct permitted
31543 operand sizes for AVX512 scatter/gather") giving all AVX512F scatter
31544 insns Dword element size. Update testcases (also their gather parts),
31545 utilizing that there previously were two identical lines each (for no
31548 2022-07-21 Alan Modra <amodra@gmail.com>
31550 PR29390, DW_CFA_AARCH64_negate_ra_state vs. DW_CFA_GNU_window_save
31553 * dwarf.c (is_aarch64, DW_CFA_GNU_window_save_name): New.
31554 (display_debug_frames): Use them.
31555 (init_dwarf_regnames_aarch64): Set is_aarch64.
31556 (init_dwarf_regnames_by_elf_machine_code): Clear is_aarch64.
31557 (init_dwarf_regnames_by_bfd_arch_and_mach): Likewise.
31559 * testsuite/gas/aarch64/pac_ab_key.d: Adjust expected output.
31560 * testsuite/gas/aarch64/pac_negate_ra_state.d: Likewise.
31562 2022-07-21 Alan Modra <amodra@gmail.com>
31564 PR29337, readelf CU/TU mixup in .gdb_index
31565 Commit 244e19c79111 changed a number of variables in display_gdb_index
31566 to count entries rather than words.
31569 * dwarf.c (display_gdb_index): Correct use of cu_list_elements.
31571 2022-07-21 Alan Modra <amodra@gmail.com>
31573 PR29370, infinite loop in display_debug_abbrev
31574 The PR29370 testcase is a fuzzed object file with multiple
31575 .trace_abbrev sections. Multiple .trace_abbrev or .debug_abbrev
31576 sections are not a violation of the DWARF standard. The DWARF5
31577 standard even gives an example of multiple .debug_abbrev sections
31578 contained in groups. Caching and lookup of processed abbrevs thus
31579 needs to be done by section and offset rather than base and offset.
31580 (Why base anyway?) Or, since section contents are kept, by a pointer
31584 * dwarf.c (struct abbrev_list): Replace abbrev_base and
31585 abbrev_offset with raw field.
31586 (find_abbrev_list_by_abbrev_offset): Delete.
31587 (find_abbrev_list_by_raw_abbrev): New function.
31588 (process_abbrev_set): Set list->raw and list->next.
31589 (find_and_process_abbrev_set): Replace abbrev list lookup with
31590 new function. Don't set list abbrev_base, abbrev_offset or next.
31592 2022-07-21 Alan Modra <amodra@gmail.com>
31594 binutils/dwarf.c: abbrev caching
31595 I'm inclined to think that abbrev caching is counter-productive. The
31596 time taken to search the list of abbrevs converted to internal form is
31597 non-zero, and it's easy to decode the raw abbrevs. It's especially
31598 silly to cache empty lists of decoded abbrevs (happens with zero
31599 padding in .debug_abbrev), or abbrevs as they are displayed when there
31600 is no further use of those abbrevs. This patch stops caching in those
31603 * dwarf.c (record_abbrev_list_for_cu): Add free_list param.
31604 Put abbrevs on abbrev_lists here.
31605 (new_abbrev_list): Delete function.
31606 (process_abbrev_set): Return newly allocated list. Move
31607 abbrev base, offset and size checking to..
31608 (find_and_process_abbrev_set): ..here, new function. Handle
31609 lookup of cached abbrevs here, and calculate start and end
31610 for process_abbrev_set. Return free_list if newly alloc'd.
31611 (process_debug_info): Consolidate cached list lookup, new list
31612 alloc and processing into find_and_process_abbrev_set call.
31613 Free list when not cached.
31614 (display_debug_abbrev): Similarly.
31616 2022-07-21 Alan Modra <amodra@gmail.com>
31618 miscellaneous dwarf.c tidies
31619 * dwarf.c: Leading and trailing whitespace fixes.
31620 (free_abbrev_list): New function.
31621 (free_all_abbrevs): Use the above. Free cu_abbrev_map here too.
31622 (process_abbrev_set): Print actual section name on error.
31623 (get_type_abbrev_from_form): Add overflow check.
31624 (free_debug_memory): Don't free cu_abbrev_map here..
31625 (process_debug_info): ..or here. Warn on another case of not
31626 finding a neeeded abbrev.
31628 2022-07-21 Alan Modra <amodra@gmail.com>
31630 PowerPC64: fix build error on 32-bit hosts
31631 elf64-ppc.c:11673:33: error: format ‘%lx’ expects argument of type ‘long unsigned int’, but argument 3 has type ‘bfd_vma’ {aka ‘long long unsigned int’} [-Werror=format=]
31632 11673 | fprintf (stderr, "offset = %#lx:", stub_entry->stub_offset);
31633 | ~~~^ ~~~~~~~~~~~~~~~~~~~~~~~
31635 | | bfd_vma {aka long long unsigned int}
31636 | long unsigned int
31639 * elf64-ppc.c (dump_stub): Use BFD_VMA_FMT.
31641 2022-07-21 Kevin Buettner <kevinb@redhat.com>
31643 Wrap python_write_bytecode with HAVE_PYTHON ifdef
31644 This commit fixes a build error on machines lacking python headers
31647 2022-07-21 GDB Administrator <gdbadmin@sourceware.org>
31649 Automatic date update in version.in
31651 2022-07-20 Kevin Buettner <kevinb@redhat.com>
31653 Handle Python 3.11 deprecation of PySys_SetPath and Py_SetProgramName
31654 Python 3.11 deprecates PySys_SetPath and Py_SetProgramName. The
31655 PyConfig API replaces these and other functions. This commit uses the
31656 PyConfig API to provide equivalent functionality while also preserving
31657 support for older versions of Python, i.e. those before Python 3.8.
31659 A beta version of Python 3.11 is available in Fedora Rawhide. Both
31660 Fedora 35 and Fedora 36 use Python 3.10, while Fedora 34 still used
31661 Python 3.9. I've tested these changes on Fedora 34, Fedora 36, and
31662 rawhide, though complete testing was not possible on rawhide due to
31663 a kernel bug. That being the case, I decided to enable the newer
31664 PyConfig API by testing PY_VERSION_HEX against 0x030a0000. This
31665 corresponds to Python 3.10.
31667 We could try to use the PyConfig API for Python versions as early as 3.8,
31668 but I'm reluctant to do this as there may have been PyConfig related
31669 bugs in earlier versions which have since been fixed. Recent linux
31670 distributions should have support for Python 3.10. This should be
31671 more than adequate for testing the new Python initialization code in
31674 Information about the PyConfig API as well as the motivation behind
31675 deprecating the old interface can be found at these links:
31677 https://github.com/python/cpython/issues/88279
31678 https://peps.python.org/pep-0587/
31679 https://docs.python.org/3.11/c-api/init_config.html
31681 The v2 commit also addresses several problems that Simon found in
31684 In v1, I had used Py_DontWriteBytecodeFlag in the new initialization
31685 code, but Simon pointed out that this global configuration variable
31686 will be deprecated in Python 3.12. This version of the patch no longer
31687 uses Py_DontWriteBytecodeFlag in the new initialization code.
31688 Additionally, both Py_DontWriteBytecodeFlag and Py_IgnoreEnvironmentFlag
31689 will no longer be used when building GDB against Python 3.10 or higher.
31690 While it's true that both of these global configuration variables are
31691 deprecated in Python 3.12, it makes sense to disable their use for
31692 gdb builds against 3.10 and higher since those are the versions for
31693 which the PyConfig API is now being used by GDB. (The PyConfig API
31694 includes different mechanisms for making the same settings afforded
31695 by use of the soon-to-be deprecated global configuration variables.)
31697 Simon also noted that PyConfig_Clear() would not have be called for
31698 one of the failure paths. I've fixed that problem and also made the
31699 rest of the "bail out" code more direct. In particular,
31700 PyConfig_Clear() will always be called, both for success and failure.
31702 The v3 patch addresses some rebase conflicts related to module
31703 initialization . Commit 3acd9a692dd ("Make 'import gdb.events' work")
31704 uses PyImport_ExtendInittab instead of PyImport_AppendInittab. That
31705 commit also initializes a struct for each module to import. Both the
31706 initialization and the call to were moved ahead of the ifdefs to avoid
31707 having to replicate (at least some of) the code three times in various
31708 portions of the ifdefs.
31710 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28668
31711 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29287
31713 2022-07-20 Christopher Di Bella <cjdb@google.com>
31715 gdb/value.c: add several headers to the include list
31716 Building GDB currently fails to build with libc++, because libc++ is
31717 stricter about which headers "leak" entities they're not guaranteed
31718 to support. The following headers have been added:
31720 * `<iterator>`, to support `std::back_inserter`
31721 * `<utility>`, to support `std::move` and `std::swap`
31722 * `<vector>`, to support `std::vector`
31724 Change-Id: Iaeb15057c5fbb43217df77ce34d4e54446dbcf3d
31726 2022-07-20 Pedro Alves <pedro@palves.net>
31728 Don't stop all threads prematurely after first step of "step N"
31729 In all-stop mode, when the target is itself in non-stop mode (like
31730 GNU/Linux), if you use the "step N" (or "stepi/next/nexti N") to step
31731 a thread a number of times:
31735 Step program until it reaches a different source line.
31737 Argument N means step N times (or till program stops for another reason).
31739 ... GDB prematurely stops all threads after the first step, and
31740 doesn't re-resume them for the subsequent N-1 steps. It's as if for
31741 the 2nd and subsequent steps, the command was running with
31742 scheduler-locking enabled.
31744 This can be observed with the testcase added by this commit, which
31747 static pthread_barrier_t barrier;
31750 thread_func (void *arg)
31752 pthread_barrier_wait (&barrier);
31762 pthread_barrier_init (&barrier, NULL, 2);
31764 /* We run to this line below, and then issue "next 3". That should
31765 step over the 3 lines below and land on the return statement. If
31766 GDB prematurely stops the thread_func thread after the first of
31767 the 3 nexts (and never resumes it again), then the join won't
31769 pthread_create (&thread, NULL, thread_func, NULL); /* set break here */
31770 pthread_barrier_wait (&barrier);
31771 pthread_join (thread, NULL);
31776 The test hangs and times out without the GDB fix:
31779 [New Thread 0x7ffff7d89700 (LWP 525772)]
31780 FAIL: gdb.threads/step-N-all-progress.exp: non-stop=off: target-non-stop=on: next 3 (timeout)
31782 The problem is a core gdb bug.
31784 When you do "step/stepi/next/nexti N", GDB internally creates a
31785 thread_fsm object and associates it with the stepping thread. For the
31786 stepping commands, the FSM's class is step_command_fsm. That object
31787 is what keeps track of how many steps are left to make. When one step
31788 finishes, handle_inferior_event calls stop_waiting and returns, and
31789 then fetch_inferior_event calls the "should_stop" method of the event
31790 thread's FSM. The implementation of that method decrements the
31791 steps-left counter. If the counter is 0, it returns true and we
31792 proceed to presenting the stop to the user. If it isn't 0 yet, then
31793 the method returns false, indicating to fetch_inferior_event to "keep
31796 Focusing now on when the first step finishes -- we're in "all-stop"
31797 mode, with the target in non-stop mode. When a step finishes,
31798 handle_inferior_event calls stop_waiting, which itself calls
31799 stop_all_threads to stop everything. I.e., after the first step
31800 completes, all threads are stopped, before handle_inferior_event
31801 returns. And after that, now in fetch_inferior_event, we consult the
31802 thread's thread_fsm::should_stop, which as we've seen, for the first
31803 step returns false -- i.e., we need to keep_going for another step.
31804 However, since the target is in non-stop mode, keep_going resumes
31805 _only_ the current thread. All the other threads remain stopped,
31808 If the target is in non-stop mode, we don't actually need to stop all
31809 threads right after each step finishes, and then re-resume them again.
31810 We can instead defer stopping all threads until all the steps are
31813 So fix this by delaying the stopping of all threads until after we
31814 called the FSM's "should_stop" method. I.e., move it from
31815 stop_waiting, to handle_inferior_events's callers,
31816 fetch_inferior_event and wait_for_inferior.
31818 New test included. Tested on x86-64 GNU/Linux native and gdbserver.
31820 Change-Id: Iaad50dcfea4464c84bdbac853a89df92ade6ae01
31822 2022-07-20 Alan Modra <amodra@gmail.com>
31824 Re: opcodes/arc: Implement style support in the disassembler
31825 * arc-dis.c (print_insn_arc): Fix thinko.
31827 2022-07-20 Dmitry Selyutin <ghostmansd@gmail.com>
31829 gas/symbols: introduce md_resolve_symbol
31830 Assuming GMSD is a special operand, marked as O_md1, the code:
31836 ...fails upon attempts to resolve the value of the symbol. This happens
31837 since machine-dependent values are not handled in the giant op switch.
31838 We introduce a custom md_resolve_symbol macro; the ports can use this
31839 macro to customize the behavior when resolve_symbol_value hits O_md
31842 2022-07-20 GDB Administrator <gdbadmin@sourceware.org>
31844 Automatic date update in version.in
31846 2022-07-19 H.J. Lu <hjl.tools@gmail.com>
31848 x86: Disallow invalid relocations against protected symbols
31849 Since glibc 2.36 will issue warnings for copy relocation against
31850 protected symbols and non-canonical reference to canonical protected
31851 functions, change the linker to always disallow such relocations.
31855 * elf32-i386.c (elf_i386_scan_relocs): Remove check for
31856 elf_has_indirect_extern_access.
31857 * elf64-x86-64.c (elf_x86_64_scan_relocs): Likewise.
31858 (elf_x86_64_relocate_section): Remove check for
31859 elf_has_no_copy_on_protected.
31860 * elfxx-x86.c (elf_x86_allocate_dynrelocs): Check for building
31861 executable instead of elf_has_no_copy_on_protected.
31862 (_bfd_x86_elf_adjust_dynamic_symbol): Disallow copy relocation
31863 against non-copyable protected symbol.
31864 * elfxx-x86.h (SYMBOL_NO_COPYRELOC): Remove check for
31865 elf_has_no_copy_on_protected.
31869 * testsuite/ld-i386/i386.exp: Expect linker error for PR ld/17709
31871 * testsuite/ld-i386/pr17709.rd: Removed.
31872 * testsuite/ld-i386/pr17709.err: New file.
31873 * testsuite/ld-x86-64/pr17709.rd: Removed.
31874 * testsuite/ld-x86-64/pr17709.err: New file.
31875 * testsuite/ld-x86-64/pr28875-func.err: Updated.
31876 * testsuite/ld-x86-64/x86-64.exp: Expect linker error for PR
31877 ld/17709 test. Add tests for function pointer against protected
31880 2022-07-19 Fangrui Song <i@maskray.me>
31882 x86: Make protected symbols local for -shared
31883 Call _bfd_elf_symbol_refs_local_p with local_protected==true. This has
31884 2 noticeable effects for -shared:
31886 * GOT-generating relocations referencing a protected data symbol no
31887 longer lead to a GLOB_DAT (similar to a hidden symbol).
31888 * Direct access relocations (e.g. R_X86_64_PC32) no longer has the
31889 confusing diagnostic below.
31891 __attribute__((visibility("protected"))) void *foo() {
31892 return (void *)foo;
31895 // gcc -fpic -shared -fuse-ld=bfd
31896 relocation R_X86_64_PC32 against protected symbol `foo' can not be used when making a shared object
31898 The new behavior matches arm, aarch64 (commit
31899 83c325007c5599fa9b60b8d5f7b84842160e1d1b), and powerpc ports, and other
31900 linkers: gold and ld.lld.
31902 Note: if some code tries to use direct access relocations to take the
31903 address of foo, the pointer equality will break, but the error should be
31904 reported on the executable link, not on the innocent shared object link.
31905 glibc 2.36 will give a warning at relocation resolving time.
31907 With this change, `#define elf_backend_extern_protected_data 1` is no
31908 longer effective. Just remove it.
31910 Remove the test "Run protected-func-1 without PIE" since -fno-pic
31911 address taken operation in the executable doesn't work with protected
31912 symbol in a shared object by default. Similarly, remove
31913 protected-data-1a and protected-data-1b. protected-data-1b can be made
31914 working by removing HAVE_LD_PIE_COPYRELOC from GCC
31915 (https://sourceware.org/pipermail/gcc-patches/2022-June/596678.html).
31917 2022-07-19 Luis Machado <luis.machado@arm.com>
31919 Reformat gdbarch-components.py to fix deviations
31920 Reformat to make sure we have a clean file with no deviations
31921 from the expected python code format.
31923 2022-07-19 Luis Machado <luis.machado@arm.com>
31925 [AArch64] MTE corefile support
31926 Teach GDB how to dump memory tags for AArch64 when using the gcore command
31927 and how to read memory tag data back from a core file generated by GDB
31928 (via gcore) or by the Linux kernel.
31930 The format is documented in the Linux Kernel documentation [1].
31932 Each tagged memory range (listed in /proc/<pid>/smaps) gets dumped to its
31933 own PT_AARCH64_MEMTAG_MTE segment. A section named ".memtag" is created for each
31934 of those segments when reading the core file back.
31936 To save a little bit of space, given MTE tags only take 4 bits, the memory tags
31937 are stored packed as 2 tags per byte.
31939 When reading the data back, the tags are unpacked.
31941 I've added a new testcase to exercise the feature.
31943 Build-tested with --enable-targets=all and regression tested on aarch64-linux
31946 [1] Documentation/arm64/memory-tagging-extension.rst (Core Dump Support)
31948 2022-07-19 Luis Machado <luis.machado@arm.com>
31950 [AArch64] Support AArch64 MTE memory tag dumps in core files
31951 The Linux kernel can dump memory tag segments to a core file, one segment
31952 per mapped range. The format and documentation can be found in the Linux
31955 The following patch adjusts bfd and binutils so they can handle this new
31956 segment type and display it accordingly. It also adds code required so GDB
31957 can properly read/dump core file data containing memory tags.
31959 Upon reading, each segment that contains memory tags gets mapped to a
31960 section named "memtag". These sections will be used by GDB to lookup the tag
31961 data. There can be multiple such sections with the same name, and they are not
31962 numbered to simplify GDB's handling and lookup.
31964 There is another patch for GDB that enables both reading
31965 and dumping of memory tag segments.
31967 Tested on aarch64-linux Ubuntu 20.04.
31969 [1] Documentation/arm64/memory-tagging-extension.rst (Core Dump Support)
31971 2022-07-19 Luis Machado <luis.machado@arm.com>
31973 [AArch64] Fix testcase compilation failure
31974 Newer distros carry newer headers that contains MTE definitions. Account
31975 for that fact in the MTE testcases (gdb.arch/aarch64-mte.exp) and define
31976 constants conditionally to prevent compilation failures.
31978 2022-07-19 H.J. Lu <hjl.tools@gmail.com>
31980 ld: Pass -nostdlib to compiler with -r
31981 Pass -nostdlib to compiler with -r to avoid unnecessary .o file and
31985 * testsuite/ld-elf/linux-x86.exp: Pass -nostdlib with -r.
31987 2022-07-19 H.J. Lu <hjl.tools@gmail.com>
31989 x86: Properly check invalid relocation against protected symbol
31990 Only check invalid relocation against protected symbol defined in shared
31996 * elf32-i386.c (elf_i386_scan_relocs): Only check invalid
31997 relocation against protected symbol defined in shared object.
31998 * elf64-x86-64.c (elf_x86_64_scan_relocs): Likewise.
32003 * testsuite/ld-elf/linux-x86.exp: Run PR ld/29377 tests.
32004 * testsuite/ld-elf/pr29377a.c: New file.
32005 * testsuite/ld-elf/pr29377b.c: Likewise.
32007 2022-07-19 GDB Administrator <gdbadmin@sourceware.org>
32009 Automatic date update in version.in
32011 2022-07-18 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
32013 gprofng: link libgprofng.so against -lpthread
32015 2022-07-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
32018 * src/Makefile.am (libgprofng_la_LIBADD): Add -lpthread.
32019 * src/Makefile.in: Rebuild.
32021 2022-07-18 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
32023 gprofng: fix regression in build and a race condition in autoreconf
32025 2022-07-14 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
32028 * libcollector/configure.ac (AC_CONFIG_HEADERS): Fix a race condition.
32029 * libcollector/configure: Rebuild.
32030 * libcollector/Makefile.in: Rebuild.
32031 * common/config.h.in: Rebuild.
32032 * common/lib-config.h.in: Created by autoreconf.
32034 2022-07-18 Tom Tromey <tromey@adacore.com>
32036 Add gdb.free_objfile event registry
32037 Currently, Python code can use event registries to detect when gdb
32038 loads a new objfile, and when gdb clears the objfile list. However,
32039 there's no way to detect the removal of an objfile, say when the
32040 inferior calls dlclose.
32042 This patch adds a gdb.free_objfile event registry and arranges for an
32043 event to be emitted in this case.
32045 2022-07-18 Pedro Alves <pedro@palves.net>
32047 Put gdb.base/bt-on-fatal-signal.exp GDB cores in output dir
32048 I noticed that gdb.base/bt-on-fatal-signal.exp was contributing four
32049 core files to the count of unexpected core files:
32051 $ make check TESTS="gdb.base/bt-on-fatal-signal.exp"
32053 === gdb Summary ===
32055 # of unexpected core files 4
32056 # of expected passes 21
32058 These are GDB core dumps. They are expected, however, because the
32059 whole point of the testcase is to crash GDB with a signal.
32061 Make GDB change its current directory to the output dir just before
32062 crashing, so that the core files end up there. The result is now:
32064 === gdb Summary ===
32066 # of expected passes 25
32070 $ find . -name "core.*"
32071 ./testsuite/outputs/gdb.base/bt-on-fatal-signal/core.gdb.1676506.nelson.1657727692
32072 ./testsuite/outputs/gdb.base/bt-on-fatal-signal/core.gdb.1672585.nelson.1657727671
32073 ./testsuite/outputs/gdb.base/bt-on-fatal-signal/core.gdb.1674833.nelson.1657727683
32074 ./testsuite/outputs/gdb.base/bt-on-fatal-signal/core.gdb.1673709.nelson.1657727676
32076 (Note the test is skipped at the top if on a remote host.)
32078 Change-Id: I79e4fb2e91330279c7a509930b1952194a72e85a
32080 2022-07-18 Tom Tromey <tromey@adacore.com>
32082 Remove array typedef assumption for Ada
32083 Currently the Ada code assumes that it can distinguish between a
32084 multi-dimensional array and an array of arrays by looking for an
32085 intervening typedef -- that is, for an array of arrays, there will be
32086 a typedef wrapping the innermost array type.
32088 A recent compiler change removes this typedef, which causes a gdb
32089 failure in the internal AdaCore test suite.
32091 This patch handles this case by checking whether the array type in
32092 question has a name.
32094 2022-07-18 Tom Tromey <tromey@adacore.com>
32096 Remove manual lifetime management from cli_interp
32097 cli_interp manually manages its cli_out object. This patch changes it
32098 to use a unique_ptr, and also changes cli_uiout to be a private
32102 cli_out_new is just a small wrapper around 'new'. This patch removes
32103 it, replacing it with uses of 'new' instead.
32105 Replace input_interactive_p with a method
32106 This replaces the global input_interactive_p function with a new
32107 method ui::input_interactive_p.
32109 Remove ui_register_input_event_handler
32110 This patch removes ui_register_input_event_handler and
32111 ui_unregister_input_event_handler, replacing them with methods on
32112 'ui'. It also changes gdb to use these methods everywhere, rather
32113 than sometimes reaching in to the ui to manage the file descriptor
32116 2022-07-18 Claudiu Zissulescu <claziss@gmail.com>
32118 opcodes/arc: Implement style support in the disassembler
32119 Update the ARC disassembler to supply style information to the
32120 disassembler output. The output formatting remains unchanged.
32123 * disassemble.c (disassemble_init_for_target): Set
32124 created_styled_output for ARC based targets.
32125 * arc-dis.c (find_format_from_table): Use fprintf_styled_ftype
32126 instead of fprintf_ftype throughout.
32127 (find_format): Likewise.
32128 (print_flags): Likewise.
32129 (print_insn_arc): Likewise.
32131 2022-07-18 Claudiu Zissulescu <claziss@synopsys.com>
32133 arc: Update missing cipher.
32134 The ciphers 5,7, and 9 are missing when parsing an assembly
32135 instruction leading to errors when those ciphers are
32139 * tc-arc.c (md_assembly): Update strspn string with the
32142 2022-07-18 Tom de Vries <tdevries@suse.de>
32144 [gdb/testsuite] Remove duplicate of supports_gnuc
32145 In commit 9d9dd861e98 ("[gdb/testsuite] Fix regression in
32146 step-indirect-call-thunk.exp with gcc 7") I accidentally committed a duplicate
32147 of supports_gnuc, which caused:
32149 DUPLICATE: gdb.base/gdb-caching-proc.exp: supports_gnuc: consistency
32152 Fix this by removing the duplicate.
32154 Tested on x86_64-linux.
32156 2022-07-18 Andrew Burgess <aburgess@redhat.com>
32158 gdb/python: look for python, then python 3 at configure time
32159 It is possible that a system might have a python3 executable, but no
32160 python executable. For example, on my Fedora system the python2
32161 package provides /usr/bin/python2, the python3 package provides
32162 /usr/bin/python3, and the python-unversioned-command package provides
32163 /usr/bin/python, which picks between python2 and python3.
32165 It is quite possible to only have python3 available on a system.
32167 Currently, when GDB configures, it looks for a 'python' executable.
32168 If non is found then GDB will be built without python support. Or the
32169 user needs to configure using --with-python=/usr/bin/python3.
32171 This commit updates GDB's configure.ac script to first look for
32172 'python', and then 'python3'. Now, on a system that only has a
32173 python3 executable, GDB will automatically find, and use that in order
32174 to provide python support, no user supplied configure arguments are
32177 I've tested this on my local machine by removing the
32178 python-unversioned-command package, confirming that there is no longer
32179 a 'python' executable in my $PATH, and then rebuilding GDB from
32180 scratch. GDB with this patch has python support.
32182 2022-07-18 Jan Beulich <jbeulich@suse.com>
32184 x86: correct VMOVSH attributes
32185 Both forms were missing VexW0 (thus allowing Evex.W=1 to be encoded by
32186 suitable means, which would cause #UD). The memory operand form further
32187 was using the wrong Masking value, thus allowing zeroing-masking to be
32188 encoded for the store form (which would again cause #UD).
32190 x86: re-order insn template fields
32191 This saves quite a number of shift instructions: The "operands" field
32192 can now be retrieved by just masking (no shift), and extracting the
32193 "extension_opcode" field now only requires a (signed) right shift, with
32194 no prereq left one. (Of course there may be architectures where, in a
32195 cross build, there might be no difference at all, e.g. when there are
32196 suitable bitfield extraction insns.)
32198 2022-07-18 Tom de Vries <tdevries@suse.de>
32200 [gdbsupport] Improve thread scheduling in parallel_for_each
32201 When running a task using parallel_for_each, we get the following
32204 Parallel for: n_elements: 7271
32205 Parallel for: minimum elements per thread: 10
32206 Parallel for: elts_per_thread: 1817
32207 Parallel for: elements on worker thread 0 : 1817
32208 Parallel for: elements on worker thread 1 : 1817
32209 Parallel for: elements on worker thread 2 : 1817
32210 Parallel for: elements on worker thread 3 : 0
32211 Parallel for: elements on main thread : 1820
32214 Note that there are 4 active threads, and scheduling elts_per_thread on each
32215 of those handles 4 * 1817 == 7268, leaving 3 "left over" elements.
32217 These leftovers are currently handled in the main thread.
32219 That doesn't seem to matter much for this example, but for say 10 threads and
32220 99 elements, you'd have 9 threads handling 9 elements and 1 thread handling 18
32223 Instead, distribute the left over elements over the worker threads, such that
32226 Parallel for: elements on worker thread 0 : 1818
32227 Parallel for: elements on worker thread 1 : 1818
32228 Parallel for: elements on worker thread 2 : 1818
32229 Parallel for: elements on worker thread 3 : 0
32230 Parallel for: elements on main thread : 1817
32233 Tested on x86_64-linux.
32235 2022-07-18 Tom de Vries <tdevries@suse.de>
32237 [gdb/testsuite] Allow override of ASAN_OPTIONS in lib/gdb.exp
32238 Use set_sanitizer_default for ASAN_OPTIONS in lib/gdb.exp.
32240 This allows us to override the default detect_leaks=0 setting, by manually
32243 $ export ASAN_OPTIONS=detect_leaks=1
32247 Tested on x86_64-linux, by building with -fsanitize=address and running
32248 test-case gdb.dwarf2/gdb-add-index.exp with and without
32249 "export ASAN_OPTIONS=detect_leaks=1".
32251 2022-07-18 Tom de Vries <tdevries@suse.de>
32253 [gdb/testsuite] Fix regression in step-indirect-call-thunk.exp with gcc 7
32254 Since commit 43127ae5714 ("Fix gdb.base/step-indirect-call-thunk.exp") I run
32257 gdb compile failed, gcc: error: unrecognized command line option \
32258 '-fcf-protection=none'; did you mean '-flto-partition=none'?
32259 UNTESTED: gdb.base/step-indirect-call-thunk.exp: failed to prepare
32262 The problem is that -fcf-protection is supported starting gcc 8, but I'm using
32265 Fix this by only adding -fcf-protection=none for gcc 8 and later.
32267 Tested on x86_64-linux, with gcc 7.5.0, 8.2.1 and 12.1.1.
32269 2022-07-18 Tom de Vries <tdevries@suse.de>
32271 [gdb/testsuite] Fix gdb.arch/i386-mpx.exp
32272 Since commit c4a3dbaf113 ("Expose current 'print' settings to Python") we
32275 (gdb) print /x $bnd0 = {0x10, 0x20}^M
32276 $22 = {lbound = 0x10, ubound = 0x20} : size 0x11^M
32277 (gdb) FAIL: gdb.arch/i386-mpx.exp: verify size for bnd0
32280 The regexp in the test-case expects "size 17".
32282 Fix this by updating the regexp.
32284 Tested on x86_64-linux.
32286 2022-07-18 Tom de Vries <tdevries@suse.de>
32288 [gdbsupport] Add parallel_for_each_debug
32289 Add a parallel_for_each_debug variable, set to false by default.
32291 With an a.out compiled from hello world, we get with
32292 parallel_for_each_debug == true:
32294 $ gdb -q -batch a.out -ex start
32296 Parallel for: n_elements: 7271
32297 Parallel for: minimum elements per thread: 10
32298 Parallel for: elts_per_thread: 1817
32299 Parallel for: elements on worker thread 0 : 1817
32300 Parallel for: elements on worker thread 1 : 1817
32301 Parallel for: elements on worker thread 2 : 1817
32302 Parallel for: elements on worker thread 3 : 0
32303 Parallel for: elements on main thread : 1820
32305 Temporary breakpoint 1, main () at /home/vries/hello.c:6
32306 6 printf ("hello\n");
32309 Tested on x86_64-linux.
32311 2022-07-18 GDB Administrator <gdbadmin@sourceware.org>
32313 Automatic date update in version.in
32315 2022-07-17 GDB Administrator <gdbadmin@sourceware.org>
32317 Automatic date update in version.in
32319 2022-07-16 GDB Administrator <gdbadmin@sourceware.org>
32321 Automatic date update in version.in
32323 2022-07-15 Aaron Merey <amerey@redhat.com>
32325 gdb-add-index always generates an error when libdebuginfod wasn't compiled in
32326 gdb-add-index runs gdb with -iex 'set debuginfod enabled off'. If gdb
32327 is not compiled against libdebuginfod this causes an unnecessary error
32328 message to be printed to stderr indicating that gdb was not built with
32329 debuginfod support.
32331 Fix this by changing the 'set debuginfod enabled off' command to a
32332 no-op when gdb isn't built with libdebuginfod.
32334 2022-07-15 Bruno Larsen <blarsen@redhat.com>
32336 gdb/testsuite: modernize gdb.base/maint.exp
32337 gdb.base/maint.exp was using several gdb_expect statements, probably
32338 because this test case predates the existance of gdb_test_multiple. This
32339 commit updates the test case to use gdb_test_multiple, making it more
32340 resilient to internal errors and such.
32342 The only gdb_expect left in the testcase is one that specifically looks
32343 for an internal error being triggered as a PASS.
32345 2022-07-15 Tom Tromey <tromey@adacore.com>
32347 Add 'nibbles' to gdb.print_options
32348 When I rebased and updated the print_options patch, I forgot to update
32349 print_options to add the new 'nibbles' feature to the result. This
32350 patch fixes the oversight. I'm checking this in.
32352 2022-07-15 Carl Love <cel@us.ibm.com>
32354 PowerPC: Add support for IEEE 128-bit format.
32355 The test gdb.base/infcall-nested-structs-c.exp fails on a gdb assert
32356 in function ppc64_sysv_abi_return_value in file gdb/ppc-sysv-tdep.c. The
32357 assert is due to the missing IEEE 128-bit support in file
32358 gdb/ppc-sysv-tdep.c.
32360 The IBM long double was the initial float 128-bit support added by IBM
32361 The IEEE 128-bit support, which is similar IBM long double support, was
32362 made the default starting with GCC 12. The floating point format
32363 differences include the number of bits used to encode the exponent
32364 and significand. Also, IBM long double values are passed in a pair of
32365 floating point registers. The IEEE 128-bit value is passed in a single
32368 This patch fixes the gdb_assert (ok); in function
32369 ppc64_sysv_abi_return_value in gdb/ppc-sysv-tdep.c by adding IEEE FLOAT
32370 128-bit type support for PowerPC.
32372 The patch has been tested on Power 10, ELFv2. It fixes the following list
32373 of regression failures on Power 10:
32375 gdb.base/infcall-nested-structs-c.exp 192
32376 gdb.base/infcall-nested-structs-c++.exp 76
32377 gdb.base/structs.exp 9
32379 The patch has been tested on Power 8 BE which is ELFv1.
32381 2022-07-15 Tom Tromey <tromey@adacore.com>
32383 Add 'summary' mode to Value.format_string
32384 This adds a 'summary' mode to Value.format_string and to
32385 gdb.print_options. For the former, it lets Python code format values
32386 using this mode. For the latter, it lets a printer potentially detect
32387 if it is being called in a backtrace with 'set print frame-arguments'
32390 I considered adding a new mode here to let a pretty-printer see
32391 whether it was being called in a 'backtrace' context at all, but I'm
32392 not sure if this is really desirable.
32394 2022-07-15 Tom Tromey <tromey@adacore.com>
32396 Expose current 'print' settings to Python
32397 PR python/17291 asks for access to the current print options. While I
32398 think this need is largely satisfied by the existence of
32399 Value.format_string, it seemed to me that a bit more could be done.
32401 First, while Value.format_string uses the user's settings, it does not
32402 react to temporary settings such as "print/x". This patch changes
32405 Second, there is no good way to examine the current settings (in
32406 particular the temporary ones in effect for just a single "print").
32407 This patch adds this as well.
32409 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=17291
32411 2022-07-15 Carl Love <cel@us.ibm.com>
32413 PowerPC: fix for gdb.base/eh_return.exp
32414 Disable the Traceback Table generation on PowerPC for this test. The
32415 Traceback Table consists of a series of bit fields to indicate things like
32416 the Traceback Table version, language, and specific information about the
32417 function. The Traceback Table is generated following the end of the code
32418 for every function by default. The Traceback Table is defined in the
32419 PowerPC ELF ABI and is intended to support debuggers and exception
32420 handlers. The Traceback Table is displayed in the disassembly of functions
32421 by default and is part of the function length. The table is typically
32422 interpreted by the disassembler as data represented by .long xxx entries.
32424 Generation of the Traceback Table is disabled in this test using the
32425 PowerPC specific gcc compiler option -mtraceback=no, the xlc option
32426 additional_flags-qtable=none and the clang optons
32427 -mllvm -xcoff-traceback-table=false. Disabling the Traceback Table
32428 generation in this test results in the gdb_test_multiple statement
32429 correctly locating the address of the bclr instruction before the statement
32430 "End of assembler dump." in the disassembly output.
32432 2022-07-15 Tom Tromey <tromey@adacore.com>
32435 Running 'black' on gdb fixed a couple of small issues. This patch is
32438 2022-07-15 GDB Administrator <gdbadmin@sourceware.org>
32440 Automatic date update in version.in
32442 2022-07-14 Tom de Vries <tdevries@suse.de>
32444 [gdb/symtab] Fix data race in cooked_index_functions::expand_symtabs_matching
32445 When building gdb with -fsanitize-threads and running test-case
32446 gdb.ada/char_enum_unicode.exp, I run into:
32448 WARNING: ThreadSanitizer: data race (pid=21301)^M
32449 Write of size 8 at 0x7b2000008080 by main thread:^M
32450 #0 free <null> (libtsan.so.2+0x4c5e2)^M
32451 #1 _dl_close_worker <null> (ld-linux-x86-64.so.2+0x4b7b)^M
32452 #2 convert_between_encodings() charset.c:584^M
32454 #21 cooked_index_functions::expand_symtabs_matching() read.c:18606
32457 This is fixed by making cooked_index_functions::expand_symtabs_matching wait
32458 for the cooked index finalization to be done.
32460 Tested on x86_64-linux.
32462 https://sourceware.org/bugzilla/show_bug.cgi?id=29311
32463 https://sourceware.org/bugzilla/show_bug.cgi?id=29286
32465 2022-07-14 Tom de Vries <tdevries@suse.de>
32467 [gdbsupport] Add sequential_for_each
32468 Add a sequential_for_each alongside the parallel_for_each, which can be used
32469 as a drop-in replacement.
32471 This can be useful when debugging multi-threading behaviour, and you want to
32472 limit multi-threading in a fine-grained way.
32474 Tested on x86_64-linux, by using it instead of the parallel_for_each in
32475 dwarf2_build_psymtabs_hard.
32477 2022-07-14 Tiezhu Yang <yangtiezhu@loongson.cn>
32479 gdb: Document floating-point support for LoongArch
32480 Update NEWS and gdb.texinfo to document floating-point support
32483 2022-07-14 Tom de Vries <tdevries@suse.de>
32485 [gdb/build] Fix gdb build with gcc 4.8.5
32486 When building gdb with gcc 4.8.5, we run into:
32488 In file included from /usr/include/c++/4.8/future:43:0,
32489 from gdbsupport/thread-pool.h:30,
32490 from gdb/dwarf2/cooked-index.h:33,
32491 from gdb/dwarf2/read.h:26,
32492 from gdb/dwarf2/abbrev-cache.c:21:
32493 /usr/include/c++/4.8/atomic: In instantiation of \
32494 ‘_Tp std::atomic<_Tp>::load(std::memory_order) const [with _Tp = \
32495 packed<dwarf_unit_type, 1ul>; std::memory_order = std::memory_order]’:
32496 gdb/dwarf2/read.h:332:44: required from here
32497 /usr/include/c++/4.8/atomic:208:13: error: no matching function for call to \
32498 ‘packed<dwarf_unit_type, 1ul>::packed()’
32503 Fix this by adding the default constructor for packed.
32505 Tested on x86_64-linux, with gdb build with gcc 4.8.5.
32507 2022-07-14 Tom de Vries <tdevries@suse.de>
32509 [gdb/symtab] Make per_cu->m_lang atomic
32510 When building gdb with -fsanitize=thread and running test-case
32511 gdb.dwarf2/inlined_subroutine-inheritance.exp, we run into a data race
32514 Read of size 1 at 0x7b2000003010 by thread T4:
32515 #0 packed<language, 1ul>::operator language() const packed.h:54
32516 #1 dwarf2_per_cu_data::set_lang(language) read.h:363
32520 Previous write of size 1 at 0x7b2000003010 by main thread:
32521 #0 dwarf2_per_cu_data::set_lang(language) read.h:365
32524 Fix this by making per_cu->m_lang atomic.
32526 Tested on x86_64-linux.
32528 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29286
32530 2022-07-14 Tom de Vries <tdevries@suse.de>
32532 [gdb/symtab] Make per_cu->unit_type atomic
32533 With gdb with -fsanitize=thread and test-case gdb.ada/array_bounds.exp, I run
32534 into a data race between:
32536 Read of size 1 at 0x7b2000025f0f by main thread:
32537 #0 packed<dwarf_unit_type, 1ul>::operator dwarf_unit_type() const packed.h:54
32538 #1 dwarf2_per_cu_data::set_unit_type(dwarf_unit_type) read.h:339
32542 Previous write of size 1 at 0x7b2000025f0f by thread T3:
32543 #0 dwarf2_per_cu_data::set_unit_type(dwarf_unit_type) read.h:341
32546 Fix this by making per_cu->unit_type atomic.
32548 Tested on x86_64-linux.
32550 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29286
32552 2022-07-14 Tom de Vries <tdevries@suse.de>
32554 [gdb/symtab] Fix data race in ~charset_vector
32557 $ gdb ./outputs/gdb.ada/char_enum_unicode/foo -batch -ex "break foo.adb:26"
32559 with a gdb build with -fsanitize=thread I run into a data race:
32561 WARNING: ThreadSanitizer: data race (pid=30917)
32562 Write of size 8 at 0x7b0400004070 by main thread:
32563 #0 free <null> (libtsan.so.2+0x4c5e2)
32564 #1 xfree<char> gdbsupport/gdb-xfree.h:37 (gdb+0x650f17)
32565 #2 charset_vector::clear() gdb/charset.c:703 (gdb+0x651354)
32566 #3 charset_vector::~charset_vector() gdb/charset.c:697 (gdb+0x6512d3)
32567 #4 <null> <null> (libtsan.so.2+0x32643)
32568 #5 captured_main_1 gdb/main.c:1310 (gdb+0xa3975a)
32571 The problem is that we're freeing the charset_vector elements in the destructor,
32572 which may still be used by a worker thread.
32574 Fix this by not freeing the charset_vector elements in the destructor.
32576 Tested on x86_64-linux.
32578 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29311
32580 2022-07-14 Alan Modra <amodra@gmail.com>
32582 Re: PowerPC: implement md_operand to parse register names
32583 I meant to make this change before committing, to let compilers know
32584 the code on the false branch of md_parse_name is dead.
32586 * config/tc-ppc.c (ppc_parse_name): Return void.
32587 * config/tc-ppc.h (md_parse_name): Always true.
32588 (ppc_parse_name): Update prototype.
32590 2022-07-14 Alan Modra <amodra@gmail.com>
32592 PowerPC: implement md_operand to parse register names
32593 Allows register names to appear in symbol assignments, so for example
32598 * gas/config/tc-ppc.c (REG_NAME_CNT): Delete, replace uses with
32600 (register_name): Rename to..
32601 (md_operand): ..this. Only handle %reg.
32602 (cr_names): Rename to..
32603 (cr_cond): ..this. Just keep conditions.
32604 (ppc_parse_name): Add mode param. Search both cr_cond and
32605 pre_defined_registers. Handle absolute and register symbol
32606 values here rather than in expr.c:operand().
32607 (md_assemble): Don't special case register name matching in
32608 operands, except to set cr_operand as appropriate.
32609 * gas/config/tc-ppc.h (md_operand): Don't define.
32610 (md_parse_name, ppc_parse_name): Update.
32611 * read.c (pseudo_set): Copy over entire O_register value.
32612 * testsuite/gas/ppc/regsyms.d.
32613 * testsuite/gas/ppc/regsyms.s: New test.
32614 * testsuite/gas/ppc/ppc.exp: Run it.
32616 2022-07-14 GDB Administrator <gdbadmin@sourceware.org>
32618 Automatic date update in version.in
32620 2022-07-13 Pedro Alves <pedro@palves.net>
32622 Tighten gdb.threads/no-unwaited-for-left.exp regexps
32623 A WIP version of a patch
32624 (https://sourceware.org/pipermail/gdb-patches/2022-June/190202.html)
32625 resulted in a bug that went unnoticed by the testuite, like so:
32627 (gdb) PASS: gdb.threads/no-unwaited-for-left.exp: enable scheduler-locking, for main thread
32630 [New Thread 1251861.1251861]
32631 No unwaited-for children left.
32632 (gdb) PASS: gdb.threads/no-unwaited-for-left.exp: continue stops when the main thread exits
32635 3 Thread 1251861.1251863 "no-unwaited-for" __pthread_clockjoin_ex (threadid=140737351558976, thread_return=0x0, clockid=<optimized out>, abstime=<optimized out>, block=<optimized out>) at pthread_join_common.c:145
32636 4 Thread 1251861.1251861 "no-unwaited-for" <unavailable> in ?? ()
32638 The current thread <Thread ID 1> has terminated. See `help thread'.
32639 (gdb) PASS: gdb.threads/no-unwaited-for-left.exp: only thread 3 left, main thread terminated
32641 Somehow, above, GDB re-added the zombie leader back before printing
32642 "No unwaited-for children left.". The "only thread 3 left, main
32643 thread terminated" test should have caught this, but didn't. That is
32644 because the test's regexp has a ".*" after the part that matches
32645 thread 3. This commit tightens that regexp to catch such a bug. It
32646 also tightens the "only main thread left, thread 2 terminated" test's
32647 regexp in the same way.
32649 Change-Id: I8744f327a0aa0e2669d1ddda88247e99b91cefff
32651 2022-07-13 Carl Love <cel@us.ibm.com>
32653 Fix for gdb.base/stap-probe.c
32654 On PowePC, the test fails on a compile error:
32656 /../binutils-gdb-current/gdb/testsuite/gdb.base/stap-probe.c:107:1: error: expected '=', ',', ';', 'asm' or 'attribute' before 'use_xmm_reg'
32657 107 | use_xmm_reg (int val)
32660 Where the source code for stap-probe.c is:
32662 static const char * __attribute__((noinline)) ATTRIBUTE_NOCLONE
32663 use_xmm_reg (int val) <-- line 107
32667 The issue is the ATTRIBUTE_NOCLONE is not defined as an attribute as
32668 expected. The #define for ATTRIBUTE_NOCLONE can be found in
32669 ../lib/attributes.h.
32671 This patch adds the missing include statement for the definition of
32674 The patch has been tested and verified on a Power10 system.
32676 2022-07-13 Carl Love <cel@us.ibm.com>
32678 Add PowerPC support to gdb.cp/call-method-register.cc
32679 This patch adds the needed define ASM_REG for PowerPC.
32681 The patch was run on a Power 10 system. The gdb Summary for the run lists
32682 2 expected passes, no unexpected failures or untested testcases.
32684 Please let me know if this patch is acceptable for mainline.
32688 2022-07-13 Carl Love <cel@us.ibm.com>
32690 Fix gdb.base/step-indirect-call-thunk.exp
32691 Due to recent changes in the default value of -fcf-protection for gcc, the
32692 test gdb.base/step-indirect-call-thunk.exp fails on Intel X86-64 with the
32695 Executing on host: gcc -fno-stack-protector -fdiagnostics-color=never
32696 -mindirect-branch=thunk -mfunction-return=thunk -c -g
32697 -o /.../gdb/testsuite/outputs/gdb.base/step-indirect-call-thunk/step-indirect-call-thunk0.o
32698 /.../gdb/testsuite/gdb.base/step-indirect-call-thunk.c
32699 (timeout = 300) builtin_spawn -ignore SIGHUP gcc -fno-stack-protector
32700 -fdiagnostics-color=never -mindirect-branch=thunk -mfunction-return=thunk -c
32701 -g -o /.../gdb/testsuite/outputs/gdb.base/step-indirect-call-thunk/step-indirect-call-thunk0.o
32702 /.../binutils-gdb-current/gdb/testsuite/gdb.base/step-indirect-call-thunk.c
32703 /.../gdb/testsuite/gdb.base/step-indirect-call-thunk.c:
32704 In function 'inc': /.../gdb/testsuite/gdb.base/step-indirect-call-thunk.c:
32705 22:1: error: '-mindirect-branch' and '-fcf-protection' are not compatible
32708 As stated in the error message the default "-fcf-protection" and
32709 "-mindirect-branch' are in compatible. The fcf-protection argument needs
32710 to be "-fcf-protection=none" for the test to compile on Intel.
32712 The gcc command line "-mindirect-branch' is an Intel specific and will give
32713 an error on other platforms. A check for X86 is added so the test will
32714 only run on X86 platforms.
32716 The patch has been tested and verified on Power 10 and Intel X86-64 systems
32717 with no regressions.
32719 2022-07-13 Pedro Alves <pedro@palves.net>
32721 Fix "until LINE" in main, when "until" runs into longjmp
32722 With a test like this:
32724 1 #include <dlfcn.h>
32728 5 dlsym (RTLD_DEFAULT, "FOO");
32732 and then "start" followed by "until 6", GDB currently incorrectly
32733 stops inside the runtime loader, instead of line 6. Vis:
32736 Temporary breakpoint 1, main () at until.c:5
32739 0x00007ffff7f0a90d in __GI__dl_catch_exception (exception=exception@entry=0x7fffffffdb00, operate=<optimized out>, args=0x7ffff7f0a90d <__GI__dl_catch_exception+109>) at dl-error-skeleton.c:206
32740 206 dl-error-skeleton.c: No such file or directory.
32743 The problem is related to longjmp handling -- dlsym internally
32744 longjmps on error. The testcase can be reduced to this:
32746 1 #include <setjmp.h>
32749 4 if (setjmp (buf) == 0)
32750 5 longjmp (buf, 1);
32755 10 return 0; /* until to here */
32758 and then with "start" followed by "until 10", GDB currently
32759 incorrectly stops at line 4 (returning from setjmp), instead of line
32762 The problem is that the BPSTAT_WHAT_CLEAR_LONGJMP_RESUME code in
32763 infrun.c fails to find the initiating frame, and so infrun thinks that
32764 the longjmp jumped somewhere outer to "until"'s originating frame.
32768 case BPSTAT_WHAT_CLEAR_LONGJMP_RESUME:
32770 struct frame_info *init_frame;
32772 /* There are several cases to consider.
32774 1. The initiating frame no longer exists. In this case we
32775 must stop, because the exception or longjmp has gone too
32780 init_frame = frame_find_by_id (ecs->event_thread->initiating_frame);
32782 if (init_frame) // this is NULL!
32787 /* For Cases 1 and 2, remove the step-resume breakpoint, if it
32789 delete_step_resume_breakpoint (ecs->event_thread);
32791 end_stepping_range (ecs); // case 1., so we stop.
32794 The initiating frame is set by until_break_command ->
32795 set_longjmp_breakpoint. The initiating frame is supposed to be the
32796 frame that is selected when the command was issued, but
32797 until_break_command instead passes the frame id of the _caller_ frame
32798 by mistake. When the "until LINE" command is issued from main, the
32799 caller frame is the caller of main. When later infrun tries to find
32800 that frame by id, it fails to find it, because frame_find_by_id
32801 doesn't unwind past main.
32803 The bug is that we passed the caller frame's id to
32804 set_longjmp_breakpoint. We should have passed the selected frame's id
32807 Change-Id: Iaae1af7cdddf296b7c5af82c3b5b7d9b66755b1c
32809 2022-07-13 Enze Li <enze.li@hotmail.com>
32811 gdbserver: remove unused variable
32812 When building with clang 15, I got this error:
32815 server.cc:2985:10: error: variable 'new_argc' set but not used [-Werror,-Wunused-but-set-variable]
32818 Remove the unused variable to eliminate the error.
32820 Tested by rebuilding on x86_64-linux with clang 15.
32822 2022-07-13 Tom de Vries <tdevries@suse.de>
32824 [gdb/symtab] Make per_cu->set_lang more strict
32825 We have in per_cu->set_lang this comment:
32827 void set_lang (enum language lang)
32829 /* We'd like to be more strict here, similar to what is done in
32830 set_unit_type, but currently a partial unit can go from unknown to
32831 minimal to ada to c. */
32834 Fix this by not setting m_lang for partial units.
32836 This requires us to move the m_unit_type initialization to ensure that
32837 m_unit_type is initialized before per_cu->m_lang.
32839 Tested on x86_64-linux, with native and target board cc-with-dwz-m.
32841 2022-07-13 GDB Administrator <gdbadmin@sourceware.org>
32843 Automatic date update in version.in
32845 2022-07-12 Pedro Alves <pedro@palves.net>
32847 Improve "set scheduler-locking" documentation
32848 This improves the "set scheduler-locking" documentation in the GDB
32851 - Use a table to describe the four available modes.
32853 - Describe "step" in terms of "on" and "off".
32855 - Tweak the "replay" mode's description to describe replay first
32856 instead of recording, and also mention how the mode behaves during
32859 - Say what is the default mode.
32861 Change-Id: Ie12140138b37534b7fc1d904da34f0f174aa11ce
32863 2022-07-12 Tom de Vries <tdevries@suse.de>
32865 [gdb/symtab] Add dwarf2_cu::lang ()
32866 The cu->per_cu->lang field was added to carry information from the initial
32867 partial symtabs phase to the symtab expansion phase, for the benefit of a
32868 particular optimization in process_imported_unit_die.
32870 Other uses have been added, but since the first phase now has been
32871 parallelized, those have become problematic and sources of race conditions.
32873 Fix this by adding dwarf2_cu::lang () and using it where we can to replace
32874 cu->per_cu->lang () with cu->lang ().
32876 Also assert in dwarf2_cu::lang () that we're not returning language_unknown.
32878 Tested on x86_64-linux.
32880 2022-07-12 Martin Liska <mliska@suse.cz>
32882 LTO plugin: sync header file with GCC
32885 * plugin-api.h (enum ld_plugin_tag): Sync with GCC.
32887 2022-07-12 Tom de Vries <tdevries@suse.de>
32889 [gdb/record] Support recording of getrandom
32890 Add missing support for recording of linux syscall getrandom.
32892 Tested on x86_64-linux with native and target board unix/-m32.
32894 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=22081
32896 2022-07-12 Tiezhu Yang <yangtiezhu@loongson.cn>
32898 gdbserver: LoongArch: Add floating-point support
32899 This commit adds floating-point support for LoongArch gdbserver.
32901 gdb: LoongArch: Add floating-point support
32902 This commit adds floating-point support for LoongArch gdb.
32904 2022-07-12 Tom de Vries <tdevries@suse.de>
32906 [gdb/testsuite] Run two test-cases with ASAN_OPTIONS=verify_asan_link_order=0
32907 When building gdb with -fsanitize=address we run into:
32909 builtin_spawn gdb -nw -nx -iex set height 0 -iex set width 0 -data-directory \
32910 build/gdb/data-directory^M
32911 ==10637==ASan runtime does not come first in initial library list; you \
32912 should either link runtime to your application or manually preload it with \
32914 ERROR: GDB process no longer exists
32917 Prevent the ASan runtime error by using
32918 ASAN_OPTIONS=verify_asan_link_order=0. This makes both test-cases pass.
32920 Tested on x86_64-linux.
32922 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29358
32924 2022-07-12 Tom de Vries <tdevries@suse.de>
32926 [gdb/testsuite] Add tsan-suppressions.txt
32927 Add a new file tsan-suppressions.txt, to suppress the "unlock unlocked mutex"
32928 problem in ncurses, filed in PR29328.
32930 The file is added to the TSAN_OPTIONS in lib/gdb.exp.
32932 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29328
32934 2022-07-12 Tom de Vries <tdevries@suse.de>
32936 [gdb/build] Fix build with gcc 4.8.5
32937 When building gdb with gcc 4.8.5, we run into problems due to unconditionally
32940 gdb_static_assert (std::is_trivially_copyable<packed>::value);
32942 in gdbsupport/packed.h.
32944 Fix this by guarding the usage with HAVE_IS_TRIVIALLY_COPYABLE.
32946 Tested by doing a full gdb build with gcc 4.8.5.
32948 2022-07-12 Tom de Vries <tdevries@suse.de>
32950 Fix -fsanitize=thread for per_cu fields
32951 When building gdb with -fsanitize=thread and gcc 12, and running test-case
32952 gdb.dwarf2/dwz.exp, we run into a data race between:
32954 Read of size 1 at 0x7b200000300d by thread T2:^M
32955 #0 cutu_reader::cutu_reader(dwarf2_per_cu_data*, dwarf2_per_objfile*, \
32956 abbrev_table*, dwarf2_cu*, bool, abbrev_cache*) gdb/dwarf2/read.c:6164 \
32961 Previous write of size 1 at 0x7b200000300d by main thread:^M
32962 #0 prepare_one_comp_unit gdb/dwarf2/read.c:23588 (gdb+0x86f973)^M
32965 In other words, between:
32967 if (this_cu->reading_dwo_directly)
32971 cu->per_cu->lang = pretend_language;
32974 Likewise, we run into a data race between:
32976 Write of size 1 at 0x7b200000300e by thread T4:
32977 #0 process_psymtab_comp_unit gdb/dwarf2/read.c:6789 (gdb+0x830720)
32981 Previous read of size 1 at 0x7b200000300e by main thread:
32982 #0 cutu_reader::cutu_reader(dwarf2_per_cu_data*, dwarf2_per_objfile*, \
32983 abbrev_table*, dwarf2_cu*, bool, abbrev_cache*) gdb/dwarf2/read.c:6164 \
32987 In other words, between:
32989 this_cu->unit_type = DW_UT_partial;
32993 if (this_cu->reading_dwo_directly)
32996 Likewise for the write to addresses_seen in cooked_indexer::check_bounds and a
32997 read from is_dwz in dwarf2_find_containing_comp_unit for test-case
32998 gdb.dwarf2/dw2-dir-file-name.exp and target board cc-with-dwz-m.
33000 The problem is that the written fields are part of the same memory location as
33001 the read fields, so executing a read and write in different threads is
33002 undefined behavour.
33004 Making the written fields separate memory locations, using the new
33005 struct packed template fixes this.
33007 The set of fields has been established experimentally to be the
33008 minimal set to get rid of this type of -fsanitize=thread errors, but
33009 more fields might require the same treatment.
33011 Looking at the properties of the lang field, unlike dwarf_version it's
33012 not available in the unit header, so it will be set the first time
33013 during the parallel cooked index reading. The same holds for
33014 unit_type, and likewise for addresses_seen.
33016 dwarf2_per_cu_data::addresses_seen is moved so that the bitfields that
33017 currently follow it can be merged in the same memory location as the
33018 bitfields that currently precede it, for better packing.
33020 Tested on x86_64-linux.
33022 Co-Authored-By: Pedro Alves <pedro@palves.net>
33024 Change-Id: Ifa94f0a2cebfae5e8f6ddc73265f05e7fd9e1532
33026 2022-07-12 Pedro Alves <pedro@palves.net>
33028 Introduce struct packed template
33029 When building gdb with -fsanitize=thread and gcc 12, and running test-case
33030 gdb.dwarf2/dwz.exp, we run into a few data races. For example, between:
33033 Write of size 1 at 0x7b200000300e by thread T4:
33034 #0 process_psymtab_comp_unit gdb/dwarf2/read.c:6789 (gdb+0x830720)
33040 Previous read of size 1 at 0x7b200000300e by main thread:
33041 #0 cutu_reader::cutu_reader(dwarf2_per_cu_data*, dwarf2_per_objfile*, \
33042 abbrev_table*, dwarf2_cu*, bool, abbrev_cache*) gdb/dwarf2/read.c:6164 \
33046 In other words, between:
33049 this_cu->unit_type = DW_UT_partial;
33055 if (this_cu->reading_dwo_directly)
33058 The problem is that the written fields are part of the same memory
33059 location as the read fields, so executing a read and write in
33060 different threads is undefined behavour.
33062 Making the written fields separate memory locations, like this:
33066 ENUM_BITFIELD (dwarf_unit_type) unit_type : 8;
33070 fixes it, however that also increases the size of struct
33071 dwarf2_per_cu_data, because it introduces padding due to alignment of
33072 these new structs, which align on the natural alignment of the
33073 specified type of their fields. We can fix that with
33074 __attribute__((packed)), like so:
33077 ENUM_BITFIELD (dwarf_unit_type) unit_type : 8 __attribute__((packed));
33080 but to avoid having to write that in several places and add suitable
33081 comments explaining how that concoction works, introduce a new struct
33082 packed template that wraps/hides this. Instead of the above, we'll be
33085 packed<dwarf_unit_type, 1> unit_type;
33087 Note that we can't change the type of dwarf_unit_type, as that is
33088 defined in include/, and shared with other projects, some of those
33091 This patch just adds the struct packed type. Following patches will
33092 make use of it. One of those patches will want to wrap a struct
33093 packed in an std::atomic, like:
33095 std::atomic<std::packed<language, 1>> m_lang;
33097 so the new gdbsupport/packed.h header adds some operators to make
33098 comparisions between that std::atomic and the type that the wrapped
33099 struct packed wraps work, like in:
33101 if (m_lang == language_c)
33103 It would be possible to implement struct packed without using
33104 __attribute__((packed)), by having it store an array of bytes of the
33105 appropriate size instead, however that would make it less convenient
33106 to debug GDB. The way it's implemented, printing a struct packed
33107 variable just prints its field using its natural type, which is
33108 particularly useful if the type is an enum. I believe that
33109 __attribute__((packed)) is supported by all compilers that are able to
33110 build GDB. Even a few BFD headers use on ATTRIBUTE_PACKED on external
33113 include/coff/external.h: } ATTRIBUTE_PACKED
33114 include/coff/external.h:} ATTRIBUTE_PACKED ;
33115 include/coff/external.h:} ATTRIBUTE_PACKED ;
33116 include/coff/pe.h:} ATTRIBUTE_PACKED ;
33117 include/coff/pe.h:} ATTRIBUTE_PACKED;
33118 include/elf/external.h:} ATTRIBUTE_PACKED Elf_External_Versym;
33120 It is not possible to build GDB with MSVC today, but if it could, that
33121 would be one compiler that doesn't support this attribute. However,
33122 it supports packing via pragmas, so there's a way to cross that bridge
33123 if we ever get to it. I believe any compiler worth its salt supports
33124 some way of packing.
33126 In any case, the worse that happens without the attribute is that some
33127 types become larger than ideal. Regardless, I've added a couple
33128 static assertions to catch such compilers in action:
33130 /* Ensure size and aligment are what we expect. */
33131 gdb_static_assert (sizeof (packed) == Bytes);
33132 gdb_static_assert (alignof (packed) == 1);
33134 Change-Id: Ifa94f0a2cebfae5e8f6ddc73265f05e7fd9e1532
33136 2022-07-12 Alan Modra <amodra@gmail.com>
33138 PowerPC md_end: Don't htab_delete(NULL)
33139 It might be possible to hit md_end before md_begin is called, don't
33140 segfault if so. Also, remove a useless check.
33142 * gas/config/tc-ppc.c (insn_calloc): Remove needless overflow
33144 (ppc_md_end): Check ppc_hash before deleting. Clear ppc_hash.
33146 2022-07-12 Alan Modra <amodra@gmail.com>
33148 PR29355, ld segfaults with -r/-q and custom-named section .rela*
33149 The bug testcase uses an output section named .rel or .rela which has
33150 input .data sections mapped to it. The input .data section has
33151 relocations. When counting output relocations SHT_REL and SHT_RELA
33152 section reloc_count is ignored, with the justification that reloc
33153 sections themselves can't have relocations and some backends use
33154 reloc_count in reloc sections. However, the test wrongly used the
33155 output section type (which normally would match input section type).
33156 Fix that. Note that it is arguably wrong for ld to leave the output
33157 .rel/.rela section type as SHT_REL/SHT_RELA when non-empty non-reloc
33158 sections are written to it, but I'm not going to change that since it
33159 might be useful to hand-craft relocs in a data section that is then
33160 written to a SHT_REL/SHT_RELA output section.
33163 * elflink.c (bfd_elf_final_link): Use input section type rather
33164 than output section type to determine whether to exclude using
33165 reloc_count from that section.
33167 2022-07-12 Jiangshuai Li <jiangshuai_li@c-sky.com>
33169 gdb/csky complete csky_dwarf_reg_to_regnum
33170 For csky arch, the correspondence between Dwarf registers and GDB
33171 registers are as follows:
33172 dwarf regnos 0~31 ==> gdb regs r0~r31
33173 dwarf regno CSKY_HI_REGNUM(36) ==> gdb reg hi
33174 dwarf regno CSKY_LO_REGNUM(37) ==> gdb reg hi
33175 dwarf regno CSKY_PC_REGNUM(72) ==> gdb reg pc
33176 dwarf regnos FV_PSEUDO_REGNO_FIRST(74)~FV_PSEUDO_REGNO_LAST(201)
33178 gdb regs s0~s127 (pseudo regs for float and vector regs)
33180 other dwarf regnos have no corresponding gdb regs to them.
33182 2022-07-12 GDB Administrator <gdbadmin@sourceware.org>
33184 Automatic date update in version.in
33186 2022-07-11 Pedro Alves <pedro@palves.net>
33188 Always emit =thread-exited notifications, even if silent
33189 [Note: the testcased added by this commit depends on
33190 https://sourceware.org/pipermail/gdb-patches/2022-June/190259.html,
33191 otherwise GDB just crashes when detaching the core]
33193 Currently, in MI, =thread-created are always emitted, like:
33195 =thread-group-started,id="i1",pid="195680"
33196 =thread-created,id="1",group-id="i1"
33199 but on teardown, if the target uses exit_inferior_silent, then you'll
33200 only see the inferior exit notification (thread-group-exited), no
33201 notification for threads.
33203 The core target is one of the few targets that use
33204 exit_inferior_silent. Here's an example session:
33206 -target-select core $corefile
33207 =thread-group-started,id="i1",pid="195680"
33208 =thread-created,id="1",group-id="i1"
33210 ^connected,frame=....
33213 =thread-group-exited,id="i1"
33217 This imbalance of emitting =thread-created but then not =thread-exited
33218 seems off to me. (And, it complicates changes I want to do to
33219 centralize emitting thread exit notifications for the CLI, which is
33220 why I'm looking at this.)
33222 And then, since most other targets use exit_inferior instead of
33223 exit_inferior_silent, MI is already emitting =thread-exited
33224 notifications when tearing down an inferior, for most targets.
33226 This commit makes MI always emit the =thread-exited notifications,
33227 even for exit_inferior_silent.
33229 Afterwards, when debugging a core, MI outputs:
33233 =thread-exited,id="1",group-id="i1" << new line
33234 =thread-group-exited,id="i1"
33238 Surprisingly, there's no MI testcase debugging a core file. This
33239 commit adds the first.
33241 Change-Id: I5100501a46f07b6bbad3e04d120c2562a51c93a4
33243 2022-07-11 Pedro Alves <pedro@palves.net>
33245 Fix core-file -> detach -> crash (corefiles/29275)
33246 After loading a core file, you're supposed to be able to use "detach"
33247 to unload the core file. That unfortunately regressed starting with
33248 GDB 11, with these commits:
33250 1192f124a308 - gdb: generalize commit_resume, avoid commit-resuming when threads have pending statuses
33251 408f66864a1a - detach in all-stop with threads running
33253 resulting in a GDB crash:
33256 Thread 1 "gdb" received signal SIGSEGV, Segmentation fault.
33257 0x0000555555e842bf in maybe_set_commit_resumed_all_targets () at ../../src/gdb/infrun.c:2899
33258 2899 if (proc_target->commit_resumed_state)
33260 #0 0x0000555555e842bf in maybe_set_commit_resumed_all_targets () at ../../src/gdb/infrun.c:2899
33261 #1 0x0000555555e848bf in scoped_disable_commit_resumed::reset (this=0x7fffffffd440) at ../../src/gdb/infrun.c:3023
33262 #2 0x0000555555e84a0c in scoped_disable_commit_resumed::reset_and_commit (this=0x7fffffffd440) at ../../src/gdb/infrun.c:3049
33263 #3 0x0000555555e739cd in detach_command (args=0x0, from_tty=1) at ../../src/gdb/infcmd.c:2791
33264 #4 0x0000555555c0ba46 in do_simple_func (args=0x0, from_tty=1, c=0x55555662a600) at ../../src/gdb/cli/cli-decode.c:95
33265 #5 0x0000555555c112b0 in cmd_func (cmd=0x55555662a600, args=0x0, from_tty=1) at ../../src/gdb/cli/cli-decode.c:2514
33266 #6 0x0000555556173b1f in execute_command (p=0x5555565c5916 "", from_tty=1) at ../../src/gdb/top.c:699
33268 The code that crashes looks like:
33271 maybe_set_commit_resumed_all_targets ()
33273 scoped_restore_current_thread restore_thread;
33275 for (inferior *inf : all_non_exited_inferiors ())
33277 process_stratum_target *proc_target = inf->process_target ();
33279 if (proc_target->commit_resumed_state)
33282 With 'proc_target' above being null. all_non_exited_inferiors filters
33283 out inferiors that have pid==0. We get here at the end of
33284 detach_command, after core_target::detach has already run, at which
33285 point the inferior _should_ have pid==0 and no process target. It is
33286 clear it no longer has a process target, but, it still has a pid!=0
33289 The reason the inferior still has pid!=0, is that core_target::detach
33290 just unpushes, and relies on core_target::close to actually do the
33291 getting rid of the core and exiting the inferior. The problem with
33292 that is that detach_command grabs an extra strong reference to the
33293 process stratum target, so the unpush_target inside
33294 core_target::detach doesn't actually result in a call to
33295 core_target::close.
33297 Fix this my moving the cleaning up the core inferior to a shared
33298 routine called by both core_target::close and core_target::detach. We
33299 still need to cleanup the inferior from within core_file::close
33300 because there are paths to it that want to get rid of the core without
33301 going through detach. E.g., "core-file" -> "run".
33303 This commit includes a new test added to gdb.base/corefile.exp to
33304 cover the "core-file core" -> "detach" scenario.
33306 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29275
33308 Change-Id: Ic42bdd03182166b19f598428b0dbc2ce6f67c893
33310 2022-07-11 Pedro Alves <pedro@palves.net>
33312 Fix non-existent "@var{thread-id}" in stop reply descriptions
33313 In the description of stop replies, where the "thread" register is
33314 explained, and where the fork and vfork stop reasons are detailed,
33315 there are references to "@var{thread-id}", but such variable does not
33316 actually exist. This commit fixes it.
33318 Change-Id: If679944aaf15f6f64aabe506339a9e7667864cab
33320 2022-07-11 Luis Machado <luis.machado@arm.com>
33322 Try a couple PAuth compilation flags for gdb.arch/aarch64-pauth.exp
33323 The -msign-return-address switch has been dropped from GCC, but some
33324 older compiler may still support it. Make sure we try both
33325 -msign-return-address and -mbranch-protection before bailing out when
33326 running gdb.arch/aarch64-pauth.exp.
33328 2022-07-11 Andrew Burgess <aburgess@redhat.com>
33330 gdb: add support for disassembler styling using libopcodes
33331 This commit extends GDB to make use of libopcodes styling support
33332 where available, currently this is just i386 based architectures, and
33335 For architectures that don't support styling using libopcodes GDB will
33336 fall back to using the Python Pygments package, when the package is
33339 The new libopcodes based styling has the disassembler identify parts
33340 of the disassembled instruction, e.g. registers, immediates,
33341 mnemonics, etc, and can style these components differently.
33342 Additionally, as the styling is now done in GDB we can add settings to
33343 allow the user to configure which colours are used right from the GDB
33346 There's some new maintenance commands:
33348 maintenance set libopcodes-styling enabled on|off
33349 maintenance show libopcodes-styling
33351 These can be used to manually disable use of libopcodes styling. This
33352 is a maintenance command as it's not anticipated that a user should
33353 need to do this. But, this could be useful for testing, or, in some
33354 rare cases, a user might want to override the Python hook used for
33355 disassembler styling, and then disable libopcode styling so that GDB
33356 falls back to using Python. Right now I would consider this second
33357 use case a rare situation, which is why I think a maintenance command
33360 When libopcodes is being used for styling then the user can make use
33361 of the following new styles:
33363 set/show style disassembler comment
33364 set/show style disassembler immediate
33365 set/show style disassembler mnemonic
33366 set/show style disassembler register
33368 The disassembler also makes use of the 'address' and 'function'
33369 styles to style some parts of the disassembler output. I have also
33370 added the following aliases though:
33372 set/show style disassembler address
33373 set/show style disassembler symbol
33375 these are aliases for:
33377 set/show style address
33378 set/show style function
33380 respectively, and exist to make it easier for users to discover
33381 disassembler related style settings. The 'address' style is used to
33382 style numeric addresses in the disassembler output, while the 'symbol'
33383 or 'function' style is used to style the names of symbols in
33384 disassembler output.
33386 As not every architecture supports libopcodes styling, the maintenance
33387 setting 'libopcodes-styling enabled' has an "auto-off" type behaviour.
33388 Consider this GDB session:
33390 (gdb) show architecture
33391 The target architecture is set to "auto" (currently "i386:x86-64").
33392 (gdb) maintenance show libopcodes-styling enabled
33393 Use of libopcodes styling support is "on".
33395 the setting defaults to "on" for architectures that support libopcodes
33398 (gdb) set architecture sparc
33399 The target architecture is set to "sparc".
33400 (gdb) maintenance show libopcodes-styling enabled
33401 Use of libopcodes styling support is "off" (not supported on architecture "sparc")
33403 the setting will show as "off" if the user switches to an architecture
33404 that doesn't support libopcodes styling. The underlying setting is
33405 still "on" at this point though, if the user switches back to
33406 i386:x86-64 then the setting would go back to being "on".
33408 (gdb) maintenance set libopcodes-styling enabled off
33409 (gdb) maintenance show libopcodes-styling enabled
33410 Use of libopcodes styling support is "off".
33412 now the setting is "off" for everyone, even if the user switches back
33413 to i386:x86-64 the setting will still show as "off".
33415 (gdb) maintenance set libopcodes-styling enabled on
33416 Use of libopcodes styling not supported on architecture "sparc".
33417 (gdb) maintenance show libopcodes-styling enabled
33418 Use of libopcodes styling support is "off".
33420 attempting to switch the setting "on" for an unsupported architecture
33421 will give an error, and the setting will remain "off".
33423 (gdb) set architecture auto
33424 The target architecture is set to "auto" (currently "i386:x86-64").
33425 (gdb) maintenance show libopcodes-styling enabled
33426 Use of libopcodes styling support is "off".
33427 (gdb) maintenance set libopcodes-styling enabled on
33428 (gdb) maintenance show libopcodes-styling enabled
33429 Use of libopcodes styling support is "on".
33431 the user will need to switch back to a supported architecture before
33432 they can one again turn this setting "on".
33434 2022-07-11 Andrew Burgess <aburgess@redhat.com>
33436 gdb: have gdb_disassemble_info carry 'this' in its stream pointer
33437 The gdb_disassemble_info class is a wrapper around the libopcodes
33438 disassemble_info struct. The 'stream' field of disassemble_info is
33439 passed as an argument to the fprintf_func and fprintf_styled_func
33440 callbacks when the disassembler wants to print anything.
33442 Previously, GDB would store a pointer to a ui_file object in the
33443 'stream' field, then, when the disassembler wanted to print anything,
33444 the content would be written to the ui_file object. An example of an
33445 fprintf_func callback, from gdb/disasm.c is:
33448 gdb_disassembler::dis_asm_fprintf (void *stream, const char *format, ...)
33450 /* Write output to STREAM here. */
33453 This is fine, but has one limitation, within the print callbacks we
33454 only have access to STREAM, we can't access any additional state
33455 stored within the gdb_disassemble_info object.
33457 Right now this isn't a problem, but in a future commit this will
33458 become an issue, how we style the output being written to STREAM will
33459 depend on the state of the gdb_disassemble_info object, and this state
33460 might need to be updated, depending on what is being printed.
33462 In this commit I propose changing the 'stream' field of the
33463 disassemble_info to carry a pointer to the gdb_disassemble_info
33464 sub-class, rather than the stream itself.
33466 We then have the two sub-classes of gdb_disassemble_info to consider,
33467 the gdb_non_printing_disassembler class never cared about the stream,
33468 previously, for this class, the stream was nullptr. With the change
33469 to make stream be a gdb_disassemble_info pointer, no further updates
33470 are needed for gdb_non_printing_disassembler.
33472 The other sub-class is gdb_printing_disassembler. In this case the
33473 sub-class now carries around a pointer to the stream object. The
33474 print callbacks are updated to cast the incoming stream object back to
33475 a gdb_printing_disassembler, and then extract the stream.
33477 This is purely a refactoring commit. A later commit will add
33478 additional state to the gdb_printing_disassembler, and update the
33479 print callbacks to access this state.
33481 There should be no user visible changes after this commit.
33483 2022-07-11 Tom de Vries <tdevries@suse.de>
33485 [gdb/symtab] Fix data race in per_cu->length
33486 With gdb build with -fsanitize=thread and test-case
33487 gdb.dwarf2/dw4-sig-types.exp and target board cc-with-dwz-m we run into a data
33490 Write of size 4 at 0x7b2800002268 by thread T4:^M
33491 #0 cutu_reader::cutu_reader(dwarf2_per_cu_data*, dwarf2_per_objfile*, \
33492 abbrev_table*, dwarf2_cu*, bool, abbrev_cache*) gdb/dwarf2/read.c:6236 \
33497 Previous read of size 4 at 0x7b2800002268 by thread T1:^M
33498 #0 dwarf2_find_containing_comp_unit gdb/dwarf2/read.c:23444 \
33502 In other words, between this write:
33504 this_cu->length = cu->header.get_length ();
33508 && mid_cu->sect_off + mid_cu->length > sect_off))
33512 Fix this similar to the per_cu->dwarf_version case, by only setting it if
33513 needed, and otherwise verifying that the same value is used. [ Note that the
33514 same code is already present in the other cutu_reader constructor. ]
33516 Move this logic into into a member function set_length to make sure it's used
33517 consistenly, and make the field private in order to enforce access through the
33518 member functions, and rename it to m_length.
33520 This exposes (running test-case gdb.dwarf2/fission-reread.exp) that in
33521 fill_in_sig_entry_from_dwo_entry, the m_length field is overwritten.
33522 For now, allow for that exception.
33524 While we're at it, make sure that the length is set before read.
33526 Tested on x86_64-linux.
33528 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29344
33530 2022-07-11 Tom de Vries <tdevries@suse.de>
33532 [gdb/symtab] Use comp_unit_head::get_length
33533 There's a spot in read_comp_units_from_section where we explictly use
33534 initial_length_size to get the total length:
33536 this_cu->length = cu_header.length + cu_header.initial_length_size;
33539 Instead, just use cu_header.get_length ().
33541 Tested on x86_64-linux.
33543 2022-07-11 GDB Administrator <gdbadmin@sourceware.org>
33545 Automatic date update in version.in
33547 2022-07-10 Luis Machado <luis.machado@arm.com>
33549 Fix include guard naming for arch/aarch64-mte-linux.h
33550 It should be ARCH_AARCH64_MTE_LINUX_H as opposed to ARCH_AARCH64_LINUX_H.
33552 2022-07-10 Youling Tang <tangyouling@loongson.cn>
33554 gdbserver: LoongArch: Add orig_a0 processing
33555 Commit 736918239b16 ("gdb: LoongArch: add orig_a0 into register set")
33556 introduced orig_a0, similar processing needs to be done in gdbserver.
33558 At the same time, add orig_a0 related comments.
33560 2022-07-10 Youling Tang <tangyouling@loongson.cn>
33562 gdbserver: LoongArch: Simplify code with register number macros
33563 Move "enum loongarch_regnum" to gdb/arch/loongarch.h so that the
33564 macro definitions can be used in gdbserver/linux-loongarch-low.cc
33565 to simplify the code.
33567 2022-07-10 GDB Administrator <gdbadmin@sourceware.org>
33569 Automatic date update in version.in
33571 2022-07-09 Alan Modra <amodra@gmail.com>
33573 gas: tc-tic54x.c hash tables
33574 Cleaning up the subsym_hash memory is a real pain. Keys and values
33575 entered into the table are quite diverse. In some cases the key is
33576 allocated and thus needs to be freed, in others the key is a const
33577 string. Values are similar, and in some cases not even a string.
33578 Tidy this by inserting a new subsym_ent_t that describes key/value
33579 type. This meant the math_hash table was no longer needed. The patch
33580 also tidies how math functions are called, those that are supposed to
33581 return int now no longer return their value in a float.
33583 * config/tc-tic54x.c (math_hash): Delete.
33584 (subsym_proc_entry): Move earlier, make proc a union, merge with..
33585 (math_proc_entry): ..this. Delete type.
33586 (math_procs): Merge into subsym_procs.
33587 (subsym_ent_t): New. Use this type in subsym_hash..
33588 (stag_add_field_symbols, tic54x_var, tic54x_macro_info): ..here..
33589 (md_begin, subsym_create_or_replace, subsym_lookup): ..and here..
33590 (subsym_substitute): ..and here. Adjust subsym_proc_entry
33591 function calls. Free replacement when not returned.
33592 (subsym_get_arg): Adjust subsym_lookup.
33593 (free_subsym_ent, subsym_htab_create ): New functions, use when
33594 creating subsym_hash.
33595 (free_local_label_ent, local_label_htab_create): Similarly.
33596 (tic54x_remove_local_label): Delete.
33597 (tic54x_clear_local_labels): Simplify.
33598 (tic54x_asg): Use notes obstack to dup strings.
33599 (tic54x_eval): Likewise.
33600 (subsym_ismember): Likewise.
33601 (math_cvi, math_int, math_sgn): Return int.
33602 (tic54x_macro_start): Decrement macro_level before calling as_fatal.
33603 (tic54x_md_end): New function.
33604 * config/tc-tic54h.c (tic54x_md_end): Declare.
33607 2022-07-09 Alan Modra <amodra@gmail.com>
33609 gas: use notes_calloc in string hash
33610 Using notes_calloc means all of the string hash table memory should
33611 now be freed before gas exits, even though htab_delete isn't called.
33612 This also means that the hash table free_f and del_f must be NULL,
33613 because freeing notes obstack memory results in all more recently
33614 allocated notes memory being freed too. So hash table resizing won't
33615 free any memory, and will be a little faster. Also, htab_delete won't
33616 do anything (and be quick about it).
33618 Since htab_traverse can also resize hash tables (to make another
33619 traversal faster if the table is largely empty), stop that happening
33620 when only one traversal is done.
33622 * as.h: Reorder hash.h after symbols.h for notes_calloc decl.
33623 * hash.h (str_htab_create): Use notes_calloc. Do not free.
33624 * symbols.c (resolve_local_symbol_values): Don't resize
33625 during hash table traversal.
33626 * config/obj-elf.c (elf_frob_file_after_relocs): Likewise.
33627 * config/tc-ia64.c (ia64_adjust_symtab, ia64_frob_file): Likewise.
33628 * config/tc-nds32.c (nds32_elf_analysis_relax_hint): Likewise.
33630 2022-07-09 Alan Modra <amodra@gmail.com>
33632 gas: target string hash tables
33633 This allocates entries added to the string hash tables on the notes
33634 obstack, so that at least those do not leak. A followup patch will
33635 switch over the str_hash allocation to notes_calloc, which is why I
33636 haven't implemented deleting all the target string hash tables.
33638 * config/obj-coff-seh.c (get_pxdata_name, alloc_pxdata_item): Use
33639 notes obstack for string hash table entries.
33640 * config/tc-alpha.c (get_alpha_reloc_tag, md_begin): Likewise.
33641 * config/tc-h8300.c (md_begin): Likewise.
33642 * config/tc-ia64.c (dot_rot, dot_pred_rel, dot_alias): Likewise.
33643 * config/tc-nds32.c (nds32_relax_hint): Likewise.
33644 * config/tc-riscv.c (riscv_init_csr_hash): Likewise.
33645 * config/tc-score.c (s3_insert_reg): Likewise.
33646 (s3_build_score_ops_hsh, s3_build_dependency_insn_hsh): Likewise.
33647 * config/tc-score7.c (s7_build_score_ops_hsh): Likewise.
33648 (s7_build_dependency_insn_hsh): Likewise.
33649 * config/tc-tic4x.c (tic4x_asg): Likewise.
33651 2022-07-09 Alan Modra <amodra@gmail.com>
33653 arc gas: don't leak arc_opcode_hash memory
33654 The arc opcode hash table has entries that have a realloc'd field.
33655 This doesn't lend itself to obstack allocation, so freeing must be
33656 done with a purpose built hashtab del_f.
33658 * config/tc-arc.c (arc_opcode_free): New function.
33659 (md_begin): Pass the above as del_f to htab_create_alloc.
33660 (arc_md_end): New function.
33661 * config/tc-arc.h (arc_md_end): Declare.
33664 2022-07-09 Alan Modra <amodra@gmail.com>
33666 i386 gas: don't leak op_hash or reg_hash memory
33667 This tidies memory used by the two x86 gas string hash tables before
33668 exiting. I'm using a two-pronged approach, firstly the obvious call
33669 to htab_delete plus telling the libiberty/hashtab.c infrastructure to
33670 free tuples generated by str_hash_insert, and secondly putting the x86
33671 core_optab memory on the notes obstack. It would be possible to free
33672 core_optab memory by using a custom hash table del_f on x86, as I do
33673 for arc, but a later patch will move all the string hash memory to the
33676 * config/tc-i386.c (md_begin): Use notes_alloc for core_optab.
33677 (386_md_end): New function.
33678 * config/tc-i386.h (386_md_end): Declare.
33680 * hash.h (str_htab_create): Pass free as del_f.
33682 2022-07-09 Alan Modra <amodra@gmail.com>
33684 ppc gas: don't leak ppc_hash memory
33685 * config/tc-ppc.c (insn_obstack): New.
33686 (insn_calloc): New function.
33687 (ppc_setup_opcodes): Use insn_obstack for ppc_hash.
33688 (ppc_md_end): New function.
33689 * config/tc-ppc.h (ppc_md_end): Declare
33692 2022-07-09 Alan Modra <amodra@gmail.com>
33695 Only inline functions should be defined in hash.h, there's no benefit
33696 in having multiple copies of hash_string_tuple and eq_string_tuple.
33697 Also, use the table alloc_f when allocating tuples to be stored, so
33698 that these functions are usable with different memory allocation
33701 * hash.h (struct string_tuple, string_tuple_t): Move earlier.
33702 (string_tuple_alloc): Add table param, allocate using table alloc_f.
33703 (str_hash_insert): Adjust to suit. Call table->free_f when
33705 (hash_string_tuple, eq_string_tuple): Move to..
33708 2022-07-09 Alan Modra <amodra@gmail.com>
33710 gas: rename md_end to md_finish
33711 Currently md_end is typically used for some final actions rather than
33712 freeing memory like other *_end functions. Rename it to md_finish,
33713 and rename target implementation. The renaming of target functions
33714 makes it possible to find them all with "grep md_finish",
33715 eg. md_mips_end is renamed to mips_md_finish, not md_mips_finish.
33716 This patch leaves a number of md_end functions unchanged, those that
33717 either do nothing or deallocate memory, and calls them late.
33719 The idea here is that target maintainers implement md_end functions to
33720 tidy memory, if anyone cares. Freeing persistent memory in gas is
33721 not at all important, except that it can hide more important memory
33722 leaks, those that happen once per some frequent gas operation, amongst
33723 these unimportant memory leaks.
33725 * as.c (main): Rename md_end to md_finish.
33726 * config/tc-alpha.c, * config/tc-alpha.h,
33727 * config/tc-arc.c, * config/tc-arc.h,
33728 * config/tc-arm.c, * config/tc-arm.h,
33729 * config/tc-csky.c, * config/tc-csky.h,
33730 * config/tc-ia64.c, * config/tc-ia64.h,
33731 * config/tc-mcore.c, * config/tc-mcore.h,
33732 * config/tc-mips.c, * config/tc-mips.h,
33733 * config/tc-mmix.c, * config/tc-mmix.h,
33734 * config/tc-msp430.c, * config/tc-msp430.h,
33735 * config/tc-nds32.c, * config/tc-nds32.h,
33736 * config/tc-ppc.c, * config/tc-ppc.h,
33737 * config/tc-pru.c, * config/tc-pru.h,
33738 * config/tc-riscv.c, * config/tc-riscv.h,
33739 * config/tc-s390.c, * config/tc-s390.h,
33740 * config/tc-sparc.c, * config/tc-sparc.h,
33741 * config/tc-tic4x.c, * config/tc-tic4x.h,
33742 * config/tc-tic6x.c, * config/tc-tic6x.h,
33743 * config/tc-v850.c, * config/tc-v850.h,
33744 * config/tc-xtensa.c, * config/tc-xtensa.h,
33745 * config/tc-z80.c, * config/tc-z80.h: Similarly.
33746 * output-file.c (output_file_close): Call md_end.
33748 2022-07-09 Alan Modra <amodra@gmail.com>
33750 gas: set up notes obstack earlier
33751 So that the notes obstack can be used for persistent storage in
33754 * as.c (parse_args): Use notes_alloc and notes_strdup.
33755 (free_notes): New function.
33756 (main): Init notes obstack, and arrange to be freed on exit.
33757 * read.c (read_begin): Don't init notes obstack.
33758 (read_end): Free cond_obstack.
33759 * subsegs.c (subsegs_end): Don't free cond_obstack or notes.
33761 2022-07-09 Alan Modra <amodra@gmail.com>
33764 itbl_files seems to be debug code. Get rid of it.
33766 * as.c (struct itbl_file_list): Delete.
33767 (itbl_files): Delete.
33768 (parse_args): Don't keep itbl_files list.
33770 2022-07-09 Alan Modra <amodra@gmail.com>
33772 gas: free sy_hash, macro_hash and po_hash
33773 * macro.c (macro_end): New function.
33774 * macro.h (macro_end): Declare.
33775 * read.c (read_end, poend): New functions.
33776 * read.h (read_end): Declare.
33777 * symbols.c (symbol_end): New function.
33778 * symbols.h (symbol_end): Declare.
33779 * output-file.c (output_file_close): Call new *_end functions.
33781 2022-07-09 Alan Modra <amodra@gmail.com>
33783 dw2gencfi.c: use notes obstack
33784 Use notes obstack for dwcfi_hash entries, and free table. Freeing the
33785 table makes memory checkers complain more about "definitely lost"
33786 memory as we've moved some from the "still reachable" category.
33787 That will be fixed with a later patch.
33789 * dw2gencfi.c (get_debugseg_name): Allocate on notes obstack.
33790 (alloc_debugseg_item): Likewise.
33791 (dwcfi_hash_find_or_make): Adjust failure path free.
33792 (cfi_finish): Delete dwfci_hash.
33794 2022-07-09 Alan Modra <amodra@gmail.com>
33796 expr.c make_expr_symbol: use notes obstack
33797 * expr.c (make_expr_symbol): Use notes_alloc.
33799 read.c assign_symbol: use notes obstack for dummy listing frag
33800 * read.c (assign_symbol): Use notes_calloc for dummy_frag.
33802 read.c s_include: use notes obstack for path
33803 * read.c (s_include): Use notes obstack for path mem.
33805 2022-07-09 Alan Modra <amodra@gmail.com>
33807 macro.c: use string hash from hash.h for macro_hash
33808 Another case of duplicated hash.h code, the only minor difference
33809 being that macro->format_hash was created with 7 entries vs. str_hash
33812 * macro.c (macro_init, define_macro): Use str_htab_create.
33813 (do_formals, define_macro, macro_expand_body): Use str_hash_insert
33814 (macro_expand_body): Use str_hash_find and str_hash_delete.
33815 (delete_macro): Likewise.
33816 (sub_actual, macro_expand, check_macro): Use str_hash_find.
33817 (expand_irp): Use str_htab_create and str_hash_insert.
33818 * macro.h (struct macro_struct): Tidy.
33819 (struct macro_hash_entry, macro_hash_entry_t, hash_macro_entry),
33820 (eq_macro_entry, macro_entry_alloc, macro_entry_find),
33821 (struct formal_hash_entry, formal_hash_entry_t),
33822 (hash_formal_entry, eq_formal_entry, formal_entry_alloc),
33823 (formal_entry_find): Delete.
33824 * config/tc-iq2000.c (iq2000_add_macro): Use str_htab_create
33825 and str_hash_insert.
33827 2022-07-09 Alan Modra <amodra@gmail.com>
33829 read.c: use string hash from hash.h for po_hash
33830 po_hash code duplicates the str_hash code in hash.h for no good reason.
33832 * read.c (struct po_entry, po_entry_t): Delete.
33833 (hash_po_entry, eq_po_entry, po_entry_alloc, po_entry_find): Delete.
33834 (pop_insert): Use str_hash_insert.
33835 (pobegin): Use str_htab_create.
33836 (read_a_source_file, s_macro): Use str_hash_find.
33838 2022-07-09 Alan Modra <amodra@gmail.com>
33840 free read_symbol_name string
33841 read_symbol_name mallocs the string it returns. Free it when done.
33843 * read.c (read_symbol_name): Free name on error path.
33844 * config/tc-ppc.c (ppc_GNU_visibility): Free name returned from
33846 (ppc_extern, ppc_globl, ppc_weak): Likewise.
33848 2022-07-09 Alan Modra <amodra@gmail.com>
33850 gas: output_file_close
33851 This is mostly a tidy with the aim of being able to free
33852 out_file_name, but it does fix a possible attempt to unlink the output
33853 file twice (not that that matters).
33855 * as.h (keep_it): New global.
33856 * as.c (keep_it): Delete.
33857 (close_output_file): Delete, merged into..
33858 * output-file.c (output_file_close): ..here. Delete parameter.
33859 * output-file.h (output_file_close): Update prototype.
33861 2022-07-09 Alan Modra <amodra@gmail.com>
33863 gas: utility notes memory alloc functions
33864 Makes it a little easier to use the notes obstack for persistent
33867 * as.h (gas_mul_overflow): Define.
33868 * symbols.h (notes_alloc, notes_calloc, notes_memdup),
33869 (notes_strdup, notes_concat, notes_free): Declare.
33870 * symbols.c (notes_alloc, notes_calloc, notes_memdup),
33871 (notes_strdup, notes_concat, notes_free): New functions.
33872 (save_symbol_name): Use notes_strdup.
33873 (symbol_create, local_symbol_make, local_symbol_convert),
33874 (symbol_clone, decode_local_label_name): Use notes_alloc.
33876 2022-07-09 Alan Modra <amodra@gmail.com>
33878 gas: arm -mwarn-syms duplicates
33879 arm gas is only supposed to warn once per symbol for -mwarn-syms, but
33880 doesn't because the str_hash_find added with commit 629310abec88
33881 always returns NULL. That's so because the str_hash_insert inserts a
33882 NULL value for the key,value pair. Let str_hash_insert do the job
33885 * config/tc-arm.c (arm_tc_equal_in_insn): Correct already_warned
33887 * testsuite/gas/arm/pr18347.s: Modify to generate duplicate
33888 warning without this patch.
33890 2022-07-09 Alan Modra <amodra@gmail.com>
33892 Regenerate with automake-1.15.1
33893 Until we update the recommended versions of autoconf/automake, files
33894 should be regenerated with automake-1.15.1 and autoconf-2.69. That's
33895 not because we think those versions are golden, and newer versions are
33896 bad. It's simply because maintainers want to be able to update
33897 configury files without trouble, and if someone regenerates files with
33898 automake-1.16.5 then --enable-maintainer-mode builds will hit errors:
33900 checking that generated files are newer than configure... configure.ac:26: error: version mismatch. This is Automake 1.15.1,
33901 configure.ac:26: but the definition used by this AM_INIT_AUTOMAKE
33902 configure.ac:26: comes from Automake 1.16.5. You should recreate
33903 configure.ac:26: aclocal.m4 with aclocal and run automake again.
33904 WARNING: 'automake-1.15' is probably too old.
33906 Correcting this requires regenerating the files by hand.
33908 2022-07-09 GDB Administrator <gdbadmin@sourceware.org>
33910 Automatic date update in version.in
33912 2022-07-08 Tom Tromey <tom@tromey.com>
33914 Accept gdb.Value in more Python APIs
33915 PR python/27000 points out that gdb.block_for_pc will accept a Python
33916 integer, but not a gdb.Value. This patch corrects this oversight.
33918 I looked at all uses of GDB_PY_LLU_ARG and fixed these up to use
33919 get_addr_from_python instead. I also looked at uses of GDB_PY_LL_ARG,
33920 but those seemed relatively unlikely to be useful with a gdb.Value, so
33921 I didn't change them. My thinking here is that a Value will typically
33922 come from inferior memory, and something like a line number is not too
33923 likely to be found this way.
33925 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=27000
33927 2022-07-08 Tom Tromey <tom@tromey.com>
33929 Handle bool specially in gdb.set_parameter
33930 PR python/29217 points out that gdb.parameter will return bool values,
33931 but gdb.set_parameter will not properly accept them. This patch fixes
33932 the problem by adding a special case to set_parameter.
33934 I looked at a fix involving rewriting set_parameter in C++. However,
33935 this one is simpler.
33937 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29217
33939 2022-07-08 Ludovic Courtès <ludo@gnu.org>
33941 [gdb/build] Handle deprecation of scm_install_gmp_memory_functions
33942 When building gdb with guile 3.0.8, we run into:
33944 gdb/guile/guile.c: In function \
33945 'void gdbscm_initialize(const extension_language_defn*)':
33946 gdb/guile/guile.c:680:5: error: 'scm_install_gmp_memory_functions' is \
33947 deprecated [-Werror=deprecated-declarations]
33948 680 | scm_install_gmp_memory_functions = 0;
33949 | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
33950 In file included from /usr/include/guile/3.0/libguile.h:128,
33951 from gdb/guile/guile-internal.h:30,
33952 from gdb/guile/guile.c:36:
33953 /usr/include/guile/3.0/libguile/deprecated.h:164:20: note: declared here
33954 164 | SCM_DEPRECATED int scm_install_gmp_memory_functions;
33955 | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
33956 cc1plus: all warnings being treated as errors
33957 make[1]: *** [Makefile:1896: guile/guile.o] Error 1
33960 The variable has been deprecated because it no longer has any effect.
33962 Fix this by disabling the specific deprecation warning.
33964 Also handle upcoming guile versions > 3.0, in which the variable will be
33965 removed, by limiting the usage of the variable to guile versions <= 3.0.
33967 This does not break anything. The variable was merely used to address a
33968 problem present in guile versions <= v3.0.5.
33970 Note that we don't limit the usage of the variable to guile versions <= 3.0.5,
33971 because we want to support f.i. building against 3.0.6 and then using a shared
33974 Tested on x86_64-linux.
33976 Co-Authored-By: Tom de Vries <tdevries@suse.de>
33978 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28994
33980 2022-07-08 Tom de Vries <tdevries@suse.de>
33982 [gdb/symtab] Fix assert in process_imported_unit_die
33983 When running test-case gdb.dwarf2/dw2-symtab-includes.exp with target board
33984 cc-with-debug-names, I run into:
33986 (gdb) maint expand-symtab dw2-symtab-includes.h^M
33987 src/gdb/dwarf2/read.h:311: internal-error: unit_type: \
33988 Assertion `m_unit_type != 0' failed.^M
33989 A problem internal to GDB has been detected,^M
33990 further debugging may prove unreliable.^M
33991 ----- Backtrace -----^M
33992 FAIL: gdb.dwarf2/dw2-symtab-includes.exp: maint expand-symtab \
33993 dw2-symtab-includes.h (GDB internal error)
33996 The assert was recently added in commit 2c474c46943 ("[gdb/symtab] Add get/set
33997 functions for per_cu->lang/unit_type").
33999 The assert is triggered here:
34001 /* We're importing a C++ compilation unit with tag DW_TAG_compile_unit
34002 into another compilation unit, at root level. Regard this as a hint,
34004 if (die->parent && die->parent->parent == NULL
34005 && per_cu->unit_type () == DW_UT_compile
34006 && per_cu->lang () == language_cplus)
34010 We're trying to access unit_type / lang which hasn't been set yet.
34012 Normally, these are set during cooked index creation, but that's not the case
34013 when using an index (or using -readnow).
34015 In other words, this lto binary reading speed optimization only works in the
34018 IWBN to have this working in all use cases, but for now, allow lang () and
34019 unit_type () to return language_unknown and 0 here.
34021 Tested on x86_64-linux.
34023 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29321
34025 2022-07-08 Tom de Vries <tdevries@suse.de>
34027 [gdb/symtab] Fix segfault in dwarf2_per_objfile::symtab_set_p
34028 When running test-case gdb.cp/cpexprs-debug-types.exp with target board
34029 cc-with-debug-names, I run into:
34031 (gdb) print base::operator new^M
34034 Fatal signal: Segmentation fault^M
34035 ----- Backtrace -----^M
34036 0x57ea46 gdb_internal_backtrace_1^M
34037 /home/vries/gdb_versions/devel/src/gdb/bt-utils.c:122^M
34038 0x57eae9 _Z22gdb_internal_backtracev^M
34039 /home/vries/gdb_versions/devel/src/gdb/bt-utils.c:168^M
34040 0x75b8ad handle_fatal_signal^M
34041 /home/vries/gdb_versions/devel/src/gdb/event-top.c:946^M
34042 0x75ba19 handle_sigsegv^M
34043 /home/vries/gdb_versions/devel/src/gdb/event-top.c:1019^M
34044 0x7f795f46a8bf ???^M
34045 0x6d3cb1 _ZNK18dwarf2_per_objfile12symtab_set_pEPK18dwarf2_per_cu_data^M
34046 /home/vries/gdb_versions/devel/src/gdb/dwarf2/read.c:1515^M
34049 The problem is in this piece of code in dw2_debug_names_iterator::next:
34051 case DW_IDX_type_unit:
34052 /* Don't crash on bad data. */
34053 if (ull >= per_bfd->tu_stats.nr_tus)
34055 complaint (_(".debug_names entry has bad TU index %s"
34056 " [in module %s]"),
34058 objfile_name (objfile));
34061 per_cu = per_bfd->get_cu (ull + per_bfd->tu_stats.nr_tus);
34065 The all_comp_units vector (which get_cu accesses) contains both CUs and TUs,
34068 So to get the nth TU we need the element at "nr_cus + n", but
34069 the code uses "nr_tus + n" instead.
34071 Fix this by using "nr_cus + n".
34073 Tested on x86_64-linux.
34075 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29334
34077 2022-07-08 Enze Li <enze.li@hotmail.com>
34079 gdb: initialize the data_head variable to eliminate compilation warnings
34080 On a machine with gcc 12, I get this warning:
34082 CXX nat/linux-btrace.o
34083 In function ‘btrace_error linux_read_bts(btrace_data_bts*, btrace_target_info*, btrace_read_type)’,
34084 inlined from ‘btrace_error linux_read_btrace(btrace_data*, btrace_target_info*, btrace_read_type)’ at ../gdb/nat/linux-btrace.c:935:29:
34085 ../gdb/nat/linux-btrace.c:865:21: warning: ‘data_head’ may be used uninitialized [-Wmaybe-uninitialized]
34086 865 | pevent->last_head = data_head;
34087 | ~~~~~~~~~~~~~~~~~~^~~~~~~~~~~
34088 ../gdb/nat/linux-btrace.c: In function ‘btrace_error linux_read_btrace(btrace_data*, btrace_target_info*, btrace_read_type)’:
34089 ../gdb/nat/linux-btrace.c:792:9: note: ‘data_head’ was declared here
34090 792 | __u64 data_head, data_tail;
34093 Fix this by initializing the 'data_head' variable.
34095 Tested by rebuilding on x86_64 openSUSE Tumbleweed with gcc 12.
34097 2022-07-08 Andrew Burgess <aburgess@redhat.com>
34099 libopcodes/s390: add support for disassembler styling
34100 This commit adds disassembler style to the libopcodes s390
34101 disassembler. This conversion was pretty straight forward, I just
34102 converted the fprintf_func calls to fprintf_styled_func calls and
34103 added an appropriate style.
34105 For testing the new styling I just assembled then disassembled the
34106 source files in gas/testsuite/gas/s390 and manually checked that the
34107 styling looked reasonable.
34109 If the user does not request styled output from objdump, then there
34110 should be no change in the disassembler output after this commit.
34112 2022-07-08 Nick Clifton <nickc@redhat.com>
34114 Fix regeneration of ld configure and makefiles
34116 Update release README with new version numbers
34118 Update version to 2.39.50 and regenerate files
34120 Add markers for 2.39 branch
34122 2022-07-08 GDB Administrator <gdbadmin@sourceware.org>
34124 Automatic date update in version.in
34126 2022-07-07 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
34128 gprofng: fix regression in testing for not yet installed version
34130 2022-07-07 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
34132 * src/Settings.cc (Settings::read_rc): Read environment variable
34133 GPROFNG_SYSCONFDIR.
34134 * testsuite/lib/Makefile.skel: Export GPROFNG_SYSCONFDIR.
34135 * testsuite/gprofng.display/display.exp: Shorten the list of tests.
34137 2022-07-07 Simon Marchi <simon.marchi@efficios.com>
34139 gdb: fix {rs6000_nat_target,aix_thread_target}::wait to not use inferior_ptid
34140 Trying to run a simple program (empty main) on AIX, I get:
34143 Starting program: /scratch/simark/build/gdb/a.out
34144 Child process unexpectedly missing: There are no child processes..
34145 ../../src/binutils-gdb/gdb/inferior.c:304: internal-error: find_inferior_pid: Assertion `pid != 0' failed.
34146 A problem internal to GDB has been detected,
34147 further debugging may prove unreliable.
34148 ----- Backtrace -----
34149 0x10ef12a8 gdb_internal_backtrace_1()
34150 ../../src/binutils-gdb/gdb/bt-utils.c:122
34151 0x10ef1470 gdb_internal_backtrace()
34152 ../../src/binutils-gdb/gdb/bt-utils.c:168
34153 0x1004d368 internal_vproblem(internal_problem*, char const*, int, char const*, char*)
34154 ../../src/binutils-gdb/gdb/utils.c:396
34155 0x1004d8a8 internal_verror(char const*, int, char const*, char*)
34156 ../../src/binutils-gdb/gdb/utils.c:476
34157 0x1004c424 internal_error(char const*, int, char const*, ...)
34158 ../../src/binutils-gdb/gdbsupport/errors.cc:55
34159 0x102ab344 find_inferior_pid(process_stratum_target*, int)
34160 ../../src/binutils-gdb/gdb/inferior.c:304
34161 0x102ab4a4 find_inferior_ptid(process_stratum_target*, ptid_t)
34162 ../../src/binutils-gdb/gdb/inferior.c:318
34163 0x1061bae8 find_thread_ptid(process_stratum_target*, ptid_t)
34164 ../../src/binutils-gdb/gdb/thread.c:519
34165 0x10319e98 handle_inferior_event(execution_control_state*)
34166 ../../src/binutils-gdb/gdb/infrun.c:5532
34167 0x10315544 fetch_inferior_event()
34168 ../../src/binutils-gdb/gdb/infrun.c:4221
34169 0x10952e34 inferior_event_handler(inferior_event_type)
34170 ../../src/binutils-gdb/gdb/inf-loop.c:41
34171 0x1032640c infrun_async_inferior_event_handler(void*)
34172 ../../src/binutils-gdb/gdb/infrun.c:9548
34173 0x10673188 check_async_event_handlers()
34174 ../../src/binutils-gdb/gdb/async-event.c:335
34175 0x1066fce4 gdb_do_one_event()
34176 ../../src/binutils-gdb/gdbsupport/event-loop.cc:214
34177 0x10001a94 start_event_loop()
34178 ../../src/binutils-gdb/gdb/main.c:411
34179 0x10001ca0 captured_command_loop()
34180 ../../src/binutils-gdb/gdb/main.c:471
34181 0x10003d74 captured_main(void*)
34182 ../../src/binutils-gdb/gdb/main.c:1329
34183 0x10003e48 gdb_main(captured_main_args*)
34184 ../../src/binutils-gdb/gdb/main.c:1344
34186 ../../src/binutils-gdb/gdb/gdb.c:32
34187 ---------------------
34188 ../../src/binutils-gdb/gdb/inferior.c:304: internal-error: find_inferior_pid: Assertion `pid != 0' failed.
34189 A problem internal to GDB has been detected,
34190 further debugging may prove unreliable.
34191 Quit this debugging session? (y or n)
34193 This is due to some bit-rot in the AIX port, still relying on the entry
34194 value of inferior_ptid in the wait methods.
34196 Problem #1 is in rs6000_nat_target::wait, here:
34198 /* Ignore terminated detached child processes. */
34199 if (!WIFSTOPPED (status) && pid != inferior_ptid.pid ())
34202 At this point, waitpid has returned an "exited" status for some pid, so
34203 pid is non-zero. Since inferior_ptid is set to null_ptid on entry, the
34204 pid returned by wait is not equal to `inferior_ptid.pid ()`, so we reset
34205 pid to -1 and go to waiting again. Since there are not more children to
34206 wait for, waitpid then returns -1 so we get here:
34210 gdb_printf (gdb_stderr,
34211 _("Child process unexpectedly missing: %s.\n"),
34212 safe_strerror (save_errno));
34214 /* Claim it exited with unknown signal. */
34215 ourstatus->set_signalled (GDB_SIGNAL_UNKNOWN);
34216 return inferior_ptid;
34219 We therefore return a "signalled" status with a null_ptid (again,
34220 inferior_ptid is null_ptid). This confuses infrun, because if the
34221 target returns a "signalled" status, it should be coupled with a ptid
34222 for an inferior that exists.
34224 So, the first step is to fix the snippets above to not use
34225 inferior_ptid. In the first snippet, use find_inferior_pid to see if
34226 we know the event process. If there is no inferior with that pid, we
34227 assume it's a detached child process to we ignore the event. That
34228 should be enough to fix the problem, because it should make it so we
34229 won't go into the second snippet. But still, fix the second snippet to
34230 return an "ignore" status. This is copied from inf_ptrace_target::wait,
34231 which is where rs6000_nat_target::wait appears to be copied from in the
34234 These changes, are not sufficient, as the aix_thread_target, which sits
34235 on top of rs6000_nat_target, also relies on inferior_ptid.
34236 aix_thread_target::wait, by calling pd_update, assumes that
34237 rs6000_nat_target has set inferior_ptid to the appropriate value (the
34238 ptid of the event thread), but that's not the case. pd_update
34239 returns inferior_ptid - null_ptid - and therefore
34240 aix_thread_target::wait returns null_ptid too, and we still hit the
34241 assert shown above.
34243 Fix this by changing pd_activate, pd_update, sync_threadlists and
34244 get_signaled_thread to all avoid using inferior_ptid. Instead, they
34245 accept as a parameter the pid of the process we are working on.
34247 With this patch, I am able to run the program to completion:
34250 Starting program: /scratch/simark/build/gdb/a.out
34251 [Inferior 1 (process 11010794) exited normally]
34253 As well as break on main:
34256 Breakpoint 1 at 0x1000036c
34258 Starting program: /scratch/simark/build/gdb/a.out
34260 Breakpoint 1, 0x1000036c in main ()
34263 [Inferior 1 (process 26083688) exited normally]
34265 Change-Id: I7c2613bbefe487d75fa1a0c0994423471d961ee9
34267 2022-07-07 Pedro Alves <pedro@palves.net>
34269 Fix pedantically invalid DWARF in gdb.trace/unavailable-dwarf-piece.exp
34270 The DWARF spec says:
34272 Any debugging information entry representing the declaration of an object,
34273 module, subprogram or type may have DW_AT_decl_file, DW_AT_decl_line and
34274 DW_AT_decl_column attributes, each of whose value is an unsigned integer
34278 Grepping around the DWARF-assembler-based testcases, I noticed that
34279 gdb.trace/unavailable-dwarf-piece.exp emits decl_line with
34280 DW_FORM_sdata, a signed integer form. This commit tweaks it to use
34281 DW_FORM_udata instead.
34283 Unsurprisingly, this:
34286 TESTS="gdb.trace/unavailable-dwarf-piece.exp" \
34287 RUNTESTFLAGS="--target_board=native-gdbserver"
34289 ... still passes cleanly for me after this change.
34291 I've noticed this because current llvm-dwarfdump crashed on an
34292 ROCm-internal DWARF-assembler-based testcase that incorrectly used
34293 signed forms for DW_AT_decl_file/DW_AT_decl_line.
34295 The older llvm-dwarfdump found on Ubuntu 20.04 (LLVM 10) reads the
34296 line numbers with signed forms as "0" instead of crashing. Here's the
34297 before/after fix for gdb.trace/unavailable-dwarf-piece.exp with that
34298 llvm-dwarfdump version:
34300 $ diff -up before.txt after.txt
34301 --- before.txt 2022-07-07 13:21:28.387690334 +0100
34302 +++ after.txt 2022-07-07 13:21:39.379801092 +0100
34305 DW_AT_byte_size (3)
34306 DW_AT_decl_file (0)
34307 - DW_AT_decl_line (0)
34308 + DW_AT_decl_line (1)
34310 0x0000002f: DW_TAG_member
34314 DW_AT_byte_size (3)
34315 DW_AT_decl_file (0)
34316 - DW_AT_decl_line (0)
34317 + DW_AT_decl_line (1)
34319 0x00000054: DW_TAG_member
34322 Change-Id: I5c866946356da421ff944019d0eca2607b2b738f
34324 2022-07-07 Tiezhu Yang <yangtiezhu@loongson.cn>
34326 gdb: LoongArch: Fix typos in code comments
34327 "it’s" should be "it's".
34329 2022-07-07 Maciej W. Rozycki <macro@embecosm.com>
34331 GDB/testsuite: Add coverage for `print -elements' command
34332 We currently have no coverage for the `print -elements ...' command (or
34333 `p -elements ...' in the shortened form), so add a couple of test cases
34334 mimicking ones using corresponding `set print elements ...' values.
34336 2022-07-07 Tiezhu Yang <yangtiezhu@loongson.cn>
34338 gdb: LoongArch: Implement the push_dummy_call gdbarch method
34339 According to "Procedure Calling Convention" in "LoongArch ELF ABI
34340 specification" [1], implement the push_dummy_call gdbarch method
34341 as clear as possible.
34343 [1] https://loongson.github.io/LoongArch-Documentation/LoongArch-ELF-ABI-EN.html#_procedure_calling_convention
34345 2022-07-07 Tsukasa OI <research_trasio@irq.a4lg.com>
34347 RISC-V: Added Zfhmin and Zhinxmin.
34348 This commit adds Zfhmin and Zhinxmin extensions (subsets of Zfh and
34349 Zhinx extensions, respectively). In the process supporting Zfhmin and
34350 Zhinxmin extension, this commit also changes how instructions are
34351 categorized considering Zfhmin, Zhinx and Zhinxmin extensions.
34355 * From INSN_CLASS_ZFH to INSN_CLASS_ZFHMIN:
34357 flh, fsh, fmv.x.h and fmv.h.x.
34359 * From INSN_CLASS_ZFH to INSN_CLASS_ZFH_OR_ZHINX:
34363 * From INSN_CLASS_ZFH_OR_ZHINX to INSN_CLASS_ZFH_OR_ZHINX:
34365 fneg.h, fabs.h, fsgnj.h, fsgnjn.h, fsgnjx.h,
34366 fadd.h, fsub.h, fmul.h, fdiv.h, fsqrt.h, fmin.h, fmax.h,
34367 fmadd.h, fnmadd.h, fmsub.h, fnmsub.h,
34368 fcvt.w.h, fcvt.wu.h, fcvt.h.w, fcvt.h.wu,
34369 fcvt.l.h, fcvt.lu.h, fcvt.h.l, fcvt.h.lu,
34370 feq.h, flt.h, fle.h, fgt.h, fge.h,
34373 * From INSN_CLASS_ZFH_OR_ZHINX to INSN_CLASS_ZFHMIN_OR_ZHINXMIN:
34375 fcvt.s.h and fcvt.h.s.
34377 * From INSN_CLASS_D_AND_ZFH_INX to INSN_CLASS_ZFHMIN_AND_D:
34379 fcvt.d.h and fcvt.h.d.
34381 * From INSN_CLASS_Q_AND_ZFH_INX to INSN_CLASS_ZFHMIN_AND_Q:
34383 fcvt.q.h and fcvt.h.q.
34387 * elfxx-riscv.c (riscv_implicit_subsets): Change implicit
34388 subsets. Zfh->Zicsr is not needed and Zfh->F is replaced with
34389 Zfh->Zfhmin and Zfhmin->F. Zhinx->Zicsr is not needed and
34390 Zhinx->Zfinx is replaced with Zhinx->Zhinxmin and
34392 (riscv_supported_std_z_ext): Added zfhmin and zhinxmin.
34393 (riscv_multi_subset_supports): Rewrite handling for new
34394 instruction classes.
34395 (riscv_multi_subset_supports_ext): Updated.
34396 (riscv_parse_check_conflicts): Change error message to include
34397 zfh and zfhmin extensions.
34401 * testsuite/gas/riscv/zfhmin-d-insn-class-fail.s: New complex
34402 error handling test.
34403 * testsuite/gas/riscv/zfhmin-d-insn-class-fail-1.d: Likewise.
34404 * testsuite/gas/riscv/zfhmin-d-insn-class-fail-1.l: Likewise.
34405 * testsuite/gas/riscv/zfhmin-d-insn-class-fail-2.d: Likewise.
34406 * testsuite/gas/riscv/zfhmin-d-insn-class-fail-2.l: Likewise.
34407 * testsuite/gas/riscv/zfhmin-d-insn-class-fail-3.d: Likewise.
34408 * testsuite/gas/riscv/zfhmin-d-insn-class-fail-3.l: Likewise.
34409 * testsuite/gas/riscv/zfhmin-d-insn-class-fail-4.d: Likewise.
34410 * testsuite/gas/riscv/zfhmin-d-insn-class-fail-4.l: Likewise.
34411 * testsuite/gas/riscv/zfhmin-d-insn-class-fail-5.d: Likewise.
34412 * testsuite/gas/riscv/zfhmin-d-insn-class-fail-5.l: Likewise.
34413 * testsuite/gas/riscv/zhinx.d: Renamed from fp-zhinx-insns.d
34415 * testsuite/gas/riscv/zhinx.s: Likewise.
34419 * opcode/riscv.h (enum riscv_insn_class): Removed INSN_CLASS_ZFH,
34420 INSN_CLASS_D_AND_ZFH_INX and INSN_CLASS_Q_AND_ZFH_INX. Added
34421 INSN_CLASS_ZFHMIN, INSN_CLASS_ZFHMIN_OR_ZHINXMIN,
34422 INSN_CLASS_ZFHMIN_AND_D and INSN_CLASS_ZFHMIN_AND_Q.
34426 * riscv-opc.c (riscv_opcodes): Change instruction classes for
34427 Zfh and Zfhmin instructions. Fix `fcvt.h.lu' instruction
34428 (two operand variant) mask.
34430 2022-07-07 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
34432 gprofng: adjust GPROFNG_VARIANT
34433 GPROFNG_VARIANT depends on compiler options, not on $(host).
34436 2022-07-06 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
34439 * libcollector/configure.ac: Adjust GPROFNG_VARIANT.
34440 * libcollector/configure: Rebuild.
34442 2022-07-07 Tsukasa OI <research_trasio@irq.a4lg.com>
34444 RISC-V: Fix disassembling Zfinx with -M numeric
34445 This commit fixes floating point operand register names from ABI ones
34446 to dynamically set ones.
34450 * testsuite/gas/riscv/zfinx-dis-numeric.s: Test new behavior of
34451 Zfinx extension and -M numeric disassembler option.
34452 * testsuite/gas/riscv/zfinx-dis-numeric.d: Likewise.
34456 * riscv-dis.c (riscv_disassemble_insn): Use dynamically set GPR
34457 names to disassemble Zfinx instructions.
34459 2022-07-07 Tsukasa OI <research_trasio@irq.a4lg.com>
34461 RISC-V: Fix requirement handling on Zhinx+{D,Q}
34462 This commit fixes how instructions are masked on Zhinx+Z{d,q}inx.
34463 fcvt.h.d and fcvt.d.h require ((D&&Zfh)||(Zdinx&&Zhinx)) and
34464 fcvt.h.q and fcvt.q.h require ((Q&&Zfh)||(Zqinx&&Zhinx)).
34468 * elfxx-riscv.c (riscv_multi_subset_supports): Fix feature gate
34469 on INSN_CLASS_{D,Q}_AND_ZFH_INX.
34470 (riscv_multi_subset_supports_ext): Fix feature gate diagnostics
34471 on INSN_CLASS_{D,Q}_AND_ZFH_INX.
34475 * testsuite/gas/riscv/fp-zhinx-insns.d: Add Zqinx to -march
34476 for proper testing.
34478 2022-07-07 Alan Modra <amodra@gmail.com>
34480 PR29320, 'struct obstack' declared inside parameter list
34482 * frags.h: Move declaration of struct obstack..
34485 2022-07-07 GDB Administrator <gdbadmin@sourceware.org>
34487 Automatic date update in version.in
34489 2022-07-06 Ruud van der Pas <ruud.vanderpas@oracle.com>
34491 gprofng: implement a functional gp-display-html
34492 This patch enables the first support for the "gprofng display html" command.
34493 This command works for C/C++ applications on x86_64. Using one or more gprofng
34494 experiment directories as input, a new directory with html files is created.
34495 Through the index.html file in this directory, the performance results may be
34496 viewed in a browser.
34499 2022-06-28 Ruud van der Pas <ruud.vanderpas@oracle.com>
34501 * gp-display-html/gp-display-html.in: implement first support for x86_64 and C/C++
34503 2022-07-06 H.J. Lu <hjl.tools@gmail.com>
34505 elf: Copy p_align of PT_GNU_STACK for stack alignment
34506 commit 74e315dbfe5200c473b226e937935fb8ce391489
34507 Author: H.J. Lu <hjl.tools@gmail.com>
34508 Date: Mon Dec 13 19:46:04 2021 -0800
34510 elf: Set p_align to the minimum page size if possible
34512 may ignore p_align of PT_GNU_STACK when copying ELF program header if
34513 the maximum page size is larger than p_align of PT_LOAD segments. Copy
34514 p_align of PT_GNU_STACK since p_align of PT_GNU_STACK describes stack
34515 alignment, not page size,
34518 * elf.c (copy_elf_program_header): Copy p_align of PT_GNU_STACK
34519 for stack alignment.
34521 2022-07-06 Jan Beulich <jbeulich@suse.com>
34523 x86: make D attribute usable for XOP and FMA4 insns
34524 This once again allows to reduce redundancy in (and size of) the opcode
34527 Don't go as far as also making D work on the two 5-operand XOP insns:
34528 This would significantly complicate the code, as there the first
34529 (immediate) operand would need special treatment in several places.
34531 Note that the .s suffix isn't being enabled to have any effect, for
34532 being deprecated. Whereas neither {load} nor {store} pseudo prefixes
34533 make sense here, as the respective operands are inputs (loads) only
34534 anyway, regardless of order. Hence there is (as before) no way for the
34535 programmer to request the alternative encoding to be used for register-
34538 Note further that it is always the first original template which is
34539 retained (and altered), to make sure the same encoding as before is
34540 used for register-only insns. This has the slightly odd (but pre-
34541 existing) effect of XOP register-only insns having XOP.W clear, but FMA4
34542 ones having VEX.W set.
34544 2022-07-06 Jan Beulich <jbeulich@suse.com>
34546 x86: fold two switch() statements in match_template()
34547 I don't see why two of them were introduced (very long ago) using
34548 similar fall-through logic.
34550 2022-07-06 Jan Beulich <jbeulich@suse.com>
34552 x86: fix 3-operand insn reverse-matching
34553 The middle operand would have gone entirely unchecked, allowing e.g.
34555 vmovss %xmm0, %esp, %xmm2
34557 to assemble successfully, or e.g.
34559 vmovss %xmm0, $4, %xmm2
34561 causing an internal error. Alongside dealing with this also drop a
34562 related comment, which hasn't been applicable anymore since the
34563 introduction of 3-operand patterns with D set (and which perhaps never
34564 had been logical to be there, as reverse-matched insns don't make it
34565 there in the first place).
34567 2022-07-06 Bhuvanendra Kumar N <Bhuvanendra.KumarN@amd.com>
34569 Descriptive DWARF operations dump support for DW_AT_rank
34570 DW_AT_rank is a dwarf-5 feature.
34572 2022-07-06 Jan Beulich <jbeulich@suse.com>
34574 x86: introduce a state stack for .arch
34575 When using just slightly non-trivial combinations of .arch, it can be
34576 quite useful to be able to go back to prior state without needing to
34577 re-invoke perhaps many earlier directives and without needing to invoke
34578 perhaps many "negative" ones. Like some other architectures allow
34579 saving (pushing) and restoring (popping) present/prior state.
34581 For now require the same .code<N> to be in effect for ".arch pop" that
34582 was in effect for the corresponding ".arch push".
34584 Also change the global "no_cond_jump_promotion" to be bool, to match the
34587 2022-07-06 Jan Beulich <jbeulich@suse.com>
34589 x86: generalize disabling of sub-architectures
34590 I never really understood upon what basis ".arch .no*" options were made
34591 available. Let's not have any "criteria" at all, and simply allow
34592 disabling of all of them. Then we also have all data for a sub-arch in
34593 a single place, as we now only need a single table.
34595 2022-07-06 Jan Beulich <jbeulich@suse.com>
34597 x86: permit "default" with .arch
34598 So far there was no way to reset the architecture to that assembly would
34599 start with in the absence of any overrides (command line or directives).
34600 Note that for Intel MCU "default" is merely an alias of "iamcu".
34602 While there also zap a stray @item from the doc section, as noticed
34603 when inspecting the generated output (which still has some quirks, but
34604 those aren't easy to address without re-flowing almost the entire
34607 2022-07-06 Jan Beulich <jbeulich@suse.com>
34609 x86: don't leak sub-architecture accumulated strings
34610 While it may not be necessary in i386_target_format() (but then setting
34611 the variable to NULL also wouldn't be necessary), at least in the other
34612 cases strings may already have accumulated.
34614 2022-07-06 GDB Administrator <gdbadmin@sourceware.org>
34616 Automatic date update in version.in
34618 2022-07-05 Tom de Vries <tdevries@suse.de>
34620 [gdb/exp] Fix internal error when printing C++ pointer-to-member
34621 When running the test-case included with this patch, we run into:
34624 $1 = gdb/gdbtypes.h:695: internal-error: loc_bitpos: \
34625 Assertion `m_loc_kind == FIELD_LOC_KIND_BITPOS' failed.^M
34627 while printing a c++ pointer-to-member.
34629 Fix this by skipping static fields in cp_find_class_member, such that we have:
34635 Tested on x86_64-linux.
34637 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29294
34639 2022-07-05 Tom Tromey <tromey@adacore.com>
34641 Add gdb.Objfile.is_file attribute
34642 Sometimes an objfile comes from memory and not from a file. It can be
34643 useful to be able to check this from Python, so this patch adds a new
34644 "is_file" attribute.
34646 2022-07-05 Tom Tromey <tromey@adacore.com>
34648 Make 'import gdb.events' work
34649 Pierre-Marie noticed that, while gdb.events is a Python module, it
34650 can't be imported. This patch changes how this module is created, so
34651 that it can be imported, while also ensuring that the module is always
34652 visible, just as it was in the past.
34654 This new approach required one non-obvious change -- when running
34655 gdb.base/warning.exp, where --data-directory is intentionally not
34656 found, the event registries can now be nullptr. Consequently, this
34657 patch probably also requires
34659 https://sourceware.org/pipermail/gdb-patches/2022-June/189796.html
34661 Note that this patch obsoletes
34663 https://sourceware.org/pipermail/gdb-patches/2022-June/189797.html
34665 2022-07-05 Xi Ruoyao <xry111@xry111.site>
34667 gdb: LoongArch: add orig_a0 into register set
34668 The basic support for LoongArch has been merged into the upstream Linux
34669 kernel since 5.19-rc1 on June 5, 2022. This commit adds orig_a0 which
34670 is added into struct user_pt_regs [1] to match the upstream kernel, and
34671 then the upstream GDB will work with the upstream kernel.
34673 Note that orig_a0 was added into struct user_pt_regs in the development
34674 cycle for merging LoongArch port into the upstream Linux kernel, so
34675 earlier kernels (notably, the product kernel with version 4.19 used in
34676 distros like UOS and Loongnix) don't have it. Inspect
34677 arch/loongarch/include/uapi/asm/ptrace.h in the kernel tree to make sure.
34678 To build upstream GDB for a kernel lacking orig_a0, it's necessary to
34679 revert this commit locally.
34681 [1]: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/loongarch/include/uapi/asm/ptrace.h#n24
34683 2022-07-05 Bhuvanendra Kumar N <Bhuvanendra.KumarN@amd.com>
34685 Support for location and range lists for split-dwarf and dwarf-5.
34686 Adding support for location and range lists for split-dwarf and dwarf-5.
34687 Following issues are taken care.
34688 1. Display of the index values for DW_FORM_loclistx and DW_FORM_rnglistx.
34689 2. Display of .debug_loclists.dwo and .debug_rnglists.dwo sections.
34691 * dwarf.c(read_and_display_attr_value): Handle DW_FORM_loclistx
34692 and DW_FORM_rnglistx for .dwo files.
34693 (process_debug_info): Load .debug_loclists.dwo and
34694 .debug_rnglists.dwo if exists.
34695 (load_separate_debug_files): Load .debug_loclists and
34696 .debug_rnglists if exists.
34697 Include 2 entries in debug_displays table.
34698 * dwarf.h (enum dwarf_section_display_enum): Include 2 entries.
34700 2022-07-05 Jan Beulich <jbeulich@suse.com>
34702 x86: introduce fake processor type to mark sub-arch entries in cpu_arch[]
34703 This is in preparation of dropping the leading . from the strings.
34705 While there also move PROCESSOR_GENERIC{32,64} from the middle of AMD
34706 entries to near the top.
34708 2022-07-05 Jan Beulich <jbeulich@suse.com>
34710 x86: macro-ize cpu_arch[] entries
34711 Putting individual elements behind macros, besides (imo) improving
34712 readability, will make subsequent (and likely also future) changes less
34715 Utilize this right away to pack the table a little more tightly, by
34716 converting "skip" to bool and putting it earlier in a group of bitfields
34717 together with "len".
34719 2022-07-05 Jan Beulich <jbeulich@suse.com>
34721 x86: de-duplicate sub-architecture strings accumulation
34722 Introduce a helper function to replace 4 instances of similar code. Use
34723 reconcat() to cover the previously explicit free().
34725 2022-07-05 GDB Administrator <gdbadmin@sourceware.org>
34727 Automatic date update in version.in
34729 2022-07-04 Nick Clifton <nickc@redhat.com>
34731 Fix snafu in rust demangler recursion limit code
34733 2022-07-04 Alan Modra <amodra@gmail.com>
34735 alloc gas seginfo on notes obstack
34736 Lots of memory used in gas should go on this obstack. The patch also
34737 frees all the gas obstacks on exit, which isn't a completely trivial
34740 * subsegs.c (alloc_seginfo): New function.
34741 (subseg_change, subseg_get): Use it.
34742 (subsegs_end): New function.
34743 * as.h (subsegs_end): Declare.
34744 * output-file.c: Include subsegs.h
34745 (stash_frchain_obs): New function.
34746 (output_file_close): Save obstacks attached to output bfd before
34747 closing. Call subsegs_end with the array of obstacks.
34749 2022-07-04 Alan Modra <amodra@gmail.com>
34751 objcopy: bfd_alloc orelocation
34752 This fixes an inconsequential objcopy memory leak. I'd normally
34753 ignore reports of leaks like this one, that are merely one block or
34754 fewer per section processed, since objcopy soon exits and frees all
34755 memory. However I thought it worth providing support for allocating
34756 memory on a bfd objalloc in objcopy and other utils.
34759 * bucomm.c (bfd_xalloc): New function.
34760 * bucomm.h (bfd_xalloc): Declare.
34761 * objcopy.c (copy_relocations_in_section): Use it to allocate
34762 array of reloc pointers. Rewrite code stripping relocs to do
34763 without extra memory allocation.
34765 2022-07-04 Nick Clifton <nickc@redhat.com>
34767 Synchronize libbierty sources with gcc.
34769 2022-07-04 Bhuvanendra Kumar N <Bhuvanendra.KumarN@amd.com>
34771 Modified changes for split-dwarf and dwarf-5.
34772 * dwarf.c(process_debug_info): Include DW_TAG_skeleton_unit.
34773 (display_debug_str_offsets): While dumping .debug_str_offsets.dwo,
34774 pass proper str_offsets_base to fetch_indexed_string().
34775 (load_separate_debug_files): Skip DWO ID dump for dwarf-5.
34777 2022-07-04 Marcus Nilsson <brainbomb@gmail.com>
34779 opcodes/avr: Implement style support in the disassembler
34780 * disassemble.c: (disassemble_init_for_target): Set
34781 created_styled_output for AVR based targets.
34782 * avr-dis.c: (print_insn_avr): Use fprintf_styled_ftype
34783 instead of fprintf_ftype throughout.
34784 (avr_operand): Pass in and fill disassembler_style when
34787 2022-07-04 Tom de Vries <tdevries@suse.de>
34789 [gdb/symtab] Add get/set functions for per_cu->lang/unit_type
34790 The dwarf2_per_cu_data fields lang and unit_type both have a dont-know
34791 initial value (respectively language_unknown and (dwarf_unit_type)0), which
34792 allows us to add certain checks, f.i. checking that that a field is not read
34795 Add get/set member functions for the two fields as a convenient location to
34796 add such checks, make the fields private to enforce using the member
34797 functions, and add the m_ prefix.
34799 Tested on x86_64-linux.
34801 2022-07-04 Jan Beulich <jbeulich@suse.com>
34803 gas/testsuite: properly exclude aout in all/weakref1u
34804 Use the (wider) predicate rather than a triplet. This eliminates the sole
34805 i386-msdos failure in the testsuite.
34807 2022-07-04 Jan Beulich <jbeulich@suse.com>
34809 x86: fold Disp32S and Disp32
34810 The only case where 64-bit code uses non-sign-extended (can also be
34811 considered zero-extended) displacements is when an address size override
34812 is in place for a memory operand (i.e. particularly excluding
34813 displacements of direct branches, which - if at all - are controlled by
34814 operand size, and then are still sign-extended, just from 16 bits).
34815 Hence the distinction in templates is unnecessary, allowing code to be
34816 simplified in a number of places. The only place where logic becomes
34817 more complicated is when signed-ness of relocations is determined in
34820 The other caveat is that Disp64 cannot be specified anymore in an insn
34821 template at the same time as Disp32. Unlike for non-64-bit mode,
34822 templates don't specify displacements for both possible addressing
34823 modes; the necessary adjustment to the expected ones has already been
34824 done in match_template() anyway (but of course the logic there needs
34825 tweaking now). Hence the single template so far doing so is split.
34827 2022-07-04 Jan Beulich <jbeulich@suse.com>
34829 x86: restore masking of displacement kinds
34830 Commit 7d5e4556a375 rendered the check near the end of what is now
34831 i386_finalize_displacement() entirely dead for AT&T mode, since for
34832 operands involving a displacement .unspecified will always be set. But
34833 the logic there is bogus anyway - Intel syntax operand size specifiers
34834 are of no interest there either. The only thing which matters in the
34835 "displacement only" determination is .baseindex.
34837 Of course when masking displacement kinds we should not at the same time
34838 also mask off other attributes.
34840 Furthermore the type mask returned by lex_got() also needs to be
34841 adjusted: The only case where we want Disp32 (rather than Disp32S) is
34842 when dealing with 32-bit addressing mode in 64-bit code.
34844 2022-07-04 Jan Beulich <jbeulich@suse.com>
34846 x86-64: improve handling of branches to absolute addresses
34847 There are two related problems here: The use of "addr32" on a direct
34848 branch would, besides causing a warning, result in operands to be
34849 permitted which mistakenly are refused without "addr32". Plus at some
34850 point not too long ago I'm afraid it may have been me who regressed the
34851 relocation addends emitted for such branches. Correct both problems,
34852 adding a testcase to guard against regressing this again.
34854 2022-07-04 Tsukasa OI <research_trasio@irq.a4lg.com>
34856 RISC-V: Update Zihintpause extension version
34857 Because ratified Zihintpause extension has a version number of 2.0
34858 (not 1.0), we should update the number.
34862 * elfxx-riscv.c (riscv_supported_std_z_ext): Update version
34863 number of Zihintpause extension.
34865 2022-07-04 GDB Administrator <gdbadmin@sourceware.org>
34867 Automatic date update in version.in
34869 2022-07-03 GDB Administrator <gdbadmin@sourceware.org>
34871 Automatic date update in version.in
34873 2022-07-02 Tom de Vries <tdevries@suse.de>
34875 [gdb/symtab] Fix data race on per_cu->dwarf_version
34876 When building gdb with -fsanitize=thread and gcc 12, and running test-case
34877 gdb.dwarf2/dwz.exp, we run into a data race between thread T2 and the main
34878 thread in the same write:
34880 Write of size 1 at 0x7b200000300c:^M
34881 #0 cutu_reader::cutu_reader(dwarf2_per_cu_data*, dwarf2_per_objfile*, \
34882 abbrev_table*, dwarf2_cu*, bool, abbrev_cache*) gdb/dwarf2/read.c:6252 \
34887 this_cu->dwarf_version = cu->header.version;
34890 Both writes are called from the parallel for in dwarf2_build_psymtabs_hard,
34893 #1 process_psymtab_comp_unit gdb/dwarf2/read.c:6774 (gdb+0x8304d7)^M
34894 #2 operator() gdb/dwarf2/read.c:7098 (gdb+0x8317be)^M
34895 #3 operator() gdbsupport/parallel-for.h:163 (gdb+0x872380)^M
34897 and this via the PU import:
34899 #1 cooked_indexer::ensure_cu_exists(cutu_reader*, dwarf2_per_objfile*, \
34900 sect_offset, bool, bool) gdb/dwarf2/read.c:17964 (gdb+0x85c43b)^M
34901 #2 cooked_indexer::index_imported_unit(cutu_reader*, unsigned char const*, \
34902 abbrev_info const*) gdb/dwarf2/read.c:18248 (gdb+0x85d8ff)^M
34903 #3 cooked_indexer::index_dies(cutu_reader*, unsigned char const*, \
34904 cooked_index_entry const*, bool) gdb/dwarf2/read.c:18302 (gdb+0x85dcdb)^M
34905 #4 cooked_indexer::make_index(cutu_reader*) gdb/dwarf2/read.c:18443 \
34907 #5 process_psymtab_comp_unit gdb/dwarf2/read.c:6812 (gdb+0x830879)^M
34908 #6 operator() gdb/dwarf2/read.c:7098 (gdb+0x8317be)^M
34909 #7 operator() gdbsupport/parallel-for.h:171 (gdb+0x8723e2)^M
34912 Fix this by setting the field earlier, in read_comp_units_from_section.
34914 The write in cutu_reader::cutu_reader() is still needed, in case
34915 read_comp_units_from_section is not used (run the test-case with say, target
34916 board cc-with-gdb-index).
34918 Make the write conditional, such that it doesn't trigger if the field is
34919 already set by read_comp_units_from_section. Instead, verify that the
34920 field already has the value that we're trying to set it to.
34922 Move this logic into into a member function set_version (in analogy to the
34923 already present member function version) to make sure it's used consistenly,
34924 and make the field private in order to enforce access through the member
34925 functions, and rename it to m_dwarf_version.
34927 While we're at it, make sure that the version is set before read, to avoid
34928 say returning true for "per_cu.version () < 5" if "per_cu.version () == 0".
34930 Tested on x86_64-linux.
34932 2022-07-02 Tom de Vries <tdevries@suse.de>
34934 [gdb/testsuite] Fix gdb.base/early-init-file.exp with -fsanitize=thread
34935 When building gdb with -fsanitize=thread, I run into:
34937 FAIL: gdb.base/early-init-file.exp: check startup version string has style \
34942 warning: Found custom handler for signal 7 (Bus error) preinstalled.^M
34943 warning: Found custom handler for signal 8 (Floating point exception) \
34945 warning: Found custom handler for signal 11 (Segmentation fault) \
34947 Some signal dispositions inherited from the environment (SIG_DFL/SIG_IGN)^M
34948 won't be propagated to spawned programs.^M
34950 appearing before the "GNU gdb (GDB) $version" line.
34952 This is similar to the problem fixed by commit f0bbba7886f
34953 ("gdb.debuginfod/fetch_src_and_symbols.exp: fix when GDB is built with
34954 AddressSanitizer").
34956 In that commit, the problem was fixed by starting gdb with -quiet, but using
34957 that would mean the "GNU gdb (GDB) $version" line that we're trying to check
34960 Fix this instead by updating the regexp to allow the message.
34962 Tested on x86_64-linux.
34964 2022-07-02 GDB Administrator <gdbadmin@sourceware.org>
34966 Automatic date update in version.in
34968 2022-07-01 Maciej W. Rozycki <macro@embecosm.com>
34970 GDB/doc: Remove indentation from `print -elements' completion example
34971 Remove indentation from the text of the manual after the example here:
34973 " Completion will in some cases guide you with a suggestion of what
34974 kind of argument an option expects. For example:
34976 (gdb) print -elements <TAB><TAB>
34979 Here, the option expects a number (e.g., '100'), not literal
34980 'NUMBER'. Such metasyntactical arguments are always presented in
34983 as this is a continuation of the same paragraph.
34985 2022-07-01 Maciej W. Rozycki <macro@embecosm.com>
34987 GDB/doc: Remove extraneous spaces from completion examples
34988 Completion results are usually different when the operation is applied
34989 to a word that is or is not followed by a space. In some cases they are
34990 equivalent, however a space would not be produced if completion was used
34991 earlier on in the word processed.
34993 However in the manual we have completion examples given using a space
34994 that actually prevents the example from working. E.g.:
34996 (gdb) info bre <TAB>
35000 (gdb) info bre <TAB><TAB>
35001 Display all 200 possibilities? (y or n)
35003 as it now goes on to propose the entire symbol table, while:
35005 (gdb) info bre<TAB>
35006 (gdb) info breakpoints
35008 does the right thing, but is not what is shown in the manual.
35010 In other cases an extraneous space is used that does not correspond to
35011 the actual completion pattern shown, which gives an impression of
35014 Remove extraneous spaces then from completion examples as appropriate.
35016 2022-07-01 Nick Clifton <nickc@redhat.com>
35018 Add newline to the end of the rnglists displsy.
35020 2022-07-01 GDB Administrator <gdbadmin@sourceware.org>
35022 Automatic date update in version.in
35024 2022-06-30 Maciej W. Rozycki <macro@embecosm.com>
35026 GDB: Add `NUMBER' completion to `set' integer commands
35027 Fix a completion consistency issue with `set' commands accepting integer
35028 values and the special `unlimited' keyword:
35030 (gdb) complete print -elements
35031 print -elements NUMBER
35032 print -elements unlimited
35037 (gdb) complete set print elements
35038 set print elements unlimited
35041 (there is a space entered at the end of both commands, not shown here)
35042 which also means if you strike <Tab> with `set print elements ' input,
35043 it will, annoyingly, complete to `set print elements unlimited' right
35044 away rather than showing a choice between `NUMBER' and `unlimited'.
35046 Add `NUMBER' then as an available completion for such `set' commands:
35048 (gdb) complete set print elements
35049 set print elements NUMBER
35050 set print elements unlimited
35053 Adjust the testsuite accordingly. Also document the feature in the
35054 Completion section of the manual in addition to the Command Options
35055 section already there.
35057 2022-06-30 Bruno Larsen <blarsen@redhat.com>
35059 gdb/testsuite: Expand gdb.cp/mb-ctor.exp to test dynamic allocation
35060 When testing GDB's ability to stop in constructors, gdb.cp/mb-ctor.exp
35061 only tested objects allocated on the stack. This commit adds a couple of
35062 dynamic allocations and tests if GDB can stop in it as well.
35064 2022-06-30 Nick Clifton <nickc@redhat.com>
35066 Fix implementation of readelf's -wE and -wN options,
35067 * dwarf.c (dwarf_select_sections_by_name): If the entry's value is
35068 zero then clear the corresponding variable.
35069 (dwarf_select_sections_by_letters): Likewise.
35070 * testsuite/binutils-all/debuginfo.exp: Expect -WE and -wE
35071 debuginfod tests to fail.
35073 2022-06-30 Tom de Vries <tdevries@suse.de>
35075 [gdb] Block SIGTERM in worker threads
35076 With gdb build with gcc-12 and -fsanitize=thread, and test-case
35077 gdb.base/gdb-sigterm.exp, I run into:
35079 WARNING: ThreadSanitizer: data race (pid=9722)^M
35080 Write of size 4 at 0x00000325bc68 by thread T1:^M
35081 #0 handle_sigterm(int) src/gdb/event-top.c:1211 (gdb+0x8ec01f)^M
35083 Previous read of size 4 at 0x00000325bc68 by main thread:^M
35084 [failed to restore the stack]^M
35086 Location is global 'sync_quit_force_run' of size 4 at \
35087 0x00000325bc68 (gdb+0x325bc68)^M
35089 SUMMARY: ThreadSanitizer: data race gdb/event-top.c:1211 in \
35090 handle_sigterm(int)^M
35092 and 3 more data races involving handle_sigterm and locations:
35095 - heap block of size 40
35096 (XNEW (async_signal_handler) in create_async_signal_handler)
35098 This was reported in PR29297.
35100 The testcase executes a "kill -TERM $gdb_pid", which generates a
35101 process-directed signal.
35103 A process-directed signal can be delivered to any thread, and what we see
35104 here is the fallout of the signal being delivered to a worker thread rather
35105 than the main thread.
35107 Fix this by blocking SIGTERM in the worker threads.
35109 [ I have not been able to reproduce this after it occurred for the first time,
35110 so unfortunately I cannot confirm that the patch fixes the problem. ]
35112 Tested on x86_64-linux, with and without -fsanitize=thread.
35114 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29297
35116 2022-06-30 Andrew Burgess <aburgess@redhat.com>
35118 gdb/doc: fix column widths in MI compatibility table
35119 In passing I noticed that the column headings for the table of MI
35120 compatibility and breaking changes, were overlapping, at least when
35121 the PDF is generated on my machine.
35123 I propose giving slightly more space to the two version number
35124 columns, this prevents the headers overlapping for me.
35126 2022-06-30 GDB Administrator <gdbadmin@sourceware.org>
35128 Automatic date update in version.in
35130 2022-06-29 Pedro Alves <pedro@palves.net>
35132 Fix GDBserver regression due to change to avoid reading shell registers
35133 Simon reported that the recent change to make GDB and GDBserver avoid
35134 reading shell registers caused a GDBserver regression, caught with
35135 ASan while running gdb.server/non-existing-program.exp:
35137 $ /home/smarchi/build/binutils-gdb/gdb/testsuite/../../gdb/../gdbserver/gdbserver stdio non-existing-program
35138 =================================================================
35139 ==127719==ERROR: AddressSanitizer: heap-use-after-free on address 0x60f0000000e9 at pc 0x55bcbfa301f4 bp 0x7ffd238a7320 sp 0x7ffd238a7310
35140 WRITE of size 1 at 0x60f0000000e9 thread T0
35141 #0 0x55bcbfa301f3 in scoped_restore_tmpl<bool>::~scoped_restore_tmpl() /home/smarchi/src/binutils-gdb/gdbserver/../gdbsupport/scoped_restore.h:86
35142 #1 0x55bcbfa2ffe9 in post_fork_inferior(int, char const*) /home/smarchi/src/binutils-gdb/gdbserver/fork-child.cc:120
35143 #2 0x55bcbf9c9199 in linux_process_target::create_inferior(char const*, std::__debug::vector<char*, std::allocator<char*> > const&) /home/smarchi/src/binutils-gdb/gdbserver/linux-low.cc:991
35144 #3 0x55bcbf954549 in captured_main /home/smarchi/src/binutils-gdb/gdbserver/server.cc:3941
35145 #4 0x55bcbf9552f0 in main /home/smarchi/src/binutils-gdb/gdbserver/server.cc:4084
35146 #5 0x7ff9d663b0b2 in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x240b2)
35147 #6 0x55bcbf8ef2bd in _start (/home/smarchi/build/binutils-gdb/gdbserver/gdbserver+0x1352bd)
35149 0x60f0000000e9 is located 169 bytes inside of 176-byte region [0x60f000000040,0x60f0000000f0)
35150 freed by thread T0 here:
35151 #0 0x7ff9d6c6f0c7 in operator delete(void*) ../../../../src/libsanitizer/asan/asan_new_delete.cpp:160
35152 #1 0x55bcbf910d00 in remove_process(process_info*) /home/smarchi/src/binutils-gdb/gdbserver/inferiors.cc:164
35153 #2 0x55bcbf9c4ac7 in linux_process_target::remove_linux_process(process_info*) /home/smarchi/src/binutils-gdb/gdbserver/linux-low.cc:454
35154 #3 0x55bcbf9cdaa6 in linux_process_target::mourn(process_info*) /home/smarchi/src/binutils-gdb/gdbserver/linux-low.cc:1599
35155 #4 0x55bcbf988dc4 in target_mourn_inferior(ptid_t) /home/smarchi/src/binutils-gdb/gdbserver/target.cc:205
35156 #5 0x55bcbfa32020 in startup_inferior(process_stratum_target*, int, int, target_waitstatus*, ptid_t*) /home/smarchi/src/binutils-gdb/gdbserver/../gdb/nat/fork-inferior.c:515
35157 #6 0x55bcbfa2fdeb in post_fork_inferior(int, char const*) /home/smarchi/src/binutils-gdb/gdbserver/fork-child.cc:111
35158 #7 0x55bcbf9c9199 in linux_process_target::create_inferior(char const*, std::__debug::vector<char*, std::allocator<char*> > const&) /home/smarchi/src/binutils-gdb/gdbserver/linux-low.cc:991
35159 #8 0x55bcbf954549 in captured_main /home/smarchi/src/binutils-gdb/gdbserver/server.cc:3941
35160 #9 0x55bcbf9552f0 in main /home/smarchi/src/binutils-gdb/gdbserver/server.cc:4084
35161 #10 0x7ff9d663b0b2 in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x240b2)
35163 previously allocated by thread T0 here:
35164 #0 0x7ff9d6c6e5a7 in operator new(unsigned long) ../../../../src/libsanitizer/asan/asan_new_delete.cpp:99
35165 #1 0x55bcbf910ad0 in add_process(int, int) /home/smarchi/src/binutils-gdb/gdbserver/inferiors.cc:144
35166 #2 0x55bcbf9c477d in linux_process_target::add_linux_process_no_mem_file(int, int) /home/smarchi/src/binutils-gdb/gdbserver/linux-low.cc:425
35167 #3 0x55bcbf9c8f4c in linux_process_target::create_inferior(char const*, std::__debug::vector<char*, std::allocator<char*> > const&) /home/smarchi/src/binutils-gdb/gdbserver/linux-low.cc:985
35168 #4 0x55bcbf954549 in captured_main /home/smarchi/src/binutils-gdb/gdbserver/server.cc:3941
35169 #5 0x55bcbf9552f0 in main /home/smarchi/src/binutils-gdb/gdbserver/server.cc:4084
35170 #6 0x7ff9d663b0b2 in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x240b2)
35172 Above we see that in the non-existing-program case, the process gets
35173 deleted before the starting_up flag gets restored to false.
35175 This happens because startup_inferior calls target_mourn_inferior
35176 before throwing an error, and in GDBserver, unlike in GDB, mourning
35177 deletes the process.
35179 Fix this by not using a scoped_restore to manage the starting_up flag,
35180 since we should only clear it when startup_inferior doesn't throw.
35182 Change-Id: I67325d6f81c64de4e89e20e4ec4556f57eac7f6c
35184 2022-06-29 Maciej W. Rozycki <macro@embecosm.com>
35186 GDB/testsuite: Tighten `set print elements' error check
35187 Match the whole error message expected to be given rather than omitting
35188 the part about the "unlimited" keyword. There's no point in omitting
35189 the missing part first, and second with an upcoming change the part in
35190 parentheses will no longer be a fixed string, so doing a full match will
35191 ensure the algorithm correctly builds the message expected here. Also
35192 avoid any wildcard matches.
35194 2022-06-29 Maciej W. Rozycki <macro@embecosm.com>
35196 GDB: Remove extraneous full stops from `set' command error messages
35197 With errors given for bad commands such as `set annotate' or `set width'
35198 we produce an extraneous full stop within parentheses:
35201 Argument required (integer to set it to.).
35203 Argument required (integer to set it to, or "unlimited".).
35206 This is grammatically incorrect, so remove the full stop and adjust the
35207 testsuite accordingly.
35209 2022-06-29 Andrew Burgess <aburgess@redhat.com>
35211 gdb/doc: improve description of --data-disassemble opcodes output
35212 Extend the description of the MI command --data-disassemble.
35213 Specifically, expand the description of the 'opcodes' field to explain
35214 how the bytes are formatted.
35216 2022-06-29 Yvan Roux <yvan.roux@foss.st.com>
35218 gdb/arm: Only stack S16..S31 when FPU registers are secure
35219 The FPCCR.TS bit is used to identify if FPU registers are considered
35220 non-secure or secure. If they are secure, then callee saved registers
35221 (S16 to S31) are stacked on exception entry or otherwise skipped.
35223 2022-06-29 Andrew Burgess <aburgess@redhat.com>
35225 opcodes/aarch64: split off creation of comment text in disassembler
35226 The function aarch64_print_operand (aarch64-opc.c) is responsible for
35227 converting an instruction operand into the textual representation of
35230 In some cases, a comment is included in the operand representation,
35231 though this (currently) only happens for the last operand of the
35234 In a future commit I would like to enable the new libopcodes styling
35235 for AArch64, this will allow objdump and GDB[1] to syntax highlight
35236 the disassembler output, however, having operands and comments
35237 combined in a single string like this makes such styling harder.
35239 In this commit, I propose to extend aarch64_print_operand to take a
35240 second buffer. Any comments for the instruction are written into this
35241 extra buffer. The two callers of aarch64_print_operand are then
35242 updated to pass an extra buffer, and print any resulting comment.
35244 In this commit no styling is added, that will come later. However, I
35245 have adjusted the output slightly. Before this commit some comments
35246 would be separated from the instruction operands with a tab character,
35247 while in other cases the comment was separated with two single spaces.
35249 After this commit I use a single tab character in all cases. This
35250 means a few test cases needed updated. If people would prefer me to
35251 move everyone to use the two spaces, then just let me know. Or maybe
35252 there was a good reason why we used a mix of styles, I could probably
35253 figure out a way to maintain the old output exactly if that is
35256 Other than that, there should be no user visible changes after this
35259 [1] GDB patches have not been merged yet, but have been posted to the
35261 https://sourceware.org/pipermail/gdb-patches/2022-June/190142.html
35263 2022-06-29 Carl Love <cel@us.ibm.com>
35264 Andrew Burgess <aburgess@redhat.com>
35266 gdb/testsuite: fix gdb.base/break-idempotent.exp on ppc
35267 When running the gdb.base/break-idempotent.exp test on ppc, I was
35268 seeing some test failures (or rather errors), that looked like this:
35271 Hardware watchpoint 2: local
35273 has_hw_wp_support: Hardware watchpoint detected
35274 ERROR: no fileid for gcc2-power8
35275 ERROR: Couldn't send delete breakpoints to GDB.
35276 ERROR OCCURED: can't read "gdb_spawn_id": no such variable
35279 -i 1000 -timeout 100
35280 -re ".*A problem internal to GDB has been detected" {
35281 fail "$message (GDB internal error)"
35282 gdb_internal_erro..."
35283 ("uplevel" body line 1)
35284 invoked from within
35286 What happens is that in break-idempotent.exp we basically do this:
35288 if {[prepare_for_testing "failed to prepare" $binfile $srcfile $opts]} {
35294 if {![skip_hw_watchpoint_tests]} {
35295 test_break $always_inserted "watch"
35298 The problem with this is that skip_hw_watchpoint_tests, includes this:
35300 if { [istarget "i?86-*-*"]
35301 || [istarget "x86_64-*-*"]
35302 || [istarget "ia64-*-*"]
35303 || [istarget "arm*-*-*"]
35304 || [istarget "aarch64*-*-*"]
35305 || ([istarget "powerpc*-*-linux*"] && [has_hw_wp_support])
35306 || [istarget "s390*-*-*"] } {
35310 For powerpc only we call has_hw_wp_support. This is a caching proc
35311 that runs a test within GDB to detect if we have hardware watchpoint
35314 Unfortunately, to run this test we restart GDB, and when the test has
35315 completed, we exit GDB. This means that in break-idempotent.exp, when
35316 we call skip_hw_watchpoint_tests for the first time on powerpc, GDB
35317 will unexpectedly be exited. When we later call delete_breakpoints we
35318 see the errors I reported above.
35320 The fix is to call skip_hw_watchpoint_tests early, before we start GDB
35321 as part of the break-idempotent.exp script, and store the result in a
35322 variable, we can then check this variable in the script as needed.
35324 After this change break-idempotent.exp runs fine on powerpc.
35326 2022-06-29 Jan Beulich <jbeulich@suse.com>
35328 x86: drop stray NoRex64 from XBEGIN
35329 Presumably this being there was a result of taking CALL as a reference
35330 when adding the RTM insns. But with No_qSuf the attribute has no effect.
35332 2022-06-29 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
35334 gprofng: fix build when BUILD_MAN is false
35336 2022-06-28 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
35339 * gp-display-html/Makefile.am: Set man_MANS only when BUILD_MAN is true.
35340 * src/Makefile.am: Likewise.
35341 * gp-display-html/Makefile.in: Rebuild.
35342 * src/Makefile.in: Rebuild.
35344 2022-06-29 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
35346 gprofng: use $(sysconfdir) instead $(prefix)/etc
35348 2022-06-28 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
35351 * src/Makefile.am: Use $(sysconfdir) instead $(prefix)/etc.
35352 * src/Settings.cc: Likewise.
35353 * src/Makefile.in: Rebuild.
35355 2022-06-29 Alan Modra <amodra@gmail.com>
35357 Re: ld/x86: skip p_align-1 tests with unsuitable compiler
35358 commit d0e0f9c87a3e results "ERROR: i586-linux-cc does not exist" if
35359 cross-building an i586-linux target without a target compiler
35362 * testsuite/ld-elf/linux-x86.exp (compiler_honours_aligned): New.
35363 Use it after first testing check_compiler_available.
35365 2022-06-29 GDB Administrator <gdbadmin@sourceware.org>
35367 Automatic date update in version.in
35369 2022-06-28 Pedro Alves <pedro@palves.net>
35371 gdb+gdbserver/Linux: avoid reading registers while going through shell
35372 For every stop, Linux GDB and GDBserver save the stopped thread's PC,
35373 in lwp->stop_pc. This is done in save_stop_reason, in both
35374 gdb/linux-nat.c and gdbserver/linux-low.cc. However, while we're
35375 going through the shell after "run", in startup_inferior, we shouldn't
35376 be reading registers, as we haven't yet determined the target's
35377 architecture -- the shell's architecture may not even be the same as
35378 the final inferior's.
35380 In gdb/linux-nat.c, lwp->stop_pc is only needed when the thread has
35381 stopped for a breakpoint, and since when going through the shell, no
35382 breakpoint is going to hit, we could simply teach save_stop_reason to
35383 only record the stop pc when the thread stopped for a breakpoint.
35385 However, in gdbserver/linux-low.cc, lwp->stop_pc is used in more cases
35386 than breakpoint hits (e.g., it's used in tracepoints & the
35387 "while-stepping" feature).
35389 So to avoid GDB vs GDBserver divergence, we apply the same approach to
35390 both implementations.
35392 We set a flag in the inferior (process in GDBserver) whenever it is
35393 being nursed through the shell, and when that flag is set,
35394 save_stop_reason bails out early. While going through the shell,
35395 we'll only ever get process exits (normal or signalled), random
35396 signals, and exec events, so nothing is lost.
35398 Change-Id: If0f01831514d3a74d17efd102875de7d2c6401ad
35400 2022-06-28 Tom de Vries <tdevries@suse.de>
35402 [gdb/build] Fix gdb build with -fsanitize=thread and gcc 7
35403 When building gdb with system gcc 7.5.0, I run into:
35405 gdb/ia64-tdep.c: In function ‘int is_float_or_hfa_type_recurse(type*, type**)’:
35406 gdb/ia64-tdep.c:3362:1: error: control reaches end of non-void function \
35407 [-Werror=return-type]
35410 This is due to PR gcc/81275 - "-fsanitize=thread produce incorrect
35411 -Wreturn-type warning", which has been fixed in gcc-8.
35413 Work around this by moving the default return outside the switch.
35415 Tested on x86_64-linux.
35417 2022-06-28 Clément Chigot <chigot@adacore.com>
35419 bfd: handle codepage when opening files on MinGW
35420 Even if MS docs say that CP_UTF8 should always be used on newer
35421 applications, forcing it might produce undefined filename if the
35422 encoding isn't UTF-8.
35423 MinGW seems to call ___lc_codepage_func() in order to retrieve the
35424 current thread codepage.
35428 * bfdio.c (_bfd_real_fopen): Retrieve codepage with
35429 ___lc_codepage_func() on MinGW.
35431 2022-06-28 Clément Chigot <chigot@adacore.com>
35433 windres: add quotes around preprocessor cmd if needed
35434 This patch ensures that the gcc binary called by windres is quoted if
35435 needed. Otherwise, errors can occur if the gcc is under a folder having
35436 a name containing a space (eg "Program Files").
35439 * resrc.c (DEFAULT_PREPROCESSOR): Split into...
35440 (DEFAULT_PREPROCESSOR_CMD): that...
35441 (DEFAULT_PREPROCESSOR_ARGS): and that.
35442 (look_for_default): Add quotes around the command if needed.
35443 (read_rc_file): Adapt to new defines.
35445 2022-06-28 Nick Clifton <nickc@redhat.com>
35447 Fix the display of the idnex values for DW_FORM_loclistx and DW_FORM_rnglistx. Correct the display of .debug.loclists sections.
35449 * dwarf.c (display_debug_rnglists): New function, broken out of..
35450 (display_debug_ranges): ... here.
35451 (read_and_display_attr_value): Correct calculation of index
35452 displayed for DW_FORM_loclistx and DW_FORM_rnglistx.
35453 * testsuite/binutils-all/x86-64/pr26808.dump: Update expected
35456 2022-06-28 Jan Beulich <jbeulich@suse.com>
35458 ld/x86: skip p_align-1 tests with unsuitable compiler
35459 When the compiler doesn't properly arrange for foo's alignment, there's
35460 no point even trying these tests. Report the situation as a single
35461 "unsupported" test.
35463 2022-06-28 Alan Modra <amodra@gmail.com>
35465 PowerPC64: align plt_branch stubs
35466 plt_branch stubs are similar to plt_call stubs in that they branch
35467 via bctr. Align them too.
35470 * elf64-ppc.c (ppc_size_one_stub): Align plt_branch stubs as for
35473 * testsuite/ld-powerpc/elfv2exe.d: Adjust for plt_branch changes.
35474 * testsuite/ld-powerpc/notoc.d: Likewise.
35475 * testsuite/ld-powerpc/notoc.wf: Likewise.
35476 * testsuite/ld-powerpc/notoc3.d: Likewise.
35477 * testsuite/ld-powerpc/pr23937.d: Likewise.
35479 2022-06-28 Alan Modra <amodra@gmail.com>
35481 PowerPC64: plt_stub_pad
35482 * elf64-ppc.c (plt_stub_pad): Simplify parameters and untangle
35483 from plt_stub_size.
35484 (ppc_size_one_stub): Call plt_stub_size before plt_stub_pad to
35485 provide size. Recalculate size if it might change.
35487 2022-06-28 Alan Modra <amodra@gmail.com>
35489 PowerPC64: Tidy stub type changes
35490 It made sense before I started using separate fields for main type and
35491 sub type to add a difference in main type to the type (thus keeping
35492 sub type unchanged). Not so much now.
35494 * elf64-ppc.c (ppc_merge_stub): Simplify stub type change.
35495 (ppc_size_one_stub): Likewise.
35497 2022-06-28 Jiangshuai Li <jiangshuai_li@c-sky.com>
35499 gdb:csky add pseudo regs for float and vector regs
35500 In the existing CSKY architecture, there are at most 32 floating
35501 and 16 vector registers. Float registers's count can be configured
35502 as 16 or 32. In the future, the vector registers's count may be
35505 The bit width of floating-point register is 64bits, and the bit
35506 width of vector register is 128bit.
35508 Special points: in fr0~fr15 and vr0~vr15, each FRx is the lower
35509 64 bits of the corresponding VRx.
35511 Here, we will split each floating-point and vector register to
35512 32bits wide, add the corresponding pseudo registers, and finally
35513 use them for the dwarf registers.
35515 There are 128 pseudo registers in total, s0~s127, including:
35516 1. s0 and s1 correspond to fr0, s4 and s5 correspond to fr1, and so on.
35517 Every two separated pseudo registers correspond to a float register.
35518 2. s0, s1, s2 and s3 correspond to vr0; s4, s5, s6 and s7 correspond to vr1,
35519 and so on. Every four pseudo registers corresponds to a vector register.
35521 Therefore, in s64~s127, there are general registers that are not actually
35522 used. This part is to prepare for the expansion of vector registers to 32
35524 Therefore, in s64~s127, half of the registers are actually unused. This
35525 part is to prepare for the expansion of the vector register to 32.
35527 2022-06-28 Pekka Seppänen <pexu@sourceware.mail.kapsi.fi>
35529 PR29293, elfnn-aarch64.c: def_protected member unintialized
35531 * elfnn-aarch64.c (elfNN_aarch64_link_hash_newfunc): Init def_protected.
35533 2022-06-28 Tsukasa OI <research_trasio@irq.a4lg.com>
35535 RISC-V: Add 'Sstc' extension and its CSRs
35536 This commit adds "stimecmp / vstimecmp" Extension (Sstc) and its CSRs.
35540 * elfxx-riscv.c (riscv_supported_std_s_ext): Add 'Sstc'
35541 extension to valid 'S' extension list.
35545 * config/tc-riscv.c (enum riscv_csr_class): Add CSR classes for
35546 'Sstc' extension. (riscv_csr_address): Add handling for new CSR
35548 * testsuite/gas/riscv/csr-dw-regnums.s: Add new CSRs.
35549 * testsuite/gas/riscv/csr-dw-regnums.d: Likewise.
35550 * testsuite/gas/riscv/csr.s: Add new CSRs.
35551 * testsuite/gas/riscv/csr-version-1p9p1.d: Likewise.
35552 * testsuite/gas/riscv/csr-version-1p9p1.l: Likewise.
35553 * testsuite/gas/riscv/csr-version-1p10.d: Likewise.
35554 * testsuite/gas/riscv/csr-version-1p10.l: Likewise.
35555 * testsuite/gas/riscv/csr-version-1p11.d: Likewise.
35556 * testsuite/gas/riscv/csr-version-1p11.l: Likewise.
35557 * testsuite/gas/riscv/csr-version-1p12.d: Likewise.
35558 * testsuite/gas/riscv/csr-version-1p12.l: Likewise.
35562 * opcode/riscv-opc.h (CSR_STIMECMP, CSR_STIMECMPH,
35563 CSR_VSTIMECMP, CSR_VSTIMECMPH): New CSR macros.
35565 2022-06-28 Tsukasa OI <research_trasio@irq.a4lg.com>
35567 RISC-V: Add 'Sscofpmf' extension with its CSRs
35568 This commit adds Count Overflow and Mode-Based Filtering Extension
35569 (Sscofpmf) and its CSRs.
35573 * elfxx-riscv.c (riscv_supported_std_s_ext): Add 'Sscofpmf'
35574 extension to valid 'S' extension list.
35578 * config/tc-riscv.c (enum riscv_csr_class): Add CSR classes for
35579 'Sscofpmf' extension. (riscv_csr_address): Add handling for new
35581 * testsuite/gas/riscv/csr-dw-regnums.s: Add new CSRs.
35582 * testsuite/gas/riscv/csr-dw-regnums.d: Likewise.
35583 * testsuite/gas/riscv/csr.s: Add new CSRs.
35584 * testsuite/gas/riscv/csr-version-1p9p1.d: Likewise.
35585 * testsuite/gas/riscv/csr-version-1p9p1.l: Likewise.
35586 * testsuite/gas/riscv/csr-version-1p10.d: Likewise.
35587 * testsuite/gas/riscv/csr-version-1p10.l: Likewise.
35588 * testsuite/gas/riscv/csr-version-1p11.d: Likewise.
35589 * testsuite/gas/riscv/csr-version-1p11.l: Likewise.
35590 * testsuite/gas/riscv/csr-version-1p12.d: Likewise.
35591 * testsuite/gas/riscv/csr-version-1p12.l: Likewise.
35595 * opcode/riscv-opc.h (CSR_SCOUNTOVF, CSR_MHPMEVENT3H,
35596 CSR_MHPMEVENT4H, CSR_MHPMEVENT5H, CSR_MHPMEVENT6H,
35597 CSR_MHPMEVENT7H, CSR_MHPMEVENT8H, CSR_MHPMEVENT9H,
35598 CSR_MHPMEVENT10H, CSR_MHPMEVENT11H, CSR_MHPMEVENT12H,
35599 CSR_MHPMEVENT13H, CSR_MHPMEVENT14H, CSR_MHPMEVENT15H,
35600 CSR_MHPMEVENT16H, CSR_MHPMEVENT17H, CSR_MHPMEVENT18H,
35601 CSR_MHPMEVENT19H, CSR_MHPMEVENT20H, CSR_MHPMEVENT21H,
35602 CSR_MHPMEVENT22H, CSR_MHPMEVENT23H, CSR_MHPMEVENT24H,
35603 CSR_MHPMEVENT25H, CSR_MHPMEVENT26H, CSR_MHPMEVENT27H,
35604 CSR_MHPMEVENT28H, CSR_MHPMEVENT29H, CSR_MHPMEVENT30H,
35605 CSR_MHPMEVENT31H): New CSR macros.
35607 2022-06-28 Tsukasa OI <research_trasio@irq.a4lg.com>
35609 RISC-V: Add 'Smstateen' extension and its CSRs
35610 This commit adds State Enable Extension (Smstateen) and its CSRs.
35614 * elfxx-riscv.c (riscv_supported_std_s_ext): Add 'Smstateen'
35615 extension to valid 'S' extension list.
35619 * config/tc-riscv.c (enum riscv_csr_class): Add CSR classes for
35620 'Smstateen' extension. (riscv_csr_address): Add handling for
35622 * testsuite/gas/riscv/csr-dw-regnums.s: Add new CSRs.
35623 * testsuite/gas/riscv/csr-dw-regnums.d: Likewise.
35624 * testsuite/gas/riscv/csr.s: Add new CSRs.
35625 * testsuite/gas/riscv/csr-version-1p9p1.d: Likewise.
35626 * testsuite/gas/riscv/csr-version-1p9p1.l: Likewise.
35627 * testsuite/gas/riscv/csr-version-1p10.d: Likewise.
35628 * testsuite/gas/riscv/csr-version-1p10.l: Likewise.
35629 * testsuite/gas/riscv/csr-version-1p11.d: Likewise.
35630 * testsuite/gas/riscv/csr-version-1p11.l: Likewise.
35631 * testsuite/gas/riscv/csr-version-1p12.d: Likewise.
35632 * testsuite/gas/riscv/csr-version-1p12.l: Likewise.
35636 * opcode/riscv-opc.h (CSR_MSTATEEN0, CSR_MSTATEEN1,
35637 CSR_MSTATEEN2, CSR_MSTATEEN3, CSR_SSTATEEN0, CSR_SSTATEEN1,
35638 CSR_SSTATEEN2, CSR_SSTATEEN3, CSR_HSTATEEN0, CSR_HSTATEEN1,
35639 CSR_HSTATEEN2, CSR_HSTATEEN3, CSR_MSTATEEN0H, CSR_MSTATEEN1H,
35640 CSR_MSTATEEN2H, CSR_MSTATEEN3H, CSR_HSTATEEN0H, CSR_HSTATEEN1H,
35641 CSR_HSTATEEN2H, CSR_HSTATEEN3H): New CSR macros.
35643 2022-06-28 Tsukasa OI <research_trasio@irq.a4lg.com>
35645 RISC-V: Add new CSR feature gate handling (RV32,H)
35646 To support feature gate like Smstateen && H, this commit adds certain
35647 CSR feature gate handling. It also changes how RV32-only CSRs are
35648 handled for cleanliness.
35652 * config/tc-riscv.c (riscv_csr_address): Add CSR feature gate
35653 handling for H. Change handling on RV32.
35655 2022-06-28 Alan Modra <amodra@gmail.com>
35657 Re: Disable execstack and rwx segments warnings for MIPS targets.
35659 * configure.ac: Fix typo.
35660 * testsuite/ld-elf/elf.exp: Add mips to targets that need
35661 --warn-execstack to pass first pr29072 test.
35663 2022-06-28 GDB Administrator <gdbadmin@sourceware.org>
35665 Automatic date update in version.in
35667 2022-06-27 Bruno Larsen <blarsen@redhat.com>
35669 gdb/testsuite: update bug numbers from Gnats to bugzilla
35670 Some tests link to outdated bug numbers when an XFAIL or a KFAIL happen.
35672 gdb.base/macscp.exp was referencing bug number 555, and the bug 7660
35673 mentions that it used to be 555 on the Gnats system and seems to relate
35674 to the issue at hand.
35676 gdb.base/annota1.exp was referencing bug number 1270, and bug 8375
35677 mentions being number 1270 on Gnats, and mentions annota1 specifically,
35678 so it seemed pretty obvious.
35680 2022-06-27 Tom de Vries <tdevries@suse.de>
35682 [gdb/build] Fix build breaker with --enable-shared
35683 When building gdb with --enable-shared, I run into:
35685 ld: build/zlib/libz.a(libz_a-inffast.o): relocation R_X86_64_32S against \
35686 `.rodata' can not be used when making a shared object; recompile with -fPIC
35687 ld: build/zlib/libz.a(libz_a-inflate.o): warning: relocation against \
35688 `inflateResetKeep' in read-only section `.text'
35689 collect2: error: ld returned 1 exit status
35690 make[3]: *** [libbfd.la] Error 1
35693 This is a regression since commit a08bdb159bb ("[gdb/build] Fix gdbserver
35694 build with -fsanitize=thread").
35696 The problem is that a single case statement in configure is shared to handle
35697 special requirements for both the host libiberty and host zlib, which has the
35698 effect that only one is handled.
35700 Fix this by handling libiberty and zlib each in its own case statement.
35702 Build on x86_64-linux, with and without --enable-shared.
35706 2022-06-27 Tom de Vries <tdevries@suse.de>
35708 * configure.ac: Set extra_host_libiberty_configure_flags and
35709 extra_host_zlib_configure_flags in separate case statements.
35710 * configure: Regenerate.
35712 2022-06-27 Pedro Alves <pedro@palves.net>
35714 Make GDBserver abort on internal error in development mode
35715 Currently, if GDBserver hits some internal assertion, it exits with
35716 error status, instead of aborting. This makes it harder to debug
35717 GDBserver, as you can't just debug a core file if GDBserver fails an
35718 assertion. I've had to hack the code to make GDBserver abort to debug
35719 something several times before.
35721 I believe the reason it exits instead of aborting, is to prevent
35722 potentially littering the filesystem of smaller embedded targets with
35723 core files. I think I recall Daniel Jacobowitz once saying that many
35724 years ago, but I can't be sure. Anyhow, that seems reasonable to me.
35726 Since we nowadays have a distinction between development and release
35727 modes, I propose to make GDBserver abort on internal error if in
35728 development mode, while keeping the status quo when in release mode.
35730 Thus, after this patch, in development mode, you get:
35732 $ ../gdbserver/gdbserver
35733 ../../src/gdbserver/server.cc:3711: A problem internal to GDBserver has been detected.
35734 captured_main: Assertion `0' failed.
35735 Aborted (core dumped)
35738 while in release mode, you'll continue to get:
35740 $ ../gdbserver/gdbserver
35741 ../../src/gdbserver/server.cc:3711: A problem internal to GDBserver has been detected.
35742 captured_main: Assertion `0' failed.
35746 I do not think that this requires a separate configure switch.
35748 A "--target_board=native-extended-gdbserver" run on Ubuntu 20.04 ends
35751 === gdb Summary ===
35753 # of unexpected core files 29
35756 for me, of which 8 are GDBserver core dumps, 7 more than without this
35759 Change-Id: I6861e08ad71f65a0332c91ec95ca001d130b0e9d
35761 2022-06-27 Nick Clifton <nickc@redhat.com>
35763 Replace a run-time assertion failure with a warning message when parsing corrupt DWARF data.
35765 * dwarf.c (display_debug_names): Replace assert with a warning
35768 Fix NULL pointer indirection when parsing corrupt DWARF data.
35770 * dwarf.c (read_and_display_attr_value): Check that debug_info_p
35771 is set before dereferencing it.
35773 Have gold's File_read::do_read() function check the start parameter
35775 * fileread.cc (File_read::do_read): Check start parameter before
35776 computing number of bytes to read.
35778 2022-06-27 Yvan Roux <yvan.roux@foss.st.com>
35780 gdb/arm: Unwind Non-Secure callbacks from Secure
35781 Without this changeset, the unwinding doesn't take into account
35782 Non-Secure to Secure stack unwinding enablement status and
35783 doesn't choose the proper SP to do the unwinding.
35785 This patch only unwinds the stack when Non-Secure to Secure
35786 unwinding is enabled, previous SP is set w/r to the current mode
35787 (Handler -> msp_s, Thread -> psp_s) and then the Secure stack is
35788 unwound. Ensure thumb bit is set in PSR when needed. Also, drop
35789 thumb bit from PC if set.
35791 2022-06-27 Nick Clifton <nickc@redhat.com>
35793 Stop bogus warnings about DWARF indexed string offsets being too big.
35794 * dwarf.c (fetch_indexed_string): Do not use length of first table
35795 in string section as the length of every table in the section.
35796 * testsuite/binutils-all/pr26112.r: Update expected output.
35798 2022-06-27 Tom de Vries <tdevries@suse.de>
35800 [gdb/testsuite] Handle older python in gdb.python/py-send-packet.py
35801 With python 3.4, I run into:
35803 Traceback (most recent call last):^M
35804 File "<string>", line 1, in <module>^M
35806 "outputs/gdb.python/py-send-packet/py-send-packet.py", line 128, in \
35807 run_set_global_var_test^M
35808 res = conn.send_packet(b"X%x,4:\x02\x02\x02\x02" % addr)^M
35809 TypeError: Could not convert Python object: b'X%x,4:\x02\x02\x02\x02'.^M
35810 Error while executing Python code.^M
35812 while with python 3.6 this works fine.
35814 The type of addr is <class 'gdb.Value'>, so the first thing to try is whether
35815 changing it into a string works:
35817 addr_str = "%x" % addr
35818 res = conn.send_packet(b"X%s,4:\x02\x02\x02\x02" % addr_str)
35820 which gets us the more detailed:
35822 TypeError: unsupported operand type(s) for %: 'bytes' and 'str'
35825 Fix this by avoiding the '%' operator in the byte literal, and use instead:
35827 def xpacket_header (addr):
35828 return ("X%x,4:" % addr).encode('ascii')
35830 res = conn.send_packet(xpacket_header(addr) + b"\x02\x02\x02\x02")
35833 Tested on x86_64-linux, with python 3.4 and 3.6, and a backported version was
35834 tested on the gdb-12-branch in combination with python 2.7.
35836 2022-06-27 Tom de Vries <tdevries@suse.de>
35838 [gdb/testsuite] Fix gdb.reverse/i387-env-reverse.exp for -pie
35839 When running test-case gdb.reverse/i387-env-reverse.exp for x86_64-linux with
35840 target board unix/-m32/-fPIE/-pie, we run into:
35842 (gdb) PASS: gdb.reverse/i387-env-reverse.exp: push st0
35843 info register eax^M
35844 eax 0x56550000 1448411136^M
35845 (gdb) FAIL: gdb.reverse/i387-env-reverse.exp: verify eax == 0x8040000
35848 The problem is that the tested instruction (fstsw) only sets $ax, not $eax.
35850 Fix this by verifying $ax instead of $eax.
35852 Tested on x86_64-linux with target boards unix/-m32 and unix/-m32/-fPIE/-pie.
35854 2022-06-27 Tom de Vries <tdevries@suse.de>
35856 [gdb/testsuite] Enable some test-cases for x86_64 -m32
35857 When trying to run test-case gdb.reverse/i387-env-reverse.exp for x86_64-linux
35858 with target board unix/-m32, it's skipped.
35860 Fix this by using is_x86_like_target instead of istarget "i?86-*linux*".
35862 This exposes a number of duplicates, fix those by making the test names unique.
35864 Likewise in a couple of other test-cases.
35866 Tested on x86_64-linux with target boards unix/-m32.
35868 2022-06-27 Tom de Vries <tdevries@suse.de>
35870 [gdb/testsuite] Workaround unnecessary .s file with gfortran 4.8
35871 After running test-case gdb.fortran/namelist.exp with gfortran 4.8.5, I'm left
35876 Your branch is up to date with 'origin/master'.
35879 (use "git add <file>..." to include in what will be committed)
35880 gdb/testsuite/lib/compiler.s
35882 nothing added to commit but untracked files present (use "git add" to track)
35885 We're running into PR gcc/60447, which was fixed in gcc 4.9.0.
35887 Workaround this by first copying the source file to the temp dir, such that
35888 the .s file is left there instead:
35890 $ ls build/gdb/testsuite/temp/<runtest pid>/
35891 compiler.c compiler.F90 compiler.s
35894 Tested on x86_64-linux.
35896 2022-06-27 Tom de Vries <tdevries@suse.de>
35898 [gdb/testsuite] Skip gdb.fortran/namelist.exp for gfortran 4.8
35899 The test-case gdb.fortran/namelist.exp uses a gfortran feature (emitting
35900 DW_TAG_namelist in the debug info) that has been supported since gfortran 4.9,
35903 Skip the test for gfortran 4.8 and earlier. Do this using gcc_major_version,
35904 and update it to be able to handle "gcc_major_version {gfortran-*} f90".
35906 Tested on x86_64-linux, with gfortran 4.8.5, 7.5.0, and 12.1.1.
35908 2022-06-27 Tom de Vries <tdevries@suse.de>
35910 [gdb/symtab] Fix parsing of .debug_str_offsets header
35911 When running test-case gdb.dwarf2/fission-mix.exp with target board dwarf64
35912 and gcc-12 (defaulting to DWARF5), I run into:
35914 (gdb) break func2^M
35915 Offset from DW_FORM_GNU_str_index or DW_FORM_strx pointing outside of \
35916 .debug_str.dwo section in CU at offset 0x0 [in module fission-mix]^M
35917 (gdb) FAIL: gdb.dwarf2/fission-mix.exp: break func2
35920 The .debug_str_offsets section has version 5, so as per the standard it has
35921 it's own header, with initial length and version:
35923 Contents of the .debug_str_offsets.dwo section (loaded from fission-mix2.dwo):
35927 Index Offset [String]
35928 0 0 build/gdb/testsuite
35930 2 8f src/gdb/testsuite/gdb.dwarf2/fission-mix-2.c
35933 But when trying to read the string offset at index 0 in the table (which
35934 is 0), we start reading at offset 8, which points in the header, at the last
35935 4 bytes of the initial length (it's 12 bytes because of 64-bit dwarf), as well
35936 at the 2-byte version field and 2 bytes of padding, so we get:
35938 (gdb) p /x str_offset
35941 which indeed is an offset that doesn't fit in the .debug_str section.
35943 The offset 8 is based on reader->cu->header.addr_size:
35945 static const char *
35946 read_dwo_str_index (const struct die_reader_specs *reader, ULONGEST str_index)
35948 ULONGEST str_offsets_base = reader->cu->header.version >= 5
35949 ? reader->cu->header.addr_size : 0;
35951 which doesn't in look in agreement with the standard.
35953 Note that this happens to give the right answer for 32-bit dwarf and
35954 addr_size == 8, because then we have header size ==
35955 (initial length (4) + version (2) + padding (2)) == 8.
35957 Conversely, for 32-bit dwarf and addr_size == 4 (target board unix/-m32)
35958 we run into a similar problem. It just happens to not trigger the warning,
35959 instead we get the wrong strings, like "func2" for DW_AT_producer and
35960 "build/gdb/testsuite" for DW_AT_name of the DW_TAG_compile_unit DIE.
35962 Fix this by parsing the .debug_str_offsets header in read_dwo_str_index.
35964 Add a FIXME that we should not parse this for every call.
35966 Tested on x86_64-linux.
35968 2022-06-27 Tom de Vries <tdevries@suse.de>
35970 [gdb/build] Fix gdbserver build with -fsanitize=thread
35971 [ Copied from gcc commit 153689603fd ("[gdb/build] Fix gdbserver build with
35972 -fsanitize=thread"). ]
35974 When building gdbserver with -fsanitize=thread (added to CFLAGS/CXXFLAGS) we
35977 ld: ../libiberty/libiberty.a(safe-ctype.o): warning: relocation against \
35978 `__tsan_init' in read-only section `.text'
35979 ld: ../libiberty/libiberty.a(safe-ctype.o): relocation R_X86_64_PC32 \
35980 against symbol `__tsan_init' can not be used when making a shared object; \
35981 recompile with -fPIC
35982 ld: final link failed: bad value
35983 collect2: error: ld returned 1 exit status
35984 make[1]: *** [libinproctrace.so] Error 1
35986 which looks similar to what is described in commit 78e49486944 ("[gdb/build]
35987 Fix gdbserver build with -fsanitize=address").
35989 The gdbserver component builds a shared library libinproctrace.so, which uses
35990 libiberty and therefore requires the pic variant. The gdbserver Makefile is
35991 setup to use this variant, if available, but it's not there.
35993 Fix this by listing gdbserver in the toplevel configure alongside libcc1, as a
35994 component that needs the libiberty pic variant, setting:
35996 extra_host_libiberty_configure_flags=--enable-shared
35999 Tested on x86_64-linux.
36003 2022-06-27 Tom de Vries <tdevries@suse.de>
36005 * configure.ac: Build libiberty pic variant for gdbserver.
36006 * configure: Regenerate.
36008 2022-06-27 Nick Clifton <nickc@redhat.com>
36010 Disable execstack and rwx segments warnings for MIPS targets.
36012 * configure.ac: Move HPPA specific code from here...
36013 * configure.tgt: ... to here. Add similar code for MIPS.
36014 Move code for CRIS, MIPS and HPPA to block at start of file.
36015 * configure: Regenerate.
36017 2022-06-27 Jan Beulich <jbeulich@suse.com>
36019 bfd: prune config.bfd's setting of targ_archs
36020 The final "match all" case can take care of a few explicit entries:
36021 Purge those. Also move s12z* into proper position (the table is
36022 otherwise sorted, after all).
36024 2022-06-27 Jan Beulich <jbeulich@suse.com>
36027 Commit 04f096fb9e25 ("Move the xc16x target to the obsolete list") moved
36028 the architecture from the "obsolete but still available" to the
36029 "obsolete / support removed" list in config.bfd, making the architecture
36030 impossible to enable (except maybe via "enable everything" options").
36032 Note that I didn't touch */po/*.po{,t} on the assumption that these
36033 would be updated by some (half)automatic means.
36035 2022-06-27 Bhuvanendra Kumar N <Bhuvanendra.KumarN@amd.com>
36037 Fix location list offset address dump under DW_AT_location (dwarf-5)
36038 For clang compiled objects with dwarf-5, location list offset address dump
36039 under DW_AT_location is corrected, where DW_FORM_loclistx is used. While
36040 dumping the location list offset, the address dumped is wrong where it was
36041 refering to .debug_addr instead of .debug_loclists
36043 * dwarf.c (fetch_indexed_value): Add base_address as parameter and
36044 use it to access the section offset.
36045 (read_and_display_attr_value): Handle DW_FORM_loclistx form separately.
36046 Pass loclists_base to fetch_indexed_value().
36048 2022-06-27 Alan Modra <amodra@gmail.com>
36050 PowerPC64 .branch_lt address
36051 .branch_lt is really an extension of .plt, as is .iplt. We'd like all
36052 of the PLT sections to be fixed relative to .TOC. after stub sizing,
36053 because changes in offset to PLT entries might mean a change in stub
36054 sizes. When -z relro, the relro layout does this by laying out
36055 sections from the end of the relro segment. So for example, a change
36056 in .eh_frame (which happens after stub sizing) will keep the same GOT
36057 to PLT offset when -z relro. Not so when -z norelro, because then the
36058 usual forward layout of section is done and .got is more aligned than
36061 * emulparams/elf64ppc.sh: Set .branch_lt address fixed relative
36063 * testsuite/ld-powerpc/elfv2exe.d: Adjust to suit.
36065 2022-06-27 Alan Modra <amodra@gmail.com>
36067 -z relro relaxation and ld script SIZEOF
36068 A number of targets use assignments like:
36069 . = DATA_SEGMENT_RELRO_END (SIZEOF (.got.plt) >= 12 ? 12 : 0, .);
36070 (from i386) in linker scripts to put the end of the relro segment past
36071 the header in .got.plt. Examination of testcases like those edited by
36072 this patch instead sees the end of the relro segment being placed at
36073 the start of .got.plt. For the i386 pie1 test:
36075 [ 9] .got.plt PROGBITS 00002000 001000 00000c 04 WA 0 0 4
36077 GNU_RELRO 0x000f90 0x00001f90 0x00001f90 0x00070 0x00070 R 0x1
36081 .dynamic 0x0000000000001f90 0x70
36083 .dynamic 0x0000000000001f90 0x70 tmpdir/pie1.o
36084 0x0000000000001f90 _DYNAMIC
36086 .got 0x0000000000002000 0x0
36088 .got 0x0000000000002000 0x0 tmpdir/pie1.o
36090 0x0000000000002ff4 . = DATA_SEGMENT_RELRO_END (., (SIZEOF (.got.plt) >= 0xc)?0xc:0x0)
36092 .got.plt 0x0000000000002000 0xc
36094 .got.plt 0x0000000000002000 0xc tmpdir/pie1.o
36095 0x0000000000002000 _GLOBAL_OFFSET_TABLE_
36097 The DATA_SEGMENT_RELRO_END value in the map file is weird too. All of
36098 this is triggered by SIZEOF (.got.plt) being evaluated wrongly as
36099 zero. Fix it by taking into account the action of
36100 lang_reset_memory_regions during relaxation.
36102 * ldexp.c (fold_name <SIZEOF>): Use rawsize if size has been reset.
36103 * ldlang.c (lang_size_sections_1): Don't reset processed_vma here.
36104 * testsuite/ld-i386/pie1.d: Adjust to suit.
36105 * testsuite/ld-x86-64/pr20830a.d: Likewise.
36106 * testsuite/ld-x86-64/pr20830b.d: Likewise.
36107 * testsuite/ld-x86-64/pr21038a.d: Likewise.
36108 * testsuite/ld-x86-64/pr21038b.d: Likewise.
36109 * testsuite/ld-x86-64/pr21038c.d: Likewise.
36111 2022-06-27 GDB Administrator <gdbadmin@sourceware.org>
36113 Automatic date update in version.in
36115 2022-06-26 GDB Administrator <gdbadmin@sourceware.org>
36117 Automatic date update in version.in
36119 2022-06-25 Fangrui Song <i@maskray.me>
36121 arm: Define elf_backend_extern_protected_data to 0 [PR 18705]
36122 Similar to commit 4fb55bf6a9606eb7b626c30a9f4e71d6c2d4fbb2 for aarch64.
36124 Commit b68a20d6675f1360ea4db50a9835c073675b9889 changed ld to produce
36125 R_ARM_GLOB_DAT but that defeated the purpose of protected visibility
36126 as an optimization. Restore the previous behavior (which matches
36127 ld.lld) by defining elf_backend_extern_protected_data to 0.
36129 2022-06-25 Tom Tromey <tom@tromey.com>
36131 Fix corrupt DWARF in dw2-double-set-die-type
36132 The dw2-double-set-die-type.exp test case caused an AddressSanitizer
36133 failure in the new DWARF scanner.
36135 The immediate cause was bad DWARF in the test -- in particular, the
36136 the sibling attribute here:
36138 <2><181>: Abbrev Number: 33 (DW_TAG_subprogram)
36139 <182> DW_AT_external : 1
36140 <183> DW_AT_name : address
36141 <18b> DW_AT_type : <0x171>
36142 <18f> DW_AT_declaration : 1
36143 <190> DW_AT_sibling : <0x1a1>
36145 <1><1a1>: Abbrev Number: 23 (DW_TAG_pointer_type)
36146 <1a2> DW_AT_byte_size : 4
36147 <1a3> DW_AT_type : <0x1a7>
36149 ...points to a "sibling" DIE that is at a different child depth.
36151 Because this test case doesn't really require sibling attributes, this
36152 patch fixes the problem by removing them from the test.
36154 Note that gdb is not generally robust against malformed DWARF.
36155 Detecting and compensating for this problem would probably be
36156 expensive and, IMO, is better left to some (still hypothetical) DWARF
36159 2022-06-25 Tom Tromey <tom@tromey.com>
36161 Fix end of CU calculation in cooked_indexer::index_dies
36162 cooked_indexer::index_dies incorrect computes the end of the current
36163 CU in the .debug_info. This isn't readily testable without writing
36164 intentionally corrupt DWARF, but it's apparent through observation: it
36165 is currently based on 'info_ptr', which does not always point to the
36166 start of the CU. This patch fixes the expression. Tested on x86-64
36169 2022-06-25 Tiezhu Yang <yangtiezhu@loongson.cn>
36171 gdb: LoongArch: Implement loongarch_linux_syscall_next_pc()
36172 When FRAME is at a syscall instruction, return the PC of the next
36173 instruction to be executed.
36175 gdb: LoongArch: Define register numbers and clean up code
36176 This commit defines register numbers of various important registers,
36177 we can use them directly in the related code, and also clean up some
36178 code to make them more clear and readable.
36180 2022-06-25 GDB Administrator <gdbadmin@sourceware.org>
36182 Automatic date update in version.in
36184 2022-06-24 Pedro Alves <pedro@palves.net>
36186 Eliminate TUI/CLI observers duplication
36187 For historical reasons, the CLI and the TUI observers are basically
36188 exact duplicates, except for the downcast:
36191 struct cli_interp *cli = as_cli_interp (interp);
36193 struct interp *tui = as_tui_interp (interp);
36195 and how they get at the interpreter's ui_out:
36200 tui->interp_ui_out ()
36202 Since interp_ui_out() is a virtual method that also works for the CLI
36203 interpreter, and, both the CLI and the TUI interpreters inherit from
36204 the same base class (cli_interp_base), we can convert the CLI
36205 observers to cast to cli_interp_base instead and use interp_ui_out()
36206 too. With that, the CLI observers will work for the TUI interpreter
36207 as well. This lets us completely eliminate the TUI observers. That's
36208 what this commit does.
36210 Change-Id: Iaf6cf12dfa200ed3ab203a895a72b69dfedbd6e0
36212 2022-06-24 Pedro Alves <pedro@palves.net>
36214 Revert "Delete delete_thread_silent"
36215 Turns out we'll be gaining a new use of this function very soon, the
36216 incoming AMDGPU port needs it. Let's add it back, as it isn't really
36219 This reverts commit 39b8a8090ed7e8967ceca3655aa5f3a2ae91219d.
36221 2022-06-24 Yvan Roux <yvan.roux@foss.st.com>
36223 gdb/arm: Update the value of active sp when base sp changes
36224 For Arm Cortex-M33 with security extensions, there are 4 different
36225 stacks pointers (msp_s, msp_ns, psp_s, psp_ns).
36226 When plain "sp" is updated during unwinding of the stack, the active
36227 stack pointer of the 4 stack pointers needs to be kept in sync.
36229 2022-06-24 Andrew Burgess <aburgess@redhat.com>
36231 gdb/testsuite: remove unneeded calls to get_compiler_info
36232 It is not necessary to call get_compiler_info before calling
36233 test_compiler_info, and, after recent commits that removed setting up
36234 the gcc_compiled, true, and false globals from get_compiler_info,
36235 there is now no longer any need for any test script to call
36236 get_compiler_info directly.
36238 As a result every call to get_compiler_info outside of lib/gdb.exp is
36239 redundant, and this commit removes them all.
36241 There should be no change in what is tested after this commit.
36243 2022-06-24 Andrew Burgess <aburgess@redhat.com>
36245 gdb/testsuite: remove global gcc_compiled from gdb.exp
36246 After this commit the gcc_compiled global is no longer exported from
36247 lib/gdb.exp. In theory we could switch over all uses of gcc_compiled
36248 to instead call test_compiler_info directly, however, I have instead
36249 added a new proc to gdb.exp: 'is_c_compiler_gcc'. I've then updated
36250 the testsuite to call this proc instead of using the global.
36252 Having a new proc specifically for this task means that we have a
36253 single consistent pattern for detecting gcc. By wrapping this logic
36254 within a proc that calls test_compiler_info, rather than using the
36255 global, means that test scripts don't need to call get_compiler_info
36256 before they read the global, simply calling the new proc does
36257 everything in one go.
36259 As a result I've been able to remove the get_compiler_info calls from
36260 all the test scripts that I've touched in this commit.
36262 In some of the tests e.g. gdb.dwarf2/*.exp, the $gcc_compiled flag was
36263 being checked at the top of the script to decide if the whole script
36264 should be skipped or not. In these cases I've called the new proc
36265 directly and removed all uses of gcc_compiled.
36267 In other cases, e.g. most of the gdb.base scripts, there were many
36268 uses of gcc_compiled. In these cases I set a new global gcc_compiled
36269 near the top of the script, and leave the rest of the script
36272 There should be no changes in what is tested after this commit.
36274 2022-06-24 Pedro Alves <pedro@palves.net>
36276 Include count of unexpected core files in gdb.sum summary
36277 If GDB, GDBserver, a testcase program, Valgrind, etc. unexpectedly
36278 crash while running the GDB testsuite, and you've setup your machine
36279 such that core files are dumped in the current directory instead of
36280 being shoved somewhere by abrt, apport, or similar (as you should for
36281 proper GDB testing), you'll end up with an unexpected core file in the
36282 $build/gdb/testsuite/ directory.
36284 It can happen that GDB, GDBserver, etc. even crashes _after_ gdb_exit,
36285 during teardown, and thus such a crash won't be noticed by looking at
36286 the gdb.sum file at all. This commit aims at improving that, by
36287 including a new "unexpected core files" line in the testrun summary.
36289 For example, here's what I get on x86-64 Ubuntu 20.04, with this
36292 === gdb Summary ===
36294 # of unexpected core files 12 << new info
36295 # of expected passes 107557
36296 # of unexpected failures 35
36297 # of expected failures 77
36298 # of unknown successes 2
36299 # of known failures 114
36300 # of untested testcases 31
36301 # of unsupported tests 139
36303 I have my core pattern setup like this:
36305 $ cat /proc/sys/kernel/core_pattern
36310 %e: executable filename
36313 %t: UNIX time of dump
36315 and so I get these core files:
36317 $ ls -1 testsuite/core.*
36318 testsuite/core.connect-with-no.216191.nelson.1656002431
36319 testsuite/core.connect-with-no.217729.nelson.1656002431
36320 testsuite/core.gdb.194247.nelson.1656002423
36321 testsuite/core.gdb.226014.nelson.1656002435
36322 testsuite/core.gdb.232078.nelson.1656002438
36323 testsuite/core.gdb.352268.nelson.1656002441
36324 testsuite/core.gdb.4152093.nelson.1656002337
36325 testsuite/core.gdb.4154515.nelson.1656002338
36326 testsuite/core.gdb.4156668.nelson.1656002339
36327 testsuite/core.gdb.4158871.nelson.1656002341
36328 testsuite/core.gdb.468495.nelson.1656002444
36329 testsuite/core.vgdb.4192247.nelson.1656002366
36331 where we can see that GDB crashed a number of times, but also
36332 Valgrind's vgdb, and a couple testcase programs. Neither of which is
36335 If your core_pattern is just "core" (but why??), then I guess that you
36336 may end up with just a single core file in testsuite/. Still, that is
36337 one core file too many.
36339 Above, we see a couple cores for "connect-with-no", which are the
36340 result of gdb.server/connect-with-no-symbol-file.exp. This is a case
36341 mentioned above -- while the program crashed, that happens during
36342 testcase teardown, and it goes unnoticed (without this commit) by
36343 gdb.sum results. Vis:
36345 $ make check TESTS="gdb.server/connect-with-no-symbol-file.exp"
36347 === gdb Summary ===
36349 # of unexpected core files 2
36350 # of expected passes 8
36355 The tests fully passed, but still the testcase program crashed
36358 $ ls -1 testsuite/core.*
36359 testsuite/core.connect-with-no.941561.nelson.1656003317
36360 testsuite/core.connect-with-no.941682.nelson.1656003317
36362 Against --target_board=native-extended-gdbserver it's even worse. I
36365 # of unexpected core files 26
36367 and note that when GDBserver hits an assertion failure, it exits with
36368 error, instead of crashing with SIGABRT. I think that should be
36369 changed, at least on development builds, but that would be for another
36370 patch. After such patch, I suspect the number of unexpected cores
36371 will be higher, as there are likely teardown GDBserver assertions that
36372 we're not noticing.
36374 I decided to put this new info in the "gdb Summary" section, as that's
36375 a place people already are used to looking at, either when looking at
36376 the tail of gdb.sum, or when diffing gdb.sum files, and we've already
36377 extended this section before, to include the count of DUPLICATE and
36378 PATH problems, so there's precedent.
36380 Implementation-wise, the new line is appended after DejaGnu is
36381 finished, with a shell script that is invoked by the Makefile. It is
36382 done this way so that serial and parallel testing work the same way.
36383 My initial cut at an implementation was in TCL, straight in
36384 testsuite/lib/check-test-names.exp, where DUPLICATES and PATH are
36387 @@ -148,6 +159,10 @@ namespace eval ::CheckTestNames {
36388 $counts(paths,$which)
36389 maybe_show_count "# of duplicate test names\t" \
36390 $counts(duplicates,$which)
36392 + set cores [glob -nocomplain -directory $::objdir core*]
36393 + maybe_show_count "# of unexpected core files\t" \
36397 But that would only work for serial testing, as in parallel testing,
36398 the final gdb.sum is generated by aggregating the results of all the
36399 individual gdb.sum files, and dg-extract-results.sh doesn't know about
36400 our new summary line. And I don't think that dg-extract-results.sh
36401 should be taught about it, since the count of core files is not
36402 something that we want to count many times, once per testcase, and
36403 then add up the subcounts at the end. Every time we count the core
36404 files, we're already counting the final count.
36406 I considered using the Tcl implementation in serial mode, and the
36407 script approach for parallel testing, but that has the obvious
36408 downside of implementing and maintaining the same thing twice. In the
36409 end, I settled on the script approach for serial mode too, which
36410 requires making the "check-single" rule print the tail end of the
36411 gdb.sum file, with a side effect being that if you look at the
36412 terminal after a run (instead of at the gdb.sum file), you'll see the
36413 "gdb Summary" section twice, once without the unexpected core lines
36414 printed, and then another with. IMO, this isn't an issue; when
36415 testing in parallel mode, if you look at the terminal after "make -jN
36416 check", you'll also see multiple "gdb Summary" sections printed.
36418 Change-Id: I190b8d41856d49ad143854b6e3e6ccd7caa04491
36420 2022-06-24 Pedro Alves <pedro@palves.net>
36422 Improve core file path detection & put cores in output dir
36423 After a testrun, I noticed that I have some kernel-produced cores for
36424 testcase programs, under build/gdb/testsuite/, which shouldn't be
36427 $ ls -1 testsuite/core.*
36428 testsuite/core.annota1.1274351.nelson.1656004407
36429 testsuite/core.annota3.1288474.nelson.1656004414
36430 testsuite/core.exitsignal.1240674.nelson.1656004391
36432 I have my core pattern setup like this:
36434 $ cat /proc/sys/kernel/core_pattern
36439 %e: executable filename
36442 %t: UNIX time of dump
36444 so it's easy to tell which program produced the core from the core
36447 From above, we can tell that the corresponding testcases are
36448 gdb.base/annota1.exp, gdb.base/annota3.exp and
36449 gdb.base/exitsignal.exp.
36451 At least gdb.base/annota1.exp and gdb.base/annota3.exp have code in
36452 them to delete the core file. However, that isn't working for me,
36453 because said code only looks for cores named exactly either "core" or
36454 "core.PID", and my core_pattern doesn't match that.
36456 Another issue I noticed, is that I have not been running
36457 gdb.base/bigcore.exp, for a similar reason. I get:
36459 Program terminated with signal SIGABRT, Aborted.
36460 The program no longer exists.
36461 (gdb) PASS: gdb.base/bigcore.exp: signal SIGABRT
36462 UNTESTED: gdb.base/bigcore.exp: can't generate a core file
36464 But I actually have a core file under the testcase's output dir:
36466 $ find . -name "core.*"
36467 ./testsuite/outputs/gdb.base/bigcore/core.bigcore.2306705.nelson.1656005213
36470 This commit fixes these things, by adding a find_core_file routine
36471 that searches core files in a way that works with my core pattern as
36472 well. This then also adds a convenience remove_core routine as a
36473 wrapper around find_core_file that removes the found core file.
36475 In addition, it changes some testcases that expect to have their
36476 program dump core, to switch the inferior's cwd to the testcase's
36477 output dir, so that the core is dumped there instead of in
36478 build/gdb/testsuite/. Some testcases were already doing that, but not
36479 all. The idea is that any core file dumped in build/gdb/testsuite/ is
36480 an unexpected core file. The next patch will add a count of such
36481 unexpected core files to gdb.sum.
36483 Another change is that the directory changing is now done with "set
36484 cwd" instead of with "cd". "set cwd" only affects the inferior cwd,
36485 while "cd" affects GDB's cwd too. By using "set cwd" instead of "cd",
36486 if GDB dumps core in these testcases, the GDB core dump will still end
36487 up in build/gdb/testsuite/, and can thus be detected as an unexpected
36490 Change-Id: I45068f21ffd4814350aaa8a3cc65cad5e3107607
36492 2022-06-24 Andrew Burgess <aburgess@redhat.com>
36494 gdb: make use of RAII in run_inferior_call
36495 In passing I noticed that there are three local variables in
36496 run_inferior_call that are used to save, and then restore some state,
36497 I think these could all be replaced with a RAII style scoped_restore
36500 Of the three locals that I've changed, the only one that I believe is
36501 now restored in a different location is ui::async, before this commit
36502 the async field was restored after a call to either delete_file_handle
36503 or ui_register_input_event_handler, and after this commit, the field
36504 is restored before these calls. However, I don't believe that either
36505 of these functions depend on the value of the async field, so I
36506 believe the commit is fine.
36508 Tested on x86-64/Linux passes with no regressions.
36510 2022-06-24 Pedro Alves <pedro@palves.net>
36512 Delete delete_thread_silent
36513 delete_thread_silent is no longer used anywhere. Delete it.
36515 Change-Id: Iafcec12339861d5ab2e29c14d7b1f884c9e11c0f
36517 2022-06-24 GDB Administrator <gdbadmin@sourceware.org>
36519 Automatic date update in version.in
36521 2022-06-23 Tom Tromey <tromey@adacore.com>
36523 Don't declare cli_set_logging
36524 cli_set_logging is declared but not defined. It's probably a leftover
36525 from whenever interpreters were changed to use inheritance. This
36526 patch removes the declaration. Tested by grep and rebuilding.
36528 Use PyBool_FromLong
36529 I noticed a few spots that were explicitly creating new references to
36530 Py_True or Py_False. It's simpler here to use PyBool_FromLong, so
36531 this patch changes all the places I found.
36533 2022-06-23 Alan Modra <amodra@gmail.com>
36535 PowerPC64: fix assertion in ppc_build_one_stub with -Os code
36536 save_res stubs aren't written in ppc_build_one_stub, their offsets
36537 (which are zero) should not be checked.
36539 * elf64-ppc.c (ppc_build_one_stub): Don't check save_res offsets.
36541 2022-06-23 Alan Modra <amodra@gmail.com>
36543 Re: PowerPC64: stub debug dump
36544 Let's show the current stub as well as the previous one. Of interest
36545 is the current offset and a new field, id. Check that the build
36546 hash table traversal is in the same order as sizing traversal too.
36548 * elf64-ppc.c (struct ppc_stub_hash_entry): Add id.
36549 (struct ppc_link_hash_table): Add stub_id.
36550 (stub_hash_newfunc): Init id and symtype.
36551 (dump_stub): New function, extracted from..
36552 (dump_previous_stub): ..here. Deleted.
36553 (ppc_build_one_stub): Sanity check stub id as well as offset.
36554 Show current stub as well as previous.
36555 (ppc_size_one_stub): Set stub id.
36556 (ppc64_elf_size_stubs): Init stub_id before traversal.
36557 (ppc64_elf_build_stubs): Likewise.
36559 2022-06-23 Fangrui Song <maskray@google.com>
36561 aarch64: Allow PC-relative relocations against protected STT_FUNC for -shared
36562 __attribute__((visibility("protected"))) void *foo() {
36563 return (void *)foo;
36566 gcc -fpic -shared -fuse-ld=bfd fails with the confusing diagnostic:
36568 relocation R_AARCH64_ADR_PREL_PG_HI21 against symbol `foo' which may bind externally can not be used when making a shared object; recompile with -fPIC
36570 Call _bfd_elf_symbol_refs_local_p with local_protected==true to suppress
36571 the error. The new behavior matches gold and ld.lld.
36573 Note: if some code tries to use direct access relocations to take the
36574 address of foo (likely due to -fno-pic), the pointer equality will
36575 break, but the error should be reported on the executable link, not on
36576 the innocent shared object link. glibc 2.36 will give a warning at
36577 relocation resolving time.
36579 2022-06-23 Fangrui Song <maskray@google.com>
36581 aarch64: Define elf_backend_extern_protected_data to 0 [PR 18705]
36582 Follow-up to commit 90b7a5df152a64d2bea20beb438e8b81049a5c30
36583 ("aarch64: Disallow copy relocations on protected data").
36585 Commit 32f573bcb3aaa1c9defcad79dbb5851fcc02ae2d changed ld to produce
36586 R_AARCH64_GLOB_DAT but that defeated the purpose of protected visibility
36587 as an optimization. Restore the previous behavior (which matches
36588 ld.lld) by defining elf_backend_extern_protected_data to 0.
36590 2022-06-23 GDB Administrator <gdbadmin@sourceware.org>
36592 Automatic date update in version.in
36594 2022-06-22 Tom Tromey <tromey@adacore.com>
36596 Use std::string for interpreter_p
36597 The global interpreter_p is a manually-managed 'char *'. This patch
36598 changes it to be a std::string instead, and removes some erroneous
36601 Move mi_interpreter to mi-interp.h
36602 I noticed that touching interps.h caused a lot of recompilation. I
36603 tracked this down to mi-common.h including this file. This patch
36604 moves the MI interpreter to mi-interp.h, which cuts down on
36605 recompilation when modifying interps.h.
36607 Use unique_xmalloc_ptr in interp
36608 This changes interp::m_name to be a unique_xmalloc_ptr, removing some
36609 manual memory management. It also cleans up the initialization of the
36610 'inited' member, and moves the 'private:' and 'public:' keywords to
36611 their proper spots.
36613 2022-06-22 Fangrui Song <i@maskray.me>
36615 aarch64: Disallow copy relocations on protected data
36616 If an executable has copy relocations for extern protected data, that
36617 can only work if the shared object containing the definition is built
36618 with assumptions (a) the compiler emits GOT-generating relocations (b)
36619 the linker produces R_*_GLOB_DAT instead of R_*_RELATIVE. Otherwise the
36620 shared object uses its own definition directly and the executable
36621 accesses a stale copy. Note: the GOT relocations defeat the purpose of
36622 protected visibility as an optimization, and it turns out this never
36625 glibc 2.36 will warn on copy relocations on protected data. Let's
36626 produce a warning at link time, matching ld.lld which has been used on
36629 Note: x86 requires GNU_PROPERTY_NO_COPY_ON_PROTECTED to have the error.
36630 This is to largely due to GCC 5's "x86-64: Optimize access to globals in
36631 PIE with copy reloc" which started to use direct access relocations for
36632 external data symbols in -fpie mode.
36634 GCC's aarch64 port does not have the change. Nowadays with most builds
36635 switching to -fpie/-fpic, aarch64 mostly doesn't need to worry about
36636 copy relocations. So for aarch64 we simply don't check
36637 GNU_PROPERTY_NO_COPY_ON_PROTECTED.
36639 2022-06-22 Kumar N, Bhuvanendra <Kavitha.Natarajan@amd.com>
36641 Binutils support for split-dwarf and dwarf-5
36642 * dwarf.c (fetch_indexed_string): Added new parameter
36643 str_offsets_base to calculate the string offset.
36644 (read_and_display_attr_value): Read DW_AT_str_offsets_base
36646 (process_debug_info): While allocating memory and initializing
36647 debug_information, do it for do_debug_info also, if its true.
36648 (load_separate_debug_files): Load .debug_str_offsets if exists.
36649 * dwarf.h (struct debug_info): Add str_offsets_base field.
36651 2022-06-22 Nelson Chu <nelson.chu@sifive.com>
36653 RISC-V: Reorder the prefixed extensions which are out of order.
36654 This patch has been pending for almost a year... However, I noticed that
36655 llvm can already re-order the extensions, even if they are out of orders.
36656 Not really sure if they can also re-order the single letter extensions,
36657 but at least we can do this for the multi-letter extensions in binutils.
36660 * elfxx-riscv.c (riscv_parse_prefixed_ext): Removed the code which are
36661 used to check the prefixed extension orders.
36663 * testsuite/gas/riscv/march-fail-order-x-z.d: Removed since we will help
36664 tp reorder the prefixed extensions for now.
36665 * testsuite/gas/riscv/march-fail-order-x-z.l: Likewise.
36666 * testsuite/gas/riscv/march-fail-order-x.d: Likewise.
36667 * testsuite/gas/riscv/march-fail-order-x.l: Likewise.
36668 * testsuite/gas/riscv/march-fail-order-z.d: Likewise.
36669 * testsuite/gas/riscv/march-fail-order-z.l: Likewise.
36671 2022-06-22 Nelson Chu <nelson.chu@sifive.com>
36673 RISC-V: Use single h extension to control hypervisor CSRs and instructions.
36674 According to the picture 28.1 in the current ISA spec, h is no larger the
36675 multi-letter extension, it is a single extension after v. Therefore, this
36676 patch fix the implementation, and use the single h to control hypervisor
36677 CSRs and instructions, which we promised to do before.
36680 * elfxx-riscv.c (riscv_supported_std_ext): Added h with version 1.0 after v.
36681 (riscv_supported_std_h_ext): Removed.
36682 (riscv_all_supported_ext): Updated since riscv_supported_std_h_ext is removed.
36683 (riscv_prefix_ext_class): Removed RV_ISA_CLASS_H.
36684 (parse_config): Updated since riscv_prefix_ext_class is removed.
36685 (riscv_recognized_prefixed_ext): Likewise.
36686 (riscv_get_default_ext_version): Likewise.
36687 (riscv_multi_subset_supports): Handle INSN_CLASS_H for hypervisor instructions.
36688 (riscv_multi_subset_supports_ext): Likewise.
36690 * config/tc-riscv.c (riscv_csr_class): Added CSR_CLASS_H and CSR_CLASS_H_32 for
36692 (riscv_csr_address): Likewise.
36693 * testsuite/gas/riscv/csr-version-1p10.d: Updated since hypervisor CSRs are
36694 controlled by single h extension for now.
36695 * testsuite/gas/riscv/csr-version-1p10.l: Likewise.
36696 * testsuite/gas/riscv/csr-version-1p11.d: Likewise.
36697 * testsuite/gas/riscv/csr-version-1p11.l: Likewise.
36698 * testsuite/gas/riscv/csr-version-1p12.d: Likewise.
36699 * testsuite/gas/riscv/csr-version-1p12.l: Likewise.
36700 * testsuite/gas/riscv/csr-version-1p9p1.d: Likewise.
36701 * testsuite/gas/riscv/csr-version-1p9p1.l: Likewise.
36702 * testsuite/gas/riscv/h-ext-32.d: Added h to architecture string.
36703 * testsuite/gas/riscv/h-ext-64.d: Likewise.
36704 * testsuite/gas/riscv/march-fail-single-prefix-h: Removed since h is no
36705 longer multi-letter extension.
36706 * testsuite/gas/riscv/march-fail-unknown-h.d: Likewise.
36708 * opcode/riscv-opc.h: Control hypervisor CSRs by h extension, rather than
36709 the privileged spec verisons.
36710 * opcode/riscv.h (riscv_insn_class): Added INSN_CLASS_H.
36712 * riscv-opc.c (riscv_opcodes): Control hypervisor instructions by h extension.
36714 2022-06-22 Tsukasa OI <research_trasio@irq.a4lg.com>
36716 RISC-V: Add 'H' to canonical extension ordering
36717 This commit adds 'H' to canonical extension ordering based on current
36718 consensus (not officially ratified as a new ISA specification manual
36719 but discussion for software compatibility is made).
36723 * elfxx-riscv.c (riscv_ext_canonical_order): Add 'H' for
36724 canonical extension ordering based on current consensus.
36726 2022-06-22 Tsukasa OI <research_trasio@irq.a4lg.com>
36728 RISC-V: Prepare i18n for required ISA extensions
36729 Some strings returned by the riscv_multi_subset_supports_ext function
36730 contain not just extension names but words like "and" and "or".
36731 This commit wraps such strings with the gettext macro (_) for
36732 internationalization in the future.
36736 * elfxx-riscv.c (riscv_multi_subset_supports_ext): Wrap some
36737 strings with the gettext macro to prepare future i18n.
36739 2022-06-22 Tsukasa OI <research_trasio@irq.a4lg.com>
36741 RISC-V: Fix inconsistent error message (range)
36742 This commit fixes inconsistent error message format involving compressed
36743 funct<n> fields. In specific, funct6 had an error message with range
36744 0..2^<n> ("0..64") unlike other funct<n> fields with 0..2^<n>-1
36745 (e.g. funct4 with "0..15").
36749 * config/tc-riscv.c (riscv_ip): Fix inconsistent error message.
36751 2022-06-22 Marcus Nilsson <brainbomb@gmail.com>
36753 readelf: replace xmalloc with malloc in slurp_relr_relocs
36754 Using xmalloc makes the null check redundant since failing allocation
36755 will exit the program. Instead use malloc and let the error be
36756 conveyed up the call chain.
36758 2022-06-22 Alan Modra <amodra@gmail.com>
36760 PowerPC64: stub debug dump
36761 powerpc64le-linux-ld is failing the assertion in ppc_build_one_stub,
36762 again apparently, which means a stub will overwrite the tail of a
36763 previous stub. The difficulty with debugging these issues is that
36764 it's not a problem with the stub that triggers the assertion, but the
36765 previous stub in that section. This patch keeps track of the last
36766 stub and adds a debug dump. Hopefully that will help.
36768 * elf64-ppc.c (enum _ppc64_sec_type): Add sec_stub.
36769 (struct _ppc64_elf_section_data): Add u.last_ent.
36770 (dump_previous_stub): New function.
36771 (ppc_build_one_stub): Keep track of previous stub, and dump it
36772 when finding an overlapping stub.
36774 2022-06-22 Alan Modra <amodra@gmail.com>
36776 PR29270, DW_FORM_udata signed output
36778 * dwarf.c (read_and_display_attr_value): Output DW_FORM_udata
36781 2022-06-22 GDB Administrator <gdbadmin@sourceware.org>
36783 Automatic date update in version.in
36785 2022-06-21 Nick Alcock <nick.alcock@oracle.com>
36787 ld: regenerate configure after recent misgeneration
36788 Things work again after this.
36792 * configure: Regenerate.
36794 2022-06-21 Nick Alcock <nick.alcock@oracle.com>
36796 libctf: tests: prune warnings from compiler output
36797 We were failing to call prune_warnings appropriately, leading to
36798 false-positive test failures on some platforms (observed on
36803 * testsuite/lib/ctf-lib.exp: Prune warnings from compiler and
36805 * testsuite/libctf-regression/libctf-repeat-cu.exp: Likewise,
36808 2022-06-21 Nick Alcock <nick.alcock@oracle.com>
36810 libctf: avoid mingw warning
36811 A missing paren led to an intended cast to avoid dependence on the size
36812 of size_t in one argument of ctf_err_warn applying to the wrong type by
36817 * ctf-serialize.c (ctf_write_mem): Fix cast.
36819 2022-06-21 Nick Alcock <nick.alcock@oracle.com>
36821 libctf: fix linking together multiple objects derived from the same source
36822 Right now, if you compile the same .c input repeatedly with CTF enabled
36823 and different compilation flags, then arrange to link all of these
36824 together, then things misbehave in various ways. libctf may conflate
36825 either inputs (if the .o files have the same name, say if they are
36826 stored in different .a archives), or per-CU outputs when conflicting
36827 types are found: the latter can lead to entirely spurious errors when
36828 it tries to produce multiple per-CU outputs with the same name
36829 (discarding all but the last, but then looking for types in the earlier
36830 ones which have just been thrown away).
36832 Fixing this is multi-pronged. Both inputs and outputs need to be
36833 differentiated in the hashtables libctf keeps them in: inputs with the
36834 same cuname and filename need to be considered distinct as long as they
36835 have different associated CTF dicts, and per-CU outputs need to be
36836 considered distinct as long as they have different associated input
36837 dicts. Right now there is nothing tying the two together other than the
36838 CU name: fix this by introducing a new field in the ctf_dict_t named
36839 ctf_link_in_out, which (for input dicts) points to the associated per-CU
36840 output dict (if any), and for output dicts points to the associated
36841 input dict. At creation time the name used is completely arbitrary:
36842 it's only important that it be distinct if CTF dicts are distinct. So,
36843 when a clash is found, adjust the CU name by sticking the number of
36844 elements in the input on the end. At output time, the CU name will
36845 appear in the linked object, so it matters a little more that it look
36846 slightly less ugly: in conflicting cases, append an incrementing
36847 integer, starting at 0.
36849 This naming scheme is not very helpful, but it's hard to see what else
36850 we can do. The input .o name may be the same. The input .a name is not
36851 even visible to ctf_link, and even *that* might be the same, because
36852 .a's can contain many members with the same name, all of which
36853 participate in the link. All we really know is that the two have
36854 distinct dictionaries with distinct types in them, and at least this way
36855 they are all represented, any any symbols, variables etc referring to
36856 those types are accurately stored.
36858 (As a side-effect this also fixes a use-after-free and double-free when
36859 errors are found during variable or symbol emission.)
36861 Use the opportunity to prevent a couple of sources of problems, to wit
36862 changing the active CU mappings when a link has already been done
36863 (no effect on ld, which doesn't use CU mappings at all), and causing
36864 multiple consecutive ctf_link's to have the same net effect as just
36865 doing the last one (no effect on ld, which only ever does one
36866 ctf_link) rather than having the links be a sort of half-incremental
36867 not-really-intended mess.
36872 * ctf-impl.h (struct ctf_dict) [ctf_link_in_out]: New.
36873 * ctf-dedup.c (ctf_dedup_emit_type): Set it.
36874 * ctf-link.c (ctf_link_add_ctf_internal): Set the input
36875 CU name uniquely when clashes are found.
36876 (ctf_link_add): Document what repeated additions do.
36877 (ctf_new_per_cu_name): New, come up with a consistent
36878 name for a new per-CU dict.
36879 (ctf_link_deduplicating): Use it.
36880 (ctf_create_per_cu): Use it, and ctf_link_in_out, and set
36881 ctf_link_in_out properly. Don't overwrite per-CU dicts with
36882 per-CU dicts relating to different inputs.
36883 (ctf_link_add_cu_mapping): Prevent per-CU mappings being set up
36884 if we already have per-CU outputs.
36885 (ctf_link_one_variable): Adjust ctf_link_per_cu call.
36886 (ctf_link_deduplicating_one_symtypetab): Likewise.
36887 (ctf_link_empty_outputs): New, delete all the ctf_link_outputs
36888 and blank out ctf_link_in_out on the corresponding inputs.
36889 (ctf_link): Clarify the effect of multiple ctf_link calls.
36890 Empty ctf_link_outputs if it already exists rather than
36891 having the old output leak into the new link. Fix a variable
36893 * testsuite/config/default.exp (AR): Add.
36894 (OBJDUMP): Likewise.
36895 * testsuite/libctf-regression/libctf-repeat-cu.exp: New test.
36896 * testsuite/libctf-regression/libctf-repeat-cu*: Main program,
36897 library, and expected results for the test.
36899 2022-06-21 Kevin Buettner <kevinb@redhat.com>
36901 Document how GDB searches for files when using -s, -e, and -se options
36902 GDB's documentation of the 'file' command says:
36904 If you do not specify a directory and the file is not found in the
36905 GDB working directory, GDB uses the environment variable PATH as a
36906 list of directories to search, just as the shell does when looking
36907 for a program to run.
36909 The same is true for files specified via commandline options -s, -e,
36912 This commit adds a cross reference to the file command for these options.
36914 2022-06-21 Nick Clifton <nickc@redhat.com>
36916 Binutils support for dwarf-5 (location and range lists related)
36917 * dwarf.h (struct debug_info): Add rnglists_base field.
36918 * dwarf.c (read_and_display_attr_value): Read attribute DW_AT_rnglists_base.
36919 (display_debug_rnglists_list): While handling DW_RLE_base_addressx,
36920 DW_RLE_startx_endx, DW_RLE_startx_length items, pass the proper parameter
36921 value to fetch_indexed_addr(), i.e. fetch the proper entry in .debug_addr section.
36922 (display_debug_ranges): Add rnglists_base to the .debug_rnglists base address.
36923 (load_separate_debug_files): Load .debug_addr section, if exists.
36925 Default to disabling the linker warnings about execstack and RWX segments if the target is the HPPA architecture.
36927 * configure.ac (ac_default_ld_warn_execstack): Default to 'no' for
36929 (ac_default_ld_warn_rwx_segments): Likewise.
36930 * configure: Regenerate.
36931 * testsuite/ld-elf/elf.exp: Add the --warn-execstack command line
36932 option to the command line when running execstack tests for the
36935 2022-06-21 GDB Administrator <gdbadmin@sourceware.org>
36937 Automatic date update in version.in
36939 2022-06-20 Tom Tromey <tromey@adacore.com>
36941 Move finish_print out of value_print_options
36942 'finish_print' does not really belong in value_print_options -- this
36943 is consulted only when deciding whether or not to print a value, and
36944 never during the course of printing. This patch removes it from the
36945 structure and makes it a static global in infcmd.c instead.
36947 Tested on x86-64 Fedora 34.
36949 2022-06-20 Alan Modra <amodra@gmail.com>
36951 PR29262, memory leak in pr_function_type
36953 * prdbg.c (pr_function_type): Free "s" on failure path.
36955 PR29261, memory leak in parse_stab_struct_fields
36957 * stabs.c (parse_stab_struct_fields): Free "fields" on failure path.
36959 2022-06-20 GDB Administrator <gdbadmin@sourceware.org>
36961 Automatic date update in version.in
36963 2022-06-19 GDB Administrator <gdbadmin@sourceware.org>
36965 Automatic date update in version.in
36967 2022-06-18 Tom Tromey <tom@tromey.com>
36969 Fix assertion failure in copy_type
36970 PR exp/20630 points out a simple way to cause an assertion failure in
36971 copy_type -- but this was found in the wild a few times as well.
36973 copy_type only works for objfile-owned types, but there isn't a deep
36974 reason for this. This patch fixes the bug by updating copy_type to
36975 work for any sort of type.
36977 Better would perhaps be to finally implement type GC, but I still
36978 haven't attempted this.
36980 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=20630
36982 2022-06-18 Tomoaki Kawada <kawada@kmckk.co.jp>
36984 Fix the sorting algorithm for reloc entries
36985 The optimized insertion sort algorithm in `elf_link_adjust_relocs`
36986 incorrectly assembled "runs" from unsorted entries and inserted them to an
36987 already-sorted prefix, breaking the loop invariants of insertion sort.
36988 This commit updates the run assembly loop to break upon encountering a
36989 non-monotonic change in the sort key.
36993 * elflink.c (elf_link_adjust_relocs): Ensure run being inserted
36996 * testsuite/ld-elf/pr29259.d,
36997 * testsuite/ld-elf/pr29259.s,
36998 * testsuite/ld-elf/pr29259.t: New test.
37000 2022-06-18 Enze Li <enze.li@hotmail.com>
37002 gdb/python: Export nibbles to python layer
37003 This patch makes it possible to allow Value.format_string() to return
37006 When we set the parameter of nibbles to True, we can achieve the
37007 displaying binary values in groups of every four bits.
37010 (gdb) py print (gdb.Value (1230).format_string (format='t', nibbles=True))
37014 Note that the parameter nibbles is only useful if format='t' is also used.
37016 This patch also includes update to the relevant testcase and
37019 Tested on x86_64 openSUSE Tumbleweed.
37021 2022-06-18 Enze Li <enze.li@hotmail.com>
37023 gdb/doc: Documentation for the new print command
37024 Document the new command "print nibbles" and add a NEWS entry.
37026 2022-06-18 Enze Li <enze.li@hotmail.com>
37028 gdb: Add new 'print nibbles' feature
37029 Make an introduction of a new print setting that can be set by 'set
37030 print nibbles [on|off]'. The default value if OFF, which can be changed
37031 by user manually. Of course, 'show print nibbles' is also included in
37034 The new feature displays binary values by group, with four bits per
37035 group. The motivation for this work is to enhance the readability of
37038 Here's a GDB session before this patch is applied.
37041 (gdb) print/t var_a
37044 With this patch applied, we can use the new print setting to display the
37045 new form of the binary values.
37048 (gdb) print/t var_a
37050 (gdb) set print nibbles on
37051 (gdb) print/t var_a
37052 $3 = 0100 1100 1110
37054 Tested on x86_64 openSUSE Tumbleweed.
37056 2022-06-18 GDB Administrator <gdbadmin@sourceware.org>
37058 Automatic date update in version.in
37060 2022-06-17 Tiezhu Yang <yangtiezhu@loongson.cn>
37062 gdb: NEWS: Move LoongArch gdbserver to the correct section
37063 commit e5ab6af52d38 ("gdbserver: Add LoongArch/Linux support")
37064 was merged into the master since GDB 12, so we should put the
37065 news in the "Changes since GDB 12" section.
37067 Thanks Tom Tromey for your correction [1], sorry for that.
37069 [1] https://sourceware.org/pipermail/gdb-patches/2022-June/190122.html
37071 2022-06-17 Alan Modra <amodra@gmail.com>
37073 PR29256, memory leak in obj_elf_section_name
37074 When handling section names in quotes obj_elf_section_name calls
37075 demand_copy_C_string, which puts the name on the gas notes obstack.
37076 Such strings aren't usually freed, since obstack_free frees all more
37077 recently allocated objects as well as its arg. When handling
37078 non-quoted names, obj_elf_section_name mallocs the name. Due to the
37079 mix of allocation strategies it isn't possible for callers to free
37080 names, if that was desirable. Partially fix this by always creating
37081 names on the obstack, which is more efficient anyway. (You still
37082 can't obstack_free on error paths due to the xtensa
37083 tc_canonicalize_section_name.) Also remove a couple of cases where
37084 the name is dup'd for no good reason as far as I know.
37087 * config/obj-elf.c (obj_elf_section_name): Create name on notes
37089 (obj_elf_attach_to_group): Don't strdup group name.
37090 (obj_elf_section): Likewise.
37091 (obj_elf_vendor_attribute): Use xmemdup0 rather than xstrndup.
37093 2022-06-17 Alan Modra <amodra@gmail.com>
37095 PR29255, memory leak in make_tempdir
37097 * bucomm.c (make_tempdir, make_tempname): Free template on all
37100 PR29254, memory leak in stab_demangle_v3_arg
37102 * stabs.c (stab_demangle_v3_arg): Free dt on failure path.
37104 2022-06-17 Pedro Alves <pedro@palves.net>
37106 Fix GDB build with GCC 4.8 & 4.9
37107 With gcc 4.8/4.9, we run into this build failure (and other similar
37110 /home/palves/gdb/binutils-gdb/src/gdb/location.h:224:59: error: could not convert ‘{0, LINE_OFFSET_UNKNOWN}’ from ‘<brace-enclosed initializer list>’ to ‘line_offset’
37111 struct line_offset line_offset = {0, LINE_OFFSET_UNKNOWN};
37114 The issue is that at around the GCC 4.8/4.9 era, a default member
37115 initializer prevented the struct from being an aggregate, so you
37116 cannot use aggregate initialization on them. That rule changed after
37117 GCC 4.9 and GCC 5 & later uses new rules.
37119 Fix this by not using aggregate initialization for struct line_offset.
37120 The default member initization already leaves line_offset as {0,
37121 LINE_OFFSET_UNKNOWN}, so initialization to those values can just go
37122 away. The remaining cases are of the form {0, LINE_OFFSET_NONE}, and
37123 those cases can be better rewritten to delay setting the sign field
37124 until we know we have a valid offset.
37126 Change-Id: I0506ea4a83c5fa2f15e159569db68b3b0a7509b4
37128 2022-06-17 Pedro Alves <pedro@palves.net>
37130 Convert set_location_spec_string to a method
37131 This converts set_location_spec_string to a method of location_spec,
37132 and makes the location_spec::as_string field protected, renaming it to
37133 m_as_string along the way.
37135 Change-Id: Iccfb1654e9fa7808d0512df89e775f9eacaeb9e0
37137 2022-06-17 Pedro Alves <pedro@palves.net>
37139 Convert location_spec_to_string to a method
37140 This converts location_spec_to_string to a method of location_spec,
37141 simplifying the code using it, as it no longer has to use
37142 std::unique_ptr::get().
37144 Change-Id: I621bdad8ea084470a2724163f614578caf8f2dd5
37146 2022-06-17 Pedro Alves <pedro@palves.net>
37148 Convert location_spec_type to a method
37149 This converts location_spec_type to location_spec::type().
37151 Change-Id: Iff4cbfafb1cf3d22adfa142ff939b4a148e52273
37153 2022-06-17 Pedro Alves <pedro@palves.net>
37155 Convert location_spec_empty_p to a method
37156 This converts location_spec_empty_p to a method of location_spec,
37157 simplifying users, as they no longer have to use
37158 std::unique_ptr::get().
37160 Change-Id: I83381a729896f12e1c5a1b4d6d4c2eb1eb6582ff
37162 2022-06-17 Pedro Alves <pedro@palves.net>
37164 Eliminate copy_location_spec
37165 copy_location_spec is just a wrapper around location_spec::clone(), so
37166 remove it and call clone() directly. This simplifies users, as they
37167 no longer have to use std::unique_ptr::get().
37169 Change-Id: I8ce8658589460b98888283b306b315a5b8f73976
37171 2022-06-17 Pedro Alves <pedro@palves.net>
37173 Eliminate the two-level data structures behind location_specs
37174 Currently, there's the location_spec hierarchy, and then some
37175 location_spec subclasses have their own struct type holding all their
37178 I.e., there is this:
37181 explicit_location_spec
37182 linespec_location_spec
37183 address_location_spec
37184 probe_location_spec
37186 and then these separate types:
37193 explicit_location_spec
37194 has-a explicit_location
37195 linespec_location_spec
37196 has-a linespec_location
37198 This patch eliminates explicit_location and linespec_location,
37199 inlining their members in the corresponding location_spec type.
37201 The location_spec subclasses were the ones currently defined in
37202 location.c, so they are moved to the header. Since the definitions of
37203 the classes are now visible, we no longer need location_spec_deleter.
37205 Some constructors that are used for cloning location_specs, like:
37207 explicit explicit_location_spec (const struct explicit_location *loc)
37209 ... were converted to proper copy ctors.
37211 In the process, initialize_explicit_location is eliminated, and some
37212 functions that returned the "data type behind a locspec", like
37213 get_linespec_location are converted to downcast functions, like
37214 as_linespec_location_spec.
37216 Change-Id: Ia31ccef9382b25a52b00fa878c8df9b8cf2a6c5a
37218 2022-06-17 Pedro Alves <pedro@palves.net>
37220 event_location -> location_spec
37221 Currently, GDB internally uses the term "location" for both the
37222 location specification the user input (linespec, explicit location, or
37223 an address location), and for actual resolved locations, like the
37224 breakpoint locations, or the result of decoding a location spec to
37225 SaLs. This is expecially confusing in the breakpoints module, as
37226 struct breakpoint has these two fields:
37228 breakpoint::location;
37231 "location" is the location spec, and "loc" is the resolved locations.
37233 And then, we have a method called "locations()", which returns the
37234 resolved locations as range...
37236 The location spec type is presently called event_location:
37238 /* Location we used to set the breakpoint. */
37239 event_location_up location;
37241 and it is described like this:
37243 /* The base class for all an event locations used to set a stop event
37244 in the inferior. */
37246 struct event_location
37249 and even that is incorrect... Location specs are used for finding
37250 actual locations in the program in scenarios that have nothing to do
37251 with stop events. E.g., "list" works with location specs.
37253 To clean all this confusion up, this patch renames "event_location" to
37254 "location_spec" throughout, and then all the variables that hold a
37255 location spec, they are renamed to include "spec" in their name, like
37256 e.g., "location" -> "locspec". Similarly, functions that work with
37257 location specs, and currently have just "location" in their name are
37258 renamed to include "spec" in their name too.
37260 Change-Id: I5814124798aa2b2003e79496e78f95c74e5eddca
37262 2022-06-17 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
37264 gprofng: fix build with -Werror=format-truncation
37266 2022-06-16 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
37268 * configure.ac: Remove -Wno-format-truncation.
37269 * src/Makefile.am: Likewise.
37270 * configure: Rebuild.
37271 * src/Makefile.in: Rebuild.
37272 * common/hwctable.c: Fix -Werror=format-truncation errors.
37273 * src/ipc.cc: Likewise.
37274 * src/parse.cc: Likewise.
37276 2022-06-17 GDB Administrator <gdbadmin@sourceware.org>
37278 Automatic date update in version.in
37280 2022-06-16 Tom de Vries <tdevries@suse.de>
37282 [gdb/testsuite] Fix have_mpx test
37283 When testing on openSUSE Leap 15.4 I ran into this FAIL:
37285 FAIL: gdb.arch/i386-mpx-map.exp: NULL address of the pointer
37287 and likewise for all the other mpx tests.
37289 The problem is that have_mpx is supposed to return 0, but it doesn't because
37290 it tries to match this output:
37292 builtin_spawn -ignore SIGHUP temp/20294/have_mpx-2-20294.x^M
37298 && ![string equal $output "No MPX support\r\n"]]
37301 Fix this by matching using a regexp instead.
37303 Tested on x86_64-linux.
37305 2022-06-16 Alan Modra <amodra@gmail.com>
37307 use of uninitialised value in input_file_open
37308 Triggered by a file containing just "#N" or "#A". fgets when hitting
37309 EOF before reading anything returns NULL and does not write to buf.
37310 strchr (buf, '\n') then is reading from uninitialised memory.
37312 * input-file.c (input_file_open): Don't assume buf contains
37313 zero string terminator when fgets returns NULL.
37315 2022-06-16 Alan Modra <amodra@gmail.com>
37317 Always free matching vector from bfd_check_format_matches
37318 At least one place calling list_matching_formats failed to free the
37319 "matching" vector from bfd_check_format_matches afterwards. Fix that
37320 by calling free inside list_matching_formats.
37323 * bucomm.c (list_matching_formats): Free arg.
37324 * addr2line.c (process_file): Adjust to suit.
37325 * ar.c (open_inarch, ranlib_touch): Likewise.
37326 * coffdump.c (main): Likewise.
37327 * nm.c (display_archive, display_file): Likewise.
37328 * objcopy.c (copy_file): Likewise.
37329 * objdump.c (display_object_bfd): Likewise.
37330 * size.c (display_bfd): Likewise.
37331 * srconv.c (main): Likewise.
37333 * ldlang.c (load_symbols): Free "matching".
37335 2022-06-16 Alan Modra <amodra@gmail.com>
37337 Revert "Revert "Fix fbsd core matching""
37338 This reverts commit 476288fa2bddecf0f0e13dee826a076309bf01fe.
37340 2022-06-16 Alan Modra <amodra@gmail.com>
37342 Restore readelf -wF
37343 Commit 94585d6d4495 resulted in readelf -wF failing with
37344 Unrecognized debug letter option 'F'
37347 * dwarf.c (debug_dump_long_opts): Add letter.
37348 (debug_option_table): New, replacing..
37349 (opts_table, letter_table): ..these.
37350 (dwarf_select_sections_by_names): Adjust to suit. Set
37351 do_debug_frames outside of loop.
37352 (dwarf_select_sections_by_letters): Similarly.
37354 * testsuite/gas/i386/ehinterp.d: Use readelf -wF.
37356 2022-06-16 Alan Modra <amodra@gmail.com>
37358 PR29250, readelf erases CIE initial register state
37361 * dwarf.c (display_debug_frames): Set col_type[reg] on sizing
37362 pass over FDE to cie->col_type[reg] if CIE specifies reg.
37363 Handle DW_CFA_restore and DW_CFA_restore_extended on second
37364 pass using the same logic. Remove unnecessary casts. Don't
37365 call frame_need_space on second pass over FDE.
37367 * testsuite/gas/i386/ehinterp.d,
37368 * testsuite/gas/i386/ehinterp.s: New test.
37369 * testsuite/gas/i386/i386.exp: Run it.
37371 2022-06-16 GDB Administrator <gdbadmin@sourceware.org>
37373 Automatic date update in version.in
37375 2022-06-15 Sergei Trofimovich <siarheit@google.com>
37377 sim: fix BFD_VMA format arguments on 32-bit hosts [PR gdb/29184]
37378 Noticed format mismatch when attempted to build gdb on i686-linux-gnu
37379 in --enable-64-bit-bfd mode:
37381 sim/../../sim/cris/sim-if.c:576:28:
37382 error: format '%lx' expects argument of type 'long unsigned int',
37383 but argument 4 has type 'bfd_size_type' {aka 'long long unsigned int'} [-Werror=format=]
37384 576 | sim_do_commandf (sd, "memory region 0x%" BFD_VMA_FMT "x,0x%lx",
37385 | ^~~~~~~~~~~~~~~~~~~
37386 577 | interp_load_addr, interpsiz);
37389 | bfd_size_type {aka long long unsigned int}
37391 While at it fixed format string for time-related types.
37393 2022-06-15 Tom Tromey <tromey@adacore.com>
37395 Check for listeners in emit_exiting_event
37396 I noticed that emit_exiting_event does not check whether there are any
37397 listeners before creating the event object. All other event emitters
37398 do this, so this patch updates this one as well.
37400 2022-06-15 Tom Tromey <tom@tromey.com>
37402 Add to documentation of Python 'dont_repeat' method
37403 PR python/28533 points out that the Python 'dont_repeat' documentation
37404 is a bit ambiguous about when the method ought to be called. This
37405 patch spells it out.
37407 2022-06-15 Yvan Roux <yvan.roux@foss.st.com>
37409 gdb/arm: Make sp alias for one of the other stack pointers
37410 For Cortex-M targets, SP register is never detached from msp or
37411 psp, it always has the same value as one of them. Let GDB treat
37412 ARM_SP_REGNUM as an alias similar to what is done in hardware.
37414 2022-06-15 Yvan Roux <yvan.roux@foss.st.com>
37416 gdb/arm: Track msp and psp
37417 For Arm Cortex-M33 with security extensions, there are 4 different
37418 stack pointers (msp_s, msp_ns, psp_s, psp_ns). To be compatible
37419 with earlier Cortex-M derivates, the msp and psp registers are
37420 aliases for one of the 4 real stack pointer registers.
37422 These are the combinations that exist:
37424 sp -> msp -> msp_ns
37426 sp -> psp -> psp_ns
37428 This means that when the GDB client is to show the value of "msp",
37429 the value should always be equal to either "msp_s" or "msp_ns".
37430 Same goes for "psp".
37432 To add a bit more context; GDB does not really use the register msp
37433 (or psp) internally, but they are part of the set of registers which
37434 are provided by the target.xml file. As a result, they will be part
37435 of the set of registers printed by the "info r" command.
37437 Without this particular patch, GDB will hit the assert in the bottom
37438 of arm_cache_get_sp_register function.
37440 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29121
37442 2022-06-15 Yvan Roux <yvan.roux@foss.st.com>
37444 gdb/arm: Fetch initial sp value prior to compare
37445 For Arm Cortex-M33 with security extensions, there are 4 different
37446 stack pointers (msp_s, msp_ns, psp_s, psp_ns). In order to
37447 identify the active one, compare the values of the different
37448 stacks. The value of the initial sp register needs to be fetched to
37449 perform this comparison.
37451 2022-06-15 Andrew Burgess <aburgess@redhat.com>
37453 gdb: unify two dis_asm_read_memory functions in disasm.c
37454 After the recent restructuring of the disassembler code, GDB has ended
37455 up with two identical class static functions, both called
37456 dis_asm_read_memory, with identical implementations.
37458 My first thought was to move these out of their respective classes,
37459 and just make them global functions, then I'd only need a single
37462 And maybe that's the right way to go. But I disliked that by doing
37463 that I loose the encapsulation of the method with the corresponding
37464 disassembler class.
37466 So, instead, I placed the static method into its own class, and had
37467 both the gdb_non_printing_memory_disassembler and gdb_disassembler
37468 classes inherit from this new class as an additional base-class.
37470 In terms of code generated, I don't think there's any significant
37471 difference with this approach, but I think this better reflects how
37472 the function is closely tied to the disassembler.
37474 There should be no user visible changes after this commit.
37476 2022-06-15 Andrew Burgess <aburgess@redhat.com>
37478 gdb: refactor the non-printing disassemblers
37479 This commit started from an observation I made while working on some
37480 other disassembler patches, that is, that the function
37481 gdb_buffered_insn_length, is broken ... sort of.
37483 I noticed that the gdb_buffered_insn_length function doesn't set up
37484 the application data field if the disassemble_info structure.
37486 Further, I noticed that some architectures, for example, ARM, require
37487 that the application_data field be set, see gdb_print_insn_arm in
37490 And so, if we ever use gdb_buffered_insn_length for ARM, then GDB will
37491 likely crash. Which is why I said only "sort of" broken. Right now
37492 we don't use gdb_buffered_insn_length with ARM, so maybe it isn't
37495 Anyway to prove to myself that there was a problem here I extended the
37496 disassembler self tests in disasm-selftests.c to include a test of
37497 gdb_buffered_insn_length. As I run the test for all architectures, I
37498 do indeed see GDB crash for ARM.
37500 To fix this we need gdb_buffered_insn_length to create a disassembler
37501 that inherits from gdb_disassemble_info, but we also need this new
37502 disassembler to not print anything.
37504 And so, I introduce a new gdb_non_printing_disassembler class, this is
37505 a disassembler that doesn't print anything to the output stream.
37507 I then observed that both ARC and S12Z also create non-printing
37508 disassemblers, but these are slightly different. While the
37509 disassembler in gdb_non_printing_disassembler reads the instruction
37510 from a buffer, the ARC and S12Z disassemblers read from target memory
37511 using target_read_code.
37513 And so, I further split gdb_non_printing_disassembler into two
37514 sub-classes, gdb_non_printing_memory_disassembler and
37515 gdb_non_printing_buffer_disassembler.
37517 The new selftests now pass, but otherwise, there should be no user
37518 visible changes after this commit.
37520 2022-06-15 Andrew Burgess <andrew.burgess@embecosm.com>
37522 gdb/python: implement the print_insn extension language hook
37523 This commit extends the Python API to include disassembler support.
37525 The motivation for this commit was to provide an API by which the user
37526 could write Python scripts that would augment the output of the
37529 To achieve this I have followed the model of the existing libopcodes
37530 disassembler, that is, instructions are disassembled one by one. This
37531 does restrict the type of things that it is possible to do from a
37532 Python script, i.e. all additional output has to fit on a single line,
37533 but this was all I needed, and creating something more complex would,
37534 I think, require greater changes to how GDB's internal disassembler
37537 The disassembler API is contained in the new gdb.disassembler module,
37538 which defines the following classes:
37542 Similar to libopcodes disassemble_info structure, has read-only
37543 properties: address, architecture, and progspace. And has methods:
37544 __init__, read_memory, and is_valid.
37546 Each time GDB wants an instruction disassembled, an instance of
37547 this class is passed to a user written disassembler function, by
37548 reading the properties, and calling the methods (and other support
37549 methods in the gdb.disassembler module) the user can perform and
37550 return the disassembly.
37554 This is a base-class which user written disassemblers should
37555 inherit from. This base class provides base implementations of
37556 __init__ and __call__ which the user written disassembler should
37561 This class can be used to hold the result of a call to the
37562 disassembler, it's really just a wrapper around a string (the text
37563 of the disassembled instruction) and a length (in bytes). The user
37564 can return an instance of this class from Disassembler.__call__ to
37565 represent the newly disassembled instruction.
37567 The gdb.disassembler module also provides the following functions:
37569 register_disassembler
37571 This function registers an instance of a Disassembler sub-class
37572 as a disassembler, either for one specific architecture, or, as a
37573 global disassembler for all architectures.
37575 builtin_disassemble
37577 This provides access to GDB's builtin disassembler. A common
37578 use case that I see is augmenting the existing disassembler output.
37579 The user code can call this function to have GDB disassemble the
37580 instruction in the normal way. The user gets back a
37581 DisassemblerResult object, which they can then read in order to
37582 augment the disassembler output in any way they wish.
37584 This function also provides a mechanism to intercept the
37585 disassemblers reads of memory, thus the user can adjust what GDB
37586 sees when it is disassembling.
37588 The included documentation provides a more detailed description of the
37591 There is also a new CLI command added:
37593 maint info python-disassemblers
37595 This command is defined in the Python gdb.disassemblers module, and
37596 can be used to list the currently registered Python disassemblers.
37598 2022-06-15 Andrew Burgess <andrew.burgess@embecosm.com>
37600 gdb: add extension language print_insn hook
37601 This commit is setup for the next commit.
37603 In the next commit I will add a Python API to intercept the print_insn
37604 calls within GDB, each print_insn call is responsible for
37605 disassembling, and printing one instruction. After the next commit it
37606 will be possible for a user to write Python code that either wraps
37607 around the existing disassembler, or even, in extreme situations,
37608 entirely replaces the existing disassembler.
37610 This commit does not add any new Python API.
37612 What this commit does is put the extension language framework in place
37613 for a print_insn hook. There's a new callback added to 'struct
37614 extension_language_ops', which is then filled in with nullptr for Python
37617 Finally, in the disassembler, the code is restructured so that the new
37618 extension language function ext_lang_print_insn is called before we
37619 delegate to gdbarch_print_insn.
37621 After this, the next commit can focus entirely on providing a Python
37622 implementation of the new print_insn callback.
37624 There should be no user visible change after this commit.
37626 2022-06-15 Andrew Burgess <andrew.burgess@embecosm.com>
37628 gdb: add new base class to gdb_disassembler
37629 The motivation for this change is an upcoming Python disassembler API
37630 that I would like to add. As part of that change I need to create a
37631 new disassembler like class that contains a disassemble_info and a
37632 gdbarch. The management of these two objects is identical to how we
37633 manage these objects within gdb_disassembler, so it might be tempting
37634 for my new class to inherit from gdb_disassembler.
37636 The problem however, is that gdb_disassembler has a tight connection
37637 between its constructor, and its print_insn method. In the
37638 constructor the ui_file* that is passed in is replaced with a member
37639 variable string_file*, and then in print_insn, the contents of the
37640 member variable string_file are printed to the original ui_file*.
37642 What this means is that the gdb_disassembler class has a tight
37643 coupling between its constructor and print_insn; the class just isn't
37644 intended to be used in a situation where print_insn is not going to be
37645 called, which is how my (upcoming) sub-class would need to operate.
37647 My solution then, is to separate out the management of the
37648 disassemble_info and gdbarch into a new gdb_disassemble_info class,
37649 and make this class a parent of gdb_disassembler.
37651 In arm-tdep.c and mips-tdep.c, where we used to cast the
37652 disassemble_info->application_data to a gdb_disassembler, we can now
37653 cast to a gdb_disassemble_info as we only need to access the gdbarch
37656 Now, my new Python disassembler sub-class will still want to print
37657 things to an output stream, and so we will want access to the
37658 dis_asm_fprintf functionality for printing.
37660 However, rather than move this printing code into the
37661 gdb_disassemble_info base class, I have added yet another level of
37662 hierarchy, a gdb_printing_disassembler, thus the class structure is
37665 struct gdb_disassemble_info {};
37666 struct gdb_printing_disassembler : public gdb_disassemble_info {};
37667 struct gdb_disassembler : public gdb_printing_disassembler {};
37669 In a later commit my new Python disassembler will inherit from
37670 gdb_printing_disassembler.
37672 The reason for adding the additional layer to the class hierarchy is
37673 that in yet another commit I intend to rewrite the function
37674 gdb_buffered_insn_length, and to do this I will be creating yet more
37675 disassembler like classes, however, these will not print anything,
37676 thus I will add a gdb_non_printing_disassembler class that also
37677 inherits from gdb_disassemble_info. Knowing that that change is
37678 coming, I've gone with the above class hierarchy now.
37680 There should be no user visible changes after this commit.
37682 2022-06-15 Andrew Burgess <aburgess@redhat.com>
37684 gdb/python: convert gdbpy_err_fetch to use gdbpy_ref
37685 Convert the gdbpy_err_fetch class to make use of gdbpy_ref, this
37686 removes the need for manual reference count management, and allows the
37687 destructor to be removed.
37689 There should be no functional change after this commit.
37691 I think this cleanup is worth doing on its own, however, in a later
37692 commit I will want to copy instances of gdbpy_err_fetch, and switching
37693 to using gdbpy_ref means that I can rely on the default copy
37694 constructor, without having to add one that handles the reference
37695 counts, so this is good preparation for that upcoming change.
37697 2022-06-15 Jan Beulich <jbeulich@suse.com>
37699 x86: drop print_operand_value()'s "hex" parameter
37700 For quite some time all callers have been passing 1 / true. While there
37701 fold the final oappend_with_style() calls.
37703 2022-06-15 Tom de Vries <tdevries@suse.de>
37705 [gdb/build] Fix build for gcc < 11
37706 When building trunk on openSUSE Leap 15.3 with system gcc 7.5.0, I run into:
37708 In file included from ../bfd/bfd.h:46:0,
37709 from gdb/defs.h:37,
37710 from gdb/debuginfod-support.c:19:
37711 gdb/debuginfod-support.c: In function ‘bool debuginfod_is_enabled()’:
37712 gdb/../include/diagnostics.h:42:3: error: unknown option after \
37713 ‘#pragma GCC diagnostic’ kind [-Werror=pragmas]
37714 _Pragma (DIAGNOSTIC_STRINGIFY (GCC diagnostic ignored option))
37716 gdb/../include/diagnostics.h:80:3: note: in expansion of macro \
37717 ‘DIAGNOSTIC_IGNORE’
37718 DIAGNOSTIC_IGNORE ("-Wstringop-overread")
37720 gdb/debuginfod-support.c:201:4: note: in expansion of macro \
37721 ‘DIAGNOSTIC_IGNORE_STRINGOP_OVERREAD’
37722 DIAGNOSTIC_IGNORE_STRINGOP_OVERREAD
37726 The problem is that the warning -Wstringop-overread has been introduced for
37727 gcc 11, and we can only tell gcc to ignore if it knows about it.
37729 Fix this by guarding the DIAGNOSTIC_IGNORE_STRINGOP_OVERREAD definition in
37730 diagnostics.c with '#if __GNUC__ >= 11'.
37732 Tested on x86_64-linux, by completing a build.
37734 2022-06-15 Alan Modra <amodra@gmail.com>
37736 PR29230, segv in lookup_symbol_in_variable_table
37737 The PR23230 testcase uses indexed strings without specifying
37738 SW_AT_str_offsets_base. In this case we left u.str with garbage (from
37739 u.val) which then led to a segfault when attempting to access the
37740 string. Fix that by clearing u.str. The patch also adds missing
37741 sanity checks in the recently committed read_indexed_address and
37742 read_indexed_string functions.
37745 * dwarf2.c (read_indexed_address): Return uint64_t. Sanity check idx.
37746 (read_indexed_string): Use uint64_t for str_offset. Sanity check idx.
37747 (read_attribute_value): Clear u.str for indexed string forms when
37748 DW_AT_str_offsets_base is not yet read or missing.
37750 2022-06-15 Mark Wielaard <mark@klomp.org>
37752 gdb: Always suppress stringop-overread warning in debuginfod-support.c
37753 Just like on s390x with g++ 11.2.1 and ppc64le with g++ 11.3.1 g++ 11
37754 on hppa produces a spurious warning for stringop-overread in
37755 debuginfod_is_enabled for url_view. Just always suppress it on all
37758 https://sourceware.org/bugzilla/show_bug.cgi?id=29198
37762 * debuginfod-support.c (debuginfod_is_enabled): Always use
37763 DIAGNOSTIC_IGNORE_STRINGOP_OVERREAD.
37765 2022-06-15 GDB Administrator <gdbadmin@sourceware.org>
37767 Automatic date update in version.in
37769 2022-06-14 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
37771 gprofng docs: provide help for <rate> == <interval>
37772 The help message from 'gprofng collect app -h', in
37773 the section on <rate> == <interval>, had a dangling
37774 reference to a non-existent manpage. Provide basic
37775 info, including reasons for caution.
37777 gprofng docs: mention HTML / PDF in the gprofng README
37778 The HTML and PDF formats are described in the gprofng tutorial (info
37779 topic "Other Document Formats"). In addition, describe them in the
37780 README because: they are important; they are easily searchable; and the
37781 README is primarily oriented to the person who is installing gprofng,
37782 who may differ from the person who follows a user tutorial.
37784 2022-06-14 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
37786 gprofng: fix build with -Werror=format-security
37788 2022-06-13 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
37791 * src/src/Hist_data.cc (print_row): Make param const.
37792 * src/src/Hist_data.h (print_row): Likewise.
37793 * src/src/Print.h: Remove unused functions and variables.
37794 * src/Print.cc: Fix -Werror=format-security errors.
37795 * src/parse.cc: Likewise.
37797 2022-06-14 Tom de Vries <tdevries@suse.de>
37799 [gdb/testsuite] Handle unordered dict in gdb.python/py-mi-cmd.exp
37800 When running test-case gdb.python/py-mi-cmd.exp on openSUSE Leap 42.3 with
37801 python 3.4, I occasionally run into:
37803 Expecting: ^(-pycmd dct[^M
37804 ]+)?(\^done,result={hello="world",times="42"}[^M
37808 ^done,result={times="42",hello="world"}^M
37810 FAIL: gdb.python/py-mi-cmd.exp: -pycmd dct (unexpected output)
37813 The problem is that the data type used here in py-mi-cmd.py:
37815 elif argv[0] == "dct":
37816 return {"result": {"hello": "world", "times": 42}}
37818 is a dictionary, and only starting version 3.6 are dictionaries insertion
37819 ordered, so using PyDict_Next in serialize_mi_result doesn't guarantee a
37822 Fix this by allowing the alternative order.
37824 Tested on x86_64-linux.
37826 2022-06-14 Tom Tromey <tromey@adacore.com>
37828 Implement lazy FPU initialization for ravenscar
37829 Some ravenscar runtimes implement lazy FPU handling. On these
37830 runtimes, the FPU is only initialized when a task tries to use it.
37831 Furthermore, the FP registers aren't automatically saved on a task
37832 switch -- instead, the save is deferred until the new task tries to
37833 use the FPU. Furthermore, each task's context area has a flag
37834 indicating whether the FPU has been initialized for this task.
37836 This patch teaches GDB to understand this implementation. When
37837 fetching or storing registers, GDB now checks to see whether the live
37838 FP registers should be used. If not, the task's saved FP registers
37839 will be used if the task has caused FPU initialization.
37841 Currently only AArch64 uses this code. bb-runtimes implements this
37842 for ARM as well, but GDB doesn't yet have an arm-ravenscar-thread.c.
37844 2022-06-14 Tom Tromey <tromey@adacore.com>
37846 Reimplement ravenscar registers using tables
37847 Currently, the ravenscar-thread implementation for each architecture
37848 is written by hand. However, these are actually written by
37849 copy-paste. It seems better to switch to a table-driven approach.
37851 The previous code also fetched all registers whenever any register was
37852 requested. This is corrected in the new implementation.
37854 2022-06-14 Tom Tromey <tromey@adacore.com>
37856 Fix bugs in aarch64-ravenscar-thread.c
37857 We found a few bugs in aarch64-ravenscar-thread.c.
37859 First, some of the register offsets were incorrect. The "bb-runtimes"
37860 file for this runtime had the wrong offsets in comments, which GDB
37861 took to be correct. However, those comments didn't account for
37862 alignment. This patch adjusts the offsets.
37864 Next, the "FPU Saved field" is not a register -- it is an
37865 implementation detail of the runtime. This is removed.
37867 Finally, I think the FP registers are actually named V0-V31, and the
37868 "Q" names are pseudo-registers. This patch fixes the comment.
37870 2022-06-14 Tom Tromey <tromey@adacore.com>
37872 Allow 'interrupt -a' in all-stop mode
37873 PR gdb/17160 points out that "interrupt -a" errors in all-stop mode,
37874 but there's no good reason for this. This patch removes the error.
37876 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=17160
37878 2022-06-14 Youling Tang <tangyouling@loongson.cn>
37880 gdbserver: Add LoongArch/Linux support
37881 Implement LoongArch/Linux support, including XML target description
37882 handling based on features determined, GPR regset support, and software
37883 breakpoint handling.
37885 In the Linux kernel code of LoongArch, ptrace implements PTRACE_POKEUSR
37886 and PTRACE_PEEKUSR in the arch_ptrace function, so srv_linux_usrregs is
37889 With this patch on LoongArch:
37891 $ make check-gdb TESTS="gdb.server/server-connect.exp"
37893 # of expected passes 18
37896 2022-06-14 Tom de Vries <tdevries@suse.de>
37898 Revert "Fix fbsd core matching"
37899 This reverts commit a7e29f797cecd5a2f73c27838b09eae1f1b6c657.
37901 I accidentally pushed this, so revert.
37903 2022-06-14 Tom de Vries <tdevries@suse.de>
37905 [gdb/testsuite] Fix regexp in gdb.ada/mi_var_access.exp
37906 With gcc-12 and target board unix/-m32, we run into:
37909 Expecting: ^(-var-create A_String_Access \* A_String_Access[^M
37910 ]+)?(\^done,name="A_String_Access",numchild="1",.*[^M
37913 -var-create A_String_Access * A_String_Access^M
37914 ^error,msg="Value out of range."^M
37916 FAIL: gdb.ada/mi_var_access.exp: Create varobj (unexpected output)
37919 What happens is easier to understand if we take things out of the mi context:
37922 outputs/gdb.ada/mi_var_access/mi_access \
37923 -ex "b mi_access.adb:19" \
37925 -ex "p A_String_Access"
37927 Breakpoint 1, mi_access () at mi_access.adb:19
37928 19 A_String : String (3 .. 5) := "345"; -- STOP
37929 $1 = (pck.string_access) <error reading variable: Value out of range.>
37931 while with target board unix we have instead:
37933 $1 = (pck.string_access) 0x431b40 <ada_main.sec_default_sized_stacks>
37936 The var-create command samples the value of the variable at a location where
37937 the variable is not yet initialized, and with target board unix we
37938 accidentally hit a valid address, but with target board unix/-m32 that's not
37941 Fix the FAIL by accepting the error message.
37943 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28464
37945 2022-06-14 Alan Modra <amodra@gmail.com>
37947 Fix fbsd core matching
37948 On Thu, Jun 09, 2022 at 08:59:37AM -0700, John Baldwin wrote:
37949 > On 6/9/22 1:58 AM, Tom de Vries via Gdb-patches wrote:
37952 > > With an --enable-targets=all build and target board unix/-m32 I run into a
37953 > > FAIL in test-case gdb.base/corefile.exp:
37955 > > (gdb) file outputs/gdb.base/corefile/corefile^M
37956 > > Reading symbols from outputs/gdb.base/corefile/corefile...^M
37957 > > (gdb) core-file outputs/gdb.base/corefile/corefile.core^M
37958 > > warning: core file may not match specified executable file.^M
37959 > > [New LWP 12011]^M
37960 > > Core was generated by `outputs/gdb.base/corefile/co'.^M
37961 > > Program terminated with signal SIGABRT, Aborted.^M
37962 > > (gdb) FAIL: gdb.base/corefile.exp: core-file warning-free
37965 > > The warning is there because of this mismatch between core and exec:
37967 > > (gdb) p core_bfd->xvec
37968 > > $3 = (const struct bfd_target *) 0x20112a0 <i386_elf32_fbsd_vec>
37969 > > (gdb) p exec_bfd->xvec
37970 > > $4 = (const struct bfd_target *) 0x2010b00 <i386_elf32_vec>
37973 > > In the exec case, the detected architecture is i386_elf32_vec because this bit
37974 > > of code in elfcode.h:elf_object_p():
37976 > > if (ebd->elf_machine_code != EM_NONE
37977 > > && i_ehdrp->e_ident[EI_OSABI] != ebd->elf_osabi
37978 > > && ebd->elf_osabi != ELFOSABI_NONE)
37979 > > goto got_wrong_format_error;
37981 > > prevents i386_elf32_fbsd from matching.
37983 > > Fix the core matching by copying that code to elfcore.h:elf_core_file_p().
37985 > > Tested on x86_64-linux.
37987 > > Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29227
37993 > Looking at elfcore.h, it seems to have not gotten changes made to elfcode.h over
37994 > time and is a bit rotted. I suspect that all of changes made in commit 0aabe54e6222
37995 > that added these lines in elfcode.h (along with several other changes) need to
37996 > be applied to this function in elfcore.h, not just adding these lines.
37998 Yes, the commit 0aabe54e6222 changes likely should go in too. I'm a
37999 little wary of adding all the sanity checks to elf_core_file_p since
38000 that might result in some core files not being recognised at all. For
38001 example, despite the FIXME I'd guess leaving out the EI_VERSION check
38002 was deliberate. The following seems reasonable to me. Please test.
38004 2022-06-14 Kavitha Natarajan <kavitha.natarajan@amd.com>
38006 Debug support for global alias variable
38007 Starting with (future) Clang 15 (since
38008 https://reviews.llvm.org/D120989), Clang emits the DWARF information
38009 of global alias variables as DW_TAG_imported_declaration. However,
38010 GDB does not handle it. It incorrectly always reads this tag as
38011 C++/Fortran imported declaration (type alias, namespace alias and
38012 Fortran module). This commit adds support to handle this tag as an
38015 This change fixes the failures in the gdb.base/symbol-alias.exp
38016 testcase with current git Clang. This testcase is also updated to
38017 test nested (recursive) aliases.
38019 2022-06-14 Alan Modra <amodra@gmail.com>
38022 MIPS should not be using BFD_RELOC_16 for its R_MIPS_16 relocation,
38023 since R_MIPS_16 specifies a 16-bit field in a 32-bit word.
38024 BFD_RELOC_16, emitted by generic code to handle fixups on 16-bit data
38025 directives, expects fixups to operate on the whole of a 16-bit word.
38027 This patch corrects the problem by using BFD_RELOC_MIPS_16, a new bfd
38028 reloc that is used to generate R_MIPS_16. BFD_RELOC_16 is handled in
38029 md_apply_fix for cases where the fixup can be applied at assembly
38030 time. Like BFD_RELOC_8, BFD_RELOC_16 now has no corresponding object
38031 file relocation, and thus .half, .hword, .short and .dc.w must be
38032 resolved at assembly time. BFD_RELOC_MIPS_REL16 is removed by this
38033 patch since it isn't used.
38037 * reloc.c (BFD_RELOC_MIPS_16): Rename from BFD_RELOC_MIPS_REL16.
38038 * elf32-mips.c (mips_reloc_map): Map BFD_RELOC_MIPS_16 to R_MIPS_16.
38039 * elf64-mips.c (mips_reloc_map): Likewise, delete BFD_RELOC_MIPS_REL16.
38040 * elfn32-mips.c (mips_reloc_map): Likewise.
38041 * libbfd.h: Regenerate.
38042 * bfd-in2.h: Regenerate.
38044 * config/tc-mips.c (append_insn): Handle BFD_RELOC_MIPS_16.
38045 (macro_build): Likewise.
38046 (mips_percent_op <%half>): Generate BFD_RELOC_MIPS_16.
38047 (md_apply_fix): Handle BFD_RELOC_16 and BFD_RELOC_MIPS_16 when fx_done.
38049 * testsuite/ld-mips-elf/reloc-local-overflow.d,
38050 * testsuite/ld-mips-elf/reloc-local-overflow.s: Rewrite.
38052 2022-06-14 Alan Modra <amodra@gmail.com>
38054 Correct R_MIPS_16 n32 howto
38055 If the howto is actually used, an all-zero dst_mask will result in
38056 unchanged section contents on attempting to apply R_MIPS_16.
38058 * elfn32-mips.c (elf_mips_howto_table_rela <R_MIPS_16>): Correct
38061 2022-06-14 Alan Modra <amodra@gmail.com>
38063 asan: applying zero offset to NULL pointer
38064 * dwarf.c (fetch_indexed_string): Move initialisation of "curr"
38065 and "end" after checking for missing section.
38067 2022-06-14 Alan Modra <amodra@gmail.com>
38069 gas dwarf2dbg.c tidy
38070 Make it a little more obvious that remap_debug_filename returns an
38071 allocated string (that should be freed) by returning a char * rather
38072 than const char *. Free a few missed cases in dwarf2dbg.c, and free
38073 other memory allocated in dwarf2dbg.c. Also remove static
38074 initialisation of variables and initialise in dwarf2_init instead,
38075 in order to ensure gas state is saner for oss-fuzz.
38077 * remap.c (remap_debug_filename): Remove const from return.
38078 * as.h (remap_debug_filename): Update prototype.
38079 * config/obj-elf.c (obj_elf_ident): Simplify free of
38080 remap_debug_filename output.
38081 * stabs.c (stabs_generate_asm_file): Likewise.
38082 * dwarf2dbg.c (dirs, dirs_in_use, dirs_allocated, current): Don't
38083 initialise statically..
38084 (dwarf2_init): ..do so here, along with most other static vars.
38085 (assign_file_to_slot): Don't set files_allocated until we
38086 succeed in allocating memory.
38087 (purge_generated_debug): Add bool param, free more stuff if true.
38088 (dwarf2_directive_filename): Adjust purge_generated_debug call.
38089 (process_entries): Don't free line_entry here..
38090 (dwarf2_cleanup): ..do so here instead, new function.
38091 (dwarf2_finish): Call dwarf2_cleanup. When chaining together
38092 subseg line entries, unhook entries from old subseg list.
38093 (dwarf2_directive_loc): Free remap_debug_filename string.
38094 (out_dir_and_file_list): Likewise.
38095 (out_debug_str): Likewise.
38097 2022-06-14 GDB Administrator <gdbadmin@sourceware.org>
38099 Automatic date update in version.in
38101 2022-06-13 Tom de Vries <tdevries@suse.de>
38103 [gdb/testsuite] Fix gdb.reverse/test_ioctl_TCSETSW.exp with libc debuginfo
38104 When running test-case gdb.reverse/test_ioctl_TCSETSW.exp with glibc debuginfo
38105 installed, I run into:
38107 (gdb) PASS: gdb.reverse/test_ioctl_TCSETSW.exp: at TCSETSW call
38109 __tcsetattr (fd=0, optional_actions=1, termios_p=0x7fffffffcf50) at \
38110 ../sysdeps/unix/sysv/linux/tcsetattr.c:45^M
38112 (gdb) FAIL: gdb.reverse/test_ioctl_TCSETSW.exp: handle TCSETSW
38115 The problem is that the step is expected to step over the call to tcsetattr,
38116 but due to glibc debuginfo being installed, we step into the call.
38118 Fix this by using next instead of step.
38120 Tested on x86_64-linux.
38122 2022-06-13 Tom de Vries <tdevries@suse.de>
38124 [gdb] Avoid warnings in cooked_{read,write}_test for m68hc11
38125 With --enable-targets=all we have:
38127 $ gdb -q -batch -ex "maint selftest"
38129 Running selftest regcache::cooked_read_test::m68hc11.
38130 warning: No frame soft register found in the symbol table.
38131 Stack backtrace will not work.
38132 Running selftest regcache::cooked_read_test::m68hc12.
38133 warning: No frame soft register found in the symbol table.
38134 Stack backtrace will not work.
38135 Running selftest regcache::cooked_read_test::m68hc12:HCS12.
38136 warning: No frame soft register found in the symbol table.
38137 Stack backtrace will not work.
38140 Likewise for regcache::cooked_write_test.
38142 The warning has no use in the selftest context.
38144 Fix this by skipping the specific selftests.
38146 Tested on x86_64-linux.
38148 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29224
38150 2022-06-13 Tiezhu Yang <yangtiezhu@loongson.cn>
38152 gdb: LoongArch: Deal with atomic sequence
38153 We can't put a breakpoint in the middle of a ll/sc atomic sequence,
38154 so look for the end of the sequence and put the breakpoint there.
38156 2022-06-13 Sam James <sam@gentoo.org>
38158 gdb: don't use bashism in configure test
38159 Results in configure output like:
38161 checking for X... no
38162 /var/tmp/portage/sys-devel/gdb-12.1/work/gdb-12.1/gdb/configure: 18837: test: yes: unexpected operator
38163 checking whether to use babeltrace... auto
38166 ... when /bin/sh is provided by a POSIX-compliant shell, like dash,
38169 2022-06-13 Jiangshuai Li <jiangshuai_li@c-sky.com>
38171 gdb:csky add support target-descriptions for CSKY arch
38172 Registers in CSKY architecture included:
38174 2. 16 ars (alternative gprs used for quick interrupt)
38176 4. fr0~fr31, fcsr, fid, fesr
38178 6. ((32 banks) * 32) cr regs (max 32 banks, 32 control regs a bank)
38180 For register names:
38181 Except over control registers, other registers, like gprs, hi, lo ...
38182 are fixed names. Among the 32*32 control registers, some used registers
38183 will have fixed names, others will have a default name "cpxcry". 'x'
38184 refers to bank, y refers index in the bank(a control register in bank
38185 4 with index 14 will has a default name cp4cr14).
38187 For register numbers in GDB:
38188 We assign a fixed number to each register in GDB, like:
38193 described in function csky_get_supported_register_by_index().
38195 Function csky_get_supported_tdesc_registers_count():
38196 To calculate the total number of registers that GDB can analyze,
38197 including those with fixed names and those with default register names.
38199 Function csky_get_supported_register_by_index():
38200 To find a supported struct csky_supported_tdesc_register, return a
38201 struct include name with regnum via index.
38203 Arrays csky_supported_tdesc_feature_names[]:
38204 Include all supported feature names in tdesc-xmls.
38206 We use the information described above to load the register description
38207 file of the target from the stub. When loading, do a little check that
38208 whether the register description file contains SP, LR and PC.
38210 2022-06-13 Tom de Vries <tdevries@suse.de>
38212 [gdb/testsuite] Handle quotes in gdb_py_module_available
38213 On openSUSE Leap 42.3 with python 3.4, I run into:
38215 (gdb) python import pygments^M
38216 Traceback (most recent call last):^M
38217 File "<string>", line 1, in <module>^M
38218 ImportError: No module named 'pygments'^M
38219 Error while executing Python code.^M
38220 (gdb) FAIL: gdb.base/style.exp: python import pygments
38221 ERROR: unexpected output from python import
38223 because gdb_py_module_available doesn't handle the single quotes around the
38224 module name in the ImportError.
38226 Fix this by allowing the single quotes.
38228 Tested on x86_64-linux.
38230 2022-06-13 Jan Beulich <jbeulich@suse.com>
38232 x86: fix incorrect indirection
38233 Commit 384e201e5aec ("x86: properly initialize struct instr_info
38234 instance(s)") was based on an improperly refreshed patch. Correct the
38237 2022-06-13 Jan Beulich <jbeulich@suse.com>
38239 x86: replace global scratch buffer
38240 With its movement to the stack, and with the subsequent desire to
38241 initialize the entire instr_info instances, this has become doubly
38242 inefficient. Individual users have better knowledge of how big a buffer
38243 they need, and in a number of cases going through an intermediate buffer
38244 can be avoided altogether.
38246 Having got confirmation that it wasn't intentional to print memory
38247 operand displacements with inconsistent style, print_displacement() is
38248 now using dis_style_address_offset consistently (eliminating the need
38249 for callers to pass in a style).
38251 While touching print_operand_value() also convert its "hex" parameter to
38252 bool. And while altering (and moving) oappend_immediate(), fold
38253 oappend_maybe_intel_with_style() into its only remaining caller. Finally
38254 where doing adjustments, use snprintf() in favor of sprintf().
38256 2022-06-13 Jan Beulich <jbeulich@suse.com>
38258 x86: avoid string copy when swapping Vex.W controlled operands
38259 Now that op_out[] is an array of pointers, there's no need anymore to
38260 copy strings. Simply swap the pointers.
38262 2022-06-13 Jan Beulich <jbeulich@suse.com>
38264 x86: shrink prefix related disassembler state fields
38265 By changing the values used for "artificial" prefix values,
38266 all_prefixes[] can be shrunk to array of unsigned char. All that
38267 additionally needs adjusting is the printing of possible apparently
38268 standalone prefixes when recovering from longjmp(): Simply check
38269 whether any prefixes were successfully decoded, to avoid converting
38270 opcode bytes matching the "artificial" values to prefix mnemonics.
38272 Similarly by re-arranging the bits assigned to PREFIX_* mask values
38273 we can fit all segment register masks in a byte and hence shrink
38274 active_seg_prefix to unsigned char.
38276 Somewhat similarly with last_*_prefix representing offsets into the
38277 opcode being disassembled, signed char is sufficient to hold all possible
38280 2022-06-13 Jan Beulich <jbeulich@suse.com>
38282 x86: properly initialize struct instr_info instance(s)
38283 Commit 39fb369834a3 ("opcodes: Make i386-dis.c thread-safe") introduced
38284 a lot of uninitialized data. Alan has in particular observed ubsan
38285 taking issue with the loop inverting the order of operands, where
38286 op_riprel[] - an array of bool - can hold values other than 0 or 1.
38288 Move instantiation of struct instr_info into print_insn() (thus having
38289 just a single central point), and make use of C99 dedicated initializers
38290 to fill fields right in the initializer where possible. This way all
38291 fields not explicitly initialized will be zero-filled, which in turn
38292 allows dropping of some other explicit initialization later in the
38293 function or in ckprefix(). Additionally this removes a lot of
38294 indirection, as all "ins->info" uses can simply become "info".
38296 Make one further arrangement though, to limit the amount of data needing
38297 (zero)initializing on every invocation: Convert the op_out structure
38298 member to just an array of pointers, with the actual arrays living
38299 inside print_insn() (and, as befoe, having just their 1st char filled
38302 While there, instead of adjusting print_insn()'s forward declaration,
38303 arrange for no such declaration to be needed in the first place.
38305 2022-06-13 GDB Administrator <gdbadmin@sourceware.org>
38307 Automatic date update in version.in
38309 2022-06-12 Tom Tromey <tom@tromey.com>
38311 Fix self-test failure in addrmap
38312 Mark pointed out that my recent addrmap C++-ficiation changes caused a
38313 regression in the self-tests. This patch fixes the problem by
38314 updating this test not to allocate the mutable addrmap on an obstack.
38316 Remove psymtab_addrmap
38317 While working on addrmaps, I noticed that psymtab_addrmap is no longer
38318 needed now. It was introduced in ancient times as an optimization for
38319 DWARF, but no other symbol reader was ever updated to use it. Now
38320 that DWARF does not use psymtabs, it can be deleted.
38322 Use malloc for mutable addrmaps
38323 Mutable addrmaps currently require an obstack. This was probably done
38324 to avoid having to call splay_tree_delete, but examination of the code
38325 shows that all mutable obstacks have a limited lifetime -- now it's
38326 simple to treat them as ordinary C++ objects, in some cases
38327 stack-allocating them, and have a destructor to make the needed call.
38328 This patch implements this change.
38330 Remove addrmap::create_fixed
38331 addrmap::create_fixed is just a simple wrapper for 'new', so remove it
38332 in favor of uses of 'new'.
38334 Remove addrmap_create_mutable
38335 This removes addrmap_create_mutable in favor of using 'new' at the
38336 spots where the addrmap is created.
38338 Remove addrmap wrapper functions
38339 This removes the various addrmap wrapper functions in favor of simple
38340 method calls on the objects themselves.
38342 Move addrmap classes to addrmap.h
38343 This moves the addrmap class definitions to addrmap.h. This is safe
38344 to do now that the contents are private.
38346 Privacy for addrmap_mutable
38347 This changes addrmap_mutable so that its data members are private.
38349 Privacy for addrmap_fixed
38350 This changes addrmap_fixed so that its data members are private.
38351 It also makes struct addrmap_transition private as well.
38353 Use inheritance for addrmap
38354 This is a simply C++-ification of the basics of addrmap: it uses
38355 virtual methods rather than a table of function pointers, and it
38356 changes the concrete implementations to be subclasses.
38358 2022-06-12 Jon Turney <jon.turney@dronecode.org.uk>
38360 Trivial fixes to Cygwin build after 8fea1a81
38361 * Remove a stray semicolon
38362 * Restore dropped nullptr program argument in use of create_process() under CYGWIN
38365 Prior to c6ca3dab dropping support for Cygwin 1.5, __USEWIDE was not
38366 defined for Cygwin 1.5. After that, it's always defined if __CYGWIN__
38367 is, so remove __USEWIDE conditionals inside __CYGWIN__ conditionals.
38369 Simplify cygwin_buf_t
38370 Prior to c6ca3dab dropping support for Cygwin 1.5, cygwin_buf_t was
38371 defined as char for Cygwin 1.5. After that, it's always wchar_t, so
38374 2022-06-12 GDB Administrator <gdbadmin@sourceware.org>
38376 Automatic date update in version.in
38378 2022-06-11 GDB Administrator <gdbadmin@sourceware.org>
38380 Automatic date update in version.in
38382 2022-06-10 Tom Tromey <tom@tromey.com>
38384 Fix warning-avoidance initialization in xcoffread.c
38385 With the registry rewrite series, on Fedora 34, I started seeing this
38386 error in xcoffread.c:
38388 ../../binutils-gdb/gdb/xcoffread.c: In function ‘void read_xcoff_symtab(objfile*, legacy_psymtab*)’:
38389 ../../binutils-gdb/gdb/xcoffread.c:948:25: error: ‘main_aux’ is used uninitialized [-Werror=uninitialized]
38390 948 | union internal_auxent fcn_aux_saved = main_aux;
38392 ../../binutils-gdb/gdb/xcoffread.c:933:25: note: ‘main_aux’ declared here
38393 933 | union internal_auxent main_aux;
38396 I don't know why this error started suddenly... that seems weird,
38397 because it's not obviously related to the changes I made.
38399 Looking into it, it seems this line was intended to avoid a similar
38400 warning -- but since 'main_aux' is uninitialized at the point where it
38401 is used, this fix was incomplete.
38403 This patch avoids the warning by initializing using "{}". I'm
38406 2022-06-10 Carl Love <cel@us.ibm.com>
38408 Fix comparison of unsigned long int to int in record_linux_system_call.
38409 The if statement in case gdb_sys_ioctl in function
38410 record_linux_system_call in file gdb/linux-record.c is as follows:
38412 if (tmpulongest == tdep->ioctl_FIOCLEX
38413 || tmpulongest == tdep->ioctl_FIONCLEX
38415 || tmpulongest == tdep->ioctl_TCSETSW
38419 The PowerPC ioctl value for ioctl_TCSETW is 0x802c7415. The variable
38420 ioctl_TCSETW is defined in gdb/linux-record.h as an int. The TCSETW value
38421 has the MSB set to one so it is a negative integer. The comparison of the
38422 unsigned long value tmpulongest to a negative integer value for
38423 ioctl_TCSETSW fails.
38425 This patch changes the declarations for the ioctl_* values in struct
38426 linux_record_tdep to unsigned long to fix the comparisons between
38427 tmpulongest and the tdep->ioctl_* values.
38429 An additional test gdb.reverse/test_ioctl_TCSETSW.exp is added to verify
38430 the gdb record_linux_system_call() if statement for the ioctl TCSETSW
38433 This patch has been tested on Power 10 and Intel with no test failures.
38435 2022-06-10 Carl Love <cel@us.ibm.com>
38437 PowerPC, correct the gdb ioctl values for TCGETS, TCSETS, TCSETSW and TCSETSF.
38438 Some of the ioctl numbers are based on the size of kernel termios structure.
38439 Currently the PowerPC GDB definitions are "hard coded" into the ioctl
38442 The current PowerPC values for TCGETS, TCSETS, TCSETSW and TCSETSF are
38443 defined in gdb/ppc-linux-tdep.c as:
38445 record_tdep->ioctl_TCGETS = 0x403c7413;
38446 record_tdep->ioctl_TCSETS = 0x803c7414;
38447 record_tdep->ioctl_TCSETSW = 0x803c7415;
38448 record_tdep->ioctl_TCSETSF = 0x803c7416;
38450 Where the termios structure size is in hex digits [5:4] as 0x3c.
38452 The definition for the PowerPC termios structure is given in:
38453 arch/powerpc/include/uapi/asm/termbits.h
38455 The size of the termios data structure in this file is 0x2c not 0x3c.
38457 This patch changes the hex digits for the size of the PowerPC termios size
38458 in the ioctl values for TCGETS, TCSETS, TCSETSW and TCSETSF to 0x2c.
38459 This patch also changes the hard coding to generate the number based on a
38460 it easier to update the ioctl numbers.
38462 2022-06-10 Andrew Burgess <aburgess@redhat.com>
38464 gdb/testsuite: remove definition of true/false from gdb_compiler_info
38465 Since pretty much forever the get_compiler_info function has included
38468 # Most compilers will evaluate comparisons and other boolean
38469 # operations to 0 or 1.
38470 uplevel \#0 { set true 1 }
38471 uplevel \#0 { set false 0 }
38473 These define global variables true (to 1) and false (to 0).
38475 It seems odd to me that these globals are defined in
38476 get_compiler_info, I guess maybe the original thinking was that if a
38477 compiler had different true/false values then we would detect it there
38478 and define true/false differently.
38480 I don't think we should be bundling this logic into get_compiler_info,
38481 it seems weird to me that in order to use $true/$false a user needs to
38482 first call get_compiler_info.
38484 It would be better I think if each test script that wants these
38485 variables just defined them itself, if in the future we did need
38486 different true/false values based on compiler version then we'd just
38489 if { [test_compiler_info "some_pattern"] } {
38490 # Defined true/false one way...
38492 # Defined true/false another way...
38495 But given the current true/false definitions have been in place since
38496 at least 1999, I suspect this will not be needed any time soon.
38498 Given that the definitions of true/false are so simple, right now my
38499 suggestion is just to define them in each test script that wants
38500 them (there's not that many). If we ever did need more complex logic
38501 then we can always add a function in gdb.exp that sets up these
38502 globals, but that seems overkill for now.
38504 There should be no change in what is tested after this commit.
38506 2022-06-10 Luis Machado <luis.machado@arm.com>
38508 Document the ARM_CC_FOR_TARGET testsuite variable
38509 This variable is useful when exercising AArch64 multi-arch support (debugging
38510 32-bit AArch32 executables).
38512 Unfortunately it isn't well documented. This patch adds information about it
38513 and explains how to use it.
38515 2022-06-10 Tom de Vries <tdevries@suse.de>
38517 [gdb/testsuite] Fix XPASS with gcc-12 in gdb.base/vla-struct-fields.exp
38518 With gcc-12, I get for test-case gdb.base/vla-struct-fields.exp:
38520 (gdb) print inner_vla_struct_object_size == sizeof(inner_vla_struct_object)^M
38522 (gdb) XPASS: gdb.base/vla-struct-fields.exp: size of inner_vla_struct_object
38525 Fix this by limiting the xfailing to gcc-11 and earlier. Also, limit the
38526 xfailing to the equality test.
38528 Tested on x86_64-linux.
38530 2022-06-10 Tom de Vries <tdevries@suse.de>
38532 [gdb/testsuite] Fix timeout in gdb.ada/ghost.exp
38533 On openSUSE Tumbleweed with gcc-12, I run into a timeout:
38535 (gdb) print value^M
38536 Multiple matches for value^M
38538 [1] ada.strings.maps.value (<ref> ada.strings.maps.character_mapping; \
38539 character) return character at a-strmap.adb:599^M
38540 [2] pck.value at src/gdb/testsuite/gdb.ada/ghost/pck.ads:17^M
38541 [3] system.object_reader.value (<ref> system.object_reader.object_symbol) \
38542 return system.object_reader.uint64 at s-objrea.adb:2279^M
38543 [4] system.traceback.symbolic.value (system.address) return string at \
38545 > FAIL: gdb.ada/ghost.exp: print value (timeout)
38546 print ghost_value^M
38547 Argument must be choice number^M
38548 (gdb) FAIL: gdb.ada/ghost.exp: print ghost_value
38551 Fix this by prefixing value (as well as the other printed values) with the
38554 (gdb) print pck.value^M
38557 Tested on x86_64-linux.
38559 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29055
38561 2022-06-10 GDB Administrator <gdbadmin@sourceware.org>
38563 Automatic date update in version.in
38565 2022-06-09 Tom Tromey <tromey@adacore.com>
38567 Minor fix to Python breakpoint event documentation
38568 I noticed that the Python event documentation referred to the event's
38569 "breakpoint" field as a function, whereas it is actually an attribute.
38570 This patch fixes the error.
38572 2022-06-09 Andrew Burgess <aburgess@redhat.com>
38574 gdb/aarch64: fix 32-bit arm compatibility
38575 GDB's ability to run 32-bit ARM processes on an AArch64 native target
38576 is currently broken. The test gdb.multi/multi-arch.exp currently
38577 fails with a timeout.
38579 The cause of these problems is the following three functions:
38581 aarch64_linux_nat_target::thread_architecture
38582 aarch64_linux_nat_target::fetch_registers
38583 aarch64_linux_nat_target::store_registers
38585 What has happened, over time, is that these functions have been
38586 modified, forgetting that any particular thread (running on the native
38587 target) might be an ARM thread, or might be an AArch64 thread.
38589 The problems always start with a line similar to this:
38591 aarch64_gdbarch_tdep *tdep
38592 = (aarch64_gdbarch_tdep *) gdbarch_tdep (inf->gdbarch);
38594 The problem with this line is that if 'inf->gdbarch' is an ARM
38595 architecture, then gdbarch_tdep will return a pointer to an
38596 arm_gdbarch_tdep object, not an aarch64_gdbarch_tdep object. The
38597 result of the above cast will, as a consequence, be undefined.
38599 In aarch64_linux_nat_target::thread_architecture, after the undefined
38600 cast we then proceed to make use of TDEP, like this:
38602 if (vq == tdep->vq)
38603 return inf->gdbarch;
38605 Obviously at this point the result is undefined, but, if this check
38606 returns false we then proceed with this code:
38608 struct gdbarch_info info;
38609 info.bfd_arch_info = bfd_lookup_arch (bfd_arch_aarch64, bfd_mach_aarch64);
38610 info.id = (int *) (vq == 0 ? -1 : vq);
38611 return gdbarch_find_by_info (info);
38613 As a consequence we will return an AArch64 gdbarch object for our ARM
38614 thread! Things go downhill from there on.
38616 There are similar problems, with similar undefined behaviour, in the
38617 fetch_registers and store_registers functions.
38619 The solution is to make use of a check like this:
38621 if (gdbarch_bfd_arch_info (inf->gdbarch)->bits_per_word == 32)
38623 If the word size is 32 then we know we have an ARM architecture. We
38624 just need to make sure that we perform this check before trying to
38625 read the tdep field.
38627 In aarch64_linux_nat_target::thread_architecture a little reordering,
38628 and the addition of the above check allows us to easily avoid the
38629 undefined behaviour.
38631 For fetch_registers and store_registers I made the decision to split
38632 each of the functions into two new helper functions, and so
38633 aarch64_linux_nat_target::fetch_registers now calls to either
38634 aarch64_fetch_registers or aarch32_fetch_registers, and there's a
38635 similar change for store_registers.
38637 One thing I had to decide was whether to place the new aarch32_*
38638 functions into the aarch32-linux-nat.c file. In the end I decided to
38639 NOT place the functions there, but instead leave them in
38640 aarch64-linux-nat.c, my reasoning was this:
38642 The existing functions in that file are shared from arm-linux-nat.c
38643 and aarch64-linux-nat.c, this generic code to support 32-bit ARM
38644 debugging from either native target.
38646 In contrast, the two new aarch32 functions I have added _only_ make
38647 sense when debugging on an AArch64 native target. These function
38648 shouldn't be called from arm-linux-nat.c at all, and so, if we places
38649 the functions into aarch32-linux-nat.c, the functions would be built
38650 into a 32-bit ARM GDB, but never used.
38652 With that said, there's no technical reason why they couldn't go in
38653 aarch32-linux-nat.c, so if that is preferred I'm happy to move them.
38655 After this commit the gdb.multi/multi-arch.exp passes.
38657 2022-06-09 Yvan Roux <yvan.roux@foss.st.com>
38659 gdb/arm: Document and fix exception stack offsets
38660 Add a description of exception entry context stacking and fix next
38661 frame offset (at 0xA8 relative to R0 location) as well as FPU
38662 registers ones (starting at 0x68 relative to R0).
38664 gdb/arm: Simplify logic for updating addresses
38665 Small performance improvement by fetching previous SP value only
38666 once before the loop and reuse it to avoid fetching at every
38669 2022-06-09 Pedro Alves <pedro@palves.net>
38671 Fix ARM_CC_FOR_TARGET handling
38672 The previous patch that introduced the arm_cc_for_target procedure
38673 moved the ARM_CC_FOR_TARGET global check to that procedure, but forgot
38674 to tell tcl that ARM_CC_FOR_TARGET is a global. As a result,
38675 specifying ARM_CC_FOR_TARGET on the command line actually does
38676 nothing. This fixes it.
38678 Change-Id: I4e33b7633fa665e2f7b8f8c9592a949d74a19153
38680 2022-06-09 Yvan Roux <yvan.roux@foss.st.com>
38682 gdb/arm: Terminate unwinding when LR is 0xffffffff
38683 ARMv7-M Architecture Reference "A2.3.1 Arm core registers" states
38684 that LR is set to 0xffffffff on reset.
38686 ARMv8-M Architecture Reference "B3.3 Registers" states that LR is set
38687 to 0xffffffff on warm reset if Main Extension is implemented,
38688 otherwise the value is unknown.
38690 2022-06-09 Andrew Burgess <aburgess@redhat.com>
38692 gdb/testsuite: solve problems with compiler_info caching
38695 commit 44d469c5f85a4243462b8966722dafa62b602bf5
38696 Date: Tue May 31 16:43:44 2022 +0200
38698 gdb/testsuite: add Fortran compiler identification to GDB
38700 Some regressions were noticed:
38702 https://sourceware.org/pipermail/gdb-patches/2022-May/189673.html
38704 The problem is associated with how compiler_info variable is cached
38705 between calls to get_compiler_info.
38707 Even before the above commit, get_compiler_info supported two
38708 language, C and C++. Calling get_compiler_info would set the global
38709 compiler_info based on the language passed as an argument to
38710 get_compiler_info, and, in theory, compiler_info would not be updated
38711 for the rest of the dejagnu run.
38713 This obviously is slightly broken behaviour. If the first call to
38714 get_compiler_info was for the C++ language then compiler_info would be
38715 set based on the C++ compiler in use, while if the first call to
38716 get_compiler_info was for the C language then compiler_info would be
38717 set based on the C compiler.
38719 This probably wasn't very noticable, assuming a GCC based test
38720 environment then in most cases the C and C++ compiler would be the
38723 However, if the user starting playing with CC_FOR_TARGET or
38724 CXX_FOR_TARGET, then they might not get the behaviour they expect.
38726 Except, to make matters worse, most of the time, the user probably
38727 would get the behaviour they expected .... except when they didn't!
38730 In gdb.exp we try to avoid global variables leaking between test
38731 scripts, this is done with the help of the two procs
38732 gdb_setup_known_globals and gdb_cleanup_globals. All known globals
38733 are recorded before a test script starts, and then, when the test
38734 script ends, any new globals are deleted.
38736 Normally, compiler_info is only set as a result of a test script
38737 calling get_compiler_info or test_compiler_info. This means that the
38738 compiler_info global will not exist when the test script starts, but
38739 will exist when the test script end, and so, the compiler_info
38740 variable is deleted at the end of each test.
38742 This means that, in reality, the compiler_info is recalculated once
38743 for each test script, hence, if a test script just checks on the C
38744 compiler, or just checks on the C++ compiler, then compiler_info will
38745 be correct and the user will get the behaviour they expect.
38747 However, if a single test script tries to check both the C and C++
38748 compiler versions then this will not work (even before the above
38751 The situation is made worse be the behaviour or the load_lib proc.
38752 This proc (provided by dejagnu) will only load each library once.
38753 This means that if a library defines a global, then this global would
38754 normally be deleted at the end of the first test script that includes
38757 As future attempts to load the library will not actually reload it,
38758 then the global will not be redefined and would be missing for later
38759 test scripts that also tried to load that library.
38761 To work around this issue we override load_lib in gdb.exp, this new
38762 version adds all globals from the newly loaded library to the list of
38763 globals that should be preserved (not deleted).
38765 And this is where things get interesting for us. The library
38766 trace-support.exp includes calls, at the file scope, to things like
38767 is_amd64_regs_target, which cause get_compiler_info to be called.
38768 This means that after loading the library the compiler_info global is
38771 Our override of load_lib then decides that this new global has to be
38772 preserved, and adds it to the gdb_persistent_globals array.
38774 From that point on compiler_info will never be recomputed!
38776 This commit addresses all the caching problems by doing the following:
38778 Change the compiler_info global into compiler_info_cache global. This
38779 new global is an array, the keys of this array will be each of the
38780 supported languages, and the values will be the compiler version for
38783 Now, when we call get_compiler_info, if the compiler information for
38784 the specific language has not been computed, then we do that, and add
38787 Next, compiler_info_cache is defined by calling
38788 gdb_persistent_global. This automatically adds the global to the list
38789 of persistent globals. Now the cache will not be deleted at the end
38790 of each test script.
38792 This means that, for a single test run, we will compute the compiler
38793 version just once for each language, this result will then be cached
38794 between test scripts.
38796 Finally, the legacy 'gcc_compiled' flag is now only set when we call
38797 get_compiler_info with the language 'c'. Without making this change
38798 the value of 'gcc_compiled' would change each time a new language is
38799 passed to get_compiler_info. If the last language was e.g. Fortran,
38800 then gcc_compiled might be left false.
38802 2022-06-09 Andrew Burgess <aburgess@redhat.com>
38804 gdb/testsuite: handle errors better in test_compiler_info
38805 Now that get_compiler_info might actually fail (if the language is not
38806 handled), then we should try to handle this failure better in
38807 test_compiler_info.
38809 After this commit, if get_compiler_info fails then we will return a
38810 suitable result depending on how the user called test_compiler_info.
38812 If the user does something like:
38814 set version [test_compiler_info "" "unknown-language"]
38816 Then test_compiler_info will return an empty string. My assumption is
38817 that the user will be trying to match 'version' against something, and
38818 the empty string hopefully will not match.
38820 If the user does something like:
38822 if { [test_compiler_info "some_pattern" "unknown-language"] } {
38826 Then test_compiler_info will return false which seems the obvious
38829 There should be no change in the test results after this commit.
38831 2022-06-09 Andrew Burgess <aburgess@redhat.com>
38833 gdb/testsuite: make 'c' default language for get/test compiler info
38834 This commit is a minor cleanup for the two functions (in gdb.exp)
38835 get_compiler_info and test_compiler_info.
38837 Instead of using the empty string as the default language, and just
38838 "knowing" that this means the C language. Make this explicit. The
38839 language argument now defaults to "c" if not specified, and the if
38840 chain in get_compiler_info that checks the language not explicitly
38841 handles "c" and gives an error for unknown languages.
38843 This is a good thing, now that the API appears to take a language, if
38846 test_compiler_info "xxxx" "rust"
38848 to check the version of the rust compiler then we will now give an
38849 error rather than just using the C compiler and leaving the user
38850 having to figure out why they are not getting the results they
38853 After a little grepping, I think the only place we were explicitly
38854 passing the empty string to either get_compiler_info or
38855 test_compiler_info was in gdb_compile_shlib_1, this is now changed to
38856 pass "c" as the default language.
38858 There should be no changes to the test results after this commit.
38860 2022-06-09 Andrew Burgess <aburgess@redhat.com>
38862 gdb/testsuite: remove get_compiler_info calls from gdb.exp and dwarf.exp
38863 We don't need to call get_compiler_info before calling
38864 test_compiler_info; test_compiler_info includes a call to
38867 This commit cleans up lib/gdb.exp and lib/dwarf.exp a little by
38868 removing some unneeded calls to get_compiler_info. We could do the
38869 same cleanup throughout the testsuite, but I'm leaving that for
38872 There should be no change in the test results after this commit.
38874 2022-06-09 Nils-Christian Kempke <nils-christian.kempke@intel.com>
38876 gdb/testsuite: use test_compiler_info in gcc_major_version
38877 The procedure gcc_major_version was earlier using the global variable
38878 compiler_info to retrieve gcc's major version. This is discouraged and
38879 (as can be read in a comment in compiler.c) compiler_info should be
38880 local to get_compiler_info and test_compiler_info.
38882 The preferred way of getting the compiler string is via calling
38883 test_compiler_info without arguments. Gcc_major_version was changed to
38886 2022-06-09 Yvan Roux <yvan.roux@foss.st.com>
38888 gdb: add Yvan Roux to gdb/MAINTAINERS
38890 2022-06-09 Andrew Burgess <aburgess@redhat.com>
38892 gdb/testsuite: resolve duplicate test names in gdb.threads/tls.exp
38893 While running the gdb.threads/tls.exp test with a GDB configured
38894 without Python, I noticed some duplicate test names.
38896 This is caused by a call to skip_python_tests that is within a proc
38897 that is called multiple times by the test script. Each call to
38898 skip_python_tests results in a call to 'unsupported', and this causes
38899 the duplicate test names.
38901 After this commit we now call skip_python_tests just once and place
38902 the result into a variable. Now, instead of calling skip_python_tests
38903 multiple times, we just check the variable.
38905 There should be no change in what is tested after this commit.
38907 2022-06-09 Andrew Burgess <aburgess@redhat.com>
38909 gdb/testsuite: resolve duplicate test name in gnu_vector.exp
38910 While testing on AArch64 I spotted a duplicate test name in the
38911 gdb.base/gnu_vector.exp test.
38913 This commit adds a 'with_test_prefix' to resolve the duplicate.
38915 While I was in the area I updated a 'gdb_test_multiple' call to make
38916 use of $gdb_test_name.
38918 There should be no change in what is tested after this commit.
38920 2022-06-09 GDB Administrator <gdbadmin@sourceware.org>
38922 Automatic date update in version.in
38924 2022-06-08 Andrew Burgess <aburgess@redhat.com>
38926 gdb: make throw_perror_with_name static
38927 The throw_perror_with_name function is not used outside of utils.c
38928 right now. And as perror_with_name is just a wrapper around
38929 throw_perror_with_name, then any future calls would be to
38932 Lets make throw_perror_with_name static.
38934 There should be no user visible changes after this commit.
38936 2022-06-08 Andrew Burgess <aburgess@redhat.com>
38938 gdb: remove trailing '.' from perror_with_name calls
38939 I ran into this error while working on AArch64 GDB:
38941 Unable to fetch VFP registers.: Invalid argument.
38943 Notice the '.:' in the middle of this error message.
38945 This is because of this call in aarch64-linux-nat.c:
38947 perror_with_name (_("Unable to fetch VFP registers."));
38949 The perror_with_name function take a string, and adds ': <message>' to
38950 the end the string, so I don't think the string that we pass to
38951 perror_with_name should end in '.'.
38953 This commit removes all of the trailing '.' characters from
38954 perror_with_name calls, which give more readable error messages.
38956 I don't believe that any of these errors are tested in the
38957 testsuite (after a little grepping).
38959 2022-06-08 Tom Tromey <tom@tromey.com>
38961 Move CU queue to dwarf2_per_objfile
38962 The CU queue is a member of dwarf2_per_bfd, but it is only used when
38963 expanding CUs. Also, the dwarf2_per_objfile destructor checks the
38964 queue -- however, if the per-BFD object is destroyed first, this will
38965 not work. This was pointed out Lancelot as fallout from the patch to
38966 rewrite the registry system.
38968 This patch avoids this problem by moving the queue to the per-objfile
38971 2022-06-08 Tom Tromey <tom@tromey.com>
38973 Change allocation of m_dwarf2_cus
38974 m_dwarf2_cus manually manages the 'dwarf2_cu' pointers it owns. This
38975 patch simplifies the code by changing it to use unique_ptr.
38977 2022-06-08 Andrew Burgess <aburgess@redhat.com>
38979 libopcodes: extend the styling within the i386 disassembler
38980 The i386 disassembler is pretty complex. Most disassembly is done
38981 indirectly; operands are built into buffers within a struct instr_info
38982 instance, before finally being printed later in the disassembly
38985 Sometimes the operand buffers are built in a different order to the
38986 order in which they will eventually be printed.
38988 Each operand can contain multiple components, e.g. multiple registers,
38989 immediates, other textual elements (commas, brackets, etc).
38991 When looking for how to apply styling I guess the ideal solution would
38992 be to move away from the operands being a single string that is built
38993 up, and instead have each operand be a list of "parts", where each
38994 part is some text and a style. Then, when we eventually print the
38995 operand we would loop over the parts and print each part with the
38998 But it feels like a huge amount of work to move from where we are
38999 now to that potentially ideal solution. Plus, the above solution
39000 would be pretty complex.
39002 So, instead I propose a .... different solution here, one that works
39003 with the existing infrastructure.
39005 As each operand is built up, piece be piece, we pass through style
39006 information. This style information is then encoded into the operand
39007 buffer (see below for details). After this the code can continue to
39008 operate as it does right now in order to manage the set of operand
39011 Then, as each operand is printed we can split the operand buffer into
39012 chunks at the style marker boundaries, with each chunk being printed
39013 with the correct style.
39015 For encoding the style information I use a single character, currently
39016 \002, followed by the style encoded as a single hex digit, followed
39017 again by the \002 character.
39019 This of course relies on there not being more than 16 styles, but that
39020 is currently true, and hopefully will remain true for the foreseeable
39023 The other major concern that has arisen around this work is whether
39024 the escape character could ever be encountered in output naturally
39025 generated by the disassembler. If this did happen then the escape
39026 characters would be stripped from the output, and the wrong styling
39029 However, I don't believe that this is currently a problem.
39030 Disassembler content comes from a number of sources. First there's
39031 content that copied directly from the i386-dis.c file, this is things
39032 like register names, and other syntax elements (brackets, commas,
39033 etc). We can easily check that the i386-dis.c file doesn't contain
39034 our special character.
39036 The next source of content are immediate operands. The text for these
39037 operands is generated by calls into libc. By selecting a
39038 non-printable character we can be confident that this is not something
39039 that libc will generate as part of an immediate representation.
39041 The other output that appears to be from the disassembler is operands
39042 that contain addresses and (possibly) symbol names. It is quite
39043 possible that a symbol name might contain any special character we
39044 could imagine, so is this a problem?
39046 I don't think it is, we don't actually print address and symbol
39047 operands through the disassembler, instead, the disassembler calls
39048 back to the user (objdump, gdb, etc) to print the address and symbol
39049 on its behalf. This content is printed directly to the output stream,
39050 it does not pass through the i386 disassembler output buffers. As a
39051 result, we never check this particular output for styling escape
39054 In some (not very scientific) benchmarking on my machine,
39055 disassembling a reasonably large (142M) shared library, I'm not seeing
39056 any significant slow down in disassembler speed with this change.
39058 Most instructions are now being fully syntax highlighted when I
39059 disassemble using the --disassembler-color=extended-color option. I'm
39060 sure that there are probably still a few corner cases that need fixing
39061 up, but we can come back to them later I think.
39063 When disassembler syntax highlighting is not being used, then there
39064 should be no user visible changes after this commit.
39066 2022-06-08 Carl Love <cel@us.ibm.com>
39068 Fix gdb.arch/powerpc-power7.exp isel disassembly output.
39069 The following commit changes the output format for the isel instruction on
39072 commit dd4832bf3efc1bd1797a6b9188260692b8b0db52 Introduces error in test
39073 Author: Dmitry Selyutin <ghostmansd@gmail.com>
39074 Date: Tue May 24 13:46:35 2022 +0000
39076 opcodes: introduce BC field; fix isel
39078 Per Power ISA Version 3.1B 3.3.12, isel uses BC field rather than CRB
39079 field present in binutils sources. Also, per 1.6.2, BC has the same
39080 semantics as BA and BB fields, so this should keep the same flags and
39081 mask, only with the different offset.
39085 (BC): Define new field, with the same definition as CRB field,
39086 but with the PPC_OPERAND_CR_BIT flag present.
39088 * testsuite/gas/ppc/476.d: Update.
39089 * testsuite/gas/ppc/a2.d: Update.
39090 * testsuite/gas/ppc/e500.d: Update.
39091 * testsuite/gas/ppc/power7.d: Update.
39093 --- a/gas/testsuite/gas/ppc/476.d
39094 +++ b/gas/testsuite/gas/ppc/476.d
39095 @@ -209,7 +209,7 @@ Disassembly of section \.text:
39096 .*: (7c 20 07 8c|8c 07 20 7c) ici 1
39097 .*: (7c 03 27 cc|cc 27 03 7c) icread r3,r4
39098 .*: (50 83 65 36|36 65 83 50) rlwimi r3,r4,12,20,27
39099 -.*: (7c 43 27 1e|1e 27 43 7c) isel r2,r3,r4,28
39100 +.*: (7c 43 27 1e|1e 27 43 7c) isel r2,r3,r4,4\*cr7\+lt
39102 The above change breaks the gdb regression test gdb.arch/powerpc-power7.exp
39103 on Power 7, Power 8, Power 9 and Power 10.
39105 This patch updates the regression test gdb.arch/powerpc-power7.exp with
39106 the new expected output for the isel instruction.
39108 The patch has been tested on Power 7 and Power 10 to verify the patch fixes
39111 2022-06-08 Pedro Alves <pedro@palves.net>
39113 aarch64: Add fallback if ARM_CC_FOR_TARGET not set
39114 On Aarch64, you can set ARM_CC_FOR_TARGET to point to the 32-bit
39115 compiler to use when testing gdb.multi/multi-arch.exp and
39116 gdb.multi/multi-arch-exec.exp. If you don't set it, then those
39117 testcases don't run.
39119 I guess that approximately nobody remembers to set ARM_CC_FOR_TARGET.
39121 This commit adds a fallback. If ARM_CC_FOR_TARGET is not set, and
39122 testing for Linux, try arm-linux-gnueabi-gcc,
39123 arm-none-linux-gnueabi-gcc, arm-linux-gnueabihf-gcc as 32-bit
39124 compilers, making sure that the produced executable runs on the target
39125 machine before claiming that the compiler produces useful executables.
39127 Change-Id: Iefe5865d5fc84b4032eaff7f4c5c61582bf75c39
39129 2022-06-08 Alan Modra <amodra@gmail.com>
39131 Don't encode reloc.size
39132 I expect the encoded reloc.size field originally came from aout
39133 r_length ecoding, but somehow went wrong for 64-bit relocs (which
39134 should have been encoded as 3). Toss all that out, just use a byte
39135 size instead. The changes outside of reloc.c in this patch should
39136 make the code independent of how reloc.size is encoded.
39138 * reloc.c (struct reloc_howto_struct): Increase size field by
39140 (HOWTO_RSIZE): Don't encode size.
39141 (bfd_get_reloc_size): Adjust, and make it an inline function.
39142 (read_reloc, write_reloc): Adjust.
39143 * bfd-in2.h: Regenerate.
39144 * aout-ns32k.c: Include libbfd.h.
39145 (put_reloc): Don't use howto->size directly. Calculate r_length
39146 using bfd_log2 and bfd_get_reloc_size.
39147 * aoutx.h (swap_std_reloc_out): Likewise.
39148 (aout_link_reloc_link_order): Likewise.
39149 * i386lynx.c (swap_std_reloc_out
39150 * mach-o-i386.c (bfd_mach_o_i386_swap_reloc_out
39151 * pdp11.c (aout_link_reloc_link_order
39152 * coff-arm.c (coff_arm_reloc): Don't use howto->size directly,
39153 use bfd_get_reloc_size instead and adjust switch cases.
39154 * coff-i386.c (coff_i386_reloc): Similarly.
39155 * coff-x86_64.c (coff_amd64_reloc): Likewise.
39156 * cpu-ns32k.c (do_ns32k_reloc): Likewise.
39157 * elf32-arc.c (arc_do_relocation): Likewise.
39158 * elf32-arm.c (elf32_arm_final_link_relocate): Likewise.
39159 * elf32-bfin.c (bfin_bfd_reloc): Likewise.
39160 * elf32-cr16.c (cr16_elf_final_link_relocate): Likewise.
39161 * elf32-cris.c (cris_elf_pcrel_reloc): Likewise.
39162 * elf32-crx.c (crx_elf_final_link_relocate): Likewise.
39163 * elf32-csky.c (csky_elf_relocate_section): Likewise.
39164 * elf32-d10v.c (extract_rel_addend, insert_rel_addend): Likewise.
39165 * elf32-i386.c (elf_i386_relocate_section): Likewise.
39166 * elf32-m32r.c (m32r_elf_generic_reloc): Likewise.
39167 * elf32-nds32.c (nds32_elf_generic_reloc): Likewise.
39168 * syms.c (_bfd_stab_section_find_nearest_line): Likewise.
39169 * coff-rs6000.c (xcoff_ppc_relocate_section): Adjust howto.size.
39170 * coff64-rs6000.c (xcoff64_ppc_relocate_section): Likewise.
39172 2022-06-08 Alan Modra <amodra@gmail.com>
39174 bfin reloc offset checks
39175 These all ought to use bfd_reloc_offset_in_range. In particular, replace
39176 the check using howto->size + 1u.
39178 * elf32-bfin.c (bfin_pcrel24_reloc): Use bfd_reloc_offset_in_range.
39179 (bfin_imm16_reloc, bfin_byte4_reloc, bfin_bfd_reloc),
39180 (bfin_final_link_relocate): Likewise.
39182 2022-06-08 Alan Modra <amodra@gmail.com>
39184 Revert reloc howto nits
39185 The "HOWTO size encoding" patch put 1 as the HOWTO size arg for
39186 numerous howtos that are unused, describe dynamic relocs, are markers,
39187 or otherwise are special purpose reloc howtos that don't care about
39188 the size. The idea was to ensure no howto changed by inspecting
39189 object files. Revert those changes, making them zero size.
39191 * coff-alpha.c: Give special purpose reloc howtos a size of zero.
39192 * coff-mcore.c, * elf-hppa.h, * elf-m10300.c, * elf32-arm.c,
39193 * elf32-csky.c, * elf32-m32c.c, * elf32-m68k.c, * elf32-mep.c,
39194 * elf32-mips.c, * elf32-ppc.c, * elf32-rx.c, * elf32-s390.c,
39195 * elf32-spu.c, * elf32-tic6x.c, * elf32-tilepro.c, *elf32-vax.c,
39196 * elf32-xtensa.c, * elf64-alpha.c, * elf64-mips.c,
39197 * elf64-mmix.c, * elf64-ppc.c, * elf64-s390.c, * elfn32-mips.c,
39198 * elfxx-loongarch.c, * elfxx-riscv.c, * elfxx-sparc.c,
39199 * elfxx-tilegx.c, * som.c, * vms-alpha.c: Likewise.
39201 2022-06-08 Alan Modra <amodra@gmail.com>
39203 HOWTO size encoding
39204 This changes the HOWTO macro to encode the howto.size field from a
39205 value given in bytes. This of course requires editing all target
39206 uses of HOWTO, a major pain, but makes it a little nicer to specify
39207 new target HOWTOs. Object files before/after this patch are
39208 unchanged in .data and .rodata.
39211 * reloc.c (HOWTO_RSIZE): Encode size in bytes.
39212 (EMPTY_HOWTO): Adjust to keep it all zero.
39213 * aout-ns32k.c, * aoutx.h, * coff-alpha.c, * coff-arm.c,
39214 * coff-i386.c, * coff-mcore.c, * coff-mips.c, * coff-rs6000.c,
39215 * coff-sh.c, * coff-tic30.c, * coff-tic4x.c, * coff-tic54x.c,
39216 * coff-x86_64.c, * coff-z80.c, * coff-z8k.c, * coff64-rs6000.c,
39217 * elf-hppa.h, * elf-m10200.c, * elf-m10300.c, * elf32-arc.c,
39218 * elf32-arm.c, * elf32-avr.c, * elf32-bfin.c, * elf32-cr16.c,
39219 * elf32-cris.c, * elf32-crx.c, * elf32-csky.c, * elf32-d10v.c,
39220 * elf32-d30v.c, * elf32-dlx.c, * elf32-epiphany.c,
39221 * elf32-fr30.c, * elf32-frv.c, * elf32-ft32.c, * elf32-gen.c,
39222 * elf32-h8300.c, * elf32-i386.c, * elf32-ip2k.c, * elf32-iq2000.c,
39223 * elf32-lm32.c, * elf32-m32c.c, * elf32-m32r.c, * elf32-m68hc11.c,
39224 * elf32-m68hc12.c, * elf32-m68k.c, * elf32-mcore.c, * elf32-mep.c,
39225 * elf32-metag.c, * elf32-microblaze.c, * elf32-mips.c,
39226 * elf32-moxie.c, * elf32-msp430.c, * elf32-mt.c, * elf32-nds32.c,
39227 * elf32-nios2.c, * elf32-or1k.c, * elf32-pj.c, * elf32-ppc.c,
39228 * elf32-pru.c, * elf32-rl78.c, * elf32-rx.c, * elf32-s12z.c,
39229 * elf32-s390.c, * elf32-score.c, * elf32-score7.c,
39230 * elf32-sh-relocs.h, * elf32-spu.c, * elf32-tic6x.c,
39231 * elf32-tilepro.c, * elf32-v850.c, * elf32-vax.c,
39232 * elf32-visium.c, * elf32-wasm32.c, * elf32-xc16x.c,
39233 * elf32-xgate.c, * elf32-xstormy16.c, * elf32-xtensa.c,
39234 * elf32-z80.c, * elf64-alpha.c, * elf64-bpf.c, * elf64-gen.c,
39235 * elf64-mips.c, * elf64-mmix.c, * elf64-nfp.c, * elf64-ppc.c,
39236 * elf64-s390.c, * elf64-x86-64.c, * elfn32-mips.c,
39237 * elfnn-aarch64.c, * elfxx-ia64.c, * elfxx-loongarch.c,
39238 * elfxx-mips.c, * elfxx-riscv.c, * elfxx-sparc.c,
39239 * elfxx-tilegx.c, * mach-o-aarch64.c, * mach-o-arm.c,
39240 * mach-o-i386.c, * mach-o-x86-64.c, * pdp11.c, * reloc.c,
39241 * som.c, * vms-alpha.c: Adjust all uses of HOWTO.
39242 * bfd-in2.h: Regenerate.
39244 * elf/arc-reloc.def: Adjust all uses of HOWTO.
39246 2022-06-08 Alan Modra <amodra@gmail.com>
39249 Define a helper macro for HOWTO.
39251 * reloc.c (HOWTO_RSIZE): Define.
39253 * bfd-in2.h: Regenerate.
39255 2022-06-08 Alan Modra <amodra@gmail.com>
39257 elf64-nfp reloc fix
39258 These are all dummy howtos, there is no reason one of them should
39259 have partial_inplace true.
39261 * elf64-nfp.c (elf_nfp_howto_table <R_NFP_IMMED_LO16_I_B>): Don't
39262 set partial_inplace.
39264 2022-06-08 Alan Modra <amodra@gmail.com>
39266 coff-z80 reloc howto fixes
39267 Mostly cosmetic unless attempting to link coff-z80 into another output
39270 * coff-z80.c (howto_table <R_IMM24, R_WORD0, R_WORD1>): Correct size.
39271 (extra_case): Use bfd_{get,put}_24 when applying R_IMM24.
39273 2022-06-08 Alan Modra <amodra@gmail.com>
39276 Make them all zero size standard do-nothing howtos.
39278 * elf32-csky.c (csky_elf_howto_table <R_CKCORE_NONE>): Correct howto.
39279 * elf32-ft32.c (ft32_elf_howto_table <R_FT32_NONE>): Likewise.
39280 * elf32-gen.c (dummy): Likewise.
39281 * elf32-nds32.c (none_howto): Likewise.
39282 * elf32-nios2.c (elf_nios2_r2_howto_table_rel <R_NIOS2_NONE>):
39284 * elf32-pru.c (elf_pru_howto_table_rel <R_PRU_NONE>): Likewise.
39285 * elf32-v850.c (v800_elf_howto_table <R_V810_NONE>): Likewise.
39286 * elf64-gen.c (dummy): Likewise.
39287 * elfn32-mips.c (elf_mips_howto_table_rela <R_MIPS_NONE): Likewise.
39288 * elfxx-mips.c (none_howto): Likewise.
39289 * reloc.c (none_howto): Likewise.
39291 2022-06-08 Alan Modra <amodra@gmail.com>
39293 asan: double free sb_kill
39294 oss-fuzz hits a flaky crash with a double-free. I think this is due
39295 to gas static state not being reinitialised between testcases, a bug
39296 with oss-fuzz not gas. Anyway, this patch should avoid the problem.
39298 * input-scrub.c (input_scrub_push): Move init of sb_index..
39299 (input_scrub_reinit): ..to here.
39301 2022-06-08 GDB Administrator <gdbadmin@sourceware.org>
39303 Automatic date update in version.in
39305 2022-06-07 Tom Tromey <tromey@adacore.com>
39307 Use subclasses of windows_process_info
39308 This changes windows_process_info to use virtual methods for its
39309 callbacks, and then changes the two clients of this code to subclass
39310 this class to implement the methods.
39312 I considered using CRTP here, but that would require making the new
39313 structures visible to the compilation of of nat/windows-nat.c. This
39314 seemed like a bit of a pain, so I didn't do it.
39316 This change then lets us change all the per-inferior globals to be
39317 members of the new subclass. Note that there can still only be a
39318 single inferior -- currently there's a single global of the new type.
39319 This is just another step toward possibly implementing multi-inferior
39322 It's possible this could be cleaned up further... ideally I'd like to
39323 move more of the data into the base class. However, because gdb
39324 supports Cygwin and gdbserver does not, and because I don't have a way
39325 to build or test Cygwin, larger refactorings are difficult.
39327 2022-06-07 Tom Tromey <tromey@adacore.com>
39329 Turn some windows-nat.c static functions into methods
39330 This patch turns some windows-nat.c static functions into methods on
39331 windows_nat_target. This avoids having to reference the
39332 windows_nat_target singleton in some more spots -- a minor code
39335 Allow ASLR to be disabled on Windows
39336 On Windows, it is possible to disable ASLR when creating a process.
39337 This patch adds code to do this, and hooks it up to gdb's existing
39338 disable-randomization feature. Because the Windows documentation
39339 cautions that this isn't available on all versions of Windows, the
39340 CreateProcess wrapper function is updated to make the attempt, and
39341 then fall back to the current approach if it fails.
39343 Introduce wrapper for CreateProcess
39344 This is a small refactoring that introduces a wrapper for the Windows
39345 CreateProcess function. This is done to make the next patch a bit
39348 2022-06-07 Enze Li <enze.li@hotmail.com>
39350 Update my email address in gdb/MAINTAINERS
39352 2022-06-07 Tom Tromey <tromey@adacore.com>
39354 Constify solib_name_from_address
39355 I noticed that solib_name_from_address returned a non-const string,
39356 but it's more appropriate to return const. This patch implements
39357 this. Tested by rebuilding.
39359 2022-06-07 Tom de Vries <tdevries@suse.de>
39361 [gdb/rust] Add missing _() for error call
39362 In commit 1390b65a1b9 ("[gdb/rust] Fix literal truncation") I forgot to add
39363 _() around a string using in an error call.
39365 Fix this by adding the missing _().
39367 Tested on x86_64-linux.
39369 2022-06-07 Tom de Vries <tdevries@suse.de>
39371 [gdb] Allow frv::fr300 in selftests
39372 In skip_arch in gdb/selftest-arch.c we skip architecture fr300 because of
39373 PR20946, but the PR has been fixed by commit 0ae60c3ef45 ("Prevent an abort in
39374 the FRV disassembler if the target bfd name is unknown.") in Januari 2017.
39376 Remove the skipping of frv::fr300.
39378 Tested on x86_64-linux.
39380 2022-06-07 GDB Administrator <gdbadmin@sourceware.org>
39382 Automatic date update in version.in
39384 2022-06-06 Tom Tromey <tromey@adacore.com>
39386 Consolidate "Python API" sections in NEWS
39387 I noticed that the gdb NEWS file had two "Python API" sections in
39388 "Changes since GDB 12". This patch consolidates the two. I chose to
39389 preserve the second one, first because it is longer, and second
39390 because I felt that user command changes should come before API
39393 Simplify varobj "change" logic
39394 varobj used to store 'print_value' as a C string, where NULL was a
39395 valid value, and so it had logic to handle this situation. However,
39396 at some point this was changed to be a std::string, and so the code
39397 can be simplified in this spot.
39399 2022-06-06 Tom Tromey <tromey@adacore.com>
39401 Remove "-break-insert -r" tests
39402 PR mi/14270 points out that mi-break.exp has some tests for an
39403 unimplemented "-r" switch for "-break-insert". This switch was never
39404 implemented, and is not documented -- though it is mentioned in a
39405 comment in the documentation. This patch removes the test and the doc
39408 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=14270
39410 2022-06-06 Tom de Vries <tdevries@suse.de>
39412 [gdb] Name arch selftests more clearly
39413 When running some all archs selftest I get:
39415 $ gdb -q -batch -ex "maint selftest unpack_field_as_long"
39416 Running selftest unpack_field_as_long::A6.
39419 By now I know that A6 is an arc architecture, but for others that's less
39422 Fix this by using unpack_field_as_long::arc::A6 instead.
39424 This then introduces redundant names like arm::arm, so try to avoid those,
39425 though I'm not entirely convinced that that's worth the trouble.
39427 This introduces the following new names:
39429 +Running selftest unpack_field_as_long::am33_2::am33-2.
39430 +Running selftest unpack_field_as_long::arc::A6.
39431 +Running selftest unpack_field_as_long::arc::A7.
39432 +Running selftest unpack_field_as_long::arc::EM.
39433 +Running selftest unpack_field_as_long::arc::HS.
39434 +Running selftest unpack_field_as_long::arm::ep9312.
39435 +Running selftest unpack_field_as_long::arm::iwmmxt.
39436 +Running selftest unpack_field_as_long::arm::iwmmxt2.
39437 +Running selftest unpack_field_as_long::arm::xscale.
39438 +Running selftest unpack_field_as_long::bpf::xbpf.
39439 +Running selftest unpack_field_as_long::frv::fr400.
39440 +Running selftest unpack_field_as_long::frv::fr450.
39441 +Running selftest unpack_field_as_long::frv::fr500.
39442 +Running selftest unpack_field_as_long::frv::fr550.
39443 +Running selftest unpack_field_as_long::frv::simple.
39444 +Running selftest unpack_field_as_long::frv::tomcat.
39445 +Running selftest unpack_field_as_long::iq2000::iq10.
39446 +Running selftest unpack_field_as_long::m32c::m16c.
39447 +Running selftest unpack_field_as_long::mep::c5.
39448 +Running selftest unpack_field_as_long::mep::h1.
39449 +Running selftest unpack_field_as_long::nds32::n1.
39450 +Running selftest unpack_field_as_long::nds32::n1h.
39451 +Running selftest unpack_field_as_long::nds32::n1h_v2.
39452 +Running selftest unpack_field_as_long::nds32::n1h_v3.
39453 +Running selftest unpack_field_as_long::nds32::n1h_v3m.
39454 +Running selftest unpack_field_as_long::z80::ez80-adl.
39455 +Running selftest unpack_field_as_long::z80::ez80-z80.
39456 +Running selftest unpack_field_as_long::z80::gbz80.
39457 +Running selftest unpack_field_as_long::z80::r800.
39458 +Running selftest unpack_field_as_long::z80::z180.
39461 Tested on x86_64-linux.
39463 2022-06-06 Tom de Vries <tdevries@suse.de>
39465 [gdb] Enable some more print_one_insn selftests
39466 In print_one_insn_test we have this cluster of skipped tests:
39468 case bfd_arch_ia64:
39470 case bfd_arch_mips:
39471 case bfd_arch_tic6x:
39472 case bfd_arch_xtensa:
39476 Enable some of these, and document in more detail why they're enabled or
39479 Likewise, document bfd_arch_or1k because it's an odd case.
39481 Tested on x86_64-linux.
39483 2022-06-06 Tom de Vries <tdevries@suse.de>
39485 [gdb] Fix maint selftest -v print_one_insn
39486 When running the print_one_insn selftests with -v, I get:
39488 $ gdb -q -batch -ex "maint selftest -v print_one_insn"
39489 Running selftest print_one_insn::A6.
39490 .shor 0x783eRunning selftest print_one_insn::A7.
39491 trap_s 0x1Running selftest print_one_insn::ARC600.
39492 .shor 0x783eRunning selftest print_one_insn::ARC601.
39493 Running selftest print_one_insn::ARC700.
39494 trap_s 0x1Running selftest print_one_insn::ARCv2.
39495 trap_s 0x1Running selftest print_one_insn::EM.
39496 trap_s 0x1Running selftest print_one_insn::HS.
39497 trap_s 0x1Running selftest print_one_insn::Loongarch32.
39500 The insn is written to gdb_stdout, and there is code in the selftest to add a
39501 newline after the insn, which writes to stream().
39503 The stream() ui_file points into a string buffer, which the disassembler uses
39504 before writing to gdb_stdout, so writing into it after the disassembler has
39505 finished has no effect.
39507 Fix this by using gdb_stdlog and debug_printf (which is what the unit test
39508 infrastructure itself uses) instead, such that we have:
39510 Running selftest print_one_insn::A6.
39512 Running selftest print_one_insn::A7.
39514 Running selftest print_one_insn::ARC600.
39516 Running selftest print_one_insn::ARC601.
39517 Running selftest print_one_insn::ARC700.
39519 Running selftest print_one_insn::ARCv2.
39521 Running selftest print_one_insn::Loongarch32.
39524 Note: I've also removed the printing of arch_name, which would give
39525 us otherwise the redundant:
39527 Running selftest print_one_insn::A6.
39529 Running selftest print_one_insn::A7.
39533 Tested on x86_64-linux.
39535 2022-06-06 Andrew Burgess <aburgess@redhat.com>
39537 gdb/testsuite: add missing skip_python_tests call in py-doc-reformat.exp
39540 commit 51e8dbe1fbe7d8955589703140ca5eba7b4f1bd7
39541 Date: Mon May 16 19:26:54 2022 +0100
39543 gdb/python: improve formatting of help text for user defined commands
39545 the test that was added (gdb.python/py-doc-reformat.exp) was missing a
39546 call to skip_python_tests. As a result, this test would fail for any
39547 GDB built within Python support.
39549 This commit adds a call to skip_python_tests.
39551 2022-06-06 GDB Administrator <gdbadmin@sourceware.org>
39553 Automatic date update in version.in
39555 2022-06-05 Tom Tromey <tom@tromey.com>
39557 Remove obsolete Python 2 comment
39558 I found a comment that referred to Python 2, but that is now obsolete
39559 -- the code it refers to is gone. I'm checking in this patch to
39560 remove the comment.
39562 There's a similar comment elsewhere, but I plan to remove that one in
39563 another patch I'm going to submit shortly.
39565 2022-06-05 GDB Administrator <gdbadmin@sourceware.org>
39567 Automatic date update in version.in
39569 2022-06-04 Alan Modra <amodra@gmail.com>
39571 asan: null dereference in coff_count_linenumbers
39572 * coffgen.c (coff_count_linenumbers): Don't segfault when asymbol
39575 asan: uninitialised write in bfd_mach_o_write_contents
39576 * mach-o.c (bfd_mach_o_write_contents): Always set
39577 bfd_mach_o_dyld_info_command *_off fields.
39579 2022-06-04 Tom de Vries <tdevries@suse.de>
39581 [gdb/ada] Fix literal truncation
39582 Make sure we error out on overflow instead of truncating in all cases.
39584 Tested on x86_64-linux, with a build with --enable-targets=all.
39586 2022-06-04 Tom de Vries <tdevries@suse.de>
39588 [gdb/m2] Fix UB and literal truncation
39589 Rewrite parse_number to use ULONGEST instead of LONGEST, to fix UB errors as
39590 mentioned in PR29163.
39592 Furthermore, make sure we error out on overflow instead of truncating in all
39595 Tested on x86_64-linux, with a build with --enable-targets=all.
39597 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29163
39599 2022-06-04 Tom de Vries <tdevries@suse.de>
39601 [gdb/rust] Fix literal truncation
39602 Make sure we error out on overflow instead of truncating in all cases.
39604 I've used as overflow string: "Integer literal is too large", based
39606 <rust-lang/rust>/src/test/ui/parser/int-literal-too-large-span.rs
39607 but perhaps someone has a better idea.
39609 Tested on x86_64-linux, with a build with --enable-targets=all.
39611 2022-06-04 Tom de Vries <tdevries@suse.de>
39613 [gdb/pascal] Fix literal truncation
39614 Make sure we error out on overflow instead of truncating in all cases.
39616 The current implementation of parse_number contains a comment about PR16377,
39617 but that's related to C-like languages. In absence of information of whether
39618 the same fix is needed for pascal, take the conservative approach and keep
39619 behaviour for decimals unchanged.
39621 Tested on x86_64-linux, with a build with --enable-targets=all.
39623 2022-06-04 Tom de Vries <tdevries@suse.de>
39625 [gdb/go] Fix literal truncation
39626 Make sure we error out on overflow instead of truncating in all cases.
39628 The current implementation of parse_number contains a comment about PR16377,
39629 but that's related to C-like languages. In absence of information of whether
39630 the same fix is needed for go, take the conservative approach and keep
39631 behaviour for decimals unchanged.
39633 Tested on x86_64-linux, with a build with --enable-targets=all.
39635 2022-06-04 Tom de Vries <tdevries@suse.de>
39637 [gdb/fortran] Fix literal truncation
39638 As mentioned in commit 5b758627a18 ("Make gdb.base/parse_number.exp test all
39641 There might be a bug that 32-bit fortran truncates 64-bit values to
39642 32-bit, given "p/x 0xffffffffffffffff" returns "0xffffffff".
39645 More concretely, we have:
39647 $ for arch in i386:x86-64 i386; do \
39648 gdb -q -batch -ex "set arch $arch" -ex "set lang fortran" \
39649 -ex "p /x 0xffffffffffffffff"; \
39651 The target architecture is set to "i386:x86-64".
39652 $1 = 0xffffffffffffffff
39653 The target architecture is set to "i386".
39657 Fix this by adding a range check in parse_number in gdb/f-exp.y.
39659 Furthermore, make sure we error out on overflow instead of truncating in all
39662 Tested on x86_64-linux.
39664 2022-06-04 Tom de Vries <tdevries@suse.de>
39666 [gdb/c] Fix type of 2147483648 and literal truncation
39667 [ Assuming arch i386:x86-64, sizeof (int) == 4,
39668 sizeof (long) == sizeof (long long) == 8. ]
39670 Currently we have (decimal for 0x80000000):
39672 (gdb) ptype 2147483648
39673 type = unsigned int
39676 According to C language rules, unsigned types cannot be used for decimal
39677 constants, so the type should be long instead (reported in PR16377).
39679 Fix this by making sure the type of 2147483648 is long.
39681 The next interesting case is (decimal for 0x8000000000000000):
39683 (gdb) ptype 9223372036854775808
39684 type = unsigned long
39687 According to the same rules, unsigned long is incorrect.
39689 Current gcc uses __int128 as type, which is allowed, but we don't have that
39690 available in gdb, so the strict response here would be erroring out with
39693 Older gcc without __int128 support, as well as clang use an unsigned type, but with
39694 a warning. Interestingly, clang uses "unsigned long long" while gcc uses
39695 "unsigned long", which seems the better choice.
39697 Given that the compilers allow this as a convience, do the same in gdb
39698 and keep type "unsigned long", and make this explicit in parser and test-case.
39700 Furthermore, make sure we error out on overflow instead of truncating in all
39703 Tested on x86_64-linux with --enable-targets=all.
39705 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=16377
39707 2022-06-04 Tom de Vries <tdevries@suse.de>
39709 [gdb/testsuite] Test more values in gdb.base/parse_numbers.exp
39710 Currently we only test value 0xffffffffffffffff in test-case
39711 gdb.base/parse_numbers.exp.
39713 Test more interesting values, both in decimal and hex format, as well as
39714 negative decimals for language modula-2.
39716 This results in an increase in total tests from 15572 to 847448 (55 times
39719 Balance out the increase in runtime by reducing the number of architectures
39720 tested: only test one architecture per sizeof longlong/long/int/short
39721 combination, while keeping the possibility intact to run with all
39722 architectures (through setting a variable in the test-case)
39724 Results in slight reduction of total tests: 15572 -> 13853.
39726 Document interesting cases in the expected results:
39727 - wrapping from unsigned to signed
39729 - PR16377: using unsigned types to represent decimal constants in C
39731 Running the test-case with a gdb build with -fsanitize=undefined, we trigger
39732 two UB errors in the modula-2 parser, filed as PR29163.
39734 Tested on x86_64-linux with --enable-targets=all.
39736 2022-06-04 Tom de Vries <tdevries@suse.de>
39738 [gdb/testsuite] Fix ERROR in gdb.ctf/funcreturn.exp
39739 On openSUSE Tumbleweed (with gcc-12, enabling ctf tests) I run into:
39741 ERROR: tcl error sourcing src/gdb/testsuite/gdb.ctf/funcreturn.exp.
39742 ERROR: tcl error code NONE
39743 ERROR: Unexpected arguments: \
39744 {print v_double_func} \
39745 {[0-9]+ = {double \(\)} 0x[0-9a-z]+.*} \
39746 {print double function} \
39750 The problem is a curly brace as fourth argument to gdb_test, which errors out
39751 due to recently introduced more strict argument checking in gdb_test.
39753 Fix the error by removing the brace.
39755 Though this fixes the error for me, due to PR29160 I get only FAILs, so I can't
39756 claim proper testing on x86_64-linux.
39758 2022-06-04 Tom de Vries <tdevries@suse.de>
39760 [gdb/testsuite] Fix gdb.threads/manythreads.exp with check-read1
39761 When running test-case gdb.threads/manythreads.exp with check-read1, I ran
39762 into this hard-to-reproduce FAIL:
39764 [New Thread 0x7ffff7318700 (LWP 31125)]^M
39765 [Thread 0x7ffff7321700 (LWP 31124) exited]^M
39768 Thread 769 "manythreads" received signal SIGINT, Interrupt.^M
39769 [Switching to Thread 0x7ffff6d66700 (LWP 31287)]^M
39770 0x00007ffff7586a81 in clone () from /lib64/libc.so.6^M
39771 (gdb) FAIL: gdb.threads/manythreads.exp: stop threads 1
39774 The matching in the failing gdb_test_multiple is done in an intricate way,
39775 trying to pass on some order and fail on another order.
39777 Fix this by rewriting the regexps to match one line at most, and detecting
39778 invalid order by setting and checking state variables.
39780 Tested on x86_64-linux.
39782 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29177
39784 2022-06-04 Tom de Vries <tdevries@suse.de>
39786 [gdb] Fix warning in print_one_insn::ez80-adl
39787 When running selftest print_one_insn::ez80-adl we run into this warning:
39789 Running selftest print_one_insn::ez80-adl.
39790 warning: Unable to determine inferior's software breakpoint type: couldn't
39791 find `_break_handler' function in inferior. Will be used default software \
39792 breakpoint instruction RST 0x08.
39795 Fix this by explicitly handling bfd_arch_z80 in print_one_insn_test.
39797 Tested on x86_64-linux.
39799 2022-06-04 GDB Administrator <gdbadmin@sourceware.org>
39801 Automatic date update in version.in
39803 2022-06-03 Tom Tromey <tromey@adacore.com>
39805 Use bool for evregpy_no_listeners_p
39806 I noticed that evregpy_no_listeners_p should return a bool. This
39807 patch makes this change. I'm checking it in.
39809 2022-06-03 Alan Modra <amodra@gmail.com>
39811 asan: heap buffer overflow in _bfd_mips_elf_section_from_shdr
39812 * elfxx-mips.c (_bfd_mips_elf_section_from_shdr): Sanity check
39813 intopt.size and remaining bytes in section for reginfo.
39815 2022-06-03 Alan Modra <amodra@gmail.com>
39817 Re: ubsan: undefined shift in frag_align_code
39818 This one needs the same fix too.
39820 * config/tc-i386.h (MAX_MEM_FOR_RS_ALIGN_CODE): Avoid signed
39823 2022-06-03 Tom de Vries <tdevries@suse.de>
39825 [gdb] Fix warning in foreach_arch selftests
39826 When running the selftests, I run into:
39828 $ gdb -q -batch -ex "maint selftest"
39830 Running selftest execute_cfa_program::aarch64:ilp32.
39831 warning: A handler for the OS ABI "GNU/Linux" is not built into this
39832 configuration of GDB. Attempting to continue with the default aarch64:ilp32
39835 and likewise for execute_cfa_program::i8086 and
39836 execute_cfa_program::ia64-elf32.
39838 The warning can easily be reproduced outside the selftests by doing:
39840 $ gdb -q -batch -ex "set arch aarch64:ilp32"
39842 and can be prevented by first doing "set osabi none".
39844 Fix the warning by setting osabi to none while doing selftests that iterate
39845 over all architectures.
39847 This causes a regression in the print_one_insn selftests for the ARC
39850 The problem is pre-existing, and can be demonstrated (already without this
39853 $ gdb -q -batch -ex "set osabi none" -ex "maint selftest print_one_insn::A6"
39854 Running selftest print_one_insn::A6.
39855 Self test failed: Cannot access memory at address 0x0
39856 Ran 1 unit tests, 1 failed
39860 For ARC, we use the generic case in print_one_insn_test, containing this code:
39862 int kind = gdbarch_breakpoint_kind_from_pc (gdbarch, &pc);
39864 insn = gdbarch_sw_breakpoint_from_kind (gdbarch, kind, &bplen);
39867 The problem is that with osabi linux we trigger:
39870 arc_linux_breakpoint_kind_from_pc (struct gdbarch *gdbarch, CORE_ADDR *pcptr)
39875 but with osabi none:
39877 arc_breakpoint_kind_from_pc (struct gdbarch *gdbarch, CORE_ADDR *pcptr)
39879 size_t length_with_limm = gdb_insn_length (gdbarch, *pcptr);
39881 which needs access to memory, and will consequently fail.
39883 Fix this in print_one_insn_test, in the default case, by iterating over
39884 supported osabi's to makes sure we trigger arc_linux_breakpoint_kind_from_pc
39885 which will give us a usable instruction to disassemble.
39887 Tested on x86_64-linux.
39889 2022-06-03 Tom de Vries <tdevries@suse.de>
39891 Revert "[gdb] Fix warning in foreach_arch selftests"
39892 This reverts commit fc18b1c5afd ("[gdb] Fix warning in foreach_arch
39895 The commit introduced regressions for an --enable-targets=all build:
39897 Running selftest print_one_insn::A6.^M
39898 Self test failed: Cannot access memory at address 0x0^M
39900 and while investigating those I realized that the commit fc18b1c5afd
39901 complicates things by trying to set the current osabi.
39903 So, revert the patch in preparation for a simpler solution.
39905 Tested on x86_64-linux.
39907 2022-06-03 Jan Beulich <jbeulich@suse.com>
39909 x86: exclude certain ISA extensions from v3/v4 ISA
39910 Like TBM and LWP, XOP and FMA4 also shouldn't be included in v3.
39912 Like AVX512-4VNNIW, AVX512-4FMAPS also shouldn't be included in v4.
39914 2022-06-03 Roland McGrath <mcgrathr@google.com>
39916 gdb: LoongArch: Remove nonportable #include
39917 Don't use gregset.h in *-tdep.c since it's not usable on
39918 hosts that don't have <sys/procfs.h>. It's not needed by
39919 this file, and should only be needed by *-nat.c files.
39921 2022-06-03 Alan Modra <amodra@gmail.com>
39923 Re: asan: mips_gprel_reloc segfault
39924 Similarly for the elf mips support.
39926 * elf32-mips.c (mips_elf_final_gp): Don't segfault on symbols
39927 in any of the bfd_is_const_section sections.
39928 * elf64-mips.c (mips_elf64_final_gp): Likewise.
39929 * elfn32-mips.c (mips_elf_final_gp): Likewise.
39931 2022-06-03 Alan Modra <amodra@gmail.com>
39933 asan: mips_gprel_reloc segfault
39934 Not just the undefined section has a NULL owner, the absolute section
39935 has too. Which means we can't find output_bfd for __gp. Also, may as
39936 well test directly for output_bfd == NULL.
39938 * coff-mips.c (mips_gprel_reloc): Don't segfault on any of
39939 bfd_is_const_section sections.
39941 2022-06-03 GDB Administrator <gdbadmin@sourceware.org>
39943 Automatic date update in version.in
39945 2022-06-02 Tom de Vries <tdevries@suse.de>
39947 [gdb/testsuite] Detect change instead of init in gdb.mi/mi-var-block.exp
39948 On openSUSE Tumbleweed with target board unix/-m32, I run into:
39950 PASS: gdb.mi/mi-var-block.exp: step at do_block_test 2
39951 Expecting: ^(-var-update \*[^M
39952 ]+)?(\^done,changelist=\[{name="foo",in_scope="true",type_changed="false",has_more="0"},
39953 {name="cb",in_scope="true",type_changed="false",has_more="0"}\][^M
39957 ^done,changelist=[{name="foo",in_scope="true",type_changed="false",has_more="0"}]^M
39959 FAIL: gdb.mi/mi-var-block.exp: update all vars: cb foo changed (unexpected output)
39962 The problem is that the test-case attempts to detect a change in the cb
39963 variable caused by this initialization:
39970 but that only works if the stack location happens to be unequal to 12 before
39971 the initialization.
39973 Fix this by first initializing to 0, and then changing the value to 12:
39979 and detecting that change.
39981 Tested on x86_64-linux.
39983 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29195
39985 2022-06-02 Eli Zaretskii <eliz@gnu.org>
39987 Rearrange and slightly reword the "Location Specification" section
39988 This rearranges and changes the wording of the "Location
39989 Specification" section of the GDB manual in minor ways.
39991 2022-06-02 Tom Tromey <tromey@adacore.com>
39993 ODR warning for "main"
39994 "main" is redeclared with a different type in maint.c. I think this
39995 might have come from my first gdb patch, many many years ago. While I
39996 wonder if this profiling code is actually useful at all any more, in
39997 the meantime it's simple to fix the declaration.
39999 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=22395
40001 2022-06-02 Tom Tromey <tromey@adacore.com>
40003 ODR warnings for "struct coff_symbol"
40004 "struct coff_symbol" is defined in multiple .c files, causing ODR
40005 warnings. This patch renames just the xcoffread.c type.
40007 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=22395
40009 2022-06-02 Tom Tromey <tromey@adacore.com>
40011 ODR warnings for "struct insn_decode_record_t"
40012 "struct insn_decode_record_t" is defined in multiple .c files, causing
40013 ODR warnings. This patch renames the types, and removes the use of
40014 "typedef" here -- this is a C-ism that's no longer needed.
40016 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=22395
40018 2022-06-02 Tom Tromey <tromey@adacore.com>
40020 ODR warnings for "struct insn_info"
40021 "struct insn_info" is defined in multiple .c files, causing ODR
40022 warnings. This patch renames the type in z80-tdep.c, leaving the
40025 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=22395
40027 2022-06-02 Tom Tromey <tromey@adacore.com>
40029 ODR warnings from overlay constants
40030 Some overlay-related constants are duplicated in z80-tdep.c, causing
40031 ODR warnings. This patch renames just the z80-specific ones.
40033 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=22395
40035 2022-06-02 Tom Tromey <tromey@adacore.com>
40037 ODR warning for "enum string_repr_result"
40038 "enum string_repr_result" is defined in multiple .c files, causing ODR
40039 warnings. This patch renames the types.
40041 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=22395
40043 2022-06-02 Tom Tromey <tromey@adacore.com>
40045 ODR warning for "struct find_targ_sec_arg"
40046 "struct find_targ_sec_arg" is defined in multiple .c files, causing
40047 ODR warnings. This patch renames the types.
40049 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=22395
40051 2022-06-02 Tom Tromey <tromey@adacore.com>
40053 ODR warning for "struct stack_item"
40054 "struct stack_item" is defined in multiple .c files, causing ODR
40055 warnings. This patch renames these types.
40057 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=22395
40059 2022-06-02 Tom Tromey <tromey@adacore.com>
40061 ODR warning for "struct instruction_type"
40062 "struct instruction_type" is defined in multiple .c files, causing an
40063 ODR warning. This patch renames the types.
40065 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=22395
40067 2022-06-02 Tom Tromey <tromey@adacore.com>
40069 ODR warning for struct ext_link_map
40070 This renames the solib-dsbt.c copy of "struct ext_link_map" to avoid
40073 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=22395
40075 2022-06-02 Tom Tromey <tromey@adacore.com>
40077 ODR warning for struct field_info
40078 This renames one of the instance of "struct field_info" to avoid an
40081 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=22395
40083 2022-06-02 Tom Tromey <tromey@adacore.com>
40085 ODR warnings for struct nextfield
40086 "struct nextfield" is defined in multiple places in GDB. This patch
40087 renames just the stabs one, leaving the DWARF one untouched.
40089 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=22395
40091 2022-06-02 Tom Tromey <tromey@adacore.com>
40093 ODR warnings for struct symloc
40094 "struct symloc" is defined in multiple spots in gdb, causing ODR
40095 warnings. This patch renames these.
40097 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=22395
40099 2022-06-02 Tom Tromey <tromey@adacore.com>
40101 Fix ODR warning in observable.h
40102 observable.h triggers an ODR warning because this line:
40104 extern observable<struct target_ops */* target */> target_changed;
40106 ... may be the only declaration of "struct target_ops" in scope
40107 (depending on the particular .c file) -- and this declares it in a
40108 namespace, resulting in confusion.
40110 This patch fixes the problem by adding a forward declaration.
40112 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=22395
40114 2022-06-02 Tiezhu Yang <yangtiezhu@loongson.cn>
40116 gdb: LoongArch: Implement the software_single_step gdbarch method
40117 When execute the following command on LoongArch:
40119 make check-gdb TESTS="gdb.base/branch-to-self.exp"
40121 there exist the following failed testcases:
40123 FAIL: gdb.base/branch-to-self.exp: single-step: si (timeout)
40124 FAIL: gdb.base/branch-to-self.exp: break-cond: side=host: continue to breakpoint: continue to break (timeout)
40125 FAIL: gdb.base/branch-to-self.exp: break-cond: side=host: p counter (timeout)
40127 Implement the software_single_step gdbarch method to decode the current
40128 branch instruction and determine the address of the next instruction on
40129 LoongArch to fix the above failed testcases.
40131 2022-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
40133 gdb: Do not add empty sections to the section map
40134 From: Ulrich Weigand <ulrich.weigand@de.ibm.com>
40136 build_objfile_section_table () creates four synthetic sections per
40137 objfile, which are collected by update_section_map () and passed to
40138 std::sort (). When there are a lot of objfiles, for example, when
40139 debugging JITs, the presence of these sections slows down the sorting
40142 The output of update_section_map () is used by find_pc_section (),
40143 which can never return any of these sections: their size is 0, so they
40144 cannot be accepted by bsearch_cmp ().
40146 Filter them (and all the other empty sections) out in
40147 insert_section_p (), which is used only by update_section_map ().
40149 2022-06-02 Jon Turney <jon.turney@dronecode.org.uk>
40151 Fix a new warning on Cygwin
40152 > ../../gdb/windows-nat.c: In function ‘windows_solib* windows_make_so(const char*, LPVOID)’:
40153 > ../../gdb/windows-nat.c:714:12: error: declaration of ‘char name [512]’ shadows a parameter [-Werror=shadow=compatible-local]
40154 > 714 | char name[SO_NAME_MAX_PATH_SIZE];
40156 > ../../gdb/windows-nat.c:655:30: note: shadowed declaration is here
40157 > 655 | windows_make_so (const char *name, LPVOID load_addr)
40158 > | ~~~~~~~~~~~~^~~~
40160 Fix Cygwin build after 85b25bd9
40161 Fix Cygwin build after 85b25bd9 ("Simplify windows-nat.c solib handling").
40163 Fix Cygwin build after 0578e87f
40164 Fix Cygwin build after 0578e87f ("Remove some globals from
40165 nat/windows-nat.c"). Update code under ifdef __CYGWIN__ for globals
40166 moved to members of struct windows_process_info.
40168 Fix Cygwin build after fcab5839
40169 Fix Cygwin build after fcab5839 ("Implement pid_to_exec_file for Windows
40170 in gdbserver"). That change moves code from gdb/windows-nat.c to
40171 gdb/nat/windows-nat.c, but doesn't add the required typedefs and
40172 includes for parts of that code under ifdef __CYGWIN__.
40174 2022-06-02 Alan Modra <amodra@gmail.com>
40176 Re: ubsan: signed integer overflow in atof_generic
40179 * atof-generic.c: Include limits.h.
40181 2022-06-02 Alan Modra <amodra@gmail.com>
40183 ubsan: signed integer overflow in atof_generic
40184 Fix the signed overflows by using unsigned variables and detect
40185 overflow at BUG! comment.
40187 * atof-generic.c (atof_generic): Avoid signed integer overflow.
40188 Return ERROR_EXPONENT_OVERFLOW if exponent overflows a long.
40190 2022-06-02 Alan Modra <amodra@gmail.com>
40192 asan: uninit write _bfd_ecoff_write_object_contents
40193 * ecoff.c (_bfd_ecoff_write_object_contents): zalloc reloc_buff.
40195 asan: null deref in coff_write_relocs
40196 * coffcode.h (coff_write_relocs): Don't deref NULL howto.
40198 ubsan: undefined shift in frag_align_code
40199 * frags.c (MAX_MEM_FOR_RS_ALIGN_CODE): Avoid signed integer
40202 2022-06-02 Alan Modra <amodra@gmail.com>
40204 gas read_a_source_file #APP processing
40205 This fixes some horrible code using do_scrub_chars. What we had ran
40206 text through do_scrub_chars twice, directly in read_a_source_file and
40207 again via the input_scrub_include_sb call. That's silly, and since
40208 do_scrub_chars is a state machine, possibly wrong. More silliness is
40209 evident in the temporary malloc'd buffer for do_scrub_chars output,
40210 which should have been written directly to sbuf.
40212 So, get rid of the do_scrub_chars call and support functions, leaving
40213 scrubbing to input_scrub_include_sb. I did wonder about #NO_APP
40214 overlapping input_scrub_next_buffer buffers, but that should only
40215 happen if the string starts in one file and finishes in another.
40217 * read.c (scrub_string, scrub_string_end): Delete.
40218 (scrub_from_string): Delete.
40219 (read_a_source_file): Rewrite #APP processing.
40221 2022-06-02 Alan Modra <amodra@gmail.com>
40223 sb_scrub_and_add_sb not draining input string buffer
40224 It is possible for sb_scrub_and_add_sb to not consume all of the input
40225 string buffer. If this happens for reasons explained in the comment,
40226 do_scrub_chars can leave pointers to the string buffer for the next
40227 call. This patch fixes that by ensuring the input is drained. Note
40228 that the behaviour for an empty string buffer is also changed,
40229 avoiding another do_scrub_chars bug where empty input and single char
40230 sized output buffers could result in a write past the end of the
40233 sb.c (sb_scrub_and_add_sb): Loop until all of input sb is
40236 2022-06-02 Alan Modra <amodra@gmail.com>
40238 asan: heap buffer overflow in dwarf2_directive_filename
40239 Seen with .file 4294967289 "xxx.c"
40241 * dwarf2dbg.c (assign_file_to_slot): Catch more cases of integer
40242 overflow. Make param i an unsigned int.
40244 2022-06-02 Alan Modra <amodra@gmail.com>
40246 asan: NULL deref in scan_unit_for_symbols
40247 Since commit b43771b045 it has been possible to look up addresses
40248 that match a unit with errors, since ranges are added to a trie while
40249 the unit is being parsed. On error, parse_comp_unit leaves
40250 first_child_die_ptr NULL which results in a NULL info_ptr being passed
40251 to scan_unit_for_symbols. Fix this by setting unit->error.
40253 Also wrap some overlong lines, and fix some formatting errors.
40255 * dwarf2.c: Formatting.
40256 (parse_comp_unit): Set unit->error on err_exit path.
40258 2022-06-02 GDB Administrator <gdbadmin@sourceware.org>
40260 Automatic date update in version.in
40262 2022-06-01 Tom de Vries <tdevries@suse.de>
40264 [gdb] Fix warning in foreach_arch selftests
40265 When running the selftests, I run into:
40267 $ gdb -q -batch -ex "maint selftest"
40269 Running selftest execute_cfa_program::aarch64:ilp32.
40270 warning: A handler for the OS ABI "GNU/Linux" is not built into this
40271 configuration of GDB. Attempting to continue with the default aarch64:ilp32
40274 and likewise for execute_cfa_program::i8086 and
40275 execute_cfa_program::ia64-elf32.
40277 The warning can easily be reproduced outside the selftests by doing:
40279 $ gdb -q -batch -ex "set arch aarch64:ilp32"
40281 and can be prevented by first doing "set osabi none".
40283 Fix the warning by setting osabi to none while doing selftests that iterate
40284 over all architectures.
40286 Tested on x86_64-linux.
40288 2022-06-01 Tom Tromey <tromey@adacore.com>
40290 Add gdb.current_language and gdb.Frame.language
40291 This adds the gdb.current_language function, which can be used to find
40292 the current language without (1) ever having the value "auto" or (2)
40293 having to parse the output of "show language".
40295 It also adds the gdb.Frame.language, which can be used to find the
40296 language of a given frame. This is normally preferable if one has a
40297 Frame object handy.
40299 2022-06-01 Yvan Roux <yvan.roux@foss.st.com>
40301 [arm] Don't use special treatment for PC
40302 In an exception frame the PC register is extracted from the stack
40303 just like other base registers, so there is no need for a special
40306 [arm] Add support for FPU registers in prologue unwinder
40307 The prologue unwinder had support for FPU registers, but only to
40308 calculate the correct offset on the stack, the values were not saved.
40310 2022-06-01 Yvan Roux <yvan.roux@foss.st.com>
40312 [arm] d0..d15 are 64-bit each, not 32-bit
40313 When unwinding the stack, the floating point registers d0 to d15
40314 need to be handled as double words, not words.
40316 Only the first 8 registers have been confirmed fixed with this patch
40317 on a STM32F407-DISC0 board, but the upper 8 registers on Cortex-M33
40318 should be handled in the same way.
40320 The test consisted of running a program compiled with float-abi=hard.
40321 In the main function, a function taking a double as an argument was
40322 called. After the function call, a hardware timer was used to
40323 trigger an interrupt.
40325 In the debug session, a breakpoint was set in the function called
40326 from main to verify the content of the registers using "info float"
40327 and another breakpoint in the interrupt handler was used to check
40328 the same registers using "info float" on frame 2 (the frame just
40329 before the dummy frame created for the signal handler in gdb).
40331 2022-06-01 Yvan Roux <yvan.roux@foss.st.com>
40333 [arm] Cleanup: use hex for offsets
40334 Changed offset from decimal to hex to match architecture reference
40335 manual terminology and keep coherency with the rest of the code.
40337 2022-06-01 Jiangshuai Li <jiangshuai_li@c-sky.com>
40339 gdb:csky save fpu and vdsp info to struct csky_gdbarch_tdep
40340 First, add three variables fpu_abi, fpu_hardfp and vdsp_version
40341 to csky_gdbarch_tdep. They will be initialized from info.abfd in
40344 Now, they are just used to find a candidate among the list of pre-declared
40347 Later, they will be used in gdbarch_return_value and gdbarch_push_dummy_call
40348 for funtions described below:
40349 fpu_abi: to check if the bfd is using VAL_CSKY_FPU_ABI_HARD or
40350 VAL_CSKY_FPU_ABI_SOFT
40351 fpu_hardfp: to check if the bfd is using VAL_CSKY_FPU_HARDFP_SINGLE
40352 or VAL_CSKY_FPU_HARDFP_DOUBLE
40353 vdsp_version: to check if a function is returned with CSKY_VRET_REGNUM
40355 2022-06-01 Alan Modra <amodra@gmail.com>
40357 Re: use libiberty xmalloc in bfd/doc/chew.c
40358 We can't use libiberty.a in chew. libiberty is a host library, chew
40359 a build program. Partly revert commit 7273d78f3f7a, instead define
40360 local versions of the libiberty functions. ansidecl.h also isn't
40363 * doc/chew.c: Don't include libiberty.h or ansidecl.h.
40364 (xmalloc, xrealloc, xstrdup): New functions.
40365 * doc/local.mk (LIBIBERTY): Don't define or use.
40366 * Makefile.in: Regenerate.
40368 2022-06-01 GDB Administrator <gdbadmin@sourceware.org>
40370 Automatic date update in version.in
40372 2022-06-01 H.J. Lu <hjl.tools@gmail.com>
40374 x86: Properly handle IFUNC function pointer reference
40377 commit 68c4956b1401de70173848a6bdf620cb42fa9358
40378 Author: H.J. Lu <hjl.tools@gmail.com>
40379 Date: Tue Apr 26 09:08:54 2022 -0700
40381 x86: Properly handle function pointer reference
40383 to properly handle IFUNC function pointer reference. Since IFUNC symbol
40384 value is only known at run-time, set pointer_equality_needed for IFUNC
40385 function pointer reference in PDE so that it will be resolved to its PLT
40391 * elf32-i386.c (elf_i386_scan_relocs): Set pointer_equality_needed
40392 for IFUNC function pointer reference in PDE.
40393 * elf64-x86-64.c (elf_x86_64_scan_relocs): Likewise.
40398 * testsuite/ld-ifunc/ifunc.exp: Run PR ld/29216 test.
40399 * testsuite/ld-ifunc/pr29216.c: New file.
40401 2022-05-31 H.J. Lu <hjl.tools@gmail.com>
40403 i386: Ajdust more tests for opcodes/i386: remove trailing whitespace
40406 FAIL: Build ifunc-1a with -z ibtplt
40407 FAIL: Build ifunc-1a with PIE -z ibtplt
40408 FAIL: Build libno-plt-1b.so
40409 FAIL: No PLT (dynamic 1a)
40410 FAIL: No PLT (dynamic 1b)
40411 FAIL: No PLT (dynamic 1c)
40412 FAIL: No PLT (static 1d)
40413 FAIL: No PLT (PIE 1e)
40414 FAIL: No PLT (PIE 1f)
40415 FAIL: No PLT (PIE 1g)
40416 FAIL: No PLT (dynamic 1h)
40417 FAIL: No PLT (dynamic 1i)
40418 FAIL: No PLT (static 1j)
40420 * ld-i386/libno-plt-1b.dd: Remove trailing whitespaces.
40421 * ld-i386/no-plt-1a.dd: Likewise.
40422 * ld-i386/no-plt-1b.dd: Likewise.
40423 * ld-i386/no-plt-1c.dd: Likewise.
40424 * ld-i386/no-plt-1d.dd: Likewise.
40425 * ld-i386/no-plt-1e.dd: Likewise.
40426 * ld-i386/no-plt-1f.dd: Likewise.
40427 * ld-i386/no-plt-1g.dd: Likewise.
40428 * ld-i386/no-plt-1h.dd: Likewise.
40429 * ld-i386/no-plt-1i.dd: Likewise.
40430 * ld-i386/no-plt-1j.dd: Likewise.
40431 * ld-i386/plt-main-ibt.dd: Likewise.
40432 * ld-i386/plt-pie-ibt.dd: Likewise.
40434 2022-05-31 Tom Tromey <tom@tromey.com>
40436 Use unique_ptr for objfiles
40437 A while back, I changed objfiles to be held via a shared_ptr. The
40438 idea at the time was that this was a step toward writing to the index
40439 cache in the background, and this would let gdb keep a reference alive
40440 to do so. However, since then we've rewritten the DWARF reader, and
40441 the new index can do this without requiring a shared pointer -- in
40442 fact there are patches pending to implement this.
40444 This patch switches objfile management to unique_ptr, which makes more
40447 Regression tested on x86-64 Fedora 34.
40449 2022-05-31 Nils-Christian Kempke <nils-christian.kempke@intel.com>
40451 gdb/testsuite: fixup common-block.exp for intel compilers
40452 The order in which the variables in info common and info locals are
40453 displayed is compiler (and dwarf) dependent. While all symbols should
40454 be displayed the order is not fixed.
40456 I added a gdb_test_multiple that lets ifx and ifort pass in cases where
40457 only the order differs.
40459 2022-05-31 Nils-Christian Kempke <nils-christian.kempke@intel.com>
40461 gdb, testsuite, fortran: fixup mixed-lang-stack for Intel/LLVM compilers
40462 When value-printing a pointer within GDB by default GDB will look for
40463 defined symbols residing at the address of the pointer. For the given
40464 test the Intel/LLVM compiler stacks both display a symbol associated
40465 with a printed pointer while the gnu stack does not. This leads to
40466 failures in the test when running the test with CC_FOR_TARGET='clang'
40467 CXX_FOR_TARGET='clang' F90_FOR_TARGET='flang'"
40477 f = 0x419ed0 "abcdef"
40478 g = 0x4041a0 <.BSS4>
40480 or CC_FOR_TARGET='icx' CXX_FOR_TARGET='icpx' F90_FOR_TARGET='ifx'"
40490 f = 0x52eee0 "abcdef"
40491 g = 0x4ca210 <mixed_func_1a_$OBJ>
40493 For the compiled binary the Intel/LLVM compilers both decide to move the
40494 local variable g into the .bss section of their executable. The gnu
40495 stack will keep the variable locally on the stack and not define a
40498 Since the behavior for Intel/LLVM is actually expected I adapted the
40499 testcase at this point to be a bit more allowing for other outputs.
40500 I added the optional "<SYMBOLNAME>" to the regex testing for g.
40502 The given changes reduce the test fails for Intel/LLVM stack by 4 each.
40504 2022-05-31 Nils-Christian Kempke <nils-christian.kempke@intel.com>
40506 gdb, testsuite, fortran: fix double free in mixed-lang-stack.exp
40507 While testing mixed-lang-stack I realized that valgrind actually
40508 complained about a double free in the test.
40512 ==2503051== HEAP SUMMARY:
40513 ==2503051== in use at exit: 0 bytes in 0 blocks
40514 ==2503051== total heap usage: 26 allocs, 27 frees, 87,343 bytes allocated
40516 ==2503051== All heap blocks were freed -- no leaks are possible
40518 ==2503051== For lists of detected and suppressed errors, rerun with: -s
40519 ==2503051== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)
40521 Reason for this is that in mixed-lang-stack.cpp in mixed_func_1f an
40522 object "derived_type obj" goes on the stack which is then passed-by-value
40523 (so copied) to mixed_func_1g. The default copy-ctor will be called but,
40524 since derived_type contains a heap allocated string and the copy
40525 constructor is not implemented it will only be able to shallow copy the
40526 object. Right after each of the functions the object gets freed - on the
40527 other hand the d'tor of derived_type actually is implemented and calls
40528 free on the heap allocated string which leads to a double free. Instead
40529 of obeying the rule of 3/5 I just got rid of all that since it does not
40530 serve the test. The string is now just a const char* = ".." object
40533 2022-05-31 Nils-Christian Kempke <nils-christian.kempke@intel.com>
40535 testsuite, fortran: allow additional completions in module.exp
40536 For ifort, ifx, and flang the tests "complete modm" and "complete
40537 modmany" fail. This is because all three emit additional completion
40538 suggestions. These additional suggestions have their origin in symbols
40539 emitted by the compilers which can also be completed from the respective
40540 incomplete word (modm or modmany). For this specific example gfortran
40541 does not emit any additional symbols.
40543 For example, in this test the linkage name for var_a in ifx is
40544 "modmany_mp_var_a_" while gfortran uses "__modmany_MOD_var_a" instead.
40545 Since modmany_mp_var_a can be completed from modm and also modmany they
40546 will get displayed, while gfortran's symbol starts with "__" and thus will
40547 be ignored (it cannot be a completion of a word starting with "m").
40549 Similar things happen in flang and ifort. Some example output is shown
40553 (gdb) complete p modm
40562 (gdb) complete p modm
40569 p modmany_mp_var_a_
40570 p modmany_mp_var_b_
40571 p modmany_mp_var_c_
40572 p modmany_mp_var_i_
40575 (gdb) complete p modm
40582 I want to emphasize: for Fortran (and also C/C++) the complete command
40583 does not actually check whether its suggestions make sense - all it does
40584 is look for any symbol (in the minimal symbols, partial symbols etc.)
40585 that a given substring can be completed to (meaning that the given substring
40586 is the beginning of the symbol). One can easily produce a similar
40587 output for the gfortran compiled executable. For this look at the
40588 slightly modified "complete p mod" in gfortran:
40590 (gdb) complete p mod
40612 Many of the displayed symbols do not actually work with print:
40615 Attempt to use a type name as an expression
40617 No symbol "mod_1" in current context.
40620 I think that in the given test the output for gfortran only looks nice
40621 "by chance" rather than is actually expected. Expected is any output
40622 that also contains the completions
40631 while anythings else can be displayed as well (depending on the
40632 compiler and its emitted symbols).
40634 This, I'd consider all three outputs as valid and expected - one is just
40635 somewhat lucky that gfortran does not produce any additional symbols that
40638 The given patch improves test performance for all three compilers
40639 by allowing additional suggested completions inbetween and after
40640 the two given blocks in the test. I did not allow additional print
40641 within the modmany_list block since the output is ordered alphabetically
40642 and there should normally not appear any additional symbols there.
40644 For flang/ifx/ifort I each see 2 failures less (which are exactly the two
40647 As a side note and since I mentioned C++ in the beginning: I also tried
40648 the gdb.cp/completion.exp. The output seems a bit more reasonable,
40649 mainly since C++ actually has a demangler in place and linkage symbols
40650 do not appear in the output of complete. Still, with a poor enough
40651 to-be-completed string one can easily produce similar results:
40655 p typeinfo name for void
40656 p typeinfo name for void const*
40657 p typeinfo name for void*
40658 p typeinfo name for wchar_t
40659 p typeinfo name for wchar_t const*
40660 p typeinfo name for wchar_t*
40661 p t *** List may be truncated, max-completions reached. ***
40662 (gdb) p typeinfo name for void*
40663 No symbol "typeinfo" in current context.
40678 p build_charclass.isra
40679 (gdb) p bad_key_err
40680 No symbol "bad_key_err" in current context.
40682 (compiled with gcc/g++ and breaking at main).
40684 This patch is only about making the referenced test more 'fair' for the
40685 other compilers. Generally, I find the behavior of complete a bit
40686 confusing and maybe one wants to change this at some point but this
40687 would be a bigger task.
40689 2022-05-31 Nils-Christian Kempke <nils-christian.kempke@intel.com>
40691 testsuite, fortran: fix info-types for intel compilers
40692 This info-types.exp test case had a few issues that this patch fixes.
40694 First, the emitted symbol character(kind=1)/character*1 (different
40695 compilers use different naming converntions here) which is checkedin the
40696 test is not actually expected given the test program. There is no
40697 variable of that type in the test. Still, gfortran emits it for every
40698 Fortran program there is. The reason is the way gfortran handles Fortran's
40699 named main program. It generates a wrapper around the Fortran program
40700 that is quite similar to a C main function. This C-like wrapper has
40701 argc and argv arguments for command line argument passing and the argv
40702 pointer type has a base type character(kind=1) DIE emitted at CU scope.
40709 the degbug info gfortran emits looks somewhat like
40711 <0><c>: Abbrev Number: 3 (DW_TAG_compile_unit)
40713 <1><2f>: Abbrev Number: 4 (DW_TAG_subprogram)
40714 <30> DW_AT_external : 1
40715 <30> DW_AT_name : (indirect string, ...): main
40717 <2><51>: Abbrev Number: 1 (DW_TAG_formal_parameter)
40718 <52> DW_AT_name : (indirect string, ...): argc
40720 <2><5d>: Abbrev Number: 1 (DW_TAG_formal_parameter)
40721 <5e> DW_AT_name : (indirect string, ...): argv
40723 <62> DW_AT_type : <0x77>
40725 <2><6a>: Abbrev Number: 0
40727 <1><77>: Abbrev Number: 6 (DW_TAG_pointer_type)
40728 <78> DW_AT_byte_size : 8
40729 <79> DW_AT_type : <0x7d>
40730 <1><7d>: Abbrev Number: 2 (DW_TAG_base_type)
40731 <7e> DW_AT_byte_size : 1
40732 <7f> DW_AT_encoding : 8 (unsigned char)
40733 <80> DW_AT_name : (indirect string, ...): character(kind=1)
40734 <1><84>: Abbrev Number: 7 (DW_TAG_subprogram)
40735 <85> DW_AT_name : (indirect string, ...): prog
40738 Ifx and flang do not emit any debug info for a wrapper main method so
40739 the type is missing here. There was the possibility of actually adding
40740 a character*1 type variable to the Fortran executable, but both, ifx and
40741 gfortran chose to emit this variable's type as a DW_TAG_string_type of
40742 length one (instead of a character(kind=1), or whatever the respective
40743 compiler naming convention is). While string types are printed as
40744 character*LENGHT in the fortran language part (e.g. when issuing a
40745 'ptype') they do not generate any symbols inside GDB. In read.c it says
40747 /* These dies have a type, but processing them does not create
40748 a symbol or recurse to process the children. Therefore we can
40749 read them on-demand through read_type_die. */
40751 So they did not add any output to 'info types'. Only flang did emit a
40752 character type here.
40753 As adding a type would have a) not solved the problem for ifx and would
40754 have b) somehow hidden the curious behavior of gfortran, instead, the
40755 check for this character type was chagened to optional with the
40756 check_optional_entry to allow for the symbols's absence and to allow
40757 flang and ifx to pass this test as well.
40759 Second, the line checked for s1 was hardcoded as 37 in the test. Given
40760 that the type is actually defined on line 41 (which is what is emitted by
40761 ifx) it even seems wrong. The line check for s1 was changed to actually
40762 check for 41 and a gfortran bug has been filed here
40764 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105454
40766 The test is now marked as xfail for gfortran.
40768 Third, the whole test of checking for the 'Type s1' in info types seemed
40769 questionable. The type s1 is declared iside the scope of the Fortran
40770 program info_types_test. Its DIE however is emitted as a child of the
40771 whole compilation unit making it visible outside of the program's scope.
40772 The 'info types' command checks for types stored in the GLOBAL_BLOCK,
40773 or STATIC_BLOCKm wgucm according to block.h
40775 The GLOBAL_BLOCK contains all the symbols defined in this compilation
40776 whose scope is the entire program linked together.
40777 The STATIC_BLOCK contains all the symbols whose scope is the
40778 entire compilation excluding other separate compilations.
40780 so for gfortran, the type shows up in the output of 'info types'. For
40781 flang and ifx on the other hand this is not the case. The two compilers
40782 emit the type (correctly) as a child of the Fortran program, thus not
40783 adding it to either, the GLOBAL_BLOCK nor the LOCAL_BLOCK. A bug has
40784 been opened for the gfortran scoping issue:
40786 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105454
40788 While the most correct change might have been removing the check for s1,
40789 the change made here was to only check for this type in case of gfortran
40790 being used as the compiler, as this check also covers the declaration
40791 line issue mentioned above. A comment was added to maybe remove this
40792 check once the scoping issue is resolved (and it starts to fail with
40793 newer gfortran versions). The one used to test these changes was 13.0.
40795 2022-05-31 Nils-Christian Kempke <nils-christian.kempke@intel.com>
40797 testsuite/lib: add check_optional_entry for GDBInfoSymbols
40798 There was already a similar functionality for the GDBInfoModuleSymbols.
40799 This just extends the GDBInfoSymbols. We will use this feature in a
40800 later commit to make a testcase less GNU specific and more flexible for
40803 Namely, in gdb.fortran/info-types.exp currenlty
40804 GDBInfoSymbols::check_entry is used to verify and test the output of the
40805 info symbols command. The test, however was written with gfortran as a
40806 basis and some of the tests are not fair with e.g. ifx and ifort as
40807 they test for symbols that are not actually required to be emitted. The
40809 GDBInfoSymbols::check_entry "${srcfile}" "" "${character1}"
40811 GDBInfoSymbols::check_entry "${srcfile}" "37" "Type s1;"
40813 check for types that are either not used in the source file (character1)
40814 or should not be emitted by the compiler at global scope (s1) thus no
40815 appearing in the info symbols command. In order to fix this we will
40816 later use the newly introduced check_optional_entry over check_entry.
40818 2022-05-31 Nils-Christian Kempke <nils-christian.kempke@intel.com>
40820 testsuite, fortran: Add '-debug-parameters all' when using ifx/ifort
40821 In order for ifx and ifort to emit all debug entries, even for unused
40822 parameters in modules we have to define the '-debug-parameters all' flag.
40824 This commit adds it to the ifx-*/ifort-* specific flags in gdb.exp.
40826 2022-05-31 Nils-Christian Kempke <nils-christian.kempke@intel.com>
40828 testsuite, fortran: add compiler dependent types to dynamic-ptype-whatis
40829 The test was earlier not using the compiler dependent type print system
40830 in fortran.exp. I changed this. It should generally improve the test
40831 performance for different compilers. For ifx and gfortran I do not see
40834 2022-05-31 Nils-Christian Kempke <nils-christian.kempke@intel.com>
40836 testsuite, fortran: add required external keyword
40837 Currenlty, ifx/ifort cannot compile the given executable as it is not
40838 valid Fortran. It is missing the external keyword on the
40839 no_arg_subroutine. Gfortran compiles the example but this is actually
40840 a bug and there is an open gcc ticket for this here:
40842 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=50377
40844 Adding the keyword does not change the gfortran compiling of the example.
40845 It will, however, prevent a future fail once 50377 has been addressed.
40847 2022-05-31 Nils-Christian Kempke <nils-christian.kempke@intel.com>
40849 gdb/testsuite: disable charset.exp for intel compilers
40850 The test specifically tests for the Fortran CHARACTER(KIND=4) which is
40851 not available in ifx/ifort.
40853 Since the other characters are also printed elsewhere, we disable this
40854 test for the unsupported compilers.
40856 2022-05-31 Nils-Christian Kempke <nils-christian.kempke@intel.com>
40858 gdb/testsuite: rename intel next gen c/cpp compilers
40859 The name for icx and icpx in the testsuite was earlier set to 'intel-*'
40860 by the compiler identification. This commit changes this to 'icx-*'.
40862 Note, that currently these names are not used within the testsuite so no
40863 tests have to be adapted here.
40865 2022-05-31 Cristian Sandu <cristian.sandu@intel.com>
40866 Nils-Christian Kempke <nils-christian.kempke@intel.com>
40868 gdb/testsuite: add Fortran compiler identification to GDB
40869 This commit adds a separate Fortran compiler identification mechanism to
40870 the testsuite, similar to the existing one for C/C++. Before this
40871 change, the options and version for the Fortran compiler specified when
40872 running the testsuite with F90_FOR_TARGET set, was detected via its
40873 respective C compiler. So running the testsuite as
40875 make check TEST=gdb.fortran/*.exp CC_FOR_TARGET=gcc F90_FOR_TARGET=ifx
40879 make check TEST=gdb.fortran/*.exp F90_FOR_TARGET=ifx
40881 would use the gcc compiler inside the procedures get_compiler_info and
40882 test_compiler_info to identify compiler flags and the compiler version.
40883 This could sometimes lead to unpredictable outputs. It also limited
40884 testsuite execution to combinations where C and Fortran compiler would
40885 come from the same family of compiers (gcc/gfortran, icc/ifort, icx/ifx,
40886 clang/flang ..). This commit enables GDB to detect C and Fortran
40887 compilers independently of each other.
40889 As most/nearly all Fortran compilers have a mechanism for preprocessing
40890 files in a C like fashion we added the exact same meachnism that already
40891 existed for C/CXX. We let GDB preprocess a file with the compilers
40892 Fortran preprocessor and evaluate the preprocessor defined macros in that
40895 This enables GDB to properly run heterogeneous combinations of C and
40896 Fortran compilers such as
40898 CC_FOR_TARGET='gcc' and F90_FOR_TARGET='ifort'
40900 or enables one to run the testsuite without specifying a C compiler as in
40902 make check TESTS=gdb.fortran/*.exp F90_FOR_TARGET='ifx'
40903 make check TESTS=gdb.fortran/*.exp F90_FOR_TARGET='flang'
40905 On the other hand this also requires one to always specify a
40906 identification mechanism for Fortran compilers in the compiler.F90 file.
40908 We added identification for GFORTRAN, FLANG (CLASSIC and LLVM) IFX,
40909 IFORT, and ARMFLANG for now.
40911 Classic and LLVM flang were each tested with their latest releases on
40912 their respective release pages. Both get recognized by the new compiler
40913 identification and we introduced the two names flang-classic and
40914 flang-llvm to distinguish the two. While LLVM flang is not quite mature
40915 enough yet for running the testsuite we still thought it would be a good
40916 idea to include it already. For this we added a case for the fortran_main
40917 procedure. LLVM flang uses 'MAIN__' as opposed to classic flang which
40920 We did not have the possibility to test ARMFLANG - the versioning scheme
40921 here was extracted from its latest online documentation.
40923 We changed the test_compiler_info procedure to take another optional
40924 argument, the language string, which will be passed though to the
40925 get_compiler_info procedure. Passing 'f90' or 'c++' here will then
40926 trigger the C++/Fortran compiler identification within
40927 get_compiler_info. The latter procedure was extended to also handle
40928 the 'f90' argument (similarly to the already existing 'c++' one).
40930 2022-05-31 Nils-Christian Kempke <nils-christian.kempke@intel.com>
40932 gdb/testsuite: move getting_compiler_info to front of gdb_compile
40933 The procedure gdb_compile queries its options via
40935 [lsearch -exact $options getting_compiler_info]
40937 to check whether or not it was called in with the option
40938 getting_compiler_info. If it was called with this option it would
40939 preprocess some test input to try and figure out the actual compiler
40940 version of the compiler used. While doing this we cannot again try to
40941 figure out the current compiler version via the 'getting_compiler_info'
40942 option as this would cause infinite recursion. As some parts of the
40943 procedure do recursively test for the compiler version to e.g. set
40944 certain flags, at several places gdb_compile there are checks for the
40945 getting_compiler_info option needed.
40947 In the procedure, there was already a variable 'getting_compiler_info'
40948 which was set to the result of the 'lsearch' query and used instead of
40949 again and again looking for getting_compiler_info in the procedure
40950 options. But, this variable was actually set too late within the code.
40951 This lead to a mixture of querying 'getting_compiler_info' or
40952 doing an lserach on the options passed to the procedure.
40954 I found this inconsistent and instead moved the variable
40955 getting_compiler_info to the front of the procedure. It is set to true
40956 or false depending on whether or not the argument is found in the
40957 procedure's options (just as before) and queried instead of doing an
40958 lsearch on the procedure options in the rest of the procedure.
40960 2022-05-31 Felix Willgerodt <felix.willgerodt@intel.com>
40961 Abdul Basit Ijaz <abdul.b.ijaz@intel.com>
40962 Nils-Christian Kempke <nils-christian.kempke@intel.com>
40964 gdb/testsuite: Fix fortran types for Intel compilers.
40965 Newer Intel compilers emit their dwarf type name in a slightly different
40966 format. Therefore, this needs adjustment to make more tests pass in the
40969 2022-05-31 Abdul Basit Ijaz <abdul.b.ijaz@intel.com>
40971 gdb/testsuite: Use -module option for Intel Fortran compilers
40972 The '-J' option is not supported in Intel compilers (ifx and ifort).
40973 The Intel version of the flag is '-module' which serves the same purpose.
40975 2022-05-31 Nils-Christian Kempke <nils-christian.kempke@intel.com>
40977 gdb/testsuite: remove F77_FOR_TARGET support
40978 The last uses of the F77_FOR_TARGET via passing f77 to GDB's compile
40979 procedure were removed in this commit
40981 commit 0ecee54cfd04a60e7ca61ae07c72b20e21390257
40982 Author: Tom Tromey <tromey@redhat.com>
40983 Date: Wed Jun 29 17:50:47 2011 +0000
40985 over 10 years ago. The last .f files in the testsuite by now are all
40986 being compiled by passing 'f90' to the GDB compile, thus only actually
40987 using F90_FOR_TARGET (array-element.f, block-data.f, subarray.f).
40988 Gfortran in this case is backwards compatible with most f77 code as
40989 claimed on gcc.gnu.org/fortran.
40991 The reason we'd like to get rid of this now is, that we'll be
40992 implementing a Fortran compiler identification mechanism, similar to the
40993 C/Cpp existing ones. It would be using the Fortran preprocessor macro
40994 defines to identify the Fortran compiler version at hand. We found it
40995 inconsequent to only implement this for f90 but, on the other hand, f77
40996 seems deprecated. So, with this commit we remove the remaining lines for
40999 2022-05-31 Pedro Alves <pedro@palves.net>
41001 Improve clear command's documentation
41002 Co-Authored-By: Eli Zaretskii <eliz@gnu.org>
41004 Change-Id: I9440052fd28f795d6f7c93a4576beadd21f28885
41006 2022-05-31 Pedro Alves <pedro@palves.net>
41008 Clarify why we unit test matching symbol names with 0xff characters
41009 In the name matching unit tests in gdb/dwarf2/read.c, explain better
41010 why we test symbols with \377 / 0xff characters (Latin1 'ÿ').
41012 Change-Id: I517f13adfff2e4d3cd783fec1d744e2b26e18b8e
41014 2022-05-31 Pedro Alves <pedro@palves.net>
41016 Improve break-range's documentation
41017 Change-Id: Iac26e1d2e7d8dc8a7d9516e6bdcc5c3fc4af45c8
41019 Explicitly mention yet-unloaded shared libraries in location spec examples
41020 Change-Id: I05639ddb3bf620c7297b57ed286adc3aa926b7b6
41022 2022-05-31 Alan Modra <amodra@gmail.com>
41024 sparc64 segfault in finish_dynamic_symbol
41025 SYMBOL_REFERENCES_LOCAL can return true for undefined symbols. This
41026 can result in a segfault when running sparc64 ld/testsuite/ld-vsb
41027 tests that expect a failure.
41029 * elfxx-sparc.c (_bfd_sparc_elf_finish_dynamic_symbol): Don't
41030 access u.def.section on non-default visibility undefined symbol.
41032 2022-05-31 Alan Modra <amodra@gmail.com>
41034 ia64 gas: Remove unnecessary init
41035 The whole struct is cleared by alloc_record.
41037 * config/tc-ia64.c (output_prologue, output_prologue_gr): Don't
41038 zero r.record.r.mask.
41040 2022-05-31 Alan Modra <amodra@gmail.com>
41042 v850_elf_set_note prototype
41043 v850_elf_set_note is declared using an unsigned int note param in
41044 elf32-v850.h but defined with enum c850_notes note in elf32-v850.c.
41045 Current mainline gcc is warning about this. Huh.
41047 * elf32-v850.c (v850_elf_set_note): Make "note" param an
41050 2022-05-31 Alan Modra <amodra@gmail.com>
41052 Import libiberty from gcc
41056 * demangle.h: Import from gcc.
41059 * testsuite/demangle-expected: Import from gcc.
41061 2022-05-31 Andrew Burgess <aburgess@redhat.com>
41063 gdb/testsuite: resolve duplicate test name in gdb.trace/signal.exp
41064 Spotted a duplicate test name in gdb.trace/signal.exp, resolved in
41065 this commit by making use of 'with_test_prefix'.
41067 2022-05-31 Alan Modra <amodra@gmail.com>
41069 Ajdust more tests for opcodes/i386: remove trailing whitespace
41070 git commit 202be274a4 also missed adjusting a few testsuite files.
41072 i686-vxworks +FAIL: VxWorks shared library test 1
41073 i686-vxworks +FAIL: VxWorks executable test 1 (dynamic)
41075 2022-05-31 Alan Modra <amodra@gmail.com>
41077 Trailing spaces in objdump -r header
41078 git commit 202be274a4 went a little wild in removing trailing spaces
41079 in gas/testsuite/gas/i386/{secidx.d,secrel.d}, causing
41080 x86_64-w64-mingw32 +FAIL: i386 secrel reloc
41081 x86_64-w64-mingw32 +FAIL: i386 secidx reloc
41083 I could have just replaced the trailing space, but let's fix the
41084 objdump output instead. Touches lots of testsuite files.
41086 2022-05-31 GDB Administrator <gdbadmin@sourceware.org>
41088 Automatic date update in version.in
41090 2022-05-30 Simon Marchi <simon.marchi@efficios.com>
41092 gdb/testsuite: fix gdb.trace/signal.exp on x86
41095 202be274a41a ("opcodes/i386: remove trailing whitespace from insns with zero operands")
41097 causes this regression:
41099 FAIL: gdb.trace/signal.exp: find syscall insn in kill
41101 It's because the test still expects to match a whitespace after the
41102 instruction, which the patch mentioned above removed. Remove the
41103 whitespaces for the regexp.
41105 Change-Id: Ie194273cc942bfd91332d4035f6eec55b7d3a428
41107 2022-05-30 Pedro Alves <pedro@palves.net>
41109 gdb/manual: Introduce location specs
41110 The current "Specify Location" section of the GDB manual starts with:
41112 "Several @value{GDBN} commands accept arguments that specify a location
41113 of your program's code."
41115 And then, such commands are documented as taking a "location"
41116 argument. For example, here's a representative subset:
41118 @item break @var{location}
41119 @item clear @var{location}
41120 @item until @var{location}
41121 @item list @var{location}
41122 @item edit @var{location}
41123 @itemx info line @var{location}
41124 @item info macros @var{location}
41125 @item trace @var{location}
41126 @item info scope @var{location}
41127 @item maint agent @r{[}-at @var{location}@r{,}@r{]} @var{expression}
41129 The issue here is that "location" isn't really correct for most of
41130 these commands. Instead, the "location" argument is really a
41131 placeholder that represent an umbrella term for all of the
41132 "linespecs", "explicit location", and "address location" input
41133 formats. GDB parses these and then finds the actual code locations
41134 (plural) in the program that match. For example, a "location"
41135 specified like "-function func" will actually match all the code
41136 locations in the program that correspond to the address/file/lineno of
41137 all the functions named "func" in all the loaded programs and shared
41138 libraries of all the inferiors. A location specified like "-function
41139 func -label lab" matches all the addresses of C labels named "lab" in
41140 all functions named "func". Etc.
41142 This means that several of the commands that claim they accept a
41143 "location", actually end up working with multiple locations, and the
41144 manual doesn't explain that all that well. In some cases, the command
41145 will work with all the resolved locations. In other cases, the
41146 command aborts with an error if the location specification resolves to
41147 multiple locations in the program. In other cases, GDB just
41148 arbitrarily and silently picks whatever is the first resolved code
41149 location (which sounds like should be improved).
41151 To clarify this, I propose we use the term "Location Specification",
41152 with shorthand "locaction spec", when we're talking about the user
41153 input, the argument or arguments that is/are passed to commands to
41154 instruct GDB how to find locations of interest. This is distinct from
41155 the actual code locations in the program, which are what GDB finds
41156 based on the user-specified location spec. Then use "location
41157 specification or the shorter "location spec" thoughout instead of
41158 "location" when we're talking about the user input.
41160 Thus, this commit does the following:
41162 - renames the "Specify Location" section of the manual to "Location
41165 - It then introduces the term "Location Specification", with
41166 corresponding shorthand "location spec", as something distinct from
41167 an actual code location in the program. It explains what a concrete
41168 code location is. It explains that a location specification may be
41169 incomplete, and that may match multiple code locations in the
41170 program, or no code location at all. It gives examples. Some
41171 pre-existing examples were moved from the "Set Breaks" section, and
41172 a few new ones that didn't exist yet were added. I think it is
41173 better to have these centralized in this "Location Specification"
41174 section, since all the other commands that accept a location spec
41175 have an xref that points there.
41177 - Goes through the manual, and where "@var{location}" was used for a
41178 command argument, updated it to say "@var{locspec}" instead. At the
41179 same time, tweaks the description of the affected commands to
41180 describe what happens when the location spec resolves to more than
41181 one location. Most commands just did not say anything about that.
41183 One command -- "maint agent -at @var{location}" -- currently says it
41184 accepts a "location", suggesting it can accept address and explicit
41185 locations too, but that's incorrect. In reality, it only accepts
41186 linespecs, so fix it accordingly.
41188 One MI command -- "-trace-find line" -- currently says it accepts a
41189 "line specification", but it can accept address and explicit
41190 locations too, so fix it accordingly.
41192 Special thanks goes to Eli Zaretskii for reviews and rewording
41195 Change-Id: Ic42ad8565e79ca67bfebb22cbb4794ea816fd08b
41197 2022-05-30 Luis Machado <luis.machado@arm.com>
41199 Move 64-bit BFD files from ALL_TARGET_OBS to ALL_64_TARGET_OBS
41200 Doing a 32-bit build with "--enable-targets=all --disable-sim" fails to link
41205 loongarch-tdep.o: In function `loongarch_gdbarch_init':
41206 binutils-gdb/gdb/loongarch-tdep.c:443: undefined reference to `loongarch_r_normal_name'
41207 loongarch-tdep.o: In function `loongarch_fetch_instruction':
41208 binutils-gdb/gdb/loongarch-tdep.c:37: undefined reference to `loongarch_insn_length'
41209 loongarch-tdep.o: In function `loongarch_scan_prologue(gdbarch*, unsigned long long, unsigned long long, frame_info*, trad_frame_cache*) [clone .isra.4]':
41210 binutils-gdb/gdb/loongarch-tdep.c:87: undefined reference to `loongarch_insn_length'
41211 binutils-gdb/gdb/loongarch-tdep.c:88: undefined reference to `loongarch_decode_imm'
41212 binutils-gdb/gdb/loongarch-tdep.c:89: undefined reference to `loongarch_decode_imm'
41213 binutils-gdb/gdb/loongarch-tdep.c:90: undefined reference to `loongarch_decode_imm'
41214 binutils-gdb/gdb/loongarch-tdep.c:91: undefined reference to `loongarch_decode_imm'
41215 binutils-gdb/gdb/loongarch-tdep.c:92: undefined reference to `loongarch_decode_imm'
41219 Given the list of 64-bit BFD files in
41220 opcodes/Makefile.am:TARGET64_LIBOPCODES_CFILES, it looks like GDB's
41221 ALL_TARGET_OBS list is including files that should be included in
41222 ALL_64_TARGET_OBS instead.
41224 This patch accomplishes this and enables a 32-bit build with
41225 "--enable-targets=all --disable-sim" to complete.
41227 Moving the bpf, tilegx and loongarch files to the correct list means GDB can
41228 find the correct disassembler function instead of finding a null pointer.
41230 We still need the "--disable-sim" switch (or "--enable-64-bit-bfd") to
41231 make a 32-bit build with "--enable-targets=all" complete correctly
41233 2022-05-30 Luis Machado <luis.machado@arm.com>
41235 Fix failing test for armeb-gnu-eabi
41236 The following test fails on the armeb-gnu-eabi target:
41238 FAIL: Unwind information for Armv8.1-M.Mainline PACBTI extension
41240 This patch adjusts the expected output for big endian.
41242 2022-05-30 Alan Modra <amodra@gmail.com>
41244 Use a union to avoid casts in bfd/doc/chew.c
41245 This fixes -Wpedantic warnings in chew.c. Conversion between function
41246 and object pointers is not guaranteed. They can even be different
41247 sizes, not that we're likely to encounter build machines like that
41251 * doc/chew.c (pcu): New union typedef.
41252 (dict_type, pc): Use it here. Adjust uses of pc.
41253 (add_to_definition): Make "word" param a pcu. Adjust all uses
41255 (stinst_type): Delete.
41257 2022-05-30 Alan Modra <amodra@gmail.com>
41259 use libiberty xmalloc in bfd/doc/chew.c
41260 Catch out of memory.
41262 * doc/chew.c: Include libibery.h.
41263 (init_string_with_size, nextword): Replace malloc with xmalloc.
41264 (newentry, add_to_definition): Likewise.
41265 (catchar, catbuf): Replace realloc with xrealloc.
41266 (add_intrinsic): Replace strdup with xstrdup.
41267 * doc/local.mk (LIBIBERTY): Define.
41268 (chew): Link against libiberty.
41269 * Makefile.in: Regenerate.
41271 2022-05-30 Alan Modra <amodra@gmail.com>
41273 Update K&R functions in bfd/doc/chew.c
41274 * doc/chew.c: Update function definitions to ISO C, remove
41275 now unnecessary prototypes.
41277 2022-05-30 Alan Modra <amodra@gmail.com>
41279 Reorganise bfd/doc/chew.c a little
41280 This also removes some unused variables, and deletes support for the
41281 "var" keyword which isn't used and was broken. (No means to set
41282 variables, and add_var used push_number inconsistent with its use
41285 * doc/chew.c: Move typedefs before variables, variables before
41287 (die): Move earlier.
41288 (word_type, sstack, ssp): Delete.
41289 (dict_type): Delete var field.
41291 (compile): Remove "var" support.
41293 2022-05-30 jiawei <jiawei@iscas.ac.cn>
41295 RISC-V: Add zhinx extension supports.
41296 The zhinx extension is a sub-extension in zfinx, corresponding to
41297 zfh extension but use GPRs instead of FPRs.
41299 This patch expanded the zfh insn class define, since zfh and zhinx
41300 use the same opcodes, thanks for Nelson's works.
41302 changelog in V2: Add missing classes of 'zfh' and 'zhinx' in
41303 "riscv_multi_subset_supports_ext".
41307 * elfxx-riscv.c (riscv_multi_subset_supports): New extensions.
41308 (riscv_multi_subset_supports_ext): New extensions.
41312 * testsuite/gas/riscv/fp-zhinx-insns.d: New test.
41313 * testsuite/gas/riscv/fp-zhinx-insns.s: New test.
41317 * opcode/riscv.h (enum riscv_insn_class): New INSN classes.
41321 * riscv-opc.c: Modify INSN_CLASS.
41323 2022-05-30 GDB Administrator <gdbadmin@sourceware.org>
41325 Automatic date update in version.in
41327 2022-05-29 GDB Administrator <gdbadmin@sourceware.org>
41329 Automatic date update in version.in
41331 2022-05-28 Andrew Burgess <aburgess@redhat.com>
41333 gdb/python: improve formatting of help text for user defined commands
41334 Consider this command defined in Python (in the file test-cmd.py):
41336 class test_cmd (gdb.Command):
41338 This is the first line.
41339 Indented second line.
41340 This is the third line.
41343 def __init__ (self):
41344 super ().__init__ ("test-cmd", gdb.COMMAND_OBSCURE)
41346 def invoke (self, arg, from_tty):
41347 print ("In test-cmd")
41351 Now, within a GDB session:
41353 (gdb) source test-cmd.py
41354 (gdb) help test-cmd
41356 This is the first line.
41357 Indented second line.
41358 This is the third line.
41362 I think there's three things wrong here:
41364 1. The leading blank line,
41365 2. The trailing blank line, and
41366 3. Every line is indented from the left edge slightly.
41368 The problem of course, is that GDB is using the Python doc string
41369 verbatim as its help text. While the user has formatted the help text
41370 so that it appears clear within the .py file, this means that the text
41371 appear less well formatted when displayed in the "help" output.
41373 The same problem can be observed for gdb.Parameter objects in their
41376 In this commit I aim to improve the "help" output for commands and
41379 To do this I have added gdbpy_fix_doc_string_indentation, a new
41380 function that rewrites the doc string text following the following
41383 1. Leading blank lines are removed,
41384 2. Trailing blank lines are removed, and
41385 3. Leading whitespace is removed in a "smart" way such that the
41386 relative indentation of lines is retained.
41388 With this commit in place the above example now looks like this:
41390 (gdb) source ~/tmp/test-cmd.py
41391 (gdb) help test-cmd
41392 This is the first line.
41393 Indented second line.
41394 This is the third line.
41397 Which I think is much neater. Notice that the indentation of the
41398 second line is retained. Any blank lines within the help text (not
41399 leading or trailing) will be retained.
41401 I've added a NEWS entry to note that there has been a change in
41402 behaviour, but I didn't update the manual. The existing manual is
41403 suitably vague about how the doc string is used, so I think the new
41404 behaviour is covered just as well by the existing text.
41406 2022-05-28 Andrew Burgess <aburgess@redhat.com>
41408 gdb: use gdb::unique_xmalloc_ptr<char> for docs in cmdpy_init
41409 Make use of gdb::unique_xmalloc_ptr<char> to hold the documentation
41410 string in cmdpy_init (when creating a custom GDB command in Python).
41411 I think this is all pretty straight forward, the only slight weirdness
41412 is the removal of the call to free toward the end of this function.
41414 Prior to this commit, if an exception was thrown after the GDB command
41415 was created then we would (I think) end up freeing the documentation
41416 string even though the command would remain registered with GDB, which
41417 would surely lead to undefined behaviour.
41419 After this commit we release the doc string at the point that we hand
41420 it over to the command creation routines. If we throw _after_ the
41421 command has been created within GDB then the doc string will be left
41422 live. If we throw during the command creation itself (either from
41423 add_prefix_cmd or add_cmd) then it is up to those functions to free
41424 the doc string (I suspect we don't, but I think in general the
41425 commands are pretty bad at cleaning up after themselves, so I don't
41426 think this is a huge problem).
41428 2022-05-28 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
41430 gprofng: fix build with -mx32
41432 2022-05-27 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
41436 * src/Experiment.cc (write_header): Fix argument for ctime.
41437 Fix -Wformat= warnings.
41438 * src/Dbe.cc: Likewise.
41439 * src/DwarfLib.h: Fix [-Wsign-compare] warnings.
41440 * src/Experiment.h: Likewise.
41441 * src/ipc.cc: Fix -Wformat= warnings.
41443 2022-05-28 GDB Administrator <gdbadmin@sourceware.org>
41445 Automatic date update in version.in
41447 2022-05-27 Tom Tromey <tom@tromey.com>
41449 Fix crash with "maint print arc"
41450 Luis noticed that "maint print arc" would crash, because the command
41451 handler did not find "show" in the command name, violating an
41452 invariant. This patch fixes the bug by changing the registration to
41453 use add_basic_prefix_cmd instead.
41455 2022-05-27 Andrew Burgess <aburgess@redhat.com>
41457 opcodes/i386: remove trailing whitespace from insns with zero operands
41458 While working on another patch[1] I had need to touch this code in
41461 ins->obufp = ins->mnemonicendp;
41462 for (i = strlen (ins->obuf) + prefix_length; i < 6; i++)
41463 oappend (ins, " ");
41464 oappend (ins, " ");
41465 (*ins->info->fprintf_styled_func)
41466 (ins->info->stream, dis_style_mnemonic, "%s", ins->obuf);
41468 What this code does is add whitespace after the instruction mnemonic
41469 and before the instruction operands.
41471 The problem I ran into when working on this code can be seen by
41472 assembling this input file:
41478 Now, when I disassemble, here's the output. I've replaced trailing
41479 whitespace with '_' so that the issue is clearer:
41481 Disassembly of section .text:
41483 0000000000000000 <.text>:
41487 Notice that there's no trailing whitespace after 'nop', but there are
41488 three spaces after 'retq'!
41490 What happens is that instruction mnemonics are emitted into a buffer
41491 instr_info::obuf, then instr_info::mnemonicendp is setup to point to
41492 the '\0' character at the end of the mnemonic.
41494 When we emit the whitespace, this is then added starting at the
41495 mnemonicendp position. Lets consider 'retq', first the buffer is
41498 'r' 'e' 't' 'q' '\0'
41500 Then we add whitespace characters at the '\0', converting the buffer
41503 'r' 'e' 't' 'q' ' ' ' ' ' ' '\0'
41505 However, 'nop' is actually an alias for 'xchg %rax,%rax', so,
41506 initially, the buffer is setup like this:
41508 'x' 'c' 'h' 'g' '\0'
41510 Then in NOP_Fixup we spot that we have an instruction that is an alias
41511 for 'nop', and adjust the buffer to this:
41513 'n' 'o' 'p' '\0' '\0'
41515 The second '\0' is left over from the original buffer contents.
41516 However, when we rewrite the buffer, we don't afjust mnemonicendp,
41517 which still points at the second '\0' character.
41519 Now, when we insert whitespace we get:
41521 'n' 'o' 'p' '\0' ' ' ' ' ' ' ' ' '\0'
41523 Notice the whitespace is inserted after the first '\0', so, when we
41524 print the buffer, the whitespace is not printed.
41526 The fix for this is pretty easy, I can change NOP_Fixup to adjust
41527 mnemonicendp, but now a bunch of tests start failing, we now produce
41528 whitespace after the 'nop', which the tests don't expect.
41530 So, I could update the tests to expect the whitespace....
41532 ...except I'm not a fan of trailing whitespace, so I'd really rather
41535 Turns out, I can pretty easily update the whitespace emitting code to
41536 spot instructions that have zero operands and just not emit any
41537 whitespace in this case. So this is what I've done.
41539 I've left in the fix for NOP_Fixup, I think updating mnemonicendp is
41540 probably a good thing, though this is not really required any more.
41542 I've then updated all the tests that I saw failing to adjust the
41543 expected patterns to account for the change in whitespace.
41545 [1] https://sourceware.org/pipermail/binutils/2022-April/120610.html
41547 2022-05-27 Alan Modra <amodra@gmail.com>
41549 Replace bfd_hostptr_t with uintptr_t
41550 bfd_hostptr_t is defined as a type large enough to hold either a long
41551 or a pointer. It mostly appears in the coff backend code in casts.
41552 include/coff/internal.h struct internal_syment and union
41553 internal_auxent have the only uses in data structures, where
41554 comparison with include/coff/external.h and other code reveals that
41555 the type only needs to be large enough for a 32-bit integer or a
41556 pointer. That should mean replacing with uintptr_t is OK.
41558 Remove much of BFD_HOST configury
41559 This patch removes the definition of bfd_uint64_t and bfd_int64_t as
41560 well as most BFD_HOST_* which are now unused.
41562 Remove use of bfd_uint64_t and similar
41563 Requiring C99 means that uses of bfd_uint64_t can be replaced with
41564 uint64_t, and similarly for bfd_int64_t, BFD_HOST_U_64_BIT, and
41565 BFD_HOST_64_BIT. This patch does that, removes #ifdef BFD_HOST_*
41566 and tidies a few places that print 64-bit values.
41568 2022-05-27 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
41570 gprofng: fix build with --disable-shared
41572 2022-05-26 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
41574 * libcollector/configure.ac: Use AC_MSG_WARN instead of AC_MSG_ERROR
41575 * libcollector/configure: Rebuild.
41577 2022-05-27 Jan Beulich <jbeulich@suse.com>
41579 x86/Intel: allow MASM representation of embedded rounding / SAE
41580 MASM doesn't support the separate operand form; the modifier belongs
41581 after the instruction instead. Accept this form alongside the original
41582 (now legacy) one. Short of having access to a MASM version to actually
41583 check in how far "after the instruction" is a precise statement in their
41584 documentation, allow both that and the SDM mandated form where the
41585 modifier is on the last register operand (with a possible immediate
41586 operand following).
41588 Sadly the split out function, at least for the time being, needs to cast
41589 away constness at some point, as the two callers disagree in this
41592 Adjust some, but not all of the testcases.
41594 2022-05-27 Jan Beulich <jbeulich@suse.com>
41596 x86: re-work AVX512 embedded rounding / SAE
41597 As a preparatory step to allowing proper non-operand forms of specifying
41598 embedded rounding / SAE, convert the internal representation to non-
41599 operand form. While retaining properties (and in a few cases perhaps
41600 providing more meaningful diagnostics), this means doing away with a few
41601 hundred standalone templates, thus - as a nice side effect - reducing
41602 memory consumption / cache occupancy.
41604 x86/Intel: adjust representation of embedded rounding / SAE
41605 MASM doesn't consider {sae} and alike a separate operand; it is attached
41606 to the last register operand instead, just like spelled out by the SDM.
41607 Make the disassembler follow this first, before also adjusting the
41608 assembler (such that it'll be easy to see that the assembler change
41609 doesn't alter generated code).
41611 2022-05-27 Jan Beulich <jbeulich@suse.com>
41613 x86/Intel: allow MASM representation of embedded broadcast
41614 MASM doesn't support the {1to<n>} form; DWORD BCST (paralleling
41615 DWORD PTR) and alike are to be used there instead. Accept these forms
41616 alongside the original (now legacy) ones.
41618 Acceptance of the original {1to<n>} operand suffix is retained both for
41619 backwards compatibility and to disambiguate VFPCLASSP{S,D,H} and vector
41620 conversions with shrinking element sizes. I have no insight (yet) into
41621 how MASM expects those to be disambiguated.
41623 Adjust some, but not all of the testcases.
41625 2022-05-27 Jan Beulich <jbeulich@suse.com>
41627 x86/Intel: adjust representation of embedded broadcast
41628 MASM doesn't support the {1to<n>} form; DWORD BCST (paralleling
41629 DWORD PTR) and alike are to be used there instead. Make the disassembler
41630 follow this first, before also adjusting the assembler (such that it'll
41631 be easy to see that the assembler change doesn't alter generated code).
41633 For VFPCLASSP{S,D,H} and vector conversions with shrinking element sizes
41634 the original {1to<n>} operand suffix is retained, to disambiguate
41635 output. I have no insight (yet) into how MASM expects those to be
41638 2022-05-27 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
41640 gprofng: fix build with -mx32
41642 2022-05-26 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
41645 * libcollector/libcol_util.h (__collector_getsp, __collector_getfp,
41646 __collector_getpc): Adapt for build with -mx32
41647 * libcollector/heaptrace.c: Fix -Wpointer-to-int-cast warnings.
41648 * libcollector/hwprofile.h: Likewise.
41649 * libcollector/mmaptrace.c: Likewise.
41650 * libcollector/synctrace.c: Likewise.
41651 * libcollector/unwind.c: Likewise.
41653 2022-05-27 GDB Administrator <gdbadmin@sourceware.org>
41655 Automatic date update in version.in
41657 2022-05-27 Hans-Peter Nilsson <hp@axis.com>
41659 ld: cris*-elf: Default to --no-warn-rwx-segment
41661 configure.tgt (cris-*-*, crisv32-*-* sans *-aout and *-linux): Unless
41662 specified through the --enable-* -option, default to
41663 --no-warn-rwx-segment.
41665 Change-Id: I846bcd3e6762da807b17215a9fe337461ea0d710
41667 2022-05-27 Hans-Peter Nilsson <hp@axis.com>
41669 cris: bfd: Correct default to no execstack
41670 In the now-historical CRIS glibc port, the default stack permission
41671 was no-exec as in "#define DEFAULT_STACK_PERMS (PF_R|PF_W)", and the
41672 gcc port only emits the executable-stack marker when needed; when
41673 emitting code needing it. In other words, the binutils setting
41674 mismatches. It doesn't matter much, except being confusing and
41675 defaulting to "off" is more sane.
41679 * testsuite/ld-elf/elf.exp (target_defaults_to_execstack): Switch to 0
41683 * elf32-cris.c (elf_backend_default_execstack): Define to 0.
41685 Change-Id: I52f37598f119b19111c7a6546c00a627fca0f396
41687 2022-05-26 John Baldwin <jhb@FreeBSD.org>
41689 aarch64-fbsd-nat: Move definition of debug_regs_probed under HAVE_DBREG.
41690 This fixes the build on older FreeBSD systems without support for
41691 hardware breakpoints/watchpoints.
41693 2022-05-26 Lancelot SIX <lancelot.six@amd.com>
41695 gdb: Change psymbol_functions::require_partial_symbols to partial_symbols
41696 The previous patch ensured that partial symbols are read before calling
41697 most of the quick_function's methods.
41699 The psymbol_functions class has the require_partial_symbols method which
41700 serves this exact purpose, and does not need to do it anymore.
41702 This patch renames this method to partial_symbols and makes it an accessor
41703 which asserts that partial symbols have been read at this point.
41705 Regression tested on x86_64-linux.
41707 2022-05-26 Lancelot SIX <lancelot.six@amd.com>
41709 gdb: Require psymtab before calling quick_functions in objfile
41710 The recent DWARF indexer rewrite introduced a regression when debugging
41713 Here is a way to reproduce the issue (there might be other ways, but one
41714 is enough and this one mimics the situation we encountered). Consider a
41715 program which forks, and the child loads a shared library and calls a
41716 function in this shared library:
41720 void *solib = dlopen (some_solib, RTLD_NOW);
41721 void (*foo) () = dlsym (some_solib, "foo");
41725 Suppose that this program is compiled without debug info, but the shared
41726 library it loads has debug info enabled.
41728 When debugging such program with the following options:
41730 - set detach-on-fork off
41731 - set follow-fork-mode child
41733 we see something like:
41736 Function "foo" not defined.
41737 Make breakpoint pending on future shared library load? (y or [n]) y
41738 Breakpoint 1 (foo) pending.
41740 Starting program: a.out
41741 [Attaching after process 19720 fork to child process 19723]
41742 [New inferior 2 (process 19723)]
41743 [Switching to process 19723]
41745 Thread 2.1 "a.out" hit Breakpoint 1, 0x00007ffff7fc3101 in foo () from .../libfoo.so
41748 Fatal signal: Segmentation fault
41749 ----- Backtrace -----
41750 0x55a278f77d76 gdb_internal_backtrace_1
41751 ../../gdb/bt-utils.c:122
41752 0x55a278f77f83 _Z22gdb_internal_backtracev
41753 ../../gdb/bt-utils.c:168
41754 0x55a27940b83b handle_fatal_signal
41755 ../../gdb/event-top.c:914
41756 0x55a27940bbb1 handle_sigsegv
41757 ../../gdb/event-top.c:987
41759 /build/glibc-sMfBJT/glibc-2.31/nptl/../sysdeps/unix/sysv/linux/x86_64/sigaction.c:0
41760 0x55a27924c9d3 _ZNKSt15__uniq_ptr_implI18dwarf2_per_cu_data26dwarf2_per_cu_data_deleterE6_M_ptrEv
41761 /usr/include/c++/9/bits/unique_ptr.h:154
41762 0x55a279248bc9 _ZNKSt10unique_ptrI18dwarf2_per_cu_data26dwarf2_per_cu_data_deleterE3getEv
41763 /usr/include/c++/9/bits/unique_ptr.h:361
41764 0x55a2792ae718 _ZN27dwarf2_base_index_functions23find_last_source_symtabEP7objfile
41765 ../../gdb/dwarf2/read.c:3164
41766 0x55a279afb93e _ZN7objfile23find_last_source_symtabEv
41767 ../../gdb/symfile-debug.c:139
41768 0x55a279aa3040 _Z20select_source_symtabP6symtab
41769 ../../gdb/source.c:365
41770 0x55a279aa22a1 _Z34set_default_source_symtab_and_linev
41771 ../../gdb/source.c:268
41772 0x55a27903c44c list_command
41773 ../../gdb/cli/cli-cmds.c:1185
41774 0x55a279051233 do_simple_func
41775 ../../gdb/cli/cli-decode.c:95
41776 0x55a27905f221 _Z8cmd_funcP16cmd_list_elementPKci
41777 ../../gdb/cli/cli-decode.c:2514
41778 0x55a279c3b0ba _Z15execute_commandPKci
41779 ../../gdb/top.c:660
41780 0x55a27940a6c3 _Z15command_handlerPKc
41781 ../../gdb/event-top.c:598
41782 0x55a27940b032 _Z20command_line_handlerOSt10unique_ptrIcN3gdb13xfree_deleterIcEEE
41783 ../../gdb/event-top.c:797
41784 0x55a279caf401 tui_command_line_handler
41785 ../../gdb/tui/tui-interp.c:278
41786 0x55a279409098 gdb_rl_callback_handler
41787 ../../gdb/event-top.c:230
41788 0x55a279ed5df2 rl_callback_read_char
41789 ../../../readline/readline/callback.c:281
41790 0x55a279408bd8 gdb_rl_callback_read_char_wrapper_noexcept
41791 ../../gdb/event-top.c:188
41792 0x55a279408de7 gdb_rl_callback_read_char_wrapper
41793 ../../gdb/event-top.c:205
41794 0x55a27940a061 _Z19stdin_event_handleriPv
41795 ../../gdb/event-top.c:525
41796 0x55a27a23771e handle_file_event
41797 ../../gdbsupport/event-loop.cc:574
41798 0x55a27a237f5f gdb_wait_for_event
41799 ../../gdbsupport/event-loop.cc:700
41800 0x55a27a235d81 _Z16gdb_do_one_eventv
41801 ../../gdbsupport/event-loop.cc:237
41802 0x55a2796c2ef0 start_event_loop
41803 ../../gdb/main.c:418
41804 0x55a2796c3217 captured_command_loop
41805 ../../gdb/main.c:478
41806 0x55a2796c717b captured_main
41807 ../../gdb/main.c:1340
41808 0x55a2796c7217 _Z8gdb_mainP18captured_main_args
41809 ../../gdb/main.c:1355
41810 0x55a278d0b381 main
41812 ---------------------
41813 A fatal error internal to GDB has been detected, further
41814 debugging is not possible. GDB will now terminate.
41816 This is a bug, please report it. For instructions, see:
41817 <https://www.gnu.org/software/gdb/bugs/>.
41819 The first issue observed is in the message printed when hitting the
41820 breakpoint. It says that there was a break in the .so file as if there
41821 was no debug info associated with it, but there is. Later, if we try to
41822 display the source where the execution stopped, we have a segfault.
41824 Note that not having the debug info on the main binary is not strictly
41825 required to encounter some issues, it only is to encounter the segfault.
41826 If the main binary has debug information, GDB shows some source form the
41827 main binary, unrelated to where we stopped.
41829 The core of the issue is that GDB never loads the psymtab for the
41830 library. It is not loaded when we first see the .so because in case of
41831 detach-on-fork off, follow-fork-mode child, infrun.c sets
41832 child_inf->symfile_flags = SYMFILE_NO_READ to delay the psymtab loading
41833 as much as possible. If we compare to what was done to handle this
41834 before the new indexer was activated, the psymatb construction for the
41835 shared library was done under
41836 psymbol_functions::expand_symtabs_matching:
41839 psymbol_functions::expand_symtabs_matching (...)
41841 for (partial_symtab *ps : require_partial_symbols (objfile))
41845 The new indexer's expand_symtabs_matching callback does not have a call
41846 to the objfile's require_partial_symbols, so if the partial symbol table
41847 is not loaded at this point, there is no mechanism to fix this.
41849 Instead of requiring each implementation of the quick_functions to check
41850 that partial symbols have been read, I think it is safer to enforce this
41851 when calling the quick functions. The general pattern for calling the
41852 quick functions is:
41854 for (auto *iter : qf)
41855 iter->the_actual_method_call (...)
41857 This patch proposes to wrap the access of the `qf` field with an accessor
41858 which ensures that partial symbols have been read before iterating:
41859 qf_require_partial_symbols. All calls to quick functions are updated
41862 - quick_functions::dump
41863 - quick_functions::read_partial_symbols (from
41864 objfile::require_partial_symbols)
41865 - quick_functions::can_lazily_read_symbols and quick_functions::has_symbols
41866 (from objfile::has_partial_symbols)
41868 Regression tested on x86_64-gnu-linux.
41870 Change-Id: I39a13a937fdbaae613a5cf68864b021000554546
41872 2022-05-26 Tom Tromey <tom@tromey.com>
41874 Fix crash in new DWARF indexer
41875 PR gdb/29128 points out a crash in the new DWARF index code. This
41876 happens if the aranges for a CU claims a PC, but the symtab that is
41877 created during CU expansion does not actually contain the PC. This
41878 can only occur due to bad debuginfo, but at the same time, gdb should
41881 This patch fixes the bug and further merges some code into
41882 dwarf2_base_index_functions. This merger helps prevent the same issue
41883 from arising from the other index implementations.
41885 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29128
41887 2022-05-26 Tom Tromey <tromey@adacore.com>
41889 Finalize each cooked index separately
41890 After DWARF has been scanned, the cooked index code does a
41891 "finalization" step in a worker thread. This step combines all the
41892 index entries into a single master list, canonicalizes C++ names, and
41893 splits Ada names to synthesize package names.
41895 While this step is run in the background, gdb will wait for the
41896 results in some situations, and it turns out that this step can be
41897 slow. This is PR symtab/29105.
41899 This can be sped up by parallelizing, at a small memory cost. Now
41900 each index is finalized on its own, in a worker thread. The cost
41901 comes from name canonicalization: if a given non-canonical name is
41902 referred to by multiple indices, there will be N canonical copies (one
41903 per index) rather than just one.
41905 This requires changing the users of the index to iterate over multiple
41906 results. However, this is easily done by introducing a new "chained
41909 When run on gdb itself, the memory cost seems rather low -- on my
41910 current machine, "maint space 1" reports no change due to the patch.
41912 For performance testing, using "maint time 1" and "file" will not show
41913 correct results. That approach measures "time to next prompt", but
41914 because the patch only affects background work, this shouldn't (and
41915 doesn't) change. Instead, a simple way to make gdb wait for the
41916 results is to set a breakpoint.
41920 $ /bin/time -f%e ~/gdb/install/bin/gdb -nx -q -batch \
41921 -ex 'break main' /tmp/gdb
41922 Breakpoint 1 at 0x43ec30: file ../../binutils-gdb/gdb/gdb.c, line 28.
41927 $ /bin/time -f%e ./gdb/gdb -nx -q -batch \
41928 -ex 'break main' /tmp/gdb
41929 Breakpoint 1 at 0x43ec30: file ../../binutils-gdb/gdb/gdb.c, line 28.
41932 Regression tested on x86-64 Fedora 34.
41934 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29105
41936 2022-05-26 Alan Modra <amodra@gmail.com>
41938 bit-rot in target before_parse function
41939 Copy initialisation over from the elf.em before_parse. Commit
41940 ba951afb999 2022-05-03 changed behaviour on arm and score regarding
41941 exec stack. This patch restores the previous behaviour.
41943 * emultempl/aarch64elf.em (before_parse): Init separate_code,
41944 warn_execstack, no_warn_rwx_segments and default_execstack.
41945 * emultempl/armelf.em (before_parse): Likewise.
41946 * emultempl/scoreelf.em (before_parse): Likewise.
41947 * testsuite/ld-elf/elf.exp (target_defaults_to_execstack): Return
41948 true for arm and nacl.
41950 2022-05-26 Richard Earnshaw <rearnsha@arm.com>
41952 arm: avoid use of GNU builtin function in s_arm_unwind_save_mixed
41953 Whilst reviewing Luis' proposed change to s_arm_unwind_save_mixed
41954 yesterday I noticed that we were making use of __builting_clzl
41955 directly within the main function, which is not guaranteed to be
41956 portable. Whilst studying the code further, I also realized that it
41957 could be rewritten without using it and also reworked to remove a lot
41958 of unnecessary iterations steps. So this patch does that (and also
41959 removes the source of the warning that Luis was trying to fix).
41960 Finally, with the rewrite we can also simplify the caller of this
41961 routine as the new version can handle all the cases directly.
41963 * config/tc-arm.c (s_arm_unwind_save_mixed): Rewrite without
41964 using __builtin_clzl.
41965 (s_arm_unwind_save): Simplify logic for simple/mixed register saves.
41967 2022-05-26 Lancelot SIX <lancelot.six@amd.com>
41969 gdb/linux-nat: xfer_memory_partial return E_IO on error
41970 When accessing /proc/PID/mem, if pread64/pwrite64/read/write encounters
41971 an error and return -1, linux_proc_xfer_memory_partial return
41974 I think it should return TARGET_XFER_E_IO in this case. TARGET_XFER_EOF
41975 is returned when pread64/pwrite64/read/frite returns 0, which indicates
41976 that the address space is gone and the whole process has exited or
41979 This patch makes this change.
41981 Regression tested on x86_64-linux-gnu.
41983 Change-Id: I6030412459663b8d7933483fdda22a6c2c5d7221
41985 2022-05-26 Lancelot SIX <lancelot.six@amd.com>
41987 gdb/testsuite: prefer gdb_test in gdb.dwarf2/calling-convention
41988 Since ed01945057c "Make gdb_test's question non-optional if specified",
41989 if the question and response parameters are given to gdb_test, the
41990 framework enforces that GDB asks the question. Before this patch, tests
41991 needed to use gdb_test_multiple to enforce this.
41993 This patch updates the gdb.dwarf2/calling-convention.exp testcase to use
41994 gdb_test to check that GDB asks a question. This replaces the more
41995 complicated gdb_test_multiple based implementation.
41997 Tested on x86_64-gnu-linux.
41999 Change-Id: I7216e822ca68f2727e0450970097d74c27c432fe
42001 2022-05-26 Potharla, Rupesh <Rupesh.Potharla@amd.com>
42003 bfd: Add Support for DW_FORM_strx* and DW_FORM_addrx*
42005 2022-05-26 Luca Boccassi <bluca@debian.org>
42007 ld: add --package-metadata
42008 Generate a .note.package FDO package metadata ELF note, following
42009 the spec: https://systemd.io/ELF_PACKAGE_METADATA/
42011 If the jansson library is available at build time (and it is explicitly
42012 enabled), link ld to it, and use it to validate that the input is
42013 correct JSON, to avoid writing garbage to the file. The
42014 configure option --enable-jansson has to be used to explicitly enable
42015 it (error out when not found). This allows bootstrappers (or others who
42016 are not interested) to seamlessly skip it without issues.
42018 2022-05-26 GDB Administrator <gdbadmin@sourceware.org>
42020 Automatic date update in version.in
42022 2022-05-26 Natarajan, Kavitha <Kavitha.Natarajan@amd.com>
42024 Re: Add bionutils support for DWARF v5's DW_OP_addrx
42025 Testsuite files belonging to commit 3ac9da49378c.
42027 2022-05-25 Pedro Alves <pedro@palves.net>
42029 Show enabled locations with disabled breakpoint parent as "y-"
42030 Currently, breakpoint locations that are enabled while their parent
42031 breakpoint is disabled are displayed with "y" in the Enb colum of
42032 "info breakpoints":
42034 (gdb) info breakpoints
42035 Num Type Disp Enb Address What
42036 1 breakpoint keep n <MULTIPLE>
42037 1.1 y 0x00000000000011b6 in ...
42038 1.2 y 0x00000000000011c2 in ...
42039 1.3 n 0x00000000000011ce in ...
42041 Such locations won't trigger a break, so to avoid confusion, show "y-"
42042 instead. For example:
42044 (gdb) info breakpoints
42045 Num Type Disp Enb Address What
42046 1 breakpoint keep n <MULTIPLE>
42047 1.1 y- 0x00000000000011b6 in ...
42048 1.2 y- 0x00000000000011c2 in ...
42049 1.3 n 0x00000000000011ce in ...
42051 The "-" sign is inspired on how the TUI represents breakpoints on the
42052 left side of the source window, with "b-" for a disabled breakpoint.
42054 Change-Id: I9952313743c51bf21b4b380c72360ef7d4396a09
42056 2022-05-25 Natarajan, Kavitha <Kavitha.Natarajan@amd.com>
42058 Add bionutils support for DWARF v5's DW_OP_addrx.
42060 2022-05-25 Pedro Alves <pedro@palves.net>
42062 gdb: Fix DUPLICATE and PATH regressions throughout
42063 The previous patch to add -prompt/-lbl to gdb_test introduced a
42064 regression: Before, you could specify an explicit empty message to
42065 indicate you didn't want to PASS, like so:
42067 gdb_test COMMAND PATTERN ""
42069 After said patch, gdb_test no longer distinguishes
42070 no-message-specified vs empty-message, so tests that previously would
42071 be silent on PASS, now started emitting PASS messages based on
42072 COMMAND. This in turn introduced a number of PATH/DUPLICATE
42073 violations in the testsuite.
42075 This commit fixes all the regressions I could see.
42077 This patch uses the new -nopass feature introduced in the previous
42078 commit, but tries to avoid it if possible. Most of the patch fixes
42079 DUPLICATE issues the usual way, of using with_test_prefix or explicit
42082 See previous commit's log for more info.
42084 In addition to looking for DUPLICATEs, I also looked for cases where
42085 we would now end up with an empty message in gdb.sum, due to a
42086 gdb_test being passed both no message and empty command. E.g., this
42087 in gdb.ada/bp_reset.exp:
42090 gdb_test "" "Breakpoint $decimal, foo\\.nested_sub \\(\\).*"
42092 was resulting in this in gdb.sum:
42094 PASS: gdb.ada/bp_reset.exp:
42096 I fixed such cases by passing an explicit message. We may want to
42097 make such cases error out.
42099 Tested on x86_64 GNU/Linux, native and native-extended-gdbserver. I
42100 see zero PATH cases now. I get zero DUPLICATEs with native testing
42101 now. I still see some DUPLICATEs with native-extended-gdbserver, but
42102 those were preexisting, unrelated to the gdb_test change.
42104 Change-Id: I5375f23f073493e0672190a0ec2e847938a580b2
42106 2022-05-25 Pedro Alves <pedro@palves.net>
42108 Add -nopass option to gdb_test/gdb_test_multiple
42109 The previous patch to add -prompt/-lbl to gdb_test introduced a
42110 regression: Before, you could specify an explicit empty message to
42111 indicate you didn't want to PASS, like so:
42113 gdb_test COMMAND PATTERN ""
42115 After said patch, gdb_test no longer distinguishes
42116 no-message-specified vs empty-message, so tests that previously would
42117 be silent on PASS, now started emitting PASS messages based on
42118 COMMAND. This in turn introduced a number of PATH/DUPLICATE
42119 violations in the testsuite.
42121 I think that not issuing a PASS should be restricted to only a few
42122 cases -- namely in shared routines exported by gdb.exp, which happen
42123 to use gdb_test internally. In tests that iterate an unknown number
42124 of tests exercising some racy scenario. In the latter case, if we
42125 emit PASSes for each iteration, we run into the situation where
42126 different testsuite runs emit a different number of PASSes.
42128 Thus, this patch preserves the current behavior, and, instead, adds a
42129 new "-nopass" option to gdb_test and gdb_test_no_output. Compared to
42130 the old way of supressing PASS with an empty message, this has the
42131 advantage that you can specify a FAIL message that is distinct from
42132 the command string, and, it's also more explicit.
42134 Change-Id: I5375f23f073493e0672190a0ec2e847938a580b2
42136 2022-05-25 Tsukasa OI <research_trasio@irq.a4lg.com>
42138 RISC-V: Fix RV32Q conflict
42139 This commit makes RV32 + 'Q' extension (version 2.2 or later) not
42140 conflicting since this combination is no longer prohibited by the
42145 * elfxx-riscv.c (riscv_parse_check_conflicts): Remove conflict
42146 detection that prohibits RV32Q on 'Q' version 2.2 or later.
42150 * testsuite/gas/riscv/march-fail-rv32iq.d: Removed.
42151 * testsuite/gas/riscv/march-fail-rv32iq.l: Likewise.
42152 * testsuite/gas/riscv/march-fail-rv32iq2p0.d: New test
42153 showing RV32IQ fails on 'Q' extension version 2.0.
42154 * testsuite/gas/riscv/march-fail-rv32iq2p0.l: Likewise.
42155 * testsuite/gas/riscv/march-fail-rv32iq2.d: Likewise.
42156 * testsuite/gas/riscv/march-fail-rv32iq-isa-2p2.d: New test
42157 showing RV32IQ fails on ISA specification version 2.2.
42158 * testsuite/gas/riscv/march-ok-rv32iq2p2.d: New test
42159 showing RV32IQ succesds on 'Q' extension version 2.2.
42160 * testsuite/gas/riscv/march-ok-rv32iq-isa-20190608.d: New test
42161 showing RV32IQ succesds on ISA specification 20190608.
42163 2022-05-25 Dmitry Selyutin <ghostmansd@gmail.com>
42165 opcodes: introduce BC field; fix isel
42166 Per Power ISA Version 3.1B 3.3.12, isel uses BC field rather than CRB
42167 field present in binutils sources. Also, per 1.6.2, BC has the same
42168 semantics as BA and BB fields, so this should keep the same flags and
42169 mask, only with the different offset.
42173 (BC): Define new field, with the same definition as CRB field,
42174 but with the PPC_OPERAND_CR_BIT flag present.
42176 * testsuite/gas/ppc/476.d: Update.
42177 * testsuite/gas/ppc/a2.d: Update.
42178 * testsuite/gas/ppc/e500.d: Update.
42179 * testsuite/gas/ppc/power7.d: Update.
42181 2022-05-25 Dmitry Selyutin <ghostmansd@gmail.com>
42183 ppc: extend opindex to 16 bits
42184 With the upcoming SVP64 extension[0] to PowerPC architecture, it became
42185 evident that PowerPC operand indices no longer fit 8 bits. This patch
42186 switches the underlying type to uint16_t, also introducing a special
42187 typedef so that any future extension goes even smoother.
42189 [0] https://libre-soc.org
42192 * opcode/ppc.h (ppc_opindex_t): New typedef.
42193 (struct powerpc_opcode): Use it.
42194 (PPC_OPINDEX_MAX): Define.
42196 * write.h (struct fix): Increase size of fx_pcrel_adjust.
42198 * config/tc-ppc.c (insn_validate): Use ppc_opindex_t for operands.
42199 (md_assemble): Likewise.
42200 (md_apply_fix): Likewise. Mask fx_pcrel_adjust with PPC_OPINDEX_MAX.
42201 (ppc_setup_opcodes): Adjust opcode index assertion.
42203 * ppc-dis.c (skip_optional_operands): Use ppc_opindex_t for
42205 (lookup_powerpc, lookup_prefix, lookup_vle, lookup_spe2): Likewise.
42206 (print_insn_powerpc): Likewise.
42208 2022-05-25 GDB Administrator <gdbadmin@sourceware.org>
42210 Automatic date update in version.in
42212 2022-05-24 Tom de Vries <tdevries@suse.de>
42214 [gdb/testsuite] Fix gdb.opt/clobbered-registers-O2.exp with clang
42215 When running test-case gdb.opt/clobbered-registers-O2.exp with clang 12.0.1, I
42219 Starting program: clobbered-registers-O2 ^M
42221 Program received signal SIGSEGV, Segmentation fault.^M
42222 gen_movsd (operand0=<optimized out>, operand1=<optimized out>) at \
42223 clobbered-registers-O2.c:31^M
42224 31 return *start_sequence(operand0, operand1);^M
42225 (gdb) FAIL: gdb.opt/clobbered-registers-O2.exp: runto: run to start_sequence
42228 The problem is that the breakpoint in start_sequence doesn't trigger, because:
42229 - the call to start_sequence in gen_movsd is optimized away, despite the
42230 __attribute__((noinline)), so the actual function start_sequence doesn't get
42232 - the debug info doesn't contain inlined function info, so there's only one
42233 breakpoint location.
42235 Adding noclone and noipa alongside the noinline attribute doesn't fix this.
42237 Adding the clang-specific attribute optnone in start_sequence does, but since
42238 it inhibits all optimization, that's not a preferred solution in a gdb.opt
42239 test-case, and it would work only for clang and not other compilers that
42240 possibly have the same issue.
42242 Fix this by moving functions start_sequence and gen_movsd into their own
42243 files, as a way of trying harder to enforce noinline/noipa/noclone.
42245 Tested on x86_64-linux.
42247 2022-05-24 Tom de Vries <tdevries@suse.de>
42249 [gdb/testsuite] Fix gdb.opt/clobbered-registers-O2.exp with gcc-12
42250 When running test-case gdb.opt/clobbered-registers-O2.exp with gcc-12, I run
42253 (gdb) PASS: gdb.opt/clobbered-registers-O2.exp: backtracing
42255 $1 = (unsigned int *) 0x7fffffffd070^M
42256 (gdb) print *operand0^M
42258 (gdb) FAIL: gdb.opt/clobbered-registers-O2.exp: print operand0
42261 The problem is that starting gcc-12, the assignments to x and y in main are
42270 return (int)gen_movsd (&x, &y);
42273 Fix this by making x and y volatile.
42275 Note that the test-case intends to check the handling of debug info for
42276 optimized code in function gen_movsd, so inhibiting optimization in main
42277 doesn't interfere with that.
42279 Tested on x86_64-linux.
42281 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29161
42283 2022-05-24 Tiezhu Yang <yangtiezhu@loongson.cn>
42285 gdb: LoongArch: Define LOONGARCH_LINUX_NUM_GREGSET as 45
42286 LOONGARCH_LINUX_NUM_GREGSET should be defined as 45 (32 + 1 + 1 + 11)
42287 due to reserved 11 for extension in glibc, otherwise when execute:
42289 make check-gdb TESTS="gdb.base/corefile.exp"
42291 there exists the following failed testcase:
42293 (gdb) core-file /home/loongson/build.git/gdb/testsuite/outputs/gdb.base/corefile/corefile.core
42295 warning: Unexpected size of section `.reg/7742' in core file.
42296 Core was generated by `/home/loongson/build.git/gdb/testsuite/outputs/gdb.base/corefile/corefile'.
42297 Program terminated with signal SIGABRT, Aborted.
42298 warning: Unexpected size of section `.reg/7742' in core file.
42299 #0 0x000000fff76f4e24 in raise () from /lib/loongarch64-linux-gnu/libc.so.6
42300 (gdb) FAIL: gdb.base/corefile.exp: core-file warning-free
42302 2022-05-24 Christophe Lyon <christophe.lyon@arm.com>
42304 AArch64: add support for DFP (Decimal Floating point)
42305 This small patch adds support for TYPE_CODE_DECFLOAT in
42306 aapcs_is_vfp_call_or_return_candidate_1 and pass_in_v_vfp_candidate,
42307 so that GDB for AArch64 knows how to pass DFP parameters and how to
42308 read DFP results when calling a function.
42310 Tested on aarch64-linux-gnu, with a GCC with DFP support in the PATH,
42311 all of GDB's DFP tests pass.
42313 2022-05-24 Christophe Lyon <christophe.lyon@arm.com>
42315 Merge config/ changes from GCC, to enable DFP on AArch64
42316 2022-04-28 Christophe Lyon <christophe.lyon@arm.com>
42319 * dfp.m4 (enable_decimal_float): Enable BID for AArch64.
42322 * configure: Regenerate.
42324 2022-05-24 Alan Modra <amodra@gmail.com>
42326 PR29171, invalid read causing SIGSEGV
42327 The fix here is to pass "section" down to read_and_display_attr_value.
42328 The test in read_and_display_attr_value is a little bit of hardening.
42331 * dwarf.c (display_debug_macro, display_debug_names): Pass section
42332 to read_and_display_attr_value2.
42333 (read_and_display_attr_value): Don't attempt to check for .dwo
42334 section name when section is NULL.
42336 2022-05-24 Alan Modra <amodra@gmail.com>
42338 PR29170, divide by zero displaying fuzzed .debug_names
42340 * dwarf.c (display_debug_names): Don't attempt to display bucket
42341 clashes when bucket count is zero.
42343 PR29169, invalid read displaying fuzzed .gdb_index
42345 * dwarf.c (display_gdb_index): Combine sanity checks. Calculate
42346 element counts, not word counts.
42348 2022-05-24 GDB Administrator <gdbadmin@sourceware.org>
42350 Automatic date update in version.in
42352 2022-05-23 John Baldwin <jhb@FreeBSD.org>
42354 Tweak the std::hash<> specialization for aarch64_features.
42355 Move the specialization into an explicit std namespace to workaround a
42356 bug in older compilers. GCC 6.4.1 at least fails to compile the previous
42357 version with the following error:
42359 gdb/arch/aarch64.h:48:13: error: specialization of 'template<class _Tp> struct std::hash' in different namespace [-fpermissive]
42361 struct std::hash<aarch64_features>
42363 2022-05-23 John Baldwin <jhb@FreeBSD.org>
42365 Fix loongarch_iterate_over_regset_sections for non-native targets.
42366 Define a constant for the number of registers stored in a register set
42367 and use this with register_size to compute the size of the
42368 general-purpose register set in core dumps.
42370 This also fixes the build on hosts such as FreeBSD that do not define
42371 an elf_gregset_t type.
42373 2022-05-23 Tiezhu Yang <yangtiezhu@loongson.cn>
42375 gdb: LoongArch: Implement the iterate_over_regset_sections gdbarch method
42376 When execute the following command on LoongArch:
42378 make check-gdb TESTS="gdb.base/auxv.exp"
42380 there exist the following unsupported and failed testcases:
42382 UNSUPPORTED: gdb.base/auxv.exp: gcore
42383 FAIL: gdb.base/auxv.exp: load core file for info auxv on native core dump
42384 FAIL: gdb.base/auxv.exp: info auxv on native core dump
42385 FAIL: gdb.base/auxv.exp: matching auxv data from live and core
42386 UNSUPPORTED: gdb.base/auxv.exp: info auxv on gcore-created dump
42387 UNSUPPORTED: gdb.base/auxv.exp: matching auxv data from live and gcore
42389 we can see the following messages in gdb/testsuite/gdb.log:
42391 gcore /home/loongson/build.git/gdb/testsuite/outputs/gdb.base/auxv/auxv.gcore
42392 Target does not support core file generation.
42393 (gdb) UNSUPPORTED: gdb.base/auxv.exp: gcore
42395 In order to fix the above issues, implement the iterate_over_regset_sections
42396 gdbarch method to iterate over core file register note sections on LoongArch.
42398 By the way, with this patch, the failed testcases in gdb.base/corefile.exp
42399 and gdb.base/gcore.exp can also be fixed.
42401 2022-05-23 Tom de Vries <tdevries@suse.de>
42403 [gdb/testsuite] Fix -prompt handling in gdb_test
42404 With check-read1 I run into:
42406 [infrun] maybe_set_commit_resumed_all_targets: not requesting
42407 commit-resumed for target native, no resumed threads^M
42408 (gdb) FAIL: gdb.base/ui-redirect.exp: debugging: continue
42409 [infrun] fetch_inferior_event: exit^M
42412 The problem is that proc gdb_test doesn't pass down the -prompt option to proc
42413 gdb_test_multiple, due to a typo making this lappend without effect:
42416 lappend "-prompt $prompt"
42419 Fix this by actually appending to opts.
42421 Tested on x86_64-linux.
42423 2022-05-23 Tom de Vries <tdevries@suse.de>
42425 [gdbsupport] Fix UB in print-utils.cc:int_string
42426 When building gdb with -fsanitize=undefined, I run into:
42428 (gdb) PASS: gdb.ada/access_to_packed_array.exp: set logging enabled on
42429 maint print symbols^M
42430 print-utils.cc:281:29:runtime error: negation of -9223372036854775808 cannot \
42431 be represented in type 'long int'; cast to an unsigned type to negate this \
42433 (gdb) FAIL: gdb.ada/access_to_packed_array.exp: maint print symbols
42436 By running in a debug session, we find that this happens during printing of:
42438 typedef system.storage_elements.storage_offset: \
42439 range -9223372036854775808 .. 9223372036854775807;
42441 Possibly, an ada test-case could be created that exercises this in isolation.
42443 The problem is here in int_string, where we negate a val with type LONGEST:
42445 return decimal2str ("-", -val, width);
42448 Fix this by, as recommend, using "-(ULONGEST)val" instead.
42450 Tested on x86_64-linux.
42452 2022-05-23 Tom de Vries <tdevries@suse.de>
42454 [gdb/exp] Fix UB in scalar_binop
42455 When building gdb with -fsanitize=undefined, I run into:
42457 $ gdb -q -batch -ex "p -(-0x7fffffffffffffff - 1)"
42458 src/gdb/valarith.c:1385:10: runtime error: signed integer overflow: \
42459 0 - -9223372036854775808 cannot be represented in type 'long int'
42460 $1 = -9223372036854775808
42463 Fix this by performing the substraction in scalar_binop using unsigned types.
42465 Tested on x86_64-linux.
42467 2022-05-23 Tom de Vries <tdevries@suse.de>
42469 [gdb/ada] Fix gdb.ada/dynamic-iface.exp with gcc 7
42470 This test in test-case gdb.ada/dynamic-iface.exp passes with gcc 8:
42473 $1 = (n => 3, a => "ABC", value => 93)^M
42474 (gdb) PASS: gdb.ada/dynamic-iface.exp: print local as interface
42476 but fails with gcc 7:
42480 (gdb) FAIL: gdb.ada/dynamic-iface.exp: print local as interface
42483 More concretely, we have trouble finding the type of obj. With gcc 8:
42485 $ gdb -q -batch main -ex "b concrete.adb:20" -ex run -ex "ptype obj"
42487 type = <ref> new concrete.intermediate with record
42493 type = <ref> tagged record null; end record
42496 The translation from tagged type to "full view" type happens in
42497 ada_tag_value_at_base_address, where we hit this code:
42499 /* Storage_Offset'Last is used to indicate that a dynamic offset to
42500 top is used. In this situation the offset is stored just after
42501 the tag, in the object itself. */
42502 if (offset_to_top == last)
42504 struct value *tem = value_addr (tag);
42505 tem = value_ptradd (tem, 1);
42506 tem = value_cast (ptr_type, tem);
42507 offset_to_top = value_as_long (value_ind (tem));
42510 resulting in an offset_to_top for gcc 8:
42512 (gdb) p offset_to_top
42517 (gdb) p offset_to_top
42521 The difference is expected, it bisects to gcc commit d0567dc0dbf ("[multiple
42522 changes]") which mentions this change.
42524 There's some code right after the code quoted above that deals with this
42527 else if (offset_to_top > 0)
42529 /* OFFSET_TO_TOP used to be a positive value to be subtracted
42530 from the base address. This was however incompatible with
42531 C++ dispatch table: C++ uses a *negative* value to *add*
42532 to the base address. Ada's convention has therefore been
42533 changed in GNAT 19.0w 20171023: since then, C++ and Ada
42534 use the same convention. Here, we support both cases by
42535 checking the sign of OFFSET_TO_TOP. */
42536 offset_to_top = -offset_to_top;
42539 but it's not activated because of the 'else'.
42541 Fix this by removing the 'else'.
42543 Tested on x86_64-linux, with gcc 7.5.0.
42545 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29057
42547 2022-05-23 Mark Harmstone <mark@harmstone.com>
42549 ld: use definitions in generate_reloc rather than raw literals
42551 2022-05-23 Tom de Vries <tdevries@suse.de>
42553 [gdb/testsuite] Skip language auto in gdb.base/parse_number.exp
42554 In test-case gdb.base/parse_number.exp, we skip architecture auto in the
42555 $supported_archs loop, to prevent duplicate testing.
42557 Likewise, skip language auto and its alias local in the $::all_languages
42558 loop. This reduces the number of tests from 17744 to 15572.
42560 Tested on x86_64-linux, with a build with --enable-targets=all.
42562 2022-05-23 GDB Administrator <gdbadmin@sourceware.org>
42564 Automatic date update in version.in
42566 2022-05-22 Alok Kumar Sharma <AlokKumar.Sharma@amd.com>
42568 Accept functions with DW_AT_linkage_name present
42569 Currently GDB is not able to debug (Binary generated with Clang) variables
42570 present in shared/private clause of OpenMP Task construct. Please note that
42571 LLVM debugger LLDB is able to debug.
42573 In case of OpenMP, compilers generate artificial functions which are not
42574 present in actual program. This is done to apply parallelism to block of
42577 For non-artifical functions, DW_AT_name attribute should contains the name
42578 exactly as present in actual program.
42579 (Ref# http://wiki.dwarfstd.org/index.php?title=Best_Practices)
42580 Since artificial functions are not present in actual program they not having
42581 DW_AT_name and having DW_AT_linkage_name instead should be fine.
42583 Currently GDB is invalidating any function not havnig DW_AT_name which is why
42584 it is not able to debug OpenMP (Clang).
42586 It should be fair to fallback to check DW_AT_linkage_name in case DW_AT_name
42589 2022-05-22 GDB Administrator <gdbadmin@sourceware.org>
42591 Automatic date update in version.in
42593 2022-05-21 GDB Administrator <gdbadmin@sourceware.org>
42595 Automatic date update in version.in
42597 2022-05-20 Pedro Alves <pedro@palves.net>
42599 Rename base_breakpoint -> code_breakpoint
42600 Even after the previous patches reworking the inheritance of several
42601 breakpoint types, the present breakpoint hierarchy looks a bit
42602 surprising, as we have "breakpoint" as the superclass, and then
42603 "base_breakpoint" inherits from "breakpoint". Like so, simplified:
42607 ordinary_breakpoint
42608 internal_breakpoint
42609 momentary_breakpoint
42611 exception_catchpoint
42618 The surprising part to me is having "base_breakpoint" being a subclass
42619 of "breakpoint". I'm just refering to naming here -- I mean, you'd
42620 expect that it would be the top level baseclass that would be called
42623 Just flipping the names of breakpoint and base_breakpoint around
42624 wouldn't be super great for us, IMO, given we think of every type of
42625 *point as a breakpoint at the user visible level. E.g., "info
42626 breakpoints" shows watchpoints, tracepoints, etc. So it makes to call
42627 the top level class breakpoint.
42629 Instead, I propose renaming base_breakpoint to code_breakpoint. The
42630 previous patches made sure that all code breakpoints inherit from
42631 base_breakpoint, so it's fitting. Also, "code breakpoint" contrasts
42632 nicely with a watchpoint also being typically known as a "data
42635 After this commit, the resulting hierarchy looks like:
42639 ordinary_breakpoint
42640 internal_breakpoint
42641 momentary_breakpoint
42643 exception_catchpoint
42650 ... which makes a lot more sense to me.
42652 I've left this patch as last in the series in case people want to
42653 bikeshed on the naming.
42655 "code" has a nice property that it's exactly as many letters as
42656 "base", so this patch didn't require any reindentation. :-)
42658 Change-Id: Id8dc06683a69fad80d88e674f65e826d6a4e3f66
42660 2022-05-20 Pedro Alves <pedro@palves.net>
42662 Test "set multiple-symbols on" creating multiple breakpoints
42663 To look for code paths that lead to create_breakpoints_sal creating
42664 multiple breakpoints, I ran the whole testsuite with this hack:
42666 --- a/gdb/breakpoint.c
42667 +++ b/gdb/breakpoint.c
42668 @@ -8377,8 +8377,7 @@ create_breakpoints_sal (struct gdbarch *gdbarch,
42670 int enabled, int internal, unsigned flags)
42672 - if (canonical->pre_expanded)
42673 - gdb_assert (canonical->lsals.size () == 1);
42674 + gdb_assert (canonical->lsals.size () == 1);
42676 surprisingly, the assert never failed...
42678 The way to get to create_breakpoints_sal with multiple lsals is to use
42679 "set multiple-symbols ask" and then select multiple options from the
42682 (gdb) set multiple-symbols ask
42683 (gdb) b overload1arg
42686 [2] /home/pedro/gdb/binutils-gdb/src/gdb/testsuite/gdb.cp/ovldbreak.cc:foo::overload1arg()
42687 [3] /home/pedro/gdb/binutils-gdb/src/gdb/testsuite/gdb.cp/ovldbreak.cc:foo::overload1arg(char)
42688 [4] /home/pedro/gdb/binutils-gdb/src/gdb/testsuite/gdb.cp/ovldbreak.cc:foo::overload1arg(double)
42689 [5] /home/pedro/gdb/binutils-gdb/src/gdb/testsuite/gdb.cp/ovldbreak.cc:foo::overload1arg(float)
42690 [6] /home/pedro/gdb/binutils-gdb/src/gdb/testsuite/gdb.cp/ovldbreak.cc:foo::overload1arg(int)
42691 [7] /home/pedro/gdb/binutils-gdb/src/gdb/testsuite/gdb.cp/ovldbreak.cc:foo::overload1arg(long)
42692 [8] /home/pedro/gdb/binutils-gdb/src/gdb/testsuite/gdb.cp/ovldbreak.cc:foo::overload1arg(short)
42693 [9] /home/pedro/gdb/binutils-gdb/src/gdb/testsuite/gdb.cp/ovldbreak.cc:foo::overload1arg(signed char)
42694 [10] /home/pedro/gdb/binutils-gdb/src/gdb/testsuite/gdb.cp/ovldbreak.cc:foo::overload1arg(unsigned char)
42695 [11] /home/pedro/gdb/binutils-gdb/src/gdb/testsuite/gdb.cp/ovldbreak.cc:foo::overload1arg(unsigned int)
42696 [12] /home/pedro/gdb/binutils-gdb/src/gdb/testsuite/gdb.cp/ovldbreak.cc:foo::overload1arg(unsigned long)
42697 [13] /home/pedro/gdb/binutils-gdb/src/gdb/testsuite/gdb.cp/ovldbreak.cc:foo::overload1arg(unsigned short)
42699 Breakpoint 2 at 0x1532: file /home/pedro/gdb/binutils-gdb/src/gdb/testsuite/gdb.cp/ovldbreak.cc, line 107.
42700 Breakpoint 3 at 0x154b: file /home/pedro/gdb/binutils-gdb/src/gdb/testsuite/gdb.cp/ovldbreak.cc, line 110.
42701 warning: Multiple breakpoints were set.
42702 Use the "delete" command to delete unwanted breakpoints.
42704 ... which would trigger the assert.
42706 This commit makes gdb.cp/ovldbreak.exp test this scenario. It does
42707 that by making set_bp_overloaded take a list of expected created
42708 breakpoints rather than just one breakpoint. It converts the
42709 procedure to use gdb_test_multiple instead of send_gdb/gdb_expect
42712 Change-Id: Id87d1e08feb6670440d926f5344e5081f5e37c8e
42714 2022-05-20 Pedro Alves <pedro@palves.net>
42716 Make sure momentary breakpoints are always thread-specific
42717 This adds a new ctor to momentary_breakpoints with a few parameters
42718 that are always necessary for momentary breakpoints.
42720 In particular, I noticed that set_std_terminate_breakpoint doesn't
42721 make the breakpoint be thread specific, which looks like a bug to me.
42723 The point of that breakpoint is to intercept std::terminate calls that
42724 happen as result of the called thread throwing an exception that won't
42725 be caught by the dummy frame. If some other thread calls
42726 std::terminate, IMO, it's no different from some other thread calling
42727 exit/_exit, for example.
42729 Change-Id: Ifc5ff4a6d6e58b8c4854d00b86725382d38a1a02
42731 2022-05-20 Pedro Alves <pedro@palves.net>
42733 Momentary breakpoints should have no breakpoint number
42734 Momentary breakpoints have no breakpoint number, their breakpoint
42735 number should be always 0, to avoid constantly incrementing (or
42736 decrementing) the internal breakpoint count.
42738 Indeed, set_momentary_breakpoint installs the created breakpoint
42741 However, momentary_breakpoint_from_master incorrectly gives an
42742 internal breakpoint number to the new breakpoint. This commit fixes
42745 Change-Id: Iedcae5432cdf232db9e9a6e1a646d358abd34f95
42747 2022-05-20 Pedro Alves <pedro@palves.net>
42749 Add/tweak intro comments of struct breakpoint and several subclasses
42750 This tweaks the intro comments of the following classes:
42752 internal_breakpoint
42753 momentary_breakpoint
42759 Change-Id: If6b31f51ebbb81705fbe5b8435f60ab2c88a98c8
42761 2022-05-20 Pedro Alves <pedro@palves.net>
42763 Move add_location(sal) to base_breakpoint
42764 After the previous patches, only base_breakpoint subclasses use
42765 add_location(sal), so we can move it to base_breakpoint (a.k.a. base
42766 class for code breakpoints).
42768 This requires a few casts here and there, but always at spots where
42769 you can see from context what the breakpoint's type actually is.
42771 I inlined new_single_step_breakpoint into its only caller exactly for
42774 I did try to propagate more use of base_breakpoint to avoid casts, but
42775 that turned out unwieldy for this patch.
42777 Change-Id: I49d959322b0fdce5a88a216bb44730fc5dd7c6f8
42779 2022-05-20 Pedro Alves <pedro@palves.net>
42781 Move common bits of catchpoint/exception_catchpoint to breakpoint's ctor
42782 Move common bits of catchpoint and exception_catchpoint to
42783 breakpoint's ctor, to avoid duplicating code.
42785 Change-Id: I3a115180f4d496426522f1d89a3875026aea3cf2
42787 2022-05-20 Pedro Alves <pedro@palves.net>
42789 Make catchpoint inherit breakpoint, eliminate init_raw_breakpoint
42790 struct catchpoint's ctor currently calls init_raw_breakpoint, which is
42791 a bit weird, as that ctor-like function takes a sal argument, but
42792 catchpoints don't have code locations.
42794 Instead, make struct catchpoint's ctor add the catchpoint's dummy
42795 location using add_dummy_location.
42797 init_raw_breakpoint uses add_location under the hood, and with a dummy
42798 sal it would ultimately use the breakpoint's gdbarch for the
42799 location's gdbarch, so replace the references to loc->gdbarch (which
42800 is now NULL) in syscall_catchpoint to references to the catchpoint's
42803 struct catchpoint's ctor was the last user of init_raw_breakpoint, so
42804 this commit eliminates the latter.
42806 Since catchpoint locations aren't code locations, make struct
42807 catchpoint inherit struct breakpoint instead of base_breakpoint. This
42808 let's us delete the tracepoint::re_set override too.
42810 Change-Id: Ib428bf71efb09fdaf399c56e4372b0f41d9c5869
42812 2022-05-20 Pedro Alves <pedro@palves.net>
42814 Make breakpoint_address_bits look at the location kind
42815 Software watchpoints allocate a special dummy location using
42816 software_watchpoint_add_no_memory_location, and then
42817 breakpoint_address_bits checks whether the location is that special
42818 location to decide whether the location has a meaninful address to
42821 Introduce a new bp_loc_software_watchpoint location kind, and make
42822 breakpoint_address_bits use bl_address_is_meaningful instead, which
42823 returns false for bp_loc_other, which is in accordance with we
42824 document for bp_location::address:
42826 /* (... snip ...) Valid for all types except
42828 CORE_ADDR address = 0;
42830 Rename software_watchpoint_add_no_memory_location to
42831 add_dummy_location, and simplify it. This will be used by catchpoints
42832 too in a following patch.
42834 Note that neither "info breakpoints" nor "maint info breakpoints"
42835 actually prints the addresses of watchpoints, but I think it would be
42836 useful to do so in "maint info breakpoints". This approach let's us
42837 implement that in the future.
42839 Change-Id: I50e398f66ef618c31ffa662da755eaba6295aed7
42841 2022-05-20 Pedro Alves <pedro@palves.net>
42843 Make exception_catchpoint inherit base_breakpoint instead of catchpoint
42844 exception_catchpoint is really a code breakpoint, with locations set
42845 by sals, re-set like other code breakpoints, etc., so make it inherit
42848 This adds a bit of duplicated code to exception_catchpoint's ctor
42849 (copied from struct catchpoint's ctor), but it will be eliminated in a
42852 Change-Id: I9fbb2927491120e9744a4f5e5cb5e6870ca07009
42854 2022-05-20 Pedro Alves <pedro@palves.net>
42856 Refactor momentary breakpoints, eliminate set_raw_breakpoint{,_without_location}
42857 This commit makes set_momentary_breakpoint allocate the breakpoint
42858 type without relying on set_raw_breakpoint, and similarly,
42859 momentary_breakpoint_from_master not rely on
42860 set_raw_breakpoint_without_location. This will let us convert
42861 init_raw_breakpoint to a ctor in a following patch.
42863 The comment about set_raw_breakpoint being used in gdbtk sources is
42864 stale. gdbtk no longer uses it.
42866 Change-Id: Ibbf77731e4b22e18ccebc1b5799bbec0aff28c8a
42868 2022-05-20 Pedro Alves <pedro@palves.net>
42870 Refactor set_internal_breakpoint / internal_breakpoint ctor
42871 This moves initialization of internal_breakpoint's breakpoint fields
42872 to internal_breakpoint's ctor, and stops using
42873 new_breakpoint_from_type for internal_breakpoint breakpoints.
42875 Change-Id: I898ed0565f47cb00e4429f1c6446e6f9a385a78d
42877 2022-05-20 Pedro Alves <pedro@palves.net>
42879 Convert init_ada_exception_catchpoint to a ctor
42880 Currently, init_ada_exception_catchpoint is defined in breakpoint.c, I
42881 presume so it can call the static describe_other_breakpoints function.
42882 I think this is a dependency inversion.
42883 init_ada_exception_catchpoint, being code specific to Ada catchpoints,
42884 should be in ada-lang.c, and describe_other_breakpoints, a core
42885 function, should be exported.
42887 And then, we can convert init_ada_exception_catchpoint to an
42888 ada_catchpoint ctor.
42890 Change-Id: I07695572dabc5a75d3d3740fd9b95db1529406a1
42892 2022-05-20 Pedro Alves <pedro@palves.net>
42894 Make ada_catchpoint_location's owner ctor parameter be ada_catchpoint
42895 This commit changes ada_catchpoint_location's ctor from:
42897 ada_catchpoint_location (breakpoint *owner)
42901 ada_catchpoint_location (ada_catchpoint *owner)
42903 just to make the code better document intention.
42905 To do this, we need to move the ada_catchpoint_location type's
42906 definition to after ada_catchpoint is defined, otherwise the compiler
42907 doesn't know that ada_catchpoint is convertible to struct breakpoint.
42909 Change-Id: Id908b2e38bde30b262381e00c5637adb9bf0129d
42911 2022-05-20 Pedro Alves <pedro@palves.net>
42913 init_breakpoint_sal -> base_breakpoint::base_breakpoint
42914 This converts init_breakpoint_sal to a base_breakpoint constructor.
42916 It removes a use of init_raw_breakpoint.
42918 To avoid manually adding a bunch of parameters to
42919 new_breakpoint_from_type, and manually passing them down to the
42920 constructors of a number of different base_breakpoint subclasses, make
42921 new_breakpoint_from_type a variable template function.
42923 Change-Id: I4cc24133ac4c292f547289ec782fc78e5bbe2510
42925 2022-05-20 Pedro Alves <pedro@palves.net>
42927 Remove "internal" parameter from a couple functions
42928 None of init_breakpoint_sal, create_breakpoint_sal, and
42929 strace_marker_create_breakpoints_sal make use of their "internal"
42930 parameter, so remove it.
42932 Change-Id: I943f3bb44717ade7a7b7547edf8f3ff3c37da435
42934 2022-05-20 Pedro Alves <pedro@palves.net>
42936 More breakpoint_ops parameter elimination
42937 Remove breakpoint_ops parameters from a few functions that don't need
42940 Change-Id: Ifcf5e1cc688184acbf5e19b8ea60138ebe63cf28
42942 2022-05-20 Pedro Alves <pedro@palves.net>
42944 Make a few functions work with base_breakpoint instead of breakpoint
42945 This makes tracepoints inherit from base_breakpoint, since their
42946 locations are code locations. If we do that, then we can eliminate
42947 tracepoint::re_set and tracepoint::decode_location, as they are doing
42948 the same as the base_breakpoint implementations.
42950 With this, all breakpoint types created by new_breakpoint_from_type
42951 are code breakpoints, i.e., base_breakpoint subclasses, and thus we
42952 can make it return a base_breakpoint pointer.
42954 Finally, init_breakpoint_sal can take a base_breakpoint pointer as
42955 "self" pointer too. This will let us convert this function to a
42956 base_breakpoint ctor in a following patch.
42958 Change-Id: I3a4073ff1a4c865f525588095c18dc42b744cb54
42960 2022-05-20 Pedro Alves <pedro@palves.net>
42962 ranged_breakpoint: move initialization to ctor
42963 Move initialization of ranged_breakpoint's fields to its ctor.
42965 Change-Id: If7b842861f3cc6a429ea329d45598b5852283ba3
42967 2022-05-20 Pedro Alves <pedro@palves.net>
42969 ranged_breakpoint: use install_breakpoint
42970 This commit replaces a chunk of code in break_range_command by an
42971 equivalent call to install_breakpoint.
42973 Change-Id: I31c06cabd36f5be91740aab029265f678aa78e35
42975 2022-05-20 Pedro Alves <pedro@palves.net>
42977 ranged_breakpoint: don't use init_raw_breakpoint
42978 ranged_breakpoint's ctor already sets the breakpoint's type to
42979 bp_hardware_breakpoint.
42981 Since this is a "regular" breakpoint, b->pspace should remain NULL.
42983 Thus, the only thing init_raw_breakpoint is needed for, is to add the
42984 breakpoint's location. Do that directly.
42986 Change-Id: I1505de94c3919881c2b300437e2c0da9b05f76bd
42988 2022-05-20 Pedro Alves <pedro@palves.net>
42990 Make structs breakpoint/base_breakpoint/catchpoint be abstract
42991 You should never instanciate these types directly.
42993 Change-Id: I8086c74c415eadbd44924bb0ef20f34b5b97ee6f
42995 2022-05-20 Pedro Alves <pedro@palves.net>
42997 add_location_to_breakpoint -> breakpoint::add_location
42998 Make add_location_to_breakpoint be a method of struct breakpoint.
43000 A patch later in the series will move this to base_breakpoint, but for
43001 now, it needs to be here.
43003 Change-Id: I5bdc2ec1a7c2d66f26f51bf6f6adc8384a90b129
43005 2022-05-20 Carl Love <cel@us.ibm.com>
43007 PowerPC: Make test gdb.arch/powerpc-power10.exp Endian independent.
43008 The .quad statement stores the 64-bit hex value in Endian order. When used
43009 to store a 64-bit prefix instructions on Big Endian (BE) systems, the .quad
43010 statement stores the 32-bit suffix followed by the 32-bit prefix rather
43011 than the expected order of prefix word followed by the suffix word. GDB
43012 fetches 32-bits at a time when disassembling instructions. The disassembly
43013 on BE gets messed up since GDB fetches the suffix first and interprets it
43014 as a word instruction not a prefixed instruction. When gdb fetches the
43015 prefix part of the instruction, following the initial suffix word, gdb
43016 associates the prefix word incorrectly with the following 32-bits as the
43017 suffix for the instruction when in fact it is the following instruction.
43019 For example on BE we have two prefixed instructions stored using the
43020 .quad statement as follows:
43022 addr word GDB action
43023 ---------------------------------------------
43024 1 suffix inst A <- GDB interprets as a word instruction
43025 2 prefix inst A <- GDB uses this prefix with
43027 3 suffix inst B <- this suffix rather than the suffix at addr 1.
43030 This patch changes the .quad statement into two .longs to explicitly store
43031 the prefix followed by the suffix of the instruction.
43033 The patch rearranges the instructions to put all of the word instructions
43034 together followed by the prefix instructions for clarity.
43036 The patch has been tested on Power 10 and Power 7 BE and LE to verify
43037 the change works as expected.
43039 2022-05-20 Tom Tromey <tromey@adacore.com>
43041 Remove obsolete text from documentation
43042 The documentation says that -enable-pretty-printing is experimental in
43043 7.0 and may change -- that's long enough ago that I think we can say
43044 that this text is no longer correct or useful.
43046 2022-05-20 Nick Clifton <nickc@redhat.com>
43048 Stop readekf and objdump from aggressively following links.
43049 * dwarf.c (dwarf_select_sections_by_names): Return zero if no
43050 sections were selected.
43051 (dwarf_select_sections_by_letters): Likewise.
43052 * dwarf.h: (dwarf_select_sections_by_names): Update prototype.
43053 (dwarf_select_sections_by_letters): Update prototype.
43054 * objdump.c (might_need_separate_debug_info): New function.
43055 (dump_bfd): Call new function before attempting to load separate
43057 (main): Do not enable dwarf section dumping for -WK or -WN.
43058 * readelf.c (parse_args): Do not enable dwarf section dumping for
43060 (might_need_separate_debug_info): New function.
43061 (process_object): Call new function before attempting to load
43062 separate debug info files.
43063 * testsuite/binutils-all/debuginfo.exp: Expect -WE and -wE
43064 debuginfod tests to pass.
43065 * testsuite/binutils-all/objdump.Wk: Add extra regexps.
43066 * testsuite/binutils-all/readelf.k: Add extra regexps.
43068 2022-05-20 Jia-Wei Chen <jiawei@iscas.ac.cn>
43070 RISC-V: Update zfinx implement with zicsr.
43071 Update zfinx implement with zicsr, fix missing fcsr use by zfinx.
43072 add zicsr imply by zfinx.
43076 * elfxx-riscv.c: New imply.
43080 * testsuite/gas/riscv/csr-insns-pseudo-zfinx.d: New test.
43084 * riscv-opc.c: Update insn class.
43086 2022-05-20 Tsukasa OI <research_trasio@irq.a4lg.com>
43088 RISC-V: Remove RV128-only fmv instructions
43089 As fmv.x.q and fmv.q.x instructions are RV128-only (not RV64-only),
43090 it should be removed until RV128 support for GNU Binutils is required
43095 * testsuite/gas/riscv/fmv.x.q-rv64-fail.d: New failure test.
43096 * testsuite/gas/riscv/fmv.x.q-rv64-fail.l: Likewise.
43097 * testsuite/gas/riscv/fmv.x.q-rv64-fail.s: Likewise.
43101 * opcode/riscv-opc.h (MATCH_FMV_X_Q, MASK_FMV_X_Q,
43102 MATCH_FMV_Q_X, MASK_FMV_Q_X): Remove RV128-only instructions.
43106 * riscv-opc.c (riscv_opcodes): Remove RV128-only instructions.
43108 2022-05-20 Aditya Vidyadhar Kamath <ADITYA.VIDYADHAR.KAMATH@ibm.com>
43110 Fix non-pointer type compilation error in aix-thread.c
43111 In aix-thread.c we use ms->value_address () to get the symbol address.
43112 This triggers the following compiler error...
43114 base operand of '->' has non-pointer type 'bound_minimal_symbol'
43116 ... because ms is not a pointer.
43118 This commit fixes this error by using ms.value_address () instead.
43120 2022-05-20 Steinar H. Gunderson <sesse@google.com>
43122 add a trie to map quickly from address range to compilation unit
43123 When using perf to profile large binaries, _bfd_dwarf2_find_nearest_line()
43124 becomes a hotspot, as perf wants to get line number information
43125 (for inline-detection purposes) for each and every sample. In Chromium
43126 in particular (the content_shell binary), this entails going through
43127 475k address ranges, which takes a long time when done repeatedly.
43129 Add a radix-256 trie over the address space to quickly map address to
43130 compilation unit spaces; for content_shell, which is 1.6 GB when some
43131 (but not full) debug information turned is on, we go from 6 ms to
43132 0.006 ms (6 µs) for each lookup from address to compilation unit, a 1000x
43135 There is a modest RAM increase of 180 MB in this binary (the existing
43136 linked list over ranges uses about 10 MB, and the entire perf job uses
43137 between 2–3 GB for a medium-size profile); for smaller binaries with few
43138 ranges, there should be hardly any extra RAM usage at all.
43140 2022-05-20 Alan Modra <amodra@gmail.com>
43142 Tidy warn-execstack handling
43143 Make ld and bfd values consistent by swapping values 0 and 2 in
43144 link_info.warn_execstack. This has the benefit of making the value an
43145 "extended" boolean, with 0 meaning no warning, 1 meaning warn, other
43146 values a conditional warning.
43148 Yes, this patch introduces fails on arm/aarch64. Not a problem with
43149 this patch but an arm/aarch64 before_parse problem.
43152 * elflink.c (bfd_elf_size_dynamic_sections): Adjust
43153 warn_execstack test.
43155 * bfdlink.h (warn_execstack): Swap 0 and 2 meaning.
43157 * configure.ac (DEFAULT_LD_WARN_EXECSTACK): Use values of 0,
43158 1, 2 consistent with link_info.warn_execstack.
43159 * ld.texi: Typo fixes.
43160 * lexsup.c (parse_args): Adjust setting of link_info.warn_execstack.
43161 (elf_static_list_options): Adjust help message conditions.
43162 * configure: Regenerate.
43164 2022-05-20 GDB Administrator <gdbadmin@sourceware.org>
43166 Automatic date update in version.in
43168 2022-05-19 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
43170 arm: Fix system register fpcxt_ns and fpcxt_s naming convention.
43171 The current assembler accepts system registers FPCXTNS and FPCXTS for Armv8.1-M
43172 Mainline Instructions VSTR, VLDR, VMRS and VMSR.
43173 Assembler should be also allowing FPCXT_NS, fpcxt_ns, fpcxtns, FPCXT_S, fpcxt_s
43174 and fpcxts. This patch fixes the issue.
43176 2022-05-19 Andrew Burgess <aburgess@redhat.com>
43178 gdb/doc: use @value{GDBP} in 'info pretty-printer' example
43179 Update the 'info pretty-printer' example in the manual to make use of
43180 @value{GDBP} instead of hard-coding '(gdb)'.
43182 2022-05-19 Andrew Burgess <aburgess@redhat.com>
43184 gdb/doc: make use of group/end group in 'info pretty-printers' example
43185 The 'info pretty-printers' example is pretty long and consists of many
43186 commands and their output.
43188 Currently, when the pdf manual is generated this example spans a
43189 page-break, with the page-break falling part way through some example
43192 This commit breaks up the example using @group .... @end group, within
43193 each group is a single GDB command and all its output.
43195 Now, when the pdf manual is created, the page-break is placed after
43196 the output of one GDB command, and before the subsequent command, this
43199 2022-05-19 Nikolaos Chatzikonstantinou <nchatz314@gmail.com>
43201 gdb/doc: fix inconsistent info pretty-printer example
43202 The example for 'info pretty-printer' in the manual passes an
43203 object-regexp in some cases, but presents output as though no
43204 object-regexp was passed.
43206 This commit fixes the two mistakes, in one case, fixing the output to
43207 filter based on object-regexp, and in the other, to remove the
43208 object-regexp from the command and leave all the output.
43210 2022-05-19 Nick Clifton <nickc@redhat.com>
43212 Fix potentially uninitialised variables in the Windows tools
43214 2022-05-19 Tiezhu Yang <yangtiezhu@loongson.cn>
43216 gdb: testsuite: Support displaced stepping on LoongArch
43217 When execute the following command on LoongArch:
43219 make check-gdb TESTS="gdb.base/async-shell.exp"
43221 we can see the following message in gdb/testsuite/gdb.sum:
43223 UNSUPPORTED: gdb.base/async-shell.exp: displaced stepping
43225 modify support_displaced_stepping to support displaced stepping
43230 PASS: gdb.base/async-shell.exp: run &
43231 PASS: gdb.base/async-shell.exp: shell echo foo
43232 PASS: gdb.base/async-shell.exp: interrupt
43233 PASS: gdb.base/async-shell.exp: process stopped
43235 I did the following tests that use support_displaced_stepping
43236 with this patch on LoongArch, there is no failed testcases.
43238 loongson@linux:~/gdb.git$ grep -r support_displaced_stepping gdb/testsuite/gdb.*
43239 gdb/testsuite/gdb.arch/disp-step-insn-reloc.exp:if { ![support_displaced_stepping] } {
43240 gdb/testsuite/gdb.base/step-over-no-symbols.exp: if { $displaced != "off" && ![support_displaced_stepping] } {
43241 gdb/testsuite/gdb.base/moribund-step.exp:if { ![support_displaced_stepping] } {
43242 gdb/testsuite/gdb.base/async-shell.exp:if { ![support_displaced_stepping] } {
43243 gdb/testsuite/gdb.base/inferior-died.exp:if { ![support_displaced_stepping] } {
43244 gdb/testsuite/gdb.base/step-over-syscall.exp: if {$displaced == "on" && ![support_displaced_stepping]} {
43245 gdb/testsuite/gdb.mi/mi-watch-nonstop.exp:if { ![support_displaced_stepping] } {
43246 gdb/testsuite/gdb.mi/mi-ns-stale-regcache.exp:if { ![support_displaced_stepping] } {
43247 gdb/testsuite/gdb.mi/mi-nonstop.exp:if { ![support_displaced_stepping] } {
43248 gdb/testsuite/gdb.mi/mi-nsmoribund.exp:if { ![support_displaced_stepping] } {
43249 gdb/testsuite/gdb.mi/mi-nsintrall.exp:if { ![support_displaced_stepping] } {
43250 gdb/testsuite/gdb.mi/mi-nsthrexec.exp:if { ![support_displaced_stepping] } {
43251 gdb/testsuite/gdb.mi/mi-nonstop-exit.exp:if { ![support_displaced_stepping] } {
43252 gdb/testsuite/gdb.multi/watchpoint-multi.exp:if [support_displaced_stepping] {
43253 gdb/testsuite/gdb.python/py-evthreads.exp:if { ![support_displaced_stepping] } {
43254 gdb/testsuite/gdb.threads/step-over-lands-on-breakpoint.exp: if { $displaced != "off" && ![support_displaced_stepping] } {
43255 gdb/testsuite/gdb.threads/interrupt-while-step-over.exp: if { ${displaced-stepping} != "off" && ![support_displaced_stepping] } {
43256 gdb/testsuite/gdb.threads/step-over-trips-on-watchpoint.exp: if { $displaced != "off" && ![support_displaced_stepping] } {
43258 2022-05-19 Simon Marchi <simon.marchi@polymtl.ca>
43260 gdbsupport: fix path_join crash with -std=c++17 and -D_GLIBCXX_DEBUG
43261 When building GDB with -std=c++17 and -D_GLIBCXX_DEBUG=1, I get:
43263 $ ./gdb -nx --data-directory=data-directory -q -ex "maint selftest path_join"
43264 /usr/include/c++/11.2.0/string_view:233: constexpr const value_type& std::basic_string_view<_CharT, _Traits>::operator[](std::basic_string_view<_CharT, _Traits>::size_type) const [with _CharT = char; _Traits = std::char_traits<char>; std::basic_string_view<_CharT, _Traits>::const_reference = const char&; std::basic_string_view<_CharT, _Traits>::size_type = long unsigned int]: Assertion '__pos < this->_M_len' failed.
43266 The problem is that we're passing an empty string_view to
43267 IS_ABSOLUTE_PATH. IS_ABSOLUTE_PATH accesses [0] on that string_view,
43268 which is out-of-bounds.
43270 The reason this is not seen with -std less than c++17 is that our local
43271 copy of string_view (used with C++ < 17) does not have the assert in
43272 operator[], as that wouldn't work in a constexpr method:
43274 https://gitlab.com/gnutools/binutils-gdb/-/blob/5890af36e5112bcbb8d7555e63570f68466e6944/gdbsupport/gdb_string_view.h#L180
43276 IS_ABSOLUTE_PATH is normally used with null-terminated string. It's
43277 fine to pass an empty null-terminated string to IS_ABSOLUTE_PATH,
43278 because index 0 in such a string is valid. But not with an empty
43281 Fix that by avoiding the "call" to IS_ABSOLUTE_PATH if the string_view
43284 Change-Id: Idf4df961b63f513b3389235e93814c02b89ea32e
43286 2022-05-19 Jan Beulich <jbeulich@suse.com>
43288 Arm64: force emission of ILP32-dependent relocs
43289 Like the placeholder types added in 04dfe7aa5217 ("Arm64: follow-on to
43290 PR gas/27217 fix"), these are also placeholders which are subsequently
43291 resolved (albeit later, hence this being a separate issue). As for the
43292 resolved types 1 is returned, these pseudo-relocs should also have 1
43293 returned to force retaining of the [eventual] relocations. This is also
43294 spelled out individually for each of them in md_apply_fix().
43296 2022-05-19 Jan Beulich <jbeulich@suse.com>
43298 COFF: use hash for string table also when copying / stripping
43299 Otherwise the string table may grow and hence e.g. change a final binary
43300 (observed with PE/COFF ones) even if really there's no change. Doing so
43301 in fact reduces the overall amount of code, and in particular the number
43302 of places which need to remain in sync.
43304 Afaics there's no real equivalent to the "traditional_format" field used
43305 when linking, so hashing is always enabled when copying / stripping.
43307 2022-05-19 Jan Beulich <jbeulich@suse.com>
43309 COFF/PE: keep linker version during objcopy / strip
43310 Neither of the tools is really a linker, so whatever was originally
43311 recorded should be retained rather than being overwritten by these
43314 COFF/PE: don't leave zero timestamp after objcopy / strip
43315 Fill the timestamp field suitably for _bfd_XXi_only_swap_filehdr_out().
43316 Instead of re-arranging the present if(), fold this logic with that of
43317 copying the optional header.
43319 2022-05-19 Jan Beulich <jbeulich@suse.com>
43321 COFF: make objcopy / strip honor --keep-file-symbols
43322 So far this option had no effect when used together with e.g.
43323 --strip-debug. Set BSF_FILE on these symbols to change that.
43325 While altering this also join two adjacent blocks of case labeled
43326 statements with identical code.
43328 2022-05-19 Jan Beulich <jbeulich@suse.com>
43330 don't over-align file positions of PE executable sections
43331 When a sufficiently small alignment was specified via --file-alignment,
43332 individual section alignment shouldn't affect placement within the file.
43333 This involves first of all clearing D_PAGED for images when section and
43334 file alignment together don't permit paging of the image. The involved
43335 comparison against COFF_PAGE_SIZE in turn helped point out (through a
43336 compiler warning) that 'page_size' should be of unsigned type (as in
43337 particular FileAlignment is). This yet in turn pointed out a dubious
43338 error condition (which is being deleted).
43340 For the D_PAGED case I think the enforced file alignment may still be
43341 too high, but I'm wary of changing that logic without knowing of
43342 possible corner cases.
43344 Furthermore file positions in PE should be independent of the alignment
43345 recorded in section headers anyway. Otherwise there are e.g. anomalies
43346 following commit 6f8f6017a0c4 ("PR27567, Linking PE files adds alignment
43347 section flags to executables") in that linking would use information a
43348 subsequent processing step (e.g. stripping) wouldn't have available
43349 anymore, and hence a binary could change in that 2nd step for no actual
43350 reason. (Similarly stripping a binary linked with a linker pre-dating
43351 that commit would change the binary again when stripping it a 2nd time.)
43353 2022-05-19 Yvan Roux <yvan.roux@foss.st.com>
43355 _bfd_real_fopen should not use ccs parameter on Windows
43357 * bfdio.c (_bfd_real_fopen): Delete ccs string.
43359 2022-05-19 Tsukasa OI <research_trasio@irq.a4lg.com>
43361 RISC-V: Fix canonical extension order (K and J)
43362 This commit fixes canonical extension order to follow the RISC-V ISA
43363 Manual draft-20210402-1271737 or later.
43367 * elfxx-riscv.c (riscv_recognized_prefixed_ext): Fix "K" extension
43368 prefix to be placed before "J".
43370 2022-05-19 GDB Administrator <gdbadmin@sourceware.org>
43372 Automatic date update in version.in
43374 2022-05-18 John Baldwin <jhb@FreeBSD.org>
43376 Use aarch64_features to describe register features in target descriptions.
43377 Replace the sve bool member of aarch64_features with a vq member that
43378 holds the vector quotient. It is zero if SVE is not present.
43380 Add std::hash<> specialization and operator== so that aarch64_features
43381 can be used as a key with std::unordered_map<>.
43383 Change the various functions that create or lookup aarch64 target
43384 descriptions to accept a const aarch64_features object rather than a
43385 growing number of arguments.
43387 Replace the multi-dimension tdesc_aarch64_list arrays used to cache
43388 target descriptions with unordered_maps indexed by aarch64_feature.
43390 2022-05-18 Jan Beulich <jbeulich@suse.com>
43392 Arm64: follow-on to PR gas/27217 fix
43395 Prior to trying to address PR gas/28888 I noticed anomalies in how
43396 certain insns would / wouldn't be affected in similar ways.
43398 Commit eac4eb8ecb26 ("Fix a problem assembling AArch64 sources when a
43399 relocation is generated against a symbol that has a defined value") had
43400 two copy-and-paste mistakes, passing the wrong type to
43401 aarch64_force_reloc().
43403 It further failed to add placeholder relocation types to that function's
43404 block of case labels leading to a return of 1. While not of interest for
43405 aarch64_force_relocation() (these placeholders are resolved right in
43406 parse_operands()), calls to aarch64_force_reloc() happen before that
43407 resolution would take place.
43409 2022-05-18 Nick Clifton <nickc@redhat.com>
43411 Fix compile time warning building gold with Clang-14.
43412 * int_encoding.cc (get_length_as_unsigned_LEB_128): Remove
43413 current_length variable.
43415 2022-05-18 Victor Do Nascimento <victor.donascimento@arm.com>
43417 oops - forgot changelog entry for the previous delta.
43419 arm: Add unwind support for mixed register lists
43420 * config/tc-arm.c (parse_reg_list): Add handling of mixed register
43422 (reg_names): Enumerate pseudoregister according to mapped physical
43424 (s_arm_unwind_save_pseudo): Modify function signature.
43425 (s_arm_unwind_save_core): Likewise.
43426 (s_arm_unwind_save_mixed): New function.
43427 (s_arm_unwind_save): Generate register list mask to pass to nested
43429 * testsuite/gas/arm/unwind-pacbti-m.s: Expand test for mixed
43430 register type lists.
43431 * testsuite/gas/arm/unwind-pacbti-m.d: Likewise.
43432 * testsuite/gas/arm/unwind-pacbti-m-readelf.d: Likewise.
43434 2022-05-18 Carl Love <cel@us.ibm.com>
43436 PowerPC: bp-permanent.exp, kill-after-signal fix
43437 Fix changes that didn't make it into commit:
43438 dd9cd55e990bcc9f8448cac38d242d53974b3604.
43440 Fix missing -wrap on gdb_test_multiple in gdb.base/kill-after-signal.exp
43441 that is causing regression test on x86_64-linux with taskset -c 0.
43443 2022-05-18 Yichao Yu <yyc1992@gmail.com>
43445 [AArch64] Return the regnum for PC (32) on aarch64
43446 This will allow the unwind info to explicitly specify a different value
43447 for the return address from the link register.
43448 Such usage, although uncommon, is valid and useful for signal frames.
43449 It is also supported by aadwarf64 from ARM (Note 9 in [1]).
43451 Ref https://sourceware.org/pipermail/gdb/2022-May/050091.html
43453 [1] https://github.com/ARM-software/abi-aa/blob/2022Q1/aadwarf64/aadwarf64.rst#dwarf-register-names
43455 2022-05-18 Jan Beulich <jbeulich@suse.com>
43457 x86: shrink op_riprel
43458 It is only ever initialized from a boolean, so it as well as related
43459 variables' types can simply be bool and there's no masking to 32 bits
43460 needed in set_op().
43462 2022-05-18 Nick Clifton <nickc@redhat.com>
43464 Add a --no-weak option to nm.
43466 * nm.c (non_weak): New variable.
43467 (filter_symbols): When non-weak is true, ignore weak symbols.
43468 (long_options): Add --no-weak.
43469 (usage): Mention --no-weak.
43470 (main): Handle -W/--no-weak.
43471 * doc/binutils.texi: Document new feature.
43472 * NEWS: Mention the new feature.
43473 * testsuite/binutils-all/nm.exp: Add test of new feature.
43474 * testsuite/binutils-all/no-weak.s: New test source file.
43476 2022-05-18 Pedro Alves <pedro@palves.net>
43478 Support -prompt and -lbl in gdb_test
43479 This teaches gdb_test to forward the -prompt and -lbl options to
43482 The option parsing is done with parse_args.
43484 As a cleanup, instead of using llength and lindex to get at the
43485 positional arguments, use lassign, and check whether the corresponding
43488 Convert gdb.base/ui-redirect.exp and gdb.xml/tdesc-reload.exp to use
43489 gdb_test -prompt/-lbl instead of gdb_test_multiple as examples.
43491 Change-Id: I243e1296d32c05a421ccef30b63d43a89eaeb4a0
43493 2022-05-18 Luis Machado <luis.machado@arm.com>
43495 Remove unused DWARF PAUTH registers
43496 The AARCH64_DWARF_PAUTH_DMASK and AARCH64_DWARF_PAUTH_CMASK DWARF registers
43497 never made their way into the aadwarf64. The following patch removes these
43498 constants and their use.
43500 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=26295
43502 2022-05-18 Luis Machado <luis.machado@arm.com>
43504 Rename PAUTH_RA_STATE to RA_SIGN_STATE
43505 The aadwarf64 [1] names this register RA_SIGN_STATE, so update the code to use
43508 [1] https://github.com/ARM-software/abi-aa/blob/main/aadwarf64/aadwarf64.rst
43510 2022-05-18 Tom de Vries <tdevries@suse.de>
43512 [gdb/testsuite] Simplify unknown lang testing in gdb.base/parse_number.exp
43513 Move testing of language unknown out of the $supported_archs loop in
43514 gdb.base/parse_number.exp. This reduces total amount of tests from 18466 to
43517 Tested on x86_64-linux.
43519 2022-05-18 Tom de Vries <tdevries@suse.de>
43521 [gdb/testsuite] Use hex_for_lang in gdb.base/parse_number.exp
43522 In gdb.base/parse_number.exp, add a new proc hex_for_lang that formats a hex
43523 number appropriately for a given language.
43525 Tested on x86_64-linux.
43527 2022-05-18 Tom de Vries <tdevries@suse.de>
43529 [gdb/tdep] Add gdb/syscalls/update-linux-from-src.sh
43530 Add a new script gdb/syscalls/update-linux-from-src.sh, that can be used to
43531 generate *-linux.xml.in files from linux kernel sources, like so:
43533 $ ./update-linux-from-src.sh ~/upstream/linux-stable.git
43534 Skipping aarch64-linux.xml.in, no syscall.tbl
43535 Generating amd64-linux.xml.in
43536 Skipping arm-linux.xml.in, use arm-linux.py instead
43537 Skipping bfin-linux.xml.in, no longer supported
43538 Generating i386-linux.xml.in
43539 Generating mips-n32-linux.xml.in
43540 Generating mips-n64-linux.xml.in
43541 Generating mips-o32-linux.xml.in
43542 Generating ppc64-linux.xml.in
43543 Generating ppc-linux.xml.in
43544 Generating s390-linux.xml.in
43545 Generating s390x-linux.xml.in
43546 Generating sparc64-linux.xml.in
43547 Generating sparc-linux.xml.in
43550 Update *-linux.xml.in and *-linux.xml using linux kernel tag v5.18-rc6.
43552 2022-05-18 Tamar Christina <tamar.christina@arm.com>
43554 AArch64: Enable FP16 by default for Armv9-A.
43555 In Armv9-A SVE is mandatory, and for SVE FP16 is mandatory. This fixes a disconnect
43556 between GCC and binutils where GCC has FP16 on by default and gas doesn't.
43560 2022-05-16 Tamar Christina <tamar.christina@arm.com>
43562 * opcode/aarch64.h (AARCH64_ARCH_V9_FEATURES): Add AARCH64_FEATURE_F16.
43564 2022-05-18 Jan Beulich <jbeulich@suse.com>
43566 gas: avoid octal numbers being accepted when processing .linefile
43567 Compilers would put decimal numbers there, so I think we should treat
43568 finding octal numbers the same as finding bignums - ignore them as
43569 actually being comments of some very specific form.
43571 2022-05-18 Jan Beulich <jbeulich@suse.com>
43573 gas: avoid bignum related errors when processing .linefile
43574 Any construct which to the scrubber looks like a C preprocessor
43575 line/file "directive" is converted to .linefile, but the amount of
43576 checking the scrubber does is minimal (albeit it does let through only
43577 decimal digits for the line part of the contruct). Since the scrubber
43578 conversion is further tied to # being a line comment character, anything
43579 which upon closer inspection turns out not to be a line/file "directive"
43580 is supposed to be treated as a comment, i.e. ignored. Therefore we
43581 cannot use get_absolute_expression(), as this may raise errors. Open-
43582 code the function instead, treating everything not resulting in
43583 O_constant as a comment as well.
43585 Furthermore also bounds-check the parsed value. This bounds check tries
43586 to avoid implementation defined behavior (which may be the raising of an
43587 implementation defined signal), but for now makes the assumption that
43588 int has less than 64 bits. The way bfd_signed_vma (which is what offsetT
43589 aliases) is defined in bfd.h for the BFD64 case I cannot really see a
43590 clean way of avoiding this assumption. Omitting the #ifdef, otoh, would
43591 risk "condition is always false" warnings by compilers.
43593 Convert get_linefile_number() to return bool at this occasion as well.
43595 2022-05-18 Jan Beulich <jbeulich@suse.com>
43597 gas: fold do_repeat{,_with_expander}()
43598 do_repeat_with_expander() already deals with the "no expander" case
43599 quite fine, so there's really little point having two functions. What it
43600 lacks compared with do_repeat() is a call to sb_build(), which can
43601 simply be moved (and the then redundant sb_new() be avoided). Along with
43602 this moving also flip if the main if()'s condition such that the "no
43603 expander" case is handled first.
43605 gas: don't ignore .linefile inside false conditionals
43606 When assembling code previously pre-processed by a C compiler, long
43607 enough comments may have been collapsed into "# <line> <file>"
43608 constructs. If we skip these, line numbers (and possibly even file
43609 names) will be off / wrong in both diagnostics and debug info.
43611 gas: simplify ignore_input()
43612 First of all convert to switch(), in preparation of adding another
43613 directive here which may not be ignored. While doing so drop dead code:
43614 A string the first two characters of which do not match "if" also wont
43615 match "ifdef" or "ifndef".
43617 gas: tweak .irp and alike file/line handling for M68K/MRI
43618 In commit 2ee1792bec22 ("gas: further adjust file/line handling for .irp
43619 and alike") I neglected the need to omit the leading . in M68K/MRI mode.
43621 2022-05-18 Xi Ruoyao <xry111@mengyan1223.wang>
43623 gold: don't invoke IA32 syscall in x86_64 assembly testcase
43624 pr17704a_test.s is a x86_64 assembly file, but it invokes IA32 exit
43625 syscall with "int 0x80". This causes a segfault on kernels with
43626 CONFIG_IA32_EMULATION disabled.
43630 * testsuite/pr17704a_test.s (_start): Invoke x86_64 exit syscall
43631 instead of its IA32 counterpart.
43633 2022-05-18 GDB Administrator <gdbadmin@sourceware.org>
43635 Automatic date update in version.in
43637 2022-05-17 Nikolaos Chatzikonstantinou <nchatz314@gmail.com>
43639 Fix typo in info page
43641 2022-05-17 Pedro Alves <pedro@palves.net>
43643 Fix gdb.python/py-connection.exp with remote targets
43644 After the patch to make gdb_test's question non-optional when
43645 specified, gdb.python/py-connection.exp started failing like so:
43647 $ make check TESTS="gdb.python/py-connection.exp" RUNTESTFLAGS="--target_board=native-gdbserver"
43648 (gdb) PASS: gdb.python/py-connection.exp: info connections while the connection is still around
43650 Ending remote debugging.^M
43651 (gdb) FAIL: gdb.python/py-connection.exp: kill the inferior
43653 The problem is that "disconnect" when debugging with the native target
43654 asks the user whether to kill the program, while with remote targets,
43657 Fix it by explicitly killing before disconnecting.
43659 Tested with --target_board unix, native-gdbserver, and native-extended-gdbserver.
43661 Change-Id: Icd85015c76deb84b71894715d43853c1087eba0b
43663 2022-05-17 Felix Willgerodt <felix.willgerodt@intel.com>
43665 gdb, btrace: Throw an error for empty recordings when replaying starts.
43666 This makes record_btrace_start_replaying() more consistent, as it already
43667 errors out e.g. on a recording with only gaps.
43669 2022-05-17 Pedro Alves <pedro@palves.net>
43671 Make gdb_test's question non-optional if specified
43672 gdb_test supports handling scenarios where GDB asks a question before
43673 finishing handling some command. The full prototype of gdb_test is:
43675 # gdb_test COMMAND PATTERN MESSAGE QUESTION RESPONSE
43677 However, QUESTION is a question that GDB _may_ ask, not one that GDB
43680 # QUESTION is a question GDB may ask in response to COMMAND, like
43682 # RESPONSE is the response to send if QUESTION appears.
43684 If GDB doesn't raise the question, the test still passes.
43686 I think that this is a misfeature. If GDB regresses and stops asking
43687 a question, the testsuite won't notice. So I think that if a QUESTION
43688 is specified, gdb_test should ensure it comes out of GDB.
43690 Running the testsuite exposed a number of tests that pass
43691 QUESTION/RESPONSE to GDB, but no question comes out. The previous
43692 commits fixed them all, so this commit changes gdb_test's behavior.
43694 A related issue is that gdb_test doesn't enforce that if you specify
43695 QUESTION, that you also specify RESPONSE. I.e., you should pass 1, 2,
43696 3, or 5 arguments to gdb_test, but never 4, or more than 5. Making
43697 gdb_test detect bogus arguments actually regressed some testcases,
43698 also all fixed in previous commits.
43700 Change-Id: I47c39c9034e6a6841129312037a5ca4c5811f0db
43702 2022-05-17 Pedro Alves <pedro@palves.net>
43704 gdb.base/skip.exp: Don't abuse gdb_test's question support
43705 gdb.base/skip.exp abuses gdb_test's support for answering a GDB
43706 question to do this:
43708 # With gcc 9.2.0 we jump once back to main before entering foo here.
43709 # If that happens try to step a second time.
43710 gdb_test "step" "foo \\(\\) at.*" "step 3" \
43711 "main \\(\\) at .*\r\n$gdb_prompt " "step"
43713 After a patch later in this series, gdb_test will FAIL if GDB does NOT
43714 issue the question, so this test would start failing on older GCCs.
43716 Switch to using gdb_test_multiple instead. There are three spots in
43717 the file that have the same pattern, and they're actually in a
43718 sequence of commands that is repeated those 3 times. Factor all that
43719 out to a procedure.
43721 I don't have gcc 9.2 handy, but I do have gcc 6.5, and that one is
43722 affected as well, so update the comment.
43724 Change-Id: If0a7e3cdf5191b4eec95ce0c8845c3a4d801c39e
43726 2022-05-17 Pedro Alves <pedro@palves.net>
43728 Avoid having to unload file in gdb.server/connect-with-no-symbol-file.exp
43729 gdb.server/connect-with-no-symbol-file.exp's connect_no_symbol_file
43732 gdb_test "file" ".*" "discard symbol table" \
43733 {Discard symbol table from `.*'\? \(y or n\) } "y"
43735 A following patch will make gdb_test expect the question out of GDB if
43736 one is passed down as argument to gdb_test. With that, this test
43737 starts failing. This is because connect_no_symbol_file is called in a
43738 loop, and the first time around, there's a loaded file, so "file" asks
43739 the "Discard symbol table ... ?" question, while in the following
43740 iterations there's no file, so there's no question.
43742 Fix this by not loading a file into GDB in the first place.
43744 Change-Id: I810c036b57842c4c5b47faf340466b0d446d1abc
43746 2022-05-17 Pedro Alves <pedro@palves.net>
43748 Fix bogus gdb_test invocations
43749 A following patch will make gdb_test error out if bogus arguments are
43750 passed, which exposed bugs in a few testcases:
43752 - gdb.python/py-parameter.exp, passing a spurious "1" as extra
43753 parameter, resulting in:
43755 ERROR: Unexpected arguments: {set test-file-param bar.txt} {The name of the file has been changed to bar.txt} {set new file parameter} 1
43757 - gdb.python/py-xmethods.exp, a missing test message, resulting in
43758 the next gdb_test being interpreted as message, question and
43759 response! With the enforcing patch, this was caught with:
43761 ERROR: Unexpected arguments: {p g.mul<char>('a')} {From Python G<>::mul.*} gdb_test {p g_ptr->mul<char>('a')} {From Python G<>::mul.*} {after: g_ptr->mul<char>('a')}
43763 - gdb.base/pointers.exp, missing a quote.
43765 Change-Id: I66f2db4412025a64121db7347dfb0b48240d46d4
43767 2022-05-17 Pedro Alves <pedro@palves.net>
43769 gdb.base/scope.exp: Remove bogus gdb_test questions
43770 This test is abusing the QUESTION/RESPONSE feature to send an
43771 alternative command to GDB if the first command fails. Like so:
43773 gdb_test "print 'scope0.c'::filelocal" \
43774 "\\\$$decimal = 1" "print 'scope0.c'::filelocal at main" \
43775 "No symbol \"scope0.c\" in current context.*" \
43776 "print '$srcdir/$subdir/scope0.c'::filelocal"
43778 So if 'scope0.c' doesn't work, we try again with
43779 '$srcdir/$subdir/scope0.c'. I strongly suspect this is really an
43780 obsolete test. I think that if '$srcdir/$subdir/scope0.c' works, then
43781 'scope0.c' should have worked too, thus I'd think that if we pass due
43782 to the question path, then it's a bug. So just remove the question
43783 part passed to gdb_test.
43785 Change-Id: I2acc99285f1d519284051b49693b5441fbdfe3cd
43787 2022-05-17 Pedro Alves <pedro@palves.net>
43789 Remove gdb_test questions that GDB doesn't ask
43790 Change-Id: Ib2616dc883e9dc9ee100f6c86d83a921a0113c16
43792 2022-05-17 Nelson Chu <nelson.chu@sifive.com>
43794 RISC-V: Added half-precision floating-point v1.0 instructions.
43796 * elfxx-riscv.c (riscv_implicit_subsets): Added implicit f
43798 (riscv_supported_std_z_ext): Added default v1.0 version for zfh.
43799 (riscv_multi_subset_supports): Handle INSN_CLASS_ZFH,
43800 INSN_CLASS_D_AND_ZFH and INSN_CLASS_Q_AND_ZFH.
43802 * config/tc-riscv.c (FLT_CHARS): Added "hH".
43803 (macro): Expand Pseudo M_FLH and M_FSH.
43804 (riscv_pseudo_table): Added .float16 directive.
43805 * testsuite/gas/riscv/float16-be.d: New testcase for .float16.
43806 * testsuite/gas/riscv/float16-le.d: Likewise.
43807 * testsuite/gas/riscv/float16.s: Likewise.
43808 * testsuite/gas/riscv/fp-zfh-insns.d: New testcase for zfh.
43809 * testsuite/gas/riscv/fp-zfh-insns.s: Likewise.
43811 * opcode/riscv-opc.h: Added MASK and MATCH encodings for zfh.
43812 * opcode/riscv.h: Added INSN_CLASS and pseudo macros for zfh.
43814 * riscv-opc.c (riscv_opcodes): Added zfh instructions.
43816 2022-05-17 GDB Administrator <gdbadmin@sourceware.org>
43818 Automatic date update in version.in
43820 2022-05-16 Ilya Leoshkevich <iii@linux.ibm.com>
43822 IBM zSystems: Fix left-shifting negative PCRel32 values (PR gas/29152)
43823 s390_insert_operand ()'s val, min and max are encoded PCRel32 values
43824 and need to be left-shifted by 1 before being shown to the user.
43825 Left-shifting negative values is undefined behavior in C, but the
43826 current code does not try to prevent it, causing UBSan to complain.
43828 Fix by casting the values to their unsigned equivalents before
43831 2022-05-16 Pedro Alves <pedro@palves.net>
43833 Reindent gdbsupport/event-loop.cc:handle_file_event
43834 The handle_file_event function has a few unnecessary {} lexical
43835 blocks, presumably because they were originally if blocks, and the
43836 conditions were removed, or something along those lines.
43838 Remove the unnecessary blocks, and reindent.
43840 Change-Id: Iaecbe5c9f4940a80b81dbbc42e51ce506f6aafb2
43842 2022-05-16 Pedro Alves <pedro@palves.net>
43844 gdbsupport/event-loop.cc: simplify !HAVE_POLL paths
43845 gdbsupport/event-loop.cc throughout handles the case of use_poll being
43846 true on a system where HAVE_POLL is not defined, by calling
43847 internal_error if that situation ever happens.
43849 Simplify this by moving the "use_poll" global itself under HAVE_POLL,
43850 so that it's way more unlikely to ever end up in such a situation.
43851 Then, move the code that checks the value of use_poll under HAVE_POLL
43852 too, and remove the internal_error calls. Like, from:
43859 internal_error (....);
43860 #endif /* HAVE_POLL */
43875 #endif /* HAVE_POLL */
43880 While at it, make use_poll be a bool. The current code is using
43881 unsigned char most probably to save space, but I don't think it really
43884 Co-Authored-By: Youling Tang <tangyouling@loongson.cn>
43885 Change-Id: I0dd74fdd4d393ccd057906df4cd75e8e83c1cdb4
43887 2022-05-16 Eli Zaretskii <eliz@gnu.org>
43889 gdb: Fix typo in last change in gdb.texinfo
43891 gdb: Document the 'metadata' styling in GDB displays.
43892 The 'metadata' styling was never documented in the GDB manual.
43893 This fills that gap.
43895 2022-05-16 Tom Tromey <tromey@adacore.com>
43897 Fix Ada exception regression on Windows
43898 The breakpoint c++-ification series introduced another bug in Ada --
43899 it caused "catch exception" and related commands to fail on Windows.
43900 The problem is that the re_set method calls the wrong superclass
43901 method, so the breakpoint doesn't get correctly re-set when the
43902 runtime offsets change. This patch fixes the problem.
43904 2022-05-16 Bruno Larsen <blarsen@redhat.com>
43906 gdb/testsuite: fix "continue outside of loop" TCL errors
43907 Many test cases had a few lines in the beginning that look like:
43913 Where conditions varied, but were mostly in the form of ![runto_main] or
43914 [skip_*_tests], making it quite clear that this code block was supposed
43915 to finish the test if it entered the code block. This generates TCL
43916 errors, as most of these tests are not inside loops. All cases on which
43917 this was an obvious mistake are changed in this patch.
43919 2022-05-16 GDB Administrator <gdbadmin@sourceware.org>
43921 Automatic date update in version.in
43923 2022-05-15 GDB Administrator <gdbadmin@sourceware.org>
43925 Automatic date update in version.in
43927 2022-05-14 GDB Administrator <gdbadmin@sourceware.org>
43929 Automatic date update in version.in
43931 2022-05-13 Tom Tromey <tromey@adacore.com>
43933 Remove unused field cooked_index::m_start
43934 cooked_index::m_start is unused and can be removed. I think this was
43935 a leftover from a previous approach in the index finalization code,
43936 and then when rewriting it I forgot to remove it.
43938 Tested by rebuilding.
43940 2022-05-13 Tom Tromey <tromey@adacore.com>
43942 Implement pid_to_exec_file for Windows in gdbserver
43943 I noticed that gdbserver did not implement pid_to_exec_file for
43944 Windows, while gdb did implement it. This patch moves the code to
43945 nat/windows-nat.c, so that it can be shared. This makes the gdbserver
43946 implementation trivial.
43948 Remove windows_process_info::id
43949 I noticed that windows_process_info::id is only used by gdbserver, and
43950 not really necessary. This patch removes it.
43952 Constify target_pid_to_exec_file
43953 This changes target_pid_to_exec_file and target_ops::pid_to_exec_file
43954 to return a "const char *". I couldn't build many of these targets,
43955 but did examine the code by hand -- also, as this only affects the
43956 return type, it's normally pretty safe. This brings gdb and gdbserver
43957 a bit closer, and allows for the removal of a const_cast as well.
43959 Put corefile-run.core into test subdirectory
43960 I noticed that corefile-run.core ends up in the 'runtest' directory.
43961 It's better, when at all possible, for test files to end up in the
43962 test's designated subdirectory. This patch makes this change.
43964 2022-05-13 Tom Tromey <tromey@adacore.com>
43966 Do not double-read minimal symbols for PE COFF
43967 This changes coffread.c to avoid re-reading minimal symbols when
43968 possible. This only works when there are no COFF symbols to be read,
43969 but at least for my mingw builds of gdb, this seems to be the case.
43971 Tested using the AdaCore internal test suite on Windows. I also did
43972 some local builds to ensure that no warnings crept in.
43974 2022-05-13 Pedro Alves <pedro@palves.net>
43976 Fix "gdb --write" with core files
43977 If you load a core file into GDB with the --write option, or "set
43978 write on" (equivalent), and then poke memory expecting it to patch the
43979 core binary, you'll notice something odd -- the write seems to
43980 succeed, but in reality, it doesn't. The value you wrote doesn't
43983 $ gdb -q --write -c testsuite/outputs/gdb.base/patch/gcore.test
43985 Core was generated by `/home/pedro/gdb/build/gdb/testsuite/outputs/gdb.base/patch/patch'.
43986 Program terminated with signal SIGTRAP, Trace/breakpoint trap.
43987 #0 0x0000555555555131 in ?? ()
43988 (gdb) p *(unsigned char *)0x0000555555555131 = 1
43990 (gdb) p *(unsigned char *)0x0000555555555131
43994 Diffing hexdumps of before/after patching, reveals that a "0x1" was
43995 actually written somewhere in the file. The problem is that the "0x1"
43996 was written at the wrong offset in the file...
43998 That happens because _bfd_elf_set_section_contents does this to seek
43999 to the section's offset:
44001 pos = hdr->sh_offset + offset;
44002 if (bfd_seek (abfd, pos, SEEK_SET) != 0
44003 || bfd_bwrite (location, count, abfd) != count)
44006 ... and 'hdr->sh_offset' is zero, so we seek to just OFFSET, which is
44007 incorrect. The reason 'hdr->sh_offset' is zero is that
44008 kernel-generated core files normally don't even have a section header
44009 table (gdb-generated ones do, but that's more an accident than a
44010 feature), and indeed elf_core_file_p doesn't even try to read sections
44013 /* Core files are simply standard ELF formatted files that partition
44014 the file using the execution view of the file (program header table)
44015 rather than the linking view. In fact, there is no section header
44016 table in a core file.
44018 The process status information (including the contents of the general
44019 register set) and the floating point register set are stored in a
44020 segment of type PT_NOTE. We handcraft a couple of extra bfd sections
44021 that allow standard bfd access to the general registers (.reg) and the
44022 floating point registers (.reg2). */
44025 elf_core_file_p (bfd *abfd)
44027 Changing _bfd_elf_set_section_contents from:
44029 pos = hdr->sh_offset + offset;
44033 pos = section->filepos + offset;
44035 fixes it. If we do that however, the tail end of
44036 _bfd_elf_set_section_contents ends up as a copy of
44037 _bfd_generic_set_section_contents, so just call the latter, thus
44038 eliminating some duplicate code.
44040 New GDB testcase included, which exercises both patching an executable
44041 and patching a core file. Patching an executable already works
44042 without this fix, because in that case BFD reads in the sections
44043 table. Still, we had no testcase for that yet. In fact, we have no
44044 "set write on" testcases at all, this is the first one.
44046 Tested on x86-64 GNU/Linux, gdb, ld, binutils, and gas.
44048 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=18227
44049 Change-Id: I0f49f58b48aabab2e269f2959b8fd8a7fe36fdce
44051 2022-05-13 Alan Modra <amodra@gmail.com>
44053 Import libiberty from gcc
44055 sim: remove use of PTR
44056 PTR will soon disappear from ansidecl.h. Remove uses in sim. Where
44057 a PTR cast is used in assignment or function args to a void* I've
44058 simply removed the unnecessary (in C) cast rather than replacing with
44061 2022-05-13 GDB Administrator <gdbadmin@sourceware.org>
44063 Automatic date update in version.in
44065 2022-05-13 Alan Modra <amodra@gmail.com>
44067 gdb: remove use of PTR
44068 PTR will disappear from ansidecl.h and libiberty on the next import
44069 from gcc. Remove current uses in gdb.
44071 2022-05-12 Tom de Vries <tdevries@suse.de>
44073 [gdb/testsuite] Fix gdb.cp/break-f-std-string.cc with older gcc
44074 When running test-case gdb.cp/break-f-std-string.exp on openSUSE Leap 15.3
44075 with system gcc 7.5.0, I run into:
44077 (gdb) whatis /r std::string^M
44078 No symbol "string" in namespace "std".^M
44079 (gdb) FAIL: gdb.cp/break-f-std-string.exp: _GLIBCXX_USE_CXX11_ABI=1: \
44080 whatis /r std::string
44082 The same for gcc 8.2.1, but it passes with gcc 9.3.1.
44084 At source level (as we can observe in the .ii file with -save-temps) we have
44088 namespace __cxx11 {
44089 typedef basic_string<char> string;
44093 while with gcc 9.3.1, we have instead:
44096 namespace __cxx11 {
44099 typedef basic_string<char> string;
44102 due to gcc commit 33b43b0d8cd ("Define std::string and related typedefs
44103 outside __cxx11 namespace").
44105 Fix this by adding the missing typedef for gcc version 5 (the first version to
44106 have the dual abi) to 8 (the last version missing aforementioned gcc commit).
44108 Tested on x86_64-linux, with:
44110 - gcc 4.8.5, 8.2.1, 9.3.1, 10.3.0, 11.2.1
44111 - clang 8.0.1, 12.0.1
44113 2022-05-12 Alan Modra <amodra@gmail.com>
44115 Fix an illegal memory access when creating DLLs.
44117 * pe-dll.c (dll_name): Delete, replacing with..
44118 (dll_filename): ..this, moved earlier in file.
44119 (generate_edata): Delete parameters. Don't set up dll_name here..
44120 (pe_process_import_defs): ..instead set up dll_filename and
44121 dll_symname here before returning.
44122 (dll_symname_len): Delete write-only variable.
44123 (pe_dll_generate_implib): Don't set up dll_symname here.
44125 2022-05-12 Mark Wielaard <mark@klomp.org>
44127 gdb: Workaround stringop-overread warning in debuginfod-support.c on powerpc64
44128 Just like on s390x with g++ 11.2.1, ppc64le with g++ 11.3.1 produces a
44129 spurious warning for stringop-overread in debuginfod_is_enabled
44130 for url_view. Also suppress it on powerpc64.
44134 * debuginfod-support.c (debuginfod_is_enabled): Use
44135 DIAGNOSTIC_IGNORE_STRINGOP_OVERREAD on powerpc64.
44137 2022-05-12 Luis Machado <luis.machado@arm.com>
44139 Make gdb.ada/float-bits.exp more generic
44140 There are assumptions in the test about the long double format
44141 being used. While the results are OK for i387 128-bit long doubles, it
44142 is not correct for IEEE quad 128-bit long doubles.
44144 Also, depending on the target (64-bit/32-bit), long doubles may not
44145 be available at all. And it may be the case that the compiler for a 64-bit
44146 target doesn't support 128-bit long doubles, but GDB might still support it
44149 Lastly, not every long double format has invalid values. Some formats
44150 consider all values as valid floating point numbers.
44152 These divergences cause the following FAIL's on aarch64/arm:
44154 FAIL: gdb.ada/float-bits.exp: print val_long_double
44155 FAIL: gdb.ada/float-bits.exp: print val_long_double after assignment
44157 With the above in mind, extend the test a little so it behaves well on
44158 different architectures and so it works with different long double
44163 - Use long double values appropriate for the long double format.
44164 - Test long double assignment to compiler-generated long
44166 - Test long double assignment to GDB internal variables.
44168 Tested on x86_64 (16 PASS), i686 (16 PASS), aarch64 (12 PASS) and arm (9 PASS).
44170 2022-05-12 Tom de Vries <tdevries@suse.de>
44172 [gdb/tdep] Improve gdb/syscalls/update-linux.sh
44173 Fix two things in update-linux.sh:
44174 - remove use of unnecessary tmp file
44175 - inline gen-header.py into update-linux.sh
44177 Tested on x86_64-linux.
44179 2022-05-12 Tom de Vries <tdevries@suse.de>
44181 [gdb/testsuite] Fix gdb.dwarf2/dw2-out-of-range-end-of-seq.exp on aarch64
44182 On aarch64-linux, with test-case gdb.dwarf2/dw2-out-of-range-end-of-seq.exp I
44186 Starting program: dw2-out-of-range-end-of-seq ^M
44188 Program received signal SIGILL, Illegal instruction.^M
44189 main () at src/gdb/testsuite/gdb.dwarf2/main.c:1^M
44190 1 /* This testcase is part of GDB, the GNU debugger.^M
44191 (gdb) FAIL: gdb.dwarf2/dw2-out-of-range-end-of-seq.exp: runto: run to main
44194 There are two problems here:
44195 - the test-case contains a hardcoded "DW_LNS_advance_pc 1" which causes the
44196 breakpoint pointing in the middle of an insn
44197 - the FAIL triggers on aarch64-linux, but not on x86_64-linux, because the
44198 test-case uses 'main_label' as the address of the first and only valid entry
44199 in the line table, and:
44200 - on aarch64-linux, there's no prologue, so main_label and main coincide,
44202 - on x86_64-linux, there's a prologue, so main_label is different from main.
44204 Fix these problems by:
44205 - eliminating the use of "DW_LNS_advance_pc 1", and using
44206 "DW_LNE_set_address $main_end" instead, and
44207 - eliminating the use of main_label, using "DW_LNE_set_address $main_start"
44210 Tested on both x86_64-linux and aarch64-linux.
44212 2022-05-12 Alan Modra <amodra@gmail.com>
44214 cgen: increase buffer for hash_insn_list
44215 As was done for hash_insn_array in commit d3d1cc7b13b4.
44217 * cgen-dis.c (hash_insn_list): Increase size of buf. Assert
44218 size is large enough.
44220 2022-05-12 Alan Modra <amodra@gmail.com>
44222 PR29142, segv in ar with empty archive and libdeps specified
44224 * ar.c (main): Properly handle libdeps for zero file_count.
44226 2022-05-12 Alan Modra <amodra@gmail.com>
44228 Re: IBM zSystems: Accept (. - 0x100000000) PCRel32 operands
44229 The new test failed on s390-linux due to bfd_sprintf_vma trimming
44230 output to 32 bits for 32-bit targets. The test was faulty anyway,
44231 expecting zero as the min end of the range is plainly wrong, but
44232 that's what you get if you cast min to int.
44234 * config/tc-s390.c (s390_insert_operand): Print range error using
44236 * testsuite/gas/s390/zarch-z900-err.l: Correct expected output.
44238 2022-05-12 GDB Administrator <gdbadmin@sourceware.org>
44240 Automatic date update in version.in
44242 2022-05-11 Tom de Vries <tdevries@suse.de>
44244 [gdb/testsuite] Fix gdb.base/catch-syscall.exp with --with-expat=no
44245 When doing a gdb build with --with-expat=no, I run into:
44247 (gdb) PASS: gdb.base/catch-syscall.exp: determine pipe syscall: \
44248 continue to breakpoint: before pipe call
44249 catch syscall pipe^M
44250 Unknown syscall name 'pipe'.^M
44251 (gdb) PASS: gdb.base/catch-syscall.exp: determine pipe syscall: \
44253 catch syscall pipe2^M
44254 Unknown syscall name 'pipe2'.^M
44255 (gdb) PASS: gdb.base/catch-syscall.exp: determine pipe syscall: \
44256 catch syscall pipe2
44259 [Detaching after vfork from child process 18538]^M
44260 [Inferior 1 (process 18537) exited normally]^M
44261 (gdb) FAIL: gdb.base/catch-syscall.exp: determine pipe syscall: continue
44264 This is a regression since recent commit 5463a15c18b ("[gdb/testsuite] Handle
44265 pipe2 syscall in gdb.base/catch-syscall.exp").
44267 Fix this by using pipe/pipe2 syscall numbers instead.
44269 Tested on x86_64-linux.
44271 2022-05-11 Nick Clifton <nickc@redhat.com>
44273 nm: use -U as an alias for --defines-only, in line with llvm-nm
44275 2022-05-11 Tom de Vries <tdevries@suse.de>
44277 [gdb/testsuite] Fix gdb.base/catch-syscall.exp without --enable-targets
44278 When doing a gdb build without --enable-targets, I run into:
44280 (gdb) UNSUPPORTED: gdb.base/catch-syscall.exp: multiple targets: \
44281 s390:31-bit vs s390:64-bit: set architecture s390:64-bit
44282 delete breakpoints^M
44283 (gdb) info breakpoints^M
44284 No breakpoints or watchpoints.^M
44285 (gdb) break -qualified main^M
44286 No symbol table is loaded. Use the "file" command.^M
44287 Make breakpoint pending on future shared library load? (y or [n]) n^M
44288 (gdb) FAIL: gdb.base/catch-syscall.exp: gdb_breakpoint: set breakpoint at main
44291 The problem is that due to recent commit e21d8399303 ("[gdb/testsuite] Remove
44292 target limits in gdb.base/catch-syscall.exp") "clean_restart $binfile" no
44293 longer is called at the end of test_catch_syscall_multi_arch.
44295 Fix this by moving "clean_restart $binfile" back to
44296 test_catch_syscall_multi_arch.
44298 Tested on x86_64-linux.
44300 2022-05-11 Tom de Vries <tdevries@suse.de>
44302 [gdb/testsuite] Fix gdb.base/maint.exp on powerpc64le
44303 On powerpc64le-linux, I ran into:
44305 FAIL: gdb.base/maint.exp: maint print objfiles: symtabs
44308 The problem is that:
44309 - the "Cooked index in use" line occurs twice in the gdb output:
44310 - once for exec maint, and
44311 - once for "Object file system-supplied DSO".
44312 - the matching of the second "Cooked index in use" also consumes
44313 the "Symtabs:" string, and consequently the corresponding
44314 clause does not trigger and $symtabs remains 0.
44316 Fix this by limiting the output of the command to the exec.
44318 Tested on x86_64-linux and powerpcle-linux.
44320 2022-05-11 Tom de Vries <tdevries@suse.de>
44322 [gdb/tdep] Update syscalls/{ppc64,ppc}-linux.xml
44323 Regenerate syscalls/{ppc64,ppc}-linux.xml on a system with 5.14 kernel.
44325 2022-05-11 Tom de Vries <tdevries@suse.de>
44327 [gdb/testsuite] Remove target limits in gdb.base/catch-syscall.exp
44328 In test-case gdb.base/catch-syscall.exp, proc test_catch_syscall_multi_arch we
44329 test for supported targets using istarget, like so:
44331 if { [istarget "i*86-*-*"] || [istarget "x86_64-*-*"] } {
44333 } elseif { [istarget "powerpc-*-linux*"] \
44334 || [istarget "powerpc64*-linux*"] } {
44337 but the tests excercised there can all be executed if gdb is configured with
44338 --enable-targets=all.
44340 Rewrite the proc to iterate over all cases, and check if the test is supported
44341 by trying "set arch $arch1" and "set arch $arch2".
44343 Tested on x86_64-linux, with:
44344 - a gdb build with --enable-targets=all, and
44345 - a gdb build build with my usual --enable-targets setting (too long to
44346 include here) which means the sparc vs sparc:v9 case is unsupported.
44348 2022-05-11 Tom de Vries <tdevries@suse.de>
44350 [gdb/record] Handle statx system call
44351 When running test-case gdb.reverse/fstatat-reverse.exp with target board
44352 unix/-m32 on openSUSE Tumbleweed, I run into:
44354 (gdb) PASS: gdb.reverse/fstatat-reverse.exp: set breakpoint at marker2
44357 Process record and replay target doesn't support syscall number 383^M
44358 Process record: failed to record execution log.^M
44361 0xf7fc5555 in __kernel_vsyscall ()^M
44362 (gdb) FAIL: gdb.reverse/fstatat-reverse.exp: continue to breakpoint: marker2
44365 The problems is that while with native we're trying to record these syscalls
44366 (showing strace output):
44368 openat(AT_FDCWD, "/", O_RDONLY|O_PATH) = 3
44369 newfstatat(3, ".", {st_mode=S_IFDIR|0755, st_size=146, ...}, 0) = 0
44371 with unix/-m32 we have instead:
44373 openat(AT_FDCWD, "/", O_RDONLY|O_PATH) = 3
44374 statx(3, ".", AT_STATX_SYNC_AS_STAT|AT_NO_AUTOMOUNT, STATX_BASIC_STATS, \
44375 {stx_mask=STATX_ALL|STATX_MNT_ID, stx_attributes=STATX_ATTR_MOUNT_ROOT, \
44376 stx_mode=S_IFDIR|0755, stx_size=146, ...}) = 0
44378 and statx is not supported.
44380 Fix this by adding support for recording syscall statx.
44382 Tested on x86_64-linux.
44384 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28461
44386 2022-05-11 Alan Modra <amodra@gmail.com>
44388 opcodes cgen: remove use of PTR
44389 Note that opcodes is regenerated with cgen commit d1dd5fcc38e reverted,
44390 due to failure of bpf to compile with that patch applied.
44392 .../opcodes/bpf-opc.c:57:11: error: conversion from ‘long unsigned int’ to ‘unsigned int’ changes value from ‘18446744073709486335’ to ‘4294902015’ [-Werror=overflow]
44393 57 | 64, 64, 0xffffffffffff00ff, { { F (F_IMM32) }, { F (F_OFFSET16) }, { F (F_SRCLE) }, { F (F_OP_CODE) }, { F (F_DSTLE) }, { F (F_OP_SRC) }, { F (F_OP_CLASS) }, { 0 } }
44394 plus other similar errors.
44397 * mep.opc (print_tpreg, print_spreg): Delete unnecessary
44398 forward declarations. Replace PTR with void *.
44399 * mt.opc (print_dollarhex, print_pcrel): Delete forward decls.
44401 * bpf-desc.c, * bpf-dis.c, * cris-desc.c,
44402 * epiphany-desc.c, * epiphany-dis.c,
44403 * fr30-desc.c, * fr30-dis.c, * frv-desc.c, * frv-dis.c,
44404 * ip2k-desc.c, * ip2k-dis.c, * iq2000-desc.c, * iq2000-dis.c,
44405 * lm32-desc.c, * lm32-dis.c, * m32c-desc.c, * m32c-dis.c,
44406 * m32r-desc.c, * m32r-dis.c, * mep-desc.c, * mep-dis.c,
44407 * mt-desc.c, * mt-dis.c, * or1k-desc.c, * or1k-dis.c,
44408 * xc16x-desc.c, * xc16x-dis.c,
44409 * xstormy16-desc.c, * xstormy16-dis.c: Regenerate.
44411 2022-05-11 GDB Administrator <gdbadmin@sourceware.org>
44413 Automatic date update in version.in
44415 2022-05-10 Youling Tang <tangyouling@loongson.cn>
44417 gdb: mips: Fix large-frame.exp test case failure
44418 $ objdump -d outputs/gdb.base/large-frame/large-frame-O2
44419 0000000120000b20 <func>:
44420 120000b20: 67bdbff0 daddiu sp,sp,-16400
44421 120000b24: ffbc4000 sd gp,16384(sp)
44422 120000b28: 3c1c0002 lui gp,0x2
44423 120000b2c: 679c8210 daddiu gp,gp,-32240
44424 120000b30: 0399e02d daddu gp,gp,t9
44425 120000b34: df998058 ld t9,-32680(gp)
44426 120000b38: ffbf4008 sd ra,16392(sp)
44427 120000b3c: 0411ffd8 bal 120000aa0 <blah>
44430 The disassembly of the above func function shows that we may use
44431 instructions such as daddiu/daddu, so add "daddiu $gp,$gp,n",
44432 "daddu $gp,$gp,$t9" and "daddu $gp,$t9,$gp" to the mips32_scan_prologue
44433 function to fix the large-frame.exp test case.
44435 Before applying the patch:
44438 #0 blah (a=0xfffffee220) at .../gdb/testsuite/gdb.base/large-frame-1.c:24
44439 #1 0x0000000120000b44 in func ()
44440 Backtrace stopped: frame did not save the PC
44441 (gdb) FAIL: gdb.base/large-frame.exp: optimize=-O2: backtrace
44443 # of expected passes 5
44444 # of unexpected failures 1
44446 After applying the patch:
44448 # of expected passes 6
44450 2022-05-10 Tiezhu Yang <yangtiezhu@loongson.cn>
44452 gdb: LoongArch: Use GDB style to check readbuf and writebuf
44453 The GDB style is to write 'if (readbuf != nullptr)', and the same for
44456 2022-05-10 Tom Tromey <tromey@adacore.com>
44458 Fix --disable-threading build
44459 PR build/29110 points out that GDB fails to build on mingw when the
44460 "win32" thread model is in use. It turns out that the Fedora cross
44461 tools using the "posix" thread model, which somehow manages to support
44462 std::future, whereas the win32 model does not.
44464 While looking into this, I found that the configuring with
44465 --disable-threading will also cause a build failure.
44467 This patch fixes this build by introducing a compatibility wrapper for
44470 I am not able to test the win32 thread model build, but I'm going to
44471 ask the reporter to try this patch.
44473 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29110
44475 2022-05-10 Pedro Alves <pedro@palves.net>
44477 Fix "b f(std::string)" when current language is C
44478 If you try to set a breakpoint at a function such as "b
44479 f(std::string)", and the current language is C, the breakpoint fails
44480 to be set, like so:
44482 (gdb) set language c
44483 break f(std::string)
44484 Function "f(std::string)" not defined.
44485 Make breakpoint pending on future shared library load? (y or [n]) n
44488 The problem is that the code in GDB that expands the std::string
44489 typedef hits this in c-typeprint.c:
44491 /* If we have "typedef struct foo {. . .} bar;" do we want to
44492 print it as "struct foo" or as "bar"? Pick the latter for
44493 C++, because C++ folk tend to expect things like "class5
44494 *foo" rather than "struct class5 *foo". We rather
44495 arbitrarily choose to make language_minimal work in a C-like
44497 if (language == language_c || language == language_minimal)
44499 if (type->code () == TYPE_CODE_UNION)
44500 gdb_printf (stream, "union ");
44501 else if (type->code () == TYPE_CODE_STRUCT)
44503 if (type->is_declared_class ())
44504 gdb_printf (stream, "class ");
44506 gdb_printf (stream, "struct ");
44508 else if (type->code () == TYPE_CODE_ENUM)
44509 gdb_printf (stream, "enum ");
44512 I.e., std::string is expanded to "class std::..." instead of just
44513 "std::...", and then the "f(class std::..." symbol doesn't exist.
44515 Fix this by making cp-support.c:inspect_type print the expanded
44516 typedef type using the language of the symbol whose type we're
44517 expanding the typedefs for -- in the example in question, the
44518 "std::string" typedef symbol, which is a C++ symbol.
44520 Use type_print_raw_options as it seems to me that in this scenario we
44521 always want raw types, to match the real symbol names.
44523 Adjust the gdb.cp/break-f-std-string.exp testcase to try setting a
44524 breakpoint at "f(std::string)" in both C and C++.
44526 Change-Id: Ib54fab4cf0fd307bfd55bf1dd5056830096a653b
44528 2022-05-10 Pedro Alves <pedro@palves.net>
44530 Always pass an explicit language down to c_type_print
44531 The next patch will want to do language->print_type(type, ...), to
44532 print a type in a given language, avoiding a dependency on the current
44533 language. That doesn't work correctly currently, however, because
44534 most language implementations of language_defn::print_type call
44535 c_print_type without passing down the language. There are two
44536 overloads of c_print_type, one that takes a language, and one that
44537 does not. The one that does not uses the current language, defeating
44538 the point of calling language->print_type()...
44540 This commit removes the c_print_type overload that does not take a
44541 language, and adjusts the codebase throughout to always pass down a
44542 language. In most places, there's already an enum language handy.
44543 language_defn::print_type implementations naturally pass down
44544 this->la_language. In a couple spots, like in ada-typeprint.c and
44545 rust-lang.c there's no enum language handy, but the code is written
44546 for a specific language, so we just hardcode the language.
44548 In gnuv3_print_method_ptr, I wasn't sure whether we could hardcode C++
44549 here, and we don't have an enum language handy, so I made it use the
44550 current language, just like today. Can always be improved later.
44552 Change-Id: Ib54fab4cf0fd307bfd55bf1dd5056830096a653b
44554 2022-05-10 Pedro Alves <pedro@palves.net>
44556 Fix "b f(std::string)", always use DMGL_VERBOSE
44557 Currently, on any remotely modern GNU/Linux system,
44558 gdb.cp/no-dmgl-verbose.exp fails like so:
44560 break 'f(std::string)'
44561 Function "f(std::string)" not defined.
44562 (gdb) FAIL: gdb.cp/no-dmgl-verbose.exp: gdb_breakpoint: set breakpoint at 'f(std::string)'
44563 break 'f(std::basic_string<char, std::char_traits<char>, std::allocator<char> >)'
44564 Function "f(std::basic_string<char, std::char_traits<char>, std::allocator<char> >)" not defined.
44565 (gdb) PASS: gdb.cp/no-dmgl-verbose.exp: DMGL_VERBOSE-demangled f(std::string) is not defined
44567 This testcase was added back in 2011, here:
44569 [patch] Remove DMGL_VERBOSE
44570 https://sourceware.org/pipermail/gdb-patches/2011-June/083081.html
44572 Back then, the testcase would pass cleanly. It turns out that the
44573 reason it fails today is that the testcase is exercising something in
44574 GDB that only makes sense if the program is built for the pre-C++11
44575 libstc++ ABI. Back then the C++11 ABI didn't exist yet, but nowadays,
44576 you need to compile with -D_GLIBCXX_USE_CXX11_ABI=0 to use the old
44577 ABI. See "Dual ABI" in the libstdc++ manual, at
44578 <https://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html>.
44580 If we tweak the gdb.cp/no-dmgl-verbose.exp testcase to force the old
44581 ABI with -D_GLIBCXX_USE_CXX11_ABI=0, then it passes cleanly.
44583 So why is it that setting a breakpoint at "f(std::string)" fails with
44584 modern ABI, but passes with old ABI?
44586 This is where libiberty demangler's DMGL_VERBOSE option comes in. The
44587 Itanium ABI mangling scheme has a shorthand form for std::string (and
44588 some other types). See
44589 <https://itanium-cxx-abi.github.io/cxx-abi/abi.html>:
44591 "In addition, the following catalog of abbreviations of the form "Sx" are used:
44593 <substitution> ::= St # ::std::
44594 <substitution> ::= Sa # ::std::allocator
44595 <substitution> ::= Sb # ::std::basic_string
44596 <substitution> ::= Ss # ::std::basic_string < char,
44597 ::std::char_traits<char>,
44598 ::std::allocator<char> >
44599 <substitution> ::= Si # ::std::basic_istream<char, std::char_traits<char> >
44600 <substitution> ::= So # ::std::basic_ostream<char, std::char_traits<char> >
44601 <substitution> ::= Sd # ::std::basic_iostream<char, std::char_traits<char> >
44604 When the libiberty demangler encounters such a abbreviation, by
44605 default, it expands it to the user-friendly typedef "std::string",
44606 "std::iostream", etc.. If OTOH DMGL_VERBOSE is specified, the
44607 abbreviation is expanded to the underlying, non-typedefed fullname
44608 "std::basic_string<char, std::char_traits<char>, std::allocator<char> >"
44609 etc. as documented in the Itanium ABI, and pasted above. You can see
44610 the standard abbreviations/substitutions in
44611 libiberty/cp-demangle.c:standard_subs.
44613 Back before Jan's patch in 2011, there were parts of GDB that used
44614 DMGL_VERBOSE, and others that did not, leading to mismatches. The
44615 solution back then was to stop using DMGL_VERBOSE throughout.
44617 GDB has code in place to let users set a breakpoint at a function with
44618 typedefs in its parameters, like "b f(uint32_t)". Demangled function
44619 names as they appear in the symbol tables almost (more on this is in a
44620 bit) never have typedefs in them, so when processing "b f(uint32_t)"
44621 GDB first replaces "uint32_t" for its underlying type, and then sets a
44622 breakpoint on the resulting prototype, in this case "f(unsigned int)".
44624 Now, if DMGL_VERBOSE is _not_ used, then the demangler demangles the
44625 mangled name of a function such as "void f(std::string)" that was
44626 mangled using the standard abbreviations mentioned above really as:
44628 "void f(std::string)".
44630 For example, the mangled name of "void f(std::string)" if you compile
44631 with old pre-C++11 ABI is "_Z1fSs". That uses the abbreviation "Ss",
44632 so if you demangle that without DMGL_VERBOSE, you get:
44634 $ echo "_Z1fSs" | c++filt --no-verbose
44637 while with DMGL_VERBOSE you'd get:
44639 $ echo "_Z1fSs" | c++filt
44640 f(std::basic_string<char, std::char_traits<char>, std::allocator<char> >)
44642 If, when the user sets a breakpoint at "f(std::string)", GDB would
44643 replace the std::string typedef for its underlying type using the same
44644 mechanism I mentioned for the "f(uint32_t)" example above, then GDB
44645 would really try to set a breakpoint at "f(std::basic_string<char,
44646 std::char_traits<char>, std::allocator<char> >)", and that would fail,
44647 as the function symbol GDB knows about for that function, given no
44648 DMGL_VERBOSE, is "f(std::string)".
44650 For this reason, the code that expands typedefs in function parameter
44651 names has an exception for std::string (and other standard
44652 abbreviation types), such that "std::string" is never
44655 And here lies the problem when you try to do "b f(std::string)" with a
44656 program compiled with the C++11 ABI. In that case, std::string
44657 expands to a different underlying type, like so:
44659 f(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >)
44661 and this symbol naturally mangles differently, as:
44663 _Z1fNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE
44665 and then because this doesn't use the shorthand mangling abbreviation
44666 for "std::string" anymore, it always demangles as:
44668 f(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >)
44670 Now, when using the C++11 ABI, and you set a breakpoint at
44671 "f(std::string)", GDB's typedefs-in-parameters expansion code hits the
44672 exception for "std::string" and doesn't expand it, so the breakpoint
44673 fails to be inserted, because the symbol that exists is really the
44675 f(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >)
44677 one, not "f(std::string)".
44679 So to fix things for C++11 ABI, clearly we need to remove the
44680 "std::string" exception from the typedef-in-parameters expansion
44681 logic. If we do just that, then "b f(std::string)" starts working
44682 with the C++11 ABI.
44684 However, if we do _just_ that, and nothing else, then we break
44687 The solution is then to in addition switch GDB to always use
44688 DMGL_VERBOSE. If we do this, then pre-C++11 ABI symbols works the
44689 same as C++11 ABI symbols overall -- the demangler expands the
44690 standard abbreviation for "std::string" as "std::basic_string<char,
44691 std::char_traits<char>, std::allocator<char> >" and letting GDB expand
44692 the "std::string" typedef (etc.) too is no longer a problem.
44694 To avoid getting in the situation where some parts of GDB use
44695 DMGL_VERBOSE and others not, this patch adds wrappers around the
44696 demangler's entry points that GDB uses, and makes those force
44699 The point of the gdb.cp/no-dmgl-verbose.exp testcase was to try to
44700 ensure that DMGL_VERBOSE doesn't creep back in:
44702 gdb_test {break 'f(std::basic_string<char, std::char_traits<char>, std::allocator<char> >)'} \
44703 {Function ".*" not defined\.} \
44704 "DMGL_VERBOSE-demangled f(std::string) is not defined"
44706 This obviously no longer makes sense to have, since we now depend on
44707 DMGL_VERBOSE. So the patch replaces gdb.cp/no-dmgl-verbose.exp with a
44708 new gdb.cp/break-f-std-string.exp testcase whose purpose is to make
44709 sure that setting a breakpoint at "f(std::string)" works. It
44710 exercises both pre-C++11 ABI and C++11 ABI.
44712 Change-Id: Ib54fab4cf0fd307bfd55bf1dd5056830096a653b
44714 2022-05-10 Nils-Christian Kempke <nils-christian.kempke@intel.com>
44716 gdb/testsuite: fix testsuite regressions for unix/-m32 board
44717 This commit fixes two regressions introduced by
44718 891e4190ba705373eec7b374209478215fff5401.
44720 Reason for the failures was, that on a 32 bit machine the maximum
44721 array length as well as the maximum allocatable memory for arrays
44722 (in bytes) both seem to be limited by the maximum value of a 4
44723 byte (signed) Fortran integer. This lead to compiler errors/unexpected
44724 behavior when compiling/running the test with the -m32 board. This
44725 behavior is compiler dependent and can differ for different compiler
44726 implementations, but generally, it seemed like a good idea to simply
44727 avoid such situations.
44729 The affected tests check for GDB's overflow behavior when using KIND
44730 parameters with GDB implemented Fortran intrinsic functions. If these
44731 KIND parameters are too small to fit the actual intrinsic function's
44732 result, an overflow is expected. This was done for 1, 2, and 4
44733 byte overflows. The last one caused problems, as it tried to allocate
44734 arrays of length/byte-size bigger than the 4 byte signed integers which
44735 would then be used with the LBOUND/UBOUND/SIZE intrinsics.
44737 The tests were adapted to only execute the 4 byte overflow tests when
44738 running on targets with 64 bit. For this, the compiled tests evaluate the
44739 byte size of a C_NULL_PTR via C_SIZEOF, both defined in the ISO_C_BINDING
44740 module. The ISO_C_BINDING constant C_NULL_PTR is a Fortran 2003, the
44741 C_SIZEOF a Fortran 2008 extension. Both have been implemented in their
44742 respective compilers for while (e.g. C_SIZEOF is available since
44743 gfortran 4.6). If this byte size evaluates to less than 8 we skip the
44744 4 byte overflow tests in the compiled tests of size.f90 and
44745 lbound-ubound.f90. Similarly, in the lbound-ubound.exp testsfile we skip
44746 the 4 byte overflow tests if the procedure is_64_target evaluates to false.
44748 In size.f90, additionally, the to-be-allocated amount of bytes did not
44749 fit into 4 byte signed integers for some of the arrays, as it was
44750 approximately 4 times the maximum size of a 4 byte signed integer. We
44751 adapted the dimensions of the arrays in question as the meaningfulness
44752 of the test does not suffer from this.
44754 With this patch both test run fine with the unix/-m32 board and
44755 gcc/gfortran (9.4) as well as the standard board file.
44757 We also thought about completely removing the affected test from the
44758 testsuite. We decided against this as the 32 bit identification comes
44759 with Fortran 2008 and removing tests would have decreased coverage.
44761 A last change that happened with this patch was due to gfortran's and
44762 ifx's type resolution when assigning big constants to Fortran Integer*8
44763 variables. Before the above changes this happened in a parameter
44764 statement. Here, both compilers happily accepted a line like
44766 integer*8, parameter :: var = 2147483647 + 5.
44768 After this change the assignment is not done as a parameter
44769 anymore, as this triggered compile time overflow errors. Instead,
44770 the assignment is done dynamically, depending on the kind of machine one
44771 is on. Sadly, just changing this line to
44774 var = 2147483647 + 5
44776 does not work with ifx (or flang for that matter, they behave similarly
44777 here). It will create an integer overflow in the addition as ifx deduces
44778 the type the additon is done in as Integer*4. So var will actually
44779 contain the value -2147483644 after this. The lines
44784 on the other hand fail to compile with gfortran (9.4.0) as the compiler
44785 identifies an Integer overflow here. Finally, to make this work with
44786 all three compilers an additional parameter has been introduced
44788 integer*8, parameter :: helper = 2147483647
44792 This works on all 3 compilers as expected.
44794 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29053
44795 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29054
44797 2022-05-10 Pedro Alves <pedro@palves.net>
44799 Move non-dependent gdb::observers::observable::visit_state outside template
44800 The other day, while looking at the symbols that end up in a GDB
44801 index, I noticed that the gdb::observers::observable::visit_state enum
44802 class appears a number of times:
44804 $ grep VISIT gdb-index-symbol-names.txt
44805 gdb::observers::observable<bpstat*, int>::visit_state::NOT_VISITED
44806 gdb::observers::observable<bpstat*, int>::visit_state::VISITED
44807 gdb::observers::observable<bpstat*, int>::visit_state::VISITING
44808 gdb::observers::observable<breakpoint*>::visit_state::NOT_VISITED
44809 gdb::observers::observable<breakpoint*>::visit_state::VISITED
44810 gdb::observers::observable<breakpoint*>::visit_state::VISITING
44811 gdb::observers::observable<char const*, char const*>::visit_state::NOT_VISITED
44812 gdb::observers::observable<char const*, char const*>::visit_state::VISITED
44813 gdb::observers::observable<char const*, char const*>::visit_state::VISITING
44814 gdb::observers::observable<char const*>::visit_state::NOT_VISITED
44815 gdb::observers::observable<char const*>::visit_state::VISITED
44816 gdb::observers::observable<char const*>::visit_state::VISITING
44817 gdb::observers::observable<enum_flags<user_selected_what_flag> >::visit_state::NOT_VISITED
44818 gdb::observers::observable<enum_flags<user_selected_what_flag> >::visit_state::VISITED
44819 gdb::observers::observable<enum_flags<user_selected_what_flag> >::visit_state::VISITING
44820 gdb::observers::observable<frame_info*, int>::visit_state::NOT_VISITED
44821 gdb::observers::observable<frame_info*, int>::visit_state::VISITED
44822 gdb::observers::observable<frame_info*, int>::visit_state::VISITING
44823 gdb::observers::observable<gdbarch*>::visit_state::NOT_VISITED
44824 gdb::observers::observable<gdbarch*>::visit_state::VISITED
44825 gdb::observers::observable<gdbarch*>::visit_state::VISITING
44826 gdb::observers::observable<gdb_signal>::visit_state::NOT_VISITED
44827 gdb::observers::observable<gdb_signal>::visit_state::VISITED
44828 gdb::observers::observable<gdb_signal>::visit_state::VISITING
44831 $ grep VISIT gdb-index-symbol-names.txt | wc -l
44834 enum class visit_state is defined inside the class template
44835 observable, but it doesn't have to be, as it does not depend on the
44836 template parameters. This commit moves it out, so that only one such
44837 type exists. This reduces the size of a -O0 -g3 build for me by
44838 around 0.6%, like so:
44840 $ du -b gdb.before gdb.after
44841 164685280 gdb.before
44842 163707424 gdb.fixed
44844 and codesize by some 0.5%.
44846 Change-Id: I405f4ef27b8358fdd22158245b145d849b45658e
44848 2022-05-10 Nick Clifton <nickc@redhat.com>
44850 Fix compiling binutils/resbin.c with Clang version 14
44852 2022-05-10 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
44854 gprofng: include percentages in default metrics list
44856 2022-05-09 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
44858 * src/gprofng.rc: Include percentages in default metrics list.
44860 2022-05-10 Alan Modra <amodra@gmail.com>
44862 gprof: remove use of PTR
44863 * basic_blocks.c: Replace uses of PTR with void * throughout.
44864 * cg_arcs.c: Likewise.
44865 * cg_print.c: Likewise.
44866 * hist.c: Likewise.
44867 * source.h: Likewise.
44868 * symtab.c: Likewise.
44870 gas: remove use of PTR
44871 * config/obj-evax.c (evax_symbol_new_hook): Don't cast to PTR.
44873 2022-05-10 Alan Modra <amodra@gmail.com>
44875 opcodes: remove use of PTR
44876 The non-cgen parts of opcodes.
44878 * cr16-dis.c (print_arg): Replace PTR with void *.
44879 * crx-dis.c (print_arg): Likewise.
44880 * rl78-dis.c (print_insn_rl78_common): Don't use PTR cast.
44881 * rx-dis.c (print_insn_rx): Likewise.
44882 * visium-dis.c (print_insn_visium): Likewise.
44883 * z8k-dis.c (print_insn_z8k): Likewise.
44885 2022-05-10 Alan Modra <amodra@gmail.com>
44887 bfd: remove use of PTR
44888 * coffcode.h (coff_write_object_contents): Don't cast to PTR.
44889 * elf32-csky.c (csky_elf_link_hash_traverse): Remove use of PTR
44891 (csky_allocate_dynrelocs): Don't use PTR cast.
44892 * elf32-nios2.c (adjust_dynrelocs, allocate_dynrelocs): Replace
44894 * elf32-visium.c (visium_elf_howto_parity_reloc): Likewise.
44895 * elfxx-ia64.c (ia64_elf_reloc): Likewise.
44896 * plugin.c (bfd_plugin_bfd_print_private_bfd_data): Likewise.
44898 include: remove use of PTR
44899 * hashtab.h (HTAB_EMPTY_ENTRY): Replace PTR with void *.
44900 (HTAB_DELETED_ENTRY): Likewise.
44902 2022-05-10 GDB Administrator <gdbadmin@sourceware.org>
44904 Automatic date update in version.in
44906 2022-05-09 Ilya Leoshkevich <iii@linux.ibm.com>
44908 IBM zSystems: Accept (. - 0x100000000) PCRel32 operands
44909 as does not accept instructions like brasl %r0,.-0x100000000, because
44910 of two problems with the generic overflow check:
44912 1. PCRel32 operands are signed, but are treated as unsigned.
44914 2. The allowed range for these operands is [-(1 << 32), (1 << 32) - 1],
44915 and not [-(1 << 31), (1 << 31) - 1].
44917 Fix both by disabling the generic overflow check - it's not needed,
44918 because s390_insert_operand () performs its own.
44922 * config/tc-s390.c (md_gather_operands): Set fx_no_overflow.
44923 * testsuite/gas/s390/s390.exp: Add zarch-z900-err.
44924 * testsuite/gas/s390/esa-z900.d: New test.
44925 * testsuite/gas/s390/esa-z900.s: New test.
44926 * testsuite/gas/s390/zarch-z900-err.l: New test.
44927 * testsuite/gas/s390/zarch-z900-err.s: New test.
44929 2022-05-09 Andrew Burgess <aburgess@redhat.com>
44931 gdb/testsuite: fix occasional failure in gdb.mi/mi-multi-commands.exp
44932 In bug PR gdb/29036, another failure was reported for the test
44933 gdb.mi/mi-multi-commands.exp. This test sends two commands to GDB as
44934 a single write, and then checks that both commands are executed.
44936 The problem that was encountered here is that the output of the first
44937 command, which looks like this:
44939 ^done,value="\"FIRST COMMAND\""
44941 Is actually produced in parts, first the '^done' is printed, then the
44942 ',value="\"FIRST COMMAND\"" is printed.
44944 What was happening is that some characters from the second command
44945 were being echoed after the '^done' had been printed, but before the
44946 value part had been printed. To avoid this issue I've relaxed the
44947 pattern that checks for the first command a little. With this fix in
44948 place the occasional failure in this test is no longer showing up.
44950 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29036
44952 2022-05-09 Tom de Vries <tdevries@suse.de>
44954 [gdb] Update syscalls/{amd64,i386}-linux.xml
44955 - Add a script syscalls/gen-header.py, based on syscalls/arm-linux.py.
44956 - Add a script syscalls/update-linux.sh (alongside update-freebsd.sh and
44958 - Use syscalls/update-linux.sh to update syscalls/{amd64,i386}-linux.xml.in.
44959 - Regenerate syscalls/{amd64,i386}-linux.xml using syscalls/Makefile.
44961 In gdb/syscalls/i386-linux.xml.in, updating has the following notable effect:
44963 - <syscall name="madvise1" number="220"/>
44964 - <syscall name="getdents64" number="221"/>
44965 - <syscall name="fcntl64" number="222"/>
44966 + <syscall name="getdents64" number="220"/>
44967 + <syscall name="fcntl64" number="221"/>
44970 I've verified in ./arch/x86/entry/syscalls/syscall_32.tbl that the numbers are
44973 Tested on x86_64-linux.
44975 2022-05-09 Tom de Vries <tdevries@suse.de>
44977 [gdb] Add gdb/syscalls/Makefile
44978 Add a Makefile in gdb/syscalls that can be used to translate
44979 gdb/syscalls/*.xml.in into gdb/syscalls/*.xml.
44981 Calling make reveals that bfin-linux.xml is missing, so add it.
44983 Tested on x86_64-linux.
44985 2022-05-09 Tiezhu Yang <yangtiezhu@loongson.cn>
44987 gdb: LoongArch: Implement the return_value gdbarch method
44988 When execute the following command on LoongArch:
44990 make check-gdb TESTS="gdb.base/async.exp"
44992 there exist the following failed testcases:
44994 FAIL: gdb.base/async.exp: finish& (timeout)
44995 FAIL: gdb.base/async.exp: jump& (timeout)
44996 FAIL: gdb.base/async.exp: until& (timeout)
44997 FAIL: gdb.base/async.exp: set exec-done-display off (GDB internal error)
44999 we can see the following messages in gdb/testsuite/gdb.log:
45002 Run till exit from #0 foo () at /home/loongson/gdb.git/gdb/testsuite/gdb.base/async.c:9
45003 (gdb) /home/loongson/gdb.git/gdb/gdbarch.c:2646: internal-error: gdbarch_return_value: Assertion `gdbarch->return_value != NULL' failed.
45004 A problem internal to GDB has been detected,
45005 further debugging may prove unreliable.
45007 In order to fix the above failed testcases, implement the return_value
45008 gdbarch method on LoongArch.
45010 2022-05-09 Andrew Burgess <aburgess@redhat.com>
45012 gdb: fix for gdb.base/eof-exit.exp test failures
45013 This fix relates to PR gdb/29032, this makes the test more stable by
45014 ensuring that the Ctrl-D is only sent once the prompt has been
45015 displayed. This issue was also discussed on the mailing list here:
45017 https://sourceware.org/pipermail/gdb-patches/2022-April/187670.html
45019 The problem identified in the bug report is that sometimes the Ctrl-D
45020 (that the test sends to GDB) arrives while GDB is processing a
45021 command. When this happens the Ctrl-D is handled differently than if
45022 the Ctrl-D is sent while GDB is waiting for input at a prompt.
45024 The original intent of the test was that the Ctrl-D be sent while GDB
45025 was waiting at a prompt, and that is the case the occurs most often,
45026 but, when the Ctrl-D arrives during command processing, then GDB will
45027 ignore the Ctrl-D, and the test will fail.
45029 This commit ensures the Ctrl-D is always sent while GDB is waiting at
45030 a prompt, which makes this test stable.
45032 But, that still leaves an open question, what should happen when the
45033 Ctrl-D arrives while GDB is processing a command? This commit doesn't
45034 attempt to answer that question, which is while bug PR gdb/29032 will
45035 not be closed once this commit is merged.
45037 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29032
45039 2022-05-09 Tom de Vries <tdevries@suse.de>
45041 [gdb] Make btrace maintainer entry more clear
45044 -record btrace <name> <email>
45046 + btrace <name> <email>
45048 to clarify that the listed maintainer is only maintainer of the btrace part of
45051 2022-05-09 Martin Liska <mliska@suse.cz>
45053 ansidecl.h: sync from GCC
45056 * ansidecl.h: Sync from GCC.
45058 2022-05-09 Tom de Vries <tdevries@suse.de>
45060 [gdb/tdep] Support catch syscall pipe2 for i386
45061 With test-case gdb.base/catch-syscall.exp and target board unix/-m32, we run
45064 (gdb) catch syscall pipe2^M
45065 Unknown syscall name 'pipe2'.^M
45066 (gdb) FAIL: gdb.base/catch-syscall.exp: determine pipe syscall: catch syscall pipe2
45070 - adding a pipe2 entry in gdb/syscalls/i386-linux.xml.in, and
45071 - regenerating gdb/syscalls/i386-linux.xml using
45072 "xsltproc --output i386-linux.xml apply-defaults.xsl i386-linux.xml.in".
45074 Tested on x86_64-linux with native and unix/-m32.
45076 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29056
45078 2022-05-09 Tom de Vries <tdevries@suse.de>
45080 [gdb/testsuite] Handle pipe2 syscall in gdb.base/catch-syscall.exp
45081 When running test-case gdb.reverse/pipe-reverse.exp on openSUSE Tumbleweed,
45087 Catchpoint 2 (returned from syscall pipe2), in pipe () from /lib64/libc.so.6^M
45088 (gdb) FAIL: gdb.base/catch-syscall.exp: without arguments: \
45089 syscall pipe has returned
45092 The current glibc on Tumbleweed is 2.35, which contains commit
45093 "linux: Implement pipe in terms of __NR_pipe2", and consequently syscall pipe2
45094 is used instead of syscall pipe.
45096 Fix this by detecting whether syscall pipe or pipe2 is used before running the
45099 Tested on x86_64-linux, specifically on:
45100 - openSUSE Tumbleweed (with glibc 2.35), and
45101 - openSUSE Leap 15.3 (with glibc 2.31).
45103 On openSUSE Tumbleweed + target board unix/-m32, this exposes:
45105 (gdb) catch syscall pipe2^M
45106 Unknown syscall name 'pipe2'.^M
45108 which will be fixed in a folllow-up patch.
45110 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29056
45112 2022-05-09 Tom de Vries <tdevries@suse.de>
45114 [gdb/tdep] Handle pipe2 syscall for amd64
45115 When running test-case gdb.reverse/pipe-reverse.exp on openSUSE Tumbleweed,
45120 Process record and replay target doesn't support syscall number 293^M
45121 Process record: failed to record execution log.^M
45124 0x00007ffff7daabdb in pipe () from /lib64/libc.so.6^M
45125 (gdb) FAIL: gdb.reverse/pipe-reverse.exp: continue to breakpoint: marker2
45128 The current glibc on Tumbleweed is 2.35, which contains commit
45129 "linux: Implement pipe in terms of __NR_pipe2", and consequently syscall pipe2
45130 is used in stead of syscall pipe.
45132 There is already support added for syscall pipe2 for aarch64 (which only has
45133 syscall pipe2, not syscall pipe), so enable the same for amd64, by:
45134 - adding amd64_sys_pipe2 in enum amd64_syscall
45135 - translating amd64_sys_pipe2 to gdb_sys_pipe2 in amd64_canonicalize_syscall
45137 Tested on x86_64-linux, specifically on:
45138 - openSUSE Tumbleweed (with glibc 2.35), and
45139 - openSUSE Leap 15.3 (with glibc 2.31).
45141 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29056
45143 2022-05-09 GDB Administrator <gdbadmin@sourceware.org>
45145 Automatic date update in version.in
45147 2022-05-08 Tom de Vries <tdevries@suse.de>
45149 [gdb/testsuite] Fix gdb.tui/scroll.exp with read1
45150 When running test-case gdb.tui/scroll.exp, I get:
45152 Box Dump (80 x 8) @ (0, 0):
45161 PASS: gdb.tui/scroll.exp: check cmd window in flip layout
45163 but with check-read1 I get instead:
45165 Box Dump (80 x 8) @ (0, 0):
45174 FAIL: gdb.tui/scroll.exp: check cmd window in flip layout
45177 The "p 19" command is handled by Term::command, which sends the command and then
45178 does Term::wait_for "^$gdb_prompt [string_to_regexp $cmd]", which:
45179 - matches the line with "(gdb) p 19", and
45180 - tries to match the following prompt "(gdb) "
45182 The problem is that scrolling results in reissuing output before the "(gdb) p
45183 19", and the second matching triggers on that. Consequently, wait_for no
45184 longer translates gdb output into screen actions, and the screen does not
45185 reflect the result of "p 19".
45187 Fix this by using a new proc wait_for_region_contents, which in contrast to
45188 wait_for can handle a multi-line regexp.
45190 Tested on x86_64-linux with make targets check and check-read1.
45192 2022-05-08 Tom de Vries <tdevries@suse.de>
45194 [gdb/testsuite] Fix gdb.cp/casts.exp with -m32
45195 When running test-case gdb.cp/casts.exp with target board unix/-m32, I run
45198 (gdb) print (unsigned long long) &gd == gd_value^M
45200 (gdb) FAIL: gdb.cp/casts.exp: print (unsigned long long) &gd == gd_value
45203 With some additional printing, we can see in more detail why the comparison
45206 (gdb) print /x &gd^M
45208 (gdb) PASS: gdb.cp/casts.exp: print /x &gd
45209 print /x (unsigned long long)&gd^M
45211 (gdb) PASS: gdb.cp/casts.exp: print /x (unsigned long long)&gd
45212 print /x gd_value^M
45213 $33 = 0xffffffffffffc5c8^M
45214 (gdb) PASS: gdb.cp/casts.exp: print /x gd_value
45215 print (unsigned long long) &gd == gd_value^M
45217 (gdb) FAIL: gdb.cp/casts.exp: print (unsigned long long) &gd == gd_value
45220 The gd_value is set by this assignment:
45222 unsigned long long gd_value = (unsigned long long) &gd;
45225 The problem here is directly casting from a pointer to a non-pointer-sized
45228 Fix this by adding an intermediate cast to std::uintptr_t.
45230 Tested on x86_64-linux with native and target board unix/-m32.
45232 2022-05-08 Tom de Vries <tdevries@suse.de>
45234 [gdb/testsuite] Handle init errors in gdb.mi/user-selected-context-sync.exp
45235 In OBS, on aarch64-linux, with a gdb 11.1 based package, I run into:
45237 (gdb) builtin_spawn -pty^M
45238 new-ui mi /dev/pts/5^M
45240 (gdb) =thread-group-added,id="i1"^M
45241 (gdb) ERROR: MI channel failed
45242 warning: Error detected on fd 11^M
45244 Unknown thread 1.1.^M
45245 (gdb) UNRESOLVED: gdb.mi/user-selected-context-sync.exp: mode=non-stop: \
45246 test_cli_inferior: reset selection to thread 1.1
45248 with many more UNRESOLVED following.
45250 The ERROR is a common problem, filed as
45251 https://sourceware.org/bugzilla/show_bug.cgi?id=28561 .
45253 But the many UNRESOLVEDs are due to not checking whether the setup as done in
45254 the test_setup function succeeds or not.
45257 - making test_setup return an error upon failure
45258 - handling test_setup error at the call site
45259 - adding a "setup done" pass/fail to be turned into an unresolved
45260 in case of error during setup.
45262 Tested on x86_64-linux, by manually triggering the error in
45263 mi_gdb_start_separate_mi_tty.
45265 2022-05-08 Tom de Vries <tdevries@suse.de>
45267 [gdb/testsuite] Fix gdb.ada/catch_ex_std.exp with remote-gdbserver-on-localhost
45268 When running test-case gdb.ada/catch_ex_std.exp on target board
45269 remote-gdbserver-on-localhost, I run into:
45273 [Inferior 1 (process 15656) exited with code 0177]^M
45274 (gdb) FAIL: gdb.ada/catch_ex_std.exp: runto: run to main
45275 Remote debugging from host ::1, port 49780^M
45276 /home/vries/foo: error while loading shared libraries: libsome_package.so: \
45277 cannot open shared object file: No such file or directory^M
45280 Fix this by adding the usual shared-library + remote-target helper
45281 "gdb_load_shlib $sofile".
45283 Tested on x86_64-linux with native and target board
45284 remote-gdbserver-on-localhost.
45286 2022-05-08 Tom de Vries <tdevries@suse.de>
45288 [gdb/testsuite] Fix gdb.threads/fork-plus-threads.exp with check-readmore
45289 When running test-case gdb.threads/fork-plus-threads.exp with check-readmore,
45292 [Inferior 11 (process 7029) exited normally]^M
45293 [Inferior 1 (process 6956) exited normally]^M
45294 FAIL: gdb.threads/fork-plus-threads.exp: detach-on-fork=off: \
45295 inferior 1 exited (timeout)
45298 The problem is that the regexp consuming the "Inferior exited normally"
45300 - consumes more than one of those messages at a time, but
45301 - counts only one of those messages.
45303 Fix this by adopting a line-by-line approach, which deals with those messages
45306 Tested on x86_64-linux with native, check-read1 and check-readmore.
45308 2022-05-08 GDB Administrator <gdbadmin@sourceware.org>
45310 Automatic date update in version.in
45312 2022-05-07 Tom Tromey <tom@tromey.com>
45314 Fix "catch syscall"
45315 Simon pointed out that some recent patches of mine broke "catch
45316 syscall". Apparently I forgot to finish the conversion of this code
45317 when removing init_catchpoint. This patch completes the conversion
45320 2022-05-07 Andrew Burgess <aburgess@redhat.com>
45322 gdb/readline: fix extra 'quit' message problem
45323 After these two commits:
45325 commit 4fb7bc4b147fd30b781ea2dad533956d0362295a
45326 Date: Mon Mar 7 13:49:21 2022 +0000
45328 readline: back-port changes needed to properly detect EOF
45330 commit 91395d97d905c31ac38513e4aaedecb3b25e818f
45331 Date: Tue Feb 15 17:28:03 2022 +0000
45333 gdb: handle bracketed-paste-mode and EOF correctly
45335 It was observed that, if a previous command is selected at the
45336 readline prompt using the up arrow key, then when the command is
45337 accepted (by pressing return) an unexpected 'quit' message will be
45338 printed by GDB. Here's an example session:
45347 In this session the second 'p 123' was entered not by typing 'p 123',
45348 but by pressing the up arrow key to select the previous command. It
45349 is important that the up arrow key is used, typing Ctrl-p will not
45352 The problem here appears to be readline's EOF detection when handling
45353 multi-character input sequences. I have raised this issue on the
45354 readline mailing list here:
45356 https://lists.gnu.org/archive/html/bug-readline/2022-04/msg00012.html
45358 a solution has been proposed here:
45360 https://lists.gnu.org/archive/html/bug-readline/2022-04/msg00016.html
45362 This patch includes a test for this issue as well as a back-port of
45363 (the important bits of) readline commit:
45365 commit 2ef9cec8c48ab1ae3a16b1874a49bd1f58eaaca1
45366 Date: Wed May 4 11:18:04 2022 -0400
45368 fix for setting RL_STATE_EOF in callback mode
45370 That commit also includes some updates to the readline documentation
45371 and tests that I have not included in this commit.
45373 With this commit in place the unexpected 'quit' messages are resolved.
45375 2022-05-07 Alan Modra <amodra@gmail.com>
45377 Fix multiple ubsan warnings in i386-dis.c
45378 Commit 39fb369834a3 "opcodes: Make i386-dis.c thread-safe" introduced
45379 a number of casts to bfd_signed_vma that cause undefined behaviour
45380 with a 32-bit libbfd. Revert those changes.
45382 * i386-dis.c (OP_E_memory): Do not cast disp to bfd_signed_vma
45384 (get32, get32s): Don't use bfd_signed_vma here.
45386 2022-05-07 Alan Modra <amodra@gmail.com>
45388 Re: Fix new linker testsuite failures due to rwx segment test problems
45392 * elfnn-loongarch.c: Remove commented out elf_backend_* defines.
45394 * testsuite/ld-elf/elf.exp (target_defaults_to_execstack): Match
45395 arm*. Delete loongarch.
45397 2022-05-07 GDB Administrator <gdbadmin@sourceware.org>
45399 Automatic date update in version.in
45401 2022-05-07 Carl Love <cel@us.ibm.com>
45403 PowerPC fix for gdb.server/sysroot.exp
45404 On PowerPC, the stop in the printf function is of the form:
45406 Breakpoint 2, 0x00007ffff7c6ab08 in printf@@GLIBC_2.17 () from /lib64/libc.so.6
45408 On other architectures the output looks like:
45410 Breakpoint 2, 0x0000007fb7ea29ac in printf () from /lib/aarch64-linux-gnu/libc.so.6
45412 The following patch modifies the printf test by matchine any character
45413 starting immediately after the printf. The test now works for PowerPC
45414 output as well as the output from other architectures.
45416 The test has been run on a Power 10 system and and Intel x86_64 system.
45418 2022-05-06 Nick Clifton <nickc@redhat.com>
45420 Fix new linker testsuite failures due to rwx segment test problems
45422 2022-05-06 Tom Tromey <tom@tromey.com>
45424 Introduce catchpoint class
45425 This introduces a catchpoint class that is used as the base class for
45426 all catchpoints. init_catchpoint is rewritten to be a constructor
45429 This changes the hierarchy a little -- some catchpoints now inherit
45430 from base_breakpoint whereas previously they did not. This isn't a
45431 problem, as long as re_set is redefined in catchpoint.
45433 2022-05-06 Tom Tromey <tom@tromey.com>
45435 Add initializers to tracepoint
45436 This adds some initializers to tracepoint. I think right now these
45437 may not be needed, due to obscure rules about zero initialization.
45438 However, this will change in the next patch, and anyway it is clearer
45441 Remove init_raw_breakpoint_without_location
45442 This removes init_raw_breakpoint_without_location, replacing it with a
45443 constructor on 'breakpoint' itself. The subclasses and callers are
45446 Disable copying for breakpoint
45447 It seems to me that breakpoint should use DISABLE_COPY_AND_ASSIGN.
45448 This patch does this.
45450 Add constructor to exception_catchpoint
45451 This adds a constructor to exception_catchpoint and simplifies the
45454 Add constructor to syscall_catchpoint
45455 This adds a constructor to syscall_catchpoint and simplifies the
45458 Add constructor to signal_catchpoint
45459 This adds a constructor to signal_catchpoint and simplifies the
45462 Add constructor to solib_catchpoint
45463 This adds a constructor to solib_catchpoint and simplifies the caller.
45465 Add constructor to fork_catchpoint
45466 This adds a constructor to fork_catchpoint and simplifies the caller.
45468 Remove unnecessary line from catch_exec_command_1
45469 catch_exec_command_1 clears the new catchpoint's "exec_pathname"
45470 field, but this is already done by virtue of calling "new".
45472 Constify breakpoint::print_recreate
45473 This constifies breakpoint::print_recreate.
45475 Constify breakpoint::print_mention
45476 This constifies breakpoint::print_mention.
45478 Constify breakpoint::print_one
45479 This constifies breakpoint::print_one.
45481 Constify breakpoint::print_it
45482 This constifies breakpoint::print_it. Doing this pointed out some
45483 code in ada-lang.c that can be simplified a little as well.
45485 Move works_in_software_mode to watchpoint
45486 works_in_software_mode is only useful for watchpoints. This patch
45487 moves it from breakpoint to watchpoint, and changes it to return bool.
45489 Boolify breakpoint::explains_signal
45490 This changes breakpoint::explains_signal to return bool.
45492 Remove breakpoint::ops
45493 The breakpoint::ops field is set but never used. This removes it.
45495 Change print_recreate_thread to a method
45496 This changes print_recreate_thread to be a method on breakpoint. This
45497 function is only used as a helper by print_recreate methods, so I
45498 thought this transformation made sense.
45500 2022-05-06 Carl Love <cel@us.ibm.com>
45502 PowerPC: bp-permanent.exp, kill-after-signal fix
45503 The break point after the stepi on Intel is the entry point of the user
45504 signal handler function test_signal_handler. The code at the break point
45507 0x<hex address> <test_signal_handler>: endbr64
45509 On PowerPC with a Linux 5.9 kernel or latter, the address where gdb stops
45510 after the stepi is in the vdso code inserted by the kernel. The code at the
45511 breakpoint looks like:
45513 0x<hex address> <__kernel_start_sigtramp_rt64>: bctrl
45515 This is different from other architectures. As discussed below, recent
45516 kernel changes involving the vdso for PowerPC have been made changes to the
45517 signal handler code flow. PowerPC is now stopping in function
45518 __kernel_start_sigtramp_rt64. PowerPC now requires an additional stepi to
45519 reach the user signal handler unlike other architectures.
45521 The bp-permanent.exp and kill-after-signal tests run fine on PowerPC with an
45522 kernel that is older than Linux 5.9.
45524 The PowerPC 64 signal handler was updated by the Linux kernel 5.9-rc1:
45526 commit id: 0138ba5783ae0dcc799ad401a1e8ac8333790df9
45527 powerpc/64/signal: Balance return predictor stack in signal trampoline
45529 An additional change to the PowerPC 64 signal handler was made in Linux
45530 kernel version 5.11-rc7 :
45532 commit id: 24321ac668e452a4942598533d267805f291fdc9
45533 powerpc/64/signal: Fix regression in __kernel_sigtramp_rt64() semantics
45535 The first kernel change, puts code into the user space signal handler (in
45536 the vdso) as a performance optimization to prevent the call/return stack
45537 from getting out of balance. The patch ensures that the entire
45538 user/kernel/vdso cycle is balanced with the addition of the "brctl"
45541 The second patch, fixes the semantics of __kernel_sigtramp_rt64(). A new
45542 symbol is introduced to serve as the jump target from the kernel to the
45543 trampoline which now consists of two parts.
45545 The above changes for PowerPC signal handler, causes gdb to stop in the
45546 kernel code not the user signal handler as expected. The kernel dispatches
45547 to the vdso code which in turn calls into the signal handler. PowerPC is
45548 special in that the kernel is using a vdso instruction (bctrl) to enter the
45551 I do not have access to a system with the first patch but not the second. I did
45552 test on Power 9 with the Linux 5.15.0-27-generic kernel. Both tests fail on
45553 this Power 9 system. The two tests also fail on Power 10 with the Linux
45554 5.14.0-70.9.1.el9_0.ppc64le kernel.
45556 The following patch fixes the issue by checking if gdb stopped at "signal
45557 handler called". If gdb stopped there, the tests verifies gdb is in the kernel
45558 function __kernel_start_sigtramp_rt64 then does an additional stepi to reach the
45559 user signal handler. With the patch below, the tests run without errors on both
45560 the Power 9 and Power 10 systems with out any failures.
45562 2022-05-06 Alan Modra <amodra@gmail.com>
45564 bfd targmatch.h makefile rule
45565 I hit this just now with a make -j build after touching config.bfd.
45566 mv: cannot stat 'targmatch.new': No such file or directory
45567 make[2]: *** [Makefile:2336: targmatch.h] Error 1
45568 make[2]: *** Waiting for unfinished jobs....
45570 Fix that by not removing the target of the rule, a practice that seems
45571 likely to cause parallel running of the rule recipe. The bug goes
45572 back to 1997, the initial c0734708814c commit.
45574 * Makefile.am (targmatch.h): rm the temp file, not targmatch.h.
45575 * Makefile.in: Regenerate.
45577 2022-05-06 Tom de Vries <tdevries@suse.de>
45579 [gdb/testsuite] Fix gdb.dwarf2/locexpr-data-member-location.exp with nopie
45580 When running test-case gdb.dwarf2/locexpr-data-member-location.exp with
45581 target board unix/-fno-PIE/-no-pie/-m32 I run into:
45585 (gdb) FAIL: gdb.dwarf2/locexpr-data-member-location.exp: step into foo
45588 The problem is that the test-case tries to mimic some gdb_compile_shlib
45591 set flags {additional_flags=-fpic debug}
45592 get_func_info foo $flags
45594 but this doesn't work with the target board setting, because we end up doing:
45596 gcc locexpr-data-member-location-lib.c -fpic -g -lm -fno-PIE -no-pie -m32 \
45597 -o func_addr23029.x
45599 while gdb_compile_shlib properly filters out the -fno-PIE -no-pie.
45601 Consequently, the address for foo determined by get_func_info doesn't match
45602 the actual address of foo.
45604 Fix this by printing the address of foo using the result of gdb_compile_shlib.
45606 2022-05-06 GDB Administrator <gdbadmin@sourceware.org>
45608 Automatic date update in version.in
45610 2022-05-05 Simon Marchi <simon.marchi@polymtl.ca>
45612 gdb: use gdb::function_view for gdbarch_iterate_over_objfiles_in_search_order callback
45613 A rather straightforward patch to change an instance of callback +
45614 void pointer to gdb::function_view, allowing pasing lambdas that
45615 capture, and eliminating the need for the untyped pointer.
45617 Change-Id: I73ed644e7849945265a2c763f79f5456695b0037
45619 2022-05-05 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
45621 gprofng: use $host instead $target
45622 By mistake, $target was used instead of $host to configure the gprogng build.
45625 2022-04-28 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
45629 * configure.ac: Use $host instead $target.
45630 * libcollector/configure.ac: Likewise.
45631 * configure: Rebuild.
45632 * libcollector/configure: Rebuild.
45634 2022-05-05 Simon Marchi <simon.marchi@efficios.com>
45636 gdb: make regcache's cooked_write_test selftest work with native-extended-gdbserver board
45639 $ make check TESTS="gdb.gdb/unittest.exp" RUNTESTFLAGS="--target_board=native-extended-gdbserver"
45641 I get some failures:
45643 Running selftest regcache::cooked_write_test::i386.^M
45644 Self test failed: target already pushed^M
45645 Running selftest regcache::cooked_write_test::i386:intel.^M
45646 Self test failed: target already pushed^M
45647 Running selftest regcache::cooked_write_test::i386:x64-32.^M
45648 Self test failed: target already pushed^M
45649 Running selftest regcache::cooked_write_test::i386:x64-32:intel.^M
45650 Self test failed: target already pushed^M
45651 Running selftest regcache::cooked_write_test::i386:x86-64.^M
45652 Self test failed: target already pushed^M
45653 Running selftest regcache::cooked_write_test::i386:x86-64:intel.^M
45654 Self test failed: target already pushed^M
45655 Running selftest regcache::cooked_write_test::i8086.^M
45656 Self test failed: target already pushed^M
45658 This is because the native-extended-gdbserver automatically connects GDB
45659 to a GDBserver on startup, and therefore pushes a remote target on the
45660 initial inferior. cooked_write_test is currently written in a way that
45661 errors out if the current inferior has a process_stratum_target pushed.
45663 Rewrite it to use scoped_mock_context, so it doesn't depend on the
45664 current inferior (the current one upon entering the function).
45666 Change-Id: I0357f989eacbdecc4bf88b043754451b476052ad
45668 2022-05-05 Luis Machado <luis.machado@arm.com>
45670 Move TILE-Gx files to TARGET64_LIBOPCODES_CFILES
45671 TILE-Gx is a 64-bit core, so we should include those files in the
45672 TARGET64_LIBOPCODES_CFILES as opposed to TARGET32_LIBOPCODES_CFILES.
45674 Don't define ARCH_cris for BFD64
45675 I believe it is a mistake to define ARCH_cris when BFD64 is defined. It is
45676 a 32-bit architecture, so should be placed outside of the BFD64 block.
45678 2022-05-05 Xi Ruoyao <xry111@mengyan1223.wang>
45680 loongarch: Don't check ABI flags if no code section
45681 Various packages (glib and gtk4 for example) produces data-only objects
45682 using `ld -r -b binary` or `objcopy`, with no elf header flags set. But
45683 these files also have no code sections, so they should be compatible
45687 * elfnn-loongarch.c (elfNN_loongarch_merge_private_bfd_data):
45688 Skip ABI checks if the input has no code sections.
45690 Reported-by: Wu Xiaotian <yetist@gmail.com>
45691 Suggested-by: Wang Xuerui <i@xen0n.name>
45693 2022-05-05 Andreas Krebbel <krebbel@linux.ibm.com>
45695 IBM zSystems: mgrk, mg first operand requires register pair
45698 * s390-opc.c (INSTR_RRF_R0RER): New instruction type.
45699 (MASK_RRF_R0RER): Define mask for new instruction type.
45700 * s390-opc.txt: Use RRF_R0RER for mgrk and RXY_RERRD for mg.
45702 2022-05-05 H.J. Lu <hjl.tools@gmail.com>
45704 bfd: Check NULL pointer before setting ref_real
45706 * linker.c (bfd_wrapped_link_hash_lookup): Check NULL pointer
45707 before setting ref_real.
45709 2022-05-05 GDB Administrator <gdbadmin@sourceware.org>
45711 Automatic date update in version.in
45713 2022-05-05 H.J. Lu <hjl.tools@gmail.com>
45715 LTO: Handle __real_SYM reference in IR
45716 When an IR symbol SYM is referenced in IR via __real_SYM, its resolution
45717 should be LDPR_PREVAILING_DEF, not PREVAILING_DEF_IRONLY, since LTO
45718 doesn't know that __real_SYM should be resolved by SYM.
45723 * linker.c (bfd_wrapped_link_hash_lookup): Mark SYM is referenced
45729 * bfdlink.h (bfd_link_hash_entry): Add ref_real.
45734 * plugin.c (get_symbols): Resolve SYM definition to
45735 LDPR_PREVAILING_DEF for __real_SYM reference.
45736 * testsuite/ld-plugin/lto.exp: Run PR ld/29086 test.
45737 * testsuite/ld-plugin/pr29086.c: New file.
45739 2022-05-04 Alan Modra <amodra@gmail.com>
45742 cris support will be built into a 32-bit bfd if using --enable-targets=all
45743 on a 32-bit host, so we may as well make targmatch.h include cris.
45745 * config.bfd (cris): Remove #idef BFD64.
45747 2022-05-04 Alan Modra <amodra@gmail.com>
45749 PowerPC64 check_relocs
45750 Tidy the dynamic reloc handling code in check_relocs, removing
45751 leftover comments and code from when check_relocs was called as each
45752 object file was read in.
45754 * elf64-ppc.c (ppc64_elf_check_relocs): Tidy dynamic reloc
45756 (dec_dynrel_count): Do the same here.
45758 2022-05-04 Tom Tromey <tromey@adacore.com>
45760 Fix crash when creating index from index
45761 My patches yesterday to unify the DWARF index base classes had a bug
45762 -- namely, I did the wholesale dynamic_cast-to-static_cast too hastily
45763 and introduced a crash. This can be seen by trying to add an index to
45764 a file that has an index, or by running a test like gdb-index-cxx.exp
45765 using the cc-with-debug-names.exp target board.
45767 This patch fixes the crash by introducing a new virtual method and
45768 removing some of the static casts.
45770 2022-05-04 Luis Machado <luis.machado@arm.com>
45772 Fix build failure for aarch64 gdbserver
45773 We're missing an argument.
45775 2022-05-04 Mark Wielaard <mark@klomp.org>
45777 gdb: Workaround stringop-overread warning in debuginfod-support.c on s390x
45778 For some reason g++ 11.2.1 on s390x produces a spurious warning for
45779 stringop-overread in debuginfod_is_enabled for url_view. Add a new
45780 DIAGNOSTIC_IGNORE_STRINGOP_OVERREAD macro to suppress this warning.
45784 * diagnostics.h (DIAGNOSTIC_IGNORE_STRINGOP_OVERREAD): New
45789 * debuginfod-support.c (debuginfod_is_enabled): Use
45790 DIAGNOSTIC_IGNORE_STRINGOP_OVERREAD on s390x.
45792 2022-05-04 Pedro Alves <pedro@palves.net>
45794 Fix GDBserver Aarch64 Linux regression
45795 Luis noticed that the recent changes to gdbserver to make it track
45796 process and threads independently regressed a few gdb.multi/*.exp
45797 tests for aarch64-linux.
45799 We started seeing the following internal error for
45800 gdb.multi/multi-target-continue.exp for example:
45802 Starting program: binutils-gdb/gdb/testsuite/outputs/gdb.multi/multi-target-continue/multi-target-continue ^M
45803 Error in re-setting breakpoint 2: Remote connection closed^M
45804 ../../../repos/binutils-gdb/gdb/thread.c:85: internal-error: inferior_thread: Assertion `current_thread_ != nullptr' failed.^M
45805 A problem internal to GDB has been detected,^M
45806 further debugging may prove unreliable.
45808 A backtrace looks like:
45810 #0 thread_regcache_data (thread=thread@entry=0x0) at ../../../repos/binutils-gdb/gdbserver/inferiors.cc:120
45811 #1 0x0000aaaaaaabf0e8 in get_thread_regcache (thread=0x0, fetch=fetch@entry=0) at ../../../repos/binutils-gdb/gdbserver/regcache.cc:31
45812 #2 0x0000aaaaaaad785c in is_64bit_tdesc () at ../../../repos/binutils-gdb/gdbserver/linux-aarch64-low.cc:194
45813 #3 0x0000aaaaaaad8a48 in aarch64_target::sw_breakpoint_from_kind (this=<optimized out>, kind=4, size=0xffffffffef04) at ../../../repos/binutils-gdb/gdbserver/linux-aarch64-low.cc:3226
45814 #4 0x0000aaaaaaabe220 in bp_size (bp=0xaaaaaab6f3d0) at ../../../repos/binutils-gdb/gdbserver/mem-break.cc:226
45815 #5 check_mem_read (mem_addr=187649984471104, buf=buf@entry=0xaaaaaab625d0 "\006", mem_len=mem_len@entry=56) at ../../../repos/binutils-gdb/gdbserver/mem-break.cc:1862
45816 #6 0x0000aaaaaaacc660 in read_inferior_memory (memaddr=<optimized out>, myaddr=0xaaaaaab625d0 "\006", len=56) at ../../../repos/binutils-gdb/gdbserver/target.cc:93
45817 #7 0x0000aaaaaaac3d9c in gdb_read_memory (len=56, myaddr=0xaaaaaab625d0 "\006", memaddr=187649984471104) at ../../../repos/binutils-gdb/gdbserver/server.cc:1071
45818 #8 gdb_read_memory (memaddr=187649984471104, myaddr=0xaaaaaab625d0 "\006", len=56) at ../../../repos/binutils-gdb/gdbserver/server.cc:1048
45819 #9 0x0000aaaaaaac82a4 in process_serial_event () at ../../../repos/binutils-gdb/gdbserver/server.cc:4307
45820 #10 handle_serial_event (err=<optimized out>, client_data=<optimized out>) at ../../../repos/binutils-gdb/gdbserver/server.cc:4520
45821 #11 0x0000aaaaaaafbcd0 in gdb_wait_for_event (block=block@entry=1) at ../../../repos/binutils-gdb/gdbsupport/event-loop.cc:700
45822 #12 0x0000aaaaaaafc0b0 in gdb_wait_for_event (block=1) at ../../../repos/binutils-gdb/gdbsupport/event-loop.cc:596
45823 #13 gdb_do_one_event () at ../../../repos/binutils-gdb/gdbsupport/event-loop.cc:237
45824 #14 0x0000aaaaaaacacb0 in start_event_loop () at ../../../repos/binutils-gdb/gdbserver/server.cc:3518
45825 #15 captured_main (argc=4, argv=<optimized out>) at ../../../repos/binutils-gdb/gdbserver/server.cc:3998
45826 #16 0x0000aaaaaaab66dc in main (argc=<optimized out>, argv=<optimized out>) at ../../../repos/binutils-gdb/gdbserver/server.cc:4084
45828 This sequence of functions is invoked due to a series of conditions:
45830 1 - The probe-based breakpoint mechanism failed (for some reason) so ...
45832 2 - ... gdbserver has to know what type of architecture it is dealing
45833 with so it can pick the right breakpoint kind, so it wants to
45834 check if we have a 64-bit target.
45836 3 - To determine the size of a register, we currently fetch the
45837 current thread's register cache, and the current thread pointer
45840 In #3, the current thread is nullptr because gdb_read_memory clears it
45841 on purpose, via set_desired_process, exactly to expose code relying on
45842 the current thread when it shouldn't. It was always possible to end
45843 up in this situation (when the current thread exits), but it was
45844 harder to reproduce before.
45846 This commit fixes it by tweaking is_64bit_tdesc to look at the current
45847 process's tdesc instead of the current thread's tdesc.
45849 Note that the thread's tdesc is itself filled from the process's
45850 tdesc, so this should be equivalent:
45853 get_thread_regcache (struct thread_info *thread, int fetch)
45855 struct regcache *regcache;
45857 regcache = thread_regcache_data (thread);
45860 if (regcache == NULL)
45862 struct process_info *proc = get_thread_process (thread);
45864 gdb_assert (proc->tdesc != NULL);
45866 regcache = new_register_cache (proc->tdesc);
45867 set_thread_regcache_data (thread, regcache);
45871 Change-Id: Ibc809d7345e70a2f058b522bdc5cdbdca97e2cdc
45873 2022-05-04 Simon Marchi <simon.marchi@efficios.com>
45875 gdb/remote: send qSymbol to all inferiors on startup
45876 start_remote_1 calls remote_check_symbols after things are set up to
45877 give the remote side a chance to look up symbols. One call to
45878 remote_check_symbols sets the "general thread", if needed, and sends one
45879 qSymbol packet. However, a remote target could have more than one
45880 process on initial connection, and this would send a qSymbol for only
45881 one of these processes.
45883 Change it to iterate on all the target's inferiors and send a qSymbol
45884 packet for each one.
45886 I tested this by changing gdbserver to spawn two processes on startup:
45888 diff --git a/gdbserver/server.cc b/gdbserver/server.cc
45889 index 33c42714e72..9b682e9f85f 100644
45890 --- a/gdbserver/server.cc
45891 +++ b/gdbserver/server.cc
45892 @@ -3939,6 +3939,7 @@ captured_main (int argc, char *argv[])
45894 /* Wait till we are at first instruction in program. */
45895 target_create_inferior (program_path.get (), program_args);
45896 + target_create_inferior (program_path.get (), program_args);
45898 /* We are now (hopefully) stopped at the first instruction of
45899 the target process. This assumes that the target process was
45901 Instead of hacking GDBserver, it should also be possible to test this by
45902 starting manually two inferiors on an "extended-remote" connection,
45903 disconnecting from GDBserver (with the disconnect command), and
45906 I was able to see qSymbol being sent for each inferior:
45908 [remote] Sending packet: $Hgp828dc.828dc#1f
45909 [remote] Packet received: OK
45910 [remote] Sending packet: $qSymbol::#5b
45911 [remote] Packet received: qSymbol:6764625f6167656e745f6764625f74705f686561705f627566666572
45912 [remote] Sending packet: $qSymbol::6764625f6167656e745f6764625f74705f686561705f627566666572#1e
45913 [remote] Packet received: qSymbol:6e70746c5f76657273696f6e
45914 [remote] Sending packet: $qSymbol::6e70746c5f76657273696f6e#4d
45915 [remote] Packet received: OK
45916 [remote] Sending packet: $Hgp828dd.828dd#21
45917 [remote] Packet received: OK
45918 [remote] Sending packet: $qSymbol::#5b
45919 [remote] Packet received: qSymbol:6764625f6167656e745f6764625f74705f686561705f627566666572
45920 [remote] Sending packet: $qSymbol::6764625f6167656e745f6764625f74705f686561705f627566666572#1e
45921 [remote] Packet received: qSymbol:6e70746c5f76657273696f6e
45922 [remote] Sending packet: $qSymbol::6e70746c5f76657273696f6e#4d
45923 [remote] Packet received: OK
45925 Note that there would probably be more work to be done to fully support
45926 this scenario, more things that need to be done for each discovered
45927 inferior instead of just for one.
45929 Change-Id: I21c4ecf6367391e2e389b560f0b4bd906cf6472f
45931 2022-05-04 Simon Marchi <simon.marchi@polymtl.ca>
45933 gdb/remote: iterate on pspace inferiors in remote_new_objfile
45934 Commit 152a17495663 ("gdb: prune inferiors at end of
45935 fetch_inferior_event, fix intermittent failure of
45936 gdb.threads/fork-plus-threads.exp") broke some tests with the
45937 native-gdbserver board, such as:
45939 (gdb) PASS: gdb.base/step-over-syscall.exp: detach-on-fork=off: follow-fork=child: break cond on target : vfork: break marker
45942 terminate called after throwing an instance of 'gdb_exception_error'^M
45944 I can manually reproduce the issue by running (just the commands that
45945 the test does as a one liner):
45947 $ ./gdb -q --data-directory=data-directory \
45948 testsuite/outputs/gdb.base/step-over-syscall/step-over-vfork \
45949 -ex "tar rem | ../gdbserver/gdbserver - testsuite/outputs/gdb.base/step-over-syscall/step-over-vfork" \
45953 -ex "set displaced-stepping off" \
45954 -ex "b *0x7ffff7d7ac5a if main == 0" \
45955 -ex "set detach-on-fork off" \
45956 -ex "set follow-fork-mode child" \
45962 ... where 0x7ffff7d7ac5a is the exact address of the vfork syscall
45963 (which can be found by looking at gdb.log).
45965 The important part of the above is that a vfork syscall creates inferior
45966 2, then inferior 2 executes until exit, then we switch back to inferior
45967 1 and try to resume it.
45969 The uncaught exception happens here:
45971 #4 0x00005596969d81a9 in error (fmt=0x559692da9e40 "Cannot execute this command while the target is running.\nUse the \"interrupt\" command to stop the target\nand then try again.")
45972 at /home/simark/src/binutils-gdb/gdbsupport/errors.cc:43
45973 #5 0x0000559695af6f66 in remote_target::putpkt_binary (this=0x617000038080, buf=0x559692da4380 "qSymbol::", cnt=9) at /home/simark/src/binutils-gdb/gdb/remote.c:9560
45974 #6 0x0000559695af6aaf in remote_target::putpkt (this=0x617000038080, buf=0x559692da4380 "qSymbol::") at /home/simark/src/binutils-gdb/gdb/remote.c:9518
45975 #7 0x0000559695ab50dc in remote_target::remote_check_symbols (this=0x617000038080) at /home/simark/src/binutils-gdb/gdb/remote.c:5141
45976 #8 0x0000559695b3cccf in remote_new_objfile (objfile=0x0) at /home/simark/src/binutils-gdb/gdb/remote.c:14600
45977 #9 0x0000559693bc52a9 in std::__invoke_impl<void, void (*&)(objfile*), objfile*> (__f=@0x61b0000167f8: 0x559695b3cb1d <remote_new_objfile(objfile*)>) at /usr/include/c++/11.2.0/bits/invoke.h:61
45978 #10 0x0000559693bb2848 in std::__invoke_r<void, void (*&)(objfile*), objfile*> (__fn=@0x61b0000167f8: 0x559695b3cb1d <remote_new_objfile(objfile*)>) at /usr/include/c++/11.2.0/bits/invoke.h:111
45979 #11 0x0000559693b8dddf in std::_Function_handler<void (objfile*), void (*)(objfile*)>::_M_invoke(std::_Any_data const&, objfile*&&) (__functor=..., __args#0=@0x7ffe0bae0590: 0x0) at /usr/include/c++/11.2.0/bits/std_function.h:291
45980 #12 0x00005596956374b2 in std::function<void (objfile*)>::operator()(objfile*) const (this=0x61b0000167f8, __args#0=0x0) at /usr/include/c++/11.2.0/bits/std_function.h:560
45981 #13 0x0000559695633c64 in gdb::observers::observable<objfile*>::notify (this=0x55969ef5c480 <gdb::observers::new_objfile>, args#0=0x0) at /home/simark/src/binutils-gdb/gdb/../gdbsupport/observable.h:150
45982 #14 0x0000559695df6cc2 in clear_symtab_users (add_flags=...) at /home/simark/src/binutils-gdb/gdb/symfile.c:2873
45983 #15 0x000055969574c263 in program_space::~program_space (this=0x6120000c8a40, __in_chrg=<optimized out>) at /home/simark/src/binutils-gdb/gdb/progspace.c:154
45984 #16 0x0000559694fc086b in delete_inferior (inf=0x61700003bf80) at /home/simark/src/binutils-gdb/gdb/inferior.c:205
45985 #17 0x0000559694fc341f in prune_inferiors () at /home/simark/src/binutils-gdb/gdb/inferior.c:390
45986 #18 0x0000559695017ada in fetch_inferior_event () at /home/simark/src/binutils-gdb/gdb/infrun.c:4293
45987 #19 0x0000559694f629e6 in inferior_event_handler (event_type=INF_REG_EVENT) at /home/simark/src/binutils-gdb/gdb/inf-loop.c:41
45988 #20 0x0000559695b3b0e3 in remote_async_serial_handler (scb=0x6250001ef100, context=0x6170000380a8) at /home/simark/src/binutils-gdb/gdb/remote.c:14466
45989 #21 0x0000559695c59eb7 in run_async_handler_and_reschedule (scb=0x6250001ef100) at /home/simark/src/binutils-gdb/gdb/ser-base.c:138
45990 #22 0x0000559695c5a42a in fd_event (error=0, context=0x6250001ef100) at /home/simark/src/binutils-gdb/gdb/ser-base.c:189
45991 #23 0x00005596969d9ebf in handle_file_event (file_ptr=0x60700005af40, ready_mask=1) at /home/simark/src/binutils-gdb/gdbsupport/event-loop.cc:574
45992 #24 0x00005596969da7fa in gdb_wait_for_event (block=0) at /home/simark/src/binutils-gdb/gdbsupport/event-loop.cc:700
45993 #25 0x00005596969d8539 in gdb_do_one_event () at /home/simark/src/binutils-gdb/gdbsupport/event-loop.cc:212
45995 If I enable "set debug infrun" just before the last continue, we see:
45999 [infrun] clear_proceed_status_thread: 965604.965604.0
46000 [infrun] proceed: enter
46001 [infrun] proceed: addr=0xffffffffffffffff, signal=GDB_SIGNAL_DEFAULT
46002 [infrun] scoped_disable_commit_resumed: reason=proceeding
46003 [infrun] start_step_over: enter
46004 [infrun] start_step_over: stealing global queue of threads to step, length = 0
46005 [infrun] operator(): step-over queue now empty
46006 [infrun] start_step_over: exit
46007 [infrun] resume_1: step=0, signal=GDB_SIGNAL_0, trap_expected=0, current thread [965604.965604.0] at 0x7ffff7d7ac5c
46008 [infrun] do_target_resume: resume_ptid=965604.0.0, step=0, sig=GDB_SIGNAL_0
46009 [infrun] prepare_to_wait: prepare_to_wait
46010 [infrun] reset: reason=proceeding
46011 [infrun] maybe_set_commit_resumed_all_targets: enabling commit-resumed for target remote
46012 [infrun] maybe_call_commit_resumed_all_targets: calling commit_resumed for target remote
46013 [infrun] proceed: exit
46014 [infrun] fetch_inferior_event: enter
46015 [infrun] scoped_disable_commit_resumed: reason=handling event
46016 [infrun] do_target_wait: Found 2 inferiors, starting at #1
46017 [infrun] random_pending_event_thread: None found.
46018 [infrun] print_target_wait_results: target_wait (-1.0.0 [process -1], status) =
46019 [infrun] print_target_wait_results: 965604.965604.0 [Thread 965604.965604],
46020 [infrun] print_target_wait_results: status->kind = VFORK_DONE
46021 [infrun] handle_inferior_event: status->kind = VFORK_DONE
46022 [infrun] context_switch: Switching context from 0.0.0 to 965604.965604.0
46023 [infrun] handle_vfork_done: not waiting for a vfork-done event
46024 [infrun] start_step_over: enter
46025 [infrun] start_step_over: stealing global queue of threads to step, length = 0
46026 [infrun] operator(): step-over queue now empty
46027 [infrun] start_step_over: exit
46028 [infrun] resume_1: step=0, signal=GDB_SIGNAL_0, trap_expected=0, current thread [965604.965604.0] at 0x7ffff7d7ac5c
46029 [infrun] do_target_resume: resume_ptid=965604.0.0, step=0, sig=GDB_SIGNAL_0
46030 [infrun] prepare_to_wait: prepare_to_wait
46031 [infrun] reset: reason=handling event
46032 [infrun] maybe_set_commit_resumed_all_targets: enabling commit-resumed for target remote
46033 [infrun] maybe_call_commit_resumed_all_targets: calling commit_resumed for target remote
46034 terminate called after throwing an instance of 'gdb_exception_error'
46038 - After doing the "continue" on inferior 1, the remote target gives us
46039 a VFORK_DONE event. The core ignores it and resumes inferior 1.
46040 - Since prune_inferiors is now called after each handled event, in
46041 fetch_inferior_event, it is called after we handled that VFORK_DONE
46042 event and resumed inferior 1.
46043 - Inferior 2 is pruned, which (see backtrace above) causes its program
46044 space to be deleted, which clears the symtabs for that program space,
46045 which calls the new_objfile observable and remote_new_objfile
46046 observer (with a nullptr objfile, to indicate that the previously
46047 loaded symbols have been discarded), which calls
46048 remote_check_symbols.
46050 remote_check_symbols is the function that sends the qSymbol packet, to
46051 let the remote side ask for symbol addresses. The problem is that the
46052 remote target is working in all-stop / sync mode and is currently
46053 resumed. It has sent a vCont packet to resume the target and is waiting
46054 for a stop reply. It can't send any packets in the mean time. That
46055 causes the exception to be thrown.
46057 This wasn't a problem before, when prune_inferiors was called in
46058 normal_stop, because it was always called at a time the target was not
46061 An important observation here is that the new_objfile observable is
46062 invoked for a change in inferior 2's program space (inferior 2's program
46063 space is the current program space). Inferior 2 isn't bound to any
46064 process on the remote side (it has exited, that's why it's being
46065 pruned). It doesn't make sense to try to send a qSymbol packet for a
46066 process that doesn't exist on the remote side. remote_check_symbols
46067 actually attempts to avoid that:
46069 /* The remote side has no concept of inferiors that aren't running
46070 yet, it only knows about running processes. If we're connected
46071 but our current inferior is not running, we should not invite the
46072 remote target to request symbol lookups related to its
46073 (unrelated) current process. */
46074 if (!target_has_execution ())
46077 The problem here is that while inferior 2's program space is the current
46078 program space, inferior 1 is the current inferior. So the check above
46079 passes, since inferior has execution. We therefore try to send a
46080 qSymbol packet for inferior 1 in reaction to a change in inferior 2's
46081 program space, that's wrong.
46083 This exposes a conceptual flaw in remote_new_objfile. The "new_objfile"
46084 event concerns a specific program space, which can concern multiple
46085 inferiors, as inferiors can share a program space. We shouldn't
46086 consider the current inferior at all, but instead all inferiors bound to
46087 the affected program space. Especially since the current inferior can
46088 be unrelated to the current program space at that point.
46090 To be clear, we are in this state because ~program_space sets itself as
46091 the current program space, but there is no more inferior having that
46092 program space to switch to, inferior 2 has already been unlinked.
46094 To fix this, make remote_new_objfile iterate on all inferiors bound to
46095 the affected program space. Remove the target_has_execution check from
46096 remote_check_symbols, replace it with an assert. All callers must
46097 ensure that the current inferior has execution before calling it.
46099 Change-Id: Ica643145bcc03115248290fd310cadab8ec8371c
46101 2022-05-04 Jan Beulich <jbeulich@suse.com>
46103 Dwarf: rename yet another instance of "index"
46104 As before, on sufficiently old glibc this conflicts with a global
46105 identifier in the library headers. While there also zap the unusual
46108 2022-05-04 Martin Liska <mliska@suse.cz>
46110 LTO plugin: sync header file with GCC
46113 * plugin-api.h (enum ld_plugin_tag): Sync with GCC.
46115 2022-05-04 Alan Modra <amodra@gmail.com>
46117 PowerPC32 treatment of absolute symbols
46118 As already done for PowerPC64, fix dynamic relocs for absolute symbols.
46119 The patch also tidies the dynamic reloc handling code in check_relocs,
46120 removing leftover comments and code from when check_relocs was called
46121 as each object file was read in.
46124 * elf32-ppc.c (ppc_elf_check_relocs): Set isym and ifunc earlier.
46125 Rearrange tests for dynamic relocs, handling absolute symbols.
46126 (allocate_dynrelocs): Don't allocate dynamic relocs for locally
46127 defined absolute symbols.
46128 (ppc_elf_size_dynamic_sections): Similarly.
46129 (ppc_elf_relocate_section): Similarly.
46131 * testsuite/ld-powerpc/abs32-pie.d,
46132 * testsuite/ld-powerpc/abs32-pie.r,
46133 * testsuite/ld-powerpc/abs32-reloc.s,
46134 * testsuite/ld-powerpc/abs32-shared.d,
46135 * testsuite/ld-powerpc/abs32-shared.r,
46136 * testsuite/ld-powerpc/abs32-static.d,
46137 * testsuite/ld-powerpc/abs32-static.r: New tests.
46138 * testsuite/ld-powerpc/powerpc.exp: Run them.
46140 2022-05-04 John Baldwin <jhb@FreeBSD.org>
46142 gdbserver: Fix build after adding tls feature to arm tdesc.
46144 2022-05-04 GDB Administrator <gdbadmin@sourceware.org>
46146 Automatic date update in version.in
46148 2022-05-04 John Baldwin <jhb@FreeBSD.org>
46150 NEWS: Add a note for TLS support on FreeBSD/arm and FreeBSD/Aarch64.
46152 Read the tpidr register from NT_ARM_TLS on Linux.
46154 gdbserver: Read the tpidr register from NT_ARM_TLS on Linux.
46156 Read the tpidr register from NT_ARM_TLS core dump notes on Linux Aarch64.
46158 Fetch the NT_ARM_TLS register set for native FreeBSD/Aarch64 processes.
46159 This permits resolving TLS variables.
46161 Support TLS variables on FreeBSD/Aarch64.
46162 Derive the pointer to the DTV array from the tpidr register.
46164 Read the tpidr register from NT_ARM_TLS core dump notes on FreeBSD/Aarch64.
46166 Add an aarch64-tls feature which includes the tpidr register.
46168 Fetch the NT_ARM_TLS register set for native FreeBSD/arm processes.
46169 This permits resolving TLS variables.
46171 Support TLS variables on FreeBSD/arm.
46172 Derive the pointer to the DTV array from the tpidruro register.
46174 Read the tpidruro register from NT_ARM_TLS core dump notes on FreeBSD/arm.
46176 Add an arm-tls feature which includes the tpidruro register from CP15.
46178 fbsd-nat: Add helper routines for register sets using PT_[G]SETREGSET.
46179 FreeBSD's kernel has recently added PT_GETREGSET and PT_SETREGSET
46180 operations to fetch a register set named by an ELF note type. These
46181 helper routines provide helpers to check for a register set's
46182 existence, fetch registers for a register set, and store registers to
46185 2022-05-03 H.J. Lu <hjl.tools@gmail.com>
46187 ld: Regenerate aclocal.m4 with automake 1.15.1
46188 * aclocal.m4: Regenerate with automake 1.15.1.
46190 2022-05-03 Pedro Alves <pedro@palves.net>
46192 gdbserver: track current process as well as current thread
46193 The recent commit 421490af33bf ("gdbserver/linux: Access memory even
46194 if threads are running") caused a regression in
46195 gdb.threads/access-mem-running-thread-exit.exp with gdbserver, which I
46196 somehow missed. Like so:
46198 (gdb) print global_var
46199 Cannot access memory at address 0x555555558010
46200 (gdb) FAIL: gdb.threads/access-mem-running-thread-exit.exp: non-stop: access mem (print global_var after writing, inf=2, iter=1)
46202 The problem starts with GDB telling GDBserver to select a thread, via
46203 the Hg packet, which GDBserver complies with, then that thread exits,
46204 and GDB, without knowing the thread is gone, tries to write to memory,
46205 through the context of the previously selected Hg thread.
46207 GDBserver's GDB-facing memory access routines, gdb_read_memory and
46208 gdb_write_memory, call set_desired_thread to make GDBserver re-select
46209 the thread that GDB has selected with the Hg packet. Since the thread
46210 is gone, set_desired_thread returns false, and the memory access
46213 Now, to access memory, it doesn't really matter which thread is
46214 selected. All we should need is the target process. Even if the
46215 thread that GDB previously selected is gone, and GDB does not yet know
46216 about that exit, it shouldn't matter, GDBserver should still know
46217 which process that thread belonged to.
46219 Fix this by making GDBserver track the current process separately,
46220 like GDB also does. Add a new set_desired_process routine that is
46221 similar to set_desired_thread, but just sets the current process,
46222 leaving the current thread as NULL. Use it in the GDB-facing memory
46223 read and write routines, to avoid failing if the selected thread is
46224 gone, but the process is still around.
46226 Change-Id: I4ff97cb6f42558efbed224b30d5c71f6112d44cd
46228 2022-05-03 Pedro Alves <pedro@palves.net>
46230 Fix gdb.threads/access-mem-running-thread-exit.exp w/ native-extended-gdbserver
46231 When testing gdb.threads/access-mem-running-thread-exit.exp with
46232 --target_board=native-extended-gdbserver, we get:
46234 Running gdb.threads/access-mem-running-thread-exit.exp ...
46235 FAIL: gdb.threads/access-mem-running-thread-exit.exp: non-stop: second inferior: runto: run to main
46236 WARNING: Timed out waiting for EOF in server after monitor exit
46238 === gdb Summary ===
46240 # of expected passes 3
46241 # of unexpected failures 1
46242 # of unsupported tests 1
46244 The problem is that the testcase spawns a second inferior with
46245 -no-connection, and then runto_main does "run", which fails like so:
46248 Don't know how to run. Try "help target".
46249 (gdb) FAIL: gdb.threads/access-mem-running-thread-exit.exp: non-stop: second inferior: runto: run to main
46251 That "run" above failed because native-extended-gdbserver forces "set
46252 auto-connect-native-target off", to prevent testcases from mistakenly
46253 running programs with the native target, which would exactly be the
46256 Fix this by letting the second inferior share the first inferior's
46257 connection everywhere except on targets that do reload on run (e.g.,
46258 --target_board=native-gdbserver).
46260 Change-Id: Ib57105a238cbc69c57220e71261219fa55d329ed
46262 2022-05-03 Andrew Burgess <aburgess@redhat.com>
46264 gdb: add some additional thread status debug output
46265 While working on this patch:
46267 https://sourceware.org/pipermail/gdb-patches/2022-January/185109.html
46269 I found it really useful to print the executing/resumed status of all
46270 threads (or all threads in a particular inferior) at various
46271 places (e.g. when a new inferior is started, when GDB attaches, etc).
46273 This debug was originally part of the above patch, but I wanted to
46274 rewrite this as a separate patch and move the code into a new function
46275 in infrun.h, which is what this patch does.
46277 Unless 'set debug infrun on' is in effect, then there should be no
46278 user visible changes after this commit.
46280 2022-05-03 Nick Clifton <nickc@redhat.com>
46282 Add a linker warning when creating potentially dangerous executable segments. Add tests, options to disabke and configure switches to choose defaults.
46284 Fix potential arithmetic overflow in the linker's plugin handling code.
46286 * libdep_plugin.c (get_libdeps): Check for overflow when computing
46287 amount of memory to allocate.
46289 2022-05-03 Andrew Burgess <aburgess@redhat.com>
46291 objdump: fix styled printing of addresses
46292 Previous work to add styled disassembler output missed a case in
46293 objdump_print_addr, which is fixed in this commit.
46295 2022-05-03 Andrew Burgess <aburgess@redhat.com>
46297 gdb/testsuite: small cleanup in mi-break-qualified.exp
46298 It is not necessary to pass an empty string to mi_gdb_start, passing
46299 the empty string is equivalent to passing no arguments, which is what
46300 we do everywhere else (that we don't need to specify an actual
46303 The only place we use 'mi_gdb_start ""' is in
46304 gdb.mi/mi-break-qualified.exp, so in this commit I just replace that
46305 with a call to 'mi_gdb_start' - just for consistency.
46307 There should be no change in what is tested after this commit.
46309 2022-05-03 Andrew Burgess <aburgess@redhat.com>
46311 gdb/testsuite: change mi_gdb_start to take a list of flags
46312 After this previous commit I was thinking about the API of
46313 mi_gdb_start. I felt that the idea of passing flags as separate
46314 arguments and using 'args' to gather these into a list, though clever,
46315 was not an intuitive API.
46317 In this commit I modify mi_gdb_start so that it expects a single
46318 argument, which should be a list of flags. Thus, where we previously
46321 mi_gdb_start separate-mi-tty separate-inferior-tty
46325 mi_gdb_start { separate-mi-tty separate-inferior-tty }
46327 However, it turns out we never actually call mi_gdb_start passing two
46328 arguments in this way at all. We do in some places do this:
46330 mi_gdb_start separate-inferior-tty
46332 But that's fine, a single string like this works equally well as a
46333 single item list, so this will not need updating.
46335 There is also one place where we do this:
46337 eval mi_gdb_start $start_ops
46339 where $start_ops is a list that might contains 0, 1, or 2 items. The
46340 eval here is used to expand the $start_ops list so mi_gdb_start sees
46341 the list contents as separate arguments. In this case we just need to
46342 drop the use of eval.
46344 I think that the new API is more intuitive, but others might
46345 disagree, in which case I can drop this change.
46347 There should be no change in what is tested after this commit.
46349 2022-05-03 Andrew Burgess <aburgess@redhat.com>
46351 gdb/testsuite: fix mi-exec-run.exp with native-extended-gdbserver board
46352 When running with the native-extended-gdbserver board, I currently see
46353 one failure in gdb.mi/mi-exec-run.exp:
46355 FAIL: gdb.mi/mi-exec-run.exp: inferior-tty=separate: mi=separate: force-fail=0: breakpoint hit reported on console (timeout)
46357 In this test the MI interface should be started in a separate tty,
46358 which means we should have a CLI tty and an MI tty, however, this is
46359 not happening. Instead GDB is just started in MI mode and there is no
46362 The test script tries to switch between the CLI an MI terminals and
46363 look for some expected output on each, however, as there is no CLI
46364 terminal the expected output never arrives, and the test times out.
46366 It turns out that this is not a GDB problem, rather, this is an issue
46367 with argument passing within the test script.
46369 The proc default_mi_gdb_start expects to take a set of flags (strings)
46370 as arguments, each of flag is expected to be a separate argument. The
46371 default_mi_gdb_start proc collects all its arguments into a list using
46372 the special 'args' parameter name, and then iterates over this list to
46373 see which flags were passed.
46375 In mi_gdb_start, which forwards to default_mi_gdb_start, the arguments
46376 are also gathered into the 'args' parameter list, but are then
46377 expanded back to be separate arguments using the eval trick, i.e.:
46379 proc mi_gdb_start { args } {
46380 return [eval default_mi_gdb_start $args]
46383 This ensures that when we arrive in default_mi_gdb_start each flag is
46384 a separate argument, rather than appearing as a single list containing
46387 When using the native-extended-gdbserver board however, the file
46388 boards/native-extended-gdbserver.exp is loaded, and this file replaces
46389 the default mi_gdb_start with its own version.
46391 This new mi_gdb_start also gathers the arguments into an 'args' list,
46392 but forgets to expand the arguments out using the eval trick.
46394 As a result, when using the native-extended-gdbserver board, by the
46395 time we get to default_mi_gdb_start, we end up with the args list
46396 containing a single item, which is a list containing all the arguments
46399 What this means is that if the user passes two arguments, then, in
46400 default_mi_gdb_start, instead of seeing two separate arguments, we see
46401 a single argument made by concatenating the two arguments together.
46403 The only place this is a problem is in the test mi-exec-run.exp,
46404 which (as far as I can see) is the only test where we might try to
46405 pass both arguments at the same time. Currently we think we passed
46406 both arguments to mi_gdb_start, but mi_gdb_start behaves as if no
46407 arguments were passed.
46409 This commit fixes the problem by making use of the eval trick within
46410 the native-extended-gdbserver version of mi_gdb_start. After this,
46411 the FAIL listed at the top of this message is resolved.
46413 2022-05-03 Andrew Burgess <aburgess@redhat.com>
46415 gdb: fix failures in gdb.mi/mi-exec-run.exp with native-extended-gdbserver
46416 When running the gdb.mi/mi-exec-run.exp test using the
46417 native-extended-gdbserver I see failures like this:
46419 FAIL: gdb.mi/mi-exec-run.exp: inferior-tty=main: mi=main: force-fail=1: run failure detected
46420 FAIL: gdb.mi/mi-exec-run.exp: inferior-tty=main: mi=separate: force-fail=1: run failure detected
46421 FAIL: gdb.mi/mi-exec-run.exp: inferior-tty=separate: mi=separate: force-fail=1: run failure detected
46423 There's a race condition here, so you might see a slightly different
46424 set of failures, but I always see some from the 'run failure detected'
46427 NOTE: I also see an additional test failure:
46429 FAIL: gdb.mi/mi-exec-run.exp: inferior-tty=separate: mi=separate: force-fail=0: breakpoint hit reported on console (timeout)
46431 but that is a completely different issue, and is not being addressed
46434 The problem for the 'run failure detected' test is that we end up
46435 in gdb_expect looking for output from two spawn-ids, one from
46436 gdbserver, and one from gdb. We're looking for one output pattern
46437 from each spawn-id, and for the test to pass we need to see both
46440 Now, if gdb exits then this is a test failure (this would indicate gdb
46441 crashing, which is bad), so we have an eof pattern associated with
46444 However, in this particular test we expect gdbserver to fail to
46445 execute the binary (the test binary is set non-executable), and so we
46446 get an error message from gdbserver (which matches the pattern), and
46447 then gdbserver exits, this is expected.
46449 The problem is that after spotting the pattern from gdbserver, we
46450 often see the eof from gdbserver before we see the pattern from gdb.
46451 If this happens then we drop out of the gdb_expect without ever seeing
46452 the pattern from gdb, and fail the test.
46454 In this commit, I place the spawn-id of gdbserver into a global
46455 variable, and then use this global variable as the -i option within
46458 Now, once we have seen the expected pattern on the gdbserver spawn-id,
46459 the global variable is cleared. After this the gdb_expect no longer
46460 checks the gdbserver spawn-id for additional output, and so never sees
46461 the eof event. This leaves the gdb_expect running, which allows the
46462 pattern from gdb to be seen, and for the test to pass.
46464 I now see no failures relating to 'run failure detected'.
46466 2022-05-03 GDB Administrator <gdbadmin@sourceware.org>
46468 Automatic date update in version.in
46470 2022-05-02 Tom de Vries <tdevries@suse.de>
46472 [gdb/testsuite] Fix gdb.cp/align.exp with gcc 12.1 / 11.3
46473 Starting with gcc 12.1 / gcc 11.3, for test-case gdb.cp/align.exp we run into:
46475 align.cc:29:23: error: invalid application of 'alignof' to a void type^M
46476 29 | unsigned a_void = alignof (void);^M
46480 Fix this by using __alignof__ instead.
46482 Tested on x86_64-linux, with gcc 7.5.0, gcc 12.1 and clang 12.0.1.
46484 2022-05-02 Aaron Merey <amerey@redhat.com>
46486 gdb/debuginfod: Whitespace-only URL should disable debuginfod
46487 Currently debuginfod is disabled when the string of server URLs
46488 is unset or set to be the empty string (via the $DEBUGINFOD_URLS
46489 environment variable or the 'set debuginfod urls' gdb command).
46491 Extend this functionality so that a whitespace-only URL also disables
46494 Modify a testcase to verify that a whitespace-only URL disables
46497 2022-05-02 Simon Marchi <simon.marchi@efficios.com>
46499 gdb: remove type_wanted parameter from a few functions
46500 The type_wanted value, passed down to the create_sals_from_location
46501 callback, is never used. Remove it.
46503 Change-Id: Ic363ee13f6af593a3e875ff7fe46de130cdc190c
46505 2022-05-02 Simon Marchi <simon.marchi@efficios.com>
46507 gnulib: update to bd11400942d6
46508 Update the gnulib import to fixes these issues:
46510 - GDB build with clang + glibc < 2.33.
46512 https://git.savannah.gnu.org/cgit/gnulib.git/commit/?id=d6a07b4dc21b3118727743142c678858df442853
46513 https://lists.gnu.org/archive/html/bug-gnulib/2022-04/msg00072.html
46515 With glibc < 2.33, gnulib (since relatively recently) enables a
46516 replacement for free (see gnulib/import/m4/free.m4). In that path,
46517 clang shows this error:
46519 make[2]: Entering directory '/home/smarchi/build/binutils-gdb-clang/gdbsupport'
46521 In file included from /home/smarchi/src/binutils-gdb/gdbsupport/agent.cc:20:
46522 In file included from /home/smarchi/src/binutils-gdb/gdbsupport/common-defs.h:95:
46523 ../gnulib/import/string.h:636:19: error: exception specification in declaration does not match previous declaration
46524 _GL_EXTERN_C void free (void *) throw ();
46526 ../gnulib/import/stdlib.h:737:17: note: expanded from macro 'free'
46527 # define free rpl_free
46529 ../gnulib/import/stdlib.h:739:1: note: previous declaration is here
46530 _GL_FUNCDECL_RPL (free, void, (void *ptr));
46532 ../gnulib/import/sys/select.h:251:23: note: expanded from macro '_GL_FUNCDECL_RPL'
46533 _GL_FUNCDECL_RPL_1 (rpl_##func, rettype, parameters_and_attributes)
46535 <scratch space>:139:1: note: expanded from here
46539 The gnulib commit mentioned fixes the exception specification of `free`.
46541 - GDB build on RHEL 7:
46543 CC libgnu_a-openat-proc.o
46544 In file included from /usr/include/string.h:633,
46545 from ./string.h:41,
46546 from ../../../binutils-gdb/gnulib/import/openat-proc.c:30:
46547 ./string.h:1105:1: error: expected identifier or '(' before '__extension__'
46548 1105 | _GL_FUNCDECL_SYS (strndup, char *,
46551 https://git.savannah.gnu.org/cgit/gnulib.git/commit/?id=84863a1c4dc8cca8fb0f6f670f67779cdd2d543b
46552 https://lists.gnu.org/archive/html/bug-gnulib/2022-04/msg00075.html
46554 Change-Id: Ibd51302feece6f385d0c53e0d08921b5d95e2776
46556 2022-05-02 Tom Tromey <tromey@adacore.com>
46558 Fix Ada catchpoint regression
46559 The breakpoint C++-ification series introduced a regression for Ada
46560 catchpoints. Specifically, commit 2b5ab5b8 ("Convert base breakpoints
46561 to vtable ops") caused these to start failing. I didn't notice this
46562 because testing Ada using a Linux distro compiler requires installing
46563 the GNAT debuginfo, which I hadn't done.
46565 This patch fixes the problem. I'm checking it in.
46567 2022-05-02 GDB Administrator <gdbadmin@sourceware.org>
46569 Automatic date update in version.in
46571 2022-05-01 Tom de Vries <tdevries@suse.de>
46573 [gdb/testsuite] Fix gdb.multi/attach-no-multi-process.exp with check-readmore
46574 When running test-case gdb.multi/attach-no-multi-process.exp with
46575 check-readmore, I get:
46577 (gdb) attach 13411^M
46578 Attaching to Remote target^M
46579 No unwaited-for children left.^M
46580 (gdb) Reading symbols from attach-no-multi-process...^M
46581 Reading symbols from /lib64/libm.so.6...^M
46582 (No debugging symbols found in /lib64/libm.so.6)^M
46583 Reading symbols from /lib64/libc.so.6...^M
46584 (No debugging symbols found in /lib64/libc.so.6)^M
46585 Reading symbols from /lib64/ld-linux-x86-64.so.2...^M
46586 (No debugging symbols found in /lib64/ld-linux-x86-64.so.2)^M
46587 0x00007f5df1fffc8a in clock_nanosleep@GLIBC_2.2.5 () from /lib64/libc.so.6^M
46588 FAIL: gdb.multi/attach-no-multi-process.exp: target_non_stop=off: \
46589 attach to the program via remote (timeout)
46592 The problem is that the attach output is matched using gdb_test, which uses
46593 the '$gdb_prompt $' regexp, and this does not handle the case that '(gdb) ' is
46594 not the last available output.
46596 Fix this by using a gdb_test_multiple instead with a '$gdb_prompt ' regexp, so
46597 without the '$' anchor.
46599 Tested on x86_64-linux with native, check-read1 and check-readmore.
46601 2022-05-01 GDB Administrator <gdbadmin@sourceware.org>
46603 Automatic date update in version.in
46605 2022-04-30 Thomas Hebb <tommyhebb@gmail.com>
46607 opcodes: don't assume ELF in riscv, csky, rl78, mep disassemblers
46608 Currently, the get_disassembler() implementations for riscv, csky, and
46609 rl78--and mep_print_insn() for mep--access ELF variants of union fields
46610 without first checking that the bfd actually represents an ELF. This
46611 causes undefined behavior and crashes when disassembling non-ELF files
46612 (the "binary" BFD, for example). Fix that.
46614 2022-04-30 GDB Administrator <gdbadmin@sourceware.org>
46616 Automatic date update in version.in
46618 2022-04-29 Tom Tromey <tom@tromey.com>
46620 Remove create_breakpoints_sal_default
46621 create_breakpoints_sal_default is just a simple wrapper, so remove it.
46623 Remove allocate_bp_location
46624 allocate_bp_location is just a small wrapper for a method call, so
46625 inline it everywhere.
46627 Constify breakpoint_ops
46628 Now that all breakpoint_ops are statically initialized, they can all
46631 Remove breakpoint ops initialization
46632 initialize_breakpoint_ops does not do much any more, so remove it in
46633 favor of statically-initialize objects.
46635 Remove vtable_breakpoint_ops
46636 There's no need to have vtable_breakpoint_ops any more, so remove it
46637 in favor of base_breakpoint_ops.
46639 Remove most fields from breakpoint_ops
46640 At this point, all implementations of breakpoints use the vtable. So,
46641 we can now remove most function pointers from breakpoint_ops and
46642 switch to using methods directly in the callers. Only the two "static
46643 virtual" methods remain in breakpoint_ops.
46645 Remove breakpoint_ops from init_catchpoint
46646 init_catchpoint is only ever passed a single breakpoint_ops pointer,
46647 so remove the parameter.
46649 Remove breakpoint_ops from init_ada_exception_breakpoint
46650 init_ada_exception_breakpoint is only ever passed a single
46651 breakpoint_ops structure, so remove the parameter.
46653 2022-04-29 Tom Tromey <tom@tromey.com>
46655 Merge probe and ordinary tracepoints
46656 Right now, probe tracepoints are handled by a separate ops object.
46657 However, they differ only in a small way from ordinary tracepoints,
46658 and furthermore can be distinguished by their event location.
46660 This patch merges the two cases, just as was done for breakpoints.
46662 2022-04-29 Tom Tromey <tom@tromey.com>
46664 Merge probe and ordinary breakpoints
46665 Right now, probe breakpoints are handled by a separate ops object.
46666 However, they differ only in a small way from ordinary breakpoints,
46667 and furthermore can be distinguished by their "probe" object.
46669 This patch merges the two cases. This avoids having to introduce a
46670 new bp_ constant (which can be quite subtle to do correctly) and a new
46673 2022-04-29 Tom Tromey <tom@tromey.com>
46675 Remove bkpt_base_breakpoint_ops
46676 An earlier patch removed the last use of bkpt_base_breakpoint_ops, so
46677 remove the object entirely.
46679 Convert static marker tracepoints to vtable ops
46680 This converts static marker tracepoints to use vtable_breakpoint_ops.
46682 Add bp_static_marker_tracepoint
46683 Because the actual construction of a breakpoint is buried deep in
46684 create_breakpoint, at present it's necessary to have a new bp_
46685 enumerator constant any time a new subclass is needed. Static marker
46686 tracepoints are one such case, so this patch introduces
46687 bp_static_marker_tracepoint and updates various spots to recognize it.
46689 Convert ranged breakpoints to vtable ops
46690 This converts ranged breakpoints to use vtable_breakpoint_ops. This
46691 requires introducing a new ranged_breakpoint type, but this is
46692 relatively simple because ranged breakpoints can only be created by
46693 break_range_command.
46695 Convert dprintf to vtable ops
46696 This converts dprintf to use vtable_breakpoint_ops.
46698 Convert Ada catchpoints to vtable ops
46699 This converts Ada catchpoints to use vtable_breakpoint_ops.
46701 Convert ordinary breakpoints to vtable ops
46702 This converts "ordinary" breakpoint to use vtable_breakpoint_ops.
46703 Recall that an ordinary breakpoint is both the kind normally created
46704 by users, and also a base class used by other classes.
46706 Change inheritance of dprintf
46707 The dprintf breakpoint ops is mostly a copy of bpkt_breakpoint_ops,
46708 except it's written out explicitly -- and, importantly, there's
46709 nothing that bpkt_breakpoint_ops overrides that dprintf does not.
46710 This changes dprintf to simply inherit directly, and updates struct
46711 dprintf_breakpoint to reflect the change as well.
46713 Convert momentary breakpoints to vtable ops
46714 This converts momentary breakpoints to use vtable_breakpoint_ops.
46716 Convert internal breakpoints to vtable ops
46717 This converts internal breakpoints to use vtable_breakpoint_ops.
46719 Convert break-catch-throw to vtable ops
46720 This converts break-catch-throw.c to use vtable_breakpoint_ops.
46722 Convert base breakpoints to vtable ops
46723 This converts base breakpoints to use vtable_breakpoint_ops.
46725 2022-04-29 Tom Tromey <tom@tromey.com>
46727 Add some new subclasses of breakpoint
46728 This adds a few new subclasses of breakpoint. The inheritance
46729 hierarchy is chosen to reflect what's already present in
46730 initialize_breakpoint_ops -- it mirrors the way that the _ops
46731 structures are filled in.
46733 This patch also changes new_breakpoint_from_type to create the correct
46734 sublcass based on bptype. This is important due to the somewhat
46735 inverted way in which create_breakpoint works; and in particular later
46736 patches will change some of these entries.
46738 2022-04-29 Tom Tromey <tom@tromey.com>
46740 Convert tracepoints to vtable ops
46741 This converts tracepoints to use vtable_breakpoint_ops.
46743 Convert watchpoints to vtable ops
46744 This converts watchpoints and masked watchpoints. to use
46745 vtable_breakpoint_ops. For masked watchpoints, a new subclass must be
46746 introduced, and watch_command_1 is changed to create one.
46748 Convert break-catch-load to vtable ops
46749 This converts break-catch-load.c to use vtable_breakpoint_ops.
46751 Convert break-catch-fork to vtable ops
46752 This converts break-catch-fork.c to use vtable_breakpoint_ops.
46754 Convert break-catch-exec to vtable ops
46755 This converts break-catch-exec.c to use vtable_breakpoint_ops.
46757 Convert break-catch-syscall to vtable ops
46758 This converts break-catch-syscall.c to use vtable_breakpoint_ops.
46760 Convert break-catch-sig to use vtable ops
46761 This converts break-catch-sig.c to use vtable_breakpoint_ops.
46763 2022-04-29 Tom Tromey <tom@tromey.com>
46765 Add a vtable-based breakpoint ops
46766 This adds methods to struct breakpoint. Each method has a similar
46767 signature to a corresponding function in breakpoint_ops, with the
46768 exceptions of create_sals_from_location and create_breakpoints_sal,
46769 which can't be virtual methods on breakpoint -- they are only used
46770 during the construction of breakpoints.
46772 Then, this adds a new vtable_breakpoint_ops structure and populates it
46773 with functions that simply forward a call from breakpoint_ops to the
46774 corresponding virtual method. These are all done with lambdas,
46775 because they are just a stepping stone -- by the end of the series,
46776 this structure will be deleted.
46778 2022-04-29 Tom Tromey <tom@tromey.com>
46780 Return bool from breakpoint_ops::print_one
46781 This changes breakpoint_ops::print_one to return bool, and updates all
46782 the implementations and the caller. The caller is changed so that a
46783 NULL check is no longer needed -- something that will be impossible
46784 with a real method.
46786 Delete some unnecessary wrapper functions
46787 This patch deletes a few unnecessary wrapper functions from
46790 Add an assertion to clone_momentary_breakpoint
46791 This adds an assertion to clone_momentary_breakpoint. This will
46792 eventually be removed, but in the meantime is is useful for helping
46793 convince oneself that momentary breakpoints will always use
46794 momentary_breakpoint_ops. This understanding will help when cleaning
46797 Boolify print_solib_event
46798 Change print_solib_event to accept a bool parameter and update the
46801 Move "catch load" to a new file
46802 The "catch load" code is reasonably self-contained, and so this patch
46803 moves it out of breakpoint.c and into a new file, break-catch-load.c.
46804 One function from breakpoint.c, print_solib_event, now has to be
46805 exposed, but this seems pretty reasonable.
46807 2022-04-29 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
46809 gprofng: assertion in gprofng/src/Expression.cc:139
46811 2022-04-28 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
46814 * src/Expression.h: Remove fixupValues.
46815 * src/Expression.cc (Expression::copy): Fix a bug.
46817 2022-04-29 Tom Tromey <tromey@adacore.com>
46819 De-duplicate .gdb_index
46820 This de-duplicates variables and types in .gdb_index, making the new
46821 index closer to what gdb generated before the new DWARF scanner
46822 series. Spot-checking the resulting index for gdb itself, it seems
46823 that the new scanner picks up some extra symbols not detected by the
46824 old one. I tested both the new and old versions of gdb on both new
46825 and old versions of the index, and startup time in all cases is
46826 roughly the same (it's worth noting that, for gdb itself, the index no
46827 longer provides any benefit over the DWARF scanner). So, I think this
46828 fixes the size issue with the new index writer.
46830 Regression tested on x86-64 Fedora 34.
46832 2022-04-29 Tom Tromey <tromey@adacore.com>
46834 Fix .debug_names regression with new indexer
46835 At AdaCore, we run the internal gdb test suite in several modes,
46836 including one using the .debug_names index. This caught a regression
46837 caused by the new DWARF indexer.
46839 First, the psymtabs-based .debug_names generator was completely wrong.
46840 However, to avoid making the rewrite series even bigger (fixing the
46841 writer will also require rewriting the .debug_names reader), it
46842 attempted to preserve the weirdness.
46844 However, this was not done properly. For example the old writer did
46847 - case STRUCT_DOMAIN:
46848 - return DW_TAG_structure_type;
46850 The new code, instead, simply preserves the actual DWARF tag -- but
46851 this makes future lookups fail, because the .debug_names reader only
46852 looks for DW_TAG_structure_type.
46854 This patch attempts to revert to the old behavior in the writer.
46856 2022-04-29 Simon Marchi <simon.marchi@polymtl.ca>
46858 gdb/infrun: make fetch_inferior_event restore thread if exited or signalled
46859 Commit 152a1749566 ("gdb: prune inferiors at end of
46860 fetch_inferior_event, fix intermittent failure of
46861 gdb.threads/fork-plus-threads.exp") introduced some follow-fork-related
46862 test failures, such as:
46865 Num Description Connection Executable ^M
46866 * 1 process 634972 1 (native) /home/simark/build/binutils-gdb-one-target/gdb/testsuite/outputs/gdb.base/foll-fork/foll-fork ^M
46867 2 process 634975 1 (native) /home/simark/build/binutils-gdb-one-target/gdb/testsuite/outputs/gdb.base/foll-fork/foll-fork ^M
46868 (gdb) PASS: gdb.base/foll-fork.exp: follow-fork-mode=parent: detach-on-fork=off: cmd=next 2: test_follow_fork: info inferiors
46870 [Switching to inferior 2 [process 634975] (/home/simark/build/binutils-gdb-one-target/gdb/testsuite/outputs/gdb.base/foll-fork/foll-fork)]^M
46871 [Switching to thread 2.1 (Thread 0x7ffff7c9a740 (LWP 634975))]^M
46872 #0 0x00007ffff7d7abf7 in _Fork () from /usr/lib/libc.so.6^M
46873 (gdb) PASS: gdb.base/foll-fork.exp: follow-fork-mode=parent: detach-on-fork=off: cmd=next 2: test_follow_fork: inferior 2
46876 [Inferior 2 (process 634975) exited normally]^M
46877 [Switching to Thread 0x7ffff7c9a740 (LWP 634972)]^M
46878 (gdb) PASS: gdb.base/foll-fork.exp: follow-fork-mode=parent: detach-on-fork=off: cmd=next 2: test_follow_fork: continue until exit at continue unfollowed inferior to end
46880 Breakpoint 2 at 0x555555555160: file /home/simark/src/binutils-gdb/gdb/testsuite/gdb.base/foll-fork.c, line 9.^M
46881 (gdb) FAIL: gdb.base/foll-fork.exp: follow-fork-mode=parent: detach-on-fork=off: cmd=next 2: test_follow_fork: break callee
46883 What happens here is:
46885 - inferior 2 is selected
46886 - we continue, leading to inferior 2's exit
46887 - we set breakpoint, expect 2 locations, but only one location is
46890 Reading between the lines, we understand that inferior 2 got pruned,
46891 when it shouldn't have been.
46893 The issue can be reproduced by hand with:
46895 $ ./gdb -q --data-directory=data-directory testsuite/outputs/gdb.base/foll-fork/foll-fork -ex "set detach-on-fork off" -ex start -ex "next 2" -ex "inferior 2" -ex "set debug infrun"
46897 Temporary breakpoint 1, main () at /home/simark/src/binutils-gdb/gdb/testsuite/gdb.base/foll-fork.c:14
46899 [New inferior 2 (process 637627)]
46900 [Thread debugging using libthread_db enabled]
46901 Using host libthread_db library "/usr/lib/../lib/libthread_db.so.1".
46902 17 if (pid == 0) /* set breakpoint here */
46903 [Switching to inferior 2 [process 637627] (/home/simark/build/binutils-gdb-one-target/gdb/testsuite/outputs/gdb.base/foll-fork/foll-fork)]
46904 [Switching to thread 2.1 (Thread 0x7ffff7c9a740 (LWP 637627))]
46905 #0 0x00007ffff7d7abf7 in _Fork () from /usr/lib/libc.so.6
46908 [infrun] clear_proceed_status_thread: 637627.637627.0
46909 [infrun] proceed: enter
46910 [infrun] proceed: addr=0xffffffffffffffff, signal=GDB_SIGNAL_DEFAULT
46911 [infrun] scoped_disable_commit_resumed: reason=proceeding
46912 [infrun] start_step_over: enter
46913 [infrun] start_step_over: stealing global queue of threads to step, length = 0
46914 [infrun] operator(): step-over queue now empty
46915 [infrun] start_step_over: exit
46916 [infrun] proceed: start: resuming threads, all-stop-on-top-of-non-stop
46917 [infrun] proceed: resuming 637627.637627.0
46918 [infrun] resume_1: step=0, signal=GDB_SIGNAL_0, trap_expected=0, current thread [637627.637627.0] at 0x7ffff7d7abf7
46919 [infrun] do_target_resume: resume_ptid=637627.637627.0, step=0, sig=GDB_SIGNAL_0
46920 [infrun] infrun_async: enable=1
46921 [infrun] prepare_to_wait: prepare_to_wait
46922 [infrun] proceed: end: resuming threads, all-stop-on-top-of-non-stop
46923 [infrun] reset: reason=proceeding
46924 [infrun] maybe_set_commit_resumed_all_targets: enabling commit-resumed for target native
46925 [infrun] maybe_call_commit_resumed_all_targets: calling commit_resumed for target native
46926 [infrun] maybe_call_commit_resumed_all_targets: calling commit_resumed for target native
46927 [infrun] proceed: exit
46928 [infrun] fetch_inferior_event: enter
46929 [infrun] scoped_disable_commit_resumed: reason=handling event
46930 [infrun] do_target_wait: Found 2 inferiors, starting at #1
46931 [infrun] random_pending_event_thread: None found.
46932 [infrun] print_target_wait_results: target_wait (-1.0.0 [process -1], status) =
46933 [infrun] print_target_wait_results: 637627.637627.0 [process 637627],
46934 [infrun] print_target_wait_results: status->kind = EXITED, exit_status = 0
46935 [infrun] handle_inferior_event: status->kind = EXITED, exit_status = 0
46936 [Inferior 2 (process 637627) exited normally]
46937 [infrun] stop_waiting: stop_waiting
46938 [infrun] stop_all_threads: start: reason=presenting stop to user in all-stop, inf=-1
46939 [infrun] stop_all_threads: pass=0, iterations=0
46940 [infrun] stop_all_threads: 637624.637624.0 not executing
46941 [infrun] stop_all_threads: pass=1, iterations=1
46942 [infrun] stop_all_threads: 637624.637624.0 not executing
46943 [infrun] stop_all_threads: done
46944 [infrun] stop_all_threads: end: reason=presenting stop to user in all-stop, inf=-1
46945 [Switching to Thread 0x7ffff7c9a740 (LWP 637624)]
46946 [infrun] infrun_async: enable=0
46947 [infrun] reset: reason=handling event
46948 [infrun] maybe_set_commit_resumed_all_targets: not requesting commit-resumed for target native, no resumed threads
46949 (gdb) [infrun] fetch_inferior_event: exit
46950 (gdb) info inferiors
46951 Num Description Connection Executable
46952 * 1 process 637624 1 (native) /home/simark/build/binutils-gdb-one-target/gdb/testsuite/outputs/gdb.base/foll-fork/foll-fork
46955 * 1 Thread 0x7ffff7c9a740 (LWP 637624) "foll-fork" main () at /home/simark/src/binutils-gdb/gdb/testsuite/gdb.base/foll-fork.c:17
46957 After handling the EXITED event for inferior 2, inferior 2 should have
46958 stayed the current inferior, which should have prevented it from getting
46959 pruned. When debugging, we find that when getting at the
46960 prune_inferiors call, the current inferior is inferior 1. Further
46961 debugging shows that prior to the call to
46962 clean_up_just_stopped_threads_fsms, the current inferior is inferior 2,
46963 and after, it's inferior 1. Then, back in fetch_inferior_event, the
46964 restore_thread object is disabled, due to:
46966 /* If we got a TARGET_WAITKIND_NO_RESUMED event, then the
46967 previously selected thread is gone. We have two
46968 choices - switch to no thread selected, or restore the
46969 previously selected thread (now exited). We chose the
46970 later, just because that's what GDB used to do. After
46971 this, "info threads" says "The current thread <Thread
46972 ID 2> has terminated." instead of "No thread
46976 && ecs->ws.kind () != TARGET_WAITKIND_NO_RESUMED)
46977 restore_thread.dont_restore ();
46979 So in the end, inferior 1 stays current, and inferior 2 gets wrongfully
46982 I'd say clean_up_just_stopped_threads_fsms is the culprit here. It
46983 actually attempts to restore the event_thread to be current at the end,
46984 after the loop (I presume the current thread on entry is always supposed
46985 to be the event thread). But in this case, the event is of kind EXITED,
46986 and ecs->event_thread is not set, so the current inferior isn't
46989 Fix that by using scoped_restore_current_thread. If there is no current
46990 thread, scoped_restore_current_thread will still restore the current
46991 inferior, and that's what we want.
46993 Random note: the thread_info object for inferior 2's thread is never
46994 freed. It is held (by refcount) by the restore_thread object in
46995 fetch_inferior_event, while the inferior's thread list gets cleared, in
46996 the exit event processing. When the refcount reaches 0 (when the
46997 restore_thread object is destroyed), there's nothing that actually
46998 deletes the thread_info object. And I think that nothing in GDB points
46999 to it anymore, so it leaks. I don't want to fix that in this patch, but
47000 thought it would be good to mention it, in case somebody has an idea for
47003 Change-Id: Ibc7df543e2c46aad5f3b9250b28c3fb5912be4e8
47005 2022-04-29 Pedro Alves <pedro@palves.net>
47007 Slightly tweak and clarify target_resume's interface
47008 The current target_resume interface is a bit odd & non-intuitive.
47009 I've found myself explaining it a couple times the recent past, while
47010 reviewing patches that assumed STEP/SIGNAL always applied to the
47011 passed in PTID. It goes like this today:
47013 - if the passed in PTID is a thread, then the step/signal request is
47016 - otherwise, if PTID is a wildcard (all threads or all threads of
47017 process), the step/signal request is for inferior_ptid, and PTID
47018 indicates which set of threads run free.
47020 Because GDB always switches the current thread to "leader" thread
47021 being resumed/stepped/signalled, we can simplify this a bit to:
47023 - step/signal are always for inferior_ptid.
47025 - PTID indicates the set of threads that run free.
47027 Still not ideal, but it's a minimal change and at least there are no
47028 special cases this way.
47030 That's what this patch does. It renames the PTID parameter to
47031 SCOPE_PTID, adds some assertions to target_resume, and tweaks
47032 target_resume's description. In addition, it also renames PTID to
47033 SCOPE_PTID in the remote and linux-nat targets, and simplifies their
47034 implementation a little bit. Other targets could do the same, but
47035 they don't have to.
47037 Change-Id: I02a2ec2ab3a3e9b191de1e9a84f55c17cab7daaf
47039 2022-04-29 GDB Administrator <gdbadmin@sourceware.org>
47041 Automatic date update in version.in
47043 2022-04-28 Tom Tromey <tromey@adacore.com>
47045 Fix libinproctrace.so build on PPC
47046 The recent gnulib import caused a build failure of libinproctrace.so
47049 alloc.c:(.text+0x20): undefined reference to `rpl_malloc'
47050 alloc.c:(.text+0x70): undefined reference to `rpl_realloc'
47052 This patch fixes the problem using the same workaround that was
47053 previously used for free.
47055 2022-04-28 H.J. Lu <hjl.tools@gmail.com>
47057 x86: Properly handle function pointer reference
47060 commit ebb191adac4ab45498dec0bfaac62f0a33537ba4
47061 Author: H.J. Lu <hjl.tools@gmail.com>
47062 Date: Wed Feb 9 15:51:22 2022 -0800
47064 x86: Disallow invalid relocation against protected symbol
47066 to allow function pointer reference and make sure that PLT entry isn't
47067 used for function reference due to function pointer reference.
47072 * elf32-i386.c (elf_i386_scan_relocs): Don't set
47073 pointer_equality_needed nor check non-canonical reference for
47074 function pointer reference.
47075 * elf64-x86-64.c (elf_x86_64_scan_relocs): Likewise.
47080 * testsuite/ld-x86-64/x86-64.exp: Run PR ld/29087 tests.
47081 * testsuite/ld-x86-64/protected-func-3.c: New file.
47083 2022-04-28 Tom Tromey <tom@tromey.com>
47085 Check OBJF_NOT_FILENAME in DWARF index code
47086 The DWARF index code currently uses 'stat' to see if an objfile
47087 represents a real file. However, I think it's more correct to check
47088 OBJF_NOT_FILENAME instead.
47090 Regression tested on x86-64 Fedora 34.
47092 2022-04-28 Tom Tromey <tromey@adacore.com>
47094 Remove "typedef enum ..."
47095 I noticed a few spots in GDB that use "typedef enum". However, in C++
47096 this isn't as useful, as the tag is automatically entered as a
47097 typedef. This patch removes most uses of "typedef enum" -- the
47098 exceptions being in some nat-* code I can't compile, and
47099 glibc_thread_db.h, which I think is more or less a copy of some C code
47102 Tested by rebuilding.
47104 2022-04-28 Andrew Burgess <aburgess@redhat.com>
47106 gdb: fix nullptr dereference in block::ranges()
47109 commit f5cb8afdd297dd68273d98a10fbfd350dff918d8
47110 Date: Sun Feb 6 22:27:53 2022 -0500
47112 gdb: remove BLOCK_RANGES macro
47114 introduces a potential nullptr dereference in block::ranges, this is
47115 breaking most tests, e.g. gdb.base/break.exp is failing for me.
47117 In the above patch BLOCK_CONTIGUOUS_P is changed from this:
47119 #define BLOCK_CONTIGUOUS_P(bl) (BLOCK_RANGES (bl) == nullptr \
47120 || BLOCK_NRANGES (bl) <= 1)
47124 #define BLOCK_CONTIGUOUS_P(bl) ((bl)->ranges ().size () == 0 \
47125 || (bl)->ranges ().size () == 1)
47127 So, before the commit we checked for the block ranges being nullptr,
47128 but afterwards we just call block::ranges() in all cases.
47130 The problem is that block::ranges() looks like this:
47132 /* Return a view on this block's ranges. */
47133 gdb::array_view<blockrange> ranges ()
47134 { return gdb::make_array_view (m_ranges->range, m_ranges->nranges); }
47138 struct blockranges *m_ranges;
47140 And so, we see that the nullptr check has been lost, and we might end
47141 up dereferencing a nullptr.
47143 My proposed fix is to move the nullptr check into block::ranges, and
47144 return an explicit empty array_view if m_ranges is nullptr.
47146 After this, everything seems fine again.
47148 2022-04-28 Stefan Liebler <stli@linux.ibm.com>
47150 s390: Add DT_JMPREL pointing to .rela.[i]plt with static-pie
47151 In static-pie case, there are IRELATIVE-relocs in
47152 .rela.iplt (htab->irelplt), which will later be grouped
47153 to .rela.plt. On s390, the IRELATIVE relocations are
47154 always located in .rela.iplt - even for non-static case.
47155 Ensure that DT_JMPREL, DT_PLTRELA, DT_PLTRELASZ is added
47156 to the dynamic section even if htab->srelplt->size == 0.
47157 See _bfd_elf_add_dynamic_tags in bfd/elflink.c.
47160 elf64-s390.c (elf_s390_size_dynamic_sections):
47161 Enforce DT_JMPREL via htab->elf.dt_jmprel_required.
47163 2022-04-28 Stefan Liebler <stli@linux.ibm.com>
47165 s390: Avoid dynamic TLS relocs in PIE
47166 No dynamic relocs are needed for TLS defined in an executable, the
47167 TP relative offset is known at link time.
47170 FAIL: Build pr22263-1
47174 * elf64-s390.c (elf_s390_tls_transition): Use bfd_link_dll
47175 instead of bfd_link_pic for TLS.
47176 (elf_s390_check_relocs): Likewise.
47177 (allocate_dynrelocs): Likewise.
47178 (elf_s390_relocate_section): Likewise.
47180 2022-04-28 Nick Alcock <nick.alcock@oracle.com>
47182 libctf: impose an ordering on conflicting types
47183 When two types conflict and they are not types which can have forwards
47184 (say, two arrays of different sizes with the same name in two different
47185 TUs) the CTF deduplicator uses a popularity contest to decide what to
47186 do: the type cited by the most other types ends up put into the shared
47187 dict, while the others are relegated to per-CU child dicts.
47189 This works well as long as one type *is* most popular -- but what if
47190 there is a tie? If several types have the same popularity count,
47191 we end up picking the first we run across and promoting it, and
47192 unfortunately since we are working over a dynhash in essentially
47193 arbitrary order, this means we promote a random one. So multiple
47194 runs of ld with the same inputs can produce different outputs!
47195 All the outputs are valid, but this is still undesirable.
47197 Adjust things to use the same strategy used to sort types on the output:
47198 when there is a tie, always put the type that appears in a CU that
47199 appeared earlier on the link line (and if there is somehow still a tie,
47200 which should be impossible, pick the type with the lowest type ID).
47202 Add a testcase -- and since this emerged when trying out extern arrays,
47203 check that those work as well (this requires a newer GCC, but since all
47204 GCCs that can emit CTF at all are unreleased this is probably OK as
47207 Fix up one testcase that has slight type ordering changes as a result
47212 * ctf-dedup.c (ctf_dedup_detect_name_ambiguity): Use
47213 cd_output_first_gid to break ties.
47217 * testsuite/ld-ctf/array-conflicted-ordering.d: New test, using...
47218 * testsuite/ld-ctf/array-char-conflicting-1.c: ... this...
47219 * testsuite/ld-ctf/array-char-conflicting-2.c: ... and this.
47220 * testsuite/ld-ctf/array-extern.d: New test, using...
47221 * testsuite/ld-ctf/array-extern.c: ... this.
47222 * testsuite/ld-ctf/conflicting-typedefs.d: Adjust for ordering
47225 2022-04-28 Nick Alcock <nick.alcock@oracle.com>
47227 libctf: add a comment explaining how to use ctf_*open
47228 Specifically, tell users what to pass to those functions that accept raw
47229 section content, since it's fairly involved and easy to get wrong.
47230 (.dynsym / .dynstr when CTF_F_DYNSTR is set, otherwise .symtab / .strtab).
47234 * ctf-api.h (ctf_*open): Improve comment.
47236 2022-04-28 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
47238 gprofng: test suite problems
47240 2022-04-27 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
47243 * testsuite/lib/Makefile.skel: Search parent dir for libs too.
47245 2022-04-28 Simon Marchi <simon.marchi@polymtl.ca>
47247 gdb: remove BLOCKVECTOR_MAP macro
47248 Replace with equivalent methods.
47250 Change-Id: I4e56c76dfc363c1447686fb29c4212ea18b4dba0
47252 2022-04-28 Simon Marchi <simon.marchi@polymtl.ca>
47254 gdb: constify addrmap_find
47255 addrmap_find shouldn't need to modify the addrmap, so constify the
47256 addrmap parameter. This helps for the following patch, where getting
47257 the map of a const blockvector will return a const addrmap.
47259 Change-Id: If670e425ed013724a3a77aab7961db50366dccb2
47261 2022-04-28 Simon Marchi <simon.marchi@efficios.com>
47263 gdb: remove BLOCKVECTOR_BLOCK and BLOCKVECTOR_NBLOCKS macros
47264 Replace with calls to blockvector::blocks, and the appropriate method
47265 call on the returned array_view.
47267 Change-Id: I04d1f39603e4d4c21c96822421431d9a029d8ddd
47269 2022-04-28 Simon Marchi <simon.marchi@efficios.com>
47271 gdb: remove BLOCK_ENTRY_PC macro
47272 Replace with equivalent method.
47274 Change-Id: I0e033095e7358799930775e61028b48246971a7d
47276 2022-04-28 Simon Marchi <simon.marchi@efficios.com>
47278 gdb: remove BLOCK_CONTIGUOUS_P macro
47279 Replace with an equivalent method.
47281 Change-Id: I60fd3be7b4c2601c2a74328f635fa48ed80eb7f5
47283 2022-04-28 Simon Marchi <simon.marchi@efficios.com>
47285 gdb: remove BLOCK_RANGE macro
47286 Replace with access through the block::ranges method.
47288 Change-Id: I50f3ed433b997c9f354e49bc6583f540ae4b6121
47290 2022-04-28 Simon Marchi <simon.marchi@efficios.com>
47292 gdb: remove BLOCK_NRANGES macro
47293 Replace with range for loops.
47295 Change-Id: Icbe04f9b6f9e6ddae2e15b2409c61f7a336bc3e3
47297 2022-04-28 Simon Marchi <simon.marchi@efficios.com>
47299 gdb: remove BLOCK_RANGES macro
47300 Replace with an equivalent method on struct block.
47302 Change-Id: I6dcf13e9464ba8a08ade85c89e7329c300fd6c2a
47304 2022-04-28 Simon Marchi <simon.marchi@efficios.com>
47306 gdb: remove BLOCK_RANGE_{START,END} macros
47307 Replace with equivalent methods on blockrange.
47309 Change-Id: I20fd8f624e0129782c36768291891e7582d77c74
47311 2022-04-28 Simon Marchi <simon.marchi@efficios.com>
47313 gdb: remove BLOCK_NAMESPACE macro
47314 Replace with equivalent methods.
47316 Change-Id: If86b8cbdfb0f52e22c929614cd53e73358bab76a
47318 2022-04-28 Simon Marchi <simon.marchi@efficios.com>
47320 gdb: remove BLOCK_MULTIDICT macro
47321 Replace with equivalent methods.
47323 Change-Id: If9a239c511a664f2a59fecb6d1cd579881b23dc2
47325 2022-04-28 Simon Marchi <simon.marchi@efficios.com>
47327 gdb: remove BLOCK_SUPERBLOCK macro
47328 Replace with equivalent methods.
47330 Change-Id: I334a319909a50b5cc5570a45c38c70e10dc00630
47332 2022-04-28 Simon Marchi <simon.marchi@efficios.com>
47334 gdb: remove BLOCK_FUNCTION macro
47335 Replace with equivalent methods.
47337 Change-Id: I31ec00f5bf85335c8b23d306ca0fe0b84d489101
47339 2022-04-28 Simon Marchi <simon.marchi@efficios.com>
47341 gdb: remove BLOCK_{START,END} macros
47342 Replace with equivalent methods.
47344 Change-Id: I10a6c8a2a86462d9d4a6a6409a3f07a6bea66310
47346 2022-04-28 GDB Administrator <gdbadmin@sourceware.org>
47348 Automatic date update in version.in
47350 2022-04-27 H.J. Lu <hjl.tools@gmail.com>
47352 x86: Disable 2 tests with large memory requirement
47355 * testsuite/gas/i386/i386.exp: Disable rept.
47359 * testsuite/ld-x86-64/x86-64.exp: Disable pr17618.
47361 2022-04-27 Pedro Alves <pedro@palves.net>
47363 Make gdb.base/parse_number.exp test all architectures
47364 There are some subtle differences between architectures, like the size
47365 of a "long" type, and this isn't currently accounted for in
47366 gdb.base/parse_number.exp.
47368 For example, on aarch64 a long type is 8 bytes, whereas a long type is
47369 4 bytes for x86_64. This causes the following FAIL's:
47371 FAIL: gdb.base/parse_number.exp: lang=asm: ptype 0xffffffffffffffff
47372 FAIL: gdb.base/parse_number.exp: lang=auto: ptype 0xffffffffffffffff
47373 FAIL: gdb.base/parse_number.exp: lang=c: ptype 0xffffffffffffffff
47374 FAIL: gdb.base/parse_number.exp: lang=c++: ptype 0xffffffffffffffff
47375 FAIL: gdb.base/parse_number.exp: lang=fortran: p/x 0xffffffffffffffff
47376 FAIL: gdb.base/parse_number.exp: lang=fortran: ptype 0xffffffffffffffff
47377 FAIL: gdb.base/parse_number.exp: lang=go: ptype 0xffffffffffffffff
47378 FAIL: gdb.base/parse_number.exp: lang=local: ptype 0xffffffffffffffff
47379 FAIL: gdb.base/parse_number.exp: lang=minimal: ptype 0xffffffffffffffff
47380 FAIL: gdb.base/parse_number.exp: lang=objective-c: ptype 0xffffffffffffffff
47381 FAIL: gdb.base/parse_number.exp: lang=opencl: ptype 0xffffffffffffffff
47382 FAIL: gdb.base/parse_number.exp: lang=pascal: ptype 0xffffffffffffffff
47384 There are some fortran-specific divergences as well, where 32-bit
47385 architectures show "unsigned int" for both 32-bit and 64-bit integers
47386 and 64-bit architectures show "unsigned int" and "unsigned long" for
47387 32-bit and 64-bit integers.
47389 There might be a bug that 32-bit fortran truncates 64-bit values to
47390 32-bit, given "p/x 0xffffffffffffffff" returns "0xffffffff".
47392 Here's what we get for aarch64:
47394 (gdb) ptype 0xffffffff
47395 type = unsigned int
47396 (gdb) ptype 0xffffffffffffffff
47397 type = unsigned long
47398 (gdb) p sizeof (0xffffffff)
47400 (gdb) p sizeof (0xffffffffffffffff)
47403 (gdb) ptype 0xffffffff
47404 type = unsigned int
47405 (gdb) ptype 0xffffffffffffffff
47406 type = unsigned long
47410 (gdb) ptype 0xffffffff
47411 type = unsigned int
47412 (gdb) ptype 0xffffffffffffffff
47414 type = unsigned long long
47415 (gdb) p sizeof (0xffffffff)
47418 (gdb) p sizeof (0xffffffffffffffff)
47421 (gdb) ptype 0xffffffff
47422 type = unsigned int
47423 (gdb) ptype 0xffffffffffffffff
47424 type = unsigned long
47428 * Makes the testcase iterate over all architectures, thus covering all
47429 the different combinations of types/sizes every time.
47431 * Adjusts the expected values and types based on the sizes of long
47432 long, long and int.
47434 A particularly curious architecture is s12z, which has 32-bit long
47435 long, and thus no way to represent 64-bit integers in C-like
47438 Co-Authored-By: Luis Machado <luis.machado@arm.com>
47439 Change-Id: Ifc0ccd33e7fd3c7585112ff6bebe7d266136768b
47441 2022-04-27 Tom Tromey <tromey@adacore.com>
47443 Fix gdbserver build for x86-64 Windows
47444 I broke the gdbserver build on x86-64 Windows a little while back.
47445 Previously, I could not build this configuration, but today I found
47446 out that if I configure with:
47448 --host=x86_64-w64-mingw32 --target=x86_64-w64-mingw32
47450 using the Fedora 34 tools, it will in fact build. I'm not certain,
47451 but maybe the gnulib update helped with this.
47453 This patch fixes the build. I'm checking it in.
47455 2022-04-27 John Baldwin <jhb@FreeBSD.org>
47457 Create pseudo sections for NT_ARM_TLS notes on FreeBSD.
47460 * elf.c (elfcore_grok_freebsd_note): Handle NT_ARM_TLS notes.
47462 2022-04-27 Christophe Lyon <christophe.lyon@arm.com>
47464 gdb/arm: Extend arm_m_addr_is_magic to support FNC_RETURN, add unwind-secure-frames command
47465 This patch makes use of the support for several stack pointers
47466 introduced by the previous patch to switch between them as needed
47469 It introduces a new 'unwind-secure-frames' arm command to enable/disable
47470 mode switching during unwinding. It is enabled by default.
47472 It has been tested using an STM32L5 board (with cortex-m33) and the
47473 sample applications shipped with the STM32Cube development
47474 environment: GTZC_TZSC_MPCBB_TrustZone in
47475 STM32CubeL5/Projects/NUCLEO-L552ZE-Q/Examples/GTZC.
47477 The test consisted in setting breakpoints in various places and check
47478 that the backtrace is correct: SecureFault_Callback (Non-secure mode),
47479 __gnu_cmse_nonsecure_call (before and after the vpush instruction),
47480 SecureFault_Handler (Secure mode).
47482 This implies that we tested only some parts of this patch (only MSP*
47483 were used), but remaining parts seem reasonable.
47485 2022-04-27 Christophe Lyon <christophe.lyon@arm.com>
47487 gdb/arm: Add support for multiple stack pointers on Cortex-M
47488 Armv8-M architecture with Security extension features four stack pointers
47489 to handle Secure and Non-secure modes.
47491 This patch adds support to switch between them as needed during
47492 unwinding, and replaces all updates of cache->prev_sp with calls to
47493 arm_cache_set_prev_sp.
47495 2022-04-27 Christophe Lyon <christophe.lyon@arm.com>
47497 gdb/arm: Introduce arm_cache_init
47498 This patch is a preparation for the rest of the series and adds two
47499 arm_cache_init helper functions. It updates every place that updates
47500 cache->saved_regs to call the helper instead.
47502 gdb/arm: Define MSP and PSP registers for M-Profile
47503 This patch removes the hardcoded access to PSP in
47504 arm_m_exception_cache() and relies on the definition with the XML
47507 2022-04-27 Christophe Lyon <christophe.lyon@arm.com>
47509 gdb/arm: Fix prologue analysis to support vpush
47510 While working on adding support for Non-secure/Secure modes unwinding,
47511 I noticed that the prologue analysis lacked support for vpush, which
47512 is used for instance in the CMSE stub routine.
47514 This patch updates thumb_analyze_prologue accordingly, adding support
47515 for vpush of D-registers.
47517 2022-04-27 Enze Li <lienze2010@hotmail.com>
47519 gdb/testsuite: fix FAIL in gdb.base/clear_non_user_bp.exp
47520 Tom and Simon feedback that there is a test failing in this commit:
47522 commit a5c69b1e49bae4d0dcb20f324cebb310c63495c6
47523 Date: Sun Apr 17 15:09:46 2022 +0800
47525 gdb: fix using clear command to delete non-user breakpoints(PR cli/7161)
47527 Then, I reproduced the same fail with Ubuntu 20.04 as Simon said, and I
47528 fixed the nit in this patch. The root of the problem is not correctly
47529 matching the presentation of internal breakpoints.
47531 In addition, as Pedro pointed out, the original testcase is not portable
47532 in some methods, so this patch fixes this issue and some other
47535 Tested on x86_64 ubuntu 20.04.4 and openSUSE Tumbleweed(VERSION_ID="20220425").
47537 2022-04-27 Jan Beulich <jbeulich@suse.com>
47539 x86: VFPCLASSSH is Evex.LLIG
47540 This also was mistakenly flagged as Evex.128.
47542 2022-04-27 Nick Clifton <nickc@redhat.com>
47544 Fix potential buffer overruns when creating DLLs.
47546 * pe-dll.c (make_head): Use asprintf to allocate and populate a
47547 buffer containing the temporary name.
47548 (make_tail, make_one, make_singleton_name_thunk): Likewise.
47549 (make_import_fixup_mark, make_import_fixup_entry): Likewise.
47550 (make_runtime_pseudo_reloc): Likewise.
47551 (pe_create_runtime_relocator_reference): Likewise.
47553 2022-04-27 Alan Modra <amodra@gmail.com>
47555 Revert pr29072 lto test changes
47556 Revert commit 65daf5bed6 testsuite changes in ld-plugin/. -z isn't
47557 supported for non-ELF targets, and isn't needed since we now prune the
47558 exec stack warning (commit 333cd559ba).
47562 2022-04-27 Simon Marchi <simon.marchi@polymtl.ca>
47564 gdb/testsuite: use with_cwd where possible
47565 I learned about with_cwd today. I spotted a few spots that could use
47566 it, to make the code more robust.
47568 Change-Id: Ia23664cb827f25e79d31948e0c006a8dc61c33e1
47570 2022-04-27 GDB Administrator <gdbadmin@sourceware.org>
47572 Automatic date update in version.in
47574 2022-04-26 Carl Love <cel@us.ibm.com>
47576 GDB PowerPC record test cases for ISA 2.06 and ISA 3.1
47577 This patch adds PowerPC specific tests to verify recording of various
47578 instructions. The first test case checks the ISA 2.06 lxvd2x instruction.
47579 The second test case tests several of the ISA 3.01 instructions. Specifically,
47580 it checks the word and prefixed instructions and some of the Matrix
47581 Multiply Assist (MMA) instructions.
47583 The patch has been run on both Power 10 and Power 9 to verify the ISA
47584 2.06 test case runs on both platforms without errors. The ISA 3.1 test
47585 runs without errors on Power 10 and is skipped as expected on Power 9.
47587 2022-04-26 Carl Love <cel@us.ibm.com>
47589 Add recording support for the ISA 3.1 PowerPC instructions.
47590 This patch adds support for the PowerPC ISA 3.1 instructions to the PowerPC
47591 gdb instruction recording routines. Case statement entries are added to a
47592 number of the existing routines for recording the 32-bit word instructions.
47593 A few new functions were added to handle the new word instructions. The 64-bit
47594 prefix instructions are all handled by a set of new routines. The function
47595 ppc_process_prefix_instruction() is the primary function to handle the
47596 prefixed instructions. It calls additional functions to handle specific
47597 sets of prefixed instructions. These new functions are:
47598 ppc_process_record_prefix_vsx_d_form(),
47599 ppc_process_record_prefix_store_vsx_ds_form(),
47600 ppc_process_record_prefix_op34(),
47601 ppc_process_record_prefix_op33(),
47602 ppc_process_record_prefix_op32(),
47603 ppc_process_record_prefix_store(),
47604 ppc_process_record_prefix_op59_XX3(),
47605 ppc_process_record_prefix_op42().
47607 2022-04-26 Tom Tromey <tromey@adacore.com>
47609 Handle encoding failures in Windows thread names
47610 Internally at AdaCore, we noticed that the new Windows thread name
47611 code could fail. First, it might return a zero-length string, but in
47612 gdb conventions it should return nullptr instead. Second, an encoding
47613 failure could wind up showing replacement characters to the user; this
47614 is confusing and not useful; it's better to recognize such errors and
47615 simply discard the name. This patch makes both of these changes.
47617 2022-04-26 H.J. Lu <hjl.tools@gmail.com>
47619 i386: Pass -z noexecstack to linker tests
47621 * testsuite/ld-i386/i386.exp: Pass -z noexecstack to gotpc1
47624 2022-04-26 John Baldwin <jhb@FreeBSD.org>
47626 bsd-kvm: Fix build after recent changes to path handling functions.
47627 Convert bsd_kvm_corefile and the local filename in bsd_kvm_open to
47628 std::string rather than simple char * pointers freed by xfree.
47630 2022-04-26 Simon Marchi <simon.marchi@polymtl.ca>
47632 gdb: make some random Python files Python 3-compatible
47633 I noticed that these files failed to format with Black, because they use
47634 print without parenthesis (which isn't Python 3 compatible).
47636 I don't know if these files are still relevant, but the change is
47637 trivial, so here it is.
47639 Change-Id: I116445c2b463486016f824d32effffc915b60766
47641 2022-04-26 Carl Love <cel@us.ibm.com>
47643 PowerPC: Update expected floating point output for gdb.arch/altivec-regs.exp and gdb.arch/vsx-regs.exp
47644 The format for printing the floating point values was changed by commit:
47646 commit 56262a931b7ca8ee3ec9104bc7e9e0b40cf3d64e
47647 Author: Tom Tromey <tromey@adacore.com>
47648 Date: Thu Feb 17 13:43:59 2022 -0700
47650 Change how "print/x" displays floating-point value
47652 Currently, "print/x" will display a floating-point value by first
47653 casting it to an integer type. This yields weird results like:
47658 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=16242
47660 The above change results in 417 regression test failures since the expected
47661 Power vector register output no longer match.
47663 This patch updates the expected Altivec floating point register prints to
47664 the hexadecimal format for both big endian and little endian systems. The
47665 patch also fixes a formatting isue with the decimal_vector expected value
47668 The expected VSX vector_register1, vector_register1_vr, vector_register2,
47669 vector_register2_vr variables are updated to include the new float128 entry.
47670 Additionally, the comment in the vsx expect file about the initialization
47671 of the vs registers is updated.
47673 The patch has been tested on Power 10, Power 8 LE and Power 8 BE.
47675 2022-04-26 John Baldwin <jhb@FreeBSD.org>
47677 gdbsupport/pathstuff.h: #include <array> explicitly for std::array<>
47678 This fixes build breakage using clang with libc++ on FreeBSD where
47679 std::array<> is not yet declared when used by the path_join variadic
47682 2022-04-26 GDB Administrator <gdbadmin@sourceware.org>
47684 Automatic date update in version.in
47686 2022-04-25 Tom Tromey <tromey@adacore.com>
47688 Do not put linkage names into .gdb_index
47689 This changes the .gdb_index writer to skip linkage names. This was
47690 always done historically (though somewhat implicitly).
47692 2022-04-25 Nick Clifton <nickc@redhat.com>
47694 Emit a note warning the user that creating an executable stack because of a missing .note.GNU-stack section is deprecated.
47696 bfd * elflink.c (bfd_elf_size_dynamic_sections): Display a note to the
47697 user that the current ehaviour of creating an executable stack
47698 because of a missing .note.GNU-stack section is deprecated and
47699 will be changed in a future release.
47701 binutils* testsuite/lib/binutils-common.exp (prune_warnings_extra): Filter
47702 out notes about the executable stacjk behaviour beign deprecated.
47704 ld * testsuite/ld-elf/pr29072.b.warn: Update to include the note
47705 about the linker's behaviour being depreccated.
47707 2022-04-25 rupothar <rupesh.potharla@amd.com>
47709 gdb/fortran: Support for assumed rank zero
47710 If a variable is passed to function in FORTRAN as an argument the
47711 variable is treated as an array with rank zero. GDB currently does
47712 not support the case for assumed rank 0. This patch provides support
47713 for assumed rank 0 and updates the testcase as well.
47716 Breakpoint 1, arank::sub1 (a=<error reading variable:
47717 failed to resolve dynamic array rank>) at assumedrank.f90:11
47718 11 PRINT *, RANK(a)
47720 failed to resolve dynamic array rank
47722 failed to resolve dynamic array rank
47725 Breakpoint 1, arank::sub1 (a=0) at assumedrank.f90:11
47726 11 PRINT *, RANK(a)
47732 2022-04-25 Lancelot SIX <lancelot.six@amd.com>
47734 gdb/infrun: assert !step_over_info_valid_p in restart_threads
47735 While working in gdb/infrun.c:restart_threads, I was wondering what are
47736 the preconditions to call the function. It seems to me that
47737 !step_over_info_valid_p should be a precondition (i.e. if we are doing
47738 an inline step over breakpoint, we do not want to resume non stepping
47739 threads as one of them might miss the breakpoint which is temporally
47742 To convince myself that this is true, I have added an assertion to
47743 enforce this, and got one regression in the testsuite:
47745 FAIL: gdb.base/step-over-syscall.exp: vfork: displaced=off: single step over vfork (GDB internal error)
47747 This call to restart_threads originates from handle_vfork_done which
47748 does not check if a step over is active when restarting other threads:
47750 if (target_is_non_stop_p ())
47752 scoped_restore_current_thread restore_thread;
47754 insert_breakpoints ();
47755 restart_threads (event_thread, event_thread->inf);
47756 start_step_over ();
47759 In this patch, I propose to:
47760 - Call start_step_over before restart_threads. If a step over is already
47761 in progress (as it is the case in the failing testcase),
47762 start_step_over return immediately, and there is no point in restarting
47763 all threads just to stop them right away for a step over breakpoint.
47764 - Only call restart_threads if no step over is in progress at this
47767 In this patch, I also propose to keep the assertion in restart_threads
47768 to help enforce this precondition, and state it explicitly.
47770 I have also checked all other places which call restart_threads, and
47771 they all seem to check that there is no step over currently active
47772 before doing the call.
47774 As for infrun-related things, I am never completely sure I did not miss
47775 something. So as usual, all feedback and thoughts are very welcome.
47777 Tested on x86_64-linux-gnu.
47779 Change-Id: If5f5f98ec4cf9aaeaabb5e3aa88ae6ffd70d4f37
47781 2022-04-25 GDB Administrator <gdbadmin@sourceware.org>
47783 Automatic date update in version.in
47785 2022-04-24 Andrew Burgess <aburgess@redhat.com>
47787 gdb: move setbuf calls out of gdb_readline_no_editing_callback
47790 commit d08cbc5d3203118da5583296e49273cf82378042
47791 Date: Wed Dec 22 12:57:44 2021 +0000
47793 gdb: unbuffer all input streams when not using readline
47795 Issues were reported with some MS-Windows hosts, see the thread
47798 https://sourceware.org/pipermail/gdb-patches/2022-March/187004.html
47800 Filed in bugzilla as: PR mi/29002
47802 The problem seems to be that calling setbuf on terminal file handles
47803 is not always acceptable, see this mail for more details:
47805 https://sourceware.org/pipermail/gdb-patches/2022-April/187310.html
47807 This commit does two things, first moving the setbuf calls out of
47808 gdb_readline_no_editing_callback so that we don't end up calling
47811 Then, for MS-Windows hosts, we don't call setbuf for terminals, this
47812 appears to resolve the issues that have been reported.
47814 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29002
47816 2022-04-24 GDB Administrator <gdbadmin@sourceware.org>
47818 Automatic date update in version.in
47820 2022-04-23 GDB Administrator <gdbadmin@sourceware.org>
47822 Automatic date update in version.in
47824 2022-04-22 Simon Marchi <simon.marchi@efficios.com>
47826 gdb: handle_no_resumed: only update thread list of event target
47829 $ make check TESTS="gdb.multi/multi-re-run.exp" RUNTESTFLAGS="--target_board=native-gdbserver"
47833 target remote localhost:2347^M
47834 Remote debugging using localhost:2347^M
47835 Reading symbols from /lib64/ld-linux-x86-64.so.2...^M
47836 Reading symbols from /usr/lib/debug//lib/x86_64-linux-gnu/ld-2.31.so...^M
47837 0x00007ffff7fd0100 in _start () from /lib64/ld-linux-x86-64.so.2^M
47840 Cannot execute this command while the target is running.^M
47841 Use the "interrupt" command to stop the target^M
47842 and then try again.^M
47843 (gdb) FAIL: gdb.multi/multi-re-run.exp: re_run_inf=1: iter=1: runto: run to all_started
47847 - Connect to a remote target with inferior 2, continue and stop on the
47848 all_started function
47849 - Connect to a separate remote target / GDBserver instance with inferior 1,
47850 continue and (expect to) stop on the all_started function
47852 The failure seen above happens when trying to continue inferior 1.
47856 - GDB tells inferior 1's remote target to continue
47857 - We go into fetch_inferior_event, try to get an event at random from
47859 - do_target_wait happens to pick inferior 2's target
47860 - That target return TARGET_WAITKIND_NO_RESUMED, which makes sense:
47861 inferior 2 is stopped, that target has no resumed thread
47862 - handle_no_resumed tries to update the thread list of all targets:
47864 for (auto *target : all_non_exited_process_targets ())
47866 switch_to_target_no_thread (target);
47867 update_thread_list ();
47870 - When trying to update the thread list of inferior 1's target, it hits
47871 the "Cannot execute this command while the target is running" error.
47872 This target is working in "remote all-stop" mode, and it is currently
47873 waiting for a stop reply, so it can't send packets to update the
47874 thread list at this time.
47876 To handle the problem described in the comment in handle_no_resumed, I
47877 don't think it is necessary to update the thread list of all targets,
47878 but only the event target. That comment describes a kind of race
47879 condition where some target reports a breakpoint hit for a thread and
47880 then its last remaining resumed thread exits, so sends a "no resumed"
47881 event. If we ended up resuming the thread that hit a breakpoint, we
47882 want to ignore the "no resumed" and carry on.
47884 But I don't really see why we need to update the thread list on the
47885 other targets. I can't really articulate this, it's more a gut feeling,
47886 maybe I just fail to imagine the situation where this is needed. But
47887 here is the patch anyway, so we can discuss it. The patch changes
47888 handle_no_resumed to only update the thread list of the event target.
47889 This fixes the test run shown above.
47891 The way I originally tried to fix this was to make
47892 remote_target::update_thread_list return early if the target is
47893 currently awaiting a stop reply, since there's no way it can update the
47894 thread list at that time. But that felt more like papering over the
47895 problem. I then thought that we probably shouldn't be asking the target
47896 to update the thread list unnecessarily.
47898 Change-Id: Ide3df22b4f556478e155ad1c67ad4b4fe7c26a58
47900 2022-04-22 Simon Marchi <simon.marchi@polymtl.ca>
47902 gdbserver/linux: free process_info_private and arch_process_info when failing to attach
47905 $ ../gdbserver/gdbserver --once --attach :1234 539436
47907 with ASan while /proc/sys/kernel/yama/ptrace_scope is set to 1 (prevents
47908 attaching) shows that we fail to free some platform-specific objects
47909 tied to the process_info (process_info_private and arch_process_info):
47911 Direct leak of 32 byte(s) in 1 object(s) allocated from:
47912 #0 0x7f6b558b3fb9 in __interceptor_calloc /usr/src/debug/gcc/libsanitizer/asan/asan_malloc_linux.cpp:154
47913 #1 0x562eaf15d04a in xcalloc /home/simark/src/binutils-gdb/gdbserver/../gdb/alloc.c:100
47914 #2 0x562eaf251548 in xcnew<process_info_private> /home/simark/src/binutils-gdb/gdbserver/../gdbsupport/poison.h:122
47915 #3 0x562eaf22810c in linux_process_target::add_linux_process_no_mem_file(int, int) /home/simark/src/binutils-gdb/gdbserver/linux-low.cc:426
47916 #4 0x562eaf22d33f in linux_process_target::attach(unsigned long) /home/simark/src/binutils-gdb/gdbserver/linux-low.cc:1132
47917 #5 0x562eaf1a7222 in attach_inferior /home/simark/src/binutils-gdb/gdbserver/server.cc:308
47918 #6 0x562eaf1c1016 in captured_main /home/simark/src/binutils-gdb/gdbserver/server.cc:3949
47919 #7 0x562eaf1c1d60 in main /home/simark/src/binutils-gdb/gdbserver/server.cc:4084
47920 #8 0x7f6b552f630f in __libc_start_call_main (/usr/lib/libc.so.6+0x2d30f)
47922 Indirect leak of 56 byte(s) in 1 object(s) allocated from:
47923 #0 0x7f6b558b3fb9 in __interceptor_calloc /usr/src/debug/gcc/libsanitizer/asan/asan_malloc_linux.cpp:154
47924 #1 0x562eaf15d04a in xcalloc /home/simark/src/binutils-gdb/gdbserver/../gdb/alloc.c:100
47925 #2 0x562eaf2a0d79 in xcnew<arch_process_info> /home/simark/src/binutils-gdb/gdbserver/../gdbsupport/poison.h:122
47926 #3 0x562eaf295e2c in x86_target::low_new_process() /home/simark/src/binutils-gdb/gdbserver/linux-x86-low.cc:723
47927 #4 0x562eaf22819b in linux_process_target::add_linux_process_no_mem_file(int, int) /home/simark/src/binutils-gdb/gdbserver/linux-low.cc:428
47928 #5 0x562eaf22d33f in linux_process_target::attach(unsigned long) /home/simark/src/binutils-gdb/gdbserver/linux-low.cc:1132
47929 #6 0x562eaf1a7222 in attach_inferior /home/simark/src/binutils-gdb/gdbserver/server.cc:308
47930 #7 0x562eaf1c1016 in captured_main /home/simark/src/binutils-gdb/gdbserver/server.cc:3949
47931 #8 0x562eaf1c1d60 in main /home/simark/src/binutils-gdb/gdbserver/server.cc:4084
47932 #9 0x7f6b552f630f in __libc_start_call_main (/usr/lib/libc.so.6+0x2d30f)
47934 Those objects are deleted by linux_process_target::mourn, but that is
47935 not called if we fail to attach, we only call remove_process. I
47936 initially fixed this by making linux_process_target::attach call
47937 linux_process_target::mourn on failure (before calling error). But this
47938 isn't done anywhere else (including in GDB) so it would just be
47939 confusing to do things differently here.
47941 Instead, add a linux_process_target::remove_linux_process helper method
47942 (which calls remove_process), and call that instead of remove_process in
47943 the Linux target. Move the free-ing of the extra data from the mourn
47944 method to that new method.
47946 Change-Id: I277059a69d5f08087a7f3ef0b8f1792a1fcf7a85
47948 2022-04-22 Andrew Burgess <aburgess@redhat.com>
47950 gdb: handle bracketed-paste-mode and EOF correctly
47951 This commit replaces an earlier commit that worked around the issues
47952 reported in bug PR gdb/28833.
47954 The previous commit just implemented a work around in order to avoid
47955 the worst results of the bug, but was not a complete solution. The
47956 full solution was considered too risky to merge close to branching GDB
47957 12. This improved fix has been applied after GDB 12 branched. See
47958 this thread for more details:
47960 https://sourceware.org/pipermail/gdb-patches/2022-March/186391.html
47962 This commit replaces this earlier commit:
47964 commit 74a159a420d4b466cc81061c16d444568e36740c
47965 Date: Fri Mar 11 14:44:03 2022 +0000
47967 gdb: work around prompt corruption caused by bracketed-paste-mode
47969 Please read that commit for a full description of the bug, and why is
47972 In this commit I extend GDB to use readline's rl_deprep_term_function
47973 hook to call a new function gdb_rl_deprep_term_function. From this
47974 new function we can now print the 'quit' message, this replaces the
47975 old printing of 'quit' in command_line_handler. Of course, we only
47976 print 'quit' in gdb_rl_deprep_term_function when we are handling EOF,
47977 but thanks to the previous commit (to readline) we now know when this
47980 There are two aspects of this commit that are worth further
47981 discussion, the first is in the new gdb_rl_deprep_term_function
47982 function. In here I have used a scoped_restore_tmpl to disable the
47983 readline global variable rl_eof_found.
47985 The reason for this is that, in rl_deprep_terminal, readline will
47986 print an extra '\n' character before printing the escape sequence to
47987 leave bracketed paste mode. You might then think that in the
47988 gdb_rl_deprep_term_function function, we could simply print "quit" and
47989 rely on rl_deprep_terminal to print the trailing '\n'. However,
47990 rl_deprep_terminal only prints the '\n' when bracketed paste mode is
47991 on. If the user has turned this feature off, no '\n' is printed.
47992 This means that in gdb_rl_deprep_term_function we need to print
47993 "quit" when bracketed paste mode is on, and "quit\n" when bracketed
47996 We could absolutely do that, no problem, but given we know how
47997 rl_deprep_terminal is implemented, it's easier (I think) to just
47998 temporarily clear rl_eof_found, this prevents the '\n' being printed
47999 from rl_deprep_terminal, and so in gdb_rl_deprep_term_function, we can
48000 now always print "quit\n" and this works for all cases.
48002 The second issue that should be discussed is backwards compatibility
48003 with older versions of readline. GDB can be built against the system
48004 readline, which might be older than the version contained within GDB's
48005 tree. If this is the case then the system readline might not contain
48006 the fixes needed to support correctly printing the 'quit' string.
48008 To handle this situation I have retained the existing code in
48009 command_line_handler for printing 'quit', however, this code is only
48010 used now if the version of readline we are using doesn't not include
48011 the required fixes. And so, if a user is using an older version of
48012 readline, and they have bracketed paste mode on, then they will see
48013 the 'quit' sting printed on the line below the prompt, like this:
48018 I think this is the best we can do when someone builds GDB against an
48019 older version of readline.
48021 Using a newer version of readline, or the patched version of readline
48022 that is in-tree, will now give a result like this in all cases:
48026 Which is what we want.
48028 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28833
48030 2022-04-22 Andrew Burgess <aburgess@redhat.com>
48032 readline: back-port changes needed to properly detect EOF
48033 This commit is a partial back-port of this upstream readline commit:
48035 commit 002d31aa5f5929eb32d0e0e2e8b8d35d99e59961
48036 Author: Chet Ramey <chet.ramey@case.edu>
48037 Date: Thu Mar 3 11:11:47 2022 -0500
48039 add rl_eof_found to public API; fix pointer aliasing problems \
48040 with history-search-backward; fix a display problem with \
48041 runs of invisible characters at the end of a physical \
48044 I have only pulled in the parts of this commit that relate to the new
48045 rl_eof_found global, and the RL_STATE_EOF state flag. These changes
48046 are needed in order to fix PR cli/28833, and are discussed in this
48047 thread to the bug-readline mailing list:
48049 https://lists.gnu.org/archive/html/bug-readline/2022-02/msg00021.html
48051 The above commit is not yet in any official readline release, but my
48052 hope is that now it has been merged into the readline tree it should
48053 be safe enough to back port this fix to GDB's tree.
48055 At some point in the future we will inevitably want to roll forward
48056 the version of readline that we maintain in the binutils-gdb
48057 repository. When that day comes the changes in this commit can be
48058 replaced with the latest upstream readline code, as I have not changed
48059 the meaning of this code at all from what is in upstream readline.
48061 This commit alone does not fix the PR cli/28833 issue, for that see
48062 the next commit, which changes GDB itself.
48064 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28833
48066 2022-04-22 Andrew Burgess <aburgess@redhat.com>
48068 gdb: improved EOF handling when using readline 7
48071 commit a6b413d24ccc5d76179bab866834e11fd6fec294
48072 Date: Fri Mar 11 14:44:03 2022 +0000
48074 gdb: work around prompt corruption caused by bracketed-paste-mode
48076 a change was made to GDB to work around bug PR gdb/28833. The
48077 consequence of this work around is that, when bracketed paste mode is
48078 enabled in readline, and GDB is quit by sending EOF, then the output
48079 will look like this:
48084 The ideal output, which is what we get when bracketed paste mode is
48089 The reason we need to make this change is explained in the original
48090 commit referenced above. What isn't mentioned in the above commit, is
48091 that the change that motivated this work around was only added in
48092 readline 8, older versions of readline don't require the change.
48094 In later commits in this series I will add a fix to GDB's in-tree copy
48095 of readline (this fix is back-ported from upstream readline), and then
48096 I will change GDB so that, when using the (patched) in-tree readline,
48097 we can have the ideal output in all cases.
48099 However, GDB can be built against the system readline. When this is
48100 done, and the system readline is version 8, then we will still have to
48101 use the work around (two line) style output.
48103 But, if GDB is built against the system readline, and the system
48104 readline is an older version 7 readline, then there's no reason why we
48105 can't have the ideal output, after all, readline 7 doesn't include the
48106 change that we need to work around.
48108 This commit changes GDB so that, when using readline 7 we get the
48109 ideal output in all cases. This change is trivial (a simple check
48110 against the readline version number) so I think this should be fine to
48113 For testing this commit, you need to configure GDB including the
48114 '--with-system-readline' flag, and build GDB on a system that uses
48115 readline 7, for example 'Ubuntu 18.04'. Then run the test
48116 'gdb.base/eof-exit.exp', you should expect everything to PASS.
48118 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28833
48120 2022-04-22 Simon Marchi <simon.marchi@efficios.com>
48122 gdb: prune inferiors at end of fetch_inferior_event, fix intermittent failure of gdb.threads/fork-plus-threads.exp
48123 This test sometimes fail like this:
48126 Id Target Id Frame ^M
48127 11.12 process 2270719 Couldn't get registers: No such process.^M
48128 (gdb) FAIL: gdb.threads/fork-plus-threads.exp: detach-on-fork=off: no threads left
48129 [Inferior 11 (process 2270719) exited normally]^M
48131 Num Description Connection Executable ^M
48132 * 1 <null> /home/smarchi/build/binutils-gdb/gdb/testsuite/outputs/gdb.threads/fork-plus-threads/fork-plus-threads ^M
48133 11 <null> /home/smarchi/build/binutils-gdb/gdb/testsuite/outputs/gdb.threads/fork-plus-threads/fork-plus-threads ^M
48134 (gdb) FAIL: gdb.threads/fork-plus-threads.exp: detach-on-fork=off: only inferior 1 left (the program exited)
48136 I can get it to fail quite reliably by pinning it to a core:
48138 $ taskset -c 5 make check TESTS="gdb.threads/fork-plus-threads.exp"
48140 The previous attempt at fixing this was:
48142 https://sourceware.org/pipermail/gdb-patches/2021-October/182846.html
48144 What we see is part due to a possible unfortunate ordering of events
48145 given by the kernel, and what could be considered a bug in GDB.
48147 The test program makes a number of forks, waits them all, then exits.
48148 Most of the time, GDB will get and process the exit event for inferior 1
48149 after the exit events of all the children. But this is not guaranteed.
48150 After the last child exits and is waited by the parent, the parent can
48151 exit quickly, such that GDB collects from the kernel the exit events for
48152 the parent and that child at the same time. It then chooses one event
48153 at random, which can be the event for the parent. This will result in
48154 the parent appearing to exit before its child. There's not much we can
48155 do about it, so I think we have to adjust the test to cope.
48157 After expect has seen the "exited normally" notification for inferior 1,
48158 it immediately does an "info thread" that it expects to come back empty.
48159 But at this point, GDB might not have processed inferior 11's (the last
48160 child) exit event, so it will look like there is still a thread. Of
48161 course that thread is dead, we just don't know it yet. But that makes
48162 the "no thread" test fail. If the test waited just a bit more for the
48163 "exited normally" notification for inferior 11, then the list of threads
48166 So, first change, make the test collect all the "exited normally"
48167 notifications for all inferiors before proceeding, that should ensure we
48168 see an empty thread list. That would fix the first FAIL above.
48170 However, we would still have the second FAIL, as we expect inferior 11
48171 to not be there, it should have been deleted automatically. Inferior 11
48172 is normally deleted when prune_inferiors is called. That is called by
48173 normal_stop, which is only called by fetch_inferior_event only if the
48174 event thread completed an execution command FSM (thread_fsm). But the
48175 FSM for the continue command completed when inferior 1 exited. At that
48176 point inferior 11 was not prunable, as it still had a thread. When
48177 inferior 11 exits, prune_inferiors is not called.
48179 I think that can be considered a GDB bug. From the user point of view,
48180 there's no reason why in one case inferior 11 would be deleted and not
48183 This patch makes the somewhat naive change to call prune_inferiors in
48184 fetch_inferior_event, so that it is called in this case. It is placed
48185 at this particular point in the function so that it is called after the
48186 user inferior / thread selection is restored. If it was called before
48187 that, inferior 11 wouldn't be pruned, because it would still be the
48190 Change-Id: I48a15d118f30b1c72c528a9f805ed4974170484a
48191 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=26272
48193 2022-04-22 Tom Tromey <tromey@adacore.com>
48195 Un-break the coff-pe-read.c build
48196 This fixes a build breakage in my recent coff-pe-read.c change.
48198 I'm sorry about this. I don't understand how it happened, because I
48199 definitely built and tested the series on Windows, and I didn't change
48200 it before pushing. Something must have gone wrong on the Windows
48201 build, but I don't know what. I'll investigate and and re-test to be
48204 2022-04-22 Tom Tromey <tromey@adacore.com>
48206 More const use and alloca avoidance in coff-pe-read.c
48207 This changes another function in coff-pe-read.c to use 'const' more,
48208 and to avoid the use of alloca by instead using std::string.
48210 Use std::string in coff-pe-read.c
48211 coff-pe-read.c uses xsnprintf and alloca, but using std::string is
48212 better, and just as easy. In general I think alloca is something to
48213 be avoided, and unbounded uses especially so.
48215 Remove a const-removing cast from coff-pe-read.c
48216 coff-pe-read.c casts away const at one spot, but this is easily
48217 replaced by calling bfd_get_filename directly in a couple of debugging
48220 Simplify BFD section iteration in coff-pe-read.c
48221 coff-pe-read.c iterates over BFD sections using bfd_map_over_sections,
48222 but it's much simpler to use a for-each loop. This allows for the
48223 removal of helper functions and types.
48225 2022-04-22 Tom Tromey <tromey@adacore.com>
48227 Fix method naming bug in new DWARF indexer
48228 Pedro pointed out that gdb-add-index is much slower with the new DWARF
48229 indexer. He also noticed that, in some cases, the generated
48230 .gdb_index would have the wrong fully-qualified name for a method.
48232 I tracked this down to a bug in the indexer. If a type could have
48233 methods but was marked as a declaration, the indexer was ignoring it.
48234 However, this meant that the internal map to find the qualified name
48235 was not updated for this container.
48237 2022-04-22 Christoph Muellner <cmuellner@gcc.gnu.org>
48239 RISC-V: Add missing DECLARE_INSNs for Zicbo{m,p,z}
48240 The recently added support for the Zicbo{m,p,z} extensions did not
48241 include DECLARE_INSN() declarations for the instructions.
48242 These declarations are needed by GDB's instruction detection code.
48243 This patch adds them.
48245 2022-04-22 GDB Administrator <gdbadmin@sourceware.org>
48247 Automatic date update in version.in
48249 2022-04-21 Carl Love <cel@us.ibm.com>
48251 Fix for gdb.base/solib-search.exp test.
48252 The variable right_lib_flags is not being set correctly to define RIGHT.
48253 The value RIGHT is needed to force the address of the library functions
48254 lib1_func3 and lib2_func4 to occur at different address in the wrong and
48257 With RIGHT defined correctly, functions lib1_func3 and lib2_func4 occur
48258 at different addresses the test runs correctly on Powerpc.
48260 The test needs the lib2 addresses to be different in the right and
48261 wrong cases. That is the point of introducing function lib2_spacer
48262 with the ifdef RIGHT compiler directive.
48264 On Intel, the ARRAY_SIZE of 1 versus 8192 is sufficient to get the
48265 dynamic linker to move the addresses of the library. You can also get
48266 the same effect on PowerPC but you must use a value much larger than
48269 The key thing is that the test was not properly setting RIGHT to
48270 defined to get the lib2_spacer function on Intel and Powerpc.
48272 Without the patch, we have the Intel backtrace for the bad libraries:
48275 #0 break_here () at /home/ ... /gdb/testsuite/gdb.base/solib-search.c:30
48276 #1 0x00007ffff7fae156 in ?? ()
48277 #2 0x00007fffffffc150 in ?? ()
48278 #3 0x00007ffff7fbb156 in ?? ()
48279 #4 0x00007fffffffc160 in ?? ()
48280 #5 0x00007ffff7fae146 in ?? ()
48281 #6 0x00007fffffffc170 in ?? ()
48282 #7 0x00007ffff7fbb146 in ?? ()
48283 #8 0x00007fffffffc180 in ?? ()
48284 #9 0x0000555555555156 in main () at /home/ ... /binutils-gdb/gdb/testsuite/gdb.base/solib-search.c:23
48285 Backtrace stopped: previous frame inner to this frame (corrupt stack?)
48286 (gdb) PASS: gdb.base/solib-search.exp: backtrace (with wrong libs) (data collection)
48288 The backtrace on Intel with the good libraries is:
48291 #0 break_here () at /.../binutils-gdb/gdb/testsuite/gdb.base/solib-search.c:30
48292 #1 0x00007ffff7fae156 in lib2_func4 () at /.../binutils-gdb/gdb/testsuite/gdb.base/solib-search-lib2.c:49
48293 #2 0x00007ffff7fbb156 in lib1_func3 () at /.../gdb.base/solib-search-lib1.c:49
48294 #3 0x00007ffff7fae146 in lib2_func2 () at /.../testsuite/gdb.base/solib-search-lib2.c:30
48295 #4 0x00007ffff7fbb146 in lib1_func1 () at /.../gdb.base/solib-search-lib1.c:30
48296 #5 0x0000555555555156 in main () at /...solib-search.c:23
48297 (gdb) PASS: gdb.base/solib-search.exp: backtrace (with right libs) (data collection)
48298 PASS: gdb.base/solib-search.exp: backtrace (with right libs)
48300 In one case the backtrace is correct and the other it
48301 is wrong on Intel. This is due to the fact that the ARRAY_SIZE caused
48302 the dynamic linker to move the library function addresses around. I
48303 believe it has to do with the default size of the data and code
48304 sections used by the dynamic linker.
48306 So without the patch the backtrace on PowerPC looks like:
48309 #0 break_here () at /.../solib-search.c:30
48310 #1 0x00007ffff7f007f4 in lib2_func4 () at /.../solib-search-lib2.c:49
48311 #2 0x00007ffff7f307f4 in lib1_func3 () at /.../solib-search-lib1.c:49
48312 #3 0x00007ffff7f007ac in lib2_func2 () at /.../solib-search-lib2.c:30
48313 #4 0x00007ffff7f307ac in lib1_func1 () at /.../solib-search-lib1.c:30
48314 #5 0x000000001000074c in main () at /.../solib-search.c:23
48316 for both the good and bad libraries.
48318 The patch fixes defining RIGHT in solib-search-lib1.c and solib-search-
48319 lib2.c. Note, without the patch the lib1_spacer and lib2_spacer
48320 functions do not show up in the object dump of the Intel or Powerpc
48321 libraries as it should. The patch fixes that by making sure RIGHT gets
48324 Now with the patch the backtrace for the bad library on PowerPC looks
48328 #0 break_here () at /.../solib-search.c:30
48329 #1 0x00007ffff7f0083c in __glink_PLTresolve () from /.../solib-search-lib2.so
48330 Backtrace stopped: frame did not save the PC
48332 And the backtrace for the good libraries on PowerPC looks like:
48335 #0 break_here () at /.../solib-search.c:30
48336 #1 0x00007ffff7f0083c in lib2_func4 () at /.../solib-search-lib2.c:49
48337 #2 0x00007ffff7f3083c in lib1_func3 () at /.../solib-search-lib1.c:49
48338 #3 0x00007ffff7f007cc in lib2_func2 () at /.../solib-search-lib2.c:30
48339 #4 0x00007ffff7f307cc in lib1_func1 () at /.../solib-search-lib1.c:30
48340 #5 0x000000001000074c in main () at /.../solib-search.c:23
48341 (gdb) PASS: gdb.base/solib-search.exp: backtrace (with right libs) (data collection)
48342 PASS: gdb.base/solib-search.exp: backtrace (with right libs)
48344 The issue then is on Power where the ARRAY_SIZE of 1 versus 8192 is not
48345 sufficient to cause the dymanic linker to allocate the libraries at
48346 different addresses. I don't claim to understand the specifics of how
48347 the dynamic linker works and what the default size is for the data and
48348 code sections are. My guess is by default PowerPC allocates a larger
48349 data size by default, which is large enough to hold array[8192]. The
48350 default size of the data section allocated by the dynamic linker on
48351 Intel is not large enough to hold array[8192] thus causing the code
48352 section on Intel to have to move when the large array is defined.
48354 Note on PowerPC, if you make ARRAY_SIZE big enough, then you will cause
48355 the library addresses to occur at different addresses as the larger
48356 data section forces the code section to a different address. That was
48357 actually my original fix for the program until I spoke with Doug Evans
48358 who originally wrote the test. Doug noticed that RIGHT was not getting
48359 defined as he originally intended in the test.
48361 With the patch to fix the definition of RIGHT, PowerPC has a bad and a
48362 good backtrace because the address of lib1_func3 and lib2_func4 both
48363 move because lib1_spacer and lib2_spacer are now defined
48364 before lib1_func3 and lib2_func4.
48366 Without the patch, the lib1_spacer and lib2_spacer function doesn't show
48367 up in the binary for the correct or incorrect library on Intel or PowerPC.
48368 With the patch, RIGHT gets defined as originally intended for the test on
48369 both architectures and lib1_spacer and lib2_spacer function show up in the
48370 binaries on both architectures changing the other function addresses as
48371 intended thus causing the test work as intended on PowerPC.
48373 2022-04-21 Simon Marchi <simon.marchi@efficios.com>
48375 gdb/dwarf: remove line_header::header_length field
48376 This can be a local in dwarf_decode_line_header.
48378 Change-Id: I2ecf4616d1a3197bd1e81ded9f999a2da9a685af
48380 2022-04-21 Simon Marchi <simon.marchi@efficios.com>
48382 gdb/dwarf: remove line_header::total_length field
48383 This doesn' have to be a field, it can simply be a local variable in
48384 dwarf_decode_line_header. Name the local variable "unit_length", since
48385 that's what the field in called in DWARF 4 and 5. It's always easier to
48386 follow the code with the standard on the side when we use the same
48389 Change-Id: I3ad1022afd9410b193ea11b9b5437686c1e4e633
48391 2022-04-21 Simon Marchi <simon.marchi@efficios.com>
48393 gdb/testsuite: fix "set temporary breakpoint" DUPLICATEs
48394 Commit c67f4e538 ("gdb/testsuite: make gdb.ada/mi_prot.exp stop at
48395 expected location") introduced some DUPLICATEs in MI tests using
48396 mi_continue_to_line, for example:
48398 DUPLICATE: gdb.ada/mi_ref_changeable.exp: mi_continue_to_line: set temporary breakpoint
48400 These test names were previously differentiated by the location passed
48401 to mi_continue_to_line. Since the location can contain a path, that
48402 commit removed the location from the test name, in favor of a hardcoded
48403 string "set temporary breakpoint", hence removing the differentiator.
48405 mi_continue_to_line receives a "test" parameter, containing a test
48406 name. Add a "with_test_prefix" with that name, so that all tests
48407 recorded during mi_continue_to_line have this in their name.
48409 mi_continue_to_line passes that "test" string to mi_get_stop_line, that
48410 is a bit superfluous. mi_get_stop_line only uses that string in case of
48411 failures (it doesn't record a pass if everything goes fine). Since it's
48412 not crucial, just remove it, and adjust all callers.
48414 Adjust three gdb.mi/mi-var-*.exp tests to use prefixes to differentiate
48415 the multiple calls to mi_run_inline_test (which calls
48416 mi_continue_to_line).
48418 Change-Id: I511c6caa70499f8657b1cde37d71068d74d56a74
48420 2022-04-21 Tom Tromey <tromey@adacore.com>
48422 Always use dwarf2_initialize_objfile
48423 Internally we noticed that some tests would fail like so on Windows:
48425 warning: Section .debug_aranges in [...] has duplicate debug_info_offset 0x0, ignoring .debug_aranges.
48427 Debugging showed that, in fact, a second CU was being created at this
48428 offset. We tracked this down to the fact that, while the ELF reader
48429 is careful to re-use the per-BFD data, other readers are not, and
48430 could re-read the DWARF data multiple times.
48432 However, since the change to allow an objfile to have multiple "quick
48433 symbol" implementations, there's no reason for this approach -- it's
48434 safe and easy for all symbol readers to reuse the per-BFD data when
48437 This patch implements this idea, simplifying dwarf2_build_psymtabs and
48438 making it private, and then switching to dwarf2_initialize_objfile as
48439 the sole way to start the DWARF reader.
48441 Note that, while I think the call to dwarf2_build_frame_info in
48442 machoread.c is also obsolete, I haven't attempted to remove it here.
48444 2022-04-21 Andrew Burgess <aburgess@redhat.com>
48446 gdb: fix 'remote show FOO-packet' aliases
48447 The following behaviour was observed in GDB:
48449 (gdb) show remote X-packet
48450 Support for the `p' packet is auto-detected, currently unknown.
48452 Note the message mentions the 'p' packet. This is a regression since
48455 commit 8579fd136a614985bd27f20539c7bb7c5a51287d
48456 Date: Mon Nov 8 14:58:46 2021 +0000
48458 gdb/gdbsupport: make xstrprintf and xstrvprintf return a unique_ptr
48460 Before this commit the behaviour was:
48462 (gdb) show remote X-packet
48463 Support for the `X' packet is auto-detected, currently unknown.
48465 The problem was caused by a failed attempt to ensure that some
48466 allocated strings were deleted when GDB exits. The code in the above
48467 commit attempted to make use of 'static' to solve this problem,
48468 however, the solution was just wrong.
48470 In this new commit I instead allocate a static vector into which all
48471 the allocated strings are stored, this ensures the strings are
48472 released when GDB exits (which makes output from tools like valgrind
48473 cleaner), but each string within the vector can be unique, which fixes
48476 2022-04-21 Simon Marchi <simon.marchi@efficios.com>
48478 gdbsupport: add path_join function
48479 In this review [1], Eli pointed out that we should be careful when
48480 concatenating file names to avoid duplicated slashes. On Windows, a
48481 double slash at the beginning of a file path has a special meaning. So
48482 naively concatenating "/" and "foo/bar" would give "//foo/bar", which
48483 would not give the desired results. We already have a few spots doing:
48485 if (first_path ends with a slash)
48486 path = first_path + second_path
48488 path = first_path + slash + second_path
48490 In general, I think it's nice to avoid superfluous slashes in file
48491 paths, since they might end up visible to the user and look a bit
48494 Introduce the path_join function that can be used to join multiple path
48495 components together (along with unit tests).
48497 I initially wanted to make it possible to join two absolute paths, to
48498 support the use case of prepending a sysroot path to a target file path,
48499 or the prepending the debug-file-directory to a target file path. But
48500 the code in solib_find_1 shows that it is more complex than this anyway
48501 (for example, when the right hand side is a Windows path with a drive
48502 letter). So I don't think we need to support that case in path_join.
48503 That also keeps the implementation simpler.
48505 Change a few spots to use path_join to show how it can be used. I
48506 believe that all the spots I changed are guarded by some checks that
48507 ensure the right hand side operand is not an absolute path.
48509 Regression-tested on Ubuntu 18.04. Built-tested on Windows, and I also
48510 ran the new unit-test there.
48512 [1] https://sourceware.org/pipermail/gdb-patches/2022-April/187559.html
48514 Change-Id: I0df889f7e3f644e045f42ff429277b732eb6c752
48516 2022-04-21 Lancelot SIX <lancelot.six@amd.com>
48518 gdb_spawn_attach_cmdline: use unsupported instead of untested
48519 In a previous commit (b750766ac96: gdb/testsuite: Introduce and use
48520 gdb_spawn_attach_cmdline), if gdb_spawn_attach_cmdline cannot have GDB
48521 attach to the process because of ptrace restrictions (operation not
48522 permitted), the proc issues UNTESTED. This should really be
48523 UNSUPPORTED, as it is done in gdb_attach.
48525 This patch fixes this oversight.
48527 Change-Id: Ib87e33b9230f3fa7a85e06220ef4c63814b71f7d
48529 2022-04-21 Enze Li <lienze2010@hotmail.com>
48531 gdb/testsuite: add binary testcases to py-format-string.exp
48532 We currently only test decimal and hexadecimal for the
48533 gdb.Value.format_string() interface, this patch adds testcases for
48536 Tested on x86_64 openSUSE Tumbleweed(VERSION_ID="20220413").
48538 2022-04-21 Pedro Alves <pedro@palves.net>
48540 gdb.debuginfod/fetch_src_and_symbols.exp: Fix "notice empty URL" test
48541 The gdb_test_multiple pattern for the "notice empty URL" test in
48542 gdb.debuginfod/fetch_src_and_symbols.exp misses expecting the prompt.
48543 Fix it by using -re -wrap.
48545 Also, by using "confirm off", the message GDB prints if Debuginfod
48546 downloading is available doesn't contain "Enable debuginfod" any
48550 (gdb) file testsuite/outputs/gdb.debuginfod/fetch_src_and_symbols/fetch_src_and_symbols
48551 Reading symbols from testsuite/outputs/gdb.debuginfod/fetch_src_and_symbols/fetch_src_and_symbols...
48553 This GDB supports auto-downloading debuginfo from the following URLs:
48554 <http://localhost:123>
48555 Enable debuginfod for this session? (y or [n])
48559 (gdb) with confirm off -- file testsuite/outputs/gdb.debuginfod/fetch_src_and_symbols/fetch_src_and_symbols
48560 Reading symbols from testsuite/outputs/gdb.debuginfod/fetch_src_and_symbols/fetch_src_and_symbols...
48562 This GDB supports auto-downloading debuginfo from the following URLs:
48563 <http://127.0.0.1:8000>
48565 Debuginfod has been disabled.
48566 To make this setting permanent, add 'set debuginfod enabled off' to .gdbinit.
48567 (No debugging symbols found in testsuite/outputs/gdb.debuginfod/fetch_src_and_symbols/fetch_src_and_symbols)
48571 I handled that correctly in the other tests that use test_urls, but
48572 had forgotten to update the "notice empty URL" one.
48574 Change-Id: I00040c83466e1494b3875574eb009c571a1504bf
48576 2022-04-21 Alan Modra <amodra@gmail.com>
48578 prune .note.GNU-stack warning from testsuite
48580 * testsuite/lib/binutils-common.exp (prune_warnings_extra): Remove
48581 .note.GNU-stack warning.
48582 (run_dump_test): Call prune_warnings for ld and objcopy output.
48584 * testsuite/ld-elf/elf.exp: Disable prune_warnings_extra temporarily
48585 around test for absent .note.GNU-stack
48586 * testsuite/ld-cris/globsymw2.s,
48587 * testsuite/ld-cris/warn3.d: Modify "is not implemented" message
48588 to avoid dejagnu prune_warnings.
48590 ld testsuite xcoff XPASS
48591 * testsuite/ld-scripts/defined5.d: Don't xfail xcoff targets.
48593 Delete unused COFF gas macro
48594 * config/obj-coff.h (sy_obj): Don't define.
48595 (OBJ_SYMFIELD_TYPE): Revise comments.
48597 2022-04-21 GDB Administrator <gdbadmin@sourceware.org>
48599 Automatic date update in version.in
48601 2022-04-20 Aaron Merey <amerey@redhat.com>
48603 gdb/debuginfod: Prevent out_of_range exception
48604 Trailing whitespace in the string of debuginfod URLs causes an
48605 out_of_range exception during the printing of URLs for the first
48608 To fix this, stop printing URLs when the substring to be printed
48609 consists only of whitespace.
48611 Also add first use notice testcases.
48613 Co-Authored-By: Pedro Alves <pedro@palves.net>
48615 2022-04-20 Lancelot SIX <lancelot.six@amd.com>
48617 gdb/testsuite: Introduce and use gdb_spawn_attach_cmdline
48618 Following a7e6a19e87f3d719ea23c65b580a6d9bca4ccab3 "gdb: testsuite: add
48619 new gdb_attach to check "attach" command", this commit proposes to
48620 introduce the gdb_spawn_attach_cmdline helper and use it in
48621 gdb.base/attach.exp.
48623 This helper starts GDB and adds the "--pid=$PID" argument.
48625 Also note that both the original and new implementation use
48626 gdb_spawn_with_cmdline_opts, which in the end uses default_gdb_spawn.
48627 This makes sure that we use $INTERNAL_GDBFLAGS, which by default already
48628 contain "-iex \"set height 0\" -iex \"set width 0\"". To avoid
48629 repetition of those arguments, gdb_spawn_attach_cmdline does not repeat
48632 To maintain a behavior similat to what gdb.base/attach.exp used to do,
48633 gdb_spawn_attach_cmdline keeps the -quiet flag.
48635 Tested on x86_64-gnu-linux
48637 Change-Id: I1fdcdb71c86d9c5d34bb28fc86fac68bcec37358
48639 2022-04-20 Tom Tromey <tom@tromey.com>
48641 Replace symbol_symtab with symbol::symtab
48642 This turns symbol_symtab into a method on symbol. It also replaces
48643 symbol_set_symtab with a method.
48645 Replace symbol_arch with symbol::arch
48646 This turns symbol_arch into a method on symbol.
48648 Replace symbol_objfile with symbol::objfile
48649 This turns symbol_objfile into a method on symbol.
48651 Remove symbol::aclass_index
48652 Symbols have an aclass_index method, but this isn't needed, because
48653 the aclass index isn't useful outside of the symbol implementation.
48655 Use array_view for symbol_impls
48656 It seemed to me that using array_view for symbol_impls would give a
48657 bit more error checking, at least when gdb is built in libstdc++ debug
48660 Add accessors for symbol's artificial field
48661 For a series I'm experimenting with, it was handy to hide a symbol's
48662 "artificial" field behind accessors. This patch is the result.
48664 Unify the DWARF index holders
48665 The dwarf2_per_bfd object has a separate field for each possible kind
48666 of index. Until an earlier patch in this series, two of these were
48667 even derived from a common base class, but still had separate slots.
48668 This patch unifies all the index fields using the common base class
48669 that was introduced earlier in this series. This makes it more
48670 obvious that only a single index can be active at a time, and also
48671 removes some code from dwarf2_initialize_objfile.
48673 Add an ad hoc version check to dwarf_scanner_base
48674 Some generic code in the DWARF reader has a special case for older
48675 versions of .gdb_index. This patch adds an ad hoc version check
48676 method so that these spots can work without specific knowledge of
48677 which index is in use.
48679 Simplify version check in dw2_symtab_iter_next
48680 This simplifies the index versio check in dw2_symtab_iter_next, by
48681 passing a reference to the index object to this function. This avoids
48682 an indirection via the per_bfd object.
48684 Introduce and use dwarf_scanner_base
48685 This introduces dwarf_scanner_base, a base class for all the index
48686 readers in the DWARF code. Then, it changes both mapped_index_base
48687 and cooked_index_vector to derive from this new base class.
48689 Introduce readnow_functions
48690 This introduces readnow_functions, a new subclass of
48691 dwarf2_base_index_functions, and changes the DWARF reader to use it.
48692 This lets us drop the "index is NULL" hack from the gdb index code.
48694 Remove some "OBJF_READNOW" code from dwarf2_debug_names_index
48695 The dwarf2_debug_names_index code treats a NULL debug_names_table as
48696 if it were from OBJF_READNOW. However, this trick is only done for
48697 gdb_index, never for debug_names -- see dwarf2_initialize_objfile.
48699 Let mapped index classes create the quick_symbol_functions object
48700 This changes the mapped index classes to create the
48701 quick_symbol_functions objects. This is a step toward having a more
48702 abstract interface to mapped indices.
48704 Give mapped_index_base a virtual destructor
48705 This changes mapped_index_base to have a virtual destructor, so it can
48706 be destroyed via its base class.
48708 Move mapped_index_base to new header file
48709 This moves mapped_index_base and the helper struct name_component to a
48710 new header file in gdb/dwarf2/.
48712 2022-04-20 Jan Beulich <jbeulich@suse.com>
48714 x86: reject all invalid SAE variants
48715 So far an SAE-only specifier was accepted for static-rounding insns,
48716 while SAE-only insns didn't accept static rounding specifiers. If
48717 anything it would make sense the other way around, allowing SAE-only
48718 insns to have the (ignored) rounding mode specified individually rather
48719 than globally via -mevexrcig=. But for now make things match the SDM.
48721 2022-04-20 Alan Modra <amodra@gmail.com>
48723 Re: xcoff: implement linker relaxation
48724 * xcofflink.c (xcoff_stub_csect_name): Increase buffer size.
48725 (xcoff_stub_get_csect_in_range, xcoff_build_one_stub): Whitespace.
48726 (bfd_xcoff_size_stubs): Cast PRIx64 arg to required type.
48727 Don't use freed stub_name.
48729 Revert "as: Reject unknown -gXXX option" testsuite
48730 This reverts the test committed as part of 6ea673e2d6.
48732 2022-04-20 Cl?ment Chigot <clement.chigot@atos.net>
48734 xcoff: implement linker relaxation
48737 * coff-rs6000.c (xcoff_reloc_type_noop): Add info argument.
48738 (xcoff_reloc_type_fail): Likewise.
48739 (xcoff_reloc_type_pos): Likewise.
48740 (xcoff_reloc_type_neg): Likewise.
48741 (xcoff_reloc_type_rel): Likewise.
48742 (xcoff_reloc_type_toc): Likewise.
48743 (xcoff_reloc_type_ba): Likewise.
48744 (xcoff_reloc_type_crel): Likewise.
48745 (xcoff_reloc_type_tls): Likewise.
48746 (xcoff_reloc_type_br): Add stub handler.
48747 (xcoff_ppc_relocate_section): Add info to
48748 xcoff_calculate_relocation.
48749 (xcoff_stub_indirect_call_code): New constant.
48750 (xcoff_stub_shared_call_code): Likewise.
48751 (bfd_xcoff_backend_data): Add stub code fields.
48752 (bfd_pmac_xcoff_backend_data): Likewise.
48753 * coff64-rs6000.c (xcoff64_reloc_type_br): Add stub handler.
48754 (xcoff64_ppc_relocate_section): Add info to
48755 xcoff64_calculate_relocation.
48756 (xcoff64_stub_indirect_call_code): New constant.
48757 (xcoff64_stub_shared_call_code): Likewise.
48758 (bfd_xcoff_backend_data): Add stub code fields.
48759 (bfd_xcoff_aix5_backend_data): Likewise.
48760 * libxcoff.h (struct xcoff_backend_data_rec): Add stub fields.
48761 (bfd_xcoff_stub_indirect_call_code): New define.
48762 (bfd_xcoff_stub_indirect_call_size): New define.
48763 (bfd_xcoff_stub_shared_call_code): New define.
48764 (bfd_xcoff_stub_shared_call_size): New define.
48765 (xcoff_reloc_function): Add info argument.
48766 (enum xcoff_stub_type): New enum.
48767 (struct xcoff_stub_hash_entry): New structure.
48768 * xcofflink.c (struct xcoff_link_hash_table): Add stub hash
48769 table and params fields.
48770 (xcoff_stub_hash_entry): New define.
48771 (xcoff_stub_hash_lookup): New define.
48772 (stub_hash_newfunc): New function.
48773 (_bfd_xcoff_bfd_link_hash_table_free): Free the new stub hash
48775 (_bfd_xcoff_bfd_link_hash_table_create): Create the new stub
48777 (xcoff_link_add_symbols): Save rawsize for XTY_SD.
48778 (bfd_xcoff_link_init): New function.
48779 (xcoff_stub_csect_name): New function.
48780 (xcoff_stub_get_csect_in_range): New function.
48781 (xcoff_stub_name): New function.
48782 (bfd_xcoff_get_stub_entry): New function.
48783 (bfd_xcoff_type_of_stub): New function.
48784 (xcoff_add_stub): New function.
48785 (xcoff_build_one_stub): New function.
48786 (bfd_xcoff_size_stubs): New function.
48787 (bfd_xcoff_build_stubs): New function.
48788 (xcoff_stub_create_relocations): New function.
48789 (xcoff_link_input_bfd): Adapt relocations to stub.
48790 (xcoff_write_global_symbol): Adapt to new TOC entries generated
48792 (_bfd_xcoff_bfd_final_link): Handle stub file.
48793 * xcofflink.h (struct bfd_xcoff_link_params): New structure.
48797 * emultempl/aix.em (params): New variable.
48798 (stub_file): New variable.
48799 (xcoff_add_stub_section): New function.
48800 (xcoff_layout_sections_again): New function
48801 (hook_in_stub): New function.
48802 (_after_allocation): Add stub creation.
48803 (_create_output_section_statements): Allocate stub file and
48804 pass params to backend.
48806 2022-04-20 Cl?ment Chigot <clement.chigot@atos.net>
48808 Stubs (added in a later patch) will generate new .loader symbols, once the allocations have been done. Thus, the .loader section cannot be layout before that.
48811 * coff-rs6000.c (_bfd_xcoff_put_ldsymbol_name): Write len in
48812 ldinfo->strings instead of directly in the output_bfd.
48813 * coff64-rs6000.c (_bfd_xcoff64_put_ldsymbol_name): Likewise.
48814 * xcofflink.c (struct xcoff_link_hash_table): Remove ldrel_count
48815 field. Add ldinfo field.
48816 (xcoff_mark_symbol): Adjust to new ldinfo field.
48817 (xcoff_mark): Likewise.
48818 (bfd_xcoff_link_count_reloc): Likewise.
48819 (xcoff_build_loader_section): Split into two functions: one that
48820 build the loader section (this function) and one that only size
48822 (xcoff_size_loader_section): ... (this function).
48823 (bfd_xcoff_size_dynamic_sections): Adapt to new ldinfo field.
48824 Move the part where the dynamic sections are build to ...
48825 (bfd_xcoff_build_dynamic_sections): ... this function.
48826 * xcofflink.h: Add bfd_xcoff_build_dynamic_sections prototype.
48830 * coff/xcoff.h (struct xcoff_loader_info): Add ldrel_count and
48835 * emultempl/aix.em (_after_allocation): New function.
48837 2022-04-20 Tom Tromey <tom@tromey.com>
48839 Use symbol_symtab accessor in compile-object-load.c
48840 I noticed that compile-object-load.c directly references owner.symtab
48841 of a symbol. However, I think it's better for all users to call
48842 symbol_symtab. This patch makes this change.
48844 2022-04-20 Nick Clifton <nickc@redhat.com>
48846 Add linker warning for when it creates an executable stack.
48849 2022-04-20 Tom Tromey <tom@tromey.com>
48851 Micro-optimize cooked_index_entry::full_name
48852 I noticed that cooked_index_entry::full_name can return the canonical
48853 string when there is no parent entry.
48855 Regression tested on x86-64 Fedora 34.
48857 2022-04-20 Tiezhu Yang <yangtiezhu@loongson.cn>
48859 gdb: LoongArch: Implement loongarch_scan_prologue()
48860 If can't determine prologue from the symbol table, need to examine
48861 instructions. Implement loongarch_scan_prologue() to analyze the
48862 function prologue from START_PC to LIMIT_PC, return the address of
48863 the first instruction past the prologue.
48865 2022-04-20 GDB Administrator <gdbadmin@sourceware.org>
48867 Automatic date update in version.in
48869 2022-04-19 H.J. Lu <hjl.tools@gmail.com>
48871 as: Reject unknown -gXXX option
48872 * as.c (parse_args): Reject unknown -gXXX option.
48873 * testsuite/gas/all/empty.s: New file.
48874 * testsuite/gas/all/pr29067.d: Likewise.
48875 * testsuite/gas/all/pr29067.err: Likewise.
48876 * testsuite/gas/all/gas.exp: Run pr29067.
48878 2022-04-19 Lancelot SIX <lancelot.six@amd.com>
48880 gdb/selftest-arch: Make register_test_foreach_arch generate arch tests lazily
48881 The register_test_foreach_arch is used to instantiate a given selftest
48882 for all architectures supported by GDB. It is used in many _initialize_*
48883 functions (under initialize_all_files, called by gdb_init).
48885 Because the call is done during GDB's initialization, and because there
48886 is no guaranty about the order in which all the _initialize_* functions
48887 are executed, when register_test_foreach_arch is called, GDB is not
48888 fully initialized. Specifically, when a particular initialize function
48889 is executed, only the architectures registered at that point are listed
48890 by gdbarch_printable_names.
48892 As a consequence, the list of selftest effectively executed depends on
48893 the order the _initialize_* functions are called. This can be observed
48894 with the following:
48897 -data-directory ./gdb/data-directory \
48898 -quiet -batch -ex "maint selftest" 2>&1 \
48899 | grep -E "Ran [0-9]+ unit tests"
48900 Ran 145 unit tests, 0 failed
48901 $ GDB_REVERSE_INIT_FUNCTIONS=1 ./gdb/gdb \
48902 -data-directory ./gdb/data-directory \
48903 -quiet -batch -ex "maint selftest" 2>&1 \
48904 | grep -E "Ran [0-9]+ unit tests"
48905 Ran 82 unit tests, 0 failed
48907 To fix this, make register_test_foreach_arch register a lazy selftest
48908 generator. This way when the test generator is eventually executed, all
48909 architectures are registered and we do not have a dependency on the
48910 order the initialize functions are executed in.
48912 Tested on x86_64-linux
48914 Change-Id: I88eefebf7d372ad672f42d3a103e89354bc8a925
48916 2022-04-19 Lancelot SIX <lancelot.six@amd.com>
48918 gdbsupport/selftest: Allow lazy registration
48919 This patch adds a way to delay the registration of tests until the
48920 latest possible moment. This is intended for situations where GDB needs
48921 to be fully initialized in order to decide if a particular selftest can
48922 be executed or not.
48924 This mechanism will be used in the next patch.
48926 Change-Id: I7f6b061f4c0a6832226c7080ab4e3a2523e1b0b0
48928 2022-04-19 Lancelot SIX <lancelot.six@amd.com>
48930 gdbsupport/selftest: Replace for_each_selftest with an iterator_range
48931 Remove the callback-based selftests::for_each_selftest function and use
48932 an iterator_range instead.
48934 Also use this iterator range in run_tests so all iterations over the
48935 selftests are done in a consistent way. This will become useful in a
48938 Change-Id: I0b3a5349a7987fbcb0071f11c394e353df986583
48940 2022-04-19 Jan Beulich <jbeulich@suse.com>
48942 x86: don't mistake ordinary immediates for SAE / rounding control
48943 The way SAE templates are constructed was always puzzling me (including
48944 the need for separate templates in the first place), and expressing the
48945 extzra attribute via Imm8 actually has a bad effect: Ordinary immediates
48946 would also be accepted, leading to an extra byte being added after the
48947 instruction (i.e. generating bad code). Before re-working this (in
48948 particular to accept proper Intel syntax there), fix the immediate issue
48949 by adding the so far missing check.
48951 x86: VCMPSH is Evex.LLIG
48952 These were mistakenly flagged as Evex.128. Getting the LLIG status right
48953 for insns allowing for SAE is a prereq for planned further work.
48955 x86: drop stray CheckRegSize from VFPCLASSPH
48956 Like VFPCLASSP{S,D} it has only a single operand allowing multiple
48957 sizes, hence there are no pairs of operands to check for consistent
48960 x86/Intel: test non-legacy VCVT{,U}SI2SH insn forms
48961 For an unclear reason corresponding AVX512F tests were apparently not
48962 cloned or used as reference here, and instead the bogus legacy forms of
48963 the insns (with the embedded rounding specifier not last) were used.
48965 2022-04-19 Jan Beulich <jbeulich@suse.com>
48967 x86: correct and simplify NOP disassembly
48968 It's not just REX.W which is ignored with opcode 0x90. The same goes for
48969 REX.R and REX.X as well as empty REX. None of these are forms of
48970 "xchg %eax,%eax" (which would mean zero-extending %eax to %rax), so they
48971 also shouldn't be disassembled this way.
48973 While there simplify things: A single hook function suffices, thus
48974 making it unnecessary to keep two expressions in sync. And checking
48975 ins->address_mode for mode_64bit also is unnecessary, as "rex" can be
48976 non-zero only in that case anyway.
48978 2022-04-19 GDB Administrator <gdbadmin@sourceware.org>
48980 Automatic date update in version.in
48982 2022-04-18 Simon Marchi <simon.marchi@efficios.com>
48984 gdb/testsuite/dwarf: don't automatically add directory and file entry for DWARF 5
48985 To support DWARF 5 in the DWARF assembler line tables, we currently copy
48986 the first user-provided directory and the first user-provided files and
48987 make them elements at indices 0 in the directory and file name tables.
48988 That was a sufficient behavior at the time (see commit 44fda089397a
48989 ("[gdb/testsuite] Support .debug_line v5 in dwarf assembler")), but in
48990 the following patches, I would need to have finer grained control on
48991 what is generated exactly. For example, I'd like to generate a DWARF 5 line
48992 table with just a single file and a single directory.
48994 Get rid of this behavior, and implement what is suggested in
48995 44fda089397a: make include_dir return the directory index that can be
48996 used to refer to that directory entry (based on the DWARF version), and
48999 Adjust dw2-lines.exp and dw2-prologue-end.exp accordingly. Their produced
49000 DWARF5 binaries will change a bit, in that they will now have a single
49001 directory and file, where they had two before. But it doesn't change
49002 the expected GDB behavior.
49004 Change-Id: I5459b16ac9b7f28c34c9693c35c9afd2ebb3aa3b
49006 2022-04-18 Simon Marchi <simon.marchi@efficios.com>
49008 gdb: use gdb_tilde_expand instead of gdb_tilde_expand_up in source_script_with_search
49009 Since this is the latest use of gdb_tilde_expand_up, remove it.
49011 Change-Id: I964c812ce55fe087876abf91e7a3577ad79c0425
49013 2022-04-18 Simon Marchi <simon.marchi@efficios.com>
49015 gdbsupport: make gdb_realpath_keepfile return an std::string
49016 I'm trying to switch these functions to use std::string instead of char
49017 arrays, as much as possible. Some callers benefit from it (can avoid
49018 doing a copy of the result), while others suffer (have to make one more
49021 Change-Id: I793aab17baaef8345488f4c40b9094e2695425bc
49023 2022-04-18 Simon Marchi <simon.marchi@efficios.com>
49025 gdbsupport: make gdb_abspath return an std::string
49026 I'm trying to switch these functions to use std::string instead of char
49027 arrays, as much as possible. Some callers benefit from it (can avoid
49028 doing a copy of the result), while others suffer (have to make one more
49031 Change-Id: Iced49b8ee2f189744c5072a3b217aab5af17a993
49033 2022-04-18 Simon Marchi <simon.marchi@polymtl.ca>
49035 gdb: call gdb_tilde_expand instead of gdb_tilde_expand_up in source_script_with_search
49036 This removes a use of gdb_tilde_expand_up, which is removed later in
49039 Change-Id: I5887d526cea987103e4ca24514a982b0a28e992a
49041 2022-04-18 Tom Tromey <tromey@adacore.com>
49044 This updates gnulib to a relatively recent commit. Most of this was
49045 done by the gnulib import script; the only change I made was to
49048 Tested on x86-64 Fedora 34. I also did a mingw cross build.
49050 2022-04-18 Tom Tromey <tom@tromey.com>
49052 Fix C++ cast of derived class to base class
49053 PR c++/28907 points out that casting from a derived class to a base
49054 class fails in some situations. The problem turned out to be a
49055 missing use of value_embedded_offset. One peculiarity here is that,
49056 if you managed to construct a pointer-to-derived with an embedded
49057 offset of 0, the cast would work -- for example, one of the two new
49058 tests here passes without the patch.
49060 This embedded offset stuff is an endless source of bugs. I wonder if
49061 it's possible to get rid of it somehow.
49063 Regression tested on x86-64 Fedora 34.
49065 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28907
49067 2022-04-18 Simon Marchi <simon.marchi@polymtl.ca>
49069 gdb/testsuite: make gdb.ada/mi_prot.exp stop at expected location
49070 This test attempts to run until the line marked "STOP", which is at
49071 prot.adb:34. It first runs until the "main" symbol, then tries to place
49072 a breakpoint by line at line 34, without specifying the source file. When looking at the logs:
49074 -break-insert -t 34^M
49075 ^done,bkpt={number="2",type="breakpoint",disp="del",enabled="y",addr="0x0000555555558a6c",func="adafinal",file="/home/simark/build/binutils-gdb-one-target/gdb/testsuite/outputs/gdb.ada/mi_pro t/b~prot.adb",fullname="/home/simark/build/binutils-gdb-one-target/gdb/testsuite/outputs/gdb.ada/mi_prot/b~prot.adb",line="44",thread-groups=["i1"],times="0",original-location="/home/simark/b uild/binutils-gdb-one-target/gdb/testsuite/outputs/gdb.ada/mi_prot/b~prot.adb:34"}^M
49077 *stopped,reason="breakpoint-hit",disp="del",bkptno="2",frame={addr="0x0000555555558a6c",func="adafinal",args=[],file="/home/simark/build/binutils-gdb-one-target/gdb/testsuite/outputs/gdb.ada/ mi_prot/b~prot.adb",fullname="/home/simark/build/binutils-gdb-one-target/gdb/testsuite/outputs/gdb.ada/mi_prot/b~prot.adb",line="44",arch="i386:x86-64"},thread-id="1",stopped-threads="all",co re="8"^M
49079 ... we see that the breakpoint is placed in some generated file, not in
49080 the test source file as we expect. The problem is that "b main" in Ada
49081 does not place a breakpoint on the "Ada main", but on some symbol in a
49082 generated source file. So when stopped at the "main" symbol, we are not
49083 stopped in the file that contains the STOP marker at line 34.
49085 The test passes anyway today, so it doesn't seem to matter that we are
49086 stopped at an unexpected location. But it starts failing with this
49087 patch [1], because b~prot.adb:34 happens to be between two functions, so
49088 the breakpoint doesn't resolve.
49090 Fix this by placing the breakpoint at "$srcfile:$line", which works
49091 regardless of what is the current source file.
49093 However, this ends up introducing a path in the test name. Modify
49094 mi_tbreak and mi_continue_to_line to avoid that.
49096 [1] https://sourceware.org/pipermail/gdb-patches/2022-April/187686.html
49098 Change-Id: I742e2a9993046dcb5e30c64fe2ad920a363baf75
49100 2022-04-18 Vignesh Balasubramanian <Vignesh.Balasubrmanian@amd.com>
49102 gdb/testsuite: add text_segment option to gdb_compile
49103 LLVM's lld linker doesn't have the "-Ttext-segment" option, but
49104 "--image-base" can be used instead.
49106 To centralize the logic of checking which option is supported, add the
49107 text_segment option to gdb_compile. Change tests that are currently
49108 using -Ttext-segment to use that new option instead.
49110 This patch fixes only compilation error, for example:
49114 $ make check TESTS="gdb.base/jit-elf.exp" RUNTESTFLAGS="CC_FOR_TARGET=clang LDFLAGS_FOR_TARGET=-fuse-ld=ld"
49115 Running /home/simark/src/binutils-gdb/gdb/testsuite/gdb.base/jit-elf.exp ...
49116 gdb compile failed, clang-13: warning: -Xlinker -Ttext-segment=0x7000000: 'linker' input unused [-Wunused-command-line-argument]
49120 $ make check TESTS="gdb.base/jit-elf.exp" RUNTESTFLAGS="CC_FOR_TARGET=clang LDFLAGS_FOR_TARGET=-fuse-ld=ld"
49121 Running /home/simark/src/binutils-gdb/gdb/testsuite/gdb.base/jit-elf.exp ...
49122 FAIL: gdb.base/jit-elf.exp: one_jit_test-1: continue to breakpoint: break here 1
49123 FAIL: gdb.base/jit-elf.exp: one_jit_test-1: continue to breakpoint: break here 2
49124 FAIL: gdb.base/jit-elf.exp: one_jit_test-2: continue to breakpoint: break here 1
49125 FAIL: gdb.base/jit-elf.exp: one_jit_test-2: info function ^jit_function
49126 FAIL: gdb.base/jit-elf.exp: one_jit_test-2: continue to breakpoint: break here 2
49127 FAIL: gdb.base/jit-elf.exp: attach: one_jit_test-2: continue to breakpoint: break here 1
49128 FAIL: gdb.base/jit-elf.exp: attach: one_jit_test-2: break here 1: attach
49129 FAIL: gdb.base/jit-elf.exp: PIE: one_jit_test-1: continue to breakpoint: break here 1
49130 FAIL: gdb.base/jit-elf.exp: PIE: one_jit_test-1: continue to breakpoint: break here 2
49132 === gdb Summary ===
49134 # of expected passes 26
49135 # of unexpected failures 9
49137 Change-Id: I3678c5c9bbfc2f80671698e28a038e6b3d14e635
49139 2022-04-18 Enze Li <lienze2010@hotmail.com>
49141 gdb: fix using clear command to delete non-user breakpoints(PR cli/7161)
49142 The clear command shouldn't delete momentary and internal breakpoints,
49143 nor internal breakpoints created via Python's gdb.Breakpoint.
49145 This patch fixes this issue and adds a testcase.
49147 Regression tested on x86_64 openSUSE Tumbleweed(VERSION_ID="20220413").
49149 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=7161
49151 2022-04-18 GDB Administrator <gdbadmin@sourceware.org>
49153 Automatic date update in version.in
49155 2022-04-17 GDB Administrator <gdbadmin@sourceware.org>
49157 Automatic date update in version.in
49159 2022-04-16 Tom Tromey <tom@tromey.com>
49161 Add comments to dwarf2/abbrev-cache.h
49162 This patch started when I noticed that the unordered_set include
49163 wasn't needed in abbrev-cache.h. (That was probably leftover from
49164 some earlier implementation of the class.) Then, I noticed that the
49165 class itself was under-commented. This patch fixes both issues.
49167 2022-04-16 GDB Administrator <gdbadmin@sourceware.org>
49169 Automatic date update in version.in
49171 2022-04-15 Tom Tromey <tom@tromey.com>
49173 Return void from gdb_putc
49174 I don't think it's very useful to return the character from gdb_putc,
49175 so this patch changes it to return void.
49177 2022-04-15 Tom Tromey <tom@tromey.com>
49179 Handle "set height 1"
49180 PR cli/17151 points out that "set height 1" has pathological behavior
49181 in gdb. What I see is that gdb will endlessly print the pagination
49182 prompt. This patch takes a simple and expedient approach to a fix:
49183 pretend that the height is really 2.
49185 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=17151
49187 2022-04-15 Tom Tromey <tom@tromey.com>
49189 Allow word wrapping even when paging is disabled
49190 PR cli/20741 points out that when pagination is disabled, this also
49191 disabled word wrapping. However, the manual documents that these
49192 settings are separate -- if you intend to disable the wrapping, you
49193 must use "set width unlimited".
49195 This patch fixes the bug by letting the pagination-disabled case fall
49196 through to the code that also handles word-wrapping.
49198 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=20741
49200 2022-04-15 Tom Tromey <tom@tromey.com>
49202 Implement value_print for Rust
49203 This adds an implementation of the value_print method to Rust. As
49204 described in PR rust/22254, this removes a bit of weird-looking output
49205 from some "print"s -- because c_value_print is bypassed. I don't have
49206 a test for the bug that inspired this patch, because I only know how
49207 to reproduce it when using a relatively old Rust compiler. However,
49208 the new "cast-printing" code in value_print is required, because
49209 omitting this causes some existing tests to fail.
49211 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=22254
49213 2022-04-15 Tom Tromey <tom@tromey.com>
49215 Reimplement Rust slice printing
49216 The current nightly Rust compiler (aka 1.61) added better DWARF
49217 representation for unsized types. Fixing this is PR rust/21466; but
49218 the code is actually the same as what is required to make slice
49219 printing more useful, which is PR rust/23871. This patch implements
49220 this. I tested this against various Rust compilers: 1.48, current
49221 stable, current beta, and current nightly.
49223 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=21466
49224 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=23871
49226 2022-04-15 Tom Tromey <tom@tromey.com>
49228 Remove some dead code from the Rust value printer
49229 This removes a bit of dead code from the Rust value printer. This
49230 code wasn't always dead -- it fixed a real bug, and a test case was
49231 added for it. However, once val_print was removed, it became
49234 2022-04-15 Tom Tromey <tom@tromey.com>
49236 Match rustc beta versions
49237 The rust_compiler_version proc extracts the Rust compiler version from
49238 the "rustc --version" output. For a beta compiler, the output looks
49241 rustc 1.60.0-beta.6 (7bccde197 2022-03-22)
49243 This patch slightly relaxes the regexp -- removing a space -- so that
49244 this can be understood by this proc.
49246 2022-04-15 Tom de Vries <tdevries@suse.de>
49248 [gdb/testsuite] Fix gdb.ada/float-bits.exp with -m32
49249 With test-case gdb.ada/float-bits.exp and native we get:
49251 (gdb) print 16llf#7FFFF7FF4054A56FA5B99019A5C8#^M
49253 (gdb) PASS: gdb.ada/float-bits.exp: print 16llf#7FFFF7FF4054A56FA5B99019A5C8#
49255 but with target board unix/-m32 we have instead:
49257 (gdb) print 16llf#7FFFF7FF4054A56FA5B99019A5C8#^M
49258 Cannot export value 2596145952482202326224873165792712 as 96-bits \
49259 unsigned integer (must be between 0 and 79228162514264337593543950335)^M
49260 (gdb) FAIL: gdb.ada/float-bits.exp: print 16llf#7FFFF7FF4054A56FA5B99019A5C8#
49263 Fix this by testing whether 16llf is supported by doing ptype long_long_float
49264 which gets us either:
49266 type = <16-byte float>^M
49270 type = <12-byte float>^M
49273 Tested on x86_64-linux with native and unix/-m32.
49275 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29041
49277 2022-04-15 Tom Tromey <tom@tromey.com>
49279 Remove WITH_SIM define
49280 Since score-tdep.c was removed, the WITH_SIM define is not used in
49281 gdb. This patch removes it.
49283 Note that re-running autoheader shows a separate change that was
49284 missed. I've kept it in this patch to avoid extra work.
49286 2022-04-15 Tom de Vries <tdevries@suse.de>
49288 [gdb/testsuite] Fix gdb.go/methods.exp with check-readmore
49289 When running test-case gdb.go/methods.exp with make check we have:
49291 (gdb) break main.T.Foo^M
49292 Function "main.T.Foo" not defined.^M
49293 Make breakpoint pending on future shared library load? (y or [n]) n^M
49294 (gdb) XFAIL: gdb.go/methods.exp: gdb_breakpoint: set breakpoint at main.T.Foo
49296 but with make check-readmore the XFAIL fails to trigger:
49298 (gdb) break main.T.Foo^M
49299 Function "main.T.Foo" not defined.^M
49300 Make breakpoint pending on future shared library load? (y or [n]) n^M
49301 (gdb) FAIL: gdb.go/methods.exp: gdb_breakpoint: set breakpoint at main.T.Foo
49304 This happens because this gdb_test_multiple "maintenance print symbols"
49307 -re "\r\n$gdb_prompt $" {
49309 matches the entire command output.
49311 Fix this by adding the missing ^ at the regexp start.
49313 Tested on x86_64-linux.
49315 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29064
49317 2022-04-15 GDB Administrator <gdbadmin@sourceware.org>
49319 Automatic date update in version.in
49321 2022-04-14 Pedro Alves <pedro@palves.net>
49323 gdbserver: Eliminate prepare_to_access_memory
49326 - The prepare_to_access_memory machinery was added for non-stop mode.
49328 - Only Linux supports non-stop.
49330 - Linux no longer needs the prepare_to_access_memory machinery. In
49331 fact, after the previous patch,
49332 linux_process_target::prepare_to_access_memory became a nop.
49334 Thus, prepare_to_access_memory can go away, simplifying core GDBserver
49337 Change-Id: I93ac8bfe66bd61c3d1c4a0e7d419335163120ecf
49339 2022-04-14 Pedro Alves <pedro@palves.net>
49341 gdbserver/linux: Access memory even if threads are running
49342 Similarly to how the native Linux target was changed
49343 and subsequently reworked in these commits:
49345 05c06f318fd9 Linux: Access memory even if threads are running
49346 8a89ddbda2ec Avoid /proc/pid/mem races (PR 28065)
49348 ... teach GDBserver to access memory even when the current thread is
49349 running, by always accessing memory via /proc/PID/mem.
49351 The existing comment:
49353 /* Neither ptrace nor /proc/PID/mem allow accessing memory through a
49356 ... is incorrect for /proc/PID/mem does allow that.
49358 Actually, from GDB's perspective, GDBserver could already access
49359 memory while threads were running, but at the expense of pausing all
49360 threads for the duration of the memory access, via
49361 prepare_to_access_memory. This new implementation does not require
49362 pausing any thread, thus
49363 linux_process_target::prepare_to_access_memory /
49364 linux_process_target::done_accessing_memory become nops. A subsequent
49365 patch will remove the whole prepare_to_access_memory infrastructure
49368 The GDBserver linux-low.cc implementation is simpler than GDB's
49369 linux-nat.c's, because GDBserver always adds the unfollowed vfork/fork
49370 children to the process list immediately when the fork/vfork event is
49371 seen out of ptrace. I.e., there's no need to keep the file descriptor
49372 stored on a side map, we can store it directly in the process
49375 Change-Id: I0abfd782ceaa4ddce8d3e5f3e2dfc5928862ef61
49377 2022-04-14 Pedro Alves <pedro@palves.net>
49379 gdbserver: special case target_write_memory len==0
49380 The next patch in this series adds a common helper routine for both
49381 memory reads and writes, like this:
49384 proc_xfer_memory (CORE_ADDR memaddr, unsigned char *readbuf,
49385 const gdb_byte *writebuf, int len)
49387 gdb_assert ((readbuf == nullptr) != (writebuf == nullptr));
49392 linux_process_target::read_memory (CORE_ADDR memaddr,
49393 unsigned char *myaddr, int len)
49395 return proc_xfer_memory (memaddr, myaddr, nullptr, len);
49398 linux_process_target::write_memory (CORE_ADDR memaddr,
49399 const unsigned char *myaddr, int len)
49401 return proc_xfer_memory (memaddr, nullptr, myaddr, len);
49404 Surprisingly, the assertion fails. That happens because it can happen
49405 that target_write_memory is called with LEN==0, due to this in
49408 /* Determine whether the remote target supports binary downloading.
49409 This is accomplished by sending a no-op memory write of zero length
49410 to the target at the specified address. (...) */
49413 remote_target::check_binary_download (CORE_ADDR addr)
49416 p = rs->buf.data ();
49418 p += hexnumstr (p, (ULONGEST) addr);
49420 p += hexnumstr (p, (ULONGEST) 0);
49424 In this scenario, in gdbserver's target_write_memory, the "myaddr"
49425 argument of the_target->write_memory is passed the data() of a local
49426 gdb::byte_vector (which is a specialized std::vector). It's valid for
49427 std::vector::data() to return NULL when the vector is empty.
49429 This commit adds an early return to target_write_memory to avoid
49430 target backends having to care about this. For good measure, do the
49431 same on the read side, in read_inferior_memory.
49433 Change-Id: Iac8f04fcf99014c624ef4036bd318ca1771ad491
49435 2022-04-14 Pedro Alves <pedro@palves.net>
49437 gdbserver/qXfer::threads, prepare_to_access_memory=>target_pause_all
49438 handle_qxfer_threads_proper needs to pause all threads even if the
49439 target can read memory when threads are running, so use
49440 target_pause_all instead, which is what the Linux implementation of
49441 prepare_to_access_memory uses. (Only Linux implements this hook.)
49443 A following patch will make the Linux backend be able to access memory
49444 when threads are running, and thus will also make
49445 prepare_to_access_memory do nothing, which would cause testsuite
49446 regressions without this change.
49448 Change-Id: I127fec7246b7c45b60dfa7341e781606bf54b5da
49450 2022-04-14 Tom Tromey <tromey@adacore.com>
49452 Ignore 0,0 entries in .debug_aranges
49453 When running the internal AdaCore test suite against the new DWARF
49454 indexer, I found one regression on RISC-V. The test in question uses
49455 --gc-sections, and winds up with an entry in the middle of a
49456 .debug_aranges that has both address and length of 0. In this
49457 scenario, gdb assumes the entries are terminated and then proceeds to
49458 reject the section because it reads a subsequent entry as if it were a
49461 It seems to me that, because each header describes the size of each
49462 .debug_aranges CU, it's better to simply ignore 0,0 entries and simply
49463 read to the end. That is what this patch does.
49465 I've patched an existing test to provide a regression test for this.
49467 2022-04-14 Tom Tromey <tromey@adacore.com>
49469 Use GetThreadDescription on Windows
49470 Windows 10 introduced SetThreadDescription and GetThreadDescription, a
49471 simpler way to set a thread's name. This changes gdb and gdbserver to
49472 use this convention when it is available.
49474 This is part of PR win32/29050.
49476 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29050
49478 2022-04-14 Tom Tromey <tromey@adacore.com>
49480 Set the worker thread name on Windows
49481 This patch is a bit different from the rest of the series, in that it
49482 is a change to gdb's behavior on the host. It changes gdb's thread
49483 pool to try to set the thread name on Windows, if SetThreadDescription
49486 This is part of PR win32/29050.
49488 This patch isn't likely to be useful to many people in the short term,
49489 because the Windows port of the libstdc++ thread code is not upstream.
49490 (AdaCore uses it, and sent it upstream, but it did not land, I don't
49491 know why.) However, if that patch does ever go in, or presumably if
49492 you build using some other C++ runtime library, then this will be
49495 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29050
49497 2022-04-14 Tom Tromey <tromey@adacore.com>
49499 Implement thread_name for gdbserver
49500 This changes gdbserver to implement thread_name method.
49502 Share handle_ms_vc_exception with gdbserver
49503 Currently, gdb's native Windows target implements the exception-based
49504 approach for setting thread names, but gdbserver does not. This patch
49505 moves handle_ms_vc_exception to the shared nat/windows-nat.c code, as
49506 preparation for adding this support to gdbserver.
49508 Move target_read_string to target/target.c
49509 This moves the two overloads of target_read_string to a new file,
49510 target/target.c, and updates both gdb and gdbserver to build this.
49512 Remove the byte order parameter to target_read_string
49513 target_read_string takes a byte order parameter, but only uses this to
49514 check whether a given character is zero. This is readily done without
49515 requiring the parameter, so remove it.
49518 This renames read_string to be an overload of target_read_string.
49519 This makes it more consistent for the eventual merger with gdbserver.
49521 Don't call QUIT in read_string
49522 read_string does not need to call QUIT, because target_read_memory
49523 already does. This change is needed to make string-reading usable by
49526 Fix possible Cygwin build problem
49527 I noticed that nat/windows-nat.c checks __USEWIDE, but nothing sets it
49528 there -- I forgot to copy over the definition when making this file.
49529 This patch tries to fix the problem. I don't have a Cygwin setup, so
49530 I don't know whether this is sufficient, but it's probably necessary.
49532 2022-04-14 Lancelot SIX <lancelot.six@amd.com>
49533 Pedro Alves <pedro@palves.net>
49535 gdb/testsuite: Fix race in gdb.dwarf2/calling-convention.exp
49536 Pedro Alves warned me that there is a race in
49537 gdb.dwarf2/calling-convention.exp making the test sometimes fail on his
49538 setup. This can be reliably reproduced using :
49540 make check-read1 TESTS="gdb.dwarf2/calling-convention.exp"
49542 The relevant part of the gdb.log file is:
49545 Function 'foo' does not follow the target calling convention.
49546 If you continue, setting the return value will probably lead to unpredictable behaviors.
49547 Make foo return now? (y or n) PASS: gdb.dwarf2/calling-convention.exp: return 35
49550 (gdb) FAIL: gdb.dwarf2/calling-convention.exp: finish
49552 The issue is that when doing the test for "return 35", the DejaGnu test
49553 sends "n" (to tell GDB not to perform the return action) but never
49554 consumes the "Not confirmed" acknowledgment sent by GDB. Later, when
49555 trying to do the next test, DejaGnu tries to match the leftover output
49556 from the "return" test. As this output is not expected, the test fails.
49558 Fix by using gdb_test to send the "n" answer and match the confirmation
49559 and consume all output to the prompt.
49561 Also do minor adjustments to the main regex:
49562 - Remove the leading ".*" which is not required.
49563 - Ensure that the "?" from the question is properly escaped.
49565 Tested on x86_64-gnu-linux, using
49567 - make check TESTS="gdb.dwarf2/calling-convention.exp"
49568 - make check-read1 TESTS="gdb.dwarf2/calling-convention.exp"
49569 - make check-readmore TESTS="gdb.dwarf2/calling-convention.exp"
49571 Change-Id: I42858b13db2cbd623c5c1739de65ad423e0c0938
49573 2022-04-14 Tom Tromey <tromey@adacore.com>
49575 Silence -Wmaybe-uninitialized warning from target_waitstatus
49576 Currently, one use of target_waitstatus yields a warning:
49578 target/waitstatus.h: In function 'void stop_all_threads()':
49579 target/waitstatus.h:175:13: warning: 'ws.target_waitstatus::m_value' may be used uninitialized in this function [-Wmaybe-uninitialized]
49580 175 | m_value = other.m_value;
49581 | ~~~~~~~~^~~~~~~~~~~~~~~
49583 This patch silences the warning. I tried the "volatile member"
49584 approach that was used for gdb::optional, but that didn't work, so
49585 this patch simply initializes the member.
49587 2022-04-14 Tom Tromey <tromey@adacore.com>
49589 Fix regression on Windows with WOW64
49590 Internally at AdaCore, we recently started testing a 64-bit gdb
49591 debugging 32-bit processes. This failed with gdb head, but not with
49594 The tests fail like this:
49596 Starting program: [...].exe
49597 warning: Could not load shared library symbols for WOW64_IMAGE_SECTION.
49598 Do you need "set solib-search-path" or "set sysroot"?
49599 warning: Could not load shared library symbols for WOW64_IMAGE_SECTION.
49600 Do you need "set solib-search-path" or "set sysroot"?
49601 warning: Could not load shared library symbols for NOT_AN_IMAGE.
49602 Do you need "set solib-search-path" or "set sysroot"?
49603 warning: Could not load shared library symbols for NOT_AN_IMAGE.
49604 Do you need "set solib-search-path" or "set sysroot"?
49606 After some debugging and bisecting, to my surprise the bug was
49607 introduced by commit 183be222 ("gdb, gdbserver: make target_waitstatus
49610 The problem occurs in handle_exception. Previously the code did:
49612 - ourstatus->kind = TARGET_WAITKIND_STOPPED;
49614 case EXCEPTION_BREAKPOINT:
49616 - ourstatus->kind = TARGET_WAITKIND_SPURIOUS;
49619 case STATUS_WX86_BREAKPOINT:
49620 DEBUG_EXCEPTION_SIMPLE ("EXCEPTION_BREAKPOINT");
49621 - ourstatus->value.sig = GDB_SIGNAL_TRAP;
49623 - last_sig = ourstatus->value.sig;
49625 However, in the new code, the fallthrough case does:
49627 + ourstatus->set_stopped (GDB_SIGNAL_TRAP);
49629 ... which changes the 'kind' in 'ourstatus' after falling through.
49631 This patch rearranges the 'last_sig' setting to more closely match
49632 what was done before (this is probably not strictly needed but also
49633 seemed harmless), and removes the fall-through in the
49634 'ignore_first_breakpoint' case when __x86_64__ is defined.
49636 2022-04-14 Tom Tromey <tromey@adacore.com>
49638 Reorganize Python events documentation
49639 This slightly reorganizes the Python events documentation. It hoists
49640 the "ThreadEvent" text out of the list of events, where it seemed to
49641 be misplaced. It tidies the formatting a little bit (adding some
49642 vertical space for easier reading in info), fixes a typo, adds some
49643 missing commas, and fixes an incorrect reference to NewInferiorEvent.
49645 2022-04-14 Simon Marchi <simon.marchi@polymtl.ca>
49647 gdb: remove move constructor and move assignment operator from cooked_index
49648 Building with clang++-14, I see:
49650 CXX dwarf2/cooked-index.o
49651 In file included from /home/smarchi/src/binutils-gdb/gdb/dwarf2/cooked-index.c:21:
49652 /home/smarchi/src/binutils-gdb/gdb/dwarf2/cooked-index.h:172:12: error: explicitly defaulted move constructor is implicitly deleted [-Werror,-Wdefaulted-function-deleted]
49653 explicit cooked_index (cooked_index &&other) = default;
49655 /home/smarchi/src/binutils-gdb/gdb/dwarf2/cooked-index.h:225:16: note: move constructor of 'cooked_index' is implicitly deleted because field 'm_storage' has a deleted move constructor
49656 auto_obstack m_storage;
49658 /home/smarchi/src/binutils-gdb/gdb/../gdbsupport/gdb_obstack.h:128:28: note: 'auto_obstack' has been explicitly marked deleted here
49659 DISABLE_COPY_AND_ASSIGN (auto_obstack);
49661 In file included from /home/smarchi/src/binutils-gdb/gdb/dwarf2/cooked-index.c:21:
49662 /home/smarchi/src/binutils-gdb/gdb/dwarf2/cooked-index.h:174:17: error: explicitly defaulted move assignment operator is implicitly deleted [-Werror,-Wdefaulted-function-deleted]
49663 cooked_index &operator= (cooked_index &&other) = default;
49665 /home/smarchi/src/binutils-gdb/gdb/dwarf2/cooked-index.h:225:16: note: move assignment operator of 'cooked_index' is implicitly deleted because field 'm_storage' has a deleted move assignment operator
49666 auto_obstack m_storage;
49668 /home/smarchi/src/binutils-gdb/gdb/../gdbsupport/gdb_obstack.h:128:3: note: 'operator=' has been explicitly marked deleted here
49669 DISABLE_COPY_AND_ASSIGN (auto_obstack);
49671 /home/smarchi/src/binutils-gdb/gdb/../include/ansidecl.h:425:8: note: expanded from macro 'DISABLE_COPY_AND_ASSIGN'
49672 void operator= (const TYPE &) = delete
49675 We explicitly make cooked_index have a default move constructor and
49676 move assignment operator. But it doesn't actually happen because
49677 cooked_index has a field of type auto_obstack, which isn't movable.
49678 We don't actually need cooked_index to be movable at the moment, so
49679 remove those lines.
49681 Change-Id: Ifc1fe3d7d67e3ae1a14363d6c1869936fe80b0a2
49683 2022-04-14 Tom Tromey <tromey@adacore.com>
49685 Let std::thread check pass even without pthreads
49686 Currently, the configure check for std::thread relies on pthreads
49687 existing. However, this means that if std::thread is implemented for
49688 a non-pthreads host, then the check will yield the wrong answer. This
49689 happened in AdaCore internal builds. Here, we have this GCC patch:
49691 https://gcc.gnu.org/legacy-ml/gcc-patches/2019-06/msg01840.html
49693 ... which adds mingw support to GCC's gthreads implementation, and
49694 also to std::thread.
49696 This configure change fixes this problem and enables threading for
49699 2022-04-14 Tiezhu Yang <yangtiezhu@loongson.cn>
49701 gdb: fix build errors in gdbsupport/thread-pool.h used with old gcc
49702 When I build gdb with gcc 8.3, there exist the following build errors,
49703 rename the typedef to task_t to fix them.
49706 In file included from /home/loongson/gdb.git/gdbsupport/thread-pool.cc:21:
49707 /home/loongson/gdb.git/gdbsupport/../gdbsupport/thread-pool.h: In member function ‘std::future<void> gdb::thread_pool::post_task(std::function<void()>&&)’:
49708 /home/loongson/gdb.git/gdbsupport/../gdbsupport/thread-pool.h:69:44: error: declaration of ‘task’ shadows a previous local [-Werror=shadow=local]
49709 std::packaged_task<void ()> task (std::move (func));
49711 /home/loongson/gdb.git/gdbsupport/../gdbsupport/thread-pool.h:102:39: note: shadowed declaration is here
49712 typedef std::packaged_task<void ()> task;
49714 /home/loongson/gdb.git/gdbsupport/../gdbsupport/thread-pool.h: In member function ‘std::future<_Res> gdb::thread_pool::post_task(std::function<T()>&&)’:
49715 /home/loongson/gdb.git/gdbsupport/../gdbsupport/thread-pool.h:80:41: error: declaration of ‘task’ shadows a previous local [-Werror=shadow=local]
49716 std::packaged_task<T ()> task (std::move (func));
49718 /home/loongson/gdb.git/gdbsupport/../gdbsupport/thread-pool.h:102:39: note: shadowed declaration is here
49719 typedef std::packaged_task<void ()> task;
49722 2022-04-14 Tom de Vries <tdevries@suse.de>
49724 [gdb/testsuite] Detect 'No MPX support'
49725 On openSUSE Leap 15.3, mpx support has been disabled for m32, so I run into:
49728 Starting program: outputs/gdb.arch/i386-mpx/i386-mpx ^M
49729 [Thread debugging using libthread_db enabled]^M
49730 Using host libthread_db library "/lib64/libthread_db.so.1".^M
49733 and eventually into all sort of fails in this and other mpx test-cases.
49735 Fix this by detecting the "No MPX support" message in have_mpx.
49737 Tested on x86_64-linux with target boards unix and unix/-m32.
49739 2022-04-14 Sergei Trofimovich <siarheit@google.com>
49741 M68K: avoid quadratic slowdlow in label alignment check
49742 Before the change tc-m68k maintained a list of seen labels.
49743 Alignment check traversed label list to resolve symbol to label.
49744 This caused quadratic slowdown as each symbol was checked against
49745 each label. Worst affected files are the ones built with debugging
49746 enabled as DWARF generates many labels.
49748 The change embeds auxiliary label information right into symbol using
49751 Before the change test from PR 29058 did not finish in 10 minutes. After
49752 the change it finishes in 2 seconds.
49757 * config/tc-m68k.h (TC_SYMFIELD_TYPE): define as m68k_tc_sy.
49758 * config/tc-m68k.c (m68k_frob_label): Use TC_SYMFIELD_TYPE to
49759 store label information.
49761 2022-04-14 caiyinyu <caiyinyu@loongson.cn>
49763 ld:LoongArch: Fix glibc fail: tst-audit25a/b.
49766 * elfnn-loongarch.c: Add new func elf_loongarch64_hash_symbol.
49768 2022-04-14 GDB Administrator <gdbadmin@sourceware.org>
49770 Automatic date update in version.in
49772 2022-04-13 Simon Marchi <simon.marchi@efficios.com>
49774 gdb: add ATTRIBUTE_PRINTF to complaint_interceptor::issue_complaint
49775 Fix this error when building with clang++-14:
49778 /home/smarchi/src/binutils-gdb/gdb/complaints.c:130:65: error: format string is not a string literal [-Werror,-Wformat-nonliteral]
49779 g_complaint_interceptor->m_complaints.insert (string_vprintf (fmt, args));
49782 Change-Id: I0ef11f970510eb8638d1651fa0d5eeecd6a9d31a
49784 2022-04-13 Simon Marchi <simon.marchi@efficios.com>
49786 gdb: fix clang build failure in msymbol_is_mips
49787 Building with clang++-14, I see:
49790 /home/smarchi/src/binutils-gdb/gdb/mips-tdep.c:453:12: error: use of bitwise '|' with boolean operands [-Werror,-Wbitwise-instead-of-logical]
49791 return !(MSYMBOL_TARGET_FLAG_MIPS16 (msym)
49792 ~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
49793 /home/smarchi/src/binutils-gdb/gdb/mips-tdep.h:54:2: note: expanded from macro 'MSYMBOL_TARGET_FLAG_MIPS16'
49794 (sym)->target_flag_1 ()
49796 /home/smarchi/src/binutils-gdb/gdb/mips-tdep.c:453:12: note: cast one or both operands to int to silence this warning
49797 /home/smarchi/src/binutils-gdb/gdb/mips-tdep.h:54:2: note: expanded from macro 'MSYMBOL_TARGET_FLAG_MIPS16'
49798 (sym)->target_flag_1 ()
49801 That's since commit e165fcef1e7 ("gdb: remove MSYMBOL_TARGET_FLAG_{1,2}
49802 macros"). Fix this by using the boolean || rather than the bitwise |,
49803 since the new methods return bool values. No change in behavior
49806 Change-Id: Ia82664135aa25db64c29c92f5c1141859d345bf7
49808 2022-04-13 Alexander von Gluck IV <kallisti5@unixzen.com>
49810 binutils: enable PE on 32bit haiku build
49811 * config.bfd (x86-haiku): Add i386_pei_vec as a selectable format.
49813 2022-04-13 Pedro Alves <pedro@palves.net>
49815 Make intrusive_list_node's next/prev private
49816 Tromey noticed that intrusive_list_node leaves its data members
49817 public, which seems sub-optimal.
49819 This commit makes intrusive_list_node's data fields private.
49820 intrusive_list_iterator, intrusive_list_reverse_iterator, and
49821 intrusive_list do need to access the fields, so they are made friends.
49823 Change-Id: Ia8b306b40344cc218d423c8dfb8355207a612ac5
49825 2022-04-13 Pedro Alves <pedro@palves.net>
49827 Tidy gdb.base/parse_number.exp
49828 Now that Ada is able to parse & print 0xffffffffffffffff (2^64-1) in
49829 hex, move it to the else branch like most other languages.
49831 Change-Id: Ib305f6bb2b6b230a1190ea783b245b865821094c
49833 2022-04-13 Alan Modra <amodra@gmail.com>
49835 ubsan: member access within null pointer of union
49836 Add some nonsense to cover "undefined behaviour".
49838 * ldlang.c (section_for_dot): Avoid UB.
49840 2022-04-13 GDB Administrator <gdbadmin@sourceware.org>
49842 Automatic date update in version.in
49844 2022-04-12 Tom Tromey <tromey@adacore.com>
49846 Fix bug in Ada number lexing
49847 On irc, Pedro pointed out that Ada couldn't properly handle
49848 0xffffffffffffffff. This used to work, but is a regression due to
49849 some patches I wrote in the Ada lexer. This patch fixes the bug.
49851 2022-04-12 Simon Marchi <simon.marchi@efficios.com>
49853 gdb: fix "passing NULL to memcpy" UBsan error in dwarf2/cooked-index.c
49854 Reading a simple file compiled with :
49856 $ gcc -DONE=1 -gdwarf-4 -g3 test.c
49858 gcc (Ubuntu 9.4.0-1ubuntu1~20.04) 9.4.0
49862 Reading symbols from /tmp/cwd/a.out...
49863 /home/smarchi/src/binutils-gdb/gdb/dwarf2/cooked-index.c:332:11: runtime error: null pointer passed as argument 2, which is declared to never be null
49865 It looks like even if the size is 0 (the size of the `entries` vector is
49866 0), we shouldn't be passing a NULL pointer to memcpy. And
49867 `entries.data ()` returns NULL.
49869 Fix that by using std::vector::insert to insert the items of entries
49870 into m_entries. I haven't checked, but it should essentially compile
49871 down to a memcpy, since the vector elements are trivially copyiable.
49873 Change-Id: I75f1c901e9b522e42e89eb5936e2c70d68eb21e5
49875 2022-04-12 Simon Marchi <simon.marchi@polymtl.ca>
49877 gdb: change subfile::line_vector to an std::vector
49878 Change this field to an std::vector to facilitate memory management.
49879 Since the linetable_entry array is copied into the symtab resulting from
49880 the subfile, it is possible to change it without changing how symtab
49881 stores the linetable entries (which would be a much larger change).
49883 There is a small change in buildsym_compunit::record_line to avoid
49884 accessing a now invalid linetable_entry. Before this patch, we keep a
49885 pointer to the last linetable entry, pop it from the vector, and then
49886 read last->line. It works with the manually-maintained array, but since
49887 we now use std::vector::pop_back, I am afraid that it could be flagged
49888 as an invalid access by the various static / dynamic analysis tools to
49889 access the linetable_entry object after popping it from the vector.
49890 Instead, record just the line number in an optional and use it.
49892 There are substantial changes in xcoffread.c that simplify the code, but
49893 I can't test them. I was hesitant to do this change because of that,
49894 but I decided to send it anyway. I don't think that an almost dead
49895 platform should hold back improving the code in the common parts of GDB.
49897 The changes in xcoffread.c are:
49899 - Make arrange_linetable "arrange" the linetable passed as a parameter,
49900 instead of returning possibly a new one, possibly the same one.
49901 - In the "Process main file's line numbers.", I'm not too sure what
49902 happens. We get the lintable from "main_subfile", "arrange" it, but
49903 then assign the result to the current subfile, obtained with
49904 get_current_subfile. I assume that the current subfile is also the
49905 main one, so now I just call arrange_linetable on the main subfile's
49907 - Remove that weird "Useless if!!!" FIXME comment. It's been there
49908 forever, but the "if" is still there, so I guess the "if" can stay
49911 Change-Id: I11799006fd85189e8cf5bd3a168f8f38c2c27a80
49913 2022-04-12 Simon Marchi <simon.marchi@efficios.com>
49915 gdb: use std::vector for temporary linetable_entry array in arrange_linetable
49916 Reduce manual memory management and make the code a bit easier to read.
49917 This helps me a bit in the following patch.
49919 I don't have a way to test this, it's best-effort.
49921 Change-Id: I64af9cd756311deabc6cd95e701dfb21234a40a5
49923 2022-04-12 Simon Marchi <simon.marchi@efficios.com>
49925 gdb: change subfile::name and buildsym_compunit::m_comp_dir to strings
49926 Change subfile::name to be a string, for easier memory management.
49927 Change buildsym_compunit::m_comp_dir as well, since we move one in to
49928 the other at some point in patch_subfile_names, so it's easier to do
49929 both at the same time. There are various NULL checks for both fields
49930 currently, replace them with empty checks, I think it ends up
49933 I can't test the change in xcoffread.c, it's best-effort.
49935 Change-Id: I62b5fb08b2089e096768a090627ac7617e90a016
49937 2022-04-12 Simon Marchi <simon.marchi@polymtl.ca>
49939 gdb: allocate subfile with new
49940 Allocate struct subfile with new, initialize its fields instead of
49941 memset-ing it to 0. Use a unique_ptr for the window after a subfile has
49942 been allocated but before it is linked in the buildsym_compunit's list
49943 of subfile (and therefore owned by the buildsym_compunit.
49945 I can't test the change in xcoffread.c, it's best-effort. I couldn't
49946 find where subfiles are freed in that file, I assume they were
49947 intentionally (or not) leaked.
49949 Change-Id: Ib3b6877de31b7e65bc466682f08dbf5840225f24
49951 2022-04-12 Simon Marchi <simon.marchi@efficios.com>
49953 gdb: use decltype instead of typeof in dwarf2/read.c
49954 When building with -std=c++11, I get:
49957 /home/smarchi/src/binutils-gdb/gdb/dwarf2/read.c: In function ‘void dwarf2_build_psymtabs_hard(dwarf2_per_objfile*)’:
49958 /home/smarchi/src/binutils-gdb/gdb/dwarf2/read.c:7130:23: error: expected type-specifier before ‘typeof’
49959 7130 | using iter_type = typeof (per_bfd->all_comp_units.begin ());
49962 This is because typeof is a GNU extension. Use C++'s decltype keyword
49965 Change-Id: Ieca2e8d25e50f71dc6c615a405a972a54de3ef14
49967 2022-04-12 Simon Marchi <simon.marchi@efficios.com>
49969 gdbsupport: use result_of_t instead of result_of in parallel-for.h
49970 When building with -std=c++11, I get:
49972 In file included from /home/smarchi/src/binutils-gdb/gdb/unittests/parallel-for-selftests.c:22: /home/smarchi/src/binutils-gdb/gdb/../gdbsupport/parallel-for.h:134:10: error: ‘result_of_t’ is not a member of ‘std’; did you mean ‘result_of’?
49973 134 | std::result_of_t<RangeFunction (RandomIt, RandomIt)>
49977 This is because result_of_t has been introduced in C++14. Use the
49978 equivalent result_of<...>::type instead.
49980 result_of and result_of_t have been removed in C++20 though, so I think
49981 we'll need some patches eventually to make the code use invoke_result
49982 instead, depending on the C++ version.
49984 Change-Id: I4817f361c0ebcdd4b32976898fc368bb302b61b9
49986 2022-04-12 Tom Tromey <tom@tromey.com>
49988 Remove dwarf2_per_cu_data::v
49989 Now that the psymtab reader has been removed, the
49990 dwarf2_per_cu_data::v union is no longer needed. Instead, we can
49991 simply move the members from dwarf2_per_cu_quick_data into
49992 dwarf2_per_cu_data and remove the "quick" object entirely.
49994 Delete DWARF psymtab code
49995 This removes the DWARF psymtab reader.
49997 2022-04-12 Tom Tromey <tom@tromey.com>
49999 Enable the new DWARF indexer
50000 This patch finally enables the new indexer. It is left until this
50001 point in the series to avoid any regressions; in particular, it has to
50002 come after the changes to the DWARF index writer to avoid this
50005 However, if you experiment with the series, this patch can be moved
50006 anywhere from the patch to wire in the new reader to this point.
50007 Moving this patch around is how I got separate numbers for the
50008 parallelization and background finalization patches.
50010 In the ongoing performance example, this reduces the time from the
50011 baseline of 1.598869 to 0.903534.
50013 2022-04-12 Tom Tromey <tom@tromey.com>
50015 Adapt .debug_names writer to new DWARF scanner
50016 This updates the .debug_names writer to work with the new DWARF
50019 2022-04-12 Tom Tromey <tom@tromey.com>
50021 Adapt .gdb_index writer to new DWARF scanner
50022 This updates the .gdb_index writer to work with the new DWARF scanner.
50023 The .debug_names writer is deferred to another patch, to make review
50026 This introduces a small hack to psyms_seen_size, but is
50027 inconsequential because this function will be deleted in a subsequent
50030 2022-04-12 Tom Tromey <tom@tromey.com>
50032 Genericize addrmap handling in the DWARF index writer
50033 This updates the DWARF index writing code to make the addrmap-writing
50034 a bit more generic. Now, it can handle multiple maps, and it can work
50035 using the maps generated by the new indexer.
50037 Note that the new addrmap_index_data::using_index field will be
50038 deleted in a future patch, when the rest of the DWARF psymtab code is
50041 2022-04-12 Tom Tromey <tom@tromey.com>
50043 Change parameters to write_address_map
50044 To support the removal of partial symtabs from the DWARF index writer,
50045 this makes a small change to have write_address_map accept the address
50046 map as a parameter, rather than assuming it always comes from the
50049 Change the key type in psym_index_map
50050 In order to change the DWARF index writer to avoid partial symtabs,
50051 this patch changes the key type in psym_index_map (and renames that
50052 type as well). Using the dwarf2_per_cu_data as the key makes it
50053 simpler to reuse this code with the new indexer.
50055 Rename write_psymtabs_to_index
50056 We'll be removing all the psymtab code from the DWARF reader. As a
50057 preparatory step, this renames write_psymtabs_to_index to avoid the
50060 2022-04-12 Tom Tromey <tom@tromey.com>
50062 "Finalize" the DWARF index in the background
50063 After scanning the CUs, the DWARF indexer merges all the data into a
50064 single vector, canonicalizing C++ names as it proceeds. While not
50065 necessarily single-threaded, this process is currently done in just
50066 one thread, to keep memory costs lower.
50068 However, this work is all done without reference to any data outside
50069 of the indexes. This patch improves the apparent performance of GDB
50070 by moving it to the background. All uses of the index are then made
50071 to wait for this process to complete.
50073 In our ongoing example, this reduces the scanning time on gdb itself
50074 to 0.173937 (wall). Recall that before this patch, the time was
50075 0.668923; and psymbol reader does this in 1.598869. That is, at the
50076 end of this series, we see about a 10x speedup.
50078 2022-04-12 Tom Tromey <tom@tromey.com>
50080 Parallelize DWARF indexing
50081 This parallelizes the new DWARF indexer. The indexer's storage was
50082 designed so that each storage object and each indexer is fully
50083 independent. This setup makes it simple to scan different CUs
50086 This patch creates a new cooked index storage object per thread, and
50087 then scans a subset of all the CUs in each such thread, using gdb's
50088 existing thread pool.
50090 In the ongoing "gdb gdb" example, this patch reduces the wall time
50091 down to 0.668923, from 0.903534. (Note that the 0.903534 is the time
50092 for the new index -- that is, when the "enable the new index" patch is
50093 rebased to before this one. However, in the final series, that patch
50094 appears toward the end. Hopefully this isn't too confusing.)
50096 2022-04-12 Tom Tromey <tom@tromey.com>
50098 Pre-read DWARF section data
50099 Because BFD is not thread-safe, we need to be sure that any section
50100 data that is needed is read before trying to do any DWARF indexing in
50103 This patch takes a simple approach to this -- it pre-reads the
50104 "info"-related sections. This is done for the main file, but also any
50105 auxiliary files as well, such as the DWO file.
50107 This patch could be perhaps enhanced by removing some now-redundant
50108 calls to dwarf2_section_info::read.
50110 2022-04-12 Tom Tromey <tom@tromey.com>
50112 Introduce thread-safe handling for complaints
50113 This introduces a new class that can be used to make the "complaint"
50114 code thread-safe. Instantiating the class installs a new handler that
50115 collects complaints, and then prints them all when the object is
50118 This approach requires some locks. I couldn't think of a better way
50119 to handle this, though, because the I/O system is not thread-safe.
50121 It seemed to me that only GDB developers are likely to enable
50122 complaints, and because the complaint macro handle this case already
50123 (before any locks are required), I reasoned that any performance
50124 degradation that would result here would be fine.
50126 As an aside about complaints -- are they useful at all? I just ignore
50127 them, myself, since mostly they seem to indicate compiler problems
50128 that can't be solved in the GDB world anyway. I'd personally prefer
50129 them to be in a separate tool, like a hypothetical 'dwarflint'.
50131 2022-04-12 Tom Tromey <tom@tromey.com>
50133 Wire in the new DWARF indexer
50134 This wires the new DWARF indexer into the existing reader code. That
50135 is, this patch makes the modification necessary to enable the new
50136 indexer. It is not actually enabled by this patch -- that will be
50139 I did a bit of performance testing for this patch and a few others. I
50140 copied my built gdb to /tmp, so that each test would be done on the
50141 same executable. Then, each time, I did:
50145 (gdb) file /tmp/gdb
50147 This patch is the baseline and on one machine came in at 1.598869 wall
50150 2022-04-12 Tom Tromey <tom@tromey.com>
50152 Implement quick_symbol_functions for cooked DWARF index
50153 This implements quick_symbol_functions for the cooked DWARF index.
50154 This is the code that interfaces between the new index and the rest of
50155 gdb. Cooked indexes still aren't created by anything.
50157 For the most part this is straightforward. It shares some concepts
50158 with the existing DWARF indices. However, because names are stored
50159 pre-split in the cooked index, name lookup here is necessarily
50160 different; see expand_symtabs_matching for the gory details.
50162 2022-04-12 Tom Tromey <tom@tromey.com>
50164 The new DWARF indexer
50165 This patch adds the code to index DWARF. This is just the scanner; it
50166 reads the DWARF and constructs the index, but nothing calls it yet.
50168 The indexer is split into two parts: a storage object and an indexer
50169 object. This is done to support the parallelization of this code -- a
50170 future patch will create a single storage object per thread.
50172 2022-04-12 Tom Tromey <tom@tromey.com>
50174 Introduce the new DWARF index class
50175 This patch introduces the new DWARF index class. It is called
50176 "cooked" to contrast against a "raw" index, which is mapped from disk
50177 without extra effort.
50179 Nothing constructs a cooked index yet. The essential idea here is
50180 that index entries are created via the "add" method; then when all the
50181 entries have been read, they are "finalize"d -- name canonicalization
50182 is performed and the entries are added to a sorted vector.
50184 Entries use the DWARF name (DW_AT_name) or linkage name, not the full
50185 name as is done for partial symbols.
50187 These two facets -- the short name and the deferred canonicalization
50188 -- help improve the performance of this approach. This will become
50189 clear in later patches, when parallelization is added.
50191 Some special code is needed for Ada, because GNAT only emits mangled
50192 ("encoded", in the Ada lingo) names, and so we reconstruct the
50193 hierarchical structure after the fact. This is also done in the
50194 finalization phase.
50196 One other aspect worth noting is that the way the "main" function is
50197 found is different in the new code. Currently gdb will notice
50198 DW_AT_main_subprogram, but won't recognize "main" during reading --
50199 this is done later, via explicit symbol lookup. This is done
50200 differently in the new code so that finalization can be done in the
50201 background without then requiring a synchronization to look up the
50204 2022-04-12 Tom Tromey <tom@tromey.com>
50206 Update skip_one_die for new abbrev properties
50207 This updates skip_one_die to speed it up in the cases where either
50208 sibling_offset or size_if_constant are set.
50210 2022-04-12 Tom Tromey <tom@tromey.com>
50212 Statically examine abbrev properties
50213 The new DIE scanner works more or less along the lines indicated by
50214 the text for the .debug_names section, disregarding the bugs in the
50217 While working on this, I noticed that whether a DIE is interesting is
50218 a static property of the DIE's abbrev. It also turns out that many
50219 abbrevs imply a static size for the DIE data, and additionally that
50220 for many abbrevs, the sibling offset is stored at a constant offset
50221 from the start of the DIE.
50223 This patch changes the abbrev reader to analyze each abbrev and stash
50224 the results on the abbrev. These combine to speed up the new indexer.
50225 If the "interesting" flag is false, GDB knows to skip the DIE
50226 immediately. If the sibling offset is statically known, skipping can
50227 be done without reading any attributes; and in some other cases, the
50228 DIE can be skipped using simple arithmetic.
50230 2022-04-12 Tom Tromey <tom@tromey.com>
50232 Introduce DWARF abbrev cache
50233 The replacement for the DWARF psymbol reader works in a somewhat
50234 different way. The current reader reads and stores all the DIEs that
50235 might be interesting. Then, if it is missing a DIE, it re-scans the
50236 CU and reads them all. This approach is used for both intra- and
50237 inter-CU references.
50239 I instrumented the partial DIE hash to see how frequently it was used:
50247 That is, most DIEs are never used, and some are looked up twice -- but
50248 this is just an artifact of the implementation of
50249 partial_die_info::fixup, which may do two lookups.
50251 Based on this, the new implementation doesn't try to store any DIEs,
50252 but instead just re-scans them on demand. In order to do this,
50253 though, it is convenient to have a cache of DWARF abbrevs. This way,
50254 if a second CU is needed to resolve an inter-CU reference, the abbrevs
50255 for that CU need only be computed a single time.
50257 2022-04-12 Tom Tromey <tom@tromey.com>
50259 Add "fullname" handling to file_and_directory
50260 This changes the file_and_directory object to be able to compute and
50261 cache the "fullname" in the same way that is done by other code, like
50262 the psymtab reader.
50264 Specialize std::hash for gdb_exception
50265 This adds a std::hash specialization for gdb_exception. This lets us
50266 store these objects in a hash table, which is used later in this
50267 series to de-duplicate the exception output from multiple threads.
50269 Return vector of results from parallel_for_each
50270 This changes gdb::parallel_for_each to return a vector of the results.
50271 However, if the passed-in function returns void, the return type
50272 remains 'void'. This functionality is used later, to parallelize the
50275 Add batching parameter to parallel_for_each
50276 parallel_for_each currently requires each thread to process at least
50277 10 elements. However, when indexing, it's fine for a thread to handle
50278 just a single CU. This patch parameterizes this, and updates the one
50281 Refactor build_type_psymtabs_reader
50282 The new DWARF scanner needs to save the entire cutu_reader object, not
50283 just parts of it. In order to make this possible, this patch
50284 refactors build_type_psymtabs_reader. This change is done separately
50285 because it is easy to review in isolation and it helps make the later
50288 Add new overload of dwarf5_djb_hash
50289 This adds a new overload of dwarf5_djb_hash. This is used in
50290 subsequent patches.
50292 2022-04-12 Tom Tromey <tom@tromey.com>
50295 The new DWARF index code works by keeping names pre-split. That is,
50296 rather than storing a symbol name like "a::b::c", the names "a", "b",
50297 and "c" will be stored separately.
50299 This patch introduces some helper code to split a full name into its
50302 2022-04-12 Tom Tromey <tom@tromey.com>
50304 Let skip_one_die not skip children
50305 This patch adds an option to skip_one_die that causes it not to skip
50306 child DIEs. This is needed in the new scanner.
50308 Allow ada_decode not to decode operators
50309 The new DWARF scanner records names as they appear in DWARF. However,
50310 because Ada is unusual, it also decodes the Ada names to synthesize
50311 package components for them. In order for this to work out properly,
50312 gdb also needs a mode where ada_decode can be instructed not to decode
50313 Ada operator names. That is what this patch implements.
50315 Refactor dwarf2_get_pc_bounds
50316 This changes dwarf2_get_pc_bounds so that it does not directly access
50317 a psymtab or psymtabs_addrmap. Instead, both the addrmap and the
50318 desired payload are passed as parameters. This makes it suitable to
50319 be used by the new indexer.
50321 Add dwarf2_per_cu_data::addresses_seen
50322 This adds a new member to dwarf2_per_cu_data that indicates whether
50323 addresses have been seen for this CU. This is then set by the
50324 .debug_aranges reader. The idea here is to detect when a CU does not
50325 have address information, so that the new indexer will know to do
50326 extra scanning in that case.
50328 Fix latent bug in read_addrmap_from_aranges
50329 Tom de Vries found a failure that we tracked down to a latent bug in
50330 read_addrmap_from_aranges (previously create_addrmap_from_aranges).
50331 The bug is that this code can erroneously reject .debug_aranges when
50332 dwz is in use, due to CUs at duplicate offsets. Because aranges can't
50333 refer to a CU coming from the dwz file, the fix is to simply skip such
50336 Split create_addrmap_from_aranges
50337 This patch splits create_addrmap_from_aranges into a wrapper function
50338 and a worker function. The worker function is then used in a later
50341 2022-04-12 Tom Tromey <tom@tromey.com>
50343 Allow thread-pool.h to work without threads
50344 thread-pool.h requires CXX_STD_THREAD in order to even be included.
50345 However, there's no deep reason for this, and during review we found
50346 that one patch in the new DWARF indexer series unconditionally
50347 requires the thread pool.
50349 Because the thread pool already allows a task to be run in the calling
50350 thread (for example if it is configured to have no threads in the
50351 pool), it seemed straightforward to make this code ok to use when host
50352 threads aren't available at all.
50354 This patch implements this idea. I built it on a thread-less host
50355 (mingw, before my recent configure patch) and verified that the result
50358 After the thread-pool change, parallel-for.h no longer needs any
50359 CXX_STD_THREAD checks at all, so this patch removes these as well.
50361 2022-04-12 Nick Clifton <nickc@redhat.com>
50363 Rebase the zlib sources to the 1.2.12 release
50365 2022-04-12 Tom de Vries <tdevries@suse.de>
50367 [gdb/testsuite] Fix gdb.base/stap-probe.exp with read1
50368 When running test-case gdb.base/stap-probe.exp with make target check-read1, I
50369 run into this and similar:
50371 FAIL: gdb.base/stap-probe.exp: without semaphore, not optimized: \
50372 info probes stap (timeout)
50375 Fix this by using gdb_test_lines instead of gdb_test.
50377 Tested on x86_64-linux.
50379 2022-04-12 Tom Tromey <tom@tromey.com>
50381 Add C++ "save gdb-index" test
50382 I found a bug in the new DWARF reader series, related to the handling
50383 of enumerator names. This bug caused a crash, so this patch adds a
50384 regression test for this particular case. I'm checking this in.
50386 2022-04-12 Tom Tromey <tromey@adacore.com>
50388 Remove "Ada Settings" node from the manual
50389 A while back, I sent a patch to unify the Ada varsize-limit setting
50390 with the more generic max-value-size:
50392 https://sourceware.org/pipermail/gdb-patches/2021-September/182004.html
50394 However, it turns out I somehow neglected to send part of the patch.
50395 Internally, I also removed the "Ada Settings" node from the manual, as
50396 it only documents the obsolete setting.
50398 This patch removes this text.
50400 2022-04-12 Tom Tromey <tromey@adacore.com>
50402 Require GNAT debug info for some Ada tests
50403 A few Ada tests require some debug info in the GNAT runtime. When run
50404 without this info, these tests can't pass. This patch changes these
50405 tests to detect this situation and stop with "untested".
50407 2022-04-12 Nick Clifton <nickc@redhat.com>
50409 Stop strip from removing debuglink sections.
50411 * objcopy.c (is_strip_section_1): Do not delete debuglink sections
50412 when stripping debug information.
50414 2022-04-12 Jan Beulich <jbeulich@suse.com>
50416 gas: new_logical_line{,_flags}() can return "void"
50417 With the sole user of the return value gone, convert the return type to
50418 void. This in turn allows simplifying another construct, by moving it
50419 slightly later in the function.
50421 gas: drop .appfile and .appline
50422 These were used originally to represent "# <line> <file>" constructs
50423 inserted by (typically) compilers when pre-processing. Quite some time
50424 ago they were replaced by .linefile though. Since the original
50425 directives were never documented, we ought to be able to remove support
50426 for them. As a result in a number of case function parameter aren't used
50427 anymore and can hence be dropped.
50429 2022-04-12 Jan Beulich <jbeulich@suse.com>
50431 gas: further adjust file/line handling for .macro
50432 Commit 7992631e8c0b ("gas/Dwarf: improve debug info generation from .irp
50433 and alike blocks"), while dealing okay with actual assembly source files
50434 not using .file/.line and alike outside but not inside of .macro, has
50435 undue effects when the logical file/line pair was already overridden:
50436 Line numbers would continuously increment while processing the expanded
50437 macro, while the goal of the PR gas/16908 workaround is to keep the
50438 expansion associated with the line invoking the macro. However, as soon
50439 as enough state was overridden _inside_ the macro to cause as_where() to
50440 no longer fall back top as_where_physical(), honor this by resuming the
50441 bumping of the logical line number.
50443 Note that from_sb_is_expansion's initializer was 1 for an unknown
50444 reason. While renaming the variable and changing its type, also change
50445 the initializer to "expanding_none", which would have been "0" in the
50446 original code. Originally the initializer value itself wasn't ever used
50447 anyway (requiring sb_index != -1), as it necessarily had changed in
50448 input_scrub_include_sb() alongside setting sb_index to other than -1.
50450 Strictly speaking input_scrub_insert_line() perhaps shouldn't use
50451 expanding_none, yet none of the other enumerators fit there either. And
50452 then strictly speaking that function probably shouldn't exist in the
50453 first place. It's used only by tic54x.
50455 2022-04-12 Jan Beulich <jbeulich@suse.com>
50457 gas: further adjust file/line handling for .irp and alike
50458 Commit 7992631e8c0b ("gas/Dwarf: improve debug info generation from .irp
50459 and alike blocks"), while dealing okay with actual assembly source files
50460 not using .file/.line and alike outside but not inside of .irp et al,
50461 has undue effects when the logical file/line pair was already
50462 overridden: Line numbers would continuously increment upon every
50463 iteration, thus potentially getting far off. Furthermore it left it to
50464 the user to actually insert .file/.line inside such constructs. Note
50465 though that before aforementioned change things weren't pretty either:
50466 Diagnostics (and debug info) would be associated with the directive
50467 terminating the iteration construct, rather than with the actual lines.
50469 Handle this automatically by simply latching the present line and then
50470 re-instating coordinates first thing on every iteration; note that the
50471 file can't change from what was previously pushed on the scrubber's
50472 state stack, and hence can be taken from there by using a new flavor of
50473 .linefile (which is far better memory-footprint-wise than recording the
50474 full path in the inserted directive). (This then leaves undisturbed any
50475 file/line control occurring in the body of the construct, as these will
50476 only be seen and processed afterwards.)
50478 2022-04-12 Jan Beulich <jbeulich@suse.com>
50480 x86: make {disp16} work similarly to {disp32}
50481 In a few places {disp32} was handled specially when really {disp16}
50482 wants handling just the same.
50484 2022-04-12 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
50486 gprofng doesn't build with gcc 5.5
50488 2022-04-07 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
50491 * configure.ac: Check version of bison.
50492 * src/Makefile.am (QLParser.yy): Run bison
50493 * src/QLParser.yy: Adapted for bison 3.04 or later.
50494 * src/DbeSession.cc: make some params const.
50495 * src/DbeSession.h: Likewise.
50496 * configure: Regenerate.
50497 * Makefile.in: Regenerate.
50498 * src/Makefile.in: Regenerate.
50499 * src/QLParser.tab.cc: Deleted.
50500 * src/QLParser.tab.hh: Deleted.
50501 * doc/Makefile.in: Regenerate.
50502 * gp-display-html/Makefile.in: Regenerate.
50503 * libcollector/configure: Regenerate.
50505 2022-04-12 GDB Administrator <gdbadmin@sourceware.org>
50507 Automatic date update in version.in
50509 2022-04-11 John Baldwin <jhb@FreeBSD.org>
50511 i386-fbsd-nat: Remove two unused variables.
50512 Earlier versions of the change in
50513 1285ce8629b37f800bf21731ee7c7a8b1b8d0233 used this variable, but not
50514 the final version that landed.
50516 2022-04-11 Simon Marchi <simon.marchi@efficios.com>
50518 gdb: remove MSYMBOL_TARGET_FLAG_{1,2} macros
50519 Replace with equivalent getter/setter macros.
50521 Change-Id: I1042564dd47347337374762bd64ec31b5c573ee2
50523 2022-04-11 Simon Marchi <simon.marchi@efficios.com>
50525 gdb: remove minimal symbol size macros
50526 Remove MSYMBOL_HAS_SIZE, MSYMBOL_SIZE and SET_MSYMBOL_SIZE, replace them
50527 with equivalent methods.
50529 Change-Id: I6ee1cf82df37e58dff52ea6568ceb4649c7d7538
50531 2022-04-11 Simon Marchi <simon.marchi@efficios.com>
50533 gdb: remove MSYMBOL_TYPE macro
50534 Add a getter and a setter for a minimal symbol's type. Remove the
50535 corresponding macro and adjust all callers.
50537 Change-Id: I89900df5ffa5687133fe1a16b2e0d4684e67a77d
50539 2022-04-11 Simon Marchi <simon.marchi@efficios.com>
50541 gdb: remove symbol value macros
50542 Remove all macros related to getting and setting some symbol value:
50544 #define SYMBOL_VALUE(symbol) (symbol)->value.ivalue
50545 #define SYMBOL_VALUE_ADDRESS(symbol) \
50546 #define SET_SYMBOL_VALUE_ADDRESS(symbol, new_value) \
50547 #define SYMBOL_VALUE_BYTES(symbol) (symbol)->value.bytes
50548 #define SYMBOL_VALUE_COMMON_BLOCK(symbol) (symbol)->value.common_block
50549 #define SYMBOL_BLOCK_VALUE(symbol) (symbol)->value.block
50550 #define SYMBOL_VALUE_CHAIN(symbol) (symbol)->value.chain
50551 #define MSYMBOL_VALUE(symbol) (symbol)->value.ivalue
50552 #define MSYMBOL_VALUE_RAW_ADDRESS(symbol) ((symbol)->value.address + 0)
50553 #define MSYMBOL_VALUE_ADDRESS(objfile, symbol) \
50554 #define BMSYMBOL_VALUE_ADDRESS(symbol) \
50555 #define SET_MSYMBOL_VALUE_ADDRESS(symbol, new_value) \
50556 #define MSYMBOL_VALUE_BYTES(symbol) (symbol)->value.bytes
50557 #define MSYMBOL_BLOCK_VALUE(symbol) (symbol)->value.block
50559 Replace them with equivalent methods on the appropriate objects.
50561 Change-Id: Iafdab3b8eefc6dc2fd895aa955bf64fafc59ed50
50563 2022-04-11 Nils-Christian Kempke <nils-christian.kempke@intel.com>
50565 gdb/doc: add section about Fortran intrinsic functions and types
50566 The earlier version of this document had no sections about the
50567 available Fortran intrinsic functions or the Fortran builtin types.
50569 I added two sections 'Fortran intrinsics' and 'Fortran types' to
50570 document the available Fortran language features. The subsection
50571 'Fortran Defaults' has been integrated into the Fortran subsection.
50573 2022-04-11 Nils-Christian Kempke <nils-christian.kempke@intel.com>
50575 gdb/fortran/testsuite: add complex from integers test
50576 When working on the files I noted that there was no actual test for a
50577 COMPLEX built from two INTEGERS. I added that now for completion.
50579 2022-04-11 Nils-Christian Kempke <nils-christian.kempke@intel.com>
50581 gdb/fortran: rewrite intrinsic handling and add some missing overloads
50582 The operators FLOOR, CEILING, CMPLX, LBOUND, UBOUND, and SIZE accept
50583 (some only with Fortran 2003) the optional parameter KIND. This
50584 parameter determines the kind of the associated return value. So far,
50585 implementation of this kind parameter has been missing in GDB.
50586 Additionally, the one argument overload for the CMPLX intrinsic function
50587 was not yet available.
50589 This patch adds overloads for all above mentioned functions to the
50590 Fortran intrinsics handling in GDB.
50592 It re-writes the intrinsic function handling section to use the helper
50593 methods wrap_unop_intrinsic/wrap_binop_intrinsic/wrap_triop_intrinsic.
50594 These methods define the action taken when a Fortran intrinsic function
50595 is called with a certain amount of arguments (1/2/3). The helper methods
50596 fortran_wrap2_kind and fortran_wrap3_kind have been added as equivalents
50597 to the existing wrap and wrap2 methods.
50599 After adding more overloads to the intrinsics handling, some of the
50600 operation names were no longer accurate. E.g. UNOP_FORTRAN_CEILING
50601 has been renamed to FORTRAN_CEILING as it is no longer a purely unary
50602 intrinsic function. This patch also introduces intrinsic functions with
50603 one, two, or three arguments to the Fortran parser and the
50604 UNOP_OR_BINOP_OR_TERNOP_INTRINSIC token has been added.
50606 2022-04-11 Nils-Christian Kempke <nils-christian.kempke@intel.com>
50608 gdb/fortran: rename f77_keywords to f_keywords
50609 Rename f77_keywords to f_keywords since some of the introduced keywords
50610 in the array are f90 only.
50612 2022-04-11 Nils-Christian Kempke <nils-christian.kempke@intel.com>
50614 gdb/fortran: Change GDB print for fortran default types
50615 Currently, when asking GDB to print the type of a Fortran default type
50616 such as INTEGER or REAL, GDB will return the default name of that type,
50617 e.g. "integer"/"real":
50619 (gdb) ptype integer
50624 For LOGICAL and COMPLEX it would return the actual underlying types
50626 (gdb) ptype logical
50628 (gdb) ptype complex
50631 Similarly, GDB would print the default integer type for the underlying
50634 (gdb) ptype integer*4
50638 (gdb) ptype logical
50640 (gdb) ptype complex*4
50643 This is inconsistent and a bit confusing. Both options somehow indicate
50644 what the internal underlying type for the default type is - but I think
50645 the logical/complex version is a bit clearer.
50649 (gdb) ptype integer
50652 This indicates to a user that the type of "integer" is Fortran's default
50653 integer type. Without examining "ptype integer*4" I would expect, that
50654 any variable declared integer in the actual code would also fit into a
50655 GDB integer. But, since we cannot adapt out internal types to the
50656 compiler flags used at compile time of a debugged binary, this might be
50657 wrong. Consider debugging Fortran code compiled with GNU and e.g. the
50658 "-fdefault-integer-8" flag. In this case the gfortran default integer
50659 would be integer*8 while GDB internally still would use a builtin_integer,
50660 so an integer of the size of an integer*4 type. On the other hand having
50663 (gdb) ptype integer
50666 makes this clearer. I would still be tempted to fit a variable declared
50667 integer in the code into a GDB integer - but at least ptype would
50668 directly tell me what is going on. Note, that when debugging a binary
50669 compiled with "-fdefault-integer-8" a user will always see the actual
50670 underlying type of any variable declared "integer" in the Fortran code.
50675 print *, a ! breakpt
50678 will, when breaking at breakpt print
50681 type = integer(kind=4)
50686 type = integer(kind=8)
50688 depending on the compiler flag.
50690 This patch changes the outputs for the REAL and INTEGER default types to
50691 actually print the internally used type over the default type name.
50693 The new behavior for the above examples is:
50695 (gdb) ptype integer
50697 (gdb) ptype integer*4
50700 Existing testcases have been adapted to reflect the new behavior.
50702 2022-04-11 Nils-Christian Kempke <nils-christian.kempke@intel.com>
50704 gdb/fortran: clean-up Fortran intrinsic types
50705 The currently implemented intrinsic type handling for Fortran missed some
50706 tokens and their parsing. While still not all Fortran type kinds are
50707 implemented this patch at least makes the currently handled types
50708 consistent. As an example for what this patch does, consider the
50709 intrinsic type INTEGER. GDB implemented the handling of the
50710 keywords "integer" and "integer_2" but missed "integer_4" and "integer_8"
50711 even though their corresponding internal types were already available as
50712 the Fortran builtin types builtin_integer and builtin_integer_s8.
50713 Similar problems applied to LOGICAL, REAL, and COMPLEX. This patch adds
50714 all missing tokens and their parsing. Whenever a section containing the
50715 type handling was touched, it also was reordered to be in a more easy to
50716 grasp order. All INTEGER/REAL/LOGICAL/COMPLEX types were grouped
50717 together and ordered ascending in their size making a missing one more
50720 Before this change GDB would print the following when tyring to use the
50723 (gdb) set language fortran
50724 (gdb) ptype integer*1
50725 unsupported kind 1 for type integer
50726 (gdb) ptype integer_1
50727 No symbol table is loaded. Use the "file" command.
50728 (gdb) ptype integer*2
50730 (gdb) ptype integer_2
50732 (gdb) ptype integer*4
50734 (gdb) ptype integer_4
50735 No symbol table is loaded. Use the "file" command.
50736 (gdb) ptype integer*8
50738 (gdb) ptype integer_8
50739 No symbol table is loaded. Use the "file" command.
50740 (gdb) ptype integer
50743 With this patch all keywords are available and the GDB prints:
50745 (gdb) set language fortran
50746 (gdb) ptype integer*1
50748 (gdb) ptype integer_1
50750 (gdb) ptype integer*2
50752 (gdb) ptype integer_2
50754 (gdb) ptype integer*4
50756 (gdb) ptype integer_4
50758 (gdb) ptype integer*8
50760 (gdb) ptype integer_8
50762 (gdb) ptype integer
50765 The described changes have been applied to INTEGER, REAL, COMPLEX,
50766 and LOGICAL. Existing testcases have been adapted to reflect the
50767 new behavior. Tests for formerly missing types have been added.
50769 2022-04-11 Nils-Christian Kempke <nils-christian.kempke@intel.com>
50771 gdb/fortran: change default logical type to builtin_logical
50772 According to the Fortran standard, logical is of the size of a
50773 'single numeric storage unit' (just like real and integer). For
50774 gfortran, flang and ifx/ifort this storage unit (or the default
50775 logical type) is of size kind 4, actually occupying 4 bytes of
50776 storage, and so the default type for logical expressions in
50777 Fortran should probably also be Logical*4 and not Logical*2. I
50778 adapted GDB's behavior to be in line with
50779 gfortran/ifort/ifx/flang.
50781 gdb/fortran: reformat build_fortran_types in f-lang.c
50782 Add a few newlines after the type definitions and remove some
50783 unnecessary linebreaks.
50785 2022-04-11 Nils-Christian Kempke <nils-christian.kempke@intel.com>
50787 gdb/fortran: fix complex type in Fortran builtin types
50788 Before this patch things like
50790 (gdb) ptype complex*8
50792 (gdb) ptype complex*4
50795 were possible in GDB, which seems confusing for a user. The reason
50796 is a mixup in the implementation of the Fortran COMPLEX type. In
50797 Fortran the "*X" after a type would normally (I don't think this
50798 is language required) specify the type's size in memory. For the
50799 COMPLEX type the kind parameters usually (at least for GNU, Intel, Flang)
50800 specify not the size of the whole type but the size of the individual
50801 two REALs used to form the COMPLEX. Thus, a COMPLEX*4 will usually
50802 consist of two REAL*4s. Internally this type was represented by a
50803 builtin_complex_s8 - but here I think the s8 actually meant the raw
50804 size of the type. This is confusing and I renamed the types (e.g.
50805 builting_complex_s8 became builtin_complex_s4 according to its most
50806 common useage) and their printed names to their language equivalent.
50807 Additionally, I added the default COMPLEX type "COMPLEX" being the same
50808 as a COMPLEX*4 (as is normally the case) and removed the latter. I added
50809 a few tests for this new behavior as well.
50811 The new behavior is
50813 (gdb) ptype complex*8
50815 (gdb) ptype complex*4
50818 2022-04-11 Nils-Christian Kempke <nils-christian.kempke@intel.com>
50820 gdb/f-lang: remove hidden ^L characters
50822 gdb/f-lang: add Integer*1 to Fortran builtin types
50823 Add builtin_integer_s1 of size TARGET_CHAR_BIT to Fortran builtin types.
50825 2022-04-11 Tom de Vries <tdevries@suse.de>
50827 [gdb/testsuite] Fix gdb.base/annota1.exp with pie
50828 Since commit 359efc2d894 ("[gdb/testsuite] Make gdb.base/annota1.exp more
50829 robust") we see this fail with target board unix/-fPIE/-pie:
50831 FAIL: gdb.base/annota1.exp: run until main breakpoint (timeout)
50834 The problem is that the commit makes the number and order of matched
50835 annotations fixed, while between target boards unix and unix/-fPIE/-pie there
50838 \032\032post-prompt
50839 Starting program: outputs/gdb.base/annota1/annota1
50841 +\032\032breakpoints-invalid
50845 \032\032frames-invalid
50848 Fix this by optionally matching the additional annotation.
50850 Tested on x86_64-linux.
50852 2022-04-11 Tom de Vries <tdevries@suse.de>
50854 [gdb/testsuite] Fix gdb.dwarf2/dw2-lines.exp for m32 pie
50855 As reported in PR29043, when running test-case gdb.dwarf2/dw2-lines.exp with
50856 target board unix/-m32/-fPIE/-pie, we run into:
50858 Breakpoint 2, 0x56555540 in bar ()^M
50859 (gdb) PASS: gdb.dwarf2/dw2-lines.exp: cv=2: cdw=32: lv=2: ldw=32: \
50860 continue to breakpoint: foo \(1\)
50862 Single stepping until exit from function bar,^M
50863 which has no line number information.^M
50864 0x56555587 in main ()^M
50865 (gdb) FAIL: gdb.dwarf2/dw2-lines.exp: cv=2: cdw=32: lv=2: ldw=32: \
50869 The problem is that the bar breakpoint ends up at an unexpected location
50871 - the synthetic debug info is incomplete and doesn't provide line info
50872 for the prologue part of the function, so consequently gdb uses the i386
50873 port prologue skipper to get past the prologue
50874 - the i386 port prologue skipper doesn't get past a get_pc_thunk call.
50876 Work around this in the test-case by breaking on bar_label instead.
50878 Tested on x86_64-linux with target boards unix, unix/-m32, unix/-fPIE/-pie and
50879 unix/-m32/-fPIE/-pie.
50881 2022-04-11 GDB Administrator <gdbadmin@sourceware.org>
50883 Automatic date update in version.in
50885 2022-04-10 GDB Administrator <gdbadmin@sourceware.org>
50887 Automatic date update in version.in
50889 2022-04-09 Tom Tromey <tom@tromey.com>
50891 Remove MSYMBOL_VALUE_CHAIN
50892 I noticed that MSYMBOL_VALUE_CHAIN is unused, so this patch removes it.
50894 2022-04-09 Alan Modra <amodra@gmail.com>
50896 Rearrange struct bfd_section a little
50897 For better packing on 64-bit hosts.
50899 * section.c (struct bfd_section): Move next and prev field earlier.
50900 Move alignment_power later.
50901 (BFD_FAKE_SECTION): Adjust to suit.
50902 * bfd-in2.h: Regenerate.
50904 2022-04-09 Alan Modra <amodra@gmail.com>
50906 Don't run pr27228 test for hppa
50907 As the comment says, hppa doesn't support use of BFD_RELOC_* in
50908 .reloc directives. Using xfail can result in a spurious XPASS result
50909 as BFD_RELOC values change.
50911 * testsuite/gas/elf/pr27228.d: Change xfail to notarget for hppa.
50913 2022-04-09 Alan Modra <amodra@gmail.com>
50915 Correct nds32 readelf reloc numbers
50916 * readelf.c (is_32bit_abs_reloc, is_16bit_abs_reloc): Comment fixes.
50917 (is_none_reloc): Correct nds32 reloc numbers.
50919 2022-04-09 GDB Administrator <gdbadmin@sourceware.org>
50921 Automatic date update in version.in
50923 2022-04-08 Fangrui Song <i@maskray.me>
50925 gas: Port "copy st_size only if unset" to aarch64 and riscv
50926 And disable the new test gas/elf/size.s for alpha which uses its own
50927 .set, for hppa*-*-hpux* which does not allow .size before declaration.
50929 2022-04-08 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
50931 gprofng: fprintf_styled_func not inizialized for disassembler
50933 2022-04-07 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
50935 * libcollector/unwind.c: inizialize fprintf_styled_func.
50936 * src/Disasm.cc: Likewise.
50938 2022-04-08 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
50940 gprofng: zlib handling
50942 2022-04-06 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
50944 * configure.ac: Add AM_ZLIB.
50945 * src/Makefile.am: Add $(ZLIBINC) and $(ZLIB).
50946 * gprofng/src/DbeSession.h: Likewise.
50947 * configure: Regenerate.
50948 * Makefile.in: Regenerate.
50949 * doc/Makefile.in: Regenerate.
50950 * gp-display-html/Makefile.in: Regenerate.
50951 * src/Makefile.in: Regenerate.
50953 2022-04-08 Pedro Alves <pedro@palves.net>
50955 gdb: Avoid undefined shifts, fix Go shifts
50956 I noticed that a build of GDB with GCC + --enable-ubsan, testing
50957 against GDBserver showed this GDB crash:
50959 (gdb) PASS: gdb.trace/trace-condition.exp: trace: 0x00abababcdcdcdcd << 46 == 0x7373400000000000: advance to trace begin
50961 ../../src/gdb/valarith.c:1365:15: runtime error: left shift of 48320975398096333 by 46 places cannot be represented in type 'long int'
50962 ERROR: GDB process no longer exists
50963 GDB process exited with wait status 269549 exp9 0 1
50964 UNRESOLVED: gdb.trace/trace-condition.exp: trace: 0x00abababcdcdcdcd << 46 == 0x7373400000000000: start trace experiment
50966 The problem is that, "0x00abababcdcdcdcd << 46" is an undefined signed
50967 left shift, because the result is not representable in the type of the
50968 lhs, which is signed. This actually became defined in C++20, and if
50969 you compile with "g++ -std=c++20 -Wall", you'll see that GCC no longer
50970 warns about it, while it warns if you specify prior language versions.
50972 While at it, there are a couple other situations that are undefined
50973 (and are still undefined in C++20) and result in GDB dying: shifting
50974 by a negative ammount, or by >= than the bit size of the promoted lhs.
50975 For the latter, GDB shifts using (U)LONGEST internally, so you have to
50976 shift by >= 64 bits to see it:
50978 $ gdb --batch -q -ex "p 1 << -1"
50979 ../../src/gdb/valarith.c:1365:15: runtime error: shift exponent -1 is negative
50982 $ gdb --batch -q -ex "p 1 << 64"
50983 ../../src/gdb/valarith.c:1365:15: runtime error: shift exponent 64 is too large for 64-bit type 'long int'
50986 Also, right shifting a negative value is implementation-defined
50987 (before C++20, after which it is defined). For this, I chose to
50988 change nothing in GDB other than adding tests, as I don't really know
50989 whether we need to do anything. AFAIK, most implementations do an
50990 arithmetic right shift, and it may be we don't support any host or
50991 target that behaves differently. Plus, this becomes defined in C++20
50992 exactly as arithmetic right shift.
50994 Compilers don't error out on such shifts, at best they warn, so I
50995 think GDB should just continue doing the shifts anyhow too.
50999 - Adjust scalar_binop to avoid the undefined paths, either by adding
51000 explicit result paths, or by casting the lhs of the left shift to
51001 unsigned, as appropriate.
51003 For the shifts by a too-large count, I made the result be what you'd
51004 get if you split the large count in a series of smaller shifts.
51007 Left shift, positive or negative lhs:
51010 => V << 16 << 16 << 16 << 16
51013 Right shift, positive lhs:
51016 => Vpos >> 16 >> 16 >> 16 >> 16
51019 Right shift, negative lhs:
51022 => Vneg >> 16 >> 16 >> 16 >> 16
51025 This is actually Go's semantics (the compiler really emits
51026 instructions to make it so that you get 0 or -1 if you have a
51027 too-large shift). So for that language GDB does the shift and
51028 nothing else. For other C-like languages where such a shift is
51029 undefined, GDB warns in addition to performing the shift.
51031 For shift by a negative count, for Go, this is a hard error. For
51032 other languages, since their compilers only warn, I made GDB warn
51033 too. The semantics I chose (we're free to pick them since this is
51034 undefined behavior) is as-if you had shifted by the count cast to
51035 unsigned, thus as if you had shifted by a too-large count, thus the
51036 same as the previous scenario illustrated above.
51040 (gdb) set language go
51052 left shift count is negative
51054 (gdb) set language c
51056 warning: right shift count >= width of type
51059 warning: left shift count >= width of type
51062 warning: left shift count is negative
51065 warning: right shift count is negative
51068 - The warnings' texts are the same as what GCC prints.
51070 - Add comprehensive tests in a new gdb.base/bitshift.exp testcase, so
51071 that we exercise all these scenarios.
51073 Change-Id: I8bcd5fa02de3114b7ababc03e65702d86ec8d45d
51075 2022-04-08 Pedro Alves <pedro@palves.net>
51077 Fix undefined behavior in the Fortran, Go and Pascal number parsers
51078 This commit ports these two fixes to the C parser:
51080 commit ebf13736b42af47c9907b5157c8e80c78dbe00e1
51081 CommitDate: Thu Sep 4 21:46:28 2014 +0100
51083 parse_number("0") reads uninitialized memory
51085 commit 20562150d8a894bc91657c843ee88c508188e32e
51086 CommitDate: Wed Oct 3 15:19:06 2018 -0600
51088 Avoid undefined behavior in parse_number
51090 ... to the Fortran, Go, and Fortran number parsers, fixing the same
51093 Also add a new testcase that exercises printing 0xffffffffffffffff
51094 (max 64-bit) in all languages, which crashes a GDB built with UBsan
51097 I moved get_set_option_choices out of all-architectures.exp.tcl to
51098 common code to be able to extract all the supported languages. I did
51099 a tweak to it to generalize it a bit -- you now have to pass down the
51100 "set" part of the command as well. This is so that the proc can be
51101 used with "maintenance set" commands as well in future.
51103 Change-Id: I8e8f2fdc1e8407f63d923c26fd55d98148b9e16a
51105 2022-04-08 Nick Clifton <nickc@redhat.com>
51107 Debug info for function in Windows PE binary on wrong instruction
51109 * coffgen.c (coff_find_nearest_line_with_names): Fix typo
51110 retrieving saved bias.
51112 2022-04-08 Simon Marchi <simon.marchi@efficios.com>
51114 Pass PKG_CONFIG_PATH down from top-level Makefile
51115 [Sending to binutils, gdb-patches and gcc-patches, since it touches the
51116 top-level Makefile/configure]
51118 I have my debuginfod library installed in a non-standard location
51119 (/opt/debuginfod), which requires me to set
51120 PKG_CONFIG_PATH=/opt/debuginfod/lib/pkg-config. If I just set it during
51123 $ PKG_CONFIG_PATH=/opt/debuginfod/lib/pkg-config ./configure --with-debuginfod
51128 $ ./configure --with-debuginfod PKG_CONFIG_PATH=/opt/debuginfod/lib/pkg-config
51131 Then PKG_CONFIG_PATH is only present (and ignored) during the top-level
51132 configure. When running make (which runs gdb's and binutils'
51133 configure), PKG_CONFIG_PATH is not set, which results in their configure
51134 script not finding the library:
51136 checking for libdebuginfod >= 0.179... no
51137 configure: error: "--with-debuginfod was given, but libdebuginfod is missing or unusable."
51139 Change the top-level configure/Makefile system to capture the value
51140 passed when configuring the top-level and pass it down to
51141 subdirectories (similar to CFLAGS, LDFLAGS, etc).
51143 I don't know much about the top-level build system, so I really don't
51144 know if I did this correctly. The changes are:
51146 - Use AC_SUBST(PKG_CONFIG_PATH) in configure.ac, so that
51147 @PKG_CONFIG_PATH@ gets replaced with the actual PKG_CONFIG_PATH value
51148 in config files (i.e. Makefile)
51149 - Add a PKG_CONFIG_PATH Makefile variable in Makefile.tpl, initialized
51150 to @PKG_CONFIG_PATH@
51151 - Add PKG_CONFIG_PATH to HOST_EXPORTS in Makefile.tpl, which are the
51152 variables set when running the sub-configures
51154 I initially added PKG_CONFIG_PATH to flags_to_pass, in Makefile.def, but
51155 I don't think it's needed. AFAIU, this defines the flags to pass down
51156 when calling "make" in subdirectories. We only need PKG_CONFIG_PATH to
51157 be passed down during configure. After that, it's captured in
51158 gdb/config.status, so even if a "make" causes a re-configure later
51159 (because gdb/configure has changed, for example), the PKG_CONFIG_PATH
51160 value will be remembered.
51164 * configure.ac: Add AC_SUBST(PKG_CONFIG_PATH).
51165 * configure: Re-generate.
51166 * Makefile.tpl (HOST_EXPORTS): Pass PKG_CONFIG_PATH.
51167 (PKG_CONFIG_PATH): New.
51168 * Makefile.in: Re-generate.
51170 Change-Id: I91138dfca41c43b05e53e445f62e4b27882536bf
51172 2022-04-08 Simon Marchi <simon.marchi@efficios.com>
51174 gdb/testsuite: use nopie in gdb.dwarf2/dw2-inline-param.exp
51175 I see this failure:
51178 Starting program: /home/smarchi/build/binutils-gdb/gdb/testsuite/outputs/gdb.dwarf2/dw2-inline-param/dw2-inline-param ^M
51180 Cannot insert breakpoint 1.^M
51181 Cannot access memory at address 0x113b^M
51183 (gdb) FAIL: gdb.dwarf2/dw2-inline-param.exp: runto: run to *0x113b
51185 The test loads the binary in GDB, grabs the address of a symbol, strips
51186 the binary, reloads it in GDB, runs the program, and then tries to place
51187 a breakpoint at that address. The problem is that the binary is built
51188 as position independent, so the address GDB grabs in the first place
51189 isn't where the code ends up after running.
51191 Fix this by linking the binary as non-position-independent. The
51192 alternative would be to compute the relocated address where to place the
51193 breakpoint, but that's not very straightforward, unfortunately.
51195 I was confused for a while, I was trying to load the binary in GDB
51196 manually to get the symbol address, but GDB was telling me the symbol
51197 could not be found. Meanwhile, it clearly worked in gdb.log. The thing
51198 is that GDB strips the binary in-place, so we don't have access to the
51199 intermediary binary with symbols. Change the test to output the
51200 stripped binary to a separate file instead.
51202 Change-Id: I66c56293df71b1ff49cf748d6784bd0e935211ba
51204 2022-04-08 Alan Modra <amodra@gmail.com>
51206 gdb maintainer commit rights
51207 Formalise what ought to be obvious. The top level of the binutils-gdb
51208 repository isn't owned by binutils.
51210 * MAINTAINERS: Spelling fix. GDB global maintainer rights.
51212 2022-04-08 Bernhard Heckel <bernhard.heckel@intel.com>
51213 Nils-Christian Kempke <nils-christian.kempke@intel.com>
51215 gdb/fortran: print fortran extended types with ptype
51216 Add the print of the base-class of an extended type to the output of
51217 ptype. This requires the Fortran compiler to emit DW_AT_inheritance
51218 for the extended type.
51220 2022-04-08 Bernhard Heckel <bernhard.heckel@intel.com>
51221 Nils-Christian Kempke <nils-christian.kempke@intel.com>
51223 gdb/fortran: add support for accessing fields of extended types
51224 Fortran 2003 supports type extension. This patch allows access
51225 to inherited members by using their fully qualified name as described
51226 in the Fortran standard.
51228 In doing so the patch also fixes a bug in GDB when trying to access the
51229 members of a base class in a derived class via the derived class' base
51232 This patch fixes PR22497 and PR26373 on GDB side.
51234 Using the example Fortran program from PR22497
51243 type, extends(my_type) :: extended_type
51244 end type extended_type
51246 type(my_type) :: foo
51247 type(extended_type) :: bar
51256 and running this with GDB and setting a BP at 17:
51259 (gdb) p bar%my_type
51260 A syntax error in expression, near `my_type'.
51262 There is no member named my_int.
51263 (gdb) p bar%my_type%my_int
51264 A syntax error in expression, near `my_type%my_int'.
51266 $1 = ( my_type = ( my_int = 1 ) )
51269 (gdb) p bar%my_type
51270 $1 = ( my_int = 1 )
51272 $2 = 1 # this line requires DW_TAG_inheritance to work
51273 (gdb) p bar%my_type%my_int
51276 $4 = ( my_type = ( my_int = 1 ) )
51278 In the above example "p bar%my_int" requires the compiler to emit
51279 information about the inheritance relationship between extended_type
51280 and my_type which gfortran and flang currently do not de. The
51281 respective issue gcc/49475 has been put as kfail.
51283 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=26373
51284 https://sourceware.org/bugzilla/show_bug.cgi?id=22497
51286 2022-04-08 Nils-Christian Kempke <nils-christian.kempke@intel.com>
51288 gdb: add Nils-Christian Kempke to gdb/MAINTAINERS
51290 2022-04-08 Simon Marchi <simon.marchi@polymtl.ca>
51292 gdb: change file_file_name to return an std::string
51293 Straightforward change, return an std::string instead of a
51294 gdb::unique_xmalloc_ptr<char>. No behavior change expected.
51296 Change-Id: Ia5e94c94221c35f978bb1b7bdffbff7209e0520e
51298 2022-04-08 GDB Administrator <gdbadmin@sourceware.org>
51300 Automatic date update in version.in
51302 2022-04-07 Andrew Burgess <aburgess@redhat.com>
51304 gdb/fortran: fix fetching assumed rank array content
51307 commit df7a7bdd9766adebc6b117c31bc617d81c1efd43
51308 Date: Thu Mar 17 18:56:23 2022 +0000
51310 gdb: add support for Fortran's ASSUMED RANK arrays
51312 Added support for Fortran assumed rank arrays. Unfortunately, this
51313 commit contained a bug that means though GDB can correctly calculate
51314 the rank of an assumed rank array, GDB can't fetch the contents of an
51315 assumed rank array.
51317 The history of this patch can be seen on the mailing list here:
51319 https://sourceware.org/pipermail/gdb-patches/2022-January/185306.html
51321 The patches that were finally committed can be found here:
51323 https://sourceware.org/pipermail/gdb-patches/2022-March/186906.html
51325 The original patches did support fetching the array contents, it was
51326 only the later series that introduced the regression.
51328 The problem is that when calculating the array rank the result is a
51329 count of the number of ranks, i.e. this is a 1 based result, 1, 2, 3,
51332 In contrast, when computing the details of any particular rank the
51333 value passed to the DWARF expression evaluator should be a 0 based
51334 rank offset, i.e. a 0 based number, 0, 1, 2, etc.
51336 In the patches that were originally merged, this was not the case, and
51337 we were passing the 1 based rank number to the expression evaluator,
51338 e.g. passing 1 when we should pass 0, 2 when we should pass 1, etc.
51339 As a result the DWARF expression evaluator was reading the
51340 wrong (undefined) memory, and returning garbage results.
51342 In this commit I have extended the test case to cover checking the
51343 array contents, I've then ensured we make use of the correct rank
51344 value, and extended some comments, and added or adjusted some asserts
51347 2022-04-07 Simon Marchi <simon.marchi@efficios.com>
51349 gdb/testsuite: add "macros" option to gdb_compile
51350 Make gdb_compile handle a new "macros" option, which makes it pass the
51351 appropriate flag to make the compiler include macro information in the
51352 debug info. This will help simplify tests using macros, reduce
51353 redundant code, and make it easier to add support for a new compiler.
51355 Right now it only handles clang specially (using -fdebug-macro) and
51356 falls back to -g3 otherwise (which works for gcc). Other compilers can
51357 be added as needed.
51359 There are some tests that are currently skipped if the compiler is nor
51360 gcc nor clang. After this patch, the tests will attempt to run (the -g3
51361 fall back will be used). That gives a chance to people using other
51362 compilers to notice something is wrong and maybe add support for their
51363 compiler. If it is needed to support a compiler that doesn't have a way
51364 to include macro information, then we can always introduce a
51365 "skip_macro_tests" that can be used to skip over them.
51367 Change-Id: I50cd6ab1bfbb478c1005486408e214b551364c9b
51369 2022-04-07 Simon Marchi <simon.marchi@polymtl.ca>
51371 gdb: remove subfile::buildsym_compunit field
51372 It is only set, never used.
51374 Change-Id: Ia46ed2f9da243b0ccfc4588c1b57be2a0f3939de
51376 2022-04-07 Tom de Vries <tdevries@suse.de>
51378 [gdb/testsuite] Make gdb.base/annota1.exp more robust
51379 On openSUSE Tumbleweed I run into:
51381 FAIL: gdb.base/annota1.exp: run until main breakpoint (timeout)
51384 The problem is that the libthread_db message occurs at a location where it's
51387 Starting program: outputs/gdb.base/annota1/annota1 ^M
51391 ^Z^Zframes-invalid^M
51392 [Thread debugging using libthread_db enabled]^M
51393 Using host libthread_db library "/lib64/libthread_db.so.1".^M
51395 ^Z^Zbreakpoints-invalid^M
51399 Fix this by making the matching more robust:
51400 - rewrite the regexp such that each annotation is on a single line,
51401 starting with \r\n\032\032 and ending with \r\n
51402 - add a regexp variable optional_re, that matches all possible optional
51403 output, and use it as a separator in the first part of the regexp
51405 Tested on x86_64-linux.
51407 2022-04-07 Simon Marchi <simon.marchi@polymtl.ca>
51409 gdb/testsuite/dwarf: simplify line number program syntax
51410 By calling `uplevel $body` in the program proc (a pattern we use at many
51411 places), we can get rid of curly braces around each line number program
51412 directive. That seems like a nice small improvement to me.
51414 Change-Id: Ib327edcbffbd4c23a08614adee56c12ea25ebc0b
51416 2022-04-07 Simon Marchi <simon.marchi@polymtl.ca>
51418 gdb/testsuite/dwarf: remove two unused variables
51419 These variables seem to be unused, remove them.
51421 Change-Id: I7d613d9d35735930ee78b2c348943c73a702afbb
51423 2022-04-07 Simon Marchi <simon.marchi@polymtl.ca>
51425 gdb: remove symtab::pspace
51426 Same idea as previous patch, but for symtab::pspace.
51428 Change-Id: I1023abe622bea75ef648c6a97a01b53775d4104d
51430 2022-04-07 Simon Marchi <simon.marchi@polymtl.ca>
51432 gdb: remove symtab::objfile
51433 Same idea as previous patch, but for symtab::objfile. I find
51434 it clearer without this wrapper, as it shows that the objfile is
51435 common to all symtabs of a given compunit. Otherwise, you could think
51436 that each symtab (of a given compunit) can have a specific objfile.
51438 Change-Id: Ifc0dbc7ec31a06eefa2787c921196949d5a6fcc6
51440 2022-04-07 Simon Marchi <simon.marchi@polymtl.ca>
51442 gdb: remove symtab::blockvector
51443 symtab::blockvector is a wrapper around compunit_symtab::blockvector.
51444 It is a bit misleadnig, as it gives the impression that a symtab has a
51445 blockvector. Remove it, change all users to fetch the blockvector
51446 through the compunit instead.
51448 Change-Id: Ibd062cd7926112a60d52899dff9224591cbdeebf
51450 2022-04-07 Simon Marchi <simon.marchi@efficios.com>
51452 gdb: remove symtab::dirname
51453 I think the symtab::dirname method is bogus, or at least very
51454 misleading. It makes you think that it returns the directory that was
51455 used to find that symtab's file during compilation (i.e. the directory
51456 the file refers to in the DWARF line header file table), or the
51457 directory part of the symtab's filename maybe. In fact, it returns the
51458 compilation unit's directory, which is the CWD of the compiler, at
51459 compilation time. At least for DWARF, if the symtab's filename is
51460 relative, it will be relative to that directory. But if the symtab's
51461 filename is absolute, then the directory returned by symtab::dirname has
51462 nothing to do with the symtab's filename.
51464 Remove symtab::dirname to avoid this confusion, change all users to
51465 fetch the same information through the compunit. At least, it will be
51466 clear that this is a compunit property, not a symtab property.
51468 Change-Id: I2894c3bf3789d7359a676db3c58be2c10763f5f0
51470 2022-04-07 Simon Marchi <simon.marchi@polymtl.ca>
51472 gdb/testsuite: make gdb_breakpoint and runto take a linespec
51473 Change gdb_breakpoint to accept a linespec, not just a function. In
51474 fact, no behavior changes are necessary, this only changes the parameter
51475 name and documentation. Change runto as well, since the two are so
51476 close (runto forwards all its arguments to gdb_breakpoint).
51478 I wrote this for a downstrean GDB port, but thought it could be
51479 useful upstream, eventually, even though not callers take advantage of
51482 Change-Id: I08175fd444d5a60df90fd9985e1b5dfd87c027cc
51484 2022-04-07 Andrew Burgess <aburgess@redhat.com>
51486 gdb: update comments throughout reggroups.{c,h} files
51487 This commit updates the comments in the gdb/reggroups.{c,h} files.
51488 Fill in some missing comments, correct a few comments that were not
51489 clear, and where we had comments duplicated between .c and .h files,
51490 update the .c to reference the .h.
51492 No user visible changes after this commit.
51494 2022-04-07 Andrew Burgess <aburgess@redhat.com>
51496 gdb: move struct reggroup into reggroups.h header
51497 Move 'struct reggroup' into the reggroups.h header. Remove the
51498 reggroup_name and reggroup_type accessor functions, and just use the
51499 name/type member functions within 'struct reggroup', update all uses
51500 of these removed functions.
51502 There should be no user visible changes after this commit.
51504 2022-04-07 Andrew Burgess <aburgess@redhat.com>
51506 gdb: convert reggroup to a C++ class with constructor, etc
51507 Convert the 'struct reggroup' into a real class, with a constructor
51508 and getter methods.
51510 There should be no user visible changes after this commit.
51512 2022-04-07 Andrew Burgess <aburgess@redhat.com>
51514 gdb: make the pre-defined register groups const
51515 Convert the 7 global, pre-defined, register groups const, and fix the
51516 fall out (a minor tweak required in riscv-tdep.c).
51518 There should be no user visible changes after this commit.
51520 2022-04-07 Andrew Burgess <aburgess@redhat.com>
51522 gdb: more 'const' in gdb/reggroups.{c,h}
51523 Convert the reggroup_new and reggroup_gdbarch_new functions to return
51524 a 'const regggroup *', and fix up all the fallout.
51526 There should be no user visible changes after this commit.
51528 2022-04-07 Andrew Burgess <aburgess@redhat.com>
51530 gdb: remove reggroup_next and reggroup_prev
51531 Add a new function gdbarch_reggroups that returns a reference to a
51532 vector containing all the reggroups for an architecture.
51534 Make use of this function throughout GDB instead of the existing
51535 reggroup_next and reggroup_prev functions.
51537 Finally, delete the reggroup_next and reggroup_prev functions.
51539 Most of these changes are pretty straight forward, using range based
51540 for loops instead of the old style look using reggroup_next. There
51541 are two places where the changes are less straight forward.
51543 In gdb/python/py-registers.c, the register group iterator needed to
51544 change slightly. As the iterator is tightly coupled to the gdbarch, I
51545 just fetch the register group vector from the gdbarch when needed, and
51546 use an index counter to find the next item from the vector when
51549 In gdb/tui/tui-regs.c the tui_reg_next and tui_reg_prev functions are
51550 just wrappers around reggroup_next and reggroup_prev respectively.
51551 I've just inlined the logic of the old functions into the tui
51552 functions. As the tui function had its own special twist (wrap around
51553 behaviour) I think this is OK.
51555 There should be no user visible changes after this commit.
51557 2022-04-07 Andrew Burgess <aburgess@redhat.com>
51559 gdb: convert reggroups to use a std::vector
51560 Replace manual linked list with a std::vector. This commit doesn't
51561 change the reggroup_next and reggroup_prev API, but that will change
51564 This commit is focused on the minimal changes needed to manage the
51565 reggroups using a std::vector, without changing the API exposed by the
51568 There should be no user visible changes after this commit.
51570 2022-04-07 Andrew Burgess <aburgess@redhat.com>
51572 gdb: always add the default register groups
51573 There's a set of 7 default register groups. If we don't add any
51574 gdbarch specific register groups during gdbarch initialisation, then
51575 when we iterate over the register groups using reggroup_next and
51576 reggroup_prev we will make use of these 7 default groups. See the use
51577 of default_groups in gdb/reggroups.c for details on this.
51579 However, if the gdbarch adds its own groups during gdbarch
51580 initialisation, then these groups will be used in preference to the
51583 A problem arises though if the particular architecture makes use of
51584 the target description mechanism. If the default target
51585 description(s) (i.e. those internal to GDB that are used when the user
51586 doesn't provide their own) don't mention any additional register
51587 groups then the default register groups will be used.
51589 But if the target description does mention additional groups then the
51590 default groups are not used, and instead, the groups from the target
51591 description are used.
51593 The problem with this is that what usually happens is that the target
51594 description will mention additional groups, e.g. groups for special
51595 registers. Most architectures that use target descriptions work
51596 around this by adding all (or most) of the default register groups in
51597 all cases. See i386_add_reggroups, aarch64_add_reggroups,
51598 riscv_add_reggroups, xtensa_add_reggroups, and others.
51600 In this patch, my suggestion is that we should just add the default
51601 register groups for every architecture, always. This change is in
51604 All the remaining changes are me updating the various architectures to
51605 not add the default groups themselves.
51607 So, where will this change be visible to the user? I think the
51608 following commands will possibly change:
51610 * info registers / info all-registers:
51612 The user can provide a register group to these commands. For example,
51613 on csky, we previously never added the 'vector' group. Now, as a
51614 default group, this will be available, but (presumably) will not
51615 contain any registers. I don't think this is necessarily a bad
51616 thing, there's something to be said for having some consistent
51617 defaults available. There are other architectures that didn't add
51618 all 7 of the defaults, which will now have gained additional groups.
51620 * maint print reggroups
51622 This prints the set of all available groups. As a maintenance
51623 command I'm less concerned with the output changing here.
51624 Obviously, for the architectures that didn't previously add all the
51625 defaults, this list just got bigger.
51627 * maint print register-groups
51629 This prints all the registers, and the groups they are in. If the
51630 defaults were not previously being added then a register (obviously)
51631 can't appear in one of the default groups. Now the groups are
51632 available then registers might be in more groups than previously.
51633 However, this is again a maintenance command, so I'm less concerned
51634 about this changing.
51636 2022-04-07 Andrew Burgess <aburgess@redhat.com>
51638 gdb/tui: fix 'tui reg next/prev' command when data window is hidden
51641 $ gdb -q executable
51644 ... tui windows are now displayed ...
51647 At this point the data (register) window should be displayed, but will
51648 contain the message 'Register Values Unavailable', and at the console
51649 you'll see the message "unknown register group 'next'".
51651 The same happens with 'tui reg prev' (but the error message is
51652 slightly different).
51654 At this point you can continue to use 'tui reg next' and/or 'tui reg
51655 prev' and you'll keep getting the error message.
51657 The problem is that when the data (register) window is first
51658 displayed, it's current register group is nullptr. As a consequence
51659 tui_reg_next and tui_reg_prev (tui/tui-regs.c) will always just return
51660 nullptr, which triggers an error in tui_reg_command.
51662 In this commit I change tui_reg_next and tui_reg_prev so that they
51663 instead return the first and last register group respectively if the
51664 current register group is nullptr.
51666 So, after this, using 'tui reg next' will (in the above case) show the
51667 first register group, while 'tui reg prev' will display the last
51670 2022-04-07 Andrew Burgess <aburgess@redhat.com>
51672 gdb/tui: avoid theoretical bug with 'tui reg' command
51673 While looking at the 'tui reg' command as part of another patch, I
51674 spotted a theoretical bug.
51676 The 'tui reg' command takes the name of a register group, but also
51677 handles partial register group matches, though the partial match has to
51678 be unique. The current command logic goes:
51680 With the code as currently written, if a target description named a
51681 register group either 'prev' or 'next' then GDB would see this as an
51682 ambiguous register name, and refuse to switch groups.
51684 Naming a register group 'prev' or 'next' seems pretty unlikely, but,
51685 by adding a single else block we can prevent this problem.
51687 Now, if there's a 'prev' or 'next' register group, the user will not
51688 be able to select the group directly, the 'prev' and 'next' names will
51689 always iterate through the available groups instead. But at least the
51690 user could select their groups by iteration, rather than direct
51693 2022-04-07 Andrew Burgess <aburgess@redhat.com>
51695 gdb: have reggroup_find return a const
51696 Update reggroup_find to return a const reggroup *.
51698 There are other function in gdb/reggroup.{c,h} files that could
51699 benefit from returning const, these will be updated in later commits.
51701 There should be no user visible changes after this commit.
51703 2022-04-07 Andrew Burgess <aburgess@redhat.com>
51705 gdb: use 'const reggroup *' in python/py-registers.c file
51706 Convert uses of 'struct reggroup *' in python/py-registers.c to be
51709 There should be no user visible changes after this commit.
51711 2022-04-07 Andrew Burgess <aburgess@redhat.com>
51713 gdb: switch to using 'const reggroup *' in tui-regs.{c,h}
51714 Make uses of 'reggroup *' const throughout tui-regs.{c,h}.
51716 There should be no user visible changes after this commit.
51718 2022-04-07 Andrew Burgess <aburgess@redhat.com>
51720 gdb: make gdbarch_register_reggroup_p take a const reggroup *
51721 Change gdbarch_register_reggroup_p to take a 'const struct reggroup *'
51722 argument. This requires a change to the gdb/gdbarch-components.py
51723 script, regeneration of gdbarch.{c,h}, and then updates to all the
51724 architectures that implement this method.
51726 There should be no user visible changes after this commit.
51728 2022-04-07 Andrew Burgess <aburgess@redhat.com>
51730 gdb: add some const in gdb/reggroups.c
51731 This commit makes the 'struct reggroup *' argument const for the
51732 following functions:
51739 There are other places that could benefit from const in the
51740 reggroup.{c,h} files, but these will be changing in further commits.
51742 There should be no user visible changes after this commit.
51744 2022-04-07 Andrew Burgess <aburgess@redhat.com>
51746 gdb: don't try to use readline before it's initialized
51747 While working on a different patch, I triggered an assertion from the
51748 initialize_current_architecture code, specifically from one of
51749 the *_gdbarch_init functions in a *-tdep.c file. This exposes a
51750 couple of issues with GDB.
51752 This is easy enough to reproduce by adding 'gdb_assert (false)' into a
51753 suitable function. For example, I added a line into i386_gdbarch_init
51754 and can see the following issue.
51756 I start GDB and immediately hit the assert, the output is as you'd
51757 expect, except for the very last line:
51759 $ ./gdb/gdb --data-directory ./gdb/data-directory/
51760 ../../src.dev-1/gdb/i386-tdep.c:8455: internal-error: i386_gdbarch_init: Assertion `false' failed.
51761 A problem internal to GDB has been detected,
51762 further debugging may prove unreliable.
51763 ----- Backtrace -----
51765 ---------------------
51766 ../../src.dev-1/gdb/i386-tdep.c:8455: internal-error: i386_gdbarch_init: Assertion `false' failed.
51767 A problem internal to GDB has been detected,
51768 further debugging may prove unreliable.
51769 Quit this debugging session? (y or n) ../../src.dev-1/gdb/ser-event.c:212:16: runtime error: member access within null pointer of type 'struct serial'
51771 Something goes wrong when we try to query the user. Note, I
51772 configured GDB with --enable-ubsan, I suspect that without this the
51773 above "error" would actually just be a crash.
51775 The backtrace from ser-event.c:212 looks like this:
51778 #0 serial_event_clear (event=0x675c020) at ../../src/gdb/ser-event.c:212
51779 #1 0x0000000000769456 in invoke_async_signal_handlers () at ../../src/gdb/async-event.c:211
51780 #2 0x000000000295049b in gdb_do_one_event () at ../../src/gdbsupport/event-loop.cc:194
51781 #3 0x0000000001f015f8 in gdb_readline_wrapper (
51782 prompt=0x67135c0 "../../src/gdb/i386-tdep.c:8455: internal-error: i386_gdbarch_init: Assertion `false' failed.\nA problem internal to GDB has been detected,\nfurther debugging may prove unreliable.\nQuit this debugg"...)
51783 at ../../src/gdb/top.c:1141
51784 #4 0x0000000002118b64 in defaulted_query(const char *, char, typedef __va_list_tag __va_list_tag *) (
51785 ctlstr=0x2e4eb68 "%s\nQuit this debugging session? ", defchar=0 '\000', args=0x7fffffffa6e0)
51786 at ../../src/gdb/utils.c:934
51787 #5 0x0000000002118f72 in query (ctlstr=0x2e4eb68 "%s\nQuit this debugging session? ")
51788 at ../../src/gdb/utils.c:1026
51789 #6 0x00000000021170f6 in internal_vproblem(internal_problem *, const char *, int, const char *, typedef __va_list_tag __va_list_tag *) (problem=0x6107bc0 <internal_error_problem>, file=0x2b976c8 "../../src/gdb/i386-tdep.c",
51790 line=8455, fmt=0x2b96d7f "%s: Assertion `%s' failed.", ap=0x7fffffffa8e8) at ../../src/gdb/utils.c:417
51791 #7 0x00000000021175a0 in internal_verror (file=0x2b976c8 "../../src/gdb/i386-tdep.c", line=8455,
51792 fmt=0x2b96d7f "%s: Assertion `%s' failed.", ap=0x7fffffffa8e8) at ../../src/gdb/utils.c:485
51793 #8 0x00000000029503b3 in internal_error (file=0x2b976c8 "../../src/gdb/i386-tdep.c", line=8455,
51794 fmt=0x2b96d7f "%s: Assertion `%s' failed.") at ../../src/gdbsupport/errors.cc:55
51795 #9 0x000000000122d5b6 in i386_gdbarch_init (info=..., arches=0x0) at ../../src/gdb/i386-tdep.c:8455
51796 (More stack frames follow...)
51798 It turns out that the problem is that the async event handler
51799 mechanism has been invoked, but this has not yet been initialized.
51801 If we look at gdb_init (in gdb/top.c) we can indeed see the call to
51802 gdb_init_signals is after the call to initialize_current_architecture.
51804 If I reorder the calls, moving gdb_init_signals earlier, then the
51805 initial error is resolved, however, things are still broken. I now
51806 see the same "Quit this debugging session? (y or n)" prompt, but when
51807 I provide an answer and press return GDB immediately crashes.
51809 So what's going on now? The next problem is that the call_readline
51810 field within the current_ui structure is not initialized, and this
51811 callback is invoked to process the reply I entered.
51813 The problem is that call_readline is setup as a result of calling
51814 set_top_level_interpreter, which is called from captured_main_1.
51815 Unfortunately, set_top_level_interpreter is called after gdb_init is
51818 I wondered how to solve this problem for a while, however, I don't
51819 know if there's an easy "just reorder some lines" solution here.
51820 Looking through captured_main_1 there seems to be a bunch of
51821 dependencies between printing various things, parsing config files,
51822 and setting up the interpreter. I'm sure there is a solution hiding
51823 in there somewhere.... I'm just not sure I want to spend any longer
51828 I propose a simpler solution, more of a hack/work-around. In utils.c
51829 we already have a function filtered_printing_initialized, this is
51830 checked in a few places within internal_vproblem. In some of these
51831 cases the call gates whether or not GDB will query the user.
51833 My proposal is to add a new readline_initialized function, which
51834 checks if the current_ui has had readline initialized yet. If this is
51835 not the case then we should not attempt to query the user.
51837 After this change GDB prints the error message, the backtrace, and
51838 then aborts (including dumping core). This actually seems pretty sane
51839 as, if GDB has not yet made it through the initialization then it
51840 doesn't make much sense to allow the user to say "no, I don't want to
51841 quit the debug session" (I think).
51843 2022-04-07 Luis Machado <luis.machado@arm.com>
51845 Recognize the NT_ARM_SYSTEM_CALL register set
51846 Update binutils to recognize the NT_ARM_SYSTEM_CALL set that is dumped by
51847 Linux to core files.
51849 2022-04-07 Mark Harmstone <mark@harmstone.com>
51851 Add support for COFF secidx relocations
51852 bfd * coff-i386.c (in_reloc_p): Add R_SECTION.
51853 (howto_table): Add R_SECTION.
51854 (coff_pe_i386_relocation_section): Add support for R_SECTION.
51855 (coff_i386_reloc_type_lookup): Add support for
51856 BFD_RELOC_16_SECCIDX.
51857 * coff-x86_64.c (in_reloc_p): Add R_SECTION.
51858 (howto_table): Add R_SECTION.
51859 (coff_pe_amd64_relocation_section): Add support for R_SECTION.
51860 (coff_amd64_reloc_type_lookup): Add support for
51861 BFD_RELOC_16_SECCIDX.
51862 * reloc.c: Add BFD_RELOC_16_SECIDX.
51863 * bfd-in2.h: Regenerate.
51864 * libbfd.h: Regenerate.
51866 gas * config/tc-i386.c (pe_directive_secidx): New function.
51867 (md_pseudo_table): Add support for secidx.
51868 (x86_cons_fix_new): Likewise.
51869 (tc_gen_reloc): Likewise.
51870 * expr.c (op_rank): Add O_secidx.
51871 * expr.h (operatorT): Likewise.
51872 * symbols.c (resolve_symbol_value): Add support for O_secidx.
51873 * testsuite/gas/i386/secidx.s: New test source file.
51874 * testsuite/gas/i386/secidx.d: New test driver file.
51875 * testsuite/gas/i386/i386.exp: Run new test.
51877 include * coff/i386.h: Define R_SECTION.
51878 * coff/x86_64.h: Likewise.
51880 ld * testsuite/ld-pe/secidx1.s: New test source file.
51881 * testsuite/ld-pe/secidx2.s: New test source file.
51882 * testsuite/ld-pe/secidx.d: New test driver file.
51883 * testsuite/ld-pe/secidx_64.d: New test driver file.
51884 * testsuite/ld-pe/pe.exp: Add new tests.
51886 2022-04-07 Jan Beulich <jbeulich@suse.com>
51888 gas/Dwarf: record functions
51889 To help tools like addr2line looking up function names, in particular
51890 when dealing with e.g. PE/COFF binaries (linked from ELF objects), where
51891 there's no ELF symbol table to fall back to, emit minimalistic
51892 information for functions marked as such and having their size
51895 Notes regarding the restriction to (pure) ELF:
51896 - I realize this is a layering violation; I don't see how to deal with
51897 that in a better way.
51898 - S_GET_SIZE(), when OBJ_MAYBE_ELF is defined, looks wrong: Unlike
51899 S_SET_SIZE() it does not check whether the hook is NULL.
51900 - symbol_get_obj(), when OBJ_MAYBE_ELF is defined, looks unusable, as
51901 its return type can only ever be one object format's type (and this
51902 may then not be ELF's).
51904 The new testcases are limited to x86 because I wanted to include the
51905 case where function size can't be determined yet at the time Dwarf2 info
51906 is generated. As .nops gains support by further targets, they could also
51907 be added here then (with, as necessary, expecations suitably relaxed to
51908 cover for insn size differences).
51910 2022-04-07 Jan Beulich <jbeulich@suse.com>
51912 Arm64: arrange for line number emission for .inst
51913 Just like insns encoded the more conventional way these should have line
51914 number info associated with them.
51916 Arm32: arrange for line number emission for .inst
51917 Just like insns encoded the more conventional way these should have line
51918 number info associated with them.
51920 RISC-V: add testcase to check line number emission for .insn
51921 Since no such test looks to exist, derive one from insn.s.
51923 2022-04-07 Andreas Krebbel <krebbel@linux.ibm.com>
51925 IBM zSystems: Add support for z16 as CPU name.
51926 So far z16 was identified as arch14. After the machine has been
51927 announced we can now add the real name.
51931 * config/tc-s390.c (s390_parse_cpu): Add z16 as alternate CPU
51933 * doc/as.texi: Add z16 and arch14 to CPU string list.
51934 * doc/c-s390.texi: Add z16 to CPU string list.
51938 * s390-mkopc.c (main): Enable z16 as CPU string in the opcode
51941 2022-04-07 GDB Administrator <gdbadmin@sourceware.org>
51943 Automatic date update in version.in
51945 2022-04-06 Youling Tang <tangyouling@loongson.cn>
51947 gdb: mips: Fix the handling of complex type of function return value
51948 $ objdump -d outputs/gdb.base/varargs/varargs
51949 00000001200012e8 <find_max_float_real>:
51951 1200013b8: c7c10000 lwc1 $f1,0(s8)
51952 1200013bc: c7c00004 lwc1 $f0,4(s8)
51953 1200013c0: 46000886 mov.s $f2,$f1
51954 1200013c4: 46000046 mov.s $f1,$f0
51955 1200013c8: 46001006 mov.s $f0,$f2
51956 1200013cc: 46000886 mov.s $f2,$f1
51957 1200013d0: 03c0e825 move sp,s8
51958 1200013d4: dfbe0038 ld s8,56(sp)
51959 1200013d8: 67bd0080 daddiu sp,sp,128
51960 1200013dc: 03e00008 jr ra
51961 1200013e0: 00000000 nop
51963 From the above disassembly, we can see that when the return value of the
51964 function is a complex type and len <= 2 * MIPS64_REGSIZE, the return value
51965 will be passed through $f0 and $f2, so fix the corresponding processing
51966 in mips_n32n64_return_value().
51968 $ make check RUNTESTFLAGS='GDB=../gdb gdb.base/varargs.exp --outdir=test'
51970 Before applying the patch:
51971 FAIL: gdb.base/varargs.exp: print find_max_float_real(4, fc1, fc2, fc3, fc4)
51972 FAIL: gdb.base/varargs.exp: print find_max_double_real(4, dc1, dc2, dc3, dc4)
51974 # of expected passes 9
51975 # of unexpected failures 2
51977 After applying the patch:
51978 # of expected passes 11
51981 FAIL: gdb.base/callfuncs.exp: call inferior func with struct - returns float _Complex
51983 Co-Authored-By: Maciej W. Rozycki <macro@orcam.me.uk>
51985 2022-04-06 Tom Tromey <tom@tromey.com>
51987 Use new and delete in jit.c
51988 This changes jit.c to use new and delete, rather than XCNEW. This
51989 simplifies the code a little. This was useful for another patch I'm
51990 working on, and I thought it would make sense to send it separately.
51992 Regression tested on x86-64 Fedora 34.
51994 2022-04-06 Simon Marchi <simon.marchi@efficios.com>
51996 gdb: don't copy entirely optimized out values in value_copy
51997 Bug 28980 shows that trying to value_copy an entirely optimized out
51998 value causes an internal error. The original bug report involves MI and
51999 some Python pretty printer, and is quite difficult to reproduce, but
52000 another easy way to reproduce (that is believed to be equivalent) was
52003 $ ./gdb -q -nx --data-directory=data-directory -ex "py print(gdb.Value(gdb.Value(5).type.optimized_out()))"
52004 /home/smarchi/src/binutils-gdb/gdb/value.c:1731: internal-error: value_copy: Assertion `arg->contents != nullptr' failed.
52006 This is caused by 5f8ab46bc691 ("gdb: constify parameter of
52007 value_copy"). It added an assertion that the contents buffer is
52008 allocated if the value is not lazy:
52010 if (!value_lazy (val))
52012 gdb_assert (arg->contents != nullptr);
52014 This was based on the comment on value::contents, which suggest that
52017 /* Actual contents of the value. Target byte-order. NULL or not
52018 valid if lazy is nonzero. */
52019 gdb::unique_xmalloc_ptr<gdb_byte> contents;
52021 However, it turns out that it can also be nullptr also if the value is
52022 entirely optimized out, for example on exit of
52023 allocate_optimized_out_value. That function creates a lazy value, marks
52024 the entire value as optimized out, and then clears the lazy flag. But
52025 contents remains nullptr.
52027 This wasn't a problem for value_copy before, because it was calling
52028 value_contents_all_raw on the input value, which caused contents to be
52029 allocated before doing the copy. This means that the input value to
52030 value_copy did not have its contents allocated on entry, but had it
52031 allocated on exit. The result value had it allocated on exit. And that
52032 we copied bytes for an entirely optimized out value (i.e. meaningless
52035 From here I see two choices:
52037 1. respect the documented invariant that contents is nullptr only and
52038 only if the value is lazy, which means making
52039 allocate_optimized_out_value allocate contents
52040 2. extend the cases where contents can be nullptr to also include
52041 values that are entirely optimized out (note that you could still
52042 have some entirely optimized out values that do have contents
52043 allocated, it depends on how they were created) and adjust
52044 value_copy accordingly
52046 Choice #1 is safe, but less efficient: it's not very useful to allocate
52047 a buffer for an entirely optimized out value. It's even a bit less
52048 efficient than what we had initially, because values coming out of
52049 allocate_optimized_out_value would now always get their contents
52052 Choice #2 would be more efficient than what we had before: giving an
52053 optimized out value without allocated contents to value_copy would
52054 result in an optimized out value without allocated contents (and the
52055 input value would still be without allocated contents on exit). But
52056 it's more risky, since it's difficult to ensure that all users of the
52057 contents (through the various_contents* accessors) are all fine with
52058 that new invariant.
52060 In this patch, I opt for choice #2, since I think it is a better
52061 direction than choice #1. #1 would be a pessimization, and if we go
52062 this way, I doubt that it will ever be revisited, it will just stay that
52065 Add a selftest to test this. I initially started to write it as a
52066 Python test (since the reproducer is in Python), but a selftest is more
52069 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28980
52070 Change-Id: I6e2f5c0ea804fafa041fcc4345d47064b5900ed7
52072 2022-04-06 Jeff Law <jeffreyalaw@gmail.com>
52074 Fix for v850e divq instruction
52075 This is the last of the correctness fixes I've been carrying around for the
52078 Like the other recent fixes, this is another case where we haven't been as
52079 careful as we should WRT host vs target types. For the divq instruction
52080 both operands are 32 bit types. Yet in the simulator code we convert them
52081 from unsigned int to signed long by assignment. So 0xfffffffb (aka -5)
52082 turns into 4294967291 and naturally that changes the result of our division.
52084 The fix is simple, insert a cast to int32_t to force interpretation as a
52087 Testcase for the simulator is included. It has a trivial dependency on the
52090 2022-04-06 Jeff Law <jeffreyalaw@gmail.com>
52092 Fix "bins" simulation for v850e3v5
52093 I've been carrying this for a few years. One test in the GCC testsuite is
52094 failing due to a bug in the handling of the v850e3v5 instruction "bins".
52096 When the "bins" instruction specifies a 32bit bitfield size, the simulator
52097 exhibits undefined behavior by trying to shift a 32 bit quantity by 32 bits.
52098 In the case of a 32 bit shift, we know what the resultant mask should be. So
52099 we can just set it.
52101 That seemed better than using 1UL for the constant (on a 32bit host unsigned
52102 long might still just be 32 bits) or needlessly forcing everything to
52105 Thankfully the case where this shows up is only bins <src>, 0, 32, <dest>
52106 which would normally be encoded as a simple move.
52108 * testsuite/v850/allinsns.exp: Add v850e3v5.
52109 * testsuite/v850/bins.cgs: New test.
52110 * v850/simops.c (v850_bins): Avoid undefined behavior on left shift.
52112 2022-04-06 Tiezhu Yang <yangtiezhu@loongson.cn>
52114 gdb: LoongArch: prepend tramp frame unwinder for signal
52115 Implement the "init" method of struct tramp_frame to prepend tramp
52116 frame unwinder for signal on LoongArch.
52118 With this patch, the following failed testcases can be fixed:
52120 FAIL: gdb.base/annota1.exp: backtrace @ signal handler (timeout)
52121 FAIL: gdb.base/annota3.exp: backtrace @ signal handler (pattern 2)
52123 2022-04-06 Andrew Burgess <aburgess@redhat.com>
52125 gdb: make interp_add static
52128 commit 8322445e0584be846f5873b9aab257dc9fbda05d
52129 Date: Tue Jun 21 01:11:45 2016 +0100
52131 Introduce interpreter factories
52133 Interpreters should be registered with GDB, not by calling interp_add,
52134 but with a call to interp_factory_register. I've checked the insight
52135 source, and it too has moved over to using interp_factory_register.
52137 In this commit I make interp_add static within interps.c.
52139 There should be no user visible change after this commit.
52141 2022-04-06 Nick Clifton <nickc@redhat.com>
52143 Add code to display the contents of .debug_loclists sections which contain offset entry tables.
52145 * dwarf.c (fetch_indexed_value): Rename to fecth_indexed_addr and
52146 return the address, rather than a string.
52147 (fetch_indexed_value): New function - returns a value indexed by a
52148 DW_FORM_loclistx or DW_FORM_rnglistx form.
52149 (read_and_display_attr_value): Add support for DW_FORM_loclistx
52150 and DW_FORM_rnglistx.
52151 (process_debug_info): Load the loclists and rnglists sections.
52152 (display_loclists_list): Add support for DW_LLE_base_addressx,
52153 DW_LLE_startx_endx, DW_LLE_startx_length and
52154 DW_LLE_default_location.
52155 (display_offset_entry_loclists): New function. Displays a
52156 .debug_loclists section that contains offset entry tables.
52157 (display_debug_loc): Call the new function.
52158 (display_debug_rnglists_list): Add support for
52159 DW_RLE_base_addressx, DW_RLE_startx_endx and DW_RLE_startx_length.
52160 (display_debug_ranges): Display the contents of the section's
52162 * dwarf.h (struct debug_info): Add loclists_base field.
52163 * testsuite/binutils-all/dw5.W: Update expected output.
52164 * testsuite/binutils-all/x86-64/pr26808.dump: Likewise.
52166 2022-04-06 Luis Machado <luis.machado@linaro.org>
52168 Enable ARMv8.1-m PACBTI support
52169 This set of changes enable support for the ARMv8.1-m PACBTI extensions [1].
52171 The goal of the PACBTI extensions is similar in scope to that of a-profile
52172 PAC/BTI (aarch64 only), but the underlying implementation is different.
52174 One important difference is that the pointer authentication code is stored
52175 in a separate register, thus we don't need to mask/unmask the return address
52176 from a function in order to produce a correct backtrace.
52178 The patch introduces the following modifications:
52180 - Extend the prologue analyser for 32-bit ARM to handle some instructions
52181 from ARMv8.1-m PACBTI: pac, aut, pacg, autg and bti. Also keep track of
52182 return address signing/authentication instructions.
52184 - Adds code to identify object file attributes that indicate the presence of
52185 ARMv8.1-m PACBTI (Tag_PAC_extension, Tag_BTI_extension, Tag_PACRET_use and
52188 - Adds support for DWARF pseudo-register RA_AUTH_CODE, as described in the
52191 - Extends the dwarf unwinder to track the value of RA_AUTH_CODE.
52193 - Decorates backtraces with the "[PAC]" identifier when a frame has signed
52194 the return address.
52196 - Makes GDB aware of a new XML feature "org.gnu.gdb.arm.m-profile-pacbti". This
52197 feature is not included as an XML file on GDB's side because it is only
52198 supported for bare metal targets.
52200 - Additional documentation.
52202 [1] https://community.arm.com/arm-community-blogs/b/architectures-and-processors-blog/posts/armv8-1-m-pointer-authentication-and-branch-target-identification-extension
52203 [2] https://github.com/ARM-software/abi-aa/blob/main/aadwarf32/aadwarf32.rst
52205 2022-04-06 Andrew Burgess <aburgess@redhat.com>
52207 gdb: move gdb_disassembly_flag into a new disasm-flags.h file
52208 While working on the disassembler I was getting frustrated. Every
52209 time I touched disasm.h it seemed like every file in GDB would need to
52210 be rebuilt. Surely the disassembler can't be required by that many
52211 parts of GDB, right?
52213 Turns out that disasm.h is included in target.h, so pretty much every
52214 file was being rebuilt!
52216 The only thing from disasm.h that target.h needed is the
52217 gdb_disassembly_flag enum, as this is part of the target_ops api.
52219 In this commit I move gdb_disassembly_flag into its own file. This is
52220 then included in target.h and disasm.h, after which, the number of
52221 files that depend on disasm.h is much reduced.
52223 I also audited all the other includes of disasm.h and found that the
52224 includes in mep-tdep.c and python/py-registers.c are no longer needed,
52225 so I've removed these.
52227 Now, after changing disasm.h, GDB rebuilds much quicker.
52229 There should be no user visible changes after this commit.
52231 2022-04-06 GDB Administrator <gdbadmin@sourceware.org>
52233 Automatic date update in version.in
52235 2022-04-05 Tom Tromey <tom@tromey.com>
52237 Introduce wrapped_file
52238 Simon pointed out that timestamped_file probably needed to implement a
52239 few more methods. This patch introduces a new file-wrapping file that
52240 forwards most of its calls, making it simpler to implement new such
52241 files. It also converts timestamped_file and pager_file to use it.
52243 Regression tested on x86-64 Fedora 34.
52245 2022-04-05 Tom Tromey <tromey@adacore.com>
52247 Don't call init_thread_list in windows-nat.c
52248 I don't think there's any need to call init_thread_list in
52249 windows-nat.c. This patch removes it. I tested this using the
52250 internal AdaCore test suite on Windows, which FWIW does include some
52251 multi-threaded inferiors.
52253 2022-04-05 Simon Marchi <simon.marchi@efficios.com>
52255 gdb/testsuite: fix intermittent failure in gdb.base/vfork-follow-parent.exp
52256 Tom de Vries reported some failures in this test:
52260 [New inferior 2 (process 14967)]
52262 Thread 1.1 "vfork-follow-pa" hit Breakpoint 2, break_parent () at /home/vries/gdb_versions/devel/src/gdb/testsuite/gdb.base/vfork-follow-parent.c:23
52264 (gdb) FAIL: gdb.base/vfork-follow-parent.exp: resolution_method=schedule-multiple: continue to end of inferior 2
52266 [Switching to inferior 1 [process 14961] (/home/vries/gdb_versions/devel/build/gdb/testsuite/outputs/gdb.base/vfork-follow-parent/vfork-follow-parent)]
52267 [Switching to thread 1.1 (process 14961)]
52268 #0 break_parent () at /home/vries/gdb_versions/devel/src/gdb/testsuite/gdb.base/vfork-follow-parent.c:23
52270 (gdb) PASS: gdb.base/vfork-follow-parent.exp: resolution_method=schedule-multiple: inferior 1
52273 [Inferior 2 (process 14967) exited normally]
52274 (gdb) FAIL: gdb.base/vfork-follow-parent.exp: resolution_method=schedule-multiple: continue to break_parent (the program exited)
52276 Here, we continue both the vfork parent and child, since
52277 schedule-multiple is on. The child exits, which un-freezes the parent
52278 and makes an exit event available to GDB. We expect GDB to consume this
52279 exit event and present it to the user. Here, we see that GDB shows the
52280 parent hitting a breakpoint before showing the child exit.
52282 Because of the vfork, we know that chronologically, the child exiting
52283 must have happend before the parent hitting a breakpoint. However,
52284 scheduling being what it is, it is possible for the parent to un-freeze
52285 and exit quickly, such that when GDB pulls events out of the kernel,
52286 exit events for both processes are available. And then, GDB may chose
52287 at random to return the one for the parent first. This is what I
52288 imagine what causes the failure shown above.
52290 We could change the test to expect both possible outcomes, but I wanted
52291 to avoid complicating the .exp file that way. Instead, add a variable
52292 that the parent loops on that we set only after we confirmed the exit of
52293 the child. That should ensure that the order is always the same.
52295 Note that I wasn't able to reproduce the failure, so I can't tell if
52296 this fix really fixes the problem.
52298 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29021
52299 Change-Id: Ibc8e527e0e00dac54b22021fe4d9d8ab0f3b28ad
52301 2022-04-05 Simon Marchi <simon.marchi@efficios.com>
52303 gdb/testsuite: fix intermittent failures in gdb.mi/mi-cmd-user-context.exp
52304 I got failures like this once on a CI:
52308 ~"#0 child_sub_function () at /home/jenkins/workspace/binutils-gdb_master_build/arch/amd64/target_board/unix/src/binutils-gdb/gdb/testsuite/gdb.mi/user-selected-context-sync.c:33\n"^M
52309 ~"33\t dummy = !dummy; /* thread loop line */\n"^M
52312 FAIL: gdb.mi/mi-cmd-user-context.exp: frame 1 (unexpected output)
52314 The problem is that the test expects the following regexp:
52318 And that typically works, when the output of the frame command looks
52321 #0 0x00005555555551bb in child_sub_function () at ...
52323 Note the lack of hexadecimal address in the failing case. Whether or
52324 not the hexadecimal address is printed (roughly) depends on whether the
52325 current PC is at the beginning of a line. So depending on where thread
52326 2 was when GDB stopped it (after thread 1 hit its breakpoint), we can
52327 get either output. Adjust the regexps to not expect an hexadecimal
52328 prefix (0x) but a function name instead (either child_sub_function or
52329 child_function). That one is always printed, and is also a good check
52330 that we are in the frame we expect.
52332 Note that for test "frame 5", we are showing a pthread frame (on my
52333 system), so the function name is internal to pthread, not something we
52334 can rely on. In that case, it's almost certain that we are not at the
52335 beginning of a line, or that we don't have debug info, so I think it's
52336 fine to expect the hex prefix.
52338 And for test "frame 6", it's ok to _not_ expect a hex prefix (what the
52339 test currently does), since we are showing thread 1, which has hit a
52340 breakpoint placed at the beginning of a line.
52342 When testing this, Tom de Vries pointed out that the current test code
52343 doesn't ensure that the child threads are in child_sub_function when
52344 they are stopped. If the scheduler chooses so, it is possible for the
52345 child threads to be still in the pthread_barrier_wait or child_function
52346 functions when they get stopped. So that would be another racy failure
52349 The only way I can think of to ensure the child threads are in the
52350 child_sub_function function when they get stopped is to synchronize the
52351 threads using some variables instead of pthread_barrier_wait. So,
52352 replace the barrier with an array of flags (one per child thread). Each
52353 child thread flips its flag in child_sub_function to allow the main
52354 thread to make progress and eventually hit the breakpoint.
52356 I copied user-selected-context-sync.c to a new mi-cmd-user-context.c and
52357 made modifications to that, to avoid interfering with
52358 user-selected-context-sync.exp.
52360 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29025
52361 Change-Id: I919673bbf9927158beb0e8b7e9e980b8d65eca90
52363 2022-04-05 Luis Machado <luis.machado@arm.com>
52365 Fix qRcmd error code parsing
52366 Someone at IRC spotted a bug in qRcmd handling. This looks like an oversight
52367 or it is that way for historical reasons.
52369 The code in gdb/remote.c:remote_target::rcmd uses isdigit instead of
52370 isxdigit. One could argue that we are expecting decimal numbers, but further
52371 below we use fromhex ().
52373 Update the function to use isxdigit instead and also update the documentation.
52375 I see there are lots of other cases of undocumented number format for error
52376 messages, mostly described as NN instead of nn. For now I'll just update
52377 this particular function.
52379 2022-04-05 Simon Marchi <simon.marchi@efficios.com>
52381 gdb: resume ongoing step after handling fork or vfork
52382 The test introduced by this patch would fail in this configuration, with
52383 the native-gdbserver or native-extended-gdbserver boards:
52385 FAIL: gdb.threads/next-fork-other-thread.exp: fork_func=fork: target-non-stop=auto: non-stop=off: displaced-stepping=auto: i=2: next to for loop
52387 The problem is that the step operation is forgotten when handling the
52388 fork/vfork. With "debug infrun" and "debug remote", it looks like this
52389 (some lines omitted for brevity). We do the next:
52391 [infrun] proceed: enter
52392 [infrun] proceed: addr=0xffffffffffffffff, signal=GDB_SIGNAL_DEFAULT
52393 [infrun] resume_1: step=1, signal=GDB_SIGNAL_0, trap_expected=0, current thread [4154304.4154304.0] at 0x5555555553bf
52394 [infrun] do_target_resume: resume_ptid=4154304.0.0, step=1, sig=GDB_SIGNAL_0
52395 [remote] Sending packet: $vCont;r5555555553bf,5555555553c4:p3f63c0.3f63c0;c:p3f63c0.-1#cd
52396 [infrun] proceed: exit
52398 We then handle a fork event:
52400 [infrun] fetch_inferior_event: enter
52401 [remote] wait: enter
52402 [remote] Packet received: T05fork:p3f63ee.3f63ee;06:0100000000000000;07:b08e59f6ff7f0000;10:bf60e8f7ff7f0000;thread:p3f63c0.3f63c6;core:17;
52403 [remote] wait: exit
52404 [infrun] print_target_wait_results: target_wait (-1.0.0 [process -1], status) =
52405 [infrun] print_target_wait_results: 4154304.4154310.0 [Thread 4154304.4154310],
52406 [infrun] print_target_wait_results: status->kind = FORKED, child_ptid = 4154350.4154350.0
52407 [infrun] handle_inferior_event: status->kind = FORKED, child_ptid = 4154350.4154350.0
52408 [remote] Sending packet: $D;3f63ee#4b
52409 [infrun] resume_1: step=0, signal=GDB_SIGNAL_0, trap_expected=0, current thread [4154304.4154310.0] at 0x7ffff7e860bf
52410 [infrun] do_target_resume: resume_ptid=4154304.0.0, step=0, sig=GDB_SIGNAL_0
52411 [remote] Sending packet: $vCont;c:p3f63c0.-1#73
52412 [infrun] fetch_inferior_event: exit
52414 In the first snippet, we resume the stepping thread with the range-stepping (r)
52415 vCont command. But after handling the fork (detaching the fork child), we
52416 resumed the whole process freely. The stepping thread, which was paused by
52417 GDBserver while reporting the fork event, was therefore resumed freely, instead
52418 of confined to the addresses of the stepped line. Note that since this
52419 is a "next", it could be that we have entered a function, installed a
52420 step-resume breakpoint, and it's ok to continue freely the stepping
52421 thread, but that's not the case here. The two snippets shown above were
52422 next to each other in the logs.
52424 For the fork case, we can resume stepping right after handling the
52427 However, for the vfork case, where we are waiting for the
52428 external child process to exec or exit, we only resume the thread that
52429 called vfork, and keep the others stopped (see patch "gdb: fix handling of
52430 vfork by multi-threaded program" prior in this series). So we can't
52431 resume the stepping thread right now. Instead, do it after handling the
52434 Change-Id: I92539c970397ce880110e039fe92b87480f816bd
52436 2022-04-05 Simon Marchi <simon.marchi@polymtl.ca>
52438 gdb/remote: remove_new_fork_children don't access target_waitstatus::child_ptid if kind == TARGET_WAITKIND_THREAD_EXITED
52439 Following the previous patch, running
52440 gdb.threads/forking-threads-plus-breakpoints.exp continuously eventually
52441 gives me an internal error.
52443 gdb/target/waitstatus.h:372: internal-error: child_ptid: Assertion `m_kind == TARGET_WAITKIND_FORKED || m_kind == TARGET_WAITKIND_VFORKED' failed.^M
52444 FAIL: gdb.threads/forking-threads-plus-breakpoint.exp: cond_bp_target=0: detach_on_fork=on: displaced=off: inferior 1 exited (GDB internal error)
52448 0x55925b679c85 internal_error(char const*, int, char const*, ...)
52449 /home/simark/src/binutils-gdb/gdbsupport/errors.cc:55
52450 0x559258deadd2 target_waitstatus::child_ptid() const
52451 /home/simark/src/binutils-gdb/gdb/target/waitstatus.h:372
52452 0x55925a7cbac9 remote_target::remove_new_fork_children(threads_listing_context*)
52453 /home/simark/src/binutils-gdb/gdb/remote.c:7311
52454 0x55925a79dfdb remote_target::update_thread_list()
52455 /home/simark/src/binutils-gdb/gdb/remote.c:3981
52456 0x55925ad79b83 target_update_thread_list()
52457 /home/simark/src/binutils-gdb/gdb/target.c:3793
52458 0x55925addbb15 update_thread_list()
52459 /home/simark/src/binutils-gdb/gdb/thread.c:2031
52460 0x559259d64838 stop_all_threads(char const*, inferior*)
52461 /home/simark/src/binutils-gdb/gdb/infrun.c:5104
52462 0x559259d88b45 keep_going_pass_signal
52463 /home/simark/src/binutils-gdb/gdb/infrun.c:8215
52464 0x559259d8951b keep_going
52465 /home/simark/src/binutils-gdb/gdb/infrun.c:8251
52466 0x559259d78835 process_event_stop_test
52467 /home/simark/src/binutils-gdb/gdb/infrun.c:6858
52468 0x559259d750e9 handle_signal_stop
52469 /home/simark/src/binutils-gdb/gdb/infrun.c:6580
52470 0x559259d6c07b handle_inferior_event
52471 /home/simark/src/binutils-gdb/gdb/infrun.c:5832
52472 0x559259d57db8 fetch_inferior_event()
52473 /home/simark/src/binutils-gdb/gdb/infrun.c:4222
52475 Indeed, the code accesses target_waitstatus::child_ptid when the kind
52476 is TARGET_WAITKIND_THREAD_EXITED, which is not right. A
52477 TARGET_WAITKIND_THREAD_EXITED event does not have a child_ptid value
52478 associated, it has an exit status, which we are not interested in. The
52479 intent is to remove from the thread list the thread that has exited.
52480 Its ptid is found in the stop reply event, get it from there.
52482 Change-Id: Icb298cbb80b8779fdf0c660dde9a5314d5591535
52484 2022-04-05 Simon Marchi <simon.marchi@efficios.com>
52486 gdbserver: report correct status in thread stop race condition
52487 The test introduced by the following patch would sometimes fail in this
52490 FAIL: gdb.threads/next-fork-other-thread.exp: fork_func=vfork: target-non-stop=on: non-stop=off: displaced-stepping=auto: i=14: next to for loop
52492 The test has multiple threads constantly forking or vforking while the
52493 main thread keep doing "next"s.
52495 (After writing the commit message, I realized this also fixes a similar
52496 failure in gdb.threads/forking-threads-plus-breakpoint.exp with the
52497 native-gdbserver and native-extended-gdbserver boards.)
52499 As stop_all_threads is called, because the main thread finished its
52500 "next", it inevitably happens at some point that we ask the remote
52501 target to stop a thread and wait() reports that this thread stopped with
52502 a fork or vfork event, instead of the SIGSTOP we sent to try to stop it.
52504 While running this test, I attached to GDBserver and stopped at
52505 linux-low.cc:3626. We can see that the status pulled from the kernel
52506 for 2742805 is indeed a vfork event:
52510 (gdb) p WIFSTOPPED(w)
52512 (gdb) p WSTOPSIG(w)
52514 (gdb) p/x (w >> 8) & (PTRACE_EVENT_VFORK << 8)
52517 However, the statement at line 3626 overrides that:
52519 ourstatus->set_stopped (gdb_signal_from_host (WSTOPSIG (w)));
52521 OURSTATUS becomes "stopped by a SIGTRAP". The information about the
52522 fork or vfork is lost.
52524 It's then all downhill from there, stop_all_threads eventually asks for
52525 a thread list update. That thread list includes the child of that
52526 forgotten fork or vfork, the remote target goes "oh cool, a new process,
52527 let's attach to it!", when in fact that vfork child's destiny was to be
52530 My reverse-engineered understanding of the code around there is that the
52531 if/else between lines 3562 and 3583 (in the original code) makes sure
52532 OURSTATUS is always initialized (not "ignore"). Either the details are
52533 already in event_child->waitstatus (in the case of fork/vfork, for
52534 example), in which case we just copy event_child->waitstatus to
52535 ourstatus. Or, if the event is a plain "stopped by a signal" or a
52536 syscall event, OURSTATUS is set to "stopped", but without a signal
52537 number. Lines 3601 to 3629 (in the original code) serve to fill in that
52538 last bit of information.
52540 The problem is that when `w` holds the vfork status, the code wrongfully
52541 takes this branch, because WSTOPSIG(w) returns SIGTRAP:
52543 else if (current_thread->last_resume_kind == resume_stop
52544 && WSTOPSIG (w) != SIGSTOP)
52546 The intent of this branch is, for example, when we sent SIGSTOP to try
52547 to stop a thread, but wait() reports that it stopped with another signal
52548 (that it must have received from somewhere else simultaneously), say
52549 SIGWINCH. In that case, we want to report the SIGWINCH. But in our
52550 fork/vfork case, we don't want to take this branch, as the thread didn't
52551 really stop because it received a signal. For the non "stopped by a
52552 signal" and non "syscall signal" cases, we would ideally skip over all
52553 that snippet that fills in the signal or syscall number.
52555 The fix I propose is to move this snipppet of the else branch of the
52556 if/else above. In addition to moving the code, the last two "else if"
52559 else if (current_thread->last_resume_kind == resume_stop
52560 && WSTOPSIG (w) != SIGSTOP)
52562 /* A thread that has been requested to stop by GDB with vCont;t,
52563 but, it stopped for other reasons. */
52564 ourstatus->set_stopped (gdb_signal_from_host (WSTOPSIG (w)));
52566 else if (ourstatus->kind () == TARGET_WAITKIND_STOPPED)
52567 ourstatus->set_stopped (gdb_signal_from_host (WSTOPSIG (w)));
52569 are changed into a single else:
52572 ourstatus->set_stopped (gdb_signal_from_host (WSTOPSIG (w)));
52574 This is the default path we take if:
52576 - W is not a syscall status
52577 - W does not represent a SIGSTOP that have sent to stop the thread and
52578 therefore want to suppress it
52580 Change-Id: If2dc1f0537a549c293f7fa3c53efd00e3e194e79
52582 2022-04-05 Simon Marchi <simon.marchi@efficios.com>
52584 gdb: fix handling of vfork by multi-threaded program (follow-fork-mode=parent, detach-on-fork=on)
52585 There is a problem with how GDB handles a vfork happening in a
52586 multi-threaded program. This problem was reported to me by somebody not
52587 using vfork directly, but using system(3) in a multi-threaded program,
52588 which may be implemented using vfork.
52590 This patch only deals about the follow-fork-mode=parent,
52591 detach-on-fork=on case, because it would be too much to chew at once to
52592 fix the bugs in the other cases as well (I tried).
52597 When a program vforks, the parent thread is suspended by the kernel
52598 until the child process exits or execs. Specifically, in a
52599 multi-threaded program, only the thread that called vfork is suspended,
52600 other threads keep running freely. This is documented in the vfork(2)
52601 man page ("Caveats" section).
52603 Let's suppose GDB is handling a vfork and the user's desire is to detach
52604 from the child. Before detaching the child, GDB must remove the software
52605 breakpoints inserted in the shared parent/child address space, in case
52606 there's a breakpoint in the path the child is going to take before
52607 exec'ing or exit'ing (unlikely, but possible). Otherwise the child could
52608 hit a breakpoint instruction while running outside the control of GDB,
52609 which would make it crash. GDB must also avoid re-inserting breakpoints
52610 in the parent as long as it didn't receive the "vfork done" event (that
52611 is, when the child has exited or execed): since the address space is
52612 shared with the child, that would re-insert breakpoints in the child
52613 process also. So what GDB does is:
52615 1. Receive "vfork" event for the parent
52616 2. Remove breakpoints from the (shared) address space and set
52617 program_space::breakpoints_not_allowed to avoid re-inserting them
52618 3. Detach from the child thread
52619 4. Resume the parent
52620 5. Wait for and receive "vfork done" event for the parent
52621 6. Clean program_space::breakpoints_not_allowed and re-insert
52623 7. Resume the parent
52625 Resuming the parent at step 4 is necessary in order for the kernel to
52626 report the "vfork done" event. The kernel won't report a ptrace event
52627 for a thread that is ptrace-stopped. But the theory behind this is that
52628 between steps 4 and 5, the parent won't actually do any progress even
52629 though it is ptrace-resumed, because the kernel keeps it suspended,
52630 waiting for the child to exec or exit. So it doesn't matter for that
52631 thread if breakpoints are not inserted.
52633 The problem is when the program is multi-threaded. In step 4, GDB
52634 resumes all threads of the parent. The thread that did the vfork stays
52635 suspended by the kernel, so that's fine. But other threads are running
52636 freely while breakpoints are removed, which is a problem because they
52637 could miss a breakpoint that they should have hit.
52639 The problem is present with all-stop and non-stop targets. The only
52640 difference is that with an all-stop targets, the other threads are
52641 stopped by the target when it reports the vfork event and are resumed by
52642 the target when GDB resumes the parent. With a non-stop target, the
52643 other threads are simply never stopped.
52648 There many combinations of settings to consider (all-stop/non-stop,
52649 target-non-stop on/off, follow-fork-mode parent/child, detach-on-fork
52650 on/off, schedule-multiple on/off), but for this patch I restrict the
52651 scope to follow-fork-mode=parent, detach-on-fork=on. That's the
52652 "default" case, where we detach the child and keep debugging the
52653 parent. I tried to fix them all, but it's just too much to do at once.
52654 The code paths and behaviors for when we don't detach the child are
52655 completely different.
52657 The guiding principle for this patch is that all threads of the vforking
52658 inferior should be stopped as long as breakpoints are removed. This is
52659 similar to handling in-line step-overs, in a way.
52661 For non-stop targets (the default on Linux native), this is what
52664 - In follow_fork, we call stop_all_threads to stop all threads of the
52666 - In follow_fork_inferior, we record the vfork parent thread in
52667 inferior::thread_waiting_for_vfork_done
52668 - Back in handle_inferior_event, we call keep_going, which resumes only
52669 the event thread (this is already the case, with a non-stop target).
52670 This is the thread that will be waiting for vfork-done.
52671 - When we get the vfork-done event, we go in the (new) handle_vfork_done
52672 function to restart the previously stopped threads.
52674 In the same scenario, but with an all-stop target:
52676 - In follow_fork, no need to stop all threads of the inferior, the
52677 target has stopped all threads of all its inferiors before returning
52679 - In follow_fork_inferior, we record the vfork parent thread in
52680 inferior::thread_waiting_for_vfork_done.
52681 - Back in handle_inferior_event, we also call keep_going. However, we
52682 only want to resume the event thread here, not all inferior threads.
52683 In internal_resume_ptid (called by resume_1), we therefore now check
52684 whether one of the inferiors we are about to resume has
52685 thread_waiting_for_vfork_done set. If so, we only resume that
52688 Note that when resuming multiple inferiors, one vforking and one not
52689 non-vforking, we could resume the vforking thread from the vforking
52690 inferior plus all threads from the non-vforking inferior. However,
52691 this is not implemented, it would require more work.
52692 - When we get the vfork-done event, the existing call to keep_going
52693 naturally resumes all threads.
52695 Testing-wise, add a test that tries to make the main thread hit a
52696 breakpoint while a secondary thread calls vfork. Without the fix, the
52697 main thread keeps going while breakpoints are removed, resulting in a
52698 missed breakpoint and the program exiting.
52700 Change-Id: I20eb78e17ca91f93c19c2b89a7e12c382ee814a1
52702 2022-04-05 Simon Marchi <simon.marchi@polymtl.ca>
52704 gdb/infrun: add logging statement to do_target_resume
52705 This helped me, it shows which ptid we actually call target_resume with.
52707 Change-Id: I2dfd771e83df8c25f39371a13e3e91dc7882b73d
52709 2022-04-05 Simon Marchi <simon.marchi@polymtl.ca>
52711 gdb/infrun: add inferior parameters to stop_all_threads and restart_threads
52712 A following patch will want to stop all threads of a given inferior (as
52713 opposed to all threads of all inferiors) while handling a vfork, and
52714 restart them after. To help with this, add inferior parameters to
52715 stop_all_threads and restart_threads. This is done as a separate patch
52716 to make sure this doesn't cause regressions on its own, and to keep the
52717 following patches more concise.
52719 No visible changes are expected here, since all calls sites pass
52720 nullptr, which should keep the existing behavior.
52722 Change-Id: I4d9ba886ce842042075b4e346cfa64bbe2580dbf
52724 2022-04-05 Simon Marchi <simon.marchi@polymtl.ca>
52726 gdb: replace inferior::waiting_for_vfork_done with inferior::thread_waiting_for_vfork_done
52727 The inferior::waiting_for_vfork_done flag indicates that some thread in
52728 that inferior is waiting for a vfork-done event. Subsequent patches
52729 will need to know which thread precisely is waiting for that event.
52731 Replace the boolean flag (waiting_for_vfork_done) with a thread_info
52732 pointer (thread_waiting_for_vfork_done).
52734 I think there is a latent buglet in that waiting_for_vfork_done is
52735 currently not reset on inferior exec or exit. I could imagine that if a
52736 thread in the parent process calls exec or exit while another thread of
52737 the parent process is waiting for its vfork child to exec or exit, we
52738 could end up with inferior::waiting_for_vfork_done without a thread
52739 actually waiting for a vfork-done event anymore. And since that flag is
52740 checked in resume_1, things could misbehave there.
52742 Since the new field points to a thread_info object, and those are
52743 destroyed on exec or exit, it could be worse now since we could try to
52744 access freed memory, if thread_waiting_for_vfork_done were to point to a
52745 stale thread_info. To avoid this, clear the field in
52746 infrun_inferior_exit and infrun_inferior_execd.
52748 Change-Id: I31b847278613a49ba03fc4915f74d9ceb228fdce
52750 2022-04-05 Simon Marchi <simon.marchi@efficios.com>
52752 gdb: make timestamped_file implement write_async_safe
52753 Trying to use "set debug linux-nat 1", I get an internal error:
52755 /home/smarchi/src/binutils-gdb/gdb/ui-file.h:70: internal-error: write_async_safe: write_async_safe
52757 The problem is that timestamped_file doesn't implement write_async_safe,
52758 which linux-nat's sigchld_handler uses. Implement it.
52760 Change-Id: I830981010c6119f13ae673605ed015cced0f5ee8
52762 2022-04-05 GDB Administrator <gdbadmin@sourceware.org>
52764 Automatic date update in version.in
52766 2022-04-04 Andrew Burgess <aburgess@redhat.com>
52768 gdb/testsuite: fix timeout in server-pipe.exp test
52769 I noticed that the gdb.server/server-pipe.exp test would sometimes
52770 timeout when my machine was more heavily loaded. Turns out the test
52771 is reading all the shared libraries over GDB's remote protocol, which
52774 We avoid this in other tests by setting the sysroot in GDBFLAGS,
52775 something which is missing from the gdb.server/server-pipe.exp test.
52777 Fix the timeouts by setting sysroot in GDBFLAGS, after this the shared
52778 libraries are no longer copied over the remote protocol, and I no
52779 longer see the test timeout.
52781 2022-04-04 John Baldwin <jhb@FreeBSD.org>
52783 Handle TLS variable lookups when using separate debug files.
52784 Commit df22c1e5d53c38f38bce6072bb46de240f9e0e2b handled the case that
52785 a separate debug file was passed as the objfile for a shared library
52786 to svr4_fetch_objfile_link_map. However, a separate debug file can
52787 also be passed for TLS variables in the main executable. In addition,
52788 frv_fetch_objfile_link_map also expects to be passed the original
52789 objfile rather than a separate debug file, so pull the code to resolve
52790 a separate debug file to the main objfile up into
52791 target_translate_tls_address.
52793 2022-04-04 Lancelot SIX <lancelot.six@amd.com>
52795 gdb: Add maint set ignore-prologue-end-flag
52796 The previous patch added support for the DWARF prologue-end flag in line
52797 table. This flag can be used by DWARF producers to indicate where to
52798 place breakpoints past a function prologue. However, this takes
52799 precedence over prologue analyzers. So if we have to debug a program
52800 with erroneous debug information, the overall debugging experience will
52803 This commit proposes to add a maintenance command to instruct GDB to
52804 ignore the prologue_end flag.
52806 Tested on x86_64-gnu-linux.
52808 Change-Id: Idda6d1b96ba887f4af555b43d9923261b9cc6f82
52810 2022-04-04 Lancelot SIX <lancelot.six@amd.com>
52812 gdb: Add support for DW_LNS_set_prologue_end in line-table
52813 Add support for DW_LNS_set_prologue_end when building line-tables. This
52814 attribute can be set by the compiler to indicate that an instruction is
52815 an adequate place to set a breakpoint just after the prologue of a
52818 The compiler might set multiple prologue_end, but considering how
52819 current skip_prologue_using_sal works, this commit modifies it to accept
52820 the first instruction with this marker (if any) to be the place where a
52821 breakpoint should be placed to be at the end of the prologue.
52823 The need for this support came from a problematic usecase generated by
52824 hipcc (i.e. clang). The problem is as follows: There's a function
52825 (lets call it foo) which covers PC from 0xa800 to 0xa950. The body of
52826 foo begins with a call to an inlined function, covering from 0xa800 to
52827 0xa94c. The issue is that when placing a breakpoint at 'foo', GDB
52828 inserts the breakpoint at 0xa818. The 0x18 offset is what GDB thinks is
52829 foo's first address past the prologue.
52831 Later, when hitting the breakpoint, GDB reports the stop within the
52832 inlined function because the PC falls in its range while the user
52833 expects to stop in FOO.
52835 Looking at the line-table for this location, we have:
52837 INDEX LINE ADDRESS IS-STMT
52839 14 293 0x000000000000a66c Y
52840 15 END 0x000000000000a6e0 Y
52841 16 287 0x000000000000a800 Y
52842 17 END 0x000000000000a818 Y
52843 18 287 0x000000000000a824 Y
52846 For comparison, let's look at llvm-dwarfdump's output for this CU:
52848 Address Line Column File ISA Discriminator Flags
52849 ------------------ ------ ------ ------ --- ------------- -------------
52851 0x000000000000a66c 293 12 2 0 0 is_stmt
52852 0x000000000000a6e0 96 43 82 0 0 is_stmt
52853 0x000000000000a6f8 102 18 82 0 0 is_stmt
52854 0x000000000000a70c 102 24 82 0 0
52855 0x000000000000a710 102 18 82 0 0
52856 0x000000000000a72c 101 16 82 0 0 is_stmt
52857 0x000000000000a73c 2915 50 83 0 0 is_stmt
52858 0x000000000000a74c 110 1 1 0 0 is_stmt
52859 0x000000000000a750 110 1 1 0 0 is_stmt end_sequence
52860 0x000000000000a800 107 0 1 0 0 is_stmt
52861 0x000000000000a800 287 12 2 0 0 is_stmt prologue_end
52862 0x000000000000a818 114 59 81 0 0 is_stmt
52863 0x000000000000a824 287 12 2 0 0 is_stmt
52864 0x000000000000a828 100 58 82 0 0 is_stmt
52867 The main difference we are interested in here is that llvm-dwarfdump's
52868 output tells us that 0xa800 is an adequate place to place a breakpoint
52869 past a function prologue. Since we know that foo covers from 0xa800 to
52870 0xa94c, 0xa800 is the address at which the breakpoint should be placed
52871 if the user wants to break in foo.
52873 This commit proposes to add support for the prologue_end flag in the
52874 line-program processing.
52876 The processing of this prologue_end flag is made in skip_prologue_sal,
52877 before it calls gdbarch_skip_prologue_noexcept. The intent is that if
52878 the compiler gave information on where the prologue ends, we should use
52879 this information and not try to rely on architecture dependent logic to
52882 The testsuite have been executed using this patch on GNU/Linux x86_64.
52883 Testcases have been compiled with both gcc/g++ (verison 9.4.0) and
52884 clang/clang++ (version 10.0.0) since at the time of writing GCC does not
52885 set the prologue_end marker. Tests done with GCC 11.2.0 (not over the
52886 entire testsuite) show that it does not emit this flag either.
52888 No regression have been observed with GCC or Clang. Note that when
52889 using Clang, this patch fixes a failure in
52890 gdb.opt/inline-small-func.exp.
52892 Change-Id: I720449a8a9b2e1fb45b54c6095d3b1e9da9152f8
52894 2022-04-04 Lancelot SIX <lancelot.six@amd.com>
52896 gdb/buildsym: Line record use a record flag
52897 Currently when recording a line entry (with
52898 buildsym_compunit::record_line), a boolean argument argument is used to
52899 indicate that the is_stmt flag should be set for this particular record.
52900 As a later commit will add support for new flags, instead of adding a
52901 parameter to record_line for each possible flag, transform the current
52902 is_stmt parameter into a enum flag. This flags parameter will allow
52903 greater flexibility in future commits.
52905 This enum flags type is not propagated into the linetable_entry type as
52906 this would require a lot of changes across the codebase for no practical
52907 gain (it currently uses a bitfield where each interesting flag only
52908 occupy 1 bit in the structure).
52910 Tested on x86_64-linux, no regression observed.
52912 Change-Id: I5d061fa67bdb34918742505ff983d37453839d6a
52914 2022-04-04 Simon Marchi <simon.marchi@polymtl.ca>
52916 gdb: make timestamped_file implement can_emit_style_escape
52917 In our AMDGPU downstream port, we use styling in some logging output.
52918 We noticed it stopped working after the gdb_printf changes. Making
52919 timestamped_file implement can_emit_style_escape (returning the value of
52920 the stream it wraps) fixes it. To show that it works, modify some
52921 logging statements in auto-load.c to output style filenames. You can
52922 see it in action by setting "set debug auto-load 1" and running a
52923 program. We can incrementally add styling to other debug statements
52924 throughout GDB, as needed.
52926 Change-Id: I78a2fd1e078f80f2263251cf6bc53b3a9de9c17a
52928 2022-04-04 Simon Marchi <simon.marchi@polymtl.ca>
52930 gdb: remove assertion in psymbol_functions::expand_symtabs_matching
52931 psymtab_to_symtab is documented as possibly returning nullptr, if the
52932 primary symtab of the partial symtab has no symbols. However,
52933 psymbol_functions::expand_symtabs_matching asserts that the result of
52934 psymtab_to_symtab as non-nullptr.
52936 I caught this assert by trying the CTF symbol reader on a library I
52939 $ ./gdb --data-directory=data-directory /tmp/babeltrace-ctf/src/lib/.libs/libbabeltrace2.so.0.0.0
52941 Reading symbols from /tmp/babeltrace-ctf/src/lib/.libs/libbabeltrace2.so.0.0.0...
52942 (gdb) maintenance expand-symtabs
52943 /home/simark/src/binutils-gdb/gdb/psymtab.c:1142: internal-error: expand_symtabs_matching: Assertion `symtab != nullptr' failed.
52945 The "symtab" in question is:
52947 $ readelf --ctf=.ctf /tmp/babeltrace-ctf/src/lib/.libs/libbabeltrace2.so.0.0.0
52949 CTF archive member: /home/simark/src/babeltrace/src/lib/graph/component-descriptor-set.c:
52952 Magic number: 0xdff2
52953 Version: 4 (CTF_VERSION_3)
52954 Flags: 0xe (CTF_F_NEWFUNCINFO, CTF_F_IDXSORTED, CTF_F_DYNSTR)
52956 Compilation unit name: /home/simark/src/babeltrace/src/lib/graph/component-descriptor-set.c
52957 Type section: 0x0 -- 0x13 (0x14 bytes)
52958 String section: 0x14 -- 0x5f (0x4c bytes)
52969 0x80000001: (kind 5) bt_bool (*) (const bt_value *) (aligned at 0x8)
52974 0x6: /home/simark/src/babeltrace/src/lib/graph/component-descriptor-set.c
52976 It contains a single type, and it is skipped by ctf_add_type_cb, because
52977 an identical type was already seen earlier in this objfile. As a
52978 result, no compunit_symtab is created.
52980 Change psymbol_functions::expand_symtabs_matching to expect that
52981 psymtab_to_symtab can return nullptr.
52983 Another possibility would be to make the CTF symbol reader always create
52984 a compunit_symtab, even if there are no symbols in it (like the DWARF
52985 parser does), but so far I don't see any advantage in doing so.
52987 Change-Id: Ic43c38202c838a5eb87630ed1fd61d33528164f4
52989 2022-04-04 Andrew Burgess <aburgess@redhat.com>
52991 sim: fixes for libopcodes styled disassembler
52994 commit 60a3da00bd5407f07d64dff82a4dae98230dfaac
52995 Date: Sat Jan 22 11:38:18 2022 +0000
52997 objdump/opcodes: add syntax highlighting to disassembler output
52999 I broke several sim/ targets by forgetting to update their uses of the
53000 libopcodes disassembler to take account of the new styled printing.
53002 These should all be fixed by this commit.
53004 I've not tried to add actual styled output to the simulator traces,
53005 instead, the styled print routines just ignore the style and print the
53008 2022-04-04 Tom Tromey <tromey@adacore.com>
53010 Remove some globals from nat/windows-nat.c
53011 nat/windows-nat.c has a number of globals that it uses to communicate
53012 with its clients (gdb and gdbserver). However, if we ever want the
53013 Windows ports to be multi-inferior, globals won't work.
53015 This patch takes a step toward that by moving most nat/windows-nat.c
53016 globals into a new struct windows_process_info. Many functions are
53017 converted to be methods on this object.
53019 A couple of globals remain, as they are needed to truly be global due
53020 to the way that the Windows debugging APIs work.
53022 The clients still have a global for the current process. That is,
53023 this patch is a step toward the end goal, but doesn't implement the
53026 2022-04-04 Tom Tromey <tromey@adacore.com>
53028 Remove windows_thread_info destructor
53029 windows_thread_info declares and defines a destructor, but this
53030 doesn't need to be explicit.
53032 Use unique_ptr in the Windows thread list
53033 windows-nat.c uses some manual memory management when manipulating the
53034 thread_list global. Changing this to use unique_ptr simplifies the
53035 code, in particular windows_init_thread_list. (Note that, while I
53036 think the the call to init_thread_list in there is wrong, I haven't
53037 removed it in this patch.)
53039 Use auto_obstack in windows-nat.c
53040 One spot in windows-nat.c can use auto_obstack, removing some manual
53043 2022-04-04 Tom Tromey <tromey@adacore.com>
53045 Simplify windows-nat.c solib handling
53046 Currently windows-nat.c uses struct so_list to record its local idea
53047 of which shared libraries have been loaded. However, many fields in
53048 this are not needed, and furthermore I found this quite confusing at
53049 first -- Windows actually uses solib-target and so the use of so_list
53052 This patch simplifies this code by changing it to use a std::vector
53053 and a new type that holds exactly what's needed for the Windows code.
53055 2022-04-04 Pedro Alves <pedro@palves.net>
53057 Avoid undefined behavior in gdbscm_make_breakpoint
53058 Running gdb.guile/scm-breakpoint.exp against an --enable-ubsan build,
53061 UNRESOLVED: gdb.guile/scm-breakpoint.exp: test_watchpoints: create a breakpoint with an invalid type number
53063 guile (define wp2 (make-breakpoint "result" #:wp-class WP_WRITE #:type 999))
53064 ../../src/gdb/guile/scm-breakpoint.c:377:11: runtime error: load of value 999, which is not a valid value for type 'bptype'
53065 ERROR: GDB process no longer exists
53067 Fix this by parsing the user/guile input as plain int, and cast to
53068 internal type only after we know we have a number that would be valid.
53070 Change-Id: I03578d07db00be01b610a8f5ce72e5521aea6a4b
53072 2022-04-04 Tom Tromey <tromey@adacore.com>
53074 Add context-sensitive field name completion to Ada parser
53075 This updates the Ada expression parser to implement context-sensitive
53076 field name completion. This is PR ada/28727.
53078 This is somewhat complicated due to some choices in the Ada lexer --
53079 it chooses to represent a sequence of "."-separated identifiers as a
53080 single token, so the parser must partially recreate the completer's
53081 logic to find the completion word boundaries.
53083 Despite the minor warts in this patch, though, it is a decent
53084 improvement. It's possible that the DWARF reader rewrite will help
53085 fix the package completion problem pointed out in this patch as well.
53087 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28727
53089 2022-04-04 Tom Tromey <tromey@adacore.com>
53091 Consolidate single-char tokens in ada-lex.l
53092 There are two rules in ada-lex.l that match single-character tokens.
53095 Also, this removes '.' from the list of such tokens. '.' is not used
53096 in any production in ada-exp.y, and removing it here helps the
53097 subsequent completion patches.
53099 2022-04-04 Tom Tromey <tromey@adacore.com>
53101 Remove the Ada DOT_ALL token
53102 The Ada parser has a DOT_ALL token to represent ".all", and another
53103 token to represent other ".<identifier>" forms. However, for
53104 completion it is a bit more convenient to unify these cases, so this
53105 patch removes DOT_ALL.
53107 Refactor ada-lex.l:processId
53108 processId in ada-lex.l is a bit funny -- it uses an "if" and a
53109 "switch", and a nested loop. This patch cleans it up a bit, changing
53110 it to use a boolean flag and a simpler "if".
53112 Implement completion for Ada attributes
53113 This adds a completer for Ada attributes. Some work in the lexer is
53114 required in order to match end-of-input correctly, as flex does not
53115 have a general-purpose way of doing this. (The approach taken here is
53116 recommended in the flex manual.)
53118 2022-04-04 Tom Tromey <tromey@adacore.com>
53120 Refactor expression completion
53121 This refactors the gdb expression completion code to make it easier to
53122 add more types of completers.
53124 In the old approach, just two kinds of completers were supported:
53125 field names for some sub-expression, or tag names (like "enum
53126 something"). The data for each kind was combined in single structure,
53127 "expr_completion_state", and handled explicitly by
53128 complete_expression.
53130 In the new approach, the parser state just holds an object that is
53131 responsible for implementing completion. This way, new completion
53132 types can be added by subclassing this base object.
53134 The structop completer is moved into structop_base_operation, and new
53135 objects are defined for use by the completion code. This moves much
53136 of the logic of expression completion out of completer.c as well.
53138 2022-04-04 Tom Tromey <tromey@adacore.com>
53140 Enable "set debug parser" for Ada
53141 I noticed that "set debug parser 1" did not affect Ada parsing. This
53142 patch fixes the problem.
53144 Because this is rarely useful, and pretty much only for maintainers, I
53145 didn't write a test case.
53147 2022-04-04 Tom Tromey <tromey@adacore.com>
53149 Fix bug in Ada attributes lexing
53150 The Ada lexer allows whitespace between the apostrophe and the
53151 attribute text, but processAttribute does not handle this. This patch
53152 fixes the problem and introduces a regression test.
53154 Remove null sentinel from 'attributes'
53155 In a subsequent patch, it's handy if the 'attributes' array in
53156 ada-lex.l does not have a NULL sentinel at the end. In C++, this is
53159 Handle ghost entities in symbol lookup
53160 Normally, SPARK ghost entities are removed from the executable.
53161 However, with -gnata, they will be preserved. In this situation, it's
53162 handy to be able to inspect them. This patch allows this by removing
53163 the "___ghost_" prefix in the appropriate places.
53165 2022-04-04 Simon Marchi <simon.marchi@efficios.com>
53167 gdb: rename start_symtab/end_symtab to start_compunit_symtab/end_compunit_symtab
53168 It's a bit confusing because we have both "compunit_symtab" and "symtab"
53169 types, and many methods and functions containing "start_symtab" or
53170 "end_symtab", which actually deal with compunit_symtabs. I believe this
53171 comes from the time before compunit_symtab was introduced, where
53172 symtab did the job of both.
53174 Rename everything I found containing start_symtab or end_symtab to use
53175 start_compunit_symtab or end_compunit_symtab.
53177 Change-Id: If3849b156f6433640173085ad479b6a0b085ade2
53179 2022-04-04 Simon Marchi <simon.marchi@efficios.com>
53181 gdb: remove some unused buildsym-legacy functions
53182 Pretty much self-explanatory.
53184 Change-Id: I5b658d017cd891ecdd1df61075eacb0f44316935
53186 2022-04-04 Fangrui Song <i@maskray.me>
53188 gas: copy st_size only if unset
53208 bar1's size is 2 while bar2, bar3, bar4's is 1. The behavior of bar1 makes sense
53209 (generally directives on the new symbol should win) and is relied upon by glibc
53210 stdio-common/errlist.c:
53213 .hidden _sys_errlist_internal
53214 .globl _sys_errlist_internal
53215 .type _sys_errlist_internal, @object
53216 .size _sys_errlist_internal, 1072
53217 _sys_errlist_internal:
53219 .globl __GLIBC_2_1_sys_errlist
53220 .set __GLIBC_2_1_sys_errlist, _sys_errlist_internal
53221 .type __GLIBC_2_1_sys_errlist, %object
53222 .size __GLIBC_2_1_sys_errlist, 125 * (64 / 8)
53224 // glibc expects that .size __GLIBC_2_1_sys_errlist, 125 * (64 / 8) wins.
53227 The behavior of bar2/bar3/bar4 seems brittle. To avoid the reordering of the two
53228 code blocks which will result in the bar3 situation, glibc compiles errlist.c
53229 with gcc -fno-toplevel-reorder (previously -fno-unit-at-a-time).
53231 To fix the inconsistency and improve robustness, make bar2/bar3/bar4 match bar1,
53232 removing the directive order sensitivity.
53234 There is a pity that `.size dest, 0` is indistinguishable from the case where
53235 dest is unset, but the compromise seems fine.
53238 * config/obj-elf.c (elf_copy_symbol_attributes): don't copy if src's size
53240 * testsuite/gas/elf/elf.exp: New test.
53241 * testsuite/gas/elf/size.d: New file.
53242 * testsuite/gas/elf/size.s: Likewise.
53244 2022-04-04 Andrew Burgess <aburgess@redhat.com>
53246 gdb: remove use of vfprintf_filtered
53249 commit 60a3da00bd5407f07d64dff82a4dae98230dfaac
53250 Date: Sat Jan 22 11:38:18 2022 +0000
53252 objdump/opcodes: add syntax highlighting to disassembler output
53254 Introduced a new use of vfprintf_filtered, which has been deprecated.
53255 This commit replaces this with gdb_vprintf instead.
53257 2022-04-04 Andrew Burgess <aburgess@redhat.com>
53259 opcodes/i386: partially implement disassembler style support
53260 This commit adds partial support for disassembler styling in the i386
53263 The i386 disassembler collects the instruction arguments into an array
53264 of strings, and then loops over the array printing the arguments out
53265 later on. The problem is that by the time we print the arguments out
53266 it's not obvious what the type of each argument is.
53268 Obviously this can be fixed, but I'd like to not do that as part of
53269 this commit, rather, I'd prefer to keep this commit as small as
53270 possible to get the basic infrastructure in place, then we can improve
53271 on this, to add additional styling, in later commits.
53273 For now then, I think this commit should correctly style mnemonics,
53274 some immediates, and comments. Everything else will be printed as
53275 plain text, which will include most instruction arguments, unless the
53276 argument is printed as a symbol, by calling the print_address_func
53279 Ignoring colours, there should be no other user visible changes in the
53280 output of the disassembler in either objdump or gdb.
53284 * disassembler.c (disassemble_init_for_target): Set
53285 created_styled_output for i386 based targets.
53286 * i386-dis.c: Changed throughout to use fprintf_styled_func
53287 instead of fprintf_func.
53289 2022-04-04 Andrew Burgess <aburgess@redhat.com>
53291 opcodes/riscv: implement style support in the disassembler
53292 Update the RISC-V disassembler to supply style information. This
53293 allows objdump to apply syntax highlighting to the disassembler
53294 output (when the appropriate command line flag is used).
53296 Ignoring colours, there should be no other user visible changes in the
53297 output of the disassembler in either objdump or gdb.
53301 * disassembler.c (disassemble_init_for_target): Set
53302 created_styled_output for riscv.
53303 * riscv-dis.c: Changed throughout to use fprintf_styled_func
53304 instead of fprintf_func.
53306 2022-04-04 Andrew Burgess <aburgess@redhat.com>
53308 objdump/opcodes: add syntax highlighting to disassembler output
53309 This commit adds the _option_ of having disassembler output syntax
53310 highlighted in objdump. This option is _off_ by default. The new
53311 command line options are:
53313 --disassembler-color=off # The default.
53314 --disassembler-color=color
53315 --disassembler-color=extended-color
53317 I have implemented two colour modes, using the same option names as we
53318 use of --visualize-jumps, a basic 8-color mode ("color"), and an
53319 extended 8bit color mode ("extended-color").
53321 The syntax highlighting requires that each targets disassembler be
53322 updated; each time the disassembler produces some output we now pass
53323 through an additional parameter indicating what style should be
53324 applied to the text.
53326 As updating all target disassemblers is a large task, the old API is
53327 maintained. And so, a user of the disassembler (i.e. objdump, gdb)
53328 must provide two functions, the current non-styled print function, and
53329 a new, styled print function.
53331 I don't currently have a plan for converting every single target
53332 disassembler, my hope is that interested folk will update the
53333 disassemblers they are interested in. But it is possible some might
53336 In this initial series I intend to convert the RISC-V disassembler
53337 completely, and also do a partial conversion of the x86 disassembler.
53338 Hopefully having the x86 disassembler at least partial converted will
53339 allow more people to try this out easily and provide feedback.
53341 In this commit I have focused on objdump. The changes to GDB at this
53342 point are the bare minimum required to get things compiling, GDB makes
53343 no use of the styling information to provide any colors, that will
53344 come later, if this commit is accepted.
53346 This first commit in the series doesn't convert any target
53347 disassemblers at all (the next two commits will update some targets),
53348 so after this commit, the only color you will see in the disassembler
53349 output, is that produced from objdump itself, e.g. from
53350 objdump_print_addr_with_sym, where we print an address and a symbol
53351 name, these are now printed with styling information, and so will have
53352 colors applied (if the option is on).
53354 Finally, my ability to pick "good" colors is ... well, terrible. I'm
53355 in no way committed to the colors I've picked here, so I encourage
53356 people to suggest new colors, or wait for this commit to land, and
53357 then patch the choice of colors.
53359 I do have an idea about using possibly an environment variable to
53360 allow the objdump colors to be customised, but I haven't done anything
53361 like that in this commit, the color choices are just fixed in the code
53364 binutils/ChangeLog:
53366 * NEWS: Mention new feature.
53367 * doc/binutils.texi (objdump): Describe --disassembler-color
53369 * objdump.c (disassembler_color): New global.
53370 (disassembler_extended_color): Likewise.
53371 (disassembler_in_comment): Likewise.
53372 (usage): Mention --disassembler-color option.
53373 (long_options): Add --disassembler-color option.
53374 (objdump_print_value): Use fprintf_styled_func instead of
53376 (objdump_print_symname): Likewise.
53377 (objdump_print_addr_with_sym): Likewise.
53378 (objdump_color_for_disassembler_style): New function.
53379 (objdump_styled_sprintf): New function.
53380 (fprintf_styled): New function.
53381 (disassemble_jumps): Use disassemble_set_printf, and reset
53382 disassembler_in_comment.
53383 (null_styled_print): New function.
53384 (disassemble_bytes): Use disassemble_set_printf, and reset
53385 disassembler_in_comment.
53386 (disassemble_data): Update init_disassemble_info call.
53387 (main): Handle --disassembler-color option.
53391 * dis-asm.h (enum disassembler_style): New enum.
53392 (struct disassemble_info): Add fprintf_styled_func field, and
53393 created_styled_output field.
53394 (disassemble_set_printf): Declare.
53395 (init_disassemble_info): Add additional parameter.
53396 (INIT_DISASSEMBLE_INFO): Add additional parameter.
53400 * dis-init.c (init_disassemble_info): Take extra parameter,
53401 initialize the new fprintf_styled_func and created_styled_output
53403 * disassembler.c (disassemble_set_printf): New function definition.
53405 2022-04-04 Tom Tromey <tom@tromey.com>
53407 Remove more Python 2 code
53408 I found another more place that still had a workaround for Python 2.
53409 This patch removes it.
53411 2022-04-04 Tom de Vries <tdevries@suse.de>
53413 [gdb/testsuite] Fix KPASS in gdb.ada/arrayptr.exp
53414 On openSUSE Leap 15.3 I run into:
53416 KPASS: gdb.ada/arrayptr.exp: scenario=minimal: print pa_ptr.all \
53417 (PRMS minimal encodings)
53418 KPASS: gdb.ada/arrayptr.exp: scenario=minimal: print pa_ptr(3) \
53419 (PRMS minimal encodings)
53420 KPASS: gdb.ada/arrayptr.exp: scenario=minimal: print pa_ptr.all(3) \
53421 (PRMS minimal encodings)
53424 The test-case KFAILs some tests. However, the analysis in the corresponding
53425 PR talks of a compiler problem, so it should use XFAILs instead.
53427 The KFAILs are setup for pre-gcc-12, but apparantly the fix has been
53428 backported to system compiler 7.5.0, hence the KPASS.
53431 - using an XFAIL instead of a KFAIL
53432 - matching the specific gdb output that corresponds to the XFAILs
53433 (reproduced on Fedora 34).
53435 Tested on x86_64-linux, specifically openSUSE Leap 15.3 and Fedora 34.
53437 2022-04-04 GDB Administrator <gdbadmin@sourceware.org>
53439 Automatic date update in version.in
53441 2022-04-03 rupothar <rupesh.potharla@amd.com>
53443 gdb: add support for Fortran's ASSUMED RANK arrays
53444 This patch adds a new dynamic property DYN_PROP_RANK, this property is
53445 read from the DW_AT_rank attribute and stored within the type just
53446 like other dynamic properties.
53448 As arrays with dynamic ranks make use of a single
53449 DW_TAG_generic_subrange to represent all ranks of the array, support
53450 for this tag has been added to dwarf2/read.c.
53452 The final piece of this puzzle is to add support in gdbtypes.c so that
53453 we can resolve an array type with dynamic rank. To do this the
53454 existing resolve_dynamic_array_or_string function is split into two,
53455 there's a new resolve_dynamic_array_or_string_1 core that is
53456 responsible for resolving each rank of the array, while the now outer
53457 resolve_dynamic_array_or_string is responsible for figuring out the
53458 array rank (which might require resolving a dynamic property) and then
53459 calling the inner core.
53461 The resolve_dynamic_range function now takes a rank, which is passed
53462 on to the dwarf expression evaluator. This rank will only be used in
53463 the case where the array itself has dynamic rank, but we now pass the
53464 rank in all cases, this should be harmless if the rank is not needed.
53466 The only small nit is that resolve_dynamic_type_internal actually
53467 handles resolving dynamic ranges itself, which now obviously requires
53468 us to pass a rank value. But what rank value to use? In the end I
53469 just passed '1' through here as a sane default, my thinking is that if
53470 we are in resolve_dynamic_type_internal to resolve a range, then the
53471 range isn't part of an array with dynamic rank, and so the range
53472 should actually be using the rank value at all.
53474 An alternative approach would be to make the rank value a
53475 gdb::optional, however, this ends up adding a bunch of complexity to
53476 the code (e.g. having to conditionally build the array to pass to
53477 dwarf2_evaluate_property, and handling the 'rank - 1' in
53478 resolve_dynamic_array_or_string_1) so I haven't done that, but could,
53479 if people think that would be a better approach.
53481 Finally, support for assumed rank arrays was only fixed very recently
53482 in gcc, so you'll need the latest gcc in order to run the tests for
53485 Here's an example test program:
53496 END SUBROUTINE sub1
53500 gcc (GCC) 12.0.0 20211122 (experimental)
53502 Compilation command:
53503 gfortran assumedrank.f90 -gdwarf-5 -o assumedrank
53508 Reading symbols from assumedrank...
53510 Breakpoint 1 at 0x4006ff: file assumedrank.f90, line 10.
53512 Starting program: /home/rupesh/STAGING-BUILD-2787/bin/assumedrank
53514 Breakpoint 1, arank::sub1 (a=<unknown type in /home/rupesh/STAGING-BUILD-2787
53515 /bin/assumedrank, CU 0x0, DIE 0xd5>) at assumedrank.f90:10
53516 10 PRINT *, RANK(a)
53517 (gdb) print RANK(a)
53518 'a' has unknown type; cast it to its declared type
53523 Reading symbols from assumedrank...
53525 Breakpoint 1 at 0x4006ff: file assumedrank.f90, line 10.
53527 Starting program: /home/rupesh/STAGING-BUILD-2787/bin/assumedrank
53529 Breakpoint 1, arank::sub1 (a=...) at assumedrank.f90:10
53530 10 PRINT *, RANK(a)
53531 (gdb) print RANK(a)
53534 type = real(kind=4) (10)
53537 Co-Authored-By: Andrew Burgess <aburgess@redhat.com>
53539 2022-04-03 Andrew Burgess <aburgess@redhat.com>
53541 gdb/dwarf: pass an array of values to the dwarf evaluator
53542 When we need to evaluate a DWARF expression in order to resolve some
53543 dynamic property of a type we call the dwarf2_evaluate_property
53544 function, which is declared in gdb/dwarf/loc.h and defined in
53547 Currently, this function takes (amongst other things) an argument of
53548 type property_addr_info called addr_stack and a boolean called
53549 push_initial_value. When push_initial_value then the top value of
53550 addr_stack is pushed onto the dwarf expression evaluation stack before
53551 the expression is evaluated.
53553 So far this has worked fine, as the only two cases we needed to handle
53554 are the case the DWARF expression doesn't require the object
53555 address (what the top of addr_stack represents), and the case where
53556 the DWARF expression does require the address.
53558 In the next commit this is going to change. As we add support for
53559 Fortran assumed rank arrays, we need to start resolving the dynamic
53560 properties of arrays. To do this, we need to push the array rank onto
53561 the dwarf expression evaluation stack before the expression is
53564 This commit is a refactoring commit aimed at making it easier to
53565 support Fortran assumed rank arrays. Instead of passing a boolean,
53566 and using this to decide if we should push the object address or not,
53567 we instead pass an array (view) of values that should be pushed to the
53568 dwarf expression evaluation stack.
53570 In the couple of places where we previously passed push_initial_value
53571 as true (mostly this was defaulting to false), we now have to pass the
53572 address from the addr_stack as an item in the array view.
53574 In the next commit, when we want to handle passing the array rank,
53575 this will easily be supported too.
53577 There should be no user visible changes after this commit.
53579 2022-04-03 Andrew Burgess <aburgess@redhat.com>
53581 gdb: small simplification in dwarf2_locexpr_baton_eval
53582 While examining the dwarf expression evaluator, I noticed that in
53583 dwarf2_locexpr_baton_eval, whenever push_initial_value is true, the
53584 addr_stack will never be nullptr.
53586 This allows for a small cleanup, replacing an if/then/else with an
53589 There should be no user visible changes after this commit.
53591 2022-04-03 Andrew Burgess <aburgess@redhat.com>
53593 gdb/testsuite: resolve some duplicate test names in gdb.base
53594 This commit resolves all the duplicate test names that I see in the
53597 gdb.base/attach-pie-misread.exp
53599 The duplicate names all come from a second call to
53600 build_executable_own_libs, so in this commit I've places the second
53601 call inside a with_test_prefix block.
53603 While I was making this change I've also modified the value being
53604 passed as the testname for the second build_executable_own_libs call.
53605 Previously we used ${test}, however, I think this was likely a
53606 mistake, the 'test' variable is setup for the previous test. I
53607 suspect that ${testfile} is a better choice - especially now we have a
53610 As the testname is only used (after various calls) from within
53611 build_executable_from_specs should the build fail, I don't think this
53612 change really makes much difference though.
53614 There should be no change in what is tested after this commit.
53616 2022-04-03 Andrew Burgess <aburgess@redhat.com>
53618 gdb/testsuite: resolve a duplicate test name in a gdb.mi test
53619 Solve two duplicate test names in the test script:
53621 gdb.mi/mi-catch-cpp-exceptions.exp
53623 by moving the call to restart_for_test inside the with_test_prefix
53624 block. There should be no difference in what is tested after this
53627 2022-04-03 Andrew Burgess <aburgess@redhat.com>
53629 gdb/Makefile.in: move ALLDEPFILES earlier in Makefile.in
53630 If I do 'make tags' in the gdb build directory, the tags target does
53631 complete, but I see these warnings:
53633 ../../src/gdb/arm.c: No such file or directory
53634 ../../src/gdb/arm-get-next-pcs.c: No such file or directory
53635 ../../src/gdb/arm-linux.c: No such file or directory
53637 The reason for this is the ordering of build rules and make variables
53638 in gdb/Makefile.in, specifically, the placement of the tags related
53639 rules, and the ALLDEPFILES variable. The ordering is like this:
53641 TAGFILES_NO_SRCDIR = .... $(ALLDEPFILES) ....
53643 TAGS: $(TAGFILES_NO_SRCDIR) ....
53644 # Recipe uses $(TAGFILES_NO_SRCDIR)
53648 ALLDEPFILES = .....
53650 When the TAGS rule is parsed TAGFILES_NO_SRCDIR is expanded, which
53651 then expands ALLDEPFILES, which, at that point in the Makefile is
53652 undefined, and so expands to the empty string. As a result TAGS does
53653 not depend on any file listed in ALLDEPFILES.
53655 However, when the TAGS recipe is invoked ALLDEPFILES is now defined.
53656 As a result, all the files in ALLDEPFILES are passed to the etags
53659 The ALLDEPFILES references three files, arm.c, arm-get-next-pcs.c, and
53660 arm-linux.c, which are actually in the gdb/arch/ directory, but, in
53661 ALLDEPFILES these files don't include the arch/ prefix. As a result,
53662 the etags program ends up looking for these files in the wrong
53665 As ALLDEPFILES is only used by the TAGS rule, this mistake was not
53666 previously noticed (the TAGS rule itself was broken until a recent
53669 In this commit I make two changes, first, I move ALLDEPFILES to be
53670 defined before TAGFILES_NO_SRCDIR, this means that the TAGS rule will
53671 depend on all the files in ALLDEPFILES. With this change the TAGS
53672 rule now breaks complaining that there's no rule to build the 3 files
53675 Next, I have added all *.c files in gdb/arch/ to ALLDEPFILES,
53676 including their arch/ prefix, and removed the incorrect (missing arch/
53677 prefix) references.
53679 With these two changes the TAGS (or tags if you prefer) target now
53680 builds without any errors or warnings.
53682 2022-04-03 Reuben Thomas <rrt@sc3d.org>
53684 gdb/Makefile.in: fix 'make tags' build target
53685 The gdb_select.h file was moved to the gdbsupport directory long ago,
53686 but a reference was accident left in gdb/Makefile.in (in the
53687 HFILES_NO_SRCDIR variable), this commit removes that reference.
53689 Before this commit, if I use 'make tags' here's what I see:
53692 make: *** No rule to make target 'gdb_select.h', needed by 'TAGS'. Stop.
53694 After this commit 'make tags' completes, but I still see these
53697 ../../src/gdb/arm.c: No such file or directory
53698 ../../src/gdb/arm-get-next-pcs.c: No such file or directory
53699 ../../src/gdb/arm-linux.c: No such file or directory
53701 These are caused by a separate issue, and will be addressed in the
53704 2022-04-03 Andrew Burgess <aburgess@redhat.com>
53706 gdb/Makefile.in: remove SOURCES variable
53707 The SOURCES variable was added to gdb/Makefile.in as part of commit:
53709 commit fb40c20903110ed8af9701ce7c2635abd3770d52
53710 Date: Wed Feb 23 00:25:43 2000 +0000
53712 Add mi/ and testsuite/gdb.mi/ subdirectories.
53714 But as far as I can tell was not used at the time it was added, and is
53719 2022-04-03 Andrew Burgess <aburgess@redhat.com>
53721 gdb/tui: fair split of delta after a resize
53722 Currently, in master gdb, when a tui window is changed in size, the
53723 screen delta is mostly just added to the next available window. We
53724 do take care to respect the min/max size, but in most cases, these
53725 limits are just "the terminal size", and so, we end up placing the
53726 whole delta on the next window.
53728 Consider these steps in an 80 column, 24 line terminal:
53734 Name Lines Columns Focus
53735 src 8 80 (has focus)
53739 (gdb) winheight cmd +2
53741 Name Lines Columns Focus
53742 src 6 80 (has focus)
53747 Notice that initially, the windows were balanced, 8 lines each for the
53748 major windows. Then, when the cmd window was adjusted, the extra two
53749 lines were given to the asm window.
53751 I think it would be nicer if the delta was spread more evenly over the
53752 available windows. In the example above, after the adjustment the
53753 layout now looks like:
53756 Name Lines Columns Focus
53757 src 7 80 (has focus)
53762 This is achieved within tui_layout_split::set_size, by just handing
53763 out the delta in increments of 1 to each window (except for the window
53764 the user adjusted), until there's no more delta left. Of course, we
53765 continue to respect the min/max window sizes.
53767 2022-04-03 Andrew Burgess <aburgess@redhat.com>
53769 gdb/tui: relax restrictions on window max height and width
53770 This commit removes some arbitrary adjustments made in
53771 tui_cmd_window::max_height, tui_win_info::max_height, and
53772 tui_win_info::max_width.
53774 These member functions all subtract some constant from the theoretical
53775 maximum height or width. I've looked back through the history a
53776 little and can see no real reason why these adjustments should be
53777 needed, with these adjustments removed all the existing tui tests
53780 However, retaining these restrictions causes some bugs, consider:
53782 (gdb) tui new-layout hsrc {-horizontal src 1 cmd 1} 1
53784 When this layout is selected with current master, gdb will leave a 4
53785 line gap at the bottom of the terminal.
53787 The problem is that the maximum height is restricted, for the cmd
53788 window, to 4 less than the terminal height.
53790 By removing this restriction gdb is able to size the windows to the
53791 complete terminal height, and the layout is done correctly.
53793 This 4 line restriction is also what prevents this layout from working
53796 (gdb) tui new-layout conly cmd 1
53798 Previously, this layout would present a cmd window only, but there
53799 would be a 4 line gap at the bottom of the terminal. This issue was
53800 mentioned in an earlier commit in this series (when a different bug
53801 was fixed), but with this commit, the above layout now correctly fills
53802 the terminal. The associated test is updated.
53804 After removing the adjustment in tui_cmd_window::max_height, the
53805 implementation is now the same as the implementation in the parent
53806 class tui_win_info, so I've completely removed the max_height call
53807 from tui_cmd_window.
53809 2022-04-03 Andrew Burgess <aburgess@redhat.com>
53811 gdb/testsuite: some additional tests in gdb.tui/scroll.exp
53812 This commit just adds an extra check of the src window size prior to
53813 sending all the commands to gdb. We also set the cmd window height to
53814 its existing height, this (obviously) shouldn't change the window
53815 layout, which we check.
53817 My main motivation was adding the initial window layout check, the
53818 winheight and recheck are just extras. All of these test pass both
53819 before and after this commit.
53821 2022-04-03 Andrew Burgess <aburgess@redhat.com>
53823 gdb/tui: support placing the cmd window into a horizontal layout
53824 This commit allows the user to place the cmd window within horizontal
53825 tui layouts. Consider this set of steps, carried out in an 80 columns
53826 by 24 lines terminal, using current master gdb:
53828 (gdb) tui new-layout hsrc { -horizontal src 1 cmd 1 } 1 status 1
53829 (gdb) tui layout hsrc
53831 What you end up with is a full width cmd window with the status bar
53832 beneath. Where's the src window gone? We then try:
53835 Name Lines Columns Focus
53836 src 23 3 (has focus)
53841 Something weird has gone on, gdb has overlapped the cmd window with
53842 the src window. If we trigger the src window to redraw is content,
53843 for example, 'list main', then we see corruption in the cmd window as
53844 the src window overwrites it.
53846 So, what's going on?
53848 The problem is some code in tui_layout_split::apply, in tui-layout.c.
53849 Within 'Step 1', there is a loop that calculates the min/max window
53850 sizes for all windows within a tui_layout_split. However, there's a
53851 special case for the cmd window.
53853 This special case is trying to have the cmd window retain its current
53854 size when a layout is re-applied, or a new layout is applied. This
53855 makes sense, consider moving from the 'src' layout to the 'asm'
53856 layout, this looks something like this (status window removed):
53858 .-------. .-------.
53860 |-------| ====> |-------|
53862 '-------' '-------'
53864 If the user has gone to the effort of adjusting the cmd window size,
53865 then, the thinking goes, we shouldn't reset the cmd window size when
53866 switching layouts like this.
53868 The problem though, is that when we do a switch more like this:
53870 .-----------. .-----------.
53872 |-----------| ====> | asm | cmd |
53874 '-----------' '-----------'
53876 Now retaining the cmd window width makes no sense; the new layout has
53877 a completely different placement for the cmd window, instead of sizing
53878 by height, we're now sizing by width. The existing code doesn't
53879 understand this though, and tried to retain the full width for the cmd
53882 To solve this problem, I propose we introduce the idea of a layout
53883 "fingerprint". The fingerprint tries to capture, in an abstract way,
53884 where the cmd window lives within the layout.
53886 Only when two layouts have the same fingerprint will we attempt to
53887 retain the cmd window size.
53889 The fingerprint for a layout is represented as a string, the string is
53890 a series of 'V' or 'H' characters, ending with a single 'C'
53891 character. The series of 'V' and 'H' characters represent the
53892 vertical or horizontal layouts that must be passed through to find the
53895 Here are a few examples:
53897 # This layout is equivalent to the builtin 'src' layout.
53899 tui new-layout example1 src 2 status 0 cmd 1
53901 # This layout is equivalent to the builtin 'split' layout.
53903 tui new-layout example2 src 1 asm 1 status 0 cmd 1
53905 # This is the same layout that was given at the top.
53907 tui new-layout hsrc { -horizontal src 1 cmd 1 } 1 status 1
53909 And so, when switching between example1 and example2, gdb knows that
53910 the cmd window is, basically, in the same sort of position within the
53911 layout, and will retain the cmd window size.
53913 In contrast, when switching to the hsrc layout, gdb understands that
53914 the position of the cmd window is different, and does not try to
53915 retain the cmd window size.
53917 2022-04-03 Andrew Burgess <aburgess@redhat.com>
53919 gdb/tui: allow cmd window to change size in tui_layout_split::apply
53920 When we switch layouts we call the tui_layout_split::apply member
53921 function to reapply the layout, and recalculate all the window sizes.
53923 One special case is the cmd window, which we try to keep at its
53926 However, in some cases it is not appropriate to keep the cmd window at
53927 its existing size. I will describe two such cases here, in one, we
53928 want the cmd window to reduce in size, and in the other, we want the
53929 cmd window to grow in size.
53931 Try these steps in a 80 columns, by 24 lines terminal:
53935 (gdb) winheight cmd 20
53938 You should see that the status window is missing from the new layout,
53939 and that the cmd window has been placed over the border of the asm
53940 window. The 'info win' output is:
53943 Name Lines Columns Focus
53944 src 3 80 (has focus)
53949 Notice that gdb has assigned 27 lines of screen space, even with the
53950 border overlap between the src and asm windows, this is still 2 lines
53953 The problem here is that after switching layouts, gdb has forced the
53954 cmd window to retain its 20 line height. Really, we want the cmd
53955 window to reduce in height so that the src and asm windows can occupy
53956 their minimum required space.
53958 This commit allows this (details on how are below). After this
53959 commit, in the above situation, we now see the status window displayed
53960 correctly, and the 'info win' output is:
53963 Name Lines Columns Focus
53964 src 3 80 (has focus)
53969 The cmd window has been reduced in size by 2 lines so that everything
53970 can fit on the screen.
53972 The second example is one which was discussed in a recent commit,
53973 consider this case (still in the 80 column, 24 line terminal):
53976 (gdb) tui new-layout conly cmd 1
53979 Name Lines Columns Focus
53980 cmd 8 80 (has focus)
53983 This layout only contains a cmd window, which we would expect to
53984 occupy the entire terminal. But instead, the cmd window only occupies
53985 the first 8 lines, and the rest of the terminal is unused!
53987 The reason is, again, that the cmd window is keeping its previous
53990 After this commit things are slightly different, the 'info win' output
53994 Name Lines Columns Focus
53995 cmd 20 80 (has focus)
53997 Which is a little better, but why only 20 lines? Turns out there's
53998 yet another bug hitting this case. That bug will be addressed in a
53999 later commit, so, for now, we're accepting the 20 lines.
54001 What this commit does is modify the phase of tui_layout_split::apply
54002 that handles any left over space. Usually, in "Step 2", each
54003 sub-layout has a size calculated. As the size is an integer, then,
54004 when all sizes are calculated we may have some space left over.
54006 This extra space is then distributed between all the windows fairly
54007 until all the space is used up.
54009 When we consider windows minimum size, or fixed size windows, then it
54010 is possible that we might try to use more space than is available,
54011 this was our first example above. The same code that added extra
54012 space to the windows, can also be used to reclaim space (in the over
54013 allocation case) to allow all windows to fit.
54015 The problem then is the cmd window, which we often force to a fixed
54016 size. Inside the loop that handles the allocation of excess space, if
54017 we find that we have tried every window, and still have space either
54018 left to give, or we need to claim back more space, then, if the cmd
54019 window was changed to a fixed size, we can change the cmd window back
54020 to a non-fixed-size window, and proceed to either give, or take space
54021 from the cmd window as needed.
54023 2022-04-03 Andrew Burgess <aburgess@redhat.com>
54025 gdb/tui: fairer distribution of excess space during apply
54026 When applying layouts gdb computes the size of each window (or rather,
54027 each sub-layout within a layout) using integer arithmetic. As this
54028 rounds down the results, then, when all sub-layouts are sized, there
54029 is the possibility that we have some space left over.
54031 Currently, this space is just assigned to an arbitrary sub-layout.
54033 This can result in some unbalanced results. Consider this set of
54034 steps with current master:
54039 Name Lines Columns Focus
54041 src 9 80 (has focus)
54045 Notice the weird split between the src and regs windows, the original
54046 layout specification has these windows given equal weight. The
54047 problem is that, with rounding, both the regs and src windows are
54048 initially sized to 7, the extra 2 lines are then arbitrarily added to
54051 In this commit, rather than add all the extra space to one single
54052 window, I instead hand out the extra space 1 line at a time, looping
54053 over all the sub-layouts. We take care to respect the min/max sizes,
54054 and so, we now get this result:
54059 Name Lines Columns Focus
54061 src 8 80 (has focus)
54065 This looks more natural to me.
54067 This is obviously a change in behaviour, and so, lots of the existing
54068 tests need to be updated to take this into account. None of the
54069 changes are huge, it's just a line or two (or column for width) moved
54072 2022-04-03 Andrew Burgess <aburgess@redhat.com>
54074 gdb/tui: avoid fp exception when applying layouts
54079 (gdb) tui new-layout conly cmd 1
54082 After this, with current master, gdb crashes with a floating-point
54085 The problem is that in tui_layout_split::apply, when we switch from
54086 'src' to 'conly', we will try to retain the cmd window height. As
54087 such, the cmd window will become a fixed size window, which decreases
54088 the available_size, but doesn't count towards the total_weight.
54090 As the cmd window is the only window, the total_weight stays at zero,
54091 and, when we move into step 2, where we attempt to size the windows,
54092 we perform a divide by zero, and crash.
54094 After this commit we avoid the divide by zero, and just directly set
54095 the window size based on the fixed size.
54097 There is still a problem after this commit, when the conly layout is
54098 selected the cmd window retains its original height, which will only
54099 be part of the terminal. The rest of the terminal is left unused.
54100 This issue will be addressed in a later commit, this commit is just
54101 about the floating-point exception.
54103 2022-04-03 Andrew Burgess <aburgess@redhat.com>
54105 gdb/tui/testsuite: refactor new-layout.exp test
54106 This commit changes the gdb.tui/new-layout.exp test to make use of a
54107 list of test descriptions, and a loop to check each description in
54108 turn. There's no change to what is actually tested after this commit.
54110 In future commits I plan to add additional tests to this file, and
54111 this will be easier now that all I have to do is add a new test
54112 description to the list.
54114 2022-04-03 Andrew Burgess <aburgess@redhat.com>
54116 gdb/tui: add left_boxed_p and right_boxed_p member functions
54117 When I initially saw this code in tui_layout_split::apply, I assumed
54118 that this must be a bug:
54120 /* Two adjacent boxed windows will share a border, making a bit
54121 more size available. */
54123 && m_splits[i - 1].layout->bottom_boxed_p ()
54124 && m_splits[i].layout->top_boxed_p ())
54127 After all, the apply might be laying out a horizontal layout, right?
54128 So checking bottom_boxed_p and top_boxed_p is clearly wrong.
54130 Well, it turns on, that due to the implementations of these things,
54131 bottom_boxed_p is equivalent to an imagined right_boxed_p, and
54132 top_boxed_p is equivalent to an imagined left_boxed_p.
54134 In this commit I've renamed both top_boxed_p and bottom_boxed_p to
54135 first_edge_has_border_p and last_edge_has_border_p respectively, and
54136 extended the comments in tui_layout_base to mention that these methods
54137 handle both horizontal and vertical layouts.
54139 Now, hopefully, the code shouldn't look like it only applies for
54142 There should be no user visible changes after this commit.
54144 2022-04-03 Andrew Burgess <aburgess@redhat.com>
54146 gdb/tui: add a tui debugging flag
54147 This commit adds 'set debug tui on|off' and 'show debug tui'. This
54148 commit adds the control variable, and the printing macros in
54149 tui/tui.h. I've then added some uses of these in tui.c and
54152 To help produce more useful debug output in tui-layout.c, I've added
54153 some helper member functions in the class tui_layout_split, and also
54154 moved the size_info struct out of tui_layout_split::apply into the
54155 tui_layout_split class.
54157 If tui debug is not turned on, then there should be no user visible
54158 changes after this commit.
54160 One thing to note is that, due to the way that the tui terminal is
54161 often cleared, the only way I've found this useful is when I do:
54164 (gdb) set logging file /path/to/file
54165 (gdb) set logging debugredirect on
54166 (gdb) set logging enable on
54168 Additionally, gdb has some quirks when it comes to setting up logging
54169 redirect and switching interpreters. Thus, the above only really
54170 works if the logging is enabled after the tui is enabled, and disabled
54171 again before the tui is disabled.
54173 Enabling logging and switching interpreters can cause undefined
54174 results, including crashes. This is an existing bug in gdb[1], and
54175 has nothing directly to do with tui debug, but it is worth mentioning
54178 [1] https://sourceware.org/bugzilla/show_bug.cgi?id=28948
54180 2022-04-03 Andrew Burgess <aburgess@redhat.com>
54182 gdb/tui: add new 'tui window width' command and 'winwidth' alias
54183 This commit adds a new command 'tui window width', and an alias
54184 'winwidth'. This command is equivalent to the old 'winheight'
54185 command (which was recently renamed 'tui window height').
54187 Even though I recently moved the old tui commands under the tui
54188 namespace, and I would strongly encourage all new tui commands to be
54189 added as 'tui ....' only (users can create their own top-level aliases
54190 if they want), I'm breaking that suggestion here, and adding a
54193 Given that we already have 'winheight' and have done for years, it
54194 just didn't seem right to no have the matching 'winwidth'.
54196 You might notice in the test that the window resizing doesn't quite
54197 work right. I setup a horizontal layout, then grow and shrink the
54198 windows. At the end of the test the windows should be back to their
54201 ... they are not. This isn't my fault, honest! GDB's window resizing
54202 is a little ... temperamental, and is prone to getting things slightly
54203 wrong during resizes, off by 1 type things. This is true for height
54204 resizing, as well as the new width resizing.
54206 Later patches in this series will rework the resizing algorithm, which
54207 should improve things in this area. For now, I'm happy that the width
54208 resizing is as good as the height resizing, given the existing quirks.
54210 For the docs side I include a paragraph that explains how multiple
54211 windows are required before the width can be adjusted. For
54212 completeness, I've added the same paragraph to the winheight
54213 description. With the predefined layouts this extra paragraph is not
54214 really needed for winheight, as there are always multiple windows on
54215 the screen. However, with custom layouts, this might not be true, so
54216 adding the paragraph seems like a good idea.
54218 As for the changes in gdb itself, I've mostly just taken the existing
54219 height adjustment code, changed the name to make it generic 'size'
54220 adjustment, and added a boolean flag to indicate if we are adjusting
54221 the width or the height.
54223 2022-04-03 Andrew Burgess <aburgess@redhat.com>
54225 gdb/tui: rename tui_layout_split:set_weights_from_heights
54226 In a following commit I'm going to add the ability to change the width
54227 of a tui window (when in a horizontal layout). As a result, some of
54228 the places where we currently hard-code references to height need to
54229 be changed to handle either height, or width, based on whether we are
54230 in a vertical, or horizontal layout.
54232 This commit renames set_weights_from_heights to
54233 set_weights_from_sizes, and makes the function use either the height,
54234 or width as appropriate.
54236 Currently, the only place that we call this function is from the
54237 tui_layout_split::set_height function, in a part of the code we will
54238 only reach for vertical layouts, so the new code is not actually being
54239 used, but, this small change will help make later patches smaller, so
54240 I'm proposing this as a stand alone change.
54242 There should be no user visible changes after this commit.
54244 2022-04-03 Andrew Burgess <aburgess@redhat.com>
54246 gdb/tui: rename tui_layout_base::adjust_size to ::set_height
54247 Rename tui_layout_base::adjust_size to tui_layout_base::set_height,
54248 the new name more accurately reflects what this member function does,
54249 and makes it easier for a later commit to add a new
54250 tui_layout_base::set_width member function.
54252 There should be no user visible changes after this commit.
54254 2022-04-03 Andrew Burgess <aburgess@redhat.com>
54256 gdb: move some commands into the tui namespace
54257 There are a lot of tui related commands that live in the top-level
54258 command name space, e.g. layout, focus, refresh, winheight.
54260 Having them at the top level means less typing for the user, which is
54261 good, but, I think, makes command discovery harder.
54263 In this commit, I propose moving all of the above mentioned commands
54264 into the tui namespace, so 'layout' becomes 'tui layout', etc. But I
54265 will then add aliases so that the old commands will still work,
54266 e.g. I'll make 'layout' an alias for 'tui layout'.
54268 The benefit I see in this work is that tui related commands can be
54269 more easily discovered by typing 'tui ' and then tab-completing. Also
54270 the "official" command is now a tui-sub-command, this is visible in,
54271 for example, the help output, e.g.:
54275 Change the layout of windows.
54276 Usage: tui layout prev | next | LAYOUT-NAME
54278 List of tui layout subcommands:
54280 tui layout asm -- Apply the "asm" layout.
54281 tui layout next -- Apply the next TUI layout.
54282 tui layout prev -- Apply the previous TUI layout.
54283 tui layout regs -- Apply the TUI register layout.
54284 tui layout split -- Apply the "split" layout.
54285 tui layout src -- Apply the "src" layout.
54287 Which I think is a good thing, it makes it clearer that this is a tui
54290 I've added a NEWS entry and updated the docs to mention the new and
54291 old command names, with the new name being mentioned first.
54293 2022-04-03 Simon Marchi <simon.marchi@polymtl.ca>
54295 gdb: fix gdb_print -> gdb_printf typo
54296 This caused a build failure with !CXX_STD_THREAD.
54298 Change-Id: I30f0c89c43a76f85c0db34809192644fa64a9d18
54300 2022-04-03 Alan Modra <amodra@gmail.com>
54302 Move microblaze relax info to target specific data
54303 Target specific data shouldn't be put in struct bfd_section.
54305 * section.c (struct bfd_section): Delete relax and relax_count.
54306 (BFD_FAKE_SECTION): Adjust to suit.
54307 (struct relax_table): Move to..
54308 * elf32-microblaze.c (struct relax_table): ..here.
54309 (struct _microblaze_elf_section_data): New.
54310 (microblaze_elf_section_data): Define.
54311 (microblaze_elf_new_section_hook): New function.
54312 (bfd_elf32_new_section_hook): Define.
54313 (calc_fixup): Return a size_t. Adjust to suit new location of
54314 relax and relax_count.
54315 (microblaze_elf_relax_section): Adjust to suit new location of
54316 relax and relax_count. Make some variables size_t.
54317 * bfd-in2.h: Regenerate.
54319 2022-04-03 Alan Modra <amodra@gmail.com>
54321 Revert commit 240d6706c6a2
54325 * dwarf2.c (lookup_address_in_line_info_table): Return bool rather
54327 (comp_unit_find_nearest_line): Likewise. Return true if function
54328 info found without line info.
54329 (_bfd_dwarf2_find_nearest_line): Revert range handling code.
54331 Regen bfd po/SRC-POTFILES.in
54333 2022-04-03 GDB Administrator <gdbadmin@sourceware.org>
54335 Automatic date update in version.in
54337 2022-04-02 Tiezhu Yang <yangtiezhu@loongson.cn>
54339 gdb: rename floatformats_ia64_quad to floatformats_ieee_quad
54340 It is better to rename floatformats_ia64_quad to floatformats_ieee_quad
54341 to reflect the reality, and then we can clean up the related code.
54343 As Tom Tromey said [1]:
54345 These files are maintained in gcc and then imported into the
54346 binutils-gdb repository, so any changes to them will have to
54347 be proposed there first.
54349 the related changes have been merged into gcc master now [2], it is time
54352 [1] https://sourceware.org/pipermail/gdb-patches/2022-March/186569.html
54353 [2] https://gcc.gnu.org/git/?p=gcc.git;a=commit;h=b2dff6b2d9d6
54355 2022-04-02 GDB Administrator <gdbadmin@sourceware.org>
54357 Automatic date update in version.in
54359 2022-04-01 John Baldwin <jhb@FreeBSD.org>
54361 Remove unused variable.
54363 2022-04-01 Aaron Merey <amerey@redhat.com>
54365 gdb/debuginfod-support.c: Always display debuginfod errors
54366 Errors encountered when downloading files from debuginfod servers
54367 are not displayed if debuginfod verbosity is set to 0 (via
54368 'set debuginfod verbose 0').
54370 Tom recommended that these errors always be displayed, regardless
54371 of the verbosity setting [1]. Fix this.
54373 [1] https://sourceware.org/pipermail/gdb-patches/2022-March/186350.html
54375 2022-04-01 John Baldwin <jhb@FreeBSD.org>
54377 Use I386_GSBASE_REGNUM in i386fbsd_get_thread_local_address.
54378 32-bit x86 arches always the I386_*BASE_REGNUM values. Only code that
54379 needs to support both 64-bit and 32-bit arches needs to use
54380 tdep->fsbase_regnum to compute a segment base register number.
54382 FreeBSD/x86: Read segment base registers from NT_X86_SEGBASES.
54383 FreeBSD kernels recently grew a new register core dump note containing
54384 the base addresses of the %fs and %gs segments (corresponding to the
54385 %fsbase and %gsbase registers). Parse this note to permit inspecting
54386 TLS variables in core dumps. Native processes already supported TLS
54387 via older ptrace() operations.
54389 2022-04-01 John Baldwin <jhb@FreeBSD.org>
54391 Use pseudosections for NT_FREEBSD_X86_SEGBASES core dump notes.
54392 This includes adding pseudosections when reading a core dump as well
54393 as support for writing out a core dump note from a pseudosection.
54397 * elf-bfd.h (elfcore_write_x86_segbases): New.
54398 * elf.c (elfcore_grok_freebsd_note): Add pseudosections for
54399 NT_FREEBSD_X86_SEGBASES register notes.
54400 (elfcore_write_x86_segbases): New.
54401 (elfcore_write_register_note): Write NT_FREEBSD_X86_SEGBASES
54404 2022-04-01 John Baldwin <jhb@FreeBSD.org>
54406 Recognize FreeBSD core dump note for x86 segment base registers.
54407 This core dump note contains the value of the base address of the %fs
54408 and %gs segments for both i386 and amd64 core dumps. It is primarily
54409 useful in resolving the address of TLS variables in core dumps.
54411 binutils/ChangeLog:
54413 * readelf.c (get_freebsd_elfcore_note_type): Handle
54414 NT_FREEBSD_X86_SEGBASES.
54418 * elf/common.h (NT_FREEBSD_X86_SEGBASES): Define.
54420 2022-04-01 John Baldwin <jhb@FreeBSD.org>
54422 elfcore_grok_freebsd_note: Remove checks of note->namesz.
54423 This function is only called if the note name is "FreeBSD", so
54424 checking the name size is unnecessary.
54428 * elf.c (elfcore_grok_freebsd_note): Remove checks for namesz.
54430 2022-04-01 Andrew Burgess <aburgess@redhat.com>
54432 gdb/testing/tui: add new _csi_{L,S,T}
54433 This patch was original part of this series:
54435 https://sourceware.org/pipermail/gdb-patches/2022-March/186429.html
54436 https://sourceware.org/pipermail/gdb-patches/2022-March/186433.html
54438 I've pulled this out as it might be useful ahead of the bigger series
54443 _csi_L - insert line
54447 2022-04-01 H.J. Lu <hjl.tools@gmail.com>
54449 x86: Remove bfd_arch_l1om and bfd_arch_k1om
54450 Remove bfd_arch_l1om and bfd_arch_k1om since L1OM/K1OM support has been
54451 removed from gas, ld and opcodes.
54455 * Makefile.am (ALL_MACHINES): Remove cpu-l1om.lo and cpu-k1om.lo.
54456 (ALL_MACHINES_CFILES): Remove cpu-l1om.c and cpu-k1om.c.
54457 * archures.c (bfd_mach_l1om): Removed.
54458 (bfd_mach_l1om_intel_syntax): Likewise.
54459 (bfd_mach_k1om): Likewise.
54460 (bfd_mach_k1om_intel_syntax): Likewise.
54461 (bfd_k1om_arch): Likewise.
54462 (bfd_l1om_arch): Likewise.
54463 (bfd_archures_list): Remove bfd_k1om_arch and bfd_l1om_arch
54465 * config.bfd (targ_selvecs): Remove l1om_elf64_vec.
54466 l1om_elf64_fbsd_vec, k1om_elf64_vec and k1om_elf64_fbsd_vec.
54467 (targ_archs): Remove bfd_l1om_arch and bfd_k1om_arch.
54468 * configure.ac (k1om_elf64_vec): Removed.
54469 (k1om_elf64_fbsd_vec): Likewise.
54470 (l1om_elf64_vec): Likewise.
54471 (l1om_elf64_fbsd_vec): Likewise.
54472 * cpu-k1om.c: Removed.
54473 * cpu-l1om.c: Likewise.
54474 * elf64-x86-64.c (elf64_l1om_elf_object_p): Removed.
54475 (elf64_k1om_elf_object_p): Likewise.
54476 (l1om_elf64_vec): Removed.
54477 (l1om_elf64_fbsd_vec): Likewise.
54478 (k1om_elf64_vec): Likewise.
54479 (k1om_elf64_fbsd_vec): Likewise.
54480 (ELF_TARGET_OS): Undefine.
54481 * targets.c (_bfd_target_vector): Remove k1om_elf64_vec,
54482 k1om_elf64_fbsd_vec, l1om_elf64_vec and l1om_elf64_fbsd_vec.
54483 * Makefile.in: Regenerate.
54484 * bfd-in2.h: Likewise.
54485 * configure: Likewise.
54489 * configure.ac: Remove bfd_arch_l1om/bfd_arch_k1om references.
54490 * disassemble.c (disassembler): Likewise.
54491 * configure: Regenerate.
54493 2022-04-01 Simon Marchi <simon.marchi@polymtl.ca>
54495 gdb/ctf: pass partial symtab's filename to buildsym_compunit
54496 I noticed that the CTF symbol reader passes the objfile's name to all
54497 buildsym_compunit instances it creates. The result is that all
54498 compunit_symtabs created have the same name, that of the objfile:
54500 { objfile /tmp/babeltrace-ctf/src/lib/.libs/libbabeltrace2.so.0.0.0 ((struct objfile *) 0x613000005d00)
54501 { ((struct compunit_symtab *) 0x621000286760)
54504 name libbabeltrace2.so.0.0.0
54506 blockvector ((struct blockvector *) 0x6210003911d0)
54507 user ((struct compunit_symtab *) (null))
54508 { symtab /tmp/babeltrace-ctf/src/lib/.libs/libbabeltrace2.so.0.0.0 ((struct symtab *) 0x6210003911f0)
54510 linetable ((struct linetable *) 0x0)
54513 { ((struct compunit_symtab *) 0x621000275c10)
54516 name libbabeltrace2.so.0.0.0
54518 blockvector ((struct blockvector *) 0x621000286710)
54519 user ((struct compunit_symtab *) (null))
54520 { symtab /tmp/babeltrace-ctf/src/lib/.libs/libbabeltrace2.so.0.0.0 ((struct symtab *) 0x621000286730)
54522 linetable ((struct linetable *) 0x0)
54526 Notice the two "name libbabeltrace2.so.0.0.0".
54528 Change it to pass the partial_symtab's filename instead. The output
54531 { objfile /tmp/babeltrace-ctf/src/lib/.libs/libbabeltrace2.so.0.0.0 ((struct objfile *) 0x613000005d00)
54532 { ((struct compunit_symtab *) 0x621000295610)
54535 name libbabeltrace2.so.0.0.0
54537 blockvector ((struct blockvector *) 0x6210003a15d0)
54538 user ((struct compunit_symtab *) (null))
54539 { symtab /tmp/babeltrace-ctf/src/lib/.libs/libbabeltrace2.so.0.0.0 ((struct symtab *) 0x6210003a15f0)
54541 linetable ((struct linetable *) 0x0)
54544 { ((struct compunit_symtab *) 0x621000288700)
54547 name current-thread.c
54549 blockvector ((struct blockvector *) 0x6210002955c0)
54550 user ((struct compunit_symtab *) (null))
54551 { symtab /home/simark/src/babeltrace/src/lib/current-thread.c ((struct symtab *) 0x6210002955e0)
54553 linetable ((struct linetable *) 0x0)
54557 Note that the first compunit_symtab still has libbabeltrace2.so.0.0.0 as
54558 its name. This is because the CTF symbol reader really creates a
54559 partial symtab named like this. It appears to be because the debug info
54560 contains information that has been factored out of all CUs and is at the
54561 "top-level" of the objfile, outside any real CU. So it creates a
54562 partial symtab and an artificial CU that's named after the objfile.
54564 Change-Id: I576316bab2a3668adf87b4e6cebda900a8159b1b
54566 2022-04-01 Simon Marchi <simon.marchi@polymtl.ca>
54568 gdb: print compunit_symtab name in "maint info symtabs"
54569 I think it would make sense to print a compunit_symtab's name in "maint
54570 info symtabs". If you are looking for a given CU in the list, that's
54571 probably the field you will be looking at. As the doc of
54572 compunit_symtab::name says, it is not meant to be a reliable file name,
54573 it is for debugging purposes (and "maint info symtabs" exists for
54574 debugging purposes).
54576 Sample output with the new field:
54578 (gdb) maintenance info symtabs
54579 { objfile /home/simark/build/binutils-gdb-one-target/gdb/a.out ((struct objfile *) 0x613000005d00)
54580 { ((struct compunit_symtab *) 0x621000131630)
54581 debugformat DWARF 5
54582 producer GNU C17 11.2.0 -mtune=generic -march=x86-64 -g3 -O0
54584 dirname /home/simark/build/binutils-gdb-one-target/gdb
54585 blockvector ((struct blockvector *) 0x621000131d10)
54586 user ((struct compunit_symtab *) (null))
54587 { symtab test.c ((struct symtab *) 0x6210001316b0)
54589 linetable ((struct linetable *) 0x621000131d40)
54591 { symtab /home/simark/build/binutils-gdb-one-target/gdb/test.h ((struct symtab *) 0x6210001316e0)
54593 linetable ((struct linetable *) 0x0)
54595 { symtab /usr/include/stdc-predef.h ((struct symtab *) 0x621000131710)
54597 linetable ((struct linetable *) 0x0)
54600 { ((struct compunit_symtab *) 0x6210001170a0)
54601 debugformat DWARF 5
54602 producer GNU C17 11.2.0 -mtune=generic -march=x86-64 -g3 -O0
54604 dirname /home/simark/build/binutils-gdb-one-target/gdb
54605 blockvector ((struct blockvector *) 0x621000131580)
54606 user ((struct compunit_symtab *) (null))
54607 { symtab foo.c ((struct symtab *) 0x621000117120)
54609 linetable ((struct linetable *) 0x6210001315b0)
54611 { symtab /usr/include/stdc-predef.h ((struct symtab *) 0x621000117150)
54613 linetable ((struct linetable *) 0x0)
54618 Change-Id: I17b87adfac2f6551cb5bda30d59f6c6882789211
54620 2022-04-01 Simon Marchi <simon.marchi@polymtl.ca>
54622 gdb/ctf: don't create a buildsym_compunit when building partial symbols
54623 I am trying to do some changes to buildsym_compunit, so I am auditing
54624 the current uses. Something seems odd with this use of
54625 buildsym_compunit (that this patch removes).
54627 A buildsym_compunit is normally used when building a compunit_symtab.
54628 That is, when expanding a partial symtab into a full compunit symtab.
54629 In ctfread.c, a buildsym_compunit is created in ctf_start_archive, which
54630 is only used when creating partial symtabs. At this moment, I don't
54631 see how that's useful. ctf_start_archive creates a new
54632 buildsym_compunit and starts a subfile. But that buildsym_compunit is
54633 never used again. It's just overriden in ctf_start_symtab, which means
54634 we leak the old buildsym_compunit, I suppose.
54636 Remove ctf_start_archive completely. Add an assert in
54637 ctf_start_symtab to verify that we are not overwriting an existing
54638 buildsym_compunit (meaning we'd leak the existing one). This assert
54639 triggers without the other part of the fix. When doing:
54641 $ ./gdb --data-directory=data-directory /tmp/babeltrace-ctf/src/lib/.libs/libbabeltrace2.so.0.0.0
54643 (gdb) maintenance expand-symtabs
54644 /home/simark/src/binutils-gdb/gdb/ctfread.c:1255: internal-error: ctf_start_symtab: Assertion `!ccp->builder' failed.
54646 Change-Id: I666d146454a019f08e7305f3a1c4a974d27b4592
54648 2022-04-01 Tom Tromey <tom@tromey.com>
54650 Style URLs in GDB output
54651 I noticed that GDB will display URLs in a few spots. This changes
54652 them to be styled. Originally I thought I'd introduce a new "url"
54653 style, but there aren't many places to use this, so I just reused
54654 filename styling instead. This patch also changes the debuginfod URL
54655 list to be printed one URL per line. I think this is probably a bit
54658 2022-04-01 GDB Administrator <gdbadmin@sourceware.org>
54660 Automatic date update in version.in
54662 2022-03-31 Simon Marchi <simon.marchi@polymtl.ca>
54664 gdb: initialize ctf_context::builder in create_partial_symtab
54665 I built a random project with -gctf, in order to test the CTF support in
54666 GDB. With my ASan/UBSan/etc-enabled build of GDB, I get:
54668 $ ./gdb --data-directory=data-directory /tmp/babeltrace-ctf/src/lib/.libs/libbabeltrace2.so.0.0.0
54670 Reading symbols from /tmp/babeltrace-ctf/src/lib/.libs/libbabeltrace2.so.0.0.0...
54671 /home/simark/src/binutils-gdb/gdb/ctfread.c:1545:31: runtime error: member call on misaligned address 0xbebebebebebebebe for type 'struct buildsym_compunit', which requires 8 byte alignment
54672 0xbebebebebebebebe: note: pointer points here
54674 The 0xbebebebebebebebe value is a sign that the ctf_context::builder
54675 field is uninitialized. The problem probably goes under the radar if
54676 the field happens to be zero-initialized, because ctf_start_archive
54677 contains this code:
54679 if (ccx->builder == nullptr)
54681 ccx->builder = new buildsym_compunit (of,
54682 of->original_name, nullptr, language_c, 0);
54684 If the field was zero-initialized (by chance), this will create a new
54685 buildsym_compunit. But if the field was purposely filled with random
54686 bytes by one of the sanitizers, we won't create a buildsym_compunit here
54687 and we'll continue with ccx->builder equal to 0xbebebebebebebebe.
54689 Fix this the easy way by initializing ccx->builder where the other
54690 ctf_context fields are initialized (yeah, this code could be made nicer
54691 C++, but I am going for the obvious fix here).
54693 With this patch, this passes cleanly on my system:
54695 $ make check TESTS="gdb.ctf/*.exp" RUNTESTFLAGS="CC_FOR_TARGET=/opt/gcc/git/bin/gcc"
54696 # of expected passes 40
54698 ... where /opt/gcc/git/bin/gcc is a gcc with CTF support, given my
54699 system gcc does not have it.
54701 Change-Id: Idea1b0cf3e3708b72ecb16b1b60222439160f9b9
54703 2022-03-31 Tom Tromey <tromey@adacore.com>
54706 This patch removes gdb's dbx mode. Regression tested on x86-64 Fedora
54709 2022-03-31 H.J. Lu <hjl.tools@gmail.com>
54711 gdb: Consolidate 32bit-pkeys.xml and 64bit-pkeys.xml
54712 1. Since 32bit-pkeys.xml and 64bit-pkeys.xml are identical, consolidate
54713 them into a single keys.xml.
54714 2. Enable PKU for x32 to fix:
54716 $ gdbserver :123456 x32-program
54718 .../gdbserver/regcache.cc:255: A problem internal to GDBserver has been detected
54720 Unknown register pkru requested
54724 2022-03-31 Simon Marchi <simon.marchi@polymtl.ca>
54726 gdb/linux-nat: remove check based on current_inferior in linux_handle_extended_wait
54727 The check removed by this patch, using current_inferior, looks wrong.
54728 When debugging multiple inferiors with the Linux native target and
54729 linux_handle_extended_wait is called, there's no guarantee about which
54730 is the current inferior. The vfork-done event we receive could be for
54731 any inferior. If the vfork-done event is for a non-current inferior, we
54732 end up wrongfully ignoring it. As a result, the core never processes a
54733 TARGET_WAITKIND_VFORK_DONE event, program_space::breakpoints_not_allowed
54734 is never cleared, and breakpoints are never reinserted. However,
54735 because the Linux native target decided to ignore the event, it resumed
54736 the thread - while breakpoints out. And that's bad.
54738 The proposed fix is to remove this check. Always report vfork-done
54739 events and let infrun's logic decide if it should be ignored. We don't
54740 save much cycles by filtering the event here.
54742 Add a test that replicates the situation described above. See comments
54743 in the test for more details.
54745 Change-Id: Ibe33c1716c3602e847be6c2093120696f2286fbf
54747 2022-03-31 Simon Marchi <simon.marchi@polymtl.ca>
54749 gdbserver/linux: set lwp !stopped when failing to resume
54750 I see some failures, at least in gdb.multi/multi-re-run.exp and
54751 gdb.threads/interrupted-hand-call.exp. Running `stress -C $(nproc)` at
54752 the same time as the test makes those tests relatively frequent.
54754 Let's take gdb.multi/multi-re-run.exp as an example. The failure looks
54755 like this, an unexpected "no resumed":
54759 No unwaited-for children left.
54760 (gdb) FAIL: gdb.multi/multi-re-run.exp: re_run_inf=2: iter=1: continue until exit
54764 - Inferior 1 is stopped somewhere, it won't really play a role here.
54765 - Inferior 2 has 2 threads, both stopped.
54766 - We resume inferior 2, the leader thread is expected to exit, making
54769 From GDB's perspective, a failing run looks like this:
54771 [infrun] fetch_inferior_event: enter
54772 [infrun] scoped_disable_commit_resumed: reason=handling event
54773 [infrun] do_target_wait: Found 2 inferiors, starting at #1
54774 [infrun] random_pending_event_thread: None found.
54775 [remote] wait: enter
54776 [remote] Packet received: T0506:20dcffffff7f0000;07:20dcffffff7f0000;10:9551555555550000;thread:pae4cd.ae4cd;core:e;
54777 [remote] wait: exit
54778 [infrun] print_target_wait_results: target_wait (-1.0.0 [process -1], status) =
54779 [infrun] print_target_wait_results: 713933.713933.0 [Thread 713933.713933],
54780 [infrun] print_target_wait_results: status->kind = STOPPED, sig = GDB_SIGNAL_TRAP
54781 [infrun] handle_inferior_event: status->kind = STOPPED, sig = GDB_SIGNAL_TRAP
54782 [infrun] clear_step_over_info: clearing step over info
54783 [infrun] context_switch: Switching context from 0.0.0 to 713933.713933.0
54784 [infrun] handle_signal_stop: stop_pc=0x555555555195
54785 [infrun] start_step_over: enter
54786 [infrun] start_step_over: stealing global queue of threads to step, length = 0
54787 [infrun] operator(): step-over queue now empty
54788 [infrun] start_step_over: exit
54789 [infrun] process_event_stop_test: no stepping, continue
54790 [remote] Sending packet: $Z0,555555555194,1#8e
54791 [remote] Packet received: OK
54792 [infrun] resume_1: step=0, signal=GDB_SIGNAL_0, trap_expected=0, current thread [713933.713933.0] at 0x555555555195
54793 [remote] Sending packet: $QPassSignals:e;10;14;17;1a;1b;1c;21;24;25;2c;4c;97;#0a
54794 [remote] Packet received: OK
54795 [remote] Sending packet: $vCont;c:pae4cd.-1#9f
54796 [infrun] prepare_to_wait: prepare_to_wait
54797 [infrun] reset: reason=handling event
54798 [infrun] maybe_set_commit_resumed_all_targets: enabling commit-resumed for target extended-remote
54799 [infrun] maybe_call_commit_resumed_all_targets: calling commit_resumed for target extended-remote
54800 [infrun] maybe_call_commit_resumed_all_targets: calling commit_resumed for target extended-remote
54801 [infrun] fetch_inferior_event: exit
54802 [infrun] fetch_inferior_event: enter
54803 [infrun] scoped_disable_commit_resumed: reason=handling event
54804 [infrun] do_target_wait: Found 2 inferiors, starting at #0
54805 [infrun] random_pending_event_thread: None found.
54806 [remote] wait: enter
54807 [remote] Packet received: N
54808 [remote] wait: exit
54809 [infrun] print_target_wait_results: target_wait (-1.0.0 [process -1], status) =
54810 [infrun] print_target_wait_results: -1.0.0 [process -1],
54811 [infrun] print_target_wait_results: status->kind = NO_RESUMED
54812 [infrun] handle_inferior_event: status->kind = NO_RESUMED
54813 [remote] Sending packet: $Hgp0.0#ad
54814 [remote] Packet received: OK
54815 [remote] Sending packet: $qXfer:threads:read::0,1000#92
54816 [remote] Packet received: l<threads>\n<thread id="pae4cb.ae4cb" core="3" name="multi-re-run-1" handle="40c7c6f7ff7f0000"/>\n<thread id="pae4cb.ae4cc" core="2" name="multi-re-run-1" handle="40b6c6f7ff7f0000"/>\n<thread id="pae4cd.ae4ce" core="1" name="multi-re-run-2" handle="40b6c6f7ff7f0000"/>\n</threads>\n
54817 [infrun] stop_waiting: stop_waiting
54818 [remote] Sending packet: $qXfer:threads:read::0,1000#92
54819 [remote] Packet received: l<threads>\n<thread id="pae4cb.ae4cb" core="3" name="multi-re-run-1" handle="40c7c6f7ff7f0000"/>\n<thread id="pae4cb.ae4cc" core="2" name="multi-re-run-1" handle="40b6c6f7ff7f0000"/>\n<thread id="pae4cd.ae4ce" core="1" name="multi-re-run-2" handle="40b6c6f7ff7f0000"/>\n</threads>\n
54820 [infrun] infrun_async: enable=0
54821 [infrun] reset: reason=handling event
54822 [infrun] maybe_set_commit_resumed_all_targets: enabling commit-resumed for target extended-remote
54823 [infrun] maybe_call_commit_resumed_all_targets: calling commit_resumed for target extended-remote
54824 [infrun] maybe_call_commit_resumed_all_targets: calling commit_resumed for target extended-remote
54825 [infrun] fetch_inferior_event: exit
54827 We can see that we resume the inferior with vCont;c, but got NO_RESUMED.
54828 When the test passes, we get an EXITED status to indicate the process
54831 From GDBserver's point of view, it looks like this. The logs contain
54832 some logging I added and that are part of this patch.
54834 [remote] getpkt: getpkt ("vCont;c:pae4cf.-1"); [no ack sent]
54835 [threads] resume: enter
54836 [threads] thread_needs_step_over: Need step over [LWP 713931]? Ignoring, should remain stopped
54837 [threads] thread_needs_step_over: Need step over [LWP 713932]? Ignoring, should remain stopped
54838 [threads] get_pc: pc is 0x555555555195
54839 [threads] thread_needs_step_over: Need step over [LWP 713935]? No, no breakpoint found at 0x555555555195
54840 [threads] get_pc: pc is 0x7ffff7d35a95
54841 [threads] thread_needs_step_over: Need step over [LWP 713936]? No, no breakpoint found at 0x7ffff7d35a95
54842 [threads] resume: Resuming, no pending status or step over needed
54843 [threads] resume_one_thread: resuming LWP 713935
54844 [threads] proceed_one_lwp: lwp 713935
54845 [threads] resume_one_lwp_throw: continue from pc 0x555555555195
54846 [threads] resume_one_lwp_throw: Resuming lwp 713935 (continue, signal 0, stop not expected)
54847 [threads] resume_one_lwp_throw: NOW ptid=713935.713935.0 stopped=0 resumed=0
54848 [threads] resume_one_thread: resuming LWP 713936
54849 [threads] proceed_one_lwp: lwp 713936
54850 [threads] resume_one_lwp_throw: continue from pc 0x7ffff7d35a95
54851 [threads] resume_one_lwp_throw: Resuming lwp 713936 (continue, signal 0, stop not expected)
54852 [threads] resume_one_lwp_throw: ptrace errno = 3 (No such process)
54853 [threads] resume: exit
54854 [threads] wait_1: enter
54855 [threads] wait_1: [<all threads>]
54856 [threads] wait_for_event_filtered: waitpid(-1, ...) returned 0, ERRNO-OK
54857 [threads] resume_stopped_resumed_lwps: resuming stopped-resumed LWP LWP 713935.713936 at 7ffff7d35a95: step=0
54858 [threads] resume_one_lwp_throw: continue from pc 0x7ffff7d35a95
54859 [threads] resume_one_lwp_throw: Resuming lwp 713936 (continue, signal 0, stop not expected)
54860 [threads] resume_one_lwp_throw: ptrace errno = 3 (No such process)
54861 [threads] operator(): check_zombie_leaders: leader_pid=713931, leader_lp!=NULL=1, num_lwps=2, zombie=0
54862 [threads] operator(): check_zombie_leaders: leader_pid=713935, leader_lp!=NULL=1, num_lwps=2, zombie=1
54863 [threads] operator(): Thread group leader 713935 zombie (it exited, or another thread execd).
54864 [threads] delete_lwp: deleting 713935
54865 [threads] wait_for_event_filtered: exit (no unwaited-for LWP)
54867 [threads] wait_1: ret = null_ptid, TARGET_WAITKIND_NO_RESUMED
54868 [threads] wait_1: exit
54872 - We resume the leader (713935) successfully.
54873 - The leader exits.
54874 - We resume the secondary thread (713936), we get ESRCH. This is
54875 expected this the leader has exited.
54876 - resume_one_lwp_throw throws, it's caught by resume_one_lwp.
54877 - resume_one_lwp checks with check_ptrace_stopped_lwp_gone that the
54878 failure can be explained by the LWP becoming zombie, and swallows the
54880 - Note that this means that the secondary lwp still has stopped==1.
54881 - wait_1 is called, probably because linux_process_target::resume marks
54882 the async pipe at the end.
54883 - The exit event isn't ready yet, probably because the machine is under
54884 load, so waitpid returns nothing.
54885 - check_zombie_leaders detects that the leader is zombie and deletes
54886 - We try to find a resumed (non-stopped) LWP to get an event from,
54887 there's none since the leader (that was resumed) is now deleted, and
54888 the secondary thread is still marked stopped.
54889 wait_for_event_filtered returns -1, causing wait_1 to return
54892 What I notice here is that there is some kind of race between the
54893 availability of the process' exit notification and the call to wait_1
54894 that results from marking the async pipe at the end of resume.
54896 I think what we want from this wait_1 invocation is to keep waiting, as
54897 we will eventually get thread exit notifications for both of our
54900 The fix I came up with is to mark the secondary thread as !stopped (or
54901 resumed) when we fail to resume it. This makes wait_1 see that there is
54902 at least one resume lwp, so it won't return NO_RESUMED. I think this
54903 makes sense to consider it resumed, because we are going to receive an
54904 exit event for it. Here's the GDBserver logs with the fix applied:
54906 [threads] resume: enter
54907 [threads] thread_needs_step_over: Need step over [LWP 724595]? Ignoring, should remain stopped
54908 [threads] thread_needs_step_over: Need step over [LWP 724596]? Ignoring, should remain stopped
54909 [threads] get_pc: pc is 0x555555555195
54910 [threads] thread_needs_step_over: Need step over [LWP 724597]? No, no breakpoint found at 0x555555555195
54911 [threads] get_pc: pc is 0x7ffff7d35a95
54912 [threads] thread_needs_step_over: Need step over [LWP 724598]? No, no breakpoint found at 0x7ffff7d35a95
54913 [threads] resume: Resuming, no pending status or step over needed
54914 [threads] resume_one_thread: resuming LWP 724597
54915 [threads] proceed_one_lwp: lwp 724597
54916 [threads] resume_one_lwp_throw: continue from pc 0x555555555195
54917 [threads] resume_one_lwp_throw: Resuming lwp 724597 (continue, signal 0, stop not expected)
54918 [threads] resume_one_lwp_throw: NOW ptid=724597.724597.0 stopped=0 resumed=0
54919 [threads] resume_one_thread: resuming LWP 724598
54920 [threads] proceed_one_lwp: lwp 724598
54921 [threads] resume_one_lwp_throw: continue from pc 0x7ffff7d35a95
54922 [threads] resume_one_lwp_throw: Resuming lwp 724598 (continue, signal 0, stop not expected)
54923 [threads] resume_one_lwp_throw: ptrace errno = 3 (No such process)
54924 [threads] resume: exit
54925 [threads] wait_1: enter
54926 [threads] wait_1: [<all threads>]
54928 [threads] wait_for_event_filtered: waitpid(-1, ...) returned 0, ERRNO-OK
54929 [threads] operator(): check_zombie_leaders: leader_pid=724595, leader_lp!=NULL=1, num_lwps=2, zombie=0
54930 [threads] operator(): check_zombie_leaders: leader_pid=724597, leader_lp!=NULL=1, num_lwps=2, zombie=1
54931 [threads] operator(): Thread group leader 724597 zombie (it exited, or another thread execd).
54932 [threads] delete_lwp: deleting 724597
54933 [threads] wait_for_event_filtered: sigsuspend'ing
54935 [threads] wait_for_event_filtered: waitpid(-1, ...) returned 724598, ERRNO-OK
54936 [threads] wait_for_event_filtered: waitpid 724598 received 0 (exited)
54937 [threads] filter_event: 724598 exited
54938 [threads] wait_for_event_filtered: waitpid(-1, ...) returned 724597, ERRNO-OK
54939 [threads] wait_for_event_filtered: waitpid 724597 received 0 (exited)
54940 [threads] wait_for_event_filtered: waitpid(-1, ...) returned 0, ERRNO-OK
54942 [threads] wait_1: ret = LWP 724597.724598, exited with retcode 0
54943 [threads] wait_1: exit
54945 Change-Id: Idf0bdb4cb0313f1b49e4864071650cc83fb3c100
54947 2022-03-31 Simon Marchi <simon.marchi@efficios.com>
54949 gdb/testsuite/tui: implement _csi_P proc
54950 Since commit 3cd522938792 ("Change the pager to a ui_file"), I see these
54951 errors when running gdb.tui/scroll.exp:
54953 ERROR: invalid command name "_csi_P"
54955 "::gdb_tcl_unknown _csi_P 2"
54956 ("uplevel" body line 1)
54957 invoked from within
54958 "uplevel 1 ::gdb_tcl_unknown $args"
54959 (procedure "::unknown" line 5)
54960 invoked from within
54962 ("eval" body line 1)
54963 invoked from within
54964 "eval _csi_$cmd $params"
54966 It looks like GDB is emitting a CSI that it did not emit before, the
54967 "Delete character" one:
54969 https://vt100.net/docs/vt510-rm/DCH.html
54973 Co-Authored-By: Andrew Burgess <aburgess@redhat.com>
54974 Change-Id: I5bf86b6104d51b0623a26a69df83d1ca9a4851b7
54976 2022-03-31 Simon Marchi <simon.marchi@polymtl.ca>
54978 gdb: fix use of fprintf_filtered in top.c
54979 A race condition in how patches were pushed causes this build failure:
54982 /home/simark/src/binutils-gdb/gdb/top.c: In function ‘void print_gdb_configuration(ui_file*)’:
54983 /home/simark/src/binutils-gdb/gdb/top.c:1622:3: error: ‘fprintf_filtered’ was not declared in this scope; did you mean ‘printf_unfiltered’?
54984 1622 | fprintf_filtered (stream, _("\
54987 fprintf_filtered has been removed, gdb_printf must be used now. Fix
54990 Change-Id: I6a172ba0d53dab2e7cc43ed0ed2696c82925245b
54992 2022-03-31 Richard Sandiford <richard.sandiford@arm.com>
54994 aarch64: Relax check for RNG system registers
54995 FEAT_RNG is an optional Armv8.5-A extension, but it can be backported
54996 to earlier architectures as well. GAS previously made the RNG registers
54997 conditional on having both armv8.5-a and +rng, but only +rng should be
55000 This seems to be the only feature that was handled like this.
55003 * aarch64-opc.c (SR_RNG): Don't require V8_5.
55006 * testsuite/gas/aarch64/rng-1.s, testsuite/gas/aarch64/rng-1.d: New
55009 2022-03-31 Eli Zaretskii <eliz@gnu.org>
55011 * gdb/top.c (print_gdb_configuration): Announce --enable-threading.
55012 This includes the reporting of --enable/disable-threading as part of
55013 the GDB configuration description.
55015 2022-03-31 Simon Marchi <simon.marchi@efficios.com>
55017 gdb/infrun: add reason parameter to stop_all_threads
55018 Add a "reason" parameter, only used to show in debug messages what is
55019 the reason for stopping all threads. This helped me understand the
55020 debug logs while adding some new uses of stop_all_threads, so I am
55021 proposing to merge it.
55023 Change-Id: I66c8c335ebf41836a7bc3d5fe1db92c195f65e55
55025 2022-03-31 Simon Marchi <simon.marchi@efficios.com>
55027 gdb/testsuite: update copyright years in gdb.base/vfork-follow-parent.*
55028 I forgot to do this before pushing the previous commit.
55030 Change-Id: Ia343f702e8357d0fd109e9ddd778973e91862805
55032 2022-03-31 Simon Marchi <simon.marchi@efficios.com>
55034 gdb: test vfork + follow-fork-mode=parent + detach-on-fork=off
55035 The particular behavior we have when using that combination of settings
55036 doesn't seem tested at all (at least, I don't find it if I grep for "Can
55037 not resume the parent process"). Add a simple test for that.
55039 Change-Id: Ib9454a615abba661b42f1b15056df73ed1bcd4c5
55041 2022-03-31 Nick Clifton <nickc@redhat.com>
55043 Accept the + character as part of filenames for MRI scripts.
55045 2022-03-31 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
55047 Fix procfs.c compilation
55048 procfs.c doesn't compile on Solaris:
55050 /vol/src/gnu/gdb/hg/master/local/gdb/procfs.c: In member function ‘virtual bool procfs_target::info_proc(const char*, info_proc_what)’:
55051 /vol/src/gnu/gdb/hg/master/local/gdb/procfs.c:3302:3: error: ‘gdb_argv’ was not declared in this scope
55052 3302 | gdb_argv built_argv (args);
55054 /vol/src/gnu/gdb/hg/master/local/gdb/procfs.c:3303:20: error: ‘built_argv’ was not declared in this scope; did you mean ‘buildargv’?
55055 3303 | for (char *arg : built_argv)
55059 Fixed by including "gdbsupport/buildargv.h".
55061 Tested on amd64-pc-solaris2.11, sparcv9-sun-solaris2.11.
55063 2022-03-31 GDB Administrator <gdbadmin@sourceware.org>
55065 Automatic date update in version.in
55067 2022-03-30 Simon Marchi <simon.marchi@polymtl.ca>
55069 gdb/testsuite: add tests for Term
55070 While trying to review Andrew's patch here [1], I thought I spotted a
55071 bug in the handling of a CSI, but I had no way to know for sure. So I
55072 thought it would be useful to have unit tests for the handling of
55073 control characters and control sequences of our toy terminal
55074 implementation. It might help avoid chasing bugs in the GDB TUI when in
55075 reality it's a problem with the testsuite's terminal implementation.
55077 Add the gdb.tui/tuiterm.exp file to do that. All currently supported
55078 control sequences and characters are tested, except _csi_m (the one that
55079 handles colors and stuff). _csi_m should probably be tested too, but it
55080 will require more work.
55082 Fix a few issues that the tests spotted:
55084 - backspace: according to [3] (table 4-1), a backspace when the cursor
55085 is at the beginning of a line should have no effect. Our
55086 implementation did wrap to the end of the previous line. Change our
55087 implementation to match the doc (and the test).
55088 - insert character: this control sequence is supposed to insert blank
55089 characters, shifting all the rest of the line right. The current
55090 implementation moves N characters right, but it overwrites the
55091 characters on the right instead of shifting them. It also doesn't
55092 insert blank characters at the cursor.
55093 - Cursor down, forward, next line: off-by-one error when reaching the
55094 end of the display.
55095 - erase in display, line: off-by-one errors.
55096 - vertical line position absolute: allowed setting the cursor outside
55097 the display, when it should clamp it to the display size.
55099 I found that this web page [2] gave some good clues on the expected
55100 behavior of some control characters or sequences that some other pages
55103 [1] https://sourceware.org/pipermail/gdb-patches/2022-March/186433.html
55104 [2] https://docs.microsoft.com/en-us/windows/console/console-virtual-terminal-sequences
55105 [3] https://vt100.net/docs/vt510-rm/chapter4.html#S4.3.3
55107 Change-Id: Iab4141fdcfb7459d1b7c45cc63bd1fcb50a78d5d
55109 2022-03-30 Tom Tromey <tom@tromey.com>
55111 Only allow QUIT on the main thread
55112 Pedro pointed out that gdb worker threads should not react to quits.
55113 While I don't think that the new DWARF reader can call QUIT from a
55114 worker thread (and I don't think the existing minsym threading code
55115 can either), it seems safest to address this before checking in the
55116 new code. This patch arranges for the QUIT macro to only work on the
55119 2022-03-30 Tom Tromey <tromey@adacore.com>
55121 Use gdb_printf and gdb_vprintf in more places
55122 Luis pointed out that I missed a spot in the gdb_printf conversion --
55123 namely aarch64-nat.c. While looking at this, I found another spot in
55124 darwin-nat.c that I also missed. I can't build either of these, but I
55125 think this patch should fix the problems.
55127 Consolidate definition of current_directory
55128 I noticed that both gdbserver and gdb define current_directory.
55129 However, as it is referenced by gdbsupport, it seemed better to define
55130 it there as well. This patch also moves the declaration to
55131 pathstuff.h. Tested by rebuilding.
55133 2022-03-30 Tom Tromey <tromey@adacore.com>
55135 Decode "dynamic" interface types in Ada
55136 In Ada, if a class implements an interface and has a dynamic
55137 superclass, then the "offset to top" -- the offset that says how to
55138 turn a pointer to the interface into a pointer to the whole object --
55139 is stored in the object itself. This patch changes GDB to understand
55142 Because this only touches Ada code, and because Joel already reviewed
55143 it internally, I am checking it in.
55145 2022-03-30 Jeff Law <jeffreyalaw@gmail.com>
55147 Fix for MUL instruction on the v850
55148 * sim/v850/simops.c (Multiply64): Properly test if we need to
55149 negate either of the operands.
55151 * sim/testsuite/v850/mul.cgs: New test.
55153 2022-03-30 GDB Administrator <gdbadmin@sourceware.org>
55155 Automatic date update in version.in
55157 2022-03-29 Tom Tromey <tom@tromey.com>
55159 Remove two unused hooks
55160 I noticed that a couple of deprecated hooks aren't ever called, so
55161 they can't really be used by Insight. This patch removes them
55162 entirely. I checked the Insight sources, and these aren't mentioned
55165 2022-03-29 Tom Tromey <tom@tromey.com>
55167 Remove unnecessary calls to wrap_here and gdb_flush
55168 Various spots in gdb currently know about the wrap buffer, and so are
55169 careful to call wrap_here to be certain that all output has been
55172 Now that the pager is just an ordinary stream, this isn't needed, and
55173 a simple call to gdb_flush is enough.
55175 Similarly, there are places where gdb prints to gdb_stderr, but first
55176 flushes gdb_stdout. stderr_file already flushes gdb_stdout, so these
55179 2022-03-29 Tom Tromey <tom@tromey.com>
55181 Minor comment updates in utils.h
55182 This patch updates some comments in utils.h to more closely reflect
55185 Remove vfprintf_styled
55186 Nothing calls vfprintf_styled any more, so remove it.
55188 Remove ui_out_flag::unfiltered_output
55189 There is no longer any need for ui_out_flag::unfiltered_output --
55190 nothing ever sets this flag. This used to be needed to make the
55191 _unfiltered output work, but now only printf_unfiltered can be used,
55192 and it uses the puts_unfiltered method. This patch removes the flag
55195 Rename fprintf_symbol_filtered
55196 fprintf_symbol_filtered is misnamed, because whether filtering happens
55197 is now up to the stream. This renames it to fprintf_symbol, which
55198 isn't a great name (the first "f" doesn't mean much and the second one
55199 is truly meaningless here), but "print_symbol" was already taken.
55201 Rename puts_filtered_tabular
55202 puts_filtered_tabular is now misnamed, because whether filtering
55203 happens is now up to the stream. So, rename it. (This function is
55204 pretty weird, and should probably be rewritten to avoid using the
55205 chars_printed global, and moved into objc-lang.c. However, I haven't
55208 Rename print_spaces_filtered
55209 print_spaces_filtered is now misnamed, because whether filtering
55210 happens is up to the stream. So, rename it.
55212 Unify gdb printf functions
55213 Now that filtered and unfiltered output can be treated identically, we
55214 can unify the printf family of functions. This is done under the name
55215 "gdb_printf". Most of this patch was written by script.
55217 Unify gdb putc functions
55218 Now that filtered and unfiltered output can be treated identically, we
55219 can unify the putc family of functions. This is done under the name
55220 "gdb_putc". Most of this patch was written by script.
55222 Unify gdb puts functions
55223 Now that filtered and unfiltered output can be treated identically, we
55224 can unify the puts family of functions. This is done under the name
55225 "gdb_puts". Most of this patch was written by script.
55227 Unify vprintf functions
55228 Now that filtered and unfiltered output can be treated identically, we
55229 can unify the vprintf family of functions: vprintf_filtered,
55230 vprintf_unfiltered, vfprintf_filtered and vfprintf_unfiltered. (For
55231 the gdb_stdout variants, recall that only printf_unfiltered gets truly
55232 unfiltered output at this point.) This removes one such function and
55233 renames the remaining two to "gdb_vprintf". All callers are updated.
55234 Much of this patch was written by script.
55236 Remove fputs_styled_unfiltered
55237 fputs_styled_unfiltered is only called from cli_ui_out, so remove it.
55238 This area will be further simplified in future patches.
55240 2022-03-29 Tom Tromey <tom@tromey.com>
55242 Change the pager to a ui_file
55243 This rewrites the output pager as a ui_file implementation.
55245 A new header is introduced to declare the pager class. The
55246 implementation remains in utils.c for the time being, because there
55247 are some static globals there that must be used by this code. (This
55248 could be cleaned up at some future date.)
55250 I went through all the text output in gdb to ensure that this change
55251 should be ok. There are a few cases:
55253 * Any existing call to printf_unfiltered is required to be avoid the
55254 pager. This is ensured directly in the implementation.
55256 * All remaining calls to the f*_unfiltered functions -- the ones that
55257 take an explicit ui_file -- either send to an unfiltered stream
55258 (e.g., gdb_stderr), which is obviously ok; or conditionally send to
55261 I investigated all such calls by searching for:
55263 grep -e '\bf[a-z0-9_]*_unfiltered' *.[chyl] */*.[ch] | grep -v gdb_stdlog | grep -v gdb_stderr
55265 This yields a number of candidates to check.
55267 * The breakpoint _print_recreate family, and
55268 save_trace_state_variables. These are used for "save" commands
55271 * Things printing to a temporary stream. Obviously ok.
55273 * Disassembly selftests.
55275 * print_gdb_help - this is non-obvious, but ok because paging isn't
55276 yet enabled at this point during startup.
55278 * serial.c - doens't use gdb_stdout
55280 * The code in compile/. This is all printing to a file.
55282 * DWARF DIE dumping - doesn't reference gdb_stdout.
55284 * Calls to the _filtered form -- these are all clearly ok, because if
55285 they are using gdb_stdout, then filtering will still apply; and if
55286 not, then filtering never applied and still will not.
55288 Therefore, at this point, there is no longer any distinction between
55289 all the other _filtered and _unfiltered calls, and they can be
55292 In this patch, take special note of the vfprintf_maybe_filtered and
55293 ui_file::vprintf change. This is one instance of the above idea,
55294 erasing the distinction between filtered and unfiltered -- in this
55295 part of the change, the "unfiltered_output" flag is never passe to
55296 cli_ui_out. Subsequent patches will go much further in this
55299 Also note the can_emit_style_escape changes in ui-file.c. Checking
55300 against gdb_stdout or gdb_stderr was always a bit of a hack; and now
55301 it is no longer needed, because this is decision can be more fully
55302 delegated to the particular ui_file implementation.
55304 ui_file::can_page is removed, because this patch removed the only call
55307 I think this is the main part of fixing PR cli/7234.
55309 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=7234
55311 2022-03-29 Tom Tromey <tom@tromey.com>
55313 Remove vfprintf_styled_no_gdbfmt
55314 This removes vfprintf_styled_no_gdbfmt, inlining it at the sole point
55317 Add style-escape methods to ui_file
55318 This adds emit_style_escape and reset_style methods to ui_file. These
55319 aren't used yet, but they will be once the pager is converted to be a
55322 Add puts_unfiltered method to ui_file
55323 When the pager is rewritten as a ui_file, gdb will still need a way to
55324 bypass the filtering. After examining a few approaches, I chose this
55325 patch, which adds a puts_unfiltered method to ui_file. For most
55326 implementations of ui_file, this will just delegate to puts. This
55327 patch also switches printf_unfiltered to use the new method.
55329 Only have one API for unfiltered output
55330 At the end of this series, the use of unfiltered output will be very
55331 restricted -- only places that definitely need it will use it. To
55332 this end, I thought it would be good to reduce the number of
55333 _unfiltered APIs that are exposed. This patch changes gdb so that
55334 only printf_unfiltered exists. (After this patch, the f* variants
55335 still exist as well, but those will be removed later.)
55337 2022-03-29 Tom Tromey <tom@tromey.com>
55339 Remove some uses of printf_unfiltered
55340 A number of spots call printf_unfiltered only because they are in code
55341 that should not be interrupted by the pager. However, I believe these
55342 cases are all handled by infrun's blanket ban on paging, and so can be
55343 converted to the default (_filtered) API.
55345 After this patch, I think all the remaining _unfiltered calls are ones
55346 that really ought to be. A few -- namely in complete_command -- could
55347 be replaced by a scoped assignment to pagination_enabled, but for the
55348 remainder, the code seems simple enough like this.
55350 2022-03-29 Tom Tromey <tom@tromey.com>
55352 Use unfiltered output in annotate.c
55353 It seems to me that annotations should not be filtered. While it
55354 might be weird for an annotation-based UI to use the pager, it's not,
55355 I think, out of the question. This patch makes this change.
55357 2022-03-29 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
55358 Aleksandar Paunovic <aleksandar.paunovic@intel.com>
55360 gdb/remote: use current_inferior in read_ptid if multi-process not supported
55361 When parsing the ptid out of a reply package, if the multi-process
55362 extensions are not supported, use current_inferior's pid as the pid of
55363 the reported thread, instead of inferior_ptid. This is needed because
55364 the inferior_ptid may be null_ptid although a legit context exists,
55365 due to a prior context switch via switch_to_inferior_no_thread.
55367 Below is a scenario that illustrates what could go wrong. First,
55368 setup a multi-target scenario. This is needed, because in a
55369 multi-target setting, the inferior_ptid is cleared out before waiting
55370 on targets. The second inferior below sits on top of a remote target.
55371 Multi-process packets are disabled.
55373 $ # First, spawn a process with PID 26253 to attach to later.
55375 Reading symbols from a.out...
55376 (gdb) maint set target-non-stop on
55377 (gdb) set remote multiprocess-feature-packet off
55380 (gdb) add-inferior -no-connection
55384 [Switching to inferior 2 [<null>] (<noexec>)]
55385 (gdb) target extended-remote | gdbserver --multi -
55386 Remote debugging using | gdbserver --multi -
55387 Remote debugging using stdio
55389 Attaching to Remote target
55390 Attached; pid = 26253
55393 Reading /tmp/a.out from remote target...
55397 Reading /usr/local/lib/debug/....debug from remote target...
55398 >>> GDB seems to hang here.
55400 After attaching to a process and reading some library files, GDB
55401 seems to hang. One interesting thing to note is that
55405 appears twice. We also see
55409 Running the same scenario with "debug infrun on" reveals more details.
55413 [infrun] scoped_disable_commit_resumed: reason=attaching
55414 Attaching to Remote target
55415 Attached; pid = 26253
55417 [infrun] infrun_async: enable=1
55418 [infrun] attach_command: immediately after attach:
55419 [infrun] attach_command: thread 26253.26253.0, executing = 1, resumed = 0, state = RUNNING
55420 [infrun] clear_proceed_status_thread: 26253.26253.0
55421 [infrun] reset: reason=attaching
55422 [infrun] maybe_set_commit_resumed_all_targets: not requesting commit-resumed for target native, no resumed threads
55423 [infrun] maybe_set_commit_resumed_all_targets: enabling commit-resumed for target extended-remote
55424 [infrun] fetch_inferior_event: enter
55425 [infrun] scoped_disable_commit_resumed: reason=handling event
55426 [infrun] do_target_wait: Found 2 inferiors, starting at #1
55427 [infrun] random_pending_event_thread: None found.
55428 [infrun] print_target_wait_results: target_wait (-1.0.0 [Thread 0], status) =
55429 [infrun] print_target_wait_results: 26253.26253.0 [Thread 26253],
55430 [infrun] print_target_wait_results: status->kind = STOPPED, sig = GDB_SIGNAL_0
55431 [infrun] handle_inferior_event: status->kind = STOPPED, sig = GDB_SIGNAL_0
55432 [infrun] start_step_over: enter
55433 [infrun] start_step_over: stealing global queue of threads to step, length = 0
55434 [infrun] operator(): step-over queue now empty
55435 [infrun] start_step_over: exit
55436 [infrun] context_switch: Switching context from 0.0.0 to 26253.26253.0
55437 [infrun] handle_signal_stop: stop_pc=0x7f849d8cf151
55438 [infrun] stop_waiting: stop_waiting
55439 [infrun] stop_all_threads: starting
55440 [infrun] stop_all_threads: pass=0, iterations=0
55442 Reading /tmp/a.out from remote target...
55443 warning: File transfers from remote targets can be slow. Use "set sysroot" to access files locally instead.
55444 Reading /tmp/a.out from remote target...
55445 Reading symbols from target:/tmp/a.out...
55447 [infrun] stop_all_threads: 4723.4723.0 not executing
55448 [infrun] stop_all_threads: 26253.26253.0 not executing
55449 [infrun] stop_all_threads: 42000.26253.0 executing, need stop
55450 [infrun] print_target_wait_results: target_wait (-1.0.0 [Thread 0], status) =
55451 [infrun] print_target_wait_results: -1.0.0 [Thread 0],
55452 [infrun] print_target_wait_results: status->kind = IGNORE
55453 [infrun] print_target_wait_results: target_wait (-1.0.0 [Thread 0], status) =
55454 [infrun] print_target_wait_results: -1.0.0 [Thread 0],
55455 [infrun] print_target_wait_results: status->kind = IGNORE
55457 GDB tried to stop Thread 42000.26253.0, which does not exist, and we
55458 are waiting for a stop event that will never happen. The PID in
55459 '42000.26253.0', namely 42000, is the PID of magic_null_ptid.
55460 It comes from gdb/remote.c:read_ptid:
55462 /* Since the stub is not sending a process id, then default to
55463 what's in inferior_ptid, unless it's null at this point. If so,
55464 then since there's no way to know the pid of the reported
55465 threads, use the magic number. */
55466 if (inferior_ptid == null_ptid)
55467 pid = magic_null_ptid.pid ();
55469 pid = inferior_ptid.pid ();
55473 return ptid_t (pid, tid);
55475 Because multi-process was turned off, GDB did not parse an explicitly
55476 specified PID. Furthermore, inferior_ptid == null_ptid, and
55477 eventually GDB picked the PID from magic_null_ptid.
55479 If target-non-stop is not turned on at the beginning, the same bug
55480 reveals itself as a duplicated thread as shown below.
55482 # Same setup as above, without 'maint set target-non-stop on'.
55485 Attaching to Remote target
55486 Attached; pid = 26253
55493 1.1 process 13517 "a.out" main () at test.c:3
55494 * 2.1 Thread 26253 "a.out" 0x00007f12750c5151 in read () from target:/lib/x86_64-linux-gnu/libc.so.6
55495 3.1 Thread 26253 "a.out" Remote 'g' packet reply is too long (expected 560 bytes, got 2496 bytes): 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
55498 Fix the problem by preferring current_inferior()'s pid instead of
55501 Regression-tested on X86-64 Linux.
55503 2022-03-29 Andrew Burgess <aburgess@redhat.com>
55505 gdb/testsuite: fix test failure when building against readline v7
55506 The test added in the commit:
55508 commit a6b413d24ccc5d76179bab866834e11fd6fec294
55509 Date: Fri Mar 11 14:44:03 2022 +0000
55511 gdb: work around prompt corruption caused by bracketed-paste-mode
55513 Was not written with readline 7 in mind, only readline 8+. Between
55514 readline 7 and 8 the escape sequence used to disable bracketed paste
55515 mode changed, an additional '\r' character was added to the end. In
55516 fact, it was the addition of this '\r' character that triggered the
55517 issue for which the above commit is part of the solution.
55519 Anyway, the test tries to spot the case where the output from GDB is
55520 not perfect, but does have the above work around applied. However,
55521 the pattern in the test assumes that the problematic '\r' will be
55522 present, and this is only true for readline 8+. With readline 7 the
55525 In this commit I generalise the pattern a little so that the test will
55526 still KFAIL with readline 7.
55528 It's a little unfortunate that the test is KFAILing with readline 7,
55529 as without the problematic '\r' there's actually no reason that GDB
55530 couldn't "do the right thing" in this case, in which case, the test
55531 would PASS, but that would require changes within GDB itself.
55533 My preference then is that initially we patch the test to get it
55534 KFAILing, then in a separate commit I can modify GDB so that it can
55535 PASS with readline 7.
55537 2022-03-29 Andrew Burgess <aburgess@redhat.com>
55539 gdb/testsuite: fix copy & paste error in gdb.python/py-format-address.exp
55540 The test gdb.python/py-format-address.exp, added in commit:
55542 commit 25209e2c6979c3838e14e099f0333609810db280
55543 Date: Sat Oct 23 09:59:25 2021 +0100
55545 gdb/python: add gdb.format_address function
55547 included 3 copy & paste errors where the wrong address was used in the
55548 expected output patterns.
55550 The test compiles two almost identical test binaries (one function
55551 changes its name, that's the only difference), two inferiors are
55552 created, each inferior using one of the test binaries.
55554 We then take the address of the name changing function in both
55555 inferiors ('foo' in inferior 1 and 'bar' in inferior 2) and the tests
55556 are carried out using these addresses.
55558 What we're checking for is that symbols 'foo' and 'bar' show up in the
55559 correct inferior, and that (as this test is for a Python API feature),
55560 the user can have one inferior selected, but ask about the other
55561 inferior, and see the correct symbol in the result.
55563 The hope is that the two binaries will be laid out identically by the
55564 compiler, and that 'foo' and 'bar' will be at the same address. This
55565 is fine, unless the executable is compiled as PIE (position
55566 independent executable), in which case there is a problem.
55568 The problem is that though inferior 1 is set running, the inferior 2
55569 never is. If the executables are compiled as PIE, then the address in
55570 the inferior 2 will not have been resolved, while the address in the
55571 inferior 1 will have been, and so the two addresses we use in the
55572 tests will be different.
55574 This issue was reported here:
55576 https://sourceware.org/pipermail/gdb-patches/2022-March/186911.html
55578 The first part of the fix is to use the correct address variable in
55579 the expected output patterns, with this change the tests pass even
55580 when the executables are compiled as PIE.
55582 A second part of this fix is to pass the 'nopie' option when we
55583 compile the tests, this should ensure that the address obtained in
55584 inferior 2 is the same as the address from inferior 1, which makes the
55587 2022-03-29 Andrew Burgess <aburgess@redhat.com>
55589 gdb/mi: fix use after free of frame_info causing spurious notifications
55592 commit a2757c4ed693cef4ecc4dcdcb2518353eb6b3c3f
55593 Date: Wed Mar 16 15:08:22 2022 +0000
55595 gdb/mi: consistently notify user when GDB/MI client uses -thread-select
55597 Changes were made to GDB to address some inconsistencies in when
55598 notifications are sent from a MI terminal to a CLI terminal (when
55599 multiple terminals are in use, see new-ui command).
55601 Unfortunately, in order to track when the currently selected frame has
55602 changed, that commit grabs a frame_info pointer before and after an MI
55603 command has executed, and compares the pointers to see if the frame
55608 If the frame cache is deleted for any reason then the frame_info
55609 pointer captured before the command started, is no longer valid, and
55610 any comparisons based on that pointer are undefined.
55612 This was leading to random test failures for some folk, see:
55614 https://sourceware.org/pipermail/gdb-patches/2022-March/186867.html
55616 This commit changes GDB so we no longer hold frame_info pointers, but
55617 instead store the frame_id and frame_level, this is safe even when the
55618 frame cache is flushed.
55620 2022-03-29 Jan Beulich <jbeulich@suse.com>
55622 bfd/Dwarf2: gas doesn't mangle names
55623 Include the language identifier emitted by gas in the set of ones where
55624 no mangled names are expected. Even if there could be "hand-mangled"
55625 names, gas doesn't emit DW_AT_linkage_name in the first place.
55627 bfd/Dwarf2: make find-nearest-line returned function name consistent
55628 Prior to entering the enclosing "else if()" the earlier associated if()
55629 checks function->is_linkage and, if set, uses function->name. The
55630 comment in patch context precedes (and explains) the setting
55631 function->is_linkage. Yet with the flag set, we should then also return
55632 the function name, just like said earlier if() would do when we came
55633 here a 2nd time for the same "addr". And indeed passing the same address
55634 twice on addr2line's command line would resolve the function for the 2nd
55635 instance, but not for the 1st (if this code path is taken). (This,
55636 obviously, is particularly relevant when there's no ELF symbol table in
55637 the first place, like would be the case - naturally - in PE/COFF
55638 binaries, for example.)
55640 gas/Dwarf: special-case .linefile only for macros
55641 Restrict the PR gas/16908 workaround to just macros, matching the
55642 original intention as well as the comment there. For constructs like
55643 .irp or .rept the reasoning doesn't apply, as there's no separate
55644 "invocation" point which may be of interest to record (for, as said
55645 there, short macros).
55647 2022-03-29 Jan Beulich <jbeulich@suse.com>
55649 RISC-V: correct FCVT.Q.L[U]
55650 While the spec isn't explicit about this, it pointing out the similarity
55651 with the D extension ought to extend to the ignoring of a meaningless
55652 rounding mode: "Note FCVT.D.W[U] always produces an exact result and is
55653 unaffected by rounding mode." Hence the chosen encodings also ought to
55656 Note that to avoid breaking existing code the forms with a 3rd operand
55657 are not removed, which means there continues to be a difference to
55660 2022-03-29 Mike Frysinger <vapier@gentoo.org>
55662 sim: add arch/.gdbinit stub scripts
55663 Make it easy to load the common gdbinit script even when running in
55664 the arch/ subdir instead of the top-level sim dir.
55666 2022-03-29 Alan Modra <amodra@gmail.com>
55668 asan: heap buffer overflow in pa_chk_field_selector
55669 The buffer overflow showed up running the gas "all macro" test.
55672 * config/tc-hppa.c (pa_chk_field_selector): Don't read past end
55675 2022-03-29 GDB Administrator <gdbadmin@sourceware.org>
55677 Automatic date update in version.in
55679 2022-03-28 Tom Tromey <tom@tromey.com>
55681 Add Rust parser check for end of expression
55682 I noticed that "print 5," passed in Rust -- the parser wasn't checking
55683 that the entire input was used. This patch fixes the problem. This
55684 in turn pointed out another bug in the parser, namely that it didn't
55685 lex the next token after handling a string token. This is also fixed
55688 2022-03-28 Tom Tromey <tom@tromey.com>
55690 Switch gdb_stdlog to use timestamped_file
55691 Currently, timestamps for logging are done by looking for the use of
55692 gdb_stdlog in vfprintf_unfiltered. This seems potentially buggy, in
55693 that during logging or other redirects (like execute_fn_to_ui_file) we
55694 might have gdb_stdout==gdb_stdlog and so, conceivably, wind up with
55695 timestamps in a log when they were not desired.
55697 It seems better, instead, for timestamps to be a property of the
55700 This patch changes gdb to use the new timestamped_file for gdb_stdlog
55701 where appropriate, and removes the special case from
55702 vfprintf_unfiltered.
55704 Note that this may somewhat change the output in some cases -- in
55705 particular, when going through execute_fn_to_ui_file (or the _string
55706 variant), timestamps won't be emitted. This could be fixed in those
55707 functions, but it wasn't clear to me whether this is really desirable.
55709 Note also that this changes the TUI to send gdb_stdlog to gdb_stderr.
55710 I imagine that the previous use of gdb_stdout here was inadvertent.
55711 (And in any case it probably doesn't matter.)
55713 2022-03-28 Tom Tromey <tom@tromey.com>
55715 Add new timestamped_file class
55716 This adds a "timestamped_file" subclass of ui_file. This class adds a
55717 timestamp to its output when appropriate. That is, it follows the
55718 rule already used in vfprintf_unfiltered of adding a timestamp at most
55721 The new class is not yet used.
55723 2022-03-28 Tom Tromey <tom@tromey.com>
55725 Use unique_ptr in CLI logging code
55726 This changes the CLI logging code to avoid manual memory management
55727 (to the extent possible) by using unique_ptr in a couple of spots.
55728 This will come in handy in a later patch.
55730 2022-03-28 Tom Tromey <tom@tromey.com>
55732 Simplify the CLI set_logging logic
55733 The CLI's set_logging logic seemed unnecessarily complicated to me.
55734 This patch simplifies it, with an eye toward changing it to use RAII
55735 objects in a subsequent patch.
55737 I did not touch the corresponding MI code. That code seems incorrect
55738 (nothing ever uses raw_stdlog, and nothing ever sets
55739 saved_raw_stdlog). I didn't attempt to fix this, because I question
55740 whether this is even useful for MI.
55742 2022-03-28 Tom Tromey <tromey@adacore.com>
55744 Handle multiple addresses in call_site_target
55745 A large customer program has a function that is partitioned into hot
55746 and cold parts. A variable in a callee of this function is described
55747 using DW_OP_GNU_entry_value, but gdb gets confused when trying to find
55748 the caller. I tracked this down to dwarf2_get_pc_bounds interpreting
55749 the function's changes so that the returned low PC is the "wrong"
55752 Intead, when processing DW_TAG_call_site, the low PC of each range in
55753 DW_AT_ranges should be preserved in the call_site_target. This fixes
55754 the variable lookup in the test case I have.
55756 I didn't write a standalone test for this as it seemed excessively
55759 2022-03-28 Tom Tromey <tromey@adacore.com>
55761 Change call_site_target to iterate over addresses
55762 In order to handle the case where a call site target might refer to
55763 multiple addresses, we change the code to use a callback style. Any
55764 spot using call_site_target::address now passes in a callback function
55765 that may be called multiple times.
55767 2022-03-28 Tom Tromey <tromey@adacore.com>
55769 Change call_site_find_chain_1 to work recursively
55770 call_site_find_chain_1 has a comment claiming that recursive calls
55771 would be too expensive. However, I doubt this is so expensive; and
55772 furthermore the explicit state management approach here is difficult
55773 both to understand and to modify. This patch changes this code to use
55774 explicit recursion, so that a subsequent patch can generalize this
55775 code without undue trauma.
55777 Additionally, I think this patch detects a latent bug in the recursion
55778 code. (It's hard for me to be completely certain.) The bug is that
55779 when a new target_call_site is entered, the code does:
55781 if (target_call_site)
55783 if (addr_hash.insert (target_call_site->pc ()).second)
55785 /* Successfully entered TARGET_CALL_SITE. */
55787 chain.push_back (target_call_site);
55792 Here, if entering the target_call_site fails, then any tail_call_next
55793 elements in this call site are not visited. However, if this code
55794 does happen to enter a call site, then the tail_call_next elements
55795 will be visited during backtracking. This applies when doing the
55796 backtracking as well -- it will only continue through a given chain as
55797 long as each element in the chain can successfully be visited.
55799 I'd appreciate some review of this. If this behavior is intentional,
55800 it can be added to the new implementation.
55802 2022-03-28 Tom Tromey <tromey@adacore.com>
55804 Constify chain_candidate
55805 While investigating this bug, I wasn't sure if chain_candidate might
55806 update 'chain'. I changed it to accept a const reference, making it
55807 clear that it cannot. This simplifies the code a tiny bit as well.
55809 Make call_site_target members private
55810 This makes the data members of call_site_target 'private'. This lets
55811 us remove most of its public API. call_site_to_target_addr is changed
55812 to be a method of this type. This is a preparatory refactoring for
55813 the fix at the end of this series.
55815 Change call_site_target to use custom type and enum
55816 call_site_target reuses field_loc_kind and field_location. However,
55817 it has never used the full range of the field_loc_kind enum. In a
55818 subsequent patch, I plan to add a new 'kind' here, so it seemed best
55819 to avoid this reuse and instead introduce new types here.
55821 2022-03-28 GDB Administrator <gdbadmin@sourceware.org>
55823 Automatic date update in version.in
55825 2022-03-27 GDB Administrator <gdbadmin@sourceware.org>
55827 Automatic date update in version.in
55829 2022-03-26 Tom Tromey <tom@tromey.com>
55831 Remove an unused declaration from value.h
55832 value.h has a declaration of value_print_array_elements that is
55833 incorrect. In C, this would have been an error, but in C++ this is a
55834 declaration of an overload that is neither defined nor used. This
55835 patch removes the declaration.
55837 2022-03-26 GDB Administrator <gdbadmin@sourceware.org>
55839 Automatic date update in version.in
55841 2022-03-25 Nick Alcock <nick.alcock@oracle.com>
55843 libtool.m4: fix the NM="/nm/over/here -B/option/with/path" case
55844 My previous nm patch handled all cases but one -- if the user set NM in
55845 the environment to a path which contained an option, libtool's nm
55846 detection tries to run nm against a copy of nm with the options in it:
55847 e.g. if NM was set to "nm --blargle", and nm was found in /usr/bin, the
55848 test would try to run "/usr/bin/nm --blargle /usr/bin/nm --blargle".
55849 This is unlikely to be desirable: in this case we should run
55850 "/usr/bin/nm --blargle /usr/bin/nm".
55852 Furthermore, as part of this nm has to detect when the passed-in $NM
55853 contains a path, and in that case avoid doing a path search itself.
55854 This too was thrown off if an option contained something that looked
55855 like a path, e.g. NM="nm -B../prev-gcc"; libtool then tries to run
55856 "nm -B../prev-gcc nm" which rarely works well (and indeed it looks
55857 to see whether that nm exists, finds it doesn't, and wrongly concludes
55858 that nm -p or whatever does not work).
55860 Fix all of these by clipping all options (defined as everything
55861 including and after the first " -") before deciding whether nm
55862 contains a path (but not using the clipped value for anything else),
55863 and then removing all options from the path-modified nm before
55864 looking to see whether that nm existed.
55866 NM=my-nm now does a path search and runs e.g.
55867 /usr/bin/my-nm -B /usr/bin/my-nm
55869 NM=/usr/bin/my-nm now avoids a path search and runs e.g.
55870 /usr/bin/my-nm -B /usr/bin/my-nm
55872 NM="my-nm -p../wombat" now does a path search and runs e.g.
55873 /usr/bin/my-nm -p../wombat -B /usr/bin/my-nm
55875 NM="../prev-binutils/new-nm -B../prev-gcc" now avoids a path search:
55876 ../prev-binutils/my-nm -B../prev-gcc -B ../prev-binutils/my-nm
55878 This seems to be all combinations, including those used by GCC bootstrap
55879 (which, before this commit, fails to bootstrap when configured
55880 --with-build-config=bootstrap-lto, because the lto plugin is now using
55881 --export-symbols-regex, which requires libtool to find a working nm,
55882 while also using -B../prev-gcc to point at the lto plugin associated
55883 with the GCC just built.)
55885 Regenerate all affected configure scripts.
55887 * libtool.m4 (LT_PATH_NM): Handle user-specified NM with
55888 options, including options containing paths.
55890 2022-03-25 Alan Modra <amodra@gmail.com>
55892 Re: gas/Dwarf: improve debug info generation from .irp and alike blocks
55893 am33_2.0-linux is a mn10300 target.
55895 * testsuite/gas/elf/dwarf-5-irp.d: xfail am33.
55897 2022-03-25 GDB Administrator <gdbadmin@sourceware.org>
55899 Automatic date update in version.in
55901 2022-03-24 Aaron Merey <amerey@redhat.com>
55903 Remove download size from debuginfod progress messages if unavailable
55904 Currently debuginfod progress update messages include the size of
55907 Downloading 7.5 MB separate debug info for /lib/libxyz.so.0
55909 This value originates from the Content-Length HTTP header of the
55910 transfer. However this header is not guaranteed to be present for
55911 each download. This can happen when debuginfod servers compress files
55912 on-the-fly at the time of transfer. In this case gdb wrongly prints
55913 "-0.00 MB" as the size.
55915 This patch removes download sizes from progress messages when they are
55916 not available. It also removes usage of the progress bar until
55917 a more thorough reworking of progress updating is implemented. [1]
55919 [1] https://sourceware.org/pipermail/gdb-patches/2022-February/185798.html
55921 2022-03-24 Reuben Thomas <rrt@sc3d.org>
55923 sim: fix a comment typo in sim-load.c
55924 Fix a typo where the documentation refers to a function parameter by the
55927 Change-Id: I99494efe62cd4aa76fb78a0bd5da438d35740ebe
55929 2022-03-24 Reuben Thomas <rrt@sc3d.org>
55931 sim: fix “alligned” typos
55932 Change-Id: Ifd574e38524dd4f1cf0fc003e0c5c7499abc84a0
55934 2022-03-24 Jan Beulich <jbeulich@suse.com>
55936 x86: mention dropped L1OM/K1OM support in ld/ as well
55937 This amends e961c696dcb2 ("x86: drop L1OM/K1OM support from ld"). Also
55938 remove the marker that I mistakenly added in c085ab00c7b2 ("x86: drop
55939 L1OM/K1OM support from gas").
55941 2022-03-24 Simon Marchi <simon.marchi@efficios.com>
55943 gdb/testsuite: remove gdb.python/pretty-print-call-by-hand.exp
55944 This test was added without a corresponding fix, with some setup_kfails.
55945 However, it results in UNRESOLVED results when GDB is built with ASan.
55947 ERROR: GDB process no longer exists
55948 GDB process exited with wait status 1946871 exp7 0 1
55949 UNRESOLVED: gdb.python/pretty-print-call-by-hand.exp: frame print: backtrace test (PRMS gdb/28856)
55951 Remove the test from the tree, I'll attach it to the Bugzilla bug
55954 [1] https://sourceware.org/bugzilla/show_bug.cgi?id=28856
55956 Change-Id: Id95d8949fb8742874bd12aeac758aa4d7564d678
55958 2022-03-24 Jan Beulich <jbeulich@suse.com>
55960 x86: drop L1OM/K1OM support from ld
55961 This was only rudimentary support anyway; none of the sub-architecture
55962 specific insns were ever supported.
55964 x86: drop L1OM special case from disassembler
55965 There wasn't any real support anyway: None of the sub-architecture
55966 specific insns were ever supported.
55968 2022-03-24 Jan Beulich <jbeulich@suse.com>
55970 MAINTAINERS: add myself
55971 I much appreciate Nick offering this role to me. Nevertheless there's
55972 still a lot for me to learn here.
55974 At this occasion also update my email address in the pre-existing, much
55977 2022-03-24 GDB Administrator <gdbadmin@sourceware.org>
55979 Automatic date update in version.in
55981 2022-03-23 Andrew Burgess <aburgess@redhat.com>
55983 gdb/testsuite: address test failures in gdb.mi/mi-multi-commands.exp
55984 The gdb.mi/mi-multi-commands.exp test was added in commit:
55986 commit d08cbc5d3203118da5583296e49273cf82378042
55987 Date: Wed Dec 22 12:57:44 2021 +0000
55989 gdb: unbuffer all input streams when not using readline
55991 And then tweaked in commit:
55993 commit 144459531dd68a1287905079aaa131b777a8cc82
55994 Date: Mon Feb 7 20:35:58 2022 +0000
55996 gdb/testsuite: relax pattern in new gdb.mi/mi-multi-commands.exp test
55998 The second of these commits was intended to address periodic test
55999 failures that I was seeing, and this change did fix some problems,
56000 but, unfortunately, introduced other issues.
56002 The problem is that the test relies on sending two commands to GDB in
56003 a single write. As the characters that make these two commands arrive
56004 they are echoed to GDB's console. However, there is a race between
56005 how quickly the characters are echoed and how quickly GDB decides to
56006 act on the incoming commands.
56008 Usually, both commands are echoed in full before GDB acts on the first
56009 command, but sometimes this is not the case, and GDB can execute the
56010 first command before both commands are fully echoed to the console.
56011 In this case, the output of the first command will be mixed in with
56012 the echoing of the second command.
56014 This mixing of the command echoing and the first command output is
56015 what was causing failures in the original version of the test.
56017 The second commit relaxed the expected output pattern a little, but
56018 was still susceptible to failures, so this commit further relaxes the
56021 Now, we look for the first command output with no regard to what is
56022 before, or after the command. Then we look for the first mi prompt to
56023 indicate that the first command has completed.
56025 I believe that this change should make the test more stable than it
56028 2022-03-23 Nick Alcock <nick.alcock@oracle.com>
56030 libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option
56031 libctf has always handled endianness differences by detecting
56032 foreign-endian CTF dicts on the input and endian-flipping them: dicts
56033 are always written in native endianness. This makes endian-awareness
56034 very low overhead, but it means that the foreign-endian code paths
56035 almost never get routinely tested, since "make check" usually reads in
56036 dicts ld has just written out: only a few corrupted-CTF tests are
56037 actually in fixed endianness, and even they only test the foreign-
56038 endian code paths when you run make check on a big-endian machine.
56039 (And the fix is surely not to add more .s-based tests like that, because
56040 they are a nightmare to maintain compared to the C-code-based ones.)
56042 To improve on this, add a new environment variable,
56043 LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally
56044 endian-flip at ctf_write time, so the output is always in the wrong
56045 endianness. This then tests the foreign-endian read paths properly
56048 Make this easier by restructuring the writeout code in ctf-serialize.c,
56049 which duplicates the maybe-gzip-and-write-out code three times (once
56050 for ctf_write_mem, with thresholding, and once each for
56051 ctf_compress_write and ctf_write just so those can avoid thresholding
56052 and/or compression). Instead, have the latter two call the former
56053 with thresholds of 0 or (size_t) -1, respectively.
56055 The endian-flipping code itself gains a bit of complexity, because
56056 one single endian-flipper (flip_types) was assuming the input to be
56057 in foreign-endian form and assuming it could pull things out of the
56058 input once they had been flipped and make sense of them. At the
56059 cost of a few lines of duplicated initializations, teach it to
56060 read before flipping if we're flipping to foreign-endianness instead
56064 * ctf-impl.h (ctf_flip_header): No longer static.
56065 (ctf_flip): Likewise.
56066 * ctf-open.c (flip_header): Rename to...
56067 (ctf_flip_header): ... this, now it is not private to one file.
56068 (flip_ctf): Rename...
56069 (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg.
56070 (flip_types): Likewise. Use it.
56071 (ctf_bufopen_internal): Adjust calls.
56072 * ctf-serialize.c (ctf_write_mem): Add flip_endian path via
56073 a newly-allocated bounce buffer.
56074 (ctf_compress_write): Move below ctf_write_mem and reimplement
56076 (ctf_write): Likewise.
56077 (ctf_gzwrite): Note that this obscure writeout function does not
56078 support endian-flipping.
56080 2022-03-23 Nick Alcock <nick.alcock@oracle.com>
56082 libctf, ld: diagnose corrupted CTF header cth_strlen
56083 The last section in a CTF dict is the string table, at an offset
56084 represented by the cth_stroff header field. Its length is recorded in
56085 the next field, cth_strlen, and the two added together are taken as the
56086 size of the CTF dict. Upon opening a dict, we check that none of the
56087 header offsets exceed this size, and we check when uncompressing a
56088 compressed dict that the result of the uncompression is the same length:
56089 but CTF dicts need not be compressed, and short ones are not.
56090 Uncompressed dicts just use the ctf_size without checking it. This
56091 field is thankfully almost unused: it is mostly used when reserializing
56092 a dict, which can't be done to dicts read off disk since they're
56095 However, when opening an uncompressed foreign-endian dict we have to
56096 copy it out of the mmaped region it is stored in so we can endian-
56097 swap it, and we use ctf_size when doing that. When the cth_strlen is
56098 corrupt, this can overrun.
56100 Fix this by checking the ctf_size in all uncompressed cases, just as we
56101 already do in the compressed case. Add a new test.
56103 This came to light because various corrupted-CTF raw-asm tests had an
56104 incorrect cth_strlen: fix all of them so they produce the expected
56109 * ctf-open.c (ctf_bufopen_internal): Always check uncompressed
56110 CTF dict sizes against the section size in case the cth_strlen is
56115 * testsuite/ld-ctf/diag-strlen-invalid.*: New test,
56116 derived from diag-cttname-invalid.s.
56117 * testsuite/ld-ctf/diag-cttname-invalid.s: Fix incorrect cth_strlen.
56118 * testsuite/ld-ctf/diag-cttname-null.s: Likewise.
56119 * testsuite/ld-ctf/diag-cuname.s: Likewise.
56120 * testsuite/ld-ctf/diag-parlabel.s: Likewise.
56121 * testsuite/ld-ctf/diag-parname.s: Likewise.
56123 2022-03-23 Nick Alcock <nick.alcock@oracle.com>
56125 include, libctf, ld: extend variable section to contain functions too
56126 The CTF variable section is an optional (usually-not-present) section in
56127 the CTF dict which contains name -> type mappings corresponding to data
56128 symbols that are present in the linker input but not in the output
56129 symbol table: the idea is that programs that use their own symbol-
56130 resolution mechanisms can use this section to look up the types of
56131 symbols they have found using their own mechanism.
56133 Because these removed symbols (mostly static variables, functions, etc)
56134 all have names that are unlikely to appear in the ELF symtab and because
56135 very few programs have their own symbol-resolution mechanisms, a special
56136 linker flag (--ctf-variables) is needed to emit this section.
56138 Historically, we emitted only removed data symbols into the variable
56139 section. This seemed to make sense at the time, but in hindsight it
56140 really doesn't: functions are symbols too, and a C program can look them
56141 up just like any other type. So extend the variable section so that it
56142 contains all static function symbols too (if it is emitted at all), with
56143 types of kind CTF_K_FUNCTION.
56145 This is a little fiddly. We relied on compiler assistance for data
56146 symbols: the compiler simply emits all data symbols twice, once into the
56147 symtypetab as an indexed symbol and once into the variable section.
56149 Rather than wait for a suitably adjusted compiler that does the same for
56150 function symbols, we can pluck unreported function symbols out of the
56151 symtab and add them to the variable section ourselves. While we're at
56152 it, we do the same with data symbols: this is redundant right now
56153 because the compiler does it, but it costs very little time and lets the
56154 compiler drop this kludge and save a little space in .o files.
56157 * ctf.h: Mention the new things we can see in the variable
56161 * testsuite/ld-ctf/data-func-conflicted-vars.d: New test.
56164 * ctf-link.c (ctf_link_deduplicating_variables): Duplicate
56165 symbols into the variable section too.
56166 * ctf-serialize.c (symtypetab_delete_nonstatic_vars): Rename
56168 (symtypetab_delete_nonstatics): ... this. Check the funchash
56169 when pruning redundant variables.
56170 (ctf_symtypetab_sect_sizes): Adjust accordingly.
56171 * NEWS: Describe this change.
56173 2022-03-23 Nick Alcock <nick.alcock@oracle.com>
56175 ld, testsuite: improve CTF-availability test
56176 The test for -gctf support in the compiler is used to determine when to
56177 run the ld-ctf tests and most of those in libctf. Unfortunately,
56178 because it uses check_compiler_available and compile_one_cc, it will
56179 fail whenever the compiler emits anything on stderr, even if it
56180 actually does support CTF perfectly well.
56182 So, instead, ask the compiler to emit assembler output and grep it for
56183 references to ".ctf": this is highly unlikely to be present if the
56184 compiler does not support CTF. (This will need adjusting when CTF grows
56185 support for non-ELF platforms that don't dot-prepend their section
56186 names, but right now the linker doesn't link CTF on any such platforms
56189 With this in place we can do things like run all the libctf tests under
56190 leak sanitizers etc even if those spray warnings on simple CTF
56191 compilations, rather than being blocked from doing so just when we would
56195 * testsuite/lib/ld-lib.exp (check_ctf_available): detect CTF
56196 even if a CTF-capable compiler emits warnings.
56198 2022-03-23 Simon Marchi <simon.marchi@efficios.com>
56200 gdb/python: remove Python 2/3 compatibility macros
56201 New in this version:
56203 - Rebase on master, fix a few more issues that appeared.
56205 python-internal.h contains a number of macros that helped make the code
56206 work with both Python 2 and 3. Remove them and adjust the code to use
56207 the Python 3 functions.
56209 Change-Id: I99a3d80067fb2d65de4f69f6473ba6ffd16efb2d
56211 2022-03-23 Simon Marchi <simon.marchi@polymtl.ca>
56213 gdb/python: remove Python 2 support
56214 New in this version:
56216 - Add a PY_MAJOR_VERSION check in configure.ac / AC_TRY_LIBPYTHON. If
56217 the user passes --with-python=python2, this will cause a configure
56218 failure saying that GDB only supports Python 3.
56220 Support for Python 2 is a maintenance burden for any patches touching
56221 Python support. Among others, the differences between Python 2 and 3
56222 string and integer types are subtle. It requires a lot of effort and
56223 thinking to get something that behaves correctly on both. And that's if
56224 the author and reviewer of the patch even remember to test with Python
56227 See this thread for an example:
56229 https://sourceware.org/pipermail/gdb-patches/2021-December/184260.html
56231 So, remove Python 2 support. Update the documentation to state that GDB
56232 can be built against Python 3 (as opposed to Python 2 or 3).
56234 Update all the spots that use:
56241 ... to only keep the Python 3 portions and drop the use of some
56242 now-removed compatibility macros.
56244 I did not update the configure script more than just removing the
56245 explicit references to Python 2. We could maybe do more there, like
56246 check the Python version and reject it if that version is not
56247 supported. Otherwise (with this patch), things will only fail at
56248 compile time, so it won't really be clear to the user that they are
56249 trying to use an unsupported Python version. But I'm a bit lost in the
56250 configure code that checks for Python, so I kept that for later.
56252 Change-Id: I75b0f79c148afbe3c07ac664cfa9cade052c0c62
56254 2022-03-23 Jan Beulich <jbeulich@suse.com>
56256 x86: reject relocations involving registers
56257 To prevent fatal or even internal errors, add a simple check to
56258 i386_validate_fix(), rejecting relocations when their target symbol is
56259 an equate of a register (or resolved to reg_section for any other
56262 x86: improve resolution of register equates
56263 Allow transitive (or recursive) equates to work in addition to direct
56264 ones. The only requirements are that
56265 - the equate being straight of a register, i.e. no expressions involved
56266 (albeit I'm afraid something like "%eax + 0" will be viewed as %eax),
56267 - at the point of use there's no forward ref left which cannot be
56270 Revert "PR28977 tc-i386.c internal error in parse_register"
56271 This reverts commit 5fac3f02edacfca458f7eeaaaa33a87e26e0e332,
56272 which was superceeded / replaced by 4faaa10f3fab.
56274 2022-03-23 Jan Beulich <jbeulich@suse.com>
56276 x86: don't attempt to resolve equates and alike from i386_parse_name()
56279 Perhaps right from its introduction in 4d1bb7955a8b it was wrong for
56280 i386_parse_name() to call parse_register(). This being a hook from the
56281 expression parser, it shouldn't be resolving e.g. equated symbols.
56282 That's relevant only for all other callers of parse_register().
56284 To compensate, in Intel syntax mode check_register() needs calling;
56285 perhaps not doing so was an oversight right when the function was
56286 introduced. This is necessary in particular to force EVEX encoding when
56287 VRex registers are used (but of course also to reject bad uses of
56288 registers, i.e. fully matching what parse_register() needs it for).
56290 2022-03-23 Luis Machado <luis.machado@arm.com>
56292 Update the list of recognized m-profile TAG_CPU_ARCH_*
56293 Check 3 additional variants previously not recognized:
56295 - TAG_CPU_ARCH_V7E_M
56296 - TAG_CPU_ARCH_V8M_BASE
56297 - TAG_CPU_ARCH_V8M_MAIN
56299 2022-03-23 Jan Beulich <jbeulich@suse.com>
56301 gas/Dwarf5: re-use file 0 line string table entry when faking file 0
56302 No need to emit the same string a 2nd time for file 1 in this case.
56304 gas/Dwarf5: adjust .debug_line file 0 checking
56305 First of all when a table entry has a NULL filename, the two inner if()s
56306 are better done the other way around: The 2nd doesn't depend on what the
56307 first does. This then renders redundant half of the conditions of the
56308 other if() and clarifies that subsequently only entry 0 is dealt with
56309 (indicating that part of the comment was wrong). Finally for there to be
56310 a usable name in slot 1, files_in_use needs to be larger than 1 and slot
56311 1's (rather than slot 0's) name needs to be non-NULL.
56313 2022-03-23 Jan Beulich <jbeulich@suse.com>
56315 gas/Dwarf5: drop dead code
56316 Commit 3417bfca676f ("GAS: DWARF-5: Ensure that the 0'th entry in the
56317 directory table contains the current... ") added a "dwarf_level < 5"
56318 check to out_dir_and_file_list(). This rendered dead that branch of the
56319 construct, due to the enclosing if()'s "DWARF2_LINE_VERSION >= 5".
56320 Delete that code as well as the corresponding part of the comment.
56322 While there also drop a redundant "dirs != NULL": "dirs" will always be
56323 non-NULL when dirs_in_use is not zero.
56325 2022-03-23 Jan Beulich <jbeulich@suse.com>
56327 gas/Dwarf: improve debug info generation from .irp and alike blocks
56328 Tying the bumping of the logical line number to reading from the
56329 original source file looks wrong: Upon finishing of the processing of an
56330 sb the original values will be restored anyway. Yet without bumping the
56331 line counter uses of .line inside e.g. an .irp construct won't have the
56332 intended effect: Such uses may be necessary to ensure proper debug info
56333 is emitted in particular when switching sections inside the .irp body,
56334 as dwarf2_gen_line_info() would bail without doing anything when it
56335 finds the line number unchanged from what it saw last.
56337 2022-03-23 Jan Beulich <jbeulich@suse.com>
56339 ELF32: don't silently truncate relocation addends
56340 At least x86-64's x32 sub-mode and RISC-V's 32-bit mode calculate
56341 addends as 64-bit values, but store them in signed 32-bit fields when
56342 generating the file without encountering any earlier error. When the
56343 relocated field is a 64-bit one, the value resulting after processing
56344 the relocation record when linking (or the latest when loading) may
56345 thus be wrong due to the truncation.
56347 With the code change in place, one x32 testcase actually triggers the
56348 new diagnostic. That one case of too large a (negative) addend is being
56349 adjusted alongside the addition of a new testcase to actually trigger
56350 the new error. (Note that due to internal BFD behavior the relocation in
56351 .data doesn't get processed anymore after the errors in .text.)
56353 Note that in principle it is possible to express 64-bit relocations in
56354 ELF32, but this would require .rel relocations, i.e. with the addend
56355 stored in the 64-bit field being relocated. But I guess it would be a
56356 lot of effort for little gain to actually support this.
56358 2022-03-23 Jan Beulich <jbeulich@suse.com>
56360 gas: retain whitespace between strings
56361 Macro arguments may be separated by commas or just whitespace. Macro
56362 arguments may also be quoted (where one level of quotes is removed in
56363 the course of determining the values for the respective formal
56364 parameters). Furthermore this quote removal knows _two_ somewhat odd
56365 escaping mechanisms: One, apparently in existence forever, is that a
56366 pair of quotes counts as the escaping of a quote, with the pair being
56367 transformed to a single quote in the course of quote removal. The other
56368 (introduced by c06ae4f232e6) looks more usual on the surface in that it
56369 deals with \" sequences, but it _retains_ the escaping \. Hence only the
56370 former mechanism is suitable when the value to be used by the macro body
56371 is to contain a quote. Yet this results in ambiguity of what "a""b" is
56372 intended to mean; elsewhere (e.g. for .ascii) it represents two
56373 successive string literals. However, in any event is the above different
56374 from "a" "b": I don't think this can be viewed the same as "a""b" when
56375 processing macro arguments.
56377 Change the scrubber to retain such whitespace, by making the processing
56378 of strings more similar to that of symbols. And indeed this appears to
56379 make sense when taking into account that for quite a while gas has been
56380 supporting quoted symbol names.
56382 Taking a more general view, however, the change doesn't go quite far
56383 enough. There are further cases where significant whitespace is removed
56384 by the scrubber. The new testcase enumerates a few in its ".if 0"
56385 section. I'm afraid the only way that I see to deal with this would be
56386 to significantly simplify the scrubber, such that it wouldn't do much
56387 more than collapse sequences of unquoted whitespace into a single blank.
56388 To be honest problems in this area aren't really surprising when seeing
56389 that there's hardly any checking of .macro use throughout the testsuite
56390 (and in particular in the [relatively] generic tests under all/).
56392 2022-03-23 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
56394 Only .so files are used in libcollector. Remove the other files.
56395 * libcollector/Makefile.am (install-data-local): Remove the *.la
56397 * libcollector/Makefile.in: Regenerate.
56399 2022-03-23 Tiezhu Yang <yangtiezhu@loongson.cn>
56401 gdb: testsuite: use gdb_attach to fix jit-elf.exp
56402 If /proc/sys/kernel/yama/ptrace_scope is 1, when execute the following
56403 command without superuser:
56405 make check-gdb TESTS="gdb.base/jit-elf.exp"
56407 we can see the following messages in gdb/testsuite/gdb.log:
56409 (gdb) attach 1650108
56410 Attaching to program: /home/yangtiezhu/build/gdb/testsuite/outputs/gdb.base/jit-elf/jit-elf-main, process 1650108
56411 ptrace: Operation not permitted.
56412 (gdb) FAIL: gdb.base/jit-elf.exp: attach: one_jit_test-2: break here 1: attach
56414 use gdb_attach to fix the above issue, at the same time, the clean_reattach
56415 proc should return a value to indicate whether it worked, and the callers
56416 should return early as well on failure.
56418 2022-03-23 Tiezhu Yang <yangtiezhu@loongson.cn>
56420 gdb: testsuite: use gdb_attach to fix attach-pie-noexec.exp
56421 If /proc/sys/kernel/yama/ptrace_scope is 1, when execute the following
56422 command without superuser:
56424 make check-gdb TESTS="gdb.base/attach-pie-noexec.exp"
56426 we can see the following messages in gdb/testsuite/gdb.log:
56429 Attaching to process 6500
56430 ptrace: Operation not permitted.
56431 (gdb) PASS: gdb.base/attach-pie-noexec.exp: attach
56433 It is obviously wrong, the expected result should be UNSUPPORTED in such
56436 With this patch, we can see "Operation not permitted" in the log info,
56437 and then we can do the following processes to test:
56438 (1) set ptrace_scope as 0
56439 $ echo 0 | sudo tee /proc/sys/kernel/yama/ptrace_scope
56440 $ make check-gdb TESTS="gdb.base/attach-pie-noexec.exp"
56442 $ sudo make check-gdb TESTS="gdb.base/attach-pie-noexec.exp"
56444 2022-03-23 Tiezhu Yang <yangtiezhu@loongson.cn>
56446 gdb: testsuite: add new gdb_attach to check "attach" command
56447 This commit adds new gdb_attach to centralize the failure checking of
56448 "attach" command. Return 0 if attach failed, otherwise return 1.
56450 2022-03-23 Tiezhu Yang <yangtiezhu@loongson.cn>
56452 gdb: testsuite: remove attach test from can_spawn_for_attach
56453 As Pedro Alves said, caching procs should not issue pass/fail [1],
56454 this commit removes attach test from can_spawn_for_attach, at the
56455 same time, use "verbose -log" instead of "unsupported" to get a
56456 trace about why a test run doesn't support spawning for attach.
56458 [1] https://sourceware.org/pipermail/gdb-patches/2022-March/186311.html
56460 2022-03-23 GDB Administrator <gdbadmin@sourceware.org>
56462 Automatic date update in version.in
56464 2022-03-22 John Baldwin <jhb@FreeBSD.org>
56466 Add support for hardware breakpoints/watchpoints on FreeBSD/Aarch64.
56467 This shares aarch64-nat.c and nat/aarch64-hw-point.c with the Linux
56468 native target. Since FreeBSD writes all of the debug registers in one
56469 ptrace op, use an unordered_set<> to track the "dirty" state for
56470 threads rather than bitmasks of modified registers.
56472 fbsd-nat: Add a low_prepare_to_resume virtual method.
56473 This method can be overridden by architecture-specific targets to
56474 perform additional work before a thread is resumed.
56476 2022-03-22 John Baldwin <jhb@FreeBSD.org>
56478 fbsd-nat: Add a low_delete_thread virtual method.
56479 This method can be overridden by architecture-specific targets to
56480 perform additional work when a thread is deleted.
56482 Note that this method is only invoked on systems supporting LWP
56483 events, but the pending use case (aarch64 debug registers) is not
56484 supported on older kernels that do not support LWP events.
56486 2022-03-22 John Baldwin <jhb@FreeBSD.org>
56488 fbsd-nat: Add helper routine to fetch siginfo_t for a ptid.
56490 2022-03-22 John Baldwin <jhb@FreeBSD.org>
56492 aarch64: Add an aarch64_nat_target mixin class.
56493 This class includes platform-independent target methods for hardware
56494 breakpoints and watchpoints using routines from
56495 nat/aarch64-hw-point.c.
56497 stopped_data_address is not platform-independent since the FAR
56498 register holding the address for a breakpoint hit must be fetched in a
56499 platform-specific manner. However, aarch64_stopped_data_address is
56500 provided as a helper routine which performs platform-independent
56501 validation given the value of the FAR register.
56503 For tracking the per-process debug register mirror state, use an
56504 unordered_map indexed by pid as recently adopted in x86-nat.c rather
56505 than a manual linked-list.
56507 2022-03-22 John Baldwin <jhb@FreeBSD.org>
56509 nat: Split out platform-independent aarch64 debug register support.
56510 Move non-Linux-specific support for hardware break/watchpoints from
56511 nat/aarch64-linux-hw-point.c to nat/aarch64-hw-point.c. Changes
56512 beyond a simple split of the code are:
56514 - aarch64_linux_region_ok_for_watchpoint and
56515 aarch64_linux_any_set_debug_regs_state renamed to drop linux_ as
56516 they are not platform specific.
56518 - Platforms must implement the aarch64_notify_debug_reg_change
56519 function which is invoked from the platform-independent code when a
56520 debug register changes for a given debug register state. This does
56521 not use the indirection of a 'low' structure as is done for x86.
56523 - The handling for kernel_supports_any_contiguous_range is not
56524 pristine. For non-Linux it is simply defined to true. Some uses of
56525 this could perhaps be implemented as new 'low' routines for the
56526 various places that check it instead?
56528 - Pass down ptid into aarch64_handle_breakpoint and
56529 aarch64_handle_watchpoint rather than using current_lwp_ptid which
56530 is only defined on Linux. In addition, pass the ptid on to
56531 aarch64_notify_debug_reg_change instead of the unused state
56534 2022-03-22 John Baldwin <jhb@FreeBSD.org>
56536 x86-fbsd-nat: Copy debug register state on fork.
56537 Use the FreeBSD native target low_new_fork hook to copy the
56538 per-process debug state from the parent to the child on fork.
56540 fbsd-nat: Add a low_new_fork virtual method.
56541 This method can be overridden by architecture-specific targets to
56542 perform additional work when a new child process is forked.
56544 2022-03-22 John Baldwin <jhb@FreeBSD.org>
56546 Add an x86_fbsd_nat_target mixin class for FreeBSD x86 native targets.
56547 This class implements debug register support common between the i386
56548 and amd64 native targets.
56550 While here, remove #ifdef's for HAVE_PT_GETDBREGS in FreeBSD-specific
56551 code. The ptrace request has been present on FreeBSD x86
56552 architectures since 4.0 (released in March 2000). The last FreeBSD
56553 release without this support is 3.5 released in June 2000.
56555 2022-03-22 John Baldwin <jhb@FreeBSD.org>
56557 x86-nat: Add x86_lookup_debug_reg_state.
56558 This function returns nullptr if debug register state does not yet
56559 exist for a given process rather than creating new state.
56561 x86-nat: Use an unordered_map to store per-pid debug reg state.
56562 This replaces a manual linked list which used O(n) lookup and removal.
56564 Remove USE_SIGTRAP_SIGINFO condition for FreeBSD/x86 debug regs support.
56565 For BSD x86 targets, stopped_by_hw_breakpoint doesn't check siginfo_t
56566 but inspects the DR6 register directly via PT_GETDBREGS.
56568 2022-03-22 Tom Tromey <tom@tromey.com>
56570 Remove two unused variables
56571 I found a couple of spots that declare a symtab_and_line but don't
56572 actually use it. I think this probably isn't detected as unused
56573 because it has a constructor.
56575 2022-03-22 Steiner H Gunderson <steinar+sourceware@gunderson.no>
56577 Fix return code in _bfd_dwarf2_find_nearest_line().
56578 * dwarf2.c (_bfd_dwarf2_find_nearest_line): if a function name is
56579 found, but no line number info, then return a result of 2.
56581 2022-03-22 Andrew Burgess <andrew.burgess@embecosm.com>
56583 gdb/python: add gdb.format_address function
56584 Add a new function, gdb.format_address, which is a wrapper around
56585 GDB's print_address function.
56587 This method takes an address, and returns a string with the format:
56589 ADDRESS <SYMBOL+OFFSET>
56591 Where, ADDRESS is the original address, formatted as hexadecimal,
56592 SYMBOL is a symbol with an address lower than ADDRESS, and OFFSET is
56593 the offset from SYMBOL to ADDRESS in decimal.
56595 If there's no SYMBOL suitably close to ADDRESS then the
56596 <SYMBOL+OFFSET> part is not included.
56598 This is useful if a user wants to write a Python script that
56599 pretty-prints addresses, the user no longer needs to do manual symbol
56600 lookup, or worry about correctly formatting addresses.
56602 Additionally, there are some settings that effect how GDB picks
56603 SYMBOL, and whether the file name and line number should be included
56604 with the SYMBOL name, the gdb.format_address function ensures that the
56605 users Python script also benefits from these settings.
56607 The gdb.format_address by default selects SYMBOL from the current
56608 inferiors program space, and address is formatted using the
56609 architecture for the current inferior. However, a user can also
56610 explicitly pass a program space and architecture like this:
56612 gdb.format_address(ADDRESS, PROGRAM_SPACE, ARCHITECTURE)
56614 In order to format an address for a different inferior.
56616 Notes on the implementation:
56618 In py-arch.c I extended arch_object_to_gdbarch to add an assertion for
56619 the type of the PyObject being worked on. Prior to this commit all
56620 uses of arch_object_to_gdbarch were guaranteed to pass this function a
56621 gdb.Architecture object, but, with this commit, this might not be the
56624 So, with this commit I've made it a requirement that the PyObject be a
56625 gdb.Architecture, and this is checked with the assert. And in order
56626 that callers from other files can check if they have a
56627 gdb.Architecture object, I've added the new function
56628 gdbpy_is_architecture.
56630 In py-progspace.c I've added two new function, the first
56631 progspace_object_to_program_space, converts a PyObject of type
56632 gdb.Progspace to the associated program_space pointer, and
56633 gdbpy_is_progspace checks if a PyObject is a gdb.Progspace or not.
56635 2022-03-22 Luis Machado <luis.machado@arm.com>
56637 Fix some stale header names from dwarf files
56638 Some of these references were not updated when they were moved to a separate
56641 2022-03-22 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
56643 Install gprofng libraries under $(pkglibdir)
56645 2022-03-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
56648 * gprofng/libcollector/Makefile.am: Rename lib_LTLIBRARIES to
56649 pkglib_LTLIBRARIES. Add install-data-local.
56650 * gprofng/src/Makefile.am: Likewise.
56651 * gprofng/src/envsets.cc (putenv_libcollector_ld_misc): New location of
56652 the gprofng libraries.
56653 * gprofng/configure.ac: Removed an unused GPROFNG_LIBDIR.
56654 * gprofng/Makefile.am: Removed an unused GPROFNG_LIBDIR. Add
56655 install-data-local.
56656 * gprofng/configure: Regenerate.
56657 * gprofng/Makefile.in: Likewise.
56658 * gprofng/doc/Makefile.in: Likewise.
56659 * gprofng/gp-display-htmllibcollector/Makefile.in: Likewise.
56660 * gprofng/libcollector/Makefile.in: Likewise.
56661 * gprofng/src/Makefile.in: Likewise.
56663 2022-03-22 GDB Administrator <gdbadmin@sourceware.org>
56665 Automatic date update in version.in
56667 2022-03-21 Roland McGrath <mcgrathr@google.com>
56669 gdb: Add missing #include in solib.h
56670 The gdb_bfd_ref_ptr type is used in solib.h declarations.
56672 2022-03-21 Aaron Merey <amerey@redhat.com>
56674 PR gdb/27570: missing support for debuginfod in core_target::build_file_mappings
56675 Add debuginfod support to core_target::build_file_mappings and
56676 locate_exec_from_corefile_build_id to enable the downloading of
56677 missing executables and shared libraries referenced in core files.
56679 Also add debuginfod support to solib_map_sections so that previously
56680 downloaded shared libraries can be retrieved from the local debuginfod
56683 When core file shared libraries are found locally, verify that their
56684 build-ids match the corresponding build-ids found in the core file.
56685 If there is a mismatch, attempt to query debuginfod for the correct
56686 build and print a warning if unsuccessful:
56688 warning: Build-id of /lib64/libc.so.6 does not match core file.
56690 Also disable debuginfod when gcore invokes gdb. Debuginfo is not
56691 needed for core file generation so debuginfod queries will slow down
56692 gcore unnecessarily.
56694 2022-03-21 Aaron Merey <amerey@redhat.com>
56696 gdb: Add soname to build-id mapping for core files
56697 Since commit aa2d5a422 gdb has been able to read executable and shared
56698 library build-ids within core files.
56700 Expand this functionality so that each core file bfd maintains a map of
56701 soname to build-id for each shared library referenced in the core file.
56703 This feature may be used to verify that gdb has found the correct shared
56704 libraries for core files and to facilitate downloading shared libaries via
56707 2022-03-21 Pedro Alves <pedro@palves.net>
56709 Watchpoint followed by catchpoint misreports watchpoint (PR gdb/28621)
56710 If GDB reports a watchpoint hit, and then the next event is not
56711 TARGET_WAITKIND_STOPPED, but instead some event for which there's a
56712 catchpoint, such that GDB calls bpstat_stop_status, GDB mistakenly
56713 thinks the watchpoint triggered. Vis, using foll-fork.c:
56716 Hardware access (read/write) watchpoint 2: v
56718 Catchpoint 3 (fork)
56722 Hardware access (read/write) watchpoint 2: v
56726 main () at gdb.base/foll-fork.c:16
56731 Hardware access (read/write) watchpoint 2: v <<<<
56732 <<<< these lines are spurious
56735 Catchpoint 3 (forked process 1712369), arch_fork (ctid=0x7ffff7fa4810) at arch-fork.h:49
56736 49 arch-fork.h: No such file or directory.
56739 The problem is that when we handle the fork event, nothing called
56740 watchpoints_triggered before calling bpstat_stop_status. Thus, each
56741 watchpoint's watchpoint_triggered field was still set to
56742 watch_triggered_yes from the previous (real) watchpoint stop.
56743 watchpoint_triggered is only current called in the handle_signal_stop
56744 path, when handling TARGET_WAITKIND_STOPPED.
56746 This fixes it by adding watchpoint_triggered calls in the other events
56747 paths that call bpstat_stop_status. But instead of adding them
56748 explicitly, it adds a new function bpstat_stop_status_nowatch that
56749 wraps bpstat_stop_status and calls watchpoint_triggered, and then
56750 replaces most calls to bpstat_stop_status with calls to
56751 bpstat_stop_status_nowatch.
56753 This required constifying watchpoints_triggered.
56755 New test included, which fails without the fix.
56757 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28621
56759 Change-Id: I282b38c2eee428d25319af3bc842f9feafed461c
56761 2022-03-21 Pedro Alves <pedro@palves.net>
56763 gdbserver: Fixup previous patch
56764 The previous prepare_resume_reply change missed updating the 'buf'
56765 reference that overwrites the 'T', so if 'buf' was advanced, we'd
56766 still overwrite the wrong character. This fixes it.
56768 Change-Id: Ia8ce433366b85af4e268c1c49e7b447da3130a4d
56770 2022-03-21 Pedro Alves <pedro@palves.net>
56772 gdbserver: Fix incorrect assertion
56773 While playing with adding a new event kind, I noticed that
56774 prepare_resume_reply TARGET_WAITKIND_FORKED, etc. advance 'buf', so if
56775 we force-disable the T packet, we'd fail the *buf == 'T' assertion.
56777 Fix it by tweaking the assertion to always look at the beginning of
56780 Change-Id: I8c38e32353db115edcde418b3b1e8ba12343c22b
56782 2022-03-21 Simon Marchi <simon.marchi@efficios.com>
56784 gdb: re-generate config.in
56785 I'm getting this diff when running `autoreconf -vf`.
56787 Change-Id: Id5f009d0f0481935c1ee9df5332cb4bf45fbd32d
56789 2022-03-21 Andrew Burgess <aburgess@redhat.com>
56791 gdb/x86: handle stap probe arguments in xmm registers
56792 On x86 machines with xmm register, and with recent versions of
56793 systemtap (and gcc?), it can occur that stap probe arguments will be
56794 placed into xmm registers.
56796 I notice this happening on a current Fedora Rawhide install with the
56797 following package versions installed:
56799 $ rpm -q glibc systemtap gcc
56800 glibc-2.35.9000-10.fc37.x86_64
56801 systemtap-4.7~pre16468670g9f253544-1.fc37.x86_64
56802 gcc-12.0.1-0.12.fc37.x86_64
56804 If I check the probe data in libc, I see this:
56806 $ readelf -n /lib64/libc.so.6
56808 stapsdt 0x0000004d NT_STAPSDT (SystemTap probe descriptors)
56810 Name: pthread_start
56811 Location: 0x0000000000090ac3, Base: 0x00000000001c65c4, Semaphore: 0x0000000000000000
56812 Arguments: 8@%xmm1 8@1600(%rbx) 8@1608(%rbx)
56813 stapsdt 0x00000050 NT_STAPSDT (SystemTap probe descriptors)
56815 Name: pthread_create
56816 Location: 0x00000000000912f1, Base: 0x00000000001c65c4, Semaphore: 0x0000000000000000
56817 Arguments: 8@%xmm1 8@%r13 8@8(%rsp) 8@16(%rsp)
56820 Notice that for both of these probes, the first argument is a uint64_t
56821 stored in the xmm1 register.
56823 Unfortunately, if I try to use this probe within GDB, then I can't
56824 view the first argument. Here's an example session:
56829 (gdb) info probes stap libc pthread_create
56831 (gdb) break *0x00007ffff729e2f1 # Use address of probe.
56834 (gdb) p $_probe_arg0
56837 What's going wrong? If I re-run my session, but this time use 'set
56838 debug stap-expression 1', this is what I see:
56840 (gdb) set debug stap-expression 1
56841 (gdb) p $_probe_arg0
56842 Operation: UNOP_CAST
56843 Operation: OP_REGISTER
56846 Operation: UNOP_CAST
56847 Operation: OP_REGISTER
56850 Operation: UNOP_CAST
56851 Operation: UNOP_IND
56852 Operation: UNOP_CAST
56853 Operation: BINOP_ADD
56856 Constant: 0x0000000000000008
56857 Operation: OP_REGISTER
56861 Operation: UNOP_CAST
56862 Operation: UNOP_IND
56863 Operation: UNOP_CAST
56864 Operation: BINOP_ADD
56867 Constant: 0x0000000000000010
56868 Operation: OP_REGISTER
56875 The important bit is this:
56877 Operation: UNOP_CAST
56878 Operation: OP_REGISTER
56882 Which is where we cast the xmm1 register to uint64_t. And the final
56883 piece of the puzzle is:
56886 type = union vec128 {
56887 v8bf16 v8_bfloat16;
56897 So, we are attempting to cast a union type to a scalar type, which is
56898 not supporting in C/C++, and as a consequence GDB's expression
56899 evaluator throws an error when we attempt to do this.
56901 The first approach I considered for solving this problem was to try
56902 and make use of gdbarch_stap_adjust_register. We already have a
56903 gdbarch method (gdbarch_stap_adjust_register) that allows us to tweak
56904 the name of the register that we access. Currently only x86
56905 architectures use this to transform things like ax to eax in some
56908 I wondered, what if we change gdbarch_stap_adjust_register to do more
56909 than just change the register names? What if this method instead
56910 became gdbarch_stap_read_register. This new method would return a
56911 operation_up, and would take the register name, and the type we are
56912 trying to read from the register, and return the operation that
56913 actually reads the register.
56915 The default implementation of this method would just use
56916 user_reg_map_name_to_regnum, and then create a register_operation,
56917 like we already do in stap_parse_register_operand. But, for x86
56918 architectures this method would fist possibly adjust the register
56919 name, then do the default action to read the register. Finally, for
56920 x86 this method would spot when we were accessing an xmm register,
56921 and, based on the type being pulled from the register, would extract
56922 the correct field from the union.
56924 The benefit of this approach is that it would work with the expression
56925 types that GDB currently supports. The draw back would be that this
56926 approach would not be very generic. We'd need code to handle each
56927 sub-field size with an xmm register. If other architectures started
56928 using vector registers for probe arguments, those architectures would
56929 have to create their own gdbarch_stap_read_register method. And
56930 finally, the type of the xmm registers comes from the type defined in
56931 the target description, there's a risk that GDB might end up
56932 hard-coding the names of type sub-fields, then if a target uses a
56933 different target description, with different field names for xmm
56934 registers, the stap probes would stop working.
56936 And so, based on all the above draw backs, I rejected this first
56939 My second plan involves adding a new expression type to GDB called
56940 unop_extract_operation. This new expression takes a value and a type,
56941 during evaluation the value contents are fetched, and then a new value
56942 is extracted from the value contents (based on type). This is similar
56943 to the following C expression:
56945 result_value = *((output_type *) &input_value);
56947 Obviously we can't actually build this expression in this case, as the
56948 input_value is in a register, but hopefully the above makes it clearer
56949 what I'm trying to do.
56951 The benefit of the new expression approach is that this code can be
56952 shared across all architectures, and it doesn't care about sub-field
56953 names within the union type.
56955 The draw-backs that I see are potential future problems if arguments
56956 are not stored within the least significant bytes of the register.
56957 However if/when that becomes an issue we can adapt the
56958 gdbarch_stap_read_register approach to allow architectures to control
56959 how a value is extracted.
56961 For testing, I've extended the existing gdb.base/stap-probe.exp test
56962 to include a function that tries to force an argument into an xmm
56963 register. Obviously, that will only work on a x86 target, so I've
56964 guarded the new function with an appropriate GCC define. In the exp
56965 script we use readelf to check if the probe exists, and is using the
56968 If the probe doesn't exist then the associated tests are skipped.
56970 If the probe exists, put isn't using the xmm register (which will
56971 depend on systemtap/gcc versions), then again, the tests are skipped.
56973 Otherwise, we can run the test. I think the cost of running readelf
56974 is pretty low, so I don't feel too bad making all the non-xmm targets
56977 I found that on a Fedora 35 install, with these packages installed, I
56978 was able to run this test and have the probe argument be placed in an
56981 $ rpm -q systemtap gcc glibc
56982 systemtap-4.6-4.fc35.x86_64
56983 gcc-11.2.1-9.fc35.x86_64
56984 glibc-2.34-7.fc35.x86_64
56986 Finally, as this patch adds a new operation type, then I need to
56987 consider how to generate an agent expression for the new operation
56990 I have kicked the can down the road a bit on this. In the function
56991 stap_parse_register_operand, I only create a unop_extract_operation in
56992 the case where the register type is non-scalar, this means that in
56993 most cases I don't need to worry about generating an agent expression
56996 In the xmm register case, when an unop_extract_operation will be
56997 created, I have sketched out how the agent expression could be
56998 handled, however, this code is currently not reached. When we try to
56999 generate the agent expression to place the xmm register on the stack,
57000 GDB hits this error:
57002 (gdb) trace -probe-stap test:xmmreg
57003 Tracepoint 1 at 0x401166
57005 Enter actions for tracepoint 1, one per line.
57006 End with a line saying just "end".
57007 >collect $_probe_arg0
57008 Value not scalar: cannot be an rvalue.
57010 This is because GDB doesn't currently support placing non-scalar types
57011 on the agent expression evaluation stack. Solving this is clearly
57012 related to the original problem, but feels a bit like a second
57013 problem. I'd like to get feedback on whether my approach to solving
57014 the original problem is acceptable or not before I start looking at
57015 how to handle xmm registers within agent expressions.
57017 2022-03-21 Steiner H Gunderson <steinar+sourceware@gunderson.no>
57019 Reduce O(n2) performance overhead when parsing DWARF unit information.
57021 * dwarf2.c (scan_unit_for_symbols): When performing second pass,
57022 check to see if the function or variable being processed is the
57023 same as the previous one.
57025 2022-03-21 Jan Beulich <jbeulich@suse.com>
57027 x86: don't suppress overflow diagnostics in x32 mode
57028 Unlike in 64-bit mode, where values wrap at the 64-bit boundary anyway,
57029 there's no wrapping at the 32-bit boundary here, and hence overflow
57030 detection shouldn't be suppressed just because rela relocations are
57033 The extra check against NO_RELOC is actually a result of an ilp32 test
57034 otherwise failing. But thinking about it, reporting overflows for
57035 not-really-relocations (typically because of earlier errors) makes
57036 little sense in general. Perhaps this should even be extended to non-
57039 2022-03-21 Simon Marchi <simon.marchi@efficios.com>
57041 gdb/testsuite: reformat gdb.python/pretty-print-call-by-hand.py
57042 Run black on the file.
57044 Change-Id: Ifb576137fb7158a0227173f61c1202f0695b3685
57046 2022-03-21 Bruno Larsen <blarsen@redhat.com>
57048 [gdb/testsuite] test a function call by hand from pretty printer
57049 The test case added here is testing the bug gdb/28856, where calling a
57050 function by hand from a pretty printer makes GDB crash. There are 6
57051 mechanisms to trigger this crash in the current test, using the commands
57052 backtrace, up, down, finish, step and continue. Since the failure happens
57053 because of use-after-free (more details below) the tests will always
57054 have a chance of passing through sheer luck, but anecdotally they seem
57055 to fail all of the time.
57057 The reason GDB is crashing is a use-after-free problem. The above
57058 mentioned functions save a pointer to the current frame's information,
57059 then calls the pretty printer, and uses the saved pointer for different
57060 reasons, depending on the function. The issue happens because
57061 call_function_by_hand needs to reset the obstack to get the current
57062 frame, invalidating the saved pointer.
57064 2022-03-21 Pedro Alves <pedro@palves.net>
57066 gdb/testsuite: Installed-GDB testing & data-directory
57067 In testsuite/README, we suggest that you can run the testsuite against
57068 some other GDB binary by using:
57070 make check RUNTESTFLAGS=GDB=/usr/bin/gdb
57072 However, that example isn't fully correct, because with that command
57073 line, the testsuite will still pass
57075 -data-directory=[pwd]/../data-directory
57077 to /usr/bin/gdb, like e.g.:
57080 builtin_spawn /usr/bin/gdb -nw -nx -data-directory /home/pedro/gdb/build/gdb/testsuite/../data-directory -iex set height 0 -iex set width 0
57083 while if you're testing an installed GDB (the system GDB being the
57084 most usual scenario), then you should normally let it use its own
57085 configured directory, not the just-built GDB's data directory.
57087 This commit improves the status quo with the following two changes:
57089 - if the user specifies GDB on the command line, then by default,
57090 don't start GDB with the -data-directory command line option.
57091 I.e., let the tested GDB use its own configured data directory.
57093 - let the user override the data directory, via a new
57094 GDB_DATA_DIRECTORY global. This replaces the existing
57095 BUILD_DATA_DIRECTORY variable in testsuite/lib/gdb.exp, which
57096 wasn't overridable, and was a bit misnamed for the new purpose.
57098 So after this, the following commands I believe behave intuitively:
57100 # Test the non-installed GDB in some build dir:
57103 RUNTESTFLAGS="GDB=/path/to/other/build/gdb \
57104 GDB_DATA_DIRECTORY=/path/to/other/build/gdb/data-directory"
57106 # Test the GDB installed in some prefix:
57109 RUNTESTFLAGS="GDB=/opt/gdb/bin/gdb"
57111 # Test the built GDB with some alternative data directory, e.g., the
57112 system GDB's data directory:
57115 RUNTESTFLAGS="GDB_DATA_DIRECTORY=/usr/share/gdb"
57117 Change-Id: Icdc21c85219155d9564a9900961997e6624b78fb
57119 2022-03-21 Nick Clifton <nickc@redhat.com>
57121 z80 assembler: Fix new unexpected overflow warning in v2.37
57123 * config/tc-z80.c (emit_data_val): Do not warn about overlarge
57124 constants generated by bit manipulation operators.
57125 * testsuite/gas/z80/pr28791.s: New test source file.
57126 * testsuite/gas/z80/pr28791.d: New test driver file.
57128 2022-03-21 Andreas Schwab <schwab@linux-m68k.org>
57130 Add support for readline 8.2
57131 In readline 8.2 the type of rl_completer_word_break_characters changed to
57134 2022-03-21 GDB Administrator <gdbadmin@sourceware.org>
57136 Automatic date update in version.in
57138 2022-03-20 Andreas Schwab <schwab@linux-m68k.org>
57140 RISC-V: Fix misplaced @end table
57141 Move the csr-check and arch items inside the table for the .option directive.
57143 2022-03-20 Alan Modra <amodra@gmail.com>
57145 PR28979, internal error in demand_empty_rest_of_line
57146 The change in read_a_source_file prevents the particular testcase in
57147 the PR from triggering the assertion in demand_empty_rest_of_line.
57148 I've also removed the assertion. Nothing much goes wrong with gas if
57149 something else triggers it, so it's not worthy of an abort.
57151 I've also changed my previous patch to ignore_rest_of_line to allow
57152 that function to increment input_line_pointer past buffer_limit, like
57153 demand_empty_rest_of_line: The two functions ought to behave the
57154 same in that respect. Finally, demand_empty_rest_of_line gets a
57155 little hardening to prevent accesses past buffer_limit plus one.
57158 * read.c (read_a_source_file): Calculate known size for sbuf
57159 rather than calling strlen.
57160 (demand_empty_rest_of_line): Remove "know" check. Expand comment.
57161 Don't dereference input_line_pointer when past buffer_limit.
57162 (ignore_rest_of_line): Allow input_line_pointer to increment to
57163 buffer_limit plus one. Expand comment.
57165 2022-03-20 Joel Brobecker <brobecker@adacore.com>
57167 Update gdb/NEWS after GDB 12 branch creation.
57168 This commit a new section for the next release branch, and renames
57169 the section of the current branch, now that it has been cut.
57171 2022-03-20 Joel Brobecker <brobecker@adacore.com>
57173 Bump version to 13.0.50.DATE-git.
57174 Now that the GDB 12 branch has been created,
57175 this commit bumps the version number in gdb/version.in to
57178 For the record, the GDB 12 branch was created
57179 from commit 2be64de603f8b3ae359d2d3fbf5db0e79869f32b.
57181 Also, as a result of the version bump, the following changes
57182 have been made in gdb/testsuite:
57184 * gdb.base/default.exp: Change $_gdb_major to 13.
57186 2022-03-20 liuzhensong <liuzhensong@loongson.cn>
57188 ld:LoongArch: Add test cases to adapt to LoongArch32 and LoongArch64
57189 ld/testsuite/ld-loongarch-elf
57191 * ld-loongarch-elf.exp: Test LoongArch32 and LoongArch64 testcases respectively.
57192 * jmp_op.d: Fix bug in test LoongArch32.
57193 * disas-jirl-32.d: New test case for LoongArch32.
57194 * disas-jirl-32.s: New test case for LoongArch32.
57195 * disas-jirl.d: Skip test case LoongArch32.
57196 * macro_op_32.d: New test case for LoongArch32.
57197 * macro_op_32.s: New test case for LoongArch32.
57198 * macro_op.d: Skip test case LoongArch32.
57200 2022-03-20 liuzhensong <liuzhensong@loongson.cn>
57202 gas:LoongArch: Fix "make check" pr21884 fail in LoongArch32.
57204 * tc-loongarch.c: Add function to select target mach.
57205 * tc-loongarch.h: Define macro TARGET_MACH.
57207 ld: loongarch: Skip unsupport test cases.
57208 ld/testsuite/ld-elf/
57209 * eh5.d Skip loongarch64 target.
57210 * pr21884.d Skip loongarch* targets.
57211 * pr26936.d Skip loongarch* targets.
57213 2022-03-20 liuzhensong <liuzhensong@loongson.cn>
57215 LoongArch: Fix LD check fails.
57216 Some test cases about ifunc.
57219 * elfnn-loongarch.c
57220 * elfxx-loongarch.h
57223 of expected passes 1430
57224 of expected failures 11
57225 of untested testcases 1
57226 of unsupported tests 154
57228 2022-03-20 liuzhensong <liuzhensong@loongson.cn>
57230 LoongArch: Update ABI eflag in elf header.
57231 Update LoongArch ABI eflag in elf header.
57240 * elfnn-loongarch.c Check object flags while ld.
57243 * tc-loongarch.c Write eflag to elf header.
57246 * loongarch.h Define ABI number.
57248 2022-03-20 liuzhensong <liuzhensong@loongson.cn>
57250 gas:LoongArch: Fix wrong line number in .debug_line
57251 The dwarf2_emit_insn() can create debuginfo of line. But it is called
57252 too late in append_fixp_and_insn. It causes extra offs when debuginfo
57253 of line sets address.
57258 2022-03-20 liuzhensong <liuzhensong@loongson.cn>
57260 gas:LoongArch: Fix segment error in compilation due to too long symbol name.
57261 Change "char buffer[8192];" into "char *buffer =
57262 (char *) malloc(1000 + 6 * len_str);" in function
57263 loongarch_expand_macro_with_format_map.
57266 * config/tc-loongarch.c
57269 * opcode/loongarch.h
57272 * loongarch-coder.c
57274 2022-03-20 liuzhensong <liuzhensong@loongson.cn>
57276 LoongArch: Use functions instead of magic numbers.
57277 Replace the magic numbers in gas(tc-loongarch.c) and
57278 bfd(elfnn-loongarch.c) with the functions defined in
57279 the howto table(elfxx-loongarch.c).
57282 * config/tc-loongarch.c: use functions.
57285 * elfnn-loongarch.c: use functions.
57286 * elfxx-loongarch.c: define functions.
57287 * elfxx-loongarch.h
57289 2022-03-20 liuzhensong <liuzhensong@loongson.cn>
57291 ubsan: loongarch : signed integer shift overflow.
57293 * loongarch-coder.c :
57295 ret <<= sizeof (ret) * 8 - len;
57296 ret >>= sizeof (ret) * 8 - len;
57298 Avoid ubsan warning.
57300 2022-03-20 GDB Administrator <gdbadmin@sourceware.org>
57302 Automatic date update in version.in
57304 2022-03-19 Simon Marchi <simon.marchi@efficios.com>
57306 gdb/python: remove gdb._mi_commands dict
57307 The motivation for this patch is the fact that py-micmd.c doesn't build
57308 with Python 2, due to PyDict_GetItemWithError being a Python 3-only
57311 CXX python/py-micmd.o
57312 /home/smarchi/src/binutils-gdb/gdb/python/py-micmd.c: In function ‘int micmdpy_uninstall_command(micmdpy_object*)’:
57313 /home/smarchi/src/binutils-gdb/gdb/python/py-micmd.c:430:20: error: ‘PyDict_GetItemWithError’ was not declared in this scope; did you mean ‘PyDict_GetItemString’?
57314 430 | PyObject *curr = PyDict_GetItemWithError (mi_cmd_dict.get (),
57315 | ^~~~~~~~~~~~~~~~~~~~~~~
57316 | PyDict_GetItemString
57318 A first solution to fix this would be to try to replace
57319 PyDict_GetItemWithError equivalent Python 2 code. But I looked at why
57320 we are doing this in the first place: it is to maintain the
57321 `gdb._mi_commands` Python dictionary that we use as a `name ->
57322 gdb.MICommand object` map. Since the `gdb._mi_commands` dictionary is
57323 never actually used in Python, it seems like a lot of trouble to use a
57324 Python object for this.
57326 My first idea was to replace it with a C++ map
57327 (std::unordered_map<std::string, gdbpy_ref<micmdpy_object>>). While
57328 implementing this, I realized we don't really need this map at all. The
57329 mi_command_py objects registered in the main MI command table can own
57330 their backing micmdpy_object (that's a gdb.MICommand, but seen from the
57331 C++ code). To know whether an mi_command is an mi_command_py, we can
57332 use a dynamic cast. Since there's one less data structure to maintain,
57333 there are less chances of messing things up.
57335 - Change mi_command_py::m_pyobj to a gdbpy_ref, the mi_command_py is
57336 now what keeps the MICommand alive.
57337 - Set micmdpy_object::mi_command in the constructor of mi_command_py.
57338 If mi_command_py manages setting/clearing that field in
57339 swap_python_object, I think it makes sense that it also takes care of
57340 setting it initially.
57341 - Move a bunch of checks from micmdpy_install_command to
57342 swap_python_object, and make them gdb_asserts.
57343 - In micmdpy_install_command, start by doing an mi_cmd_lookup. This is
57344 needed to know whether there's a Python MI command already registered
57345 with that name. But we can already tell if there's a non-Python
57346 command registered with that name. Return an error if that happens,
57347 rather than waiting for insert_mi_cmd_entry to fail. Change the
57348 error message to "name is already in use" rather than "may already be
57349 in use", since it's more precise.
57351 I asked Andrew about the original intent of using a Python dictionary
57352 object to hold the command objects. The reason was to make sure the
57353 objects get destroyed when the Python runtime gets finalized, not later.
57354 Holding the objects in global C++ data structures and not doing anything
57355 more means that the held Python objects will be decref'd after the
57356 Python interpreter has been finalized. That's not desirable. I tried
57357 it and it indeed segfaults.
57359 Handle this by adding a gdbpy_finalize_micommands function called in
57360 finalize_python. This is the mirror of gdbpy_initialize_micommands
57361 called in do_start_initialization. In there, delete all Python MI
57362 commands. I think it makes sense to do it this way: if it was somehow
57363 possible to unload Python support from GDB in the middle of a session
57364 we'd want to unregister any Python MI command. Otherwise, these MI
57365 commands would be backed with a stale PyObject or simply nothing.
57367 Delete tests that were related to `gdb._mi_commands`.
57369 Co-Authored-By: Andrew Burgess <aburgess@redhat.com>
57370 Change-Id: I060d5ebc7a096c67487998a8a4ca1e8e56f12cd3
57372 2022-03-19 GDB Administrator <gdbadmin@sourceware.org>
57374 Automatic date update in version.in
57376 2022-03-18 Pedro Alves <pedro@palves.net>
57378 Fix crash with stepi, no debug info, and "set debug infrun 1"
57379 A stepi in a function without debug info with "set debug infrun 1"
57380 crashes GDB since commit c8353d682f69 (gdb/infrun: some extra infrun
57381 debug print statements), due to a reference to
57382 "tp->current_symtab->filename" when tp->current_symtab is null.
57384 This commit adds the missing null check. The output in this case
57387 [infrun] set_step_info: symtab = <null>, line = 0, step_frame_id = {stack=0x7fffffffd980,code=0x0000000000456c30,!special}, step_stack_frame_id = {stack=0x7fffffffd980,code=0x0000000000456c30,!special}
57389 Change-Id: I5171a9d222bc7e15b9dba2feaba7d55c7acd99f8
57391 2022-03-18 Tom Tromey <tromey@adacore.com>
57393 Implement gdbarch_stack_frame_destroyed_p for aarch64
57394 The internal AdaCore testsuite has a test that checks that an
57395 out-of-scope watchpoint is deleted. This fails on some aarch64
57396 configurations, reporting an extra stop:
57401 Thread 3 hit Watchpoint 2: result
57405 0x0000000040021648 in pck.get_val (seed=0, off_by_one=false) at [...]/pck.adb:13
57408 I believe what is happening here is that the variable is stored at:
57410 <efa> DW_AT_location : 2 byte block: 91 7c (DW_OP_fbreg: -4)
57412 and the extra stop is reported just before a return, when the ldp
57413 instruction is executed:
57415 0x0000000040021644 <+204>: ldp x29, x30, [sp], #48
57416 0x0000000040021648 <+208>: ret
57418 This instruction modifies the frame base calculation, and so the test
57419 picks up whatever memory is pointed to in the callee frame.
57421 Implementing the gdbarch hook gdbarch_stack_frame_destroyed_p fixes
57424 As usual with this sort of patch, it has passed internal testing, but
57425 I don't have a good way to try it with dejagnu. So, I don't know
57426 whether some existing test covers this. I suspect there must be one,
57427 but it's also worth noting that this test passes for aarch64 in some
57428 configurations -- I don't know what causes one to fail and another to
57431 2022-03-18 Nick Clifton <nickc@redhat.com>
57433 Fix Build issues due to patch "gprofng: a new GNU profiler"
57434 Find and fix more places where clock_gettime() and CLOCK_MONOTONIC_RAW are used.
57436 2022-03-18 Simon Marchi <simon.marchi@efficios.com>
57438 gdb: run black to format some Python files
57439 Seems like some leftovers from previous commits.
57441 Change-Id: I7155ccdf7a4fef83bcb3d60320252c3628efdb83
57443 2022-03-18 Viorel Preoteasa <viorel.preoteasa@gmail.com>
57445 Fix ld-arm bug in encoding of blx calls jumping from thumb to arm instructions
57447 * elf32-arm.c (THM_MAX_FWD_BRANCH_OFFSET): Fix definition.
57448 (THM2_MAX_FWD_BRANCH_OFFSET): Likewise.
57450 2022-03-18 Jan Beulich <jbeulich@suse.com>
57452 x86: also fold remaining multi-vector-size shift insns
57453 By slightly relaxing the checking in operand_type_register_match() we
57454 can fold the vector shift insns with an XMM source as well. While
57455 strictly speaking an overlap in just one size (see the code comment) is
57456 not enough (both operands could have multiple sizes with just a single
57457 common one), this is good enough for all templates we have, or which
57458 could sensibly / usefully appear (within the scope of the present
57459 operand matching model).
57461 Tightening this a little would be possible, but would require broadcast
57462 related information to be passed into the function.
57464 2022-03-18 Jan Beulich <jbeulich@suse.com>
57466 x86: drop stray CheckRegSize from VEXTRACT{F,I}32X4
57467 They have only a single operand allowing multiple sizes, hence there are
57468 no pairs of operands to check for consistent size.
57470 x86: fold certain AVX2 templates into their AVX counterparts
57471 Like for AVX512VL we can make the handling of operand sizes a little
57472 more flexible to allow reducing the number of templates we have.
57474 2022-03-18 Tsukasa OI <research_trasio@irq.a4lg.com>
57476 RISC-V: Cache management instructions
57477 This commit adds 'Zicbom' / 'Zicboz' instructions.
57481 * elfxx-riscv.c (riscv_multi_subset_supports): Add handling for
57482 new instruction classes.
57486 * opcode/riscv-opc.h (MATCH_CBO_CLEAN, MASK_CBO_CLEAN,
57487 MATCH_CBO_FLUSH, MASK_CBO_FLUSH, MATCH_CBO_INVAL,
57488 MASK_CBO_INVAL, MATCH_CBO_ZERO, MASK_CBO_ZERO): New macros.
57489 * opcode/riscv.h (enum riscv_insn_class): Add new instruction
57490 classes INSN_CLASS_ZICBOM and INSN_CLASS_ZICBOZ.
57494 * riscv-opc.c (riscv_opcodes): Add cache-block management
57497 2022-03-18 Tsukasa OI <research_trasio@irq.a4lg.com>
57499 RISC-V: Prefetch hint instructions and operand set
57500 This commit adds 'Zicbop' hint instructions.
57504 * elfxx-riscv.c (riscv_multi_subset_supports): Add handling for
57505 new instruction class.
57509 * config/tc-riscv.c (riscv_ip): Add handling for new operand
57510 type 'f' (32-byte aligned pseudo S-type immediate for prefetch
57512 (validate_riscv_insn): Likewise.
57516 * opcode/riscv-opc.h (MATCH_PREFETCH_I, MASK_PREFETCH_I,
57517 MATCH_PREFETCH_R, MASK_PREFETCH_R, MATCH_PREFETCH_W,
57518 MASK_PREFETCH_W): New macros.
57519 * opcode/riscv.h (enum riscv_insn_class): Add new instruction
57520 class INSN_CLASS_ZICBOP.
57524 * riscv-dis.c (print_insn_args): Add handling for new operand
57526 * riscv-opc.c (riscv_opcodes): Add prefetch hint instructions.
57528 2022-03-18 Alan Modra <amodra@gmail.com>
57530 PR28977 tc-i386.c internal error in parse_register
57532 * config/tc-i386.c (parse_register): Handle X_op not O_register
57533 as for a non-reg_section symbol. Simplify array bounds check.
57535 2022-03-18 Alan Modra <amodra@gmail.com>
57537 Tidy gas current_frame before exit
57538 Releases some obstack memory on an error path.
57540 * cond.c (cond_finish_check): Call cond_exit_macro.
57542 2022-03-18 Alan Modra <amodra@gmail.com>
57544 ubsan: logical_input_line signed integer overflow
57545 To avoid a completely useless fuzzing ubsan "bug" report, I decided to
57546 make logical_input_line unsigned.
57548 * input-scrub.c (logical_input_line): Make unsigned.
57549 (struct input_save): Here too.
57550 (input_scrub_reinit, input_scrub_close, bump_line_counters),
57551 (as_where): Adjust to suit.
57553 2022-03-18 GDB Administrator <gdbadmin@sourceware.org>
57555 Automatic date update in version.in
57557 2022-03-17 H.J. Lu <hjl.tools@gmail.com>
57559 gprofng: Skip jsynprog with a broken javac
57560 On CET enabled Linux/x86-64 machines, one can get
57562 $ javac simple.java
57563 Error: dl failure on line 894
57564 Error: failed /usr/lib/jvm/java-1.8.0-openjdk-1.8.0.322.b06-6.fc35.x86_64/jre/lib/amd64/server/libjvm.so, because /usr/lib/jvm/java-1.8.0-openjdk-1.8.0.322.b06-6.fc35.x86_64/jre/lib/amd64/server/libjvm.so: rebuild shared object with SHSTK support enabled
57566 Set GPROFNG_BROKEN_JAVAC to "yes" only with a broken javac and skip the
57567 jsynprog test with a broken javac.
57570 * Makefile.am (GPROFNG_BROKEN_JAVAC): New.
57571 (check-DEJAGNU): Pass GPROFNG_BROKEN_JAVAC to runtest.
57572 * configure.ac (GPROFNG_BROKEN_JAVAC): New AC_SUBST. Set to yes
57573 with a broken javac.
57574 * Makefile.in: Regenerate.
57575 * configure: Likewise.
57576 * testsuite/gprofng.display/display.exp: Skip jsynprog with a
57579 2022-03-17 John Baldwin <jhb@FreeBSD.org>
57581 Remove fall throughs in core_target::xfer_partial.
57582 The cases for TARGET_OBJECT_LIBRARIES and TARGET_OBJECT_LIBRARIES_AIX
57583 can try to fetch different data objects (such as
57584 TARGET_OBJECT_SIGNAL_INFO) if gdbarch methods for the requested data
57585 aren't present. Return with TARGET_XFER_E_IO if the gdbarch method
57586 isn't present instead.
57588 2022-03-17 Pedro Alves <pedro@palves.net>
57590 gdb: Remove support for S+core
57591 GCC removed support for score back in 2014 already. Back then, we
57592 basically agreed about removing it from GDB too, but it ended up being
57595 https://sourceware.org/pipermail/gdb/2014-October/044643.html
57597 Following through this time around.
57599 Change-Id: I5b25a1ff7bce7b150d6f90f4c34047fae4b1f8b4
57601 2022-03-17 Tom Tromey <tromey@adacore.com>
57603 Add another test for Ada Wide_Wide_String
57604 In an earlier patch, I had written that I wanted to add this test:
57606 ptype Wide_Wide_String'("literal")
57608 ... but that it failed with the distro GNAT. Further investigation
57609 showed that it could be made to work by adding a function using
57610 Wide_Wide_String to the program -- this caused the type to end up in
57613 This patch adds the test. I'm checking this in.
57615 2022-03-17 Alan Modra <amodra@gmail.com>
57617 ubsan: Null dereference in parse_module
57618 * vms-alpha.c (parse_module): Sanity check that DST__K_RTNBEG
57619 has set module->func_table for DST__K_RTNEND. Check return
57622 2022-03-17 Alan Modra <amodra@gmail.com>
57624 asan: Buffer overflow in evax_bfd_print_dst
57625 With "name" a char*, the length at name[0] might be negative, escaping
57626 buffer limit checks.
57628 * vms-alpha.c (evax_bfd_print_dst): Make name an unsigned char*.
57629 (evax_bfd_print_emh): Likewise.
57631 2022-03-17 Alan Modra <amodra@gmail.com>
57633 asan: Buffer overflow in som_set_reloc_info
57634 * som.c (som_set_reloc_info): Add symcount parameter. Don't
57635 access symbols past symcount. Don't access fixup past end_fixups.
57636 (som_slurp_reloc_table): Adjust som_set_reloc_info calls.
57638 2022-03-17 Alan Modra <amodra@gmail.com>
57640 asan: use of uninitialized value in buffer_and_nest
57641 More occurences of the same as commit d12b8d620c6a.
57643 * macro.c (buffer_and_nest): Sanity check length in buffer
57644 before calling strncasecmp.
57646 2022-03-17 Alan Modra <amodra@gmail.com>
57648 gprofng configure target tests
57649 ${target} in configure.ac should be the canonical target, so that for
57650 example, someone configuring with --target=x86_64-linux will match
57653 * configure.ac: Invoke AC_CANONICAL_TARGET.
57654 * libcollector/configure.ac: Likewise.
57655 * Makefile.in: Regenerate.
57656 * configure: Regenerate.
57657 * doc/Makefile.in: Regenerate.
57658 * gp-display-html/Makefile.in: Regenerate.
57659 * libcollector/Makefile.in: Regenerate.
57660 * libcollector/configure: Regenerate.
57661 * src/Makefile.in: Regenerate.
57663 2022-03-17 Alan Modra <amodra@gmail.com>
57665 Re: asan: buffer overflow in peXXigen.c
57666 In the process of fixing a buffer overflow in commit fe69d4fcf0194a,
57667 I managed to introduce a fairly obvious NULL pointer dereference..
57669 * peXXigen.c (_bfd_XX_bfd_copy_private_bfd_data_common): Don't
57670 segfault on not finding section. Wrap overlong lines.
57672 2022-03-17 Alan Modra <amodra@gmail.com>
57674 asan: buffer overflows after calling ignore_rest_of_line
57675 operand() is not a place that should be calling ignore_rest_of_line.
57676 ignore_rest_of_line shouldn't increment input_line_pointer if already
57679 * expr.c (operand): Don't call ignore_rest_of_line.
57680 * read.c (s_mri_common): Likewise.
57681 (ignore_rest_of_line): Don't increment input_line_pointer if
57682 already at buffer_limit.
57684 2022-03-17 Alan Modra <amodra@gmail.com>
57686 Re: bfd: add AMDGCN architecture
57687 * po/SRC-POTFILES.in: Regenerate.
57689 2022-03-17 Jan Beulich <jbeulich@suse.com>
57691 x86: don't accept base architectures as extensions
57692 The -march= intentions are quite clear: A base architecture may be
57693 followed by any number of extensions. Accepting a base architecture in
57694 place of an extension will at best result in confusion, as the first of
57695 the two (or more) items specified simply would not take effect, due to
57696 being overridden by the later one(s).
57698 2022-03-17 Jan Beulich <jbeulich@suse.com>
57700 x86: never set i386_cpu_flags' "unused" field
57701 Setting this field risks cpu_flags_all_zero() mistakenly returning
57702 "false" when the object passed in was e.g. the result of ANDing together
57703 two objects which had the bit set, or ANDNing together an object with
57704 the field set and one with the field clear.
57706 While there also avoid setting CpuNo64: Like Cpu64 this is driven
57707 differently anyway and hence shouldn't be set anywhere by default.
57709 Note that the moving of the two items in i386-gen.c's cpu_flags[] is
57710 only for documentation purposes (and slight reducing of overhead), as
57711 the fields are sorted anyway upon program start.
57713 2022-03-17 Jan Beulich <jbeulich@suse.com>
57715 x86: unify CPU flag on/off processing
57716 There's no need for the arbitrary special "unknown" token: Simply
57717 recognize the leading ~ and process everything else the same, merely
57718 recording whether to set individual fields to 1 or 0.
57720 While there exclude CpuIAMCU from CPU_UNKNOWN_FLAGS - CPU_IAMCU_FLAGS
57721 override cpu_arch_flags anyway when -march=iamcu is passed, and there's
57722 no reason to have the stray flag set even if no insn actually is keyed
57725 2022-03-17 Jan Beulich <jbeulich@suse.com>
57727 x86: add another IAMCU testcase
57728 Now that {L,K}1OM support is gone, and with it the brokenness in
57729 check_cpu_arch_compatible(), put in place a test making sure that only
57730 extensions can be enabled via .arch for IAMCU, and that the base
57731 architecture cannot be changed.
57733 x86: drop L1OM/K1OM support from gas
57734 This was only rudimentary support anyway; none of the sub-architecture
57735 specific insns were ever supported.
57737 x86: assorted IAMCU CPU checking fixes
57738 The checks done by check_cpu_arch_compatible() were halfway sensible
57739 only at the time where only L1OM support was there. The purpose,
57740 however, has always been to prevent bad uses of .arch (turning off the
57741 base CPU "feature" flag) while at the same time permitting extensions to
57742 be enabled / disabled. In order to achieve this (and to prevent
57743 regressions when L1OM and K1OM support are removed)
57744 - set CpuIAMCU in CPU_IAMCU_FLAGS,
57745 - adjust the IAMCU check in the function itself (the other two similarly
57746 broken checks aren't adjusted as they're slated to be removed anyway),
57747 - avoid calling the function for extentions (which would never have the
57748 base "feature" flag set),
57749 - add a new testcase actually exercising ".arch iamcu" (which would also
57750 regress with the planned removal).
57752 2022-03-17 GDB Administrator <gdbadmin@sourceware.org>
57754 Automatic date update in version.in
57756 2022-03-16 Andrew Burgess <aburgess@redhat.com>
57758 gdb: work around prompt corruption caused by bracketed-paste-mode
57761 commit b4f26d541aa7224b70d363932e816e6e1a857633
57762 Date: Tue Mar 2 13:42:37 2021 -0700
57764 Import GNU Readline 8.1
57766 We imported readline 8.1 into GDB. As a consequence bug PR cli/28833
57767 was reported. This bug spotted that, when the user terminated GDB by
57768 sending EOF (usually bound to Ctrl+d), the last prompt would become
57769 corrupted. Here's what happens, the user is sat at a prompt like
57774 And then the user sends EOF (Ctrl+d), we now see this:
57777 ... gdb terminates, and we return to the shell ...
57779 Notice the 'quit' was printed over the prompt.
57781 This problem is a result of readline 8.1 enabling bracketed paste mode
57782 by default. This problem is present in readline 8.0 too, but in that
57783 version of readline bracketed paste mode is off by default, so a user
57784 will not see the bug unless they specifically enable the feature.
57786 Bracketed paste mode is available in readline 7.0 too, but the bug
57787 is not present in this version of readline, see below for why.
57789 What causes this problem is how readline disables bracketed paste
57790 mode. Bracketed paste mode is a terminal feature that is enabled and
57791 disabled by readline emitting a specific escape sequence. The problem
57792 for GDB is that the escape sequence to disable bracketed paste mode
57793 includes a '\r' character at the end, see this thread for more
57796 https://lists.gnu.org/archive/html/bug-bash/2018-01/msg00097.html
57798 The change to add the '\r' character to the escape sequence used to
57799 disable bracketed paste mode was introduced between readline 7.0 and
57800 readline 8.0, this is why the bug would not occur when using older
57801 versions of readline (note: I don't know if its even possible to build
57802 GDB using readline 7.0. That really isn't important, I'm just
57803 documenting the history of this issue).
57805 So, the escape sequence to disable bracketed paste mode is emitted
57806 from the readline function rl_deprep_terminal, this is called after
57807 the user has entered a complete command and pressed return, or, if the
57810 However, these two cases are slightly different. In the first case,
57811 when the user has entered a command and pressed return, the cursor
57812 will have moved to the next, empty, line, before readline emits the
57813 escape sequence to leave bracketed paste mode. The final '\r'
57814 character moves the cursor back to the beginning of this empty line,
57817 For the EOF case though, this is not what happens. Instead, the
57818 escape sequence to leave bracketed paste mode is emitted on the same
57819 line as the prompt. The final '\r' moves the cursor back to the start
57820 of the prompt line. This leaves us ready to override the prompt.
57822 It is worth noting, that this is not the intended behaviour of
57823 readline, in rl_deprep_terminal, readline should emit a '\n' character
57824 when EOF is seen. However, due to a bug in readline this does not
57825 happen (the _rl_eof_found flag is never set). This is the first
57826 readline bug that effects GDB.
57828 GDB prints the 'quit' message from command_line_handler (in
57829 event-top.c), this function is called (indirectly) from readline to
57830 process the complete command line, but also in the EOF case (in which
57831 case the command line is set to nullptr). As this is part of the
57832 callback to process a complete command, this is called after readline
57833 has disabled bracketed paste mode (by calling rl_deprep_terminal).
57835 And so, when bracketed paste mode is in use, rl_deprep_terminal leaves
57836 the cursor at the start of the prompt line (in the EOF case), and
57837 command_line_handler then prints 'quit', which overwrites the prompt.
57839 The solution to this problem is to print the 'quit' message earlier,
57840 before rl_deprep_terminal is called. This is easy to do by using the
57841 rl_deprep_term_function hook. It is this hook that usually calls
57842 rl_deprep_terminal, however, if we replace this with a new function,
57843 we can print the 'quit' string, and then call rl_deprep_terminal
57844 ourselves. This allows the 'quit' to be printed before
57845 rl_deprep_terminal is called.
57847 The problem here is that there is no way in rl_deprep_terminal to know
57848 if readline is processing EOF or not, and as a result, we don't know
57849 when we should print 'quit'. This is the second readline bug that
57852 Both of these readline issues are discussed in this thread:
57854 https://lists.gnu.org/archive/html/bug-readline/2022-02/msg00021.html
57856 The result of that thread was that readline was patched to address
57857 both of these issues.
57859 Now it should be easy to backport the readline fix to GDB's in tree
57860 copy of readline, and then change GDB to make use of these fixes to
57861 correctly print the 'quit' string.
57863 However, we are just about to branch GDB 12, and there is concern from
57864 some that changing readline this close to a new release is a risky
57865 idea, see this thread:
57867 https://sourceware.org/pipermail/gdb-patches/2022-March/186391.html
57869 So, this commit doesn't change readline at all. Instead, this commit
57870 is the smallest possible GDB change in order to avoid the prompt
57873 In this commit I change GDB to print the 'quit' string on the line
57874 after the prompt, but only when bracketed paste mode is on. This
57875 avoids the overwriting issue, the user sees this:
57879 ... gdb terminates, and returns to the shell ...
57881 This isn't ideal, but is better than the existing behaviour. After
57882 GDB 12 has branched, we can backport the readline fix, and apply a
57885 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28833
57887 2022-03-16 Fangrui Song <i@maskray.me>
57889 objcopy --weaken-symbol: apply to STB_GNU_UNIQUE symbols
57891 * objcopy.c (filter_symbols): Apply weaken to STB_GNU_UNIQUE symbols
57892 * NEWS: Mention feature.
57893 * testsuite/binutils-all/objcopy.exp (objcopy_test_symbol_manipulation): New test.
57894 * testsuite/binutils-all/weaken-gnu-unique.s: New.
57896 2022-03-16 Tom Tromey <tromey@adacore.com>
57898 Reimplement array concatenation for Ada and D
57899 This started as a patch to implement string concatenation for Ada.
57900 However, while working on this, I looked at how this code could
57901 possibly be called. It turns out there are only two users of
57902 concat_operation: Ada and D. So, in addition to implementing this for
57903 Ada, this patch rewrites value_concat, removing the odd "concatenate
57904 or repeat" semantics, which were completely unused. As Ada and D both
57905 seem to represent strings using TYPE_CODE_ARRAY, this removes the
57906 TYPE_CODE_STRING code from there as well.
57908 Remove eval_op_concat
57909 eval_op_concat has code to search for an operator overload of
57910 BINOP_CONCAT. However, the operator overloading code is specific to
57911 C++, which does not have this operator. And,
57912 binop_types_user_defined_p rejects this case right at the start, and
57913 value_x_binop does not handle this case. I think this code has been
57914 dead for a very long time. This patch removes it and hoists the
57915 remaining call into concatenation::evaluate, removing eval_op_concat
57918 2022-03-16 Tom Tromey <tromey@adacore.com>
57920 Ada support for wide strings
57921 This adds some basic support for Wide_String and Wide_Wide_String to
57922 the Ada expression evaluator. In particular, a string literal may be
57923 converted to a wide or wide-wide string depending on context.
57925 The patch updates an existing test case. Note that another test,
57926 namely something like:
57928 ptype Wide_Wide_String'("literal")
57930 ... would be nice to add, but when tested against a distro GNAT, this
57931 did not work (probably due to lack of debuginfo); so, I haven't
57934 2022-03-16 Tom Tromey <tromey@adacore.com>
57936 Remove eval_op_string
57937 eval_op_string is only used in a single place -- the implementation of
57938 string_operation. This patch turns it into the
57939 string_operation::evaluate method, removing a bit of extraneous code.
57941 2022-03-16 Carl Love <cel@us.ibm.com>
57943 Powerpc fix for gdb.base/ending-run.exp
57944 The last two tests in gdb.base/ending-run.exp case fail on Powerpc when the
57945 system does not have the needed glibc debug-info files loaded. In this
57946 case, gdb is not able to determine where execution stopped. This behavior
57947 looks as follows for the test case:
57949 The next to the last test does a next command when the program is stopped
57950 at the closing bracket for main. The message printed is:
57952 0x00007ffff7d01524 in ?? () from /lib/powerpc64le-linux-gnu/libc.so.6
57954 which fails to match any of the test_multiple options.
57956 The test then does another next command. On Powerpc, the
57957 message printed it:
57959 Cannot find bounds of current function
57961 The test fails as the output does not match any of the options for the
57964 I checked the behavior on Powerpc to see if this is typical.
57965 I ran gdb on the following simple program as shown below.
57971 printf("Hello, world!\n");
57976 <snip the gdb start info>
57978 Type "apropos word" to search for commands related to "word"...
57979 Reading symbols from ./hello_world...
57980 (No debugging symbols found in ./hello_world)
57982 Breakpoint 1 at 0x818
57985 Starting program: /home/carll/hello_world
57986 [Thread debugging using libthread_db enabled]
57987 Using host libthread_db library "/lib/powerpc64le-linux-gnu/libthread_db.so.1".
57989 Breakpoint 1, 0x0000000100000818 in main ()
57991 Single stepping until exit from function main,
57992 which has no line number information.
57994 0x00007ffff7d01524 in ?? () from /lib/powerpc64le-linux-gnu/libc.so.6
57996 Cannot find bounds of current function
57998 So it would seem that the messages seen from the test case are
57999 "normal" output for Powerpc when the debug-info is not available.
58001 The following patch adds the output from Powerpc as an option
58002 to the gdb_test_multiple statement, identifying the output as the expected
58003 output on Powerpc without the needed debug-info files installed.
58005 The patch has been tested on a Power 10 system and an Intel
58006 64-bit system. No additional regression failures were seen on
58009 2022-03-16 Martin Storsj? <martin@martin.st>
58011 dlltool: Use the output name as basis for deterministic temp prefixes
58013 * dlltool.c (main): use imp_name rather than dll_name when
58014 generating a temporary file name.
58016 2022-03-16 Jan Vrany <jan.vrany@labware.com>
58017 Andrew Burgess <aburgess@redhat.com>
58019 gdb/mi: consistently notify user when GDB/MI client uses -thread-select
58020 GDB notifies users about user selected thread changes somewhat
58021 inconsistently as mentioned on gdb-patches mailing list here:
58023 https://sourceware.org/pipermail/gdb-patches/2022-February/185989.html
58025 Consider GDB debugging a multi-threaded inferior with both CLI and GDB/MI
58026 interfaces connected to separate terminals.
58028 Assuming inferior is stopped and thread 1 is selected, when a thread
58029 2 is selected using '-thread-select 2' command on GDB/MI terminal:
58032 ^done,new-thread-id="2",frame={level="0",addr="0x00005555555551cd",func="child_sub_function",args=[],file="/home/jv/Projects/gdb/users_jv_patches/gdb/testsuite/gdb.mi/user-selected-context-sync.c",fullname="/home/uuu/gdb/gdb/testsuite/gdb.mi/user-selected-context-sync.c",line="30",arch="i386:x86-64"}
58035 and on CLI terminal we get the notification (as expected):
58037 [Switching to thread 2 (Thread 0x7ffff7daa640 (LWP 389659))]
58038 #0 child_sub_function () at /home/uuu/gdb/gdb/testsuite/gdb.mi/user-selected-context-sync.c:30
58039 30 volatile int dummy = 0;
58041 However, now that thread 2 is selected, if thread 1 is selected
58042 using 'thread-select --thread 1 1' command on GDB/MI terminal
58045 -thread-select --thread 1 1
58046 ^done,new-thread-id="1",frame={level="0",addr="0x0000555555555294",func="main",args=[],file="/home/jv/Projects/gdb/users_jv_patches/gdb/testsuite/gdb.mi/user-selected-context-sync.c",fullname="/home/jv/Projects/gdb/users_jv_patches/gdb/testsuite/gdb.mi/user-selected-context-sync.c",line="66",arch="i386:x86-64"}
58049 but no notification is printed on CLI terminal, despite the fact
58050 that user selected thread has changed.
58052 The problem is that when `-thread-select --thread 1 1` is executed
58053 then thread is switched to thread 1 before mi_cmd_thread_select () is
58054 called, therefore the condition "inferior_ptid != previous_ptid"
58055 there does not hold.
58057 To address this problem, we have to move notification logic up to
58058 mi_cmd_execute () where --thread option is processed and notify
58059 user selected contents observers there if context changes.
58061 However, this in itself breaks GDB/MI because it would cause context
58062 notification to be sent on MI channel. This is because by the time
58063 we notify, MI notification suppression is already restored (done in
58064 mi_command::invoke(). Therefore we had to lift notification suppression
58065 logic also up to mi_cmd_execute (). This change in made distinction
58066 between mi_command::invoke() and mi_command::do_invoke() unnecessary
58067 as all mi_command::invoke() did (after the change) was to call
58068 do_invoke(). So this patches removes do_invoke() and moves the command
58069 execution logic directly to invoke().
58071 With this change, all gdb.mi tests pass, tested on x86_64-linux.
58073 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=20631
58075 2022-03-16 H.J. Lu <hjl.tools@gmail.com>
58077 gprofng: Use symver attribute if available
58078 Use symver attribute if available, instead of asm statement, to support
58082 * libcollector/dispatcher.c (timer_create@@GLIBC_2.3.3): Use
58084 (timer_create@GLIBC_2.2): Likewise.
58085 (timer_create@GLIBC_2.2.5): Likewise.
58086 (pthread_create@@GLIBC_2.1): Likewise.
58087 (pthread_create@GLIBC_2.0): Likewise.
58088 * libcollector/iotrace.c (open64@@GLIBC_2.2): Likewise.
58089 (open64@GLIBC_2.1): Likewise.
58090 (fopen@@GLIBC_2.1): Likewise.
58091 (fopen@GLIBC_2.0): Likewise.
58092 (fclose@@GLIBC_2.1): Likewise.
58093 (fclose@GLIBC_2.0): Likewise.
58094 (fdopen@@GLIBC_2.1): Likewise.
58095 (fdopen@GLIBC_2.0): Likewise.
58096 (pread@@GLIBC_2.2): Likewise.
58097 (pread@GLIBC_2.1): Likewise.
58098 (pwrite@@GLIBC_2.2): Likewise.
58099 (pwrite@GLIBC_2.1): Likewise.
58100 (pwrite64@@GLIBC_2.2): Likewise.
58101 (pwrite64@GLIBC_2.1): Likewise.
58102 (fgetpos@@GLIBC_2.2): Likewise.
58103 (fgetpos@GLIBC_2.0): Likewise.
58104 (fgetpos64@@GLIBC_2.2): Likewise.
58105 (fgetpos64@GLIBC_2.1): Likewise.
58106 (fsetpos@@GLIBC_2.2): Likewise.
58107 (fsetpos@GLIBC_2.0): Likewise.
58108 (fsetpos64@@GLIBC_2.2): Likewise.
58109 (fsetpos64@GLIBC_2.1): Likewise.
58110 * libcollector/linetrace.c (posix_spawn@@GLIBC_2.15): Likewise.
58111 (posix_spawn@GLIBC_2.2): Likewise.
58112 (posix_spawn@GLIBC_2.2.5): Likewise.
58113 (posix_spawnp@@GLIBC_2.15): Likewise.
58114 (posix_spawnp@GLIBC_2.2): Likewise.
58115 (posix_spawnp@GLIBC_2.2.5): Likewise.
58116 (popen@@GLIBC_2.1): Likewise.
58117 (popen@GLIBC_2.0): Likewise.
58118 (_popen@@GLIBC_2.1): Likewise.
58119 (_popen@GLIBC_2.0): Likewise.
58120 * libcollector/mmaptrace.c (dlopen@@GLIBC_2.1): Likewise.
58121 (dlopen@GLIBC_2.0): Likewise.
58122 * libcollector/synctrace.c (pthread_cond_wait@@GLIBC_2.3.2):
58124 (pthread_cond_wait@GLIBC_2.0): Likewise.
58125 (pthread_cond_wait@GLIBC_2.2.5): Likewise.
58126 (pthread_cond_wait@GLIBC_2.2): Likewise.
58127 (pthread_cond_timedwait@@GLIBC_2.3.2): Likewise.
58128 (pthread_cond_timedwait@GLIBC_2.0): Likewise.
58129 (pthread_cond_timedwait@GLIBC_2.2.5): Likewise.
58130 (pthread_cond_timedwait@GLIBC_2.2): Likewise.
58131 (sem_wait@@GLIBC_2.1): Likewise.
58132 (sem_wait@GLIBC_2.0): Likewise.
58133 * src/collector_module.h (SYMVER_ATTRIBUTE): New.
58135 2022-03-16 H.J. Lu <hjl.tools@gmail.com>
58137 gprofng: Don't hardcode -Wno-format-truncation/-Wno-switch
58138 Use -Wno-format-truncation and -Wno-switch only if they are supported.
58141 * configure.ac (GPROFNG_NO_FORMAT_TRUNCATION_CFLAGS): New
58142 AC_SUBST for -Wno-format-truncation.
58143 (GPROFNG_NO_SWITCH_CFLAGS): New AC_SUBST for -Wno-switch.
58144 * Makefile.in: Regenerate.
58145 * configure: Likewise.
58146 * src/Makefile.am (AM_CFLAGS): Replace -Wno-format-truncation
58147 and -Wno-switch with GPROFNG_NO_FORMAT_TRUNCATION_CFLAGS and
58148 GPROFNG_NO_SWITCH_CFLAGS.
58149 * src/Makefile.in: Regenerate.
58151 2022-03-16 H.J. Lu <hjl.tools@gmail.com>
58153 gprofng: Don't hardcode -Wno-nonnull-compare
58154 Use -Wno-nonnull-compare only if it is supported.
58157 * libcollector/Makefile.am (AM_CFLAGS): Replace
58158 -Wno-nonnull-compare with GPROFNG_NO_NONNULL_COMPARE_CFLAGS.
58159 * libcollector/configure.ac (GPROFNG_NO_NONNULL_COMPARE_CFLAGS):
58160 New AC_SUBST for -Wno-nonnull-compare.
58161 * libcollector/Makefile.in: Regenerate.
58162 * libcollector/aclocal.m4: Likewise.
58163 * libcollector/configure: Likewise.
58165 2022-03-16 H.J. Lu <hjl.tools@gmail.com>
58167 gprofng: Define ATTRIBUTE_FALLTHROUGH
58168 Define ATTRIBUTE_FALLTHROUGH to __attribute__ ((fallthrough)) only for
58172 * common/gp-defs.h (ATTRIBUTE_FALLTHROUGH): New.
58173 * src/gp-collect-app.cc (collect::check_args): Replace
58174 /* FALLTHROUGH */ with ATTRIBUTE_FALLTHROUGH.
58176 2022-03-16 Simon Marchi <simon.marchi@efficios.com>
58178 binutils/readelf: handle AMDGPU relocation types
58179 Make readelf recognize AMDGPU relocation types, as documented here:
58181 https://llvm.org/docs/AMDGPUUsage.html#amdgpu-relocation-records
58183 The user-visible change looks like:
58185 -000000000004 000400000001 unrecognized: 1 0000000000000000 SCRATCH_RSRC_DWORD0
58186 -00000000000c 000500000001 unrecognized: 1 0000000000000000 SCRATCH_RSRC_DWORD1
58187 -000000000014 000600000007 unrecognized: 7 0000000000000000 global_var0
58188 -00000000001c 000700000008 unrecognized: 8 0000000000000000 global_var1
58189 -000000000024 000800000009 unrecognized: 9 0000000000000000 global_var2
58190 -00000000002c 00090000000a unrecognized: a 0000000000000000 global_var3
58191 -000000000034 000a0000000b unrecognized: b 0000000000000000 global_var4
58192 +000000000004 000400000001 R_AMDGPU_ABS32_LO 0000000000000000 SCRATCH_RSRC_DWORD0
58193 +00000000000c 000500000001 R_AMDGPU_ABS32_LO 0000000000000000 SCRATCH_RSRC_DWORD1
58194 +000000000014 000600000007 R_AMDGPU_GOTPCREL 0000000000000000 global_var0
58195 +00000000001c 000700000008 R_AMDGPU_GOTPCREL 0000000000000000 global_var1
58196 +000000000024 000800000009 R_AMDGPU_GOTPCREL 0000000000000000 global_var2
58197 +00000000002c 00090000000a R_AMDGPU_REL32_LO 0000000000000000 global_var3
58198 +000000000034 000a0000000b R_AMDGPU_REL32_HI 0000000000000000 global_var4
58200 binutils/ChangeLog:
58202 * readelf.c (dump_relocations): Handle EM_AMDGPU.
58206 * elf/amdgpu.h: Add relocation values.
58208 Change-Id: I2ed4589f4cd37ea11ad2e0cb38d4b682271e1334
58210 2022-03-16 Simon Marchi <simon.marchi@efficios.com>
58212 binutils/readelf: build against msgpack, dump NT_AMDGPU_METADATA note contents
58213 The AMDGPU HSA OS ABI (code object v3 and above) defines the
58214 NT_AMDGPU_METADATA ELF note [1]. The content is a msgpack object
58215 describing, among other things, the kernels present in the code object
58216 and how to call them.
58218 I think it would be useful for readelf to be able to display the content
58219 of those notes. msgpack is a structured format, a bit like JSON, except
58220 not text-based. It is therefore possible to dump the contents in
58221 human-readable form without knowledge of the specific layout of the
58224 Add configury to binutils to optionally check for the msgpack C library
58225 [2]. Add There is a new --with{,out}-msgpack configure flag, and the actual
58226 library lookup is done using pkg-config.
58228 If msgpack support is enabled, dumping a NT_AMDGPU_METADATA note looks
58231 $ readelf --notes amdgpu-code-object
58232 Displaying notes found in: .note
58233 Owner Data size Description
58234 AMDGPU 0x0000040d NT_AMDGPU_METADATA (code object metadata)
58236 "amdhsa.kernels": [
58240 ".address_space": "global",
58241 ".name": "out.coerce",
58244 ".value_kind": "global_buffer",
58248 If msgpack support is disabled, dump the contents as hex, as is done
58249 with notes that are not handled in a special way. This allows one to
58250 decode the contents manually (maybe using a command-line msgpack
58251 decoder) if really needed.
58253 [1] https://llvm.org/docs/AMDGPUUsage.html#code-object-metadata
58254 [2] https://github.com/msgpack/msgpack-c/tree/c_master
58256 binutils/ChangeLog:
58258 * Makefile.am (readelf_CFLAGS): New.
58259 (readelf_LDADD): Add MSGPACK_LIBS.
58260 * Makefile.in: Re-generate.
58261 * config.in: Re-generate.
58262 * configure: Re-generate.
58263 * configure.ac: Add --with-msgpack flag and check for msgpack
58265 * readelf.c: Include msgpack.h if HAVE_MSGPACK.
58266 (print_note_contents_hex): New.
58267 (print_indents): New.
58268 (dump_msgpack_obj): New.
58269 (dump_msgpack): New.
58270 (print_amdgpu_note): New.
58271 (process_note): Handle NT_AMDGPU_METADATA note contents.
58272 Use print_note_contents_hex.
58274 Change-Id: Ia60a654e620bc32dfdb1bccd845594e2af328b84
58276 2022-03-16 Simon Marchi <simon.marchi@efficios.com>
58278 binutils/readelf: handle NT_AMDGPU_METADATA note name
58279 Handle the NT_AMDGPU_METADATA note, which is described here:
58281 https://llvm.org/docs/AMDGPUUsage.html#code-object-v3-note-records
58283 As of this patch, just print out the name, not the contents, which is in
58284 the msgpack format.
58286 binutils/ChangeLog:
58288 * readelf.c (get_amdgpu_elf_note_type): New.
58289 (process_note): Handle "AMDGPU" notes.
58293 * elf/amdgcn.h (NT_AMDGPU_METADATA): New.
58295 Change-Id: Id2dba2e2aeaa55ef7464fb35aee9c7d5f96ddb23
58297 2022-03-16 Simon Marchi <simon.marchi@efficios.com>
58299 binutils/readelf: decode AMDGPU-specific e_flags
58300 Decode and print the AMDGPU-specific fields of e_flags, as documented
58303 https://llvm.org/docs/AMDGPUUsage.html#header
58307 - The specific GPU model
58308 - Whether the xnack and sramecc features are enabled
58310 The result looks like:
58313 + Flags: 0x52f, gfx906, xnack any, sramecc any
58315 The flags for the "HSA" OS ABI are properly versioned and documented on
58316 that page. But the NONE, PAL and MESA3D OS ABIs are not well documented
58317 nor versioned. Taking a peek at the LLVM source code, we see that they
58318 encode their flags the same way as HSA v3. For example, for PAL:
58320 https://github.com/llvm/llvm-project/blob/c8b614cd74a92d85936aed5ac7c642af75ffdc29/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp#L601
58322 So for those other OS ABIs, we read them the same as HSA v3.
58324 binutils/ChangeLog:
58326 * readelf.c: Include elf/amdgcn.h.
58327 (decode_AMDGPU_machine_flags): New.
58328 (get_machine_flags): Handle flags for EM_AMDGPU machine type.
58332 * elf/amdgcn.h: Add EF_AMDGPU_MACH_AMDGCN_* and
58333 EF_AMDGPU_FEATURE_* defines.
58335 Change-Id: Ib5b94df7cae0719a22cf4e4fd0629330e9485c12
58337 2022-03-16 Simon Marchi <simon.marchi@efficios.com>
58339 binutils/readelf: handle AMDGPU OS ABIs
58340 When the machine is EM_AMDGPU, handle the various OS ABIs described
58343 https://llvm.org/docs/AMDGPUUsage.html#header
58345 For a binary with the HSA OS ABI, the change looks like:
58347 - OS/ABI: <unknown: 40>
58350 binutils/ChangeLog:
58352 * readelf.c (get_osabi_name): Handle EM_AMDGPU OS ABIs.
58356 * elf/common.h (ELFOSABI_AMDGPU_PAL, ELFOSABI_AMDGPU_MESA3D):
58359 Change-Id: I383590c390f7dc2fe0f902f50038735626d71863
58361 2022-03-16 Simon Marchi <simon.marchi@efficios.com>
58363 opcodes: handle bfd_amdgcn_arch in configure script
58364 There isn't an actual opcodes implementation for the AMDGCN arch (yet),
58365 this is just the bare minimum to get
58367 $ ./configure --target=amdgcn-hsa-amdhsa --disable-gas
58368 $ make all-binutils
58370 working later in this series.
58374 * configure.ac: Handle bfd_amdgcn_arch.
58375 * configure: Re-generate.
58377 Change-Id: Ib7d7c5533a803ed8b2a293e9275f667ed781ce79
58379 2022-03-16 Simon Marchi <simon.marchi@efficios.com>
58381 bfd: add AMDGCN architecture
58382 Add support for the AMDGCN architecture to BFD.
58384 This is the bare minimum to get
58386 $ ./configure --target=amdgcn-hsa-amdhsa --disable-gas
58387 $ make all-binutils
58389 working later in this series.
58391 The specific AMDGCN models added here are a bit arbitrary, based on
58392 what we intend to initially support in GDB. This list will need to be
58393 updated in the future anyway. The complete up-to-date list of existing
58394 AMDGPU models can be found here:
58396 https://llvm.org/docs/AMDGPUUsage.html#processors
58398 The ELF format for this architecture is documented here:
58400 https://llvm.org/docs/AMDGPUUsage.html#elf-code-object
58402 The flags for the "HSA" OS ABI are properly versioned and documented on
58403 that page. But the NONE, PAL and MESA3D OS ABIs are not well documented
58404 nor versioned. Taking a peek at the LLVM source code, we see that they
58405 encode their flags the same way as HSA v3. For example, for PAL:
58407 https://github.com/llvm/llvm-project/blob/c8b614cd74a92d85936aed5ac7c642af75ffdc29/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp#L601
58409 So at least, we know that all AMDGPU objects (of which AMDGCN objects
58410 are a subset of) at the time of writing encode the specific GPU model in
58411 the EF_AMDGPU_MACH field of e_flags.
58415 * Makefile.am (ALL_MACHINES, ALL_MACHINES_CFILES):
58417 (BFD64_BACKENDS): Add elf64-amdgcn.lo.
58418 (BFD64_BACKENDS_CFILES): Add elf64-amdgcn.c.
58419 * Makefile.in: Re-generate.
58420 * cpu-amdgcn.c: New.
58421 * elf64-amdgcn.c: New.
58422 * archures.c (bfd_architecture): Add bfd_arch_amdgcn and related
58424 (bfd_amdgcn_arch): New.
58425 (bfd_archures_list): Add bfd_amdgcn_arch.
58426 * bfd-in2.h: Re-generate.
58427 * config.bfd: Handle amdgcn* target.
58428 * configure.ac: Handle amdgcn_elf64_le_vec.
58429 * configure: Re-generate.
58430 * elf-bfd.h (elf_target_id): Add AMDGCN_ELF_DATA.
58431 * targets.c (amdgcn_elf64_le_vec): New.
58432 (_bfd_target_vector): Add amdgcn_elf64_le_vec.
58436 * elf/amdgpu.h: New.
58437 * elf/common.h (ELFOSABI_AMDGPU_HSA): Add.
58439 Change-Id: I969f7b14960797e88891c308749a6e341eece5b2
58441 2022-03-16 Nick Clifton <nickc@redhat.com>
58443 Updated Serbian (for binutils/) and Russian (for gprof/) translations
58445 2022-03-16 Pedro Alves <pedro@palves.net>
58447 Make gdb.fortran/{array-slices,lbound-ubound} work against gdbserver
58448 gdb.fortran/array-slices.exp and gdb.fortran/lbound-ubound.exp were
58449 recently disabled unless testing with the native target, because they
58450 rely on inferior I/O. However, when testing against gdbserver using
58451 the native-gdbserver/native-extended-gdbserver boards, we do have
58452 access to inferior I/O.
58454 The right way to check whether the board can do I/O, is via checking
58455 the gdb,noinferiorio board variable. Switch to using that.
58457 And then, tweak the testcases to expect output to appear in
58458 inferior_spawn_id, instead of gdb_spawn_id. When testing against the
58459 native target, inferior_spawn_id is the same as gdb_spawn_id. When
58460 testing against gdbserver, it maps to gdbserver_spawn_id.
58462 This exposed a buglet in gdb.fortran/array-slices.f90's show_1d
58463 subroutine -- it was missing printing newline at the end of the
58464 "Expected GDB Output" text, leading to a test timeout. All other
58465 subroutines end with advance=yes, except this one. Fix it by using
58466 advance=yes here too.
58468 Change-Id: I4640729f334431cfc24b0917e7d3977b677c6ca5
58470 2022-03-16 GDB Administrator <gdbadmin@sourceware.org>
58472 Automatic date update in version.in
58474 2022-03-15 Alan Modra <amodra@gmail.com>
58476 Delete PowerPC macro insn support
58477 Let's hope this stays dead, but it's here as a patch separate from
58478 those that removed use of powerpc_macros just in case it needs to be
58482 * opcode/ppc.h (struct powerpc_macro): Delete declaration.
58483 (powerpc_macros, powerpc_num_macros): Likewise..
58485 * ppc-opc.c (powerpc_macros, powerpc_num_macros): Delete.
58487 * config/tc-ppc.c (ppc_macro): Delete function.
58488 (ppc_macro_hash): Delete.
58489 (ppc_setup_opcodes, md_assemble): Delete macro support.
58491 2022-03-15 Alan Modra <amodra@gmail.com>
58493 PowerPC SPE/SPE2 aliases in powerpc_macros
58494 * ppc-opc.c (powerpc_macros): Move "evsadd", "evssub", "evsabs",
58495 "evsnabs", "evsneg", "evsmul", "evsdiv", "evscmpgt", "evsgmplt",
58496 "evsgmpeq", "evscfui", "evscfsi", "evscfuf", "evscfsf", "evsctui",
58497 "evsctsi", "evsctuf", "evsctsf", "evsctuiz", "evsctsiz",
58498 "evststgt", "evststlt", "evststeq"..
58499 (powerpc_opcodes): ..to here.
58500 (powerpc_macros): Move "evdotphsssi", "evdotphsssia", "evdotpwsssi",
58501 and "evdotpwsssia"..
58502 (spe2_opcodes): ..to here.
58504 2022-03-15 Alan Modra <amodra@gmail.com>
58506 PowerPC VLE extended instructions in powerpc_macros
58507 This moves VLE insn out of the macro table. "e_slwi" and "e_srwi"
58508 already exist in vle_opcodes as distinct instructions rather than
58509 encodings of e_rlwinm.
58512 * ppc-opc.c (vle_opcodes): Typo fix e_rlwinm operand.
58513 Add "e_inslwi", "e_insrwi", "e_rotlwi", "e_rotrwi", "e_clrlwi",
58514 "e_clrrwi", "e_extlwi", "e_extrwi", and "e_clrlslwi".
58515 (powerpc_macros): Delete same. Delete "e_slwi" and "e_srwi" too.
58517 * testsuite/gas/ppc/vle-simple-5.d: Update.
58519 2022-03-15 Alan Modra <amodra@gmail.com>
58521 PowerPC32 extended instructions in powerpc_macros
58522 As for PowerPC64, move instructions to the main opcode table.
58525 * ppc-opc.c (insert_crwn, extract_crwn, insert_elwn, extract_elwn),
58526 (insert_erwn, extract_erwn, insert_erwb, extract_erwb),
58527 (insert_cslwn, extract_cslwb, insert_ilwb, extract_ilwn),
58528 (insert_irwb, extract_irwn, insert_rrwn, extract_rrwn),
58529 (insert_slwn, extract_slwn, insert_srwn, extract_srwn): New functions.
58530 (CRWn, ELWn, ERWn, ERWb, CSLWb, CSLWn, ILWn, ILWb, IRWn, IRWb),
58531 (RRWn, SLWn, SRWn): Define and add powerpc_operands entries.
58532 (MMB_MASK, MME_MASK, MSHMB_MASK): Define.
58533 (powerpc_opcodes): Add "inslwi", "insrwi", "rotrwi", "clrrwi",
58534 "slwi", "srwi", "extlwi", "extrwi", "sli", "sri" and corresponding
58535 record (ie. dot suffix) forms.
58536 (powerpc_macros): Delete same.
58538 * testsuite/gas/ppc/476.d: Update.
58539 * testsuite/gas/ppc/simpshft.d: Update.
58541 2022-03-15 Alan Modra <amodra@gmail.com>
58543 PowerPC64 extended instructions in powerpc_macros
58544 The extended instructions implemented in powerpc_macros aren't used by
58545 the disassembler. That means instructions like "sldi r3,r3,2" appear
58546 in disassembly as "rldicr r3,r3,2,61", which is annoying since many
58547 other extended instructions are shown.
58549 Note that some of the instructions moved out of the macro table to the
58550 opcode table won't appear in disassembly, because they are aliases
58551 rather than a subset of the underlying raw instruction. If enabled,
58552 rotrdi, extrdi, extldi, clrlsldi, and insrdi would replace all
58553 occurrences of rotldi, rldicl, rldicr, rldic and rldimi. (Or many
58554 occurrences in the case of clrlsldi if n <= b was added to the extract
58557 The patch also fixes a small bug in opcode sanity checking.
58560 * opcode/ppc.h (PPC_OPSHIFT_SH6): Define.
58562 * ppc-opc.c (insert_erdn, extract_erdn, insert_eldn, extract_eldn),
58563 (insert_crdn, extract_crdn, insert_rrdn, extract_rrdn),
58564 (insert_sldn, extract_sldn, insert_srdn, extract_srdn),
58565 (insert_erdb, extract_erdb, insert_csldn, extract_csldb),
58566 (insert_irdb, extract_irdn): New functions.
58567 (ELDn, ERDn, ERDn, RRDn, SRDn, ERDb, CSLDn, CSLDb, IRDn, IRDb):
58568 Define and add associated powerpc_operands entries.
58569 (powerpc_opcodes): Add "rotrdi", "srdi", "extrdi", "clrrdi",
58570 "sldi", "extldi", "clrlsldi", "insrdi" and corresponding record
58571 (ie. dot suffix) forms.
58572 (powerpc_macros): Delete same from here.
58574 * config/tc-ppc.c (insn_validate): Don't modify value passed
58575 to operand->insert for PPC_OPERAND_PLUS1 when calculating mask.
58576 Handle PPC_OPSHIFT_SH6.
58577 * testsuite/gas/ppc/prefix-reloc.d: Update.
58578 * testsuite/gas/ppc/simpshft.d: Update.
58580 * testsuite/ld-powerpc/elfv2so.d: Update.
58581 * testsuite/ld-powerpc/notoc.d: Update.
58582 * testsuite/ld-powerpc/notoc3.d: Update.
58583 * testsuite/ld-powerpc/tlsdesc2.d: Update.
58584 * testsuite/ld-powerpc/tlsget.d: Update.
58585 * testsuite/ld-powerpc/tlsget2.d: Update.
58586 * testsuite/ld-powerpc/tlsopt5.d: Update.
58587 * testsuite/ld-powerpc/tlsopt6.d: Update.
58589 2022-03-15 Tom Tromey <tom@tromey.com>
58591 Do not capture updated 'pc' in add_local_symbols
58592 Simon pointed out that commit 13835d88 ("Use function view when
58593 iterating over block symbols") caused a regression. The bug is that
58594 the new closure captures 'pc' by reference, but later code updates
58595 this variable -- but the earlier code did not update the callback
58596 structure with the new value.
58598 This patch restores the old behavior by using a new varible name in an
58601 2022-03-15 Jose E. Marchesi <jose.marchesi@oracle.com>
58603 gprofng: avoid using `fallthrough' attributes
58604 gprofng didn't build with gcc 6.3 due to the usage of __attribute__
58605 ((fallthrough)). This patch uses /* FALLTHROUGH */ instead.
58607 2022-03-15 Jose E. Marchesi <jose.marchesi@oracle.com>
58609 * gprofng/src/gp-collect-app.cc (collect::check_args): Use
58610 fallthrough comment instead of attribute.
58612 2022-03-15 Tom Tromey <tromey@adacore.com>
58614 Fix bug in dwarf-mode.el
58615 I noticed that, occasionally, dwarf-mode would think that the objdump
58616 subprocess was still running after it had clearly exited. I managed
58617 to reliably reproduce this today and learned that a process sentinel
58618 is not guaranteed to be run with the current buffer set to the process
58619 buffer. This patch fixes the problem.
58621 I've bumped the version number of dwarf-mode.el to make it easier to
58622 install for users who already have an earlier one installed.
58624 I'm checking this in.
58626 2022-03-15 Tom Tromey <tromey@adacore.com>
58628 * dwarf-mode.el: Now 1.7.
58629 (dwarf--sentinel): Switch to the process buffer.
58631 2022-03-15 Andrew Burgess <aburgess@redhat.com>
58633 gdb/testsuite: rename a proc and fix a typo
58634 Rename a proc in gdb.mi/user-selected-context-sync.exp, I think the
58635 old name was most likely a typo. The old name
58636 match_re_or_ensure_not_output seems (to me) to imply we're in some way
58637 checking that the regexp was not output. But that's not what we are
58638 doing, we're checking either for the regexp, or for no output, hence
58639 the new name match_re_or_ensure_no_output.
58641 Additionally, I found a definite typo in one of the comments that I've
58644 I also updated some test names. These tests (probably due to copy &
58645 paste errors) has 'on MI' on their name, when they were actually
58646 checking CLI output. For these test I changed the name to use 'on
58649 There should be no change in what is tested after this commit.
58651 2022-03-15 Nick Clifton <nickc@redhat.com>
58653 gprofng: Add a configure test for clock_gettime and a use of the test in getthrtime.c
58655 2022-03-15 H.J. Lu <hjl.tools@gmail.com>
58657 gprofng: Don't generate gprofng.info in source
58658 Add info-in-builddir to AUTOMAKE_OPTIONS.
58661 * doc/Makefile.am (AUTOMAKE_OPTIONS): Add info-in-builddir.
58662 * doc/Makefile.in: Regenerate.
58664 2022-03-15 Tiezhu Yang <yangtiezhu@loongson.cn>
58666 gdb: LoongArch: fix failed testcases in gdb.base/align-c.exp
58667 When execute the following command on LoongArch:
58669 make check-gdb TESTS="gdb.base/align-c.exp"
58671 there exist some failed testcases:
58674 FAIL: gdb.base/align-c.exp: print _Alignof(struct align_pair_long_double_x_float)
58675 FAIL: gdb.base/align-c.exp: print _Alignof(struct align_pair_long_double_x_double)
58676 FAIL: gdb.base/align-c.exp: print _Alignof(struct align_pair_long_double_x_long_double)
58679 According to LoongArch ELF ABI specification [1], set the target data types
58680 of floating-point to fix the above failed testcases.
58682 [1] https://loongson.github.io/LoongArch-Documentation/LoongArch-ELF-ABI-EN.html
58684 2022-03-15 GDB Administrator <gdbadmin@sourceware.org>
58686 Automatic date update in version.in
58688 2022-03-14 Andrew Burgess <aburgess@redhat.com>
58690 gdb/python/mi: create MI commands using python
58691 This commit allows a user to create custom MI commands using Python
58692 similarly to what is possible for Python CLI commands.
58694 A new subclass of mi_command is defined for Python MI commands,
58695 mi_command_py. A new file, gdb/python/py-micmd.c contains the logic
58696 for Python MI commands.
58698 This commit is based on work linked too from this mailing list thread:
58700 https://sourceware.org/pipermail/gdb/2021-November/049774.html
58702 Which has also been previously posted to the mailing list here:
58704 https://sourceware.org/pipermail/gdb-patches/2019-May/158010.html
58706 And was recently reposted here:
58708 https://sourceware.org/pipermail/gdb-patches/2022-January/185190.html
58710 The version in this patch takes some core code from the previously
58711 posted patches, but also has some significant differences, especially
58712 after the feedback given here:
58714 https://sourceware.org/pipermail/gdb-patches/2022-February/185767.html
58716 A new MI command can be implemented in Python like this:
58718 class echo_args(gdb.MICommand):
58719 def invoke(self, args):
58720 return { 'args': args }
58722 echo_args("-echo-args")
58724 The 'args' parameter (to the invoke method) is a list
58725 containing (almost) all command line arguments passed to the MI
58726 command (--thread and --frame are handled before the Python code is
58727 called, and removed from the args list). This list can be empty if
58728 the MI command was passed no arguments.
58730 When used within gdb the above command produced output like this:
58734 ^done,args=["a","b","c"]
58737 The 'invoke' method of the new command must return a dictionary. The
58738 keys of this dictionary are then used as the field names in the mi
58739 command output (e.g. 'args' in the above).
58741 The values of the result returned by invoke can be dictionaries,
58742 lists, iterators, or an object that can be converted to a string.
58743 These are processed recursively to create the mi output. And so, this
58746 class new_command(gdb.MICommand):
58747 def invoke(self,args):
58748 return { 'result_one': { 'abc': 123, 'def': 'Hello' },
58749 'result_two': [ { 'a': 1, 'b': 2 },
58750 { 'c': 3, 'd': 4 } ] }
58752 Which produces output like:
58756 ^done,result_one={abc="123",def="Hello"},result_two=[{a="1",b="2"},{c="3",d="4"}]
58759 I have required that the fields names used in mi result output must
58760 match the regexp: "^[a-zA-Z][-_a-zA-Z0-9]*$" (without the quotes).
58761 This restriction was never written down anywhere before, but seems
58762 sensible to me, and we can always loosen this rule later if it proves
58763 to be a problem. Much harder to try and add a restriction later, once
58764 people are already using the API.
58766 What follows are some details about how this implementation differs
58767 from the original patch that was posted to the mailing list.
58769 In this patch, I have changed how the lifetime of the Python
58770 gdb.MICommand objects is managed. In the original patch, these object
58771 were kept alive by an owned reference within the mi_command_py object.
58772 As such, the Python object would not be deleted until the
58773 mi_command_py object itself was deleted.
58775 This caused a problem, the mi_command_py were held in the global mi
58776 command table (in mi/mi-cmds.c), which, as a global, was not cleared
58777 until program shutdown. By this point the Python interpreter has
58778 already been shutdown. Attempting to delete the mi_command_py object
58779 at this point was causing GDB to try and invoke Python code after
58780 finalising the Python interpreter, and we would crash.
58782 To work around this problem, the original patch added code in
58783 python/python.c that would search the mi command table, and delete the
58784 mi_command_py objects before the Python environment was finalised.
58786 In contrast, in this patch, I have added a new global dictionary to
58787 the gdb module, gdb._mi_commands. We already have several such global
58788 data stores related to pretty printers, and frame unwinders.
58790 The MICommand objects are placed into the new gdb.mi_commands
58791 dictionary, and it is this reference that keeps the objects alive.
58792 When GDB's Python interpreter is shut down gdb._mi_commands is deleted,
58793 and any MICommand objects within it are deleted at this point.
58795 This change avoids having to make the mi_cmd_table global, and walk
58796 over it from within GDB's python related code.
58798 This patch handles command redefinition entirely within GDB's python
58799 code, though this does impose one small restriction which is not
58800 present in the original code (detailed below), I don't think this is a
58801 big issue. However, the original patch relied on being able to
58802 finish executing the mi_command::do_invoke member function after the
58803 mi_command object had been deleted. Though continuing to execute a
58804 member function after an object is deleted is well defined, it is
58805 also (IMHO) risky, its too easy for someone to later add a use of the
58806 object without realising that the object might sometimes, have been
58807 deleted. The new patch avoids this issue.
58809 The one restriction that is added to avoid this, is that an MICommand
58810 object can't be reinitialised with a different command name, so:
58812 (gdb) python cmd = MyMICommand("-abc")
58813 (gdb) python cmd.__init__("-def")
58814 can't reinitialize object with a different command name
58816 This feels like a pretty weird edge case, and I'm happy to live with
58819 I have also changed how the memory is managed for the command name.
58820 In the most recently posted patch series, the command name is moved
58821 into a subclass of mi_command, the python mi_command_py, which
58822 inherits from mi_command is then free to use a smart pointer to manage
58823 the memory for the name.
58825 In this patch, I leave the mi_command class unchanged, and instead
58826 hold the memory for the name within the Python object, as the lifetime
58827 of the Python object always exceeds the c++ object stored in the
58828 mi_cmd_table. This adds a little more complexity in py-micmd.c, but
58829 leaves the mi_command class nice and simple.
58831 Next, this patch adds some extra functionality, there's a
58832 MICommand.name read-only attribute containing the name of the command,
58833 and a read-write MICommand.installed attribute that can be used to
58834 install (make the command available for use) and uninstall (remove the
58835 command from the mi_cmd_table so it can't be used) the command. This
58836 attribute will be automatically updated if a second command replaces
58837 an earlier command.
58839 This patch adds additional error handling, and makes more use the
58840 gdbpy_handle_exception function.
58842 Co-Authored-By: Jan Vrany <jan.vrany@labware.com>
58844 2022-03-14 Andrew Burgess <aburgess@redhat.com>
58846 gdb/gdbarch: compare some fields against 0 verify_gdbarch
58847 After the previous commit, which removes the predicate function
58848 gdbarch_register_type_p, I noticed that the gdbarch->register_type
58849 field was not checked at in the verify_gdbarch function.
58851 More than not being checked, the field wasn't mentioned at all.
58853 I find this strange, I would expect that every field would at least be
58854 mentioned - we already generate comments for some fields saying that
58855 this field is _not_ being checked, so the fact that this field isn't
58856 being checked looks (to me), like this field is somehow slipping
58857 through the cracks.
58859 The comment at the top of gdbarch-components.py tries to explain how
58860 the validation is done. I didn't understand this comment completely,
58861 but, I think this final sentence:
58863 "Otherwise, the check is done against 0 (really NULL for function
58864 pointers, but same idea)."
58866 Means that, if non of the other cases apply, then the field should be
58867 checked against 0, with 0 indicating that the field is invalid (was
58868 not set by the tdep code). However, this is clearly not being done.
58870 Looking in gdbarch.py at the code to generate verify_gdbarch we do
58871 find that there is a case that is not handled, the case where the
58872 'invalid' field is set true True, but non of the other cases apply.
58874 In this commit I propose two changes:
58876 1. Handle the case where the 'invalid' field of a property is set to
58877 True, this should perform a check for the field of gdbarch still
58878 being set to 0, and
58880 2. If the if/else series that generates verify_gdbarch doesn't handle
58881 a property then we should raise an exception. This means that if a
58882 property is added which isn't handled, we should no longer silently
58885 After doing this, I re-generated the gdbarch files and saw that the
58886 following gdbarch fields now had new validation checks:
58889 believe_pcc_promotion
58892 frame_red_zone_size
58893 displaced_step_restore_all_in_ptid
58894 solib_symbols_extension
58896 Looking at how these are currently set in the various -tdep.c files, I
58897 believe the only one of these that is required to be set for all
58898 architectures is the register_type field.
58900 And so, for all of the other fields, I've changed the property
58901 definition on gdbarch-components.py, setting the 'invalid' field to
58904 Now, after re-generation, the register_type field is checked against
58905 0, thus an architecture that doesn't set_gdbarch_register_type will
58906 now fail during validation. For all the other fields we skip the
58907 validation, in which case, it is find for an architecture to not set
58910 My expectation is that there should be no user visible changes after
58911 this commit. Certainly for all fields except register_type, all I've
58912 really done is cause some extra comments to be generated, so I think
58913 that's clearly fine.
58915 For the register_type field, my claim is that any architecture that
58916 didn't provide this would fail when creating its register cache, and I
58917 couldn't spot an architecture that doesn't provide this hook. As
58918 such, I think this change should be fine too.
58920 2022-03-14 Andrew Burgess <aburgess@redhat.com>
58922 gdb/gdbarch: remove the predicate function for gdbarch_register_type
58923 I don't believe that the gdbarch_register_type_p predicate is called
58924 anywhere in GDB, and the gdbarch_register_type function is called
58925 without checking the gdbarch_register_type_p predicate function
58926 everywhere it is used, for example in
58927 init_regcache_descr (regcache.c).
58929 My claim is that the gdbarch_register_type function is required for
58930 every architecture, and GDB will not work if this function is not
58933 And so, in this commit, I remove the 'predicate=True' from
58934 gdbarch-components.py for the 'register_type' field, and regenerate
58937 There should be no user visible changes after this commit.
58939 2022-03-14 Patrick Monnerat <patrick@monnerat.net>
58941 Replace deprecated_target_wait_hook by observers
58942 Commit b60cea7 (Make target_wait options use enum flags) broke
58943 deprecated_target_wait_hook usage: there's a commit comment telling
58944 this hook has not been converted.
58946 Rather than trying to mend it, this patch replaces the hook by two
58947 target_wait observers:
58949 target_pre_wait (ptid_t ptid)
58950 target_post_wait (ptid_t event_ptid)
58952 Upon target_wait entry, target_pre_wait is notified with the ptid
58953 passed to target_wait. Upon exit, target_post_wait is notified with
58954 the event ptid returned by target_wait. Should an exception occur,
58955 event_ptid is null_ptid.
58957 This change benefits to Insight (out-of-tree): there's no real use of the
58958 late hook in gdb itself.
58960 2022-03-14 Tom Tromey <tromey@adacore.com>
58962 Correctly print subrange types in generic_value_print
58963 I noticed that generic_value_print assumes that a subrange type is
58964 always a subrange of an integer type. However, this isn't necessarily
58965 the case. In Ada, for example, one has subranges of character and
58968 This code isn't often exercised, I think, because languages with real
58969 subrange types tend to implement their own printers. However, it
58970 still seemed worth fixing.
58972 2022-03-14 Luis Machado <luis.machado@arm.com>
58974 [aarch64/arm] Properly extract the return value returned in memory
58975 When running gdb.cp/non-trivial-retval.exp, the following shows up for
58976 both aarch64-linux and armhf-linux:
58978 Breakpoint 3, f1 (i1=23, i2=100) at src/gdb/testsuite/gdb.cp/non-trivial-retval.cc:35
58981 Run till exit from #0 f1 (i1=23, i2=100) at src/gdb/testsuite/gdb.cp/non-trivial-retval.cc:35
58982 main () at /src/gdb/testsuite/gdb.cp/non-trivial-retval.cc:163
58983 163 B b = f2 (i1, i2);
58984 Value returned is $6 = {a = -11952}
58987 The return value should be {a = 123} instead. This happens because the
58988 backends don't extract the return value from the correct location. GDB should
58989 fetch a pointer to the memory location from X8 for aarch64 and r0 for armhf.
58991 With the patch, gdb.cp/non-trivial-retval.exp has full passes on
58992 aarch64-linux and armhf-linux on Ubuntu 20.04/18.04.
58994 The problem only shows up with the "finish" command. The "call" command
58995 works correctly and displays the correct return value.
58997 This is also related to PR gdb/28681
58998 (https://sourceware.org/bugzilla/show_bug.cgi?id=28681) and fixes FAIL's in
58999 gdb.ada/mi_var_array.exp.
59001 A new testcase is provided, and it exercises GDB's ability to "finish" a
59002 function that returns a large struct (> 16 bytes) and display the
59003 contents of this struct correctly. This has always been incorrect for
59004 these backends, but no testcase exercised this particular scenario.
59006 2022-03-14 GDB Administrator <gdbadmin@sourceware.org>
59008 Automatic date update in version.in
59010 2022-03-13 Alan Modra <amodra@gmail.com>
59012 PR28959, obdump doesn't disassemble mftb instruction
59013 Without a -M cpu option given, powerpc objdump defaults currently to
59014 -Mpower10 but -Many is also given. Commit 1ff6a3b8e562 regressed
59015 -Many disassembly of instructions that are encoded differently
59016 depending on cpu, such as mftb which has pre- and post-power4
59020 * ppc-dis.c (lookup_powerpc): Revert 2021-05-28 change. Instead
59021 only look at deprecated PPC_OPCODE_RAW bit when -Many.
59023 2022-03-13 GDB Administrator <gdbadmin@sourceware.org>
59025 Automatic date update in version.in
59027 2022-03-12 Tom Tromey <tom@tromey.com>
59029 Relax regexp in gdb.rust/unsized.exp
59030 With nightly rustc, gdb.rust/unsized.exp fails:
59033 Structure has no component named operator*.
59035 rustc changed to emit a bit more debug info for unsized types.
59036 Because the original test is just to make sure that ptype of an
59037 unsized array looks right, this patch relaxes the regexp and changes
59038 the expression. I think this keeps the original test meaning, but
59039 also works with nightly. I also tested stable and 1.48.
59041 2022-03-12 GDB Administrator <gdbadmin@sourceware.org>
59043 Automatic date update in version.in
59045 2022-03-11 Tom Tromey <tromey@adacore.com>
59047 Avoid crash with cross-linux core file
59048 An internal test case creates a core file using gcore, then restarts
59049 gdb with that core. When run with a cross-linux gdb (in this case,
59050 x86-64 host with ppc64-linux target), the test fails:
59054 | warning: `/lib64/libc.so.6': Shared library architecture unknown is not compatible with target architecture powerpc:common64.
59055 | warning: Could not load shared library symbols for /lib64/ld64.so.1.
59056 | Do you need "set solib-search-path" or "set sysroot"?
59057 | ../../src/gdb/gdbarch.c:3388: internal-error: int gdbarch_elf_make_msymbol_special_p(gdbarch*): Assertion `gdbarch != NULL' failed.
59058 | A problem internal to GDB has been detected,
59059 | further debugging may prove unreliable.
59060 | Quit this debugging session? (y or n) y
59062 What's happening here is that the core file lists some shared
59063 libraries. These aren't available via the solib search path, and so
59064 gdb finds the local (x86-64) libraries. This is not ideal, but on the
59065 other hand, it is what was asked for -- while the test does set
59066 solib-search-path, it does not set the sysroot.
59068 But, because gdb isn't configured to handle these libraries, it
59071 It seems to me that it's better to avoid the crash by having
59072 solib_bfd_open fail in the case where a library is incompatible. That
59073 is what this patch does. Now it looks like:
59076 | Error while mapping shared library sections:
59077 | `/lib64/libc.so.6': Shared library architecture unknown is not compatible with target architecture powerpc:common64.
59079 ... and does not crash gdb.
59081 I don't have a good setup for testing this using dejagnu, so I don't
59082 know whether an existing gdb test covers this scenario.
59084 2022-03-11 Andrew Burgess <aburgess@redhat.com>
59086 gdb/testsuite: remove duplicates from gdb.base/stap-probe.exp
59087 Remove the duplicate test names from gdb.base/stap-probe.exp, this is
59088 done by actually passing a unique test name in a couple of
59089 places (rather than using the command as the test name), and in
59090 another couple of places, a test has a duplicate name due to a cut &
59091 paste error, which I've fixed.
59093 There's no change in what is actually being tested after this commit.
59095 2022-03-11 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
59097 gprofng: a new GNU profiler
59099 * Makefile.def: Add gprofng module.
59100 * configure.ac: Add --enable-gprofng option.
59101 * src-release.sh: Add gprofng.
59102 * Makefile.in: Regenerate.
59103 * configure: Regenerate.
59104 * gprofng: New directory.
59107 * MAINTAINERS: Add gprofng maintainer.
59108 * README-how-to-make-a-release: Add gprofng.
59111 * collectorAPI.h: New file.
59112 * libcollector.h: New file.
59113 * libfcollector.h: New file.
59115 2022-03-11 GDB Administrator <gdbadmin@sourceware.org>
59117 Automatic date update in version.in
59119 2022-03-10 Aaron Merey <amerey@redhat.com>
59121 gdb/auto-load: Remove repeating "auto-load" from debug message
59122 Remove "auto-load:" from a format string passed to auto_load_debug_printf.
59123 It is unnecessary since this function will prefix the string with "[auto-load]"
59126 2022-03-10 Tom Tromey <tromey@adacore.com>
59128 Change how "print/x" displays floating-point value
59129 Currently, "print/x" will display a floating-point value by first
59130 casting it to an integer type. This yields weird results like:
59135 This has confused users multiple times -- see PR gdb/16242, where
59136 there are several dups. I've also seen some confusion from this
59137 internally at AdaCore.
59142 Regard the bits of the value as an integer, and print the integer
59145 ... which seems more useful. So, perhaps what happened is that this
59146 was incorrectly implemented (or maybe correctly implemented and then
59147 regressed, as there don't seem to be any tests).
59149 This patch fixes the bug.
59151 There was a previous discussion where we agreed to preserve the old
59154 https://sourceware.org/legacy-ml/gdb-patches/2017-06/msg00314.html
59156 However, I think it makes more sense to follow the manual.
59158 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=16242
59160 2022-03-10 Tom Tromey <tromey@adacore.com>
59162 Simplify the ui-out progress API
59163 I noticed that 'progress' is a method on ui-out, but it seems to me
59164 that it would be better if the only API were via the progress_meter
59165 class. This patch makes this change, changing progress to be a method
59166 on the meter itself.
59168 2022-03-10 Andrew Burgess <aburgess@redhat.com>
59170 gdb/gdbarch: fix typo in gdbarch-components.py
59171 Fixes a minor typo, in a comment, in the gdbarch-components.py script.
59173 There should be no user visible changes after this commit.
59175 2022-03-10 Lancelot SIX <lancelot.six@amd.com>
59177 Process exit status is leader exit status testcase
59178 This adds a multi-threaded testcase that has all threads in the
59179 process exit with a different exit code, and ensures that GDB reports
59180 the thread group leader's exit status as the whole-process exit
59181 status. Before this set of patches, this would randomly report the
59182 exit code of some other thread, and thus fail.
59184 Tested on Linux-x86_64, native and gdbserver.
59186 Co-Authored-By: Pedro Alves <pedro@palves.net>
59187 Change-Id: I30cba2ff4576fb01b5169cc72667f3268d919557
59189 2022-03-10 Pedro Alves <pedro@palves.net>
59191 Re-add zombie leader on exit, gdbserver/linux
59192 Same as the previous patch, but for GDBserver.
59194 In summary, the current zombie leader detection code in linux-low.cc
59195 has a race -- if a multi-threaded inferior exits just before
59196 check_zombie_leaders finds that the leader is now zombie via checking
59197 /proc/PID/status, check_zombie_leaders deletes the leader, assuming we
59198 won't get an event for that exit (which we won't in some scenarios,
59199 but not in this one), which is a false-positive scenario, where the
59200 whole process is simply exiting. Later when we see the last LWP in
59201 our list exit, we report that LWP's exit status as exit code, even
59202 though for the (real) parent process, the exit code that counts is the
59203 child's leader thread's exit code.
59205 Like for GDB, the solution here is to:
59207 - only report whole-process exit events for the leader.
59208 - re-add the leader back to the LWP list when we finally see it
59211 Change-Id: Id2d7bbb51a415534e1294fff1d555b7ecaa87f1d
59213 2022-03-10 Pedro Alves <pedro@palves.net>
59215 Re-add zombie leader on exit, gdb/linux
59216 The current zombie leader detection code in linux-nat.c has a race --
59217 if a multi-threaded inferior exits just before check_zombie_leaders
59218 finds that the leader is now zombie via checking /proc/PID/status,
59219 check_zombie_leaders deletes the leader, assuming we won't get an
59220 event for that exit (which we won't in some scenarios, but not in this
59221 one). That might seem mostly harmless, but it has some downsides:
59223 - later when we continue pulling events out of the kernel, we will
59224 collect the exit event of the non-leader threads, and once we see
59225 the last lwp in our list exit, we return _that_ lwp's exit code as
59226 whole-process exit code to infrun, instead of the leader's exit
59229 - this can cause a hang in stop_all_threads in infrun.c. Say there
59230 are 2 threads in the process. stop_all_threads stops each of those
59231 threads, and then waits for two stop or exit events, one for each
59232 thread. If the whole process exits, and check_zombie_leaders hits
59233 the false-positive case, linux-nat.c will only return one event to
59234 GDB (the whole-process exit returned when we see the last thread,
59235 the non-leader thread, exit), making stop_all_threads hang forever
59236 waiting for a second event that will never come.
59238 However, in this false-positive scenario, where the whole process is
59239 exiting, as opposed to just the leader (with pthread_exit(), for
59240 example), we _will_ get an exit event shortly for the leader, after we
59241 collect the exit event of all the other non-leader threads. Or put
59242 another way, we _always_ get an event for the leader after we see it
59245 I tried a number of approaches to fix this:
59247 #1 - My first thought to address the race was to make GDB always
59248 report the whole-process exit status for the leader thread, not for
59249 whatever is the last lwp in the list. We _always_ get a final exit
59250 (or exec) event for the leader, and when the race triggers, we're not
59253 #2 - My second thought was to try to plug the race in the first place.
59255 I thought of making GDB call waitpid/WNOHANG for all non-leader
59256 threads immediately when the zombie leader is detected, assuming there
59257 would be an exit event pending for each of them waiting to be
59258 collected. Turns out that that doesn't work -- you can see the leader
59259 become zombie _before_ the kernel kills all other threads. Waitpid in
59260 that small time window returns 0, indicating no-event. Thankfully we
59261 hit that race window all the time, which avoided trading one race for
59262 another. Looking at the non-leader thread's status in /proc doesn't
59263 help either, the threads are still in running state for a bit, for the
59266 #3 - My next attempt, which seemed promising, was to synchronously
59267 stop and wait for the stop for each of the non-leader threads. For
59268 the scenario in question, this will collect all the exit statuses of
59269 the non-leader threads. Then, if we are left with only the zombie
59270 leader in the lwp list, it means we either have a normal while-process
59271 exit or an exec, in which case we should not delete the leader. If
59272 _only_ the leader exited, like in gdb.threads/leader-exit.exp, then
59273 after pausing threads, we will still have at least one live non-leader
59274 thread in the list, and so we delete the leader lwp. I got this
59275 working and polished, and it was only after staring at the kernel code
59276 to convince myself that this would really work (and it would, for the
59277 scenario I considered), that I realized I had failed to account for
59278 one scenario -- if any non-leader thread is _already_ stopped when
59279 some thread triggers a group exit, like e.g., if you have some threads
59280 stopped and then resume just one thread with scheduler-locking or
59281 non-stop, and that thread exits the process. I also played with
59282 PTRACE_EVENT_EXIT, see if it would help in any way to plug the race,
59283 and I couldn't find a way that it would result in any practical
59284 difference compared to looking at /proc/PID/status, with respect to
59287 So I concluded that there's no way to plug the race, we just have to
59288 deal with it. Which means, going back to approach #1. That is the
59289 approach taken by this patch.
59291 Change-Id: I6309fd4727da8c67951f9cea557724b77e8ee979
59293 2022-03-10 Pedro Alves <pedro@palves.net>
59295 gdbserver: Reindent check_zombie_leaders
59296 This fixes the indentation of
59297 linux_process_target::check_zombie_leaders, which will help with
59298 keeping its comments in sync with the gdb/linux-nat.c counterpart.
59300 Change-Id: I37332343bd80423d934249e3de2d04feefad1891
59302 2022-03-10 Pedro Alves <pedro@palves.net>
59304 gdbserver: Reorganize linux_process_target::filter_event
59305 Reorganize linux-low.cc:linux_process_target::filter_event such that
59306 all the handling for events for LWPs not in the LWP list is together.
59307 This helps make a following patch clearer. The comments and debug
59308 messages have also been tweaked to have them synchronized with the GDB
59311 Change-Id: If9019635f63a846607cfda44b454b4254a404019
59313 2022-03-10 Pedro Alves <pedro@palves.net>
59315 gdb: Reorganize linux_nat_filter_event
59316 Reorganize linux-nat.c:linux_nat_filter_event such that all the
59317 handling for events for LWPs not in the LWP list is together. This
59318 helps make a following patch clearer. The comments and debug messages
59319 have also been tweaked - the end goal is to have them synchronized
59320 with the gdbserver counterpart.
59322 Change-Id: I8586d8dcd76d8bd3795145e3056fc660e3b2cd22
59324 2022-03-10 Pedro Alves <pedro@palves.net>
59326 Fix gdb.threads/current-lwp-dead.exp race
59327 If we make GDB report the process EXIT event for the leader thread, as
59328 will be done in a latter patch of this series, then
59329 gdb.threads/current-lwp-dead.exp starts failing:
59331 (gdb) break fn_return
59332 Breakpoint 2 at 0x5555555551b5: file /home/pedro/rocm/gdb/build/gdb/testsuite/../../../src/gdb/testsuite/gdb.threads/current-lwp-dead.c, line 45.
59336 [Inferior 1 (process 2138459) exited normally]
59337 (gdb) FAIL: gdb.threads/current-lwp-dead.exp: continue to breakpoint: fn_return (the program exited)
59339 The inferior exit reported is actually correct. The main thread has
59340 indeed exited, and that's the thread that has the right exit code to
59341 report to the user, as that's the exit code that is reported to the
59342 program's parent. In this case, GDB managed to collect the exit code
59343 for the leader thread before reaping the other thread, because in
59344 reality, the testcase isn't creating standard threads, it is using raw
59345 clone, and the new clones are put in their own thread group.
59347 Fix it by making the main "thread" not exit until the scenario we're
59348 exercising plays out. Also, run the program to completion for
59351 The original program really wanted the leader thread to exit before
59352 the fn_return function was reached -- it was important that the
59353 current thread as pointed by inferior_ptid was gone when infrun got
59354 the breakpoint event. I've tweaked the testcase to ensure that that
59355 condition is still held, though it is no longer the main thread that
59356 exits. This required a bit of synchronization between the threads,
59357 which required using CLONE_VM unconditionally. The #ifdef guards were
59359 https://sourceware.org/bugzilla/show_bug.cgi?id=11214, though I don't
59360 think they were necessary because the program is not using TLS. If it
59361 turns out they were necessary, we can link the testcase with "-z now"
59362 instead, which was mentioned as an alternative workaround in that
59365 Change-Id: I7be2f0da4c2fe8f80a60bdde5e6c623d8bd5a0aa
59367 2022-03-10 Pedro Alves <pedro@palves.net>
59369 Fix gdb.threads/clone-new-thread-event.exp race
59370 If we make GDB report the process EXIT event for the leader thread,
59371 instead of whatever is the last thread in the LWP list, as will be
59372 done in a latter patch of this series, then
59373 gdb.threads/current-lwp-dead.exp starts failing:
59375 (gdb) FAIL: gdb.threads/clone-new-thread-event.exp: catch SIGUSR1 (the program exited)
59377 This is a testcase race -- the main thread does not wait for the
59378 spawned clone "thread" to finish before exiting, so the main program
59379 may exit before the second thread is scheduled and reports its
59380 SIGUSR1. With the change to make GDB report the EXIT for the leader,
59381 the race is 100% reproducible by adding a sleep(), like so:
59383 --- c/gdb/testsuite/gdb.threads/clone-new-thread-event.c
59384 +++ w/gdb/testsuite/gdb.threads/clone-new-thread-event.c
59385 @@ -51,6 +51,7 @@ local_gettid (void)
59390 tkill (local_gettid (), SIGUSR1);
59396 Breakpoint 1, main (argc=1, argv=0x7fffffffd418) at gdb.threads/clone-new-thread-event.c:65
59397 65 stack = malloc (STACK_SIZE);
59401 [Inferior 1 (process 3715555) exited normally]
59402 (gdb) FAIL: gdb.threads/clone-new-thread-event.exp: catch SIGUSR1 (the program exited)
59404 That inferior exit reported is actually correct. The main thread has
59405 indeed exited, and that's the thread that has the right exit code to
59406 report to the user, as that's the exit code that is reported to the
59407 program's parent. In this case, GDB managed to collect the exit code
59408 for the leader thread before reaping the other thread, because in
59409 reality, the testcase isn't creating standard threads, it is using raw
59410 clone, and the new clones are put in their own thread group.
59412 Fix it by making the main thread wait for the child to exit. Also,
59413 run the program to completion for completeness.
59415 Change-Id: I315cd3dc2b9e860395dcab9658341ea868d7a6bf
59417 2022-03-10 Pedro Alves <pedro@palves.net>
59419 Fix gdbserver/linux target_waitstatus logging assert
59420 Turning on debug output in gdbserver leads to an assertion failure if
59421 gdbserver reports a non-signal event:
59423 [threads] wait_1: LWP 3273770: extended event with waitstatus status->kind = EXECD, execd_pathname = gdb.threads/non-ldr-exc-1/non-ldr-exc-1
59424 [threads] wait_1: Hit a non-gdbserver trap event.
59425 ../../src/gdbserver/../gdb/target/waitstatus.h:365: A problem internal to GDBserver has been detected.
59426 sig: Assertion `m_kind == TARGET_WAITKIND_STOPPED || m_kind == TARGET_WAITKIND_SIGNALLED' failed.
59428 Fix it in the obvious way, using target_waitstatus::to_string(),
59429 resulting in, for example:
59431 [threads] wait_1: ret = LWP 1542412.1542412, status->kind = STOPPED, sig = GDB_SIGNAL_TRAP
59433 Change-Id: Ia4832f9b4fa39f4af67fcaf21fd4d909a285a645
59435 2022-03-10 Nick Clifton <nickc@redhat.com>
59437 Add option to objdump/readelf to disable access to debuginfod servers.
59438 * dwarf.c (use_debuginfod): New variable. Set to 1.
59439 (load_separate_debug_info): Only call
59440 debuginfod_fetch_separate_debug_info is use_debuginfod is true.
59441 (dwarf_select_sections_by_names): Add do-not-use-debuginfod and
59442 use-debuginfod options.
59443 (dwarf_select_sections_by_letters): Add D and E options.
59444 * dwarf.h (use_debuginfod): New extern.
59445 * objdump.c (usage): Mention the new options.
59446 * readelf.c (usage): Likewise.
59447 * doc/binutils.texi: Document the new options.
59448 * doc/debug-options.texi: Describe the new options.
59449 * NEWS: Mention the new feature.
59450 * testsuite/binutils-all/debuginfod.exp: Add tests of the new
59453 2022-03-10 Alan Modra <amodra@gmail.com>
59455 Re: ld: Add a before_plugin_all_symbols_read hook
59456 * testsuite/ld-plugin/pr28849.d: Adjust for powerpc64 function
59459 2022-03-10 H.J. Lu <hjl.tools@gmail.com>
59461 ld: Add a before_plugin_all_symbols_read hook
59462 Add a before_plugin_all_symbols_read hook to load symbol references from
59463 DT_NEEDED entries, included from --copy-dt-needed-entries, before reading
59464 plugin symbols to properly resolve plugin symbol references.
59469 * elf-bfd.h (elf_link_hash_table): Add handling_dt_needed.
59470 * elflink.c (_bfd_elf_merge_symbol): Don't set non_ir_ref_dynamic
59471 before plugin 'all symbols read' hook is called.
59476 * ldelf.c (ldelf_handle_dt_needed): New function.
59477 (ldelf_before_plugin_all_symbols_read): Likewise.
59478 (ldelf_after_open): Call ldelf_handle_dt_needed.
59479 * ldelf.h (ldelf_before_plugin_all_symbols_read): New.
59480 * ldemul.c (ldemul_before_plugin_all_symbols_read): Likewise.
59481 * ldemul.h (ldemul_before_plugin_all_symbols_read): Likewise.
59482 (ld_emulation_xfer_struct): Add before_plugin_all_symbols_read.
59483 * ldlang.c (lang_process): Call
59484 ldemul_before_plugin_all_symbols_read before calling
59485 plugin_call_all_symbols_read.
59487 (gld${EMULATION_NAME}_before_plugin_all_symbols_read): New.
59488 (LDEMUL_BEFORE_PLUGIN_ALL_SYMBOLS_READ): New.
59489 * emultempl/emulation.em (ld_${EMULATION_NAME}_emulation):
59490 Initialize the before_plugin_all_symbols_read field.
59491 * testsuite/ld-plugin/lto.exp: Run PR ld/28849 tests.
59492 * testsuite/ld-plugin/pr28849.d: New file.
59493 * testsuite/ld-plugin/pr28849a.c: Likewise.
59494 * testsuite/ld-plugin/pr28849b.c: Likewise.
59496 2022-03-10 GDB Administrator <gdbadmin@sourceware.org>
59498 Automatic date update in version.in
59500 2022-03-09 Hans-Peter Nilsson <hp@axis.com>
59502 toplevel: Makefile.def: Make configure-sim depend on all-readline
59503 Without this, a "make all-sim" without the equivalent of
59504 libreadline-dev installed on the build system, won't
59505 properly pick up the in-tree readline build, and you'll see:
59508 Configuring in ./sim
59509 configure: creating cache ./config.cache
59510 checking build system type... x86_64-pc-linux-gnu
59511 checking host system type... x86_64-pc-linux-gnu
59512 checking target system type... cris-axis-elf
59513 checking for x86_64-pc-linux-gnu-gcc... gcc
59514 checking whether the C compiler works... yes
59516 checking for library containing tgetent... -ltermcap
59517 checking for readline in -lreadline... no
59518 configure: error: the required "readline" library is missing
59519 make[1]: *** [Makefile:11188: configure-sim] Error 1
59520 make[1]: Leaving directory '/home/hp/sim/b'
59522 The sim dependency on readline is apparently (nominally)
59523 valid as there's a readline call in sim/erc32/sis.c.
59525 2022-02-21 Hans-Peter Nilsson <hp@axis.com>
59527 * Makefile.def (dependencies): Make configure-sim depend on
59530 2022-03-09 Maciej W. Rozycki <macro@embecosm.com>
59532 GDB/testsuite: Fix a "displayed" typo in gdb.base/default.exp
59533 Fix a typo, s/dislayed/displayed/ in default.exp at the top level.
59535 GDB/testsuite: Remove a stray backslash from gdb.base/settings.exp
59536 Remove a stray trailing backslash from `test-integer' in settings.exp.
59537 It is harmless as only white space follows in the next line before the
59538 closing brace, so it merely swallows the newline character, but it may
59539 look confusing to the reader.
59541 2022-03-09 Christina Schimpe <christina.schimpe@intel.com> (tiny change)
59543 * gdb/doc/gdb.texinfo (Requirements): Fix a typo.
59545 2022-03-09 Alan Modra <amodra@gmail.com>
59547 Constant fold view increment expressions
59548 The idea here is to replace expressions like v + 1 + 1 + 1 with v + 3.
59550 * dwarf2dbg.c (set_or_check_view): Remove useless assertion.
59551 Resolve multiple view increments.
59552 * testsuite/gas/elf/dwarf2-18.d: Don't xfail mep.
59554 2022-03-09 Alan Modra <amodra@gmail.com>
59556 Reduce duplicated symbol_clone_if_forward_ref work
59557 * symbol.c (struct symbol_flags): Add forward_resolved.
59558 (symbol_entry_find): Update needle initialisation.
59559 (symbol_clone_if_forward_ref): Do no work when forward_resolved
59560 is already set. Set forward_resolved.
59562 2022-03-09 Aaron Merey <amerey@redhat.com>
59564 gdb: Try searching for auto-load script using .gnu_debuglink
59565 If an auto-load script cannot be found and objfile is a separate
59566 debuginfo whose filename does not match the name found in the parent
59567 file's .gnu_debuglink section, then repeat the search using the
59568 parent's filename where the last component is replaced with the
59569 .gnu_debuglink name.
59571 For example if the parent's filename is "/usr/lib/libxyz.so" and the
59572 name in its .gnu_debuglink section is "libxyz.so.debug", then
59573 if no auto-load script is otherwise found the search will be
59574 repeated with the filename "/usr/lib/libxyz.so.debug".
59576 This helps gdb locate auto-load scripts when debuginfo files do not have
59577 the expected filename, such as when they are aquired from debuginfod.
59579 2022-03-09 GDB Administrator <gdbadmin@sourceware.org>
59581 Automatic date update in version.in
59583 2022-03-08 Aaron Merey <amerey@redhat.com>
59585 PR gdb/27876 - debuginfod-downloaded source files don't pass proper fullname across mi / (gdb)info source
59586 Source files downloaded from debuginfod currently use their original DWARF
59587 filename as their "fullname". This causes a mismatch between the fullname
59588 and the actual location of the source file in the debuginfod client cache.
59590 MI consumers such as VSCode will fail to open debuginfod-downloaded
59591 source files due to this. Also 'info source' will fail to include the
59592 true paths of these files.
59594 To fix this, use the debuginfod cache path as the fullname for debuginfod-
59595 downloaded source files.
59597 2022-03-08 Jan Vrany <jan.vrany@labware.com>
59599 gdb/mi: preserve user selected thread and frame when invoking MI commands
59600 Fix for PR gdb/20684. When invoking MI commands with --thread and/or
59601 --frame, the user selected thread and frame was not preserved:
59606 ~" Id Target Id Frame \n"
59607 ~"* 1 Thread 0x7ffff7c30740 (LWP 19302) \"user-selected-c\" main () at /home/uuu/gdb/gdb/testsuite/gdb.mi/user-selected-context-sync.c:60\n"
59608 ~" 2 Thread 0x7ffff7c2f700 (LWP 19306) \"user-selected-c\" child_sub_function () at /home/uuu/gdb/gdb/testsuite/gdb.mi/user-selected-context-sync.c:30\n"
59609 ~" 3 Thread 0x7ffff742e700 (LWP 19307) \"user-selected-c\" child_sub_function () at /home/uuu/gdb/gdb/testsuite/gdb.mi/user-selected-context-sync.c:30\n"
59614 ~"Stack level 0, frame at 0x7fffffffdf90:\n"
59615 ~" rip = 0x555555555207 in main (/home/uuu/gdb/gdb/testsuite/gdb.mi/user-selected-context-sync.c:60); saved rip = 0x7ffff7c5709b\n"
59616 ~" source language c.\n"
59617 ~" Arglist at 0x7fffffffdf80, args: \n"
59618 ~" Locals at 0x7fffffffdf80, Previous frame's sp is 0x7fffffffdf90\n"
59619 ~" Saved registers:\n "
59620 ~" rbp at 0x7fffffffdf80, rip at 0x7fffffffdf88\n"
59623 -stack-info-depth --thread 3
59628 ~" Id Target Id Frame \n"
59629 ~" 1 Thread 0x7ffff7c30740 (LWP 19302) \"user-selected-c\" main () at /home/uuu/gdb/gdb/testsuite/gdb.mi/user-selected-context-sync.c:60\n"
59630 ~" 2 Thread 0x7ffff7c2f700 (LWP 19306) \"user-selected-c\" child_sub_function () at /home/uuu/gdb/gdb/testsuite/gdb.mi/user-selected-context-sync.c:30\n"
59631 ~"* 3 Thread 0x7ffff742e700 (LWP 19307) \"user-selected-c\" child_sub_function () at /home/uuu/gdb/gdb/testsuite/gdb.mi/user-selected-context-sync.c:30\n"
59636 ~"Stack level 0, frame at 0x7ffff742dee0:\n"
59637 ~" rip = 0x555555555169 in child_sub_function (/home/uuu/gdb/gdb/testsuite/gdb.mi/user-selected-context-sync.c:30); saved rip = 0x555555555188\n"
59638 ~" called by frame at 0x7ffff742df00\n"
59639 ~" source language c.\n"
59640 ~" Arglist at 0x7ffff742ded0, args: \n"
59641 ~" Locals at 0x7ffff742ded0, Previous frame's sp is 0x7ffff742dee0\n"
59642 ~" Saved registers:\n "
59643 ~" rbp at 0x7ffff742ded0, rip at 0x7ffff742ded8\n"
59647 This caused problems for frontends that provide access to CLI because UI
59648 may silently change the context for CLI commands (as demonstrated above).
59650 This commit fixes the problem by restoring thread and frame in
59651 mi_cmd_execute (). With this change, there are only two GDB/MI commands
59652 that can change user selected context: -thread-select and -stack-select-frame.
59653 This allows us to remove all and rather complicated logic of notifying
59654 about user selected context change from mi_execute_command (), leaving it
59655 to these two commands themselves to notify.
59657 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=20684
59659 2022-03-08 Andrew Burgess <aburgess@redhat.com>
59661 gdb: announce upcoming removal of Python 2 support from gdb
59662 As has been discussed here:
59664 https://sourceware.org/pipermail/gdb-patches/2022-January/184910.html
59666 Python 2 support will be removed from GDB after GDB 12 has branched.
59667 This commit places an entry in the NEWS file to inform users of this
59670 2022-03-08 GDB Administrator <gdbadmin@sourceware.org>
59672 Automatic date update in version.in
59674 2022-03-07 Andrew Burgess <aburgess@redhat.com>
59676 gdb/testsuite: add new test for comparing char types in Python
59677 There's an interesting property of the 'char' type in C and C++, the
59678 three types 'char', 'unsigned char', and 'signed char', are all
59679 considered distinct.
59681 In contrast, and 'int' is signed by default, and so 'int' and 'signed
59682 int' are considered the same type.
59684 This commit adds a test to ensure that this edge case is visible to a
59687 It is worth noting that for any particular compiler implementation (or
59688 the flags a compiler was invoked with), a 'char' will be either signed
59689 or unsigned; it has to be one or the other, and a user can access this
59690 information by using the Type.is_signed property. However, for
59691 something like function overload resolution, the 'char' type is
59692 considered distinct from the signed and unsigned variants.
59694 There's no change to GDB with this commit, this is just adding a new
59695 test to guard some existing functionality.
59697 2022-03-07 Andrew Burgess <aburgess@redhat.com>
59699 gdb/python: add Type.is_signed property
59700 Add a new read-only property, Type.is_signed, which is True for signed
59701 types, and False otherwise.
59703 This property should only be read on types for which Type.is_scalar is
59704 true, attempting to read this property for non-scalar types will raise
59707 I chose 'is_signed' rather than 'is_unsigned' in order to match the
59708 existing Architecture.integer_type method, which takes a 'signed'
59709 parameter. As far as I could find, that was the only existing
59710 signed/unsigned selector in the Python API, so it seemed reasonable to
59713 2022-03-07 Andrew Burgess <aburgess@redhat.com>
59715 gdb/python: add Type.is_scalar property
59716 Add a new read-only property which is True for scalar types,
59717 otherwise, it's False.
59719 2022-03-07 Andrew Burgess <aburgess@redhat.com>
59721 gdb/mi: add --no-connection to MI -add-inferior command
59722 Following on from the previous commit, where the -add-inferior command
59723 now uses the same connection as the current inferior, this commit adds
59724 a --no-connection option to -add-inferior.
59726 This new option matches the existing option of the same name for the
59727 CLI version of add-inferior; the new inferior is created with no
59730 I've added a new 'connection' field to the MI output of -add-inferior,
59731 which includes the connection number and short name. I haven't
59732 included the longer description field, this is the MI after all. My
59733 expectation would be that if the frontend wanted to display all the
59734 connection details then this would be looked up from 'info
59735 connection' (or the MI equivalent if/when such a command is added).
59737 The existing -add-inferior tests are updated, as are the docs.
59739 2022-03-07 Umair Sair <umair_sair@hotmail.com>
59741 gdb/mi: fix regression in mi -add-inferior command
59742 Prior to the multi-target support commit:
59744 commit 5b6d1e4fa4fc6827c7b3f0e99ff120dfa14d65d2
59745 Date: Fri Jan 10 20:06:08 2020 +0000
59747 Multi-target support
59749 When a new inferior was added using the MI -add-inferior command, the
59750 new inferior would be using the same target as all the other
59751 inferiors. This makes sense, GDB only supported a single target stack
59754 After the above commit, each inferior has its own target stack.
59756 To maintain backward compatibility, for the CLI add-inferior command,
59757 when a new inferior is added the above commit has the new inferior
59758 inherit a copy of the target stack from the current inferior.
59760 Unfortunately, this same backward compatibility is missing for the MI.
59762 This commit fixes this oversight.
59764 Now, when the -add-inferior MI command is used, the new inferior will
59765 inherit a copy of the target stack from the current inferior.
59767 2022-03-07 Tom Tromey <tromey@adacore.com>
59770 GDB has a dbx emulation mode that adds a few aliases and helper
59771 commands. This mode is barely documented and is very superficial
59772 besides. I suspect it is rarely used, and I would like to propose
59773 deprecating it for GDB 12, and then removing it in GDB 13.
59775 2022-03-07 Pedro Alves <pedro@palves.net>
59777 Remove unnecessary inferior lookup in infrun:handle_one
59778 infrun.c:handle_one calls find_inferior_ptid unnecessarily, since we
59779 already have a thread pointer handy, and the thread has a pointer to
59780 the inferior. This commit removes the unnecessary lookup.
59782 Change-Id: I2ae18601dd75346c6c91068e9a4f9a6484fb3339
59784 2022-03-07 Tom Tromey <tromey@adacore.com>
59786 Fix bug in ada_print_floating
59787 ada_print_floating rewrites a floating-point string representation to
59788 conform to Ada syntax. However, if you managed to get a floating
59789 point error, you might see:
59791 (gdb) print whatever
59792 $2 = <invalid float valu.0e>
59794 What's happening here is that ada_print_floating doesn't recognize
59795 this error case, and proceeds to modify the error text.
59797 This patch fixes this problem.
59799 2022-03-07 Tom Tromey <tromey@adacore.com>
59801 Implement real literal extension for Ada
59802 Sometimes it is convenient to be able to specify the exact bits of a
59803 floating-point literal. For example, you may want to set a
59804 floating-point register to a denormalized value, or to a particular
59807 In C, you can do this by combining the "{}" cast with an array
59810 (gdb) p {double}{0x576488BDD2AE9FFE}
59811 $1 = 9.8765449999999996e+112
59813 This patch adds a somewhat similar idea to Ada. It extends the lexer
59814 to allow "l" and "f" suffixes in a based literal. The "f" indicates a
59815 floating-point literal, and the "l"s control the size of the
59816 floating-point type.
59818 Note that this differs from Ada's based real literals. I believe
59819 those can also be used to control the bits of a floating-point value,
59820 but they are a bit more cumbersome to use (simplest is binary but
59821 that's also very lengthy). Also, these aren't implemented in GDB.
59823 I chose not to allow this extension to work with based integer
59824 literals with exponents. That didn't seem very useful.
59826 2022-03-07 Tom Tromey <tromey@adacore.com>
59828 Fix Ada integer literals with exponents
59829 While working on another patch, I noticed that Ada integer literals
59830 with exponents did not work. For example, with one form you get an
59834 Invalid digit `e' in based literal
59836 And with another form you get an incorrect value:
59838 (gdb) print 16#8#e2
59841 This patch fixes the bugs and adds tests.
59843 2022-03-07 Tom Tromey <tromey@adacore.com>
59845 Fix gdb.ada/arrayptr.exp results
59846 PR ada/28115 points out that gdb.ada/arrayptr.exp works with GNAT 12,
59847 but fails with minimal encodings in earlier versions.
59849 This patch updates the test to try to report the results correctly. I
59850 tried this with the Fedora 34 system gcc (GCC 11) and with a GCC 12
59851 built from git trunk sometime relatively recently.
59853 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28115
59855 2022-03-07 Tom Tromey <tromey@adacore.com>
59857 Handle non-ASCII identifiers in Ada
59858 Ada allows non-ASCII identifiers, and GNAT supports several such
59859 encodings. This patch adds the corresponding support to gdb.
59861 GNAT encodes non-ASCII characters using special symbol names.
59863 For character sets like Latin-1, where all characters are a single
59864 byte, it uses a "U" followed by the hex for the character. So, for
59865 example, thorn would be encoded as "Ufe" (0xFE being lower case
59868 For wider characters, despite what the manual says (it claims
59869 Shift-JIS and EUC can be used), in practice recent versions only
59870 support Unicode. Here, characters in the base plane are represented
59871 using "Wxxxx" and characters outside the base plane using
59874 GNAT has some further quirks here. Ada is case-insensitive, and GNAT
59875 emits symbols that have been case-folded. For characters in ASCII,
59876 and for all characters in non-Unicode character sets, lower case is
59877 used. For Unicode, however, characters that fit in a single byte are
59878 converted to lower case, but all others are converted to upper case.
59880 Furthermore, there is a bug in GNAT where two symbols that differ only
59881 in the case of "Y WITH DIAERESIS" (and potentially others, I did not
59882 check exhaustively) can be used in one program. I chose to omit
59883 handling this case from gdb, on the theory that it is hard to figure
59884 out the logic, and anyway if the bug is ever fixed, we'll regret
59885 having a heuristic.
59887 This patch introduces a new "ada source-charset" setting. It defaults
59888 to Latin-1, as that is GNAT's default. This setting controls how "U"
59889 characters are decoded -- W/WW are always handled as UTF-32.
59891 The ada_tag_name_from_tsd change is needed because this function will
59892 read memory from the inferior and interpret it -- and this caused an
59893 encoding failure on PPC when running a test that tries to read
59894 uninitialized memory.
59896 This patch implements its own UTF-32-based case folder. This avoids
59897 host platform quirks, and is relatively simple. A short Python
59898 program to generate the case-folding table is included. It simply
59899 relies on whatever version of Unicode is used by the host Python,
59900 which seems basically acceptable.
59902 Test cases for UTF-8, Latin-1, and Latin-3 are included. This
59903 exercises most of the new code paths, aside from Y WITH DIAERESIS as
59906 2022-03-07 Tom Tromey <tromey@adacore.com>
59908 Define HOST_UTF32 in charset.h
59909 rust-parse.c has a #define for the host-specific UTF-32 charset name.
59910 A later patch needs the same thing, so this patch moves the definition
59911 to charset.h for easier reuse.
59913 2022-03-07 Tom Tromey <tromey@adacore.com>
59915 Let phex and phex_nz handle sizeof_l==1
59916 Currently, neither phex nor phex_nz handle sizeof_l==1 -- they let
59917 this case fall through to the default case. However, a subsequent
59918 patch in this series needs this case to work correctly.
59920 I looked at all calls to these functions that pass a 1 for the
59921 sizeof_l parameter. The only such case seems to be correct with this
59924 2022-03-07 Tom Tromey <tromey@adacore.com>
59926 Don't pre-size result string in ada_decode
59927 Currently, ada_decode pre-sizes the output string, filling it with 'X'
59928 characters. However, it's a bit simpler and more flexible to let
59929 std::string do the work here, and simply append characters to the
59930 string as we go. This turns out to be useful for a subsequent patch.
59932 Simplify a regular expression in ada-lex.l
59933 ada-lex.l uses "%option case-insensitive", so there is no need for
59934 regular expressions to match upper case.
59936 2022-03-07 GDB Administrator <gdbadmin@sourceware.org>
59938 Automatic date update in version.in
59940 2022-03-06 Maciej W. Rozycki <macro@orcam.me.uk>
59942 MIPS/opcodes: Fix alias annotation for branch instructions
59943 Correct issues with INSN2_ALIAS annotation for branch instructions:
59945 - regular MIPS BEQZ/L and BNEZ/L assembly instructions are idioms for
59946 BEQ/L and BNE/L respectively with the `rs' operand equal to $0,
59948 - microMIPS 32-bit BEQZ and BNEZ assembly instructions are idioms for
59949 BEQ and BNE respectively with the `rt' operand equal to $0,
59951 - regular MIPS BAL assembly instruction is an idiom for architecture
59952 levels of up to the MIPSr5 ISA and a machine instruction on its own
59953 from the MIPSr6 ISA up.
59955 Add missing annotation to BEQZ/L and BNEZ/L accordingly then and add a
59956 new entry for BAL for the MIPSr6 ISA, correcting a disassembly bug:
59958 $ mips-linux-gnu-objdump -m mips:isa64r6 -M no-aliases -d bal.o
59960 bal.o: file format elf32-tradlittlemips
59962 Disassembly of section .text:
59965 0: 04110000 0x4110000
59969 Add test cases accordingly.
59971 Parts for regular MIPS BEQZ/L and BNEZ/L instructions from Sagar Patel.
59973 2022-03-06 Maciej W. Rozycki <macro@orcam.me.uk>
59976 * testsuite/binutils-all/mips/mips1-branch-alias.d: New test.
59977 * testsuite/binutils-all/mips/mips1-branch-noalias.d: New test.
59978 * testsuite/binutils-all/mips/mips2-branch-alias.d: New test.
59979 * testsuite/binutils-all/mips/mips2-branch-noalias.d: New test.
59980 * testsuite/binutils-all/mips/mips32r6-branch-alias.d: New test.
59981 * testsuite/binutils-all/mips/mips32r6-branch-noalias.d: New
59983 * testsuite/binutils-all/mips/micromips-branch-alias.d: New
59985 * testsuite/binutils-all/mips/micromips-branch-noalias.d: New
59987 * testsuite/binutils-all/mips/mips-branch-alias.s: New test
59989 * testsuite/binutils-all/mips/micromips-branch-alias.s: New test
59991 * testsuite/binutils-all/mips/mips.exp: Run the new tests.
59993 2022-03-06 Sagar Patel <sagarmp@cs.unc.edu>
59994 Maciej W. Rozycki <macro@orcam.me.uk>
59997 * mips-opc.c (mips_builtin_opcodes): Fix INSN2_ALIAS annotation
59998 for "bal", "beqz", "beqzl", "bnez" and "bnezl" instructions.
59999 * micromips-opc.c (micromips_opcodes): Likewise for "beqz" and
60000 "bnez" instructions.
60002 2022-03-06 Tom Tromey <tom@tromey.com>
60004 Use function view when iterating over block symbols
60005 This changes iterate_over_block_local_vars and
60006 iterate_over_block_arg_vars to take a gdb::function_view rather than a
60007 function pointer and a user-data. In one spot, this allows us to
60008 remove a helper structure and helper function. In another spot, this
60009 looked more complicated, so I changed the helper function to be an
60010 "operator()" -- also a simplification, just not as big.
60012 2022-03-06 Tom Tromey <tom@tromey.com>
60014 Simplify hppa-tdep.c use of objfile_key
60015 I happened to notice a couple of unnecessary casts in hppa-tdep.c, and
60016 then I saw that the use of objfile_key could be simplified -- removing
60017 some code and using the default deleter rather than noop_deleter.
60019 Tested by rebuilding. Let me know what you think.
60021 2022-03-06 Simon Marchi <simon.marchi@polymtl.ca>
60023 gdb: constify parameter of value_copy
60024 In a following patch, I have a const value I want to copy using a
60025 value_copy. However, value_copy takes a non-const source value, at the
60026 moment. Change the paramter to be const,
60028 If the source value is not lazy, we currently call
60029 value_contents_all_raw, which calls allocate_value_contents, to get a
60030 view on the contents. They both take a non-const value, that's a
60031 problem. My first attempt at solving it was to add a const version of
60032 value_contents_all_raw, make allocate_value_contents take a const value,
60035 - make value::contents mutable
60036 - make allocate_value_contents cast away the const
60038 The idea being that allocating the value contents buffer does modify the
60039 value at the bit level, but logically that doesn't change its state.
60041 That was getting a bit complicated, so what I ended up doing is make
60042 value_copy not call value_contents_all_raw. We know at this point that
60043 the value is not lazy, so value::contents must have been allocate
60046 Change-Id: I3741ab362bce14315f712ec24064ccc17e3578d4
60048 2022-03-06 Simon Marchi <simon.marchi@polymtl.ca>
60050 gdb: remove internalvar_funcs::destroy
60051 No kind of internal var uses it remove it. This makes the transition to
60052 using a variant easier, since we don't need to think about where this
60053 should be called (in a destructor or not), if it can throw, etc.
60055 Change-Id: Iebbc867d1ce6716480450d9790410d6684cbe4dd
60057 2022-03-06 GDB Administrator <gdbadmin@sourceware.org>
60059 Automatic date update in version.in
60061 2022-03-05 GDB Administrator <gdbadmin@sourceware.org>
60063 Automatic date update in version.in
60065 2022-03-04 Tom Tromey <tromey@adacore.com>
60067 Mark vDSO as not a file
60068 The vDSO objfile is not a real file, so mark it as such. I noticed
60069 this because, when playing with debuginfod, I saw:
60071 Downloading 0.01 MB separate debug info for /tmp/system-supplied DSO at 0x7ffff7fc9000
60073 That "/tmp" is wrong -- it's just gdb's cwd. This patch corrects the
60074 problem, resulting in:
60076 Downloading 0.01 MB separate debug info for system-supplied DSO at 0x7ffff7fc9000
60078 Regression tested on x86-64 Fedora 34.
60080 2022-03-04 Simon Marchi <simon.marchi@polymtl.ca>
60082 binutils/readelf: fix indentation in process_dynamic_section
60083 Clangd shows a warning about misleading indentation in this file, fix
60086 binutils/ChangeLog:
60088 * readelf.c (process_dynamic_section): Fix indentation.
60090 Change-Id: I43a7f4f4c75dd080af614222b980526f5debf297
60092 2022-03-04 Christina Schimpe <christina.schimpe@intel.com>
60094 gdb: Use a typedef's scoped type name to identify local typedefs
60095 GDB prints the wrong type for typedefs in case there is another typedef
60096 available for the same raw type (gdb/16040). The reason is that the
60097 current hashmap based substitution mechanism always compares the target
60098 type of a typedef and not its scoped name.
60100 The original output of GDB for a program like
60105 typedef double scoped_double;
60108 typedef double global_double;
60110 class TypedefHolder
60114 ns::scoped_double b;
60118 typedef double class_double;
60121 double method1(ns::scoped_double) { return 24.0; }
60122 double method2(global_double) { return 24.0; }
60136 Breakpoint 1 at 0x1131: file TypedefHolder.cc, line 27.
60138 Starting program: /tmp/typedefholder
60140 Breakpoint 1, main () at TypedefHolder.cc:27
60143 type = class TypedefHolder {
60151 class_double method1(class_double);
60152 class_double method2(class_double);
60154 typedef double class_double;
60158 Basically all attributes of a class which have the raw type "double" are
60159 substituted by "class_double".
60161 With the patch the output is the following
60164 type = class TypedefHolder {
60167 ns::scoped_double b;
60172 double method1(ns::scoped_double);
60173 double method2(global_double);
60175 typedef double class_double;
60179 2022-03-04 Jan Beulich <jbeulich@suse.com>
60181 RISC-V: make .insn actually work for 64-bit insns
60182 Presently in this case, due to an undefined behavior shift, at least
60183 with x86 cross builds I'm observing:
60185 Error: value conflicts with instruction length `8,0x0000003f'
60187 Eliminate the UB and extend the respective testcase.
60189 2022-03-04 Jan Beulich <jbeulich@suse.com>
60191 x86: drop redundant x86-64-code16-2 test
60192 The code16-2 test is already meaningless enough as a gas test, identical
60193 to this one, and is run uniformly for all ELF targets anyway.
60195 2022-03-04 GDB Administrator <gdbadmin@sourceware.org>
60197 Automatic date update in version.in
60199 2022-03-03 Roland McGrath <mcgrathr@google.com>
60201 Fix typo in last change.
60203 2022-03-03 Roland McGrath <mcgrathr@google.com>
60205 Avoid conflict with gnulib open/close macros.
60206 On some systems, the gnulib configuration will decide to define open
60207 and/or close as macros to replace the POSIX C functions. This
60208 interferes with using those names in C++ class or namespace scopes.
60211 * event-pipe.cc (event_pipe::open): Renamed to ...
60212 (event_pipe::open_pipe): ... this.
60213 (event_pipe::close): Renamed to ...
60214 (event_pipe::close_pipe): ... this.
60215 * event-pipe.h (class event_pipe): Updated.
60217 * inf-ptrace.h (async_file_open, async_file_close): Updated.
60219 * gdbserver/linux-low.cc (linux_process_target::async): Likewise.
60221 2022-03-03 Alan Modra <amodra@gmail.com>
60223 Adjust ld ctf test for 32-bit targets
60224 powerpc-linux, and I suspect other 32-bit targets, report "aligned at
60225 0x4" for this test.
60227 * testsuite/ld-ctf/nonrepresentable.d: Accept any alignment.
60229 2022-03-03 Luis Machado <luis.machado@arm.com>
60231 Update my e-mail address in the MAINTAINERS file
60232 Update the information accordingly.
60234 2022-03-03 Tiezhu Yang <yangtiezhu@loongson.cn>
60236 gdb: testsuite: fix failed testcases in gdb.base/gdb-caching-proc.exp
60237 When execute the following command:
60239 make check-gdb TESTS="gdb.base/gdb-caching-proc.exp"
60241 we can see there exist some failed testcases:
60243 FAIL: gdb.base/gdb-caching-proc.exp: can_spawn_for_attach: 0: can spawn for attach (got interactive prompt)
60244 FAIL: gdb.base/gdb-caching-proc.exp: can_spawn_for_attach: 1: can spawn for attach (got interactive prompt)
60245 FAIL: gdb.base/gdb-caching-proc.exp: can_spawn_for_attach: 2: can spawn for attach (got interactive prompt)
60246 FAIL: gdb.base/gdb-caching-proc.exp: can_spawn_for_attach: 3: can spawn for attach (got interactive prompt)
60247 FAIL: gdb.base/gdb-caching-proc.exp: can_spawn_for_attach: 4: can spawn for attach (got interactive prompt)
60248 FAIL: gdb.base/gdb-caching-proc.exp: can_spawn_for_attach: 5: can spawn for attach (got interactive prompt)
60249 FAIL: gdb.base/gdb-caching-proc.exp: can_spawn_for_attach: 6: can spawn for attach (got interactive prompt)
60250 FAIL: gdb.base/gdb-caching-proc.exp: can_spawn_for_attach: 7: can spawn for attach (got interactive prompt)
60251 FAIL: gdb.base/gdb-caching-proc.exp: can_spawn_for_attach: 8: can spawn for attach (got interactive prompt)
60252 FAIL: gdb.base/gdb-caching-proc.exp: can_spawn_for_attach: 9: can spawn for attach (got interactive prompt)
60254 here are the detailed messages in gdb/testsuite/gdb.log:
60257 A program is being debugged already. Kill it? (y or n) n
60259 (gdb) FAIL: gdb.base/gdb-caching-proc.exp: can_spawn_for_attach: 0: can spawn for attach (got interactive prompt)
60261 so handle the case "A program is being debugged already. Kill it" in
60262 can_spawn_for_attach to fix the failed testcases.
60264 2022-03-03 Alan Modra <amodra@gmail.com>
60268 2022-03-03 Alan Modra <amodra@gmail.com>
60270 PowerPC64 DT_RELR relative reloc addresses
60271 Section addresses can change between ppc64_elf_size_stubs and
60272 ppc64_elf_build_stubs due to .eh_frame editing. The idea of stashing
60273 r_offset final addresses calculated in ppc64_elf_size_stubs for use by
60274 ppc64_elf_build_stubs was never a good idea. Instead, we need to keep
60275 section/offset pairs.
60277 * elf64-ppc.c (struct ppc_link_hash_table): Delete relr_addr.
60278 Add relr section/offset array.
60279 (append_relr_off): Rewrite. Update all callers.
60280 (sort_relr): New function.
60281 (ppc64_elf_size_stubs): Adjust to suit new relative reloc stash.
60282 (ppc64_elf_build_stubs): Likewise.
60284 2022-03-03 GDB Administrator <gdbadmin@sourceware.org>
60286 Automatic date update in version.in
60288 2022-03-02 John Baldwin <jhb@FreeBSD.org>
60290 configure: Stop checking for PT_GETXMMREGS.
60291 This request is present on all modern *BSD/i386 systems (those
60292 released since mid-2006), and the *BSD/i386 targets now assume it is
60293 present unconditionally.
60295 i386-bsd-nat: Assume PT_GETXMMREGS is present.
60296 NetBSD has included PT_GETXMMREGS since 1.6 released in September
60297 2002. OpenBSD has included PT_GETXMMREGS since 3.8 released in
60300 i386-fbsd-nat: Assume PT_GETXMMREGS is present.
60301 PT_GETXMMREGS was first added in FreeBSD 6.0 released in November 2005.
60302 The last FreeBSD release without support was 5.5 released in May 2006.
60304 2022-03-02 John Baldwin <jhb@FreeBSD.org>
60306 fbsd-tdep: Implement the vsyscall_range gdbarch hook.
60307 FreeBSD recently added a real vDSO in its shared page for the amd64
60308 architecture. The vDSO is mapped at the address given by the
60309 AT_KPRELOAD ELF auxiliary vector entry. To find the end of the
60310 mapping range, parse the list of virtual map entries used by 'info
60311 proc mappings' either from the NT_PROCSTAT_VMMAP core dump note, or
60312 via the kinfo_getvmmap function for native targets (fetched from the
60313 native target as the TARGET_OBJECT_FREEBSD_VMMAP object).
60315 This silences warnings on recent FreeBSD/amd64 kernels due to not
60316 finding symbols for the vdso:
60318 warning: Could not load shared library symbols for [vdso].
60319 Do you need "set solib-search-path" or "set sysroot"?
60321 2022-03-02 Tom Tromey <tromey@adacore.com>
60323 Rewrite make-target-delegates in Python
60324 I think gdb is probably better off having fewer languages involved
60325 when generating code. 'sh' is unavoidable for build-time generation,
60326 but for other things, let's use Python.
60328 This rewrites make-target-delegates in Python. I've stuck pretty
60329 closely to the original code in this rewrite, so it may look slightly
60330 weird from a Python perspective.
60332 The only output difference is that a copyright header is now
60333 generated, using the code introduced in the previous patch.
60335 make-target-delegates.py is simpler to invoke, as it knows the correct
60336 input file to scan and it creates the output file itself.
60338 2022-03-02 Tom Tromey <tromey@adacore.com>
60340 Move copyright code from gdbarch.py to new file
60341 This moves the copyright code from gdbarch.py to a new Python source
60342 file, gdbcopyright.py. The function in this file will find the
60343 copyright dates by scanning the calling script. This will be reused
60346 This involved minor changes to the output of gdbarch.py. Also, I've
60347 updated copyright.py to remove the reference to gdbarch.sh. We don't
60348 need to mention gdbarch.py there, either.
60350 2022-03-02 GDB Administrator <gdbadmin@sourceware.org>
60352 Automatic date update in version.in
60354 2022-03-01 Tom Tromey <tom@tromey.com>
60356 Some "distclean" fixes in gdb
60357 PR build/12440 points out that "make distclean" is broken in gdb.
60358 Most of the breakage comes from other projects in the tree, but we can
60359 fix some of the issues, which is what this patch does.
60361 Note that the yacc output files, like c-exp.c, are left alone. In a
60362 source distribution, these are included in the tarball, and if the
60363 user builds in-tree, we would not want to remove them.
60365 While that seems a bit obscure, it seems to me that "distclean" is
60366 only really useful for in-tree builds anyway -- out-of-tree I simply
60367 delete the entire build directory and start over.
60369 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=12440
60371 2022-03-01 Tom Tromey <tom@tromey.com>
60373 Fix typo in the "alias" example
60374 PR cli/17332, filed around 8 years ago, points out a typo in the docs
60375 -- in one example, the command and its output are obviously out of
60376 sync. This patch fixes it. I'm checking this in as obvious.
60378 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=17332
60380 2022-03-01 Nick Clifton <nickc@redhat.com>
60382 Fix a typo in the previous delta to bfdio.c.
60384 * bfdio.c (_bfd_real_fopen): Fix typo.
60386 2022-03-01 Alan Modra <amodra@gmail.com>
60388 Revert "Check thin archive element file size against archive header"
60389 This reverts commit 48e3e6aec8a4f37d00ea6c0da3ab45e76490e3db.
60392 * archive.c (_bfd_get_elt_at_filepos): Don't check thin archive
60395 2022-03-01 Nick Clifton <nickc@redhat.com>
60397 Fix linker tests to compile with gcc-12.
60399 * testsuite/ld-elf/pr21964-1a.c: Fix array comparisons.
60400 * testsuite/ld-elf/pr21964-1b.c: Likewise.
60401 * testsuite/ld-elf/pr21964-1c.c: Likewise.
60402 * testsuite/ld-elf/pr21964-2a.c: Likewise.
60403 * testsuite/ld-elf/pr21964-2b.c: Likewise.
60404 * testsuite/ld-elf/pr21964-3a.c: Likewise.
60406 Prevent an assertion from being triggered when linking an ARM object file with incorrectly set build attributes.
60409 * elf32-arm.c (elf32_arm_merge_eabi_attributes): If the first
60410 input bfd has a Tag_ABI_HardFP_use set to 3 but does not also have
60411 TAG_FP_arch set then reset the TAG_ABI_HardFP_use.
60413 2022-03-01 Tiezhu Yang <yangtiezhu@loongson.cn>
60415 gdb: testsuite: fix wrong expected result in attach-pie-noexec.exp
60416 If /proc/sys/kernel/yama/ptrace_scope is 1, when execute the test case
60417 gdb.base/attach-pie-noexec.exp without superuser, the gdb.log shows the
60421 Attaching to process 6500
60422 ptrace: Operation not permitted.
60423 (gdb) PASS: gdb.base/attach-pie-noexec.exp: attach
60425 It is obviously wrong, the expected result should be UNSUPPORTED in such
60428 It is better to make can_spawn_for_attach to return false for this case.
60429 It would have to setup a small test program, compile it to exec, spawn it
60430 and try to attach to it.
60432 With this patch, we can see "Operation not permitted" in the log info,
60433 and then we can do the following processes to test:
60434 (1) set ptrace_scope as 0
60435 $ echo 0 | sudo tee /proc/sys/kernel/yama/ptrace_scope
60436 $ make check-gdb TESTS="gdb.base/attach-pie-noexec.exp"
60438 $ sudo make check-gdb TESTS="gdb.base/attach-pie-noexec.exp"
60440 Additionally, handle the other cases when test with RUNTESTFLAGS=
60441 "--target_board=native-extended-gdbserver".
60443 2022-03-01 Tiezhu Yang <yangtiezhu@loongson.cn>
60445 gdb: testsuite: print explicit test result in can_spawn_for_attach
60446 In the current code, there is no test result when execute the following
60449 $ make check-gdb TESTS="gdb.base/attach-pie-noexec.exp" RUNTESTFLAGS="--target_board=remote-gdbserver-on-localhost"
60450 $ make check-gdb TESTS="gdb.base/attach-pie-noexec.exp" RUNTESTFLAGS="--target_board=native-gdbserver"
60452 It is better to print explicit test result in can_spawn_for_attach.
60454 2022-03-01 Tiezhu Yang <yangtiezhu@loongson.cn>
60456 gdb: add Tiezhu Yang as LoongArch maintainer
60457 The patch series "gdb: Add basic support for LoongArch" has been
60458 merged into master, list Tiezhu Yang as LoongArch maintainer.
60460 2022-03-01 GDB Administrator <gdbadmin@sourceware.org>
60462 Automatic date update in version.in
60464 2022-02-28 Keith Seitz <keiths@redhat.com>
60466 Fix "spawn id XYZ not open" errors in gdb.mi/mi-exec-run.exp
60467 Running mi-exec-run.exp on native-extended-gdbserver/-m{32,64}
60468 causes several Tcl errors to appear. For example,
60471 ERROR: : spawn id exp20 not open
60474 -i exp11 -timeout 10
60475 -i "$inferior_spawn_id"
60476 -re ".*Cannot exec.*Permission denied" {
60477 set saw_perm_error 1
60478 verbose -log "saw..."
60479 ("uplevel" body line 1)
60480 invoked from within
60481 "uplevel $body" NONE : spawn id exp20 not open
60482 UNRESOLVED: gdb.mi/mi-exec-run.exp: inferior-tty=separate: mi=separate: force-fail=1: run failure detected (eof)
60484 This is happening because of the way this test is implemented:
60488 -i "$inferior_spawn_id"
60489 -re ".*Cannot exec.*Permission denied" {
60490 set saw_perm_error 1
60491 verbose -log "saw mi error"
60494 -re "\\^error,msg=\"During startup program exited with code 127" {
60496 verbose -log "saw mi error"
60502 The first time this loop is executed, `inferior_spawn_id' is valid. When the
60503 first branch of the expect statement is reached, gdbserver has exited, closing
60504 the spawn_id. Since we haven't seen the gdb-side error yet, the loop is executed
60505 again. The first branch now refers to a non-existent spawn_id, leading to the error.
60507 This can be fixed by using exp_continue to loop in expect instead of looping around
60508 expect, which is the approach I have used[1]. Note I've had to update the expected
60509 message for the "During startup..." error message when running with gdbserver.
60511 One other small change I've made is to add a log entry which spills the values of
60512 the two variables, saw_mi_error and saw_perm_error (and updated the log output
60513 for the later). This should make the log output clearer about why the test failed.
60515 With this patch installed, all the ERRORs disappear, leaving previously masked
60516 FAILs (which I have not attempted to fix).
60518 [1] Anyone know why this test doesn't simply use gdb_test_multiple? I can only
60519 assume that it was intentionally written this way, and I've modified the code with
60520 that assumption. I have tested a version using gdb_test_multiple, and that appears
60521 to work fine, too, if that is preferred. [It still employs exp_continue to fix the
60524 2022-02-28 Tom Tromey <tromey@adacore.com>
60526 Add more filename styling
60527 I found a few spots where filename styling ought to be applied, but is
60530 2022-02-28 Tom Tromey <tromey@adacore.com>
60532 Fix maybe-uninitialized warning in py-infthread.c
60533 I got this warning from py-infthread.c using the Fedora 34 system GCC:
60535 ../../binutils-gdb/gdb/python/py-infthread.c:102:30: warning: ‘extra_info’ may be used uninitialized in this function [-Wmaybe-uninitialized]
60537 I think this happens because GDB_PY_HANDLE_EXCEPTION expands to an
60538 'if' whose condition is always true -- but GCC can't know this. This
60539 patch avoids the warning by adding a harmless initialization.
60541 2022-02-28 Tom Tromey <tromey@adacore.com>
60543 Handle multi-byte bracket sequences in Ada lexer
60544 As noted in an earlier patch, the Ada lexer does not handle multi-byte
60545 bracket sequences. This patch adds support for these for character
60546 literals. gdb does not generally seem to handle the Ada wide string
60547 types, so for the time being these continue to be excluded -- but an
60548 explicit error is added to make this more clear.
60550 2022-02-28 Tom Tromey <tromey@adacore.com>
60552 Handle 'QWW' encoding case in Ada enums
60553 In Ada, an enum can contain character literals. GNAT encodes these
60554 values in a special way. For example, the Unicode character U+0178
60555 would be represented as 'QW0178' in the DWARF:
60557 <3><112f>: Abbrev Number: 2 (DW_TAG_enumerator)
60558 <1130> DW_AT_name : (indirect string, offset: 0x19ff): QW0178
60559 <1134> DW_AT_const_value : 2
60561 gdb handles this reasonably well, but failed to handle the 'QWW'
60562 encoding, which is used for characters outside the base plane.
60564 Also, while working on this, I noticed that gdb will print the decimal
60565 value for an enum character constant:
60570 This is a nice feature, IMO, because in this situation the 'x' enum
60571 constant does not have its usual decimal value -- it has the value
60572 that's assigned based on the enumeration type.
60574 However, gdb did not do this when it decided to print the constant
60575 using the bracket notation:
60577 (gdb) print Char_Thorn
60580 This patch changes gdb to print the decimal value here as well, and to
60581 put the bracket notation in single quotes -- otherwise gdb will be
60582 printing something that it can't then read. Now it looks like:
60584 (gdb) print Char_Thorn
60587 Note that gdb can't read longer bracket notations, like the other ones
60588 printed in this test case:
60590 (gdb) print Char_King
60591 $4 = 3 '["01fa00"]'
60593 While I think this is a bug, I plan to fix it separately.
60595 Finally, in the new test case, the copyright dates are chosen this way
60596 because this all started as a copy of an existing test.
60598 2022-02-28 Andrew Burgess <aburgess@redhat.com>
60600 gdb/python: Add gdb.InferiorThread.details attribute
60601 This adds a new read-only attribute gdb.InferiorThread.details, this
60602 attribute contains a string, the results of target_extra_thread_info
60603 for the thread, or None, if target_extra_thread_info returns nullptr.
60605 As the string returned by target_extra_thread_info is unstructured,
60606 this attribute is only really useful for echoing straight through to
60607 the user, but, if a user wants to write a command that displays the
60608 same, or a similar 'Thread Id' to the one seen in 'info threads', then
60609 they need access to this string.
60611 Given that the string produced by target_extra_thread_info varies by
60612 target, there's only minimal testing of this attribute, I check that
60613 the attribute can be accessed, and that the return value is either
60616 2022-02-28 Keith Seitz <keiths@redhat.com>
60618 Error when gdb_is_target_1 is called without running gdb instance
60619 This is a snafu that I encountered while implementing the previous
60620 patch, which attempted to use gdb_is_target_native. This proc and
60621 gdb_is_target_remote both rely on gdb_is_target_1, which actually
60622 cannot be called without gdb already running.
60624 This patch adds appropriate warning comments to these procs and
60625 causes gdb_is_target_1 to issue a Tcl error if it is called without a
60626 gdb instance already running. This should prevent unwitting callers
60627 from using this at the wrong time.
60629 2022-02-28 Keith Seitz <keiths@redhat.com>
60631 Fix gdb.fortran "failed to extract expected results" errors
60632 When running the gdb.fortran tests array-slices.exp and lbound-ubound.exp,
60633 the test suite throws several ERRORs on native-gdbserver/-m{32,64},
60634 and native-extended-gdbsever/-m{32,64}:
60636 [on native-extended-gdbserver/-m64]
60637 Running /home/keiths/work/gdb/branches/testsuite-errors/linux/gdb/testsuite/../../../src/gdb/testsuite/gdb.fortran/array-slices.exp ...
60638 ERROR: failed to extract expected results
60639 ERROR: failed to extract expected results
60640 Running /home/keiths/work/gdb/branches/testsuite-errors/linux/gdb/testsuite/../../../src/gdb/testsuite/gdb.fortran/lbound-ubound.exp ...
60641 ERROR: failed to extract expected results for lbound
60643 This occurs because the tests require inferior I/O which we do not have
60644 access to while using these targets.
60646 This patch skips these tests when running on non-native targets.
60648 2022-02-28 Torbj?rn Svensson <torbjorn.svensson@st.com>
60650 Further correct the handling of long pathnames on Windows hosts.
60652 * bfdio.c (_bfd_real_fopen): Fix handling of parhs longer than 260
60653 characters on Windows hosts.
60655 2022-02-28 Nick Clifton <nickc@redhat.com>
60657 Clarify the wording of the error message when an obsolete configuration is encountered.
60659 * config.bfd: Update error message for obsolete configurations.
60661 2022-02-28 GDB Administrator <gdbadmin@sourceware.org>
60663 Automatic date update in version.in
60665 2022-02-27 GDB Administrator <gdbadmin@sourceware.org>
60667 Automatic date update in version.in
60669 2022-02-26 Kevin Buettner <kevinb@redhat.com>
60671 Handle recursive internal problem in gdb_internal_error_resync
60672 I came across this problem when testing gdb.base/gdb-sigterm.exp
60673 on a machine with a pre-release version of glib-2.34 installed:
60675 A problem internal to GDB has been detected,
60676 further debugging may prove unreliable.
60677 Quit this debugging session? (y or n) Recursive internal problem.
60678 FAIL: gdb.base/gdb-sigterm.exp: expect eof #0 (GDB internal error)
60679 Resyncing due to internal error.
60680 ERROR: : spawn id exp11 not open
60683 -i exp11 -timeout 10
60684 -re "Quit this debugging session\\? \\(y or n\\) $" {
60685 send_gdb "n\n" answer
60689 ("uplevel" body line 1)
60690 invoked from within
60691 "uplevel $body" NONE : spawn id exp11 not open
60692 ERROR: Could not resync from internal error (timeout)
60693 gdb.base/gdb-sigterm.exp: expect eof #0: stepped 9 times
60694 UNRESOLVED: gdb.base/gdb-sigterm.exp: 50 SIGTERM passes
60696 I don't have a problem with the latter ERROR nor the UNRESOLVED
60697 messages. However the first ERROR regarding the exp11 spawn id
60698 not being open is not especially useful.
60700 This commit handles the "Recursive internal problem" case, avoiding
60701 the problematic ERROR shown above.
60703 With this commit in place, the log messages look like this instead:
60705 A problem internal to GDB has been detected,
60706 further debugging may prove unreliable.
60707 Quit this debugging session? (y or n) Recursive internal problem.
60708 FAIL: gdb.base/gdb-sigterm.exp: expect eof #15 (GDB internal error)
60709 Resyncing due to internal error.
60710 ERROR: Could not resync from internal error (recursive internal problem)
60711 gdb.base/gdb-sigterm.exp: expect eof #15: stepped 12 times
60712 UNRESOLVED: gdb.base/gdb-sigterm.exp: 50 SIGTERM passes
60714 gdb/testsuite/ChangeLog:
60716 * lib/gdb.exp (gdb_internal_error_resync): Handle "Recursive
60719 2022-02-26 GDB Administrator <gdbadmin@sourceware.org>
60721 Automatic date update in version.in
60723 2022-02-25 Aaron Merey <amerey@redhat.com>
60725 gdb-add-index: disable debuginfod
60726 gdb-add-index may trigger debuginfod's first-use notice. The notice
60727 is misleading in this case. It instructs the user to modify .gdbinit
60728 in order to permanently enable/disable debuginfod but gdb-add-index
60729 invokes gdb with -nx which ignores .gdbinit.
60731 Additionally debuginfod is not needed for gdb-add-index since the
60732 symbol file is given as an argument and should already be present
60735 Fix this by disabling debuginfod when gdb-add-index invokes gdb.
60737 2022-02-25 Andrew Burgess <aburgess@redhat.com>
60739 gdb: add operator+= and operator+ overload for std::string
60740 This commit adds operator+= and operator+ overloads for adding
60741 gdb::unique_xmalloc_ptr<char> to a std::string. I could only find 3
60742 places in GDB where this was useful right now, and these all make use
60745 I've also added a self test for gdb::unique_xmalloc_ptr<char>, which
60746 makes use of both operator+= and operator+, so they are both getting
60749 There should be no user visible changes after this commit, except when
60750 running 'maint selftest', where the new self test is visible.
60752 2022-02-25 Tom Tromey <tromey@adacore.com>
60754 Print MI prompt on interrupted command
60755 Joel noticed that if the remote dies unexpectedly during a command --
60756 you can simulate this by using "continue" and then killing gdbserver
60757 -- then the CLI will print a new prompt, but MI will not. Later, we
60758 found out that this was also filed in bugzilla as PR mi/23820.
60760 The output looks something like this:
60767 | *running,thread-id="all"
60769 | [... some output from GDB during program startup...]
60770 | =thread-exited,id="1",group-id="i1"
60771 | =thread-group-exited,id="i1"
60772 | &"Remote connection closed\n"
60774 Now, what about that "(gdb)" in the middle?
60776 That prompt comes from this questionable code in
60777 mi-interp.c:mi_on_resume_1:
60779 /* This is what gdb used to do historically -- printing prompt
60780 even if it cannot actually accept any input. This will be
60781 surely removed for MI3, and may be removed even earlier. */
60782 if (current_ui->prompt_state == PROMPT_BLOCKED)
60783 fputs_unfiltered ("(gdb) \n", mi->raw_stdout);
60785 ... which seems like something to remove. But maybe the intent here
60786 is that this prompt is sufficient, and MI clients must be ready to
60787 handle output coming after a prompt. On the other hand, if this code
60788 *is* removed, then nothing would print a prompt in this scenario.
60790 Anyway, the CLI and the TUI handle emitting the prompt here by hooking
60791 into gdb::observers::command_error, but MI doesn't install an observer
60794 This patch adds the missing observer and arranges to show the MI
60795 prompt. Regression tested on x86-64 Fedora 34.
60797 It seems like this area could be improved a bit, by having
60798 start_event_loop call the prompt-displaying code directly, rather than
60799 indirecting through an observer. However, I haven't done this.
60801 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=23820
60803 2022-02-25 Andrew Burgess <aburgess@redhat.com>
60804 Tom Tromey <tromey@adacore.com>
60806 gdb/testsuite: fix list.exp test cases
60807 PR testsuite/7142 -- old enough to have been converted from Gnats --
60808 points out that test_list_filename_and_function in gdb.base/list.exp
60809 has "fails" that are unmatched with passes. This patch cleans this up
60812 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=7142
60814 2022-02-25 Tsukasa OI <research_trasio@irq.a4lg.com>
60816 RISC-V: Remove a loop in the ISA parser
60817 Since commit e601909a3287bf541c6a7d82214bb387d2c76d82 ("RISC-V: Support
60818 to parse the multi-letter prefix in the architecture string.") changed
60819 so that all prefixed extensions are parsed in single
60820 riscv_parse_prefixed_ext call, a "while" loop on riscv_parse_subset
60821 is no longer required.
60825 * elfxx-riscv.c (riscv_parse_subset): Remove unnecessary loop.
60827 2022-02-25 Tsukasa OI <research_trasio@irq.a4lg.com>
60829 RISC-V: Fix mask for some fcvt instructions
60830 This commit fixes incorrect uses of mask values in 'fcvt' instruction
60835 * riscv-opc.c (riscv_opcodes): Fix incorrect uses of mask values
60836 in 'fcvt' instruction family.
60838 2022-02-25 Keith Seitz <keiths@redhat.com>
60840 Support template lookups in strncmp_iw_with_mode
60841 This patch adds support for wild template parameter list matches, similar
60842 to how ABI tags or function overloads are now handled.
60844 With this patch, users will be able to "gloss over" the details of matching
60845 template parameter lists. This is accomplished by adding (yet more) logic
60846 to strncmp_iw_with_mode to skip parameter lists if none is explicitly given
60849 Here's a simple example using gdb.linespec/cpls-ops.exp:
60853 (gdb) ptype test_op_call
60854 type = struct test_op_call {
60856 void operator()(void);
60857 void operator()(int);
60858 void operator()(long);
60859 void operator()<int>(int *);
60861 (gdb) b test_op_call::operator()
60862 Breakpoint 1 at 0x400583: test_op_call::operator(). (3 locations)
60864 Num Type Disp Enb Address What
60865 1 breakpoint keep y <MULTIPLE>
60866 1.1 y 0x400583 in test_op_call::operator()(int)
60868 1.2 y 0x40058e in test_op_call::operator()()
60870 1.3 y 0x40059e in test_op_call::operator()(long)
60873 The breakpoint at test_op_call::operator()<int> was never set.
60877 (gdb) b test_op_call::operator()
60878 Breakpoint 1 at 0x400583: test_op_call::operator(). (4 locations)
60880 Num Type Disp Enb Address What
60881 1 breakpoint keep y <MULTIPLE>
60882 1.1 y 0x400583 in test_op_call::operator()(int)
60884 1.2 y 0x40058e in test_op_call::operator()()
60886 1.3 y 0x40059e in test_op_call::operator()(long)
60888 1.4 y 0x4008d0 in test_op_call::operator()<int>(int*)
60891 Similar to how scope lookups work, passing "-qualified" to the break command
60892 will cause a literal lookup of the symbol. In the example immediately above,
60893 this will cause GDB to only find the three non-template functions.
60895 2022-02-25 Keith Seitz <keiths@redhat.com>
60897 Unit tests for strncmp_iw_with_mode
60898 This patch attempts to make a start at adding unit tests for
60899 strncmp_iw_with_mode. While there is quite a bit of testing
60900 of this function in other tests, these are currently end-to-end
60903 This patch attempts to cover the basics of string matching, white
60904 space, C++ ABI tags, and several other topics. However, one area
60905 that is ostensibly missing is testing the `match_for_lcd' feature.
60906 This is otherwise tested as part of our end-to-end DejaGNU-based
60909 2022-02-25 Keith Seitz <keiths@redhat.com>
60911 Move find_toplevel_char to cp-support.[ch]
60912 find_toplevel_char is being used more and more outside of linespec.c, so
60913 this patch moves it into cp-support.[ch].
60915 2022-02-25 GDB Administrator <gdbadmin@sourceware.org>
60917 Automatic date update in version.in
60919 2022-02-24 Tom Tromey <tom@tromey.com>
60920 Andrew Burgess <aburgess@redhat.com>
60922 Fix crash in Fortran code
60923 PR fortran/28801 points out a gdb crash that can be provoked by
60924 certain Fortran code. The bug is that f77_get_upperbound assumes the
60925 property is either a constant or undefined, but in this case it is
60928 This patch fixes the crash by making this function (and the
60929 lower-bound one as well) do the correct check before calling
60932 Thanks to Andrew for writing the test case.
60934 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28801
60936 2022-02-24 John Baldwin <jhb@FreeBSD.org>
60938 Revert "do_target_wait_1: Clear TARGET_WNOHANG if the target isn't async."
60939 Commit 14b3360508b1 ("do_target_wait_1: Clear
60940 TARGET_WNOHANG if the target isn't async.") broke some multi-target
60941 tests, such as gdb.multi/multi-target-info-inferiors.exp. The symptom
60942 is that execution just hangs at some point. What happens is:
60944 1. One remote inferior is started, and now sits stopped at a breakpoint.
60945 It is not "async" at this point (but it "can async").
60947 2. We run a native inferior, the event loop gets woken up by the native
60950 3. In do_target_wait, we randomly choose an inferior to call target_wait
60951 on first, it happens to be the remote inferior.
60953 4. Because the target is currently not "async", we clear
60954 TARGET_WNOHANG, resulting in synchronous wait. We therefore block
60957 #0 0x00007fe9540dbb4d in select () from /usr/lib/libc.so.6
60958 #1 0x000055fc7e821da7 in gdb_select (n=15, readfds=0x7ffdb77c1fb0, writefds=0x0, exceptfds=0x7ffdb77c2050, timeout=0x7ffdb77c1f90) at /home/simark/src/binutils-gdb/gdb/posix-hdep.c:31
60959 #2 0x000055fc7ddef905 in interruptible_select (n=15, readfds=0x7ffdb77c1fb0, writefds=0x0, exceptfds=0x7ffdb77c2050, timeout=0x7ffdb77c1f90) at /home/simark/src/binutils-gdb/gdb/event-top.c:1134
60960 #3 0x000055fc7eda58e4 in ser_base_wait_for (scb=0x6250002e4100, timeout=1) at /home/simark/src/binutils-gdb/gdb/ser-base.c:240
60961 #4 0x000055fc7eda66ba in do_ser_base_readchar (scb=0x6250002e4100, timeout=-1) at /home/simark/src/binutils-gdb/gdb/ser-base.c:365
60962 #5 0x000055fc7eda6ff6 in generic_readchar (scb=0x6250002e4100, timeout=-1, do_readchar=0x55fc7eda663c <do_ser_base_readchar(serial*, int)>) at /home/simark/src/binutils-gdb/gdb/ser-base.c:444
60963 #6 0x000055fc7eda718a in ser_base_readchar (scb=0x6250002e4100, timeout=-1) at /home/simark/src/binutils-gdb/gdb/ser-base.c:471
60964 #7 0x000055fc7edb1ecd in serial_readchar (scb=0x6250002e4100, timeout=-1) at /home/simark/src/binutils-gdb/gdb/serial.c:393
60965 #8 0x000055fc7ec48b8f in remote_target::readchar (this=0x617000038780, timeout=-1) at /home/simark/src/binutils-gdb/gdb/remote.c:9446
60966 #9 0x000055fc7ec4da82 in remote_target::getpkt_or_notif_sane_1 (this=0x617000038780, buf=0x6170000387a8, forever=1, expecting_notif=1, is_notif=0x7ffdb77c24f0) at /home/simark/src/binutils-gdb/gdb/remote.c:9928
60967 #10 0x000055fc7ec4f045 in remote_target::getpkt_or_notif_sane (this=0x617000038780, buf=0x6170000387a8, forever=1, is_notif=0x7ffdb77c24f0) at /home/simark/src/binutils-gdb/gdb/remote.c:10037
60968 #11 0x000055fc7ec354d4 in remote_target::wait_ns (this=0x617000038780, ptid=..., status=0x7ffdb77c33c8, options=...) at /home/simark/src/binutils-gdb/gdb/remote.c:8147
60969 #12 0x000055fc7ec38aa1 in remote_target::wait (this=0x617000038780, ptid=..., status=0x7ffdb77c33c8, options=...) at /home/simark/src/binutils-gdb/gdb/remote.c:8337
60970 #13 0x000055fc7f1409ce in target_wait (ptid=..., status=0x7ffdb77c33c8, options=...) at /home/simark/src/binutils-gdb/gdb/target.c:2612
60971 #14 0x000055fc7e19da98 in do_target_wait_1 (inf=0x617000038080, ptid=..., status=0x7ffdb77c33c8, options=...) at /home/simark/src/binutils-gdb/gdb/infrun.c:3636
60972 #15 0x000055fc7e19e26b in operator() (__closure=0x7ffdb77c2f90, inf=0x617000038080) at /home/simark/src/binutils-gdb/gdb/infrun.c:3697
60973 #16 0x000055fc7e19f0c4 in do_target_wait (ecs=0x7ffdb77c33a0, options=...) at /home/simark/src/binutils-gdb/gdb/infrun.c:3716
60974 #17 0x000055fc7e1a31f7 in fetch_inferior_event () at /home/simark/src/binutils-gdb/gdb/infrun.c:4061
60976 Before the aforementioned commit, we would not have cleared
60977 TARGET_WNOHANG, the remote target's wait would have returned nothing,
60978 and we would have consumed the native target's event.
60980 After applying this revert, the testsuite state looks as good as before
60981 for me on Ubuntu 20.04 amd64.
60983 Change-Id: Ic17a1642935cabcc16c25cb6899d52e12c2f5c3f
60985 2022-02-24 Andrew Burgess <aburgess@redhat.com>
60987 gdb: use a range based for loop when iterating over an array
60988 Make use of a range based for loop to iterate over a static global
60989 array, removing the need to have a null entry at the end of the
60992 There should be no user visible changes after this commit.
60994 2022-02-24 Dominique Quatravaux <dominique.quatravaux@epfl.ch>
60995 Louis-He <1726110778@qq.com>
60996 Philippe Blain <levraiphilippeblain@gmail.com>
60998 gdb/darwin: skip over WIFSTOPPED wait4 status
60999 On modern Darwin's, there appears to be a new circumstance in which a
61000 MACH_NOTIFY_DEAD_NAME message can be received, and which was not
61001 previously accounted for: to signal the WIFSTOPPED condition in the
61002 debuggee. In that case the debuggee is not dead yet (and in fact,
61003 counting it as dead would cause a zombie leak - A process in such a
61004 state reparents to PID 1, but cannot be killed).
61006 - Read and ignore such messages (counting on the next exception message
61007 to let us know of the inferior's new state again)
61008 - Refactor logging so as to clearly distinguish between the
61009 MACH_NOTIFY_DEAD_NAME cases (WIFEXITED, WIFSTOPPED, signal, or
61010 something else), and warn in the last case
61012 Change-Id: Ie86904a894e9bd154e6b674b1bfbfbaee7fde3e1
61014 2022-02-24 Simon Marchi <simon.marchi@polymtl.ca>
61016 gdb/linux-tdep: move "Perms" column right
61017 Commit 29ef4c0699e1 ("gdb/linux-tdep.c: Add Perms to the 'info proc
61018 mappings' output") has broken test gdb.base/info-proc.exp on Linux,
61019 because it changes the output of "info proc mappings" in a way that the
61020 test does not expect (my bad for not testing before pushing).
61022 I looked at how FreeBSD handles this, since I remembered it did show
61023 permission flags. It looks like this:
61025 Start Addr End Addr Size Offset Flags File
61026 0x200000 0x243000 0x43000 0x0 r-- CN-- /usr/local/bin/tmux
61028 (I think that `Flags` and the flags not being aligned is not
61031 The test passes on FreeBSD, because the test looks for four hex numbers
61032 in a row and ignores the rest:
61034 ".*Mapped address spaces:.*${hex}${ws}${hex}${ws}${hex}${ws}${hex}.*"
61036 I suggest fixing it on Linux by moving the flags column to the same
61037 place as in the FreeBSD output. It makes things a bit more consistent
61038 between OSes, and we don't have to touch the test.
61040 At the same time, make use of the actual length of the permission's
61041 string to specify the number of characters to print.
61043 Before this patch, the output looks like:
61045 Start Addr End Addr Perms Size Offset objfile
61046 0x55dd4b544000 0x55dd4b546000 r--p 0x2000 0x0 /usr/bin/sleep
61048 and after, it looks like:
61050 Start Addr End Addr Size Offset Perms objfile
61051 0x5622ae662000 0x5622ae664000 0x2000 0x0 r--p /usr/bin/sleep
61053 Change-Id: If0fc167b010b25f97a3c54e2f491df4973ccde8f
61055 2022-02-24 Simon Marchi <simon.marchi@polymtl.ca>
61057 gdb/linux-tdep: make read_mapping return a structure
61058 Change read_mapping to return a structure instead of taking many output
61059 parameters. Change the string + length output parameters (permissions
61060 and device) to be gdb::string_view, since that's what string_view is
61061 for (a non-NULL terminated view on a string). No changes in behavior
61064 Change-Id: I86e627d84d3dda8c9b835592b0f4de8d90d12112
61066 2022-02-24 GDB Administrator <gdbadmin@sourceware.org>
61068 Automatic date update in version.in
61070 2022-02-23 Tom Tromey <tromey@adacore.com>
61072 Fix bug in C++ overload resolution
61073 PR c++/28901 points out a bug in C++ overload resolution. When
61074 comparing two overloads, one might be better than the other for
61075 certain parameters -- but, if that one also has some invalid
61076 conversion, then it should never be considered the better choice.
61077 Instead, a valid-but-not-apparently-quite-as-good overload should be
61080 This patch fixes this problem by changing how overload comparisons are
61081 done. I don't believe it should affect any currently valid overload
61082 resolution; nor should it affect resolutions where all the choices are
61085 2022-02-23 Dominik 'Disconnect3d' Czarnota <dominik.b.czarnota@gmail.com>
61087 gdb/linux-tdep.c: Add Perms to the 'info proc mappings' output
61088 Fixes #28914 and so it adds a 'Perms' (permissions) column to the
61089 'info proc mappings' command output. This will allow users to know
61090 the memory pages permissions right away from GDB instead of having
61091 to fetch them from the /proc/$pid/maps file (which is also what GDB
61092 does internally, but it just did not print that column).
61094 Below I am also showing how an example output looks like before and
61095 after this commit in case someone wonders.
61097 On i386 targets - before this commit:
61099 (gdb) info proc mappings
61101 Mapped address spaces:
61103 Start Addr End Addr Size Offset objfile
61104 0x56555000 0x56556000 0x1000 0x0 /home/dc/src/binutils-gdb/build/a.out
61105 0x56556000 0x56557000 0x1000 0x1000 /home/dc/src/binutils-gdb/build/a.out
61106 0x56557000 0x56558000 0x1000 0x2000 /home/dc/src/binutils-gdb/build/a.out
61107 0x56558000 0x5655a000 0x2000 0x2000 /home/dc/src/binutils-gdb/build/a.out
61108 0xf7fc4000 0xf7fc8000 0x4000 0x0 [vvar]
61109 0xf7fc8000 0xf7fca000 0x2000 0x0 [vdso]
61110 0xf7fca000 0xf7fcb000 0x1000 0x0 /usr/lib/i386-linux-gnu/ld-2.33.so
61111 0xf7fcb000 0xf7fee000 0x23000 0x1000 /usr/lib/i386-linux-gnu/ld-2.33.so
61112 0xf7fee000 0xf7ffb000 0xd000 0x24000 /usr/lib/i386-linux-gnu/ld-2.33.so
61113 0xf7ffb000 0xf7ffe000 0x3000 0x30000 /usr/lib/i386-linux-gnu/ld-2.33.so
61114 0xfffdc000 0xffffe000 0x22000 0x0 [stack]
61118 On i386 targets - after this commit:
61120 (gdb) info proc mappings
61122 Mapped address spaces:
61124 Start Addr End Addr Perms Size Offset objfile
61125 0x56555000 0x56556000 r--p 0x1000 0x0 /home/dc/src/binutils-gdb/build/a.out
61126 0x56556000 0x56557000 r-xp 0x1000 0x1000 /home/dc/src/binutils-gdb/build/a.out
61127 0x56557000 0x56558000 r--p 0x1000 0x2000 /home/dc/src/binutils-gdb/build/a.out
61128 0x56558000 0x5655a000 rw-p 0x2000 0x2000 /home/dc/src/binutils-gdb/build/a.out
61129 0xf7fc4000 0xf7fc8000 r--p 0x4000 0x0 [vvar]
61130 0xf7fc8000 0xf7fca000 r-xp 0x2000 0x0 [vdso]
61131 0xf7fca000 0xf7fcb000 r--p 0x1000 0x0 /usr/lib/i386-linux-gnu/ld-2.33.so
61132 0xf7fcb000 0xf7fee000 r-xp 0x23000 0x1000 /usr/lib/i386-linux-gnu/ld-2.33.so
61133 0xf7fee000 0xf7ffb000 r--p 0xd000 0x24000 /usr/lib/i386-linux-gnu/ld-2.33.so
61134 0xf7ffb000 0xf7ffe000 rw-p 0x3000 0x30000 /usr/lib/i386-linux-gnu/ld-2.33.so
61135 0xfffdc000 0xffffe000 rw-p 0x22000 0x0 [stack]
61139 On amd64 targets - after this commit:
61141 (gdb) info proc mappings
61143 Mapped address spaces:
61145 Start Addr End Addr Perms Size Offset objfile
61146 0x555555554000 0x555555555000 r--p 0x1000 0x0 /home/dc/src/binutils-gdb/build/a.out
61147 0x555555555000 0x555555556000 r-xp 0x1000 0x1000 /home/dc/src/binutils-gdb/build/a.out
61148 0x555555556000 0x555555557000 r--p 0x1000 0x2000 /home/dc/src/binutils-gdb/build/a.out
61149 0x555555557000 0x555555559000 rw-p 0x2000 0x2000 /home/dc/src/binutils-gdb/build/a.out
61150 0x7ffff7fc3000 0x7ffff7fc7000 r--p 0x4000 0x0 [vvar]
61151 0x7ffff7fc7000 0x7ffff7fc9000 r-xp 0x2000 0x0 [vdso]
61152 0x7ffff7fc9000 0x7ffff7fca000 r--p 0x1000 0x0 /usr/lib/x86_64-linux-gnu/ld-2.33.so
61153 0x7ffff7fca000 0x7ffff7ff1000 r-xp 0x27000 0x1000 /usr/lib/x86_64-linux-gnu/ld-2.33.so
61154 0x7ffff7ff1000 0x7ffff7ffb000 r--p 0xa000 0x28000 /usr/lib/x86_64-linux-gnu/ld-2.33.so
61155 0x7ffff7ffb000 0x7ffff7fff000 rw-p 0x4000 0x31000 /usr/lib/x86_64-linux-gnu/ld-2.33.so
61156 0x7ffffffdd000 0x7ffffffff000 rw-p 0x22000 0x0 [stack]
61157 0xffffffffff600000 0xffffffffff601000 --xp 0x1000 0x0 [vsyscall]
61161 Change-Id: I4991f6cc758cd532eae3ae98c29d22e7bd9d9c36
61163 2022-02-23 Patrick O'Neill <patrick@rivosinc.com>
61165 RISC-V: PR28733, add missing extension info to 'unrecognized opcode' error
61166 Currently we report errors as "unrecognized opcode `fence.i'" when the
61167 opcode isn't part of the selected extensions.
61168 This patch expands that error message to include the missing extension
61169 information. For example, now the error message would be "unrecognized
61170 opcode `fence.i', extension `zifencei' required".
61171 If the opcode is not a part of any extension, the error message reverts
61172 to "unrecognized opcode `<op statement>'".
61177 * elfxx-riscv.c (riscv_multi_subset_supports_ext): New function,
61178 used to return the extension string for each INSN_CLASS_*.
61179 * elfxx-riscv.h: Added extern riscv_multi_subset_supports_ext.
61182 * config/tc-riscv.c (struct riscv_ip_error): New structure,
61183 contains information about errors that occur within the riscv_ip.
61184 (riscv_ip): Use struct riscv_ip_error to report more detailed errors.
61185 * testsuite/gas/riscv/c-fld-fsd-fail.l: Updated.
61186 * testsuite/gas/riscv/march-imply-i2p1-01.: Likewise.
61188 2022-02-23 Patrick O'Neill <patrick@rivosinc.com>
61190 RISC-V: PR28733, add missing extension info to 'invalid CSR' error
61191 Currently we report errors as "invalid CSR 'fscr' for the current ISA"
61192 when the instruction isn't valid.
61194 This patch expands that error message to include the missing extension
61195 information. For example, now the error message would be "invalid CSR
61196 'fscr' for the current ISA, CSR 'fscr' needs 'f' extension".
61201 * config/tc-riscv.c (riscv_csr_address): Report more details
61202 when the CSR is invalid.
61203 * testsuite/gas/riscv/csr-version-1p10.l: Updated detailed errors.
61204 * testsuite/gas/riscv/csr-version-1p11.l: Likewise.
61205 * testsuite/gas/riscv/csr-version-1p12.l: Likewise.
61206 * testsuite/gas/riscv/csr-version-1p9p1.l: Likewise.
61208 2022-02-23 Alan Modra <amodra@gmail.com>
61210 binutils 2.38 vs. ppc32 linux kernel
61211 Commit b25f942e18d6 made .machine more strict. Weaken it again.
61213 * config/tc-ppc.c (ppc_machine): Treat an early .machine specially,
61214 keeping sticky options to work around gcc bugs.
61216 2022-02-23 Nelson Chu <nelson.chu@sifive.com>
61218 RISC-V: Updated CSRs to privileged spec v1.12 and debug spec v1.0.
61219 * Removed N extension CSRs,
61220 ustatus, uie, utvec, uscratch, uepc, ucause, utval and uip.
61222 * Removed two supervisor CSRs,
61223 sedeleg and sideleg.
61225 * Changed debug CSR address of scontext from 0x7aa to 0x5a8. We cannot support
61226 different versions of debug specs for now, so only supporting the latest one is
61227 the only way to move forward.
61229 * Added debug CSRs,
61230 mscontext (0x7aa), mcontrol6 (0x7a1, tdata1) and tmexttrigger ((0x7a1, tdata1).
61232 * Regarded hcontext as a debug CSR.
61235 * opcode/riscv-opc.h: Updated CSRs to privileged spec v1.12 and
61238 * testsuite/gas/riscv/csr.s: Updated CSRs to privileged spec v1.12
61239 and debug spec v1.0.
61240 * testsuite/gas/riscv/csr-dw-regnums.d: Likewise.
61241 * testsuite/gas/riscv/csr-version-1p10.d: Likewise.
61242 * testsuite/gas/riscv/csr-version-1p10.l: Likewise.
61243 * testsuite/gas/riscv/csr-version-1p11.d: Likewise.
61244 * testsuite/gas/riscv/csr-version-1p11.l: Likewise.
61245 * testsuite/gas/riscv/csr-version-1p12.d: Likewise.
61246 * testsuite/gas/riscv/csr-version-1p12.l: Likewise.
61247 * testsuite/gas/riscv/csr-version-1p9p1.d: Likewise.
61248 * testsuite/gas/riscv/csr-version-1p9p1.l: Likewise.
61249 * testsuite/gas/riscv/csr-dw-regnums.d: Likewise.
61250 * testsuite/gas/riscv/csr-dw-regnums.s: Likewise.
61252 2022-02-23 Tsukasa OI <research_trasio@irq.a4lg.com>
61254 RISC-V: Add Privileged Architecture 1.12 CSRs
61257 * Most of CSRs as listed in the Privileged Architecture,
61258 version 1.12 (except scontext and mscontext).
61260 * Testcases for most CSRs added on the Privileged
61261 Architecture, version 1.12 (except moved "scontext" and
61266 * opcode/riscv-opc.h (CSR_SENVCFG, CSR_MCONFIGPTR, CSR_MENVCFG,
61267 CSR_MSTATUSH, CSR_MENVCFGH, CSR_MTINST, CSR_MTVAL2, CSR_MSECCFG,
61268 CSR_MSECCFGH, CSR_PMPCFG4, CSR_PMPCFG5, CSR_PMPCFG6,
61269 CSR_PMPCFG7, CSR_PMPCFG8, CSR_PMPCFG9, CSR_PMPCFG10,
61270 CSR_PMPCFG11, CSR_PMPCFG12, CSR_PMPCFG13, CSR_PMPCFG14,
61271 CSR_PMPCFG15, CSR_PMPADDR16, CSR_PMPADDR17, CSR_PMPADDR18,
61272 CSR_PMPADDR19, CSR_PMPADDR20, CSR_PMPADDR21, CSR_PMPADDR22,
61273 CSR_PMPADDR23, CSR_PMPADDR24, CSR_PMPADDR25, CSR_PMPADDR26,
61274 CSR_PMPADDR27, CSR_PMPADDR28, CSR_PMPADDR29, CSR_PMPADDR30,
61275 CSR_PMPADDR31, CSR_PMPADDR32, CSR_PMPADDR33, CSR_PMPADDR34,
61276 CSR_PMPADDR35, CSR_PMPADDR36, CSR_PMPADDR37, CSR_PMPADDR38,
61277 CSR_PMPADDR39, CSR_PMPADDR40, CSR_PMPADDR41, CSR_PMPADDR42,
61278 CSR_PMPADDR43, CSR_PMPADDR44, CSR_PMPADDR45, CSR_PMPADDR46,
61279 CSR_PMPADDR47, CSR_PMPADDR48, CSR_PMPADDR49, CSR_PMPADDR50,
61280 CSR_PMPADDR51, CSR_PMPADDR52, CSR_PMPADDR53, CSR_PMPADDR54,
61281 CSR_PMPADDR55, CSR_PMPADDR56, CSR_PMPADDR57, CSR_PMPADDR58,
61282 CSR_PMPADDR59, CSR_PMPADDR60, CSR_PMPADDR61, CSR_PMPADDR62,
61283 CSR_PMPADDR63): New CSR macros.
61287 * testsuite/gas/riscv/csr-dw-regnums.s: Add new CSRs.
61288 * testsuite/gas/riscv/csr-dw-regnums.d: Likewise.
61289 * testsuite/gas/riscv/csr.s: Add new CSRs.
61290 * testsuite/gas/riscv/csr-version-1p9p1.d: Likewise.
61291 * testsuite/gas/riscv/csr-version-1p9p1.l: Likewise.
61292 * testsuite/gas/riscv/csr-version-1p10.d: Likewise.
61293 * testsuite/gas/riscv/csr-version-1p10.l: Likewise.
61294 * testsuite/gas/riscv/csr-version-1p11.d: Likewise.
61295 * testsuite/gas/riscv/csr-version-1p11.l: Likewise.
61296 * testsuite/gas/riscv/csr-version-1p12.d: Likewise.
61297 * testsuite/gas/riscv/csr-version-1p12.l: Likewise.
61299 2022-02-23 Tsukasa OI <research_trasio@irq.a4lg.com>
61301 RISC-V: Reorganize testcases for CFI directives
61302 This commit reorganizes and adds some CSRs to csr-dw-regnums.[sd] to
61303 make it test the same CSRs as csr.s.
61307 * testsuite/gas/riscv/csr-dw-regnums.s: Reorganize and add
61308 defined CSRs tested in csr.s.
61309 * testsuite/gas/riscv/csr-dw-regnums.d: Likewise.
61311 2022-02-23 GDB Administrator <gdbadmin@sourceware.org>
61313 Automatic date update in version.in
61315 2022-02-22 John Baldwin <jhb@FreeBSD.org>
61317 NEWS: Note that the FreeBSD async target supports async mode.
61319 2022-02-22 John Baldwin <jhb@FreeBSD.org>
61321 inf-ptrace: Add an event_pipe to be used for async mode in subclasses.
61322 Subclasses of inf_ptrace_target have to opt-in to using the event_pipe
61323 by implementing the can_async_p and async methods. For subclasses
61324 which do this, inf_ptrace_target provides is_async_p, async_wait_fd
61325 and closes the pipe in the close target method.
61327 inf_ptrace_target also provides wrapper routines around the event pipe
61328 (async_file_open, async_file_close, async_file_flush, and
61329 async_file_mark) for use in target methods such as async.
61330 inf_ptrace_target also exports a static async_file_mark_if_open
61331 function which can be used in SIGCHLD signal handlers.
61333 2022-02-22 John Baldwin <jhb@FreeBSD.org>
61335 Enable async mode in the target in attach_cmd.
61336 If the attach target supports async mode, enable it after the
61337 attach target's ::attach method returns.
61339 fbsd-nat: Return nullptr rather than failing ::thread_name.
61340 ptrace on FreeBSD cannot be used against running processes and instead
61341 fails with EBUSY. This meant that 'info threads' would fail if any of
61342 the threads were running (for example when using schedule-multiple=on
61343 in gdb.base/fork-running-state.exp). Instead of throwing errors, just
61344 return nullptr as no thread name is better than causing info threads to
61347 2022-02-22 John Baldwin <jhb@FreeBSD.org>
61349 fbsd-nat: Various cleanups to the ::resume entry debug message.
61350 Move the message from 'show debug fbsd-lwp' to 'show debug fbsd-nat'
61351 since it is helpful for debugging async target support and not just
61354 Use target_pid_to_str to format the ptid and log the step and signo
61357 2022-02-22 John Baldwin <jhb@FreeBSD.org>
61359 fbsd-nat: Include ptrace operation in error messages.
61361 2022-02-22 John Baldwin <jhb@FreeBSD.org>
61363 fbsd-nat: Implement async target support.
61364 This is a fairly simple version of async target support.
61366 Synchronous mode still uses blocking waitpid() calls in
61367 inf_ptrace::wait() unlike the Linux native target which always uses
61368 WNOHANG and uses sigsuspend() for synchronous operation.
61370 Asynchronous mode registers an event pipe with the core as a file
61371 handle and writes to the pipe when SIGCHLD is raised. TARGET_WNOHANG
61372 is handled by inf_ptrace::wait().
61374 2022-02-22 John Baldwin <jhb@FreeBSD.org>
61376 inf-ptrace: Support async targets in inf_ptrace_target::wait.
61377 - Handle TARGET_WNOHANG by passing WNOHANG to waitpid and returning
61378 TARGET_WAITKIND_IGNORE if there are no events to report.
61380 - Handle a race in async mode where SIGCHLD might signal the event
61381 pipe for an event that has already been reported. If the event was
61382 the exit of the last child process, waitpid() will fail with ECHILD
61383 rather than returning a pid of 0. For this case, return
61384 TARGET_WAITKIND_NO_RESUMED.
61386 2022-02-22 John Baldwin <jhb@FreeBSD.org>
61388 inf-ptrace: Return an IGNORE event if waitpid() fails.
61389 Previously this returned a TARGET_WAITKIND_SIGNALLED event for
61390 inferior_ptid. However, inferior_ptid is invalid during ::wait()
61391 methods after the multi-target changes, so this was triggering an
61392 assertion further up the stack.
61394 do_target_wait_1: Clear TARGET_WNOHANG if the target isn't async.
61395 Previously, TARGET_WNOHANG was cleared if a target supported async
61396 mode even if async mode wasn't currently enabled. This change only
61397 permits TARGET_WNOHANG if async mode is enabled.
61399 2022-02-22 John Baldwin <jhb@FreeBSD.org>
61401 Don't enable async mode at the end of target ::resume methods.
61402 Now that target_resume always enables async mode after target::resume
61403 returns, these calls are redundant.
61405 The other place that target resume methods are invoked outside of
61406 target_resume are as the beneath target in record_full_wait_1. In
61407 this case, async mode should already be enabled when supported by the
61408 target before the resume method is invoked due to the following:
61410 In general, targets which support async mode run as async until
61411 ::wait returns TARGET_WAITKIND_NO_RESUMED to indicate that there are
61412 no unwaited for children (either they have exited or are stopped).
61413 When that occurs, the loop in wait_one disables async mode. Later
61414 if a stopped child is resumed, async mode is re-enabled in
61415 do_target_resume before waiting for the next event.
61417 In the case of record_full_wait_1, this function is invoked from the
61418 ::wait target method when fetching an event. If the underlying
61419 target supports async mode, then an earlier call to do_target_resume
61420 to resume the child reporting an event in the loop in
61421 record_full_wait_1 would have already enabled async mode before
61422 ::wait was invoked. In addition, nothing in the code executed in
61423 the loop in record_full_wait_1 disables async mode. Async mode is
61424 only disabled higher in the call stack in wait_one after ::wait
61427 It is also true that async mode can be disabled by an
61428 INF_EXEC_COMPLETE event passed to inferior_event_handle, but all of
61429 the places that invoke that are in the gdb core which is "above" a
61430 target ::wait method.
61432 Note that there is an earlier call to enable async mode in
61433 linux_nat_target::resume. That call also marks the async event pipe
61434 to report an existing event after enabling async mode, so it needs to
61437 2022-02-22 John Baldwin <jhb@FreeBSD.org>
61439 Enable async mode on supported targets in target_resume.
61440 Enabling async mode above the target layer removes duplicate code in
61441 ::resume methods of async-capable targets. Commit 5b6d1e4fa4f
61442 ("Multi-target support") enabled async mode in do_target_resume after
61443 target_resume returns which is a step in this direction. However,
61444 other callers of target_resume such as target_continue do not enable
61445 async mode. Rather than enabling async mode in each of the callers
61446 after target_resume returns, enable async mode at the end of
61449 gdbserver linux-low: Convert linux_event_pipe to the event_pipe class.
61450 Use event_pipe from gdbsupport in place of the existing file
61453 gdb linux-nat: Convert linux_nat_event_pipe to the event_pipe class.
61454 Use event_pipe from gdbsupport in place of the existing file
61457 2022-02-22 John Baldwin <jhb@FreeBSD.org>
61459 gdbsupport: Add an event-pipe class.
61460 This pulls out the implementation of an event pipe used to implement
61461 target async support in both linux-low.cc (gdbserver) and linux-nat.c
61464 This will be used to replace the existing event pipe in linux-low.cc
61465 and linux-nat.c in future commits.
61467 Co-Authored-By: Lancelot SIX <lsix@lancelotsix.com>
61469 2022-02-22 Ruslan Kabatsayev <b7.10110111@gmail.com>
61471 gdb: fix detection of compilation and linking flags for source-highlight
61472 Currently there are two problems with the detection of
61473 source-highlight via pkg-config in GDB's configure script:
61475 1. The LDFLAGS variable is used to pass the 'pkg-config --libs' output
61476 to AC_LINK_IFELSE, which results in the "-L/some/path
61477 -lsource-highlight" preceding the conftest.cpp, which can result in a
61478 failure to find symbols referenced in conftest.cpp, if the linker is
61479 using --as-needed by default.
61481 2. The CFLAGS variable is used to pass the 'pkg-config --cflags'
61482 output to AC_LINK_IFELSE. However, as the current language is C++,
61483 AC_LINK_IFELSE will actuall use CXXFLAGS, not CFLAGS, so any flags
61484 returned from pkg-config will not be seen.
61486 This patch fixes both of these mistakes, allowing GDB to correctly
61487 configure and build using source-highlight installed into a custom
61488 prefix, e.g. ~/opt/gdb-git (because the system version of
61489 source-highlight is too old).
61491 2022-02-22 Philippe Blain <levraiphilippeblain@gmail.com>
61493 gdb/testsuite/README: point to default value of INTERNAL_GDBFLAGS
61494 The INTERNAL_GDBFLAGS runtest variable was updated in 55c3ad88013
61495 ([gdb/testsuite] Prevent pagination in GDB_INTERNALFLAGS, 2020-10-26) to
61496 disable pagination, and in aae1c79a03a (PR python/12227..., 2010-12-07)
61497 to point to the data directory, but its default value mentioned in the
61498 testsuite's README was not kept up to date.
61500 To avoid it getting out of sync even more, point the reader to the
61501 definition of the variable in lib/gdb.exp, and move the explanation of
61502 the different flags there. Also adjust the example in the README
61503 so it follows the flags added in 55c3ad88013.
61505 Change-Id: I3533608a7d6ae5198af09c7dc7743bde24c19ed7
61507 2022-02-22 Kito Cheng <kito.cheng@sifive.com>
61509 RISC-V: Maintain a string to hold the canonical order
61510 Using dummy entry in riscv_supported_std_ext cause confusing and wrongly
61511 support `b` and `k` extensions.
61514 * elfxx-riscv.c (riscv_supported_std_ext): Drop unsupported
61516 (riscv_ext_canonical_order): New.
61517 (riscv_init_ext_order): Use riscv_ext_canonical_order rather
61518 than riscv_supported_std_ext to compute canonical order.
61522 - Use `*ext` rather than `*ext != NULL` for checking is reach end of
61525 2022-02-22 GDB Administrator <gdbadmin@sourceware.org>
61527 Automatic date update in version.in
61529 2022-02-21 Alan Modra <amodra@gmail.com>
61531 Re: ld: Support customized output section type
61532 "DO NOT EDIT!" says the comment at the top of bfd-in2.h. Move the new
61533 type field where it belongs.
61536 * section.c (struct bfd_section): Add type. Formatting.
61537 (BFD_FAKE_SECTION): Formatting.
61538 * bfd-in2.h: Regenerate.
61540 2022-02-21 Mike Frysinger <vapier@gentoo.org>
61542 sim: gdbinit: hoist setup to common code
61543 This was left in subdirs because of the dynamic cgen usage. However,
61544 we can move this breakpoint call to runtime and let gdb detect whether
61547 2022-02-21 Andrew Burgess <aburgess@redhat.com>
61549 gdb/testsuite: relax pattern in new gdb.mi/mi-multi-commands.exp test
61550 I saw some failures in the test gdb.mi/mi-multi-commands.exp that I
61551 added recently. This test was added in commit:
61553 commit d08cbc5d3203118da5583296e49273cf82378042
61554 Date: Wed Dec 22 12:57:44 2021 +0000
61556 gdb: unbuffer all input streams when not using readline
61558 The failures I see only occurred when my machine was very heavily
61561 In this test I send multiple commands from dejagnu to gdb with a
61562 single send_gdb call. In a well behaving world what I want to happen
61563 is that the gdb console sees both commands arrive and echos the text
61564 of those commands. Then gdb starts processing the first command,
61565 prints the result, and then processes the second command, and prints
61568 However, what I saw in my loaded environment was that only after
61569 sending the two commands, only the first command was echoed to gdb's
61570 terminal. Then gdb started processing the first command, and started
61571 to write the output. Now, mixed in with the first command output, the
61572 second command was echoed to gdb's terminal. Finally, gdb would
61573 finish printing the first command output, and would read and handle
61574 the second command.
61576 This mixing of command echoing with the first command output was
61577 causing the test matching patterns to fail.
61579 In this commit I change the command I use in the test from a CLI
61580 command to an MI command, this reduces the number of lines of output
61581 that come from the test, CLI commands sent through the MI interpreter
61582 are echoed back like this:
61585 set $a = "FIRST COMMAND"
61586 &"set $a = \"FIRST COMMAND\"\n"
61590 While this is not the case for true MI command:
61593 -data-evaluate-expression $a
61594 ^done,value="\"FIRST COMMAND\""
61597 Less output makes for simpler patterns to match against.
61599 Next, when sending two command to gdb I was previously trying to spot
61600 the output of the first command followed by the prompt with nothing
61601 between. This is not really needed, for the first command I can look
61602 for just the ^done,value="\"FIRST COMMAND\"" string, then I can start
61603 looking for the output of the second command.
61605 So long as the second pattern matches up to the gdb prompt, then I can
61606 be sure than nothing is left over in the expect buffer to muck up
61609 As to see the second command output gdb must have read in the second
61610 command, the second command output never suffers from the corruption
61611 that the first command output does.
61613 Since making this change, I've not seen a failure in this test.
61615 2022-02-21 Andrew Burgess <aburgess@redhat.com>
61617 gdb: avoid nullptr access in dbxread.c from read_dbx_symtab
61618 This fixes a GDB crash reported in bug pr/28900, related to reading in
61619 some stabs debug information.
61621 In this commit my goal is to stop GDB crashing. I am not trying to
61622 ensure that GDB makes the best possible use of the available stabs
61623 debug information. At this point I consider stabs a legacy debug
61624 format, with only limited support in GDB.
61626 So, the problem appears to be that, when reading in the stabs data, we
61627 need to find a N_SO entry, this is the entry that defines the start of
61628 a compilation unit (or at least the location of a corresponding source
61631 It is while handling an N_SO that GDB creates a psymtab to hold the
61632 incoming debug information (symbols, etc).
61634 The problem we hit in the bug is that we encounter some symbol
61635 information (an N_PC entry) outside of an N_SO entry - that is we find
61636 some symbol information that is not associated with any source file.
61638 We already have some protection for this case, look (in
61639 read_dbx_symtab) at the handling of N_PC entries of type 'F' and 'f',
61640 if we have no psymtab (the pst variable is nullptr) then we issue a
61641 complaint. However, for whatever reason, in both 'f' and 'F'
61642 handling, there is one place where we assume that the pst
61643 variable (the psymtab) is not nullptr. This is a mistake.
61645 In this commit, I guard these two locations (in 'f' and 'F' handling)
61646 so we no longer assume pst is not nullptr.
61648 While I was at it, I audited all the other uses of pst in
61649 read_dbx_symtab, and in every potentially dangerous case I added a
61650 nullptr check, and issue a suitable complaint if pst is found to be
61653 It might well be true that we could/should do something smarter if we
61654 see a debug symbol outside of an N_SO entry, and if anyone wanted to
61655 do that work, they're welcome too. But this commit is just about
61656 preventing the nullptr access, and the subsequent GDB crash.
61658 I don't have any tests for this change, I have no idea how to generate
61659 weird stabs data for testing. The original binary from the bug report
61660 now loads just fine without GDB crashing.
61662 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28900
61664 2022-02-21 Andrew Burgess <aburgess@redhat.com>
61666 gdb: make use of std::string in dbxread.c and xcoffread.c
61667 While taking a look through dbxread.c I spotted a couple of places
61668 where making use of std::string would remove the need for manual
61669 memory allocation and memcpy.
61671 During review Simon pointed out that the same code exists in
61672 xcoffread.c, so I've applied the same fix there too.
61674 There should be no user visible changes after this commit.
61676 2022-02-21 GDB Administrator <gdbadmin@sourceware.org>
61678 Automatic date update in version.in
61680 2022-02-20 Lancelot SIX <lancelot.six@amd.com>
61682 gdb: Only paginate for filtered output in fputs_maybe_filtered
61683 A have had situation where a unfiltered output (done using
61684 fputs_unfiltered) ended up triggering pagination. The backtrace for this was:
61687 #24 0x000055839377ee4e in check_async_event_handlers () at ../../gdb/async-event.c:335
61688 #25 0x0000558394b67b57 in gdb_do_one_event () at ../../gdbsupport/event-loop.cc:216
61689 #26 0x0000558394587454 in gdb_readline_wrapper (prompt=0x7ffd907712d0 "--Type <RET> for more, q to quit, c to continue without paging--") at ../../gdb/top.c:1148
61690 #27 0x0000558394707270 in prompt_for_continue () at ../../gdb/utils.c:1438
61691 #28 0x00005583947088b3 in fputs_maybe_filtered (linebuffer=0x60c0000f4000 " [...quite big message...]", stream=0x60300028e9d0, filter=0) at ../../gdb/utils.c:1752
61692 #29 0x0000558394708e57 in fputs_unfiltered (linebuffer=0x60c0000f4000 " [...quite big message...]", stream=0x60300028e9d0) at ../../gdb/utils.c:1811
61695 This comes from what appears to be a oversight in fputs_maybe_filtered. This
61696 function has a FILTER parameter which if true makes the function pause after
61697 every screenful (i.e. triggers pagination).
61699 The filter parameter is correctly used to guard the first place where
61700 prompt_for_continue. There is a second place in the function which can call
61701 prompt_for_continue, but is currently unguarded. I believe that this is an
61702 oversight, this patch fixes that.
61704 Tested on Linux-x86_64, no regression observed.
61706 Change-Id: Iad8ffd50a87cf20077500878e2564b5a7dc81ece
61708 2022-02-20 GDB Administrator <gdbadmin@sourceware.org>
61710 Automatic date update in version.in
61712 2022-02-19 Dominique Quatravaux <dominique.quatravaux@epfl.ch>
61714 gdb/darwin: remove not-so-harmless spurious call to `wait4`
61715 As seen in https://sourceware.org/bugzilla/show_bug.cgi?id=24069 this
61716 code will typically wait4() a second time on the same process that was
61717 already wait4()'d a few lines above. While this used to be
61718 harmless/idempotent (when we assumed that the process already exited),
61719 this now causes a deadlock in the WIFSTOPPED case.
61721 The early (~2019) history of bug #24069 cautiously suggests to use
61722 WNOHANG instead of outright deleting the call. However, tests on the
61723 current version of Darwin (Big Sur) demonstrate that gdb runs just fine
61724 without a redundant call to wait4(), as would be expected.
61725 Notwithstanding the debatable value of conserving bug compatibility with
61726 an OS release that is more than a decade old, there is scant evidence of
61727 what that double-wait4() was supposed to achieve in the first place - A
61728 cursory investigation with `git blame` pinpoints commits bb00b29d7802
61729 and a80b95ba67e2 from the 2008-2009 era, but fails to answer the
61730 "why" question conclusively.
61732 Co-Authored-By: Philippe Blain <levraiphilippeblain@gmail.com>
61733 Change-Id: Id4e4415d66d6ff6b3552b60d761693f17015e4a0
61735 2022-02-19 GDB Administrator <gdbadmin@sourceware.org>
61737 Automatic date update in version.in
61739 2022-02-18 Tom Tromey <tromey@adacore.com>
61741 Add constructor to bound_minimal_symbol
61742 This adds a constructor to bound_minimal_symbol, to avoid a build
61743 failure with clang that Simon pointed out.
61745 I also took the opportunity to remove some redundant initializations,
61746 and to change one use of push_back to emplace_back, as suggested by
61749 2022-02-18 Roland McGrath <mcgrathr@google.com>
61751 Fix typo in ld.texi
61753 * ld.texi (Output Section Type): Fix typo in @code syntax.
61755 2022-02-18 Simon Marchi <simon.marchi@polymtl.ca>
61757 gdb: remove newlines from some linux_nat_debug_printf calls
61758 Change-Id: I80328fab7096221356864b5a4fb30858b48d2c10
61760 2022-02-18 GDB Administrator <gdbadmin@sourceware.org>
61762 Automatic date update in version.in
61764 2022-02-17 Nick Clifton <nickc@redhat.com>
61766 Updated Serbian translations for the bfd, gold, ld and opcodes directories
61768 2022-02-17 GDB Administrator <gdbadmin@sourceware.org>
61770 Automatic date update in version.in
61772 2022-02-16 Fangrui Song <maskray@google.com>
61774 ld: Support customized output section type
61777 * bfd-in2.h (struct bfd_section): Add type.
61778 (discarded_section): Add field.
61779 * elf.c (elf_fake_sections): Handle bfd_section::type.
61780 * section.c (BFD_FAKE_SECTION): Add field.
61781 * mri.c (mri_draw_tree): Update function call.
61785 * ld.texi: Document new output section type.
61786 * ldlex.l: Add new token TYPE.
61787 * ldgram.y: Handle TYPE=exp.
61788 * ldlang.h: Add type_section to list of section types.
61789 * ldlang.c (lang_add_section): Handle type_section.
61790 (map_input_to_output_sections): Handle type_section.
61791 * testsuite/ld-scripts/output-section-types.t: Add tests.
61792 * testsuite/ld-scripts/output-section-types.d: Update.
61794 2022-02-16 Andrew Burgess <aburgess@redhat.com>
61796 gdb/tui: add a missing white space character
61797 Just adds a missing space. There should be no user visible changes
61800 gdb: convert callback_handler_installed from int to bool
61801 Simple int to bool conversion on callback_handler_installed in
61802 event-top.c. There should be no user visible changes after this
61805 2022-02-16 Alan Modra <amodra@gmail.com>
61807 gas local label and dollar label handling
61808 Much of the gas source and older BFD source use "long" for function
61809 parameters and variables, when other types would be more appropriate.
61810 This patch fixes one of those cases. Dollar labels and numeric local
61811 labels do not need large numbers. Small positive itegers are usually
61812 all that is required. Due to allowing longs, it was possible for
61813 fb_label_name and dollar_label_name to overflow their buffers.
61815 * symbols.c: Delete unnecessary forward declarations.
61816 (dollar_labels, dollar_label_instances): Use unsigned int.
61817 (dollar_label_defined, dollar_label_instance): Likewise.
61818 (define_dollar_label): Likewise.
61819 (fb_low_counter, fb_labels, fb_label_instances): Likewise.
61820 (fb_label_instance_inc, fb_label_instance): Likewise.
61821 (fb_label_count, fb_label_max): Make them size_t.
61822 (dollar_label_name, fb_label_name): Rewrite using sprintf.
61823 * symbols.h (dollar_label_defined): Update prototype.
61824 (define_dollar_label, dollar_label_name): Likewise.
61825 (fb_label_instance_inc, fb_label_name): Likewise.
61826 * config/bfin-lex.l (yylex): Remove unnecessary casts.
61827 * expr.c (integer_constant): Likewise.
61828 * read.c (read_a_source_file): Limit numeric label range to int.
61830 2022-02-16 Alan Modra <amodra@gmail.com>
61832 ubsan: s_app_line integer overflow
61833 There are quite a few ubsan warnings in gas. This one disappears with
61836 * read.c (s_app_line): Rename 'l' to 'linenum'. Avoid ubsan
61839 2022-02-16 Alan Modra <amodra@gmail.com>
61841 pe_ILF_make_a_symbol_reloc segfault
61842 pei-aarch64-little apparently lacks support for BFD_RELOC_RVA.
61844 * peicode.h (pe_ILF_make_a_symbol_reloc): Don't segfault on
61847 2022-02-16 Alan Modra <amodra@gmail.com>
61849 What to do when sh_addralign isn't a power of two
61850 BFD generally doesn't handle anything but a power of two section
61851 alignment, and ELF sh_addralign is required to be an integral power of
61852 two (or zero) by the ELF spec. Of course this is ignored by fuzzers,
61853 and because bfd_log2 rounds up, we can end up with alignment_power
61854 being 32 on a 32-bit object or 64 on a 64-bit object. That then
61855 triggers ubsan warnings in places like bfd_update_compression_header
61856 where we want to convert from alignment_power back to an alignment.
61857 I suppose we could reject object files that have non-compliant
61858 sh_addralign, but I think it's also reasonable to use the greatest
61859 power of two divisor of sh_addralign, ie. the rightmost 1 bit.
61861 * elf.c (_bfd_elf_make_section_from_shdr): Use greatest power
61862 of two divisor of sh_addralign.
61863 (_bfd_elf_assign_file_position_for_section): Likewise.
61864 (assign_file_positions_for_non_load_sections): Likewise.
61866 2022-02-16 Alan Modra <amodra@gmail.com>
61868 asan: buffer overflow in vms-alpha.c
61869 * vms-alpha.c (evax_bfd_print_dst): Sanity check another place
61872 asan : use of uninitialized value in buffer_and_nest
61873 * macro.c (buffer_and_nest): Don't read past end of string buffer.
61875 asan: buffer overflow in peXXigen.c
61876 * peXXigen.c (_bfd_XX_bfd_copy_private_bfd_data_common): Properly
61877 sanity check DataDirectory[PE_DEBUG_DATA].Size.
61879 2022-02-16 Hans-Peter Nilsson <hp@axis.com>
61881 sim/common: Improve sim_dump_memory head comment
61882 As requested by Mike.
61884 * sim-memopt.c: Improve head comment.
61886 2022-02-16 Hans-Peter Nilsson <hp@axis.com>
61888 sim/testsuite/cris/c/stat3.c: Fix formatting nit
61889 * c/stat3.c (main): Fix formatting nit.
61891 2022-02-16 Mike Frysinger <vapier@gentoo.org>
61893 sim: testsuite: cleanup the istarget * logic
61894 Now that the multitarget testing has settled, clean up the cases where
61895 istarget * is used. This ends up being mostly style unindenting.
61897 2022-02-16 GDB Administrator <gdbadmin@sourceware.org>
61899 Automatic date update in version.in
61901 2022-02-15 H.J. Lu <hjl.tools@gmail.com>
61903 i386: Update I386_NEED_DYNAMIC_RELOC_TYPE_P for DT_TEXTREL
61904 Update I386_NEED_DYNAMIC_RELOC_TYPE_P to allow R_386_TLS_IE for relocation
61905 in read-only section.
61910 * elfxx-x86.h (I386_NEED_DYNAMIC_RELOC_TYPE_P): Allow
61915 * testsuite/ld-i386/i386.exp: Run pr28894.
61916 * testsuite/ld-i386/pr28894.d: New file.
61917 * testsuite/ld-i386/pr28894.s: Likewise.
61919 2022-02-15 Hans-Peter Nilsson <hp@axis.com>
61921 sim/testsuite: Default global_cc_os and global_cc_works properly
61922 There was an omission on 3e6dc39ed7a8 "sim/testsuite: Set
61923 global_cc_os also when no compiler is found"; global_cc_os
61924 wasn't set for other than the primary target, which means
61925 that the "unguarded" use of global_cc_os in
61926 testsuite/cris/c/c.exp caused the dreaded "ERROR: can't read
61927 "global_cc_os": no such variable" when e.g. configuring for
61928 pru-elf and doing "make check-sim". Better initializing
61929 both variables at the top to default values, rather than
61930 adding another single 'set global_cc_os ""', to reduce the
61931 risk of not setting them properly if or when that
61932 if-statement-chain is made longer.
61935 * lib/sim-defs.exp (sim_init_toolchain): Default
61936 global_cc_os and global_cc_works properly, before if-chain.
61938 2022-02-15 H.J. Lu <hjl.tools@gmail.com>
61940 x86: Add has_sib to struct instr_info
61941 Add has_sib to struct instr_info and use SIB info only if ins->has_sib
61945 * i386-dis.c (instr_info): Add has_sib.
61946 (get_sib): Set has_sib.
61947 (OP_E_memory): Replace havesib with ins->has_sib.
61948 (OP_VEX): Use ins->sib.index only if ins->has_sib is true.
61950 2022-02-15 Lancelot SIX <lancelot.six@amd.com>
61952 gdb: Respect the DW_CC_nocall attribute
61953 It is possible for a compiler to optimize a function in a such ways that
61954 the function does not follow the calling convention of the target. In
61955 such situation, the compiler can use the DW_AT_calling_convention
61956 attribute with the value DW_CC_nocall to tell the debugger that it is
61957 unsafe to call the function. The DWARF5 standard states, in 3.3.1.1:
61959 > If the value of the calling convention attribute is the constant
61960 > DW_CC_nocall, the subroutine does not obey standard calling
61961 > conventions, and it may not be safe for the debugger to call this
61964 Non standard calling convention can affect GDB's assumptions in multiple
61965 ways, including how arguments are passed to the function, how values are
61966 returned, and so on. For this reason, it is unsafe for GDB to try to do
61967 the following operations on a function with marked with DW_CC_nocall:
61969 - call / print an expression requiring the function to be evaluated,
61970 - inspect the value a function returns using the 'finish' command,
61971 - force the value returned by a function using the 'return' command.
61973 This patch ensures that if a command which relies on GDB's knowledge of
61974 the target's calling convention is used on a function marked nocall, GDB
61975 prints an appropriate message to the user and does not proceed with the
61976 operation which is unreliable.
61978 Note that it is still possible for someone to use a vendor specific
61979 value for the DW_AT_calling_convention attribute for example to indicate
61980 the use of an alternative calling convention. This commit does not
61981 prevent this, and target dependent code can be adjusted if one wanted to
61982 support multiple calling conventions.
61984 Tested on x86_64-Linux, with no regression observed.
61986 Change-Id: I72970dae68234cb83edbc0cf71aa3d6002a4a540
61988 2022-02-15 Lancelot SIX <lancelot.six@amd.com>
61989 Simon Marchi <simon.marchi@polymtl.ca>
61991 gdb: add a symbol* argument to get_return_value
61992 Add an argument to the get_return_value function to indicate the symbol
61993 of the function the debuggee is returning from. This will be used by
61994 the following patch.
61996 Since the function return type can be deduced from the symbol remove the
61997 value_type argument which becomes redundant.
61999 No user visible change after this patch.
62001 Tested on x86_64-linux.
62003 Change-Id: Idf1279f1f7199f5022738a6679e0fa63fbd22edc
62005 2022-02-15 H.J. Lu <hjl.tools@gmail.com>
62007 x86-64: Use MAXPAGESIZE for the relro segment alignment
62008 Adjust x86-64 linker tests after reverting
62010 commit 31b4d3a16f200bf04db8439a63b72bba7af4e1be
62011 Author: Alan Modra <amodra@gmail.com>
62012 Date: Thu Feb 3 08:57:47 2022 +1030
62014 PR28824, relro security issues, x86 keep COMMONPAGESIZE relro
62016 to use MAXPAGESIZE for the end of the relro segment alignment, like other
62019 * testsuite/ld-x86-64/plt-main-bnd.dd: Updated.
62020 * testsuite/ld-x86-64/plt-main-ibt-x32.dd: Likewise.
62021 * testsuite/ld-x86-64/plt-main-ibt.dd: Likewise.
62022 * testsuite/ld-x86-64/pr14207.d: Likewise.
62023 * testsuite/ld-x86-64/pr18176.d: Likewise.
62024 * testsuite/ld-x86-64/pr20830a-now.d: Likewise.
62025 * testsuite/ld-x86-64/pr20830a.d: Likewise.
62026 * testsuite/ld-x86-64/pr20830b-now.d: Likewise.
62027 * testsuite/ld-x86-64/pr20830b.d: Likewise.
62028 * testsuite/ld-x86-64/pr21038a-now.d: Likewise.
62029 * testsuite/ld-x86-64/pr21038a.d: Likewise.
62030 * testsuite/ld-x86-64/pr21038b-now.d: Likewise.
62031 * testsuite/ld-x86-64/pr21038b.d: Likewise.
62032 * testsuite/ld-x86-64/pr21038c-now.d: Likewise.
62033 * testsuite/ld-x86-64/pr21038c.d: Likewise.
62035 2022-02-15 H.J. Lu <hjl.tools@gmail.com>
62037 Revert "PR28824, relro security issues, x86 keep COMMONPAGESIZE relro"
62038 This reverts commit 31b4d3a16f200bf04db8439a63b72bba7af4e1be.
62040 2022-02-15 GDB Administrator <gdbadmin@sourceware.org>
62042 Automatic date update in version.in
62044 2022-02-14 Hans-Peter Nilsson <hp@axis.com>
62046 sim/testsuite/cris: If failing compilation, mark C tests as errors
62047 ...when we know we have a working compiler. This will reduce
62048 the risk of faulty edits by exposing them rather than hiding
62049 them as "unresolved". It also harmonizes behavior with that of
62052 * c/c.exp: Mark C tests failing compilation test errors.
62054 2022-02-14 Hans-Peter Nilsson <hp@axis.com>
62056 sim/testsuite/cris: Remove faulty use of basename in C tests
62057 Calls to basename were added here as part of commit
62058 e1e1ae6e9b5e "sim: testsuite: fix objdir handling", but that
62059 commit missed adding "#include <libgen.h>" or the equivalent
62060 GNU extension, see basename(3). Fixing that shows a logical
62061 error in the change to openpf1.c; the non-/-prefixed
62062 code-path was changed instead of the "/"-prefixed code-path,
62063 which is the one executed after that commit.
62065 For "newlib" these tests failed linking after that commit.
62066 Recent newlib has the (asm-renamed) GNU-extension-variant of
62067 basename, but we're better off not using it at all.
62069 Unfortunately, compilation failures for C tests run by the
62070 machinery in c.exp are currently just marked "unresolved",
62071 in contrast to C and assembler tests run by calling
62074 The interaction of calling with the full program-path vs.
62075 use of --sysroot exposes a consistency problem: when
62076 --sysroot is used, argv[0] isn't the path by which the
62077 program can find itself. It's undecided whether argv[0] for
62078 the program running in the simulator should be edited
62079 (related to the naked argument to the simulator before
62080 passing on to the simulated program) to remove a leading
62081 --sysroot. Either way, such a change would be out of scope
62084 * c/stat3.c (mybasename): New macro. Use it instead of basename.
62085 * c/openpf1.c: Correct basename-related change and update related
62088 2022-02-14 Hans-Peter Nilsson <hp@axis.com>
62090 sim: Add sim_dump_memory for debugging
62091 Intended to be called from the debugger tool.
62094 * sim-memopt.c (sim_dump_memory): New function.
62096 2022-02-14 Hans-Peter Nilsson <hp@axis.com>
62098 sim: Fix use of out-of-tree assembler and linker when testing
62099 With commit 7a259895bb2d "sim: testsuite: expand arch specific
62100 toolchain settings", trying to use out-of-tree ld and as at test-time
62101 broke for the "primary target", like when testing a release-tarball.
62103 Subsequent to that commit, all assembler tests without in-tree-built
62104 tools FAIL, getting errors when trying to call
62105 $(abs_builddir)/../gas/as-new. But, that isn't the actual culprint;
62106 it's actually it's its immediate predecessor, commit 8996c21067373
62107 "sim: testsuite: setup per-port toolchain settings for multitarget
62108 build", which hardcodes in-tree-paths to those tools instead of
62109 considering e.g. $(<X>_FOR_TARGET), the preferred overridable variable
62110 for single-target builds, as set up by the toplevel Makefile.
62112 This commit calls GCC_TARGET_TOOL (a deceptive name; gcc-specific
62113 features aren't used) from toplev/config/acx.m4, somewhat like calls
62114 in toplev/configure.ac but without the NCN_STRICT_CHECK_TARGET_TOOLS
62115 step, for each X to find a value for $(<X>_FOR_TARGET). N.B.: in-tree
62116 tools still override any ${target}-${tool} found in $PATH, i.e. only
62117 previously broken builds are affected.
62119 The variables $(<X>_FOR_TARGET) are usually overridden by the toplevel
62120 Makefile to the same value or better, but has to be set here too, as
62121 automake "wants" Makefiles to be self-contained (you get an error
62122 pointing out that the variable may be empty). If it hadn't been for
62123 that, SIM_AC_CHECK_TOOLCHAIN_FOR_PRIMARY_TARGET would not be needed.
62124 This detail should only (positively) affect users invoking "make
62125 check" in sim/ instead of "make check-sim" (or "make check") at the
62126 toplevel. Now the output from "configure" matches the target tools
62127 actually used by sim at test-time, for the "primary target".
62129 Using $(CC) for "example-" targets CC_FOR_TARGET is not changed, as
62130 that appears to be a deliberate special-case.
62132 Note that all tools still have to be installed and present in
62133 $PATH at configure-time to be properly used at test-time.
62136 * m4/sim_ac_toolchain.m4 (SIM_AC_CHECK_TOOLCHAIN_FOR_PRIMARY_TARGET):
62138 (SIM_TOOLCHAIN_VARS): Call it using AC_REQUIRE, and use variables
62139 AS_FOR_TARGET, LD_FOR_TARGET and CC_FOR_TARGET instead of hard-coded
62141 * Makefile.in, configure: Regenerate.
62143 2022-02-14 Hans-Peter Nilsson <hp@axis.com>
62145 sim cris: Unbreak --disable-sim-hardware builds
62146 With --disable-sim-hardware (--enable-sim-hardware=no),
62147 whose default was changed to --enable-sim-hardware(=yes) in
62148 commit 34cf51120683, building for cris-elf fails as
62149 sim_hw_parse then doesn't exist.
62151 A cris-elf simulator configured for --enable-sim-hardware
62152 (or the default after to the mentioned commit) runs about
62153 2.5x slower than one configured --disable-sim-hardware.
62154 A further 2-5% performance regression was not investigated.
62156 When sim_hw_parse doesn't exist, --cris-900000xx can't be
62157 supported. The best action here is to remove it completely,
62158 so its absence can be identified through --help, but
62159 avoiding littering the code with "#if WITH_HW".
62162 * sim-if.c (cris_options) [WITH_HW]: Conditionalize
62163 support of option --cris-900000xx.
62164 (sim_open) [WITH_HW]: Conditionalize sim_hw_parse
62167 2022-02-14 Hans-Peter Nilsson <hp@axis.com>
62169 sim/testsuite/cris: As applicable, require simoption --cris-900000xx
62170 Apply the new run_sim_test option "require" as in "#require
62171 simoption --cris-900000xx" for all tests using that option.
62172 This allows a clean test-suite-run for a build with
62173 --disable-sim-hardware, where that option is not supported,
62174 by skipping those tests as "untested".
62176 sim/testsuite/cris:
62177 * asm/io1.ms, asm/io2.ms, asm/io3.ms, asm/io6.ms,
62178 asm/io7.ms: Call "#require: simoption --cris-900000xx".
62180 2022-02-14 Hans-Peter Nilsson <hp@axis.com>
62182 sim/testsuite: Support "requires: simoption <--name-of-option>"
62183 Simulator features can be present or not, typically
62184 depending on different-valued configure options, like
62185 --enable-sim-hardware[=off|=on]. To avoid failures in
62186 test-suite-runs when testing such configurations, a new
62187 predicate is needed, as neither "target", "progos" nor
62188 "mach" fits cleanly.
62190 The immediate need was to check for presence of a simulator
62191 option, but rather than a specialized "requires-simoption:"
62192 predicate I thought I'd handle the general (parametrized)
62193 need, so here's a generic predicate machinery and a (first)
62194 predicate to use together with it; checking whether a
62195 particular option is supported, by looking at "run --help"
62196 output. This was inspired by the check_effective_target_
62197 machinery in the gcc test-suite.
62199 Multiple "requires: <requirement> <parameter>" form a list of
62200 predicates (with parameters), to be used as a conjunction.
62203 * lib/sim-defs.exp (sim_check_requires_simoption): New function.
62204 (run_sim_test): Support "requires: <requirement> <parameter>".
62206 2022-02-14 Hans-Peter Nilsson <hp@axis.com>
62208 sim/testsuite/cris/hw/rv-n-cris/irq1.ms: Disable due to randomness
62209 For reasons that remain largely to be investigated (besides
62210 the apparent lack of synchronization between two processes),
62211 this test fails randomly, with two different sets of common
62212 outputs. Curiously, that doesn't happen for the other
62213 similar tests. There's a comment that mentions this, though
62214 that doesn't make it a sustainable part of a test-suite.
62215 (Known-blinking tests should be disabled until fixed.)
62217 sim/testsuite/cris:
62218 * hw/rv-n-cris/irq1.ms: Disable by use of a never-matched
62221 2022-02-14 Hans-Peter Nilsson <hp@axis.com>
62223 sim/testsuite/cris/c: Use -sim3 but only for newlib targets
62224 Commit a39487c6685f "sim: cris: use -sim with C tests for cris-elf
62225 targets" caused " -sim" to be appended to CFLAGS_FOR_TARGET for
62226 cris*-*-elf, where testing had until then relied on
62227 "RUNTESTFLAGS=--target_board=cris-sim" being passed when running "make
62228 check-sim", adding the right options. While "-sim" happens to work,
62229 the baseboard-file cris-sim.exp uses "-sim3" so for consistency use
62232 Then commit b42f20d2ac72 "sim: testsuite: drop most specific istarget
62233 checks" caused " -sim" to be appended for *all* targets, which just
62234 doesn't work. For example, for crisv32-linux-gnu, that's not a
62235 recognized option and will cause a dejagnu error and further testing
62236 in c.exp will be aborted.
62238 While cris-sim.exp appends "-static" for *-linux-gnu, further changes
62239 in the test-suite have caused "linux"-specific tests to break, so that
62240 part will be tended to separately.
62242 But, save and restore CFLAGS_FOR_TARGET around the modification and
62243 use where needed, to not have the CRIS-specific modification affect a
62244 continuing test-run (possibly for other targets).
62246 sim/testsuite/cris:
62247 * c/c.exp (CFLAGS_FOR_TARGET): Replace appended option " -sim"
62248 with " -sim3", but do it conditionally for newlib targets. Save
62249 and restore CFLAGS_FOR_TARGET in saved_CFLAGS_FOR_TARGET such
62250 that it doesn't affect the value of CFLAGS_FOR_TARGET outside
62253 2022-02-14 Hans-Peter Nilsson <hp@axis.com>
62255 sim/testsuite: Set global_cc_os also when no compiler is found
62256 If we don't set this variable, it doesn't exist, and using "#progos:"
62257 in an assembler-file will cause an error rather than just skipping the
62260 Running /src/sim/testsuite/cris/hw/rv-n-cris/rvc.exp ...
62261 ERROR: tcl error sourcing /src/sim/testsuite/cris/hw/rv-n-cris/rvc.exp.
62262 ERROR: can't read "global_cc_os": no such variable
62264 "if { $opts(progos) != "" && $opts(progos) != $global_cc_os } {
62265 untested $subdir/$name
62268 (procedure "run_sim_test" line 102)
62270 Neither the commit introducing progos, nor the top comment
62271 in run_sim_test, mentions progos as intended only for C
62272 tests, or that its use must be gated on $global_cc_works !=
62273 0, so (not) setting it in the no-working-compiler path seems
62276 Allowing it to be used for assembler tests makes it usable
62277 for e.g. an always-false predicate and in expressions in
62278 .exp files without gating on $global_cc_works != 0.
62280 With this patch, global_cc_os is set to "", just as for "unknown OS".
62283 * lib/sim-defs.exp (sim_init_toolchain): Set global_cc_os also when
62284 no working target C compiler is found.
62286 2022-02-14 Hans-Peter Nilsson <hp@axis.com>
62288 sim/testsuite/cris: Assembler testcase for PRIx32 usage bug
62289 Several C test-cases exposed the bug, but let's have one for
62290 people who test using just the assembler and linker.
62292 * asm/endmem1.ms: New test.
62294 2022-02-14 Hans-Peter Nilsson <hp@axis.com>
62296 sim cris: Correct PRIu32 to PRIx32
62297 In 5ee0bc23a68f "sim: clean up bfd_vma printing" there was
62298 an additional introduction of PRIx32 and PRIu32 but just in
62299 sim/cris/sim-if.c. One type of bug was fixed in commit
62300 d16ce6e4d581 "sim: cris: fix memory setup typos" but one
62301 remained; the PRIu32 usage is wrong, as hex output is
62302 desired; note the 0x prefix.
62304 Without this fix, you'll see output like:
62305 memory map 0:0x4000..0x5fff (8192 bytes) overlaps 0:0x0..0x16383 (91012 bytes)
62306 program stopped with signal 6 (Aborted).
62307 for some C programs, like some of the ones in the sim/cris/c
62308 testsuite from where the example is taken (freopen2.c).
62310 The bug behavior was with memory allocation. With an
62311 attempt to allocate memory using the brk syscall such that
62312 the room up to the next 8192-byte "page boundary" wasn't
62313 sufficient, the simulator memory allocation machinery horked
62314 on a consistency error when trying to allocate a memory
62315 block to raise the "end of the data segment": there was
62316 already memory allocated at that address.
62318 Unfortunately, none of the programs in sim/cris/asm exposed
62319 this bug at the time, but an assembler test-case is
62320 committed after this fix.
62323 * sim-if.c (sim_open): Correct PRIu32 to PRIx32.
62325 2022-02-14 Sergei Trofimovich <siarheit@google.com>
62327 microblaze: fix fsqrt collicion to build on glibc-2.35
62328 * microblaze-opcm.h: Renamed 'fsqrt' to 'microblaze_fsqrt'.
62329 * microblaze-opc.h: Follow 'fsqrt' rename.
62331 2022-02-14 Tom Tromey <tom@tromey.com>
62333 Remove LA_PRINT_STRING
62334 This removes the LA_PRINT_STRING macro, in favor of using ordinary
62337 Remove LA_PRINT_CHAR
62338 This removes the LA_PRINT_CHAR macro, in favor of using ordinary
62341 Remove LA_PRINT_TYPE
62342 This removes the LA_PRINT_TYPE macro, in favor of using ordinary
62345 2022-02-14 Andrew Burgess <andrew.burgess@embecosm.com>
62347 gdb/python: move styling support to gdb.styling
62348 This commit moves the two Python functions that are used for styling
62349 into a new module, gdb.styling, there's then a small update in
62350 python.c so GDB can find the functions in their new location.
62352 The motivation for this change is purely to try and reduce the clutter
62353 in the top-level gdb module, and encapsulate related functions into
62354 modules. I did ponder documenting these functions as part of the
62355 Python API, however, doing so would effectively "fix" the API, and I'm
62356 still wondering if there's improvements that could be made, also, the
62357 colorize function is only called in some cases now that GDB prefers
62358 libsource-highlight, so it's not entirely sure how this would work as
62359 part of a user facing API.
62361 Still, despite these functions never having been part of a documented
62362 API, it is possible that a user out there has overridden these to, in
62363 some way, customize how GDB performs styling. Moving the function as
62364 I propose in this patch could break things for that user, however,
62365 fixing this breakage is trivial, and, as these functions were never
62366 documented, I don't think we should be obliged to not break user code
62367 that relies on them.
62369 2022-02-14 Andrew Burgess <andrew.burgess@embecosm.com>
62371 gdb: use python to colorize disassembler output
62372 This commit adds styling support to the disassembler output, as such
62373 two new commands are added to GDB:
62375 set style disassembler enabled on|off
62376 show style disassembler enabled
62378 In this commit I make use of the Python Pygments package to provide
62379 the styling. I did investigate making use of libsource-highlight,
62380 however, I found the highlighting results to be inferior to those of
62381 Pygments; only some mnemonics were highlighted, and highlighting of
62382 register names such as r9d and r8d (on x86-64) was incorrect.
62384 To enable disassembler highlighting via Pygments, I've added a new
62385 extension language hook, which is then implemented for Python. This
62386 hook is very similar to the existing hook for source code
62389 One possibly odd choice I made with the new hook is to pass a
62390 gdb.Architecture through, even though this is currently unused. The
62391 reason this argument is not used is that, currently, styling is
62392 performed identically for all architectures.
62394 However, even though the Python function used to perform styling of
62395 disassembly output is not part of any documented API, I don't want
62396 to close the door on a user overriding this function to provide
62397 architecture specific styling. To do this, the user would inevitably
62398 require access to the gdb.Architecture, and so I decided to add this
62401 The styling is applied within gdb_disassembler::print_insn, to achieve
62402 this, gdb_disassembler now writes its output into a temporary buffer,
62403 styling is then applied to the contents of this buffer. Finally the
62404 gdb_disassembler buffer is copied out to its final destination stream.
62406 There's a new test to check that the disassembler output includes some
62407 escape sequences, though I don't check for specific colours; the
62408 precise colors will depend on which instructions are in the
62409 disassembler output, and, I guess, how pygments is configured.
62411 The only negative change with this commit is how we currently style
62414 Currently, when the disassembler wants to print an address, we call
62415 back into GDB, and GDB prints the address value using the `address`
62416 styling, and the symbol name using `function` styling. After this
62417 commit, if pygments is used, then all disassembler styling is done
62418 through pygments, and this include the address and symbol name parts
62419 of the disassembler output.
62421 I don't know how much of an issue this will be for people. There's
62422 already some precedent for this in GDB when we look at source styling.
62423 For example, function names in styled source listings are not styled
62424 using the `function` style, but instead, either GNU Source Highlight,
62425 or pygments gets to decide how the function name should be styled.
62427 If the Python pygments library is not present then GDB will continue
62428 to behave as it always has, the disassembler output is mostly
62429 unstyled, but the address and symbols are styled using the `address`
62430 and `function` styles, as they are today.
62432 However, if the user does `set style disassembler enabled off`, then
62433 all disassembler styling is switched off. This obviously covers the
62434 use of pygments, but also includes the minimal styling done by GDB
62435 when pygments is not available.
62437 2022-02-14 H.J. Lu <hjl.tools@gmail.com>
62439 ld: Keep indirect symbol from IR if referenced from shared object
62440 Don't change indirect symbol defined in IR to undefined if it is
62441 referenced from shared object.
62446 * elflink.c (_bfd_elf_merge_symbol): Don't change indirect
62447 symbol defined in IR to undefined if it is referenced from
62453 * testsuite/ld-plugin/lto.exp: Run PR ld/28879 tests.
62454 * testsuite/ld-plugin/pr28879a.cc: New file.
62455 * testsuite/ld-plugin/pr28879b.cc: Likewise.
62457 2022-02-14 GDB Administrator <gdbadmin@sourceware.org>
62459 Automatic date update in version.in
62461 2022-02-13 Alan Modra <amodra@gmail.com>
62463 PR28882, build failure with gcc-4.2 due to use of 0b literals
62465 * elf/loongarch.h: Replace binary literals with hex.
62467 2022-02-13 Alan Modra <amodra@gmail.com>
62469 Don't pass around expld.dataseg pointer
62470 The better to see any code that accesses expld.dataseg.
62472 * ldexp.c (fold_segment_end): Remove seg parameter. Adjust calls.
62473 (fold_segment_align, fold_segment_relro_end): Likewise.
62474 * ldlang.c (lang_size_segment): Likewise.
62475 (lang_size_relro_segment_1, lang_find_relro_sections_1): Likewise.
62477 2022-02-13 Alan Modra <amodra@gmail.com>
62479 Remove bfd ELF_RELROPAGESIZE
62480 Now that ld properly aligns the end of the relro segment, the hack to
62481 make relro work on powerpc can disappear.
62484 * bfd.c (bfd_emul_get_commonpagesize): Remove relro param.
62485 Don't return bed->relropagesize.
62486 * elf-bfd.h (struct elf_backend_data): Remove relropagesize.
62487 * elfxx-target.h (ELF_RELROPAGESIZE): Remove.
62488 * elf32-ppc.c (ELF_RELROPAGESIZE): Don't define.
62489 * elf64-ppc.c: Likewise.
62490 * bfd-in2.h: Regenerate.
62492 * ldemul.c (after_parse_default): Adjust
62493 bfd_emul_get_commonpagesize call.
62495 2022-02-13 Alan Modra <amodra@gmail.com>
62497 PR28824, relro security issues, x86 keep COMMONPAGESIZE relro
62498 x86 treats MAXPAGESIZE as a memory optimisation parameter, actual
62499 hardware paging is always COMMPAGESIZE of 4k. Use COMMONPAGESIZE for
62500 the end of the relro segment alignment.
62502 The previous patch regresses pr18176, increasing the testcase file
62503 size from 322208 to 2099872 bytes. Fixing this on x86 will require
62504 introducing a gap after the end of the relro segment (of up to
62505 relropagesize-1 bytes).
62509 * ld.h (ld_config_type): Add relro_use_commonpagesize field.
62510 * ldexp.c (fold_segment_align): Set relropagesize depending on
62511 relro_use_commonpagesize.
62512 * emultempl/elf-x86.em (elf_x86_create_output_section_statements):
62513 Set relro_use_commonpagesize.
62514 * testsuite/ld-x86-64/pr18176.d: xfail.
62516 2022-02-13 Alan Modra <amodra@gmail.com>
62518 PR28824, relro security issues
62521 There are constraints on layout of binaries to meet demand paging and
62522 memory protection requirements. Demand paged binaries must have file
62523 offset mod pagesize equal to vma mod pagesize. Memory protection
62524 (executable, read, write status) can only change at page boundaries.
62525 The linker's MAXPAGESIZE variable gives the page size for these layout
62528 In a typical basic executable with two memory segments, text (RE) and
62529 data (RW), the data segment must start on a different page to the
62530 last text segment page. For example, with 64k pages and a small
62531 executable of 48k text and 1k data, the text segment might start at
62532 address 0x10000 and data at 0x20000 for a total of two 64k memory
62533 pages. Demand paging would require the image on disk to be 64k+1k
62534 in size. We can do better than that. If the data segment instead
62535 starts at 0x2c000 (the end of the text segment plus one 64k page) then
62536 there are still only two memory pages, but the disk image is now
62537 smaller, 48k+1k in size. This is why the linker normally starts the
62538 data segment at the end of the text segment plus one page. That
62539 simple heuristic isn't ideal in all cases. Changing our simple
62540 example to one with 64k-1 text size, following that heuristic would
62541 result in data starting at 0x2ffff. Now we have two 64k memory data
62542 pages for a data segment of 1k! If the data segment instead started
62543 at 0x30000 we'd get a single data segment page at the cost of 1 byte
62544 extra in the disk image, which is likely a good trade-off. So the
62545 linker does adjust the simple heuristic. Just how much disk image
62546 size increase is allowed is controlled by the linker's COMMONPAGESIZE
62549 A PT_GNU_RELRO segment overlays the initial part of the data segment,
62550 saying that those pages should be made read-only after relocation by
62551 the dynamic loader. Page granularity for memory protection means that
62552 the end of the relro segment must be at a page boundary.
62556 Unfortunately most targets currently only align the end of the relro
62557 segment to COMMONPAGESIZE. That results in only partial relro
62558 protection if an executable is running with MAXPAGESIZE pages, since
62559 any part of the relro segment past the last MAXPAGESIZE boundary can't
62560 be made read-only without also affecting sections past the end of the
62561 relro segment. I believe this problem arose because x86 always runs
62562 with 4k (COMMPAGESIZE) memory pages, and therefore using a larger
62563 MAXPAGESIZE on x86 is for reasons other than the demand paging and
62564 memory page protection boundary requirements.
62568 Always end the relro segment on a MAXPAGESIZE boundary, except for
62569 x86. Note that the relro segment, comprising of sections at the start
62570 of the data segment, is sized according to how those sections are laid
62571 out. That means the start of the relro segment is fixed relative to
62572 its end. Which also means the start of the data segment must be at a
62573 fixed address mod MAXPAGESIZE. So for relro the linker can't play
62574 games with the start of the data segment to save disk space. At
62575 least, not without introducing gaps between the relro sections. In
62576 fact, because the linker was starting layout using its simple
62577 heuristic of starting the data segment at the end of the text segment
62578 plus one page, it was sometimes introducing page gaps for no reason.
62583 * ldexp.c (fold_segment_align): When relro, don't adjust up by
62584 offset within page. Set relropagesize.
62585 (fold_segment_relro_end): Align to relropagesize.
62586 * ldexp.h (seg_align_type): Rename pagesize to commonpagesize.
62587 Add relropagesize. Comment.
62588 * ldlang.c (lang_size_segment): Adjust to suit field renaming.
62589 (lang_size_relro_segment_1): Align relro_end using relropagesize.
62591 2022-02-13 GDB Administrator <gdbadmin@sourceware.org>
62593 Automatic date update in version.in
62595 2022-02-12 GDB Administrator <gdbadmin@sourceware.org>
62597 Automatic date update in version.in
62599 2022-02-11 H.J. Lu <hjl.tools@gmail.com>
62601 x86: Disallow invalid relocation against protected symbol
62602 I am checking this into master and will backport it to 2.38 branch.
62606 On x86, GCC 12 supports -mno-direct-extern-access to enable canonical
62607 reference to protected function and disable copy relocation. With
62608 -mno-direct-extern-access, the canonical protected function symbols must
62609 be accessed via canonical reference and the protected data symbols in
62610 shared libraries are non-copyable. Under glibc 2.35, non-canonical
62611 reference to the canonical protected function will get the run-time error:
62613 ./y: internal_f: ./libfoo.so: non-canonical reference to canonical protected function
62615 and copy relocations against the non-copyable protected symbols will get
62616 the run-time error:
62618 ./x: internal_i: ./libfoo.so: copy relocation against non-copyable protected symbol
62620 Update x86 linker to disallow non-canonical reference to the canonical
62621 protected function:
62623 ld: plt.o: non-canonical reference to canonical protected function `internal_f' in libfoo.so
62624 ld: failed to set dynamic section sizes: bad value
62626 and copy relocation against the non-copyable protected symbol:
62628 ld: main.o: copy relocation against non-copyable protected symbol `internal_i' in libfoo.so
62635 * elf-properties.c (_bfd_elf_parse_gnu_properties): Don't skip
62636 shared libraries for GNU_PROPERTY_1_NEEDED_INDIRECT_EXTERN_ACCESS.
62637 * elf32-i386.c (elf_i386_scan_relocs): Disallow non-canonical
62638 reference to canonical protected function.
62639 * elf64-x86-64.c (elf_x86_64_scan_relocs): Likewise.
62640 * elfxx-x86.c (elf_x86_allocate_dynrelocs): Don't allow copy
62641 relocation against non-copyable protected symbol.
62646 * testsuite/ld-i386/i386.exp: Check non-canonical reference to
62647 canonical protected function and check copy relocation against
62648 non-copyable protected symbol.
62649 * testsuite/ld-i386/pr21997-1.err: New file.
62650 * testsuite/ld-i386/pr28875.err: Likewise.
62651 * testsuite/ld-i386/pr28875a.c: Likewise.
62652 * testsuite/ld-i386/pr28875b.c: Likewise.
62653 * testsuite/ld-x86-64/pr21997-1a.err: Updated.
62654 * testsuite/ld-x86-64/pr21997-1b.err: Likewise.
62655 * testsuite/ld-x86-64/pr28875-data.err: New file.
62656 * testsuite/ld-x86-64/pr28875-func.err: Likewise.
62657 * testsuite/ld-x86-64/x86-64.exp: Check non-canonical reference
62658 to canonical protected function and check copy relocation against
62659 non-copyable protected symbol.
62661 2022-02-11 Tom Tromey <tromey@adacore.com>
62663 Add initializers to bound_minimal_symbol
62664 This adds initializers to bound_minimal_symbol, allowing for the
62665 removal of some calls to memset.
62667 2022-02-11 Bhuvanendra Kumar N <Bhuvanendra.KumarN@amd.com>
62669 gdb/fortran: support ptype and print commands for namelist variables
62670 Gfortran supports namelists (a Fortran feature); it emits
62671 DW_TAG_namelist and DW_TAG_namelist_item dies. But gdb does not
62672 process these dies and does not support 'print' or 'ptype' commands on
62673 namelist variables.
62675 An attempt to print namelist variables results in gdb bailing out with
62676 the error message as shown below.
62679 No symbol "nml" in current context.
62681 This commit is to make the print and ptype commands work for namelist
62682 variables and its items. Sample output of these commands is shared
62683 below, with fixed gdb.
62687 integer(kind=4) :: a
62688 integer(kind=4) :: b
62691 $1 = ( a = 10, b = 20 )
62693 2022-02-11 Bruno Larsen <blarsen@redhat.com>
62695 gdb: fix until behavior with trailing !is_stmt lines
62696 When using the command "until", it is expected that GDB will exit a
62697 loop if the current instruction is the last one related to that loop.
62698 However, if there were trailing non-statement instructions, "until"
62699 would just behave as "next". This was noticeable in clang-compiled
62700 code, but might happen with gcc-compiled as well. PR gdb/17315 relates
62701 to this problem, as running gdb.base/watchpoint.exp with clang
62702 would fail for this reason.
62704 To better understand this issue, consider the following source code,
62705 with line numbers marked on the left:
62707 10: for (i = 0; i < 10; ++i)
62709 12: other_stuff ();
62711 If we transform this to pseudo-assembler, and generate a line table,
62712 we could end up with something like this:
62714 Address | Pseudo-Assembler | Line | Is-Statement?
62716 0x100 | i = 0 | 10 | Yes
62717 0x104 | loop_body () | 11 | Yes
62718 0x108 | i = i + 1 | 10 | Yes
62719 0x10c | if (i < 10): | 10 | No
62720 0x110 | goto 0x104 | 10 | No
62721 0x114 | other_stuff () | 12 | Yes
62723 Notice the two non-statement instructions at the end of the loop.
62725 The problem is that when we reach address 0x108 and use 'until',
62726 hoping to leave the loop, GDB sets up a stepping range that runs from
62727 the start of the function (0x100 in our example) to the end of the
62728 current line table entry, that is 0x10c in our example. GDB then
62729 starts stepping forward.
62731 When 0x10c is reached GDB spots that we have left the stepping range,
62732 that the new location is not a statement, and that the new location is
62733 associated with the same source line number as the previous stepping
62734 range. GDB then sets up a new stepping range that runs from 0x10c to
62735 0x114, and continues stepping forward.
62737 Within that stepping range the inferior hits the goto (at 0x110) and
62738 loops back to address 0x104.
62740 At 0x104 GDB spots that we have left the previous stepping range, that
62741 the new address is marked as a statement, and that the new address is
62742 for a different source line. As a result, GDB stops and returns
62743 control to the user. This is not what the user was expecting, they
62744 expected GDB to exit the loop.
62746 The fix proposed in this patch, is that, when the user issues the
62747 'until' command, and GDB sets up the initial stepping range, GDB will
62748 check subsequent SALs (symtab_and_lines) to see if they are
62749 non-statements associated with the same line number. If they are then
62750 the end of the initial stepping range is extended to the end of the
62751 non-statement SALs.
62753 In our example above, the user is at 0x108 and uses 'until', GDB now
62754 sets up a stepping range from the start of the function 0x100 to
62755 0x114, the first address associated with a different line.
62757 Now as GDB steps around the loop it never leaves the initial stepping
62758 range. It is only when GDB exits the loop that we leave the stepping
62759 range, and the stepping finishes at address 0x114.
62761 This patch also adds a test case that can be run with gcc to test that
62762 this functionality is not broken in the future.
62764 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=17315
62766 2022-02-11 Richard Sandiford <richard.sandiford@arm.com>
62768 gas/doc: Fix "a true results" typo
62770 2022-02-11 Jan Vrany <jan.vrany@labware.com>
62772 gdb: extend the information printed by 'maint info jit'
62773 This commit updates the output of 'maint info jit' to print not just
62774 the jit_code_entry address, but also the symfile address, and the
62777 The new information could be obtained by looking into target memory at
62778 the contents of the jit_code_entry, but, by storing this information
62779 within gdb at the time the jit object is loaded, it is now possible to
62780 check if the jit_code_entry has been modified in target memory behind
62783 Additionally, the symfile address is the same address that is now used
62784 in the objfile names after commit 4a620b7e.
62786 One test that relies on the output of 'maint info jit' was updated to
62787 allow for the new output format.
62789 2022-02-11 Michael Forney <mforney@mforney.org>
62791 bfd: Remove return with expression in void function
62792 * bfd.c (bfd_set_gp_value): Remove return with expression
62795 2022-02-11 Tiezhu Yang <yangtiezhu@loongson.cn>
62797 gdb: LoongArch: Add Makefile, configure and NEWS
62798 This commit adds Makefile, configure and NEWS for LoongArch.
62800 gdb: LoongArch: Add initial native Linux support
62801 This commit adds initial native Linux support for LoongArch.
62803 gdb: LoongArch: Add initial Linux target support
62804 This commit adds initial Linux target support for LoongArch.
62806 gdb: LoongArch: Add initial baremetal support
62807 This commit adds initial baremetal support for LoongArch.
62809 gdb: LoongArch: Add initial target description support
62810 This commit adds initial target description support for LoongArch.
62812 2022-02-11 Mike Frysinger <vapier@gentoo.org>
62814 libctf: delete unused libctf_TEXINFOS
62815 It's not clear what this was meant for, but it's not used by anything,
62816 and the info pages still generate fine without it.
62818 2022-02-11 Simon Marchi <simon.marchi@polymtl.ca>
62820 gdb/linux: remove ptrace support check for exec, fork, vfork, vforkdone, clone, sysgood
62821 I think it's safe to remove checking support for these ptrace features,
62822 they have all been added in what is now ancient times (around the
62823 beginning of Linux 2.6). This allows removing a bit of complexity in
62824 linux-nat.c and nat/linux-ptrace.c.
62826 It also allows saving one extra fork every time we start debugging on
62827 Linux: linux_check_ptrace_features forks a child process to test if some
62828 ptrace features are supported. That child process forks a grand-child,
62829 to test whether ptrace reports an event for the fork by the child. This
62830 is no longer needed, if we assume the kernel supports reporting forks.
62832 PTRACE_O_TRACEVFORKDONE was introduced in Linux in this change, in 2003:
62834 https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git/commit/?id=45c1a159b85b3b30afd26a77b4be312226bba416
62836 PTRACE_O_TRACESYSGOOD was supported at least as of this change, in 2002:
62838 https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git/commit/?id=acc7088569c8eef04eeed0eff51d23bb5bcff964
62840 PTRACE_O_TRACEFORK, PTRACE_O_TRACEVFORK, PTRACE_O_TRACEEXEC and
62841 PTRACE_O_TRACECLONE were introduced in this change, in 2002:
62843 https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git/commit/?id=a0691b116f6a4473f0fa264210ab9b95771a2b46
62845 Change-Id: Iffb906549a89cc6b619427f976ec044706ab1e8d
62847 2022-02-11 GDB Administrator <gdbadmin@sourceware.org>
62849 Automatic date update in version.in
62851 2022-02-10 Andrew Burgess <aburgess@redhat.com>
62853 gdb/infrun: some extra infrun debug print statements
62854 While reviewing a different patch I wanted to know more about what was
62855 going on during GDB's stepping. I added some extra infrun debug print
62856 calls, and I thought these might be useful to others.
62858 2022-02-10 GDB Administrator <gdbadmin@sourceware.org>
62860 Automatic date update in version.in
62862 2022-02-09 Nick Clifton <nickc@redhat.com>
62864 Update the obsolete list and how-to-make-a-release documentation now that the 2.38 release is out.
62866 2022-02-09 Alan Modra <amodra@gmail.com>
62868 PR28763, SIGSEGV during processing of program headers via readelf
62870 * readelf.c (process_file_header): Discard any cached program
62871 headers if there is an extension field for e_phnum in first
62874 2022-02-09 Alan Modra <amodra@gmail.com>
62876 Work around gcc-4 warnings in elf64-ppc.c
62877 elf64-ppc.c: In function 'ppc64_elf_size_dynamic_sections':
62878 elf64-ppc.c:10309:45: error: value computed is not used [-Werror=unused-value]
62879 ++lgot_ents, ++lgot_masks, isym != NULL && isym++)
62881 It is of course a silly warning, fixed in later versions of gcc. I
62882 wrote "isym != NULL && isym++" rather than the simpler "isym++" to
62883 stop sanitisers complaining about incrementing a NULL pointer. isym
62884 is of course unused in any code path where it might start off as
62885 NULL. Sometimes you can't win. So don't try to be clever in reading
62886 local symbols only when needed. 99 times out of 100 they will be
62889 * elf64-ppc.c (ppc64_elf_size_dynamic_sections): Avoid annoying
62890 warnings by always reading local syms.
62891 (ppc64_elf_layout_multitoc): Likewise.
62893 2022-02-09 Peilin Ye <peilin.ye@bytedance.com>
62895 Test --only-keep-debug on ELF relocatables
62896 Add a test for commit 7c4643efe7be, which fixed --only-keep-debug for ELF
62899 * testsuite/binutils-all/objcopy.exp
62900 (keep_debug_symbols_for_elf_relocatable): New test.
62902 2022-02-09 GDB Administrator <gdbadmin@sourceware.org>
62904 Automatic date update in version.in
62906 2022-02-08 Palmer Dabbelt <palmer@rivosinc.com>
62908 RISC-V: Stop reporting warnings for mismatched extension versions
62909 The extension version checking logic is really just too complicated to
62910 encode into the linker, trying to do so causes more harm than good.
62911 This removes the checks and the associated tests, leaving the logic to
62912 keep the largest version of each extension linked into the target.
62916 * elfnn-riscv.c (riscv_version_mismatch): Rename to
62917 riscv_update_subset_version, and stop reporting warnings on
62918 version mismatches.
62919 (riscv_merge_std_ext): Adjust calls to riscv_version_mismatch.
62920 (riscv_merge_multi_letter_ext): Likewise.
62923 * testsuite/ld-riscv-elf/attr-merge-arch-failed-01.d: Remove
62924 * testsuite/ld-riscv-elf/attr-merge-arch-failed-01a.s: Likewise
62925 * testsuite/ld-riscv-elf/attr-merge-arch-failed-01b.s: Likewise
62926 * testsuite/ld-riscv-elf/attr-merge-arch-failed-02.d: Likewise
62927 * testsuite/ld-riscv-elf/attr-merge-arch-failed-02a.s: Likewise
62928 * testsuite/ld-riscv-elf/attr-merge-arch-failed-02b.s: Likewise
62929 * testsuite/ld-riscv-elf/attr-merge-arch-failed-02c.s: Likewise
62930 * testsuite/ld-riscv-elf/attr-merge-arch-failed-02d.s: Likewise
62931 * testsuite/ld-riscv-elf/attr-merge-user-ext-01.d: New test.
62932 * testsuite/ld-riscv-elf/attr-merge-user-ext-rv32i21_m2p0.s:
62934 * testsuite/ld-riscv-elf/attr-merge-user-ext-rv32i21_m2p1.s:
62936 * testsuite/ld-riscv-elf/ld-riscv-elf.exp: Remove obselete
62937 attr-merge-arch-failed-{01,02}, replace with
62938 attr-merge-user-ext-01.
62940 2022-02-08 Alan Modra <amodra@gmail.com>
62942 PR28862, heap-buffer-overflow in parse_stab_string
62943 I have no info on the format of a "SUNPRO C++ Namespace" stab, so am
62944 relying on the previous code being correct in parsing these stabs.
62945 Just don't allow NULs anywhere in the stab.
62948 * stabs.c (parse_stab_string): Don't overrun buffer when parsing
62951 2022-02-08 Alan Modra <amodra@gmail.com>
62953 Re: elf: Check symbol version without any symbols
62954 * testsuite/ld-elf/pr24718-1.d: Don't xfail for hppa64.
62956 2022-02-08 Andrew Burgess <aburgess@redhat.com>
62958 gdb: remove tailing newlines from index_cache_debug calls
62959 I noticed that most of the calls to index_cache_debug include a
62960 trailing newline. As the new debug mechanism already adds a newline,
62961 that means all of these debug calls result in a blank line being
62962 printed, which I think is a mistake.
62964 Remove all the trailing newlines.
62966 I also reformatted one of the index_cache_debug where a string will
62967 now fit onto a single line.
62969 Unless 'set debug index-cache on' is used, there should be no visible
62970 change in output after this commit.
62972 2022-02-08 H.J. Lu <hjl.tools@gmail.com>
62974 i386: Allow GOT32 relocations against ABS symbols
62975 GOT32 relocations are allowed since absolute value + addend is stored in
62978 Tested on glibc 2.35 build with GCC 11.2 and -Os.
62983 * elfxx-x86.c (_bfd_elf_x86_valid_reloc_p): Also allow GOT32
62989 * testsuite/ld-i386/i386.exp: Run pr28870.
62990 * testsuite/ld-i386/pr28870.d: New file.
62991 * testsuite/ld-i386/pr28870.s: Likewise.
62993 2022-02-08 GDB Administrator <gdbadmin@sourceware.org>
62995 Automatic date update in version.in
62997 2022-02-07 Andrew Burgess <aburgess@redhat.com>
62999 gdb/python: allow Value.format_string to return styled output
63000 Add a new argument to the gdb.Value.format_string method, 'styling'.
63001 This argument is False by default.
63003 When this argument is True, then the returned string can contain output
63004 styling escape sequences.
63006 When this argument is False, then the returned string will not contain
63007 any styling escape sequences.
63009 If the returned string is going to be printed to the user, then it is
63010 often nice to retain the GDB styling.
63012 For the testing, we need to adjust the TERM environment variable, as
63013 we do for all the styling tests. I'm now running all of the C tests
63014 in gdb.python/py-format-string.exp in an environment where styling
63015 could be generated, but only my new test should actually produce
63016 styled output, hopefully this will catch the case where a bug might
63017 cause format_string to always produce styled output.
63019 2022-02-07 Lancelot SIX <lancelot.six@amd.com>
63021 gdb: make thread_info::m_thread_fsm a std::unique_ptr
63022 While working on function calls, I realized that the thread_fsm member
63023 of struct thread_info is a raw pointer to a resource it owns. This
63024 commit changes the type of the thread_fsm member to a std::unique_ptr in
63025 order to signify this ownership relationship and slightly ease resource
63026 management (no need to manually call delete).
63028 To ensure consistent use, the field is made a private member
63029 (m_thread_fsm). The setter method (set_thread_fsm) can then check
63030 that it is incorrect to associate a FSM to a thread_info object if
63031 another one is already in place. This is ensured by an assertion.
63033 The function run_inferior_call takes an argument as a pointer to a
63034 call_thread_fsm and installs it in it in a thread_info instance. Also
63035 change this function's signature to accept a unique_ptr in order to
63036 signify that the ownership of the call_thread_fsm is transferred during
63039 No user visible change expected after this commit.
63041 Tested on x86_64-linux with no regression observed.
63043 Change-Id: Ia1224f72a4afa247801ce6650ce82f90224a9ae8
63045 2022-02-07 Andrew Burgess <aburgess@redhat.com>
63047 gdb: unbuffer all input streams when not using readline
63048 This commit should fix PR gdb/28711. What's actually going on is
63049 pretty involved, and there's still a bit of the story that I don't
63050 understand completely, however, from my observed results, I think that
63051 the change I propose making here (or something very similar) is going
63054 The original bug report involves using eclipse to drive gdb using mi
63055 commands. A separate tty is spun off in which to send gdb the mi
63056 commands, this tty is created using the new-ui command.
63058 The behaviour observed is that, given a particular set of mi commands
63059 being sent to gdb, we sometimes see an ESPIPE error from a lseek
63060 call, which ultimately results in gdb terminating.
63062 The problems all originate from gdb_readline_no_editing_callback in
63063 gdb/event-top.c, where we can (sometimes) perform calls to fgetc, and
63064 allow glibc to perform buffering on the FILE object being used.
63066 I say sometime, because, gdb_readline_no_editing_callback already
63067 includes a call to disable the glibc buffering, but this is only done
63068 if the input stream is not a tty. In our case the input stream is a
63069 tty, so the buffering is left in place.
63071 The first step to understanding why this problem occurs is to
63072 understand that eclipse sends multiple commands to gdb very quickly
63073 without waiting for and answer to each command, eclipse plans to
63074 collect all of the command results after sending all the commands to
63075 gdb. In fact, eclipse sends the commands to gdb that they appear to
63076 arrive in the gdb process as a single block of data. When reproducing
63077 this issue within the testsuite I find it necessary to send multiple
63078 commands using a single write call.
63080 The next bit of the story gets a little involved, and this is where my
63081 understanding is not complete. I can describe the behaviour that I
63082 observe, and (for me at least) I'm happy that what I'm seeing, if a
63083 little strange, is consistent. In order to fully understand what's
63084 going on I think I would likely need to dive into kernel code, which
63085 currently seems unnecessary given that I'm happy with the solution I'm
63088 The following description all relates to input from a tty in which I'm
63089 not using readline. I see the same problems either when using a
63090 new-ui tty, or with gdb's standard, non-readline, mi tty.
63092 Here's what I observe happening when I send multiple commands to gdb
63093 using a single write, if I send gdb this:
63095 command_1\ncommand_2\ncommand_3
63097 then gdb's event loop will wake up (from its select) as it sees there
63098 is input available. We call into gdb_readline_no_editing_callback,
63099 where we call fgetc, glibc will do a single big read, and get back
63104 that is, despite there being multiple lines of input available, I
63105 consistently get just a single line. From glibc a single character is
63106 returned from the fgetc call, and within gdb we accumulate characters,
63107 one at a time, into our own buffer. Eventually gdb sees the '\n'
63108 character, and dispatches the whole 'command_1' into gdb's command
63109 handler, which processes the command and prints the result. We then
63110 return to gdb_readline_no_editing_callback, which in turn returns to
63111 gdb's event loop where we re-enter the select.
63113 Inside the select we immediately see that there is more input waiting
63114 on the input stream, drop out of the select, and call back into
63115 gdb_readline_no_editing_callback. In this function we again call
63116 fgetc where glibc performs another big read. This time glibc gets:
63120 that is, we once again get just a single line, despite there being a
63121 third line available. Just like the first command we copy the whole
63122 string, character by character into gdb's buffer, then handle the
63123 command. After handling the command we go to the event loop, enter,
63124 and then exit the select, and call back to the function
63125 gdb_readline_no_editing_callback.
63127 In gdb_readline_no_editing_callback we again call fgetc, this time
63128 glibc gets the string:
63132 like before, we copy this to gdb's buffer and handle the command, then
63133 we return to the event loop. At this point the select blocks while we
63134 wait for more input to arrive.
63136 The important bit of this is that someone, somewhere is, it appears,
63137 taking care to split the incoming write into lines.
63139 My next experiment is to try something like:
63141 this_is_a_very_long_command\nshort_command\n
63143 However, I actually make 'this_is_a_very_long_command' very long, as
63144 in many hundreds of characters long. One way to do this is:
63146 echo xxxxxx.....xxxxx
63148 and just adding more and more 'x' characters as needed. What I'm
63149 aiming for is to have the first command be longer than glibc's
63150 internal read buffer, which, on my machine, is 1024 characters.
63152 However, for this discussion, lets imagine that glibc's buffer is just
63153 8 characters (we can create just this situation by adding a suitable
63154 setbuf call into gdb_readline_no_editing_callback).
63156 Now, if I send gdb this data:
63160 The first read from glibc will get 'abcdefgh', that is a full 8
63161 character buffer. Once gdb has copied these to its buffer we call
63162 fgetc again, and now glibc will get 'ij\n', that is, just like before,
63163 multiple lines are split at the '\n' character. The full command,
63164 which is now in gdb's buffer can be handled 'abcdefghij', after which
63165 we go (via the event loop) back to gdb_readline_no_editing_callback.
63166 Now we call fgetc, and glibc will get 'kl\n', which is then handled in
63169 So far, so good. However, there is, apparently, one edge case where
63170 the above rules don't apply.
63172 If the '\n' character is the first character read from the kernel,
63173 then the incoming lines are not split up. So, given glibc's 8
63174 character buffer, if I send gdb this:
63178 that is the first command is 8 characters plus a newline, then, on the
63179 first read (from within glibc) we get 'abcdefgh' in a single buffer.
63180 As there's no newline gdb calls fgetc again, and glibc does another
63181 large read, now we get:
63185 which doesn't follow the above pattern - the lines are not split into
63188 So, gdb reads the first character from glibc using fgetc, this is the
63189 newline. Now gdb has a complete command, and so the command is
63190 handled. We then return to the event loop and enter the select.
63192 The problem is that, as far as the kernel is concerned, there is no
63193 more input pending, it's all been read into glibc's buffer, and so the
63194 select doesn't return. The second command is basically stuck in
63197 If I send another command to gdb, or even just send an empty
63198 command (a lone newline) then the select returns, we call into
63199 gdb_readline_no_editing_callback, and now gdb sees the second
63202 OK, so the above is interesting, but it doesn't explain the ESPIPE
63205 Well, that's a slightly different, but related issue. The ESPIPE
63206 case will _only_ show up when using new-ui to create the separate tty
63207 for mi commands, and is a consequence of this commit:
63209 commit afe09f0b6311a4dd1a7e2dc6491550bb228734f8
63210 Date: Thu Jul 18 17:20:04 2019 +0100
63212 Fix for using named pipes on Windows
63214 Prior to this commit, the new-ui command would open the tty three
63215 times, once each for stdin, stderr, and stdout. After this commit we
63216 open the tty just once and reuse the FILE object for all three roles.
63218 Consider the problem case, where glibc has (unexpectedly) read the
63219 second command into its internal buffer. When we handle the first
63220 command we usually end up having to write something to the mi output
63223 After the above commit the same FILE object represents both the input
63224 and output streams, so, when gdb tries to write to the FILE object,
63225 glibc spots that there is input pending within the input buffer, and
63226 so assumes that we have read ahead of where we should be in the input
63227 file. To correct for this glibc tries to do an lseek call to
63228 reposition the file offset of the output stream prior to writing to
63229 it. However, as the output stream is a tty, and seeking is not
63230 supported on a tty, this lseek call fails, this results in the ESPIPE,
63231 which ultimately causes gdb to terminate.
63233 So, now we understand why the ESPIPE triggers (which was what caused
63234 the gdb crash in the original bug report), and we also understand that
63235 sometime gdb will not handle the second command in a timely
63236 fashion (if the first command is just the wrong length). So, what to
63239 We could revert the commit mentioned above (and implement its
63240 functionality another way). This would certainly resolve the ESPIPE
63241 issue, the buffered input would now only be on the input stream, the
63242 output stream would have no buffered input, and so glibc would never
63243 try to lseek, and so we'd never get the ESPIPE error.
63245 However, this only solves one of the two problems. We would still
63246 suffer from the problem where, if the first command is just the wrong
63247 length, the second command will not (immediately) get handled.
63249 The only solution I can see to this problem is to unbuffer the input
63250 stream. If glibc is not buffering the input, but instead, we read
63251 incoming data character by character from the kernel, then everything
63252 will be fine. As soon as we see the newline at the end of the first
63253 command we will handle the first command. As glibc will have no
63254 buffered input it will not be tempted to lseek, so no ESPIPE error.
63255 When we go have to the event loop there will be more data pending in
63256 the kernel, so the select will immediately return, and the second
63257 command will be processed.
63259 I'm tempted to suggest that we should move the unbuffering of the
63260 input stream out of gdb_readline_no_editing_callback and do it
63261 somewhere earlier, more like when we create the input streams.
63262 However, I've not done that in this commit for a couple of reasons:
63264 1. By keeping the unbuffering in gdb_readline_no_editing_callback
63265 I'm making the smallest possible change that fixes the bug. Moving
63266 the unbuffering somewhere better can be done as a refactor later, if
63267 that 's felt to be important,
63269 2. I don't think making repeated calls to unbuffer the input will
63270 have that much performance impact. We only make the unbuffer call
63271 once per call to gdb_readline_no_editing_callback, and, if the input
63272 stream is already unbuffered we'll return pretty quickly, so I don't
63273 see this as being massively costly,
63275 3. Tom is currently doing lots of gdb stream management changes and
63276 I want to minimise the chances we'll conflict.
63278 So, this commit just changes gdb_readline_no_editing_callback to
63279 always unbuffer the input stream.
63281 The test for this issue sends two commands in a loop, with the first
63282 command growing bigger each time around the loop. I actually make the
63283 first command bigger by just adding whitespace to the front, as gdb
63284 still has to read the complete command (including whitespace) via
63285 glibc, so this is enough to trigger the bug.
63287 The original bug was reported when using a virtual machine, and in
63288 this situation we see this in the strace output:
63290 read(9, "70-var-info-path-expression var1.aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa", 1024) = 64
63291 read(9, "\n71-var-info-path-expression var1.aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa\n", 1024) = 67
63293 I'm not completely sure what's going on here, but it appears that the
63294 kernel on the virtual machine is delivering the input to glibc slower
63295 than I see on my real hardware; glibc asks for 1024 bytes, but only
63296 gets 64 bytes the first time. In the second read we see the problem
63297 case, the first character is the newline, but then the entire second
63298 command is included.
63300 If I run this exact example on my real hardware then the first command
63301 would not be truncated at 64 bytes, instead, I'd expect to see the
63302 newline included in the first read, with the second command split into
63305 So, for testing, I check cases where the first command is just a few
63306 characters (starting at 8 character), all the way up to 2048
63307 characters. Hopefully, this should mean we hit the problem case for
63308 most machine setups.
63310 The only last question relates to commit afe09f0b6311a4d that I
63311 mentioned earlier. That commit was intended to provide support for
63312 Microsoft named pipes:
63314 https://docs.microsoft.com/en-us/windows/win32/ipc/named-pipes
63316 I know next to nothing about this topic beyond a brief scan of the
63317 above link, but I think these windows named pipe are closer in
63318 behaviour to unix sockets than to unix named fifos.
63320 I am a little nervous that, after the above commit, we now use the
63321 same FILE for in, err, and out streams. In contrast, in a vanilla C
63322 program, I would expect different FILE objects for each stream.
63324 Still, I'm reluctant to revert the above commit (and provide the same
63325 functionality a different way) without a specific bug to point at,
63326 and, now that the streams are unbuffered, I expect a lot of the read
63327 and write calls are going straight to the kernel with minimal glibc
63328 involvement, so maybe it doesn't really matter. Anyway, I haven't
63329 touched the above patch, but it is something to keep in mind when
63330 working in this area.
63332 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28711
63334 2022-02-07 Andrew Burgess <aburgess@redhat.com>
63336 gdb/disasm: combine the no printing disassembler setup code
63337 We have three places in gdb where we initialise a disassembler that
63338 will not print anything (used for figuring out the length of
63339 instructions, or collecting other information from the disassembler).
63341 Each of these places has its own stub function to act as a print like
63342 callback, the stub function is identical in each case, and just does
63345 In this commit I create a new function to initialise a disassembler
63346 that doesn't print anything, and have all three locations use this new
63347 function. There's now only one non-printing stub function.
63349 There should be no user visible changes after this commit.
63351 2022-02-07 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
63353 gdb: add the 'set/show suppress-cli-notifications' command
63354 GDB already has a flag to suppress printing notification events, such
63355 as thread and inferior context switches, on the CLI. This is used
63356 internally when executing commands. Make the flag available to the
63357 user via a new command. This is expected to be useful in scripts.
63359 For instance, suppose that when Inferior 1 gets to a certain state,
63360 you want to add and set up a new inferior using the commands below,
63361 but you also want to have a reduced/clean output.
63364 printf "Setting up Inferior 2...\n"
63365 add-inferior -exec a.out
63373 Currently, GDB prints
63376 Setting up Inferior 2...
63378 Added inferior 2 on connection 1 (native)
63379 [Switching to inferior 2 [<null>] (/tmp/a.out)]
63380 Breakpoint 2 at 0x1155: file file.c, line 3.
63382 Thread 2.1 "a.out" hit Breakpoint 2, main () at file.c:3
63384 [Switching to inferior 1 [process 7670] (/tmp/test)]
63385 [Switching to thread 1.1 (process 7670)]
63386 #0 main () at test.c:2
63390 GDB's Python API make it possible to capture and return GDB's output,
63391 but this does not work for all the streams. In particular, CLI
63392 notification events are not captured:
63394 (gdb) python gdb.execute("do-setup", False, True)
63395 [Switching to inferior 2 [<null>] (/tmp/a.out)]
63397 Thread 2.1 "a.out" hit Breakpoint 2, main () at file.c:3
63399 [Switching to inferior 1 [process 8263] (/tmp/test)]
63400 [Switching to thread 1.1 (process 8263)]
63401 #0 main () at test.c:2
63404 You can use the new "set suppress-cli-notifications" command to
63405 suppress the output:
63407 (gdb) set suppress-cli-notifications on
63409 Setting up Inferior 2...
63411 Added inferior 2 on connection 1 (native)
63412 Breakpoint 2 at 0x1155: file file.c, line 3.
63415 2022-02-07 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
63417 gdb/cli: add a 'normal_stop' option to 'cli_suppress_notification'
63418 Extend the 'cli_suppress_notification' struct with a new field,
63419 'normal_stop', that can be used for checking if printing normal stop
63420 events on the CLI should be suppressed.
63422 This patch only introduces the flag. The subsequent patch adds a user
63423 command to turn the flag off/on.
63425 2022-02-07 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
63427 gdb/cli: convert cli_suppress_notification from int to bool
63428 Convert the suppress_notification flag for the CLI from int to bool.
63430 2022-02-07 Alan Modra <amodra@gmail.com>
63432 Revert "elf: Remove the 1-page gap before the RELRO segment"
63433 This reverts commit 2f83249c13d86065b4c7cdb198ea871017b4bba1.
63436 * ldlang.c (lang_size_relro_segment_1): Revert 2022-01-10 changes.
63437 * testsuite/ld-i386/pr20830.d: Likewise.
63438 * testsuite/ld-s390/gotreloc_64-relro-1.dd: Likewise.
63439 * testsuite/ld-x86-64/pr14207.d: Likewise.
63440 * testsuite/ld-x86-64/pr18176.d: Likewise.
63441 * testsuite/ld-x86-64/pr20830a-now.d: Likewise.
63442 * testsuite/ld-x86-64/pr20830a.d: Likewise.
63443 * testsuite/ld-x86-64/pr20830b-now.d: Likewise.
63444 * testsuite/ld-x86-64/pr20830b.d: Likewise.
63445 * testsuite/ld-x86-64/pr21038a-now.d: Likewise.
63446 * testsuite/ld-x86-64/pr21038a.d: Likewise.
63447 * testsuite/ld-x86-64/pr21038b-now.d: Likewise.
63448 * testsuite/ld-x86-64/pr21038c-now.d: Likewise.
63449 * testsuite/ld-x86-64/pr21038c.d: Likewise.
63451 2022-02-07 Alan Modra <amodra@gmail.com>
63453 Revert "ld: Rewrite lang_size_relro_segment_1"
63454 This reverts commit c804c6f98d342c3d46f73d7a7ec6229d5ab1c9f3.
63458 * ldlang.c (lang_size_relro_segment_1): Revert 2022-01-14 change.
63459 * testsuite/ld-x86-64/pr28743-1.d: Likewise.
63460 * testsuite/ld-x86-64/pr28743-1.s: Likewise.
63461 * testsuite/ld-x86-64/x86-64.exp: Likewise.
63463 2022-02-07 GDB Administrator <gdbadmin@sourceware.org>
63465 Automatic date update in version.in
63467 2022-02-06 Alan Modra <amodra@gmail.com>
63469 A more elegant pr28827-1 testcase
63470 Use .irpc macros in pr28827-1.s
63472 * testsuite/ld-powerpc/pr28827-1.s: Make the testcase more
63475 2022-02-06 Tom Tromey <tom@tromey.com>
63477 Merge do_val_print and common_val_print
63478 The only caller of do_val_print just does a small bit of work before
63479 the call. This patch merges the two functions, and removes an
63480 unnecessary local variable, making gdb a bit simpler.
63482 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63484 gdb: remove SYMBOL_LINE macro
63485 Add a getter and a setter for a symbol's line. Remove the corresponding macro
63486 and adjust all callers.
63488 Change-Id: I229f2b8fcf938c07975f641361313a8761fad9a5
63490 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63492 gdb: remove SYMBOL_TYPE macro
63493 Add a getter and a setter for a symbol's type. Remove the corresponding
63494 macro and adjust all callers.
63496 Change-Id: Ie1a137744c5bfe1df4d4f9ae5541c5299577c8de
63498 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63500 gdb: remote SYMBOL_IS_CPLUS_TEMPLATE_FUNCTION macro
63501 Add a getter for a whether a symbol is a C++ template function. Remove
63502 the corresponding macro and adjust all callers.
63504 Change-Id: I89abc2802a952b77b0e0eb73a25c2306cb8e8bcc
63506 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63508 gdb: remove SYMBOL_INLINED macro
63509 Add a getter and a setter for whether a symbol is inlined. Remove the
63510 corresponding macro and adjust all callers.
63512 Change-Id: I934468da3b5a32dd6b161a6f252a6b1b94737279
63514 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63516 gdb: remove SYMBOL_IS_ARGUMENT macro
63517 Add a getter and a setter for whether a symbol is an argument. Remove
63518 the corresponding macro and adjust all callers.
63520 Change-Id: I71b4f0465f3dfd2ed8b9e140bd3f7d5eb8d9ee81
63522 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63524 gdb: remove SYMBOL_OBJFILE_OWNED macro
63525 Add a getter and a setter for whether a symbol is objfile owned. Remove
63526 the corresponding macro and adjust all callers.
63528 Change-Id: Ib7ef3718d65553ae924ca04c3fd478b0f4f3147c
63530 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63532 gdb: remove SYMBOL_DOMAIN macro
63533 Add a getter and a setter for a symbol's domain. Remove the
63534 corresponding macro and adjust all callers.
63536 Change-Id: I54465b50ac89739c663859a726aef8cdc6e4b8f3
63538 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63540 gdb: remove SYMBOL_CLASS macro, add getter
63541 Change-Id: I83211d5a47efc0564386e5b5ea4a29c00b1fd46a
63543 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63545 gdb: remove SYMBOL_IMPL macro, add method
63546 Add a getter for a symbol's "impl". Remove the corresponding macro and
63547 adjust all callers.
63549 Change-Id: Ibe26ed442f0f99a0f5cddafca30bd96ec7fb9fa8
63551 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63553 gdb: remove SYMBOL_ACLASS_INDEX macro, add getter/setter
63554 Add a getter and a setter for a symbol's aclass index. Remove the
63555 corresponding macro and adjust all callers.
63557 Change-Id: Ie8c8d732624cfadb714aba5ddafa3d29409b3d39
63559 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63561 gdb: remove SYMBOL_MATCHES_SEARCH_NAME
63562 It seems like this macro is not needed at all anymore, it just wraps the
63563 function of the same name with the same arguments.
63565 Change-Id: I3c342ac8d89c27af5aee1a819dc32cc6396fd41b
63567 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63569 gdb: remove SYMTAB_DIRNAME macro
63570 Remove the macro, replace with an equivalent method.
63572 Change-Id: I46ec36b91bb734331138eb9cd086b2db01635aed
63574 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63576 gdb: remove SYMTAB_PSPACE macro
63577 Remove the macro, replace with an equivalent method.
63579 Change-Id: Icccc20e7e8ae03ac4dac1c7514c25a12a9a0ac69
63581 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63583 gdb: remove SYMTAB_OBJFILE macro
63584 Remove the macro, replace with an equivalent method.
63586 Change-Id: I8f9ecd290ad28502e53c1ceca5006ba78bf042eb
63588 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63590 gdb: remove SYMTAB_BLOCKVECTOR macro
63591 Remove the macro, replace with an equivalent method.
63593 Change-Id: Id6fe2a79c04bcd6c69ccaefb7a69bc06a476288c
63595 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63597 gdb: remove SYMTAB_LANGUAGE macro, add getter/setter
63598 Add a getter and a setter for a symtab's language. Remove the
63599 corresponding macro and adjust all callers.
63601 Change-Id: I9f4d840b11c19f80f39bac1bce020fdd1739e11f
63603 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63605 gdb: remove SYMTAB_LINETABLE macro, add getter/setter
63606 Add a getter and a setter for a symtab's linetable. Remove the
63607 corresponding macro and adjust all callers.
63609 Change-Id: I159183fc0ccd8e18ab937b3c2f09ef2244ec6e9c
63611 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63613 gdb: remove SYMTAB_COMPUNIT macro, add getter/setter
63614 Add a getter and a setter for a symtab's compunit_symtab. Remove the
63615 corresponding macro and adjust all callers.
63617 For brevity, I chose the name "compunit" instead of "compunit_symtab"
63618 the the field, getter and setter names. Since we are already in symtab
63619 context, the _symtab suffix seems redundant.
63621 Change-Id: I4b9b731c96e3594f7733e75af1e3d01bc0e4fe92
63623 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63625 gdb: remove COMPUNIT_MACRO_TABLE macro, add getter/setter
63626 Add a getter and a setter for a compunit_symtab's macro table. Remove the
63627 corresponding macro and adjust all callers.
63629 Change-Id: I00615ea72d5ac43d9a865e941cb2de0a979c173a
63631 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63633 gdb: remove COMPUNIT_EPILOGUE_UNWIND_VALID macro, add getter/setter
63634 Add a getter and a setter for a compunit_symtab's epilogue unwind valid flag.
63635 Remove the corresponding macro and adjust all callers.
63637 Change-Id: If3b68629d987767da9be7041a95d96dc34367a9a
63639 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63641 gdb: remove COMPUNIT_LOCATIONS_VALID macro, add getter/setter
63642 Add a getter and a setter for a compunit_symtab's locations valid flag.
63643 Remove the corresponding macro and adjust all callers.
63645 Change-Id: I3e3cfba926ce62993d5b61814331bb3244afad01
63647 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63649 gdb: remove COMPUNIT_BLOCK_LINE_SECTION macro, add getter/setter
63650 Add a getter and a setter for a compunit_symtab's block line section. Remove
63651 the corresponding macro and adjust all callers.
63653 Change-Id: I3eb1a323388ad55eae8bfa45f5bc4a08dc3df455
63655 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63657 gdb: remove COMPUNIT_BLOCKVECTOR macro, add getter/setter
63658 Add a getter and a setter for a compunit_symtab's blockvector. Remove
63659 the corresponding macro and adjust all callers.
63661 Change-Id: I99484c6619dcbbea7c5d89c72aa660316ca62f64
63663 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63665 gdb: remove COMPUNIT_DIRNAME macro, add getter/setter
63666 Add a getter and a setter for a compunit_symtab's dirname. Remove the
63667 corresponding macro and adjust all callers.
63669 Change-Id: If2f39b295fd26822586485e04a8b8b5aa5cc9b2e
63671 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63673 gdb: remove COMPUNIT_PRODUCER macro, add getter/setter
63674 Add a getter and a setter for a compunit_symtab's producer. Remove the
63675 corresponding macro and adjust all callers.
63677 Change-Id: Ia1d6d8a0e247a08a21af23819d71e49b37d8931b
63679 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63681 gdb: remove COMPUNIT_DEBUGFORMAT macro, add getter/setter
63682 Add a getter and a setter for a compunit_symtab's debugformat. Remove
63683 the corresponding macro and adjust all callers.
63685 Change-Id: I1667b02d5322346f8e23abd9f8a584afbcd75975
63687 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63689 gdb: remove COMPUNIT_FILETABS macro
63690 I think that most remaining uses of COMPUNIT_FILETABS intend to get the
63691 primary filetab of the compunit_symtab specifically (and not to iterate
63692 over all filetabs, for example, those cases would use compunit_filetabs,
63693 which has been converted to compunit_symtab::filetabs), so replace mosts
63694 uses with compunit_symtab::primary_filetab.
63696 In jit.c, function finalize_symtab, we can save the symtab object
63697 returned by allocate_symtab and use it, it makes things simpler.
63699 Change-Id: I4e51d6d4b40759de8768b61292e5e13c8eae2e38
63701 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63703 gdb: move compunit_filetabs to compunit_symtab::filetabs
63704 Make compunit_filetabs, used to iterate a compunit_symtab's filetabs, a
63705 method of compunit_symtab. The name filetabs conflicts with the current
63706 name of the field. Rename the field to m_filetabs, since at this point
63707 nothing outside of compunit_symtab uses it, so we should treat it as
63708 private (even though it's not actually private). Rename the
63709 last_filetab field to m_last_filetab as well (it's only used on
63710 compunit_symtab::add_filetab).
63712 Adjust the COMPUNIT_FILETABS macro to keep its current behavior of
63713 returning the first filetab.
63715 Change-Id: I537b553a44451c52d24b18ee1bfa47e23747cfc3
63717 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63719 gdb: add compunit_symtab::set_primary_filetab method
63720 Add a method to set the primary filetab of the CU. This is currently
63721 done in buildsym_compunit::end_symtab_with_blockvector.
63723 Change-Id: I16c51a6b90a4cb4c0c5f183b0f2e12bc64b6fd74
63725 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63727 gdb: add compunit_symtab::add_filetab method
63728 Add a method to append a filetab/symtab to a compunit_symtab. There is
63729 a single place where this is done currently, in allocate_symtab.
63731 Change-Id: Ie86c6e34d175728173d1cffdce44acd6cff6c31d
63733 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63735 gdb: rename compunit_primary_filetab to compunit_symtab::primary_filetab
63736 Make compunit_primary_filetab a method of compunit_symtab.
63738 Change-Id: Iee3c4f7e36d579bf763c5bba146e5e10d6766768
63740 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63742 gdb: remove COMPUNIT_OBJFILE macro
63743 Remove the macro, update all users to use the getter directly.
63745 Change-Id: I3f0fd6f4455d1c4ebd5da73b561eb18a979ef1f6
63747 2022-02-06 Simon Marchi <simon.marchi@efficios.com>
63749 gdb: add getter/setter for compunit_symtab::objfile
63750 Rename the field to m_objfile, and add a getter and a setter. Update
63753 Change-Id: If7e2f763ee3e70570140d9af9261b1b056253317
63755 2022-02-06 Tom Tromey <tom@tromey.com>
63757 Allow non-ASCII characters in Rust identifiers
63758 Rust 1.53 (quite a while ago now) ungated the support for non-ASCII
63759 identifiers. This didn't work in gdb. This is PR rust/20166.
63761 This patch fixes the problem by allowing non-ASCII characters to be
63762 considered as identifier components. It seemed simplest to just pass
63763 them through -- doing any extra checking didn't seem worthwhile.
63765 The new test also verifies that such characters are allowed in strings
63766 and character literals as well. The latter also required a bit of
63769 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=20166
63771 2022-02-06 Tom Tromey <tom@tromey.com>
63773 Fix Rust parser bug with function fields
63774 In Rust, 'obj.f()' is a method call -- but '(obj.f)()' is a call of a
63775 function-valued field 'f' in 'obj'. The Rust parser in gdb currently
63776 gets this wrong. This is PR rust/24082.
63778 The expression and Rust parser rewrites made this simple to fix --
63779 simply wrapping a parenthesized expression in a new operation handles
63780 it. This patch has a slight hack because I didn't want to introduce a
63781 new exp_opcode enumeration constant just for this. IMO this doesn't
63782 matter, since we should work toward removing dependencies on these
63783 opcodes anyway; but let me know what you think of this.
63785 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=24082
63787 2022-02-06 H.J. Lu <hjl.tools@gmail.com>
63789 ld: Add emultempl/emulation.em
63790 Add emultempl/emulation.em to define ld_${EMULATION_NAME}_emulation so
63791 that new emulation hooks can be added easily.
63793 * emultempl/aix.em (LDEMUL_AFTER_OPEN): New.
63794 (LDEMUL_SET_OUTPUT_ARCH): Likewise.
63795 (LDEMUL_CHOOSE_TARGET): Likewise.
63796 (LDEMUL_BEFORE_ALLOCATION): Likewise.
63797 (LDEMUL_CREATE_OUTPUT_SECTION_STATEMENTS): Likewise.
63798 (LDEMUL_OPEN_DYNAMIC_ARCHIVE): Likewise.
63799 (LDEMUL_PARSE_ARGS): Likewise.
63800 (LDEMUL_ADD_OPTIONS): Likewise.
63801 (LDEMUL_HANDLE_OPTION): Likewise.
63802 (LDEMUL_UNRECOGNIZED_FILE): Likewise.
63803 (LDEMUL_PRINT_SYMBOL): Likewise.
63804 (ld_${EMULATION_NAME}_emulation): Removed.
63805 Source ${srcdir}/emultempl/emulation.em.
63806 * emultempl/beos.em (gld_${EMULATION_NAME}_before_parse):
63808 (gld${EMULATION_NAME}_before_parse): This.
63809 (gld_${EMULATION_NAME}_set_symbols): Renamed to ...
63810 (gld${EMULATION_NAME}_set_symbols): This.
63811 (gld_${EMULATION_NAME}_after_open): Renamed to ...
63812 (gld${EMULATION_NAME}_after_open): This.
63813 (gld_${EMULATION_NAME}_before_allocation): Renamed to ...
63814 (gld${EMULATION_NAME}_before_allocation): This.
63815 (gld_${EMULATION_NAME}_get_script): Renamed to ...
63816 (gld${EMULATION_NAME}_get_script): This.
63817 (LDEMUL_AFTER_OPEN): New.
63818 (LDEMUL_BEFORE_ALLOCATION): Likewise.
63819 (LDEMUL_PLACE_ORPHAN): Likewise.
63820 (LDEMUL_SET_SYMBOLS): Likewise.
63821 (LDEMUL_ADD_OPTIONS): Likewise.
63822 (LDEMUL_HANDLE_OPTION): Likewise.
63823 (ld_${EMULATION_NAME}_emulation): Removed.
63824 Source ${srcdir}/emultempl/emulation.em.
63825 * emultempl/elf.em (LDEMUL_AFTER_PARSE): New.
63826 (LDEMUL_AFTER_OPEN): Likewise.
63827 (LDEMUL_BEFORE_PLACE_ORPHANS): Likewise.
63828 (LDEMUL_AFTER_ALLOCATION): Likewise.
63829 (LDEMUL_SET_OUTPUT_ARCH): Likewise.
63830 (LDEMUL_BEFORE_ALLOCATION): Likewise.
63831 (LDEMUL_OPEN_DYNAMIC_ARCHIVE): Likewise.
63832 (LDEMUL_PLACE_ORPHAN): Likewise.
63833 (LDEMUL_ADD_OPTIONS): Likewise.
63834 (LDEMUL_HANDLE_OPTION): Likewise.
63835 (LDEMUL_LIST_OPTIONS): Likewise.
63836 (LDEMUL_UNRECOGNIZED_FILE): Likewise.
63837 (ld_${EMULATION_NAME}_emulation): Removed.
63838 Source ${srcdir}/emultempl/emulation.em.
63839 * emultempl/emulation.em: New file.
63840 * emultempl/generic.em (ld_${EMULATION_NAME}_emulation): Removed.
63841 Source ${srcdir}/emultempl/emulation.em.
63842 * emultempl/msp430.em (LDEMUL_AFTER_OPEN): New.
63843 (LDEMUL_AFTER_ALLOCATION): Likewise.
63844 (LDEMUL_PLACE_ORPHAN): Likewise.
63845 (LDEMUL_FINISH): Likewise.
63846 (LDEMUL_ADD_OPTIONS): Likewise.
63847 (LDEMUL_HANDLE_OPTION): Likewise.
63848 (LDEMUL_LIST_OPTIONS): Likewise.
63849 (ld_${EMULATION_NAME}_emulation): Removed.
63850 Source ${srcdir}/emultempl/emulation.em.
63851 * emultempl/pe.em (gld_${EMULATION_NAME}_before_parse): Renamed
63853 (gld${EMULATION_NAME}_before_parse): This.
63854 (gld_${EMULATION_NAME}_list_options): Renamed to ...
63855 (gld${EMULATION_NAME}_list_options): This.
63856 (gld_${EMULATION_NAME}_set_symbols): Renamed to ...
63857 (gld${EMULATION_NAME}_set_symbols): This.
63858 (gld_${EMULATION_NAME}_after_parse): Renamed to ...
63859 (gld${EMULATION_NAME}_after_parse): This.
63860 (gld_${EMULATION_NAME}_after_open): Renamed to ...
63861 (gld${EMULATION_NAME}_after_open): This.
63862 (gld_${EMULATION_NAME}_before_allocation): Renamed to ...
63863 (gld${EMULATION_NAME}_before_allocation): This.
63864 (gld_${EMULATION_NAME}_unrecognized_file): Renamed to ...
63865 (gld${EMULATION_NAME}_unrecognized_file): This.
63866 (gld_${EMULATION_NAME}_recognized_file): Renamed to ...
63867 (gld${EMULATION_NAME}_recognized_file): This.
63868 (gld_${EMULATION_NAME}_finish): Renamed to ...
63869 (gld${EMULATION_NAME}_finish): This.
63870 (gld_${EMULATION_NAME}_place_orphan): Renamed to ...
63871 (gld${EMULATION_NAME}_place_orphan): This.
63872 (gld_${EMULATION_NAME}_open_dynamic_archive): Renamed to ...
63873 (gld${EMULATION_NAME}_open_dynamic_archive): This.
63874 (gld_${EMULATION_NAME}_find_potential_libraries): Renamed to ...
63875 (gld${EMULATION_NAME}_find_potential_libraries): This.
63876 (gld_${EMULATION_NAME}_get_script): Renamed to ...
63877 (gld${EMULATION_NAME}_get_script): This.
63878 (LDEMUL_AFTER_PARSE): New.
63879 (LDEMUL_AFTER_OPEN): Likewise.
63880 (LDEMUL_BEFORE_ALLOCATION): Likewise.
63881 (LDEMUL_FINISH=): Likewise.
63882 (LDEMUL_OPEN_DYNAMIC_ARCHIVE): Likewise.
63883 (LDEMUL_PLACE_ORPHAN): Likewise.
63884 (LDEMUL_SET_SYMBOLS): Likewise.
63885 (LDEMUL_ADD_OPTIONS): Likewise.
63886 (LDEMUL_HANDLE_OPTION): Likewise.
63887 (LDEMUL_UNRECOGNIZED_FILE): Likewise.
63888 (LDEMUL_LIST_OPTIONS): Likewise.
63889 (LDEMUL_RECOGNIZED_FILE): Likewise.
63890 (LDEMUL_FIND_POTENTIAL_LIBRARIES): Likewise.
63891 (ld_${EMULATION_NAME}_emulation): Removed.
63892 Source ${srcdir}/emultempl/emulation.em.
63893 * emultempl/pep.em (gld_${EMULATION_NAME}_before_parse): Renamed
63895 (gld${EMULATION_NAME}_before_parse): This.
63896 (gld_${EMULATION_NAME}_list_options): Renamed to ...
63897 (gld${EMULATION_NAME}_list_options): This.
63898 (gld_${EMULATION_NAME}_set_symbols): Renamed to ...
63899 (gld${EMULATION_NAME}_set_symbols): This.
63900 (gld_${EMULATION_NAME}_after_parse): Renamed to ...
63901 (gld${EMULATION_NAME}_after_parse): This.
63902 (gld_${EMULATION_NAME}_after_open): Renamed to ...
63903 (gld${EMULATION_NAME}_after_open): This.
63904 (gld_${EMULATION_NAME}_before_allocation): Renamed to ...
63905 (gld${EMULATION_NAME}_before_allocation): This.
63906 (gld_${EMULATION_NAME}_unrecognized_file): Renamed to ...
63907 (gld${EMULATION_NAME}_unrecognized_file): This.
63908 (gld_${EMULATION_NAME}_recognized_file): Renamed to ...
63909 (gld${EMULATION_NAME}_recognized_file): This.
63910 (gld_${EMULATION_NAME}_finish): Renamed to ...
63911 (gld${EMULATION_NAME}_finish): This.
63912 (gld_${EMULATION_NAME}_place_orphan): Renamed to ...
63913 (gld${EMULATION_NAME}_place_orphan): This.
63914 (gld_${EMULATION_NAME}_open_dynamic_archive): Renamed to ...
63915 (gld${EMULATION_NAME}_open_dynamic_archive): This.
63916 (gld_${EMULATION_NAME}_find_potential_libraries): Renamed to ...
63917 (gld${EMULATION_NAME}_find_potential_libraries): This.
63918 (gld_${EMULATION_NAME}_get_script): Renamed to ...
63919 (gld${EMULATION_NAME}_get_script): This.
63920 (LDEMUL_AFTER_PARSE): New.
63921 (LDEMUL_AFTER_OPEN): Likewise.
63922 (LDEMUL_BEFORE_ALLOCATION): Likewise.
63923 (LDEMUL_FINISH=): Likewise.
63924 (LDEMUL_OPEN_DYNAMIC_ARCHIVE): Likewise.
63925 (LDEMUL_PLACE_ORPHAN): Likewise.
63926 (LDEMUL_SET_SYMBOLS): Likewise.
63927 (LDEMUL_ADD_OPTIONS): Likewise.
63928 (LDEMUL_HANDLE_OPTION): Likewise.
63929 (LDEMUL_UNRECOGNIZED_FILE): Likewise.
63930 (LDEMUL_LIST_OPTIONS): Likewise.
63931 (LDEMUL_RECOGNIZED_FILE): Likewise.
63932 (LDEMUL_FIND_POTENTIAL_LIBRARIES): Likewise.
63933 (ld_${EMULATION_NAME}_emulation): Removed.
63934 Source ${srcdir}/emultempl/emulation.em.
63935 * emultempl/ticoff.em (gld_${EMULATION_NAME}_list_options):
63937 (gld${EMULATION_NAME}_list_options): This.
63938 (gld_${EMULATION_NAME}_before_parse): Renamed to ...
63939 (gld_${EMULATION_NAME}_get_script): Renamed to ...
63940 (gld${EMULATION_NAME}_get_script): This.
63941 (LDEMUL_ADD_OPTIONS): New.
63942 (LDEMUL_HANDLE_OPTION): Likewise.
63943 (LDEMUL_LIST_OPTIONS): Likewise.
63944 (ld_${EMULATION_NAME}_emulation): Removed.
63945 Source ${srcdir}/emultempl/emulation.em.
63946 * emultempl/vanilla.em (LDEMUL_BEFORE_PARSE): New.
63947 (LDEMUL_SET_OUTPUT_ARCH): Likewise.
63948 (LDEMUL_GET_SCRIPT): Likewise.
63949 (EMULATION_NAME): Likewise.
63950 (OUTPUT_FORMAT): Likewise.
63951 (ld_vanilla_emulation): Removed.
63952 Source ${srcdir}/emultempl/emulation.em.
63954 2022-02-06 Andrew Burgess <aburgess@redhat.com>
63956 gdb/doc: update docs for 'info win' and 'winheight' commands
63957 This started by noticing that the docs for 'winheight' are out of
63958 date, the docs currently give a specific list of possible window
63959 names. However, now that windows can be implemented in Python, it is
63960 not possible to list all possible names.
63962 I now link the user to a mechanism by which they can discover the
63963 valid names for themselves at run time (by using 'info win'). That,
63964 and the fact that gdb provides tab-completion of the name at the
63965 command line, feels good enough.
63967 Finally, I noticed that the docs for 'win info' don't explicitly say
63968 that the name of the window is given in the output. This could
63969 probably have been inferred, but given I'm now linking to this as a
63970 mechanism to find the window name, I'd prefer to mention that the name
63971 can be found in the output.
63973 2022-02-06 Andrew Burgess <aburgess@redhat.com>
63975 gdb/tui: add window width information to 'info win' output
63976 Now that we support horizontal window placement in the tui, it makes
63977 sense to have 'info win' include the width, as well as the height, of
63978 the currently visible windows.
63980 That's what this commit does. Example output is now:
63983 Name Lines Columns Focus
63984 src 12 40 (has focus)
63989 I've added a NEWS entry, but the documentation was already suitably
63990 vague, it just says that 'info win' displays the size of the visible
63991 windows, so I don't think anything needs to be added there.
63993 I've also added some tests, as far as I could find, the 'info win'
63994 command was previously untested.
63996 2022-02-06 GDB Administrator <gdbadmin@sourceware.org>
63998 Automatic date update in version.in
64000 2022-02-05 H.J. Lu <hjl.tools@gmail.com>
64002 x86: Skip undefined symbol when finishing DT_RELR
64003 Don't abort for undefined symbol when finishing DT_RELR. Instead, skip
64004 undefined symbol. Undefined symbol will be reported by relocate_section.
64006 * elfxx-x86.c (elf_x86_size_or_finish_relative_reloc): Skip
64007 undefined symbol in finishing phase.
64009 2022-02-05 Alan Modra <amodra@gmail.com>
64011 Tweak assembler invocation for pr28827-1 test
64013 * testsuite/ld-powerpc/pr28827-1.d: Pass -a64 to gas.
64015 2022-02-05 Alan Modra <amodra@gmail.com>
64018 This testcase triggers a stub sizing error with the patches applied
64019 for PR28743 (commit 2f83249c13d8 and c804c6f98d34).
64022 * testsuite/ld-powerpc/pr28827-1.s,
64023 * testsuite/ld-powerpc/pr28827-1.d: New test.
64024 * testsuite/ld-powerpc/powerpc.exp: Run it.
64026 2022-02-05 Alan Modra <amodra@gmail.com>
64028 Enable "size" as a dumpprog in ld
64030 * testsuite/lib/binutils-common.exp (run_dump_test): Reference
64031 global SIZE and SIZEFLAGS.
64033 * testsuite/config/default.exp: Define SIZE and SIZEFLAGS.
64035 2022-02-05 Alan Modra <amodra@gmail.com>
64037 Detect .eh_frame_hdr earlier for SIZEOF_HEADERS
64038 Current code detects the need for PT_GNU_EH_FRAME using a field set by
64039 _bfd_elf_discard_section_eh_frame_hdr, which is called fairly late in
64040 the linking process. Use the elf hash table eh_info instead, which is
64041 set up earlier by size_dynamic_sections.
64043 * elf-bfd.h (struct output_elf_obj_tdata): Delete eh_frame_hdr.
64044 (elf_eh_frame_hdr): Don't define.
64045 (_bfd_elf_discard_section_eh_frame_hdr): Update prototype.
64046 * elf-eh-frame.c (_bfd_elf_discard_section_eh_frame_hdr): Delete
64047 abfd parameter. Don't set elf_eh_frame_hdr.
64048 * elf.c (elf_eh_frame_hdr): New function.
64049 (get_program_header_size): Adjust elf_eh_frame_hdr call.
64050 (_bfd_elf_map_sections_to_segments): Likewise.
64052 2022-02-05 Faraz Shahbazker <fshahbazker@wavecomp.com>
64054 sim: mips: Add simulator support for mips32r6/mips64r6
64055 2022-02-01 Ali Lown <ali.lown@imgtec.com>
64056 Andrew Bennett <andrew.bennett@imgtec.com>
64057 Dragan Mladjenovic <dragan.mladjenovic@rt-rk.com>
64058 Faraz Shahbazker <fshahbazker@wavecomp.com>
64060 sim/common/ChangeLog:
64061 * sim-bits.h (EXTEND9, EXTEND18 ,EXTEND19, EXTEND21,
64062 EXTEND26): New macros.
64064 sim/mips/ChangeLog:
64065 * Makefile.in (IGEN_INCLUDE): Add mips3264r6.igen.
64066 * configure: Regenerate.
64067 * configure.ac: Support mipsisa32r6 and mipsisa64r6.
64068 (sim_engine_run): Pick simulator model from processor specified
64070 * cp1.c (value_fpr): Handle fmt_dc32.
64071 (fp_unary, fp_binary): Zero initialize locals.
64072 (update_fcsr, fp_classify, fp_rint, fp_r6_cmp, inner_fmac,
64073 fp_fmac, fp_min, fp_max, fp_mina, fp_maxa, fp_fmadd, fp_fmsub):
64075 (sim_fpu_class_mips_mapping): New.
64076 * cp1.h (fcsr_ABS2008_mask, fcsr_ABS2008_shift): New define.
64077 * interp.c (MIPSR6_P): New.
64078 (load_word): Allow unaligned memory access for MIPSR6.
64079 * micromips.igen (sc, scd): Adapt to new do_sc* helper signature.
64080 * mips.igen: Add *r6 models.
64081 (signal_if_cti, forbiddenslot32): New helpers.
64082 (delayslot32): Use signal_if_cti.
64083 (do_sc, do_scd); Add store_ll_bit parameter.
64084 (sc, scd): Adapt to previous change.
64085 (nal, beq, bal): New definitions for *r6.
64086 (sll): Split nop and ssnop cases into ...
64087 (nop, ssnop): New definitions.
64088 (loadstore_ea): Use the 32-bit compatibility adressing.
64089 (cache): Split logic into ...
64090 (do_cache): New helper.
64091 (check_fpu): Select IEEE 754-2008 mode for R6.
64092 (not_word_value, unpredictable, check_mt_hilo, check_mf_hilo,
64093 check_multi_hilo, check_div_hilo, check_u64, do_dmfc1b, add,
64094 li, addu, and, andi, bgez, bgtz, blez, bltz, bne, break, dadd,
64095 daddiu, daddu, dror, dror32, drorv, dsll, dsll32, dsllv, dsra,
64096 dsra32, dsrav, dsrl, dsrl32, dsub, dsubu, j, jal, jalr,
64097 jalr.hb, lb, lbu, ld, lh, lhu, lui, lw, lwu, nor, or, ori, ror,
64098 rorv, sb, sd, sh, sll, sllv, slt, slti, sltiu, sltu, sra, srav,
64099 srl, srlv, sub, subu, sw, sync, syscall, teq, tge, tgeu, tlt,
64100 tltu, tne, xor, xori, check_fmt_p, do_load_double,
64101 do_store_double, abs.FMT, add.FMT, ceil.l.FMT, ceil.w.FMT,
64102 cfc1, ctc1, cvt.d.FMT, cvt.l.FMT, cvt.w.FMT, div.FMT, dfmc1,
64103 dmtc1, floor.l.FMT, floor.w.FMT, ldc1, lwc1, mfc1, mov.FMT,
64104 mtc1, mul.FMT, recip.FMT, round.l.FMT, round.w.FMT, rsqrt.FMT,
64105 sdc1, sqrt.FMT, sub.FMT, swc1, trunc.l.FMT, trunc.w.FMT, bc0f,
64106 bc0fl, bc0t, bc0tl, dmfc0, dmtc0, eret, mfc0, mtc0, cop, tlbp,
64107 tlbr, tlbwi, tlbwr): Enable on *r6 models.
64108 * mips3264r2.igen (dext, dextm, dextu, di, dins, dinsm, dinsu,
64109 dsbh, dshd, ei, ext, mfhc1, mthc1, ins, seb, seh, synci, rdhwr,
64111 * mips3264r6.igen: New file.
64112 * sim-main.h (FP_formats): Add fmt_dc32.
64113 (FORBIDDEN_SLOT): New macros.
64114 (simFORBIDDENSLOT, FP_R6CMP_*, FP_R6CLASS_*): New defines.
64115 (fp_r6_cmp, fp_classify, fp_rint, fp_min, fp_max, fp_mina,
64116 fp_maxa, fp_fmadd, fp_fmsub): New declarations.
64117 (R6Compare, Classify, RoundToIntegralExact, Min, Max, MinA,
64118 MaxA, FusedMultiplyAdd, FusedMultiplySub): New macros. Wrapping
64119 previous declarations.
64121 sim/testsuite/mips/ChangeLog:
64122 * basic.exp: Add r6-*.s tests.
64123 (run_r6_removed_test): New function.
64124 (run_endian_tests): New function.
64125 * hilo-hazard-3.s: Skip for mips*r6.
64126 * r2-fpu.s: New test.
64127 * r6-64.s: New test.
64128 * r6-branch.s: New test.
64129 * r6-forbidden.s: New test.
64130 * r6-fpu.s: New test.
64131 * r6-llsc-dp.s: New test.
64132 * r6-llsc-wp.s: New test.
64133 * r6-removed.csv: New test.
64134 * r6-removed.s: New test.
64136 * utils-r6.inc: New inc.
64138 2022-02-05 Faraz Shahbazker <fshahbazker@wavecomp.com>
64140 sim: Add partial support for IEEE 754-2008
64141 2022-02-01 Faraz Shahbazker <fshahbazker@wavecomp.com>
64143 sim/common/ChangeLog:
64144 * sim-fpu.c (sim_fpu_minmax_nan): New.
64145 (sim_fpu_max): Add variant behaviour for IEEE 754-2008.
64146 (sim_fpu_min): Likewise.
64147 (sim_fpu_is_un, sim_fpu_is_or): New.
64148 (sim_fpu_un, sim_fpu_or): New.
64149 (sim_fpu_is_ieee754_2008, sim_fpu_is_ieee754_1985): New.
64150 (sim_fpu_set_mode): New.
64151 (sim_fpu_classify): New.
64152 * sim-fpu.h (sim_fpu_minmax_nan): New declaration.
64153 (sim_fpu_un, sim_fpu_or): New declarations.
64154 (sim_fpu_is_un, sim_fpu_is_or): New declarations.
64155 (sim_fpu_mode): New enum.
64156 [sim_fpu_state](current_mode): New field.
64157 (sim_fpu_current_mode): New define.
64158 (sim_fpu_is_ieee754_2008): New declaration.
64159 (sim_fpu_is_ieee754_1985): New declaration.
64160 (sim_fpu_set_mode): New declaration.
64161 (sim_fpu_classify): New declaration.
64163 2022-02-05 Faraz Shahbazker <fshahbazker@wavecomp.com>
64165 sim: Factor out NaN handling in floating point operations
64166 2022-02-01 Faraz Shahbazker <fshahbazker@wavecomp.com>
64168 sim/common/ChangeLog:
64169 * sim-fpu.c (sim_fpu_op_nan): New.
64170 (sim_fpu_add): Factor out NaN operand handling with
64171 a call to sim_fpu_op_nan.
64172 (sim_fpu_sub, sim_fpu_mul, sim_fpu_div): Likewise.
64173 (sim_fpu_rem, sim_fpu_max, sim_fpu_min): Likewise.
64174 * sim-fpu.h (sim_fpu_op_nan): New declaration.
64176 2022-02-05 Faraz Shahbazker <fshahbazker@wavecomp.com>
64178 sim: Allow toggling of quiet NaN-bit semantics
64179 IEEE754-1985 specifies the top bit of the mantissa as an indicator
64180 of signalling vs. quiet NaN, but does not define the precise semantics.
64181 Most architectures treat this bit as indicating quiet NaN, but legacy
64182 (pre-R6) MIPS goes the other way and treats it as signalling NaN.
64184 This used to be controlled by a macro that was only defined for MIPS.
64185 This patch replaces the macro with a variable to track the current
64186 semantics of the NaN bit and allows differentiation between older
64187 (pre-R6) and and newer MIPS cores.
64189 2022-02-01 Faraz Shahbazker <fshahbazker@wavecomp.com>
64191 sim/common/ChangeLog:
64192 * sim-fpu.c (_sim_fpu): New.
64193 (pack_fpu, unpack_fpu): Allow reversal of quiet NaN semantics.
64194 * sim-fpu.h (sim_fpu_state): New struct.
64195 (_sim_fpu): New extern.
64196 (sim_fpu_quiet_nan_inverted): New define.
64198 sim/mips/ChangeLog:
64199 * cp1.h (fcsr_NAN2008_mask, fcsr_NAN2008_shift): New.
64200 * mips.igen (check_fpu): Select default quiet NaN mode
64202 * sim-main.h (SIM_QUIET_NAN_NEGATED): Remove.
64204 2022-02-05 GDB Administrator <gdbadmin@sourceware.org>
64206 Automatic date update in version.in
64208 2022-02-04 H.J. Lu <hjl.tools@gmail.com>
64210 ld: Remove emultempl/armcoff.em
64211 Remove emultempl/armcoff.em which has been unused after
64213 commit 2ac93be706418f3b2aebeb22159a328023faed52
64214 Author: Alan Modra <amodra@gmail.com>
64215 Date: Mon Apr 16 20:33:36 2018 +0930
64217 Remove arm-aout and arm-coff support
64219 This also removes arm-netbsd (not arm-netbsdelf!), arm-openbsd, and
64220 arm-riscix. Those targets weren't on the obsolete list but they are
64221 all aout, and it doesn't make all that much sense to remove arm-aout
64222 without removing them too.
64224 * emultempl/armcoff.em: Removed.
64226 2022-02-04 Simon Marchi <simon.marchi@efficios.com>
64228 gdb: include jit_code_entry::symfile_addr value in names of objfiles created by jit reader API
64229 This commit includes the JIT object's symfile address in the names of
64230 objfiles created by JIT reader API (e.g., << JIT compiled code at
64231 0x7ffd8a0c77a0 >>). This allows one to at least differentiate one from
64234 The address is the one that the debugged program has put in
64235 jit_code_entry::symfile_addr, and that the JIT reader's read function
64236 receives. As we can see in gdb.base/jit-reader-host.c and
64237 gdb.base/jit-reader.c, that may not be the actual value of where the
64238 JIT-ed code is. But it is a value chosen by the author of the JIT
64239 engine and the JIT reader, so including this value in the objfile name
64240 may help them correlate the JIT objfiles created by with their logs /
64243 To access this field, we need to pass down a reference to the
64244 jit_code_entry. So make jit_dbg_reader_data a structure (instead of an
64245 alias for a CORE_ADDR) that includes the address of the code entry in
64246 the inferior's address space (the previous meaning of
64247 jit_dbg_reader_data) plus a reference to the jit_code_entry as read into
64248 GDB's address space. And while at it, pass down the gdbarch, so that we
64249 don't have to call target_gdbarch.
64251 Co-Authored-By: Jan Vrany <jan.vrany@labware.com>
64252 Change-Id: Ib26c4d1bd8de503d651aff89ad2e500cb312afa5
64254 2022-02-04 Tom Tromey <tromey@adacore.com>
64256 Improve Ada unchecked union type printing
64257 Currently, "ptype" of an Ada unchecked union may show a
64258 compiler-generated wrapper structure in its output. It's more
64259 Ada-like to elide this structure, which is what this patch implements.
64260 It turned out to be simplest to reuse a part of print_variant_clauses
64263 As this is Ada-specific, and Joel already reviewed it internally, I am
64264 going to check it in.
64266 2022-02-04 Tom Tromey <tromey@adacore.com>
64268 Remove host_hex_value
64269 I noticed that host_hex_value is redundant, because gdbsupport already
64270 has fromhex. This patch removes the former in favor of the latter.
64272 Regression tested on x86-64 Fedora 34.
64274 2022-02-04 Andi Kleen <andi@firstfloor.org>
64276 Support symbol+offset lookup in addr2line
64277 The Linux kernel usually ouputs symbol+offset instead of plain code
64278 addresses these days, to avoid leaking ASLR secrets and to handle
64279 dynamically loaded modules.
64281 Converting those with addr2line is somewhat involved: it requires
64282 looking up the symbol first using nm and then manually compute the
64283 offset, and then pass it to addr2line.
64285 This patch implements the necessary steps directly in addr2line,
64286 by looking up the symbol (with demangling if needed) and computing
64289 It's possible that a symbol is ambigious with a hex number. In this
64290 case it uses the symbol lookup if the string contains a +. When it isn't
64291 ambigious the + is optional.
64293 2022-02-04 GDB Administrator <gdbadmin@sourceware.org>
64295 Automatic date update in version.in
64297 2022-02-03 Cary Coutant <ccoutant@gmail.com>
64299 Rename EM_56800V4 to EM_56800EF.
64301 * common.h: Rename EM_56800V4 to EM_56800EF.
64303 2022-02-03 H.J. Lu <hjl.tools@gmail.com>
64305 x86: Update X86_64_GOT_TYPE_P to cover more GOT relocations
64306 Add R_X86_64_GOT32, R_X86_64_GOT64, R_X86_64_GOTPCREL64 and
64307 R_X86_64_GOTPLT64 to X86_64_GOT_TYPE_P to cover more GOT relocations.
64310 * elfxx-x86.h (X86_64_GOT_TYPE_P): Add R_X86_64_GOT32,
64311 R_X86_64_GOT64, R_X86_64_GOTPCREL64 and R_X86_64_GOTPLT64.
64313 2022-02-03 Cary Coutant <ccoutant@gmail.com>
64315 Add new e_machine values.
64317 * common.h: Add EM_U16_U8CORE, EM_TACHYUM, EM_56800V4.
64319 2022-02-03 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
64321 testsuite: fix failure in gdb.threads/killed-outside.exp
64322 Starting with commit
64324 commit 1da5d0e664e362857153af8682321a89ebafb7f6
64325 Date: Tue Jan 4 08:02:24 2022 -0700
64327 Change how Python architecture and language are handled
64329 we see a failure in gdb.threads/killed-outside.exp:
64332 Executing on target: kill -9 16622 (timeout = 300)
64333 builtin_spawn -ignore SIGHUP kill -9 16622
64336 Couldn't get registers: No such process.
64337 (gdb) [Thread 0x7ffff77c2700 (LWP 16626) exited]
64339 Program terminated with signal SIGKILL, Killed.
64340 The program no longer exists.
64341 FAIL: gdb.threads/killed-outside.exp: prompt after first continue (timeout)
64343 This is not a regression but a failure due to a change in GDB's
64344 output. Prior to the aforementioned commit, GDB has been printing the
64345 "Couldn't get registers: No such process." message twice. The second
64349 #0 amd64_linux_nat_target::fetch_registers (this=0x555557f31440 <the_amd64_linux_nat_target>, regcache=0x555558805ce0, regnum=16) at /gdb-up/gdb/amd64-linux-nat.c:225
64350 #1 0x000055555640ac5f in target_ops::fetch_registers (this=0x555557d636d0 <the_thread_db_target>, arg0=0x555558805ce0, arg1=16) at /gdb-up/gdb/target-delegates.c:502
64351 #2 0x000055555641a647 in target_fetch_registers (regcache=0x555558805ce0, regno=16) at /gdb-up/gdb/target.c:3945
64352 #3 0x0000555556278e68 in regcache::raw_update (this=0x555558805ce0, regnum=16) at /gdb-up/gdb/regcache.c:587
64353 #4 0x0000555556278f14 in readable_regcache::raw_read (this=0x555558805ce0, regnum=16, buf=0x555558881950 "") at /gdb-up/gdb/regcache.c:601
64354 #5 0x00005555562792aa in readable_regcache::cooked_read (this=0x555558805ce0, regnum=16, buf=0x555558881950 "") at /gdb-up/gdb/regcache.c:690
64355 #6 0x000055555627965e in readable_regcache::cooked_read_value (this=0x555558805ce0, regnum=16) at /gdb-up/gdb/regcache.c:748
64356 #7 0x0000555556352a37 in sentinel_frame_prev_register (this_frame=0x555558181090, this_prologue_cache=0x5555581810a8, regnum=16) at /gdb-up/gdb/sentinel-frame.c:53
64357 #8 0x0000555555fa4773 in frame_unwind_register_value (next_frame=0x555558181090, regnum=16) at /gdb-up/gdb/frame.c:1235
64358 #9 0x0000555555fa420d in frame_register_unwind (next_frame=0x555558181090, regnum=16, optimizedp=0x7fffffffd570, unavailablep=0x7fffffffd574, lvalp=0x7fffffffd57c, addrp=0x7fffffffd580,
64359 realnump=0x7fffffffd578, bufferp=0x7fffffffd5b0 "") at /gdb-up/gdb/frame.c:1143
64360 #10 0x0000555555fa455f in frame_unwind_register (next_frame=0x555558181090, regnum=16, buf=0x7fffffffd5b0 "") at /gdb-up/gdb/frame.c:1199
64361 #11 0x00005555560178e2 in i386_unwind_pc (gdbarch=0x5555587c4a70, next_frame=0x555558181090) at /gdb-up/gdb/i386-tdep.c:1972
64362 #12 0x0000555555cd2b9d in gdbarch_unwind_pc (gdbarch=0x5555587c4a70, next_frame=0x555558181090) at /gdb-up/gdb/gdbarch.c:3007
64363 #13 0x0000555555fa3a5b in frame_unwind_pc (this_frame=0x555558181090) at /gdb-up/gdb/frame.c:948
64364 #14 0x0000555555fa7621 in get_frame_pc (frame=0x555558181160) at /gdb-up/gdb/frame.c:2572
64365 #15 0x0000555555fa7706 in get_frame_address_in_block (this_frame=0x555558181160) at /gdb-up/gdb/frame.c:2602
64366 #16 0x0000555555fa77d0 in get_frame_address_in_block_if_available (this_frame=0x555558181160, pc=0x7fffffffd708) at /gdb-up/gdb/frame.c:2665
64367 #17 0x0000555555fa5f8d in select_frame (fi=0x555558181160) at /gdb-up/gdb/frame.c:1890
64368 #18 0x0000555555fa5bab in lookup_selected_frame (a_frame_id=..., frame_level=-1) at /gdb-up/gdb/frame.c:1720
64369 #19 0x0000555555fa5e47 in get_selected_frame (message=0x0) at /gdb-up/gdb/frame.c:1810
64370 #20 0x0000555555cc9c6e in get_current_arch () at /gdb-up/gdb/arch-utils.c:848
64371 #21 0x000055555625b239 in gdbpy_before_prompt_hook (extlang=0x555557451f20 <extension_language_python>, current_gdb_prompt=0x555557f4d890 <top_prompt+16> "(gdb) ")
64372 at /gdb-up/gdb/python/python.c:1063
64373 #22 0x0000555555f7cfbb in ext_lang_before_prompt (current_gdb_prompt=0x555557f4d890 <top_prompt+16> "(gdb) ") at /gdb-up/gdb/extension.c:922
64374 #23 0x0000555555f7d442 in std::_Function_handler<void (char const*), void (*)(char const*)>::_M_invoke(std::_Any_data const&, char const*&&) (__functor=...,
64375 __args#0=@0x7fffffffd900: 0x555557f4d890 <top_prompt+16> "(gdb) ") at /usr/include/c++/7/bits/std_function.h:316
64376 #24 0x0000555555f752dd in std::function<void (char const*)>::operator()(char const*) const (this=0x55555817d838, __args#0=0x555557f4d890 <top_prompt+16> "(gdb) ")
64377 at /usr/include/c++/7/bits/std_function.h:706
64378 #25 0x0000555555f75100 in gdb::observers::observable<char const*>::notify (this=0x555557f49060 <gdb::observers::before_prompt>, args#0=0x555557f4d890 <top_prompt+16> "(gdb) ")
64379 at /gdb-up/gdb/../gdbsupport/observable.h:150
64380 #26 0x0000555555f736dc in top_level_prompt () at /gdb-up/gdb/event-top.c:444
64381 #27 0x0000555555f735ba in display_gdb_prompt (new_prompt=0x0) at /gdb-up/gdb/event-top.c:411
64382 #28 0x00005555564611a7 in tui_on_command_error () at /gdb-up/gdb/tui/tui-interp.c:205
64383 #29 0x0000555555c2173f in std::_Function_handler<void (), void (*)()>::_M_invoke(std::_Any_data const&) (__functor=...) at /usr/include/c++/7/bits/std_function.h:316
64384 #30 0x0000555555e10c20 in std::function<void ()>::operator()() const (this=0x5555580f9028) at /usr/include/c++/7/bits/std_function.h:706
64385 #31 0x0000555555e10973 in gdb::observers::observable<>::notify() const (this=0x555557f48d20 <gdb::observers::command_error>) at /gdb-up/gdb/../gdbsupport/observable.h:150
64386 #32 0x00005555560e9b3f in start_event_loop () at /gdb-up/gdb/main.c:438
64387 #33 0x00005555560e9bcc in captured_command_loop () at /gdb-up/gdb/main.c:481
64388 #34 0x00005555560eb616 in captured_main (data=0x7fffffffddd0) at /gdb-up/gdb/main.c:1348
64389 #35 0x00005555560eb67c in gdb_main (args=0x7fffffffddd0) at /gdb-up/gdb/main.c:1363
64390 #36 0x0000555555c1b6b3 in main (argc=12, argv=0x7fffffffded8) at /gdb-up/gdb/gdb.c:32
64392 Commit 1da5d0e664 eliminated the call to 'get_current_arch'
64393 in 'gdbpy_before_prompt_hook'. Hence, the second instance of
64394 "Couldn't get registers: No such process." does not appear anymore.
64396 Fix the failure by updating the regular expression in the test.
64398 2022-02-03 Alan Modra <amodra@gmail.com>
64400 PowerPC64 treatment of absolute symbols
64401 Supporting -static-pie on PowerPC64 requires the linker to properly
64402 treat SHN_ABS symbols for cases like glibc's _nl_current_LC_CTYPE_used
64403 absolute symbol. I've been slow to fix the linker on powerpc because
64404 there is some chance that this will break some shared libraries or
64408 * elf64-ppc.c (ppc64_elf_check_relocs): Consolidate local sym
64409 handling code. Don't count dyn relocs against non-dynamic
64411 (dec_dynrel_count): Adjust to suit.
64412 (ppc64_elf_edit_toc): Don't remove entries for absolute symbols
64414 (allocate_got): Don't allocate space for got relocs against
64415 non-dynamic absolute syms.
64416 (ppc64_elf_layout_multitoc): Likewise.
64417 (got_and_plt_relr): Likewise.
64418 (ppc64_elf_size_dynamic_sections): Likewise for local got.
64419 (got_and_plt_relr_for_local_syms): Likewise.
64420 (ppc64_elf_size_stubs): Don't allocate space for relr either.
64421 (ppc64_elf_relocate_section): Don't write relocs against non-dynamic
64422 absolute symbols. Don't optimise got and toc code sequences
64423 loading absolute symbol entries.
64425 * testsuite/ld-powerpc/abs-reloc.s,
64426 * testsuite/ld-powerpc/abs-static.d,
64427 * testsuite/ld-powerpc/abs-static.r,
64428 * testsuite/ld-powerpc/abs-pie.d,
64429 * testsuite/ld-powerpc/abs-pie.r,
64430 * testsuite/ld-powerpc/abs-shared.d,
64431 * testsuite/ld-powerpc/abs-shared.r,
64432 * testsuite/ld-powerpc/abs-pie-relr.d,
64433 * testsuite/ld-powerpc/abs-pie-relr.r,
64434 * testsuite/ld-powerpc/abs-shared-relr.d,
64435 * testsuite/ld-powerpc/abs-shared-relr.r: New tests.
64436 * testsuite/ld-powerpc/powerpc.exp: Run them.
64438 2022-02-03 GDB Administrator <gdbadmin@sourceware.org>
64440 Automatic date update in version.in
64442 2022-02-02 Nick Clifton <nickc@redhat.com>
64444 Stop the BFD library complaining about compressed dwarf debug string sections being too big.
64446 * dwarf2.c (read_section): Change the heuristic that checks for
64447 overlarge dwarf debug info sections.
64449 2022-02-02 Andrew Burgess <aburgess@redhat.com>
64451 gdb: fix formatting for help set/show extended-prompt
64452 The formatting of the help text for 'help set extended-prompt' and
64453 'help show extended-prompt' is a little off.
64455 Here's the offending snippet:
64457 Substitutions are applied to VALUE to compute the real prompt.
64459 The currently defined substitutions are:
64460 \[ Begins a sequence of non-printing characters.
64462 \] Ends a sequence of non-printing characters.
64463 \e The ESC character.
64465 Notice that the line for '\[' is indented more that the others.
64467 Turns out this is due to how we build this help text, something which
64468 is done in Python. We extended a classes __doc__ string with some
64469 dynamically generated text.
64471 The classes doc string looks like this:
64473 """Set the extended prompt.
64475 Usage: set extended-prompt VALUE
64477 Substitutions are applied to VALUE to compute the real prompt.
64479 The currently defined substitutions are:
64482 Notice the closing """ are in a line of their own, and include some
64483 white space just before. It's this extra white space that's causing
64486 Fix the formatting issue by moving the """ to the end of the previous
64487 line. I then add the extra newline in at the point where the doc
64488 string is merged with the dynamically generated text.
64490 Now everything lines up correctly.
64492 2022-02-02 Andrew Burgess <aburgess@redhat.com>
64494 gdb: test to check one aspect of the linespec parsing code
64495 While working on the fix for PR cli/28665 (see previous couple of
64496 commits), I was playing with making a change in the linespec parsing
64497 code. Specifically, I was thinking about whether the spec_string for
64498 LINESPEC_LOCATION locations should ever be nullptr.
64500 I made a change to prevent the spec_string from ever being nullptr,
64501 tested gdb, and saw no regressions.
64503 However, as part of this work I was reviewing how the breakpoint code
64504 handles this case (spec_string being nullptr), and spotted that in
64505 parse_breakpoint_sals the nullptr case is specifically handled, so
64506 changing this should have caused a regression. But I didn't see one.
64508 So, this commit adds a comment in location.c mentioning that the
64509 nullptr case is (a) not an oversight, and (b) is required. Then I add
64510 a new test to gdb.base/break.exp that ensures a change in this area
64511 will cause a regression.
64513 This test passes on current gdb, but with my modified (and broken)
64514 gdb, the test would fail.
64516 2022-02-02 Andrew Burgess <aburgess@redhat.com>
64518 gdb: handle calls to edit command passing only a linespec condition
64519 While working on the previous commit to fix PR cli/28665, I noticed
64520 that the 'edit' command would suffer from the same problem. That is,
64523 (gdb) edit task 123
64525 would cause GDB to break. For a full explanation of what's going on
64526 here, see the commit message for the previous commit.
64528 As with the previous commit, this issue can be prevented by detecting,
64529 and throwing, a junk at the end of the line error earlier, before
64530 calling decode_line_1.
64532 So, that's what this commit does. I've also added some tests for this
64535 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28665
64537 2022-02-02 Andrew Burgess <aburgess@redhat.com>
64539 gdb: handle calls to list command passing only a linespec condition
64540 In PR cli/28665, it was reported that GDB would crash when given a
64543 (gdb) list task 123
64545 The problem here is that in cli/cli-cmd.c:list_command, the string
64546 'task 123' is passed to string_to_event_location in find a location
64547 specification. However, this location parsing understands about
64548 breakpoint conditions, and so, will stop parsing when it sees
64549 something that looks like a condition, in this case, the 'task 123'
64550 looks like a breakpoint condition.
64552 As a result, the location we get back from string_to_event_location
64553 has no actual location specification attached to it. The actual call
64557 string_to_event_location
64558 string_to_event_location_basic
64559 new_linespec_location
64561 In new_linespec_location we call linespec_lex_to_end, which looks at
64562 'task 123' and decides that there's nothing there that describes a
64563 location. As such, in new_linespec_location, the spec_string field of
64564 the location is left as nullptr.
64566 Back in list_command we then call decode_line_1, which calls
64567 event_location_to_sals, which calls parse_linespec, which takes the
64568 spec_string we found earlier, and tries to converts this into a list
64571 However, parse_linespec is not intended to be passed a nullptr, for
64572 example, calling is_ada_operator will try to access through the
64573 nullptr, causing undefined behaviour. But there are other cases
64574 within parse_linespec which don't expect to see a nullptr.
64576 When looking at how to fix this issue, I first considered having
64577 linespec_lex_to_end detect the problem. That function understands
64578 when the first thing in the linespec is a condition keyword, and so,
64579 could throw an error saying something like: "no linespec before
64580 condition keyword", however, this is not going to work, at least, not
64581 without additional changes to GDB, it is valid to place a breakpoint
64584 (gdb) break task 123
64586 This will place a breakpoint at the current location with the
64587 condition 'task 123', and changing linespec_lex_to_end breaks this
64590 So, next, I considered what would happen if I added a condition to an
64591 otherwise valid list command, this is what I see:
64593 (gdb) list file.c:1 task 123
64594 Junk at end of line specification.
64597 So, then I wondered, could we just pull the "Junk" detection forward,
64598 so that we throw the error earlier, before we call decode_line_1?
64600 It turns out that yes we can. Well, sort of.
64602 It is simpler, I think, to add a separate check into the list_command
64603 function, after calling string_to_event_location, but before calling
64604 decode_line_1. We know when we call string_to_event_location that the
64605 string in question is not empty, so, after calling
64606 string_to_event_location, if non of the string has been consumed, then
64607 the content of the string must be junk - it clearly doesn't look like
64608 a location specification.
64610 I've reused the same "Junk at end of line specification." error for
64611 consistency, and added a few tests to cover this issue.
64613 While the first version of this patch was on the mailing list, a
64614 second bug PR gdb/28797 was raised. This was for a very similar
64615 issue, but this time the problem command was:
64619 Here the list command understands about the first comma, list can have
64620 two arguments separated by a comma, and the first argument can be
64621 missing. So we end up trying to parse the second command "," as a
64624 However, in linespec_lex_to_end, we will stop parsing a linespec at a
64625 comma, so, in the above case we end up with an empty linespec (between
64626 the two commas), and, like above, this results in the spec_string
64629 As with the previous case, I've resolved this issue by adding an extra
64630 check for junk at the end of the line - after parsing (or failing to
64631 parse) the nothing between the two commas, we still have the "," left
64632 at the end of the list command line - when we see this we can throw
64633 the same "junk at the end of the line" error, and all is good.
64635 I've added tests for this case too.
64637 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28665
64638 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28797
64640 2022-02-02 Andrew Burgess <aburgess@redhat.com>
64642 gdb/testsuite: move linespec test into gdb.linespec/ directory
64643 The gdb.base/linespecs.exp test should really live in the gdb.linespec
64644 directory, so lets move it there.
64646 As we already have gdb.linespec/linespec.exp, I've renamed the test to
64647 gdb.linespec/errors.exp, as this better reflects what the test is
64650 Finally, the test script doesn't have its own source file, it was
64651 reusing a random other source file, gdb.base/memattr.c. Now the tests
64652 script is in gdb.linespec/, I've updated the test to use a different
64653 source file from that directory.
64655 2022-02-02 Andrew Burgess <aburgess@redhat.com>
64657 gdb: add empty string check in parse_linespec
64658 If parse_linespec (linespec.c) is passed ARG as an empty string then
64659 we end up calling `strchr (linespec_quote_characters, '\0')`, which
64660 will return a pointer to the '\0' at the end of
64661 linespec_quote_characters. This then results in GDB calling
64662 skip_quote_char with `ARG + 1`, which is undefined behaviour (as ARG
64663 only contained a single character, the '\0').
64665 Fix this by checking for the first character of ARG being '\0' before
64666 the call to strchr.
64668 I have additionally added an assertion that ARG can't itself be
64669 nullptr, as calling is_ada_operator with nullptr can end up calling
64670 'startswith' on the nullptr, which is undefined behaviour.
64672 Finally, I moved the declaration of TOKEN into the body of
64673 parse_linespec, to where TOKEN is defined.
64675 This patch came about while I was working on fixes for PR cli/28665
64676 and PR gdb/28797. The actual fixes for these two issues will be in a
64677 later commit in this series, but, with this patch in place, both of
64678 the above bugs would hit the new assertion rather than accessing
64679 invalid memory and crashing. The '\0' check is not currently ever
64680 hit, but just makes the code a little safer.
64682 Because this patch only changes the nature of the failure for the
64683 above two bugs, there's no tests here. A later commit will fix the
64684 above two issues, at which point I'll add some tests.
64686 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28665
64687 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28797
64689 2022-02-02 Andrew Burgess <aburgess@redhat.com>
64691 gdb: update the comment on string_to_event_location
64692 The comment on string_to_event_location is (I believe) out of date.
64693 This commit fixes the two issues I see:
64695 1. This function can't return NULL any more. The implementation
64696 calls string_to_explicit_location which can return NULL, but if this
64697 is the case we then call string_to_event_location_basic, which I
64698 don't believe can ever return NULL.
64700 2. I've removed the mention that the returned string is malloc'd,
64701 though this is true, now that we return a managed pointer, I believe
64702 the source of the memory allocation is irrelevant, and so, shouldn't
64703 be discussed in the header comment.
64705 There should be no user visible changes after this commit.
64707 2022-02-02 Nick Clifton <nickc@redhat.com>
64709 Updated French translation for the ld/ and gold/ sub-directories
64711 2022-02-02 Stafford Horne <shorne@gmail.com>
64713 or1k: Avoid R_OR1K_GOT16 signed overflow by using special howto
64714 Previously when fixing PR 21464 we masked out upper bits of the
64715 relocation value in order to avoid overflow complaints when acceptable.
64716 It turns out this does not work when the relocation value ends up being
64719 To fix this this patch introduces a special howto with
64720 complain_on_overflow set to complain_overflow_dont. This is used in
64721 place of the normal R_OR1K_GOT16 howto when we detect R_OR1K_GOT_AHI16
64727 * elf32-or1k.c (or1k_elf_got16_no_overflow_howto): Define.
64728 (or1k_elf_relocate_section): Use new howto instead of trying to
64729 mask out relocation bits.
64731 2022-02-02 GDB Administrator <gdbadmin@sourceware.org>
64733 Automatic date update in version.in
64735 2022-02-01 Tom Tromey <tromey@adacore.com>
64737 Fix flex rule in gdb
64738 Currently, if flex fails, it will leave the resulting .c file in the
64739 tree. This will cause a cascade of errors, and requires the manual
64740 deletion of the .c file in order to recreate the problem.
64742 It's better for the rule to fail such that the .c file is not updated.
64743 This way, 'make' will fail the same way every time -- which is much
64744 handier for debugging syntax errors.
64746 This fix just updates the Makefile rule to follow the way that the
64747 "yacc" rule already works.
64749 2022-02-01 Markus Metzger <markus.t.metzger@intel.com>
64751 gdb, btrace: improve error messages
64752 When trying to use 'record btrace' on a system that does not support it,
64753 the error message isn't as clear as it could be. See
64754 https://sourceware.org/pipermail/gdb/2022-January/049870.html.
64756 Improve the error message in a few cases.
64758 Reported-by: Simon Sobisch <simonsobisch@gnu.org>
64760 2022-02-01 Jan Vrany <jan.vrany@labware.com>
64762 gdb/python: fix gdb.Objfile.__repr__ () for dynamically compiled code
64763 While experimenting with JIT reader API I realized that calling repr ()
64764 on objfile created by JIT reader crashes GDB.
64766 The problem was that objfpy_repr () called objfile_filename () which
64767 returned NULL, causing PyString_FromFormat () to crash.
64769 This commit fixes this problem by using objfile_name () instead of
64770 objfile_filename (). This also makes consistent with the value of gdb.Objfile.filename variable.
64772 2022-02-01 Samuel Thibault <samuel.thibault@gnu.org>
64774 hurd: Fix RPC prototypes
64775 The last updates of MIG introduced qualifying strings and arrays with
64776 const as appropriate. We thus have to update the protypes in gdb too.
64778 Change-Id: I3f72aac1dfa6e58d1394d5776b822d7c8f2409df
64780 2022-02-01 Samuel Thibault <samuel.thibault@gnu.org>
64782 hurd: Fix RPC link names
64783 The RPC stub code expects to be calling a C function, not a C++
64786 Change-Id: Idd7549fc118f2addc7fb4975667a011cacacc03f
64788 2022-02-01 GDB Administrator <gdbadmin@sourceware.org>
64790 Automatic date update in version.in
64792 2022-01-31 H.J. Lu <hjl.tools@gmail.com>
64794 elf: Check symbol version without any symbols
64795 VER_FLG_WEAK doesn't indicate that all symbol references of the symbol
64796 version have STB_WEAK. VER_FLG_WEAK indicates a weak symbol version
64797 definition with no symbols associated with it. It is used to verify
64798 the existence of a particular implementation without any symbol references
64799 to the weak symbol version.
64802 * testsuite/ld-elf/pr24718-1.d: New file.
64803 * testsuite/ld-elf/pr24718-1.s: Likewise.
64804 * testsuite/ld-elf/pr24718-1.t: Likewise.
64806 2022-01-31 H.J. Lu <hjl.tools@gmail.com>
64808 Load debug section only when dumping debug sections
64809 Don't load debug sections if we aren't dumping any debug sections.
64812 * objdump.c (dump_any_debugging): New.
64813 (load_debug_section): Return false if dump_any_debugging isn't
64815 (main): Set dump_any_debugging when dumping any debug sections.
64816 * readelf (dump_any_debugging): New.
64817 (parse_args): Set dump_any_debugging when dumping any debug
64819 (load_debug_section): Return false if dump_any_debugging isn't
64822 2022-01-31 Simon Marchi <simon.marchi@efficios.com>
64824 gdb: fix some clang-tidy readability-misleading-indentation warnings
64825 I have warnings like these showing in my editor all the time, so I
64826 thought I'd run clang-tidy with this diagnostic on all the files (that I
64827 can compile) and fix them.
64829 There is still one warning, in utils.c, but that's because some code
64830 is mixed up with preprocessor macros (#ifdef TUI), so I think there no
64831 good solution there.
64833 Change-Id: I345175fc7dd865318f0fbe61ac026c88c3b6a96b
64835 2022-01-31 Nils-Christian Kempke <nils-christian.kempke@intel.com>
64837 gdb, testsuite, fortran: adapt info symbol expected output for intel compilers
64838 Info symbol is expected to print the symbol table name of a symbol, since
64839 symbol lookup happens via the minimal symbol table. This name
64840 corresponds to the linkage name in the full symbol table.
64842 For gfortran (and maybe others) these names currently have the form
64843 XXXX.NUMBER where XXXX is the symbol name and NUMBER a compiler
64844 generated appendix for mangling.
64845 An example taken from the modified nested-funcs-2.exp would be
64848 $ objdump -t ./outputs/gdb.fortran/nested-funcs-2/nested-funcs-2 | grep \
64850 00000000000014ab l F .text 0000000000000095 increment.3883
64851 000000000000141c l F .text 000000000000008f increment_program_global.3881
64854 This mangled name gets recognized by the Ada demangler/decoder and decoded as
64855 Ada to XXXX (setting the symbol language to Ada). This leads to output
64856 of XXXX over XXXX.NUMBER for info symbol on gfortran symbols.
64858 For ifort and ifx the generated linkage names have the form
64859 SCOPEA_SCOPEB_XXXX_ which are not recognized by the Ada decoder (or any
64860 other demangler for that matter) and thus printed as is.
64861 The respective objdump in the above case looks like
64864 $ objdump -t ./outputs/gdb.fortran/nested-funcs-2/nested-funcs-2 | grep \
64866 0000000000403a44 l F .text 0000000000000074 contains_keyword_IP_increment_
64867 0000000000403ab8 l F .text 0000000000000070
64868 contains_keyword_IP_increment_program_global_
64871 In the unmodified testcase this results in 'fails' when ran with the intel
64875 >> make check RUNTESTFLAGS="gdb.fortran/nested-funcs-2.exp \
64876 GDBFLAGS='$GDBFLAGS' CC_FOR_TARGET='icpc' F90_FOR_TARGET='ifort'"
64880 === gdb Summary ===
64882 \# of expected passes 80
64883 \# of unexpected failures 14
64886 Note that there is no Fortran mangling standard. We keep the gfortran
64887 behavior as is and modify the test to reflect ifx and ifort mangled
64888 names which fixes above fails.
64890 2022-01-31 Nick Clifton <nickc@redhat.com>
64892 Import patch from mainline GCC to fix an infinite recusion in the Rust demangler.
64895 * rust-demangle.c (struct rust_demangler): Add a recursion
64897 (demangle_path): Increment/decrement the recursion counter upon
64898 entry and exit. Fail if the counter exceeds a fixed limit.
64899 (demangle_type): Likewise.
64900 (rust_demangle_callback): Initialise the recursion counter,
64901 disabling if requested by the option flags.
64903 2022-01-31 Alan Modra <amodra@gmail.com>
64905 Re: PR28827, assertion building LLVM 9 on powerpc64le-linux-gnu
64906 In trying to find a testcase for PR28827, I managed to hit a linker
64907 error in bfd_set_section_contents with a .branch_lt input section
64908 being too large for the output .branch_lt.
64912 * elf64-ppc.c (ppc64_elf_size_stubs): Set section size to
64913 maxsize past STUB_SHRINK_ITER before laying out. Remove now
64914 unnecessary conditional setting of maxsize at start of loop.
64916 * testsuite/ld-powerpc/pr28827-2.d,
64917 * testsuite/ld-powerpc/pr28827-2.lnk,
64918 * testsuite/ld-powerpc/pr28827-2.s: New test.
64919 * testsuite/ld-powerpc/powerpc.exp: Run it.
64921 2022-01-31 Tom Tromey <tom@tromey.com>
64923 Remove unused variables in fbsd-tdep.c files
64924 i386-fbsd-tdep.c and amd64-fbsd-tdep.c failed to build on my x86-64
64925 Fedora 34 machine, using the system gcc, after a recent patch. These
64926 two files now have unused variables, which provokes a warning in this
64929 I'm checking in this patch to remove the unused variables.
64931 2022-01-31 GDB Administrator <gdbadmin@sourceware.org>
64933 Automatic date update in version.in
64935 2022-01-30 GDB Administrator <gdbadmin@sourceware.org>
64937 Automatic date update in version.in
64939 2022-01-29 Alan Modra <amodra@gmail.com>
64941 Re: PR28827, assertion building LLVM 9 on powerpc64le-linux-gnu
64942 The previous patch wasn't quite correct. The size and padding depends
64943 on offset used in the current iteration, and if we're fudging the
64944 offset past STUB_SHRINK_ITER then we'd better use that offset. We
64945 can't have plt_stub_pad using stub_sec->size as the offset.
64948 * elf64-ppc.c (plt_stub_pad): Add stub_off param.
64949 (ppc_size_one_stub): Set up stub_offset to value used in this
64950 iteration before sizing the stub. Adjust plt_stub_pad calls.
64952 2022-01-29 Alan Modra <amodra@gmail.com>
64954 objcopy --only-keep-debug
64955 From: Peilin Ye <peilin.ye@bytedance.com>
64956 objcopy's --only-keep-debug option has been broken for ELF files since
64957 commit 8c803a2dd7d3.
64959 1. binutils/objcopy.c:setup_section() marks non-debug sections as
64960 SHT_NOBITS, then calls bfd_copy_private_section_data();
64961 2. If ISEC and OSEC share the same section flags,
64962 bfd/elf.c:_bfd_elf_init_private_section_data() restores OSEC's
64963 section type back to ISEC's section type, effectively undoing
64966 * objcopy.c (setup_section): Act on make_nobits after calling
64967 bfd_copy_private_section_data.
64969 2022-01-29 GDB Administrator <gdbadmin@sourceware.org>
64971 Automatic date update in version.in
64973 2022-01-28 John Baldwin <jhb@FreeBSD.org>
64975 gdb: fix ppc-sysv-tdep.c build on 32-bit platforms
64976 The previous code triggered the following error on an i386 host:
64978 /git/gdb/gdb/ppc-sysv-tdep.c:1764:34: error: non-constant-expression cannot be narrowed from type 'ULONGEST' (aka 'unsigned long long') to 'size_t' (aka 'unsigned int') in initializer list [-Wc++11-narrowing]
64979 unscaled.read ({writebuf, TYPE_LENGTH (valtype)},
64980 ^~~~~~~~~~~~~~~~~~~~~
64981 /git/gdb/gdb/gdbtypes.h:2043:31: note: expanded from macro 'TYPE_LENGTH'
64983 /git/gdb/gdb/ppc-sysv-tdep.c:1764:34: note: insert an explicit cast to silence this issue
64984 unscaled.read ({writebuf, TYPE_LENGTH (valtype)},
64985 ^~~~~~~~~~~~~~~~~~~~~
64986 static_cast<size_t>( )
64987 /git/gdb/gdb/gdbtypes.h:2043:31: note: expanded from macro 'TYPE_LENGTH'
64991 Fix this by using gdb::make_array_view.
64993 2022-01-28 John Baldwin <jhb@FreeBSD.org>
64995 FreeBSD x86 nat: Use register maps for GP register sets.
64996 Rather than using the x86-specific register offset tables, use
64997 register maps to describe the layout of the general purpose registers
64998 fetched via PT_GETREGS. The sole user-visible difference is that
64999 FreeBSD/amd64 will now report additional segment registers ($ds, $es,
65000 $fs, and $gs) for both 32-bit and 64-bit processes.
65002 As part of these changes, the FreeBSD x86 native targets no longer use
65003 amd64-bsd-nat.c or i386-bsd-nat.c. Remove FreeBSD-specific register
65004 handling (for $fs_base, $gs_base, and XSAVE state) from these files.
65005 Similarly, remove the global x86bsd_xsave_len from x86-bsd-nat.c. The
65006 FreeBSD x86 native targets use a static xsave_len instead.
65008 While here, rework the probing of PT_GETXMMREGS on FreeBSD/i386.
65009 Probe the ptrace op once in the target read_description method and
65010 cache the result for the future similar to the way the status of XSAVE
65011 support is probed in the read_description method. In addition, return
65012 the proper xcr0 mask (X87-only) for old kernels or systems without
65013 either XSAVE or XMM support.
65015 2022-01-28 John Baldwin <jhb@FreeBSD.org>
65017 fbsd-nat: Return a bool from fetch_register_set and store_register_set.
65018 Change these helper functions to return true if they did any work.
65020 2022-01-28 John Baldwin <jhb@FreeBSD.org>
65022 FreeBSD x86: Use tramp-frame for signal frames.
65023 Use a register map to describe the registers in mcontext_t as part of
65024 the signal frame as is done on several other FreeBSD arches. This
65025 permits fetching the fsbase and gsbase register values from the signal
65026 frame for both amd64 and i386 and permits fetching additional segment
65027 registers stored as 16-bit values on amd64.
65029 While signal frames on FreeBSD do contain floating point/XSAVE state,
65030 these unwinders do not attempt to supply those registers. The
65031 existing x86 signal frame uwinders do not support these registers, and
65032 the only existing functions which handle FSAVE/FXSAVE/XSAVE state all
65033 work with regcaches. In the future these unwinders could create a
65034 tempory regcache, collect floating point registers, and then supply
65035 values out of the regcache into the trad-frame.
65037 2022-01-28 John Baldwin <jhb@FreeBSD.org>
65039 Use register maps for gp regsets on FreeBSD/x86 core dumps.
65040 In particular, this permits reporting the value of the $ds, $es, $fs,
65041 and $gs segment registers from amd64 core dumps since they are stored
65042 as 16-bit values rather than the 32-bit size assumed by i386_gregset.
65044 2022-01-28 John Baldwin <jhb@FreeBSD.org>
65046 regcache: Zero-extend small registers described by a register map.
65047 When registers are supplied via regcache_supply_register from a
65048 register block described by a register map, registers may be stored in
65049 slots smaller than GDB's native register size (e.g. x86 segment
65050 registers are 16 bits, but the GDB registers for those are 32-bits).
65051 regcache_collect_regset is careful to zero-extend slots larger than a
65052 register size, but regcache_supply_regset just used
65053 regcache::raw_supply_part and did not initialize the upper bytes of a
65056 trad_frame_set_reg_regmap assumes these semantics (zero-extending
65057 short registers). Upcoming patches also require these semantics for
65058 handling x86 segment register values stored in 16-bit slots on
65059 FreeBSD. Note that architecturally x86 segment registers are 16 bits,
65060 but the x86 gdb architectures treat these registers as 32 bits.
65062 2022-01-28 John Baldwin <jhb@FreeBSD.org>
65064 FreeBSD x86: Remove fallback for detecting signal trampolines by address.
65065 A few FreeBSD releases did not include the page holding the signal
65066 code in core dumps. As a workaround, a sysctl was used to fetch the
65067 default location of the signal code instead. The youngest affected
65068 FreeBSD release is 10.1 released in November 2014 and EOLed in
65069 December 2016. The fallback only works for native processes and would
65070 require a separate unwinder once the FreeBSD arches are converted to
65071 use tramp_frame for signal frames.
65073 Remove support for pre-5.0 FreeBSD/i386 signal trampolines.
65074 The last relevant release (FreeBSD 4.11) was released in January of
65077 Remove vestigal FreeBSD/i386 3.x support.
65078 This was orphaned when a.out support was removed as the FreeBSD/i386
65079 ELF support always used the register layouts from 4.0+.
65081 2022-01-28 Bruno Larsen <blarsen@redhat.com>
65083 Add Bruno Larsen to gdb/MAINTAINERS
65085 2022-01-28 Enze Li <lienze2010@hotmail.com>
65087 gdb/build: Fix Wpessimizing-move in clang build
65088 When building with clang, I run into an error:
65091 tui/tui-disasm.c:138:25: error: moving a temporary object prevents copy
65092 elision [-Werror,-Wpessimizing-move]
65093 tal.addr_string = std::move (gdb_dis_out.release ());
65095 tui/tui-disasm.c:138:25: note: remove std::move call here
65096 tal.addr_string = std::move (gdb_dis_out.release ());
65100 The error above is caused by the recent commit 5d10a2041eb8 ("gdb: add
65101 string_file::release method").
65103 Fix this by removing std::move.
65105 Build on x86_64-linux with clang 13.0.0.
65107 2022-01-28 Simon Marchi <simon.marchi@polymtl.ca>
65109 Add top-level .editorconfig file
65110 Add a .editorconfig [1] file. This helps configure editors
65111 automatically with the right whitespace settings. It will help me,
65112 since I need to juggle with different whitespace settings for different
65113 projects. But I think it can also help newcomers get things right from
65116 Some editors have native support for reading these files, while others
65117 require a plug-in [2]. And if you don't want to use it, then this file
65118 doesn't change anything to your life.
65120 I added rules for the kinds of files I edit most often, but more can be
65121 added later. I assumed that the rules were the same for GDB and the
65122 other projects, but if that's not the case, we can always put
65123 .editorconfig files in project subdirectories to override settings.
65125 [1] https://editorconfig.org/
65126 [2] https://editorconfig.org/#download
65128 Change-Id: Ifda136d13877fafcf0d137fec8501f6a34e1367b
65130 2022-01-28 Nick Clifton <nickc@redhat.com>
65132 Updated French translation for the gas sub-directory.
65134 2022-01-28 Alan Modra <amodra@gmail.com>
65136 Set __ehdr_start rel_from_abs earlier
65137 This is just a tidy, making the __ehdr_start symbol flag tweaks all in
65140 * ldelf.c (ldelf_before_allocation): Don't set rel_from_abs
65142 * ldlang.c (lang_symbol_tweaks): Set it here instead.
65144 2022-01-28 Alan Modra <amodra@gmail.com>
65146 PowerPC64 handling of @tocbase
65147 * elf64-ppc.c (ppc64_elf_relocate_section): Warn if the symbol
65148 on R_PPC64_TOC isn't local.
65150 2022-01-28 Alan Modra <amodra@gmail.com>
65152 Update PowerPC64 symtocbase test
65153 Using a symbol other than .TOC. with @tocbase is an extension to the
65154 ABI. It is never valid to use a symbol without a definition in the
65155 binary, and symbols on these expressions cannot be overridden. Make
65156 this explicit by using ".hidden" in the testcase.
65158 * testsuite/ld-powerpc/symtocbase-1.s: Align data. Make function
65159 entry symbol hidden.
65160 * testsuite/ld-powerpc/symtocbase-2.s: Likewise.
65161 * testsuite/ld-powerpc/symtocbase.d: Adjust expected output.
65163 2022-01-28 Alan Modra <amodra@gmail.com>
65165 PR28827, assertion building LLVM 9 on powerpc64le-linux-gnu
65166 The assertion is this one in ppc_build_one_stub
65167 BFD_ASSERT (stub_entry->stub_offset >= stub_entry->group->stub_sec->size);
65168 It is checking that a stub doesn't overwrite the tail of a previous
65169 stub, so not something trivial.
65171 Normally, stub sizing iterates until no stubs are added, detected by
65172 no change in stub section size. Iteration also continues if no stubs
65173 are added but one or more stubs increases in size, which also can be
65174 detected by a change in stub section size. But there is a
65175 pathological case where stub section sizing decreases one iteration
65176 then increases the next. To handle that situation, stub sizing also
65177 stops at more than STUB_SHRINK_ITER (20) iterations when calculated
65178 stub section size is smaller. The previous larger size is kept for
65179 the actual layout (so that building the stubs, which behaves like
65180 another iteration of stub sizing, will see the stub section sizes
65181 shrink). The problem with that stopping condition is that it assumes
65182 that stub sizing is only affected by addresses external to the stub
65183 sections, which isn't always true.
65185 This patch fixes that by also keeping larger individual stub_offset
65186 addresses past STUB_SHRINK_ITER. It also catches a further
65187 pathological case where one stub shrinks and another expands in such a
65188 way that no stub section size change is seen.
65191 * elf64-ppc.c (struct ppc_link_hash_table): Add stub_changed.
65192 (STUB_SHRINK_ITER): Move earlier in file.
65193 (ppc_size_one_stub): Detect any change in stub_offset. Keep
65194 larger one if past STUB_SHRINK_ITER.
65195 (ppc64_elf_size_stubs): Iterate on stub_changed too.
65197 2022-01-28 Alan Modra <amodra@gmail.com>
65199 PR28826 x86_64 ld segfaults building xen
65200 Fallout from commit e86fc4a5bc37
65203 * coffgen.c (coff_write_alien_symbol): Init dummy to zeros.
65205 2022-01-28 Alan Modra <amodra@gmail.com>
65207 PR28753, buffer overflow in read_section_stabs_debugging_info
65209 * rddbg.c (read_section_stabs_debugging_info): Don't read past
65210 end of section when concatentating stab strings.
65212 2022-01-28 GDB Administrator <gdbadmin@sourceware.org>
65214 Automatic date update in version.in
65216 2022-01-27 Simon Marchi <simon.marchi@polymtl.ca>
65218 gdb: work around negative DW_AT_data_member_location GCC 11 bug
65219 g++ 11.1.0 has a bug where it will emit a negative
65220 DW_AT_data_member_location in some cases:
65228 std::unique_ptr<int> ptr;
65231 $ llvm-dwarfdump -F a.out
65233 0x00000964: DW_TAG_member
65234 DW_AT_name [DW_FORM_strp] ("_M_head_impl")
65235 DW_AT_decl_file [DW_FORM_data1] ("/usr/include/c++/11.1.0/tuple")
65236 DW_AT_decl_line [DW_FORM_data1] (125)
65237 DW_AT_decl_column [DW_FORM_data1] (0x27)
65238 DW_AT_type [DW_FORM_ref4] (0x0000067a "default_delete<int>")
65239 DW_AT_data_member_location [DW_FORM_sdata] (-1)
65242 This leads to a GDB crash (when built with ASan, otherwise probably
65243 garbage results), since it tries to read just before (to the left, in
65244 ASan speak) of the value's buffer:
65246 ==888645==ERROR: AddressSanitizer: heap-buffer-overflow on address 0x6020000c52af at pc 0x7f711b239f4b bp 0x7fff356bd470 sp 0x7fff356bcc18
65247 READ of size 1 at 0x6020000c52af thread T0
65248 #0 0x7f711b239f4a in __interceptor_memcpy /build/gcc/src/gcc/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc:827
65249 #1 0x555c4977efa1 in value_contents_copy_raw /home/simark/src/binutils-gdb/gdb/value.c:1347
65250 #2 0x555c497909cd in value_primitive_field(value*, long, int, type*) /home/simark/src/binutils-gdb/gdb/value.c:3126
65251 #3 0x555c478f2eaa in cp_print_value_fields(value*, ui_file*, int, value_print_options const*, type**, int) /home/simark/src/binutils-gdb/gdb/cp-valprint.c:333
65252 #4 0x555c478f63b2 in cp_print_value /home/simark/src/binutils-gdb/gdb/cp-valprint.c:513
65253 #5 0x555c478f02ca in cp_print_value_fields(value*, ui_file*, int, value_print_options const*, type**, int) /home/simark/src/binutils-gdb/gdb/cp-valprint.c:161
65254 #6 0x555c478f63b2 in cp_print_value /home/simark/src/binutils-gdb/gdb/cp-valprint.c:513
65255 #7 0x555c478f02ca in cp_print_value_fields(value*, ui_file*, int, value_print_options const*, type**, int) /home/simark/src/binutils-gdb/gdb/cp-valprint.c:161
65256 #8 0x555c478f63b2 in cp_print_value /home/simark/src/binutils-gdb/gdb/cp-valprint.c:513
65257 #9 0x555c478f02ca in cp_print_value_fields(value*, ui_file*, int, value_print_options const*, type**, int) /home/simark/src/binutils-gdb/gdb/cp-valprint.c:161
65258 #10 0x555c4760d45f in c_value_print_struct /home/simark/src/binutils-gdb/gdb/c-valprint.c:383
65259 #11 0x555c4760df4c in c_value_print_inner(value*, ui_file*, int, value_print_options const*) /home/simark/src/binutils-gdb/gdb/c-valprint.c:438
65260 #12 0x555c483ff9a7 in language_defn::value_print_inner(value*, ui_file*, int, value_print_options const*) const /home/simark/src/binutils-gdb/gdb/language.c:632
65261 #13 0x555c49758b68 in do_val_print /home/simark/src/binutils-gdb/gdb/valprint.c:1048
65262 #14 0x555c49759b17 in common_val_print(value*, ui_file*, int, value_print_options const*, language_defn const*) /home/simark/src/binutils-gdb/gdb/valprint.c:1151
65263 #15 0x555c478f2fcb in cp_print_value_fields(value*, ui_file*, int, value_print_options const*, type**, int) /home/simark/src/binutils-gdb/gdb/cp-valprint.c:335
65264 #16 0x555c478f63b2 in cp_print_value /home/simark/src/binutils-gdb/gdb/cp-valprint.c:513
65265 #17 0x555c478f02ca in cp_print_value_fields(value*, ui_file*, int, value_print_options const*, type**, int) /home/simark/src/binutils-gdb/gdb/cp-valprint.c:161
65266 #18 0x555c4760d45f in c_value_print_struct /home/simark/src/binutils-gdb/gdb/c-valprint.c:383
65267 #19 0x555c4760df4c in c_value_print_inner(value*, ui_file*, int, value_print_options const*) /home/simark/src/binutils-gdb/gdb/c-valprint.c:438
65268 #20 0x555c483ff9a7 in language_defn::value_print_inner(value*, ui_file*, int, value_print_options const*) const /home/simark/src/binutils-gdb/gdb/language.c:632
65269 #21 0x555c49758b68 in do_val_print /home/simark/src/binutils-gdb/gdb/valprint.c:1048
65270 #22 0x555c49759b17 in common_val_print(value*, ui_file*, int, value_print_options const*, language_defn const*) /home/simark/src/binutils-gdb/gdb/valprint.c:1151
65271 #23 0x555c478f2fcb in cp_print_value_fields(value*, ui_file*, int, value_print_options const*, type**, int) /home/simark/src/binutils-gdb/gdb/cp-valprint.c:335
65272 #24 0x555c4760d45f in c_value_print_struct /home/simark/src/binutils-gdb/gdb/c-valprint.c:383
65273 #25 0x555c4760df4c in c_value_print_inner(value*, ui_file*, int, value_print_options const*) /home/simark/src/binutils-gdb/gdb/c-valprint.c:438
65274 #26 0x555c483ff9a7 in language_defn::value_print_inner(value*, ui_file*, int, value_print_options const*) const /home/simark/src/binutils-gdb/gdb/language.c:632
65275 #27 0x555c49758b68 in do_val_print /home/simark/src/binutils-gdb/gdb/valprint.c:1048
65276 #28 0x555c49759b17 in common_val_print(value*, ui_file*, int, value_print_options const*, language_defn const*) /home/simark/src/binutils-gdb/gdb/valprint.c:1151
65277 #29 0x555c4760f04c in c_value_print(value*, ui_file*, value_print_options const*) /home/simark/src/binutils-gdb/gdb/c-valprint.c:587
65278 #30 0x555c483ff954 in language_defn::value_print(value*, ui_file*, value_print_options const*) const /home/simark/src/binutils-gdb/gdb/language.c:614
65279 #31 0x555c49759f61 in value_print(value*, ui_file*, value_print_options const*) /home/simark/src/binutils-gdb/gdb/valprint.c:1189
65280 #32 0x555c48950f70 in print_formatted /home/simark/src/binutils-gdb/gdb/printcmd.c:337
65281 #33 0x555c48958eda in print_value(value*, value_print_options const&) /home/simark/src/binutils-gdb/gdb/printcmd.c:1258
65282 #34 0x555c48959891 in print_command_1 /home/simark/src/binutils-gdb/gdb/printcmd.c:1367
65283 #35 0x555c4895a3df in print_command /home/simark/src/binutils-gdb/gdb/printcmd.c:1458
65284 #36 0x555c4767f974 in do_simple_func /home/simark/src/binutils-gdb/gdb/cli/cli-decode.c:97
65285 #37 0x555c47692e25 in cmd_func(cmd_list_element*, char const*, int) /home/simark/src/binutils-gdb/gdb/cli/cli-decode.c:2475
65286 #38 0x555c4936107e in execute_command(char const*, int) /home/simark/src/binutils-gdb/gdb/top.c:670
65287 #39 0x555c485f1bff in catch_command_errors /home/simark/src/binutils-gdb/gdb/main.c:523
65288 #40 0x555c485f249c in execute_cmdargs /home/simark/src/binutils-gdb/gdb/main.c:618
65289 #41 0x555c485f6677 in captured_main_1 /home/simark/src/binutils-gdb/gdb/main.c:1317
65290 #42 0x555c485f6c83 in captured_main /home/simark/src/binutils-gdb/gdb/main.c:1338
65291 #43 0x555c485f6d65 in gdb_main(captured_main_args*) /home/simark/src/binutils-gdb/gdb/main.c:1363
65292 #44 0x555c46e41ba8 in main /home/simark/src/binutils-gdb/gdb/gdb.c:32
65293 #45 0x7f71198bcb24 in __libc_start_main (/usr/lib/libc.so.6+0x27b24)
65294 #46 0x555c46e4197d in _start (/home/simark/build/binutils-gdb-one-target/gdb/gdb+0x77f197d)
65296 0x6020000c52af is located 1 bytes to the left of 8-byte region [0x6020000c52b0,0x6020000c52b8)
65297 allocated by thread T0 here:
65298 #0 0x7f711b2b7459 in __interceptor_calloc /build/gcc/src/gcc/libsanitizer/asan/asan_malloc_linux.cpp:154
65299 #1 0x555c470acdc9 in xcalloc /home/simark/src/binutils-gdb/gdb/alloc.c:100
65300 #2 0x555c49b775cd in xzalloc(unsigned long) /home/simark/src/binutils-gdb/gdbsupport/common-utils.cc:29
65301 #3 0x555c4977bdeb in allocate_value_contents /home/simark/src/binutils-gdb/gdb/value.c:1029
65302 #4 0x555c4977be25 in allocate_value(type*) /home/simark/src/binutils-gdb/gdb/value.c:1040
65303 #5 0x555c4979030d in value_primitive_field(value*, long, int, type*) /home/simark/src/binutils-gdb/gdb/value.c:3092
65304 #6 0x555c478f6280 in cp_print_value /home/simark/src/binutils-gdb/gdb/cp-valprint.c:501
65305 #7 0x555c478f02ca in cp_print_value_fields(value*, ui_file*, int, value_print_options const*, type**, int) /home/simark/src/binutils-gdb/gdb/cp-valprint.c:161
65306 #8 0x555c478f63b2 in cp_print_value /home/simark/src/binutils-gdb/gdb/cp-valprint.c:513
65307 #9 0x555c478f02ca in cp_print_value_fields(value*, ui_file*, int, value_print_options const*, type**, int) /home/simark/src/binutils-gdb/gdb/cp-valprint.c:161
65308 #10 0x555c478f63b2 in cp_print_value /home/simark/src/binutils-gdb/gdb/cp-valprint.c:513
65309 #11 0x555c478f02ca in cp_print_value_fields(value*, ui_file*, int, value_print_options const*, type**, int) /home/simark/src/binutils-gdb/gdb/cp-valprint.c:161
65310 #12 0x555c4760d45f in c_value_print_struct /home/simark/src/binutils-gdb/gdb/c-valprint.c:383
65311 #13 0x555c4760df4c in c_value_print_inner(value*, ui_file*, int, value_print_options const*) /home/simark/src/binutils-gdb/gdb/c-valprint.c:438
65312 #14 0x555c483ff9a7 in language_defn::value_print_inner(value*, ui_file*, int, value_print_options const*) const /home/simark/src/binutils-gdb/gdb/language.c:632
65313 #15 0x555c49758b68 in do_val_print /home/simark/src/binutils-gdb/gdb/valprint.c:1048
65314 #16 0x555c49759b17 in common_val_print(value*, ui_file*, int, value_print_options const*, language_defn const*) /home/simark/src/binutils-gdb/gdb/valprint.c:1151
65315 #17 0x555c478f2fcb in cp_print_value_fields(value*, ui_file*, int, value_print_options const*, type**, int) /home/simark/src/binutils-gdb/gdb/cp-valprint.c:335
65316 #18 0x555c478f63b2 in cp_print_value /home/simark/src/binutils-gdb/gdb/cp-valprint.c:513
65317 #19 0x555c478f02ca in cp_print_value_fields(value*, ui_file*, int, value_print_options const*, type**, int) /home/simark/src/binutils-gdb/gdb/cp-valprint.c:161
65318 #20 0x555c4760d45f in c_value_print_struct /home/simark/src/binutils-gdb/gdb/c-valprint.c:383
65319 #21 0x555c4760df4c in c_value_print_inner(value*, ui_file*, int, value_print_options const*) /home/simark/src/binutils-gdb/gdb/c-valprint.c:438
65320 #22 0x555c483ff9a7 in language_defn::value_print_inner(value*, ui_file*, int, value_print_options const*) const /home/simark/src/binutils-gdb/gdb/language.c:632
65321 #23 0x555c49758b68 in do_val_print /home/simark/src/binutils-gdb/gdb/valprint.c:1048
65322 #24 0x555c49759b17 in common_val_print(value*, ui_file*, int, value_print_options const*, language_defn const*) /home/simark/src/binutils-gdb/gdb/valprint.c:1151
65323 #25 0x555c478f2fcb in cp_print_value_fields(value*, ui_file*, int, value_print_options const*, type**, int) /home/simark/src/binutils-gdb/gdb/cp-valprint.c:335
65324 #26 0x555c4760d45f in c_value_print_struct /home/simark/src/binutils-gdb/gdb/c-valprint.c:383
65325 #27 0x555c4760df4c in c_value_print_inner(value*, ui_file*, int, value_print_options const*) /home/simark/src/binutils-gdb/gdb/c-valprint.c:438
65326 #28 0x555c483ff9a7 in language_defn::value_print_inner(value*, ui_file*, int, value_print_options const*) const /home/simark/src/binutils-gdb/gdb/language.c:632
65327 #29 0x555c49758b68 in do_val_print /home/simark/src/binutils-gdb/gdb/valprint.c:1048
65329 Since there are some binaries with this in the wild, I think it would be
65330 useful for GDB to work around this. I did the obvious simple thing, if
65331 the DW_AT_data_member_location's value is -1, replace it with 0. I
65332 added a producer check to only apply this fixup for GCC 11. The idea is
65333 that if some other compiler ever uses a DW_AT_data_member_location value
65334 of -1 by mistake, we don't know (before analyzing the bug at least) if
65335 they did mean 0 or some other value. So I wouldn't want to apply the
65336 fixup in that case.
65338 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28063
65339 Change-Id: Ieef3459b0b9bbce8bdad838ba83b4b64e7269d42
65341 2022-01-27 Kevin Buettner <kevinb@redhat.com>
65343 Fix GDB internal error by using text (instead of data) section offset
65344 Fedora Rawhide is now using gcc-12.0. As part of updating to the
65345 gcc-12.0 package set, Rawhide is also now using a version of libgcc_s
65346 which lacks a .data section. This causes gdb to fail in the following
65347 fashion while debugging a program (such as gdb) which uses libgcc_s:
65350 Starting program: rawhide-master/bld/gdb/gdb
65352 objfiles.h:467: internal-error: sect_index_data not initialized
65353 A problem internal to GDB has been detected,
65354 further debugging may prove unreliable.
65357 I snipped the backtrace from the above output. Instead, here's a
65358 portion of a backtrace obtained using GDB's backtrace command.
65359 (Obviously, in order to obtain it, I used a GDB which has been patched
65362 #0 internal_error (
65363 file=0xc6a508 "gdb/objfiles.h", line=467,
65364 fmt=0xc6a4e8 "sect_index_data not initialized")
65365 at gdbsupport/errors.cc:51
65366 #1 0x00000000005f9651 in objfile::data_section_offset (this=0x4fa48f0)
65367 at gdb/objfiles.h:467
65368 #2 0x000000000097c5f8 in relocate_address (address=0x17244, objfile=0x4fa48f0)
65369 at gdb/stap-probe.c:1333
65370 #3 0x000000000097c630 in stap_probe::get_relocated_address (this=0xa1a17a0,
65372 at gdb/stap-probe.c:1341
65373 #4 0x00000000004d7025 in create_exception_master_breakpoint_probe (
65375 at gdb/breakpoint.c:3505
65376 #5 0x00000000004d7426 in create_exception_master_breakpoint ()
65377 at gdb/breakpoint.c:3575
65378 #6 0x00000000004efcc1 in breakpoint_re_set ()
65379 at gdb/breakpoint.c:13407
65380 #7 0x0000000000956998 in solib_add (pattern=0x0, from_tty=0, readsyms=1)
65381 at gdb/solib.c:1001
65382 #8 0x00000000009576a8 in handle_solib_event ()
65383 at gdb/solib.c:1269
65386 The function 'relocate_address' in gdb/stap-probe.c attempts to do
65387 its "relocation" by using objfile->data_section_offset(). That
65388 method, data_section_offset() is defined as follows in objfiles.h:
65390 CORE_ADDR data_section_offset () const
65392 return section_offsets[SECT_OFF_DATA (this)];
65395 The internal error occurs when the SECT_OFF_DATA macro finds that the
65396 'sect_index_data' field is -1:
65398 #define SECT_OFF_DATA(objfile) \
65399 ((objfile->sect_index_data == -1) \
65400 ? (internal_error (__FILE__, __LINE__, \
65401 _("sect_index_data not initialized")), -1) \
65402 : objfile->sect_index_data)
65404 relocate_address() is obtaining the section offset in order to compute
65405 a relocated address. For some ABIs, such as the System V ABI, the
65406 section offsets will all be the same. So for those ABIs, it doesn't
65407 matter which offset is used. However, other ABIs, such as the FDPIC
65408 ABI, will have different offsets for the various sections. Thus, for
65409 those ABIs, it is vital that this and other relocation code use the
65412 In stap_probe::get_relocated_address, the address to which to add the
65413 offset (thus forming the relocated address) is obtained via
65414 this->get_address (); get_address is a getter for m_address in
65415 probe.h. It's documented/defined as follows (also in probe.h):
65417 /* The address where the probe is inserted, relative to
65419 CORE_ADDR m_address;
65421 (Thanks to Tom Tromey for this observation.)
65423 So, based on this, the current use of data_section_offset /
65424 SECT_OFF_DATA is wrong. This relocation code should have been using
65425 text_section_offset / SECT_OFF_TEXT all along. That being the
65426 case, I've adjusted the stap-probe.c relocation code accordingly.
65428 Searching the sources turned up one other use of data_section_offset,
65429 in gdb/dtrace-probe.c, so I've updated that code as well. The same
65430 reasoning presented above applies to this case too.
65434 * gdb/dtrace-probe.c (dtrace_probe::get_relocated_address):
65435 Use method text_section_offset instead of data_section_offset.
65436 * gdb/stap-probe.c (relocate_address): Likewise.
65438 2022-01-27 Markus Metzger <markus.t.metzger@intel.com>
65440 gdb, remote, btrace: move switch_to_thread call right before xfer call
65441 In remote_target::remote_btrace_maybe_reopen, we switch to the currently
65442 iterated thread in order to set inferior_ptid for a subsequent xfer.
65444 Move the switch_to_thread call directly before the target_read_stralloc
65445 call to clarify why we need to switch threads.
65447 2022-01-27 Markus Metzger <markus.t.metzger@intel.com>
65449 gdb, gdbserver: update thread identifier in enable_btrace target method
65450 The enable_btrace target method takes a ptid_t to identify the thread on
65451 which tracing shall be enabled.
65453 Change this to thread_info * to avoid translating back and forth between
65454 the two. This will be used in a subsequent patch.
65456 2022-01-27 Markus Metzger <markus.t.metzger@intel.com>
65458 gdb, btrace: switch threads in remote_btrace_maybe_reopen()
65459 In remote_btrace_maybe_reopen() we iterate over threads and use
65460 set_general_thread() to set the thread from which to transfer the btrace
65463 This sets the remote general thread but does not affect inferior_ptid. On
65464 the xfer request later on, remote_target::xfer_partial() again sets the
65465 remote general thread to inferior_ptid, overwriting what
65466 remote_btrace_maybe_reopen() had done.
65468 In one case, this led to inferior_ptid being null_ptid when we tried to
65469 enable tracing on a newly created thread inside a newly created process
65472 This, in turn, led to find_inferior_pid() asserting when we iterated over
65473 threads in record_btrace_is_replaying(), which was called from
65474 record_btrace_target::xfer_partial() when reading the btrace configuration
65475 of the new thread to check whether it was already being recorded.
65477 The bug was exposed by
65479 0618ae41497 gdb: optimize all_matching_threads_iterator
65483 FAIL: gdb.btrace/enable-new-thread.exp: ... (GDB internal error)
65485 Use switch_to_thread() in remote_btrace_maybe_reopen().
65487 2022-01-27 Markus Metzger <markus.t.metzger@intel.com>
65489 gdb, btrace: rename record_btrace_enable_warn()
65490 We use record_btrace_enable_warn() as the new-thread observer callback.
65491 It is not used in other contexts.
65493 Rename it to record_btrace_on_new_thread() to make its role more clear.
65495 2022-01-27 Nick Clifton <nickc@redhat.com>
65497 Updated Swedish translation for the binutils subdirectory
65499 2022-01-27 GDB Administrator <gdbadmin@sourceware.org>
65501 Automatic date update in version.in
65503 2022-01-26 Andrew Burgess <aburgess@redhat.com>
65505 gdb/python: handle non utf-8 characters when source highlighting
65506 This commit adds support for source files that contain non utf-8
65507 characters when performing source styling using the Python pygments
65508 package. This does not change the behaviour of GDB when the GNU
65509 Source Highlight library is used.
65511 For the following problem description, assume that either GDB is built
65512 without GNU Source Highlight support, of that this has been disabled
65513 using 'maintenance set gnu-source-highlight enabled off'.
65515 The initial problem reported was that a source file containing non
65516 utf-8 characters would cause GDB to print a Python exception, and then
65517 display the source without styling, e.g.:
65519 Python Exception <class 'UnicodeDecodeError'>: 'utf-8' codec can't decode byte 0xc0 in position 142: invalid start byte
65520 /* Source code here, without styling... */
65522 Further, as the user steps through different source files, each time
65523 the problematic source file was evicted from the source cache, and
65524 then later reloaded, the exception would be printed again.
65526 Finally, this problem is only present when using Python 3, this issue
65527 is not present for Python 2.
65529 What makes this especially frustrating is that GDB can clearly print
65530 the source file contents, they're right there... If we disable
65531 styling completely, or make use of the GNU Source Highlight library,
65532 then everything is fine. So why is there an error when we try to
65533 apply styling using Python?
65535 The problem is the use of PyString_FromString (which is an alias for
65536 PyUnicode_FromString in Python 3), this function converts a C string
65537 into a either a Unicode object (Py3) or a str object (Py2). For
65538 Python 2 there is no unicode encoding performed during this function
65539 call, but for Python 3 the input is assumed to be a uft-8 encoding
65540 string for the purpose of the conversion. And here of course, is the
65541 problem, if the source file contains non utf-8 characters, then it
65542 should not be treated as utf-8, but that's what we do, and that's why
65545 My first thought when looking at this was to spot when the
65546 PyString_FromString call failed with a UnicodeDecodeError and silently
65547 ignore the error. This would mean that GDB would print the source
65548 without styling, but would also avoid the annoying exception message.
65550 However, I also make use of `pygmentize`, a command line wrapper
65551 around the Python pygments module, which I use to apply syntax
65552 highlighting in the output of `less`. And this command line wrapper
65553 is quite happy to syntax highlight my source file that contains non
65554 utf-8 characters, so it feels like the problem should be solvable.
65556 It turns out that inside the pygments module there is already support
65557 for guessing the encoding of the incoming file content, if the
65558 incoming content is not already a Unicode string. This is what
65559 happens for Python 2 where the incoming content is of `str` type.
65561 We could try and make GDB smarter when it comes to converting C
65562 strings into Python Unicode objects; this would probably require us to
65563 just try a couple of different encoding schemes rather than just
65564 giving up after utf-8.
65566 However, I figure, why bother? The pygments module already does this
65567 for us, and the colorize API is not part of the documented external
65568 API of GDB. So, why not just change the colorize API, instead of the
65569 content being a Unicode string (for Python 3), lets just make the
65570 content be a bytes object. The pygments module can then take
65571 responsibility for guessing the encoding.
65573 So, currently, the colorize API receives a unicode object, and returns
65574 a unicode object. I propose that the colorize API receive a bytes
65575 object, and return a bytes object.
65577 2022-01-26 Tom Tromey <tom@tromey.com>
65579 Remove global wrap_here function
65580 This removes the global wrap_here function, so that future calls
65581 cannot be introduced. Instead, all callers must use the method on the
65582 appropriate ui_file.
65584 This temporarily moves the implementation of this method to utils.c.
65585 This will change once the remaining patches to untangle the pager have
65588 2022-01-26 Tom Tromey <tom@tromey.com>
65590 Always call the wrap_here method
65591 This changes all existing calls to wrap_here to call the method on the
65592 appropriate ui_file instead. The choice of ui_file is determined by
65595 2022-01-26 Tom Tromey <tom@tromey.com>
65597 Add ui_file::wrap_here
65598 Right now, wrap_here is a global function. In the long run, we'd like
65599 output streams to be relatively self-contained objects, and having a
65600 global function like this is counter to that goal. Also, existing
65601 code freely mixes writes to some parameterized stream with calls to
65602 wrap_here -- but wrap_here only really affects gdb_stdout, so this is
65605 This step is a patch toward making wrap_here more sane. It adds a
65606 wrap_here method to ui_file and changes ui_out implementations to use
65609 2022-01-26 Tom Tromey <tom@tromey.com>
65611 Convert wrap_here to use integer parameter
65612 I think it only really makes sense to call wrap_here with an argument
65613 consisting solely of spaces. Given this, it seemed better to me that
65614 the argument be an int, rather than a string. This patch is the
65615 result. Much of it was written by a script.
65617 2022-01-26 Andrew Burgess <aburgess@redhat.com>
65619 gdb/python: improve the auto help text for gdb.Parameter
65620 This commit attempts to improve the help text that is generated for
65621 gdb.Parameter objects when the user fails to provide their own
65624 Documentation for a gdb.Parameter is currently pulled from two
65625 sources: the class documentation string, and the set_doc/show_doc
65626 class attributes. Thus, a fully documented parameter might look like
65629 class Param_All (gdb.Parameter):
65630 """This is the class documentation string."""
65632 show_doc = "Show the state of this parameter"
65633 set_doc = "Set the state of this parameter"
65635 def get_set_string (self):
65637 if (self.value == False):
65639 return "Test Parameter has been set to " + val
65641 def __init__ (self, name):
65642 super (Param_All, self).__init__ (name, gdb.COMMAND_DATA, gdb.PARAM_BOOLEAN)
65645 Param_All ('param-all')
65647 Then in GDB we see this:
65649 (gdb) help set param-all
65650 Set the state of this parameter
65651 This is the class documentation string.
65653 Which is fine. But, if the user skips both of the documentation parts
65656 class Param_None (gdb.Parameter):
65658 def get_set_string (self):
65660 if (self.value == False):
65662 return "Test Parameter has been set to " + val
65664 def __init__ (self, name):
65665 super (Param_None, self).__init__ (name, gdb.COMMAND_DATA, gdb.PARAM_BOOLEAN)
65668 Param_None ('param-none')
65670 Now in GDB we see this:
65672 (gdb) help set param-none
65673 This command is not documented.
65674 This command is not documented.
65676 That's not great, the duplicated text looks a bit weird. If we drop
65677 different parts we get different results. Here's what we get if the
65678 user drops the set_doc and show_doc attributes:
65680 (gdb) help set param-doc
65681 This command is not documented.
65682 This is the class documentation string.
65684 That kind of sucks, we say it's undocumented, then proceed to print
65685 the documentation. Finally, if we drop the class documentation but
65686 keep the set_doc and show_doc:
65688 (gdb) help set param-set-show
65689 Set the state of this parameter
65690 This command is not documented.
65694 So, I think there's room for improvement.
65696 With this patch, for the four cases above we now see this:
65698 # All values provided by the user, no change in this case:
65699 (gdb) help set param-all
65700 Set the state of this parameter
65701 This is the class documentation string.
65703 # Nothing provided by the user, the first string is now different:
65704 (gdb) help set param-none
65705 Set the current value of 'param-none'.
65706 This command is not documented.
65708 # Only the class documentation is provided, the first string is
65709 # changed as in the previous case:
65710 (gdb) help set param-doc
65711 Set the current value of 'param-doc'.
65712 This is the class documentation string.
65714 # Only the set_doc and show_doc are provided, this case is unchanged
65715 # from before the patch:
65716 (gdb) help set param-set-show
65717 Set the state of this parameter
65718 This command is not documented.
65720 The one place where this change might be considered a negative is when
65721 dealing with prefix commands. If we create a prefix command but don't
65722 supply the set_doc / show_doc strings, then this is what we saw before
65725 (gdb) python Param_None ('print param-none')
65726 (gdb) help set print
65727 set print, set pr, set p
65728 Generic command for setting how things print.
65730 List of set print subcommands:
65733 set print param-none -- This command is not documented.
65736 And after my patch:
65738 (gdb) python Param_None ('print param-none')
65739 (gdb) help set print
65740 set print, set pr, set p
65741 Generic command for setting how things print.
65743 List of set print subcommands:
65746 set print param-none -- Set the current value of 'print param-none'.
65749 This seems slightly less helpful than before, but I don't think its
65752 Additionally, I've changed what we print when the get_show_string
65753 method is not provided in Python.
65755 Back when gdb.Parameter was first added to GDB, we didn't provide a
65756 show function when registering the internal command object within
65757 GDB. As a result, GDB would make use of its "magic" mangling of the
65758 show_doc string to create a sentence that would display the current
65759 value (see deprecated_show_value_hack in cli/cli-setshow.c).
65761 However, when we added support for the get_show_string method to
65762 gdb.Parameter, there was an attempt to maintain backward compatibility
65763 by displaying the show_doc string with the current value appended, see
65764 get_show_value in py-param.c. Unfortunately, this isn't anywhere
65765 close to what deprecated_show_value_hack does, and the results are
65766 pretty poor, for example, this is GDB before my patch:
65768 (gdb) show param-none
65769 This command is not documented. off
65771 I think we can all agree that this is pretty bad.
65773 After my patch, we how show this:
65775 (gdb) show param-none
65776 The current value of 'param-none' is "off".
65778 Which at least is a real sentence, even if it's not very informative.
65780 This patch does change the way that the Python API behaves slightly,
65781 but only in the cases when the user has missed providing GDB with some
65782 information. In most cases I think the new behaviour is a lot better,
65783 there's the one case (noted above) which is a bit iffy, but I think is
65786 I've updated the existing gdb.python/py-parameter.exp test to cover
65787 the modified behaviour.
65789 Finally, I've updated the documentation to (I hope) make it clearer
65790 how the various bits of help text come together.
65792 2022-01-26 Andrew Burgess <aburgess@redhat.com>
65794 gdb/python: add gdb.history_count function
65795 Add a new function gdb.history_count to the Python api, this function
65796 returns an integer, the number of items in GDB's value history.
65798 This is useful if you want to pull items from the history by their
65799 absolute number, for example, if you wanted to show a complete history
65800 list. Previously we could figure out how many items are in the
65801 history list by trying to fetch the items, and then catching the
65802 exception when the item is not available, but having this function
65805 2022-01-26 Tom Tromey <tromey@adacore.com>
65807 Remove unused declaration
65808 This removes an unused declaration from top.h. This type is not
65811 2022-01-26 Simon Marchi <simon.marchi@efficios.com>
65813 gdb: convert maintenance target-async and target-non-stop settings to callbacks
65814 This simplifies things a bit, as we don't need two variables and think
65815 about reverting target_async_permitted_1 and target_non_stop_enabled_1
65816 values if we can't change the setting.
65818 Change-Id: I36acab045dacf02ae1988486cfdb27c1dff309f6
65820 2022-01-26 Keith Seitz <keiths@redhat.com>
65822 Reference array of structs instead of first member during memcpy
65823 aarch64-tdep.c defines the following macro:
65825 #define MEM_ALLOC(MEMS, LENGTH, RECORD_BUF) \
65828 unsigned int mem_len = LENGTH; \
65831 MEMS = XNEWVEC (struct aarch64_mem_r, mem_len); \
65832 memcpy(&MEMS->len, &RECORD_BUF[0], \
65833 sizeof(struct aarch64_mem_r) * LENGTH); \
65838 This is simlpy allocating a new array and copying it. However, for
65839 the destination address, it is actually copying into the first member
65840 of the first element of the array (`&MEMS->len"). This elicits a
65841 warning with GCC 12:
65843 ../../binutils-gdb/gdb/aarch64-tdep.c: In function ‘int aarch64_process_record(gdbarch*, regcache*, CORE_ADDR)’:
65844 ../../binutils-gdb/gdb/aarch64-tdep.c:3711:23: error: writing 16 bytes into a region of size 8 [-Werror=stringop-overflow=]
65845 3711 | memcpy(&MEMS->len, &RECORD_BUF[0], \
65847 ../../binutils-gdb/gdb/aarch64-tdep.c:4394:3: note: in expansion of macro ‘MEM_ALLOC’
65848 4394 | MEM_ALLOC (aarch64_insn_r->aarch64_mems, aarch64_insn_r->mem_rec_count,
65850 ../../binutils-gdb/gdb/aarch64-tdep.c:3721:12: note: destination object ‘aarch64_mem_r::len’ of size 8
65851 3721 | uint64_t len; /* Record length. */
65854 The simple fix is to reference the array, `MEMS' as the destination of the copy.
65856 Tested by rebuilding.
65859 # Please enter the commit message for your changes. Lines starting
65860 # with '#' will be kept; you may remove them yourself if you want to.
65861 # An empty message aborts the commit.
65863 # Date: Tue Jan 25 08:28:32 2022 -0800
65866 # Your branch is ahead of 'origin/master' by 1 commit.
65867 # (use "git push" to publish your local commits)
65869 # Changes to be committed:
65870 # modified: aarch64-tdep.c
65873 2022-01-26 Simon Marchi <simon.marchi@polymtl.ca>
65875 gdb: add string_file::release method
65876 A common pattern for string_file is to want to move out the internal
65877 string buffer, because it is the result of the computation that we want
65878 to return. It is the reason why string_file::string returns a non-const
65879 reference, as explained in the comment. I think it would make sense to
65880 have a dedicated method for that instead and make string_file::string
65881 return a const reference.
65883 This allows removing the explicit std::move in the typical case. Note
65884 that compile_program::compute was missing a move, meaning that the
65885 resulting string was copied. With the new version, it's not possible to
65888 Change-Id: Ieaefa35b73daa7930b2f3a26988b6e3b4121bb79
65890 2022-01-26 Tom Tromey <tromey@adacore.com>
65892 Add a way to temporarily set a gdb parameter from Python
65893 It's sometimes useful to temporarily set some gdb parameter from
65894 Python. Now that the 'endian' crash is fixed, and now that the
65895 current language is no longer captured by the Python layer, it seems
65896 reasonable to add a helper function for this situation.
65898 This adds a new gdb.with_parameter function. This creates a context
65899 manager which temporarily sets some parameter to a specified value.
65900 The old value is restored when the context is exited. This is most
65901 useful with the Python "with" statement:
65903 with gdb.with_parameter('language', 'ada'):
65906 This also adds a simple function to set a parameter,
65907 gdb.set_parameter, as suggested by Andrew.
65909 This is PR python/10790.
65911 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=10790
65913 2022-01-26 Tom Tromey <tromey@adacore.com>
65915 Fix another crash with gdb parameters in Python
65916 While looking into the language-capturing issue, I found another way
65917 to crash gdb using parameters from Python:
65919 (gdb) python print(gdb.parameter('endian'))
65921 (This is related to PR python/12188, though this patch isn't going to
65922 fix what that bug is really about.)
65924 The problem here is that the global variable that underlies the
65925 "endian" parameter is initialized to NULL. However, that's not a
65926 valid value for an "enum" set/show parameter.
65928 My understanding is that, in gdb, an "enum" parameter's underlying
65929 variable must have a value that is "==" (not just strcmp-equal) to one
65930 of the values coming from the enum array. This invariant is relied on
65933 I started this patch by fixing the problem with "endian". Then I
65934 added some assertions to add_setshow_enum_cmd to try to catch other
65935 problems of the same type.
65937 This patch fixes all the problems that I found. I also looked at all
65938 the calls to add_setshow_enum_cmd to ensure that they were all
65939 included in the gdb I tested. I think they are: there are no calls in
65940 nat-* files, or in remote-sim.c; and I was trying a build with all
65941 targets, Python, and Guile enabled.
65943 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=12188
65945 2022-01-26 Tom Tromey <tromey@adacore.com>
65947 Change how Python architecture and language are handled
65948 Currently, gdb's Python layer captures the current architecture and
65949 language when "entering" Python code. This has some undesirable
65950 effects, and so this series changes how this is handled.
65952 First, there is code like this:
65954 gdbpy_enter enter_py (python_gdbarch, python_language);
65956 This is incorrect, because both of these are NULL when not otherwise
65957 assigned. This can cause crashes in some cases -- I've added one to
65958 the test suite. (Note that this crasher is just an example, other
65959 ones along the same lines are possible.)
65961 Second, when the language is captured in this way, it means that
65962 Python code cannot affect the current language for its own purposes.
65963 It's reasonable to want to write code like this:
65965 gdb.execute('set language mumble')
65966 ... stuff using the current language
65967 gdb.execute('set language previous-value')
65969 However, this won't actually work, because the language is captured on
65970 entry. I've added a test to show this as well.
65972 This patch changes gdb to try to avoid capturing the current values.
65973 The Python concept of the current gdbarch is only set in those few
65974 cases where a non-default value is computed or needed; and the
65975 language is not captured at all -- instead, in the cases where it's
65976 required, the current language is temporarily changed.
65978 2022-01-26 H.J. Lu <hjl.tools@gmail.com>
65980 bfd: Make bfd.stamp depend on source bfd.texi
65981 Make bfd.stamp depend on source bfd.texi to avoid regenerating
65982 doc/bfd.info for each make run.
65985 * Makefile.in: Regenerate.
65986 * doc/local.mk (%D%/bfd.stamp): Depend on $(srcdir)/%D%/bfd.texi.
65988 2022-01-26 H.J. Lu <hjl.tools@gmail.com>
65990 ld: Rewrite lang_size_relro_segment_1
65991 1. Compute the desired PT_GNU_RELRO segment base and find the maximum
65992 section alignment of sections starting from the PT_GNU_RELRO segment.
65993 2. Find the first preceding load section.
65994 3. Don't add the 1-page gap between the first preceding load section and
65995 the relro segment if the maximum page size >= the maximum section
65996 alignment. Align the PT_GNU_RELRO segment first. Subtract the maximum
65997 page size if therer is still a 1-page gap.
66001 * ldlang.c (lang_size_relro_segment_1): Rewrite.
66002 * testsuite/ld-x86-64/pr28743-1.d: New file.
66003 * testsuite/ld-x86-64/pr28743-1.s: Likewise.
66004 * testsuite/ld-x86-64/x86-64.exp: Run pr28743-1.
66006 2022-01-26 Lancelot SIX <lancelot.six@amd.com>
66008 gdb/testsuite: Ensure constant test name in gdb.base/break-interp.exp
66009 When running the testsuite, I have lines similar to the following in the
66013 PASS: gdb.base/break-interp.exp: ldprelink=NO: ldsepdebug=NO: first backtrace: p /x 0x7f283d2f0fd1
66015 PASS: gdb.base/break-interp.exp: ldprelink=NO: ldsepdebug=NO: binprelink=NO: binsepdebug=NO: binpie=NO: INNER: first backtrace: p /x 0x7f00de0317a5
66019 The address part of the command might change between execution of the
66020 test, which adds noise to a diff between two .sum files.
66022 This patch changes to test name to "p /x $pc" in order to have constant
66025 Tested on x86_64-Linux.
66027 Change-Id: I973c1237a084dd6d424276443cbf0920533c9a21
66029 2022-01-26 GDB Administrator <gdbadmin@sourceware.org>
66031 Automatic date update in version.in
66033 2022-01-25 Tom Tromey <tom@tromey.com>
66035 Always print the "host libthread-db" message to stdout
66036 linux-thread-db.c has a bit of unusual code that unconditionally
66037 prints a message, but decides whether to print to gdb_stdout or
66038 gdb_stdlog based on a debug flag. It seems better to me to simply
66039 always print this; and this is the only spot in gdb where we
66040 conditionally pass gdb_stdout to one of the f*_unfiltered functions.
66042 2022-01-25 Tom Tromey <tom@tromey.com>
66044 Reduce explicit use of gdb_stdout
66045 In an earlier version of the pager rewrite series, it was important to
66046 audit unfiltered output calls to see which were truly necessary.
66048 This is no longer necessary, but it still seems like a decent cleanup
66049 to change calls to avoid explicitly passing gdb_stdout. That is,
66050 rather than using something like fprintf_unfiltered with gdb_stdout,
66051 the code ought to use plain printf_unfiltered instead.
66053 This patch makes this change. I went ahead and converted all the
66054 _filtered calls I could find, as well, for the same clarity.
66056 2022-01-25 Tom Tromey <tom@tromey.com>
66058 Sent timing stats to gdb_stdlog
66059 This changes the time / space / symtab per-command statistics code to
66060 send its output to gdb_stdlog rather than gdb_stdout. This seems
66061 slightly more correct to me.
66063 Send some error output to gdb_stderr
66064 This changes some code to send some error messages to gdb_stderr
66065 rather than gdb_stdout.
66067 2022-01-25 Klaus Ziegler <klausz@haus-gisela.de>
66069 Fix a probem building the binutils on SPARC/amd64
66071 * elf/common.h (AT_SUN_HWCAP): Make definition conditional.
66073 2022-01-25 H.J. Lu <hjl.tools@gmail.com>
66075 bfd: Regenerate Makefile.in
66076 * Makefile.in: Regenerate.
66078 2022-01-25 Mike Frysinger <vapier@gentoo.org>
66080 gold: drop old cygnus install hack
66081 The gold subdir doesn't actually have a manual, so this hack doesn't
66082 do anything. Plus the automake cygnus option was removed years ago
66083 by Simon in d0ac1c44885daf68f631befa37e ("Bump to autoconf 2.69 and
66084 automake 1.15.1"). So delete it here.
66086 gas: drop old cygnus install hack
66087 This was needed when gas was using the automake cygnus option, but
66088 this was removed years ago by Simon in d0ac1c44885daf68f631befa37e
66089 ("Bump to autoconf 2.69 and automake 1.15.1"). So delete it here.
66090 The info pages are already & still installed by default w/out it.
66092 2022-01-25 GDB Administrator <gdbadmin@sourceware.org>
66094 Automatic date update in version.in
66096 2022-01-24 H.J. Lu <hjl.tools@gmail.com>
66098 bfd: Update doc/local.mk
66100 * Makefile.in: Regenerate.
66101 * doc/local.mk (AM_MAKEINFOFLAGS): Add -I "$(srcdir)/%D%" -I %D%.
66103 (%D%/bfd.texi): Removed.
66104 (doc/bfd/index.html): Remove -I$(srcdir). Replace bfd.texi with
66107 2022-01-24 Roland McGrath <mcgrathr@google.com>
66109 bfd/doc: Fix racy build failure from missing mkdir
66111 * doc/local.mk (%D%/bfdver.texi): Add mkdir command.
66113 2022-01-24 Martin Sebor <msebor@redhat.com>
66115 Fix a proble building the libiberty library with gcc-12.
66117 * regex.c: Suppress -Wuse-after-free.
66119 2022-01-24 Andrew Burgess <aburgess@redhat.com>
66121 gdb/doc: improve description for Window.click on Python TUI windows
66122 The description of the Window.click method doesn't mention where the
66123 coordinates are anchored (it's the top left corner).
66125 This minor tweak just mentions this point.
66127 2022-01-24 Nick Clifton <nickc@redhat.com>
66129 Update Bulgarian, French, Romaniam and Ukranian translation for some of the sub-directories
66131 2022-01-24 GDB Administrator <gdbadmin@sourceware.org>
66133 Automatic date update in version.in
66135 2022-01-23 Tom Tromey <tom@tromey.com>
66137 Simplify some Rust expression-evaluation code
66138 A few Rust operations do a bit of work in their 'evaluate' functions
66139 and then call another function -- but are also the only caller. This
66140 patch simplifies this code by removing the extra layer.
66142 Tested on x86-64 Fedora 34. I'm checking this in.
66144 2022-01-23 H.J. Lu <hjl.tools@gmail.com>
66146 bfd: Partially revert commit 0e3839bde6f
66149 commit 0e3839bde6f93e1e3eefce815be3636e3d81054d
66150 Author: H.J. Lu <hjl.tools@gmail.com>
66151 Date: Sun Jan 23 07:29:27 2022 -0800
66153 bfd: Properly install library and header files
66156 * Makefile.am: Revert bfdlib_LTLIBRARIES and bfdinclude_HEADERS
66158 * Makefile.in: Regenerate.
66160 2022-01-23 H.J. Lu <hjl.tools@gmail.com>
66162 bfd: Properly install library and header files
66163 Rename bfdlib_LTLIBRARIES and bfdinclude_HEADERS to lib_LTLIBRARIES and
66164 include_HEADERS to fix the missing installed library and header files in
66167 commit bd32be01c997f686ab0b53f0640eaa0aeb61fbd3
66168 Author: Mike Frysinger <vapier@gentoo.org>
66169 Date: Fri Dec 3 00:23:20 2021 -0500
66171 bfd: merge doc subdir up a level
66174 * Makefile.am (bfdlib_LTLIBRARIES): Renamed to ...
66175 (lib_LTLIBRARIES): This.
66176 (bfdinclude_HEADERS): Renamed to ...
66177 (include_HEADERS): This.
66178 * Makefile.in: Regenerate.
66179 * doc/local.mk (install): Removed.
66181 2022-01-23 H.J. Lu <hjl.tools@gmail.com>
66183 Regenerate Makefile.in files with automake 1.15.1
66184 Regenerate Makefile.in files with the unmodified automake 1.15.1 to
66187 runstatedir = @runstatedir@
66191 * Makefile.in: Regenerate.
66195 * Makefile.in: Regenerate.
66199 * Makefile.in: Regenerate.
66203 * Makefile.in: Regenerate.
66204 * testsuite/Makefile.in: Likewise.
66208 * Makefile.in: Regenerate.
66212 * Makefile.in: Regenerate.
66216 * Makefile.in: Regenerate.
66218 2022-01-23 H.J. Lu <hjl.tools@gmail.com>
66220 Regenerate configure files with autoconf 2.69
66221 Regenerate configure files with the unmodified autoconf 2.69 to remove
66223 --runstatedir=DIR modifiable per-process data [LOCALSTATEDIR/run]
66227 * configure: Regenerate.
66231 * configure: Regenerate.
66235 * configure: Regenerate.
66239 * configure: Regenerate.
66243 * configure: Regenerate.
66247 * configure: Regenerate.
66251 * configure: Regenerate.
66253 2022-01-23 GDB Administrator <gdbadmin@sourceware.org>
66255 Automatic date update in version.in
66257 2022-01-22 Mike Frysinger <vapier@gentoo.org>
66259 bfd: merge doc subdir up a level
66260 This avoids a recursive make into the doc subdir and speeds up the
66261 build slightly. It also allows for more parallelism.
66263 bfd: rename core.texi to corefile.texi
66264 This is a generated file name from a correspondingly named C file.
66265 Rename it to avoid unique build rules since there's no difference
66266 to the generated manual.
66268 bfd: replace doc header generation with pattern rules
66269 This unifies boilerplate rules for most files with pattern rules.
66271 2022-01-22 Martin Storsj? <martin@martin.st>
66273 Allow inferring tmp_prefix from the dll name from a def file.
66275 2022-01-22 Alexander von Gluck IV <kallisti5@unixzen.com>
66277 Adjust default page sizes for haiku arm.
66278 * configure.tgt (arm-haiku): Fix typo.
66279 * emulparams/armelf_haiku.su (MAXPAGESIZE): Use the default value.
66280 (COMMONPAGESIZE): Likewise.
66282 2022-01-22 Nick Clifton <nickc@redhat.com>
66284 Update release makeing script with new release numbers
66286 Change version number to 2.38.50 and regenerate files
66288 Add markers for 2.38 branch
66290 2022-01-22 Lifang Xia <lifang_xia@linux.alibaba.com>
66292 RISC-V: create new frag after alignment.
66295 The alignment may be removed in linker. We need to create new frag after
66296 alignment to prevent the assembler from computing static offsets.
66299 * config/tc-riscv.c (riscv_frag_align_code): Create new frag.
66301 2022-01-22 GDB Administrator <gdbadmin@sourceware.org>
66303 Automatic date update in version.in
66305 2022-01-21 Simon Marchi <simon.marchi@polymtl.ca>
66307 gdb: include gdbsupport/buildargv.h in ser-mingw.c
66311 /home/simark/src/binutils-gdb/gdb/ser-mingw.c: In function ‘int pipe_windows_open(serial*, const char*)’:
66312 /home/simark/src/binutils-gdb/gdb/ser-mingw.c:870:3: error: ‘gdb_argv’ was not declared in this scope
66313 870 | gdb_argv argv (name);
66316 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28802
66317 Change-Id: I7f3e8ec5f9ca8582d587545fdf6b69901259f199
66319 2022-01-21 Nick Clifton <nickc@redhat.com>
66321 Updated Serbian translation for the ld sub-directory
66323 2022-01-21 Andrew Burgess <aburgess@redhat.com>
66325 gdb/doc: fill in two missing @r
66326 I noticed two places in the docs where we appear to be missing @r.
66327 makeinfo seems to do the correct things despite these being
66328 missing (at least, I couldn't see any difference in the pdf or info
66329 output), but it doesn't hurt to have the @r in place.
66331 2022-01-21 Mike Frysinger <vapier@gentoo.org>
66333 drop old unused stamp-h.in file
66334 This was needed by ancient versions of automake, but that hasn't been
66335 the case since at least automake-1.5, so punt this from the tree.
66337 2022-01-21 Simon Marchi <simon.marchi@polymtl.ca>
66339 gdbsupport/gdb_regex.cc: replace defs.h include with common-defs.h
66340 This was forgotten when gdb_regex was moved from gdb to gdbsupport.
66342 Change-Id: I73b446f71861cabbf7afdb7408ef9d59fa64b804
66344 2022-01-21 GDB Administrator <gdbadmin@sourceware.org>
66346 Automatic date update in version.in
66348 2022-01-20 Tom Tromey <tromey@adacore.com>
66350 Avoid bad breakpoints with --gc-sections
66351 We found a case where --gc-sections can cause gdb to set an invalid
66352 breakpoint. In the included test case, gdb will set a breakpoint with
66353 two locations, one of which is 0x0.
66355 The code in lnp_state_machine::check_line_address is intended to
66356 filter out this sort of problem, but in this case, the entire CU is
66357 empty, causing unrelocated_lowpc==0x0 -- which circumvents the check.
66359 It seems to me that if a CU is empty like this, then it is ok to
66360 simply ignore the line table, as there won't be any locations anyway.
66362 2022-01-20 GDB Administrator <gdbadmin@sourceware.org>
66364 Automatic date update in version.in
66366 2022-01-19 Maciej W. Rozycki <macro@embecosm.com>
66368 Add `set print array-indexes' tests for C/C++ arrays
66369 Add `set print array-indexes' tests for C/C++ arrays, complementing one
66370 for Fortran arrays.
66372 2022-01-19 Maciej W. Rozycki <macro@embecosm.com>
66374 Respect `set print array-indexes' with Fortran arrays
66375 Add `set print array-indexes' handling for Fortran arrays. Currently
66376 the setting is ignored and indices are never shown.
66378 Keep track of the most recent index handled so that any outstanding
66379 repeated elements printed when the limit set by `set print elements' is
66380 hit have the correct index shown.
66382 Output now looks like:
66384 (gdb) set print array-indexes on
66385 (gdb) print array_1d
66386 $1 = ((-2) = 1, (-1) = 1, (0) = 1, (1) = 1, (2) = 1)
66387 (gdb) set print repeats 4
66388 (gdb) set print elements 12
66389 (gdb) print array_2d
66390 $2 = ((-2) = ((-2) = 2, <repeats 5 times>) (-1) = ((-2) = 2, <repeats 5 times>) (0) = ((-2) = 2, (-1) = 2, ...) ...)
66393 for a 5-element vector and a 5 by 5 array filled with the value of 2.
66395 2022-01-19 Maciej W. Rozycki <macro@embecosm.com>
66397 Add `set print repeats' tests for C/C++ arrays
66398 Add `set print repeats' tests for C/C++ arrays, complementing one for
66399 Fortran arrays and covering the different interpretation of the `set
66400 print elements' setting in particular where the per-dimension count of
66401 the elements handled is matched against the trigger rather than the
66402 total element count as with Fortran arrays.
66404 2022-01-19 Maciej W. Rozycki <macro@embecosm.com>
66406 Respect `set print repeats' with Fortran arrays
66407 Implement `set print repeats' handling for Fortran arrays. Currently
66408 the setting is ignored and always treated as if no limit was set.
66410 Unlike the generic array walker implemented decades ago the Fortran one
66411 is a proper C++ class. Rather than trying to mimic the old walker then,
66412 which turned out a bit of a challenge where interacting with the `set
66413 print elements' setting, write it entirely from scratch, by adding an
66414 extra specialization handler method for processing dimensions other than
66415 the innermost one and letting the specialization class call the `walk_1'
66416 method from the handler as it sees fit. This way repeats can be tracked
66417 and the next inner dimension recursed into as a need arises only, or
66418 unconditionally in the base class.
66420 Keep track of the dimension number being handled in the class rather as
66421 a parameter to the walker so that it does not have to be passed across
66422 by the specialization class.
66424 Use per-dimension element count tracking, needed to terminate processing
66425 early when the limit set by `set print elements' is hit. This requires
66426 extra care too where the limit triggers exactly where another element
66427 that is a subarray begins. In that case rather than recursing we need
66428 to terminate processing or lone `(...)' would be printed. Additionally
66429 if the skipped element is the last one in the current dimension we need
66430 to print `...' by hand, because `continue_walking' won't print it at the
66431 upper level, because it can see the last element has already been taken
66434 Preserve the existing semantics of `set print elements' where the total
66435 count of the elements handled is matched against the trigger level which
66436 is unlike with the C/C++ array printer where the per-dimension element
66437 count is used instead.
66439 Output now looks like:
66441 (gdb) set print repeats 4
66442 (gdb) print array_2d
66443 $1 = ((2, <repeats 5 times>) <repeats 5 times>)
66444 (gdb) set print elements 12
66445 (gdb) print array_2d
66446 $2 = ((2, <repeats 5 times>) (2, <repeats 5 times>) (2, 2, ...) ...)
66449 for a 5 by 5 array filled with the value of 2.
66451 Amend existing test cases accordingly that rely on the current incorrect
66452 behavior and explicitly request that there be no limit for printing
66453 repeated elements there.
66455 Add suitable test cases as well covering sliced arrays in particular.
66457 Co-Authored-By: Andrew Burgess <andrew.burgess@embecosm.com>
66459 2022-01-19 John Baldwin <jhb@FreeBSD.org>
66461 fbsd-nat: Add include for gdb_argv.
66463 2022-01-19 Alan Modra <amodra@gmail.com>
66465 PowerPC64 DT_RELR ELFv1
66466 More fun with R_PPC64_NONE found in .opd. Fixed by the
66467 allocate_dynrelocs and ppc64_elf_size_dynamic_sections changes, and
66468 since we are doing ifunc, opd and SYMBOL_REFERENCES_LOCAL tests later,
66469 don't duplicate that work in check_relocs.
66471 * elf64-ppc.c (ppc64_elf_check_relocs): Remove opd and ifunc
66472 conditions for rel_count.
66473 (dec_dynrel_count): Likewise.
66474 (allocate_dynrelocs): Test for opd and ifunc when allocating
66476 (ppc64_elf_size_dynamic_sections): Likewise.
66478 2022-01-19 Alan Modra <amodra@gmail.com>
66480 PowerPC64 DT_RELR local PLT
66481 Similarly to the local GOT case.
66483 * elf64-ppc.c (ppc64_elf_size_dynamic_sections): Don't allocate
66484 space for PLT relocs against local syms when enable_dt_relr.
66486 2022-01-19 Alan Modra <amodra@gmail.com>
66488 PowerPC64 DT_RELR local GOT
66489 Fixes another case where we end up with superfluous R_PPC64_NONE.
66491 * elf64-ppc.c (ppc64_elf_size_dynamic_sections): Don't allocate
66492 space for GOT relocs against non-TLS local syms when enable_dt_relr.
66493 (ppc64_elf_layout_multitoc): Likewise.
66495 2022-01-19 GDB Administrator <gdbadmin@sourceware.org>
66497 Automatic date update in version.in
66499 2022-01-18 Alan Modra <amodra@gmail.com>
66501 Re: PowerPC64 DT_RELR
66502 HJ: "There are 238 R_PPC64_NONEs in libc.so.6 alone."
66503 Indeed, let's make them go away. I had the SYMBOL_REFERENCES_LOCAL
66504 test in the wrong place. check_relocs is too early to know whether a
66505 symbol is dynamic in a shared library. Lots of glibc symbols are made
66506 local by version script, but that doesn't happen until
66507 size_dynamic_sections.
66509 * elf64-ppc.c (ppc64_elf_check_relocs): Don't count relative relocs
66510 here depending on SYMBOL_REFERENCES_LOCAL.
66511 (dec_dynrel_count): Likewise.
66512 (allocate_dynrelocs): Do so here instead.
66514 2022-01-18 Tom Tromey <tom@tromey.com>
66516 Fix the remote-sim.c build
66517 My earlier patch to move gdb_argv broke the remote-sim.c build. This
66518 patch fixes the bug. I'm checking it in.
66520 2022-01-18 Simon Marchi <simon.marchi@polymtl.ca>
66522 gdbserver: introduce remote_debug_printf
66523 Add remote_debug_printf, and use it for all debug messages controlled by
66526 Change remote_debug to be a bool, which is trivial in this case.
66528 Change-Id: I90de13cb892faec3830047b571661822b126d6e8
66530 2022-01-18 Simon Marchi <simon.marchi@polymtl.ca>
66532 gdbserver: introduce threads_debug_printf, THREADS_SCOPED_DEBUG_ENTER_EXIT
66533 Add the threads_debug_printf and THREADS_SCOPED_DEBUG_ENTER_EXIT, which
66534 use the logging infrastructure from gdbsupport/common-debug.h. Replace
66535 all debug_print uses that are predicated by debug_threads with
66536 threads_dethreads_debug_printf. Replace uses of the debug_enter and
66537 debug_exit macros with THREADS_SCOPED_DEBUG_ENTER_EXIT, which serves
66538 essentially the same purpose, but allows showing what comes between the
66539 enter and the exit in an indented form.
66541 Note that "threads" debug is currently used for a bit of everything in
66542 GDBserver, not only threads related stuff. It should ideally be cleaned
66543 up and separated logically as is done in GDB, but that's out of the
66544 scope of this patch.
66546 Change-Id: I2d4546464462cb4c16f7f1168c5cec5a89f2289a
66548 2022-01-18 Simon Marchi <simon.marchi@polymtl.ca>
66550 gdbserver: turn debug_threads into a boolean
66551 debug_threads is always used as a boolean. Except in ax.cc and
66552 tracepoint.cc. These files have their own macros that use
66553 debug_threads, and have a concept of verbosity level. But they both
66554 have a single level, so it's just a boolean in the end.
66556 Remove this concept of level. If we ever want to re-introduce it, I
66557 think it will be better implemented in a more common location.
66559 Change debug_threads to bool and adjust some users that were treating it
66562 Change-Id: I137f596eaf763a08c977dd74417969cedfee9ecf
66564 2022-01-18 Tom Tromey <tom@tromey.com>
66566 Simplify Ada catchpoints
66567 All the Ada catchpoints use the same breakpoint_ops contents, because
66568 the catchpoint itself records its kind. This patch simplifies the
66569 code by removing the redundant ops structures.
66571 Move "catch exec" to a new file
66572 The "catch exec" code is reasonably self-contained, and so this patch
66573 moves it out of breakpoint.c (the second largest source file in gdb)
66574 and into a new file, break-catch-exec.c.
66576 Move "catch fork" to a new file
66577 The "catch fork" code is reasonably self-contained, and so this patch
66578 moves it out of breakpoint.c (the second largest source file in gdb)
66579 and into a new file, break-catch-fork.c.
66581 Unify "catch fork" and "catch vfork"
66582 I noticed that "catch fork" and "catch vfork" are nearly identical.
66583 This patch simplifies the code by unifying these two cases.
66585 Move gdb_regex to gdbsupport
66586 This moves the gdb_regex convenience class to gdbsupport.
66588 Introduce gdb-hashtab module in gdbsupport
66589 gdb has some extensions and helpers for working with the libiberty
66590 hash table. This patch consolidates these and moves them to
66593 Move gdb obstack code to gdbsupport
66594 This moves the gdb-specific obstack code -- both extensions like
66595 obconcat and obstack_strdup, and things like auto_obstack -- to
66598 Move gdb_argv to gdbsupport
66599 This moves the gdb_argv class to a new header in gdbsupport.
66601 Simplify event_location_probe
66602 event_location_probe currently stores two strings, but really only
66603 needs one. This patch simplifies it and removes some unnecessary
66606 Use std::string in event_location
66607 This changes event_location to use std::string, removing some manual
66608 memory management, and an unnecessary string copy.
66610 Split event_location into subclasses
66611 event_location uses the old C-style discriminated union approach.
66612 However, it's better to use subclassing, as this makes the code
66613 clearer and removes some chances for error. This also enables future
66614 cleanups to avoid manual memory management and copies.
66616 Remove EL_* macros from location.c
66617 This patch removes the old-style EL_* macros from location.c. This
66618 cleans up the code by itself, IMO, but also enables further cleanups
66619 in subsequent patches.
66621 Boolify explicit_to_string_internal
66622 This changes explicit_to_string_internal to use 'bool' rather than
66625 Remove a use of xfree in location.c
66626 This small cleanup removes a use of xfree from location.c, by
66627 switching to unique_xmalloc_ptr. One function is only used in
66628 location.c, so it is made static. And, another function is changed to
66631 2022-01-18 Simon Marchi <simon.marchi@polymtl.ca>
66633 gdb: use ptid_t::to_string instead of target_pid_to_str in debug statements
66634 Same idea as 0fab79556484 ("gdb: use ptid_t::to_string in infrun debug
66635 messages"), but throughout GDB.
66637 Change-Id: I62ba36eaef29935316d7187b9b13d7b88491acc1
66639 2022-01-18 Andrew Burgess <aburgess@redhat.com>
66641 gdb: preserve `|` in connection details string
66642 Consider this GDB session:
66645 (gdb) target remote | gdbserver - ~/tmp/hello.x
66646 Remote debugging using | gdbserver - ~/tmp/hello.x
66648 (gdb) info connections
66649 Num What Description
66650 * 1 remote gdbserver - ~/tmp/hello.x Remote target using gdb-specific protocol
66651 (gdb) python conn = gdb.selected_inferior().connection
66652 (gdb) python print(conn.details)
66653 gdbserver - ~/tmp/hello.x
66656 I think there are two things wrong here, first in the "What" column of
66657 the 'info connections' output, I think the text should be:
66659 remote | gdbserver - ~/tmp/hello.x
66661 to correctly show the user how the connection was established. And in
66662 a similar fashion, I think that the `details` string of the
66663 gdb.TargetConnection object should be:
66665 | gdbserver - ~/tmp/hello.x
66667 This commit makes this change. Currently the '|' is detected and
66668 removed in gdb/serial.c. The string passed to the pipe_ops
66669 structure (from gdb/ser-pipe.c), doesn't then, contain the `|`, this
66670 is instead implied by the fact that it is a pipes based implementation
66671 of the serial_ops interface.
66673 After this commit we still detect the `|` in gdb/serial.c, but we now
66674 store the full string (including the `|`) in the serial::name member
66677 For pipe based serial connections, this name is only used for
66678 displaying the two fields I mention above, and in pipe_open (from
66679 gdb/ser-pipe.c), and in pipe_open, we now know to skip over the `|`.
66681 The benefit I see from this change is that GDB's output now more
66682 accurately reflects the commands used to start a target, thus making
66683 it easier for a user to understand what is going on.
66685 2022-01-18 Tiezhu Yang <yangtiezhu@loongson.cn>
66687 gdb: testsuite: print explicit test result for gdb.base/dfp-test.exp
66688 In the current code, if decimal floating point is not supported for
66689 this target, there is no binary file dfp-test, and also there is no
66690 test result after execute the following commands:
66692 $ make check-gdb TESTS="gdb.base/dfp-test.exp"
66693 $ grep error gdb/testsuite/gdb.log
66694 /home/loongson/gdb.git/gdb/testsuite/gdb.base/dfp-test.c:39:1: error: decimal floating point not supported for this target
66696 $ cat gdb/testsuite/gdb.sum
66698 Running target unix
66699 Running /home/loongson/gdb.git/gdb/testsuite/gdb.base/dfp-test.exp ...
66701 === gdb Summary ===
66706 $ make check-gdb TESTS="gdb.base/dfp-test.exp"
66707 $ cat gdb/testsuite/gdb.sum
66709 Running target unix
66710 Running /home/loongson/gdb.git/gdb/testsuite/gdb.base/dfp-test.exp ...
66711 UNSUPPORTED: gdb.base/dfp-test.exp: decimal floating point not supported for this target.
66713 === gdb Summary ===
66715 # of unsupported tests 1
66718 2022-01-18 Simon Marchi <simon.marchi@efficios.com>
66720 bfd/elf64-ppc.c: fix clang -Wbitwise-instead-of-logical warning in ppc64_elf_check_init_fini
66721 I see this error with clang-14:
66724 /home/smarchi/src/binutils-gdb/bfd/elf64-ppc.c:13131:11: error: use of bitwise '&' with boolean operands [-Werror,-Wbitwise-instead-of-logical]
66725 return (check_pasted_section (info, ".init")
66726 ~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
66728 Fix by replacing & with &&. But given that the check_pasted_section
66729 function has side-effects and we want to make sure both calls are made,
66730 assign to temporary variables before evaluating the `&&`.
66732 Change-Id: I849e1b2401bea5f4d8ef3ab9af99ba9e3ef42490
66734 2022-01-18 Alan Modra <amodra@gmail.com>
66736 PR28029, debuginfod tests
66737 binutils/NEWS says of the change in --process-links semantics:
66738 If other debug section display options are also enabled (eg
66739 --debug-dump=info) then the contents of matching sections in both the main
66740 file and the separate debuginfo file *will* be displayed. This is because in
66741 most cases the debug section will only be present in one of the files.
66743 Implying that debug info is dumped without --process-links. Indeed
66744 that appears to be the case for readelf. This does the same for
66748 * objdump.c (dump_bfd): Do not exit early when !is_mainfile
66749 && !processlinks, instead just exclude non-debug output.
66750 (dump_dwarf): Add is_mainfile parameter and pass to
66751 dump_dwarf_section.
66752 (dump_dwarf_section): Only display debug sections when
66753 !is_mainfile and !process_links.
66755 2022-01-18 Alan Modra <amodra@gmail.com>
66757 Check thin archive element file size against archive header
66758 Makes it a little less likely for someone to break their thin archives.
66760 * archive.c (_bfd_get_elt_at_filepos): Check thin archive
66763 2022-01-18 Alan Modra <amodra@gmail.com>
66765 lang_size_relro_segment tidy
66766 This function has seen too many minimal change style edits.
66767 No functional changes in this patch.
66769 * ldlang.c (lang_size_relro_segment): Tidy.
66771 2022-01-18 Alan Modra <amodra@gmail.com>
66774 PowerPC64 takes a more traditional approach to DT_RELR than x86. Count
66775 relative relocs in check_relocs, allocate space for them and output in
66776 the usual places but not doing so when enable_dt_relr. DT_RELR is
66777 sized in the existing ppc stub relaxation machinery, run via the
66778 linker's ldemul_after_allocation hook. DT_RELR is output in the same
66779 function that writes ppc stubs, run via ldemul_finish.
66781 This support should be considered experimental.
66784 * elf64-ppc.c (struct ppc_local_dyn_relocs): Renamed from
66785 ppc_dyn_relocs. Add rel_count field. Update uses.
66786 (struct ppc_dyn_relocs): New. Replace all uses of elf_dyn_relocs.
66787 (struct ppc_link_hash_table): Add relr_alloc, relr_count and
66789 (ppc64_elf_copy_indirect_symbol): Merge rel_count.
66790 (ppc64_elf_check_relocs): Init rel_count for global and local syms.
66791 (dec_dynrel_count): Change r_info param to reloc pointer. Update
66792 all callers. Handle decrementing rel_count.
66793 (allocate_got): Don't allocate space for relative relocs when
66795 (allocate_dynrelocs): Likewise.
66796 (ppc64_elf_size_dynamic_sections): Likewise. Handle srelrdyn.
66797 (ppc_build_one_stub): Don't emit relative relocs on .branch_lt.
66798 (compare_relr_address, append_relr_off): New functions.
66799 (got_and_plt_relr_for_local_syms, got_and_plt_relr): Likewise.
66800 (ppc64_elf_size_stubs): Size .relr.syn.
66801 (ppc64_elf_build_stubs): Emit .relr.dyn.
66802 (build_global_entry_stubs_and_plt): Don't output relative relocs
66803 when enable_dt_relr.
66804 (write_plt_relocs_for_local_syms): Likewise.
66805 (ppc64_elf_relocate_section): Likewise.
66807 * testsuite/lib/binutils-common.exp (supports_dt_relr): Add
66810 * emulparams/elf64ppc.sh: Source dt-relr.sh.
66811 * testsuite/ld-elf/dt-relr-2b.d: Adjust for powerpc.
66812 * testsuite/ld-elf/dt-relr-2c.d: Likewise.
66813 * testsuite/ld-elf/dt-relr-2d.d: Likewise.
66814 * testsuite/ld-elf/dt-relr-2e.d: Likewise.
66816 2022-01-18 Alan Modra <amodra@gmail.com>
66818 tweak __ehdr_start visibility and flags for check_relocs
66820 * elf-bfd.h (UNDEFWEAK_NO_DYNAMIC_RELOC): Test linker_def.
66822 * ldelf.c (ldelf_before_allocation): Don't force __ehdr_start
66823 local and hidden here..
66824 * ldlang.c (lang_symbol_tweaks): ..do so here instead and set
66825 def_regular and linker_def for check_relocs. New function
66826 extracted from lang_process.
66828 2022-01-18 GDB Administrator <gdbadmin@sourceware.org>
66830 Automatic date update in version.in
66832 2022-01-17 Nick Clifton <nickc@redhat.com>
66834 Update the config.guess and config.sub files from the master repository and regenerate files.
66836 2022-01-17 Sergey Belyashov <sergey.belyashov@gmail.com>
66838 Fix Z80 assembly failure.
66840 * app.c (do_scrub_chars): Correct handling when the symbol is not 'af'.
66842 2022-01-17 Simon Marchi <simon.marchi@polymtl.ca>
66844 gdb/infrun: rename variable and move to more specific scope
66845 Move the "started" variable to the scope it's needed, and rename it to
66846 "step_over_started".
66848 Change-Id: I56f3384dbd328f55198063bb855edda10f1492a3
66850 2022-01-17 Jan Beulich <jbeulich@suse.com>
66852 x86: adjust struct instr_info field types
66853 Now that this lives on the stack, let's have it be a little less
66854 wasteful in terms of space. Switch boolean fields to "bool" (also when
66855 this doesn't change their size) and also limit the widths of "rex",
66856 "rex_used", "op_ad", and "op_index". Do a little bit of re-ordering as
66857 well to limit the number of padding holes.
66859 x86: drop index16 field
66860 There's a single use on a generally infrequently taken code path. Put
66861 the necessary conditional there instead.
66863 x86: drop most Intel syntax register name arrays
66864 By making use of, in particular, oappend_maybe_intel() there's no need
66865 for this redundant set of static data.
66867 x86: fold variables in memory operand index handling
66868 There's no real need for the pseudo-boolean "haveindex" or for separate
66869 32-bit / 64-bit index pointers. Fold them into a single "indexes" and
66870 set that uniformly to AT&T names, compensating by emitting the register
66871 name via oappend_maybe_intel().
66873 x86: constify disassembler static data
66874 Now that the code is intended to be largely thread-safe, we'd better not
66875 have any writable static objects.
66877 2022-01-17 GDB Administrator <gdbadmin@sourceware.org>
66879 Automatic date update in version.in
66881 2022-01-16 Joel Brobecker <brobecker@adacore.com>
66883 gdb/copyright.py: Do not update gdbsupport/Makefile.in
66884 This file is generated, so we should not modify it (any modification
66885 we make is going to be undone at the next re-generation anyway).
66887 2022-01-16 GDB Administrator <gdbadmin@sourceware.org>
66889 Automatic date update in version.in
66891 2022-01-15 GDB Administrator <gdbadmin@sourceware.org>
66893 Automatic date update in version.in
66895 2022-01-14 Simon Marchi <simon.marchi@efficios.com>
66897 gdb.dlang/demangle.exp: update expected output for _D8demangle4testFnZv
66898 Since commit ce2d3708bc8b ("Synchronize binutils libiberty sources with
66899 gcc version."), I see this failure:
66901 demangle _D8demangle4testFnZv^M
66902 demangle.test(typeof(null))^M
66903 (gdb) FAIL: gdb.dlang/demangle.exp: _D8demangle4testFnZv
66905 The commit imported the commit 0e32a5aa8bc9 ("libiberty: Add support for
66906 D `typeof(*null)' types") from the gcc repository. That commit includes
66907 an update to libiberty/testsuite/d-demangle-expected, which updates a
66908 test for the exact same mangled name:
66910 _D8demangle4testFnZv
66911 -demangle.test(none)
66912 +demangle.test(typeof(null))
66914 I don't know anything about D, but give that the change was made by Iain
66915 Buclaw, the D language maintainer, I trust him on that.
66917 Fix our test by updating the expected output in the same way.
66919 Note: it's not really useful to have all these D demangling tests in the
66920 GDB testsuite, since there are demangling tests in libiberty. We should
66921 consider removing them, but we first need to make sure that everything
66922 that is covered in gdb/testsuite/gdb.dlang/demangle.exp is also covered
66923 in libiberty/testsuite/d-demangle-expected.
66925 Change-Id: If2b290ea8367b8e1e0b90b20d4a6e0bee517952d
66927 2022-01-14 Nils-Christian Kempke <nils-christian.kempke@intel.com>
66929 gdb/testsuite: enable __INTEL_LLVM_COMPILER preprocessor in get_compiler_info
66930 Intel Next Gen compiler defines preprocessor __INTEL_LLVM_COMPILER and provides
66931 version info in __clang_version__ e.g. value: 12.0.0 (icx 2020.10.0.1113).
66933 gdb/testsuite/ChangeLog:
66934 2020-12-07 Abdul Basit Ijaz <abdul.b.ijaz@intel.com>
66936 * lib/compiler.c: Add Intel next gen compiler pre-processor check.
66937 * lib/compiler.cc: Ditto.
66938 * lib/fortran.exp (fortran_main): Check Intel next gen compiler in
66939 test_compiler_info.
66941 2022-01-14 Alan Modra <amodra@gmail.com>
66943 PR28751 mbind2a / mbind2b regressions on powerpc*-linux
66945 * bfdlink.h (struct bfd_link_info): Add commonpagesize_is_set.
66948 * emultempl/elf.em (handle_option): Set commonpagesize_is_set.
66949 * ldelf.c (ldelf_after_parse): Don't error when only one of
66950 -z max-page-size or -z common-page-size is given, correct the
66951 other value to make it sane.
66952 * testsuite/ld-elf/elf.exp (mbind2a, mbind2b): Do not pass
66955 2022-01-14 Jan Beulich <jbeulich@suse.com>
66957 x86: drop ymmxmm_mode
66958 This enumerator is not used by any table entry.
66960 x86: share yet more VEX table entries with EVEX decoding
66961 On top of prior similar work more opportunities have appeared in the
66962 meantime. Note that this also happens to address the prior lack of
66963 decoding of EVEX.L'L for VMOV{L,H}P{S,D} and VMOV{LH,HL}PS.
66965 x86: consistently use scalar_mode for AVX512-FP16 scalar insns
66966 For some reason the original AVFX512F insns were not taken as a basis
66967 here, causing unnecessary divergence. While not an active issue, it is
66968 still relevant to note that OP_XMM() has special treatment of e.g.
66969 scalar_mode (marking broadcast as invalid). Such would better be
66970 consistent for all sufficiently similar insns.
66972 x86: record further wrong uses of EVEX.b
66973 For one EVEX.W set does not imply EVEX.b is uniformly valid. Reject it
66974 for modes which occur for insns allowing for EVEX.W to be set (noticed
66975 with VMOV{H,L}PD and VMOVDDUP, and only in AT&T mode, but not checked
66976 whether further insns would also have been impacted; I expect e.g.
66977 VCMPSD would have had the same issue). And then the present concept of
66978 broadcast makes no sense at all when the memory operand of an insn is
66981 2022-01-14 Jan Beulich <jbeulich@suse.com>
66983 x86: reduce AVX512 FP set of insns decoded through vex_w_table[]
66984 Like for AVX512-FP16, there's not that many FP insns where going through
66985 this table is easier / cheaper than using suitable macros. Utilize %XS
66986 and %XD more to eliminate a fair number of table entries.
66988 While doing this I noticed a few anomalies. Where lines get touched /
66989 moved anyway, these are being addressed right here:
66990 - vmovshdup used EXx for its 2nd operand, thus displaying seemingly
66991 valid broadcast when EVEX.b is set with a memory operand; use
66992 EXEvexXNoBcst instead just like vmovsldup already does
66993 - vmovlhps used EXx for its 3rd operand, when all sibling entries use
66994 EXq; switch to EXq there for consistency (the two differ only for
66997 2022-01-14 Jan Beulich <jbeulich@suse.com>
66999 x86: reduce AVX512-FP16 set of insns decoded through vex_w_table[]
67000 Like already indicated during review of the original submission, there's
67001 really only very few insns where going through this table is easier /
67002 cheaper than using suitable macros. Utilize %XH more and introduce
67003 similar %XS and %XD (which subsequently can be used for further table
67006 While there also switch to using oappend() in 'XH' macro processing.
67008 2022-01-14 GDB Administrator <gdbadmin@sourceware.org>
67010 Automatic date update in version.in
67012 2022-01-13 H.J. Lu <hjl.tools@gmail.com>
67014 ld: Disable DT_RELR in some -z relro tests
67015 Disable DT_RELR in the following -z relro tests which don't expect
67016 DT_RELR in linker outputs.
67018 * testsuite/ld-i386/pr20830.d: Pass $NO_DT_RELR_LDFLAGS to ld.
67019 * testsuite/ld-x86-64/pr20830a-now.d: Likewise.
67020 * testsuite/ld-x86-64/pr20830a.d: Likewise.
67021 * testsuite/ld-x86-64/pr20830b-now.d: Likewise.
67022 * testsuite/ld-x86-64/pr20830b.d: Likewise.
67023 * testsuite/ld-x86-64/pr21038a-now.d: Likewise.
67024 * testsuite/ld-x86-64/pr21038a.d: Likewise.
67025 * testsuite/ld-x86-64/pr21038b-now.d: Likewise.
67026 * testsuite/ld-x86-64/pr21038c-now.d: Likewise.
67027 * testsuite/ld-x86-64/pr21038c.d: Likewise.
67029 2022-01-13 H.J. Lu <hjl.tools@gmail.com>
67031 Reapply libiberty: Pass --plugin to AR and RANLIB
67032 Reapply the patch to detect GCC LTO plugin used for libiberty build to
67033 support LTO build in libiberty.
67035 * Makefile.in (AR): Add @AR_PLUGIN_OPTION@
67036 (RANLIB): Add @RANLIB_PLUGIN_OPTION@.
67037 (configure_deps): Depend on ../config/gcc-plugin.m4.
67038 * aclocal.m4: Include ../config/gcc-plugin.m4.
67039 * configure.ac: AC_SUBST AR_PLUGIN_OPTION and
67040 RANLIB_PLUGIN_OPTION.
67041 * configure: Regenerate.
67043 2022-01-13 H.J. Lu <hjl.tools@gmail.com>
67045 elf: Remove the 1-page gap before the RELRO segment
67046 The existing RELRO scheme may leave a 1-page gap before the RELRO segment
67047 and align the end of the RELRO segment to the page size:
67049 [18] .eh_frame PROGBITS 408fa0 008fa0 005e80 00 A 0 0 8
67050 [19] .init_array INIT_ARRAY 410de0 00fde0 000008 08 WA 0 0 8
67051 [20] .fini_array FINI_ARRAY 410de8 00fde8 000008 08 WA 0 0 8
67052 [21] .dynamic DYNAMIC 410df0 00fdf0 000200 10 WA 7 0 8
67053 [22] .got PROGBITS 410ff0 00fff0 000010 08 WA 0 0 8
67054 [23] .got.plt PROGBITS 411000 010000 000048 08 WA 0 0 8
67056 Instead, we can remove the 1-page gap if the maximum page size >= the
67057 maximum section alignment:
67059 [18] .eh_frame PROGBITS 408fa0 008fa0 005e80 00 A 0 0 8
67060 [19] .init_array INIT_ARRAY 40fde0 00fde0 000008 08 WA 0 0 8
67061 [20] .fini_array FINI_ARRAY 40fde8 00fde8 000008 08 WA 0 0 8
67062 [21] .dynamic DYNAMIC 40fdf0 00fdf0 000200 10 WA 7 0 8
67063 [22] .got PROGBITS 40fff0 00fff0 000010 08 WA 0 0 8
67064 [23] .got.plt PROGBITS 410000 010000 000048 08 WA 0 0 8
67066 Because the end of the RELRO segment is always aligned to the page size
67067 and may not be moved, the RELRO segment size may be increased:
67069 [ 3] .dynstr STRTAB 000148 000148 000001 00 A 0 0 1
67070 [ 4] .eh_frame PROGBITS 000150 000150 000000 00 A 0 0 8
67071 [ 5] .init_array INIT_ARRAY 200150 000150 000010 08 WA 0 0 1
67072 [ 6] .fini_array FINI_ARRAY 200160 000160 000010 08 WA 0 0 1
67073 [ 7] .jcr PROGBITS 200170 000170 000008 00 WA 0 0 1
67074 [ 8] .data.rel.ro PROGBITS 200180 000180 000020 00 WA 0 0 16
67075 [ 9] .dynamic DYNAMIC 2001a0 0001a0 0001c0 10 WA 3 0 8
67076 [10] .got PROGBITS 200360 000360 0002a8 00 WA 0 0 8
67077 [11] .bss NOBITS 201000 000608 000840 00 WA 0 0 1
67079 vs the old section layout:
67081 [ 3] .dynstr STRTAB 000148 000148 000001 00 A 0 0 1
67082 [ 4] .eh_frame PROGBITS 000150 000150 000000 00 A 0 0 8
67083 [ 5] .init_array INIT_ARRAY 200b48 000b48 000010 08 WA 0 0 1
67084 [ 6] .fini_array FINI_ARRAY 200b58 000b58 000010 08 WA 0 0 1
67085 [ 7] .jcr PROGBITS 200b68 000b68 000008 00 WA 0 0 1
67086 [ 8] .data.rel.ro PROGBITS 200b70 000b70 000020 00 WA 0 0 16
67087 [ 9] .dynamic DYNAMIC 200b90 000b90 0001c0 10 WA 3 0 8
67088 [10] .got PROGBITS 200d50 000d50 0002a8 00 WA 0 0 8
67089 [11] .bss NOBITS 201000 000ff8 000840 00 WA 0 0 1
67091 But there is no 1-page gap.
67094 * ldlang.c (lang_size_relro_segment_1): Remove the 1-page gap
67095 before the RELRO segment if the maximum page size >= the maximum
67097 * testsuite/ld-i386/pr20830.d: Adjusted.
67098 * testsuite/ld-s390/gotreloc_64-relro-1.dd: Likewise.
67099 * testsuite/ld-x86-64/pr14207.d: Likewise.
67100 * testsuite/ld-x86-64/pr18176.d: Likewise.
67101 * testsuite/ld-x86-64/pr20830a-now.d: Likewise.
67102 * testsuite/ld-x86-64/pr20830a.d: Likewise.
67103 * testsuite/ld-x86-64/pr20830b-now.d: Likewise.
67104 * testsuite/ld-x86-64/pr20830b.d: Likewise.
67105 * testsuite/ld-x86-64/pr21038a-now.d: Likewise.
67106 * testsuite/ld-x86-64/pr21038a.d: Likewise.
67107 * testsuite/ld-x86-64/pr21038b-now.d: Likewise.
67108 * testsuite/ld-x86-64/pr21038c-now.d: Likewise.
67109 * testsuite/ld-x86-64/pr21038c.d: Likewise.
67111 2022-01-13 Nick Clifton <nickc@redhat.com>
67113 Synchronize binutils libiberty sources with gcc version.
67114 +2021-12-30 Lancelot SIX <lsix@lancelotsix.com>
67116 + * cp-demangle.c (d_clone_suffix): Support digits in clone tag
67118 + * testsuite/demangle-expected: Check demangling of clone symbols
67119 + with digits in name.
67121 +2021-12-16 H.J. Lu <hjl.tools@gmail.com>
67124 + 2021-12-16 H.J. Lu <hjl.tools@gmail.com>
67126 + * Makefile.in (AR): Add @AR_PLUGIN_OPTION@
67127 + (RANLIB): Add @RANLIB_PLUGIN_OPTION@.
67128 + (configure_deps): Depend on ../config/gcc-plugin.m4.
67129 + * configure.ac: AC_SUBST AR_PLUGIN_OPTION and
67130 + RANLIB_PLUGIN_OPTION.
67131 + * aclocal.m4: Regenerated.
67132 + * configure: Likewise.
67134 +2021-12-15 H.J. Lu <hjl.tools@gmail.com>
67136 + * Makefile.in (AR): Add @AR_PLUGIN_OPTION@
67137 + (RANLIB): Add @RANLIB_PLUGIN_OPTION@.
67138 + (configure_deps): Depend on ../config/gcc-plugin.m4.
67139 + * configure.ac: AC_SUBST AR_PLUGIN_OPTION and
67140 + RANLIB_PLUGIN_OPTION.
67141 + * aclocal.m4: Regenerated.
67142 + * configure: Likewise.
67144 +2021-11-29 Eric Gallager <egallager@gcc.gnu.org>
67147 + * Makefile.in: Use ETAGS variable in TAGS target.
67148 + * configure: Regenerate.
67149 + * configure.ac: Allow ETAGS variable to be overridden.
67151 +2021-11-29 Andrew Pinski <apinski@marvell.com>
67153 + * make-temp-file.c (try_dir): Check to see if the dir
67154 + is actually a directory.
67156 +2021-10-22 Eric Gallager <egallager@gcc.gnu.org>
67159 + * Makefile.in: Allow dvi-formatted documentation
67162 +2021-10-17 Lu?s Ferreira <contact@lsferreira.net>
67165 + * d-demangle.c (dlang_parse_qualified): Handle anonymous
67166 + symbols correctly.
67167 + * testsuite/d-demangle-expected: New tests to cover anonymous
67170 +2021-10-14 Lu?s Ferreira <contact@lsferreira.net>
67172 + * testsuite/d-demangle-expected: Add test case for function literals.
67174 +2021-10-14 Lu?s Ferreira <contact@lsferreira.net>
67176 + * testsuite/d-demangle-expected: Add test cases for simple special
67179 +2021-10-12 Lu?s Ferreira <contact@lsferreira.net>
67181 + * d-demangle.c (dlang_parse_qualified): Remove redudant parenthesis
67182 + around lhs and rhs of assignments.
67184 +2021-10-01 Lu?s Ferreira <contact@lsferreira.net>
67186 + * testsuite/d-demangle-expected: Add missing format for new test
67188 +2021-09-23 Lu?s Ferreira <contact@lsferreira.net>
67190 + * d-demangle.c (dlang_Type): Validate MANGLED is nonnull.
67191 + * testsuite/d-demangle-expected: New test.
67193 +2021-09-23 Lu?s Ferreira <contact@lsferreira.net>
67195 + * d-demangle.c (dlang_symbol_backref): Ensure strlen of
67196 + string is less than length computed by dlang_number.
67198 +2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
67200 * configure: Regenerate.
67201 + * configure.ac: Do not search for sbrk on Darwin.
67202 + * xmalloc.c: Do not declare sbrk unless it has been found
67205 +2021-08-29 Iain Buclaw <ibuclaw@gdcproject.org>
67207 + * d-demangle.c (dlang_identifier): Skip over fake parent manglings.
67208 + * testsuite/d-demangle-expected: Add tests.
67210 +2021-08-29 Iain Buclaw <ibuclaw@gdcproject.org>
67212 + * d-demangle.c (dlang_parse_arrayliteral): Add 'info' parameter.
67213 + (dlang_parse_assocarray): Likewise.
67214 + (dlang_parse_structlit): Likewise.
67215 + (dlang_value): Likewise. Handle function literal symbols.
67216 + (dlang_template_args): Pass 'info' to dlang_value.
67217 + * testsuite/d-demangle-expected: Add new test.
67219 +2021-08-29 Iain Buclaw <ibuclaw@gdcproject.org>
67221 + * d-demangle.c (dlang_attributes): Handle typeof(*null).
67222 + (dlang_type): Likewise. Demangle 'n' as typeof(null).
67223 + * testsuite/d-demangle-expected: Update tests.
67225 +2021-08-23 Iain Sandoe <iain@sandoe.co.uk>
67227 + * simple-object-mach-o.c (simple_object_mach_o_write_segment):
67228 + Cast the first argument to set_32 as needed.
67230 -2021-07-03 Nick Clifton <nickc@redhat.com>
67231 +2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
67233 + * simple-object-mach-o.c (simple_object_mach_o_write_segment):
67234 + Arrange to swap the LTO index tables where needed.
67235 # Please enter the commit message for your changes. Lines starting
67237 2022-01-13 Andrew Burgess <aburgess@redhat.com>
67239 gdb: don't use -Wmissing-prototypes with g++
67240 This commit aims to not make use of -Wmissing-prototypes when
67241 compiling with g++.
67243 Use of -Wmissing-prototypes was added with this commit:
67245 commit a0761e34f054767de6d6389929d27e9015fb299b
67246 Date: Wed Mar 11 15:15:12 2020 -0400
67248 gdb: enable -Wmissing-prototypes warning
67250 Because clang can provide helpful warnings with this flag.
67251 Unfortunately, g++ doesn't accept this flag, and will give this
67254 cc1plus: warning: command line option ‘-Wmissing-prototypes’ is valid for C/ObjC but not for C++
67256 In theory the fact that this flag is not supported should be detected
67257 by the configure check in gdbsupport/warning.m4, but for users of
67258 ccache, this check doesn't work due to a long standing ccache issue:
67260 https://github.com/ccache/ccache/issues/738
67262 The ccache problem is that -W... options are reordered on the command
67263 line, and so -Wmissing-prototypes is seen before -Werror. Usually
67264 this doesn't matter, but the above warning (about the flag not being
67265 valid) is issued before the -Werror flag is processed, and so is not
67268 There have been two previous attempts to fix this that I'm aware of.
67271 https://sourceware.org/pipermail/gdb-patches/2021-September/182148.html
67273 In this attempt, instead of just relying on a compile to check if a
67274 flag is valid, the proposal was to both compile and link. As linking
67275 doesn't go through ccache, we don't suffer from the argument
67276 reordering problem, and the link phase will correctly fail when using
67277 -Wmissing-prototypes with g++. The configure script will then disable
67278 the use of this flag.
67280 This approach was rejected, and the suggestion was to only add the
67281 -Wmissing-prototypes flag if we are compiling with gcc.
67283 The second attempt, attempts this approach, and can be found here:
67285 https://sourceware.org/pipermail/gdb-patches/2021-November/183076.html
67287 This attempt only adds the -Wmissing-prototypes flag is the value of
67288 GCC is not 'yes'. This feels like it is doing the right thing,
67289 unfortunately, the GCC flag is really a 'is gcc like' flag, not a
67290 strict, is gcc check. As such, GCC is set to 'yes' for clang, which
67291 would mean the flag was not included for clang or gcc. The entire
67292 point of the original commit was to add this flag for clang, so
67293 clearly the second attempt is not sufficient either.
67295 In this new attempt I have added gdbsupport/compiler-type.m4, this
67296 file defines AM_GDB_COMPILER_TYPE. This macro sets the variable
67297 GDB_COMPILER_TYPE to either 'gcc', 'clang', or 'unknown'. In future
67298 the list of values might be extended to cover other compilers, if this
67301 I've then modified gdbsupport/warning.m4 to only add the problematic
67302 -Wmissing-prototypes flag if GDB_COMPILER_TYPE is not 'gcc'.
67304 I've tested this with both gcc and clang and see the expected results,
67305 gcc no longer attempts to use the -Wmissing-prototypes flag, while
67306 clang continues to use it.
67308 When compiling using ccache, I am no longer seeing the warning.
67310 2022-01-13 Andrew Burgess <aburgess@redhat.com>
67312 gdb: add some extra debug information to attach_command
67313 While working on another patch I wanted to add some extra debug
67314 information to the attach_command function. This required me to add a
67315 new function to convert the thread_info::state variable to a string.
67317 The new debug might be useful to others, and the state to string
67318 function might be useful in other locations, so I thought I'd merge
67321 2022-01-13 Alan Modra <amodra@gmail.com>
67323 Re: gas: add visibility support using GNU syntax on XCOFF
67324 tc-ppc.c: In function 'ppc_comm':
67325 tc-ppc.c:4560:40: error: 'visibility' may be used uninitialized in this function [-Werror=maybe-uninitialized]
67327 With that fixed we hit lots of segfaults in the ld testsuite.
67331 * xcofflink.c (xcoff_link_input_bfd): Don't segfault on NULL
67334 * config/tc-ppc.c (ppc_comm): Init visibility.
67336 2022-01-13 Alan Modra <amodra@gmail.com>
67338 dt-relr.exp --no-as-needed
67339 Otherwise the very simple test may not be linked with libc.so at all,
67340 and thus correctly have no version reference added. Causing failure
67341 of the dt-relr-glibc-1b.so test.
67343 * testsuite/ld-elf/dt-relr.exp: Link with --no-as-needed.
67345 2022-01-13 Alan Modra <amodra@gmail.com>
67347 Correct .relr.dyn nocombreloc script
67348 * scripttempl/elf.sc (.relr.dyn): Don't depend on $COMBRELOC.
67350 2022-01-13 Alan Modra <amodra@gmail.com>
67352 testsuite supports_dt_relr
67353 Tidy, and fix "FAIL: Build dt-relr-glibc-1b.so" on all non-x86
67357 * binutils-common.exp (supports_dt_relr): New proc.
67359 * testsuite/config/default.exp (DT_RELR_LDFLAGS, NO_DT_RELR_LDFLAGS),
67360 (DT_RELR_CC_LDFLAGS, NO_DT_RELR_CC_LDFLAGS): Use supports_dt_relr.
67361 * testsuite/ld-elf/dt-relr.exp: Don't run unless supports_dt_relr.
67362 * testsuite/ld-elf/dt-relr-1a.d: Likewise.
67363 * testsuite/ld-elf/dt-relr-1b.d: Likewise.
67364 * testsuite/ld-elf/dt-relr-1c.d: Likewise.
67365 * testsuite/ld-elf/dt-relr-2a.d: Likewise.
67366 * testsuite/ld-elf/dt-relr-2b.d: Likewise.
67367 * testsuite/ld-elf/dt-relr-2c.d: Likewise.
67368 * testsuite/ld-elf/dt-relr-2d.d: Likewise.
67369 * testsuite/ld-elf/dt-relr-2e.d: Likewise.
67370 * testsuite/ld-elf/dt-relr-2f.d: Likewise.
67371 * testsuite/ld-elf/dt-relr-2g.d: Likewise.
67372 * testsuite/ld-elf/dt-relr-2h.d: Likewise.
67373 * testsuite/ld-elf/dt-relr-3a.d: Likewise.
67374 * testsuite/ld-elf/dt-relr-3b.d: Likewise.
67376 2022-01-13 Alan Modra <amodra@gmail.com>
67378 Don't use C++ comments in assembly
67379 It might seem to work, but only if '/' is a start of comment char.
67381 * testsuite/ld-elf/dt-relr-1.s: Use # for comment.
67382 * testsuite/ld-elf/dt-relr-2.s: Likewise.
67383 * testsuite/ld-elf/dt-relr-3.s: Likewise.
67385 2022-01-13 Alan Modra <amodra@gmail.com>
67387 Move DT_RELR tag setting to elflink.c
67388 This makes the code setting DT_RELR tags generally available. Many
67389 targets will be able to use the defaults. Those that can't should set
67390 up sh_entsize for .relr.dyn output section before reaching the dynamic
67391 tag code in bfd_elf_final_link.
67393 * elflink.c (bfd_elf_final_link): Set up DT_RELR tags and sh_entsize.
67394 * elfxx-x86.c (_bfd_x86_elf_finish_dynamic_sections): Don't do any
67397 2022-01-13 Alan Modra <amodra@gmail.com>
67399 Re: Set SEC_ELF_REVERSE_COPY earlier
67400 Let's not rely on .init/.fini having relocs for the size sanity check.
67401 This is mainly to squash reports of "my fuzzed object made ld hang".
67403 2022-01-13 Tiezhu Yang <yangtiezhu@loongson.cn>
67405 gdb: testsuite: make string[] type as char in gdb.base/charset.c
67406 This reverts the commit ff656e2e1cb1 ("gdb: testsuite: fix failed
67407 testcases in gdb.base/charset.exp").
67409 The original test code has no problem. On an architecture where
67410 char is signed, then both 'A' and ebcdic_us_string[7] will yield
67411 -63, which makes the equality true. On an architecture where char
67412 is unsigned, then both 'A' and ebcdic_us_string[7] will yield 193,
67413 which also makes the equality true.
67415 The test cases only failed on LoongArch. The default type of char
67416 is signed char on LoongArch, like x86-64. But when use gdb print
67417 command on LoongArch, the default type of char is unsigned char,
67418 this is wrong, I will look into it later, sorry for that.
67428 unsigned char c2 = 193;
67430 printf("%d\n", c1);
67431 printf("%d\n", c1 == c2);
67435 $ gcc test_char.c -o test_char
67440 (gdb) set target-charset EBCDIC-US
67452 2022-01-13 GDB Administrator <gdbadmin@sourceware.org>
67454 Automatic date update in version.in
67456 2022-01-12 Carl Love <cel@us.ibm.com>
67458 gdb Power 9 add test for HW watchpoint support.
67459 The Power 9 processor revision 2.2 has HW watchpoint support disabled due
67460 to a HW bug. The support is fixed in Power 9 processor revision 2.3. This
67461 patch add a test to lib/gdb.exp for Power to determine if the processor
67462 supports HW watchpoints or not. If the Power processor doesn't support HW
67463 watchpoints the proceedure skip_hw_watchpoint_tests will return 1 to
67464 disable the various HW watchpoint tests.
67466 The patch has been tested on Power 9, processor revesions 2.2 and 2.3. The
67467 patch has also been tested on Power 10. No regression test failures were
67470 2022-01-12 Andrew Burgess <aburgess@redhat.com>
67472 gdb/python: add gdb.host_charset function
67473 We already have gdb.target_charset and gdb.target_wide_charset. This
67474 commit adds gdb.host_charset along the same lines.
67476 2022-01-12 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
67478 gdb/testsuite: fix gdb.python/py-events.exp for finding process id
67479 When executed with --target_board=native-extended-gdbserver, the
67480 gdb.python/py-events.exp test errors out with
67482 ERROR: tcl error sourcing /path/to/gdb/testsuite/gdb.python/py-events.exp.
67483 ERROR: can't read "process_id": no such variable
67485 "lappend expected "ptid: \\($process_id, $process_id, 0\\)" "address: $addr""
67486 (file "/path/to/gdb/testsuite/gdb.python/py-events.exp" line 103)
67487 invoked from within
67488 "source /path/to/gdb/testsuite/gdb.python/py-events.exp"
67489 ("uplevel" body line 1)
67490 invoked from within
67491 "uplevel #0 source /path/to/gdb/testsuite/gdb.python/py-events.exp"
67492 invoked from within
67493 "catch "uplevel #0 source $test_file_name""
67495 There are multiple problems around this:
67497 1. The process_id variable is not initialized to a default value.
67499 2. The test attempts to find the PID of the current thread, but the
67500 regexp that it uses is not tailored for the output printed by the
67503 3. The test uses "info threads" to find the current thread PID.
67504 Using the "thread" command instead is simpler.
67506 Fix these problems.
67508 2022-01-12 Tom Tromey <tromey@adacore.com>
67510 Don't mention "serial" in target remote description
67511 PR remote/9177 points out that "info files" mentions "serial" a couple
67514 Remote serial target in gdb-specific protocol:
67515 Debugging a target over a serial line.
67517 However, often the remote target isn't really a serial connection.
67519 It seems to me that this text could be a bit clearer; and furthermore
67520 since "info files" prints the target's long description,
67521 remote_target::files_info doesn't really add much and can simply be
67524 Regression tested on x86-64 Fedora 34.
67526 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=9177
67528 2022-01-12 H.J. Lu <hjl.tools@gmail.com>
67530 ld: Add glibc dependency for DT_RELR
67531 When DT_RELR is enabled, to avoid random run-time crash with older glibc
67532 binaries without DT_RELR support, add a GLIBC_ABI_DT_RELR symbol version,
67533 which is provided by glibc with DT_RELR support, dependency on the shared
67534 C library if it provides a GLIBC_2.XX symbol version.
67538 * elflink.c (elf_link_add_dt_relr_dependency): New function.
67539 (bfd_elf_size_dynamic_sections): Call
67540 elf_link_add_dt_relr_dependency if DT_RELR is enabled.
67544 * ld.texi: Mention GLIBC_ABI_DT_RELR in -z pack-relative-relocs
67546 * testsuite/ld-elf/dt-relr-glibc-1.c: New file.
67547 * testsuite/ld-elf/dt-relr-glibc-1a.rd: Likewise.
67548 * testsuite/ld-elf/dt-relr-glibc-1b.rd: Likewise.
67549 * testsuite/ld-elf/dt-relr.exp: Likewise.
67551 2022-01-12 H.J. Lu <hjl.tools@gmail.com>
67553 ld: Add simple DT_RELR tests
67554 * testsuite/ld-elf/dt-relr-1.s: New file.
67555 * testsuite/ld-elf/dt-relr-1a.d: Likewise.
67556 * testsuite/ld-elf/dt-relr-1b.d: Likewise.
67557 * testsuite/ld-elf/dt-relr-1c.d: Likewise.
67558 * testsuite/ld-elf/dt-relr-2.s: Likewise.
67559 * testsuite/ld-elf/dt-relr-2a.d: Likewise.
67560 * testsuite/ld-elf/dt-relr-2b.d: Likewise.
67561 * testsuite/ld-elf/dt-relr-2c.d: Likewise.
67562 * testsuite/ld-elf/dt-relr-2d.d: Likewise.
67563 * testsuite/ld-elf/dt-relr-2e.d: Likewise.
67564 * testsuite/ld-elf/dt-relr-2f.d: Likewise.
67565 * testsuite/ld-elf/dt-relr-2g.d: Likewise.
67566 * testsuite/ld-elf/dt-relr-2h.d: Likewise.
67567 * testsuite/ld-elf/dt-relr-3.s: Likewise.
67568 * testsuite/ld-elf/dt-relr-3a.d: Likewise.
67569 * testsuite/ld-elf/dt-relr-3b.d: Likewise.
67570 * testsuite/ld-i386/dt-relr-1.s: Likewise.
67571 * testsuite/ld-i386/dt-relr-1a.d: Likewise.
67572 * testsuite/ld-i386/dt-relr-1b.d: Likewise.
67573 * testsuite/ld-x86-64/dt-relr-1a-x32.d: Likewise.
67574 * testsuite/ld-x86-64/dt-relr-1a.d: Likewise.
67575 * testsuite/ld-x86-64/dt-relr-1b-x32.d: Likewise.
67576 * testsuite/ld-x86-64/dt-relr-1b.d: Likewise.
67577 * testsuite/ld-x86-64/dt-relr-1.s: Likewise.
67578 * testsuite/ld-i386/i386.exp: Run dt-relr-1a and dt-relr-1b.
67579 * testsuite/ld-x86-64/x86-64.exp: Run dt-relr-1a, dt-relr-1a-x32
67580 dt-relr-1b and dt-relr-1b-x32.
67582 2022-01-12 H.J. Lu <hjl.tools@gmail.com>
67584 x86: Add DT_RELR support
67585 DT_RELR is implemented with linker relaxation:
67587 1. During linker relaxation, we scan input relocations with the same
67588 logic in relocate_section to determine if a relative relocation should
67589 be generated and save the relative relocation candidate information for
67590 sizing the DT_RELR section later after all symbols addresses can be
67591 determined. For these relative relocations which can't be placed in
67592 the DT_RELR section, they will be placed in the rela.dyn/rel.dyn
67594 2. When DT_RELR is enabled, _bfd_elf_map_sections_to_segments calls a
67595 backend function to size the DT_RELR section which will compute the
67596 DT_RELR section size and tell ldelf_map_segments to layout sections
67597 again when the DT_RELR section size has been increased.
67598 3. After regular symbol processing is finished, bfd_elf_final_link calls
67599 a backend function to finish the DT_RELR section.
67601 * elf32-i386.c (elf_i386_relocate_section): Don't generate
67602 relative relocation when DT_RELR is enabled.
67603 (elf_i386_finish_dynamic_symbol): Likewise.
67604 * elf64-x86-64.c (elf_x86_64_relocate_section): Don't generate
67605 relative relocation when DT_RELR is enabled.
67606 (elf_x86_64_finish_dynamic_symbol): Likewise.
67607 * elfxx-x86.c (_bfd_x86_elf_link_hash_table_create): Initialize
67608 relative_r_type, relative_r_name, elf_append_reloc,
67609 elf_write_addend and elf_write_addend_in_got.
67610 (elf_x86_relative_reloc_record_add): New function.
67611 (_bfd_x86_elf_link_relax_section): Likewise.
67612 (elf64_dt_relr_bitmap_add): Likewise.
67613 (elf32_dt_relr_bitmap_add): Likewise.
67614 (_bfd_elf32_write_addend): Likewise.
67615 (_bfd_elf64_write_addend): Likewise.
67616 (elf_x86_size_or_finish_relative_reloc): Likewise.
67617 (elf_x86_compute_dl_relr_bitmap): Likewise.
67618 (elf_x86_write_dl_relr_bitmap): Likewise.
67619 (elf_x86_relative_reloc_compare ): Likewise.
67620 (_bfd_elf_x86_size_relative_relocs): Likewise.
67621 (_bfd_elf_x86_finish_relative_relocs): Likewise.
67622 (_bfd_x86_elf_size_dynamic_sections): Skip the .relr.dyn section.
67623 (_bfd_x86_elf_finish_dynamic_sections): Convert 3 spare dynamic
67624 tags to DT_RELR, DT_RELRSZ and for compact relative relocation.
67625 * elfxx-x86.h (X86_64_GOT_TYPE_P): New.
67626 (I386_GOT_TYPE_P): Likewise.
67627 (X86_GOT_TYPE_P): Likewise.
67628 (X86_64_RELATIVE_RELOC_TYPE_P): Likewise.
67629 (I386_RELATIVE_RELOC_TYPE_P): Likewise.
67630 (X86_RELATIVE_RELOC_TYPE_P): Likewise.
67631 (X86_LOCAL_GOT_RELATIVE_RELOC_P): Likewise.
67632 (I386_PCREL_TYPE_P): Likewise.
67633 (X86_64_PCREL_TYPE_P): Likewise.
67634 (X86_64_NEED_DYNAMIC_RELOC_TYPE_P): Rewrite.
67635 (I386_NEED_DYNAMIC_RELOC_TYPE_P): Likewise.
67636 (GENERATE_DYNAMIC_RELOCATION_P): Also check rel_from_abs.
67637 (elf_x86_link_hash_entry): Add got_relative_reloc_done.
67638 (elf_x86_relative_reloc_record): New.
67639 (elf_x86_relative_reloc_data): Likewise.
67640 (elf_dt_relr_bitmap): Likewise.
67641 (elf_x86_link_hash_table): Add dt_relr_bitmap, relative_reloc,
67642 unaligned_relative_reloc, relative_r_type, relative_r_name,
67643 elf_append_reloc, elf_write_addend, elf_write_addend_in_got and
67644 relative_reloc_done.
67645 (elf_x86_relative_reloc_done): New.
67646 (relative_reloc_packed): Likewise.
67647 (_bfd_x86_elf_link_relax_section): Likewise.
67648 (_bfd_elf_x86_size_relative_relocs): Likewise.
67649 (_bfd_elf_x86_finish_relative_relocs): Likewise.
67650 (_bfd_elf32_write_addend): Likewise.
67651 (_bfd_elf64_write_addend): Likewise.
67652 (bfd_elf32_bfd_relax_section): Likewise.
67653 (bfd_elf64_bfd_relax_section): Likewise.
67654 (elf_backend_size_relative_relocs): Likewise.
67655 (elf_backend_finish_relative_relocs): Likewise.
67656 (elf_x86_allocate_local_got_info): Also allocate
67657 relative_reloc_done.
67659 2022-01-12 H.J. Lu <hjl.tools@gmail.com>
67661 elf: Support DT_RELR in linker tests
67662 Allow eabling and disabling DT_RELR in linker tests. Disable DT_RELR in
67663 linker tests which don't expect DT_RELR in linker outputs.
67667 * testsuite/lib/binutils-common.exp (run_dump_test): Make
67668 DT_RELR_LDFLAGS and NO_DT_RELR_LDFLAGS global.
67672 * testsuite/config/default.exp (DT_RELR_LDFLAGS): New.
67673 (DT_RELR_CC_LDFLAGS): Likewise.
67674 (NO_DT_RELR_LDFLAGS): Likewise.
67675 (NO_DT_RELR_CC_LDFLAGS): Likewise.
67676 * testsuite/ld-elf/shared.exp: Pass $NO_DT_RELR_LDFLAGS to
67677 linker for some tests.
67678 * testsuite/ld-i386/export-class.exp: Likewise.
67679 * testsuite/ld-i386/i386.exp: Likewise.
67680 * testsuite/ld-i386/ibt-plt-2a.d: Pass $NO_DT_RELR_LDFLAGS to
67682 * testsuite/ld-i386/ibt-plt-3a.d: Likewise.
67683 * testsuite/ld-i386/ibt-plt-3c.d: Likewise.
67684 * testsuite/ld-i386/pr26869.d: Likewise.
67685 * testsuite/ld-i386/report-reloc-1.d: Likewise.
67686 * testsuite/ld-ifunc/ifunc-2-i386-now.d: Likewise.
67687 * testsuite/ld-ifunc/ifunc-2-local-i386-now.d: Likewise.
67688 * testsuite/ld-ifunc/ifunc-2-local-x86-64-now.d: Likewise.
67689 * testsuite/ld-ifunc/ifunc-2-x86-64-now.d: Likewise.
67690 * testsuite/ld-ifunc/pr17154-x86-64.d: Likewise.
67691 * testsuite/ld-x86-64/bnd-branch-1-now.d: Likewise.
67692 * testsuite/ld-x86-64/bnd-ifunc-1-now.d: Likewise.
67693 * testsuite/ld-x86-64/bnd-ifunc-2-now.d: Likewise.
67694 * testsuite/ld-x86-64/bnd-ifunc-2.d: Likewise.
67695 * testsuite/ld-x86-64/bnd-plt-1-now.d: Likewise.
67696 * testsuite/ld-x86-64/bnd-plt-1.d: Likewise.
67697 * testsuite/ld-x86-64/ibt-plt-2a-x32.d: Likewise.
67698 * testsuite/ld-x86-64/ibt-plt-2a.d: Likewise.
67699 * testsuite/ld-x86-64/ibt-plt-3a-x32.d: Likewise.
67700 * testsuite/ld-x86-64/ibt-plt-3a.d: Likewise.
67701 * testsuite/ld-x86-64/ilp32-4.d: Likewise.
67702 * testsuite/ld-x86-64/load1c.d: Likewise.
67703 * testsuite/ld-x86-64/load1d.d: Likewise.
67704 * testsuite/ld-x86-64/pr13082-2b.d: Likewise.
67705 * testsuite/ld-x86-64/pr14207.d: Likewise.
67706 * testsuite/ld-x86-64/pr18176.d: Likewise.
67707 * testsuite/ld-x86-64/pr19162.d: Likewise.
67708 * testsuite/ld-x86-64/pr19636-2d.d: Likewise.
67709 * testsuite/ld-x86-64/pr19636-2l.d: Likewise.
67710 * testsuite/ld-x86-64/pr20253-1d.d: Likewise.
67711 * testsuite/ld-x86-64/pr20253-1f.d: Likewise.
67712 * testsuite/ld-x86-64/pr20253-1j.d: Likewise.
67713 * testsuite/ld-x86-64/pr20253-1l.d: Likewise.
67714 * testsuite/ld-x86-64/report-reloc-1-x32.d: Likewise.
67715 * testsuite/ld-x86-64/report-reloc-1.d: Likewise.
67716 * testsuite/ld-x86-64/export-class.exp (x86_64_export_class_test):
67717 Pass $NO_DT_RELR_LDFLAGS to linker.
67718 * testsuite/ld-x86-64/x86-64.exp: Pass $NO_DT_RELR_LDFLAGS to
67719 linker for some tests.
67721 2022-01-12 H.J. Lu <hjl.tools@gmail.com>
67723 elf: Add size_relative_relocs and finish_relative_relocs
67724 On some targets, the DT_RELR section size can be computed only after all
67725 symbols addresses can be determined. Set the preliminary DT_RELR section
67726 size before mapping sections to segments and set the final DT_RELR section
67727 size after regular symbol processing is done.
67729 * elf-bfd.h (elf_backend_data): Add size_relative_relocs and
67730 finish_relative_relocs.
67731 * elf.c (_bfd_elf_map_sections_to_segments): Call
67732 size_relative_relocs if DT_RELR is enabled.
67733 * elflink.c (bfd_elf_final_link): Call finish_relative_relocs
67734 after regular symbol processing is finished if DT_RELR is enabled.
67735 * elfxx-target.h (elf_backend_size_relative_relocs): New.
67736 (elf_backend_finish_relative_relocs): Likewise.
67737 (elfNN_bed): Add elf_backend_size_relative_relocs and
67738 elf_backend_finish_relative_relocs.
67740 2022-01-12 H.J. Lu <hjl.tools@gmail.com>
67742 ld: Initial DT_RELR support
67743 Add a -z pack-relative-relocs option to enable DT_RELR and create a
67744 relr.dyn section for DT_RELR. DT_RELR is implemented with the linker
67745 relaxation infrastructure, but it doesn't require the --relax option
67746 enabled. -z pack-relative-relocs implies -z combreloc. -z nocombreloc
67747 implies -z nopack-relative-relocs.
67749 -z pack-relative-relocs is chosen over the similar option in lld,
67750 --pack-dyn-relocs=relr, to implement a glibc binary lockout mechanism
67751 with a special glibc version symbol, to avoid random crashes of DT_RELR
67752 binaries with the existing glibc binaries.
67756 * elf-bfd.h (elf_link_hash_table): Add srelrdyn.
67757 * elflink.c (_bfd_elf_link_create_dynamic_sections): Create a
67758 .relr.dyn section for DT_RELR.
67762 * bfdlink.h (bfd_link_info): Add enable_dt_relr.
67766 * News: Mention -z pack-relative-relocs and
67767 -z nopack-relative-relocs.
67768 * ld.texi: Document -z pack-relative-relocs and
67769 -z nopack-relative-relocs.
67770 * ldelf.c (ldelf_after_parse): Disable DT_RELR if not building
67771 PIE nor shared library. Add 3 spare dynamic tags for DT_RELR,
67772 DT_RELRSZ and DT_RELRENT.
67773 * ldlang.c (lang_relax_sections): Also enable relaxation if
67774 DT_RELR is enabled.
67775 * emulparams/elf32_x86_64.sh: Source dt-relr.sh.
67776 * emulparams/elf_i386.sh: Likewise.
67777 * emulparams/elf_x86_64.sh: Likewise.
67778 * emulparams/dt-relr.sh: New file.
67779 * scripttempl/elf.sc: Support .relr.dyn.
67781 2022-01-12 H.J. Lu <hjl.tools@gmail.com>
67783 elf: Pass need_layout to _bfd_elf_map_sections_to_segments
67784 On some targets, the DT_RELR section size can be computed only after all
67785 symbols addresses can be determined. Update ldelf_map_segments to pass
67786 need_layout to _bfd_elf_map_sections_to_segments which will size DT_RELR
67787 section and set need_layout to true if the DT_RELR section size is changed.
67791 * elf-bfd.h (_bfd_elf_map_sections_to_segments): Add a bool
67793 * elf.c (_bfd_elf_map_sections_to_segments): Add a bool pointer
67794 argument to indicate if section layout needs update.
67795 (assign_file_positions_for_load_sections): Pass NULL to
67796 _bfd_elf_map_sections_to_segments.
67797 * elflink.c (_bfd_elf_strip_zero_sized_dynamic_sections): Pass
67798 NULL to _bfd_elf_map_sections_to_segments.
67802 * ldelfgen.c (ldelf_map_segments): Pass &need_layout to
67803 _bfd_elf_map_sections_to_segments.
67805 2022-01-12 H.J. Lu <hjl.tools@gmail.com>
67807 elf: Add .relr.dyn to special_sections_r
67808 * elf.c (special_sections_r): Add .relr.dyn.
67810 2022-01-12 Andrew Burgess <aburgess@redhat.com>
67812 gdb: add 'maint set/show gnu-source-highlight enabled' command
67813 In a later commit I want to address an issue with the Python pygments
67814 based code styling solution. As this approach is only used when the
67815 GNU Source Highlight library is not available, testing bugs in this
67816 area can be annoying, as it requires GDB to be rebuilt with use of GNU
67817 Source Highlight disabled.
67819 This commit adds a pair of new maintenance commands:
67821 maintenance set gnu-source-highlight enabled on|off
67822 maintenance show gnu-source-highlight enabled
67824 these commands can be used to disable use of the GNU Source Highlight
67825 library, allowing me, in a later commit, to easily test bugs that
67826 would otherwise be masked by GNU Source Highlight being used.
67828 I made this a maintenance command, rather than a general purpose
67829 command, as it didn't seem like this was something a general user
67830 would need to adjust. We can always convert the maintenance command
67831 to a general command later if needed.
67833 There's no test for this here, but this feature will be used in a
67836 2022-01-12 Andrew Burgess <aburgess@redhat.com>
67838 gdb: erase items from the source_cache::m_offset_cache
67839 The source_cache class has two member variables m_source_map, which
67840 stores the file contents, and m_offset_cache, which stores offsets
67841 into the file contents.
67843 As source files are read the contents of the file, as well as the
67844 offset data, are stored in the cache using these two member variables.
67846 Whenever GDB needs either the files contents, or the offset data,
67847 source_cache::ensure is called. This function looks for the file in
67848 m_source_map, and if it's found then this implies the file is also in
67849 m_offset_cache, and we're done.
67851 If the file is not in m_source_map then GDB calls
67852 source_cache::get_plain_source_lines to open the file and read its
67853 contents. ::get_plain_source_lines also calculates the offset data,
67854 which is then inserted into m_offset_cache.
67856 Back in ::ensure, the file contents are added into m_source_map. And
67857 finally, if m_source_map contains more than MAX_ENTRIES, an entry is
67858 removed from m_source_map.
67860 The problem is entries are not removed from m_offset_cache at the same
67863 This means that if a program contains enough source files, GDB will
67864 hold at most MAX_ENTRIES cached source file contents, but can contain
67865 offsets data for every source file.
67867 Now, the offsets data is going to be smaller than the cached file
67868 contents, so maybe there's no harm here. But, when we reload the file
67869 contents we always recalculate the offsets data. And, when we
67870 ::get_line_charpos asking for offset data we still call ::ensure which
67871 will ends up loading and caching the file contents.
67873 So, given the current code does the work of reloading the offset data
67874 anyway, we may as well save memory by capping m_offset_cache to
67875 MAX_ENTRIES just like we do m_source_map.
67877 That's what this commit does.
67879 There should be no user visible changes after this commit, except for
67880 ever so slightly lower memory usage in some cases.
67882 2022-01-12 Andrew Burgess <aburgess@redhat.com>
67884 gdb: new 'maint flush source-cache' command
67885 This commit adds a new 'maint flush source-cache' command, this
67886 flushes the cache of source file contents.
67888 After flushing GDB is forced to reread source files the next time any
67889 source lines are to be displayed.
67891 I've added a test for this new feature. The test is a little weird,
67892 in that it modifies a source file after compilation, and makes use of
67893 the cache flush so that the changes show up when listing the source
67894 file. I'm not sure when such a situation would ever crop up in real
67895 life, but maybe we can imagine such cases.
67897 In reality, this command is useful for testing the syntax highlighting
67898 within GDB, we can adjust the syntax highlighting settings, flush the
67899 cache, and then get the file contents re-highlighted using the new
67902 2022-01-12 Andrew Burgess <aburgess@redhat.com>
67904 gdb: rename lin-lwp to linux-nat in set/show debug
67905 Rename 'set debug lin-lwp' to 'set debug linux-nat' and 'show debug
67906 lin-lwp' to 'show debug linux-nat'.
67908 I've updated the documentation and help text to match, as well as
67909 making it clear that the debug that is coming out relates to all
67910 aspects of Linux native inferior support, not just the LWP aspect of
67913 The boundary between general "native" target debug, and the lwp
67914 specific part of that debug was always a little blurry, but the actual
67915 debug variable inside GDB is debug_linux_nat, and the print routine
67916 linux_nat_debug_printf, is used throughout the linux-nat.c file, not
67917 just for lwp related debug, so the new name seems to make more sense.
67919 2022-01-12 Clément Chigot <clement.chigot@atos.net>
67921 ld: add hidden and internal visibility support for XCOFF
67922 This patch adds a primary support for hidden and internal visibility in
67923 GNU linker for XCOFF format.
67924 The protected visibility isn't yet supported.
67930 * xcofflink.c (xcoff_dynamic_definition_p): Add hidden
67931 and internal visibility support.
67932 (xcoff_link_add_symbols): Likewise.
67933 (xcoff_auto_export_p): Likewise.
67934 (bfd_xcoff_export_symbol): Likewise.
67935 (xcoff_link_input_bfd): Likewise.
67939 * testsuite/ld-vsb/main.c: Adapt for XCOFF.
67940 * testsuite/ld-vsb/sh1.c: Likewse.
67941 * testsuite/ld-vsb/vsb.exp: Likewise.
67942 * testsuite/ld-vsb/visibility-1-xcoff-32.d: New test.
67943 * testsuite/ld-vsb/visibility-1-xcoff-64.d: New test.
67944 * testsuite/ld-vsb/visibility-2-xcoff-32.d: New test.
67945 * testsuite/ld-vsb/visibility-2-xcoff-64.d: New test.
67946 * testsuite/ld-vsb/xcoffvsb.dat: New test.
67948 2022-01-12 Clément Chigot <clement.chigot@atos.net>
67950 ld/testsuite: prepare ld-elfvsb to support XCOFF
67951 A following patch will add visibility support in ld for XCOFF. Thus,
67952 ld-elfvsb is renamed ld-vsb and a suffix is added to files targeting only
67957 * testsuite/ld-elfvsb: rename as ld-vsb.
67958 * testsuite/ld-elfvsb/hidden0.d: move to ld-vsb and rename with
67960 * testsuite/ld-elfvsb/hidden1.d: Likewise.
67961 * testsuite/ld-elfvsb/hidden2.d: Likewise.
67962 * testsuite/ld-elfvsb/internal0.d: Likewise.
67963 * testsuite/ld-elfvsb/internal1.d: Likewise.
67964 * testsuite/ld-elfvsb/protected0.d: Likewise.
67965 * testsuite/ld-elfvsb/protected1.d: Likewise.
67967 2022-01-12 Clément Chigot <clement.chigot@atos.net>
67969 gas: add visibility support using GNU syntax on XCOFF
67970 In order to ease port of GNU assembly code and especially ld testsuite,
67971 this patch allows XCOFF to accept the usual GNU syntax for visibility.
67977 * config/tc-ppc.c (ppc_GNU_visibility): New function.
67978 * testsuite/gas/ppc/aix.exp: Add new tests.
67979 * testsuite/gas/ppc/xcoff-visibility-2-32.d: New test.
67980 * testsuite/gas/ppc/xcoff-visibility-2-64.d: New test.
67981 * testsuite/gas/ppc/xcoff-visibility-2.s: New test.
67983 2022-01-12 Clément Chigot <clement.chigot@atos.net>
67985 gas: add visibility support for XCOFF
67986 XCOFF assembly defines the visibility using an additional argument
67987 on several pseudo-ops: .globl, .weak, .extern and .comm.
67988 This implies that .globl and .weak syntax is different than the
67989 usual GNU syntax. But we want to provide compatibility with AIX
67990 assembler, especially because GCC is generating the visibility
67991 using this XCOFF syntax.
67997 * coffcode.h (coff_write_object_contents): Change XCOFF header
67999 * coffgen.c (coff_print_symbol): Increase the size for n_type.
68003 * config/tc-ppc.c (ppc_xcoff_get_visibility): New function.
68004 (ppc_globl): New function.
68005 (ppc_weak): New function.
68006 (ppc_comm): Add visibility field support.
68007 (ppc_extern): Likewise.
68008 * testsuite/gas/all/cofftag.d: Adjust to new n_type size
68009 providing by objdump.
68010 * testsuite/gas/ppc/test1xcoff32.d: Likewise.
68011 * testsuite/gas/ppc/aix.exp: Add new tests.
68012 * testsuite/gas/ppc/xcoff-visibility-1-32.d: New test.
68013 * testsuite/gas/ppc/xcoff-visibility-1-64.d: New test.
68014 * testsuite/gas/ppc/xcoff-visibility-1.s: New test.
68018 * coff/internal.h (SYM_V_INTERNAL, SYM_V_HIDDEN,
68019 SYM_V_PROTECTED, SYM_V_EXPORTED, SYM_V_MASK): New defines.
68020 * coff/xcoff.h (struct xcoff_link_hash_entry): Add visibility
68025 * testsuite/ld-pe/pr19803.d: Adjust to new n_type size
68026 providing by objdump.
68028 2022-01-12 Hans-Peter Nilsson <hp@axis.com>
68030 objdump, readelf: Emit "CU:" format only when wide output is requested
68031 As pre-approved by Alan in
68032 https://sourceware.org/pipermail/binutils/2021-September/118019.html
68033 and I believe people have run into getting testsuite failures for
68034 test-environments with "long" directory names, at least once more
68035 since that time. Enough. I grepped the gas, binutils and ld
68036 testsuites for "CU:" to catch target-specific occurrences, but I
68037 noticed none. I chose to remove "CU:" on the objdump tests instead of
68038 changing options to get the wide format, so as to keep the name of the
68039 test consistent with actual options; but added it to the readelf
68040 options for the gas test as I believe the "CU:" format is preferable.
68042 Tested for cris-elf and native x86_64-pc-linux-gnu.
68045 * dwarf.c (display_debug_lines_decoded): Don't check the
68046 string length of the directory, instead emit the "CU: dir/name"
68047 format only if wide output is requested.
68048 * testsuite/binutils-all/dw5.W, testsuite/binutils-all/objdump.WL:
68049 Adjust accordingly.
68052 * testsuite/gas/elf/dwarf-5-loc0.d: Add -W to readelf options.
68054 2022-01-12 Alan Modra <amodra@gmail.com>
68056 Set SEC_ELF_REVERSE_COPY earlier
68057 For the sake of DT_RELR.
68060 * elflink.c (elf_link_input_bfd): Don't set SEC_ELF_REVERSE_COPY
68061 here. Move sanity checks to reverse copying code.
68063 * ldlang.c (lang_add_section): Set SEC_ELF_REVERSE_COPY for
68064 .ctors/.dtors in .init_array/.fini_array.
68066 2022-01-12 Tiezhu Yang <yangtiezhu@loongson.cn>
68068 gdb: testsuite: fix wrong comment in gdb.base/charset.c
68069 In gdb/testsuite/gdb.base/charset.c, use "IBM1047" instead of "EBCDIC"
68070 to fix the wrong comment.
68072 2022-01-12 Tiezhu Yang <yangtiezhu@loongson.cn>
68074 gdb: testsuite: fix failed testcases in gdb.base/charset.exp
68075 In gdb/testsuite/gdb.base/charset.c, the last argument is greater than 127
68076 when call fill_run() in EBCDIC-US and IBM1047, but the type of string[] is
68077 char, this will change the value due to sign extension.
68079 For example, ebcdic_us_string[7] will be -63 instead of the original 193 in
68082 Make the type of string[] as unsigned char to fix the following six failed
68085 $ grep FAIL gdb/testsuite/gdb.sum
68086 FAIL: gdb.base/charset.exp: check value of parsed character literal in EBCDIC-US
68087 FAIL: gdb.base/charset.exp: check value of parsed string literal in EBCDIC-US
68088 FAIL: gdb.base/charset.exp: check value of escape that doesn't exist in EBCDIC-US
68089 FAIL: gdb.base/charset.exp: check value of parsed character literal in IBM1047
68090 FAIL: gdb.base/charset.exp: check value of parsed string literal in IBM1047
68091 FAIL: gdb.base/charset.exp: check value of escape that doesn't exist in IBM1047
68093 2022-01-12 GDB Administrator <gdbadmin@sourceware.org>
68095 Automatic date update in version.in
68097 2022-01-11 Fangrui Song <maskray@google.com>
68099 ar: Add --thin for creating thin archives
68100 In many ar implementations (FreeBSD, elfutils, etc), -T has the X/Open
68101 System Interface specified semantics. Therefore -T for thin archives is
68102 not recommended for portability. -T is deprecated without diagnostics.
68105 * ar.c (long_options): Add --thin.
68106 (usage) Add --thin. Deprecate -T without diagnostics.
68107 * doc/binutils.texi: Add doc.
68108 * NEWS: Mention --thin.
68109 * binutils/testsuite/binutils-all/ar.exp: Add tests.
68111 2022-01-11 Martin Storsj <martin@martin.st>
68113 Fix multiple problems with DLL generation.
68114 ld * pe-dll.c (make_head): Prefix the symbol name with the dll name.
68115 (make_tail, make_one, make_singleton_name_thunk): Likewise.
68116 (make_import_fixup_entry, make_runtime_pseudo_reloc): Likewise.
68117 (pe_create_runtime_relocator_reference): Likewise.
68118 (pe_dll_generate_implib): Set dll_symname_len.
68119 (pe_process_import_defs): Likewise.
68122 * dlltool.c (main): If a prefix has not been provided, attempt to
68123 use a deterministic one based upon the dll name.
68125 2022-01-11 Jan Beulich <jbeulich@suse.com>
68127 gas/doc: mention quoted symbol names
68129 2022-01-11 Andrew Burgess <aburgess@redhat.com>
68131 gdbsupport: regenerate Makefile.in
68132 I had cause to regenerate gdbsupport/Makefile.in, and noticed some
68133 unexpected changes in the copyright header dates.
68135 I suspect that this was caused by the end of year date range update
68138 The Makefile.in contains two date ranges. The first range appears to
68139 be the date range for the version of automake being used, that is the
68140 range runs up to 2017 only, when automake 1.15.1 was released.
68142 The second date range in Makefile.in represents the date range for the
68143 generated file, and so, now runs up to 2022.
68145 Anyway, this is the result of running autoreconf (using automake
68146 1.15.1) in the gdbsupport directory.
68148 2022-01-11 GDB Administrator <gdbadmin@sourceware.org>
68150 Automatic date update in version.in
68152 2022-01-10 Clément Chigot <clement.chigot@atos.net>
68154 XCOFF: add support for TLS relocations on hidden symbols
68155 This patch adds support for TLS relocation targeting C_HIDEXT symbols.
68156 In gas, TLS relocations, except R_TLSM and R_TLMSL, must keep the value
68157 of their target symbol.
68158 In ld, it simply ensures that internal TLS symbols are added to the
68159 linker hash table for xcoff_reloc_type_tls.
68161 It also improves the tests made by both.
68165 * coff-rs6000.c (xcoff_howto_table): Fix name of R_TLSML.
68166 (xcoff_reloc_type_tls): Replace the error when h is NULL by
68168 (xcoff_complain_overflow_unsigned_func): Adjust comments.
68169 * coff64-rs6000.c (xcoff64_howto_table): Fix name of R_TLSML.
68170 * xcofflink.c (xcoff_link_add_symbols_to_hash_table): New
68172 (xcoff_link_add_symbols): Add C_HIDEXT TLS symbols to the linker
68177 * config/tc-ppc.c (md_apply_fix): Enable support for TLS
68178 relocation over internal symbols.
68179 * testsuite/gas/ppc/aix.exp: Replace xcoff-tlms by xcoff-tls.
68180 * testsuite/gas/ppc/xcoff-tlsm-32.d: Removed.
68181 * testsuite/gas/ppc/xcoff-tlsm-64.d: Removed.
68182 * testsuite/gas/ppc/xcoff-tlsm.s: Removed.
68183 * testsuite/gas/ppc/xcoff-tls-32.d: New test.
68184 * testsuite/gas/ppc/xcoff-tls-64.d: New test.
68185 * testsuite/gas/ppc/xcoff-tls.s: New test.
68189 * testsuite/ld-powerpc/aix52.exp: Improve aix-tls-reloc test.
68190 * testsuite/ld-powerpc/aix-tls-reloc.s: Likewise.
68191 * testsuite/ld-powerpc/aix-tls-reloc-32.d: Removed.
68192 * testsuite/ld-powerpc/aix-tls-reloc-64.d: Removed.
68193 * testsuite/ld-powerpc/aix-tls-reloc-32.dd: New test.
68194 * testsuite/ld-powerpc/aix-tls-reloc-32.dt: New test.
68195 * testsuite/ld-powerpc/aix-tls-reloc-64.dd: New test.
68196 * testsuite/ld-powerpc/aix-tls-reloc-64.dt: New test.
68198 2022-01-10 Tiezhu Yang <yangtiezhu@loongson.cn>
68200 gdb: add Tiezhu Yang to MAINTAINERS
68202 2022-01-10 Tom Tromey <tom@tromey.com>
68204 Reduce use of unfiltered output in Darwin code
68205 The Darwin code uses unfiltered output liberally. This patch changes
68206 this code to send some output to gdb_stdlog (in some cases via the use
68207 of debug_prefixed_printf_cond_nofunc), or to gdb_stderr, or to simply
68208 switch to filtered output.
68210 Note that I didn't switch inferior_debug to use
68211 debug_prefixed_printf_cond_nofunc, because that would affect the
68212 output by removing the information about the inferior. I wasn't sure
68213 if this was important or not, so I left it in.
68215 v2 of this patch uses warning rather than prints to gdb_stderr, and
68216 removes some trailing whitespace.
68218 I can't compile this patch, so it's "best effort".
68220 2022-01-10 GDB Administrator <gdbadmin@sourceware.org>
68222 Automatic date update in version.in
68224 2022-01-09 GDB Administrator <gdbadmin@sourceware.org>
68226 Automatic date update in version.in
68228 2022-01-08 Andrew Burgess <aburgess@redhat.com>
68230 gdb/hurd: handle inferiors exiting
68231 While testing on GNU/Hurd (i386) I noticed that GDB crashes when an
68232 inferior exits, with this error:
68234 inferior.c:293: internal-error: inferior* find_inferior_pid(process_stratum_target*, int): Assertion `pid != 0' failed.
68236 The problem appears to be in gnu_nat_target::wait.
68238 We always set inferior_ptid to null_ptid before calling target_wait,
68239 this has been the case since the multi-target changes were made to GDB
68242 commit 5b6d1e4fa4fc6827c7b3f0e99ff120dfa14d65d2
68243 Date: Fri Jan 10 20:06:08 2020 +0000
68245 Multi-target support
68247 With follow up changes in commit:
68249 commit 24ed6739b699f329c2c45aedee5f8c7d2f54e493
68250 Date: Thu Jan 30 14:35:40 2020 +0000
68252 gdb/remote: Restore support for 'S' stop reply packet
68254 Unfortunately, the GNU/Hurd target is still relying on the value of
68255 inferior_ptid in the case where an inferior exits - we return the
68256 value of inferior_ptid as the pid of the process that exited. This
68257 was fine in the single target world, where inferior_ptid identified
68258 the one running inferior, but this is no longer good enough.
68260 Instead, we should return a ptid containing the pid of the process
68261 that exited, as obtained from the wait event, and this is what this
68264 I've not run the full testsuite on GNU/Hurd as there appear to be lots
68265 of other issues with this target that makes running the full testsuite
68266 very painful, but I think this looks like a small easy improvement.
68268 2022-01-08 Tom Tromey <tom@tromey.com>
68270 Add explicit check for nullptr to target_announce_attach
68271 Lancelot pointed out that target_announce_attach was missing an
68272 explicit check against nullptr. This patch adds it.
68274 2022-01-08 Hannes Domani <ssbssa@yahoo.de>
68276 Add _sigsys info to siginfo struct
68277 This patch adds information about _sigsys structure from newer
68278 kernels, so that $_siginfo decoding can show information about
68279 _sigsys, making it easier for developers to debug seccomp failures.
68280 Requested in PR gdb/24283.
68282 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=24283
68284 2022-01-08 Tiezhu Yang <yangtiezhu@loongson.cn>
68286 gdb: testsuite: show print array-indexes after set in arrayidx.exp
68287 Add "show print array-indexes" testcases after set print array-indexes
68290 Without this patch:
68292 PASS: gdb.base/arrayidx.exp: set print array-indexes to off
68293 PASS: gdb.base/arrayidx.exp: print array with array-indexes off
68294 PASS: gdb.base/arrayidx.exp: set print array-indexes to on
68295 PASS: gdb.base/arrayidx.exp: print array with array-indexes on
68299 PASS: gdb.base/arrayidx.exp: set print array-indexes to off
68300 PASS: gdb.base/arrayidx.exp: show print array-indexes is off
68301 PASS: gdb.base/arrayidx.exp: print array with array-indexes off
68302 PASS: gdb.base/arrayidx.exp: set print array-indexes to on
68303 PASS: gdb.base/arrayidx.exp: show print array-indexes is on
68304 PASS: gdb.base/arrayidx.exp: print array with array-indexes on
68306 2022-01-08 H.J. Lu <hjl.tools@gmail.com>
68308 ld: Extract _bfd_elf_link_iterate_on_relocs
68309 DT_RELR encodes consecutive R_*_RELATIVE relocations in GOT (the global
68310 offset table) and data sections in a compact format:
68312 https://groups.google.com/g/generic-abi/c/bX460iggiKg
68314 On some targets, R_*_RELATIVE relocations are counted and the GOT offsets
68315 are allocated when setting the dynamic section sizes after seeing all
68316 relocations. R_*_RELATIVE relocations are generated while relocating
68317 sections after section layout has been finalized.
68319 To prepare for DT_RELR implementation on these targets, extract
68320 _bfd_elf_link_iterate_on_relocs from _bfd_elf_link_check_relocs so
68321 that a backend can scan relocations in elf_backend_always_size_sections
68323 For x86 targets, the old check_relocs is renamed to scan_relocs and a
68324 new check_relocs is added to chek input sections and create dynamic
68325 relocation sections so that they will be mapped to output sections.
68326 scan_relocs is now called from elf_backend_always_size_sections.
68328 Since relocations are scanned after __start, __stop, .startof. and
68329 .sizeof. symbols have been finalized on x86, __[start|stop]_SECNAME for
68330 --gc-sections -z start-stop-gc are now zero when all SECNAME sections
68331 been garbage collected. This is no need for elf_x86_start_stop_gc_p.
68335 * elf-bfd.h (_bfd_elf_link_iterate_on_relocs): New.
68336 * elf32-i386.c (elf_i386_convert_load_reloc): Don't call
68337 elf_x86_start_stop_gc_p.
68338 (elf_i386_check_relocs): Renamed to ...
68339 (elf_i386_scan_relocs): This. Don't call
68340 _bfd_elf_make_dynamic_reloc_section.
68341 (elf_i386_always_size_sections): New.
68342 (elf_backend_check_relocs): Removed.
68343 (elf_backend_always_size_sections): New.
68344 * elf64-x86-64.c (elf_x86_64_convert_load_reloc): Don't call
68345 elf_x86_start_stop_gc_p.
68346 (elf_x86_64_check_relocs): Renamed to ...
68347 (elf_x86_64_scan_relocs): This. Don't call
68348 _bfd_elf_make_dynamic_reloc_section.
68349 (elf_x86_64_always_size_sections): New.
68350 (elf_backend_check_relocs): Removed.
68351 (elf_backend_always_size_sections): New.
68352 * elflink.c (elf_link_check_or_scan_relocs):
68353 New. Extracted from _bfd_elf_link_check_relocs.
68354 (_bfd_elf_link_check_relocs): Call elf_link_check_or_scan_relocs.
68355 * elfxx-x86.c (_bfd_x86_elf_check_relocs): New.
68356 * elfxx-x86.h (X86_64_NEED_DYNAMIC_RELOC_TYPE_P): New.
68357 (I386_NEED_DYNAMIC_RELOC_TYPE_P): Likewise.
68358 (X86_NEED_DYNAMIC_RELOC_TYPE_P): Likewise.
68359 (_bfd_x86_elf_check_relocs): Likewise.
68360 (elf_backend_check_relocs): Likewise.
68361 (elf_backend_always_size_sections): Removed.
68362 (elf_x86_start_stop_gc_p): Likewise.
68366 * testsuite/ld-i386/pr27491-1a.d: Updated.
68367 * testsuite/ld-x86-64/pr27491-1a.d: Likewise.
68369 2022-01-08 GDB Administrator <gdbadmin@sourceware.org>
68371 Automatic date update in version.in
68373 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68375 gdb/testsuite: Remove duplicates from gdb.mi/mi-catch-load.exp
68376 When I run the testsuite, I have:
68378 Running .../gdb/testsuite/gdb.mi/mi-catch-load.exp ...
68379 DUPLICATE: gdb.mi/mi-catch-load.exp: breakpoint at main
68380 DUPLICATE: gdb.mi/mi-catch-load.exp: mi runto main
68382 Fix by grouping the various phases in with_test_prefix blocks. Since
68383 the tests now have a prefix, remove the manually written prefixes in
68386 Also change some messages with the pattern "(timeout) $testname" into
68387 "$estname (timeout)" since tools will handle this as $testname[1] (which
68388 is what we want in this particular scenario).
68390 Tested on x86_64-linux.
68392 [1] https://sourceware.org/gdb/wiki/GDBTestcaseCookbook#Do_not_use_.22tail_parentheses.22_on_test_messages
68394 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68396 gdb/testsuite: Remove duplicates from gdb.threads/staticthreads.ex
68397 When running the testsuite, I have:
68399 Running .../gdb/testsuite/gdb.threads/staticthreads.exp ...
68400 DUPLICATE: gdb.threads/staticthreads.exp: couldn't compile staticthreads.c: unrecognized error
68402 Fix by using foreach_with_prefix instead of foreach when preparing the
68405 Testeed on x86_64-linux both in a setup where the test fails to prepare
68406 and in a setup where the test fails to setup.
68408 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68410 gdb/testsuite: Remove duplicates from gdb.mi/mi-language.exp
68411 When running the testsuite, I have:
68413 Running .../gdb/testsuite/gdb.mi/mi-language.exp ...
68414 DUPLICATE: gdb.mi/mi-language.exp: set lang ada
68416 This is due to an erroneous explicit test name. This explicit test name
68417 also happens to be useless (at least it would have been if it was
68418 correct) since it only repeats the command, so just remove the explicit
68419 test name and let the command be used as default test name. Also remove
68420 explicit test name at another location in the file since it also just
68421 repeat the command.
68423 Tested on x86_64-linux.
68425 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68427 gdb/testsuite: Remove duplicates from gdb.mi/mi-nonstop-exit.exp
68428 When running the testsuite, I have:
68430 Running .../gdb/testsuite/gdb.mi/mi-nonstop-exit.exp ...
68431 DUPLICATE: gdb.mi/mi-nonstop-exit.exp: breakpoint at main
68432 DUPLICATE: gdb.mi/mi-nonstop-exit.exp: mi runto main
68434 This test runs the same sequence of operations twice. Refactor the code
68435 by running both of those sequences within a foreach_with_prefix block to
68436 ensure that the commands have unique test names.
68438 Tested on x86_64-linux.
68440 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68442 gdb/testsuite: Remove duplicates from gdb.mi/mi-nonstop.exp
68443 When running the testsuite, I have:
68445 Running .../gdb/testsuite/gdb.mi/mi-nonstop.exp ...
68446 DUPLICATE: gdb.mi/mi-nonstop.exp: check varobj, w1, 1
68447 DUPLICATE: gdb.mi/mi-nonstop.exp: stacktrace of stopped thread
68449 Fix by adjusting the problematic test names.
68451 Tested on x86_64-linux.
68453 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68455 gdb/testsuite: Remove duplicates from gdb.mi/mi-nsthrexec.exp
68456 When running the testsuite, I have:
68458 Running .../gdb/testsuite/gdb.mi/mi-nsthrexec.exp ...
68459 DUPLICATE: gdb.mi/mi-nsthrexec.exp: breakpoint at main
68461 Fix by adjusting the duplicated test name.
68463 Tested on x86_64-linux.
68465 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68467 gdb/testsuite: Remove duplicates from gdb.base/watchpoints.exp
68468 When running the testsuite, I have:
68470 Running ../gdb/testsuite/gdb.base/watchpoints.exp ...
68471 DUPLICATE: gdb.base/watchpoints.exp: watchpoint hit, first time
68473 Fix by adjusting the test names where appropriate.
68475 Tested on x86_64-linux.
68477 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68479 gdb/testsuite: Remove duplicates from gdb.base/nested-subp2.exp
68480 When running the testsuite, I have:
68482 Running .../gdb/testsuite/gdb.base/nested-subp2.exp ...
68483 DUPLICATE: gdb.base/nested-subp2.exp: continue to the STOP marker
68484 DUPLICATE: gdb.base/nested-subp2.exp: print c
68485 DUPLICATE: gdb.base/nested-subp2.exp: print count
68487 Fix by using with_test_prefix to differentiate the test that are
68488 performed at different points during the execution of the debuggee.
68490 Tested on x86_64-linux.
68492 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68494 gdb/testsuite: Remove duplicates from gdb.base/call-signal-resume.exp
68495 When running the testsuite, I have:
68497 Running .../gdb/testsuite/gdb.base/call-signal-resume.exp ...
68498 DUPLICATE: gdb.base/call-signal-resume.exp: dummy stack frame number
68499 DUPLICATE: gdb.base/call-signal-resume.exp: set confirm off
68500 DUPLICATE: gdb.base/call-signal-resume.exp: return
68502 This is due to the fact that a pattern was probably copy/pasted to
68503 re-use the logic while not adjusting the test names to avoid the
68506 Fix by removing the redundant tests ('set confirm off' only needs to be
68507 used once) and adjusting the test names where appropriate.
68509 Tested on x86_64-linux.
68511 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68513 gdb/testsuite: Remove duplicates from gdb.base/pointers.exp
68514 When I run the testsuite, I have :
68516 Running .../gdb/testsuite/gdb.base/pointers.exp ...
68517 DUPLICATE: gdb.base/pointers.exp: pointer assignment
68519 Fix by placing the sections with duplication in with_test_prefix blocks.
68520 This removes the duplication and gives a better organization the file.
68522 Tested on x86_64-linux.
68523 Co-Authored-By: Pedro Alves <pedro@palves.net>
68525 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68527 gdb/testsuite: Remove duplicates from gdb.base/unload.exp
68528 When running the testsuite, I have:
68530 Running .../gdb/testsuite/gdb.base/unload.exp ...
68531 DUPLICATE: gdb.base/unload.exp: continuing to unloaded libfile
68533 Fix by adjusting the test name.
68535 Tested on x86_64-linux.
68537 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68539 gdb/testsuite: Remove duplicates from gdb.base/define-prefix.exp
68540 When running the testsuite, I have:
68542 Running .../gdb/testsuite/gdb.base/define-prefix.exp ...
68543 DUPLICATE: gdb.base/define-prefix.exp: define user command: ghi-prefix-cmd
68545 Fix by adjusting test names.
68547 Tested on x86_64-linux.
68549 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68551 gdb/testsuite: Remove duplicates from gdb.base/funcargs.exp
68552 When running the testsuite, I have:
68554 Running .../gdb/testsuite/gdb.base/funcargs.exp ...
68555 DUPLICATE: gdb.base/funcargs.exp: run to call2a
68557 Fix by using proc_with_prefix instead on plain proc to create logical
68560 Tested on x86_64-linux.
68562 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68564 gdb/testsuite: Remove duplicates from gdb.base/shlib-call.exp
68565 When I run the testsuite, I have:
68567 Running .../gdb/testsuite/gdb.base/shlib-call.exp ...
68568 DUPLICATE: gdb.base/shlib-call.exp: print g
68569 DUPLICATE: gdb.base/shlib-call.exp: set print sevenbit-strings
68570 DUPLICATE: gdb.base/shlib-call.exp: set print address off
68571 DUPLICATE: gdb.base/shlib-call.exp: set width 0
68572 DUPLICATE: gdb.base/shlib-call.exp: continue until exit
68574 Fix by adjusting the test names when required, and by removing
68575 un-necessary commands.
68577 While at it, do some cleanup:
68578 - Replace an explicit GDB restart sequence with a call to clean_restart.
68579 - Remove trailing whitespaces.
68580 - Use $gdb_test_name in gdb_test_multiple.
68582 Tested on x86_64-linux.
68584 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68586 gdb/testsuite: Remove duplicates from gdb.base/set-cfd.exp
68587 When running the testsuite, I have:
68589 Running .../gdb/testsuite/gdb.base/set-cwd.exp ...
68590 DUPLICATE: gdb.base/set-cwd.exp: test_cwd_reset: continue to breakpoint: break-here
68592 Fix by moving the tests after the 'runto_main' within the same
68593 with_test_prefix scope.
68595 While at it, I fix some indentation issues.
68597 Tested on x86_64-linux.
68599 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68601 gdb/testsuite: Remove duplicates from gdb.base/exprs.exp
68602 When running the testsuite, I have:
68604 Running .../gdb/testsuite/gdb.base/exprs.exp ...
68605 DUPLICATE: gdb.base/exprs.exp: \$[0-9]* = red (setup)
68607 Fix by using with_test_prefix where appropriate.
68609 Tested on x86_64-linux.
68611 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68613 gdb/testsuite: Remove duplicates from gdb.base/readline.exp
68614 When running the testsuite, I have:
68616 Running .../gdb/testsuite/gdb.base/readline.exp ...
68617 DUPLICATE: gdb.base/readline.exp: Simple operate-and-get-next - final prompt
68619 Fix by adjusting the prefix given to the second 'simple' call to
68620 operate_and_get_next.
68622 Tested on x86_64-linux.
68624 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68626 gdb/testsuite: Remove duplicates from gdb.base/pretty-array.exp
68627 When I run the testsuite, I have:
68629 Running .../gdb/testsuite/gdb.base/pretty-array.exp ...
68630 DUPLICATE: gdb.base/pretty-array.exp: print nums
68631 DUPLICATE: gdb.base/pretty-array.exp: print nums
68633 Fix by giving a name to the test cases.
68635 Tested on x86_64-linux.
68637 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68639 gdb/testsuite: Remove duplicates from gdb.base/ui-redirect.exp
68640 When running the testsuite, I have:
68642 Running .../gdb/testsuite/gdb.base/ui-redirect.exp ...
68643 DUPLICATE: gdb.base/ui-redirect.exp: redirect while already logging: set logging redirect off
68645 Fix by moving the first 'set logging redirect off' to the end of the
68646 previous [with_test_prefix] test block. The statement's purpose is to
68647 clean the on flag set in this previous block, so moving it there makes
68648 sense and does not change the sequence of commands in the test file.
68650 Tested on x86_64-linux.
68652 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68654 gdb: completion-support.exp: improve leading whitespace support
68655 There is a expect support library in the source tree designed to help
68656 developers test the auto-completion capabilities of GDB.
68658 One of the functions is test_gdb_complete_unique_re. It is used
68659 (usually indirectly via test_gdb_complete_unique) to test that a given
68660 input line is completed as a given output line. The test checks for two
68661 ways to do the completion: using tab-completion, or using the
68662 'complete' command. To do this, calls to two dedicated functions are
68663 performed. If we omit few details, we can consider that a call to
68665 test_gdb_complete_unique $input $expected
68667 is equivalent to the two following calls:
68669 test_gdb_complete_tab_unique $input $expected
68670 test_gdb_complete_cmd_unique $input $expected
68672 When using the tab-completion, everything works as expected, but some
68673 care must be taken when using the 'complete' command if the given input
68674 has leading whitespaces. In such situation, the output of the
68675 'complete' command will drop the leading whitespaces.
68677 The current approach is that in such situation, the input and expected
68678 outputs are right trimmed (i.e. all leading whitespaces are removed)
68679 when performing the command completion check.
68681 This means that the following call:
68683 test_gdb_complete_unique " $input" " $expected"
68685 is almost equivalent to (again, omitting few details and arguments):
68687 test_gdb_complete_tab_unique " $input" " $expected"
68688 test_gdb_complete_cmd_unique "$input" "$expected"
68690 This approach comes with a problem that we encounter when running the
68691 tests in complete-empty.exp. When doing so, we have:
68693 Running .../gdb/testsuite/gdb.base/complete-empty.exp ...
68694 DUPLICATE: gdb.base/complete-empty.exp: empty-input-line: cmd complete ""
68696 This is because the test file does something like:
68698 test_gdb_complete_unique "" "!" " " 1
68699 test_gdb_complete_unique " " " !" " " 1¬
68701 which, if we do the substitution introduced above is equivalent to:
68703 test_gdb_complete_tab_unique "" "!"
68704 test_gdb_complete_cmd_unique "" "!"
68705 test_gdb_complete_tab_unique " " " !"
68706 test_gdb_complete_cmd_unique "" "!"
68708 We see that the lines 2 and 4 are now the same, and for this reason the
68709 testing framework complains about DUPLICATE test names.
68711 To fix that, this commit proposes that instead of left trimming both
68712 input and expected outputs, only the expected output is trimmed.
68714 Care must be taken in the case the completion gives more possibilities
68715 than allowed by the max-completions setting. In this case, the input
68716 will be repeated in the output in its left trimmed version. This commit
68717 also ensures that this is taken care of.
68719 With this commit, the gdb.base/complete-empty.exp still passes all its
68720 tests but does not report the DUPLICATE anymore.
68722 Tested on x86_64-linux.
68724 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68726 gdb/testsuite: Remove duplicates from gdb.base/subst.exp
68727 When I run the testsuite, I have:
68729 Running .../gdb/testsuite/gdb.base/subst.ex ...
68730 DUPLICATE: gdb.base/subst.exp: unset substitute-path from, no rule entered yet
68732 Fix by adjusting the problematic test name.
68734 Tested on x86_64-linux.
68736 2022-01-07 Pedro Alves <pedro@palves.net>
68738 gdb/testsuite: Remove duplicates from gdb.base/dfp-exprs.exp
68739 When I run the testsuite, I have:
68741 Running ../gdb/testsuite/gdb.base/dfp-exprs.exp ...
68742 DUPLICATE: gdb.base/dfp-exprs.exp: p 1.2dl < 1.3df
68744 Replace hand-written tests checking various comparison operators between
68745 various decimal floating point types with a loop to programmatically
68746 generate all the combinations. This removes the need to eyeball for all
68747 suffixes, which lead to the original duplication.
68749 Also add a lot more combinations, testing all comparison operators
68750 comprehensively. The result is 262 unique tests vs 104 before this
68753 Tested on x86_86-linux.
68755 Change-Id: Id215a3d610aa8e032bf06ee160b5e3aed4a92d1e
68757 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68759 gdb/testsuite: Remove duplicates from gdb.base/ptype.exp
68760 When running the testsuite, I have:
68762 Running .../gdb/testsuite/gdb.base/ptype.exp ...
68763 DUPLICATE: gdb.base/ptype.exp: ptype the_highest
68764 DUPLICATE: gdb.base/ptype.exp: list intfoo
68765 DUPLICATE: gdb.base/ptype.exp: list charfoo
68767 Fix by adjusting the offending test names.
68769 Tested on x86_64-linux.
68771 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68773 gdb/testsuite: Remove duplicates from gdb.base/dfp-test.exp
68774 When running the testsuite, I have:
68776 Running .../gdb/testsuite/gdb.base/dfp-test.exp ...
68777 DUPLICATE: gdb.base/dfp-test.exp: 1.23E is an invalid number
68778 DUPLICATE: gdb.base/dfp-test.exp: 1.23E45A is an invalid number
68779 DUPLICATE: gdb.base/dfp-test.exp: 1.23E is an invalid number
68780 DUPLICATE: gdb.base/dfp-test.exp: 1.23E45A is an invalid number
68782 Fix by using proc_with_prefix where appropriate.
68784 Tested on x86_64-linux.
68785 Co-Authored-By: Andrew Burgess <aburgess@redhat.com>
68787 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68789 gdb/testsuite: Remove duplicates from gdb.base/del.exp
68790 When running the testsuite, I have:
68792 Running .../gdb/testsuite/gdb.base/del.exp ...
68793 DUPLICATE: gdb.base/del.exp: info break after removing break on main
68795 Refactor slightly this test to run the various configurations under
68796 foreach_with_prefix so each variant is automatically prefixed, ensuring
68797 that the forgotten custom test name cannot happen.
68799 Tested on x86_64-linux.
68801 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68803 gdb/testsuite: Remove duplicates from gdb.base/solib-display.exp
68804 When running the testsuite, I have:
68806 Running .../gdb/testsuite/gdb.base/solib-display.exp ...
68807 DUPLICATE: gdb.base/solib-display.exp: NO: break 25
68808 DUPLICATE: gdb.base/solib-display.exp: NO: continue
68809 DUPLICATE: gdb.base/solib-display.exp: IN: break 25
68810 DUPLICATE: gdb.base/solib-display.exp: IN: continue
68811 DUPLICATE: gdb.base/solib-display.exp: SEP: break 25
68812 DUPLICATE: gdb.base/solib-display.exp: SEP: continue
68814 The 'break 25' appears because the test inserts two breakpoints at the
68815 same location. Fix this by only inserting the breakpoint once.
68817 Fix the 'continue' DUPLICATE by giving a phony name to the second
68818 continue: 'continue two'.
68820 While at it, this commit also removes a trailing space.
68822 Tested on x86_64-linux.
68824 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68826 gdb/testsuite: Remove duplicates from gdb.base/decl-before-def.exp
68827 When running the testsuite, I have:
68829 Running .../gdb/testsuite/gdb.base/decl-before-def.exp ...
68830 DUPLICATE: gdb.base/decl-before-def.exp: p a
68832 Fix by giving explicit names to the two tests that use the same command.
68834 Tested on x86_64-linux.
68836 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68838 gdb/testsuite: Remove duplicates from gdb.base/pending.exp
68839 When running the testsuite, I have:
68841 Running .../gdb/testsuite/gdb.base/pending.exp ...
68842 DUPLICATE: gdb.base/pending.exp: disable other breakpoints
68844 Fix by adjusting the test names.
68846 Tested on x86_64-linux.
68848 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68850 gdb/testsuite: Remove duplicates from gdb.base/checkpoint.exp
68851 When running the testsuite, I have:
68853 Running .../gdb/testsuite/gdb.base/checkpoint.exp ...
68854 DUPLICATE: gdb.base/checkpoint.exp: verify lines 5 two
68855 DUPLICATE: gdb.base/checkpoint.exp: restart 0 one
68857 This patch fixes the various erroneous incorrect test names.
68859 While at it, this patch also remove some trailing white spaces across
68862 Tested on x86_64-linux.
68864 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68866 gdb/testsuite: Remove duplicates from gdb.base/pie-fork.exp
68867 When running the testsuite, I have:
68869 Running .../gdb/testsuite/gdb.base/pie-fork.exp ...
68870 DUPLICATE: gdb.base/pie-fork.exp: test_no_detach_on_fork: continue
68872 Fix by giving explicit names to the 'continue' commands that cause the
68875 Tested on x86_64-linux.
68877 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68879 gdb/testsuite: Remove duplicates from gdb.base/realname-expand.exp
68880 When running the testsuite, I have:
68882 Running .../gdb/testsuite/gdb.base/realname-expand.exp ...
68883 DUPLICATE: gdb.base/realname-expand.exp: set basenames-may-differ on
68885 This is due to the fact that the test restarts GDB twice and each time
68886 sets the basenames-may-differ setting. This patch proposes to fix this
68887 by not restarting GDB so the setting is maintained. It just clears the
68888 breakpoints between the two tests and updates the breakpoints number as
68891 This patch also perform some minor refactorings to improve visibility.
68893 Tested on x86_64-linux.
68895 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68897 gdb/testsuite: Remove duplicates from gdb.base/interp.exp
68898 When running the testsuite I have:
68900 Running .../gdb/testsuite/gdb.base/interp.exp ...
68901 DUPLICATE: gdb.base/interp.exp: interpreter-exec mi "-var-update *"
68903 This is due to the fact that multiple successive instances of
68904 gdb_test_multiple use 'pass $cmd', but one of them forgets to reset $cmd
68905 to a new test name.
68907 Fix by using 'pass $gdb_test_name', given that the gdb_test_name is set
68908 by gdb_test_multiple.
68910 While fixing this, this patch refactors all occurrences of the following
68914 gdb_test_multiple $cmd $cmd {
68922 gdb_test_multiple foo "" {
68924 pass $gdb_test_name
68928 This makes this test file coherent in its use of $gdb_test_name.
68930 Tested on x86_64-linux.
68932 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68934 gdb/testsuite: Remove duplicates from gdb.base/miscexprs.exp
68935 When running the testsuite I see:
68937 Running .../gdb/testsuite/gdb.base/miscexprs.exp ...
68938 DUPLICATE: gdb.base/miscexprs.exp: print value of !ibig.i[100]
68939 DUPLICATE: gdb.base/miscexprs.exp: print value of !ibig.i[100]
68941 This is due to an explicit test name repeated across multiple tests.
68942 The actual test explicit names do not add much over the command from
68943 wich default test names are derived.
68945 Fix by removing the explicit test names across the file where they do
68946 not add value. While at doing some cleaning, also use $gdb_test_name in
68947 the various uses of gdb_test_multiple.
68949 Tested on x86_64-linux.
68951 2022-01-07 Lancelot SIX <lsix@lancelotsix.com>
68953 gdb/testsuite: Remove duplicates from gdb.base/stack-checking.exp
68954 When running the testsuite I have:
68956 Running .../gdb/testsuite/gdb.base/stack-checking.exp ...
68957 DUPLICATE: gdb.base/stack-checking.exp: bt
68958 DUPLICATE: gdb.base/stack-checking.exp: bt
68960 Fix by using with_test_prefix.
68962 Tested on x86_64-linux.
68964 2022-01-07 Philipp Tomsich <philipp.tomsich@vrull.eu>
68966 RISC-V: update docs to reflect privileged spec v1.9 has been dropped
68967 After commit d8af286fffa ("RISC-V: Drop the privileged spec v1.9
68968 support.") has removed support for privileged spec v1.9, this removes
68969 it from the documentation.
68971 References: d8af286fffa ("RISC-V: Drop the privileged spec v1.9 support.")
68975 * configure: Regenerate.
68976 * configure.ac: Remove reference to priv spec 1.9.
68981 2022-01-07 Philipp Tomsich <philipp.tomsich@vrull.eu>
68983 RISC-V: update docs for -mpriv-spec/--with-priv-spec for 1.12
68984 While support for the privileged spec was added in a63375ac337
68985 ("RISC-V: Hypervisor ext: support Privileged Spec 1.12"), the
68986 documentation has not been updated. Add 1.12 to the relevant
68989 References: a63375ac337 ("RISC-V: Hypervisor ext: support Privileged Spec 1.12")
68993 * config/tc-riscv.c: Add 1.12 to the usage message.
68994 * configure: Regenerate.
68995 * configure.ac: Add 1.12 to the help/usage message.
69000 2022-01-07 Tom Tromey <tromey@adacore.com>
69002 Do not use CC_HAS_LONG_LONG
69003 ax.cc checks CC_HAS_LONG_LONG, but nothing defines this. However,
69004 PRINTF_HAS_LONG_LONG is checked in configure. This patch removes the
69005 former and keeps the latter. This is PR remote/14976 (filed by me in
69008 I'm checking this in.
69010 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=14976
69012 2022-01-07 Andrew Burgess <aburgess@redhat.com>
69014 gdb/doc: shorten some source lines, and prevent some line breaks
69015 Building on the previous commit, this makes use of a trailing @ to
69016 split long @deffn lines in the guile.texi source file. This splitting
69017 doesn't change how the document is laid out by texinfo.
69019 I have also wrapped keyword and argument name pairs in @w{...} to
69020 prevent line breaks appearing between the two. I've currently only
69021 done this for the longer @deffn lines, where a line break is
69022 possible. This makes the @deffn lines much nicer to read in the
69025 2022-01-07 Andrew Burgess <aburgess@redhat.com>
69027 gdb/doc: Remove (...) around guile procedure names in @deffn lines
69028 Most guile procedures in the guile.texi file are defined like:
69030 @deffn {Scheme Procedure} name arg1 arg2 arg3
69032 But there are two places where we do this:
69034 @deffn {Scheme Procedure} (name arg1 arg2 arg3)
69036 Notice the added (...). Though this does represent how a procedure
69037 call is written in scheme, it's not the normal style throughout the
69038 manual. I also checked the 'info guile' info page to see how they
69039 wrote there declarations, and they use the first style too.
69041 The second style also has the drawback that index entries are added as
69042 '(name', and so they are grouped in the '(' section of the index,
69043 which is not very user friendly.
69045 In this commit I've changed the definitions of make-command and
69046 make-parameter to use the first style.
69048 The procedure declaration lines can get pretty long with all of the
69049 arguments, and this was true for both of the procedures I am changing
69050 in this commit. I have made use of a trailing '@' to split the deffn
69051 lines, and keep them under 80 characters in the texi source. This
69052 makes no difference to how the final document looks.
69054 Finally, our current style for keyword arguments, appears to be:
69056 [#:keyword-name argument-name]
69058 I don't really understand the reason for this, 'info guile' just seems
69063 which seems just as good to me. But I don't propose to change
69064 that just now. What I do notice though, is that sometimes, texinfo
69065 will place a line break between the keyword-name and the
69066 argument-name, for example, the pdf of make-command is:
69068 make-command name [#:invoke invoke] [#:command-class
69069 command-class] [#:completer-class completer] [#:prefix? prefix] [#:doc
69072 Notice the line break after '#:command-class' and after '#:doc',
69073 neither of which are ideal. And so, for the two commands I am
69074 changing in this commit, I have made use of @w{...} to prevent line
69075 breaks between the keyword-name and the argument-name. Now the pdf
69078 make-command name [#:invoke invoke]
69079 [#:command-class command-class] [#:completer-class completer]
69080 [#:prefix? prefix] [#:doc doc-string]
69082 Which seems much better. I'll probably update the other deffn lines
69085 2022-01-07 Pavel Mayorov <pmayorov@cloudlinux.com>
69087 Revert previous delta to debug.c. Replace with patch to reject indirect types that point to indirect types.
69089 * dwarf.c: Revert previous delta.
69090 (debug_get_real_type): Reject indirect types that point to
69092 (debug_get_type_name, debug_get_type_size, debug_write_type):
69095 2022-01-07 Nelson Chu <nelson.chu@sifive.com>
69097 RISC-V: Updated the default ISA spec to 20191213.
69098 Update the default ISA spec from 2.2 to 20191213 will change the default
69099 version of i from 2.0 to 2.1. Since zicsr and zifencei are separated
69100 from i 2.1, users need to add them in the architecture string if they need
69101 fence.i and csr instructions. Besides, we also allow old ISA spec can
69102 recognize zicsr and zifencei, but we won't output them since they are
69103 already included in the i extension when i's version is less than 2.1.
69106 * elfxx-riscv.c (riscv_parse_add_subset): Allow old ISA spec can
69107 recognize zicsr and zifencei.
69109 * config/tc-riscv.c (DEFAULT_RISCV_ISA_SPEC): Updated to 20191213.
69110 * testsuite/gas/riscv/csr-version-1p10.d: Added zicsr to -march since
69111 the default version of i is 2.1.
69112 * testsuite/gas/riscv/csr-version-1p11.d: Likewise.
69113 * testsuite/gas/riscv/csr-version-1p12.d: Likewise.
69114 * testsuite/gas/riscv/csr-version-1p9p1.d: Likewise.
69115 * testsuite/gas/riscv/option-arch-03.d: Updated i's version to 2.1.
69116 * testsuite/gas/riscv/option-arch-03.s: Likewise.
69118 * testsuite/ld-riscv-elf/call-relax.d: Added zicsr to -march since
69119 the default version of i is 2.1.
69120 * testsuite/ld-riscv-elf/attr-merge-arch-01.d: Updated i's version to 2.1.
69121 * testsuite/ld-riscv-elf/attr-merge-arch-01a.s: Likewise.
69122 * testsuite/ld-riscv-elf/attr-merge-arch-01b.: Likewise.
69123 * testsuite/ld-riscv-elf/attr-merge-arch-02.d: Likewise.
69124 * testsuite/ld-riscv-elf/attr-merge-arch-02a.s: Likewise.
69125 * testsuite/ld-riscv-elf/attr-merge-arch-02b.s: Likewise.
69126 * testsuite/ld-riscv-elf/attr-merge-arch-03.d: Likewise.
69127 * testsuite/ld-riscv-elf/attr-merge-arch-03a.s: Likewise.
69128 * testsuite/ld-riscv-elf/attr-merge-arch-03b.s: Likewise.
69129 * testsuite/ld-riscv-elf/attr-merge-arch-failed-02.d: Added zifencei
69130 into Tag_RISCV_arch since it is added implied when i's version is
69133 2022-01-07 Alan Modra <amodra@gmail.com>
69135 Move elf_backend_always_size_sections earlier
69136 * elflink.c (bfd_elf_size_dynamic_sections): Move plt/got init
69137 earlier and call elf_backend_always_size_sections at the start
69140 2022-01-07 GDB Administrator <gdbadmin@sourceware.org>
69142 Automatic date update in version.in
69144 2022-01-06 H.J. Lu <hjl.tools@gmail.com>
69146 ldelfgen.c: Add missing newlines when calling einfo
69147 * ldelfgen.c (ldelf_map_segments): Add the missing newline to
69150 2022-01-06 Nick Clifton <nickc@redhat.com>
69152 Fix a stack exhaustion bug parsing malicious STABS format debug information.
69154 * debug.c (debug_write_type): Allow for malicious recursion via
69155 indirect debug types.
69157 2022-01-06 Richard Sandiford <richard.sandiford@arm.com>
69159 aarch64: Add support for new SME instructions
69160 This patch adds support for three new SME instructions: ADDSPL,
69161 ADDSVL and RDSVL. They behave like ADDPL, ADDVL and RDVL, but read
69162 the streaming vector length instead of the current vector length.
69165 * aarch64-tbl.h (aarch64_opcode_table): Add ADDSPL, ADDSVL and RDSVL.
69166 * aarch64-dis-2.c: Regenerate.
69169 * testsuite/gas/aarch64/sme.s, testsuite/gas/aarch64/sme.d: Add tests
69170 for ADDSPL, ADDSVL and RDSVL.
69172 2022-01-06 Tom Tromey <tom@tromey.com>
69174 Use target_announce_detach in more targets
69175 target_announce_detach was added in commit 0f48b757 ("Factor out
69176 "Detaching from program" message printing"). There, Pedro wrote:
69178 (For now, I left the couple targets that print this a bit differently
69179 alone. Maybe this could be further pulled out into infcmd.c. If we
69180 did that, and those targets want to continue printing differently,
69181 this new function could be converted to a target method.)
69183 It seems to me that the differences aren't very big, and in some cases
69184 other targets handled the output a bit more nicely. In particular,
69185 some targets will print a different message when exec_file==NULL,
69186 rather than printing the same output with an empty string as
69189 This patch incorporates the nicer output into target_announce_detach,
69190 then changes the remaining ports to use this function.
69192 2022-01-06 Tom Tromey <tom@tromey.com>
69194 Introduce target_announce_attach
69195 This introduces target_announce_attach, by analog with
69196 target_announce_detach. Then it converts existing targets to use
69197 this, rather than emitting their own output by hand.
69199 2022-01-06 Andrew Burgess <aburgess@redhat.com>
69201 gdb: make use add_setshow_prefix_cmd in gnu-nat.c
69202 In gnu-nat.c we currently implement some set/show prefix commands
69203 "manually", that is, we call add_prefix_cmd, and assign a set and show
69204 function to each prefix command.
69206 These set/show functions print an error indicating that the user
69207 didn't type a complete command.
69209 If we instead switch to using add_setshow_prefix_cmd then we can
69210 delete the set/show functions, GDB provides some default functions,
69211 which give a nice help style summary that lists all of the available
69212 sub-commands, along with a one line summary of what each does.
69214 Though this clearly changes the existing behaviour, I think this
69215 change is acceptable as the new behaviour is more inline with other
69216 set/show prefix commands, and the new behaviour is more informative.
69218 This change will conflict with Tom's change here:
69220 https://sourceware.org/pipermail/gdb-patches/2022-January/184724.html
69222 Where Tom changes the set/show functions that I delete. My suggestion
69223 is that the set/show functions still be deleted even after Tom's
69224 patch (or instead of Tom's patch).
69226 For testing I've build GDB on GNU/Hurd, and manually tested these
69227 functions. I did a grep over the testsuite, and don't believe the
69228 existing error messages are being checked for in any tests.
69230 2022-01-06 Tom Tromey <tom@tromey.com>
69232 Use warning in windows-nat error messages
69233 A warning in windows-nat.c can be converted to use the warning
69234 function. As a side effect, this arranges for the output to be sent
69237 2022-01-06 Tom Tromey <tom@tromey.com>
69239 Clean up some dead code in windows-tdep.c
69240 windows-tdep.c checks the result of xmalloc, which isn't necessary. I
69241 initially removed this dead check, but then went a bit further and
69242 modified the code so that some "goto"s and explicit memory management
69243 could be removed. Then, I added a couple of missing bounds checks.
69245 I believe this also fixes a possible bug with a missing 0-termination
69246 of a string. I am not certain, but that is why I think the existing
69247 code allocates a buffer that is 1 byte too long -- but then it fails
69248 to set this byte to 0.
69250 2022-01-06 Tom Tromey <tromey@adacore.com>
69252 Avoid crash in language_info
69253 language_info calls:
69255 show_language_command (NULL, 1, NULL, NULL);
69257 ... "knowing" that show_language_command does not use its ui_file
69258 parameter. However, this was changed in commit 7514a661
69259 ("Consistently Use ui_file parameter to show callbacks").
69261 This patch changes language_info to pass a ui_file.
69263 It took a while to write the test -- this function is only called when
69264 'verbose' is on and when switching the "expected" language in auto
69267 2022-01-06 Tom Tromey <tromey@adacore.com>
69269 Fix some failures in langs.exp
69270 langs.exp currently has some fails for me because the stack trace
69271 includes full paths to the source files.
69273 FAIL: gdb.base/langs.exp: up to foo in langs.exp
69274 FAIL: gdb.base/langs.exp: up to cppsub_ in langs.exp
69275 FAIL: gdb.base/langs.exp: up to fsub in langs.exp
69277 This fixes the failures by making the filename regexps a bit more lax.
69279 2022-01-06 Jan Beulich <jbeulich@suse.com>
69281 x86: drop NoAVX insn attribute
69282 To avoid issues like that addressed by 6e3e5c9e4181 ("x86: extend SSE
69283 check to PCLMULQDQ, AES, and GFNI insns"), base the check on opcode
69284 attributes and operand types.
69286 x86: drop NoAVX from POPCNT
69287 With the introduction of CpuPOPCNT the NoAVX attribute has become
69288 meaningless for POPCNT.
69290 x86: drop some "comm" template parameters
69291 As already indicated in a remark when introducing these templates, the
69292 "commutative" attribute is ignored for legacy encoding templates. Hence
69293 it is possible to shorten a number of templates by specifying C directly
69294 rather than through a template parameter. I think this helps readability
69297 x86: templatize FMA insn templates
69298 The operand ordering portion of the mnemonics repeats, causing a flurry
69299 of almost identical templates. Abstract this out.
69301 x86-64: restrict PC32 -> PLT32 conversion
69302 Neither non-64-bit code nor uses with a non-zero offset from a symbol
69303 should be converted to PLT32, as an eventual PLT entry would not express
69304 what was requested.
69306 2022-01-06 Lancelot SIX <lancelot.six@amd.com>
69308 gdb: Fix copyright year in gdb/testsuite/gdb.base/inferior-clone.exp
69309 I just realized that I forgot to update the year before pushing the
69310 patch that created this file. Since it landed after the global
69311 copyright year update have been done, this file’s copyright year is
69314 This patch fixes that.
69316 Change-Id: I280f7d86e02d38425f7afdcf19a1c3500d51c23f
69318 2022-01-06 Mike Frysinger <vapier@gentoo.org>
69320 sim: ppc: migrate to standard uintXX_t types
69321 Drop the sim-specific unsignedXX types and move to the standard uintXX_t
69322 types that C11 provides.
69324 sim: common: migrate to standard uintXX_t types
69325 Drop the sim-specific unsignedXX types and move to the standard uintXX_t
69326 types that C11 provides.
69328 sim: igen: migrate to standard uintXX_t types
69329 Move off the custom local 64-bit types and to the standard uintXX_t
69330 types that C11 provides.
69332 sim: mips: migrate to standard uintXX_t types
69333 Move off the sim-specific unsignedXX types and to the standard uintXX_t
69334 types that C11 provides.
69336 sim: cris: migrate to standard uintXX_t types
69337 Move off the sim-specific unsignedXX types and to the standard uintXX_t
69338 types that C11 provides.
69340 sim: iq2000: migrate to standard uintXX_t types
69341 Move off the sim-specific unsignedXX types and to the standard uintXX_t
69342 types that C11 provides.
69344 sim: synacor: migrate to standard uintXX_t types
69345 Move off the sim-specific unsignedXX types and to the standard uintXX_t
69346 types that C11 provides.
69348 sim: msp430: migrate to standard uintXX_t types
69349 Move off the sim-specific unsignedXX types and to the standard uintXX_t
69350 types that C11 provides.
69352 sim: riscv: migrate to standard uintXX_t types
69353 Move off the sim-specific unsignedXX types and to the standard uintXX_t
69354 types that C11 provides.
69356 sim: bfin: migrate to standard uintXX_t types
69357 Move off the sim-specific unsignedXX types and to the standard uintXX_t
69358 types that C11 provides.
69360 sim: testsuite: migrate to standard uintXX_t types
69361 This old code setup its own uintXX types, but since we require C11
69362 now, we can assume the standard uintXX_t types exist and use them.
69364 sim: erc32: migrate to standard uintXX_t types
69365 This old port setup its own uintXX types, but since we require C11
69366 now, we can assume the standard uintXX_t types exist and use them.
69368 sim: mn10300: migrate to standard uintXX_t types
69369 This old port setup its own uintXX types, but since we require C11
69370 now, we can assume the standard uintXX_t types exist and use them.
69372 sim: v850: migrate to standard uintXX_t types
69373 This old port setup its own uintXX types, but since we require C11
69374 now, we can assume the standard uintXX_t types exist and use them.
69376 2022-01-06 Mike Frysinger <vapier@gentoo.org>
69378 sim: m68hc11: migrate to standard uintXX_t types
69379 This old port setup its own uintXX types, but since we require C11
69380 now, we can assume the standard uintXX_t types exist and use them.
69382 Also migrate off the sim-specific unsignedXX types.
69384 2022-01-06 Mike Frysinger <vapier@gentoo.org>
69386 sim: d10v: migrate to standard uintXX_t types
69387 This old port setup its own uintXX types, but since we require C11
69388 now, we can assume the standard uintXX_t types exist and use them.
69390 Also migrate off the sim-specific unsignedXX types.
69392 2022-01-06 Mike Frysinger <vapier@gentoo.org>
69394 sim: cr16: migrate to standard uintXX_t types
69395 This old port setup its own uintXX types, but since we require C11
69396 now, we can assume the standard uintXX_t types exist and use them.
69398 Also migrate off the sim-specific unsignedXX types.
69400 2022-01-06 GDB Administrator <gdbadmin@sourceware.org>
69402 Automatic date update in version.in
69404 2022-01-05 H.J. Lu <hjl.tools@gmail.com>
69406 x86: Add elf_x86_allocate_local_got_info
69407 Add elf_x86_allocate_local_got_info to allocate x86 GOT info for local
69410 * elf32-i386.c (elf_i386_check_relocs): Call
69411 elf_x86_allocate_local_got_info.
69412 * elf64-x86-64.c (elf_x86_64_check_relocs): Likewise.
69413 * elfxx-x86.h (elf_x86_allocate_local_got_info): New.
69415 2022-01-05 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
69417 opcodes: Make i386-dis.c thread-safe
69418 Improve thread safety in print_insn_i386_att, print_insn_i386_intel and
69419 print_insn_i386 by removing the use of static variables.
69421 Tested on x86_64-pc-linux-gnu.
69423 2022-01-04 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
69425 * i386-dis.c: Make print_insn_i386_att, print_insn_i386_intel
69426 and print_insn_i386 thread-safe
69428 2022-01-05 H.J. Lu <hjl.tools@gmail.com>
69430 doc: Replace =frame-interp with =frames-interp
69431 The actual objdump and readelf option name is =frames-interp, not
69435 * doc/debug.options.texi: Replace =frame-interp with
69438 2022-01-05 Tom Tromey <tromey@adacore.com>
69440 Change riscv_return_value to use RETURN_VALUE_ABI_PRESERVES_ADDRESS
69441 Internally, AdaCore has a test that is equivalent to (really a direct
69442 translation of) gdb.base/gnu_vector.exp. On 32-bit RISC-V, the
69443 "return" part of this test fails.
69445 Joel tracked this down to riscv_return_value returning
69446 RETURN_VALUE_ABI_RETURNS_ADDRESS. Using
69447 RETURN_VALUE_ABI_PRESERVES_ADDRESS is more correct here, and fixes the
69450 I tested this for both 32- and 64-bit RISC-V using the AdaCore
69451 internal test suite, and Andrew Burgess tested it using
69454 2022-01-05 Tom Tromey <tom@tromey.com>
69456 Filtered output cleanup in expression dumping
69457 Most of the expression-dumping code uses filtered output, but a few
69458 functions did not. This patch cleans up these instance.
69460 Note that this won't cause any behavior change, because the only calls
69461 to dump_prefix_expression pass in gdb_stdlog. However, in the long
69462 run it's easier to audit the code if the number of uses of _unfiltered
69465 2022-01-05 Tom Tromey <tom@tromey.com>
69467 Use filtered output in terminal_info implementations
69468 This changes one terminal_info implementation, and
69469 default_terminal_info, to use filtered output. Other implementations
69470 of this method already use filtered output.
69472 I can't compile go32-nat.c, so this is a 'best effort' patch.
69474 2022-01-05 Tom Tromey <tom@tromey.com>
69476 Use filtered output in gnu-nat.c commands
69477 gnu-nat.c has a number of ordinary commands that should use filtered
69478 output. In a few cases, I changed the output to use gdb_stderr as
69479 well. I can't compile this file, so this patch is split out as a
69482 Use filtered output in *-tdep commands
69483 Various targets introduce their own commands, which then use
69484 unfiltered output. It's better to use filtered output by default, so
69485 this patch fixes the instances I found.
69487 Use filtered output in btrace-related commands
69488 This changes btrace.c and record-btrace.c to use filtered output in
69489 the commands implemented there.
69491 Use filtered output in some dumping commands
69492 There are several commands that may optionally send their output to a
69493 file -- they take an optional filename argument and open a file. This
69494 patch changes these commands to use filtered output. The rationale
69495 here is that, when printing to gdb_stdout, filtering is appropriate --
69496 it is, and should be, the default for all commands. And, when writing
69497 to a file, paging will not happen anyway (it only happens when the
69498 stream==gdb_stdout), so using the _filtered form will not change
69501 Use filtered output in kill command
69502 This changes the kill command to use filtered output. I split this
69503 one into its own patch because, out of an abundance of caution, I
69504 changed the function to call bfd_cache_close_all a bit earlier, in
69505 case pagination caused an exception.
69507 2022-01-05 Tom Tromey <tom@tromey.com>
69509 Use filtered output in ordinary commands
69510 Many otherwise ordinary commands choose to use unfiltered output
69511 rather than filtered. I don't think there's any reason for this, so
69512 this changes many such commands to use filtered output instead.
69514 Note that complete_command is not touched due to a comment there
69515 explaining why unfiltered output is believed to be used.
69517 2022-01-05 Tom Tromey <tom@tromey.com>
69519 Use filtered output in language_info
69520 Change language_info to use filtered output. This is ok because the
69521 sole caller uses filtered output elsewhere, and because this function
69522 calls show_language_command, which also uses filtered output.
69524 Use filtered output in files_info implementations
69525 This changes the implementations of the target files_info method to
69526 use filtered output. This makes sense because the sole caller of this
69527 method is an ordinary command (info_program_command). This patch
69528 changes this command to use filtered output as well.
69530 Use filtered output in target-descriptions.c
69531 target-descriptions.c uses unfiltered output. However, if you happen
69532 to invoke this command interactively, it's probably better for it to
69533 use filtering. For non-interactive use, this doesn't matter.
69535 Use filtered output for gdbarch dump
69536 This changes gdbarch dumping to use filtered output. This seems a bit
69537 better to me, both on the principle that this is an ordinary command,
69538 and because the output can be voluminous, so it may be nice to stop in
69541 2022-01-05 Tom Tromey <tom@tromey.com>
69543 Implement putstr and putstrn in ui_file
69544 In my tour of the ui_file subsystem, I found that fputstr and fputstrn
69545 can be simplified. The _filtered forms are never used (and IMO
69546 unlikely to ever be used) and so can be removed. And, the interface
69547 can be simplified by removing a callback function and moving the
69548 implementation directly to ui_file.
69550 A new self-test is included. Previously, I think nothing was testing
69553 Regression tested on x86-64 Fedora 34.
69555 2022-01-05 Tom Tromey <tromey@adacore.com>
69557 Change how versioned symbols are recorded
69558 A change to BFD caused a gdb regression when using the Ada "catch
69559 exception" feature. The bug is visible when a shared library throws
69560 an exception that is caught in the main executable.
69562 This was discussed here:
69564 https://sourceware.org/pipermail/binutils/2021-July/117538.html
69566 This patch implements Alan's proposed fix, namely to use VERSYM_HIDDEN
69567 rather than the name when deciding to install a version-less symbol.
69569 The internal test case is identical to the catch_ex_std.exp that is
69570 in-tree, so I haven't added a new test. I could not make that one
69571 fail on x86-64 Linux, though. It's possible that maybe I'd have to
69572 update the system linker first, but I didn't want to try that.
69574 Regression tested on x86-64 Fedora 32.
69576 2022-01-05 Hannes Domani <ssbssa@yahoo.de>
69578 Fix inferior_thread attribute in new_thread event
69579 Commit 72ee03ff58 fixed a use-after-move bug in add_thread_object, but
69580 it changed the inferior_thread attribute to contain the inferior instead
69581 of the actual thread.
69582 This now uses the thread_obj in its new location instead.
69584 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28429
69586 2022-01-05 Tom Tromey <tom@tromey.com>
69588 Simplify execute_control_commands_to_string
69589 execute_control_commands_to_string can be rewritten in terms of
69590 execute_fn_to_string, which consolidates some knowledge about which
69591 streams to redirect.
69593 Regression tested on x86-64 Fedora 34.
69595 2022-01-05 Tom Tromey <tromey@adacore.com>
69597 Do not print anything when self-backtrace unavailable
69598 Right now, gdb's self-backtrace feature will still print something
69599 when a backtrace is unavailable:
69601 sig_write (_("----- Backtrace -----\n"));
69603 sig_write (_("Backtrace unavailable\n"));
69604 sig_write ("---------------------\n");
69606 However, if GDB_PRINT_INTERNAL_BACKTRACE is undefined, it seems better
69607 to me to print nothing at all.
69609 This patch implements this change. It also makes a couple of other
69610 small changes in this same module: it adds a header guard to
69611 bt-utils.h, and it protects the definitions of
69612 gdb_internal_backtrace_1 with a check of GDB_PRINT_INTERNAL_BACKTRACE.
69614 2022-01-05 Tom Tromey <tromey@adacore.com>
69616 Fix pager regression
69617 The patch to fix paging with redirection caused a regression in the
69618 internal AdaCore test suite. The problem occurs when running an MI
69619 command from the CLI using interpreter-exec, when paging is enabled.
69620 This scenario isn't covered by the current test suite, so this patch
69621 includes a new test.
69623 The problem is that, in this situation, MI does:
69625 fputs_unfiltered (strcmp (context->command, "target-select") == 0
69626 ? "^connected" : "^done", mi->raw_stdout);
69628 Here raw_stdout is a stdio_file wrapping stdout, so the pager thinks
69629 that it is ok to buffer the output. However, in this setup, it isn't
69630 ok, and flushing the wrap buffer doesn't really work properly. Also,
69633 mi_out_put (uiout, mi->raw_stdout);
69635 ... but this uses ui_file::write, which also doesn't flush the wrap
69638 I think all this will be fixed by the pager rewrite series I'm working
69639 on. However, in the meantime, adding the old gdb_stdout check back to
69640 the pager fixes this problem.
69642 Regression tested on x86-64 Fedora 34.
69644 2022-01-05 H.J. Lu <hjl.tools@gmail.com>
69646 elf: Set p_align to the minimum page size if possible
69647 Currently, on 32-bit and 64-bit ARM, it seems that ld generates p_align
69648 values of 0x10000 even if no section alignment is greater than 0x1000.
69649 The issue is more general and probably affects other targets with multiple
69652 While file layout absolutely must take 64K page size into account, that
69653 does not have to be reflected in the p_align value. If running on a 64K
69654 kernel, the file will be loaded at a 64K page boundary by necessity. On
69655 a 4K kernel, 64K alignment is not needed.
69657 The glibc loader has been fixed to honor p_align:
69659 https://sourceware.org/bugzilla/show_bug.cgi?id=28676
69663 commit ce81bb256a224259ab686742a6284930cbe4f1fa
69664 Author: Chris Kennelly <ckennelly@google.com>
69665 Date: Thu Oct 15 20:12:32 2020 -0700
69667 fs/binfmt_elf: use PT_LOAD p_align values for suitable start address
69669 This means that on 4K kernels, we will start to do extra work for 64K
69670 p_align, but this pointless for pretty much all binaries (whose section
69671 alignment rarely exceeds 16).
69673 The minimum page size is used, instead of the maximum section alignment
69674 due to this glibc bug:
69676 https://sourceware.org/bugzilla/show_bug.cgi?id=28688
69678 It has been fixed in glibc 2.35. But linker output must work on existing
69681 1. Set p_align to the minimum page size while laying out segments aligning
69682 to the maximum page size or section alignment. The run-time loader can
69683 align segments to the minimum page size or above, depending on system page
69685 2. If -z max-page-size=NNN is used, p_align will be set to the maximum
69686 page size or the largest section alignment.
69687 3. If a section requires alignment higher than the minimum page size,
69688 don't set p_align to the minimum page size.
69689 4. If a section requires alignment higher than the maximum page size,
69690 set p_align to the section alignment.
69691 5. For objcopy, when the minimum page size != the maximum page size,
69692 p_align may be set to the minimum page size while segments are aligned
69693 to the maximum page size. In this case, the input p_align will be
69694 ignored and the maximum page size will be used to align the ouput
69696 6. Update linker to disallow the common page size > the maximum page size.
69697 7. Update linker to avoid the common page size > the maximum page size.
69698 8. Adjust pru_irq_map-1.d to expect p_align == sh_addralign:
69701 [Nr] Name Type Addr Off Size ES Flg Lk Inf Al
69702 [ 0] NULL 00000000 000000 000000 00 0 0 0
69703 [ 1] .text PROGBITS 20000000 00007c 000004 00 AX 0 0 4
69706 Type Offset VirtAddr PhysAddr FileSiz MemSiz Flg Align
69707 LOAD 0x000074 0x00000000 0x00000000 0x00008 0x00008 RW 0x1
69708 LOAD 0x00007c 0x20000000 0x20000000 0x00004 0x00004 R E 0x4
69713 [Nr] Name Type Addr Off Size ES Flg Lk Inf Al
69714 [ 0] NULL 00000000 000000 000000 00 0 0 0
69715 [ 1] .text PROGBITS 20000000 00007c 000004 00 AX 0 0 4
69718 Type Offset VirtAddr PhysAddr FileSiz MemSiz Flg Align
69719 LOAD 0x000074 0x00000000 0x00000000 0x00008 0x00008 RW 0x1
69720 LOAD 0x00007c 0x20000000 0x20000000 0x00004 0x00004 R E 0x1
69722 To enable this linker optimization, the backend should define ELF_P_ALIGN
69723 to ELF_MINPAGESIZE.
69729 * elf-bfd.h (elf_backend_data): Add p_align.
69730 * elf.c (assign_file_positions_for_load_sections): Set p_align
69731 to the default p_align value while laying out segments aligning
69732 to maximum page size or section alignment.
69733 (elf_is_p_align_valid): New function.
69734 (copy_elf_program_header): Call elf_is_p_align_valid to determine
69735 if p_align is valid.
69736 * elfxx-target.h (ELF_P_ALIGN): New. Default to 0.
69737 (elfNN_bed): Add ELF_P_ALIGN.
69738 * elfxx-x86.h (ELF_P_ALIGN): New. Set to ELF_MINPAGESIZE.
69744 * bfdlink.h (bfd_link_info): Add maxpagesize_is_set.
69750 * emultempl/elf.em (gld${EMULATION_NAME}_handle_option): Set
69751 link_info.maxpagesize_is_set for -z max-page-size=NNN.
69752 * ldelf.c (ldelf_after_parse): Disallow link_info.commonpagesize
69753 > link_info.maxpagesize.
69754 * testsuite/ld-elf/elf.exp: Pass -z max-page-size=0x4000 to
69755 linker to build mbind2a and mbind2b.
69756 * testsuite/ld-elf/header.d: Add -z common-page-size=0x100.
69757 * testsuite/ld-elf/linux-x86.exp: Add PR ld/28689 tests.
69758 * testsuite/ld-elf/p_align-1.c: New file.
69759 * testsuite/ld-elf/page-size-1.d: New test.
69760 * testsuite/ld-elf/pr26936.d: Add -z common-page-size=0x1000.
69761 * testsuite/ld-elf/seg.d: Likewise.
69762 * testsuite/ld-scripts/rgn-at5.d: Likewise.
69763 * testsuite/ld-pru/pru_irq_map-1.d: Append 1 to name. Adjust
69764 expected PT_LOAD segment alignment.
69765 * testsuite/ld-pru/pru_irq_map-2.d: Append 2 to name.
69766 * testsuite/ld-scripts/pr23571.d: Add -z max-page-size=0x1000.
69768 2022-01-05 Alan Modra <amodra@gmail.com>
69770 Adjust quoted-sym-names test
69771 Some targets restrict symbol addresses in .text to instruction
69774 * testsuite/gas/all/quoted-sym-names.s: Define syms in .data.
69775 * testsuite/gas/all/quoted-sym-names.d: Adjust to suit.
69777 2022-01-05 Alan Modra <amodra@gmail.com>
69779 infinite recursion detected in gold testcase
69780 gold/testsuite/icf_test.cc:32:5: error: infinite recursion detected [-Werror=infinite-recursion]
69781 32 | int kept_func()
69784 * testsuite/icf_test.cc: Avoid infinite recursion error.
69786 2022-01-05 GDB Administrator <gdbadmin@sourceware.org>
69788 Automatic date update in version.in
69790 2022-01-04 H.J. Lu <hjl.tools@gmail.com>
69792 ld/x86: Update -z report-relative-reloc
69793 Use 0x%v, instead of bfd_sprintf_vma, to report relative relocations.
69794 Change linker relative relocations report from
69796 tmpdir/dump: R_X86_64_IRELATIVE (offset: 0x0000000000002000, info: 0x0000000000000025, addend: 0x0000000000001007) against 'ifunc' for section '.data.rel.ro.local' in tmpdir/report-reloc-1.o
69800 tmpdir/dump: R_X86_64_IRELATIVE (offset: 0x2000, info: 0x25, addend: 0x1007) against 'ifunc' for section '.data.rel.ro.local' in tmpdir/report-reloc-1.o
69804 * elfxx-x86.c (_bfd_x86_elf_link_report_relative_reloc): Use
69805 0x%v instead of bfd_sprintf_vma.
69809 * testsuite/ld-i386/report-reloc-1.l: Updated.
69810 * testsuite/ld-x86-64/report-reloc-1.l: Likewise.
69812 2022-01-04 H.J. Lu <hjl.tools@gmail.com>
69814 ld: Improve thin archive member error message
69815 Improve thin archive member error message with:
69817 ld: libbar.a(bar.o): error opening thin archive member: No such file or directory
69821 ld: libbar.a: error adding symbols: No such file or directory
69824 * archive.c (_bfd_get_elt_at_filepos): Add a pointer argument
69825 for struct bfd_link_info. Call linker callback when failing to
69826 open thin archive member.
69827 (_bfd_generic_get_elt_at_index): Pass NULL to
69828 _bfd_get_elt_at_filepos.
69829 (bfd_generic_openr_next_archived_file): Likewise.
69830 * coff-alpha.c (alpha_ecoff_get_elt_at_filepos): Add a pointer
69831 argument for struct bfd_link_info and pass it to
69832 _bfd_get_elt_at_filepos.
69833 (alpha_ecoff_openr_next_archived_file): Pass NULL to
69834 _bfd_get_elt_at_filepos.
69835 (alpha_ecoff_get_elt_at_index): Likewise.
69836 * coff-rs6000.c (_bfd_xcoff_openr_next_archived_file): Likewise.
69837 * ecoff.c (ecoff_link_add_archive_symbols): Pass info to
69838 backend->get_elt_at_filepos.
69839 * elflink.c (elf_link_is_defined_archive_symbol): info to
69840 _bfd_get_elt_at_filepos.
69841 * libbfd-in.h (_bfd_get_elt_at_filepos): Add a pointer argument
69842 for struct bfd_link_info.
69843 * libbfd.h: Regenerate.
69844 * libecoff.h (ecoff_backend_data): Add a pointer argument for
69845 struct bfd_link_info to get_elt_at_filepos.
69846 * linker.c (_bfd_generic_link_add_archive_symbols): Pass info to
69847 _bfd_get_elt_at_filepos.
69849 2022-01-04 Lancelot SIX <lancelot.six@amd.com>
69851 gdb/testsuite: fix inferior-clone.exp for native-extended-gdbserver
69852 003aae076207dbf32f98ba846158fc32669ef85f (gdb: Copy inferior properties
69853 in clone-inferior) introduced a testcase that fails when testing with
69854 the native-extended-gdbserver board:
69856 Running ../gdb/testsuite/gdb.base/inferior-clone.exp ...
69857 FAIL: gdb.base/inferior-clone.exp: inferior 2: clone-inferior
69858 FAIL: gdb.base/inferior-clone.exp: inferior 3: clone-inferior
69860 The error is as follows:
69864 Added inferior 2 on connection 1 (extended-remote localhost:2346)
69865 (gdb) FAIL: gdb.base/inferior-clone.exp: inferior 2: clone-inferior
69867 This fails because the testcase only expect the 'Added inferior 2' part
69868 of the message. The 'on connection 1 [...]' part is unexpected.
69870 Fix by adjusting the testcase to a account for the possible trailing
69871 part of the message.
69873 Tested on x86_64-linux with native-extende-gdbserver and unix boards.
69875 Change-Id: Ie3d6f04c9ffe9cab1fbda8ddf4935ee09b858c7a
69877 2022-01-04 Nick Clifton <nickc@redhat.com>
69879 Add ATTRIBUTE_UNUSED to load_build_id_debug_file()'s main_filename parameter.
69881 2022-01-04 Andrew Burgess <aburgess@redhat.com>
69883 gdb: don't pass nullptr to sigwait
69884 I tried building GDB on GNU/Hurd, and ran into this warning:
69886 gdbsupport/scoped_ignore_signal.h:78:16: error: null argument where non-null required (argument 2) [-Werror=nonnull]
69888 This is because in this commit:
69890 commit 99624310dd82542c389c89c2e55d8cae36bb74e1
69891 Date: Sun Jun 27 15:13:14 2021 -0400
69893 gdb: fall back on sigpending + sigwait if sigtimedwait is not available
69895 A call to sigwait was introduced that passes nullptr as the second
69896 argument, this call is only reached if sigtimedwait is not supported.
69898 The original patch was written for macOS, I assume on that target
69899 passing nullptr as the second argument is fine.
69901 On my GNU/Linux box, the man-page for sigwait doesn't mention that
69902 nullptr is allowed for the second argument, so my assumption would be
69903 that nullptr is not OK, and, if I change the '#ifdef
69904 HAVE_SIGTIMEDWAIT' introduced by the above patch to '#if 0', and
69905 rebuild on GNU/Linux, I see the same warning that I see on GNU/Hurd.
69907 I propose that we stop passing nullptr as the second argument to
69908 sigwait, and instead pass a valid int pointer. The value returned in
69909 the int can then be used in an assert.
69911 For testing, I (locally) made the change to the #ifdef I mentioned
69912 above, compiled GDB, and ran the usual tests, this meant I was using
69913 sigwait instead on sigtimedwait on GNU/Linux, I saw no regressions.
69915 2022-01-04 Nick Clifton <nickc@redhat.com>
69917 Remove a spurious debugging message.
69919 * dwarf.c (load_build_id_debug_file): Remove spurious printf.
69921 2022-01-04 Tom de Vries <tdevries@suse.de>
69923 [gdb/build] Fix build breaker in gdb/cli/cli-logging.c
69924 Fix build breaker in gdb/cli/cli-logging.c:
69926 gdb/cli/cli-logging.c: In function \
69927 ‘void show_logging_enabled(ui_file*, int, cmd_list_element*, const char*)’:
69928 gdb/gdbsupport/gdb_locale.h:28:28: error: cannot convert ‘char*’ to ‘ui_file*’
69929 28 | # define _(String) gettext (String)
69933 gdb/cli/cli-logging.c:202:25: note: in expansion of macro ‘_’
69934 202 | fprintf_unfiltered (_("on: Logging is enabled.\n"));
69938 Build and tested on x86_64-linux.
69940 Fixes: 45aec4e5ed8 ("[gdb/cli] Improve show logging output")
69942 2022-01-04 Jan Beulich <jbeulich@suse.com>
69944 x86/Intel: correct VFPCLASSP{S,D} handling when displacement is present
69945 fits_in_disp8() can be called before ambiguous operands get resolved
69946 or rejected (in process_suffix()), which requires that i.memshift be
69947 non-negative to avoid an internal error. This case wasn't covered by
69948 6c0946d0d28d ("x86: correct VFPCLASSP{S,D} operand size handling").
69950 2022-01-04 Jan Beulich <jbeulich@suse.com>
69952 gas: rework handling of backslashes in quoted symbol names
69953 Strange effects can result from the present handling, e.g.:
69959 yields first (correctly) "missing closing `"'" but then also "invalid
69960 character '\' in mnemonic" and further "end of file inside conditional".
69961 Symbols names ending in \ are in principle not expressable with that
69964 Instead of recording whether a backslash was seen, inspect the
69965 subsequent character right away. Only accept \\ (meaning a single
69966 backslash in the resulting symbol name) and \" (meaning an embedded
69967 double quote in the resulting symbol name) for now, warning about any
69970 While perhaps not necessary immediately, also permit concatenated
69971 strings to form a symbol name. This may become useful if going forward
69972 we would want to support \<octal> or \x<hex> sequences, where closing
69973 and re-opening quotes can be useful to delimit such sequences.
69975 The ELF "Multibyte symbol names" test gets switched away from using
69976 .set, as that would now also mean excluding nios2 and pru. By using
69977 .equiv instead, even the existing #notarget can be dropped. (For h8300
69978 the .section directive additionally needs attributes specified, to avoid
69979 a target specific warning.)
69981 2022-01-04 GDB Administrator <gdbadmin@sourceware.org>
69983 Automatic date update in version.in
69985 2022-01-03 Tom de Vries <tdevries@suse.de>
69987 [gdb/cli] Improve show logging output
69988 Before commit 3b6acaee895 "Update more calls to add_prefix_cmd" we had the
69989 following output for "show logging":
69991 $ gdb -q -batch -ex "set trace-commands on" \
69992 -ex "set logging off" \
69993 -ex "show logging" \
69994 -ex "set logging on" \
69998 Future logs will be written to gdb.txt.
69999 Logs will be appended to the log file.
70000 Output will be logged and displayed.
70001 Debug output will be logged and displayed.
70004 Currently logging to "gdb.txt".
70005 Logs will be appended to the log file.
70006 Output will be logged and displayed.
70007 Debug output will be logged and displayed.
70010 After that commit we have instead:
70014 debugredirect: The logging output mode is off.
70015 file: The current logfile is "gdb.txt".
70016 overwrite: Whether logging overwrites or appends to the log file is off.
70017 redirect: The logging output mode is off.
70020 debugredirect: The logging output mode is off.
70021 file: The current logfile is "gdb.txt".
70022 overwrite: Whether logging overwrites or appends to the log file is off.
70023 redirect: The logging output mode is off.
70025 which gives less clear output for some subcommands.
70027 OTOH, it's explicit about whether boolean values are on or off.
70029 The new text seems to have been chosen to match the set/show help texts:
70031 (gdb) help show logging
70032 Show logging options.
70034 List of show logging subcommands:
70036 show logging debugredirect -- Show the logging debug output mode.
70037 show logging file -- Show the current logfile.
70038 show logging overwrite -- \
70039 Show whether logging overwrites or appends to the log file.
70040 show logging redirect -- Show the logging output mode.
70043 Make the show logging messages more clear, while still keep the boolean
70044 values explicit, such that we have:
70046 $ ./gdb.sh -q -batch -ex "show logging"
70047 logging debugredirect: off: \
70048 Debug output will go to both the screen and the log file.
70049 logging enabled: off: Logging is disabled.
70050 logging file: The current logfile is "gdb.txt".
70051 logging overwrite: off: Logging appends to the log file.
70052 logging redirect: off: Output will go to both the screen and the log file.
70055 Tested on x86_64-linux.
70057 2022-01-03 Tom Tromey <tromey@adacore.com>
70059 Fix use of 'printf' in gdbtypes.c
70060 An earlier patch of mine, commit 64b7cc50 ("Remove
70061 gdb_print_host_address") inadvertently changed a function in
70062 gdbtypes.c to use printf rather than printf_filtered. This patch
70065 Fix regression in page-logging.exp
70066 Simon and Tom pointed out that page-logging.exp failed on their
70067 machines. Tom tracked this down to the "width" setting. Since
70068 there's no need in the test to change the width, it seems simplest to
70069 remove the setting. I confirmed that the test still fails if the fix
70070 is backed out, ensuring that the test is still testing what it
70073 Small indentation fix in eval.c
70074 I noticed that the AdaCore tree had a small divergence in eval.c -- it
70075 had a fix for an indentation problem in binop_promote. I'm checking
70076 in this small fix as obvious.
70078 2022-01-03 Tom de Vries <tdevries@suse.de>
70080 [gdb/testsuite] Handle for loop initial decl with gcc 4.8.5
70081 When running test-case gdb.threads/schedlock-thread-exit.exp on a system with
70082 system compiler gcc 4.8.5, I run into:
70084 src/gdb/testsuite/gdb.threads/schedlock-thread-exit.c:33:3: error: \
70085 'for' loop initial declarations are only allowed in C99 mode
70089 - using -std=c99, or
70090 - using -std=gnu99, in case that's required, or
70091 - in the case of the jit test-cases, rewriting the for loops.
70093 Tested on x86_64-linux, both with gcc 4.8.5 and gcc 7.5.0.
70095 2022-01-03 GDB Administrator <gdbadmin@sourceware.org>
70097 Automatic date update in version.in
70099 2022-01-02 Tom Tromey <tom@tromey.com>
70101 Update copying.awk for _initialize declaration patch
70102 Commit 6c265988 ("gdb: add back declarations for _initialize
70103 functions") modified copying.c, but not copying.awk. This patch
70104 updates copying.awk to backport the appropriate fix. This way, if
70105 copying.awk is run again, it will create the correct output.
70107 I'm checking this in as obvious.
70109 2022-01-02 Tom Tromey <tom@tromey.com>
70111 Use filtered output in print_i387_ext
70112 print_i387_ext mostly uses filtered output, but one call in the middle
70113 of the function uses the _unfiltered form. This patch fixes this
70114 call. I'm checking this in as obvious.
70116 2022-01-02 Alan Modra <amodra@gmail.com>
70118 Update year range in copyright notice of binutils files
70119 The result of running etc/update-copyright.py --this-year, fixing all
70120 the files whose mode is changed by the script, plus a build with
70121 --enable-maintainer-mode --enable-cgen-maint=yes, then checking
70122 out */po/*.pot which we don't update frequently.
70124 The copy of cgen was with commit d1dd5fcc38ead reverted as that commit
70125 breaks building of bfp opcodes files.
70127 2022-01-02 GDB Administrator <gdbadmin@sourceware.org>
70129 Automatic date update in version.in
70131 2022-01-01 Mike Frysinger <vapier@gentoo.org>
70133 gdb: copyright: fix a few comment typos
70135 sim: ppc: drop natural types
70136 These are almost entirely unused. For the very few places using them,
70137 replace with explicit signed types. This matches what was done in the
70140 sim: mips: clean up bad style/whitespace
70141 This doesn't fix all the problems, but grabs a bunch of the more
70144 sim: tweak copyright lines for gnulib update-copyright
70145 The regex it uses does not like so many leading spaces which causes
70146 it to think the files lack copyright. Trim them down so the script
70147 can find & update them accordingly.
70149 gdb: update sim mips testsuite copyright exemption
70150 The sim testsuite was reorganized last year, so update the path.
70152 2022-01-01 Mike Frysinger <vapier@gentoo.org>
70154 unify 64-bit bfd checks
70155 Move the 64-bit bfd logic out of bfd/configure.ac and into bfd64.m4
70156 under config so it can be shared between all the other subdirs.
70158 This replaces want64 with enable_64_bit_bfd which was already being
70159 declared, but not used directly.
70161 2022-01-01 Joel Brobecker <brobecker@adacore.com>
70163 Update Copyright year in gdb/testsuite/gdb.arch/powerpc-power10.exp
70164 This commit updates the copyright year range in the script
70165 gdb/testsuite/gdb.arch/powerpc-power10.exp. The update was
70166 performed by running gdb/copyright.py again, to make sure
70167 that the copyright year range will be automatically updated
70170 2022-01-01 Joel Brobecker <brobecker@adacore.com>
70172 Fix copyright header in gdb/testsuite/gdb.arch/powerpc-power10.exp
70173 The copyright year and holder line is slight malformed, missing
70174 a space after a comma, and this is sufficient for gdb's
70175 copyright.py script to miss this file during its automated
70176 copyright year update.
70178 This commit fixes this.
70180 2022-01-01 Joel Brobecker <brobecker@adacore.com>
70182 gdb/copyright.py: Add update-netbsd.sh to MULTIPLE_COPYRIGHT_HEADERS
70183 Add gdb/syscalls/update-netbsd.sh to the reminder printed
70184 at the end of the execution listing all the files where
70185 a manual update of the copyright header is needed. This
70186 scripts contains some inline code which includes a copyright
70189 Manual copyright year update of various GDB files
70190 This commit updates the copyright year in some files where
70191 we have a copyright year outside of the copyright year,
70192 and thus are not included in gdb's copyright.py script.
70194 2022-01-01 Joel Brobecker <brobecker@adacore.com>
70196 Automatic Copyright Year update after running gdb/copyright.py
70197 This commit brings all the changes made by running gdb/copyright.py
70198 as per GDB's Start of New Year Procedure.
70200 For the avoidance of doubt, all changes in this commits were
70201 performed by the script.
70203 2022-01-01 Joel Brobecker <brobecker@adacore.com>
70205 Update Copyright Year in gdb, gdbserver and gdbreplay version output
70206 This commit changes the copyright year printed by gdb, gdbserver
70207 and gdbreplay when printing the tool's version.
70209 2022-01-01 Alan Modra <amodra@gmail.com>
70211 ubsan: next_char_of_string signed integer overflow
70212 Squash another totally useless fuzz report that I should have ignored.
70214 * read.c (next_char_of_string): Avoid integer overflow.
70216 2022-01-01 Alan Modra <amodra@gmail.com>
70218 ubsan: bfd_mach_o_build_commands shift exponent 64 is too large
70219 * mach-o.c (bfd_mach_o_read_section_32): Limit alignment further.
70220 (bfd_mach_o_read_section_64): Likewise.
70222 2022-01-01 Alan Modra <amodra@gmail.com>
70224 ubsan: signed integer multiply overflow
70225 9223371018427387904 * 2 cannot be represented in type 'long', yes, but
70228 * expr.c (expr): Avoid signed overflow.
70230 2022-01-01 Alan Modra <amodra@gmail.com>
70232 asan: Null-dereference in _bfd_xcoff_copy_private_bfd_data
70233 sec->output_section will be NULL when objcopy removes sections.
70235 * coff-rs6000.c (_bfd_xcoff_copy_private_bfd_data): Protect against
70236 objcopy removing sections.
70238 2022-01-01 Alan Modra <amodra@gmail.com>
70240 ubsan: integer overflow in section filepos subtraction
70241 * elf.c (assign_file_positions_for_non_load_sections): Avoid
70242 signed integer overflow.
70244 2022-01-01 Alan Modra <amodra@gmail.com>
70246 Remove unnecessary ELF_MINPAGESIZE defines
70247 The idea of this patch is to make it easy to see which targets (just
70248 sparc) have ELF_MINPAGESIZE != ELF_COMMONPAGESIZE.
70250 * elf32-arm.c (ELF_MINPAGESIZE): Don't define.
70251 * elf32-metag.c: Likewise.
70252 * elfnn-aarch64.c: Likewise.
70253 * elf64-x86-64.c: Likewise. Also don't redefine a bunch of other
70254 macros for l1om elf64-target.h use that are unchanged from default.
70256 2022-01-01 H.J. Lu <hjl.tools@gmail.com>
70258 ld-x86-64: Pass options to linker with "-Wl,"
70259 * testsuite/ld-x86-64/x86-64.exp: Pass options to linker with
70262 2022-01-01 GDB Administrator <gdbadmin@sourceware.org>
70264 Automatic date update in version.in
70266 2021-12-31 Tom Tromey <tom@tromey.com>
70268 Do not call reinitialize_more_filter from avr_io_reg_read_command
70269 avr_io_reg_read_command is an ordinary gdb command, and so should not
70270 be calling reinitialize_more_filter. This patch removes it. I'm
70271 checking this in as obvious. Tested by rebuilding.
70273 2021-12-31 H.J. Lu <hjl.tools@gmail.com>
70275 x86: Define check_relocs_failed in elfxx-x86.h
70276 * elf32-i386.c (check_relocs_failed): Moved to ...
70277 * elfxx-x86.h (check_relocs_failed): Here. New.
70278 * elf64-x86-64.c (check_relocs_failed): Removed.
70280 Define X86_PCREL_TYPE_P/X86_SIZE_TYPE_P in elfxx-x86.h
70281 * elf32-i386.c: Don't include "elf/i386.h".
70282 (X86_PCREL_TYPE_P): Removed.
70283 (X86_SIZE_TYPE_P): Likewise.
70284 (elf_i386_check_relocs): Pass false to NEED_DYNAMIC_RELOCATION_P.
70285 (elf_i386_relocate_section): Pass false to
70286 GENERATE_DYNAMIC_RELOCATION_P and COPY_INPUT_RELOC_P.
70287 * elf64-x86-64.c: Don't include "elf/x86-64.h".
70288 (X86_PCREL_TYPE_P): Removed.
70289 (X86_SIZE_TYPE_P): Likewise.
70290 (elf_x86_64_check_relocs): Pass true to NEED_DYNAMIC_RELOCATION_P
70291 and X86_PCREL_TYPE_P.
70292 (elf_x86_64_relocate_section): Pass true to X86_PCREL_TYPE_P,
70293 X86_SIZE_TYPE_P, GENERATE_DYNAMIC_RELOCATION_P and
70294 COPY_INPUT_RELOC_P.
70295 * elfxx-x86.c: Don't include "elf/i386.h" nor "elf/x86-64.h".
70296 * elfxx-x86.h (X86_64_PCREL_TYPE_P): New.
70297 (I386_PCREL_TYPE_P): Likewise.
70298 (X86_PCREL_TYPE_P): Likewise.
70299 (X86_64_SIZE_TYPE_P): Likewise.
70300 (I386_SIZE_TYPE_P): Likewise.
70301 (X86_SIZE_TYPE_P): Likewise.
70302 (NEED_DYNAMIC_RELOCATION_P): Add IS_X86_64 and pass it to
70304 (COPY_INPUT_RELOC_P): Likewise.
70305 (GENERATE_DYNAMIC_RELOCATION_P): Add IS_X86_64, pass it to
70306 X86_PCREL_TYPE_P and X86_SIZE_TYPE_P.
70308 2021-12-31 Tamar Christina <tamar.christina@arm.com>
70310 ld: fix coff PE SEH
70311 COFF_WITH_pex64 and COFF_WITH_peAArch64 can't be true at the same time.
70312 That means that two conditionals that control the sorting of the .pdata section
70315 The testsuite doesn't catch this because the linker does the sorting and to link
70316 you require library support from the unwinder so we can't test from binutils in
70321 2021-12-31 Tamar Christina <tamar.christina@arm.com>
70324 * peXXigen.c: Fix conditional.
70326 2021-12-31 GDB Administrator <gdbadmin@sourceware.org>
70328 Automatic date update in version.in
70330 2021-12-30 GDB Administrator <gdbadmin@sourceware.org>
70332 Automatic date update in version.in
70334 2021-12-29 Tom Tromey <tom@tromey.com>
70336 Use filtered output in show callbacks
70337 "show" command callbacks, like most ordinary gdb commands, should use
70338 filtered output. I found a few that did not, so this patch changes
70339 them to use the filtered form.
70341 2021-12-29 Tom Tromey <tom@tromey.com>
70343 Consistently Use ui_file parameter to show callbacks
70344 I happened to notice that one "show" callback was printing to
70345 gdb_stdout rather than to the passed-in ui_file parameter. I went
70346 through all such callbacks and fixed them to consistently use the
70349 Regression tested on x86-64 Fedora 34.
70351 2021-12-29 Tom Tromey <tom@tromey.com>
70353 Use gdb_stdlog for MI debugging
70354 When MI debugging is enabled, the logging output should be sent to
70355 gdb_stdlog. This is part of PR gdb/7233.
70357 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=7233
70359 2021-12-29 Tom Tromey <tom@tromey.com>
70361 Use debug_prefixed_printf_cond_nofunc in index-cache
70362 This changes index-cache.c to use debug_prefixed_printf_cond_nofunc.
70363 As a side effect, logs are now written to gdb_stdlog. This is part of
70366 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=7233
70368 2021-12-29 Tom Tromey <tom@tromey.com>
70370 Send minsym logging to gdb_stdlog
70371 This changes minsyms.c to send logging output to gdb_stdlog. This is
70372 part of PR gdb/7233.
70374 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=7233
70376 2021-12-29 Tom Tromey <tom@tromey.com>
70378 Use gdb_stdlog for separate debug file logging
70379 This changes the separate debug file logging code (spread across two
70380 files) to use gdb_stdlog for its output. This is part of PR gdb/7233.
70382 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=7233
70384 2021-12-29 Tom Tromey <tom@tromey.com>
70386 Use debug_prefixed_printf_cond_nofunc in machoread
70387 This changes machoread.c to use debug_prefixed_printf_cond_nofunc. As
70388 a side effect, the logs are now written to gdb_stdlog. This is part
70391 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=7233
70393 2021-12-29 Tom Tromey <tom@tromey.com>
70395 Use debug_prefixed_printf_cond_nofunc in microblaze.c
70396 This changes microblaze.c to use the standard logging macro. As a
70397 side effect, logs will now go to gdb_stdlog. This is part of PR gdb/7233.
70399 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=7233
70401 2021-12-29 Tom Tromey <tom@tromey.com>
70403 Send debugging data to gdb_stdlog in mips-linux-nat.c
70404 This changes mips-linux-nat.c to send some logging output to
70405 gdb_stdlog, rather than stdout. This is part of PR gdb/7233.
70407 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=7233
70409 2021-12-29 Tom Tromey <tom@tromey.com>
70411 Send arch-utils error messages to gdb_stderr
70412 This changes arch-utils.c to send some error messages to gdb_stderr.
70413 This is part of PR gdb/7233.
70415 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=7233
70417 2021-12-29 Tom Tromey <tom@tromey.com>
70419 Use correct stream for process record output
70420 The process record code often emits unfiltered output. In some cases,
70421 this output ought to go to gdb_stderr (but see below). In other
70422 cases, the output is guarded by a logging variable and so ought to go
70423 to gdb_stdlog. This patch makes these changes.
70425 Note that in many cases, the output to stderr is followed by a
70426 "return -1", which is how process record indicates an error. It seems
70427 to me that calling error here would be preferable, because, in many
70428 cases, that's all the caller does when it sees a -1. However, I
70429 haven't made this change.
70431 This is part of PR gdb/7233.
70433 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=7233
70435 2021-12-29 Tom Tromey <tom@tromey.com>
70437 Send jit.c errors to gdb_stderr
70438 jit.c writes some error messages to gdb_stdout, but using gdb_stderr
70439 is better. This is part of PR gdb/7233.
70441 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=7233
70443 2021-12-29 Tom Tromey <tom@tromey.com>
70445 Fix logging redirection bug with pager
70446 I noticed yesterday that if gdb output is redirected to a file, the
70447 pager will still be active. This is irritating, because the output
70448 isn't actually visible -- just the pager prompt. Looking in bugzilla,
70449 I found that this had been filed 17 years ago, as PR cli/8798.
70451 This patch fixes the bug. It changes the pagination code to query the
70452 particular ui-file to see if paging is allowable. The ui-file
70453 implementations are changed so that only the stdout implementation and
70454 a tee (where one sub-file is stdout) can page.
70456 Regression tested on x86-64 Fedora 34.
70458 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=8798
70460 2021-12-29 Tom Tromey <tom@tromey.com>
70462 Remove unusual use of core_addr_eq and core_addr_hash
70463 gdbtypes.h uses core_addr_eq and core_addr_hash in a weird way: taking
70464 the address of a member and then passing this (as a void*) to these
70467 It seems better to simply inline the ordinary code here. CORE_ADDR is
70468 a scalar so it can be directly compared, and the identity hash
70469 function seems safe to assume as well.
70471 After this, core_addr_eq and core_addr_hash are unused, so this patch
70474 2021-12-29 Lancelot SIX <lancelot.six@amd.com>
70476 gdb: Copy inferior properties in clone-inferior
70477 This commit ensures that the following settings are cloned from one
70478 inferior to the new one when processing the clone-inferior command:
70480 - environment variables
70484 Some of those parameters can be passed as command line arguments to GDB
70485 (-args and -tty), so one could expect the clone-inferior to respect
70486 those flags. The following debugging session illustrates that:
70488 gdb -nx -quiet -batch \
70490 -ex "show inferior-tty" \
70491 -ex "clone-inferior" \
70494 -ex "show inferior-tty" \
70497 Argument list to give program being debugged when it is started is "foo bar".
70498 Terminal for future runs of program being debugged is "/some/tty".
70501 [Switching to inferior 2 [<null>] (/bin/echo)]
70502 Argument list to give program being debugged when it is started is "".
70503 Terminal for future runs of program being debugged is "".
70505 The other properties this commit copies on clone (i.e. CWD and the
70506 environment variables) are included since they are related (in the sense
70507 that they influence the runtime behavior of the program) even if they
70508 cannot be directly set using command line switches.
70510 There is a chance that this patch changes existing user workflow. I
70511 think that this change is mostly harmless. If users want to start a new
70512 inferior based on an existing one, they probably already propagate those
70513 settings to the new inferior in some way.
70515 Tested on x86_64-linux.
70517 Change-Id: I3b1f28b662f246228b37bb24c2ea1481567b363d
70519 2021-12-29 GDB Administrator <gdbadmin@sourceware.org>
70521 Automatic date update in version.in
70523 2021-12-28 H.J. Lu <hjl.tools@gmail.com>
70525 elf32-i386: Fix a typo in GOT comments
70526 Entry offsets in the global offset table are multiples of 4, not 8.
70528 * elf32-i386.c (elf_i386_relocate_section): Fix a typo in GOT
70531 2021-12-28 H.J. Lu <hjl.tools@gmail.com>
70533 bfd: Don't check non-thin archive member file size
70534 There is no need to check member file size for thin archive member.
70536 * bfdio.c (bfd_bread): Don't check non-thin archive member file
70539 2021-12-28 Alan Modra <amodra@gmail.com>
70542 In some cases, eg. riscv_pre_output_hook, gas generates out-of-order
70543 relocations. Various places in the linker assume relocs are sorted
70544 by increasing r_offset, which is normally the case. Provide
70545 GAS_SORT_RELOCS to handle unsorted relocs.
70549 * elf32-nds32.c (nds32_insertion_sort): Make static.
70550 * elf32-nds32.h (nds32_insertion_sort): Delete declaration.
70553 * write.c (write_relocs): Implement reloc sorting by r_offset
70554 when GAS_SORT_RELOCS.
70555 * config/tc-nds32.c (compar_relent, nds32_set_section_relocs): Delete.
70556 * config/tc-nds32.h (nds32_set_section_relocs): Don't declare.
70557 (SET_SECTION_RELOCS): Don't define.
70558 (GAS_SORT_RELOCS): Define.
70559 * config/tc-riscv.h (GAS_SORT_RELOCS): Define.
70561 2021-12-28 jiawei <jiawei@iscas.ac.cn>
70563 ld: Fix testcase errors due to -shared not support.
70564 Reviewed-by: Jim Wilson <jim.wilson.gcc@gmail.com>
70568 * testsuite/ld-ctf/ctf.exp: Add shared lib check.
70569 * testsuite/ld-plugin/lto.exp: Add lto shared check.
70571 2021-12-28 GDB Administrator <gdbadmin@sourceware.org>
70573 Automatic date update in version.in
70575 2021-12-27 H.J. Lu <hjl.tools@gmail.com>
70577 elf: Update comments for check_relocs in elf_backend_data
70580 commit 5c3261b0e834647cf9eb555320e20871b7854f98
70581 Author: H.J. Lu <hjl.tools@gmail.com>
70582 Date: Mon Oct 16 03:49:54 2017 -0700
70584 ELF: Call check_relocs after opening all inputs
70586 check_relocs is called after opening all inputs.
70588 * elf-bfd.h (elf_backend_data::check_relocs): Update comments.
70590 2021-12-27 H.J. Lu <hjl.tools@gmail.com>
70592 ld: Remove emultempl/linux.em
70593 Remove emultempl/linux.em whose last usage was removed by
70595 commit c65c21e1ffd1e02d9970a4bca0b7e384788a50f0
70596 Author: Alan Modra <amodra@gmail.com>
70597 Date: Mon Apr 16 22:14:01 2018 +0930
70599 various i386-aout and i386-coff target removal
70601 Also tidies some other aout leftovers in binutils-common.exp.
70603 2021-12-27 GDB Administrator <gdbadmin@sourceware.org>
70605 Automatic date update in version.in
70607 2021-12-26 GDB Administrator <gdbadmin@sourceware.org>
70609 Automatic date update in version.in
70611 2021-12-25 GDB Administrator <gdbadmin@sourceware.org>
70613 Automatic date update in version.in
70615 2021-12-24 Tom Tromey <tom@tromey.com>
70617 Remove gdb_print_host_address
70618 gdb_print_host_address is just a simple wrapper around
70619 fprintf_filtered. However, it is readily replaced in all callers by a
70620 combination of %s and call to host_address_to_string. This also
70621 simplifies the code, so I think it's worthwhile to remove this
70624 Regression tested on x86-64 Fedora 64.
70626 2021-12-24 Tom Tromey <tom@tromey.com>
70628 Move gdb_bfd_errmsg to gdb_bfd.c
70629 gdb_bfd.c contains most of gdb's BFD-related utility functions.
70630 However, gdb_bfd_errmsg is in utils.c. It seemed better to me to move
70631 this out of util.[ch] and into the BFD-related file instead.
70633 Tested by rebuilding.
70635 2021-12-24 Nelson Chu <nelson.chu@sifive.com>
70637 RISC-V: Rewrite the csr testcases.
70638 Maskray (Fangrui Song) had suggested me before that we should combine
70639 multiple testcases into one file as possible as we can. So that we can
70640 more easily understand what these test cases are testing, and easier to
70641 maintain. Therefore, this patch rewrites all csr testcases, to make them
70645 * testsuite/gas/riscv/csr-fail-nonexistent.d: Renamed from
70646 priv-reg-fail-nonexistent testcase.
70647 * testsuite/gas/riscv/csr-fail-nonexistent.: Likewise.
70648 * testsuite/gas/riscv/csr-fail-nonexistent.s: Likewise.
70649 * testsuite/gas/riscv/csr-insns-pseudo-noalias.d: Renamed from
70650 priv-reg-pseudo testcase.
70651 * testsuite/gas/riscv/csr-insns-pseudo.d: Likewise.
70652 * testsuite/gas/riscv/csr-insns-pseudo.s: Likewise.
70653 * testsuite/gas/riscv/csr-insns-read-only.d: Renamed from
70654 priv-reg-fail-read-only-02 testcase.
70655 * testsuite/gas/riscv/csr-insns-read-only.l: Likewise.
70656 * testsuite/gas/riscv/csr-insns-read-only.s: Likewise.
70657 * testsuite/gas/riscv/h-ext-32.d: Moved hypervisor csrs to csr.s.
70658 * testsuite/gas/riscv/h-ext-32.s: Likewise.
70659 * testsuite/gas/riscv/h-ext-64.d: Likewise.
70660 * testsuite/gas/riscv/h-ext-64.s: Likewise.
70661 * testsuite/gas/riscv/csr.s: Renamed from priv-reg.s, and then
70662 added the hypervisor csrs.
70663 * testsuite/gas/riscv/csr-version-1p9p1.d: The csr testcase when
70664 the privileged spec is 1.9.1. Also tested all invalid csr warnings
70665 when -mcsr-check is enabled.
70666 * testsuite/gas/riscv/csr-version-1p9p1.l: Likewise.
70667 * testsuite/gas/riscv/csr-version-1p10.d: Likewise, but the
70668 privileged spec is 1.10..
70669 * testsuite/gas/riscv/csr-version-1p10.l: Likewise.
70670 * testsuite/gas/riscv/csr-version-1p11.d: Likewise, but the
70671 privileged spec is 1.11.
70672 * testsuite/gas/riscv/csr-version-1p11.l: Likewise.
70673 * testsuite/gas/riscv/csr-version-1p12.d: Likewise, but the
70674 privileged spec is 1.12.
70675 * testsuite/gas/riscv/csr-version-1p12.l: Likewise.
70676 * testsuite/gas/riscv/priv-reg*: Removed or Renamed.
70678 2021-12-24 Vineet Gupta <vineetg@rivosinc.com>
70680 RISC-V: Hypervisor ext: support Privileged Spec 1.12
70681 This is the Hypervisor Extension 1.0
70683 - Hypervisor Memory-Management Instructions
70684 HFENCE.VVMA, HFENCE.GVMA,
70686 - Hypervisor Virtual Machine Load and Store Instructions
70687 HLV.B, HLV.BU, HSV.B,
70688 HLV.H, HLV.HU, HLVX.HU, HSB.H,
70689 HLV.W, HLV.WU, HLVX.WU, HSV.W,
70692 - Hypervisor CSRs (some new, some address changed)
70693 hstatus, hedeleg, hideleg, hie, hcounteren, hgeie, htval, hip, hvip,
70694 htinst, hgeip, henvcfg, henvcfgh, hgatp, hcontext, htimedelta, htimedeltah,
70695 vsstatus, vsie, vstvec, vsscratch, vsepc, vscause, vstval, vsip, vsatp,
70697 Note that following were added already as part of svinval extension
70699 HINVAL.GVMA, HINVAL.VVMA
70701 Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
70702 Reviewed-by: Nelson Chu <nelson.chu@sifive.com>
70705 * cpu-riscv.c (riscv_priv_specs): Added entry for 1.12.
70706 * cpu-riscv.h (enum riscv_spec_class): Added PRIV_SPEC_CLASS_1P12.
70708 * config/tc-riscv.c (abort_version): Updated comment.
70709 (validate_riscv_insn): Annotate switch-break.
70710 * testsuite/gas/riscv/h-ext-32.d: New testcase for hypervisor.
70711 * testsuite/gas/riscv/h-ext-32.s: Likewise.
70712 * testsuite/gas/riscv/h-ext-64.d: Likewise.
70713 * testsuite/gas/riscv/h-ext-64.s: Likewise.
70715 * opcode/riscv-opc.h: Added encodings for hypervisor csrs and
70718 * riscv-opc.c (riscv_opcodes): Added hypervisor instrcutions.
70720 2021-12-24 Vineet Gupta <vineetg@rivosinc.com>
70722 RISC-V: Hypervisor ext: drop Privileged Spec 1.9.1 implementation/tests
70723 This makes way for a clean 1.12 based Hypervisor Ext support.
70725 There are no known implementors of 1.9.1 H-ext. (Per Jim, kendryte k210
70726 is based on priv spec 1.9.1, but it seems unlikely that they implemented
70729 Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
70730 Reviewed-by: Nelson Chu <nelson.chu@sifive.com>
70733 * testsuite/gas/riscv/csr-dw-regnums.d: Drop the hypervisor csrs
70734 defined in the privileged spec 1.9.1.
70735 * testsuite/gas/riscv/csr-dw-regnums.s: Likewise.
70736 * testsuite/gas/riscv/priv-reg-fail-read-only-01.s: Likewise.
70737 * testsuite/gas/riscv/priv-reg-fail-version-1p10.l: Likewise.
70738 * testsuite/gas/riscv/priv-reg-fail-version-1p11.l: Likewise.
70739 * testsuite/gas/riscv/priv-reg-version-1p10.d: Likewise.
70740 * testsuite/gas/riscv/priv-reg-version-1p11.d: Likewise.
70741 * testsuite/gas/riscv/priv-reg-version-1p9p1.d: Likewise.
70742 * testsuite/gas/riscv/priv-reg.s: Likewise.
70744 * opcode/riscv-opc.h: Drop the hypervisor csrs defined in the
70745 privileged spec 1.9.1.
70747 2021-12-24 GDB Administrator <gdbadmin@sourceware.org>
70749 Automatic date update in version.in
70751 2021-12-23 Andrew Burgess <aburgess@redhat.com>
70753 gdb/testsuite: resolve some duplicate testnames in gdb.mi
70754 Set of fixes to resolve some duplicate test names in the gdb.mi/
70755 directory. There should be no real test changes after this set of
70756 fixes, they are all either:
70758 - Adding with_test_prefix type constructs to make test names unique,
70761 - Changing the test name to be more descriptive, or better reflect
70762 the test being run.
70764 2021-12-23 Andrew Burgess <andrew.burgess@embecosm.com>
70766 gdb/remote: handle attach when stop packet lacks thread-id
70767 Bug PR gdb/28405 reports a regression when using attach with an
70768 extended-remote target. In this case the target is not including a
70769 thread-id in the stop packet it sends back after the attach.
70771 The regression was introduced with this commit:
70773 commit 8f66807b98f7634c43149ea62e454ea8f877691d
70774 Date: Wed Jan 13 20:26:58 2021 -0500
70776 gdb: better handling of 'S' packets
70778 The problem is that when GDB processes the stop packet, it sees that
70779 there is no thread-id and so has to "guess" which thread the stop
70782 In this case the target only has one thread, so really, there's no
70783 guessing needed, but GDB still runs through the same process, this
70784 shouldn't cause us any problems.
70786 However, after the above commit, GDB now expects itself to be more
70787 internally consistent, specifically, only a thread that GDB thinks is
70788 resumed, can be a candidate for having stopped.
70790 It turns out that, when GDB attaches to a process through an
70791 extended-remote target, the threads of the process being attached too,
70792 are not, initially, marked as resumed.
70794 And so, when GDB tries to figure out which thread the stop might apply
70795 too, it finds no threads in the processes marked resumed, and so an
70798 In extended_remote_target::attach we create a new thread with a call
70799 to add_thread_silent, rather than remote_target::remote_add_thread,
70800 the reason is that calling the latter will result in a call to
70801 'add_thread' rather than 'add_thread_silent'. However,
70802 remote_target::remote_add_thread includes additional
70803 actions (i.e. calling remote_thread_info::set_resumed and set_running)
70804 which are missing from extended_remote_target::attach. These missing
70805 calls are what would serve to mark the new thread as resumed.
70807 In this commit I propose that we add an extra parameter to
70808 remote_target::remote_add_thread. This new parameter will force the
70809 new thread to be added with a call to add_thread_silent. We can now
70810 call remote_add_thread from the ::attach method, the extra
70811 actions (listed above) will now be performed, and the thread will be
70812 left in the correct state.
70814 Additionally, in PR gdb/28405, a segfault is reported. This segfault
70815 triggers when 'set debug remote 1' is used before trying to reproduce
70816 the original assertion failure. The cause of this is in
70817 remote_target::select_thread_for_ambiguous_stop_reply, where we do
70820 remote_debug_printf ("first resumed thread is %s",
70821 pid_to_str (first_resumed_thread->ptid).c_str ());
70822 remote_debug_printf ("is this guess ambiguous? = %d", ambiguous);
70824 gdb_assert (first_resumed_thread != nullptr);
70826 Notice that when debug printing is on we dereference
70827 first_resumed_thread before we assert that the pointer is not
70828 nullptr. This is the cause of the segfault, and is resolved by moving
70829 the assert before the debug printing code.
70831 I've extended an existing test, ext-attach.exp, so that the original
70832 test is run multiple times; we run in the original mode, as normal,
70833 but also, we now run with different packets disabled in gdbserver. In
70834 particular, disabling Tthread would trigger the assertion as it was
70835 reported in the original bug. I also run the test in all-stop and
70836 non-stop modes now for extra coverage, we also run the tests with
70837 target-async enabled, and disabled.
70839 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28405
70841 2021-12-23 Andrew Burgess <aburgess@redhat.com>
70843 gdb: on x86-64 non-trivial C++ objects are returned in memory
70844 Fixes PR gdb/28681. It was observed that after using the `finish`
70845 command an incorrect value was displayed in some cases. Specifically,
70846 this behaviour was observed on an x86-64 target.
70848 Consider this test program:
70875 And this GDB session:
70878 Reading symbols from ex.x...
70880 Breakpoint 1 at 0x401115: file ex.cc, line 14.
70882 Starting program: /home/andrew/tmp/ex.x
70884 Breakpoint 1, func (i=3) at ex.cc:14
70887 Run till exit from #0 func (i=3) at ex.cc:14
70888 main () at ex.cc:23
70890 Value returned is $1 = {
70899 Notice how after the `finish` the contents of $1 are junk, but, when I
70900 immediately ask for the value of `a`, I get back the correct value.
70902 The problem here is that after the finish command GDB calls the
70903 function amd64_return_value to figure out where the return value can
70904 be found (on x86-64 targets anyway).
70906 This function makes the wrong choice for the struct A in our case, as
70907 sizeof(A) <= 8, then amd64_return_value decides that A will be
70908 returned in a register. GDB then reads the return value register an
70909 interprets the contents as an instance of A.
70911 Unfortunately, A is not trivially copyable (due to its copy
70912 constructor), and the sys-v specification for argument and return
70913 value passing, says that any non-trivial C++ object should have space
70914 allocated for it by the caller, and the address of this space is
70915 passed to the callee as a hidden first argument. The callee should
70916 then return the address of this space as the return value.
70918 And so, the register that GDB is treating as containing an instance of
70919 A, actually contains the address of an instance of A (in this case on
70920 the stack), this is why GDB shows the incorrect result.
70922 The call stack within GDB for where we actually go wrong is this:
70926 amd64_classify_aggregate
70928 And it is in amd64_classify_aggregate that we should be classifying
70929 the type as AMD64_MEMORY, instead of as AMD64_INTEGER as we currently
70930 do (via a call to amd64_classify_aggregate_field).
70932 At the top of amd64_classify_aggregate we already have this logic:
70934 if (TYPE_LENGTH (type) > 16 || amd64_has_unaligned_fields (type))
70936 theclass[0] = theclass[1] = AMD64_MEMORY;
70940 Which handles some easy cases where we know a struct will be placed
70941 into memory, that is (a) the struct is more than 16-bytes in size,
70942 or (b) the struct has any unaligned fields.
70944 All we need then, is to add a check here to see if the struct is
70945 trivially copyable. If it is not then we know the struct will be
70948 I originally structured the code like this:
70950 if (TYPE_LENGTH (type) > 16
70951 || amd64_has_unaligned_fields (type)
70952 || !language_pass_by_reference (type).trivially_copyable)
70954 theclass[0] = theclass[1] = AMD64_MEMORY;
70958 This solved the example from the bug, and my small example above. So
70959 then I started adding some more extensive tests to the GDB testsuite,
70960 and I ran into a problem. I hit this error:
70962 gdbtypes.h:676: internal-error: loc_bitpos: Assertion `m_loc_kind == FIELD_LOC_KIND_BITPOS' failed.
70964 This problem is triggered from:
70966 amd64_classify_aggregate
70967 amd64_has_unaligned_fields
70970 Inside the unaligned field check we try to get the bit position of
70971 each field. Unfortunately, in some cases the field location is not
70972 FIELD_LOC_KIND_BITPOS, but is FIELD_LOC_KIND_DWARF_BLOCK.
70974 An example that shows this bug is:
70981 struct A : virtual public B
71006 It is the virtual base class, B, that causes the problem. The base
71007 class is represented, within GDB, as a field within A. However, the
71008 location type for this field is a DWARF_BLOCK.
71010 I spent a little time trying to figure out how to convert the
71011 DWARF_BLOCK to a BITPOS, however, I realised that, in this case at
71012 least, conversion is not needed.
71014 The C++ standard says that a class is not trivially copyable if it has
71015 any virtual base classes. And so, in this case, even if I could
71016 figure out the BITPOS for the virtual base class fields, I know for
71017 sure that I would immediately fail the trivially_copyable check. So,
71018 lets just reorder the checks in amd64_classify_aggregate to:
71020 if (TYPE_LENGTH (type) > 16
71021 || !language_pass_by_reference (type).trivially_copyable
71022 || amd64_has_unaligned_fields (type))
71024 theclass[0] = theclass[1] = AMD64_MEMORY;
71028 Now, if we have a class with virtual bases we will fail quicker, and
71029 avoid the unaligned fields check completely.
71031 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28681
71033 2021-12-23 Andrew Burgess <aburgess@redhat.com>
71035 gdb: make use of SCOPE_EXIT to manage thread executing state
71036 While working on another patch relating to how GDB manages threads
71037 executing and resumed state, I spotted the following code in
71040 executing = tp->executing ();
71041 set_executing (proc_target, inferior_ptid, false);
71043 id = null_frame_id;
71046 id = get_frame_id (get_current_frame ());
71048 catch (const gdb_exception &except)
71050 /* Restore the previous execution state. */
71051 set_executing (proc_target, inferior_ptid, executing);
71056 /* Restore the previous execution state. */
71057 set_executing (proc_target, inferior_ptid, executing);
71061 I notice that we only catch the exception so we can call
71062 set_executing, and this is the same call to set_executing that we need
71063 to perform in the non-exception return path.
71065 This would be much cleaner if we could use SCOPE_EXIT to avoid the
71066 try/catch, so lets do that.
71068 While cleaning this up, I also applied a similar patch to
71069 record-full.c, though there's no try/catch in that case, but using
71070 SCOPE_EXIT makes the code safe if, in the future, we do start throwing
71073 There should be no user visible changes after this commit.
71075 2021-12-23 GDB Administrator <gdbadmin@sourceware.org>
71077 Automatic date update in version.in
71079 2021-12-22 Andrew Burgess <aburgess@redhat.com>
71081 gdb/doc: add some index entries relating to mi-async setting
71082 I noticed that the mi-async setting was not referenced from the index
71083 in any way, this commit tries to rectify that a bit.
71085 The @cindex lines I think are not controversial, these same index
71086 entries are used elsewhere in the manual for async related topics (see
71087 @node Background Execution).
71089 The only bit that might be controversial is that I've added a @kindex
71090 entry for 'set mi-async' when the command is documented as '-gdb-set
71091 mi-async' (with a similar difference for the show/-gdb-show).
71093 My reasoning here is that nothing else is indexed under -gdb-set or
71094 -gdb-show, and as -gdb-set/-gdb-show are just the MI equivalent for
71095 set/show anything that is documented under set/show can be adjusted
71096 using -gdb-set/-gdbshow, and so, I've tried to keep the index
71097 consistent for mi-async.
71099 2021-12-22 Andrew Burgess <aburgess@redhat.com>
71101 gdb: convert 'set debug lin-lwp' to a boolean command
71102 Convert the 'set debug lin-lwp' command to a boolean. Adds a new
71103 LINUX_NAT_SCOPED_DEBUG_ENTER_EXIT macro, and makes use of it in one
71104 place (linux_nat_target::stop).
71106 The manual entry for 'set debug lin-lwp' is already vague about
71107 exactly what arguments this command takes, and the description talks
71108 about turning debug on and off, so I don't think there's any updates
71111 I have updated the doc strings shown when the users enters 'help show
71112 debug lin-lwp' or 'help show debug lin-lwp'. The old title lines used
71113 to talk about the 'GNU/Linux lwp module', but this debug flag is now
71114 used for any native linux target debug, so we now talk about
71115 'GNU/Linux native target'. The body string for this setting has been
71116 changed from 'Enables printf debugging output.' to 'When on, print
71117 debug messages relating to the GNU/Linux native target.', the old
71118 value looks like a cut&paste error to me.
71120 2021-12-22 Andrew Burgess <aburgess@redhat.com>
71122 gdb: add threads debugging switch
71125 set debug threads on|off
71128 Prints additional debug information relating to thread creation and
71131 GDB already announces when threads are created of course.... most of
71132 the time, but sometimes threads are added silently, in which case this
71133 debug message is the only mechanism to see the thread being added.
71134 Also, though GDB does announce when a thread exits, it doesn't
71135 announce when the thread object is deleted, I've added a debug message
71138 Additionally, having message printed through the debug system will
71139 cause the messages to be nested to an appropriate depth when other
71140 debug sub-systems are turned on (especially things like `infrun` and
71143 2021-12-22 jiawei <jiawei@iscas.ac.cn>
71145 RISC-V: Update Scalar Crypto testcases.
71146 Add opcodes in testcases to make sure every instruction generate
71147 right opcode after disassemble.
71151 * testsuite/gas/riscv/k-ext-64.d: Add opcode detect.
71152 * testsuite/gas/riscv/k-ext.d: Ditto.
71153 * testsuite/gas/riscv/zbkb-32.d: Ditto.
71154 * testsuite/gas/riscv/zbkb-64.d: Ditto.
71155 * testsuite/gas/riscv/zbkc-32.d: Ditto.
71156 * testsuite/gas/riscv/zbkc-64.d: Ditto.
71157 * testsuite/gas/riscv/zbkx-32.d: Ditto.
71158 * testsuite/gas/riscv/zbkx-64.d: Ditto.
71159 * testsuite/gas/riscv/zknd-32.d: Ditto.
71160 * testsuite/gas/riscv/zknd-64.d: Ditto.
71161 * testsuite/gas/riscv/zkne-32.d: Ditto.
71162 * testsuite/gas/riscv/zkne-64.d: Ditto.
71163 * testsuite/gas/riscv/zknh-32.d: Ditto.
71164 * testsuite/gas/riscv/zknh-64.d: Ditto.
71165 * testsuite/gas/riscv/zksed-32.d: Ditto.
71166 * testsuite/gas/riscv/zksed-64.d: Ditto.
71167 * testsuite/gas/riscv/zksh-32.d: Ditto.
71168 * testsuite/gas/riscv/zksh-64.d: Ditto.
71170 2021-12-22 Simon Marchi <simon.marchi@polymtl.ca>
71172 gdbarch-components.py: change empty "params" tuples to empty lists
71173 During review, it was suggested to change the "params" parameter from a
71174 tuple to a list, for esthetic reasons. The empty ones are still tuples
71175 though, they should probably be changed to be empty lists, for
71176 consistency. It does not change anything in the script result.
71178 Change-Id: If13c6c527aa167a5ee5b45740e5f1bda1e9517e4
71180 2021-12-22 GDB Administrator <gdbadmin@sourceware.org>
71182 Automatic date update in version.in
71184 2021-12-21 Luis Machado <luis.machado@linaro.org>
71186 [AArch64] Fix typo in error messages
71187 Fix mispelling of PROT_ME to PROT_MTE in the error messages.
71189 2021-12-21 Joel Sherrill <joel@rtems.org>
71191 Obsolete m32c-rtems and m32r-rtems
71192 2020-12-20 Joel Sherrill <joel@rtems.org>
71195 * config.bfd (m32c-*-rtems*): Remove target.
71198 * configure.tgt (m32c-*-rtems*): Remove target.
71199 * configure.tgt (m32r-*-rtems*): Remove target.
71201 2021-12-21 Jan Beulich <jbeulich@suse.com>
71203 x86: -mfence-as-lock-add=yes doesn't work for 16-bit mode
71204 Rather than trying to fix this (which would require making an assumption
71205 on the upper half of %esp being zero), simply issue an error. While at
71206 it, since the generated code is in conflict with -momit-lock-prefix=yes,
71207 issue an error in that case as well.
71209 gas/ELF: avoid below-base ref in obj_elf_parse_section_letters()
71210 We would better be prepared for 'm' being the first character of the
71213 2021-12-21 Alan Modra <amodra@gmail.com>
71215 Typo fixes in binutils doc
71216 * doc/binutils.texi: Fix typos.
71218 2021-12-21 GDB Administrator <gdbadmin@sourceware.org>
71220 Automatic date update in version.in
71222 2021-12-20 Tom Tromey <tom@tromey.com>
71224 Remove print_spaces
71225 This removes the print_spaces helper function, in favor of using the
71226 "*%s" idiom that's already used in many places in gdb. One spot (in
71227 symmisc.c) is changed to use print_spaces_filtered, because the rest
71228 of that function is using filtered output. (This highlights one way
71229 that the printf idiom is better -- this error is harder to make when
71232 Regression tested on x86-64 Fedora 34.
71234 2021-12-20 Tom Tromey <tom@tromey.com>
71237 I noticed that puts_debug isn't used in the tree. git log tells me
71238 that the last use was removed in 2015:
71240 commit 40e0b27177e747600d3ec186458fe0e482a1cf77
71241 Author: Pedro Alves <palves@redhat.com>
71242 Date: Mon Aug 24 15:40:26 2015 +0100
71244 Delete the remaining ROM monitor targets
71246 ... and this commit mentions that the code being removed here probably
71247 hadn't worked for 6 years prior to that.
71249 Based on this, I'm removing puts_debug. I don't think it's useful.
71250 Tested by rebuilding.
71252 2021-12-20 Tom Tromey <tom@tromey.com>
71254 Make n_spaces return a const char *
71255 n_spaces keeps the spaces in a static buffer. If a caller overwrites
71256 these, it may give an incorrect result to a subsequent caller. So,
71257 make the return type const to help avoid this outcome.
71259 2021-12-20 Enze Li <lienze2010@hotmail.com>
71261 Add Enze Li to gdb/MAINTAINERS
71263 2021-12-20 Joel Brobecker <brobecker@adacore.com>
71265 gdb/ada-exp.y: Reformat comment to follow GDB's coding standards
71266 This commit reformats a comment in gdb/ada-exp.y to avoid
71267 the leading '*' at the beginning of each line of the comment.
71269 gdb/ada-lang.h: Reformat comment to follow coding standards
71270 This commit reformats a comment in gdb/ada-lang.h to avoid
71271 the leading '*' at the beginning of each line of the comment.
71273 2021-12-20 GDB Administrator <gdbadmin@sourceware.org>
71275 Automatic date update in version.in
71277 2021-12-19 Alan Modra <amodra@gmail.com>
71279 Obsolete m32c-rtems
71281 readelf: avoid a possible divide by zero
71282 * readelf.c (process_section_headers): Check SHT_RELR entsize.
71284 2021-12-19 GDB Administrator <gdbadmin@sourceware.org>
71286 Automatic date update in version.in
71288 2021-12-18 Enze Li <lienze2010@hotmail.com>
71290 gdb: add "exit" command as an alias for "quit"
71291 This command adds the "exit" command as an alias for the "quit"
71292 command, as requested in PR gdb/28406.
71294 The documentation is also updated to mention this new command.
71296 Tested on x86_64-linux.
71298 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28406
71300 2021-12-18 Andrew Burgess <aburgess@redhat.com>
71302 gdb: add assert in remote_target::wait relating to async being off
71303 While working on another patch I ended up in a situation where I had
71304 async mode disabled (with 'maint set target-async off'), but the async
71305 event token got marked anyway.
71307 In this situation GDB was continually calling into
71308 remote_target::wait, however, the async token would never become
71309 unmarked as the unmarking is guarded by target_is_async_p.
71311 We could just unconditionally unmark the token, but that would feel
71312 like just ignoring a bug, so, instead, lets assert that if
71313 !target_is_async_p, then the async token should not be marked.
71315 This assertion would have caught my earlier mistake.
71317 There should be no user visible changes with this commit.
71319 2021-12-18 Andrew Burgess <aburgess@redhat.com>
71321 gdb/remote: some fixes for 'maint set target-async off'
71322 While working on another patch relating to remote targets, I wanted to
71323 test with 'maint set target-async off' in place. Unfortunately I ran
71324 into some problems. This commit is an attempt to fix one of the
71327 In my particular case I was actually running with:
71329 maint set target-async off
71330 maint set target-non-stop off
71332 that is, we're telling GDB to force the targets to operate in
71333 non-async mode, and in all-stop mode. Here's my GDB session showing
71336 (gdb) maintenance set target-async off
71337 (gdb) maintenance set target-non-stop off
71338 (gdb) target extended-remote :54321
71339 Remote debugging using :54321
71340 (gdb) attach 2365960
71341 Attaching to process 2365960
71342 No unwaited-for children left.
71345 Notice the 'No unwaited-for children left.' error, this is the
71346 problem. There's no reason why GDB should not be able to attach to
71349 The problem is this:
71351 1. The user runs 'attach PID' and this sends GDB into attach_command
71352 in infcmd.c. From here we call the ::attach method on the attach
71353 target, which will be the extended_remote_target.
71355 2. In extended_remote_target::attach, we attach to the remote target
71356 and get the first reply (which is a stop packet). We put off
71357 processing the stop packet until the end of ::attach. We setup the
71358 inferior and thread to represent the process we attached to, and
71359 download the target description. Finally, we process the initial
71362 If '!target_is_non_stop_p ()' and '!target_can_async_p ()', which is
71363 the case for us given the maintenance commands we used, we cache the
71364 stop packet within the remote_state::buf for later processing.
71366 3. Back in attach_command, if 'target_is_non_stop_p ()' then we
71367 request that the target stops. This will either process any cached
71368 stop replies, or request that the target stops, and process the stop
71369 replies. However, this code is not what we use due to non-stop mode
71370 being disabled. So, we skip to the next step which is to call
71371 validate_exec_file.
71373 4. Calling validate_exec_file can cause packets to be sent to the
71374 remote target, and replies received, the first path I hit is the
71375 call to target_pid_to_exec_file, which calls
71376 remote_target::pid_to_exec_file, which can then try to read the
71377 executable from the remote. Sending an receiving packets will make
71378 use of the remote_state::buf object.
71380 5. The attempt to attach continues, but the damage is already done...
71382 So, the problem is that, in step #2 we cache a stop reply in the
71383 remote_state::buf, and then in step #4 we reuse the remote_state::buf
71384 object, discarding any cached stop reply. As a result, the initial
71385 stop, which is sent when GDB first attaches to the target, is lost.
71387 This problem can clearly be seen, I feel, by looking at the
71388 remote_state::cached_wait_status flag. This flag tells GDB if there
71389 is a wait status cached in remote_state::buf. However, in
71390 remote_target::putpkt_binary and remote_target::getpkt_or_notif_sane_1
71391 this flag is just set back to 0, doing this immediately discards any
71394 I don't know if this scheme ever made sense, looking at commit
71395 2d717e4f8a54, where the cached_wait_status flag was added, it appears
71396 that there was nothing between where the stop was cached, and where
71397 the stop was consumed, so, I suspect, there never was a situation
71398 where we ended up in putpkt_binary or getpkt_or_notif_sane_1 and
71399 needed to clear to the flag, maybe the clearing was added "just in
71400 case". Whatever the history, I claim that this clearing this flag is
71401 no longer a good idea.
71403 So, my first step toward fixing this issue was to replace the two
71404 instances of 'rs->cached_wait_status = 0;' in ::putpkt_binary and
71405 ::getpkt_or_notif_sane_1 with 'gdb_assert (rs->cached_wait_status ==
71406 0);', this, at least would show me when GDB was doing something
71407 dangerous, and indeed, this assert is now hit in my test case above.
71409 I did play with using some kind of scoped restore to backup, and
71410 restore the remote_state::buf object in all the places within remote.c
71411 that I was hitting where the ::buf was being corrupted. The first
71412 problem with this is that, where the ::cached_wait_status flag is
71413 reset is _not_ where ::buf is corrupted. For the ::putpkt_binary
71414 case, by the time we get to the method the buffer has already been
71415 corrupted in many cases, so we end up needing to add the scoped
71416 save/restore within the callers, which means we need the save/restore
71417 in _lots_ of places.
71419 Plus, using this save/restore model feels like the wrong solution. I
71420 don't think that it's obvious that the buffer might be holding cached
71421 data, and I think it would be too easy for new corruptions of the
71422 buffer to be introduced, which could easily go unnoticed for a long
71425 So, I really wanted a solution that didn't require us to cache data in
71428 Luckily, I think we already have such a solution in place, the
71429 remote_state::stop_reply_queue, it seems like this does exactly the
71430 same task, just in a slightly different way. With the
71431 ::stop_reply_queue, the stop packets are processed upon receipt and
71432 the stop_reply object is added to the queue. With the ::buf cache
71433 solution, the unprocessed stop reply is cached in the ::buf, and
71436 So, finally, in this commit, I propose to remove the
71437 remote_state::cached_wait_status flag and to stop using the ::buf to
71438 cache stop replies. Instead, stop replies will now always be stored
71439 in the ::stop_reply_queue.
71441 There are two places where we use the ::buf to hold a cached stop
71442 reply, the first is in the ::attach method, and the second is in
71443 remote_target::start_remote, however, the second of these cases is far
71444 less problematic, as after caching the stop reply in ::buf we call the
71445 global start_remote function, which does very little work before
71446 calling normal_stop, which processes the cached stop reply. However,
71447 my plan is to switch both users over to using ::stop_reply_queue so
71448 that the old (unsafe) ::cached_wait_status mechanism can be completely
71451 The next problem is that the ::stop_reply_queue is currently only used
71452 for async-mode, and so, in remote_target::push_stop_reply, where we
71453 push stop_reply objects into the ::stop_reply_queue, we currently also
71454 mark the async event token. I've modified this so we only mark the
71455 async event token if 'target_is_async_p ()' - note, _is_, not _can_
71456 here. The ::push_stop_reply method is called in places where async
71457 mode has been temporarily disabled, but, when async mode is switched
71458 back on (see remote_target::async) we will mark the event token if
71459 there are events in the queue.
71461 Another change of interest is in remote_target::remote_interrupt_as.
71462 Previously this code checked ::cached_wait_status, but didn't check
71463 for events in the ::stop_reply_queue. Now that ::cached_wait_status
71464 has been removed we now check the queue length instead, which should
71465 have the same result.
71467 Finally, in remote_target::wait_as, I've tried to merge the processing
71468 of the ::stop_reply_queue with how we used to handle the
71469 ::cached_wait_status flag.
71471 Currently, when processing the ::stop_reply_queue we call
71472 process_stop_reply and immediately return. However, when handling
71473 ::cached_wait_status we run through the whole of ::wait_as, and return
71474 at the end of the function.
71476 If we consider a standard stop packet, the two differences I see are:
71478 1. Resetting of the remote_state::waiting_for_stop_reply, flag; this
71479 is not currently done when processing a stop from the
71480 ::stop_reply_queue.
71482 2. The final return value has the possibility of being adjusted at
71483 the end of ::wait_as, as well as there being calls to
71484 record_currthread, non of which are done if we process a stop from
71485 the ::stop_reply_queue.
71487 After discussion on the mailing list:
71489 https://sourceware.org/pipermail/gdb-patches/2021-December/184535.html
71491 it was suggested that, when an event is pushed into the
71492 ::stop_reply_queue, the ::waiting_for_stop_reply flag is never going
71493 to be set. As a result, we don't need to worry about the first
71494 difference. I have however, added a gdb_assert to validate the
71495 assumption that the flag is never going to be set. If in future the
71496 situation ever changes, then we should find out pretty quickly.
71498 As for the second difference, I have resolved this by having all stop
71499 packets taken from the ::stop_reply_queue, pass through the return
71500 value adjustment code at the end of ::wait_as.
71502 An example of a test that reveals the benefits of this commit is:
71505 RUNTESTFLAGS="--target_board=native-extended-gdbserver \
71506 GDBFLAGS='-ex maint\ set\ target-async\ off \
71507 -ex maint\ set\ target-non-stop\ off' \
71508 gdb.base/attach.exp"
71510 For testing I've been running test on x86-64/GNU Linux, and run with
71511 target boards unix, native-gdbserver, and native-extended-gdbserver.
71512 For each board I've run with the default GDBFLAGS, as well as with:
71514 GDBFLAGS='-ex maint\ set\ target-async\ off \
71515 -ex maint\ set\ target-non-stop\ off' \
71517 Though running with the above GDBFLAGS is clearly a lot more unstable
71518 both before and after my patch, I'm not seeing any consistent new
71519 failures with my patch, except, with the native-extended-gdbserver
71520 board, where I am seeing new failures, but only because more tests are
71521 now running. For that configuration alone I see the number of
71522 unresolved go down by 49, the number of passes goes up by 446, and the
71523 number of failures also increases by 144. All of the failures are new
71524 tests as far as I can tell.
71526 2021-12-18 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
71528 x86: Terminate mnemonicendp in swap_operand()
71529 Tested on x86_64-pc-linux-gnu.
71532 2021-12-17 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
71534 * i386-dis.c (swap_operand): Terminate mnemonicendp.
71537 2021-12-17 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
71539 * testsuite/gas/i386/opts-intel.d: Updated expected disassembly.
71540 * testsuite/gas/i386/opts.d: Likewise.
71541 * testsuite/gas/i386/sse2avx-opts-intel.d: Likewise.
71542 * testsuite/gas/i386/sse2avx-opts.d: Likewise.
71543 * testsuite/gas/i386/x86-64-opts-intel.d: Likewise.
71544 * testsuite/gas/i386/x86-64-opts.d: Likewise.
71545 * testsuite/gas/i386/x86-64-sse2avx-opts-intel.d: Likewise.
71546 * testsuite/gas/i386/x86-64-sse2avx-opts.d: Likewise.
71548 2021-12-18 GDB Administrator <gdbadmin@sourceware.org>
71550 Automatic date update in version.in
71552 2021-12-17 Tom Tromey <tom@tromey.com>
71554 Document gdbarch-components.py
71555 This adds a comment to document how to update gdbarch.
71558 This patch runs gdbarch.py and removes gdbarch.sh.
71560 2021-12-17 Simon Marchi <simon.marchi@efficios.com>
71562 Add new gdbarch generator
71563 The new gdbarch generator is a Python program. It reads the
71564 "components.py" that was created in the previous patch, and generates
71565 gdbarch.c and gdbarch-gen.h.
71567 This is a relatively straightforward translation of the existing .sh
71568 code. It doesn't try very hard to be idiomatic Python or to be
71571 It is, however, incredibly faster:
71573 $ time ./gdbarch.sh
71579 $ time ./gdbarch.py
71585 Co-Authored-By: Tom Tromey <tom@tromey.com>
71587 2021-12-17 Tom Tromey <tom@tromey.com>
71589 Generate new gdbarch-components.py from gdbarch.sh
71590 The new gdbarch.sh approach will be to edit a Python file, rather than
71591 adding a line to a certain part of gdbarch.sh. We use the existing sh
71592 code, though, to generate the first draft of this .py file.
71594 Documentation on the format will come in a subsequent patch.
71596 Note that some info (like "staticdefault") in the current code is
71597 actually unused, and so is ignored by this new generator.
71599 2021-12-17 Tom Tromey <tom@tromey.com>
71601 Do not sort the fields in gdbarch_dump
71602 This changes gdbarch.sh so that it no longer sorts the fields in
71603 gdbarch_dump. This sorting isn't done anywhere else by gdbarch.sh,
71604 and this simplifies the new generator a little bit.
71606 Do not generate gdbarch.h
71607 Now that gdbarch.h has been split, we no longer need the generator
71608 code in gdbarch.sh, so remove it.
71610 2021-12-17 Tom Tromey <tom@tromey.com>
71612 Split gdbarch.h into two files
71613 This patch splits gdbarch.h into two files -- gdbarch.h now is
71614 editable and hand-maintained, and the new gdbarch-gen.h file is the
71615 only thing generated by gdbarch.sh. This lets us avoid maintaining
71616 boilerplate in the gdbarch.sh file.
71618 Note that gdbarch.sh still generates gdbarch.h after this patch. This
71619 makes it easier to re-run when rebasing. This code is removed in a
71622 2021-12-17 Tom Tromey <tom@tromey.com>
71624 Move ordinary gdbarch code to arch-utils
71625 While I think it makes sense to generate gdbarch.c, at the same time I
71626 think it is better for ordinary code to be editable in a C file -- not
71627 as a hunk of C code embedded in the generator.
71629 This patch moves this sort of code out of gdbarch.sh and gdbarch.c and
71630 into arch-utils.c, then has arch-utils.c include gdbarch.c.
71632 2021-12-17 Maciej W. Rozycki <macro@embecosm.com>
71634 Avoid redundant operations in `fortran_array_walker'
71635 Move inner dimension's element type determination outside the respective
71636 loops in `fortran_array_walker'. The operation is exactly the same with
71637 each iteration, so there is no point in redoing it for each element and
71638 while a smart compiler might be able to move it outside the loop it is
71639 regardless a bad coding style. No functional change.
71641 Initialize `m_ndimensions' in the member initializer list
71642 Following our coding convention initialize the `m_ndimensions' member in
71643 the member initializer list rather than in the body of the constructor
71644 of the `fortran_array_walker' class. No functional change.
71646 2021-12-17 Lancelot SIX <lancelot.six@amd.com>
71647 Pedro Alves <pedro@palves.net>
71649 gdb/tui: install SIGWINCH only when connected to a TTY
71650 PR26056 reports that when GDB is connected to non-TTY stdin/stdout, it
71651 crashes when it receives a SIGWINCH signal.
71653 This can be reproduced as follows:
71655 $ gdb/gdb -nx -batch -ex 'run' --args sleep 60 </dev/null 2>&1 | cat
71657 # from another terminal:
71658 $ kill -WINCH %(pidof gdb)
71660 When doing so, the process crashes in a call to rl_resize_terminal:
71663 rl_resize_terminal (void)
71665 _rl_get_screen_size (fileno (rl_instream), 1);
71669 The problem is that at this point rl_instream has the value NULL.
71671 The rl_instream variable is supposed to be initialized during a call to
71672 readline_initialize_everything, which in a normal startup sequence is
71673 called under this call chain:
71676 tui_ensure_readline_initialized
71678 readline_initialize_everything
71680 In tui_interp::init, we have the following sequence:
71682 tui_initialize_io ();
71683 tui_initialize_win (); // <- Installs SIGWINCH
71684 if (gdb_stdout->isatty ())
71685 tui_ensure_readline_initialized (); // <- Initializes rl_instream
71687 This function unconditionally installs the SIGWINCH signal handler (this
71688 is done by tui_initialize_win), and then if gdb_stdout is a TTY it
71689 initializes readline. Therefore, if stdout is not a TTY, SIGWINCH is
71690 installed but readline is not initialized. In such situation
71691 rl_instream stays NULL, and when GDB receives a SIGWINCH it calls its
71692 handler and in fine tries to access rl_instream leading to the crash.
71694 This patch proposes to fix this issue by installing the SIGWINCH signal
71695 handler only if GDB is connected to a TTY. Given that this
71696 initialization it the only task of tui_initialize_win, this patch moves
71697 tui_initialize_win just after the call to
71698 tui_ensure_readline_initialized.
71700 Tested on x86_64-linux.
71702 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=26056
71703 Change-Id: I6458acef7b0d9beda2a10715d0345f02361076d9
71705 2021-12-17 Alan Modra <amodra@gmail.com>
71707 asan: NULL dereference in bfd_elf_set_group_contents
71708 * elf-bfd.h (struct output_elf_obj_tdata): Make num_section_syms
71710 * elf.c (bfd_elf_set_group_contents): Bounds check sec->index
71711 and check that entry in elf_section_syms for sec is non-NULL.
71712 (_bfd_elf_symbol_from_bfd_symbol): Adjust.
71714 2021-12-17 Alan Modra <amodra@gmail.com>
71716 asan: use after free in _bfd_elf_mips_get_relocated_section_contents
71717 Leaving entries on mips_hi16_list from a previous pass over relocs
71718 leads to confusing bugs.
71720 * elfxx-mips.c (_bfd_elf_mips_get_relocated_section_contents):
71721 Free mips_hi16_list entries on error exit.
71723 2021-12-17 Alan Modra <amodra@gmail.com>
71725 asan: abort in wasm_scan_name_function_section
71726 Macros like READ_LEB128 in wasm-module.c that alter control flow are
71727 evil. Maintainers will break your code if you have hidden ways to
71730 * wasm-module.c (wasm_scan_name_function_section): Don't
71731 attempt to bfd_release NULL.
71733 2021-12-17 Alan Modra <amodra@gmail.com>
71735 asan: heap-buffer-overflow in bpf_elf_generic_reloc
71736 The bpf reloc howtos are a bit weird, using bitpos to specify an
71737 offset from r_offset that is outside the size of the reloc as given by
71738 howto.size. That means bfd_get_reloc_size gives the wrong answer for
71739 range checking, and thus bfd_reloc_offset_in_range can't be used.
71741 * elf64-bpf.c (bpf_elf_generic_reloc): Handle bitpos offset reloc
71744 2021-12-17 Alan Modra <amodra@gmail.com>
71746 ubsan: bfd.c:2519:8: shift exponent 34 is too large
71747 * bfd.c (bfd_update_compression_header): Avoid integer overflow.
71749 asan: buffer overflow in mmo_get_symbols
71750 * mmo.c (mmo_get_symbols): Error on symbol name exceeding max length.
71752 2021-12-17 Alan Modra <amodra@gmail.com>
71754 asan: buffer overflow in elfnn-aarch64.c get_plt_type
71755 We can't assume .dynamic is a multiple of ElfNN_External_Dyn, at least
71756 not when presented with fuzzed object files.
71758 * elfnn-aarch64.c (get_plt_type): Don't access past end of
71759 improperly sized .dynamic.
71761 2021-12-17 Alan Modra <amodra@gmail.com>
71763 try_build_id_prefix gcc-10 -Wformat-security errors
71764 dwarf.c:11300:3: error: format not a string literal and no format arguments [-Werror=format-security]
71765 11300 | f += sprintf (f, prefix);
71768 * dwarf.c (try_build_id_prefix): Avoid -Wformat-security error.
71770 2021-12-17 GDB Administrator <gdbadmin@sourceware.org>
71772 Automatic date update in version.in
71774 2021-12-16 Nick Clifton <nickc@redhat.com>
71776 Fix AVR assembler so that it creates relocs that will work with linker relaxation.
71778 gas * config/tc-avr.h (tc_fix_adjustable): Define.
71779 * config/tc-avr.c (avr_fix_adjustable): New function.
71780 * testsuite/gas/all/gas.exp: Skip tests that need adjustable fixups.
71781 * testsuite/gas/elf/elf.exp: Likewise.
71782 * testsuite/gas/avr/diffreloc_withrelax.d: Adjust expected output.
71783 * testsuite/gas/avr/pc-relative-reloc.d: Adjust expected output.
71785 ld * testsuite/ld-avr/avr-prop-7.d: Adjust expected output.
71786 * testsuite/ld-avr/avr-prop-8.d: Likewise.
71787 * testsuite/ld-avr/pr13402.d: Likewise.
71789 2021-12-16 Nick Clifton <nickc@redhat.com>
71791 When loading separate debug info files, also attempt to locate a file based upon the build-id.
71793 * dwarf.c (load_build_id_debug_file): New function.
71794 (try_build_id_prefix): New function.
71795 (check_for_and_load_links): Call load_build_id_debug_file.
71796 (debug_displays): Add entry for .note.gnu.build-id.
71797 * dwarf.h (enum dwarf_section_display_enum): Add
71799 * testsuite/binutils-all/debuginfod.exp (test_fetch_debuglink):
71800 Fix regexp for loads via debuglink section.
71802 2021-12-16 Richard Sandiford <richard.sandiford@arm.com>
71804 arm: Add support for Armv9.1-A to Armv9.3-A
71805 This patch adds AArch32 support for -march=armv9.[123]-a.
71806 The behaviour of the new options can be expressed using a
71807 combination of existing feature flags and tables.
71809 The cpu_arch_ver entries for ARM_ARCH_V9_2A and ARM_ARCH_V9_3A
71810 are technically redundant but it seemed less surprising to include
71814 * opcode/arm.h (ARM_ARCH_V9_1A, ARM_ARCH_V9_2A): New macros.
71815 (ARM_ARCH_V9_3A): Likewise.
71818 * doc/c-arm.texi: Add armv9.1-a, armv9.2-a and armv9.3-a.
71819 * config/tc-arm.c (armv91a_ext_table, armv92a_ext_table): New macros.
71820 (armv93a_ext_table): Likewise.
71821 (arm_archs): Add armv9.1-a, armv9.2-a and armv9.3-a.
71822 (cpu_arch_ver): Add ARM_ARCH_V9_1A, ARM_ARCH_V9_2A and ARM_ARCH_V9_3A.
71823 * NEWS: Mention the above.
71824 * testsuite/gas/arm/attr-march-armv9_1-a.d: New test.
71825 * testsuite/gas/arm/attr-march-armv9_2-a.d: Likewise.
71826 * testsuite/gas/arm/attr-march-armv9_3-a.d: Likewise.
71827 * testsuite/gas/arm/bfloat16-armv9.1-a.d: Likewise.
71828 * testsuite/gas/arm/bfloat16-armv9.2-a.d: Likewise.
71829 * testsuite/gas/arm/bfloat16-armv9.3-a.d: Likewise.
71830 * testsuite/gas/arm/i8mm-armv9.1-a.d: Likewise.
71831 * testsuite/gas/arm/i8mm-armv9.2-a.d: Likewise.
71832 * testsuite/gas/arm/i8mm-armv9.3-a.d: Likewise.
71834 2021-12-16 Richard Sandiford <richard.sandiford@arm.com>
71836 arm: Add support for Armv8.7-A and Armv8.8-A
71837 This patch adds AArch32 support for -march=armv8.[78]-a.
71838 The behaviour of the new options can be expressed using a
71839 combination of existing feature flags and tables.
71841 The cpu_arch_ver entries are technically redundant but
71842 it seemed less surprising to include them anyway.
71845 * opcode/arm.h (ARM_ARCH_V8_7A, ARM_ARCH_V8_8A): New macros.
71848 * doc/c-arm.texi: Add armv8.7-a and armv8.8-a.
71849 * config/tc-arm.c (armv87a_ext_table, armv88a_ext_table): New macros.
71850 (arm_archs): Add armv8.7-a and armv8.8-a.
71851 (cpu_arch_ver): Add ARM_ARCH_V8_7A and ARM_ARCH_V8_8A.
71852 * NEWS: Mention the above.
71853 * testsuite/gas/arm/attr-march-armv8_7-a.d: New test.
71854 * testsuite/gas/arm/attr-march-armv8_8-a.d: Likewise.
71855 * testsuite/gas/arm/bfloat16-armv8.7-a.d: Likewise.
71856 * testsuite/gas/arm/bfloat16-armv8.8-a.d: Likewise.
71857 * testsuite/gas/arm/i8mm-armv8.7-a.d: Likewise.
71858 * testsuite/gas/arm/i8mm-armv8.8-a.d: Likewise.
71860 2021-12-16 Richard Sandiford <richard.sandiford@arm.com>
71862 aarch64: Add support for Armv9.1-A to Armv9.3-A
71863 This patch adds AArch64 support for -march=armv9.[123]-a.
71864 The behaviour of the new options can be expressed using a
71865 combination of existing feature flags, so we don't need to
71866 eat into the vanishing number of spare AARCH64_FEATURE_* bits.
71867 Hoewver, it was more convenient to separate out the |s of
71868 feature flags so that Armv9.1-A could reuse the set for
71869 Armv8.6-A, and so on.
71872 * opcode/aarch64.h (AARCH64_ARCH_V8_FEATURES): New macro,
71874 (AARCH64_ARCH_V8): ...here.
71875 (AARCH64_ARCH_V8_1_FEATURES): New macro, split out from...
71876 (AARCH64_ARCH_V8_1): ...here.
71877 (AARCH64_ARCH_V8_2_FEATURES): New macro, split out from...
71878 (AARCH64_ARCH_V8_2): ...here.
71879 (AARCH64_ARCH_V8_3_FEATURES): New macro, split out from...
71880 (AARCH64_ARCH_V8_3): ...here.
71881 (AARCH64_ARCH_V8_4_FEATURES): New macro, split out from...
71882 (AARCH64_ARCH_V8_4): ...here.
71883 (AARCH64_ARCH_V8_5_FEATURES): New macro, split out from...
71884 (AARCH64_ARCH_V8_5): ...here.
71885 (AARCH64_ARCH_V8_6_FEATURES): New macro, split out from...
71886 (AARCH64_ARCH_V8_6): ...here.
71887 (AARCH64_ARCH_V8_7_FEATURES): New macro, split out from...
71888 (AARCH64_ARCH_V8_7): ...here.
71889 (AARCH64_ARCH_V8_8_FEATURES): New macro, split out from...
71890 (AARCH64_ARCH_V8_8): ...here.
71891 (AARCH64_ARCH_V9_FEATURES): New macro, split out from...
71892 (AARCH64_ARCH_V9): ...here.
71893 (AARCH64_ARCH_V9_1_FEATURES, AARCH64_ARCH_V9_1): New macros.
71894 (AARCH64_ARCH_V9_2_FEATURES, AARCH64_ARCH_V9_2): New macros.
71895 (AARCH64_ARCH_V9_3_FEATURES, AARCH64_ARCH_V9_3): New macros.
71898 * doc/c-aarch64.texi: Add armv9.1-a, armv9-2-a and armv9.3-a.
71899 * config/tc-aarch64.c (aarch64_archs): Likewise.
71900 * NEWS: Mention the above.
71901 * testsuite/gas/aarch64/armv9_invalid.d,
71902 testsuite/gas/aarch64/armv9_invalid.s,
71903 testsuite/gas/aarch64/armv9_invalid.l: New test.
71904 * testsuite/gas/aarch64/armv9_1.d,
71905 testsuite/gas/aarch64/armv9_1.s: Likewise.
71906 * testsuite/gas/aarch64/armv9_1_invalid.d,
71907 testsuite/gas/aarch64/armv9_1_invalid.s,
71908 testsuite/gas/aarch64/armv9_1_invalid.l: Likewise.
71909 * testsuite/gas/aarch64/armv9_2.d,
71910 testsuite/gas/aarch64/armv9_2.s: Likewise.
71911 * testsuite/gas/aarch64/armv9_2_invalid.d,
71912 testsuite/gas/aarch64/armv9_2_invalid.s,
71913 testsuite/gas/aarch64/armv9_2_invalid.l: Likewise.
71914 * testsuite/gas/aarch64/armv9_3.d,
71915 testsuite/gas/aarch64/armv9_3.s: Likewise.
71917 2021-12-16 Nelson Chu <nelson.chu@sifive.com>
71919 RISC-V: Support svinval extension with frozen version 1.0.
71920 According to the privileged spec, there are five new instructions for
71921 svinval extension. Two of them (HINVAL.VVMA and HINVAL.GVMA) need to
71922 enable the hypervisor extension. But there is no implementation of
71923 hypervisor extension in mainline for now, so let's consider the related
71926 31..25 24..20 19..15 14..12 11...7 6..2 1..0
71927 sinval.vma 0001011 rs2 rs1 000 00000 11100 11
71928 sfence.w.inval 0001100 00000 00000 000 00000 11100 11
71929 sfence.inval.ir 0001100 00001 00000 000 00000 11100 11
71930 hinval.vvma 0010011 rs2 rs1 000 00000 11100 11
71931 hinval.gvma 0110011 rs2 rs1 000 00000 11100 11
71933 This patch is cherry-picked from the riscv integration branch since the
71934 svinval extension is frozen for now. Besides, we fix the funct7 encodings
71935 of hinval.vvma and hinval.gvma, from 0x0011011 and 0x0111011 to 0x0010011
71939 * elfxx-riscv.c (riscv_supported_std_s_ext): Added svinval.
71940 (riscv_multi_subset_supports): Handle INSN_CLASS_SVINVAL.
71942 * testsuite/gas/riscv/svinval.d: New testcase.
71943 * testsuite/gas/riscv/svinval.s: Likewise.
71945 * opcode/riscv-opc.h: Added encodings for svinval.
71946 * opcode/riscv.h (enum riscv_insn_class): Added INSN_CLASS_SVINVAL.
71948 * riscv-opc.c (riscv_opcodes): Added svinval instructions.
71950 2021-12-16 Mike Frysinger <vapier@gentoo.org>
71952 sim: mips/or1k: drop redundant arg to bitsize macro
71953 These are just using the default behavior for the 3rd arg, so drop
71954 it to make it more clear. This also makes them match all other
71955 ports that only use the first 2 arguments.
71957 2021-12-16 Mike Frysinger <vapier@gentoo.org>
71959 bfd: unify texi generation rules
71960 The logic between these rules are extremely similar, so unify them
71961 into a single variable by leveraging make $@ and $< variables.
71963 Also add automake silent rule support while we're here.
71965 2021-12-16 Mike Frysinger <vapier@gentoo.org>
71967 sim: fix mingw builds with replacement gnulib open
71968 The header shuffling in here broke the workaround for gnulib defining
71969 "open". Move it back before the sim-specific includes to fix. This
71970 is because the callback struct in the headers has an "open" member and
71971 this file tries to call that.
71973 2021-12-16 Sandra Loosemore <sandra@codesourcery.com>
71975 Adjust compare_link_order for unstable qsort
71976 In a cross toolchain for nios2-elf target and x86_64-w64-mingw32 host
71977 using binutils 2.37, we observed a failure that didn't show up on
71978 x86_64-linux-gnu host: testcase pr25490-5.s was failing with
71980 C:\path\to\nios2-elf-ld.exe: looping in map_segments
71981 FAIL: __patchable_function_entries section 5
71983 * ldelfgen.c (compare_link_order): Don't use section id in
71984 sorting. Keep original ordering instead. Update comments.
71986 2021-12-16 Alan Modra <amodra@gmail.com>
71988 Re: Fix an undefined behaviour in the BFD library's DWARF parser
71989 Using an unsigned int cast (to 32 bits) on a pointer difference (of
71990 possibly 64 bits) is wrong. Even though it will work on all real
71991 object files, the fuzzers will eventually find this hole.
71994 * dwarf1.c (parse_die): Cast pointer difference to size_t.
71995 Catch another possible pointer overflow.
71997 2021-12-16 Simon Marchi <simon.marchi@polymtl.ca>
71999 gdb: re-format with black 21.12b0
72000 Run black 21.12b0 on gdb/, there is a single whitespace change. I will
72001 update the wiki [1] in parallel to bump the version of black to 21.12b0.
72003 [1] https://sourceware.org/gdb/wiki/Internals%20GDB-Python-Coding-Standards
72005 Change-Id: Ib3b859e3506c74a4f15d16f1e44ef402de3b98e2
72007 2021-12-16 Simon Marchi <simon.marchi@polymtl.ca>
72009 gdb: re-format with black 21.9b0
72010 Run black 21.9b0 on gdb/ (this is the version currently mentioned on the
72011 wiki [1], the subsequent commit will bump that version).
72013 [1] https://sourceware.org/gdb/wiki/Internals%20GDB-Python-Coding-Standards
72015 Change-Id: I5ceaab42c42428e053e2572df172aa42a88f0f86
72017 2021-12-16 GDB Administrator <gdbadmin@sourceware.org>
72019 Automatic date update in version.in
72021 2021-12-15 Alan Modra <amodra@gmail.com>
72023 PR28691, validate dwarf attribute form
72024 PR28691 is a fuzzing PR that triggers a non-problem of "output changes
72025 per run" with PIEs and/or different compilers. I've closed similar
72026 PRs before as wontfix, but I guess there will be no end of this type
72027 of PR. The trigger is an attribute that usually takes one of the
72028 offset/constant reference DW_FORMs being given an indexed string
72029 DW_FORM. The bfd reader doesn't support indexed strings and returns
72030 an error string instead. The address of the string varies with PIE
72031 runs and/or compiler, and we allow that address to appear in output.
72032 Fix this by validating integer attribute forms, as we do for string
72036 * dwarf2.c (is_str_attr): Rename to..
72037 (is_str_form): ..this. Change param type. Update calls.
72038 (is_int_form): New function.
72039 (read_attribute_value): Handle DW_FORM_addrx2.
72040 (find_abstract_instance): Validate form when using attr.u.val.
72041 (scan_unit_for_symbols, parse_comp_unit): Likewise.
72043 2021-12-15 Luis Machado <luis.machado@linaro.org>
72045 New --enable-threading configure option to control use of threads in GDB/GDBserver
72046 Add the --enable-threading configure option so multithreading can be disabled
72047 at configure time. This is useful for statically-linked builds of
72048 GDB/GDBserver, since the thread library doesn't play well with that setup.
72050 If you try to run a statically-linked GDB built with threading, it will crash
72051 when setting up the number of worker threads.
72053 This new option is also convenient when debugging GDB in a system with lots of
72054 threads, where the thread discovery code in GDB will emit too many messages,
72057 [New Thread 0xfffff74d3a50 (LWP 2625599)]
72059 If you have X threads, that message will be repeated X times.
72061 The default for --enable-threading is "yes".
72063 2021-12-15 Nikita Popov <npv1310@gmail.com>
72065 Fix an undefined behaviour in the BFD library's DWARF parser.
72067 * dwarf1.c (parse_die): Fix undefined behaviour in range tests.
72069 2021-12-15 Alan Modra <amodra@gmail.com>
72071 PR28694, Out-of-bounds write in stab_xcoff_builtin_type
72073 * stabs.c (stab_xcoff_builtin_type): Make typenum unsigned.
72074 Negate typenum earlier, simplifying bounds checking. Correct
72075 off-by-one indexing. Adjust switch cases.
72077 2021-12-15 GDB Administrator <gdbadmin@sourceware.org>
72079 Automatic date update in version.in
72081 2021-12-14 Alan Modra <amodra@gmail.com>
72083 loongarch32 build failure on 32-bit host
72084 gas/config/tc-loongarch.c: In function ‘assember_macro_helper’:
72085 gas/config/tc-loongarch.c:915:28: error: right shift count >= width of type [-Werror=shift-count-overflow]
72086 915 | hi32 = insn->args[1] >> 32;
72089 One possible fix is to make offsetT a 64-bit type for loongarch32.
72090 This also makes bfd/targmatch.h (generated from bfd/config.bfd)
72091 consistent since the loongarch32 match is inside #ifdef BFD64.
72093 * config.bfd (loongarch32-*): Set want64.
72095 2021-12-14 Alan Modra <amodra@gmail.com>
72097 loongarch64 build failure on 32-bit host
72098 gas/config/tc-loongarch.c: In function ‘loongarch_args_parser_can_match_arg_helper’:
72099 gas/config/tc-loongarch.c:661:13: error: cast from pointer to integer of different size [-Werror=pointer
72101 661 | imm = (offsetT) str_hash_find (r_htab, arg);
72104 Cast it to the correct size int, relying on normal integer promotions
72105 if offsetT is larger than a pointer.
72107 * config/tc-loongarch.c (loongarch_args_parser_can_match_arg_helper):
72108 Cast return from str_hash_find to intptr_t, not offsetT.
72110 2021-12-14 Alan Modra <amodra@gmail.com>
72112 XCOFF C_STSYM test failure on 32-bit host
72113 This test was failing here and on another similar symbol:
72114 [ 4](sec 1)(fl 0x00)(ty 0)(scl 143) (nx 0) 0x05d1745d11745d21 .bs
72115 where correct output is
72116 [ 4](sec 1)(fl 0x00)(ty 0)(scl 143) (nx 0) 0x000000000000000a .bs
72118 The problem is caused by a 32-bit host pointer being sign-extended
72119 when stored into a 64-bit bfd_vma, and then that value not being
72120 trimmed back to 32 bits when used. The following belt-and-braces
72121 patch fixes both the store and subsequent reads.
72123 * coffcode.h (coff_slurp_symbol_table): Do not sign extend
72124 when storing a host pointer to syment.n_value.
72125 * coffgen.c (coff_get_symbol_info): Cast syment.n_value to a
72126 bfd_hostptr_t before using in arithmetic.
72127 (coff_print_symbol): Likewise.
72129 2021-12-14 Simon Marchi <simon.marchi@efficios.com>
72131 gdbserver/tracepoint.cc: use snprintf in gdb_agent_socket_init
72132 If we modify tracepoint.cc to try to use a too long unix socket name,
72133 for example by modifying SOCK_DIR to be:
72135 #define SOCK_DIR "/tmp/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut/salut"
72137 ... trying to start an application with libinproctrace.so loaded
72140 $ LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libasan.so.6:./libinproctrace.so /bin/ls
72141 /home/smarchi/src/binutils-gdb/gdbserver/../gdbsupport/common-utils.cc:69: A problem internal to GDBserver in-process agent has been detected.
72142 xsnprintf: Assertion `ret < size' failed.
72144 Looking at the rest of the socket initialization code, the intent seems
72145 to be that if something goes wrong, we warn but let the program
72146 execute. So crashing on this failed assertions seems against the intent.
72148 Commit 6cebaf6e1ae4 ("use xsnprintf instead of snprintf.") changed this
72149 code to use xsnprintf instead of snprintf, introducing this assertion.
72150 Before that, snprintf would return a value bigger that UNIX_PATH_MAX and
72151 the "if" after would catch it and emit a warning, which is exactly what
72152 we want. That change was done because LynxOS didn't have snprintf.
72153 Since LynxOS isn't supported anymore, we can simply revert to use
72156 With this patch, we get a warning (printed by the caller of
72157 gdb_agent_socket_init), but the program keeps executing:
72159 $ LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libasan.so.6:./libinproctrace.so /bin/ls
72160 ipa: could not create sync socket
72163 Change-Id: I78bca52d5dc3145335abeae45a42052701e3f5dd
72165 2021-12-14 Simon Marchi <simon.marchi@efficios.com>
72167 gdbserver/tracepoint.cc: work around -Wstringop-truncation error
72168 When building gdb with on AArch64 with g++ 11.1.0 (and some preceding
72169 versions too), -O2 and -fsanitize=address, I get this error.
72171 CXX tracepoint-ipa.o
72172 cc1plus: warning: command-line option ‘-Wmissing-prototypes’ is valid for C/ObjC but not for C++
72173 In file included from /usr/include/string.h:519,
72174 from ../gnulib/import/string.h:41,
72175 from /home/simark/src/binutils-gdb/gdbserver/../gdbsupport/common-defs.h:95,
72176 from /home/simark/src/binutils-gdb/gdbserver/server.h:22,
72177 from /home/simark/src/binutils-gdb/gdbserver/tracepoint.cc:19:
72178 In function ‘char* strncpy(char*, const char*, size_t)’,
72179 inlined from ‘int init_named_socket(const char*)’ at /home/simark/src/binutils-gdb/gdbserver/tracepoint.cc:6902:11,
72180 inlined from ‘int gdb_agent_socket_init()’ at /home/simark/src/binutils-gdb/gdbserver/tracepoint.cc:6953:26,
72181 inlined from ‘void* gdb_agent_helper_thread(void*)’ at /home/simark/src/binutils-gdb/gdbserver/tracepoint.cc:7204:41:
72182 /usr/include/bits/string_fortified.h:95:34: error: ‘char* __builtin_strncpy(char*, const char*, long unsigned int)’ output may be truncated copying 107 bytes from a string of length 107 [-Werror=stringop-truncation]
72183 95 | return __builtin___strncpy_chk (__dest, __src, __len,
72184 | ~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~
72185 96 | __glibc_objsize (__dest));
72186 | ~~~~~~~~~~~~~~~~~~~~~~~~~
72188 Note that _FORTIFY_SOURCE changes the message a bit, but I get a similar
72189 error if I use -D_FORTIFY_SOURCE=0.
72191 I am pretty sure it's spurious and might be related to this GCC bug:
72193 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88780
72195 From what I can see, we are copying from a static 108 bytes long buffer
72196 (the global array agent_socket_name) to a 108 bytes long array,
72197 sockaddr_un::sun_path. I don't see anything wrong.
72199 Still, it would make things easier if we didn't see this error. Change
72200 the code to check that the source string length is smaller than the
72201 destination buffer (including space for the NULL byte) and use strcpy.
72203 For anybody who would like to try to reproduce, the full command line
72206 g++ -I. -I/home/simark/src/binutils-gdb/gdbserver -I/home/simark/src/binutils-gdb/gdbserver/../gdb/regformats -I/home/simark/src/binutils-gdb/gdbserver/.. -I/home/simark/src/binutils-gdb/gdbserver/../include -I/home/simark/src/binutils-gdb/gdbserver/../gdb -I/home/simark/src/binutils-gdb/gdbserver/../gnulib/import -I../gnulib/import -I/home/simark/src/binutils-gdb/gdbserver/.. -I.. -pthread -Wall -Wpointer-arith -Wno-unused -Wunused-value -Wunused-variable -Wunused-function -Wno-switch -Wno-char-subscripts -Wempty-body -Wunused-but-set-parameter -Wunused-but-set-variable -Wno-sign-compare -Wno-error=maybe-uninitialized -Wno-mismatched-tags -Wsuggest-override -Wimplicit-fallthrough=3 -Wduplicated-cond -Wshadow=local -Wdeprecated-copy -Wdeprecated-copy-dtor -Wredundant-move -Wmissing-declarations -Wmissing-prototypes -Wstrict-null-sentinel -Wformat -Wformat-nonliteral -Werror -DGDBSERVER -DCONFIG_UST_GDB_INTEGRATION -Drpl_strerror_r=strerror_r -Drpl_free=free -fPIC -DIN_PROCESS_AGENT -fvisibility=hidden -g3 -O2 -fsanitize=address -c -o tracepoint-ipa.o -MT tracepoint-ipa.o -MMD -MP -MF ./.deps/tracepoint-ipa.Tpo /home/simark/src/binutils-gdb/gdbserver/tracepoint.cc
72208 Change-Id: I18e86c0487feead7e7677e69398405e7057cf464
72210 2021-12-14 Simon Marchi <simon.marchi@polymtl.ca>
72212 bfd: fix -Wunused errors with clang 13+
72213 Clang 13 and 14 produce some -Wunused-but-set-{variable,parameter} for
72214 situations where gcc doesn't. In particular, when a variable is set and
72215 then used in a way to update its own value. For example, if `i` is only
72222 gcc won't warn, but clang will.
72224 Fix all such errors found in an --enable-targets=all build. It would be
72225 important for somebody who knows what they're doing to just make sure
72226 that these variables can indeed be deleted, and that there a no cases
72227 where it's a bug, and the variable should actually be used.
72229 The first instance of this error fix by this patch is:
72232 /home/simark/src/binutils-gdb/bfd/elf32-score.c:450:11: error: variable 'relocation' set but not used [-Werror,-Wunused-but-set-variable]
72233 bfd_vma relocation;
72236 Change-Id: I2f233ce20352645cf388aff3dfa08a651d21a6b6
72238 2021-12-14 Andrew Burgess <aburgess@redhat.com>
72240 gdb/mi: rename build_table to add_builtin_mi_commands
72241 Just give the function build_table a more descriptive name. There
72242 should be no user visible changes after this commit.
72244 2021-12-14 Jan Vrany <jan.vrany@labware.com>
72246 gdb/mi: rename mi_cmd to mi_command
72247 Just give this class a new name, more inline with the name of the
72248 sub-classes. I've also updated mi_cmd_up to mi_command_up in
72249 mi-cmds.c inline with this new naming scheme.
72251 There should be no user visible changes after this commit.
72253 2021-12-14 Jan Vrany <jan.vrany@labware.com>
72255 gdb/mi: use separate classes for different types of MI command
72256 This commit changes the infrastructure in mi-cmds.{c,h} to add new
72257 sub-classes for the different types of MI command. Instances of these
72258 sub-classes are then created and added into mi_cmd_table.
72260 The existing mi_cmd class becomes the abstract base class, this has an
72261 invoke method and takes care of the suppress notifications handling,
72262 before calling a do_invoke virtual method which is implemented by all
72263 of the sub-classes.
72265 There's currently two different sub-classes, one of pure MI commands,
72266 and a second for MI commands that delegate to CLI commands.
72268 There should be no user visible changes after this commit.
72270 2021-12-14 Andrew Burgess <aburgess@redhat.com>
72272 gdb/mi: int to bool conversion in mi_execute_cli_command
72273 Change an argument of mi_execute_cli_command from int to bool. Update
72274 the callers to take this into account. Within mi_execute_cli_command,
72275 update a comparison of a pointer to 0 with a comparison to nullptr,
72276 and add an assert, if we are not using the argument string then the
72277 string should be nullptr. Also removed a cryptic 'gdb_????' comment,
72278 which isn't really helpful.
72280 There should be no user visible changes after this commit.
72282 2021-12-14 Jan Vrany <jan.vrany@labware.com>
72284 gdb/mi: use std::map for MI commands in mi-cmds.c
72285 This changes the hashmap used in mi-cmds.c from a custom structure to
72286 std::map. Not only is replacing a custom container with a standard
72287 one an improvement, but using std::map will make it easier to
72288 dynamically add commands; which is something that is planned for a
72289 later series, where we will allow MI commands to be implemented in
72292 There should be no user visible changes after this commit.
72294 2021-12-14 Jan Vrany <jan.vrany@labware.com>
72296 gdb/mi: rename mi_lookup to mi_cmd_lookup
72297 Lets give this function a more descriptive name. I've also improved
72298 the comments in the header and source files.
72300 There should be no user visible changes after this commit.
72302 2021-12-14 Nelson Chu <nelson.chu@sifive.com>
72304 RISC-V: Added ld testcases for the medlow and medany code models.
72305 There are two linker scripts, code-model-01.ld and code-model-02.ld,
72306 which are corresponding to the two different memory layouts,
72308 * code-model-01.ld: the text section is in the 32-bit address range, but
72309 the data section is far away from the text section, which means the data
72310 section is over the 32-bit address range.
72312 * code-model-02.ld: the text section is over the 32-bit address range, but
72313 the data section is placed nearly zero address.
72315 We use the two linker scripts, to test the current medlow and medany behaviors
72316 of GNU ld, including the weak symbol references and the relaxations behaviors.
72317 Besides, these testcases also show the limits of the current medlow and medany
72318 code models, that is - we may get the truncated to fit errors when linking
72319 with the above two linker scripts.
72322 * testsuite/ld-riscv-elf/code-model-01.ld: New testcases to test the
72323 behaviors of the current medlow and medany code models.
72324 * testsuite/ld-riscv-elf/code-model-02.ld: Likewise.
72325 * testsuite/ld-riscv-elf/code-model-medany-01.d: Likewise.
72326 * testsuite/ld-riscv-elf/code-model-medany-02.d: Likewise.
72327 * testsuite/ld-riscv-elf/code-model-medany-weakref-01.d: Likewise.
72328 * testsuite/ld-riscv-elf/code-model-medany-weakref-02.d: Likewise.
72329 * testsuite/ld-riscv-elf/code-model-medlow-01.d: Likewise.
72330 * testsuite/ld-riscv-elf/code-model-medlow-02.d: Likewise.
72331 * testsuite/ld-riscv-elf/code-model-medlow-weakref-01.d: Likewise.
72332 * testsuite/ld-riscv-elf/code-model-medlow-weakref-02.d: Likewise.
72333 * testsuite/ld-riscv-elf/code-model-relax-medany-01.d: Likewise.
72334 * testsuite/ld-riscv-elf/code-model-relax-medany-02.d: Likewise.
72335 * testsuite/ld-riscv-elf/code-model-relax-medany-weakref-01.d: Likewise.
72336 * testsuite/ld-riscv-elf/code-model-relax-medany-weakref-02.d: Likewise.
72337 * testsuite/ld-riscv-elf/code-model-relax-medlow-01.d: Likewise.
72338 * testsuite/ld-riscv-elf/code-model-relax-medlow-02.d: Likewise.
72339 * testsuite/ld-riscv-elf/code-model-relax-medlow-weakref-01.d: Likewise.
72340 * testsuite/ld-riscv-elf/code-model-relax-medlow-weakref-02.d: Likewise.
72341 * testsuite/ld-riscv-elf/code-model.s: Likewise.
72342 * testsuite/ld-riscv-elf/ld-riscv-elf.exp: Updated.
72344 2021-12-14 GDB Administrator <gdbadmin@sourceware.org>
72346 Automatic date update in version.in
72348 2021-12-13 H.J. Lu <hjl.tools@gmail.com>
72350 x86: Adjust linker tests for --disable-separate-code
72351 Adjust linker tests for linker configured with --disable-separate-code:
72353 1. Update expected outputs.
72354 2. Pass -z max-page-size=0x1000 -z separate-code" to linker.
72356 * testsuite/ld-i386/report-reloc-1.l: Updated.
72357 * testsuite/ld-x86-64/report-reloc-1.l: Likewise.
72358 * testsuite/ld-x86-64/pe-x86-64.exp: Pass
72359 "-z max-page-size=0x1000 -z separate-code" to linker.
72360 * testsuite/ld-x86-64/pr19609-4e.d: Likewise.
72361 * testsuite/ld-x86-64/pr19609-6a.d: Likewise.
72362 * testsuite/ld-x86-64/pr19609-6b.d: Likewise.
72363 * testsuite/ld-x86-64/pr19609-7b.d: Likewise.
72364 * testsuite/ld-x86-64/pr19609-7d.d: Likewise.
72366 2021-12-13 Carl Love <cel@us.ibm.com>
72368 gdb: Powerpc mark xfail in gdb.base/catch-syscall.exp
72369 Powerpc is not reporting the
72371 Catchpoint 1 (returned from syscall execve), ....
72373 as expected. The issue appears to be with the kernel not returning the
72374 expected result. This patch marks the test failure as an xfail.
72376 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28623
72378 2021-12-13 Andrew Burgess <andrew.burgess@embecosm.com>
72380 gdb: improve reuse of value contents when fetching array elements
72381 While working on a Python script, which was interacting with a remote
72382 target, I noticed some weird slowness in GDB. In my program I had a
72383 structure something like this:
72390 struct foo_t global_foo;
72392 Then in the Python script I was fetching a complete copy of global
72395 val = gdb.parse_and_eval('global_foo')
72398 Then I would work with items in foo_t.array, like:
72400 print(val['array'][1])
72402 I called the fetch_lazy method specifically because I knew I was going
72403 to end up accessing almost all of the contents of val, and so I wanted
72404 GDB to do a single remote protocol call to fetch all the contents in
72405 one go, rather than trying to do lazy fetches for a couple of bytes at
72408 What I observed was that, after the fetch_lazy call, GDB does,
72409 correctly, fetch the entire contents of global_foo, including all of
72410 the contents of array, however, when I access val.array[1], GDB still
72411 goes and fetches the value of this element from the remote target.
72413 What's going on is that in valarith.c, in value_subscript, for C like
72414 languages, we always end up treating the array value as a pointer, and
72415 then doing value_ptradd, and value_ind, the second of these calls
72416 always returns a lazy value.
72418 My guess is that this approach allows us to handle indexing off the
72419 end of an array, when working with zero element arrays, or when
72420 indexing a raw pointer as an array. And, I agree, that in these
72421 cases, where, even when the original value is non-lazy, we still will
72422 not have the content of the array loaded, we should be using the
72423 value_ind approach.
72425 However, for cases where we do have the array contents loaded, and we
72426 do know the bounds of the array, I think we should be using
72427 value_subscripted_rvalue, which is what we use for non C like
72430 One problem I did run into, exposed by gdb.base/charset.exp, was that
72431 value_subscripted_rvalue stripped typedefs from the element type of
72432 the array, which means the value returned will not have the same type
72433 as an element of the array, but would be the raw, non-typedefed,
72434 type. In charset.exp we got back an 'int' instead of a
72435 'wchar_t' (which is a typedef of 'int'), and this impacts how we print
72436 the value. Removing typedefs from the resulting value just seems
72437 wrong, so I got rid of that, and I don't see any test regressions.
72439 With this change in place, my original Python script is now doing no
72440 additional memory accesses, and its performance increases about 10x!
72442 2021-12-13 Andrew Burgess <aburgess@redhat.com>
72444 gdb: update gdb-gdb.py.in for latest changes to struct field
72445 This commit updates uses of 'loc' and 'loc_kind' to 'm_loc' and
72446 'm_loc_kind' respectively, in gdb-gdb.py.in, which is required after
72449 commit cd3f655cc7a55437a05aa8e7b1fcc9051b5fe404
72450 Date: Thu Sep 30 22:38:29 2021 -0400
72452 gdb: add accessors for field (and call site) location
72454 I have also incorporated this change:
72456 https://sourceware.org/pipermail/gdb-patches/2021-September/182171.html
72458 Which means we print 'm_name' instead of 'name' when displaying the
72459 'm_name' member variable.
72461 Finally, I have also added support for the new TYPE_SPECIFIC_INT
72462 fields, which were added with this commit:
72464 commit 20a5fcbd5b28cca88511ac5a9ad5e54251e8fa6d
72465 Date: Wed Sep 23 09:39:24 2020 -0600
72467 Handle bit offset and bit size in base types
72469 I updated the gdb.gdb/python-helper.exp test to cover all of these
72472 2021-12-13 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
72474 gdbserver/linux-low: replace direct assignment to current_thread
72475 Use scoped_restore_current_thread and switch_to_thread in
72476 linux_process_target::wait_for_sigstop.
72478 2021-12-13 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
72480 gdbserver: replace direct assignments to current_thread
72481 Replace the direct assignments to current_thread with
72482 switch_to_thread. Use scoped_restore_current_thread when appropriate.
72483 There is one instance remaining in linux-low.cc's wait_for_sigstop.
72484 This will be handled in a separate patch.
72486 Regression-tested on X86-64 Linux using the native-gdbserver and
72487 native-extended-gdbserver board files.
72489 2021-12-13 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
72491 gdbserver: introduce scoped_restore_current_thread and switch_to_thread
72492 Introduce a class for restoring the current thread and a function to
72493 switch to the given thread. This is a preparation for a refactoring
72494 that aims to remove direct assignments to 'current_thread'.
72496 2021-12-13 Andrew Burgess <aburgess@redhat.com>
72498 gdb: make post_startup_inferior a virtual method on inf_ptrace_target
72499 While working on a later patch that required me to understand how GDB
72500 starts up inferiors, I was confused by the
72501 target_ops::post_startup_inferior method.
72503 The post_startup_inferior target function is only called from
72504 inf_ptrace_target::create_inferior.
72506 Part of the target class hierarchy looks like this:
72510 '-- inf_ptrace_target
72512 |-- linux_nat_target
72514 |-- fbsd_nat_target
72516 |-- nbsd_nat_target
72518 |-- obsd_nat_target
72520 '-- rs6000_nat_target
72522 Every sub-class of inf_ptrace_target, except rs6000_nat_target,
72523 implements ::post_startup_inferior. The rs6000_nat_target picks up
72524 the implementation of ::post_startup_inferior not from
72525 inf_ptrace_target, but from inf_child_target.
72527 No descendent of inf_child_target, outside the inf_ptrace_target
72528 sub-tree, implements ::post_startup_inferior, which isn't really
72529 surprising, as they would never see the method called (remember, the
72530 method is only called from inf_ptrace_target::create_inferior).
72532 What I find confusing is the role inf_child_target plays in
72533 implementing, what is really a helper function for just one of its
72536 In this commit I propose that we formally make ::post_startup_inferior
72537 a helper function of inf_ptrace_target. To do this I will remove the
72538 ::post_startup_inferior from the target_ops API, and instead make this
72539 a protected, pure virtual function on inf_ptrace_target.
72541 I'll remove the empty implementation of ::post_startup_inferior from
72542 the inf_child_target class, and add a new empty implementation to the
72543 rs6000_nat_target class.
72545 All the other descendents of inf_ptrace_target already provide an
72546 implementation of this method and so don't need to change beyond
72547 making the method protected within their class declarations.
72549 To me, this makes much more sense now. The helper function, which is
72550 only called from within the inf_ptrace_target class, is now a part of
72551 the inf_ptrace_target class.
72553 The only way in which this change is visible to a user is if the user
72554 turns on 'set debug target 1'. With this debug flag on, prior to this
72555 patch the user would see something like:
72557 -> native->post_startup_inferior (...)
72558 <- native->post_startup_inferior (2588939)
72560 After this patch these lines are no longer present, as the
72561 post_startup_inferior is no longer a top level target method. For me,
72562 this is an acceptable change.
72564 2021-12-13 Andrew Burgess <aburgess@redhat.com>
72566 gdb: have mips_nbsd_nat_target inherit from nbsd_nat_target
72567 While working on another patch I had reason to look at
72568 mips-netbsd-nat.c, and noticed that the class mips_nbsd_nat_target
72569 inherits directly from inf_ptrace_target.
72571 This is a little strange as alpha_bsd_nat_target,
72572 arm_netbsd_nat_target, hppa_nbsd_nat_target, m68k_bsd_nat_target,
72573 ppc_nbsd_nat_target, sh_nbsd_nat_target, and vax_bsd_nat_target all
72574 inherit from nbsd_nat_target.
72576 Originally, in this commit:
72578 commit f6ac5f3d63e03a81c4ff3749aba234961cc9090e
72579 Date: Thu May 3 00:37:22 2018 +0100
72581 Convert struct target_ops to C++
72583 When the target tree was converted to C++, all of the above classes
72584 inherited from inf_ptrace_target except for hppa_nbsd_nat_target,
72585 which was the only class that originally inherited from
72588 Later on all the remaining targets (except mips) were converted to
72589 inherit from nbsd_nat_target, these are the commits:
72591 commit 4fed520be264b60893aa674071947890f8172915
72592 Date: Sat Mar 14 16:05:24 2020 +0100
72594 Inherit alpha_netbsd_nat_target from nbsd_nat_target
72596 commit 6018d381a00515933016c539d2fdc18ad0d304b8
72597 Date: Sat Mar 14 14:50:51 2020 +0100
72599 Inherit arm_netbsd_nat_target from nbsd_nat_target
72601 commit 01a801176ea15ddfc988cade2e3d84c3b0abfec3
72602 Date: Sat Mar 14 16:54:42 2020 +0100
72604 Inherit m68k_bsd_nat_target from nbsd_nat_target
72606 commit 9faa006d11a5e08264a007463435f84b77864c9c
72607 Date: Thu Mar 19 14:52:57 2020 +0100
72609 Inherit ppc_nbsd_nat_target from nbsd_nat_target
72611 commit 9809762324491b851332ce600ae9bde8dd34f601
72612 Date: Tue Mar 17 15:07:39 2020 +0100
72614 Inherit sh_nbsd_nat_target from nbsd_nat_target
72616 commit d5be5fa4207da00d039a1d5a040ba316e7092cbd
72617 Date: Sat Mar 14 13:21:58 2020 +0100
72619 Inherit vax_bsd_nat_target from nbsd_nat_target
72621 I could only find mailing list threads for ppc and sh in the archive ,
72622 and unfortunately, none of the commits has any real detail that might
72623 explain why mips was missed out, the only extra context I could find
72626 https://sourceware.org/pipermail/gdb-patches/2020-March/166853.html
72628 Which says that "proper" OS support is going to be added to
72629 nbsd_nat_target, hence the need to inherit from that class.
72631 My guess is that leaving mips_nbsd_nat_target unchanged was an
72632 oversight, so, in this commit, I propose changing mips_nbsd_nat_target
72633 to inherit from nbsd_nat_target just like all the other nbsd targets.
72635 My motivation for this patch relates to the post_startup_inferior
72636 target method. In a future commit I want to change how this method is
72637 handled. Currently the mips_nbsd_nat_target will pick up the empty
72638 implementation of inf_child_target::post_startup_inferior rather than
72639 the version in netbsd-nat.c. This feels like a bug to me, as surely,
72640 enabling of proc events is something that would need to be done for
72641 all netbsd targets, regardless of architecture.
72643 In my future patch I have a choice then, either (a) add a new, empty
72644 implementation of post_startup_inferior to mips_nbsd_nat_target,
72645 or (b) this commit, have mips_nbsd_nat_target inherit from
72646 nbsd_nat_target. Option (b) seems like the right way to go, hence,
72649 I've done absolutely no testing for this change, not even building it,
72650 as that would require at least an environment in which I can x-build
72651 mips-netbsd applications, which I have no idea how to set up.
72653 2021-12-13 Andrew Burgess <aburgess@redhat.com>
72655 gdb: only include mips and riscv targets if building with 64-bit bfd
72656 While testing another patch I was trying to build different
72657 configurations of GDB, and, during one test build I ran into a
72658 problem, I configured with `--enable-targets=all
72659 --host=i686-w64-mingw32` and saw this error while linking GDB:
72661 .../i686-w64-mingw32/bin/ld: mips-tdep.o: in function `mips_gdbarch_init':
72662 .../src/gdb/mips-tdep.c:8730: undefined reference to `disassembler_options_mips'
72663 .../i686-w64-mingw32/bin/ld: riscv-tdep.o: in function `riscv_gdbarch_init':
72664 .../src/gdb/riscv-tdep.c:3851: undefined reference to `disassembler_options_riscv'
72666 So the `disassembler_options_mips` and `disassembler_options_riscv`
72667 symbols are missing.
72669 This turns out to be because mips-dis.c and riscv-dis.c, in which
72670 these symbols are defined, are in the TARGET64_LIBOPCODES_CFILES list
72671 in opcodes/Makefile.am, these files are only built when we are
72672 building with a 64-bit bfd.
72674 If we look further, into bfd/Makefile.am, we can see that all the
72675 files matching elf*-riscv.lo are in the BFD64_BACKENDS list, as are
72676 the elf*-mips.lo files, and (I know because I tried), the two
72677 disassemblers do, not surprisingly, depend on features supplied from
72680 So, though we can build most of GDB just fine for riscv and mips with
72681 a 32-bit bfd, if I understand correctly, the final GDB
72682 executable (assuming we could get it to link) would not understand
72683 these architectures at the bfd level, nor would there be any
72684 disassembler available. This sounds like a pretty useless GDB to me.
72686 So, in this commit, I move the riscv and mips targets into GDB's list
72687 of targets that require a 64-bit bfd. After this I can build GDB just
72688 fine with the configure options given above.
72690 This was discussed on the mailing list in a couple of threads:
72692 https://sourceware.org/pipermail/gdb-patches/2021-December/184365.html
72693 https://sourceware.org/pipermail/binutils/2021-November/118498.html
72695 and it is agreed, that it is unfortunate that the 32-bit riscv and
72696 32-bit mips targets require a 64-bit bfd. If in the future this
72697 situation ever changes then it would be expected that some (or all) of
72698 this patch would be reverted. Until then though, this patch allows
72699 GDB to build when configured with --enable-targets=all, and when using
72702 2021-12-13 GDB Administrator <gdbadmin@sourceware.org>
72704 Automatic date update in version.in
72706 2021-12-12 Tom Tromey <tom@tromey.com>
72708 C++-ify path substitution code
72709 I found some uses of xfree in the path substitution code in source.c.
72710 C++-ifying struct substitute_path_rule both simplifies the code and
72711 removes manual memory management.
72713 Regression tested on x86-64 Fedora 34.
72715 2021-12-12 GDB Administrator <gdbadmin@sourceware.org>
72717 Automatic date update in version.in
72719 2021-12-11 Alan Modra <amodra@gmail.com>
72721 [GOLD] PowerPC64 @notoc in non-power10 code
72722 Gold version of commit 7aba54da42.
72725 * powerpc.h (R_PPC64_REL24_P9NOTOC): Define.
72727 * powerpc.cc (Target_powerpc::maybe_skip_tls_get_addr_call,
72728 is_branch_reloc, max_branch_delta): Handle R_PPC64_REL24_P9NOTOC.
72729 (Target_powerpc::Branch_info::make_stub): Likewise.
72730 (struct Plt_stub_ent): Add p9notoc_, p9off_, tsize_.
72731 (struct Branch_stub_ent): Add p9notoc_, p9off_.
72732 (Stub_table::add_plt_call_entry): Handle R_PPC64_REL24_P9NOTOC.
72733 (Stub_table::add_long_branch_entry): Likewise.
72734 (Stub_table::add_eh_frame): Likewise.
72735 (Stub_table::plt_call_size): Return aligned size. Adjust callers.
72736 Handle p9notoc_ sizing.
72737 (Stub_table::do_write): Write out p9notoc_ stubs.
72738 (Target_powerpc::Scan::get_reference_flags, local, global):
72739 Handle R_PPC64_REL24_P9NOTOC.
72740 (Target_powerpc::Relocate::relocate): Likewise.
72742 2021-12-11 H.J. Lu <hjl.tools@gmail.com>
72744 Don't return the main file as the separate debug info
72747 $ readelf -d /usr/bin/npc
72749 caused readelf to run out of stack since load_separate_debug_info
72750 returned the input main file as the separate debug info:
72753 #0 load_separate_debug_info (
72754 main_filename=main_filename@entry=0x510f50 "/export/home/hjl/.cache/debuginfod_client/dcc33c51c49e7dafc178fdb5cf8bd8946f965295/debuginfo",
72755 xlink=xlink@entry=0x4e5180 <debug_displays+4480>,
72756 parse_func=parse_func@entry=0x431550 <parse_gnu_debuglink>,
72757 check_func=check_func@entry=0x432ae0 <check_gnu_debuglink>,
72758 func_data=func_data@entry=0x7fffffffdb60, file=file@entry=0x51d430)
72759 at /export/gnu/import/git/sources/binutils-gdb/binutils/dwarf.c:11057
72760 #1 0x000000000043328d in check_for_and_load_links (file=0x51d430,
72761 filename=0x510f50 "/export/home/hjl/.cache/debuginfod_client/dcc33c51c49e7dafc178fdb5cf8bd8946f965295/debuginfo")
72762 at /export/gnu/import/git/sources/binutils-gdb/binutils/dwarf.c:11381
72763 #2 0x00000000004332ae in check_for_and_load_links (file=0x51b070,
72764 filename=0x518dd0 "/export/home/hjl/.cache/debuginfod_client/dcc33c51c49e7dafc178fdb5cf8bd8946f965295/debuginfo")
72766 Return NULL if the separate debug info is the same as the input main
72767 file to avoid infinite recursion.
72770 * dwarf.c (load_separate_debug_info): Don't return the input
72773 2021-12-11 Alan Modra <amodra@gmail.com>
72775 Don't edit bogus sh_link on reading relocatable objects (Oracle fix)
72776 This reverts a 1995 fix to handle bogus object files. Presumably such
72777 object files have long gone.
72779 * elf.c (bfd_section_from_shdr): Remove old hack for Oracle
72782 2021-12-11 GDB Administrator <gdbadmin@sourceware.org>
72784 Automatic date update in version.in
72786 2021-12-10 Jan Vrany <jan.vrany@labware.com>
72788 gdb/testsuite: respect GDBSERVER variable in remote-stdio-gdbserver "board"
72789 The comment on top of gdb/testsuite/boards/remote-stdio-gdbserver.exp says
72790 that user can specify path to gdbserver on remote system by setting
72791 GDBSERVER variable. However, this variable was ignored and /usr/bin/gdbserver
72792 was used unconditionally.
72794 This commit updates the code to respect GDBSERVER if set and fall back to
72795 /usr/bin/gdbserver if not.
72797 2021-12-10 Simon Marchi <simon.marchi@polymtl.ca>
72799 Revert "gdbsupport: remove unnecessary `#ifndef IN_PROCESS_AGENT`"
72800 This reverts commit fe72c32765e1190c8a17d309fc3a7e1882d6a430.
72802 It turns out it was wrong, libinproctrace.so does build its own
72803 gdbsupport/tdesc.cc. This broke the build:
72805 make[1]: Entering directory '/home/simark/build/binutils-gdb-one-target/gdbserver'
72806 CXXLD libinproctrace.so
72807 /usr/bin/ld: gdbsupport/tdesc-ipa.o: in function `print_xml_feature::visit_pre(target_desc const*)':
72808 /home/simark/src/binutils-gdb/gdbserver/../gdbsupport/tdesc.cc:407: undefined reference to `tdesc_architecture_name(target_desc const*)'
72809 /usr/bin/ld: /home/simark/src/binutils-gdb/gdbserver/../gdbsupport/tdesc.cc:408: undefined reference to `tdesc_architecture_name(target_desc const*)'
72810 /usr/bin/ld: /home/simark/src/binutils-gdb/gdbserver/../gdbsupport/tdesc.cc:411: undefined reference to `tdesc_osabi_name(target_desc const*)'
72811 /usr/bin/ld: /home/simark/src/binutils-gdb/gdbserver/../gdbsupport/tdesc.cc:416: undefined reference to `tdesc_compatible_info_list(target_desc const*)'
72812 /usr/bin/ld: /home/simark/src/binutils-gdb/gdbserver/../gdbsupport/tdesc.cc:418: undefined reference to `tdesc_compatible_info_arch_name(std::unique_ptr<tdesc_compatible_info, std::default_delete<tdesc_compatible_info> > const&)'
72814 2021-12-10 GDB Administrator <gdbadmin@sourceware.org>
72816 Automatic date update in version.in
72818 2021-12-09 Alan Modra <amodra@gmail.com>
72820 PR28674, objdump crash
72821 Not returning an error indication here leaves the attribute
72822 uninitialised, which then leads to intemperate behaviour.
72825 * dwarf2.c (read_attribute_value): Return NULL on trying to read
72826 past end of attributes.
72828 2021-12-09 Alan Modra <amodra@gmail.com>
72830 Set sh_link for reloc sections created as normal sections
72831 binutils-all/strip-13 and binutils-all/strip-14 tests create
72832 SHT_REL/SHT_RELA sections by hand. These don't have sh_link set to
72833 the .symtab section as they should, leading to readelf warnings if you
72834 happen to be looking at the object files.
72836 * elf.c (assign_section_numbers): Formatting. Set sh_link for
72837 reloc sections created as normal sections in relocatable
72840 2021-12-09 Simon Marchi <simon.marchi@efficios.com>
72842 gdbsupport: remove unnecessary `#ifndef IN_PROCESS_AGENT`
72843 I suppose this code was copied from GDBserver and this ifndef was left
72844 there. As far as I know, IN_PROCESS_AGENT will never be defined when
72845 building this file, so we can remove this.
72847 Change-Id: I84fc408e330b3a29106df830a09342861cadbaf6
72849 2021-12-09 Simon Marchi <simon.marchi@polymtl.ca>
72851 gdb/microblaze-tdep.c: fix -Wunused-but-set-variable
72852 Fix this, seen when building with clang 14:
72854 CXX microblaze-tdep.o
72855 /home/simark/src/binutils-gdb/gdb/microblaze-tdep.c:207:7: error: variable 'flags' set but not used [-Werror,-Wunused-but-set-variable]
72859 Change-Id: I59f726ed33e924912748bc475b6fd9a9394fc0d0
72861 2021-12-09 Simon Marchi <simon.marchi@polymtl.ca>
72863 gdb/csky-tdep.c: fix -Wunused-but-set-variable error
72864 Fix these, seen when building with clang 14:
72867 /home/simark/src/binutils-gdb/gdb/csky-tdep.c:332:7: error: variable 'need_dummy_stack' set but not used [-Werror,-Wunused-but-set-variable]
72868 int need_dummy_stack = 0;
72870 /home/simark/src/binutils-gdb/gdb/csky-tdep.c:805:12: error: variable 'offset' set but not used [-Werror,-Wunused-but-set-variable]
72874 Change-Id: I6703bcb50e83c50083f716f4084ef6aa30d659c4
72876 2021-12-09 Simon Marchi <simon.marchi@polymtl.ca>
72878 gdb/testsuite: fix default behavior of runto
72879 The documented behavior of proc runto is to not emit a PASS when
72880 succeeding to to run to the specified location, but emit a FAIL when
72881 failing to do so. I suppose the intent is that it won't pollute the
72882 results normally passing tests (although I don't see why we would care),
72883 but make visible any problems.
72885 However, it seems like the implementation makes it default to never
72886 print anything. "no-message" is prependend to "args", so if "message"
72887 is not passed, we will always take the path that sets print_fail to 0,
72888 which will silence any failure.
72890 This unfortunately means that tests relying on runto_main won't emit a
72891 FAIL if failing to run to main. And since commit 4dfef5be6812
72892 ("gdb/testsuite: make runto_main not pass no-message to runto"), tests
72893 don't emit a FAIL themselves when failing to run to main. This means
72894 that tests failing to run to main just fail silently, and that's bad.
72896 This can be reproduced by hacking gdb.base/template.exp like so:
72898 diff --git a/gdb/testsuite/gdb.base/template.c b/gdb/testsuite/gdb.base/template.c
72899 index bcf39c377d92..052be5b79d73 100644
72900 --- a/gdb/testsuite/gdb.base/template.c
72901 +++ b/gdb/testsuite/gdb.base/template.c
72903 You should have received a copy of the GNU General Public License
72904 along with this program. If not, see <http://www.gnu.org/licenses/>. */
72906 +#include <stdlib.h>
72908 +__attribute__((constructor))
72909 +static void c (void)
72918 Running the modified gdb.base/template.exp shows that it exits without
72919 printing any result.
72921 Remove the line that prepends no-message to args, that should make
72922 runto's behavior match its documentation.
72924 This patch will appear to add many failures, but in reality they already
72925 existed, they were just silenced.
72927 Change-Id: I2a730d5bc72b6ef0698cd6aad962d9902aa7c3d6
72929 2021-12-09 Carl Love <cel@us.ibm.com>
72931 gdb fix elfv1 Powerpc gdb.dwarf2/frame-inlined-in-outer-frame.exp
72932 On ELFv1, the _start symbol must point to the *function descriptor* (in
72933 the .opd section), not to the function code (in the .text section) like
72934 with ELFv2 and other architectures.
72936 2021-12-09 Tom de Vries <tdevries@suse.de>
72938 [gdb/testsuite] Fix gdb.base/maint.exp with -readnow
72939 With test-case gdb.base/maint.exp and target board -readnow, I run into:
72941 FAIL: gdb.base/maint.exp: maint info line-table w/o a file name
72944 The problem is that this and other regexps anchored using '^':
72946 -re "^$gdb_prompt $" {
72948 don't trigger because other regexps don't consume the entire preceding line.
72950 This is partly due to the addition of the IS-STMT column.
72952 Fix this by making the regexps consume entire lines.
72954 Tested on x86_64-linux with native and target board readnow, as well as
72955 check-read1 and check-readmore.
72957 2021-12-09 Tom de Vries <tdevries@suse.de>
72959 [gdb/testsuite] Fix gdb.base/include-main.exp with -readnow
72960 With test-case gdb.base/include-main.exp and target board readnow, I run into:
72962 FAIL: gdb.base/include-main.exp: maint info symtab
72965 The corresponding check in gdb.base/include-main.exp:
72967 gdb_test_no_output "maint info symtab"
72969 checks that no CU was expanded, while -readnow ensures that all CUs are
72972 Fix this by skipping the check with -readnow.
72974 Tested on x86_64-linux, with native and target board readnow.
72976 2021-12-09 Nelson Chu <nelson.chu@sifive.com>
72978 RISC-V: Clarify the behavior of .option arch directive.
72979 * To be consistent with -march option, removed the "=" operator when
72980 user want to reset the whole architecture string. So the formats are,
72982 .option arch, +<extension><version>, ...
72983 .option arch, -<extension>
72984 .option arch, <ISA string>
72986 * Don't allow to add or remove the base extensions in the .option arch
72987 directive. Instead, users should reset the whole architecture string
72988 while they want to change the base extension.
72990 * The operator "+" won't update the version of extension, if the
72991 extension is already in the subset list.
72994 * elfxx-riscv.c (riscv_add_subset): Don't update the version
72995 if the extension is already in the subset list.
72996 (riscv_update_subset): To be consistent with -march option,
72997 removed the "=" operator when user want to reset the whole
72998 architecture string. Besides, Don't allow to add or remove
72999 the base extensions in the .option arch directive.
73001 * testsuite/gas/riscv/option-arch-01.s: Updated since we cannot
73002 add or remove the base extensions in the .option arch directive.
73003 * testsuite/gas/riscv/option-arch-02.s: Likewise.
73004 * testsuite/gas/riscv/option-arch-fail.l: Likewise.
73005 * testsuite/gas/riscv/option-arch-fail.s: Likewise.
73006 * testsuite/gas/riscv/option-arch-01a.d: Set -misa-spec=2.2.
73007 * testsuite/gas/riscv/option-arch-01b.d: Likewise.
73008 * testsuite/gas/riscv/option-arch-02.d: Updated since the .option
73009 arch, + won't change the version of extension, if the extension is
73010 already in the subset list.
73011 * testsuite/gas/riscv/option-arch-03.s: Removed the "=" operator
73012 when resetting the whole architecture string.
73014 2021-12-09 Mike Frysinger <vapier@gentoo.org>
73016 sim: use ## for automake comments
73017 The ## marker tells automake to not include the comment in its
73018 generated output, so use that in most places where the comment
73019 only makes sense in the inputs.
73021 2021-12-09 Simon Marchi <simon.marchi@efficios.com>
73023 gdb, gdbserver: detach fork child when detaching from fork parent
73024 While working with pending fork events, I wondered what would happen if
73025 the user detached an inferior while a thread of that inferior had a
73026 pending fork event. What happens with the fork child, which is
73027 ptrace-attached by the GDB process (or by GDBserver), but not known to
73028 the core? Sure enough, neither the core of GDB or the target detach the
73029 child process, so GDB (or GDBserver) just stays ptrace-attached to the
73030 process. The result is that the fork child process is stuck, while you
73031 would expect it to be detached and run.
73033 Make GDBserver detach of fork children it knows about. That is done in
73034 the generic handle_detach function. Since a process_info already exists
73035 for the child, we can simply call detach_inferior on it.
73037 GDB-side, make the linux-nat and remote targets detach of fork children
73038 known because of pending fork events. These pending fork events can be
73041 - thread_info::pending_waitstatus, if the core has consumed the event
73042 but then saved it for later (for example, because it got the event
73043 while stopping all threads, to present an all-stop stop on top of a
73045 - thread_info::pending_follow: if we ran to a "catch fork" and we
73046 detach at that moment
73048 Additionally, pending fork events can be in target-specific fields:
73050 - For linux-nat, they can be in lwp_info::status and
73051 lwp_info::waitstatus.
73052 - For the remote target, they could be stored as pending stop replies,
73053 saved in `remote_state::notif_state::pending_event`, if not
73054 acknowledged yet, or in `remote_state::stop_reply_queue`, if
73055 acknowledged. I followed the model of remove_new_fork_children for
73056 this: call remote_notif_get_pending_events to process /
73057 acknowledge any unacknowledged notification, then look through
73060 Update the gdb.threads/pending-fork-event.exp test (and rename it to
73061 gdb.threads/pending-fork-event-detach.exp) to try to detach the process
73062 while it is stopped with a pending fork event. In order to verify that
73063 the fork child process is correctly detached and resumes execution
73064 outside of GDB's control, make that process create a file in the test
73065 output directory, and make the test wait $timeout seconds for that file
73066 to appear (it happens instantly if everything goes well).
73068 This test catches a bug in linux-nat.c, also reported as PR 28512
73069 ("waitstatus.h:300: internal-error: gdb_signal target_waitstatus::sig()
73070 const: Assertion `m_kind == TARGET_WAITKIND_STOPPED || m_kind ==
73071 TARGET_WAITKIND_SIGNALLED' failed.). When detaching a thread with a
73072 pending event, get_detach_signal unconditionally fetches the signal
73073 stored in the waitstatus (`tp->pending_waitstatus ().sig ()`). However,
73074 that is only valid if the pending event is of type
73075 TARGET_WAITKIND_STOPPED, and this is now enforced using assertions (iit
73076 would also be valid for TARGET_WAITKIND_SIGNALLED, but that would mean
73077 the thread does not exist anymore, so we wouldn't be detaching it). Add
73078 a condition in get_detach_signal to access the signal number only if the
73079 wait status is of kind TARGET_WAITKIND_STOPPED, and use GDB_SIGNAL_0
73080 instead (since the thread was not stopped with a signal to begin with).
73082 Add another test, gdb.threads/pending-fork-event-ns.exp, specifically to
73083 verify that we consider events in pending stop replies in the remote
73084 target. This test has many threads constantly forking, and we detach
73085 from the program while the program is executing. That gives us some
73086 chance that we detach while a fork stop reply is stored in the remote
73087 target. To verify that we correctly detach all fork children, we ask
73088 the parent to exit by sending it a SIGUSR1 signal and have it write a
73089 file to the filesystem before exiting. Because the parent's main thread
73090 joins the forking threads, and the forking threads wait for their fork
73091 children to exit, if some fork child is not detach by GDB, the parent
73092 will not write the file, and the test will time out. If I remove the
73093 new remote_detach_pid calls in remote.c, the test fails eventually if I
73096 There is a known limitation: we don't remove breakpoints from the
73097 children before detaching it. So the children, could hit a trap
73098 instruction after being detached and crash. I know this is wrong, and
73099 it should be fixed, but I would like to handle that later. The current
73100 patch doesn't fix everything, but it's a step in the right direction.
73102 Change-Id: I6d811a56f520e3cb92d5ea563ad38976f92e93dd
73103 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28512
73105 2021-12-09 Simon Marchi <simon.marchi@efficios.com>
73107 gdb: move clearing of tp->pending_follow to follow_fork_inferior
73108 A following patch will change targets so that when they detach an
73109 inferior, they also detach any pending fork children this inferior may
73110 have. While doing this, I hit a case where we couldn't differentiate
73111 two cases, where in one we should detach the fork detach but not in the
73114 Suppose we continue past a fork with "follow-fork-mode == child" &&
73115 "detach-on-fork on". follow_fork_inferior calls target_detach to detach
73116 the parent. In that case the target should not detach the fork
73117 child, as we'll continue debugging the child. As of now, the
73118 tp->pending_follow field of the thread who called fork still contains
73119 the details about the fork.
73121 Then, suppose we run to a fork catchpoint and the user types "detach".
73122 In that case, the target should detach the fork child in addition to the
73123 parent. In that case as well, the tp->pending_follow field contains
73124 the details about the fork.
73126 To allow targets to differentiate the two cases, clear
73127 tp->pending_follow a bit earlier, when following a fork. Targets will
73128 then see that tp->pending_follow contains TARGET_WAITKIND_SPURIOUS, and
73129 won't detach the fork child.
73131 As of this patch, no behavior changes are expected.
73133 Change-Id: I537741859ed712cb531baaefc78bb934e2a28153
73135 2021-12-09 Simon Marchi <simon.marchi@efficios.com>
73137 gdb/remote.c: refactor pending fork status functions
73138 In preparation for a following patch, refactor a few things that I did
73139 find a bit awkward, and to make them a bit more reusable.
73141 - Pass an inferior to kill_new_fork_children instead of a pid. That
73142 allows iterating on only this inferior's threads and avoid further
73143 filtering on the thread's pid.
73144 - Change thread_pending_fork_status to return a non-nullptr value only
73145 if the thread does have a pending fork status.
73146 - Remove is_pending_fork_parent_thread, as one can just use
73147 thread_pending_fork_status and check for nullptr.
73148 - Replace is_pending_fork_parent with is_fork_status, which just
73149 returns if the given target_waitkind if a fork or a vfork. Push
73150 filtering on the pid to the callers, when it is necessary.
73152 Change-Id: I0764ccc684d40f054e39df6fa5458cc4c5d1cd7b
73154 2021-12-09 Simon Marchi <simon.marchi@efficios.com>
73156 gdb/remote.c: move some things up
73157 Move the stop_reply and a few functions up. Some code above them in the
73158 file will need to use them in a following patch. No behavior changes
73161 Change-Id: I3ca57d0e3ec253f56e1ba401289d9d167de14ad2
73163 2021-12-09 Simon Marchi <simon.marchi@efficios.com>
73165 gdb/linux-nat: factor ptrace-detach code to new detach_one_pid function
73166 The following patch will add some code paths that need to ptrace-detach
73167 a given PID. Factor out the code that does this and put it in its own
73168 function, so that it can be re-used.
73170 Change-Id: Ie65ca0d89893b41aea0a23d9fc6ffbed042a9705
73172 2021-12-09 Simon Marchi <simon.marchi@efficios.com>
73174 gdbserver: hide fork child threads from GDB
73175 This patch aims at fixing a bug where an inferior is unexpectedly
73176 created when a fork happens at the same time as another event, and that
73177 other event is reported to GDB first (and the fork event stays pending
73178 in GDBserver). This happens for example when we step a thread and
73179 another thread forks at the same time. The bug looks like (if I
73180 reproduce the included test by hand):
73182 (gdb) show detach-on-fork
73183 Whether gdb will detach the child of a fork is on.
73184 (gdb) show follow-fork-mode
73185 Debugger response to a program call of fork or vfork is "parent".
73188 Reading /home/simark/build/binutils-gdb/gdb/testsuite/outputs/gdb.threads/step-while-fork-in-other-thread/step-while-fork-in-other-thread from remote target...
73189 Reading /home/simark/build/binutils-gdb/gdb/testsuite/outputs/gdb.threads/step-while-fork-in-other-thread/step-while-fork-in-other-thread from remote target...
73190 Reading symbols from target:/home/simark/build/binutils-gdb/gdb/testsuite/outputs/gdb.threads/step-while-fork-in-other-thread/step-while-fork-in-other-thread...
73191 [New Thread 965190.965190]
73192 [Switching to Thread 965190.965190]
73193 Remote 'g' packet reply is too long (expected 560 bytes, got 816 bytes): ... <long series of bytes>
73195 The sequence of events leading to the problem is:
73197 - We are using the all-stop user-visible mode as well as the
73198 synchronous / all-stop variant of the remote protocol
73199 - We have two threads, thread A that we single-step and thread B that
73200 calls fork at the same time
73201 - GDBserver's linux_process_target::wait pulls the "single step
73202 complete SIGTRAP" and the "fork" events from the kernel. It
73203 arbitrarily choses one event to report, it happens to be the
73204 single-step SIGTRAP. The fork stays pending in the thread_info.
73205 - GDBserver send that SIGTRAP as a stop reply to GDB
73206 - While in stop_all_threads, GDB calls update_thread_list, which ends
73207 up querying the remote thread list using qXfer:threads:read.
73208 - In the reply, GDBserver includes the fork child created as a result
73209 of thread B's fork.
73210 - GDB-side, the remote target sees the new PID, calls
73211 remote_notice_new_inferior, which ends up unexpectedly creating a new
73212 inferior, and things go downhill from there.
73214 The problem here is that as long as GDB did not process the fork event,
73215 it should pretend the fork child does not exist. Ultimately, this event
73216 will be reported, we'll go through follow_fork, and that process will be
73219 The remote target (GDB-side), has some code to remove from the reported
73220 thread list the threads that are the result of forks not processed by
73221 GDB yet. But that only works for fork events that have made their way
73222 to the remote target (GDB-side), but haven't been consumed by the core
73223 yet, so are still lingering as pending stop replies in the remote target
73224 (see remove_new_fork_children in remote.c). But in our case, the fork
73225 event hasn't made its way to the GDB-side remote target. We need to
73226 implement the same kind of logic GDBserver-side: if there exists a
73227 thread / inferior that is the result of a fork event GDBserver hasn't
73228 reported yet, it should exclude that thread / inferior from the reported
73231 This was actually discussed a while ago, but not implemented AFAIK:
73233 https://pi.simark.ca/gdb-patches/1ad9f5a8-d00e-9a26-b0c9-3f4066af5142@redhat.com/#t
73234 https://sourceware.org/pipermail/gdb-patches/2016-June/133906.html
73236 Implementation details-wise, the fix for this is all in GDBserver. The
73237 Linux layer of GDBserver already tracks unreported fork parent / child
73238 relationships using the lwp_info::fork_relative, in order to avoid
73239 wildcard actions resuming fork childs unknown to GDB. This information
73240 needs to be made available to the handle_qxfer_threads_worker function,
73241 so it can filter the reported threads. Add a new thread_pending_parent
73242 target function that allows the Linux target to return the parent of an
73243 eventual fork child.
73245 Testing-wise, the test replicates pretty-much the sequence of events
73246 shown above. The setup of the test makes it such that the main thread
73247 is about to fork. We stepi the other thread, so that the step completes
73248 very quickly, in a single event. Meanwhile, the main thread is resumed,
73249 so very likely has time to call fork. This means that the bug may not
73250 reproduce every time (if the main thread does not have time to call
73251 fork), but it will reproduce more often than not. The test fails
73252 without the fix applied on the native-gdbserver and
73253 native-extended-gdbserver boards.
73255 At some point I suspected that which thread called fork and which thread
73256 did the step influenced the order in which the events were reported, and
73257 therefore the reproducibility of the bug. So I made the test try both
73258 combinations: main thread forks while other thread steps, and vice
73259 versa. I'm not sure this is still necessary, but I left it there
73260 anyway. It doesn't hurt to test a few more combinations.
73262 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28288
73263 Change-Id: I2158d5732fc7d7ca06b0eb01f88cf27bf527b990
73265 2021-12-09 GDB Administrator <gdbadmin@sourceware.org>
73267 Automatic date update in version.in
73269 2021-12-08 Tom Tromey <tom@tromey.com>
73271 Use for-each more in gdb
73272 There are some loops in gdb that use ARRAY_SIZE (or a wordier
73273 equivalent) to loop over a static array. This patch changes some of
73274 these to use foreach instead.
73276 Regression tested on x86-64 Fedora 34.
73278 2021-12-08 Tom Tromey <tromey@adacore.com>
73280 Fix error in file_and_directory patch
73281 In my earlier C++-ization patch for file_and_directory, I introduced
73284 - if (strcmp (fnd.name, "<unknown>") != 0)
73285 + if (fnd.is_unknown ())
73287 This change inverted the sense of the test, which causes failures with
73290 This patch fixes the bug. Regression tested on x86-64 Fedora 34. I
73291 also tested it using the AdaCore internal test suite, with
73292 .debug_names -- this was failing before, and now it works.
73294 2021-12-08 Andrew Burgess <aburgess@redhat.com>
73296 gdb/python: Use tp_init instead of tp_new to setup gdb.Value
73297 The documentation suggests that we implement gdb.Value.__init__,
73298 however, this is not currently true, we really implement
73299 gdb.Value.__new__. This will cause confusion if a user tries to
73300 sub-class gdb.Value. They might write:
73302 class MyVal (gdb.Value):
73303 def __init__ (self, val):
73304 gdb.Value.__init__(self, val)
73307 print ("Got: %s" % obj)
73309 But, when they source this code they'll see:
73311 (gdb) source ~/tmp/value-test.py
73312 Traceback (most recent call last):
73313 File "/home/andrew/tmp/value-test.py", line 7, in <module>
73315 File "/home/andrew/tmp/value-test.py", line 5, in __init__
73316 gdb.Value.__init__(self, val)
73317 TypeError: object.__init__() takes exactly one argument (the instance to initialize)
73320 The reason for this is that, as we don't implement __init__ for
73321 gdb.Value, Python ends up calling object.__init__ instead, which
73322 doesn't expect any arguments.
73324 The Python docs suggest that the reason why we might take this
73325 approach is because we want gdb.Value to be immutable:
73327 https://docs.python.org/3/c-api/typeobj.html#c.PyTypeObject.tp_new
73329 But I don't see any reason why we should require gdb.Value to be
73330 immutable when other types defined in GDB are not. This current
73331 immutability can be seen in this code:
73333 obj = gdb.Value(1234)
73334 print("Got: %s" % obj)
73335 obj.__init__ (5678)
73336 print("Got: %s" % obj)
73338 Which currently runs without error, but prints:
73343 In this commit I propose that we switch to using __init__ to
73344 initialize gdb.Value objects.
73346 This does introduce some additional complexity, during the __init__
73347 call a gdb.Value might already be associated with a gdb value object,
73348 in which case we need to cleanly break that association before
73349 installing the new gdb value object. However, the cost of doing this
73350 is not great, and the benefit - being able to easily sub-class
73351 gdb.Value seems worth it.
73353 After this commit the first example above works without error, while
73354 the second example now prints:
73359 In order to make it easier to override the gdb.Value.__init__ method,
73360 I have tweaked the definition of gdb.Value.__init__. The second,
73361 optional argument to __init__ is a gdb.Type, if this argument is not
73362 present then GDB figures out a suitable type.
73364 However, if we want to override the __init__ method in a sub-class,
73365 and still support the default argument, it is easier to write:
73367 class MyVal (gdb.Value):
73368 def __init__ (self, val, type=None):
73369 gdb.Value.__init__(self, val, type)
73371 Currently, passing None for the Type will result in an error:
73373 TypeError: type argument must be a gdb.Type.
73375 After this commit I now allow the type argument to be None, in which
73376 case GDB figures out a suitable type just as if the type had not been
73379 Unless a user is trying to reinitialize a value, or create sub-classes
73380 of gdb.Value, there should be no user visible changes after this
73383 2021-12-08 Andrew Burgess <aburgess@redhat.com>
73385 gdb: use try/catch around a gdb_disassembler::print_insn call
73386 While investigating some disassembler problems I ran into this case;
73387 GDB compiled on a 32-bit arm target, with --enable-targets=all. Then
73390 (gdb) set architecture i386
73391 (gdb) disassemble 0x0,+4
73392 unknown disassembler error (error = -1)
73394 This is interesting because it shows a case where the libopcodes
73395 disassembler is returning -1 without first calling the
73396 memory_error_func callback. Indeed, the return from libopcodes
73397 happens from this code snippet in i386-dis.c in the print_insn
73400 if (address_mode == mode_64bit && sizeof (bfd_vma) < 8)
73402 (*info->fprintf_func) (info->stream,
73403 _("64-bit address is disabled"));
73407 Notice how, prior to the return the disassembler tries to print a
73408 helpful message out, but GDB doesn't print this message.
73410 The reason this message goes missing is the call stack, it looks like
73413 gdb_pretty_print_disassembler::pretty_print_insn
73414 gdb_disassembler::print_insn
73417 i386-dis.c:print_insn
73419 When i386-dis.c:print_insn returns -1 this is handled in
73420 gdb_disassembler::print_insn, where an exception is thrown. However,
73421 the actual printing of the disassembler output is done in
73422 gdb_pretty_print_disassembler::pretty_print_insn, and is only done if
73423 an exception is not thrown.
73425 In this commit I change this. The pretty_print_insn now uses
73426 try/catch around the call to gdb_disassembler::print_insn, if we catch
73427 an error then we first print any pending output in the instruction
73428 buffer, before rethrowing the exception. As a result, even if an
73429 exception is thrown we still print any pending disassembler output to
73430 the screen; in the above case the helpful message will now be shown.
73432 Before my patch we might expect to see this output:
73434 (gdb) disassemble 0x0,+4
73435 Dump of assembler code from 0x0 to 0x4:
73436 0x0000000000000000: unknown disassembler error (error = -1)
73439 But now we see this:
73441 (gdb) disassemble 0x0,+4
73442 Dump of assembler code from 0x0 to 0x4:
73443 0x0000000000000000: 64-bit address is disabled
73444 unknown disassembler error (error = -1)
73446 If the disassembler returns -1 without printing a helpful message then
73447 we would still expect a change in output, something like:
73449 (gdb) disassemble 0x0,+4
73450 Dump of assembler code from 0x0 to 0x4:
73451 0x0000000000000000:
73452 unknown disassembler error (error = -1)
73454 Which I think is still acceptable, though at this point I think a
73455 strong case can be made that this is a disassembler bug (not printing
73456 anything, but still returning -1).
73458 Notice however, that the error message is always printed on a new line
73459 now. This is also true for the memory error case, where before we
73462 (gdb) disassemble 0x0,+4
73463 Dump of assembler code from 0x0 to 0x4:
73464 0x00000000: Cannot access memory at address 0x0
73468 (gdb) disassemble 0x0,+4
73469 Dump of assembler code from 0x0 to 0x4:
73471 Cannot access memory at address 0x0
73473 For me, I'm happy to accept this change, having the error on a line by
73474 itself, rather than just appended to the end of the previous line,
73475 seems like an improvement, but I'm aware others might feel
73476 differently, so I'd appreciate any feedback.
73478 2021-12-08 Jan Vrany <jan.vrany@labware.com>
73480 ppc: recognize all program traps
73481 Permanent program breakpoints (ones inserted into the code) other than
73482 the one GDB uses for POWER (0x7fe00008) did not result in stop but
73483 caused GDB to loop infinitely.
73485 This was because GDB did not recognize trap instructions other than
73486 "trap". For example, "tw 12, 4, 4" was not be recognized, causing GDB
73489 This commit fixes this by providing POWER specific hook
73490 (gdbarch_program_breakpoint_here_p) recognizing all tw, twi, td and tdi
73493 Tested on Linux on PowerPC e500 and on QEMU PPC64le.
73495 2021-12-08 Jan Vrany <jan.vrany@labware.com>
73497 ppc: use "trap" ("tw, 31, 0, 0") as breakpoint instruction
73498 Power ISA 3.0 B spec [1], sections 3.3.11 "Fixed-Point Trap Instructions"
73499 and section C.6 "Trap Mnemonics" specify "tw, 31, 0, 0" (encoded as
73500 0x7fe00008) as canonical unconditional trap instruction.
73502 This commit changes the breakpoint instruction used by GDB from
73503 "tw 12, r2, r2" to unconditional "trap".
73505 [1]: https://openpowerfoundation.org/?resource_lib=power-isa-version-3-0
73507 2021-12-08 Fangrui Song <maskray@google.com>
73509 bfd_section_from_shdr: Support SHT_RELR sections
73510 If a.so contains an SHT_RELR section, objcopy a.so will fail with:
73512 a.so: unknown type [0x13] section `.relr.dyn'
73514 This change allows objcopy to work.
73517 * elf.c (bfd_section_from_shdr): Support SHT_RELR.
73519 2021-12-08 Alan Modra <amodra@gmail.com>
73521 PR28673, input file 'gcov' is the same as output file
73523 * ldlang.c (open_output): Use local_sym_name when checking
73524 output against input files rather than filename.
73526 2021-12-08 Tom Tromey <tom@tromey.com>
73528 Fix bug in source.c change
73529 My earlier change to source.c ("Remove an xfree from add_path")
73530 introduced a regression. This patch fixes the problem.
73532 2021-12-08 Simon Marchi <simon.marchi@polymtl.ca>
73534 gdb: make struct linespect contain vectors, not pointers to vectors
73535 struct linespec contains pointers to vectors, instead of containing
73536 vectors directly. This is probably historical, when linespec_parser
73537 (which contains a struct linespec field) was not C++-ified yet. But it
73538 seems easy to change the pointers to vectors to just vectors today.
73539 This simplifies the code, we don't need to manually allocate and delete
73540 the vectors and there's no pointer that can be NULL.
73542 As far as I understand, there was not meaningful distinction between a
73543 NULL pointer to vector and an empty vector. So all NULL checks are
73544 changed for !empty checks.
73546 Change-Id: Ie759707da14d9d984169b93233343a86e2de9ee6
73548 2021-12-08 GDB Administrator <gdbadmin@sourceware.org>
73550 Automatic date update in version.in
73552 2021-12-07 Tom Tromey <tom@tromey.com>
73554 Remove an xfree from add_path
73555 This removes a temporary \0 assignment and an xfree from add_path,
73556 replacing it with a simpler use of std::string.
73558 2021-12-07 Simon Marchi <simon.marchi@polymtl.ca>
73560 gdb/linespec.c: simplify condition
73561 We can remove the empty check: if the vector has size 1, it is obviously
73562 not empty. This code ended up like this because the empty check used to
73565 Change-Id: I1571bd0228818ca93f6a6b444e9b010dc2da4c08
73567 2021-12-07 Simon Marchi <simon.marchi@polymtl.ca>
73569 gdb: rename "maint agent" functions
73570 Functions agent_eval_command and agent_command are used to implement
73571 maintenance commands, rename them accordingly (with the maint_ prefix),
73572 as well as the agent_command_1 helper function.
73574 Change-Id: Iacf96d4a0a26298e8dd4648a0f38da649ea5ef61
73576 2021-12-07 Simon Marchi <simon.marchi@polymtl.ca>
73578 gdb: make set_raw_breakpoint static
73579 set_raw_breakpoint is only used in breakpoint.c, make it static.
73581 Change-Id: I7fbeda067685309a30b88aceaf957eff7a28e310
73583 2021-12-07 John Baldwin <jhb@FreeBSD.org>
73585 Support AT_FXRNG and AT_KPRELOAD on FreeBSD.
73586 FreeBSD's kernel has recently added two new ELF auxiliary vector
73587 entries. AT_FXRNG points to a root seed version for the kernel's
73588 PRNG. Userland can use this to reseed a userland PRNG after the
73589 kernel's PRNG has reseeded. AT_KPRELOAD is the base address of a
73590 kernel-provided vDSO.
73592 This change displays the proper name and description of these entries
73597 * elf/common.h (AT_FREEBSD_FXRNG, AT_FREEBSD_KPRELOAD): Define.
73599 2021-12-07 Tom Tromey <tromey@adacore.com>
73601 Avoid extra work in global_symbol_searcher::expand_symtabs
73602 I noticed that global_symbol_searcher::expand_symtabs always passes a
73603 file matcher to expand_symtabs_matching. However, if 'filenames' is
73604 empty, then this always returns true. It's slightly more efficient to
73605 pass a null file matcher in this case, because that lets the "quick"
73606 symbol implementations skip any filename checks.
73608 Regression tested on x86-64 Fedora 34.
73610 2021-12-07 Tom de Vries <tdevries@suse.de>
73612 [gdb/testsuite] Fix options arg handling in compile_jit_elf_main_as_so
73613 In commit 80ad340c902 ("[gdb/testsuite] use -Ttext-segment for jit-elf tests")
73614 the following change was made:
73616 proc compile_jit_elf_main_as_so {main_solib_srcfile main_solib_binfile options} {
73617 - set options [concat $options debug]
73618 + global jit_load_address jit_load_increment
73620 + set options [list \
73621 + additional_flags="-DMAIN=jit_dl_main" \
73622 + additional_flags=-DLOAD_ADDRESS=$jit_load_address \
73623 + additional_flags=-DLOAD_INCREMENT=$jit_load_increment \
73627 Before the change, the options argument was used, but after the change not
73630 Fix this by reverting back to using "set options [concat $options ...]".
73632 Fixing this gets us twice the -DMAIN=jit_dl_main bit, once from a caller, and
73633 once from compile_jit_elf_main_as_so. Fix this by removing the bit from
73634 compile_jit_elf_main_as_so, which makes the code similar to compile_jit_main.
73636 Tested on x86_64-linux.
73638 2021-12-07 Tom de Vries <tdevries@suse.de>
73640 [gdb/testsuite] Fix FAIL in gdb.tui/basic.exp
73641 On openSUSE Leap 15.2 aarch64 I ran into:
73643 FAIL: gdb.tui/basic.exp: check main is where we expect on the screen
73645 while this is passing on x86_64.
73647 On x86_64-linux we have at the initial screen dump for "list -q main":
73649 0 +-/home/vries/gdb_versions/devel/src/gdb/testsuite/gdb.tui/tui-layout.c--+
73650 1 | 15 You should have received a copy of the GNU General Public |
73651 2 | 16 along with this program. If not, see <http://www.gnu.org/|
73662 0 +-/home/tdevries/gdb/src/gdb/testsuite/gdb.tui/tui-layout.c--------------+
73663 1 | 16 along with this program. If not, see <http://www.gnu.org/|
73674 The cause of the diffferent placement is that we have as line number for main
73677 $ gdb -q -batch outputs/gdb.tui/basic/basic -ex "info line main"
73678 Line 20 of "tui-layout.c" starts at address 0x4004a7 <main> \
73679 and ends at 0x4004ab <main+4>.
73681 and on aarch64 instead:
73683 $ gdb -q -batch outputs/gdb.tui/basic/basic -ex "info line main"
73684 Line 21 of "tui-layout.c" starts at address 0x4005f4 <main> \
73685 and ends at 0x4005f8 <main+4>.
73688 Fix this by using a new source file main-one-line.c, that implements the
73689 entire main function on a single line, in order to force the compiler to use
73692 Also try to do less hard-coding in the test-case.
73694 Tested on x86_64-linux and aarch64-linux.
73696 2021-12-07 Tom de Vries <tdevries@suse.de>
73698 [gdb/tdep] Fix inferior plt calls in PIE for i386
73699 Consider test-case test.c:
73702 void *p = malloc (10);
73707 When compiled to a non-PIE exec:
73711 the call sequence looks like:
73713 8048447: 83 ec 0c sub $0xc,%esp
73714 804844a: 6a 0a push $0xa
73715 804844c: e8 bf fe ff ff call 8048310 <malloc@plt>
73719 08048310 <malloc@plt>:
73720 8048310: ff 25 0c a0 04 08 jmp *0x804a00c
73721 8048316: 68 00 00 00 00 push $0x0
73722 804831b: e9 e0 ff ff ff jmp 8048300 <.plt>
73724 where the first insn at 0x8048310 initially jumps to the following address
73725 0x8048316, read from the .got.plt @ 0x804a00c:
73727 804a000 0c9f0408 00000000 00000000 16830408 ................
73728 804a010 26830408 &...
73731 Likewise, when compiled as a PIE:
73733 $ gcc -m32 -fPIE -pie test.c
73735 we have this call sequence (with %ebx setup to point to the .got.plt):
73738 579: 83 ec 0c sub $0xc,%esp
73739 57c: 6a 0a push $0xa
73740 57e: 89 c3 mov %eax,%ebx
73741 580: e8 6b fe ff ff call 3f0 <malloc@plt>
73745 000003f0 <malloc@plt>:
73746 3f0: ff a3 0c 00 00 00 jmp *0xc(%ebx)
73747 3f6: 68 00 00 00 00 push $0x0
73748 3fb: e9 e0 ff ff ff jmp 3e0 <.plt>
73750 where the insn at 0x3f0 initially jumps to following address 0x3f6, read from
73751 the .got.plt at offset 0xc:
73753 2000 f41e0000 00000000 00000000 f6030000 ................
73757 When instead doing an inferior call to malloc (with nosharedlib to force
73758 malloc to resolve to malloc@plt rather than the functions in ld.so or libc.so)
73759 with the non-PIE exec, we have the expected:
73761 $ gdb -q -batch a.out -ex start -ex nosharedlib -ex "p /x (void *)malloc (10)"
73762 Temporary breakpoint 1 at 0x8048444
73764 Temporary breakpoint 1, 0x08048444 in main ()
73768 But with the PIE exec, we run into:
73770 $ gdb -q -batch a.out -ex start -ex nosharedlib -ex "p /x (void *)malloc (10)"
73771 Temporary breakpoint 1 at 0x56c
73773 Temporary breakpoint 1, 0x5655556c in main ()
73775 Program received signal SIGSEGV, Segmentation fault.
73776 0x565553f0 in malloc@plt ()
73779 The segfault happens because:
73780 - the inferior call mechanism doesn't setup %ebx
73781 - %ebx instead is 0
73782 - the jump to "*0xc(%ebx)" reads from memory at 0xc
73784 Fix this by setting up %ebx properly in i386_thiscall_push_dummy_call.
73786 Fixes this failure with target board unix/-m32/-pie/-fPIE reported in
73789 FAIL: gdb.base/nodebug.exp: p/c (int) array_index("abcdef",2)
73792 Tested on x86_64-linux, with target board unix/-m32 and unix/-m32/-fPIE/-pie.
73794 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28467
73796 2021-12-07 Tom de Vries <tdevries@suse.de>
73798 [gdb/symtab] Support -readnow during reread
73799 When running test-case gdb.base/cached-source-file.exp with target board
73800 readnow, we run into:
73802 FAIL: gdb.base/cached-source-file.exp: rerun program (the program exited)
73805 The problem is that when rereading, the readnow is ignored.
73807 Fix this by copying the readnow handling code from symbol_file_add_with_addrs
73810 Tested on x86_64-linux.
73812 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=26800
73814 2021-12-07 Tom de Vries <tdevries@suse.de>
73816 [gdb/ada] Fix assert in ada_is_unconstrained_packed_array_type
73817 On openSUSE Leap 42.3, with system compiler gcc 4.8.5 I run into:
73819 (gdb) print u_one_two_three^M
73820 src/gdb/gdbtypes.h:1050: internal-error: field: \
73821 Assertion `idx >= 0 && idx < num_fields ()' failed.^M
73824 We run into trouble while doing this in
73825 ada_is_unconstrained_packed_array_type:
73827 1953 return TYPE_FIELD_BITSIZE (type, 0) > 0;
73829 which tries to get field 0 from a type without fields:
73831 (gdb) p type->num_fields ()
73834 which is the case because the type is a typedef:
73836 (gdb) p type->code ()
73837 $7 = TYPE_CODE_TYPEDEF
73840 Fix this by using the type referenced by the typedef instead.
73842 Tested on x86_64-linux.
73844 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28323
73846 2021-12-07 Alan Modra <amodra@gmail.com>
73848 Re: Add support for AArch64 EFI (efi-*-aarch64)
73849 Commit b69c9d41e8 was broken in multiple ways regarding the realloc
73850 of the target string, most notably in that "-little" wasn't actually
73851 appended to the input_target or output_target. This caused asan
73852 errors and "FAIL: Check if efi app format is recognized". I also
73853 noticed that the input_target string wasn't being copied but rather
73854 the output_target when dealing with the input target. Fix that too.
73857 * objcopy.c (convert_efi_target): Rewrite. Allocate modified
73858 target strings here..
73859 (copy_main): ..rather than here. Do handle input_target,
73860 not output_target for input.
73862 2021-12-07 Alan Modra <amodra@gmail.com>
73864 Error on ld output file name matching input file name
73865 It's not foolproof, for example we don't catch output to a linker
73866 script, to a library specified with -l, or to an element of a thin
73869 * ldlang.c (open_output): Exit with error on output file matching
73871 * testsuite/ld-misc/just-symbols.exp: Adjust ld -r test to suit.
73873 2021-12-07 GDB Administrator <gdbadmin@sourceware.org>
73875 Automatic date update in version.in
73877 2021-12-06 Carl Love <cel@us.ibm.com>
73879 gdb: Add PowerPC support to gdb.dwarf2/frame-inlined-in-outer-frame
73880 This patch adds an #elif defined for PowerPC to setup the exit_0 macro.
73881 This patch addes the needed macro definitionald logic to handle both elfV1
73884 The patch has been successfully tested on both PowerPC BE, Powerpc LE and
73885 X86_64 with no regressions.
73887 2021-12-06 Tom de Vries <tdevries@suse.de>
73889 [gdb/testsuite] Use precise align in gdb.arch/i386-{avx,sse}.exp
73890 Test-cases gdb.arch/i386-{avx,sse}.exp use assembly instructions that require
73891 the memory operands to be aligned to a certain boundary, and the test-cases
73892 use C11's _Alignas to make that happen.
73894 The draw-back of using _Alignas is that while it does enforce a minimum
73895 alignment, the actual alignment may be bigger, which makes the following
73897 - copy say, gdb.arch/i386-avx.c as basis for a new test-case
73898 - run the test-case and observe a PASS
73899 - commit the new test-case in the supposition that the test-case is correct
73901 - run later into a failure on a different test setup (which may be a setup
73902 where reproduction and investigation is more difficult and time-consuming),
73903 and find out that the specified alignment was incorrect and should have been
73904 updated to say, 64 bytes. The initial PASS occurred only because the actual
73905 alignment happened to be greater than required.
73907 The idea of having precise alignment as a means of having more predictable
73908 execution which allows flushing out bugs earlier, has been filed as PR
73911 Add a new file lib/precise-aligned-alloc.c with functions
73912 precise_aligned_alloc and precise_aligned_dup, to support precise alignment.
73914 Use precise_aligned_dup in aforementioned test-cases to:
73915 - verify that the specified alignment is indeed sufficient, rather
73916 than too little but accidentally over-aligned.
73917 - prevent the same type of problems in any new test-cases based on these
73919 Tested on x86_64-linux, with both gcc and clang.
73921 2021-12-06 Tom de Vries <tdevries@suse.de>
73923 [gdb/testsuite] Fix data alignment in gdb.arch/i386-{avx,sse}.exp
73924 When running test-case gdb.arch/i386-avx.exp with clang I ran into:
73926 (gdb) PASS: gdb.arch/i386-avx.exp: set first breakpoint in main
73930 Program received signal SIGSEGV, Segmentation fault.^M
73931 0x000000000040052b in main (argc=1, argv=0x7fffffffd3c8) at i386-avx.c:54^M
73932 54 asm ("vmovaps 0(%0), %%ymm0\n\t"^M
73933 (gdb) FAIL: gdb.arch/i386-avx.exp: continue to breakpoint: \
73934 continue to first breakpoint in main
73937 The problem is that the vmovaps insn requires an 256-bit (or 32-byte) aligned
73938 address, and it's only 16-byte aligned:
73944 Fix this by using a sufficiently aligned address, using _Alignas.
73946 Compile using -std=gnu11 to support _Alignas.
73948 Likewise in gdb.arch/i386-sse.exp.
73950 Tested on x86_64-linux, with both gcc and clang.
73952 2021-12-06 Alan Modra <amodra@gmail.com>
73954 [GOLD] PowerPC64 inline plt sequences
73955 The fixes gold failures to handle inline PLT sequences properly.
73956 PowerPC gold was always turning these back into direct calls due to
73957 gsym->use_plt_offset() returning false. This is fixed for dynamic
73958 linking by correcting get_reference_flags, and for static linking by
73959 overriding use_plt_offset() in relocate(). The rest of the patch
73960 revolves around needing to create PLT entries for inline PLT calls
73961 when statically linking (for gcc -mlongcall). The lplt section
73962 handled that for local symbols, now it does globals too.
73964 * powerpc.cc (Target_powerpc::plt_off): Return proper section
73966 (Target_powerpc::symval_for_branch): Make public.
73967 (Target_powerpc::make_lplt_section): Add Symbol_table* param.
73969 (Target_powerpc::make_local_plt_entry): Likewise.
73970 (Target_powerpc::make_local_plt_entry): New variant for global syms.
73971 (Powerpc_relobj::do_relocate_sections): Don't write lplt contents.
73972 (Output_data_plt_powerpc::do_write): Write lplt contents here.
73973 (Output_data_plt_powerpc::Output_data_plt_powerpc): Save
73974 symbol table pointer. Adjust all uses.
73975 (Output_data_plt_powerpc::add_entry): Add stash parameter. Don't
73976 do dynamic reloc handling when no reloc section. Save symbol
73977 for local plt entries.
73978 (Output_data_plt_powerpc::add_local_entry): Save symbol.
73979 (Output_data_plt_powerpc::Local_plt_ent): New class.
73980 (Output_data_plt_powerpc::sym_ents_): New vector.
73981 (Target_powerpc::Scan::get_reference_flags): Return
73982 FUNCTION_CALL|RELATIVE_REF for inline plt relocs.
73983 (Target_powerpc::Scan::global): Make entries in lplt for inline
73984 plt call relocation symbols.
73985 (Target_powerpc::Relocate::relocate): Rename has_plt_offset to
73986 use_plt_offset. Set use_plt_offset for inline plt relocs.
73988 2021-12-06 Clément Chigot <clement.chigot@atos.net>
73990 ld: improve shared tests for AIX
73991 It's now possible to refer symbols in the main program from the
73992 shared library. However, it still impossible to have the same
73993 overriden features between shared objects and mains than ELF,
73994 without using the runtime linking feature which isn't yet fully
73999 * testsuite/ld-shared/shared.exp: Improve XCOFF support
74000 * testsuite/ld-shared/main.c: Likewise.
74001 * testsuite/ld-shared/sh1.c: Likewise.
74002 * testsuite/ld-shared/xcoff.dat: Likewise.
74004 2021-12-06 GDB Administrator <gdbadmin@sourceware.org>
74006 Automatic date update in version.in
74008 2021-12-05 Tom Tromey <tom@tromey.com>
74010 Preserve artificial CU name in process_psymtab_comp_unit_reader
74011 This fixes a use-after-free that Simon pointed out.
74012 process_psymtab_comp_unit_reader was allocating an artificial name for
74013 a CU, and then discarding it. However, this name was preserved in the
74014 cached file_and_directory. This patch arranges for the allocated name
74015 to be preserved there.
74017 2021-12-05 Mike Frysinger <vapier@gentoo.org>
74019 sim: include ansidecl.h when needed
74020 Avoid implicit include deps with this to help untangle sim headers
74021 so we can get rid of arch-specific sim-main.h.
74023 sim: include stdint.h when needed
74024 Avoid implicit include deps with this to help untangle sim headers
74025 so we can get rid of arch-specific sim-main.h.
74027 sim: include stdarg.h when used
74028 Avoid implicit include deps with this to help untangle sim headers
74029 so we can get rid of arch-specific sim-main.h.
74031 2021-12-05 Mike Frysinger <vapier@gentoo.org>
74033 sim: reorder header includes
74034 We're including system headers after local headers in a bunch of
74035 places, but this leads to conflicts when our local headers happen
74036 to define symbols that show up in the system headers.
74038 Use the more standard order of:
74039 * config.h (via defs.h)
74041 * local library headers (e.g. bfd & libiberty)
74042 * sim specific headers
74044 2021-12-05 Simon Marchi <simon.marchi@polymtl.ca>
74046 gdbsupport: fix memory leak in create_file_handler when re-using file handler
74047 ASan made me notice a memory leak, where the memory tied to the file
74048 handle name string wasn't freed. When register a file handler with an
74049 fd that is already registered, we re-use the file_handler object, so we
74050 ended up creating a new std::string object and overwriting the
74051 file_handler::name pointer, without free-ing the old std::string.
74053 Fix this by allocating file_handler with new, deleting it with
74054 delete, and making file_handler::name not a pointer.
74056 Change-Id: Ie304cc78ab5ae5dfad9a1366e9890c09de651f43
74058 2021-12-05 GDB Administrator <gdbadmin@sourceware.org>
74060 Automatic date update in version.in
74062 2021-12-04 Mike Frysinger <vapier@gentoo.org>
74064 sim: moxie: hoist dtb rules up to common builds
74065 These rules don't depend on the target compiler settings, so hoist
74066 the build logic up to the common builds for better parallelization.
74068 sim: m68hc11: delete unused profile flags
74069 These were moved to the common configure script a while ago and have
74070 the same default as these, so just delete it.
74072 sim: msp430: delete redundant comments & settings
74073 These were copied from the example docs, so aren't adding any value.
74075 sim: erc32: drop old configure target
74076 There is no configure script in here anymore to regenerate.
74078 sim: m32c/rl78: drop redundant -Wall settings
74079 We already turn these on in the configure script.
74081 2021-12-04 Tom Tromey <tom@tromey.com>
74083 Cache the result of find_file_and_directory
74084 This changes the DWARF reader to cache the result of
74085 find_file_and_directory. This is not especially important now, but it
74086 will help the new DWARF indexer.
74088 Move file_and_directory to new file and C++-ize
74089 This moves file_and_directory to a new file, and then C++-izes it --
74090 replacing direct assignments with methods, and arranging for it to own
74091 any string that must be computed. Finally, the CU's objfile will only
74092 be used on demand; this is an important property for the new DWARF
74093 indexer's parallel mode.
74095 Remove Irix case from find_file_and_directory
74096 find_file_and_directory has a special case for the Irix 6.2 compiler.
74097 Since this is long obsolete, this patch removes it.
74099 2021-12-04 Mike Frysinger <vapier@gentoo.org>
74101 sim: frv: split up testsuite a bit
74102 Running frv's allinsn in serial is quite slow due to the sheer number
74103 of tests it contains. By splitting it up and running in parallel, the
74104 execution time on my system goes from ~100sec to ~60sec.
74106 2021-12-04 Simon Marchi <simon.marchi@polymtl.ca>
74108 gdb: don't show deprecated aliases
74109 I don't think it's very useful to show deprecated aliases to the
74110 user. It encourages the user to use them, when the goal is the
74113 For example, before:
74115 (gdb) help set index-cache enabled
74116 set index-cache enabled, set index-cache off, set index-cache on
74117 alias set index-cache off = set index-cache enabled off
74118 alias set index-cache on = set index-cache enabled on
74119 Enable the index cache.
74120 When on, enable the use of the index cache.
74122 (gdb) help set index-cache on
74123 Warning: 'set index-cache on', an alias for the command 'set index-cache enabled', is deprecated.
74124 Use 'set index-cache enabled on'.
74126 set index-cache enabled, set index-cache off, set index-cache on
74127 alias set index-cache off = set index-cache enabled off
74128 alias set index-cache on = set index-cache enabled on
74129 Enable the index cache.
74130 When on, enable the use of the index cache.
74134 (gdb) help set index-cache enabled
74135 Enable the index cache.
74136 When on, enable the use of the index cache.
74137 (gdb) help set index-cache on
74138 Warning: 'set index-cache on', an alias for the command 'set index-cache enabled', is deprecated.
74139 Use 'set index-cache enabled on'.
74141 Enable the index cache.
74142 When on, enable the use of the index cache.
74144 Change-Id: I989b618a5ad96ba975367e9d16db95523cd57a4c
74146 2021-12-04 Simon Marchi <simon.marchi@polymtl.ca>
74148 gdb/testsuite: fix two "maint info line-table"-related tests
74149 Commit 92228a334ba2 ("gdb: small "maintenance info line-table"
74150 readability improvements") change the output format of "maint info
74151 line-table" slightly, adding some empty lines between each
74152 line-table. This causes two tests to start failing, update them to
74153 account for those empty lines.
74155 Change-Id: I9d33a58fce3e860ba0554b25f5582e8066a5c519
74157 2021-12-04 Simon Marchi <simon.marchi@efficios.com>
74159 gdb: revert one array_view copy change in ada-lang.c
74160 Commit 4bce7cdaf481 ("gdbsupport: add array_view copy function") caused
74161 an internal error when running gdb.ada/packed_array_assign.exp:
74163 print pra(1) := pr^M
74164 /home/smarchi/src/binutils-gdb/gdb/../gdbsupport/array-view.h:217: internal-error: copy: Assertion `dest.size () == src.size ()' failed.^M
74166 I am not sure what's the root cause of this, whether it is a GDB bug
74167 exposed by using the array_view copy function or not. Back out the
74168 change that triggers the internal error for now, while we investigate
74171 Change-Id: I055ab14143e4cfd3ca7ce8f4855c6c3c05db52a7
74173 2021-12-04 Mike Frysinger <vapier@gentoo.org>
74175 bfd: unify header generation rules
74176 The logic between these rules are extremely similar, so unify them
74177 into a single variable.
74179 2021-12-04 Mike Frysinger <vapier@gentoo.org>
74181 bfd: move header updates up a directory
74182 The rules for rebuilding the bfd headers live in the doc/ subdir
74183 (most likely) because they rely on the chew & related tools. But
74184 we can collapse them into the main Makefile while keeping the tools
74185 in the doc subdir easily enough. This makes the code simpler and
74186 allows for rebuilding them in parallel.
74188 Also add automake silent rule support while we're here.
74190 2021-12-04 Mike Frysinger <vapier@gentoo.org>
74192 bfd: convert bfdver.h to silent automake rules
74194 2021-12-04 GDB Administrator <gdbadmin@sourceware.org>
74196 Automatic date update in version.in
74198 2021-12-03 Simon Marchi <simon.marchi@polymtl.ca>
74200 gdb: small "maintenance info line-table" readability improvements
74201 - separate each entry with a newline, to visually separate them
74202 - style filenames with the filename style
74203 - print the name of the compunit_symtab
74205 A header now looks like this, with the compunit_symtab name added (and
74206 the coloring, but you can't really see it here):
74208 objfile: /home/simark/build/babeltrace/src/cli/.libs/babeltrace2 ((struct objfile *) 0x613000005980)
74209 compunit_symtab: babeltrace2-cfg-cli-args.c ((struct compunit_symtab *) 0x62100da1ed10)
74210 symtab: /usr/include/glib-2.0/glib/gdatetime.h ((struct symtab *) 0x62100d9ee530)
74211 linetable: ((struct linetable *) 0x0):
74213 Change-Id: Idc23e10aaa66e2e692adb0a6a74144f72c4fa1c7
74215 2021-12-03 Simon Marchi <simon.marchi@polymtl.ca>
74217 gdb: change some alias functions parameters to const-reference
74218 Now that we use intrusive list to link aliases, it becomes easier to
74219 pass cmd_list_element arguments by const-reference rather than by
74220 pointer to some functions, change a few.
74222 Change-Id: Id0df648ed26e9447da0671fc2c858981cda31df8
74224 2021-12-03 Simon Marchi <simon.marchi@polymtl.ca>
74226 gdb: use intrusive_list for cmd_list_element aliases list
74227 Change the manually-implemented linked list to use intrusive_list. This
74228 is not strictly necessary, but it makes the code much simpler.
74230 Change-Id: Idd08090ebf2db8bdcf68e85ef72a9635f1584ccc
74232 2021-12-03 Simon Marchi <simon.marchi@polymtl.ca>
74234 gdb: trivial changes to use array_view
74235 Change a few relatively obvious spots using value contents to propagate
74236 the use array_view a bit more.
74238 Change-Id: I5338a60986f06d5969fec803d04f8423c9288a15
74240 2021-12-03 Simon Marchi <simon.marchi@polymtl.ca>
74242 gdb: make extract_integer take an array_view
74243 I think it would make sense for extract_integer, extract_signed_integer
74244 and extract_unsigned_integer to take an array_view. This way, when we
74245 extract an integer, we can validate that we don't overflow the buffer
74246 passed by the caller (e.g. ask to extract a 4-byte integer but pass a
74249 - Change extract_integer to take an array_view
74250 - Add overloads of extract_signed_integer and extract_unsigned_integer
74251 that take array_views. Keep the existing versions so we don't
74252 need to change all callers, but make them call the array_view
74255 This shortens some places like:
74257 result = extract_unsigned_integer (value_contents (result_val).data (),
74258 TYPE_LENGTH (value_type (result_val)),
74263 result = extract_unsigned_integer (value_contents (result_val), byte_order);
74265 value_contents returns an array view that is of length
74266 `TYPE_LENGTH (value_type (result_val))` already, so the length is
74267 implicitly communicated through the array view.
74269 Change-Id: Ic1c1f98c88d5c17a8486393af316f982604d6c95
74271 2021-12-03 Simon Marchi <simon.marchi@polymtl.ca>
74273 gdbsupport: add array_view copy function
74274 An assertion was recently added to array_view::operator[] to ensure we
74275 don't do out of bounds accesses. However, when the array_view is copied
74276 to or from using memcpy, it bypasses that safety.
74278 To address this, add a `copy` free function that copies data from an
74279 array view to another, ensuring that the destination and source array
74280 views have the same size. When copying to or from parts of an
74281 array_view, we are expected to use gdb::array_view::slice, which does
74282 its own bounds check. With all that, any copy operation that goes out
74283 of bounds should be caught by an assertion at runtime.
74285 copy is implemented using std::copy and std::copy_backward, which, at
74286 least on libstdc++, appears to pick memmove when copying trivial data.
74287 So in the end there shouldn't be much difference vs using a bare memcpy,
74288 as we do right now. When copying non-trivial data, std::copy and
74289 std::copy_backward assigns each element in a loop.
74291 To properly support overlapping ranges, we must use std::copy or
74292 std::copy_backward, depending on whether the destination is before the
74293 source or vice-versa. std::copy and std::copy_backward don't support
74294 copying exactly overlapping ranges (where the source range is equal to
74295 the destination range). But in this case, no copy is needed anyway, so
74298 The order of parameters of the new copy function is based on std::copy
74299 and std::copy_backward, where the source comes before the destination.
74301 Change a few randomly selected spots to use the new function, to show
74302 how it can be used.
74304 Add a test for the new function, testing both with arrays of a trivial
74305 type (int) and of a non-trivial type (foo). Test non-overlapping
74306 ranges as well as three kinds of overlapping ranges: source before dest,
74307 dest before source, and dest == source.
74309 Change-Id: Ibeaca04e0028410fd44ce82f72e60058d6230a03
74311 2021-12-03 Simon Marchi <simon.marchi@efficios.com>
74313 gdb: change store_waitstatus to return a target_waitstatus by value
74314 store_waitstatus is basically a translation function between a status
74315 integer and an equivalent target_waitstatus object. It would make sense
74316 for it to take the integer as a parameter and return the
74317 target_waitstatus by value. Do that, and rename to
74318 host_status_to_waitstatus. Users can then do:
74320 ws = host_status_to_waitstatus (status)
74322 which does the right thing, given the move constructor of
74325 Change-Id: I7a07d59d3dc19d3ed66929642f82f44f3e85d61b
74327 2021-12-03 Simon Marchi <simon.marchi@efficios.com>
74329 gdb: return *this in target_waitstatus setters
74330 While playing with some code creating target_waitstatus objects, I was
74331 mildly annoyed by the fact that we can't just return a new
74332 target_waitstatus object. We have to do:
74334 target_waitstatus ws;
74335 ws.set_exited (123);
74338 Make the setters return the "this" object as a reference, such that it's
74341 return target_waitstatus ().set_exited (123);
74343 I initially thought of adding static creation functions, which you would
74346 return target_waitstatus::make_exited (123);
74348 However, making the setters return a reference to the object achieves
74349 pretty much the same thing, with less new code.
74351 Change-Id: I45159b7f9fcd9db5b20603480e323020b14ed147
74353 2021-12-03 Simon Marchi <simon.marchi@polymtl.ca>
74355 gdb: make saved_filename an std::string
74356 Make this variable an std::string, avoiding manual memory management.
74358 Change-Id: Ie7a8d7381449ab9c4dfc4cb8b99e63b9ffa8f947
74360 2021-12-03 Richard Sandiford <richard.sandiford@arm.com>
74362 aarch64: Fix uninitialised memory
74363 AARCH64_OPDE_EXPECTED_A_AFTER_B and AARCH64_OPDE_A_SHOULD_FOLLOW_B
74364 are not paired with an error string, but we had an assert that the
74365 error was nonnull. Previously this assert was testing uninitialised
74366 memory and so could pass or fail arbitrarily.
74369 * aarch64-opc.c (verify_mops_pme_sequence): Initialize the error
74370 field to null for AARCH64_OPDE_EXPECTED_A_AFTER_B and
74371 AARCH64_OPDE_A_SHOULD_FOLLOW_B.
74372 * aarch64-dis.c (print_verifier_notes): Move assert.
74374 2021-12-03 Andrew Burgess <andrew.burgess@embecosm.com>
74376 gdb: make value_subscripted_rvalue static
74377 The function value_subscripted_rvalue is only used in valarith.c, so
74378 lets make it a static function.
74380 There should be no user visible change after this commit.
74382 2021-12-03 Andrew Burgess <aburgess@redhat.com>
74384 gdb/testsuite: give a test a real name
74385 A test in gdb.python/py-send-packet.exp added in this commit:
74387 commit 24b2de7b776f8f23788d855b1eec290c6e208821
74388 Date: Tue Aug 31 14:04:36 2021 +0100
74390 gdb/python: add gdb.RemoteTargetConnection.send_packet
74392 included a large amount of binary data in the command sent to GDB. As
74393 this test didn't have a real test name the binary data was included in
74394 the gdb.sum file. The contents of the binary data could change
74395 between different runs of GDB, and this makes comparing results
74398 This commit gives the test a real test name.
74400 2021-12-03 Andrew Burgess <aburgess@redhat.com>
74402 gdb/remote: fix use after free bug
74405 commit 288712bbaca36bff6578bc839ebcdc3707662f81
74406 Date: Mon Nov 22 15:16:27 2021 +0000
74408 gdb/remote: use scoped_restore to control starting_up flag
74410 introduced a use after free bug. The scoped restore added in the
74411 above commit resets a flag within a remote_target's remote_state
74414 However, in some situations, the remote_target can be unpushed before
74415 the error is thrown. If the only reference to the target is the one
74416 in the target stack, then unpushing the target will cause the
74417 remote_target to be deleted, which, in turn, will delete the
74418 remote_state object. The scoped restore will then try to reset the
74419 flag within a deleted object.
74421 This problem was caught in the gdb.server/server-connect.exp test,
74422 which, when run with the address sanitizer enabled, highlights the
74423 write after free bug described above.
74425 This commit resolves this issue by adding a new class specifically for
74426 the purpose of managing the starting_up flag. As well as setting, and
74427 then clearing the starting_up flag, this new class increments, and
74428 then decrements the reference count on the remote_target object. This
74429 prevents the remote_target from being deleted until after the flag has
74432 The gdb.server/server-connect.exp now runs cleanly with the address
74435 2021-12-03 Mike Frysinger <vapier@gentoo.org>
74437 libctf: workaround automake bug with conditional info pages
74438 It looks like automake makes assumptions about its ability to build info
74439 pages based on the GNU standard behavior of shipping info pages with the
74440 distributions. So even though the info pages were conditionalized, and
74441 automake disabled some of the targets, it was still creeping in by way
74442 of unconditional INFO_DEPS settings.
74444 We can workaround this by adding a stub target for the info page when
74445 building info pages are disabled. This tricks automake into disabling
74446 its own extended generation target. I'll follow up with the automake
74447 folks to see what they think.
74449 2021-12-03 Chenghua Xu <xuchenghua@loongson.cn>
74451 Add myself and Zhensong Liu as the LoongArch port maintainer.
74453 2021-12-03 Alan Modra <amodra@gmail.com>
74455 Revert "Re: Don't compile some opcodes files when bfd is 32-bit only"
74456 This reverts commit 7a53275579e7cec9389ccb924f5ecf69e8d89d41.
74457 The bpf sim doesn't work with a 32-bit bfd after all.
74459 2021-12-03 GDB Administrator <gdbadmin@sourceware.org>
74461 Automatic date update in version.in
74463 2021-12-02 Simon Marchi <simon.marchi@polymtl.ca>
74465 gdb: remove unexpected xstrdup in _initialize_maint_test_settings
74466 That xstrdup is not correct, since we are assigning an std::string. The
74467 result of xstrdup is used to initialize the string, and then lost
74468 forever. Remove it.
74470 Change-Id: Ief7771055e4bfd643ef3b285ec9fb7b1bfd14335
74472 2021-12-02 Nick Clifton <nickc@redhat.com>
74474 Fix illegal memory access whilst parsing corrupt DWARF debug information.
74476 * dwarf.c (process_cu_tu_index): Add test for overruning section
74477 whilst processing slots.
74479 2021-12-02 Tom de Vries <tdevries@suse.de>
74481 [gdb/tdep] Fix avx512 -m32 support in gdbserver
74482 PR27257 reports a problem that can be reproduced as follows:
74483 - use x86_64 machine with avx512 support
74484 - compile a hello world with -m32 to a.out
74485 - start a gdbserver session with a.out
74486 - use gdb to connect to the gdbserver session
74488 This makes us run into:
74490 Listening on port 2346
74491 Remote debugging from host ::1, port 34940
74492 src/gdbserver/regcache.cc:257: \
74493 A problem internal to GDBserver has been detected.
74494 Unknown register zmm16h requested
74497 The problem is that i387_xsave_to_cache in gdbserver/i387-fp.cc can't find a
74498 register zmm16h in the register cache.
74500 To understand how this happens, first some background.
74502 SSE has 16 128-bit wide xmm registers.
74504 AVX extends the SSE registers set as follows:
74505 - it extends the 16 existing 128-bit wide xmm registers to 256-bit wide ymm
74508 AVX512 extends the AVX register set as follows:
74509 - it extends the 16 existing 256-bit wide ymm registers to 512-bit wide zmm
74511 - it adds 16 additional 512-bit wide zmm registers (with corresponding ymm and
74512 xmm subregisters added as well)
74514 However, in 32-bit mode, there are only 8 xmm/ymm/zmm registers.
74516 The problem we're running into is that gdbserver/i387-fp.cc uses these
74517 constants to describe the size of the register file:
74519 static const int num_avx512_zmmh_low_registers = 16;
74520 static const int num_avx512_zmmh_high_registers = 16;
74521 static const int num_avx512_ymmh_registers = 16;
74522 static const int num_avx512_xmm_registers = 16;
74524 which are all incorrect for the 32-bit case.
74526 Fix this by replacing the constants with variables that have the appropriate
74527 values in 64-bit and 32-bit mode.
74529 Tested on x86_64-linux with native and unix/-m32.
74531 2021-12-02 Simon Marchi <simon.marchi@polymtl.ca>
74533 gdb/testsuite: update tests looking for "DWARF 2" debug format
74534 Commit ab557072b8ec ("gdb: use actual DWARF version in compunit's
74535 debugformat field") changes the debug format string in "info source" to
74536 show the actual DWARF version, rather than always show "DWARF 2".
74538 However, it failed to consider that some tests checked for the "DWARF 2"
74539 string to see if the test program is compiled with DWARF debug
74540 information. Since everything is compiled with DWARF 4 or 5 nowadays,
74541 that changed the behavior of those tests. Notably, it prevent the
74542 tests using skip_inline_var_tests to run.
74544 Grep through the testsuite for "DWARF 2" and change all occurrences I
74545 could find to use "DWARF [0-9]" instead (that string is passed to TCL's
74548 Change-Id: Ic7fb0217fb9623880c6f155da6becba0f567a885
74550 2021-12-02 Joel Brobecker <brobecker@adacore.com>
74552 (PPC64) fix handling of fixed-point values when using "return" command
74553 In the gdb.ada/fixed_points_function.exp testcase, we have the following
74556 type FP1_Type is delta 0.1 range -1.0 .. +1.0; -- Ordinary
74557 function Call_FP1 (F : FP1_Type) return FP1_Type is
74563 ... used as follow:
74565 F1 : FP1_Type := 1.0;
74566 F1 := Call_FP1 (F1);
74568 The testcase, among other things, verifies that "return" works
74569 properly as follow:
74572 | Make pck.call_fp1 return now? (y or n) y
74574 | 9 F1 := Call_FP1 (F1);
74579 The output of the last command shows that we returned the wrong
74580 value. The value printed gives a clue about the problem, since
74581 it is 1/16th of the value we expected, where 1/16 is FP1_Type's
74584 The problem, here, comes from the fact that the function
74585 handling return values for base types (ppc64_sysv_abi_return_value_base)
74586 writes the return value using unpack_long which, upon seeing that
74587 the value being unpacked is a fixed point type, applies the scaling
74588 factor, to get the integer-representation of our fixed-point value
74589 (similar to what it does with floats, for instance).
74591 So, the fix consists in teaching ppc64_sysv_abi_return_value_base
74592 about fixed-point types, and to avoid the unwanted application
74593 of the scaling factor.
74595 Note that the "finish" function, on the other hand, does not
74596 suffer from this issue, simply becaue the value returned by
74597 the function is read from register without the use of a type,
74598 thus avoiding an unwanted application of a scaling factor.
74600 No test added, as this change is already tested by
74601 gdb.ada/fixed_points_function.exp.
74603 Co-Authored-By: Tristan Gingold <gingold@adacore.com>
74605 2021-12-02 Joel Brobecker <brobecker@adacore.com>
74607 (RISCV) fix handling of fixed-point type return values
74608 This commit adds support for TYPE_CODE_FIXED_POINT types for
74609 "finish" and "return" commands.
74611 Consider the following Ada code...
74613 type FP1_Type is delta 0.1 range -1.0 .. +1.0; -- Ordinary
74614 function Call_FP1 (F : FP1_Type) return FP1_Type is
74620 ... used as follow:
74622 F1 : FP1_Type := 1.0;
74623 F1 := Call_FP1 (F1);
74625 "finish" currently behaves as follow:
74629 | Value returned is $1 = 0
74631 We expect the returned value to be "1".
74633 Similarly, "return" makes the function return the wrong value:
74636 | Make pck.call_fp1 return now? (y or n) y
74638 | 9 F1 := Call_FP1 (F1);
74643 (we expect it to print "1" instead).
74645 This problem comes from the handling of integral return values
74646 when the return value is actually fixed point type. Our type
74647 here is actually a range of a fixed point type, but the same
74648 principles should also apply to pure fixed-point types. For
74649 the record, here is what the debugging info looks like:
74651 <1><238>: Abbrev Number: 2 (DW_TAG_subrange_type)
74652 <239> DW_AT_lower_bound : -16
74653 <23a> DW_AT_upper_bound : 16
74654 <23b> DW_AT_name : pck__fp1_type
74655 <23f> DW_AT_type : <0x248>
74657 <1><248>: Abbrev Number: 4 (DW_TAG_base_type)
74658 <249> DW_AT_byte_size : 1
74659 <24a> DW_AT_encoding : 13 (signed_fixed)
74660 <24b> DW_AT_binary_scale: -4
74661 <24c> DW_AT_name : pck__Tfp1_typeB
74662 <250> DW_AT_artificial : 1
74664 ... where the scaling factor is 1/16.
74666 Looking at the "finish" command, what happens is that riscv_arg_location
74667 determines that our return value should be returned by parameter using
74668 an integral convention (via builtin type long). And then,
74669 riscv_return_value uses a cast to that builtin type long to
74670 store the value of into a buffer with the right register size.
74671 This doesn't work in our case, because the underlying value
74672 returned by the function is unscaled, which means it is 16,
74673 and thus the cast is like doing:
74675 arg_val = (FP1_Type) 16
74677 ... In other words, it is trying to create an FP1_Type enty whose
74678 value is 16. Applying the scaling factor, that's 256, and because
74679 the size of FP1_Type is 1 byte, we overflow and thus it ends up
74682 The same happen with the "return" function, but the other way around.
74684 The fix consists in handling fixed-point types separately from
74687 2021-12-02 Joel Brobecker <brobecker@adacore.com>
74689 (ARM/fixed-point) wrong value shown by "finish" command:
74690 Consider the following Ada code:
74692 type FP1_Type is delta 0.1 range -1.0 .. +1.0; -- Ordinary
74693 FP1_Arg : FP1_Type := 0.0;
74695 function Call_FP1 (F : FP1_Type) return FP1_Type is
74701 After having stopped inside function Call_FP1 as follow:
74703 Breakpoint 1, pck.call_fp1 (f=1) at /[...]/pck.adb:5
74706 Returning from that function call using "finish" should show
74707 that the function return "1.0" (the same value as was passed
74708 as an argument). However, this is not the case:
74711 Run till exit from #0 pck.call_fp1 (f=1)
74713 9 F1 := Call_FP1 (F1);
74714 Value returned is $1 = 0
74716 This patch enhances the extraction of the return value to know about
74719 2021-12-02 Xavier Roirand <roirand@adacore.com>
74721 (Ada/AArch64) fix fixed point argument passing in inferior funcall
74722 Consider the following code:
74724 type FP1_Type is delta 0.1 range -1.0 .. +1.0; -- Ordinary
74726 function Call_FP1 (F : FP1_Type) return FP1_Type is
74731 When the default in GCC is to generate proper DWARF info for fixed point
74732 types, then in gdb, printing the result of a call to call_fp1 with a
74733 decimal parameter leads to:
74735 (gdb) p call_fp1(0.5)
74738 The displayed value is wrong, and we actually expected:
74740 (gdb) p call_fp1(0.5)
74743 What happened is that our fixed point type parameter got promoted to a
74744 32bit integer because we detected that the length of that object was less
74745 than 4 bytes. The compiler does not perform this promotion and therefore
74746 GDB should not either.
74748 This patch fixes the behavior described above.
74750 2021-12-02 Tom Tromey <tromey@adacore.com>
74752 Implement 'task apply'
74753 This adds a 'task apply' command, which is the Ada tasking analogue of
74754 'thread apply'. Unlike 'thread apply', it doesn't offer the
74755 'ascending' flag; but otherwise it's essentially the same.
74757 Add "task" keyword to the "watch" command
74758 Breakpoints in gdb can be made specific to an Ada task using the
74759 "task" qualifier. This patch applies this same idea to watchpoints.
74761 2021-12-02 Richard Sandiford <richard.sandiford@arm.com>
74763 aarch64: Update gas/NEWS for recent changes
74765 * NEWS: Mention support for Armv8.8-A and for new system registers.
74767 2021-12-02 Richard Sandiford <richard.sandiford@arm.com>
74769 aarch64: Add BC instruction
74770 This patch adds support for the Armv8.8-A BC instruction.
74771 [https://developer.arm.com/documentation/ddi0596/2021-09/Base-Instructions/BC-cond--Branch-Consistent-conditionally-?lang=en]
74774 * opcode/aarch64.h (AARCH64_FEATURE_HBC): New macro.
74775 (AARCH64_ARCH_V8_8): Make armv8.8-a imply AARCH64_FEATURE_HBC.
74778 * aarch64-tbl.h (aarch64_feature_hbc): New variable.
74779 (HBC, HBC_INSN): New macros.
74780 (aarch64_opcode_table): Add BC.C.
74781 * aarch64-dis-2.c: Regenerate.
74784 * doc/c-aarch64.texi: Document +hbc.
74785 * config/tc-aarch64.c (aarch64_features): Add "hbc".
74786 * testsuite/gas/aarch64/hbc.s, testsuite/gas/aarch64/hbc.d: New test.
74787 * testsuite/gas/aarch64/hbc-invalid.s,
74788 testsuite/gas/aarch64/hbc-invalid.l,
74789 testsuite/gas/aarch64/hbc-invalid.d: New test.
74791 2021-12-02 Richard Sandiford <richard.sandiford@arm.com>
74793 aarch64: Enforce P/M/E order for MOPS instructions
74794 The MOPS instructions should be used as a triple, such as:
74796 cpyfp [x0]!, [x1]!, x2!
74797 cpyfm [x0]!, [x1]!, x2!
74798 cpyfe [x0]!, [x1]!, x2!
74800 The registers should also be the same for each writeback operand.
74801 This patch adds a warning for code that doesn't follow this rule,
74802 along similar lines to the warning that we already emit for
74803 invalid uses of MOVPRFX.
74806 * opcode/aarch64.h (C_SCAN_MOPS_P, C_SCAN_MOPS_M, C_SCAN_MOPS_E)
74807 (C_SCAN_MOPS_PME): New macros.
74808 (AARCH64_OPDE_A_SHOULD_FOLLOW_B): New aarch64_operand_error_kind.
74809 (AARCH64_OPDE_EXPECTED_A_AFTER_B): Likewise.
74810 (aarch64_operand_error): Make each data value a union between
74811 an int and a string.
74814 * aarch64-tbl.h (MOPS_CPY_OP1_OP2_INSN): Add scan flags.
74815 (MOPS_SET_OP1_OP2_INSN): Likewise.
74816 * aarch64-opc.c (set_out_of_range_error): Update after change to
74817 aarch64_operand_error.
74818 (set_unaligned_error, set_reg_list_error): Likewise.
74819 (init_insn_sequence): Use a 3-instruction sequence for
74820 MOPS P instructions.
74821 (verify_mops_pme_sequence): New function.
74822 (verify_constraints): Call it.
74823 * aarch64-dis.c (print_verifier_notes): Handle
74824 AARCH64_OPDE_A_SHOULD_FOLLOW_B and AARCH64_OPDE_EXPECTED_A_AFTER_B.
74827 * config/tc-aarch64.c (operand_mismatch_kind_names): Add entries
74828 for AARCH64_OPDE_A_SHOULD_FOLLOW_B and AARCH64_OPDE_EXPECTED_A_AFTER_B.
74829 (operand_error_higher_severity_p): Check that
74830 AARCH64_OPDE_A_SHOULD_FOLLOW_B and AARCH64_OPDE_EXPECTED_A_AFTER_B
74831 come between AARCH64_OPDE_RECOVERABLE and AARCH64_OPDE_SYNTAX_ERROR;
74832 their relative order is not significant.
74833 (record_operand_error_with_data): Update after change to
74834 aarch64_operand_error.
74835 (output_operand_error_record): Likewise. Handle
74836 AARCH64_OPDE_A_SHOULD_FOLLOW_B and AARCH64_OPDE_EXPECTED_A_AFTER_B.
74837 * testsuite/gas/aarch64/mops_invalid_2.s,
74838 testsuite/gas/aarch64/mops_invalid_2.d,
74839 testsuite/gas/aarch64/mops_invalid_2.l: New test.
74841 2021-12-02 Richard Sandiford <richard.sandiford@arm.com>
74843 aarch64: Add support for +mops
74844 This patch adds support for FEAT_MOPS, an Armv8.8-A extension
74845 that provides memcpy and memset acceleration instructions.
74847 I took the perhaps controversial decision to generate the individual
74848 instruction forms using macros rather than list them out individually.
74849 This becomes useful with a follow-on patch to check that code follows
74850 the correct P/M/E sequence.
74851 [https://developer.arm.com/documentation/ddi0596/2021-09/Base-Instructions?lang=en]
74854 * opcode/aarch64.h (AARCH64_FEATURE_MOPS): New macro.
74855 (AARCH64_ARCH_V8_8): Make armv8.8-a imply AARCH64_FEATURE_MOPS.
74856 (AARCH64_OPND_MOPS_ADDR_Rd): New aarch64_opnd.
74857 (AARCH64_OPND_MOPS_ADDR_Rs): Likewise.
74858 (AARCH64_OPND_MOPS_WB_Rn): Likewise.
74861 * aarch64-asm.h (ins_x0_to_x30): New inserter.
74862 * aarch64-asm.c (aarch64_ins_x0_to_x30): New function.
74863 * aarch64-dis.h (ext_x0_to_x30): New extractor.
74864 * aarch64-dis.c (aarch64_ext_x0_to_x30): New function.
74865 * aarch64-tbl.h (aarch64_feature_mops): New feature set.
74866 (aarch64_feature_mops_memtag): Likewise.
74867 (MOPS, MOPS_MEMTAG, MOPS_INSN, MOPS_MEMTAG_INSN)
74868 (MOPS_CPY_OP1_OP2_PME_INSN, MOPS_CPY_OP1_OP2_INSN, MOPS_CPY_OP1_INSN)
74869 (MOPS_CPY_INSN, MOPS_SET_OP1_OP2_PME_INSN, MOPS_SET_OP1_OP2_INSN)
74870 (MOPS_SET_INSN): New macros.
74871 (aarch64_opcode_table): Add MOPS instructions.
74872 (aarch64_opcode_table): Add entries for AARCH64_OPND_MOPS_ADDR_Rd,
74873 AARCH64_OPND_MOPS_ADDR_Rs and AARCH64_OPND_MOPS_WB_Rn.
74874 * aarch64-opc.c (aarch64_print_operand): Handle
74875 AARCH64_OPND_MOPS_ADDR_Rd, AARCH64_OPND_MOPS_ADDR_Rs and
74876 AARCH64_OPND_MOPS_WB_Rn.
74877 (verify_three_different_regs): New function.
74878 * aarch64-asm-2.c: Regenerate.
74879 * aarch64-dis-2.c: Likewise.
74880 * aarch64-opc-2.c: Likewise.
74883 * doc/c-aarch64.texi: Document +mops.
74884 * config/tc-aarch64.c (parse_x0_to_x30): New function.
74885 (parse_operands): Handle AARCH64_OPND_MOPS_ADDR_Rd,
74886 AARCH64_OPND_MOPS_ADDR_Rs and AARCH64_OPND_MOPS_WB_Rn.
74887 (aarch64_features): Add "mops".
74888 * testsuite/gas/aarch64/mops.s, testsuite/gas/aarch64/mops.d: New test.
74889 * testsuite/gas/aarch64/mops_invalid.s,
74890 * testsuite/gas/aarch64/mops_invalid.d,
74891 * testsuite/gas/aarch64/mops_invalid.l: Likewise.
74893 2021-12-02 Richard Sandiford <richard.sandiford@arm.com>
74895 aarch64: Add Armv8.8-A system registers
74896 Armv8.8-A defines two new system registers: allint and icc_nmiar1_el1.
74897 Both of them were previously unmapped. allint supports a 0/1 immediate.
74898 [https://developer.arm.com/documentation/ddi0595/2021-09/AArch64-Registers/ALLINT--All-Interrupt-Mask-Bit?lang=en]
74899 [https://developer.arm.com/documentation/ddi0595/2021-09/AArch64-Registers/ICC-NMIAR1-EL1--Interrupt-Controller-Non-maskable-Interrupt-Acknowledge-Register-1?lang=en]
74902 * aarch64-opc.c (SR_V8_8): New macro.
74903 (aarch64_sys_regs): Add allint and icc_nmiar1_el1.
74904 (aarch64_pstatefields): Add allint.
74907 * testsuite/gas/aarch64/armv8_8-a-sysregs.s,
74908 * testsuite/gas/aarch64/armv8_8-a-sysregs.d: New test.
74909 * testsuite/gas/aarch64/armv8_8-a-sysregs-invalid.s,
74910 * testsuite/gas/aarch64/armv8_8-a-sysregs-invalid.l,
74911 * testsuite/gas/aarch64/armv8_8-a-sysregs-invalid.d: New test.
74913 2021-12-02 Richard Sandiford <richard.sandiford@arm.com>
74915 aarch64: Add id_aa64isar2_el1
74916 Armv8.8-A defines a read-only system register called id_aa64isar2_el1.
74917 The register was previously RES0 and should therefore be accepted
74918 at all architecture levels.
74919 [https://developer.arm.com/documentation/ddi0595/2021-09/AArch64-Registers/ID-AA64ISAR2-EL1--AArch64-Instruction-Set-Attribute-Register-2?lang=en]
74922 * aarch64-opc.c (aarch64_sys_regs): Add id_aa64isar2_el1.
74925 * testsuite/gas/aarch64/sysreg-diagnostic.s: Test writes to
74927 * testsuite/gas/aarch64/sysreg-diagnostic.d: Update accordingly.
74928 * testsuite/gas/aarch64/sysreg-diagnostic.l: Likewise.
74929 * testsuite/gas/aarch64/sysreg.s: Test reads from
74931 * testsuite/gas/aarch64/sysreg.d: Update accordingly.
74933 2021-12-02 Richard Sandiford <richard.sandiford@arm.com>
74935 aarch64: Add support for Armv8.8-A
74936 This patch adds skeleton support for -march=armv8.8-a, testing only
74937 that it correctly inherits from armv8.7-a.
74940 * opcode/aarch64.h (AARCH64_FEATURE_V8_8): New macro.
74941 (AARCH64_ARCH_V8_8): Likewise.
74944 * doc/c-aarch64.texi: Document armv8.8-a.
74945 * config/tc-aarch64.c (aarch64_archs): Add armv8-8-a
74946 * testsuite/gas/aarch64/v8-8-a.s,
74947 * testsuite/gas/aarch64/v8-8-a.d: New test.
74949 2021-12-02 Richard Sandiford <richard.sandiford@arm.com>
74951 aarch64: Provide line info for unclosed sequences
74952 We warn about MOVPRFX instructions that have no following
74953 instruction. This patch adds a line number to the message,
74954 which is useful if the assembly code has multiple text sections.
74956 The new code is unconditional since OBJ_ELF is always defined
74960 * config/tc-aarch64.h (aarch64_segment_info_type): Add last_file
74962 * config/tc-aarch64.c (now_instr_sequence): Delete.
74963 (force_automatic_sequence_close): Provide a line number when
74964 reporting unclosed sequences.
74965 (md_assemble): Record the location of the instruction in
74967 * testsuite/gas/aarch64/sve-movprfx_4.l: Add line number to error
74969 * testsuite/gas/aarch64/sve-movprfx_7.l: Likewise.
74970 * testsuite/gas/aarch64/sve-movprfx_8.l: Likewise.
74972 2021-12-02 Richard Sandiford <richard.sandiford@arm.com>
74974 aarch64: Tweak insn sequence code
74975 libopcodes has some code to check constraints across sequences
74976 of consecutive instructions. It was added to support MOVPRFX
74977 sequences but is going to be useful for the Armv8.8-A MOPS
74980 Currently the structure has one field to record the instruction
74981 that started a sequence and another to record the remaining
74982 instructions in the sequence. It's more convenient for the
74983 MOPS code if we put the instructions into a single array instead.
74985 No functional change intended.
74988 * opcode/aarch64.h (aarch64_instr_sequence): Replace num_insns
74989 and current_insns with num_added_insns and num_allocated_insns.
74992 * aarch64-opc.c (add_insn_to_sequence): New function.
74993 (init_insn_sequence): Update for new aarch64_instr_sequence layout.
74994 Add the first instruction to the inst array.
74995 (verify_constraints): Update for new aarch64_instr_sequence layout.
74996 Don't add the last instruction to the array.
74998 2021-12-02 Richard Sandiford <richard.sandiford@arm.com>
75000 aarch64: Add maximum immediate value to aarch64_sys_reg
75001 The immediate form of MSR has a 4-bit immediate field (in CRm).
75002 However, many forms of MSR require a smaller immediate. These cases
75003 are identified by value in operand_general_constraint_met_p,
75004 but they're now the common case rather than the exception.
75006 This patch therefore adds the maximum value to the sys_reg
75007 description and gets the range from there. It also enforces
75008 the minimum of 0, which avoids a situation in which:
75012 would give the expected:
75014 Error: immediate value out of range 0 to 1
75022 Error: immediate value out of range 0 to 15
75024 (from the later UIMM4 checking).
75028 - we were reporting the first error above against the wrong operand
75029 - TCO takes a single-bit immediate, but we previously allowed
75031 [https://developer.arm.com/documentation/ddi0596/2021-09/Base-Instructions/MSR--immediate---Move-immediate-value-to-Special-Register-?lang=en]
75034 * aarch64-opc.h (F_REG_MAX_VALUE, F_GET_REG_MAX_VALUE): New macros.
75035 * aarch64-opc.c (operand_general_constraint_met_p): Read the
75036 maximum MSR immediate value from aarch64_pstatefields.
75037 (aarch64_pstatefields): Add the maximum immediate value
75041 * testsuite/gas/aarch64/sysreg-4.s: Use an immediate value of 1
75042 rather than 8 for the TCO test.
75043 * testsuite/gas/aarch64/sysreg-4.d: Update accordingly.
75044 * testsuite/gas/aarch64/armv8_2-a-illegal.l: Fix operand number
75045 in MSR immediate error messages.
75046 * testsuite/gas/aarch64/diagnostic.l: Likewise.
75047 * testsuite/gas/aarch64/pan-illegal.l: Likewise.
75048 * testsuite/gas/aarch64/ssbs-illegal1.l: Likewise.
75049 * testsuite/gas/aarch64/illegal-sysreg-4b.s,
75050 * testsuite/gas/aarch64/illegal-sysreg-4b.d,
75051 * testsuite/gas/aarch64/illegal-sysreg-4b.l: New test.
75053 2021-12-02 Marcus Nilsson <brainbomb@gmail.com>
75055 Allow the --visualize-jumps feature to work with the AVR disassembler.
75056 * avr-dis.c (avr_operand); Pass in disassemble_info and fill
75057 in insn_type on branching instructions.
75059 2021-12-02 Simon Marchi <simon.marchi@polymtl.ca>
75061 gdb, include: replace pragmas with DIAGNOSTIC macros, fix build with g++ 4.8
75062 When introducing this code, I forgot that we had some macros for this.
75063 Replace some "manual" pragma diagnostic with some DIAGNOSTIC_* macros,
75064 provided by include/diagnostics.h.
75068 - Add DIAGNOSTIC_ERROR, to enable a diagnostic at error level.
75069 - Add DIAGNOSTIC_ERROR_SWITCH, to enable -Wswitch at error level, for
75070 both gcc and clang.
75072 Additionally, using DIAGNOSTIC_PUSH, DIAGNOSTIC_ERROR_SWITCH and
75073 DIAGNOSTIC_POP seems to misbehave with g++ 4.8, where we see these
75077 /home/smarchi/src/binutils-gdb/gdb/ada-tasks.c: In function void read_known_tasks():
75078 /home/smarchi/src/binutils-gdb/gdb/ada-tasks.c:998:10: error: enumeration value ADA_TASKS_UNKNOWN not handled in switch [-Werror=switch]
75079 switch (data->known_tasks_kind)
75082 Because of the POP, the diagnostic should go back to being disabled,
75083 since it was disabled in the beginning, but that's not what we see
75084 here. Versions of GCC >= 5 compile correctly.
75086 Work around this by making DIAGNOSTIC_ERROR_SWITCH a no-op for GCC < 5.
75088 Note that this code (already as it exists in master today) enables
75089 -Wswitch at the error level even if --disable-werror is passed. It
75090 shouldn't be a problem, as it's not like a new enumerator will appear
75091 out of nowhere and cause a build error if building with future
75092 compilers. Still, for correctness, we would ideally want to ask the
75093 compiler to enable -Wswitch at its default level (as if the user had
75094 passed -Wswitch on the command-line). There doesn't seem to be a way to
75097 Change-Id: Id33ebec3de39bd449409ea0bab59831289ffe82d
75099 2021-12-02 Simon Marchi <simon.marchi@polymtl.ca>
75101 gas: re-generate configure
75102 When configuring gas, I get:
75104 config.status: error: cannot find input file: `doc/Makefile.in'
75106 This is because configure is out-of-date, re-generate it.
75108 Change-Id: Iaa5980c282900d9fd23b90f0df25bf8ba3676498
75110 2021-12-02 Simon Marchi <simon.marchi@polymtl.ca>
75112 libctf: re-generate configure
75113 When configuring libctf, I get:
75115 config.status: error: cannot find input file: `doc/Makefile.in'
75117 This is because configure is out-of-date, re-generate it.
75119 Change-Id: Ie69acd33012211a4620661582c7d24ad6d2cd169
75121 2021-12-02 H.J. Lu <hjl.tools@gmail.com>
75123 x86: Skip __[start|stop]_SECNAME for --gc-sections -z start-stop-gc
75124 Don't convert memory load to immediate load on __start_SECNAME and
75125 __stop_SECNAME for --gc-sections -z start-stop-gc if all SECNAME
75126 sections been garbage collected.
75131 * elf32-i386.c (elf_i386_convert_load_reloc): Skip __start_SECNAME
75132 and __stop_SECNAME for --gc-sections -z start-stop-gc if the input
75133 section been garbage collected.
75134 * elf64-x86-64.c (elf_x86_64_convert_load_reloc): Likewise.
75135 * elfxx-x86.h (elf_x86_start_stop_gc_p): New function.
75139 * testsuite/ld-i386/i386.exp: Run PR ld/27491 tests.
75140 * testsuite/ld-x86-64/x86-64.exp: Likewise.
75141 * testsuite/ld-i386/pr27491-1.s: New file.
75142 * testsuite/ld-i386/pr27491-1a.d: Likewise.
75143 * testsuite/ld-i386/pr27491-1b.d: Likewise.
75144 * testsuite/ld-i386/pr27491-1c.d: Likewise.
75145 * testsuite/ld-i386/pr27491-2.d: Likewise.
75146 * testsuite/ld-i386/pr27491-2.s: Likewise.
75147 * testsuite/ld-i386/pr27491-3.d: Likewise.
75148 * testsuite/ld-i386/pr27491-3.s: Likewise.
75149 * testsuite/ld-i386/pr27491-4.d: Likewise.
75150 * testsuite/ld-i386/pr27491-4a.s: Likewise.
75151 * testsuite/ld-i386/pr27491-4b.s: Likewise.
75152 * testsuite/ld-x86-64/pr27491-1.s: Likewise.
75153 * testsuite/ld-x86-64/pr27491-1a.d: Likewise.
75154 * testsuite/ld-x86-64/pr27491-1b.d: Likewise.
75155 * testsuite/ld-x86-64/pr27491-1c.d: Likewise.
75156 * testsuite/ld-x86-64/pr27491-2.d: Likewise.
75157 * testsuite/ld-x86-64/pr27491-2.s: Likewise.
75158 * testsuite/ld-x86-64/pr27491-3.d: Likewise.
75159 * testsuite/ld-x86-64/pr27491-3.s: Likewise.
75160 * testsuite/ld-x86-64/pr27491-4.d: Likewise.
75161 * testsuite/ld-x86-64/pr27491-4a.s: Likewise.
75162 * testsuite/ld-x86-64/pr27491-4b.s: Likewise.
75164 2021-12-02 Mike Frysinger <vapier@gentoo.org>
75166 bfd: delete unused proto settings
75167 These have been around for decades but don't appear to be used, and
75168 trying to build them (e.g. `make archive.p archive.ip`) doesn't work,
75169 so just delete it all.
75171 gas: merge doc subdir up a level
75172 This avoids a recursive make into the doc subdir and speeds up the
75173 build slightly. It also allows for more parallelism.
75175 libctf: merge doc subdir up a level
75176 This avoids a recursive make into the doc subdir and speeds up the
75177 build slightly. It also allows for more parallelism.
75179 2021-12-02 Simon Marchi <simon.marchi@polymtl.ca>
75181 gdb: use actual DWARF version in compunit's debugformat field
75182 The "info source" command, with a DWARF-compile program, always show
75183 that the debug info is "DWARF 2":
75186 Current source file is test.c
75187 Compilation directory is /home/smarchi/build/binutils-gdb/gdb
75188 Located in /home/smarchi/build/binutils-gdb/gdb/test.c
75190 Source language is c.
75191 Producer is GNU C17 9.3.0 -mtune=generic -march=x86-64 -g3 -gdwarf-5 -O0 -fasynchronous-unwind-tables -fstack-protector-strong -fstack-clash-protection -fcf-protection.
75192 Compiled with DWARF 2 debugging format.
75193 Includes preprocessor macro info.
75195 Change it to display the actual DWARF version:
75198 Current source file is test.c
75199 Compilation directory is /home/smarchi/build/binutils-gdb/gdb
75200 Located in /home/smarchi/build/binutils-gdb/gdb/test.c
75202 Source language is c.
75203 Producer is GNU C17 9.3.0 -mtune=generic -march=x86-64 -g3 -gdwarf-5 -O0 -fasynchronous-unwind-tables -fstack-protector-strong -fstack-clash-protection -fcf-protection.
75204 Compiled with DWARF 5 debugging format.
75205 Includes preprocessor macro info.
75207 The comp_unit_head::version field is guaranteed to be between 2 and 5,
75208 thanks to the check in read_comp_unit_head. So we can still use static
75209 strings to pass to record_debugformat, and keep it efficient.
75211 In the future, when somebody will update GDB to support DWARF 6, they'll
75212 hit this assert and have to update this code.
75214 Change-Id: I3270b7ebf5e9a17b4215405bd2e365662a4d6172
75216 2021-12-02 H.J. Lu <hjl.tools@gmail.com>
75218 elf: Discard input .note.gnu.build-id sections
75219 1. Discard input .note.gnu.build-id sections.
75220 2. Clear the build ID field before writing.
75221 3. Use bfd_make_section_anyway_with_flags to create the output
75222 .note.gnu.build-id section.
75225 * ldelf.c (ldelf_after_open): Discard input .note.gnu.build-id
75226 sections, excluding the first one.
75227 (write_build_id): Clear the build ID field before writing.
75228 (ldelf_setup_build_id): Use bfd_make_section_anyway_with_flags
75229 to create the output .note.gnu.build-id section.
75230 * testsuite/ld-elf/build-id.exp: New file.
75231 * testsuite/ld-elf/pr28639a.rd: Likewise.
75232 * testsuite/ld-elf/pr28639b.rd: Likewise.
75233 * testsuite/ld-elf/pr28639c.rd: Likewise.
75234 * testsuite/ld-elf/pr28639d.rd: Likewise.
75236 2021-12-02 GDB Administrator <gdbadmin@sourceware.org>
75238 Automatic date update in version.in
75240 2021-12-01 Mike Frysinger <vapier@gentoo.org>
75242 binutils: add missing prefix for binutils/index.html rule
75244 2021-12-01 Luca Boccassi <luca.boccassi@gmail.com>
75246 readelf: recognize FDO Packaging Metadata ELF note. (Correcting snafu during patch application)
75248 2021-12-01 Luca Boccassi <luca.boccassi@gmail.com>
75250 readelf: recognize FDO Packaging Metadata ELF note
75251 As defined on: https://systemd.io/COREDUMP_PACKAGE_METADATA/
75252 this note will be used starting from Fedora 36. Allow
75253 readelf --notes to pretty print it:
75255 Displaying notes found in: .note.package
75256 Owner Data size Description
75257 FDO 0x00000039 FDO_PACKAGING_METADATA
75258 Packaging Metadata: {"type":"deb","name":"fsverity-utils","version":"1.3-1"}
75260 2021-12-01 Tom de Vries <tdevries@suse.de>
75262 [gdb/testsuite] Fix typo in gdb.multi/multi-arch-exec.exp
75263 With gdb.multi/multi-arch-exec.exp I run into:
75265 Running src/gdb/testsuite/gdb.multi/multi-arch-exec.exp ...
75266 ERROR: tcl error sourcing src/gdb/testsuite/gdb.multi/multi-arch-exec.exp.
75267 ERROR: wrong # args: extra words after "else" clause in "if" command
75269 "if [istarget "powerpc64*-*-*"] {
75271 } else if [istarget "s390*-*-*"] {
75278 Fix the else if -> elseif typo.
75280 Tested on x86_64-linux.
75282 2021-12-01 Tom de Vries <tdevries@suse.de>
75284 [gdb/testsuite] Fix gdb.arch/i386-pkru.exp on linux
75285 When running test-case gdb.arch/i386-pkru.exp on a machine with "Memory
75286 Protection Keys for Userspace" support, we run into:
75288 (gdb) PASS: gdb.arch/i386-pkru.exp: probe PKRU support
75291 (gdb) FAIL: gdb.arch/i386-pkru.exp: pkru register
75294 The test-case expects the $pkru register to have the default value 0, matching
75295 the "init state" of 0 defined by the XSAVE hardware.
75297 Since linux kernel version v4.9 containing commit acd547b29880 ("x86/pkeys:
75298 Default to a restrictive init PKRU"), the register is set to 0x55555554 by
75299 default (which matches the printed decimal value above).
75301 Fix the FAIL by accepting this value for linux.
75303 Tested on x86_64-linux.
75305 2021-12-01 Nick Clifton <nickc@redhat.com>
75307 Fix the fields in the x_n union inside the the x_file structure so that pointers can be stored.
75309 * coff/internal.h (x_n): Use bfd_hostptr_t for the fields in this
75312 2021-12-01 Andrew Burgess <aburgess@redhat.com>
75314 gdb/remote: use scoped_restore to control starting_up flag
75315 This commit makes use of a scoped_restore object to control the
75316 remote_state::starting_up flag within the remote_target::start_remote
75319 Ideally I would have liked to create the scoped_restore inside
75320 start_remote and just leave the restore in place until the end of the
75321 scope, however, I'm worried that doing this would change the behaviour
75322 of GDB. Specifically, in start_remote, the following code is executed
75323 once the starting_up flag has been restored to its previous value:
75325 if (breakpoints_should_be_inserted_now ())
75326 insert_breakpoints ();
75328 I think (but am not 100% sure) that calling install_breakpoints could
75329 end up back inside remote_target::can_download_tracepoint, which does
75330 check the value of remote_state::starting_up. And so, I'm concerned
75331 that leaving the scoped_restore in place until the end of start_remote
75332 will cause a possible change in behaviour.
75334 To avoid this, and to leave things as close to the current behaviour
75335 as possible, I've split remote_target::start_remote into two, there's
75336 the main function body which moves into remote_target::start_remote_1,
75337 this function uses the scoped_restore to change the ::starting_up
75338 flag, then there's the old remote_target::start_remote, which now just
75339 calls ::start_remote_1, and then does the insert_breakpoints call.
75341 There should be no user visible changes after this commit, unless
75342 there's a situation where the ::starting_up flag could previously have
75343 been left set, if this was the case, then this situation should no
75344 longer be possible.
75346 2021-12-01 Simon Marchi <simon.marchi@polymtl.ca>
75348 gdb.base/corefile-buildid.exp: fix DUPLICATEs when failing to generate a core file
75349 When my system isn't properly configured to generate core files in the
75350 local directory, I see these DUPLICATEs:
75352 DUPLICATE: gdb.base/corefile-buildid.exp: could not generate core file
75354 Fix that by having a single with_test_prefix around that message and
75357 Change-Id: I4ac245fcce1c666db56e3bad3582aa17f183dcba
75359 2021-12-01 Mike Frysinger <vapier@gentoo.org>
75361 gold: enable silent build rules
75363 2021-12-01 GDB Administrator <gdbadmin@sourceware.org>
75365 Automatic date update in version.in
75367 2021-11-30 Carl Love <cel@us.ibm.com>
75369 gdb: Powerpc fix gdb.multi/multi-arch-exec.exp test
75370 The expect file has a procedure append_arch_options which sets march based
75371 the istarget. The current if / else statement does not check for
75372 powerpc64. The else statement is hit which sets march to -m32. This
75373 results in compilation errors on 64-bit PowerPC.
75375 This patch adds an if statement to check for powerpc64 and if true sets mach
75378 The patch was tested on a Power 10 system. No compile errors were generated.
75379 The test completes with 1 expected pass and no failures.
75381 2021-11-30 Mike Frysinger <vapier@gentoo.org>
75383 binutils: regenerate Makefile.in after doc/ changes
75385 2021-11-30 Roland McGrath <mcgrathr@google.com>
75387 Fix missing build dependency for binutils man pages
75389 * doc/local.mk: Give each man page target its missing dependency on
75390 doc/$(am__dirstamp).
75392 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
75394 aarch64: Add missing system registers [PR27145]
75395 This patch adds support for various system registers, up to Armv8.7-A.
75396 This includes all the registers that were mentioned in the PR and that
75397 hadn't become supported since.
75401 * aarch64-opc.c (SR_V8_4): Remove duplicate definition.
75402 (SR_V8_6, SR_V8_7, SR_GIC, SR_AMU): New macros.
75403 (aarch64_sys_regs): Add missing entries (up to Armv8.7-A).
75407 * testsuite/gas/aarch64/sysreg-8.s,
75408 * testsuite/gas/aarch64/sysreg-8.d,
75409 * testsuite/gas/aarch64/illegal-sysreg-8.s,
75410 * testsuite/gas/aarch64/illegal-sysreg-8.d,
75411 * testsuite/gas/aarch64/illegal-sysreg-8.l,
75412 * testsuite/gas/aarch64/illegal-sysreg-8b.s,
75413 * testsuite/gas/aarch64/illegal-sysreg-8b.d,
75414 * testsuite/gas/aarch64/illegal-sysreg-8b.l: New tests.
75415 * testsuite/gas/aarch64/sysreg.s: Change system register numbers
75416 to ones that are still unallocated.
75417 * testsuite/gas/aarch64/sysreg.d: Update accordingly.
75419 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
75421 aarch64: Make LOR registers conditional on +lor
75422 We have a +lor feature flag for the Limited Ordering Regions
75423 extension, but the associated registers didn't use it.
75426 * aarch64-opc.c (SR_LOR): New macro.
75427 (aarch64_sys_regs): Use it for lorc_el1, lorea_el1, lorn_el1 and
75431 * testsuite/gas/aarch64/sysreg-7.s: Enable +lor.
75432 * testsuite/gas/aarch64/illegal-sysreg-7.s: Test for LOR registers
75434 * testsuite/gas/aarch64/illegal-sysreg-7.d: Update accordingly.
75435 * testsuite/gas/aarch64/illegal-sysreg-7.l: Likewise.
75437 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
75439 aarch64: Remove ZIDR_EL1
75440 ZIDR_EL1 was part of an early version of SVE, but didn't make
75441 it to the final release.
75444 * aarch64-opc.c (aarch64_sys_regs): Remove zidr_el1 entry.
75447 * testsuite/gas/aarch64/sve-sysreg.s: Remove zidr_el1.
75448 * testsuite/gas/aarch64/sve-sysreg.d: Update accordingly.
75449 * testsuite/gas/aarch64/sve-sysreg-invalid.l: Likewise.
75451 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
75453 aarch64: Allow writes to MFAR_EL3
75454 MFAR_EL3 is a read/write register, but was incorrectly marked as
75456 [https://developer.arm.com/documentation/ddi0601/2021-09/AArch64-Registers/MFAR-EL3--PA-Fault-Address-Register?lang=en]
75459 * aarch64-opc.c (aarch64_sys_regs): Mark mfar_el3 as read-write.
75462 * testsuite/gas/aarch64/rme.s: Test writing to mfar_el3.
75463 * testsuite/gas/aarch64/rme.d: Update accordingly.
75464 * testsuite/gas/aarch64/rme-invalid.s: Delete.
75465 * testsuite/gas/aarch64/rme-invalid.l: Likewise.
75466 * testsuite/gas/aarch64/rme-invalid.d: Likewise.
75468 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
75470 aarch64: Mark PMSIDR_EL1 as read-only
75471 We were incorrectly allowing writes to PMSIDR_EL1, which is
75472 a read-only register.
75473 [https://developer.arm.com/documentation/ddi0595/2021-09/AArch64-Registers/PMSIDR-EL1--Sampling-Profiling-ID-Register?lang=en]
75476 * aarch64-opc.c (aarch64_sys_regs): Make pmsidr_el1 as F_REG_READ.
75479 * testsuite/gas/aarch64/msr.s: Remove write to pmsidr_el1.
75480 * testsuite/gas/aarch64/msr.d: Update accordingly.
75481 * testsuite/gas/aarch64/illegal-sysreg-2.s,
75482 * testsuite/gas/aarch64/illegal-sysreg-2.d,
75483 * testsuite/gas/aarch64/illegal-sysreg-2.l: New test.
75485 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
75487 aarch64: Remove duplicate system register entries
75488 There is a lot of overlap between the ETM and ETE system registers,
75489 so some registers were listed twice.
75491 Already tested by etm.[sd] and ete.[sd].
75494 * aarch64-opc.c (aarch64_sys_regs): Combine ETE and ETM blocks
75495 and remove redundant entries.
75498 * testsuite/gas/aarch64/etm.s: Remove duplicated test.
75499 * testsuite/gas/aarch64/etm.d: Update accordingly.
75501 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
75503 aarch64: Check for register aliases before mnemonics
75504 Previously we would not accept:
75508 if A happened to be the name of an instruction. Adding new
75509 instructions could therefore invalidate existing register aliases.
75511 I noticed this with a test that used "zero" as a register alias
75512 for "xzr", where "zero" is now also the name of an SME instruction.
75513 I don't have any evidence that "real" code is doing this, but it
75514 seems at least plausible.
75516 This patch switches things so that we check for register aliases
75517 first. It might slow down parsing slightly, but the difference
75518 is unlikely to be noticeable.
75524 still work, since create_register_alias checks for " .req ",
75525 and with the input scrubber, we'll only keep whitespace after
75526 .req if it's followed by another name. If there's some valid
75527 expression that I haven't thought about that is scrubbed to
75528 " .req ", users could avoid the ambiguity by wrapping .req
75531 The new test for invalid aliases already passed. I just wanted
75532 something to exercise the !dot condition.
75534 I can't find a way of exercising the (existing) p == base condition,
75535 but I'm not brave enough to say that it can never happen. If it does
75536 happen, get_mnemonic_name would return an empty string.
75539 * config/tc-aarch64.c (opcode_lookup): Move mnemonic extraction
75541 (md_assemble): ...here. Check for register aliases first.
75542 * testsuite/gas/aarch64/register_aliases.d,
75543 testsuite/gas/aarch64/register_aliases.s: Test for a register
75544 alias called "zero".
75545 * testsuite/gas/aarch64/register_aliases_invalid.d,
75546 testsuite/gas/aarch64/register_aliases_invalid.l,
75547 testsuite/gas/aarch64/register_aliases_invalid.s: New test.
75549 2021-11-30 Andrew Burgess <aburgess@redhat.com>
75551 gdb/python: don't use the 'p' format for parsing args
75552 When running the gdb.python/py-arch.exp tests on a GDB built
75553 against Python 2 I ran into some errors. The problem is that this
75554 test script exercises the gdb.Architecture.integer_type method, and
75555 this method uses 'p' as an argument format specifier in a call to
75556 gdb_PyArg_ParseTupleAndKeywords.
75558 Unfortunately this specified was only added in Python 3.3, so will
75559 cause an error for earlier versions of Python.
75561 This commit switches to use the 'O' specifier to collect a PyObject,
75562 and then uses PyObject_IsTrue to convert the object to a boolean.
75564 An earlier version of this patch incorrectly switched from using 'p'
75565 to use 'i', however, it was pointed out during review that this would
75566 cause some changes in behaviour, for example both of these will work
75567 with 'p', but not with 'i':
75569 gdb.selected_inferior().architecture().integer_type(32, None)
75570 gdb.selected_inferior().architecture().integer_type(32, "foo")
75572 The new approach of using 'O' works fine with these cases. I've added
75573 some new tests to cover both of the above.
75575 There should be no user visible changes after this commit.
75577 2021-11-30 Tom de Vries <tdevries@sdflex.arch.suse.de>
75579 [gdb/testsuite] Fix gdb.base/style.exp with stub-termcap
75580 When running test-case gdb.base/style.exp with a gdb build using
75581 stub-termcap.c, we run into:
75583 (gdb) PASS: gdb.base/style.exp: all styles enabled: frame when width=20
75585 (gdb) FAIL: gdb.base/style.exp: all styles enabled: set width 30
75588 The problem is that we're trying to issue the command "set width 30" while
75589 width is set to 20, which causes horizontal scrolling.
75591 Fix this by resetting the width to 0 before issuing the "set width 30"
75594 Tested on x86_64-linux.
75596 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=24582
75598 2021-11-30 Nick Clifton <nickc@redhat.com>
75600 Use dwarf_vma type for offsets, ranges and section sizes in DWARF decoder.
75601 * dwarf.c (find_debug_info_for_offset): Use dwarf_vma type for
75602 offsets, sizes and ranges.
75603 (display_loc_list): Likewise. Also use print_dwarf_vma to print
75605 (display_loclists_list): Likewise.
75606 (display_loc_list_dwo): Likewise.
75607 (display_debug_str): Likewise.
75608 (display_debug_aranges): Likewise.
75609 (display_debug_ranges_list): Likewise.
75610 (display_debug_rnglists_list): Likewise.
75611 (display_debug_ranges): Likewise.
75613 ld: pru: Add pru_irq_map output section
75614 * scripttempl/pru.sc (.pru_irq_map): Define output section.
75615 * testsuite/ld-pru/pru_irq_map-1.d: New test.
75616 * testsuite/ld-pru/pru_irq_map-2.d: New test.
75617 * testsuite/ld-pru/pru_irq_map.s: New test.
75619 2021-11-30 Andrew Burgess <aburgess@redhat.com>
75621 gdb/testsuite: check the python module is available before using it
75622 The gdb.python/py-inferior-leak.exp test makes use of the tracemalloc
75623 module. When running the Python tests with a GDB built against Python
75624 2 I ran into a test failure due to the tracemalloc module not being
75627 This commit adds a new helper function to lib/gdb-python.exp that
75628 checks if a named module is available. Using this we can then skip
75629 the py-inferior-leak.exp test when the tracemalloc module is not
75632 2021-11-30 Andrew Burgess <aburgess@redhat.com>
75634 gdb: fix disassembler regressions for 32-bit arm
75637 commit 76b43c9b5c2b275cbf4f927bfc25984410cb5dd5
75638 Date: Tue Oct 5 15:10:12 2021 +0100
75640 gdb: improve error reporting from the disassembler
75642 We started seeing FAILs in the gdb.base/all-architectures*.exp tests,
75643 when running on a 32-bit ARM target, though I suspect running on any
75644 target that compiles such that bfd_vma is 32-bits would also trigger
75647 The problem is that the test is expected GDB's disassembler to print
75648 an error like this:
75650 Cannot access memory at address 0x0
75652 However, after the above commit we see an error like:
75654 unknown disassembler error (error = -1)
75656 The reason for this is this code in opcodes/i386-dis.c (in the
75657 print_insn function):
75659 if (address_mode == mode_64bit && sizeof (bfd_vma) < 8)
75661 (*info->fprintf_func) (info->stream,
75662 _("64-bit address is disabled"));
75666 This code effectively disallows us from ever disassembling 64-bit x86
75667 code if we compiled GDB with a 32-bit bfd_vma. Notice we return
75668 -1 (indicating a failure to disassemble), but never call the
75669 memory_error_func callback.
75671 Prior to the above commit GDB, when it received the -1 return value
75672 would assume that a memory error had occurred and just print whatever
75673 value happened to be in the memory error address variable, the default
75674 value of 0 just happened to be fine because the test had asked GDB to
75675 do this 'disassemble 0x0,+4'.
75677 If we instead change the test to do 'disassemble 0x100,+4' then GDB
75678 would (previously) have still reported:
75680 Cannot access memory at address 0x0
75682 which makes far less sense.
75684 In this commit I propose to fix this issue by changing the test to
75685 accept either the "Cannot access memory ..." string, or the newer
75686 "unknown disassembler error ..." string. With this change done the
75689 However, there is one weakness with this strategy; if GDB broke such
75690 that we _always_ reported "unknown disassembler error ..." we would
75691 never notice. This clearly would be bad. To avoid this issue I have
75692 adjusted the all-architectures*.exp tests so that, when we disassemble
75693 for the default architecture (the one selected by "auto") we _only_
75694 expect to get the "Cannot access memory ..." error string.
75696 [ Note: In an ideal world we should be able to disassemble any
75697 architecture at all times. There's no reason why the 64-bit x86
75698 disassembler requires a 64-bit bfd_vma, other than the code happens
75699 to be written that way. We could rewrite the disassemble to not
75700 have this requirement, but, I don't plan to do that any time soon. ]
75702 Further, I have changed the all-architectures*.exp test so that we now
75703 disassemble at address 0x100, this should avoid us being able to pass
75704 by printing a default address of 0x0. I did originally change the
75705 address we disassembled at to 0x4, however, some architectures,
75706 e.g. ia64, have a default instruction alignment that is greater than
75707 4, so would still round down to 0x0. I could have just picked 0x8 as
75708 an address, but I figured that 0x100 was likely to satisfy most
75709 architectures alignment requirements.
75711 2021-11-30 Andrew Burgess <aburgess@redhat.com>
75713 gdb/python: add gdb.RemoteTargetConnection.send_packet
75714 This commits adds a new sub-class of gdb.TargetConnection,
75715 gdb.RemoteTargetConnection. This sub-class is created for all
75716 'remote' and 'extended-remote' targets.
75718 This new sub-class has one additional method over its base class,
75719 'send_packet'. This new method is equivalent to the 'maint
75720 packet' CLI command, it allows a custom packet to be sent to a remote
75723 The outgoing packet can either be a bytes object, or a Unicode string,
75724 so long as the Unicode string contains only ASCII characters.
75726 The result of calling RemoteTargetConnection.send_packet is a bytes
75727 object containing the reply that came from the remote.
75729 2021-11-30 Andrew Burgess <aburgess@redhat.com>
75731 gdb: make packet_command function available outside remote.c
75732 In a later commit I will add a Python API to access the 'maint packet'
75733 functionality, that is, sending a user specified packet to the target.
75735 To make implementing this easier, this commit refactors how this
75736 command is currently implemented so that the packet_command function
75739 The new global send_remote_packet function takes an object that is an
75740 implementation of an abstract interface. Two functions within this
75741 interface are then called, one just before a packet is sent to the
75742 remote target, and one when the reply has been received from the
75743 remote target. Using an interface object in this way allows (1) for
75744 the error checking to be done before the first callback is made, this
75745 means we only print out what packet it being sent once we know we are
75746 going to actually send it, and (2) we don't need to make a copy of the
75747 reply if all we want to do is print it.
75749 One user visible changes after this commit are the error
75750 messages, which I've changed to be less 'maint packet' command
75751 focused, this will make them (I hope) better for when
75752 send_remote_packet can be called from Python code.
75754 So: "command can only be used with remote target"
75755 Becomes: "packets can only be sent to a remote target"
75757 And: "remote-packet command requires packet text as argument"
75758 Becomes: "a remote packet must not be empty"
75760 Additionally, in this commit, I've added support for packet replies
75761 that contain binary data. Before this commit, the code that printed
75762 the reply treated the reply as a C string, it assumed that the string
75763 only contained printable characters, and had a null character only at
75766 One way to show the problem with this is if we try to read the auxv
75767 data from a remote target, the auxv data is binary, so, before this
75770 (gdb) target remote :54321
75772 (gdb) maint packet qXfer:auxv:read::0,1000
75773 sending: "qXfer:auxv:read::0,1000"
75777 And after this commit:
75779 (gdb) target remote :54321
75781 (gdb) maint packet qXfer:auxv:read::0,1000
75782 sending: "qXfer:auxv:read::0,1000"
75783 received: "l!\x00\x00\x00\x00\x00\x00\x00\x00\xf0\xfc\xf7\xff\x7f\x00\x00\x10\x00\x00\x00\x00\x00\x00\x00\xff\xf>
75786 The binary contents of the reply are now printed as escaped hex.
75788 2021-11-30 Andrew Burgess <aburgess@redhat.com>
75790 gdb/python: introduce gdb.TargetConnection object type
75791 This commit adds a new object type gdb.TargetConnection. This new
75792 type represents a connection within GDB (a connection as displayed by
75793 'info connections').
75795 There's three ways to find a gdb.TargetConnection, there's a new
75796 'gdb.connections()' function, which returns a list of all currently
75797 active connections.
75799 Or you can read the new 'connection' property on the gdb.Inferior
75800 object type, this contains the connection for that inferior (or None
75801 if the inferior has no connection, for example, it is exited).
75803 Finally, there's a new gdb.events.connection_removed event registry,
75804 this emits a new gdb.ConnectionEvent whenever a connection is removed
75805 from GDB (this can happen when all inferiors using a connection exit,
75806 though this is not always the case, depending on the connection type).
75807 The gdb.ConnectionEvent has a 'connection' property, which is the
75808 gdb.TargetConnection being removed from GDB.
75810 The gdb.TargetConnection has an 'is_valid()' method. A connection
75811 object becomes invalid when the underlying connection is removed from
75812 GDB (as discussed above, this might be when all inferiors using a
75813 connection exit, or it might be when the user explicitly replaces a
75814 connection in GDB by issuing another 'target' command).
75816 The gdb.TargetConnection has the following read-only properties:
75818 'num': The number for this connection,
75820 'type': e.g. 'native', 'remote', 'sim', etc
75822 'description': The longer description as seen in the 'info
75823 connections' command output.
75825 'details': A string or None. Extra details for the connection, for
75826 example, a remote connection's details might be
75829 2021-11-30 Nelson Chu <nelson.chu@sifive.com>
75831 RISC-V: The vtype immediate with more than the defined 8 bits are preserved.
75832 According the rvv spec,
75833 https://github.com/riscv/riscv-v-spec/blob/master/vtype-format.adoc
75835 The bits of vtype immediate from 8 to (xlen - 1) should be reserved.
75836 Therefore, we should also dump the vtype immediate as numbers, when
75837 they are set over 8-bits. I think this is a bug that we used to support
75838 vediv extension and use the bit 8 and 9 of vtype, but forgot to update
75839 the behavior when removing the vediv.
75841 Consider the testcases,
75843 vsetvli a0, a1, 0x700 # the reserved bit 10, 9 and 8 are used.
75844 vsetvli a0, a1, 0x400 # the reserved bit 10 is used.
75845 vsetvli a0, a1, 0x300 # the reserved bit 9 and 8 are used.
75846 vsetvli a0, a1, 0x100 # the reserved bit 8 is used.
75847 vsetivli a0, 0xb, 0x300 # the reserved bit 9 and 8 are used.
75848 vsetivli a0, 0xb, 0x100 # the reserved bit 8 is used.
75850 The original objdump shows the following result,
75852 0000000000000000 <.text>:
75853 0: 7005f557 vsetvli a0,a1,1792
75854 4: 4005f557 vsetvli a0,a1,1024
75855 8: 3005f557 vsetvli a0,a1,e8,m1,tu,mu
75856 c: 1005f557 vsetvli a0,a1,e8,m1,tu,mu
75857 10: f005f557 vsetivli a0,11,e8,m1,tu,mu
75858 14: d005f557 vsetivli a0,11,e8,m1,tu,mu
75860 But in fact the correct result should be,
75862 0000000000000000 <.text>:
75863 0: 7005f557 vsetvli a0,a1,1792
75864 4: 4005f557 vsetvli a0,a1,1024
75865 8: 3005f557 vsetvli a0,a1,768
75866 c: 1005f557 vsetvli a0,a1,256
75867 10: f005f557 vsetivli a0,11,768
75868 14: d005f557 vsetivli a0,11,256
75871 * testsuite/gas/riscv/vector-insns.d: Added testcases to
75872 test the reserved bit 8 to (xlen-1) of vtype.
75873 * testsuite/gas/riscv/vector-insns.s: Likewise.
75875 * opcode/riscv.h: Removed OP_MASK_VTYPE_RES and OP_SH_VTYPE_RES,
75876 since they are different for operand Vc and Vb.
75878 * riscv-dis.c (print_insn_args): Updated imm_vtype_res to
75879 extract the reserved immediate of vtype correctly.
75881 2021-11-30 Nelson Chu <nelson.chu@sifive.com>
75883 RISC-V: Dump vset[i]vli immediate as numbers once vsew or vlmul is reserved.
75884 Consider the following case,
75886 vsetvli a0, a1, 0x4 # unrecognized vlmul
75887 vsetvli a0, a1, 0x20 # unrecognized vsew
75888 vsetivli a0, 0xb, 0x4 # unrecognized vlmul
75889 vsetivli a0, 0xb, 0x20 # unrecognized vsew
75891 For the current dis-assembler, we get the result,
75893 0000000000000000 <.text>:
75894 0: 0045f557 vsetvli a0,a1,e8,(null),tu,mu
75895 4: 0205f557 vsetvli a0,a1,e128,m1,tu,mu
75896 8: c045f557 vsetivli a0,11,e8,(null),tu,mu
75897 c: c205f557 vsetivli a0,11,e128,m1,tu,mu
75899 The vsew e128 and vlmul (null) are preserved according to the spec,
75900 so dump these fields looks wrong. Consider that we are used to dump
75901 the unrecognized csr as csr numbers directly, we should also dump
75902 the whole vset[i]vli immediates as numbers, once the vsew or vlmul
75903 is reserved. Therefore, following is what I expected,
75905 0000000000000000 <.text>:
75906 0: 0045f557 vsetvli a0,a1,4
75907 4: 0205f557 vsetvli a0,a1,32
75908 8: c045f557 vsetivli a0,11,4
75909 c: c205f557 vsetivli a0,11,32
75912 * testsuite/gas/riscv/vector-insns.d: Rewrite the vset[i]vli
75913 testcases since we should dump the immediate as numbers once
75914 the vsew or vlmul is reserved.
75915 * testsuite/gas/riscv/vector-insns.s: Likewise.
75917 * riscv-dis.c (print_insn_args): The reserved vsew and vlmul
75918 are NULL string in the riscv_vsew and riscv_vlmul, so dump the
75919 whole imm as numbers once one of them is NULL.
75920 * riscv-opc.c (riscv_vsew): Set the reserved vsew to NULL.
75921 (riscv_vlmul): Set the reserved vlmul to NULL.
75923 2021-11-30 Mike Frysinger <vapier@gentoo.org>
75925 zlib: enable silent build rules
75927 ld: enable silent build rules
75928 Also add $(AM_V_xxx) to various manual rules in here.
75930 libctf: enable silent build rules
75931 Also add $(AM_V_xxx) to various manual rules in here.
75933 gprof: enable silent build rules
75934 Also add $(AM_V_xxx) to various manual rules in here.
75936 binutils: merge doc subdir up a level
75937 This avoids a recursive make into the doc subdir and speeds up the
75938 build slightly. It also allows for more parallelism.
75940 binutils: enable silent build rules
75941 Also add $(AM_V_xxx) to various manual rules in here.
75943 bfd: enable silent build rules
75944 Also add $(AM_V_xxx) to various manual rules in here.
75946 opcodes: enable silent build rules
75947 Also add $(AM_V_xxx) to various manual rules in here.
75949 2021-11-30 GDB Administrator <gdbadmin@sourceware.org>
75951 Automatic date update in version.in
75953 2021-11-29 Tom Tromey <tom@tromey.com>
75955 Allow DW_ATE_UTF for Rust characters
75956 The Rust compiler plans to change the encoding of a Rust 'char' type
75957 to use DW_ATE_UTF. You can see the discussion here:
75959 https://github.com/rust-lang/rust/pull/89887
75961 However, this fails in gdb. I looked into this, and it turns out that
75962 the handling of DW_ATE_UTF is currently fairly specific to C++. In
75963 particular, the code here assumes the C++ type names, and it creates
75966 This comes from commit 53e710acd ("GDB thinks char16_t and char32_t
75967 are signed in C++"). The message says:
75969 Both places need fixing. But since I couldn't tell why dwarf2read.c
75970 needs to create a new type, I've made it use the per-arch built-in
75971 types instead, so that the types are only created once per arch
75972 instead of once per objfile. That seems to work fine.
75974 ... which is fine, but it seems to me that it's also correct to make a
75975 new character type; and this approach is better because it preserves
75976 the type name as well. This does use more memory, but first we
75977 shouldn't be too concerned about the memory use of types coming from
75978 debuginfo; and second, if we are, we should implement type interning
75981 Changing this code to use a character type revealed a couple of
75982 oddities in the C/C++ handling of TYPE_CODE_CHAR. This patch fixes
75985 I filed PR rust/28637 for this issue, so that this patch can be
75986 backported to the gdb 11 branch.
75988 2021-11-29 Aaron Merey <amerey@redhat.com>
75989 Tom de Vries <tdevries@suse.de>
75991 [PR gdb/27026] CTRL-C is ignored when debug info is downloaded
75992 During debuginfod downloads, ctrl-c should result in the download
75993 being cancelled and skipped. However in some cases, ctrl-c fails to
75994 get delivered to gdb during downloading. This can result in downloads
75997 Fix this by ensuring that target_terminal::ours is in effect for the
75998 duration of each download.
76000 https://sourceware.org/bugzilla/show_bug.cgi?id=27026#c3
76002 2021-11-29 Nick Clifton <nickc@redhat.com>
76004 strings: Replace references to -u option with references to -U.
76007 2021-11-29 Tom de Vries <tdevries@suse.de>
76009 [gdb/symtab] Fix segfault in search_one_symtab
76010 PR28539 describes a segfault in lambda function search_one_symtab due to
76011 psymbol_functions::expand_symtabs_matching calling expansion_notify with a
76014 struct compunit_symtab *symtab =
76015 psymtab_to_symtab (objfile, ps);
76017 if (expansion_notify != NULL)
76018 if (!expansion_notify (symtab))
76022 This happens as follows. The partial symtab ps is a dwarf2_include_psymtab
76023 for some header file:
76025 (gdb) p ps.filename
76026 $5 = 0x64fcf80 "/usr/include/c++/11/bits/stl_construct.h"
76029 The includer of ps is a shared symtab for a partial unit, with as user:
76031 (gdb) p ps.includer().user.filename
76033 "/usr/src/debug/llvm13-13.0.0-1.2.x86_64/tools/clang/lib/AST/Decl.cpp"
76036 The call to psymtab_to_symtab expands the Decl.cpp symtab (and consequently
76037 the shared symtab), but returns nullptr because:
76039 struct dwarf2_include_psymtab : public partial_symtab
76042 compunit_symtab *get_compunit_symtab (struct objfile *objfile) const override
76048 Fix this by returning the Decl.cpp symtab instead, which fixes the segfault
76051 Tested on x86_64-linux.
76053 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28539
76055 2021-11-29 Nick Clifton <nickc@redhat.com>
76057 Update description of string's -n option.
76059 * strings.c (usage): Update desciption of -n option.
76060 * doc/binutils.texi: Likewise.
76062 2021-11-29 Tom de Vries <tdevries@suse.de>
76064 [gdb/testsuite] Fix typo in proc lines
76065 Proc lines contains a typo:
76067 string_form { set $_line_string_form $value }
76070 Remove the incorrect '$' in '$_line_string_form'.
76072 Tested on x86_64-linux.
76074 2021-11-29 Tom de Vries <tdevries@suse.de>
76076 [gdb/testsuite] Use unique files in gdb.dwarf2/dw2-lines.exp
76077 While debugging a problem in gdb.dwarf2/dw2-lines.exp, I realized that the
76078 test-case generates all executables and associated temporary files using the
76081 Fix this by adding a new proc prefix_id in lib/gdb.exp, and using it in the
76084 Tested on x86_64-linux.
76086 2021-11-29 Tom de Vries <tdevries@suse.de>
76088 [gdb/testsuite] Fix gdb.dwarf2/dw2-lines.exp with -m32
76089 When running test-case gdb.dwarf2/dw2-lines.exp with target board -unix/-m32,
76090 we run into another instance of PR28383, where the dwarf assembler generates
76091 64-bit relocations which are not supported by the 32-bit assembler:
76093 dw2-lines-dw.S: Assembler messages:^M
76094 outputs/gdb.dwarf2/dw2-lines/dw2-lines-dw.S:76: Error: \
76095 cannot represent relocation type BFD_RELOC_64^M
76098 Fix this by using _op_offset in _line_finalize_header.
76100 Tested on x86_64-linux.
76102 2021-11-29 Mike Frysinger <vapier@gentoo.org>
76104 sim: testsuite: drop most specific istarget checks
76105 We'll rely on the toolchain probing to determine whether each arch's
76106 tests can be run rather the current configure target. This allows
76107 testing all of the ports in a multitarget configuration.
76109 For now, we don't reformat the files entirely to make it easier to
76110 review, and in case we need to make adjustments. Once this feels
76111 like it's stable, we can flatten the code a bit by removing the if
76112 statement entirely.
76114 2021-11-29 Mike Frysinger <vapier@gentoo.org>
76116 sim: testsuite: support parallel execution
76117 Break up the dejagnu logic so that we can parallelize the testsuite.
76118 This takes a page from gcc & gdb where each .exp is run in isolation
76119 instead of in serial.
76121 For most targets, this doesn't make much of a difference as they only
76122 have a single .exp. A few (like cris & frv) have multiple .exp though
76123 and will see a bit of a speed up.
76125 The real gain is when testing a multitarget build. This way we can
76126 run all the targets in parallel and cut the execution time a bit.
76127 On my system, it goes from ~155sec to ~100sec.
76129 We can gain further speedups by splitting up some of the larger .exp
76130 files into smaller groups. We'll do that in a followup though.
76132 2021-11-29 Mike Frysinger <vapier@gentoo.org>
76134 sim: testsuite: expand arch specific toolchain settings
76135 Leverage the new per-port toolchain settings to initialize the env
76136 for eeach set of tests. This allows us to run all the tests in a
76137 multitarget build if the user sets up the vars. If they don't, we
76138 can still skip all the tests.
76140 2021-11-29 Mike Frysinger <vapier@gentoo.org>
76142 sim: testsuite: setup per-port toolchain settings for multitarget build
76143 Gas does not support multitarget builds -- it still only supports
76144 a single input & output format. ld is a bit better, but requires
76145 manual flags to select the right output. This makes it impossible
76146 to run the complete testsuite in a multitarget build.
76148 To address this limitation, create a suite of FOR_TARGET variables
76149 so these can be set to precompiled as & ld programs. It requires
76150 a bit of setup ahead of time, but it's a one-time cost, and makes
76151 running the full testsuite at once much easier.
76153 2021-11-29 GDB Administrator <gdbadmin@sourceware.org>
76155 Automatic date update in version.in
76157 2021-11-28 Alan Modra <amodra@gmail.com>
76159 PR28629 NIOS2 fallout
76160 The test exactly matched wrong output.
76163 * testsuite/gas/nios2/relax.d: Update expected output.
76165 2021-11-28 Mike Frysinger <vapier@gentoo.org>
76167 sim: add checks to core headers to prevent incorrect common building
76168 Some of the core sim headers rely on the SIM_AC_OPTION_BITSIZE macro
76169 which can change the size of core types. Since these haven't been
76170 unified across ports, add checks to make sure they aren't accidentally
76171 included when building for all ports. This caught the sim-load file
76172 using poisoned headers that it didn't actually need.
76174 sim: unify syscall.o building
76175 Now that we've unified all the syscall tables, this file does not rely
76176 on any port-specific settings, so move it up to building as part of the
76177 common step so we only do it once in a multibuild.
76179 sim: drop unused gentmap & nltvals.def logic
76180 Now that all ports have switched to target-newlib-* files, there's
76181 no need for these files & generating things at build time. So punt
76182 the logic and make target-newlib-syscall a hard requirement.
76184 sim: mcore: switch to new target-newlib-syscall
76185 Use the new target-newlib-syscall module. This is needed to merge all
76186 the architectures into a single build, and mcore has a custom syscall
76187 table for its newlib/libgloss port.
76189 sim: riscv: switch to new target-newlib-syscall
76190 Use the new target-newlib-syscall module. This is needed to merge all
76191 the architectures into a single build, and riscv has a custom syscall
76192 table for its newlib/libgloss port.
76194 2021-11-28 Mike Frysinger <vapier@gentoo.org>
76196 sim: cr16: switch to new target-newlib-syscall
76197 Use the new target-newlib-syscall module. This is needed to merge all
76198 the architectures into a single build, and cr16 has a custom syscall
76199 table for its newlib/libgloss port.
76201 This allows cleaning up the syscall ifdef logic. We know these will
76204 2021-11-28 Mike Frysinger <vapier@gentoo.org>
76206 sim: d10v: switch to new target-newlib-syscall
76207 Use the new target-newlib-syscall module. This is needed to merge all
76208 the architectures into a single build, and d10v has a custom syscall
76209 table for its newlib/libgloss port.
76211 This allows cleaning up the syscall ifdef logic. We know these will
76214 2021-11-28 Mike Frysinger <vapier@gentoo.org>
76216 sim: sh: switch to new target-newlib-syscall
76217 Use the new target-newlib-syscall module. This is needed to merge all
76218 the architectures into a single build, and sh has a custom syscall
76219 table for its newlib/libgloss port.
76221 2021-11-28 Mike Frysinger <vapier@gentoo.org>
76223 sim: v850: switch to new target-newlib-syscall
76224 Use the new target-newlib-syscall module. This is needed to merge all
76225 the architectures into a single build, and v850 has a custom syscall
76226 table for its newlib/libgloss port.
76228 This allows cleaning up the syscall ifdef logic. We know these will
76231 2021-11-28 Mike Frysinger <vapier@gentoo.org>
76233 sim: iq2000/lm32/m32c/moxie/rx: switch to new target-newlib-syscall.h
76234 Use the new target-newlib-syscall.h to provide the target syscall
76235 defines. These code paths are written specifically for the newlib
76236 ABI rather than being generalized, so switching them to the defines
76237 rather than trying to go through the dynamic callback conversion
76238 seems like the best trade-off for now. Might have to reconsider
76239 this in the future.
76241 2021-11-28 Mike Frysinger <vapier@gentoo.org>
76243 sim: nltvals: pull target syscalls out into a dedicated source file
76244 Like we just did for pulling out the errno map, pull out the syscall
76245 maps into a dedicated common file. Most newlib ports are using the
76246 same syscall map, but not all, which means we have to do a bit more
76249 This commit adds the maps and switches the ports using the common
76250 default syscall table over to it. Ports using unique syscall tables
76251 are still using the old targ-map.c logic.
76253 Switching common ports over is easy by checking NL_TARGET, but the
76254 ppc code needs a bit more cleanup here hence its larger diff.
76256 2021-11-28 Mike Frysinger <vapier@gentoo.org>
76258 sim: frv: resolve syscalls dynamically
76259 Avoid use of TARGET_<syscall> defines and rely on the callback layers
76260 to resolve these dynamically so we can support multiple syscall layers
76261 instead of assuming the newlib/libgloss numbers all the time.
76263 sim: mn10300: resolve syscalls dynamically
76264 Avoid use of TARGET_<syscall> defines and rely on the callback layers
76265 to resolve these dynamically so we can support multiple syscall layers
76266 instead of assuming the newlib/libgloss numbers all the time.
76268 sim: nltvals: drop i960
76269 This port was dropped from gdb/bfd/sim years ago, so stop including
76270 its syscall constants too.
76272 sim: moxie: fix datadir handling
76273 Expand the value at `make` time rather than configure generation time
76274 so that we handle $(datarootdir) setting properly.
76276 2021-11-28 GDB Administrator <gdbadmin@sourceware.org>
76278 Automatic date update in version.in
76280 2021-11-27 Simon Marchi <simon.marchi@polymtl.ca>
76282 gdb: fix typos in configure
76283 The variable names used to restore CFLAGS and LDFLAGS here don't quite
76284 match the names used above, resulting in losing the original CFLAGS and
76287 Change-Id: I9cc2c3b48b1dc30c31a7143563c893fd6f426a0a
76289 2021-11-27 Mike Frysinger <vapier@gentoo.org>
76291 sim: hw: mark hw_descriptors const
76293 sim: testsuite: add dedicated flag for init toolchain tests
76294 As we setup more reliable CC_FOR_TARGET variables for each target, the
76295 bfin way of overriding it to stuff custom CFLAGS doesn't scale well.
76296 Add a dedicated CFLAGS_FOR_TARGET_init setting that each set of tests
76297 can setup if they want to add custom options.
76299 sim: testsuite: clean up arch specific toolchain settings
76300 In a multitarget build, we process all targets in order, so make sure
76301 the toolchain settings from one don't leak into the next.
76303 2021-11-27 Mike Frysinger <vapier@gentoo.org>
76305 sim: cris: always search for local rvdummy tool
76306 If the board info sets the sim to a basename that is found via $PATH
76307 (which is the default dejagnu behavior), the logic here to use its
76308 dirname to find rvdummy fails because it looks for `./rvdummy`. So
76309 switch it to always use the local build of rvdummy which is the one
76310 we want to be testing against in the first place.
76312 If we get a request for testing against a different setup, we can
76313 figure out & document the needs at that point, and then setup some
76314 config knobs to control it.
76316 2021-11-27 Tom de Vries <tdevries@suse.de>
76318 [gdb/testsuite] Fix FAIL in gdb.base/list-missing-source.exp
76319 In commit f8080fb7a44 "[gdb/testsuite] Add gdb.base/include-main.exp" a
76320 file gdb.base/main.c was added, which caused the following regression:
76324 (gdb) FAIL: gdb.base/list-missing-source.exp: list
76327 The problem is that the test-case does not expect to find a file main.c, but
76328 now it finds gdb.base/main.c.
76330 Fix this by using the more specific file name list-missing-source.c.
76332 Tested on x86_64-linux.
76334 2021-11-27 Mike Frysinger <vapier@gentoo.org>
76336 sim: testsuite: fix bits-gen EXEEXT handling
76337 Add missing $(EXEEXT) to dependencies on bits-gen. These are actually
76338 build-only tools, but automake doesn't allow for build & host tools, so
76339 the rules are re-using EXEEXT.
76341 2021-11-27 Mike Frysinger <vapier@gentoo.org>
76343 sim: testsuite: initial support for OS-specific tests
76344 We usually test against the newlib/libgloss environment, but for a
76345 few ports that also support Linux apps, we want to test that logic
76346 too. A lot of the C code is written such that it works with either
76347 newlib/libgloss or glibc/linux toolchains, but we have some tests
76348 that end up being Linux-specific. Cris has been using the target
76349 tuple as a rough proxy for this (where cris*-*-elf is assumed to be
76350 newlib/libgloss, and everything else is glibc/linux), but that is a
76351 bit too rough, and it doesn't work in a multitarget build.
76353 So lets create a few stub files that we can do compile tests with
76354 to detect the different setups, and then let tests declare which
76355 one they require (if they require any at all).
76357 2021-11-27 Mike Frysinger <vapier@gentoo.org>
76359 sim: testsuite: unify basic C compiler checks
76360 Both bfin & cris ports test the C compiler to see if it works, but in
76361 their own way. Unify the checks in the common code so we can leverage
76362 them in more ports in the future, and collapse the bfin & cris code.
76364 2021-11-27 Mike Frysinger <vapier@gentoo.org>
76366 sim: testsuite: rework sim_init usage
76367 The sim_init function was called by runtest for each test when --tool
76368 was set to sim. When we changed to --tool '' to collapse the testsuite
76369 dir, the init function was no longer called on every test. However, it
76370 was still being called explicitly by config/default.exp. It's not clear
76371 why that explicit call ever existed since, in the past, it meant it was
76374 Lets drop the single sim_init call in config/default.exp and move it out
76375 to all our tests. This replicates the runtest behavior so we can setup
76376 variables on a per-test basis which allows us to recollapse the sim_path
76377 logic back. We'll also leverage this in the future for toolchain setup.
76379 Also add a few comments clarifying the overall runtime behavior.
76381 2021-11-27 Mike Frysinger <vapier@gentoo.org>
76383 sim: cris: fix testsuite hang when sim is missing
76384 If the cris sim hasn't been built yet, trying to run its testsuite
76385 will hang indefinitely. The common sim APIs already have this, so
76386 copy it over to the cris forks of the test+run functions.
76388 2021-11-27 Mike Frysinger <vapier@gentoo.org>
76390 sim: testsuite: fix objdir handling
76391 The tests assume that the cwd is the objdir directory and write its
76392 intermediates to there all the time. When using runtest's --objdir
76393 setting though, this puts the files in the wrong place. This isn't
76394 a big problem currently as we never change --objdir, but in order to
76395 support parallel test execution, we're going to start setting that
76396 option, so clean up the code ahead of time.
76398 We also have to tweak some of the cris tests which were making
76399 assumptions about the argv[0] value.
76401 2021-11-27 Mike Frysinger <vapier@gentoo.org>
76403 sim: testsuite: rename global_sim_options to SIMFLAGS_FOR_TARGET
76404 Now that all the other toolchain settings have been renamed to match
76405 the dejagnu settings of XXX_FOR_TARGET, rename global_sim_options to
76406 SIMFLAGS_FOR_TARGET too.
76408 sim: testsuite: replace global_ld_options with LDFLAGS_FOR_TARGET
76409 Only a few tests actually use global_ld_options, but we can replace the
76410 sim-specific settings with the dejagnu common LDFLAGS_FOR_TARGET and get
76413 2021-11-27 GDB Administrator <gdbadmin@sourceware.org>
76415 Automatic date update in version.in
76417 2021-11-26 John David Anglin <danglin@gcc.gnu.org>
76419 Fix ifunc test fails on hppa*-*-*
76420 2021-11-26 John David Anglin <danglin@gcc.gnu.org>
76426 * ld/testsuite/ld-ifunc/ifunc.exp (contains_irelative_reloc): Adjust
76428 Skip static ifunc-using executable test on hppa*-*-*.
76430 2021-11-26 H.J. Lu <hjl.tools@gmail.com>
76432 gas: Update commit 4780e5e4933
76435 commit 4780e5e4933a2497a5aecc4ceabbbb8e82aaf822
76436 Author: Tom de Vries <tdevries@suse.de>
76437 Date: Fri Nov 26 09:59:45 2021 +0100
76439 [gas] Fix file 0 dir with -gdwarf-5
76441 1. Replace i with j in
76443 for (j = 0; i < NUM_MD5_BYTES; ++j)
76445 2. Pass -W to readelf to force CU: in output due to:
76447 if (do_wide || strlen (directory) < 76)
76448 printf (_("CU: %s/%s:\n"), directory, file_table[0].name);
76450 printf ("%s:\n", file_table[0].name);
76453 * dwarf2dbg.c (out_dir_and_file_list): Fix a typo in commit
76455 * testsuite/gas/elf/dwarf-5-nop-for-line-table.d: Pass -W to
76458 2021-11-26 Mike Frysinger <vapier@gentoo.org>
76460 sim: testsuite: replace global_as_options with ASFLAGS_FOR_TARGET
76461 Only a few tests actually use global_as_options, but we can replace the
76462 sim-specific settings with the dejagnu common ASFLAGS_FOR_TARGET and get
76465 2021-11-26 Tom de Vries <tdevries@suse.de>
76467 [gdb/testsuite] Add gdb.base/include-main.exp
76468 The test-case gdb.ada/dgopt.exp uses the -gnatD switch, in combination with
76471 This causes the source file $src/gdb/testsuite/gdb.ada/dgopt/x.adb to be
76472 expanded into $build/gdb/testsuite/outputs/gdb.ada/dgopt/x.adb.dg, and the
76473 debug information should refer to the x.adb.dg file.
76475 That is the case for the .debug_line part:
76477 The Directory Table is empty.
76479 The File Name Table (offset 0x1c):
76480 Entry Dir Time Size Name
76483 but not for the .debug_info part:
76485 <11> DW_AT_name : $src/gdb/testsuite/gdb.ada/dgopt/x.adb
76486 <15> DW_AT_comp_dir : $build/gdb/testsuite/outputs/gdb.ada/dgopt
76489 Filed as PR gcc/103436.
76491 In C we can generate similar debug information, using a source file that does
76492 not contain any code, but includes another one that does:
76494 $ cat gdb/testsuite/gdb.base/include-main.c
76497 such that in the .debug_line part we have:
76499 The Directory Table (offset 0x1c):
76500 1 /home/vries/gdb_versions/devel/src/gdb/testsuite/gdb.base
76502 The File Name Table (offset 0x57):
76503 Entry Dir Time Size Name
76506 and in the .debug_info part:
76508 <11> DW_AT_name : $src/gdb/testsuite/gdb.base/include-main.c
76509 <15> DW_AT_comp_dir : $build/gdb/testsuite
76512 Add a C test-case that mimics gdb.ada/dgopt.exp, that is:
76513 - generate debug info as described above,
76514 - issue a list of a line in include-main.c, while the corresponding
76515 CU is not expanded yet.
76517 Tested on x86_64-linux.
76519 2021-11-26 Mike Frysinger <vapier@gentoo.org>
76521 sim: testsuite: drop unused global_cc_options
76522 Nothing in the testsuite is using this setting, so let's drop it.
76523 Any code that wants to set compiler flags can use CFLAGS_FOR_TARGET
76524 instead to get the same effect.
76526 sim: testsuite: punt unused toolchain variables
76527 These haven't been used in over 20 years. The sim testsuite used to
76528 run these tools itself directly, but back in ~1999 it switched to the
76529 dejagnu helpers (e.g. target_assemble & target_link), and the dejagnu
76530 logic only utilizes XXX_FOR_TARGET variables. Punt them here to avoid
76531 confusion with dead code.
76533 2021-11-26 Andrew Burgess <aburgess@redhat.com>
76534 Simon Cook <simon.cook@embecosm.com>
76536 gdb: add risc-v disassembler options support
76537 This commit adds support for RISC-V disassembler options to GDB. This
76538 commit is based on this patch which was never committed:
76540 https://sourceware.org/pipermail/binutils/2021-January/114944.html
76542 All of the binutils refactoring has been moved to a separate, earlier,
76543 commit, so this commit is pretty straight forward, just registering
76544 the required gdbarch hooks.
76546 2021-11-26 Andrew Burgess <aburgess@redhat.com>
76547 Simon Cook <simon.cook@embecosm.com>
76549 opcodes/riscv: add disassembler options support to libopcodes
76550 In preparation for the next commit, which will add GDB support for
76551 RISC-V disassembler options, this commit restructures how the
76552 disassembler options are managed within libopcodes.
76554 The implementation provided here is based on this mailing list patch
76555 which was never committed:
76557 https://sourceware.org/pipermail/binutils/2021-January/114944.html
76559 which in turn took inspiration from the MIPS implementation of the
76562 The biggest changes from the original mailing list post are:
76564 1. The GDB changes have been split into a separate patch, and
76566 2. The `riscv_option_args_privspec` variable, which held the valid
76567 priv-spec values is now gone, instead we use the `riscv_priv_specs`
76568 array from bfd/cpu-riscv.c instead.
76573 * dis-asm.h (disassembler_options_riscv): Declare.
76577 * riscv-dis.c (enum riscv_option_arg_t): New enum typedef.
76578 (riscv_options): New static global.
76579 (disassembler_options_riscv): New function.
76580 (print_riscv_disassembler_options): Rewrite to use
76581 disassembler_options_riscv.
76583 2021-11-26 Tom de Vries <tdevries@suse.de>
76585 [gas] Fix file 0 dir with -gdwarf-5
76586 In out_dir_and_file_list, if file 0 is copied from file 1, only the filename
76587 is copied, and the dir and md5 fields are left to their default values.
76589 Fix this by adding the copy of the dir and md5 fields.
76593 2021-11-26 Tom de Vries <tdevries@suse.de>
76596 * dwarf2dbg.c (out_dir_and_file_list): When copying file 1 to file 0,
76597 also copy dir and md5 fields.
76598 * testsuite/gas/i386/dwarf5-line-4.d: Adjust expected output.
76600 2021-11-26 Mike Frysinger <vapier@gentoo.org>
76602 sim: mips: avoid _ namespace
76603 Some C libraries export _P symbols in their headers (like older
76604 newlib and its ctype.h), so use P_ instead to avoid conflicts.
76606 ld: fix POSIX shell test usage
76607 POSIX test uses = for compares, not == which is a bashism.
76609 gas: enable silent build rules
76611 ld: fix --disable-multiple-abs-defs alignment in help
76613 2021-11-26 GDB Administrator <gdbadmin@sourceware.org>
76615 Automatic date update in version.in
76617 2021-11-25 Enze Li <lienze2010@hotmail.com>
76619 gdb: ensure extension_language_python is always defined
76622 commit c6a6aad52d9e839d6a84ac31cabe2b7e1a2a31a0
76623 Date: Mon Oct 25 17:25:45 2021 +0100
76625 gdb/python: make some global variables static
76627 building without Python was broken. The extension_language_python
76628 global was moved from being always defined, to only being defined when
76629 the HAVE_PYTHON macro was defined. As a consequence, building without
76630 Python support would result in errors like:
76632 /usr/bin/ld: extension.o:(.rodata+0x120): undefined reference to `extension_language_python'
76634 This commit fixes the problem by moving the definition of
76635 extension_language_python outside of the HAVE_PYTHON macro protection.
76637 2021-11-25 Andrew Burgess <aburgess@redhat.com>
76639 Revert "gdb: add assert in remote_target::wait relating to async being off"
76640 This commit introduced a test failure in gdb.server/attach-flag.exp.
76641 I didn't spot this failure originally as the problem is fixed by this,
76642 as yet unpushed patch:
76644 https://sourceware.org/pipermail/gdb-patches/2021-November/183768.html
76646 I unfortunately didn't test each patch in the original series
76647 independently. I'll repost this patch after the above patch has been
76650 This reverts commit 32b1f5e8d6b8ddd3be6e471c26dd85a1dac31dda.
76652 2021-11-25 Nick Clifton <nickc@redhat.com>
76654 Fix building the AArch64 assembler and disassembler when assertions are disabled.
76656 * aarch64-asm.c: Replace assert(0) with real code.
76657 * aarch64-dis.c: Likewise.
76658 * aarch64-opc.c: Likewise.
76660 2021-11-25 Bruno Larsen <blarsen@redhat.com>
76662 PR gdb/28480: Improve ambiguous member detection
76663 Basic ambiguity detection assumes that when 2 fields with the same name
76664 have the same byte offset, it must be an unambiguous request. This is not
76665 always correct. Consider the following code:
76671 [[no_unique_address]] empty e;
76679 class C: public A, public B { };
76681 if we tried to use c.e in code, the compiler would warn of an ambiguity,
76682 however, since A::e does not demand an unique address, it gets the same
76683 address (and thus byte offset) of the members, making A::e and B::e have the
76684 same address. however, "print c.e" would fail to report the ambiguity,
76685 and would instead print it as an empty class (first path found).
76687 The new code solves this by checking for other found_fields that have
76688 different m_struct_path.back() (final class that the member was found
76689 in), despite having the same byte offset.
76691 The testcase gdb.cp/ambiguous.exp was also changed to test for this
76694 2021-11-25 Jan W. Jagersma <jwjagersma@gmail.com>
76696 coff-go32: consistent 16-byte section alignment
76697 Section alignment for coff-go32 is inconsistent - The '.text' and
76698 '.data' sections are 16-byte aligned, but named sections '.text.*' and
76699 '.data.*' are only 4-byte aligned. '.gnu.linkonce.r.*' is aligned to
76700 16 bytes, yet '.rodata' and '.rodata.*' are aligned to 4 bytes. For
76701 '.bss' all input sections are only aligned to 4 bytes.
76703 This primarily can cause trouble when using SSE instructions, which
76704 require their memory operands to be aligned to 16-byte boundaries.
76706 This patch solves the issue simply by setting the section alignment
76707 to 16 bytes, for all code and data sections referenced in the default
76710 * coff-go32.c (COFF_SECTION_ALIGNMENT_ENTRIES): Use partial
76711 name match for .text, .data. Add entries for .const, .rodata,
76712 .bss, .gnu.linkonce.b.
76714 2021-11-25 Alan Modra <amodra@gmail.com>
76716 Re: AArch64: Add support for AArch64 EFI (efi-*-aarch64)
76717 Commit b69c9d41e8 edited bfd/Makefile.in rather than using automake,
76718 which meant a typo in Makefile.am was not discovered and other
76719 differences in Makefile.in are seen with a proper regeneration. One
76720 difference was lack of an empty line between the pe-aarch64igen.c rule
76721 and the following $(BFD32_LIBS) etc. dependency rule, in the
76722 regenerated file. Not that it matters for proper "make" behaviour,
76723 but it's nicer with a line between those rules. Moving the rule
76724 earlier seems to cure the missing empty line.
76726 * Makefile.am (BFD64_BACKENDS): Correct typo.
76727 (BFD_H_DEPS, LOCAL_H_DEPS): Move earlier. Move rule using these
76729 * Makefile.in: Regenerate.
76730 * po/BLD-POTFILES.in: Regenerate.
76731 * po/SRC-POTFILES.in: Regenerate.
76733 2021-11-25 Nick Clifton <nickc@redhat.com>
76735 Updated French translation for the opcodes directory.
76736 * po/fr.po; Updated French translation.
76738 2021-11-25 Andrew Burgess <aburgess@redhat.com>
76740 gdb: rename source_styling_changed observer
76741 In a later commit I plan to add disassembler styling. In the same way
76742 that we have a source_styling_changed observer I would need to add a
76743 disassembler_styling_changed observer.
76745 However, currently, these observers would only be notified from
76746 cli-style.c:set_style_enabled, and observed in tui-winsource.c,
76747 tui_source_window::style_changed, as a result, having two observers
76748 seems unnecessary right now, so, in this commit, I plan to rename
76749 source_styling_changed to just styling_changed, then, in the later
76750 commit, when disassembler styling is added, I can use the same
76751 observer for both source styling, and disassembler styling.
76753 There should be no user visible changes after this commit.
76755 2021-11-25 Andrew Burgess <andrew.burgess@embecosm.com>
76757 gdb/python: make some global variables static
76758 Make a couple of global variables static in python/python.c. To do
76759 this I had to move the definition of extension_language_python to
76762 There should be no user visible changes after this commit.
76764 2021-11-25 Andrew Burgess <aburgess@redhat.com>
76766 gdb: add assert in remote_target::wait relating to async being off
76767 While working on another patch I ended up in a situation where I had
76768 async mode disabled (with 'maint set target-async off'), but the async
76769 event token got marked anyway.
76771 In this situation GDB was continually calling into
76772 remote_target::wait, however, the async token would never become
76773 unmarked as the unmarking is guarded by target_is_async_p.
76775 We could just unconditionally unmark the token, but that would feel
76776 like just ignoring a bug, so, instead, lets assert that if
76777 !target_is_async_p, then the async token should not be marked.
76779 This assertion would have caught my earlier mistake.
76781 There should be no user visible changes with this commit.
76783 2021-11-25 Andrew Burgess <aburgess@redhat.com>
76785 gdb: simplify remote_target::is_async_p
76786 This commit simplifies remote_target::is_async_p by removing the
76787 target_async_permitted check.
76789 In previous commits I have added additional assertions around the
76790 target_async_permitted flag into target.c, as a result we should now
76791 be confident that if target_can_async_p returns false, a target will
76792 never have async mode enabled. Given this, it should not be necessary
76793 to check target_async_permitted in remote_target::is_async_p, if this
76794 flag is false ::is_async_p should return false anyway. There is an
76795 assert to this effect in target_is_async_p.
76797 There should be no user visible change after this commit.
76799 2021-11-25 Andrew Burgess <aburgess@redhat.com>
76801 gdb: add asserts in target.c for target_async_permitted
76802 The target_async_permitted flag allows a user to override whether a
76803 target can act in async mode or not. In previous commits I have moved
76804 the checking of this flag out of the various ::can_async_p methods and
76805 into the common target.c code.
76807 In this commit I will add some additional assertions into
76808 target_is_async_p and target_async. The rules these assertions are
76811 1. A target that returns false for target_can_async_p should never
76812 become "async enabled", and so ::is_async_p should always return
76813 false. This is being checked in target_is_async_p.
76815 2. GDB should never try to enable async mode for a target that
76816 returns false for target_can_async_p, this is checked in
76819 There are a few places where we call the ::is_async_p method directly,
76820 in these cases we will obviously not pass through the assert in
76821 target_is_async_p, however, there are also plenty of places where we
76822 do call target_is_async_p so if GDB starts to misbehave we should
76823 catch it quickly enough.
76825 There should be no user visible changes after this commit.
76827 2021-11-25 Andrew Burgess <aburgess@redhat.com>
76829 gdb: hoist target_async_permitted checks into target.c
76830 This commit moves the target_async_permitted check out of each targets
76831 ::can_async_p method and into the target_can_async_p wrapper function.
76833 I've left some asserts in the two ::can_async_p methods that I
76834 changed, which will hopefully catch any direct calls to these methods
76835 that might be added in the future.
76837 There should be no user visible changes after this commit.
76839 2021-11-25 Andrew Burgess <aburgess@redhat.com>
76841 gdb: introduce a new overload of target_can_async_p
76842 There are a few places where we call the target_ops::can_async_p
76843 member function directly, instead of using the target_can_async_p
76846 In some of these places this is because we need to ask before the
76847 target has been pushed, and in another location (in target.c) it seems
76848 unnecessary to go through the wrapper when we are already in target.c
76851 However, in the next commit I'd like to hoist some common checks out
76852 of target specific code into target.c. To achieve this, in this
76853 commit, I introduce a new overload of target_can_async_p which takes a
76854 target_ops pointer, and calls the ::can_async_p method directly. I
76855 then make use of the new overload where appropriate.
76857 There should be no user visible changes after this commit.
76859 2021-11-25 Clément Chigot <clement.chigot@atos.net>
76861 ld/testsuite/ld-elfvsb: correctly test "weak hidden symbol DSO last"
76862 The test must be done with the shared object and not with the object
76863 file which is already being tested above.
76866 * testsuite/ld-elfvsb/elfvsb.exp: use .so file in "weak hidden
76869 2021-11-25 Tom de Vries <tdevries@suse.de>
76871 [gdb/cli] Add "set logging enabled", deprecate "set logging on/off"
76872 Before commit 3b6acaee895 "Update more calls to add_prefix_cmd" we had the
76873 following output for "show logging file":
76875 $ gdb -q -batch -ex "set trace-commands on" \
76876 -ex "set logging off" \
76877 -ex "show logging file" \
76878 -ex "set logging on" \
76879 -ex "show logging file"
76882 Future logs will be written to gdb.txt.
76885 Currently logging to "gdb.txt".
76888 After that commit we have instead:
76892 The current logfile is "gdb.txt".
76895 The current logfile is "gdb.txt".
76898 Before the commit, whether logging is enabled or not can be deduced from the
76899 output of the command. After the commit, the message is unified and it's no
76900 longer clear whether logging is enabled or not.
76903 - adding a new command "show logging enabled"
76904 - adding a corresponding new command "set logging enabled on/off"
76905 - making the commands "set logging on/off" deprecated aliases of the
76906 "set logging enabled on/off" command.
76908 Update the docs and testsuite to use "set logging enabled". Mention the new
76909 and deprecated commands in NEWS.
76911 Tested on x86_64-linux.
76913 2021-11-25 Tom de Vries <tdevries@suse.de>
76915 [gdb/cli] Fix typo in logging overwrite help text
76918 $ gdb -q -batch -ex "help set logging overwrite"
76919 Set whether logging overwrites or appends to the log file.
76920 If set, logging overrides the log file.
76923 Fix overrides -> overwrites typo.
76925 2021-11-25 GDB Administrator <gdbadmin@sourceware.org>
76927 Automatic date update in version.in
76929 2021-11-24 Simon Marchi <simon.marchi@efficios.com>
76931 gdb: fix help doc for "set index-cache enabled"
76932 When implementing this command, I put "help doc" as a placeholder for
76933 the help string, and forgot to update it. Change it for a real help
76936 Change-Id: Id23c2142c5073dc570bd8a706e9ec6fa8c40eb09
76938 2021-11-24 Simon Marchi <simon.marchi@efficios.com>
76940 Revert (part of) "gdb fix for catch-syscall.exp"
76941 This reverts (par of) commit ab198279120fe7937c0970a8bb881922726678f9.
76942 This commit changed what the test expects when catching the execve
76943 syscall based on the behavior seen on a Linux PowerPC machine. That is,
76944 we get an "entry" event, but no "return" event. This is not what we get
76945 on Linux with other architectures, though, and it seems like a
76946 PowerPC-specific bug.
76948 Revert the part of the patch related to this, but not the other hunk.
76950 Change-Id: I4248776e4299f10999487be96d4acd1b33639996
76952 2021-11-24 Nick Clifton <nickc@redhat.com>
76954 Fix an illegal memory access parsing a corrupt sysroff file.
76956 * sysdump.c (getCHARS): Check for an out of bounds read.
76958 2021-11-24 Andrew Burgess <aburgess@redhat.com>
76960 gdb: fix crash when reading ECOFF debug information
76963 commit 633cf2548bcd3dafe297e21a1dd3574240280d48
76964 Date: Wed May 9 15:42:28 2018 -0600
76966 Remove cleanups from mdebugread.c
76968 the following change was made in the function parse_partial_symbols in
76971 - fdr_to_pst = XCNEWVEC (struct pst_map, hdr->ifdMax + 1);
76972 - old_chain = make_cleanup (xfree, fdr_to_pst);
76973 + gdb::def_vector<struct pst_map> fdr_to_pst_holder (hdr->ifdMax + 1);
76974 + fdr_to_pst = fdr_to_pst_holder.data ();
76976 The problem with this change is that XCNEWVEC calls xcalloc, which in
76977 turn calls calloc, and calloc zero initializes the allocated memory.
76978 In contrast, the new line gdb::def_vector<struct pst_map> specifically
76979 does not initialize the underlying memory.
76981 This is a problem because, later on in this same function, we
76982 increment the n_globals field within 'struct pst_map' objects stored
76983 in the vector. The incrementing is now being done from an
76984 uninitialized starting point.
76986 In this commit we switch from using gdb::def_vector to using
76987 std::vector, this alone should be enough to ensure that the fields are
76988 initialized to zero.
76990 However, for extra clarity, I have also added initial values in the
76991 'struct pst_map' to make it crystal clear how the struct will start
76994 This issue was reported on the mailing list here:
76996 https://sourceware.org/pipermail/gdb-patches/2021-November/183693.html
76998 Co-Authored-By: Lightning <lightningth@gmail.com>
77000 2021-11-24 GDB Administrator <gdbadmin@sourceware.org>
77002 Automatic date update in version.in
77004 2021-11-23 Alexandra Hájková <ahajkova@redhat.com>
77006 configure.ac: Check for the readline.h explicitly
77007 When readline development package is missing make fails with
77008 "configure: error: system readline is not new enough" which
77009 might be confusing. This patch checks for the readline.h explicitly
77010 and makes make to warn about the missing package.
77012 2021-11-23 Tamar Christina <tamar.christina@arm.com>
77014 AArch64: Add support for AArch64 EFI (efi-*-aarch64).
77015 This adds support for efi-*-aarch64 by virtue of adding a new PEI target
77016 pei-aarch64-little. This is not a full target and only exists to support EFI
77019 This means that this target does not support relocation processing and is mostly
77020 a container format. This format has been added to elf based aarch64 targets
77021 such that efi images can be made natively on Linux.
77023 However this target is not valid for use with gas but only with objcopy.
77025 With these changes the resulting file is recognized as an efi image by
77028 > pecli info hello.efi
77031 ================================================================================
77032 MD5: 598c32a778b0f0deebe977fef8578c4e
77033 SHA1: 4580121edd5cb4dc40f51b28f171fd15250df84c
77034 SHA256: 3154bd7cf42433d1c957f6bf55a17ad8c57ed41b29df2d485703349fd6ff1d5c
77037 Type: PE32+ executable (EFI application) (stripped to external PDB), for MS Windows
77038 Compile Time: 1970-01-01 00:00:00 (UTC - 0x0 )
77039 Entry point: 0x2000 (section .text)
77042 ================================================================================
77043 Name RWX VirtSize VirtAddr RawAddr RawSize Entropy md5
77044 .text R-X 0x5bb0 0x2000 0x400 0x5c00 6.39 551fbc264256a3f387de8a891500ae0d
77045 .reloc R-- 0xc 0x8000 0x6000 0x200 0.02 0c45f6d812d079821c1d54c09ab89e1d
77046 .data RW- 0x1d88 0x9000 0x6200 0x1e00 4.18 5d1137c09f01289dc62bf754f7290db3
77047 .dynamic RW- 0xf0 0xb000 0x8000 0x200 0.34 5c94ed3206f05a277e6f04fbf131f131
77048 .rela R-- 0xe58 0xc000 0x8200 0x1000 1.87 8b5c6bc30f3acb7ca7bf2e6789d68519
77049 .dynsym R-- 0x138 0xd000 0x9200 0x200 0.96 bdcf5101da51aadc663ca8859f88138c
77052 ================================================================================
77054 Any magic number is based on the Microsoft PE specification [1].
77056 [1] https://docs.microsoft.com/en-us/windows/win32/debug/pe-format
77060 2021-10-21 Tamar Christina <tamar.christina@arm.com>
77063 * .gitignore (pe-aarch64igen.c): New.
77064 * Makefile.am (pei-aarch64.lo, pe-aarch64igen.lo, pei-aarch64.c,
77065 pe-aarch64igen.c): Add support.
77066 * Makefile.in: Likewise.
77067 * bfd.c (bfd_get_sign_extend_vma): Add pei-aarch64-little.
77068 * coff-aarch64.c: New file.
77069 * coffcode.h (coff_set_arch_mach_hook, coff_set_flags,
77070 coff_write_object_contents) Add aarch64 (aarch64_pei_vec) support.
77071 * config.bfd: Likewise.
77072 * configure: Likewise.
77073 * configure.ac: Likewise.
77074 * libpei.h (GET_OPTHDR_IMAGE_BASE, PUT_OPTHDR_IMAGE_BASE,
77075 GET_OPTHDR_SIZE_OF_STACK_RESERVE, PUT_OPTHDR_SIZE_OF_STACK_RESERVE,
77076 GET_OPTHDR_SIZE_OF_STACK_COMMIT, PUT_OPTHDR_SIZE_OF_STACK_COMMIT,
77077 GET_OPTHDR_SIZE_OF_HEAP_RESERVE, PUT_OPTHDR_SIZE_OF_HEAP_RESERVE,
77078 GET_OPTHDR_SIZE_OF_HEAP_COMMIT, PUT_OPTHDR_SIZE_OF_HEAP_COMMIT,
77079 GET_PDATA_ENTRY, _bfd_peAArch64_bfd_copy_private_bfd_data_common,
77080 _bfd_peAArch64_bfd_copy_private_section_data,
77081 _bfd_peAArch64_get_symbol_info, _bfd_peAArch64_only_swap_filehdr_out,
77082 _bfd_peAArch64_print_private_bfd_data_common,
77083 _bfd_peAArch64i_final_link_postscript,
77084 _bfd_peAArch64i_only_swap_filehdr_out, _bfd_peAArch64i_swap_aouthdr_in,
77085 _bfd_peAArch64i_swap_aouthdr_out, _bfd_peAArch64i_swap_aux_in,
77086 _bfd_peAArch64i_swap_aux_out, _bfd_peAArch64i_swap_lineno_in,
77087 _bfd_peAArch64i_swap_lineno_out, _bfd_peAArch64i_swap_scnhdr_out,
77088 _bfd_peAArch64i_swap_sym_in, _bfd_peAArch64i_swap_sym_out,
77089 _bfd_peAArch64i_swap_debugdir_in, _bfd_peAArch64i_swap_debugdir_out,
77090 _bfd_peAArch64i_write_codeview_record,
77091 _bfd_peAArch64i_slurp_codeview_record,
77092 _bfd_peAArch64_print_ce_compressed_pdata): New.
77093 * peXXigen.c (_bfd_XXi_swap_aouthdr_in, _bfd_XXi_swap_aouthdr_out,
77094 pe_print_pdata, _bfd_XX_print_private_bfd_data_common,
77095 _bfd_XX_bfd_copy_private_section_data, _bfd_XXi_final_link_postscript):
77096 Support COFF_WITH_peAArch64,
77097 * pei-aarch64.c: New file.
77098 * peicode.h (coff_swap_scnhdr_in, pe_ILF_build_a_bfd, pe_ILF_object_p):
77099 Support COFF_WITH_peAArch64.
77100 (jtab): Add dummy entry that traps.
77101 * targets.c (aarch64_pei_vec): New.
77103 binutils/ChangeLog:
77105 2021-10-21 Tamar Christina <tamar.christina@arm.com>
77108 * NEWS: Add new support.
77109 * objcopy.c (convert_efi_target): Add efi-*-aarch64 support.
77110 * testsuite/binutils-all/aarch64/pei-aarch64-little.d: New test.
77111 * testsuite/binutils-all/aarch64/pei-aarch64-little.s: New test.
77115 2021-10-21 Tamar Christina <tamar.christina@arm.com>
77118 * coff/aarch64.h: New file.
77119 * coff/pe.h (IMAGE_FILE_MACHINE_ARM64): New.
77121 2021-11-23 Alan Modra <amodra@gmail.com>
77123 binutils debuginfod test
77124 A missing "return" resulted in this non-ELF fail:
77125 x86_64-w64-mingw32 +FAIL: debuginfod (create separate debug info file)
77127 Also, the debuginfod I have installed does not appear to handle
77128 non-native ELF objects, so only run the test when native.
77130 * testsuite/binutils-all/debuginfod.exp: Don't run test unless
77133 2021-11-23 Alan Modra <amodra@gmail.com>
77135 Update bug reporting address
77136 https://sourceware.org/bugzilla/ everywhere
77139 * configure.ac (ACX_BUGURL): Set to https://sourceware.org/bugzilla/
77140 * po/Make-in (msgid-bugs-address): Likewise.
77141 * README: Report bugs to the above.
77142 * configure: Regenerate.
77144 * po/Make-in (msgid-bugs-address): Update.
77146 * README: Update bug address. Delete mention of gcc.
77147 * po/Make-in: Update bug address.
77149 * po/Make-in: Update bug address.
77151 * po/Make-in: Update bug address.
77153 * po/Make-in: Update bug address.
77155 * po/Make-in: Update bug address.
77157 2021-11-23 Jan (janneke) Nieuwenhuizen <janneke@gnu.org>
77159 gdb: more compile fixes for gnu-nat.c
77160 This fixes compile errors like
77162 ../../gdb-11.1/gdb/gnu-nat.c: In function void add_task_commands():
77163 ../../gdb-11.1/gdb/gnu-nat.c:3204:17: error: no matching function for call to add_cmd(const char [8], command_class, cmd_list_element*&, char*, cmd_list_element**)
77166 In file included from ../../gdb-11.1/gdb/completer.h:21,
77167 from ../../gdb-11.1/gdb/symtab.h:36,
77168 from ../../gdb-11.1/gdb/infrun.h:21,
77169 from ../../gdb-11.1/gdb/target.h:42,
77170 from ../../gdb-11.1/gdb/inf-child.h:23,
77171 from ../../gdb-11.1/gdb/gnu-nat.h:38,
77172 from ../../gdb-11.1/gdb/gnu-nat.c:24:
77173 ../../gdb-11.1/gdb/command.h:160:33: note: candidate: cmd_list_element* add_cmd(const char*, command_class, void (*)(const char*, int), const char*, cmd_list_element**)
77174 160 | extern struct cmd_list_element *add_cmd (const char *, enum command_class,
77176 ../../gdb-11.1/gdb/command.h:161:30: note: no known conversion for argument 3 from cmd_list_element* to void (*)(const char*, int)
77177 161 | cmd_const_cfunc_ftype *fun,
77178 | ~~~~~~~~~~~~~~~~~~~~~~~^~~
77179 ../../gdb-11.1/gdb/command.h:167:33: note: candidate: cmd_list_element* add_cmd(const char*, command_class, const char*, cmd_list_element**)
77180 167 | extern struct cmd_list_element *add_cmd (const char *, enum command_class,
77182 ../../gdb-11.1/gdb/command.h:167:33: note: candidate expects 4 arguments, 5 provided
77183 ../../gdb-11.1/gdb/gnu-nat.c:3210:18: error: no matching function for call to add_cmd(const char [8], command_class, cmd_list_element*&, char*, cmd_list_element**)
77187 Change-Id: Ie9029363d3fb40e34e8f5b1ab503745bc44bfe3f
77189 2021-11-23 Andrea Monaco <andrea.monaco@autistici.org>
77191 gnu-nat.c: fix calls to add_info_alias
77192 Some time ago add_info_alias was changed (commit
77193 e0f25bd9717c7973197095523db7c1cdc956cea2). These calls were not updated
77194 and caused errors on compilation.
77196 Change-Id: I354ae4e8b8926d785abc94ec7142471ffd76d2de
77198 2021-11-23 GDB Administrator <gdbadmin@sourceware.org>
77200 Automatic date update in version.in
77202 2021-11-22 Simon Marchi <simon.marchi@efficios.com>
77204 gdb: pass more const target_waitstatus by reference
77205 While working on target_waitstatus changes, I noticed a few places where
77206 const target_waitstatus objects could be passed by reference instead of
77207 by pointers. And in some cases, places where a target_waitstatus could
77208 be passed as const, but was not. Convert them as much as possible.
77210 Change-Id: Ied552d464be5d5b87489913b95f9720a5ad50c5a
77212 2021-11-22 Simon Marchi <simon.marchi@efficios.com>
77214 gdb: introduce target_waitkind_str, use it in target_waitstatus::to_string
77215 I would like to print target_waitkind values in debug messages, so I
77216 think that a target_waitkind-to-string function would be useful. While
77217 at it, use it in target_waitstatus::to_string. This changes the output
77218 of target_waitstatus::to_string a bit, but I think it is for the better.
77219 The debug messages will show a string matching exactly the
77220 target_waitkind enumerator (minus the TARGET_WAITKIND prefix).
77222 As a convenience, make string_appendf return the same reference to
77223 string it got as a parameter. This allows doing this:
77225 return string_appendf (str, "foo");
77227 ... keeping the code concise.
77229 Change-Id: I383dffc9c78614e7d0668b1516073905e798eef7
77231 2021-11-22 Simon Marchi <simon.marchi@efficios.com>
77233 gdb: rename target_waitstatus_to_string to target_waitstatus::to_string
77234 Make target_waitstatus_to_string a "to_string" method of
77235 target_waitstatus, a bit like we have ptid_t::to_string already. This
77236 will save a bit of typing.
77238 Change-Id: Id261b7a09fa9fa3c738abac131c191a6f9c13905
77240 2021-11-22 Nelson Chu <nelson.chu@sifive.com>
77242 RISC-V: Removed the redundant NULL pointer check in the riscv_update_subset.
77243 If we always use the .option arch to call the riscv_update_subset, then
77244 it is almost impossible that the input string will be NULL. Therefore,
77245 just remove the redundant NULL pointer check in the riscv_update_subset.
77248 * elfxx-riscv.c (riscv_update_subset): Removed the redundant NULL
77251 2021-11-22 Nelson Chu <nelson.chu@sifive.com>
77253 RISC-V: Replace .option rvc/norvc with .option arch, +c/-c.
77254 Since the .option rvc/norvc directives are obsolete, replace them with
77255 the new proposed diretives: .option arch, +c/-c. And also reset the
77256 riscv_opts.rvc flag for the .option arch directives.
77259 * config/tc-riscv.c (s_riscv_option): Reset the riscv_opts.rvc
77260 for the .option arch directives.
77261 * testsuite/gas/riscv/align-1.s: Replace the obsolete .option
77262 rvc/norvc with .option arch, +c/-c.
77263 * testsuite/gas/riscv/c-add-addi.s: Likewise.
77264 * testsuite/gas/riscv/c-nonzero-imm.s: Likewise.
77265 * testsuite/gas/riscv/c-nonzero-reg.s: Likewise.
77266 * testsuite/gas/riscv/c-zero-imm-64.s: Likewise.
77267 * testsuite/gas/riscv/c-zero-imm.s: Likewise.
77268 * testsuite/gas/riscv/c-zero-reg.s: Likewise.
77269 * testsuite/gas/riscv/ext.s: Likewise.
77270 * testsuite/gas/riscv/mapping-01.s: Likewise.
77271 * testsuite/gas/riscv/mapping-02.s: Likewise.
77272 * testsuite/gas/riscv/mapping-03.s: Likewise.
77273 * testsuite/gas/riscv/mapping-04.s: Likewise.
77274 * testsuite/gas/riscv/no-relax-align-2.s: Likewise.
77275 * testsuite/gas/riscv/shamt-32.s: Likewise.
77276 * testsuite/gas/riscv/shamt-64.s: Likewise.
77278 2021-11-22 Tom de Vries <tdevries@suse.de>
77280 [gdb/build] Fix x86_64 x32 build
77281 A build error on x86_64 with x32 abi was reported here (
77282 https://sourceware.org/pipermail/gdb/2021-November/049787.html ):
77284 gdb/nat/amd64-linux-siginfo.c:280:42: error: \
77285 'struct compat_x32_siginfo_t::<unnamed union>::<unnamed>' has no member \
77286 named 'si_addr_bnd'
77287 280 | #define cpt_si_lower _sifields._sigfault.si_addr_bnd._lower
77289 gdb/nat/amd64-linux-siginfo.c:337:38: note: in expansion of macro 'cpt_si_lower'
77290 337 | to->cpt_si_lower = from_ptrace.cpt_si_lower;
77294 The problem is that code added in commit d3d7d1ba3bb "[gdb/tdep] Handle
77295 si_addr_bnd in compat_siginfo_from_siginfo" doesn't compile on an x86_64 x32
77296 setup, because compat_x32_siginfo_t doesn't have the si_addr_bnd fields.
77298 Fix this conservatively by disabling the code for x32.
77300 Tested on x86_64-linux.
77302 2021-11-22 Nelson Chu <nelson.chu@sifive.com>
77304 RISC-V: PR28610, Fix ASAN heap-buffer-overflow error in riscv_update_subset.
77305 The architecture parser in riscv_update_subset shouldn't check (or access)
77306 the pointer space which doesn't exist.
77310 * elfxx-riscv.c (riscv_update_subset): The architecture parser
77311 shouldn't access the pointer space which doesn't exist.
77313 2021-11-22 Tom de Vries <tdevries@suse.de>
77315 [gdb/symtab] Support .debug_line with DW_FORM_line_strp
77316 I noticed a new gcc option -gdwarf64 and tried it out (using gcc 11.2.1).
77318 With a test-case hello.c:
77323 printf ("hello\n");
77327 compiled like this:
77329 $ gcc -g -gdwarf64 ~/hello.c
77333 $ gdb -q -batch a.out
77334 DW_FORM_line_strp pointing outside of .debug_line_str section \
77338 Debugging gdb revealed that the string offset is:
77341 objfile=0x182ab70, str_offset=1378684502312,
77342 form_name=0xeae9b5 "DW_FORM_line_strp")
77343 at src/gdb/dwarf2/section.c:208
77344 208 error (_("%s pointing outside of %s section [in module %s]"),
77345 (gdb) p /x str_offset
77349 which is read when parsing a .debug_line entry at 0x1e0.
77351 Looking with readelf at the 0x1e0 entry, we have:
77353 The Directory Table (offset 0x202, lines 2, columns 1):
77355 0 (indirect line string, offset: 0x128): /data/gdb_versions/devel
77356 1 (indirect line string, offset: 0x141): /home/vries
77358 which in a hexdump looks like:
77360 0x00000200 1f022801 00004101 00000201 1f020f02
77363 What happens is the following:
77364 - readelf interprets the DW_FORM_line_strp reference to .debug_line_str as
77365 a 4 byte value, and sees entries 0x00000128 and 0x00000141.
77366 - gdb instead interprets it as an 8 byte value, and sees as first entry
77367 0x0000014100000128, which is too big so it bails out.
77369 AFAIU, gdb is wrong. It assumes DW_FORM_line_strp is 8 bytes on the basis
77370 that the corresponding CU is 64-bit DWARF. However, the .debug_line
77371 contribution has it's own initial_length field, and encodes there that it's
77374 Fix this by using the correct offset size for DW_FORM_line_strp references
77377 Note: the described test-case does trigger this complaint (both with and
77378 without this patch):
77380 $ gdb -q -batch -iex "set complaints 10" a.out
77381 During symbol reading: intermixed 32-bit and 64-bit DWARF sections
77384 The reason that the CU has 64-bit dwarf is because -gdwarf64 was passed to
77385 gcc. The reason that the .debug_line entry has 32-bit dwarf is because that's
77386 what gas generates. Perhaps this is complaint-worthy, but I don't think it
77389 Tested on x86_64-linux, using native and target board dwarf64.exp.
77391 2021-11-22 Tom de Vries <tdevries@suse.de>
77393 [gdb/testsuite] Add target board dwarf64.exp
77394 Add a new target board dwarf64.exp, that runs test with -gdwarf64.
77396 Tested on x86_64-linux.
77398 2021-11-22 Tom de Vries <tdevries@suse.de>
77400 [gdb/testsuite] Support .debug_line v5 in dwarf assembler
77401 The v5 section version for .debug_line has:
77402 - two new fields address_size and segment_selector_size
77403 - a different way to encode the directory and filename tables.
77405 Add support for this in the dwarf assembler.
77407 For now, make the v5 directory and filename tables work with the v4 type of
77408 specification in the test-cases by adding duplicate entries at position 0.
77410 This will need to be properly fixed with an intrusive fix that changes how
77411 directory and filename entries are specified in the test-cases, f.i:
77413 set diridx [include_dir "${srcdir}/${subdir}"]
77414 set fileidx [file_name "$srcfile" $diridx]
77417 Tested on x86_64-linux.
77419 2021-11-22 Tom de Vries <tdevries@suse.de>
77421 [gdb/testsuite] Factor out _line_finalize_header
77422 Rather than generate dwarf immediately in procs include_dir and file_name,
77423 postpone generation and store the data in variables. Then handle the
77424 generation in a new proc _line_finalize_header.
77426 Tested on x86-64-linux.
77428 2021-11-22 Tom de Vries <tdevries@suse.de>
77430 [gdb/testsuite] Support .debug_line v4 in dwarf assembler
77431 The .debug_line header got a new field in v4:
77432 maximum_operations_per_instruction.
77434 Generate this field in the dwarf assembler, for now hardcoding the value to 1,
77437 Tested on x86_64-linux.
77439 2021-11-22 Tom de Vries <tdevries@suse.de>
77441 [gdb/testsuite] Add test-case gdb.dwarf2/dw2-lines.exp
77442 Add a new test-case gdb.dwarf2/dw2-lines.exp that tests various .debug_line
77445 Tested on x86_64-linux.
77447 2021-11-22 Tom de Vries <tdevries@suse.de>
77449 [gdb/testsuite] Speed up MACRO_AT_* calls
77450 Currently, for each MACRO_AT_range or MACRO_AT_func in dwarf assembly the
77452 - $srcdir/$subdir/$srcfile is compiled to an executable using
77454 - a new gdb instance is started
77455 - the new executable is loaded.
77457 This is inefficient, because the executable is identical within the same
77458 Dwarf::assemble call.
77460 Share the gdb instance in the same Dwarf::assemble invocation, which speeds
77461 up a make check with RUNTESTFLAGS like this to catch all dwarf assembly
77464 rtf=$(echo $(cd src/gdb/testsuite; find gdb.* -type f -name "*.exp" \
77465 | xargs grep -l Dwarf::assemble))
77480 Tested on x86_64-linux.
77482 2021-11-22 GDB Administrator <gdbadmin@sourceware.org>
77484 Automatic date update in version.in
77486 2021-11-21 Lancelot SIX <lsix@lancelotsix.com>
77488 gdb/testsuite: Remove duplicates in gdb.base/catch-signal.exp
77489 When running the testsuite I have the following:
77491 Running .../gdb/testsuite/gdb.base/catch-signal.exp ...
77492 DUPLICATE: gdb.base/catch-signal.exp: SIGHUP: continue
77493 DUPLICATE: gdb.base/catch-signal.exp: SIGHUP: continue
77494 DUPLICATE: gdb.base/catch-signal.exp: 1: continue
77495 DUPLICATE: gdb.base/catch-signal.exp: 1: continue
77496 DUPLICATE: gdb.base/catch-signal.exp: SIGHUP SIGUSR2: continue
77497 DUPLICATE: gdb.base/catch-signal.exp: SIGHUP SIGUSR2: continue
77499 This patch removes DUPLICATE in gdb.base/catch-signal.exp by explicitly
77500 giving names to the offending 'gdb_test "continue"' statements to make
77503 Tested on x86_64-linux.
77505 2021-11-21 Mike Frysinger <vapier@gentoo.org>
77507 sim: v850: fix cpu_option testsuite handling
77508 The v850 testsuite code has been testing the $opt variable, but this
77509 was never actually set anywhere globally or v850-specific. Instead,
77510 this was a random variable leaking out of the sh testsuite code. As
77511 far as I can tell, it has always been this way. That means the code
77512 only ever tested the v850 cpu target (which is the default).
77514 This failure can be easily seen in practice by running the v850 code
77515 in isolation and seeing it crash:
77516 $ runtest v850/allinsns.exp
77518 Running target unix
77519 Using /usr/share/dejagnu/baseboards/unix.exp as board description file for target.
77520 Using /usr/share/dejagnu/config/unix.exp as generic interface file for target.
77521 Using ../../../sim/testsuite/config/default.exp as tool-and-target-specific interface file.
77522 WARNING: Assuming target board is the local machine (which is probably wrong).
77523 You may need to set your DEJAGNU environment variable.
77524 Running ../../../sim/testsuite/v850/allinsns.exp ...
77525 ERROR: tcl error sourcing ../../../sim/testsuite/v850/allinsns.exp.
77526 ERROR: tcl error code TCL LOOKUP VARNAME opt
77527 ERROR: can't read "opt": no such variable
77529 "switch -regexp -- $opt {
77531 Backing up a bit, the reason for this logic in the first place is
77532 because the common sim testsuite code makes an assumption about the
77533 assembler options with cpu_option -- the option and its value are
77534 always separated by an =. This is not the case with v850. So tweak
77535 the core sim logic a bit to support omitting the = so that we can
77536 switch v850 to the standard all_machs setting and avoid opt entirely.
77538 2021-11-21 GDB Administrator <gdbadmin@sourceware.org>
77540 Automatic date update in version.in
77542 2021-11-20 Jeff Law <jeffreyalaw@gmail.com>
77544 Fix intermittent failures on the H8, particularly H8/SX tests.
77545 The upstream GCC tester has showed spurious execution failures on the
77546 H8 target for the H8/SX multilibs. I suspected memory corruption or an
77547 uninitialized variable early as the same binary would sometimes work and
77548 sometimes it got the wrong result. Worse yet, the point where the test
77549 determined it was getting the wrong result would change.
77551 Because it only happened on the H8/SX variant I was able to zero in on
77552 the "mova" support and the "short form" of those instructions in particular.
77554 As the code stands it checks if code->op3.type == 0 to try and identify cases
77555 where op3 wasn't filled in and thus we've got the short form of the mova
77558 But for the short-form of those instructions we never set any of the "op3"
77559 data structure. We get whatever was lying around -- it's usually zero and
77560 thus things usually work, but if the stale data was nonzero, then we'd
77561 fail to recognize the instruction as a short-form and fail to set up the
77562 various fields appropriately.
77564 I initially initialized the op3.type field to zero, but didn't like that
77565 because it was inconsistent with how other operands were initialized.
77566 Bringing consistency meant using -1 as the initializer value and adjusting
77567 the check for short form mova appropriately.
77569 I've had this in the upstream GCC tester for perhaps a year at this point
77570 and haven't seen any of the intermittent failures again.
77572 2021-11-20 Simon Marchi <simon.marchi@efficios.com>
77574 gdbsupport: fix array-view compilation with c++11 && _GLIBCXX_DEBUG
77575 When building with -std=c++11 and -D_GLIBCXX_DEBUG=1, we get some errors
77578 CXX unittests/array-view-selftests.o
77579 In file included from /home/smarchi/src/binutils-gdb/gdb/utils.h:25,
77580 from /home/smarchi/src/binutils-gdb/gdb/defs.h:630,
77581 from /home/smarchi/src/binutils-gdb/gdb/unittests/array-view-selftests.c:20:
77582 /home/smarchi/src/binutils-gdb/gdb/../gdbsupport/array-view.h: In instantiation of constexpr gdb::array_view<T> gdb::array_view<T>::slice(gdb::array_view<T>::size_type, gdb::array_view<T>::size_type) const [with T = unsigned char; gdb::array_view<T>::size_type = long unsigned int:
77583 /home/smarchi/src/binutils-gdb/gdb/unittests/array-view-selftests.c:532:29: required from here
77584 /home/smarchi/src/binutils-gdb/gdb/../gdbsupport/array-view.h:192:3: error: body of constexpr function constexpr gdb::array_view<T> gdb::array_view<T>::slice(gdb::array_view<T>::size_type, gdb::array_view<T>::size_type) const [with T = unsigned char; gdb::array_view<T>::size_type = long unsigned int not a return-statement
77588 This is because constexpr functions in c++11 can only consist of a
77589 single return statement, so we can't have the gdb_assert in there. Make
77590 the gdb_assert presence conditional to the __cplusplus version, to
77591 enable it only for c++14 and later.
77593 Change-Id: I2ac33f7b4bd1765ddc3ac8d07445b16ac1f340f0
77595 2021-11-20 Tom de Vries <tdevries@suse.de>
77597 [gdb/build] Check if libsource-highlight is usable
77598 When building gdb with g++ 4.8.5, I ran into:
77600 ld: source-cache.o: in function `source_cache::ensure(symtab*)':
77601 source-cache.c:207: undefined reference to \
77602 srchilite::SourceHighlight::SourceHighlight(std::string const&)
77605 [ I configured gdb without explicit settings related to source-highlight, so
77606 we're excercising the enable_source_highlight=auto scenario. ]
77608 The problem is that:
77609 - the source-highlight library is build with system compiler
77610 g++ 7.5.0 which uses the new libstdc++ library abi (see
77611 https://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html )
77612 - gdb is build using g++ 4.8.5 which uses the old abi.
77614 [ There's a compatibility macro _GLIBCXX_USE_CXX11_ABI, but that doesn't work
77615 for this case. Instead, it enables the opposite case where the
77616 source-highlight library is build with g++ 4.8.5 and gdb is build with
77619 Fix this by checking whether the source-highlight library is usable during
77622 In the enable_source_highlight=auto scenario, this allows the build to skip
77623 the unusable library and finish successfully.
77625 In the enable_source_highlight=yes scenario, this allows the build to error
77628 Tested on x86_64-linux.
77630 2021-11-20 Clément Chigot <clement.chigot@atos.net>
77632 bfd: remove wrong comment in xcofflink.c
77633 This comment was long time ago associated to the function
77634 "xcoff_build_ldsyms" which have since been replaced by
77635 "xcoff_build_ldsym".
77637 * xcofflink.c: Remove wrong comment.
77639 2021-11-20 Mike Frysinger <vapier@gentoo.org>
77641 sim: bfin: fix short --env usage in testsuite
77642 Now that we have more than one option that matches "--env", the test
77643 config here doesn't work. Use the explicit --environment.
77645 2021-11-20 GDB Administrator <gdbadmin@sourceware.org>
77647 Automatic date update in version.in
77649 2021-11-19 H.J. Lu <hjl.tools@gmail.com>
77651 elfedit: Align --[in|out]put-abiversion usage
77654 --input-abiversion [0-255] Set input ABIVERSION
77655 --output-abiversion [0-255] Set output ABIVERSION
77659 --input-abiversion [0-255]
77660 Set input ABIVERSION
77661 --output-abiversion [0-255]
77662 Set output ABIVERSION
77664 * elfedit.c (usage): Align --[in|out]put-abiversion usage.
77666 2021-11-19 Tom de Vries <tdevries@suse.de>
77668 [gdb/testsuite] Handle runto fail in gdb.mi/mi-var-cp.exp
77671 PASS: gdb.mi/mi-var-cp.exp: run to mi-var-cp.cc:81 (set breakpoint)
77672 UNRESOLVED: gdb.mi/mi-var-cp.exp: unable to start target
77674 followed by 81 FAILs and two more UNRESOLVEDs.
77676 I didn't manage to reproduce this, but I did notice that the initial
77677 problem causing the UNRESOLVED caused all subsequent UNRESOLVEDs and FAILs.
77679 I emulated the problem by commenting out the send_gdb "run\n" in
77683 - handling mi_run_cmd failure in mi_get_inline_test
77684 - handling mi_run_inline_test failure in gdb.mi/mi-var-cp.exp, and
77685 other test-cases using mi_get_inline_test
77687 Tested on x86_64-linux.
77689 2021-11-19 Tom de Vries <tdevries@suse.de>
77691 [gdb/testsuite] Fix 64-bit dwarf test-cases with -m32
77692 When running test-case gdb.dwarf2/loc-sec-offset.exp with target board -m32,
77695 builtin_spawn -ignore SIGHUP gcc -fno-stack-protector -m32 \
77696 -fdiagnostics-color=never -c -o loc-sec-offset-dw641.o \
77697 loc-sec-offset-dw64.S^M
77698 as: loc-sec-offset-dw641.o: unsupported relocation type: 0x1^M
77699 loc-sec-offset-dw64.S: Assembler messages:^M
77700 loc-sec-offset-dw64.S:29: Error: cannot represent relocation type \
77704 Looking at line 29, we have:
77706 .8byte .Labbrev1_begin /* Abbrevs */
77709 It would be nice if the assembler could handle this somehow. But I guess
77710 it's not unreasonable that an assembler for a 32-bit architecture will object
77711 to handling 64-bit labels.
77713 Instead, work around this in the dwarf assembler by emitting:
77715 .4byte .Labbrev1_begin /* Abbrevs (lsw) */
77716 .4byte 0 /* Abbrevs (msw) */
77719 Tested on x86_64-linux with target board unix/-m32.
77721 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28383
77723 2021-11-19 Tom de Vries <tdevries@suse.de>
77725 [gdb/testsuite] Fix gdb.threads/thread-specific-bp.exp
77726 On OBS I ran into a failure in test-case gdb.threads/thread-specific-bp.exp:
77728 (gdb) PASS: gdb.threads/thread-specific-bp.exp: non-stop: continue to end
77730 Num Type Disp Enb Address What^M
77731 1 breakpoint keep y 0x0000555555555167 in main at $src:36^M
77732 breakpoint already hit 1 time^M
77733 2 breakpoint keep y 0x0000555555555151 in start at $src:23^M
77734 breakpoint already hit 1 time^M
77735 3 breakpoint keep y 0x0000555555555167 in main at $src:36 thread 2^M
77736 stop only in thread 2^M
77737 4 breakpoint keep y 0x000055555555515c in end at $src:29^M
77738 breakpoint already hit 1 time^M
77739 (gdb) [Thread 0x7ffff7db1640 (LWP 19984) exited]^M
77740 Thread-specific breakpoint 3 deleted - thread 2 no longer in the thread list.^M
77741 FAIL: gdb.threads/thread-specific-bp.exp: non-stop: \
77742 thread-specific breakpoint was deleted (timeout)
77745 Fix this by waiting for the "[Thread 0x7ffff7db1640 (LWP 19984) exited]"
77746 message before issuing the "info breakpoint command".
77748 Tested on x86_64-linux.
77750 2021-11-19 Christina Schimpe <christina.schimpe@intel.com>
77752 gdb/testsuite: Extend tests for print of cv qualifiers
77753 This commit supplements whatis and ptype command tests for print of
77754 const-volatile qualifiers.
77756 gdb/testsuite/ChangeLog:
77757 2021-11-16 Christina Schimpe <christina.schimpe@intel.com>
77759 * gdb.cp/ptype-cv-cp.cc: New const and volatile typedef
77761 * gdb.cp/ptype-cv-cp.exp: Add new tests.
77763 2021-11-19 Christina Schimpe <christina.schimpe@intel.com>
77765 gdb: Print cv qualifiers if class attributes are substituted
77766 Make ptype print const/volatile qualifiers when template or typedef
77767 attributes are substituted.
77769 For a programm like
77771 template<typename DataT>
77774 typedef float myfloat;
77778 const myfloat me2=2.0;
77788 gdb outputs the following type for cfoo's attributes:
77792 Breakpoint 1 at 0x1170: file tmp.cc, line 14.
77794 Starting program: /tmp
77796 Breakpoint 1, main () at tmp.cc:14
77799 type = class Cfoo<int> [with DataT = int] {
77806 typedef float myfloat;
77811 The cv qualifiers (const in this case) are ignored for me1 and me2.
77816 type = class Cfoo<int> [with DataT = int] {
77823 typedef float myfloat;
77828 2021-11-16 Christina Schimpe <christina.schimpe@intel.com>
77830 * gdb/c-typeprint.c: Print cv qualifiers in case of parameter
77833 gdb/testsuite/ChangeLog:
77834 2021-11-16 Christina Schimpe <christina.schimpe@intel.com>
77836 * gdb.cp/templates.cc: New template class Cfoo with const,
77837 template, typdef and integer attributes.
77838 * gdb.cp/templates.exp: Add new test using ptype and ptype/r
77839 commmands for template class CFoo.
77841 2021-11-19 Nelson Chu <nelson.chu@sifive.com>
77843 RISC-V: Support new .option arch directive.
77844 https://github.com/riscv/riscv-asm-manual/pull/67
77847 .option arch, +<extension><version>, ...
77848 .option arch, -<extension>
77849 .option arch, =<ISA string>
77851 The new direcitve is used to enable/disable extensions for the specific
77852 code region. For example,
77854 .attribute arch, "rv64ic" # arch = rv64i2p0_c2p0
77856 .option arch, +d2p0, -c # arch = rv64i2p0_f2p0_d2p0, f is added implied
77857 .option arch, =rv32gc # arch = rv32i2p0_m2p0_a2p0_f2p0_d2p0_c2p0
77858 .option pop # arch = rv64i2p0_c2p0
77861 1. ".option rvc/norvc" have the same behavior as ".option arch +c/-c".
77862 2. ".option arch -i" is illegal, since we cannot remove base i extension.
77863 3. If arch=rv64i2p0, then ".option arch, +i3p0" will update the i's version
77865 4. If arch=rv64i3p0, then ".option arch, +i" will update the i's version
77866 from 2.0 to the default one according to the chosen isa spec.
77869 * elfxx-riscv.c (riscv_add_subset): If the subset is already added,
77870 and the new versions are not RISCV_UNKNOWN_VERSION, then update the
77871 versions to the subset list.
77872 (riscv_copy_subset): New function. Copy the subset from list.
77873 (riscv_copy_subset_list): New function. Return the new copyed list.
77874 (riscv_update_subset): Updated to make .option arch directives workable.
77875 * elfxx-riscv.h: Updated.
77877 * config/tc-riscv.c (riscv_subsets): Defined as a pointer.
77878 (riscv_rps_as): Init the subset_list to NULL, we will set it later
77879 once riscv_opts_stack is created or updated.
77880 (struct riscv_option_stack, riscv_opts_stack): Moved forward.
77881 (riscv_set_arch): Updated.
77882 (s_riscv_option): Support new .option arch directive, to add, remove
77883 or update subsets for the specific code region.
77884 (riscv_write_out_attrs): Updated.
77885 * doc/c-riscv.texi: Added document for new .option arch directive.
77886 * testsuite/gas/riscv/option-arch-01a.d: New testcase.
77887 * testsuite/gas/riscv/option-arch-01b.d: Likewise.
77888 * testsuite/gas/riscv/option-arch-01.s: Likewise..
77889 * testsuite/gas/riscv/option-arch-02.d: Likewise.
77890 * testsuite/gas/riscv/option-arch-02.s: Likewise.
77891 * testsuite/gas/riscv/option-arch-fail.d: Likewise.
77892 * testsuite/gas/riscv/option-arch-fail.l: Likewise.
77893 * testsuite/gas/riscv/option-arch-fail.s: Likewise.
77895 2021-11-19 Alan Modra <amodra@gmail.com>
77897 Re: Add multibyte character warning option to the assembler.
77898 On hppa*-hp-hpux* run_dump_test edits the test file, adjusting .comm
77899 directives to suit those target's unusual syntax. Thus gas is passed
77900 a temporary file name.
77902 * testsuite/gas/all/multibyte1.l: Ignore file name.
77904 2021-11-19 Mike Frysinger <vapier@gentoo.org>
77906 sim: install various doc files
77908 2021-11-19 Nelson Chu <nelson.chu@sifive.com>
77910 RISC-V: Support STO_RISCV_VARIANT_CC and DT_RISCV_VARIANT_CC.
77911 This is the original discussion,
77912 https://github.com/riscv/riscv-elf-psabi-doc/pull/190
77914 And here is the glibc part,
77915 https://sourceware.org/pipermail/libc-alpha/2021-August/129931.html
77917 For binutils part, we need to support a new direcitve: .variant_cc.
77918 The function symbol marked by .variant_cc means it need to be resolved
77919 directly without resolver for dynamic linker. We also add a new dynamic
77920 entry, STO_RISCV_VARIANT_CC, to indicate there are symbols with the
77921 special attribute in the dynamic symbol table of the object.
77923 I heard that llvm already have supported this in their mainline, so
77924 I think it's time to commit this.
77927 * elfnn-riscv.c (riscv_elf_link_hash_table): Added variant_cc
77928 flag. It is used to check if relocations for variant CC symbols
77930 (allocate_dynrelocs): If the symbol has STO_RISCV_VARIANT_CC
77931 flag, then raise the variant_cc flag of riscv_elf_link_hash_table.
77932 (riscv_elf_size_dynamic_sections): Added dynamic entry for
77934 (riscv_elf_merge_symbol_attribute): New function, used to merge
77935 non-visibility st_other attributes, including STO_RISCV_VARIANT_CC.
77937 * readelf.c (get_riscv_dynamic_type): New function.
77938 (get_dynamic_type): Called get_riscv_dynamic_type for riscv targets.
77939 (get_riscv_symbol_other): New function.
77940 (get_symbol_other): Called get_riscv_symbol_other for riscv targets.
77942 * config/tc-riscv.c (s_variant_cc): Marked symbol that it follows a
77943 variant CC convention.
77944 (riscv_elf_copy_symbol_attributes): Same as elf_copy_symbol_attributes,
77945 but without copying st_other. If a function symbol has special st_other
77946 value set via directives, then attaching an IFUNC resolver to that symbol
77947 should not override the st_other setting.
77948 (riscv_pseudo_table): Support variant_cc diretive.
77949 * config/tc-riscv.h (OBJ_COPY_SYMBOL_ATTRIBUTES): Defined.
77950 * testsuite/gas/riscv/variant_cc-set.d: New testcase.
77951 * testsuite/gas/riscv/variant_cc-set.s: Likewise.
77952 * testsuite/gas/riscv/variant_cc.d: Likewise.
77953 * testsuite/gas/riscv/variant_cc.s: Likewise.
77955 * elf/riscv.h (DT_RISCV_VARIANT_CC): Defined to (DT_LOPROC + 1).
77956 (STO_RISCV_VARIANT_CC): Defined to 0x80.
77958 * testsuite/ld-riscv-elf/variant_cc-1.s: New testcase.
77959 * testsuite/ld-riscv-elf/variant_cc-2.s: Likewise.
77960 * testsuite/ld-riscv-elf/variant_cc-now.d: Likewise.
77961 * testsuite/ld-riscv-elf/variant_cc-r.d: Likewise.
77962 * testsuite/ld-riscv-elf/variant_cc-shared.d: Likewise.
77963 * testsuite/ld-riscv-elf/ld-riscv-elf.exp: Updated.
77965 2021-11-19 Mike Frysinger <vapier@gentoo.org>
77967 sim: use program_transform_name for libsim
77968 Instead of always using target_alias as a prefix on the name, use
77969 program_transform_name instead so that the library is scoped in the
77970 same way as the run program.
77972 sim: avoid installing headers when there is no sim
77973 If we aren't building any sims, don't install the sim headers as they
77974 won't be useful to anyone.
77976 2021-11-19 GDB Administrator <gdbadmin@sourceware.org>
77978 Automatic date update in version.in
77980 2021-11-18 Kevin Buettner <kevinb@redhat.com>
77982 dprintf-execution-x-script.exp: Adjust test for native-extended-gdbserver
77983 Without this commit, doing...
77985 make check RUNTESTFLAGS="--target_board=native-extended-gdbserver" \
77986 TESTS="gdb.base/dprintf-execution-x-script.exp"
77988 ...will show one failure.
77990 Here's a snippet from gdb.log showing the circumstances - I've trimmed
77991 the paths for readability:
77993 builtin_spawn gdb -nw -nx -data-directory data-directory -iex set height 0 -iex set width 0 -iex set auto-connect-native-target off -iex set sysroot -ex set height unlimited -x testsuite/gdb.base/dprintf-execution-x-script.gdb --args testsuite/outputs/gdb.base/dprintf-execution-x-script/dprintf-execution-x-script
77995 Reading symbols from testsuite/outputs/gdb.base/dprintf-execution-x-script/dprintf-execution-x-script...
77996 Dprintf 1 at 0x40116e: file testsuite/gdb.base/dprintf-execution-x-script.c, line 38.
77997 Breakpoint 2 at 0x40113a: file testsuite/gdb.base/dprintf-execution-x-script.c, line 26.
77998 testsuite/gdb.base/dprintf-execution-x-script.gdb:21: Error in sourced command file:
77999 Don't know how to run. Try "help target".
78000 (gdb) FAIL: gdb.base/dprintf-execution-x-script.exp: load and run script with -x
78002 GNU gdb (GDB) 12.0.50.20211118-git
78003 Copyright (C) 2021 Free Software Foundation, Inc.
78007 (gdb) builtin_spawn gdbserver/gdbserver --once --multi localhost:2346
78008 Listening on port 2346
78009 target extended-remote localhost:2346
78010 Remote debugging using localhost:2346
78012 [Tests after this point will pass.]
78014 Note that the command which spawns gdb prevents the gdb script from
78015 using the native target via "-iex set auto-connect-native-target off".
78017 Moreover, the script in question contains a "run" command, so GDB
78018 doesn't know how to run (since it's prevented from using the native
78019 target and no alternate "target" command has been issued. But, once
78020 GDB finishes starting up, the test will spawn a gdbserver and then
78021 connect to it. The other (two) tests after this point both pass.
78023 I've fixed this by using gdb_test_multiple instead of gdb_test.
78024 When a "Don't know how to run message" is received, the test is
78027 I've also added a comment explaining the reason for needing to check
78028 for "Don't know how to run" despite bailing out at the top of the test
78031 if ![target_can_use_run_cmd] {
78035 2021-11-18 Simon Marchi <simon.marchi@polymtl.ca>
78037 gdb: fix array-view-selftests.c build with g++ 4.8
78038 When building with g++ 4.8, I get:
78040 CXX unittests/array-view-selftests.o
78041 /home/smarchi/src/binutils-gdb/gdb/unittests/array-view-selftests.c:123:42: error: expected 'class' before 'Container'
78042 template<template<typename ...> typename Container>
78045 I am no C++ template expert, but it looks like if I change "typename" for
78046 "class", as the compiler kind of suggests, the code compiles.
78048 Change-Id: I9c3edd29fb2b190069f0ce0dbf3bc3604d175f48
78050 2021-11-18 Simon Marchi <simon.marchi@polymtl.ca>
78052 gdb: fix ia64-tdep.c build with g++ 4.8
78053 When building with g++ 4.8, I get:
78056 /home/smarchi/src/binutils-gdb/gdb/ia64-tdep.c:3862:1: error: could not convert '{ia64_allocate_new_rse_frame, ia64_store_argument_in_slot, ia64_set_function_addr}' from '<brace
78057 -enclosed initializer list>' to 'const ia64_infcall_ops'
78061 This happens since commit 345bd07cce3 ("gdb: fix gdbarch_tdep ODR
78062 violation"), which added default values for ia64_infcall_ops fields. It
78063 looks like g++ 4.8 doesn't like initializing the ia64_infcall_ops object
78064 using the brace-enclosed initializer list when the ia64_infcall_ops
78065 fields are assigned default values.
78067 Later compilers don't have a problem with that, so I suppose that the
78068 code is correct, but still, change it to make gcc 4.8 happy. Don't
78069 initialize the fields of ia64_infcall_ops directly, instead
78070 default-initialize ia64_gdbarch_tdep::infcall_ops.
78072 Change-Id: I35e3a61abd7b7bbcafe6cb207078c738c5266d76
78074 2021-11-18 Simon Marchi <simon.marchi@polymtl.ca>
78076 gdb: move AIX_TEXT_SEGMENT_BASE to rs6000-aix-tdep.c, remove rs6000-tdep.h
78077 The contents of rs6000-tdep.h (AIX_TEXT_SEGMENT_BASE) is AIX-specific,
78078 so I thought that this file should be named rs6000-aix-tdep.h. But
78079 there's already a rs6000-aix-tdep.h, so then I though
78080 AIX_TEXT_SEGMENT_BASE should simply be moved there, and rs6000-tdep.h
78081 deleted. But then I realized that AIX_TEXT_SEGMENT_BASE is only used in
78082 rs6000-aix-tdep.c, so move it to the beginning of that file.
78084 Change-Id: Ia212c6fae202f31aedb46575821cd642beeda7a3
78086 2021-11-18 Simon Marchi <simon.marchi@polymtl.ca>
78088 gdb: rename rs6000-nat.c to rs6000-aix-nat.c
78089 This file seems to be AIX-specific, according to its contents and
78090 configure.nat. Rename it to rs6000-aix-nat.c, to make that clear (and
78091 to follow the convention).
78093 Change-Id: Ib418dddc6b79b2e28f64431121742b5e87f5f4f5
78095 2021-11-18 Tom de Vries <tdevries@suse.de>
78097 [gdb/doc] Fix negative repeat count examining memory example
78098 The documentation for the examining memory command x contains an example:
78100 You can also specify a negative repeat count to examine memory backward from
78101 the given address. For example, 'x/-3uh 0x54320' prints three halfwords (h)
78102 at 0x54314, 0x54328, and 0x5431c.
78105 The 0x54328 looks like a typo, which was intended to be 0x54318.
78107 But the series uses a 4-byte distance, while the halfword size used in the
78108 command means a 2-byte distance, so the series should be:
78110 0x5431a, 0x5431c, and 0x5431e.
78113 Fix this by updating the addresses in the example accordingly.
78115 Reported here ( https://sourceware.org/pipermail/gdb/2021-November/049784.html
78118 2021-11-18 Nick Clifton <nickc@redhat.com>
78120 Add multibyte character warning option to the assembler.
78121 * as.c (parse_args): Add support for --multibyte-handling.
78122 * as.h (multibyte_handling): Declare.
78123 * app.c (scan_for_multibyte_characters): New function.
78124 (do_scrub_chars): Call the new function if multibyte warning is
78126 * input-scrub,c (input_scrub_next_buffer): Call the multibyte
78127 scanning function if multibyte warnings are enabled.
78128 * symbols.c (struct symbol_flags): Add multibyte_warned bit.
78129 (symbol_init): Call the multibyte scanning function if multibyte
78130 symbol warnings are enabled.
78131 (S_SET_SEGMENT): Likewise.
78132 * NEWS: Mention the new feature.
78133 * doc/as.texi: Document the new feature.
78134 * testsuite/gas/all/multibyte.s: New test source file.
78135 * testsuite/gas/all/multibyte1.d: New test driver file.
78136 * testsuite/gas/all/multibyte1.l: New test expected output.
78137 * testsuite/gas/all/multibyte2.d: New test driver file.
78138 * testsuite/gas/all/multibyte2.l: New test expected output.
78139 * testsuite/gas/all/gas.exp: Run the new tests.
78141 2021-11-18 Simon Marchi <simon.marchi@polymtl.ca>
78143 gdb: include gdbarch.h in all files extending gdbarch_tdep
78144 Commit 345bd07cce33 ("gdb: fix gdbarch_tdep ODR violation") made a bunch
78145 of files define a *_gdbarch_tdep class that inherits from a gdbarch_tdep
78146 base. But some of these files don't include gdbarch.h, where
78147 gdbarch_tdep is defined. This may cause build errors if gdbarch.h isn't
78148 already included by chance by some other header file. Avoid this by
78149 making them include gdbarch.h.
78151 Change-Id: If433d302007e274daa4f656cfc94f769cf1aa68a
78153 2021-11-18 Simon Marchi <simon.marchi@polymtl.ca>
78155 gdbsupport: make gdb_assert_not_reached accept a format string
78156 Change gdb_assert_not_reached to accept a format string plus
78157 corresponding arguments. This allows giving more precise messages.
78159 Because the format string passed by the caller is prepended with a "%s:"
78160 to add the function name, the callers can no longer pass a translated
78161 string (`_(...)`). Make the gdb_assert_not_reached include the _(),
78162 just like the gdb_assert_fail macro just above.
78164 Change-Id: Id0cfda5a57979df6cdaacaba0d55dd91ae9efee7
78166 2021-11-18 Carl Love <cel@us.ibm.com>
78168 gdb fix for catch-syscall.exp
78169 Remove check_continue "execve" from Proc test_catch_syscall_execve.
78171 The check_continue proceedure checs that the command, execve, starts and
78172 checks for the return from the command. The execve command starts a new
78173 program and thus the return from the command causing the test to fail.
78175 The call to proc check_continue "execve" is removed and replaced with
78176 just the call to check_call_to_syscall "execve" to verify the command
78177 executed. The next test in proc test_catch_syscall_execve verifies that
78178 the new program started and hit the break point in main.
78180 Update the check for the PowerPC architecture. Power Little Endian systems
78181 include "le" in the name. The istarget "power64-*-linux*" check fails to
78182 match LE sytems. The expected string is updated to capture both Big Endian
78183 and Little Endian systems. Power 10 LE istarget prints as:
78184 powerpc64le-unknown-linux-gnu.
78186 This patch fixes three failures and the error:
78188 ERROR: can't read "arch1": no such variable
78190 Patch tested on Power 10 ppc64le GNU/Linux platform.
78192 2021-11-18 Carl Love <cel@us.ibm.com>
78194 gdb: PowerPC fix gdb.base/break-interp.exp
78195 This patch fixes eight test failures on PowerPC for the test
78196 gdb.base/break-interp.exp. The patch adds a funtion and registers it to
78197 setup the displaced stepping for ppc-linux platform. The patch moves the
78198 struct ppc_inferior_data to the ppc-tdep.h include file to make it visible
78199 to the ppc-linux-tdep.c and rs6000-tdep.c files. Additionally the function
78200 get_ppc_per_inferior is made external in ppc-tdep.h to make it visible in
78203 Tested on Power 10 ppc64le-linux with no regressions.
78205 2021-11-18 Carl Love <cel@us.ibm.com>
78207 gdb fix PowerPC test gdb.arch/ppc-longdouble.exp
78208 The test complains of duplicate tests.
78210 DUPLICATE: gdb.arch/ppc-longdouble.exp: continue to breakpoint: return
78212 The do_test calls gdb_continue_to_breakpoint "return". The duplicates
78213 are the result of calling do_test three times with different arguments.
78215 This patch fixes the duplicate tests by adding $name to the
78216 gdb_continue_to_breakpoint argument.
78218 Patch tested on Power 10 ppc64le GNU/Linux, no duplicate tests reported,
78219 no new regression errors.
78221 2021-11-18 H.J. Lu <hjl.tools@gmail.com>
78223 elf/x86: Issue an error on discarded output .plt section
78224 Issue an error, instead of crash, on discarded output .plt section.
78229 * elf32-i386.c (elf_i386_finish_dynamic_sections): Issue an error
78230 on discarded output .plt section.
78231 * elf64-x86-64.c (elf_x86_64_finish_dynamic_sections): Likewise.
78236 * testsuite/ld-elf/pr28597.d: New file.
78237 * testsuite/ld-elf/pr28597.s: Likewise.
78238 * testsuite/ld-elf/pr28597.t: Likewise.
78240 2021-11-18 Tom de Vries <tdevries@suse.de>
78242 [gdb/testsuite] Add missing wait in gdb.base/signals-state-child.exp
78245 (gdb) shell diff -s outputs/gdb.base/signals-state-child/standalone.txt \
78246 outputs/gdb.base/signals-state-child/gdb.txt^M
78247 diff: outputs/gdb.base/signals-state-child/standalone.txt: \
78248 No such file or directory^M
78249 (gdb) FAIL: gdb.base/signals-state-child.exp: signals states are identical
78252 I managed to reproduce this by adding "sleep (5)" at the start of main in
78253 signals-state-child.c.
78255 Fix this by waiting on the result of the spawned command.
78257 Tested on x86_64-linux.
78259 2021-11-18 Alan Modra <amodra@gmail.com>
78261 Re: Don't compile some opcodes files when bfd is 32-bit only
78262 Put bpf back in the 32-bit targets, even though bpf requires a 64-bit
78263 bfd. bpf sim support apparently works without being 64-bit.
78265 * Makefile.am (TARGET64_LIBOPCODES_CFILES): Move bpf files..
78266 (TARGET32_LIBOPCODES_CFILES): ..to here.
78267 * Makefile.in: Regenerate.
78269 2021-11-18 Alan Modra <amodra@gmail.com>
78271 Pass DEBUGINFOD_CFLAGS when compiling dwarf.c
78272 Pick up the elfutils/debuginfod.h install location -I flags from
78273 a variable set by debuginfod.m4 (via pkg.m4 and pkg-config).
78275 * Makefile.am (DEBUGINFOD_CFLAGS): Define.
78276 (dwarf.@OBJECT@): New rule.
78278 2021-11-18 jiawei <jiawei@iscas.ac.cn>
78280 RISC-V: Add testcases for z[fdq]inx
78281 Use gpr when the zfinx enable, the testcases contain float
78282 instructions that reuse by z[fdq]inx.
78286 * testsuite/gas/riscv/zdinx.d: New test.
78287 * testsuite/gas/riscv/zdinx.s: New test.
78288 * testsuite/gas/riscv/zfinx.d: New test.
78289 * testsuite/gas/riscv/zfinx.s: New test.
78290 * testsuite/gas/riscv/zqinx.d: New test.
78291 * testsuite/gas/riscv/zqinx.s: New test.
78293 Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
78295 2021-11-18 jiawei <jiawei@iscas.ac.cn>
78297 RISC-V: Add instructions and operand set for z[fdq]inx
78298 Reuse float instructions in INSN_CLASS_F/D/Q, use riscv_subset_supports to
78299 verify if z*inx enabled and use gpr instead of fpr when z*inx is enable.
78303 * elfxx-riscv.c (riscv_multi_subset_supports): Added support for
78308 * config/tc-riscv.c (riscv_ip): Added register choice for z*inx.
78312 * opcode/riscv.h (enum riscv_insn_class): Reused INSN_CLASS_* for z*inx.
78316 * riscv-dis.c (riscv_disassemble_insn): Added disassemble check for
78318 * riscv-opc.c: Reused INSN_CLASS_* for z*inx.
78320 Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
78322 2021-11-18 jiawei <jiawei@iscas.ac.cn>
78324 RISC-V: Add mininal support for z[fdq]inx
78325 Minimal support for zfinx, zdinx, zqinx. Like f/d/q, the zqinx
78326 imply zdinx and zdinx imply zfinx, where zfinx are not compatible
78331 * elfxx-riscv.c (riscv_implicit_subsets): Added implicit rules
78332 for z*inx extensions.
78333 (riscv_supported_std_z_ext): Added entries for z*inx.
78334 (riscv_parse_check_conflicts): Added conflict check for z*inx.
78336 Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
78338 2021-11-18 GDB Administrator <gdbadmin@sourceware.org>
78340 Automatic date update in version.in
78342 2021-11-17 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
78344 aarch64: [SME] SVE2 instructions added to support SME
78345 This patch is adding new SVE2 instructions added to support SME extension.
78346 The following SVE2 instructions are added by the SME architecture:
78348 * REVD, SCLAMP and UCLAMP.
78352 * config/tc-aarch64.c (parse_sme_pred_reg_with_index):
78354 (parse_operands): New parser.
78355 * testsuite/gas/aarch64/sme-9-illegal.d: New test.
78356 * testsuite/gas/aarch64/sme-9-illegal.l: New test.
78357 * testsuite/gas/aarch64/sme-9-illegal.s: New test.
78358 * testsuite/gas/aarch64/sme-9.d: New test.
78359 * testsuite/gas/aarch64/sme-9.s: New test.
78363 * opcode/aarch64.h (enum aarch64_opnd): New operand
78364 AARCH64_OPND_SME_PnT_Wm_imm.
78368 * aarch64-asm.c (aarch64_ins_sme_pred_reg_with_index):
78370 * aarch64-dis.c (aarch64_ext_sme_pred_reg_with_index):
78372 * aarch64-opc.c (aarch64_print_operand): Printout of
78373 OPND_SME_PnT_Wm_imm.
78374 * aarch64-opc.h (enum aarch64_field_kind): New bitfields
78375 FLD_SME_Rm, FLD_SME_i1, FLD_SME_tszh, FLD_SME_tszl.
78376 * aarch64-tbl.h (OP_SVE_NN_BHSD): New qualifier.
78377 (OP_SVE_QMQ): New qualifier.
78378 (struct aarch64_opcode): New instructions PSEL, REVD,
78380 aarch64-asm-2.c: Regenerate.
78381 aarch64-dis-2.c: Regenerate.
78382 aarch64-opc-2.c: Regenerate.
78384 2021-11-17 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
78386 aarch64: [SME] Add new SME system registers
78387 This patch is adding miscellaneous SME related system registers.
78391 * testsuite/gas/aarch64/sme-sysreg.d: New test.
78392 * testsuite/gas/aarch64/sme-sysreg.s: New test.
78393 * testsuite/gas/aarch64/sme-sysreg-illegal.d: New test.
78394 * testsuite/gas/aarch64/sme-sysreg-illegal.l: New test.
78395 * testsuite/gas/aarch64/sme-sysreg-illegal.s: New test.
78399 * aarch64-opc.c: New system registers id_aa64smfr0_el1,
78400 smcr_el1, smcr_el12, smcr_el2, smcr_el3, smpri_el1,
78401 smprimap_el2, smidr_el1, tpidr2_el0 and mpamsm_el1.
78403 2021-11-17 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
78405 aarch64: [SME] Add SME mode selection and state access instructions
78406 This patch is adding new SME mode selection and state access instructions:
78407 * Add SMSTART and SMSTOP instructions.
78408 * Add SVCR system register.
78412 * config/tc-aarch64.c (parse_sme_sm_za): New parser.
78413 (parse_operands): New parser.
78414 * testsuite/gas/aarch64/sme-8-illegal.d: New test.
78415 * testsuite/gas/aarch64/sme-8-illegal.l: New test.
78416 * testsuite/gas/aarch64/sme-8-illegal.s: New test.
78417 * testsuite/gas/aarch64/sme-8.d: New test.
78418 * testsuite/gas/aarch64/sme-8.s: New test.
78422 * opcode/aarch64.h (enum aarch64_opnd): New operand
78423 AARCH64_OPND_SME_SM_ZA.
78424 (enum aarch64_insn_class): New instruction classes
78425 sme_start and sme_stop.
78429 * aarch64-asm.c (aarch64_ins_pstatefield): New inserter.
78430 (aarch64_ins_sme_sm_za): New inserter.
78431 * aarch64-dis.c (aarch64_ext_imm): New extractor.
78432 (aarch64_ext_pstatefield): New extractor.
78433 (aarch64_ext_sme_sm_za): New extractor.
78434 * aarch64-opc.c (operand_general_constraint_met_p):
78435 New pstatefield value for SME instructions.
78436 (aarch64_print_operand): Printout for OPND_SME_SM_ZA.
78437 (SR_SME): New register SVCR.
78438 * aarch64-opc.h (F_REG_IN_CRM): New register endcoding.
78439 * aarch64-opc.h (F_IMM_IN_CRM): New immediate endcoding.
78440 (PSTATE_ENCODE_CRM): Encode CRm field.
78441 (PSTATE_DECODE_CRM): Decode CRm field.
78442 (PSTATE_ENCODE_CRM_IMM): Encode CRm immediate field.
78443 (PSTATE_DECODE_CRM_IMM): Decode CRm immediate field.
78444 (PSTATE_ENCODE_CRM_AND_IMM): Encode CRm and immediate
78446 * aarch64-tbl.h (struct aarch64_opcode): New SMSTART
78447 and SMSTOP instructions.
78448 aarch64-asm-2.c: Regenerate.
78449 aarch64-dis-2.c: Regenerate.
78450 aarch64-opc-2.c: Regenerate.
78452 2021-11-17 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
78454 aarch64: [SME] Add LD1x, ST1x, LDR and STR instructions
78455 This patch is adding new loads and stores defined by SME instructions.
78459 * config/tc-aarch64.c (parse_sme_address): New parser.
78460 (parse_sme_za_hv_tiles_operand_with_braces): New parser.
78461 (parse_sme_za_array): New parser.
78462 (output_operand_error_record): Print error details if
78464 (parse_operands): Support new operands.
78465 * testsuite/gas/aarch64/sme-5-illegal.d: New test.
78466 * testsuite/gas/aarch64/sme-5-illegal.l: New test.
78467 * testsuite/gas/aarch64/sme-5-illegal.s: New test.
78468 * testsuite/gas/aarch64/sme-5.d: New test.
78469 * testsuite/gas/aarch64/sme-5.s: New test.
78470 * testsuite/gas/aarch64/sme-6-illegal.d: New test.
78471 * testsuite/gas/aarch64/sme-6-illegal.l: New test.
78472 * testsuite/gas/aarch64/sme-6-illegal.s: New test.
78473 * testsuite/gas/aarch64/sme-6.d: New test.
78474 * testsuite/gas/aarch64/sme-6.s: New test.
78475 * testsuite/gas/aarch64/sme-7-illegal.d: New test.
78476 * testsuite/gas/aarch64/sme-7-illegal.l: New test.
78477 * testsuite/gas/aarch64/sme-7-illegal.s: New test.
78478 * testsuite/gas/aarch64/sme-7.d: New test.
78479 * testsuite/gas/aarch64/sme-7.s: New test.
78483 * opcode/aarch64.h (enum aarch64_opnd): New operands.
78484 (enum aarch64_insn_class): Added sme_ldr and sme_str.
78485 (AARCH64_OPDE_UNTIED_IMMS): New operand error kind.
78489 * aarch64-asm.c (aarch64_ins_sme_za_hv_tiles): New inserter.
78490 (aarch64_ins_sme_za_list): New inserter.
78491 (aarch64_ins_sme_za_array): New inserter.
78492 (aarch64_ins_sme_addr_ri_u4xvl): New inserter.
78493 * aarch64-asm.h (AARCH64_DECL_OPD_INSERTER): Added
78494 ins_sme_za_list, ins_sme_za_array and ins_sme_addr_ri_u4xvl.
78495 * aarch64-dis.c (aarch64_ext_sme_za_hv_tiles): New extractor.
78496 (aarch64_ext_sme_za_list): New extractor.
78497 (aarch64_ext_sme_za_array): New extractor.
78498 (aarch64_ext_sme_addr_ri_u4xvl): New extractor.
78499 * aarch64-dis.h (AARCH64_DECL_OPD_EXTRACTOR): Added
78500 ext_sme_za_list, ext_sme_za_array and ext_sme_addr_ri_u4xvl.
78501 * aarch64-opc.c (operand_general_constraint_met_p):
78502 (aarch64_match_operands_constraint): Handle sme_ldr, sme_str
78504 (aarch64_print_operand): New operands supported.
78505 * aarch64-tbl.h (OP_SVE_QUU): New qualifier.
78506 (OP_SVE_QZU): New qualifier.
78507 aarch64-asm-2.c: Regenerate.
78508 aarch64-dis-2.c: Regenerate.
78509 aarch64-opc-2.c: Regenerate.
78511 2021-11-17 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
78513 aarch64: [SME] Add ZERO instruction
78514 This patch is adding ZERO (a list of 64-bit element ZA tiles)
78519 * config/tc-aarch64.c (parse_sme_list_of_64bit_tiles):
78521 (parse_operands): Handle OPND_SME_list_of_64bit_tiles.
78522 * testsuite/gas/aarch64/sme-4-illegal.d: New test.
78523 * testsuite/gas/aarch64/sme-4-illegal.l: New test.
78524 * testsuite/gas/aarch64/sme-4-illegal.s: New test.
78525 * testsuite/gas/aarch64/sme-4.d: New test.
78526 * testsuite/gas/aarch64/sme-4.s: New test.
78530 * opcode/aarch64.h (enum aarch64_opnd): New operand
78531 AARCH64_OPND_SME_list_of_64bit_tiles.
78535 * aarch64-opc.c (print_sme_za_list): New printing function.
78536 (aarch64_print_operand): Handle OPND_SME_list_of_64bit_tiles.
78537 * aarch64-opc.h (enum aarch64_field_kind): New bitfield
78539 * aarch64-tbl.h (struct aarch64_opcode): New ZERO instruction.
78540 aarch64-asm-2.c: Regenerate.
78541 aarch64-dis-2.c: Regenerate.
78542 aarch64-opc-2.c: Regenerate.
78544 2021-11-17 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
78546 aarch64: [SME] Add MOV and MOVA instructions
78547 This patch is adding new MOV (alias) and MOVA SME instruction.
78551 * config/tc-aarch64.c (enum sme_hv_slice): new enum.
78552 (struct reloc_entry): Added ZAH and ZAV registers.
78553 (parse_sme_immediate): Immediate parser.
78554 (parse_sme_za_hv_tiles_operand): ZA tile parser.
78555 (parse_sme_za_hv_tiles_operand_index): Index parser.
78556 (parse_operands): Added ZA tile parser calls.
78557 (REGNUMS): New macro. Regs with suffix.
78558 (REGSET16S): New macro. 16 regs with suffix.
78559 * testsuite/gas/aarch64/sme-2-illegal.d: New test.
78560 * testsuite/gas/aarch64/sme-2-illegal.l: New test.
78561 * testsuite/gas/aarch64/sme-2-illegal.s: New test.
78562 * testsuite/gas/aarch64/sme-2.d: New test.
78563 * testsuite/gas/aarch64/sme-2.s: New test.
78564 * testsuite/gas/aarch64/sme-2a.d: New test.
78565 * testsuite/gas/aarch64/sme-2a.s: New test.
78566 * testsuite/gas/aarch64/sme-3-illegal.d: New test.
78567 * testsuite/gas/aarch64/sme-3-illegal.l: New test.
78568 * testsuite/gas/aarch64/sme-3-illegal.s: New test.
78569 * testsuite/gas/aarch64/sme-3.d: New test.
78570 * testsuite/gas/aarch64/sme-3.s: New test.
78571 * testsuite/gas/aarch64/sme-3a.d: New test.
78572 * testsuite/gas/aarch64/sme-3a.s: New test.
78576 * opcode/aarch64.h (enum aarch64_opnd): New enums
78577 AARCH64_OPND_SME_ZA_HV_idx_src and
78578 AARCH64_OPND_SME_ZA_HV_idx_dest.
78579 (struct aarch64_opnd_info): New ZA tile vector struct.
78583 * aarch64-asm.c (aarch64_ins_sme_za_hv_tiles):
78585 * aarch64-asm.h (AARCH64_DECL_OPD_INSERTER):
78586 New inserter ins_sme_za_hv_tiles.
78587 * aarch64-dis.c (aarch64_ext_sme_za_hv_tiles):
78589 * aarch64-dis.h (AARCH64_DECL_OPD_EXTRACTOR):
78590 New extractor ext_sme_za_hv_tiles.
78591 * aarch64-opc.c (aarch64_print_operand):
78592 Handle SME_ZA_HV_idx_src and SME_ZA_HV_idx_dest.
78593 * aarch64-opc.h (enum aarch64_field_kind): New enums
78594 FLD_SME_size_10, FLD_SME_Q, FLD_SME_V and FLD_SME_Rv.
78595 (struct aarch64_operand): Increase fields size to 5.
78596 * aarch64-tbl.h (OP_SME_BHSDQ_PM_BHSDQ): New qualifiers
78597 aarch64-asm-2.c: Regenerate.
78598 aarch64-dis-2.c: Regenerate.
78599 aarch64-opc-2.c: Regenerate.
78601 2021-11-17 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
78603 aarch64: [SME] Add SME instructions
78604 Patch is adding new SME matrix instructions. Please note additional
78605 instructions will be added in following patches.
78609 * config/tc-aarch64.c (parse_sme_zada_operand):
78611 * config/tc-aarch64.c (parse_reg_with_qual):
78613 * config/tc-aarch64.c (R_ZA): New egister type.
78614 (parse_operands): New parser.
78615 * testsuite/gas/aarch64/sme-illegal.d: New test.
78616 * testsuite/gas/aarch64/sme-illegal.l: New test.
78617 * testsuite/gas/aarch64/sme-illegal.s: New test.
78618 * testsuite/gas/aarch64/sme.d: New test.
78619 * testsuite/gas/aarch64/sme.s: New test.
78620 * testsuite/gas/aarch64/sme-f64.d: New test.
78621 * testsuite/gas/aarch64/sme-f64.s: New test.
78622 * testsuite/gas/aarch64/sme-i64.d: New test.
78623 * testsuite/gas/aarch64/sme-i64.s: New test.
78627 * opcode/aarch64.h (enum aarch64_opnd): New operands
78628 AARCH64_OPND_SME_ZAda_2b, AARCH64_OPND_SME_ZAda_3b and
78629 AARCH64_OPND_SME_Pm.
78630 (enum aarch64_insn_class): New instruction class sme_misc.
78634 * aarch64-opc.c (aarch64_print_operand):
78635 Print OPND_SME_ZAda_2b and OPND_SME_ZAda_3b operands.
78636 (verify_constraints): Handle OPND_SME_Pm.
78637 * aarch64-opc.h (enum aarch64_field_kind):
78638 New bit fields FLD_SME_ZAda_2b, FLD_SME_ZAda_3b and FLD_SME_Pm.
78639 * aarch64-tbl.h (OP_SME_ZADA_PN_PM_ZN_S): New qualifier set.
78640 (OP_SME_ZADA_PN_PM_ZN_D): New qualifier.
78641 (OP_SME_ZADA_PN_PM_ZN_ZM): New qualifier.
78642 (OP_SME_ZADA_S_PM_PM_S_S): New qualifier.
78643 (OP_SME_ZADA_D_PM_PM_D_D): New qualifier.
78644 (OP_SME_ZADA_S_PM_PM_H_H): New qualifier.
78645 (OP_SME_ZADA_S_PM_PM_B_B): New qualifier.
78646 (OP_SME_ZADA_D_PM_PM_H_H): New qualifier.
78647 (SME_INSN): New instruction macro.
78648 (SME_F64_INSN): New instruction macro.
78649 (SME_I64_INSN): New instruction macro.
78650 (SME_INSNC): New instruction macro.
78651 (struct aarch64_opcode): New SME instructions.
78652 aarch64-asm-2.c: Regenerate.
78653 aarch64-dis-2.c: Regenerate.
78654 aarch64-opc-2.c: Regenerate.
78656 2021-11-17 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
78658 aarch64: [SME] Add +sme option to -march
78659 This series of patches (tagged [SME]) add support for the Scalable
78660 Matrix Extension. Patch introduces new command line options: +sme, +sme-f64 and
78661 +sme-i64 to -march command line options.
78665 * NEWS: Updated docs.
78666 * config/tc-aarch64.c: New SME command line options.
78667 * doc/c-aarch64.texi: Update docs.
78671 * opcode/aarch64.h (AARCH64_FEATURE_SME): New flag.
78672 (AARCH64_FEATURE_SME_F64): New flag.
78673 (AARCH64_FEATURE_SME_I64): New flag.
78677 * aarch64-tbl.h (SME): New feature object.
78679 2021-11-17 Jeremy Drake <cygwin@jdrake.com>
78681 Set the default DLL chracteristics to 0 for Cygwin based targets.
78682 * emultempl/pep.em (DEFAULT_DLL_CHARACTERISTICS): Set to 0 for
78684 * emultempl/pep.em (DEFAULT_DLL_CHARACTERISTICS): Likewise.
78686 2021-11-17 Nick Clifton <nickc@redhat.com>
78688 Fix the linker script parser so that it will recognise the PT_GNU_RELRO segment type, and the linker itself so that it will gracefully handle being unable to assign any sections to such a segment.
78690 bfd * elf.c (assign_file_positions_for_non_load_sections): Replace
78691 assertion with a warning message.
78693 ld * ldgram.y: Add support for PT_GNU_RELRO and PT_GNU_PROPERTY.
78694 * ldgram.c: Regenerate.
78696 2021-11-17 Andreas Arnez <arnez@linux.ibm.com>
78698 [gdb/build, s390x] Fix build after gdbarch_tdep changes
78699 Commit 345bd07cce33 ("gdb: fix gdbarch_tdep ODR violation") changes a
78700 declaration in s390-tdep.h from
78702 struct gdbarch_tdep { ... };
78706 struct s390_gdbarch_tdep : gdbarch_tdep { ... };
78708 and now requires that gdbarch_tdep has been declared before. Which is
78709 usually the case, except when compiling s390-linux-nat.c, where
78710 s390-tdep.h is included before gdbarch.h. Thus the s390x build errors out
78711 with the compiler complaining about a missing class name after the colon.
78713 Fix this in s390-linux-nat.c, by including gdbarch.h before s390-tdep.h.
78715 2021-11-17 Luis Machado <luis.machado@linaro.org>
78717 Expose the BTI BTYPE more explicitly in the registers
78718 Augment the register description XML to expose the BTI BTYPE field contained
78719 in the CPSR register. It will be displayed like so:
78721 cpsr 0x60001000 [ EL=0 BTYPE=0 SSBS C Z ]
78723 2021-11-17 H.J. Lu <hjl.tools@gmail.com>
78725 elfedit: Add --output-abiversion option to update ABIVERSION
78726 * NEWS: Mention --output-abiversion.
78727 * elfedit.c (input_elf_abiversion): New.
78728 (output_elf_abiversion): Likewise.
78729 (update_elf_header): Update EI_ABIVERSION.
78730 (command_line_switch): Add OPTION_INPUT_ABIVERSION and
78731 OPTION_OUTPUT_ABIVERSION.
78732 (options): Add --input-abiversion and --output-abiversion.
78734 (main): Handle --input-abiversion and --output-abiversion.
78735 * doc/binutils.texi: Document --input-abiversion and
78736 --output-abiversion.
78737 * testsuite/binutils-all/elfedit.exp: Run elfedit-6.
78738 * testsuite/binutils-all/elfedit-6.d: New file.
78740 2021-11-17 Nelson Chu <nelson.chu@sifive.com>
78742 RISC-V: Support rvv extension with released version 1.0.
78743 2021-11-17 Jim Wilson <jimw@sifive.com>
78744 Kito Cheng <kito.cheng@sifive.com>
78745 Nelson Chu <nelson.chu@sifive.com>
78747 This patch is porting from the following riscv github,
78748 https://github.com/riscv/riscv-binutils-gdb/tree/rvv-1.0.x
78750 And here is the vector spec,
78751 https://github.com/riscv/riscv-v-spec
78754 * elfxx-riscv.c (riscv_implicit_subsets): Added imply rules
78755 of v, zve and zvl extensions.
78756 (riscv_supported_std_ext): Updated verison of v to 1.0.
78757 (riscv_supported_std_z_ext): Added zve and zvl extensions.
78758 (riscv_parse_check_conflicts): The zvl extensions need to
78759 enable either v or zve extension.
78760 (riscv_multi_subset_supports): Check the subset list to know
78761 if the INSN_CLASS_V and INSN_CLASS_ZVEF instructions are supported.
78763 * config/tc-riscv.c (enum riscv_csr_class): Added CSR_CLASS_V.
78764 (enum reg_class): Added RCLASS_VECR and RCLASS_VECM.
78765 (validate_riscv_insn): Check whether the rvv operands are valid.
78766 (md_begin): Initialize register hash for rvv registers.
78767 (macro_build): Added rvv operands when expanding rvv pseudoes.
78768 (vector_macro): Expand rvv macros into one or more instructions.
78770 (my_getVsetvliExpression): Similar to my_getVsetvliExpression,
78771 but used for parsing vsetvli operands.
78772 (riscv_ip): Parse and encode rvv operands. Besides, The rvv loads
78773 and stores with EEW 64 cannot be used when zve32x is enabled.
78774 * testsuite/gas/riscv/priv-reg-fail-version-1p10.d: Updated -march
78776 * testsuite/gas/riscv/priv-reg-fail-version-1p11.d: Likewise.
78777 * testsuite/gas/riscv/priv-reg-fail-version-1p9p1.d: Likewise.
78778 * testsuite/gas/riscv/priv-reg.s: Added rvv csr testcases.
78779 * testsuite/gas/riscv/priv-reg-version-1p10.d: Likewise.
78780 * testsuite/gas/riscv/priv-reg-version-1p11.d: Likewise.
78781 * testsuite/gas/riscv/priv-reg-version-1p9p1.d: Likewise.
78782 * testsuite/gas/riscv/march-imply-v.d: New testcase.
78783 * testsuite/gas/riscv/vector-insns-fail-zve32xf.d: Likewise.
78784 * testsuite/gas/riscv/vector-insns-fail-zve32xf.l: Likewise.
78785 * testsuite/gas/riscv/vector-insns-fail-zvl.d: Likewise.
78786 * testsuite/gas/riscv/vector-insns-fail-zvl.l: Likewise.
78787 * testsuite/gas/riscv/vector-insns-vmsgtvx.d: Likewise.
78788 * testsuite/gas/riscv/vector-insns-vmsgtvx.s: Likewise.
78789 * testsuite/gas/riscv/vector-insns-zero-imm.d: Likewise.
78790 * testsuite/gas/riscv/vector-insns-zero-imm.s: Likewise.
78791 * testsuite/gas/riscv/vector-insns.d: Likewise.
78792 * testsuite/gas/riscv/vector-insns.s: Likewise.
78794 * opcode/riscv-opc.h: Defined mask/match encodings and csrs for rvv.
78795 * opcode/riscv.h: Defined rvv immediate encodings and fields.
78796 (enum riscv_insn_class): Added INSN_CLASS_V and INSN_CLASS_ZVEF.
78797 (INSN_V_EEW64): Defined.
78798 (M_VMSGE, M_VMSGEU): Added for the rvv pseudoes.
78800 * riscv-dis.c (print_insn_args): Dump the rvv operands.
78801 * riscv-opc.c (riscv_vecr_names_numeric): Defined rvv registers.
78802 (riscv_vecm_names_numeric): Likewise.
78803 (riscv_vsew): Likewise.
78804 (riscv_vlmul): Likewise.
78805 (riscv_vta): Likewise.
78806 (riscv_vma): Likewise.
78807 (match_vs1_eq_vs2): Added for rvv Vu operand.
78808 (match_vd_eq_vs1_eq_vs2): Added for rvv Vv operand.
78809 (riscv_opcodes): Added rvv v1.0 instructions.
78811 2021-11-17 Sergei Trofimovich <siarheit@google.com>
78813 gdb/nat/linux-osdata.c: fix build on gcc-12 (string overfow)
78814 On gcc-12 build fails as:
78816 ../../gdbserver/../gdb/nat/linux-osdata.c: In function 'void linux_xfer_osdata_processes(buffer*)':
78817 ../../gdbserver/../gdb/nat/linux-osdata.c:330:39: error:
78818 '__builtin___sprintf_chk' may write a terminating nul past the end of the destination [-Werror=format-overflow=]
78819 330 | sprintf (core_str, "%d", i);
78822 It's an off-by-one case in an infeasible scenario for negative
78823 huge core count. The change switches to std::string for memory
78826 Tested by running 'info os processes' and checking CPU cores column.
78828 2021-11-17 Aaron Merey <amerey@redhat.com>
78830 gdb: Add aliases for read_core_file_mappings callbacks
78831 Add aliases read_core_file_mappings_loop_ftype and
78832 read_core_file_mappings_pre_loop_ftype. Intended for use with
78833 read_core_file_mappings.
78835 Also add build_id parameter to read_core_file_mappings_loop_ftype.
78837 2021-11-17 Mike Frysinger <vapier@gentoo.org>
78839 sim: testsuite: add support for $pwd replacements
78840 Extend the common test framework to support $pwd replacements in
78841 settings. This allows replacing the custom cris @exedir@ with it.
78843 sim: cris: replace @srcdir@ test extension with $srcdir/$subdir
78844 The common framework supports $srcdir & $subdir replacements already,
78845 so replace the custom @srcdir@ logic with those. Since the replace
78846 happens in slurp_options that cris already uses, we don't have any
78847 logic to port over there. We have to duplicate that into the cris
78848 slurp_rv helper though.
78850 2021-11-17 Mike Frysinger <vapier@gentoo.org>
78852 sim: cris: drop custom "dynamic" test field
78853 This tag is used to force tests to be built dynamically (i.e. without
78854 -static linking). This is because cris-sim.exp in dejagnu turns on
78855 static linking in ldflags.
78857 The default configs and runtest flags shouldn't load these boards.
78858 If these settings are still needed, we should figure out a different
78859 way of suppressing the stock settings wholesale. We want these to
78860 all pass out of the box with little to no configuration so that they
78861 can run in a multitarget build.
78863 With dropping "dynamic", it'll be easier to merge the custom cris
78864 test logic with the common sim test logic.
78866 2021-11-17 Mike Frysinger <vapier@gentoo.org>
78868 sim: testsuite: add more silent build rules
78869 site.exp is still verbose, but that comes from automake, so have
78870 to get it fixed upstream.
78872 2021-11-17 GDB Administrator <gdbadmin@sourceware.org>
78874 Automatic date update in version.in
78876 2021-11-16 Sergei Trofimovich <siarheit@google.com>
78878 sim: cr16: fix build on gcc-12 (NULL comparison)
78879 On gcc-12 build fails as:
78881 sim/cr16/interp.c: In function 'lookup_hash':
78882 sim/cr16/interp.c:89:25: error:
78883 the comparison will always evaluate as 'true'
78884 for the address of 'mnimonic' will never be NULL [-Werror=address]
78885 89 | if ((h->ops->mnimonic != NULL) &&
78888 'mnimonic' is a sharr array within ops. It can never be NULL.
78890 While at it renamed 'mnimonic' to 'mnemonic'.
78892 2021-11-16 Simon Marchi <simon.marchi@polymtl.ca>
78894 gdb: fix length of array view returned by some value_contents functions
78895 In commit 50888e42dcd3 ("gdb: change functions returning value contents
78896 to use gdb::array_view"), I believe I made a mistake with the length of
78897 the array views returned by some functions. All functions return a view
78898 of `TYPE_LENGTH (value_type (type))` length. This is not correct when
78899 the value's enclosing type is larger than the value's type. In that
78900 case, the value's contents buffer is of the size of the enclosing type,
78901 and the value's actual contents is a slice of that (as returned by
78902 value_contents). So, functions value_contents_all_raw,
78903 value_contents_for_printing and value_contents_for_printing_const are
78904 not correct. Since they are meant to return the value's contents buffer
78905 as a whole, they should have the size of the enclosing type.
78907 There is nothing that uses the returned array view size at the moment,
78908 so this didn't cause a problem. But it became apparent when trying to
78909 adjust some callers.
78911 Change-Id: Ib4e8837e1069111d2b2784d3253d5f3002419e68
78913 2021-11-16 Fangrui Song <maskray@google.com>
78915 readelf: Support SHT_RELR/DT_RELR for -r
78916 The -r output for SHT_RELR looks like:
78918 Relocation section '.relr.dyn' at offset 0x530 contains 4 entries:
78928 For --use-dynamic, the header looks like
78930 'RELR' relocation section at offset 0x530 contains 32 bytes:
78933 * elf/common.h (DT_ENCODING): Bump to 38.
78934 * elf/external.h (Elf32_External_Relr): New.
78935 (Elf64_External_Relr): New.
78937 * readelf.c (enum relocation_type): New.
78938 (slurp_relr_relocs): New.
78939 (dump_relocations): Change is_rela to rel_type.
78941 (dynamic_relocations): Add DT_RELR.
78942 (process_relocs): Check SHT_RELR and DT_RELR.
78943 (process_dynamic_section): Store into dynamic_info for
78944 DT_RELR/DT_RELRENT/DT_RELRSZ.
78946 2021-11-16 Simon Marchi <simon.marchi@polymtl.ca>
78948 gdbsupport: remove FUNCTION_NAME
78949 __func__ is standard C++11:
78951 https://en.cppreference.com/w/cpp/language/function
78953 Also, in C++11, __func__ expands to the demangled function name, so the
78954 mention in the comment above FUNCTION_NAME doesn't apply anymore.
78955 Finally, in places where FUNCTION_NAME is used, I think it's enough to
78956 print the function name, no need to print the whole signature.
78957 Therefore, I propose to just remove FUNCTION_NAME and update users to
78958 use the standard __func__.
78960 Change-Id: I778f28155422b044402442dc18d42d0cded1017d
78962 2021-11-16 Andrew Burgess <aburgess@redhat.com>
78964 gdb/gdbsupport: make xstrprintf and xstrvprintf return a unique_ptr
78965 The motivation is to reduce the number of places where unmanaged
78966 pointers are returned from allocation type routines. All of the
78967 callers are updated.
78969 There should be no user visible changes after this commit.
78971 2021-11-16 Andrew Burgess <aburgess@redhat.com>
78973 gdbsupport: move xfree into its own file
78974 In the next commit I'd like to reference gdb_unique_ptr within the
78975 common-utils.h file. However, this requires that I include
78976 gdb_unique_ptr.h, which requires that xfree be defined.
78978 Interestingly, gdb_unique_ptr.h doesn't actually include anything that
78979 defines xfree, but I was finding that when I added a gdb_unique_ptr.h
78980 include to common-utils.h I was getting a dependency cycle; before my
78981 change xfree was defined when gdb_unique_ptr.h was processed, while
78982 after my change it was not, and this made g++ unhappy.
78984 To break this cycle, I propose to move xfree into its own header file,
78985 gdb-xfree.h, which I'll then include into gdb_unique_ptr.h and
78988 2021-11-16 Andrew Burgess <aburgess@redhat.com>
78990 gdb: throw OPTIMIZED_OUT_ERROR rather than GENERIC_ERROR
78991 While reviewing this patch:
78993 https://sourceware.org/pipermail/gdb-patches/2021-November/183227.html
78995 I spotted that the patch could be improved if we threw
78996 OPTIMIZED_OUT_ERROR rather than GENERIC_ERROR in a few places.
78998 This commit updates error_value_optimized_out and
78999 require_not_optimized_out to throw OPTIMIZED_OUT_ERROR.
79001 I ran the testsuite and saw no regressions. This doesn't really
79002 surprise me, we don't usually write code like:
79004 catch (const gdb_exception_error &ex)
79006 (if ex.error == GENERIC_ERROR)
79012 There are a three places where we write something like:
79014 catch (const gdb_exception_error &ex)
79016 (if ex.error == OPTIMIZED_OUT_ERROR)
79020 In frame.c:unwind_pc, stack.c:info_frame_command_core, and
79021 value.c:value_optimized_out, but if we are hitting these cases then
79022 it's not significantly changing GDB's behaviour.
79024 2021-11-16 Tom Tromey <tromey@adacore.com>
79026 Remove config.cache in gdbserver's "distclean"
79027 PR gdb/28586 points out that "make distclean" fails to delete
79028 config.cache from gdbserver/. This patch fixes the bug, and removes a
79029 duplicate "Makefile" deletion that was also pointed out in the PR.
79031 2021-11-16 Tom de Vries <tdevries@suse.de>
79033 [gdb/testsuite] Remove inferior output in gdb.base/foll-vfork.exp
79034 Test-case gdb.base/foll-vfork.exp has inferior output that is not needed, but
79035 which makes the regexp matching more difficult (see commit 1f28b70def1
79036 "[gdb/testsuite] Fix regexp in gdb.base/foll-vfork.exp").
79038 Remove the inferior output, and revert commit 1f28b70def1 to make the matching
79041 Tested on x86_64-linux.
79043 2021-11-16 H.J. Lu <hjl.tools@gmail.com>
79045 x86: Don't allow KMOV in TLS code sequences
79046 Don't allow KMOV in TLS code sequences which require integer MOV
79050 * config/tc-i386.c (match_template): Don't allow KMOV in TLS
79052 * testsuite/gas/i386/i386.exp: Run inval-tls and x86-64-inval-tls
79054 * testsuite/gas/i386/inval-tls.l: New file.
79055 * testsuite/gas/i386/inval-tls.s: Likewise.
79056 * testsuite/gas/i386/x86-64-inval-tls.l: Likewise.
79057 * testsuite/gas/i386/x86-64-inval-tls.s: Likewise.
79059 2021-11-16 Mike Frysinger <vapier@gentoo.org>
79061 sim: run: support concise env var settings
79062 Support the same syntax as other common utilities where env vars can
79063 be specified before the program to be run without an explicit option.
79065 This behavior can be suppressed by using the -- marker.
79067 2021-11-16 Mike Frysinger <vapier@gentoo.org>
79069 sim: nrun: add --env-{set,unset,clear} command line options
79070 Provide explicit control over the program's environment with the
79071 basic set/unset/clear options. These are a bit clunky to use,
79072 but they're functional.
79074 The env set operation is split out into a separate function as it'll
79075 be used in the next commit.
79077 With these in place, we can adjust the custom cris testsuite to use
79078 the now standard options and not its one-off hack.
79080 2021-11-16 Mike Frysinger <vapier@gentoo.org>
79082 sim: syscall: hoist argc/argn/argnlen to common code
79083 Now that the callback framework supports argv & envp, we can move
79084 the Blackfin implementation of these syscalls to the common code.
79086 sim: syscall: fix argvlen & argv implementation
79087 Now that we have access to the argv & envp strings, finish implementing
79088 these syscalls. Delete unused variables, fix tbuf by incrementing the
79089 pointer instead of setting to the length, and make sure we don't write
79090 more data than the bufsize says is available.
79092 sim: callback: expose argv & environ
79093 Pass the existing strings data to the callbacks so that common
79094 libgloss syscalls can be implemented (which we'll do shortly).
79096 2021-11-16 Mike Frysinger <vapier@gentoo.org>
79098 sim: keep track of program environment strings
79099 We've been passing the environment strings to sim_create_inferior,
79100 but most ports don't do anything with them. A few will use ad-hoc
79101 logic to stuff the stack for user-mode programs, but that's it.
79103 Let's formalize this across the board by storing the strings in the
79104 normal sim state. This will allow (in future commits) supporting
79105 more functionality in the run interface, and to unify some of the
79108 2021-11-16 Mike Frysinger <vapier@gentoo.org>
79110 sim: iq2000: fix some missing prototypes warnings
79111 Turns out some of these were hiding real bugs like not passing the
79114 2021-11-16 jiawei <jiawei@iscas.ac.cn>
79116 RISC-V: Scalar crypto instruction and entropy source CSR testcases.
79117 Add testcases for Scalar Crypto extension, with total testcase contain all
79118 instructions in k-ext/k-ext-64 and sub-extension testcase for zbk* zk*. Also
79119 add testcase for new CSR name 'seed' which is the Entropy Source in zkr.
79121 In fact these whole testcases can be combined into one file, after we have
79122 supported the .option arch +-= directives.
79125 * testsuite/gas/riscv/k-ext-64.d: New testcase for crypto instructions.
79126 * testsuite/gas/riscv/k-ext-64.s: Likewise.
79127 * testsuite/gas/riscv/k-ext.d: Likewise.
79128 * testsuite/gas/riscv/k-ext.s: Likewise.
79129 * testsuite/gas/riscv/zbkb-32.d: Likewise.
79130 * testsuite/gas/riscv/zbkb-32.s: Likewise.
79131 * testsuite/gas/riscv/zbkb-64.d: Likewise.
79132 * testsuite/gas/riscv/zbkb-64.s: Likewise.
79133 * testsuite/gas/riscv/zbkc-32.d: Likewise.
79134 * testsuite/gas/riscv/zbkc-64.d: Likewise.
79135 * testsuite/gas/riscv/zbkc.s: Likewise.
79136 * testsuite/gas/riscv/zbkx-32.d: Likewise.
79137 * testsuite/gas/riscv/zbkx-64.d: Likewise.
79138 * testsuite/gas/riscv/zbkx.s: Likewise.
79139 * testsuite/gas/riscv/zknd-32.d: Likewise.
79140 * testsuite/gas/riscv/zknd-32.s: Likewise.
79141 * testsuite/gas/riscv/zknd-64.d: Likewise.
79142 * testsuite/gas/riscv/zknd-64.s: Likewise.
79143 * testsuite/gas/riscv/zkne-32.d: Likewise.
79144 * testsuite/gas/riscv/zkne-32.s: Likewise.
79145 * testsuite/gas/riscv/zkne-64.d: Likewise.
79146 * testsuite/gas/riscv/zkne-64.s: Likewise.
79147 * testsuite/gas/riscv/zknh-32.d: Likewise.
79148 * testsuite/gas/riscv/zknh-32.s: Likewise.
79149 * testsuite/gas/riscv/zknh-64.d: Likewise.
79150 * testsuite/gas/riscv/zknh-64.s: Likewise.
79151 * testsuite/gas/riscv/zksed-32.d: Likewise.
79152 * testsuite/gas/riscv/zksed-64.d: Likewise.
79153 * testsuite/gas/riscv/zksed.s: Likewise.
79154 * testsuite/gas/riscv/zksh-32.d: Likewise.
79155 * testsuite/gas/riscv/zksh-64.d: Likewise.
79156 * testsuite/gas/riscv/zksh.s: Likewise.
79157 * testsuite/gas/riscv/priv-reg-fail-zkr.d: New testcase for zkr
79159 * testsuite/gas/riscv/priv-reg-fail-zkr.l: Likewise.
79160 * testsuite/gas/riscv/priv-reg-fail-version-1p10.d: Updated march to
79162 * testsuite/gas/riscv/priv-reg-fail-version-1p11.d: Likewise.
79163 * testsuite/gas/riscv/priv-reg-fail-version-1p9p1.d: Likewise.
79164 * testsuite/gas/riscv/priv-reg-version-1p10.d: Added Crypto seed csr.
79165 * testsuite/gas/riscv/priv-reg-version-1p11.d: Likewise.
79166 * testsuite/gas/riscv/priv-reg-version-1p9p1.d: Likewise.
79167 * testsuite/gas/riscv/priv-reg.s: Likewise.
79169 2021-11-16 jiawei <jiawei@iscas.ac.cn>
79171 RISC-V: Scalar crypto instructions and operand set.
79172 Add instructions in k-ext, some instruction in zbkb, zbkc is reuse from
79173 zbb,zbc, we just change the class attribute to make them both support.
79174 The 'aes64ks1i' and 'aes64ks2' instructions are present in both the Zknd
79175 and Zkne extensions on rv64. Add new operand letter 'y' to present 'bs'
79176 symbol and 'Y' to present 'rnum' symbolc for zkn instructions. Also add
79177 a new Entropy Source CSR define 'seed' located at address 0x015.
79180 * elfxx-riscv.c (riscv_multi_subset_supports): Added support for
79183 *config/tc-riscv.c (enum riscv_csr_class): Added CSR_CLASS_ZKR.
79184 (riscv_csr_address): Checked for CSR_CLASS_ZKR.
79185 (validate_riscv_insn): Added y and Y for bs and rnum operands.
79186 (riscv_ip): Handle y and Y operands.
79188 * opcode/riscv-opc.h: Added encodings of crypto instructions.
79189 Also defined new csr seed, which address is 0x15.
79190 * opcode/riscv.h: Defined OP_* and INSN_CLASS_* for crypto.
79192 * riscv-dis.c (print_insn_args): Recognized new y and Y operands.
79193 * riscv-opc.c (riscv_opcodes): Added crypto instructions.
79195 2021-11-16 jiawei <jiawei@iscas.ac.cn>
79197 RISC-V: Minimal support of scalar crypto extension.
79198 Minimal support of scalar crypto extension, add "k" in the
79199 riscv_supported_std_ext, to make the order check right with
79203 * elfxx-riscv.c (riscv_implicit_subsets): Added implicit
79204 rules for zk* extensions.
79205 (riscv_supported_std_ext): Added entry for k.
79206 (riscv_supported_std_z_ext): Added entries for zk*.
79208 2021-11-16 Simon Marchi <simon.marchi@efficios.com>
79210 gdb: rework "set debuginfod" commands
79211 As discussed here [1], do some re-work in the "set debuginfod commands".
79213 First, use "set debuginfod enabled on/off/ask" instead of "set
79214 debuginfod on/off/ask". This is more MI-friendly, and it gives an
79215 output that makes more sense in "info set", for example.
79217 Then, make the show commands not call "error" when debuginfod support is
79218 not compiled in. This makes the commands "show" and "show debuginfod"
79219 stop early, breaking gdb.base/default.exp:
79221 Running /home/smarchi/src/binutils-gdb/gdb/testsuite/gdb.base/default.exp ...
79222 FAIL: gdb.base/default.exp: info set
79223 FAIL: gdb.base/default.exp: show
79225 - Make the "debuginfod enabled" setting default to "off" when debuginfod
79226 support is not compiled in, and "ask" otherwise.
79227 - Make the setter of "debuginfod enabled" error out when debuginfod
79228 support is not compiled in, so that "debuginfod enabled" will always
79229 remain "off" in that case.
79230 - Make the setter of "debuginfod verbose" work in any case. I don't
79231 see the harm in letting the user change that setting, since the user will
79232 hit an error if they try to enable the use of debuginfod.
79233 - I would do the same for the "debuginfod urls" setter, but because
79234 this one needs to see the DEBUGINFOD_URLS_ENV_VAR macro, provided by
79235 libdebuginfod, I made that one error out as well if debuginfod
79236 support is not compiled it (otherwise, I would have left it like
79237 "debuginfod verbose". Alternatively, we could hard-code
79238 "DEBUGINFOD_URLS" in the code (in fact, it was prior to this patch,
79239 but I think it was an oversight, as other spots use
79240 DEBUGINFOD_URLS_ENV_VAR), or use a dummy string to store the setting,
79241 but I don't really see the value in that.
79243 Rename debuginfod_enable to debuginfod_enabled, just so it matches the
79246 [1] https://sourceware.org/pipermail/gdb-patches/2021-October/182937.html
79248 Change-Id: I45fdb2993f668226a5639228951362b7800f09d5
79249 Co-Authored-By: Aaron Merey <amerey@redhat.com>
79251 2021-11-16 Simon Marchi <simon.marchi@polymtl.ca>
79253 gdb: adjust gdbarch_tdep calls in nat files
79254 Commit 345bd07cce33 ("gdb: fix gdbarch_tdep ODR violation") forgot to
79255 update the gdbarch_tdep calls in the native files other than x86-64
79256 Linux. This patch updates them all (to the best of my knowledge).
79257 These are the files I was able to build-test:
79259 aarch64-linux-nat.c
79266 And these are the ones I could not build-test:
79272 ia64-tdep.c (the part that needs libunwind)
79276 If there are still some build problems related to gdbarch_tdep in them,
79277 they should be pretty obvious to fix.
79279 Change-Id: Iaa3d791a850e4432973757598e634e3da6061428
79281 2021-11-16 Simon Marchi <simon.marchi@polymtl.ca>
79283 gdb: remove unused variables in xtensa-linux-nat.c
79284 While build-testing this file, the compiler complained about these two
79285 unused variables, remove them.
79287 Change-Id: I3c54f779f12c16ef6184af58aca75eaad042ce4e
79289 2021-11-16 Simon Marchi <simon.marchi@polymtl.ca>
79291 gdb: add arc-newlib-tdep.c to ALL_TARGET_OBS
79292 This file is currently not compiled in an --enable-targets=all build,
79293 but it should be. Add it to ALL_TARGET_OBS.
79295 Update the gdbarch_tdep call that commit 345bd07cce33 ("gdb: fix
79296 gdbarch_tdep ODR violation") forgot to update.
79298 Change-Id: I86248a01493eea5e70186e9c46a298ad3994b034
79300 2021-11-16 Jim Wilson <wilson@tuliptree.org>
79302 Update my email address.
79303 I've left SiFive and have a new gmail account because it is convenient
79304 to use with git send-email. I'm planning to use this for my RISC-V
79305 work. My tuliptree address still works, it just isn't as convenient.
79308 * MAINTAINERS (RISC-V): Update my address.
79310 2021-11-16 GDB Administrator <gdbadmin@sourceware.org>
79312 Automatic date update in version.in
79314 2021-11-15 Tom de Vries <tdevries@suse.de>
79316 [gdb] Don't use gdb_stdlog for inferior-events
79317 The test-case gdb.base/foll-vfork.exp contains:
79319 if [gdb_debug_enabled] {
79320 untested "debug is enabled"
79325 To understand what it does, I disabled this bit and ran with GDB_DEBUG=infrun,
79328 $ cd $build/gdb/testsuite
79329 $ make check GDB_DEBUG=infrun RUNTESTFLAGS=gdb.base/foll-vfork.exp
79333 (gdb) PASS: gdb.base/foll-vfork.exp: exec: \
79334 vfork parent follow, through step: set follow-fork parent
79336 33 if (pid == 0) {^M
79337 (gdb) FAIL: gdb.base/foll-vfork.exp: exec: \
79338 vfork parent follow, through step: step
79341 The problem is that the test-case expects:
79343 (gdb) PASS: gdb.base/foll-vfork.exp: exec: \
79344 vfork parent follow, through step: set follow-fork parent
79346 [Detaching after vfork from child process 28169]^M
79347 33 if (pid == 0) {^M
79348 (gdb) PASS: gdb.base/foll-vfork.exp: exec: \
79349 vfork parent follow, through step: step
79351 but the "Detaching" line has been redirected to
79352 $outputs/gdb.base/foll-vfork/gdb.debug.
79354 I looked at the documentation of "set logging debugredirect [on|off]":
79356 By default, GDB debug output will go to both the terminal and the logfile.
79357 Set debugredirect if you want debug output to go only to the log file.
79359 and my interpretation of it was that "debug output" did not match the
79360 "messages" description of inferior-events:
79362 The set print inferior-events command allows you to enable or disable printing
79363 of messages when GDB notices that new inferiors have started or that inferiors
79364 have exited or have been detached.
79367 Fix the discrepancy by not using gdb_stdlog for inferior-events.
79369 Update the gdb.base/foll-vfork.exp test-case to not require
79370 gdb_debug_enabled == 0.
79372 Tested on x86_64-linux.
79374 Tested test-case gdb.base/foll-vfork.exp with and without GDB_DEBUG=infrun.
79376 2021-11-15 Roland McGrath <mcgrathr@google.com>
79378 ld: Fix testsuite failures under --enable-textrel-check=error
79380 * testsuite/ld-aarch64/dt_textrel.d: Pass explicit -z notext in
79381 case ld was configured with --enable-textrel-check=error.
79382 * testsuite/ld-aarch64/pr22764.d: Likewise.
79383 * testsuite/ld-aarch64/pr20402.d: Likewise.
79385 2021-11-15 Luis Machado <luis.machado@linaro.org>
79387 Extend the prologue analyzer to handle the bti instruction
79388 Handle the BTI instruction in the prologue analyzer. The patch handles all
79389 the variations of the BTI instruction.
79391 2021-11-15 Simon Marchi <simon.marchi@polymtl.ca>
79393 gdb: fix gdbarch_tdep ODR violation
79394 I would like to be able to use non-trivial types in gdbarch_tdep types.
79395 This is not possible at the moment (in theory), because of the one
79398 To allow it, rename all gdbarch_tdep types to <arch>_gdbarch_tdep, and
79399 make them inherit from a gdbarch_tdep base class. The inheritance is
79400 necessary to be able to pass pointers to all these <arch>_gdbarch_tdep
79401 objects to gdbarch_alloc, which takes a pointer to gdbarch_tdep.
79403 These objects are never deleted through a base class pointer, so I
79404 didn't include a virtual destructor. In the future, if gdbarch objects
79405 deletable, I could imagine that the gdbarch_tdep objects could become
79406 owned by the gdbarch objects, and then it would become useful to have a
79407 virtual destructor (so that the gdbarch object can delete the owned
79408 gdbarch_tdep object). But that's not necessary right now.
79410 It turns out that RISC-V already has a gdbarch_tdep that is
79411 non-default-constructible, so that provides a good motivation for this
79414 Most changes are fairly straightforward, mostly needing to add some
79415 casts all over the place. There is however the xtensa architecture,
79416 doing its own little weird thing to define its gdbarch_tdep. I did my
79417 best to adapt it, but I can't test those changes.
79419 Change-Id: Ic001903f91ddd106bd6ca09a79dabe8df2d69f3b
79421 2021-11-15 Clément Chigot <clement.chigot@atos.net>
79423 COFF: avoid modifications over C_FILE filename aux entries.
79424 Commit e86fc4a5bc37 ("PR 28447: implement multiple parameters for .file
79425 on XCOFF") introduces C_FILE entries which can store additional
79427 However, some modifications are needed by them but not by the original
79428 C_FILE entries, usually representing the filename.
79429 This patch ensures that filename entries are kept as is, in order to
79430 protect targets not supporting the additional entries.
79432 * coffgen.c (coff_write_symbol): Protect filename entries
79433 (coff_write_symbols): Likewise.
79434 (coff_print_symbol): Likewise.
79436 2021-11-15 Eric Botcazou <ebotcazou@gcc.gnu.org>
79438 Deal with full path in .file 0 directive
79439 Gas uses the directory part, if present, of the .file 0 directive to set
79440 entry 0 of the directory table in DWARF 5, which represents the "current
79443 Now Gas also uses the file part of the same directive to set entry 0 of the
79444 file table, which represents the "current compilation file". But the latter
79445 need not be located in the former so GCC will use a full path in the file
79446 part when it is passed a full path:
79448 gcc -c /full/path/test.c -save-temps
79452 .file 0 "/current/directory" "/full/path/test.c"
79454 in the assembly file and:
79456 The Directory Table (offset 0x22, lines 2, columns 1):
79458 0 (indirect line string, offset: 0x25): /current/directory
79459 1 (indirect line string, offset: 0x38): /full/path
79461 The File Name Table (offset 0x30, lines 2, columns 2):
79463 0 0 (indirect line string, offset: 0x43): /full/path/test.c
79465 in the object file. Note the full path and the questionable Dir value in
79466 the 0 entry of the file table.
79468 2021-11-15 Mike Frysinger <vapier@gentoo.org>
79470 sim: cris: make error message test a little more flexible
79471 The point of this test is to just make sure the usage text is shown,
79472 not the exact details of the usage text. So shorten the output test
79473 to match the beginning. This fixes breakage when the output changed
79474 slightly to include [--].
79476 sim: run: fix crash in argc==0 error situation
79477 The new argv processing code assumed that we were always passed a
79478 command line. If we weren't, make sure we don't crash before we
79479 get a chance to output an error message about incorrect usage.
79481 sim: cris: touch up rvdummy handling
79482 Add quiet build support and make sure it's removed with `make clean`.
79484 sim: cris: replace custom "dest" test field with new --argv0
79485 The #dest field used in the cris testsuite is a bit of hack to set the
79486 argv[0] for the tests to read out later on. Now that the sim has an
79487 option to set argv[0] explicitly, we don't need this custom field, so
79488 let's drop it to harmonize the testsuites a little.
79490 sim: run: add --argv0 option to control argv[0]
79491 We default argv[0] to the program we run which is a standard *NIX
79492 convention, but sometimes we want to be able to control the argv[0]
79493 setting independently (especially for programs that inspect argv[0]
79494 to change their behavior or output). Add an option to control it.
79496 2021-11-15 Mike Frysinger <vapier@gentoo.org>
79498 sim: split program path out of argv vector
79499 We use the program argv to both find the program to run (argv[0]) and
79500 to hold the arguments to the program. Most of the time this is fine,
79501 but if we want to let programs specify argv[0] independently (which is
79502 possible in standard *NIX programs), this double duty doesn't work.
79504 So let's split the path to the program to run out into a separate
79505 field by itself. This simplifies the various sim_open funcs too.
79507 By itself, this code is more of a logical cleanup than something that
79508 is super useful. But it will open up customization of argv[0] in a
79509 follow up commit. Split the changes to make it easier to review.
79511 2021-11-15 Mike Frysinger <vapier@gentoo.org>
79513 sim: bfin: fix mach/xfail usage in tests
79514 Set the mach to the right value all the time, and update xfail to
79515 say the test fails on all targets. WIth multitarget testing, the
79516 idea of target here doesn't make much sense.
79518 2021-11-15 Alan Modra <amodra@gmail.com>
79520 -Waddress fixes for gold testsuite
79521 Current mainline gcc.
79522 common_test_1.c: In function 'main':
79523 common_test_1.c:56:14: error: comparison between two arrays [-Werror=array-compare]
79524 56 | assert (c5 > c4);
79526 common_test_1.c:56:14: note: use '&c5[0] > &c4[0]' to compare the addresses
79528 * testsuite/common_test_1.c: Avoid -Waddress warnings.
79529 * testsuite/common_test_1_v1.c: Likewise.
79530 * testsuite/common_test_1_v2.c: Likewise.
79531 * testsuite/script_test_2.cc: Likewise.
79533 2021-11-15 Alan Modra <amodra@gmail.com>
79535 PowerPC64 @notoc in non-power10 code
79536 R_PPC64_REL24_P9NOTOC is a variant of R_PPC64_REL24_NOTOC for use on
79537 @notoc cals from non-power10 code in the rare case that using such a
79538 construct is useful. R_PPC64_REL24_P9NOTOC will be emitted by gas
79539 rather than R_PPC64_REL24_NOTOC when @notoc is used in a branch
79540 instruction if power10 instructions are not enabled at that point.
79541 The new relocation tells the linker to not use power10 instructions on
79542 any stub emitted for that branch, unless overridden by
79543 --power10-stubs=yes.
79545 The current linker heuristic of only generating power10 instructions
79546 for stubs if power10-only relocations are detected, continues to be
79550 * elf/ppc64.h (R_PPC64_REL24_P9NOTOC): Define.
79552 * reloc.c (BFD_RELOC_PPC64_REL24_P9NOTOC): Define.
79553 * elf64-ppc.c (ppc64_elf_howto_raw): Add entry for new reloc.
79554 (ppc64_elf_reloc_type_lookup): Handle it.
79555 (enum ppc_stub_type): Delete.
79556 (enum ppc_stub_main_type, ppc_stub_sub_type): New.
79557 (struct ppc_stub_type): New.
79558 (struct ppc_stub_hash_entry): Use the above new type.
79559 (struct ppc_link_hash_table): Update stub_count.
79560 (is_branch_reloc, ppc64_elf_check_relocs),
79561 (toc_adjusting_stub_needed): Handle new reloc.
79562 (stub_hash_newfunc, select_alt_stub, ppc_merge_stub),
79563 (ppc_type_of_stub, plt_stub_size, build_plt_stub),
79564 (build_tls_get_addr_head, build_tls_get_addr_tail),
79565 (ppc_build_one_stub, ppc_size_one_stub, ppc64_elf_size_stubs),
79566 (ppc64_elf_build_stubs, ppc64_elf_relocate_section): Handle new
79567 reloc. Modify stub handling to suit new scheme.
79568 * bfd-in2.h: Regenerate.
79569 * libbfd.h: Regenerate.
79571 * config/tc-ppc.c (ppc_elf_suffix): When power10 is not enabled
79572 return BFD_RELOC_PPC64_REL24_P9NOTOC for @notoc.
79573 (fixup_size, ppc_force_relocation, ppc_fix_adjustable): Handle
79574 BFD_RELOC_PPC64_REL24_P9NOTOC.
79576 * testsuite/ld-powerpc/callstub-2.s: Add .machine power10.
79578 2021-11-15 Alan Modra <amodra@gmail.com>
79580 Regenerate a couple of files
79581 A couple of files changed on my latest --enable-maintainer-mode
79582 build. ld/Makefile.in had a missing dependency but better sorting of
79583 the loongson entries.
79586 * configure: Regenerate.
79588 * Makefile.am: Sort loongson entries.
79589 * Makefile.in: Regenerate.
79591 2021-11-15 Pedro Alves <pedro@palves.net>
79593 Fix build with current GCC: EL_EXPLICIT(location) always non-NULL
79594 Compiling GDB with current GCC (1b4a63593b) runs into this:
79596 src/gdb/location.c: In function 'int event_location_empty_p(const event_location*)':
79597 src/gdb/location.c:963:38: error: the address of 'event_location::<unnamed union>::explicit_loc' will never be NULL [-Werror=address]
79598 963 | return (EL_EXPLICIT (location) == NULL
79600 src/gdb/location.c:57:30: note: 'event_location::<unnamed union>::explicit_loc' declared here
79601 57 | struct explicit_location explicit_loc;
79604 GCC is right, EL_EXPLICIT is defined as returning the address of an
79607 /* An explicit location. */
79608 struct explicit_location explicit_loc;
79609 #define EL_EXPLICIT(P) (&((P)->u.explicit_loc))
79611 and thus must always be non-NULL.
79613 Change-Id: Ie74fee7834495a93affcefce03c06e4d83ad8191
79615 2021-11-15 GDB Administrator <gdbadmin@sourceware.org>
79617 Automatic date update in version.in
79619 2021-11-14 Lancelot SIX <lsix@lancelotsix.com>
79621 [PR gdb/16238] Add completer for the show user command
79622 The 'show user' command (which shows the definition of non-python/scheme
79623 user defined commands) is currently missing a completer. This is
79624 mentioned in PR 16238. Having one can improve the user experience.
79626 In this commit I propose an implementation for such completer as well as
79627 the associated tests.
79629 Tested on x86_64 GNU/Linux.
79631 All feedbacks are welcome.
79633 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=16238
79635 2021-11-14 Alan Modra <amodra@gmail.com>
79637 sync libbacktrace from gcc
79639 2021-11-14 GDB Administrator <gdbadmin@sourceware.org>
79641 Automatic date update in version.in
79643 2021-11-13 H.J. Lu <hjl.tools@gmail.com>
79645 Sync Makefile.tpl with GCC
79646 * Makefile.tpl: Sync with GCC.
79647 * Makefile.in: Regenerate.
79649 2021-11-13 Mike Frysinger <vapier@gentoo.org>
79651 sim: sh: fix switch-bool warnings
79652 This code triggers -Werror=switch-bool warnings with <=gcc-5 versions.
79653 Rework it to use if statements instead as it also simplifies a bit.
79655 sim: sh: rework carry checks to not rely on integer overflows
79656 In <=gcc-7 versions, -fstrict-overflow is enabled by default, and that
79657 triggers warnings in this code that relies on integer overflows to test
79658 for carries. Change the logic to test against the limit directly.
79660 2021-11-13 GDB Administrator <gdbadmin@sourceware.org>
79662 Automatic date update in version.in
79664 2021-11-12 Carl Love <cel@us.ibm.com>
79666 Fix gdb.base/sigstep.exp test for ppc
79667 The test stops at <signal_handler called> which is the call to the handler
79668 rather than in the handler as intended. This patch replaces the
79669 gdb_test "$enter_cmd to handler" with a gdb_test_multiple test. The multiple
79670 test looks for the stop at <signal_handler called>. If found, the command
79671 is issued again. The test passes if gdb stops in the handler as expected.
79673 (gdb) PASS: gdb.base/sigstep.exp: stepi to handler, nothing in handler, step
79674 from handler: continue to signal
79676 <signal handler called>
79678 => 0x7ffff7f80440 <__kernel_start_sigtramp_rt64>: bctrl
79680 handler (sig=551) at sigstep.c:32
79683 => 0x10000097c <handler>: addis r2,r12,2
79684 (gdb) PASS: gdb.base/sigstep.exp: stepi to handler, nothing in handler,
79685 step from handler: stepi to handler
79687 Patch has been tested on x86_64-linux and ppc64le-linux with no test failures.
79689 2021-11-12 Tom de Vries <tdevries@suse.de>
79691 [gdb/testsuite] Fix regexp in gdb.base/foll-vfork.exp
79694 (gdb) PASS: gdb.base/foll-vfork.exp: exit: \
79695 vfork relations in info inferiors: continue to child exit
79697 Num Description Connection Executable ^M
79698 1 <null> foll-vfork-exit ^M
79699 * 2 <null> foll-vfork-exit ^M
79700 (gdb) I'm the proud parent of child #5044!^M
79701 FAIL: gdb.base/foll-vfork.exp: exit: vfork relations in info inferiors: \
79702 vfork relation no longer appears in info inferiors (timeout)
79705 Fix this by removing the '$' anchor in the corresponding '$gdb_prompt $'
79708 Tested on x86_64-linux.
79710 2021-11-12 Alan Modra <amodra@gmail.com>
79712 Don't compile some opcodes files when bfd is 32-bit only
79713 * Makefile.am (TARGET_LIBOPCODES_CFILES): Split into..
79714 (TARGET64_LIBOPCODES_CFILES): ..this and..
79715 (TARGET32_LIBOPCODES_CFILES): ..this.
79716 (ALL_MACHINES): Likewise split to
79717 (ALL64_MACHINES, ALL32_MACHINES): ..this.
79718 * disassemble.c: Define some ARCH_* when ARCH_all only if BFD64.
79719 * configure.ac (BFD_MACHINES): Defined depending on BFD_ARCH_SIZE.
79720 * Makefile.in: Regenerate.
79721 * configure: Regenerate.
79723 Import Makefile.def from gcc
79724 * Makefile.def: Import from gcc.
79725 * Makefile.in: Regenerate.
79727 2021-11-12 Alan Modra <amodra@gmail.com>
79729 Fix demangle style usage info
79730 Extract allowed styles from libiberty, so we don't have to worry about
79731 our help messages getting out of date. The function probably belongs
79732 in libiberty/cplus-dem.c but it can be here for a while to iron out
79736 * demanguse.c: New file.
79737 * demanguse.h: New file.
79738 * nm.c (usage): Break up output. Use display_demangler_styles.
79739 * objdump.c (usage): Use display_demangler_styles.
79740 * readelf.c (usage): Likewise.
79741 * Makefile.am: Add demanguse.c and demanguse.h.
79742 * Makefile.in: Regenerate.
79743 * po/POTFILESin: Regenerate.
79745 2021-11-12 GDB Administrator <gdbadmin@sourceware.org>
79747 Automatic date update in version.in
79749 2021-11-11 Simon Marchi <simon.marchi@efficios.com>
79751 gdb: fix "set scheduler-locking" thread exit hang
79752 GDB hangs when doing this:
79754 - launch inferior with multiple threads
79755 - multiple threads hit some breakpoint(s)
79756 - one breakpoint hit is presented as a stop, the rest are saved as
79757 pending wait statuses
79758 - "set scheduler-locking on"
79759 - resume the currently selected thread (because of scheduler-locking,
79760 it's the only one resumed), let it execute until exit
79761 - GDB hangs, not showing the prompt, impossible to interrupt with ^C
79763 When the resumed thread exits, we expect the target to return a
79764 TARGET_WAITKIND_NO_RESUMED event, and that's what we see:
79766 [infrun] fetch_inferior_event: enter
79767 [infrun] scoped_disable_commit_resumed: reason=handling event
79768 [infrun] random_pending_event_thread: None found.
79769 [Thread 0x7ffff7d9c700 (LWP 309357) exited]
79770 [infrun] print_target_wait_results: target_wait (-1.0.0 [process -1], status) =
79771 [infrun] print_target_wait_results: -1.0.0 [process -1],
79772 [infrun] print_target_wait_results: status->kind = no-resumed
79773 [infrun] handle_inferior_event: status->kind = no-resumed
79774 [infrun] handle_no_resumed: TARGET_WAITKIND_NO_RESUMED (ignoring: found resumed)
79775 [infrun] prepare_to_wait: prepare_to_wait
79776 [infrun] reset: reason=handling event
79777 [infrun] maybe_set_commit_resumed_all_targets: not requesting commit-resumed for target native, no resumed threads
79778 [infrun] fetch_inferior_event: exit
79780 The problem is in handle_no_resumed: we check if some other thread is
79781 actually resumed, to see if we should ignore that event (see comments in
79782 that function for more info). If this condition is true:
79784 (thread->executing () || thread->has_pending_waitstatus ())
79786 ... then we ignore the event. The problem is that there are some non-resumed
79787 threads with a pending event, which makes us ignore the event. But these
79788 threads are not resumed, so we end up waiting while nothing executes, hence
79791 My first fix was to change the condition to:
79793 (thread->executing ()
79794 || (thread->resumed () && thread->has_pending_waitstatus ()))
79796 ... but then it occured to me that we could simply check for:
79798 (thread->resumed ())
79800 Since "executing" implies "resumed", checking simply for "resumed"
79801 covers threads that are resumed and executing, as well as threads that
79802 are resumed with a pending status, which is what we want.
79804 Change-Id: Ie796290f8ae7f34c026ca3a8fcef7397414f4780
79806 2021-11-11 Tom de Vries <tdevries@suse.de>
79808 [gdb/build] Fix Wimplicit-exception-spec-mismatch in clang build
79809 When building with clang 13 (and -std=gnu++17 to work around an issue in
79810 string_view-selftests.c), we run into a few Wimplicit-exception-spec-mismatch
79813 src/gdbsupport/new-op.cc:102:1: error: function previously declared with an \
79814 explicit exception specification redeclared with an implicit exception \
79815 specification [-Werror,-Wimplicit-exception-spec-mismatch]
79816 operator delete (void *p)
79818 /usr/include/c++/11/new:130:6: note: previous declaration is here
79819 void operator delete(void*) _GLIBCXX_USE_NOEXCEPT
79823 These are due to recent commit 5fff6115fea "Fix
79824 LD_PRELOAD=/usr/lib64/libasan.so.6 gdb".
79826 Fix this by adding the missing noexcept.
79828 Build on x86_64-linux, using gcc 7.5.0 and clang 13.0.0.
79830 2021-11-11 Tom de Vries <tdevries@suse.de>
79832 [gdb/build] Fix build with -std=c++11
79833 When building with -std=c++11, we run into two Werror=missing-declarations:
79835 new-op.cc: In function 'void operator delete(void*, std::size_t)':
79836 new-op.cc:114:1: error: no previous declaration for \
79837 'void operator delete(void*, std::size_t)' [-Werror=missing-declarations]
79838 operator delete (void *p, std::size_t) noexcept
79840 new-op.cc: In function 'void operator delete [](void*, std::size_t)':
79841 new-op.cc:132:1: error: no previous declaration for \
79842 'void operator delete [](void*, std::size_t)' [-Werror=missing-declarations]
79843 operator delete[] (void *p, std::size_t) noexcept
79847 These are due to recent commit 5fff6115fea "Fix
79848 LD_PRELOAD=/usr/lib64/libasan.so.6 gdb".
79850 The declarations are provided by <new> (which is included) for c++14 onwards,
79851 but they are missing for c++11.
79853 Fix this by adding the missing declarations.
79855 Tested on x86_64-linux, with gcc 7.5.0, both without (implying -std=gnu++14) and
79858 2021-11-11 Tom de Vries <tdevries@suse.de>
79860 [gdb/testsuite] Add gdb.arch/ppc64-break-on-_exit.exp
79861 Add a regression test-case for commit a50bdb99afe "[gdb/tdep, rs6000] Don't
79862 skip system call in skip_prologue":
79863 - set a breakpoint on a local copy of glibc's _exit, and
79864 - verify that it triggers.
79866 The test-case uses an assembly file by default, but also has the possibility
79867 to use a C source file instead.
79869 Tested on ppc64le-linux. Verified that the test-case fails without
79870 aforementioned commit, and passes with the commit. Both with assembly
79873 2021-11-11 Nelson Chu <nelson.chu@sifive.com>
79875 RISC-V: Dump objects according to the elf architecture attribute.
79876 For now we should always generate the elf architecture attribute both for
79877 elf and linux toolchains, so that we could dump the objects correctly
79878 according to the generated architecture string. This patch resolves the
79879 problem that we probably dump an object with c.nop instructions, but
79880 in fact the c extension isn't allowed. Consider the following case,
79882 nelson@LAPTOP-QFSGI1F2:~/test$ cat temp.s
79889 nelson@LAPTOP-QFSGI1F2:~/test$ ~/binutils-dev/build-elf32-upstream/build-install/bin/riscv32-unknown-elf-as temp.s -o temp.o
79890 nelson@LAPTOP-QFSGI1F2:~/test$ ~/binutils-dev/build-elf32-upstream/build-install/bin/riscv32-unknown-elf-objdump -d temp.o
79892 temp.o: file format elf32-littleriscv
79894 Disassembly of section .text:
79897 0: 00a50533 add a0,a0,a0
79903 nelson@LAPTOP-QFSGI1F2:~/test$ ~/binutils-dev/build-elf32-upstream/build-install/bin/riscv32-unknown-elf-readelf -A temp.o
79904 Attribute Section: riscv
79906 Tag_RISCV_arch: "rv32i2p0_m2p0_a2p0_f2p0_d2p0"
79908 The c.nop at address 0x6 is generated for alignment, but since the rvc isn't
79909 allowed for this object, dump it as a c.nop instruction looks wrong. After
79910 applying this patch, I get the following result,
79912 nelson@LAPTOP-QFSGI1F2:~/test$ ~/binutils-dev/build-elf32-upstream/build-install/bin/riscv32-unknown-elf-objdump -d temp.o
79914 temp.o: file format elf32-littleriscv
79916 Disassembly of section .text:
79919 0: 00a50533 add a0,a0,a0
79926 For the current objdump, we dump data to .byte/.short/.word/.dword, and
79927 dump the unknown or unsupported instructions to .2byte/.4byte/.8byte, which
79928 respectively are 2, 4 and 8 bytes instructions. Therefore, we shouldn't
79929 dump the 0x0001 as a c.nop instruction in the above case, we should dump
79930 it to .2byte 0x1 as a unknown instruction, since the rvc is disabled.
79932 However, consider that some people may use the new objdump to dump the old
79933 objects, which don't have any elf attributes. We usually set the default
79934 architecture string to rv64g by bfd/elfxx-riscv.c:riscv_set_default_arch.
79935 But this will cause rvc instructions to be unrecognized. Therefore, we
79936 set the default architecture string to rv64gc for disassembler, to keep
79937 the previous behavior.
79939 This patch pass the riscv-gnu-toolchain gcc/binutils regressions for
79940 rv32emc-elf, rv32gc-linux, rv32i-elf, rv64gc-elf and rv64gc-linux
79941 toolchains. Also, tested by --enable-targets=all and can build
79942 riscv-gdb successfully.
79945 * elfnn-riscv.c (riscv_merge_arch_attr_info): Tidy the
79946 codes for riscv_parse_subset_t setting.
79947 * elfxx-riscv.c (riscv_get_default_ext_version): Updated.
79948 (riscv_subset_supports): Moved from gas/config/tc-riscv.c.
79949 (riscv_multi_subset_supports): Likewise.
79950 * elfxx-riscv.h: Added extern for riscv_subset_supports and
79951 riscv_multi_subset_supports.
79953 * config/tc-riscv.c (riscv_subset_supports): Moved to
79955 (riscv_multi_subset_supports): Likewise.
79956 (riscv_rps_as): Defined for architectrue parser.
79957 (riscv_set_arch): Updated.
79958 (riscv_set_abi_by_arch): Likewise.
79959 (riscv_csr_address): Likewise.
79960 (reg_lookup_internal): Likewise.
79961 (riscv_ip): Likewise.
79962 (s_riscv_option): Updated.
79963 * testsuite/gas/riscv/mapping-04b.d: Updated.
79964 * testsuite/gas/riscv/mapping-norelax-03b.d: Likewise.
79965 * testsuite/gas/riscv/mapping-norelax-04b.d: Likewise.
79967 * riscv-dis.c: Include elfxx-riscv.h since we need the
79968 architecture parser. Also removed the cpu-riscv.h, it
79969 is already included in elfxx-riscv.h.
79970 (default_isa_spec): Defined since the parser need this
79971 to set the default architecture string.
79972 (xlen): Moved out from riscv_disassemble_insn as a global
79973 variable, it is more convenient to initialize riscv_rps_dis.
79974 (riscv_subsets): Defined to recoed the supported
79976 (riscv_rps_dis): Defined for architectrue parser.
79977 (riscv_disassemble_insn): Call riscv_multi_subset_supports
79978 to make sure if the instructions are valid or not.
79979 (print_insn_riscv): Initialize the riscv_subsets by parsing
79980 the elf architectrue attribute. Otherwise, set the default
79981 architectrue string to rv64gc.
79983 2021-11-11 Mike Frysinger <vapier@gentoo.org>
79985 sim: testsuite: drop sim_compile cover function
79986 Most code isn't using this, and the only call site (in one cris file)
79987 can use target_compile directly. So switch it over to simplify.
79989 sim: cris: stop testing a.out explicitly [ld/13900]
79990 Since gcc dropped support for a.out starting with 4.4.0 in 2009, it's
79991 been impossible to verify this code actually still works. Since it
79992 crashes in ld, and it uses a config option that no other tests uses
79993 and we want to remove, drop the test to avoid all the trouble.
79995 sim: io: tweak compiler workaround with error output
79996 Outputting an extra space broke a cris test. Change the workaround
79997 to use %s with an empty string to avoid the compiler warning but not
79998 output an extra space.
80000 sim: testsuite: delete unused arm remote host logic
80001 There's no need to sync testutils.inc with remote hosts. The one
80002 we have in the source tree is all we need and only thing we test.
80003 Delete it to simplify.
80005 sim: synacor: simplify test generation
80006 Objcopy was used to create a binary file of just the executable code
80007 since the environment requires code to based at address 0. We can
80008 accomplish the same thing with the -Ttext=0 flag, so switch to that
80009 to get rid of custom logic.
80011 2021-11-11 GDB Administrator <gdbadmin@sourceware.org>
80013 Automatic date update in version.in
80015 2021-11-10 Tom Tromey <tromey@adacore.com>
80017 Handle PIE in .debug_loclists
80018 Simon pointed out that my recent patches to .debug_loclists caused
80019 some regressions. After a brief discussion we realized it was because
80020 his system compiler defaults to PIE.
80022 This patch changes this code to unconditionally apply the text offset
80023 here. It also changes loclist_describe_location to work more like
80024 dwarf2_find_location_expression.
80026 I tested this by running the gdb.dwarf2 tests both with and without
80029 2021-11-10 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
80031 arm: enable Cortex-A710 CPU
80032 This patch is adding support for Cortex-A710 CPU in Arm.
80036 * cpu-arm.c (processors): Add cortex-a710.
80040 * NEWS: Update docs.
80041 * config/tc-arm.c (arm_cpus): Add cortex-a710 to -mcpu.
80042 * doc/c-arm.texi: Update docs.
80043 * testsuite/gas/arm/cpu-cortex-a710.d: New test.
80045 2021-11-10 Clément Chigot <clement.chigot@atos.net>
80047 gdb: adjust x_file fields on COFF readers
80048 Commit e86fc4a5bc37 ("PR 28447: implement multiple parameters for .file
80049 on XCOFF") changes the structure associated to the internal
80050 representation of files in COFF formats. However, gdb directory update
80051 has been forgotten, leading to compilation errors of this kind:
80054 /home/simark/src/binutils-gdb/gdb/coffread.c: In function 'const char* coff_getfilename(internal_auxent*)':
80055 /home/simark/src/binutils-gdb/gdb/coffread.c:1343:29: error: 'union internal_auxent::<unnamed struct>::<unnamed>' has no member named 'x_zeroes'
80056 1343 | if (aux_entry->x_file.x_n.x_zeroes == 0)
80059 Fix it by adjusting the COFF code in GDB.
80061 Change-Id: I703fa134bc722d47515efbd72b88fa5650af6c3c
80063 2021-11-10 Tom de Vries <tdevries@suse.de>
80065 [gdb/testsuite] Add gdb.opt/break-on-_exit.exp
80066 Add a test-case to excercise the problem scenario reported in PR28527 and
80067 fixed in commit a50bdb99afe "[gdb/tdep, rs6000] Don't skip system call in
80069 - set a breakpoint on _exit, and
80070 - verify that it triggers.
80072 Note that this is not a regression test for that commit. Since the actual
80073 code in _exit may vary across os instances, we cannot guarantee that the
80074 problem will always trigger with this test-case.
80076 Rather, this test-case is a version of the original test-case
80077 (gdb.threads/process-dies-while-detaching.exp) that is minimal while still
80078 reproducing the problem reported in PR28527, in that same setting.
80080 The benefit of this test-case is that it exercise real-life code and may
80081 expose similar problems in other settings. Also, it provides a much easier
80082 test-case to investigate in case a similar problem occurs.
80084 Tested on x86_64-linux and ppc64le-linux.
80086 2021-11-10 Mike Frysinger <vapier@gentoo.org>
80088 sim: frv: flip trapdump default back to off
80089 When I refactored this by scoping it to sim-frv-xxx in commit
80090 e7954ef5e5ed90fb7d28c013518f4c2e6bcd20a1 ("sim: frv: scope the
80091 unique configure flag"), I changed the default from off to on.
80092 While the feature is nice for developers, it breaks a bunch of
80093 tests which aren't expecting this extra output. So flip it back
80096 2021-11-10 Pekka Seppänen <pexu@sourceware.mail.kapsi.fi>
80098 PR28575, readelf.c and strings.c use undefined type uint
80099 Since --unicode support (commit b3aa80b45c4) both binutils/readelf.c
80100 and binutils/strings.c use 'uint' in a few locations. It likely
80101 should be 'unsigned int' since there isn't anything defining 'uint'
80102 within binutils (besides zlib) and AFAIK it isn't a standard type.
80104 * readelf.c (print_symbol): Replace uint with unsigned int.
80105 * strings.c (string_min, display_utf8_char): Likewise.
80106 (print_unicode_stream_body, print_unicode_stream): Likewise.
80107 (print_strings): Likewise.
80108 (get_unicode_byte): Wrap long line.
80110 2021-11-10 Clément Chigot <clement.chigot@atos.net>
80112 ld: set correct flags for AIX shared tests
80113 Previous flags were aimed to be run with XLC.
80114 Nowadays, only GCC is being tested with GNU toolchain. Moreover,
80115 recent XLC versions might also accept "-shared".
80117 * testsuite/ld-shared/shared.exp: Adjust shared flags.
80119 2021-11-10 Clément Chigot <clement.chigot@atos.net>
80121 PR 28447: implement multiple parameters for .file on XCOFF
80122 On XCOFF, ".file" pseudo-op allows 3 extras parameters to provide
80123 additional information to AIX linker, or its debugger. These are
80124 stored in auxiliary entries of the C_FILE symbol.
80128 * coffcode.h (combined_entry_type): Add extrap field.
80129 (coff_bigobj_swap_aux_in): Adjust names of x_file fields.
80130 (coff_bigobj_swap_aux_out): Likewise.
80131 * coffgen.c (coff_write_auxent_fname): New function.
80132 (coff_fix_symbol_name): Write x_file using
80133 coff_write_auxent_fname.
80134 (coff_write_symbol): Likewise.
80135 (coff_write_symbols): Add C_FILE auxiliary entries to
80136 string table if needed.
80137 (coff_get_normalized_symtab): Adjust names of x_file fields.
80138 Normalize C_FILE auxiliary entries.
80139 (coff_print_symbol): Print C_FILE auxiliary entries.
80140 * coff-rs6000.c (_bfd_xcoff_swap_aux_in): Adjust names of
80142 (_bfd_xcoff_swap_aux_out): Likewise.
80143 * coff64-rs6000.c (_bfd_xcoff64_swap_aux_in): Likewise.
80144 (_bfd_xcoff64_swap_aux_out): Likewise.
80145 * cofflink.c (_bfd_coff_final_link): Likewise.
80146 (_bfd_coff_link_input_bfd): Likewise.
80147 * coffswap.h (coff_swap_aux_in): Likewise.
80148 * peXXigen.c (_bfd_XXi_swap_aux_in): Likewise.
80149 (_bfd_XXi_swap_aux_out): Likewise.
80150 * xcofflink.c (xcoff_link_input_bfd): Likewise.
80151 * libcoff.h: Regenerate.
80153 * config/tc-ppc.c (ppc_file): New function.
80154 * config/tc-ppc.h (OBJ_COFF_MAX_AUXENTRIES): Change to 4.
80155 * testsuite/gas/ppc/aix.exp: Add tests.
80156 * testsuite/gas/ppc/xcoff-file-32.d: New test.
80157 * testsuite/gas/ppc/xcoff-file-64.d: New test.
80158 * testsuite/gas/ppc/xcoff-file.s: New test.
80160 * coff/internal.h (union internal_auxent): Change x_file to be a
80161 struct instead of a union. Add x_ftype field.
80162 * coff/rs6000.h (union external_auxent): Add x_resv field.
80163 * coff/xcoff.h (XFT_FN): New define.
80164 (XFT_CT): Likewise.
80165 (XFT_CV): Likewise.
80166 (XFT_CD): Likewise.
80168 2021-11-10 Kevin Buettner <kevinb@redhat.com>
80170 Test case for Bug 28308
80171 The purpose of this test is described in the comments in
80172 dprintf-execution-x-script.exp.
80174 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28308
80176 The name of this new test was based on that of an existing test,
80177 bp-cmds-execution-x-script.exp. I started off by copying that test,
80178 adding to it, and then rewriting almost all of it. It's different
80179 enough that I decided that listing the copyright year as 2021
80182 2021-11-10 Kevin Buettner <kevinb@redhat.com>
80184 Fix PR 28308 - dprintf breakpoints not working when run from script
80185 This commit fixes Bug 28308, titled "Strange interactions with
80186 dprintf and break/commands":
80188 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28308
80190 Since creating that bug report, I've found a somewhat simpler way of
80191 reproducing the problem. I've encapsulated it into the GDB test case
80192 which I've created along with this bug fix. The name of the new test
80193 is gdb.base/dprintf-execution-x-script.exp, I'll demonstrate the
80194 problem using this test case, though for brevity, I've placed all
80195 relevant files in the same directory and have renamed the files to all
80196 start with 'dp-bug' instead of 'dprintf-execution-x-script'.
80198 The script file, named dp-bug.gdb, consists of the following commands:
80200 dprintf increment, "dprintf in increment(), vi=%d\n", vi
80207 Note that the final command in this script is 'run'. When 'run' is
80208 instead issued interactively, the bug does not occur. So, let's look
80209 at the interactive case first in order to see the correct/expected
80212 $ gdb -q -x dp-bug.gdb dp-bug
80213 ... eliding buggy output which I'll discuss later ...
80215 Starting program: /mesquite2/sourceware-git/f34-master/bld/gdb/tmp/dp-bug
80217 dprintf in increment(), vi=0
80219 Breakpoint 2, inc_vi () at dprintf-execution-x-script.c:26
80220 26 in dprintf-execution-x-script.c
80222 dprintf in increment(), vi=1
80224 Breakpoint 2, inc_vi () at dprintf-execution-x-script.c:26
80225 26 in dprintf-execution-x-script.c
80227 dprintf in increment(), vi=2
80229 Breakpoint 2, inc_vi () at dprintf-execution-x-script.c:26
80230 26 in dprintf-execution-x-script.c
80232 [Inferior 1 (process 1539210) exited normally]
80234 In this run, in which 'run' was issued from the gdb prompt (instead
80235 of at the end of the script), there are three dprintf messages along
80236 with three 'Breakpoint 2' messages. This is the correct output.
80238 Now let's look at the output that I snipped above; this is the output
80239 when 'run' is issued from the script loaded via GDB's -x switch:
80241 $ gdb -q -x dp-bug.gdb dp-bug
80242 Reading symbols from dp-bug...
80243 Dprintf 1 at 0x40116e: file dprintf-execution-x-script.c, line 38.
80244 Breakpoint 2 at 0x40113a: file dprintf-execution-x-script.c, line 26.
80246 dprintf in increment(), vi=0
80248 Breakpoint 2, inc_vi () at dprintf-execution-x-script.c:26
80249 26 dprintf-execution-x-script.c: No such file or directory.
80252 Breakpoint 2, inc_vi () at dprintf-execution-x-script.c:26
80253 26 in dprintf-execution-x-script.c
80256 Breakpoint 2, inc_vi () at dprintf-execution-x-script.c:26
80257 26 in dprintf-execution-x-script.c
80259 [Inferior 1 (process 1539175) exited normally]
80261 In the output shown above, only the first dprintf message is printed.
80262 The 2nd and 3rd dprintf messages are missing! However, all three
80263 'Breakpoint 2...' messages are still printed.
80265 Why does this happen?
80267 bpstat_do_actions_1() in gdb/breakpoint.c contains the following
80268 comment and code near the start of the function:
80270 /* Avoid endless recursion if a `source' command is contained
80271 in bs->commands. */
80272 if (executing_breakpoint_commands)
80275 scoped_restore save_executing
80276 = make_scoped_restore (&executing_breakpoint_commands, 1);
80278 Also, as described by this comment prior to the 'async' field
80279 in 'struct ui' in top.h, the main UI starts off in sync mode
80280 when processing command line arguments:
80282 /* True if the UI is in async mode, false if in sync mode. If in
80283 sync mode, a synchronous execution command (e.g, "next") does not
80284 return until the command is finished. If in async mode, then
80285 running a synchronous command returns right after resuming the
80286 target. Waiting for the command's completion is later done on
80287 the top event loop. For the main UI, this starts out disabled,
80288 until all the explicit command line arguments (e.g., `gdb -ex
80289 "start" -ex "next"') are processed. */
80291 This combination of things, the state of the static global
80292 'executing_breakpoint_commands' plus the state of the async
80293 field in the main UI causes this behavior.
80295 This is a backtrace after hitting the dprintf breakpoint for
80296 the second time when doing 'run' from the script file, i.e.
80299 Thread 1 "gdb" hit Breakpoint 3, bpstat_do_actions_1 (bsp=0x7fffffffc2b8)
80300 at /ironwood1/sourceware-git/f34-master/bld/../../worktree-master/gdb/breakpoint.c:4431
80301 4431 if (executing_breakpoint_commands)
80303 #0 bpstat_do_actions_1 (bsp=0x7fffffffc2b8)
80304 at gdb/breakpoint.c:4431
80305 #1 0x00000000004d8bc6 in dprintf_after_condition_true (bs=0x1538090)
80306 at gdb/breakpoint.c:13048
80307 #2 0x00000000004c5caa in bpstat_stop_status (aspace=0x116dbc0, bp_addr=0x40116e, thread=0x137f450, ws=0x7fffffffc718,
80308 stop_chain=0x1538090) at gdb/breakpoint.c:5498
80309 #3 0x0000000000768d98 in handle_signal_stop (ecs=0x7fffffffc6f0)
80310 at gdb/infrun.c:6172
80311 #4 0x00000000007678d3 in handle_inferior_event (ecs=0x7fffffffc6f0)
80312 at gdb/infrun.c:5662
80313 #5 0x0000000000763cd5 in fetch_inferior_event ()
80314 at gdb/infrun.c:4060
80315 #6 0x0000000000746d7d in inferior_event_handler (event_type=INF_REG_EVENT)
80316 at gdb/inf-loop.c:41
80317 #7 0x00000000007a702f in handle_target_event (error=0, client_data=0x0)
80318 at gdb/linux-nat.c:4207
80319 #8 0x0000000000b8cd6e in gdb_wait_for_event (block=block@entry=0)
80320 at gdbsupport/event-loop.cc:701
80321 #9 0x0000000000b8d032 in gdb_wait_for_event (block=0)
80322 at gdbsupport/event-loop.cc:597
80323 #10 gdb_do_one_event () at gdbsupport/event-loop.cc:212
80324 #11 0x00000000009d19b6 in wait_sync_command_done ()
80326 #12 0x00000000009d1a3f in maybe_wait_sync_command_done (was_sync=0)
80328 #13 0x00000000009d2033 in execute_command (p=0x7fffffffcb18 "", from_tty=0)
80330 #14 0x0000000000560d5b in execute_control_command_1 (cmd=0x13b9bb0, from_tty=0)
80331 at gdb/cli/cli-script.c:547
80332 #15 0x000000000056134a in execute_control_command (cmd=0x13b9bb0, from_tty=0)
80333 at gdb/cli/cli-script.c:717
80334 #16 0x00000000004c3bbe in bpstat_do_actions_1 (bsp=0x137f530)
80335 at gdb/breakpoint.c:4469
80336 #17 0x00000000004c3d40 in bpstat_do_actions ()
80337 at gdb/breakpoint.c:4533
80338 #18 0x00000000006a473a in command_handler (command=0x1399ad0 "run")
80339 at gdb/event-top.c:624
80340 #19 0x00000000009d182e in read_command_file (stream=0x113e540)
80342 #20 0x0000000000563697 in script_from_file (stream=0x113e540, file=0x13bb0b0 "dp-bug.gdb")
80343 at gdb/cli/cli-script.c:1642
80344 #21 0x00000000006abd63 in source_gdb_script (extlang=0xc44e80 <extension_language_gdb>, stream=0x113e540,
80345 file=0x13bb0b0 "dp-bug.gdb") at gdb/extension.c:188
80346 #22 0x0000000000544400 in source_script_from_stream (stream=0x113e540, file=0x7fffffffd91a "dp-bug.gdb",
80347 file_to_open=0x13bb0b0 "dp-bug.gdb")
80348 at gdb/cli/cli-cmds.c:692
80349 #23 0x0000000000544557 in source_script_with_search (file=0x7fffffffd91a "dp-bug.gdb", from_tty=1, search_path=0)
80350 at gdb/cli/cli-cmds.c:750
80351 #24 0x00000000005445cf in source_script (file=0x7fffffffd91a "dp-bug.gdb", from_tty=1)
80352 at gdb/cli/cli-cmds.c:759
80353 #25 0x00000000007cf6d9 in catch_command_errors (command=0x5445aa <source_script(char const*, int)>,
80354 arg=0x7fffffffd91a "dp-bug.gdb", from_tty=1, do_bp_actions=false)
80356 #26 0x00000000007cf85d in execute_cmdargs (cmdarg_vec=0x7fffffffd1b0, file_type=CMDARG_FILE, cmd_type=CMDARG_COMMAND,
80357 ret=0x7fffffffd18c) at gdb/main.c:615
80358 #27 0x00000000007d0c8e in captured_main_1 (context=0x7fffffffd3f0)
80360 #28 0x00000000007d0eba in captured_main (data=0x7fffffffd3f0)
80362 #29 0x00000000007d0f25 in gdb_main (args=0x7fffffffd3f0)
80364 #30 0x00000000004186dd in main (argc=5, argv=0x7fffffffd508)
80367 There are two frames for bpstat_do_actions_1(), one at frame #16 and
80368 the other at frame #0. The one at frame #16 is processing the actions
80369 for Breakpoint 2, which is a 'continue'. The one at frame #0 is attempting
80370 to process the dprintf breakpoint action. However, at this point,
80371 the value of 'executing_breakpoint_commands' is 1, forcing an early
80372 return, i.e. prior to executing the command(s) associated with the dprintf
80375 For the sake of comparison, this is what the stack looks like when hitting
80376 the dprintf breakpoint for the second time when issuing the 'run'
80377 command from the GDB prompt.
80379 Thread 1 "gdb" hit Breakpoint 3, bpstat_do_actions_1 (bsp=0x7fffffffccd8)
80380 at /ironwood1/sourceware-git/f34-master/bld/../../worktree-master/gdb/breakpoint.c:4431
80381 4431 if (executing_breakpoint_commands)
80383 #0 bpstat_do_actions_1 (bsp=0x7fffffffccd8)
80384 at gdb/breakpoint.c:4431
80385 #1 0x00000000004d8bc6 in dprintf_after_condition_true (bs=0x16b0290)
80386 at gdb/breakpoint.c:13048
80387 #2 0x00000000004c5caa in bpstat_stop_status (aspace=0x116dbc0, bp_addr=0x40116e, thread=0x13f0e60, ws=0x7fffffffd138,
80388 stop_chain=0x16b0290) at gdb/breakpoint.c:5498
80389 #3 0x0000000000768d98 in handle_signal_stop (ecs=0x7fffffffd110)
80390 at gdb/infrun.c:6172
80391 #4 0x00000000007678d3 in handle_inferior_event (ecs=0x7fffffffd110)
80392 at gdb/infrun.c:5662
80393 #5 0x0000000000763cd5 in fetch_inferior_event ()
80394 at gdb/infrun.c:4060
80395 #6 0x0000000000746d7d in inferior_event_handler (event_type=INF_REG_EVENT)
80396 at gdb/inf-loop.c:41
80397 #7 0x00000000007a702f in handle_target_event (error=0, client_data=0x0)
80398 at gdb/linux-nat.c:4207
80399 #8 0x0000000000b8cd6e in gdb_wait_for_event (block=block@entry=0)
80400 at gdbsupport/event-loop.cc:701
80401 #9 0x0000000000b8d032 in gdb_wait_for_event (block=0)
80402 at gdbsupport/event-loop.cc:597
80403 #10 gdb_do_one_event () at gdbsupport/event-loop.cc:212
80404 #11 0x00000000007cf512 in start_event_loop ()
80406 #12 0x00000000007cf631 in captured_command_loop ()
80408 #13 0x00000000007d0ebf in captured_main (data=0x7fffffffd3f0)
80410 #14 0x00000000007d0f25 in gdb_main (args=0x7fffffffd3f0)
80412 #15 0x00000000004186dd in main (argc=5, argv=0x7fffffffd508)
80415 This relatively short backtrace is due to the current UI's async field
80418 Yet another thing to be aware of regarding this problem is the
80419 difference in the way that commands associated to dprintf breakpoints
80420 versus regular breakpoints are handled. While they both use a command
80421 list associated with the breakpoint, regular breakpoints will place
80422 the commands to be run on the bpstat chain constructed in
80423 bp_stop_status(). These commands are run later on. For dprintf
80424 breakpoints, commands are run via the 'after_condition_true' function
80425 pointer directly from bpstat_stop_status(). (The 'commands' field in
80426 the bpstat is cleared in dprintf_after_condition_true(). This
80427 prevents the dprintf commands from being run again later on when other
80428 commands on the bpstat chain are processed.)
80430 Another thing that I noticed is that dprintf breakpoints are the only
80431 type of breakpoint which use 'after_condition_true'. This suggests
80432 that one possible way of fixing this problem, that of making dprintf
80433 breakpoints work more like regular breakpoints, probably won't work.
80434 (I must admit, however, that my understanding of this code isn't
80435 complete enough to say why. I'll trust that whoever implemented it
80436 had a good reason for doing it this way.)
80438 The comment referenced earlier regarding 'executing_breakpoint_commands'
80439 states that the reason for checking this variable is to avoid
80440 potential endless recursion when a 'source' command appears in
80441 bs->commands. We know that a dprintf command is constrained to either
80442 1) execution of a GDB printf command, 2) an inferior function call of
80443 a printf-like function, or 3) execution of an agent-printf command.
80444 Therefore, infinite recursion due to a 'source' command cannot happen
80445 when executing commands upon hitting a dprintf breakpoint.
80447 I chose to fix this problem by having dprintf_after_condition_true()
80448 directly call execute_control_commands(). This means that it no
80449 longer attempts to go through bpstat_do_actions_1() avoiding the
80450 infinite recursion check for potential 'source' commands on the
80451 command chain. I think it simplifies this code a little bit too, a
80456 * breakpoint.c (dprintf_after_condition_true): Don't call
80457 bpstat_do_actions_1(). Call execute_control_commands()
80460 2021-11-10 Alan Modra <amodra@gmail.com>
80462 Re: Add --unicode option
80463 * objdump: Whitespace fixes.
80464 (long_options): Correct "ctf" entry.
80466 2021-11-10 Alan Modra <amodra@gmail.com>
80468 Re: Add --unicode option
80469 At low optimisation levels gcc may warn.
80471 * strings.c (print_unicode_stream_body): Avoid bogus "may be
80472 used unitialised" warning.
80474 2021-11-10 GDB Administrator <gdbadmin@sourceware.org>
80476 Automatic date update in version.in
80478 2021-11-09 Alan Modra <amodra@gmail.com>
80480 PR28543, readelf entered an infinite loop
80481 This little tweak terminates fuzzed binary readelf output a little
80485 * dwarf.c (read_and_display_attr_value): Consume a byte when
80486 form is unrecognized.
80488 2021-11-09 Alan Modra <amodra@gmail.com>
80490 PR28542, Undefined behaviours in readelf.c
80492 * readelf.c (dump_relocations): Check that section headers have
80493 been read before attempting to access section name.
80494 (print_dynamic_symbol): Likewise.
80495 (process_mips_specific): Delete dead code.
80497 2021-11-09 Pedro Alves <pedro@palves.net>
80499 gdb::array_view slicing/container selftest - test std::array too
80500 Change-Id: I2141b0b8a09f6521a59908599eb5ba1a19b18dc6
80502 2021-11-09 Simon Marchi <simon.marchi@efficios.com>
80504 gdb.debuginfod/fetch_src_and_symbols.exp: fix when GDB is built with AddressSanitizer
80505 This test fails for me, showing:
80507 ERROR: tcl error sourcing /home/smarchi/src/binutils-gdb/gdb/testsuite/gdb.debuginfod/fetch_src_and_symbols.exp.
80508 ERROR: This GDB was configured as follows:
80509 configure --host=x86_64-pc-linux-gnu --target=x86_64-pc-linux-gnu
80510 --with-auto-load-dir=$debugdir:$datadir/auto-load
80511 --with-auto-load-safe-path=$debugdir:$datadir/auto-load
80512 ... and much more ...
80514 The problem is that TCL's exec throws an error as soon as the exec'ed
80515 process outputs on stderr. When GDB is built with ASan, it prints some
80516 warnings about pre-existing signal handlers:
80518 warning: Found custom handler for signal 7 (Bus error) preinstalled.
80519 warning: Found custom handler for signal 8 (Floating point exception) preinstalled.
80520 warning: Found custom handler for signal 11 (Segmentation fault) preinstalled.
80522 Pass --quiet to GDB to avoid these warnings.
80524 Change-Id: I3751d89b9b1df646da19149d7cb86775e2d3e80f
80526 2021-11-09 Tom Tromey <tromey@adacore.com>
80528 Correctly handle DW_LLE_start_end
80529 When the code to handle DW_LLE_start_end was added (as part of some
80530 DWARF 5 work), it was written to add the base address. However, this
80531 seems incorrect -- the DWARF standard describes this as an address,
80532 not an offset from the base address.
80534 This patch changes a couple of spots in dwarf2/loc.c to fix this
80535 problem. It then changes decode_debug_loc_addresses to return
80536 DEBUG_LOC_OFFSET_PAIR instead, which preserves the previous semantics.
80538 This only showed up on the RISC-V target internally, due to the
80539 combination of DWARF 5 and a newer version of GCC. I've updated a
80540 couple of existing loclists test cases to demonstrate the bug.
80542 2021-11-09 Tom Tromey <tromey@adacore.com>
80545 The rhES5 build failed due to an upstream import a while back. The
80546 bug here is that, while the 'personality' function exists,
80547 ADDR_NO_RANDOMIZE is only defined in <linux/personality.h>, not
80548 <sys/personality.h>.
80550 However, <linux/personality.h> does not declare the 'personality'
80551 function, and <sys/personality.h> and <linux/personality.h> cannot
80554 This patch restores one of the removed configure checks and updates
80555 the code to check it.
80557 We had this as a local patch at AdaCore, because it seemed like there
80558 was no interest upstream. However, now it turns out that this fixes
80559 PR build/28555, so I'm sending it now.
80561 2021-11-09 H.J. Lu <hjl.tools@gmail.com>
80563 doc/ctf-spec.texi: Remove "@validatemenus off"
80564 Remove @validatemenus from ctf-spec.texi, which has been removed from
80567 commit a16dd1a9ece08568a1980b9a65a3a9090717997f
80568 Author: Gavin Smith <gavinsmith0123@gmail.com>
80569 Date: Mon Oct 12 16:32:37 2020 +0100
80572 (Writing a Menu, Customization Variables for @-Commands)
80574 * doc/refcard/txirefcard.tex
80575 Remove @validatemenus.
80576 * tp/Texinfo/XS/Makefile.am (command_ids.h): Use gawk instead
80577 of awk. Avoid discouraged "$p" usage, using "$(p)" instead.
80578 * tp/Texinfo/XS/configure.ac: Check for gawk.
80580 commit 128acab3889b51809dc3bd3c6c74b61d13f7f5f4
80581 Author: Gavin Smith <gavinsmith0123@gmail.com>
80582 Date: Thu Jan 3 14:51:53 2019 +0000
80586 * doc/refcard/txirefcard.tex: @setfilename is no longer
80587 mandatory. Do not mention @validatemenus or explicitly giving
80588 @node pointers, as these are not very important features.
80591 * doc/ctf-spec.texi: Remove "@validatemenus off".
80593 2021-11-09 Nick Clifton <nickc@redhat.com>
80595 Add --unicode option to control how unicode characters are handled by display tools.
80596 * nm.c: Add --unicode option to control how unicode characters are
80598 * objdump.c: Likewise.
80599 * readelf.c: Likewise.
80600 * strings.c: Likewise.
80601 * binutils.texi: Document the new feature.
80602 * NEWS: Document the new feature.
80603 * testsuite/binutils-all/unicode.exp: New file.
80604 * testsuite/binutils-all/nm.hex.unicode
80605 * testsuite/binutils-all/strings.escape.unicode
80606 * testsuite/binutils-all/objdump.highlight.unicode
80607 * testsuite/binutils-all/readelf.invalid.unicode
80609 2021-11-09 Mike Frysinger <vapier@gentoo.org>
80611 sim: sh: simplify testsuite a bit
80612 Switch from the centralized list in the exp file to each test declaring
80613 its own requirements which they're already (mostly) doing. This will
80614 increase coverage slightly by running more tests in more configurations
80615 since the hardcoded exp list was a little out of date.
80617 We have to mark the psh* tests as shdsp only (to match what the exp
80618 file was doing), mark the fsca & fsrra tests as failing (since they
80619 weren't even being run by the exp file), and to fix the expected
80620 output & status of the fail test.
80622 2021-11-09 Mike Frysinger <vapier@gentoo.org>
80624 sim: cris: clean up missing func prototype warnings
80625 Move some unused funcs under existing #if 0 protection, mark a few
80626 local funcs as static, and add missing prototypes for the rest which
80627 are used from other files. This fixes all the fatal warnings in the
80628 mloop files so we can turn -Werror on here fully.
80630 2021-11-09 GDB Administrator <gdbadmin@sourceware.org>
80632 Automatic date update in version.in
80634 2021-11-08 Lancelot SIX <lsix@lancelotsix.com>
80636 Improve gdb::array_view ctor from contiguous containers
80637 While reading the interface of gdb::array_view, I realized that the
80638 constructor that builds an array_view on top of a contiguous container
80639 (such as std::vector, std::array or even gdb::array_view) can be
80642 Lets consider the following code sample:
80646 Parent (int a): a { a } {}
80650 std::ostream &operator<< (std::ostream& os, const Parent & p)
80651 { os << "Parent {a=" << p.a << "}"; return os; }
80653 struct Child : public Parent
80655 Child (int a, int b): Parent { a }, b { b } {}
80659 std::ostream &operator<< (std::ostream& os, const Child & p)
80660 { os << "Child {a=" << p.a << ", b=" << p.b << "}"; return os; }
80662 template <typename T>
80663 void print (const gdb::array_view<const T> &p)
80665 std::for_each (p.begin (), p.end (), [](const T &p) { std::cout << p << '\n'; });
80668 Then with the current interface nothinng prevents this usage of
80669 array_view to be done:
80671 const std::array<Child, 3> elts = {
80676 print_all<Parent> (elts);
80678 This compiles fine and produces the following output:
80684 which is obviously wrong. There is nowhere in memory a Parent-like
80685 object for which the A member is 2 and this call to print_all<Parent>
80686 shold not compile at all (calling print_all<Child> is however fine).
80688 This comes down to the fact that a Child* is convertible into a Parent*,
80689 and that an array view is constructed to a pointer to the first element
80690 and a size. The valid type pointed to that can be used with this
80691 constructor are restricted using SFINAE, which requires that a
80692 pointer to a member into the underlying container can be converted into a
80693 pointer the array_view's data type.
80695 This patch proposes to change the constraints on the gdb::array_view
80696 ctor which accepts a container now requires that the (decayed) type of
80697 the elements in the container match the (decayed) type of the array_view
80700 Applying this change required minimum adjustment in GDB codebase, which
80701 are also included in this patch.
80703 Tested by rebuilding.
80705 2021-11-08 Lancelot SIX <lsix@lancelotsix.com>
80707 Add a const version of gdb_argv:as_array_view
80708 This commits adds const versions for the GET and AS_ARRAX_VIEW methods
80709 of gdb_argv. Those methods will be required in the following patch of
80712 2021-11-08 Simon Marchi <simon.marchi@polymtl.ca>
80714 gdb: fix nulltr -> nullptr typo
80715 Change-Id: I04403bd85ec3fa75ea14130d68daba675a2a8aeb
80717 2021-11-08 Simon Marchi <simon.marchi@polymtl.ca>
80719 gdb: tweak scoped_disable_commit_resumed uses when resuming all threads in non-stop
80720 When doing "continue -a" in non-stop mode, each thread is individually
80721 resumed while the commit resumed state is enabled. This forces the
80722 target to commit each resumption immediately, instead of being able to
80725 The reason is that there is no scoped_disable_commit_resumed around the
80726 loop over threads in continue_1, when "non_stop && all_threads" is true.
80727 Since the proceed function is called once for each thread, the
80728 scoped_disable_commit_resumed in proceed therefore forces commit-resumed
80729 between each thread resumption. Add the necessary
80730 scoped_disable_commit_resumed in continue_1 to avoid that.
80732 I looked at the MI side of things, the function exec_continue, and found
80733 that it was correct. There is a similar iteration over threads, and
80734 there is a scoped_disable_commit_resumed at the function scope. This is
80735 not wrong, but a bit more than we need. The branches that just call
80736 continue_1 do not need it, as continue_1 takes care of disabling commit
80737 resumed. So, move the scoped_disable_commit_resumed to the inner scope
80738 where we iterate on threads and proceed them individually.
80740 Here's an example debugging a multi-threaded program attached by
80741 gdbserver (debug output trimmed for brevity):
80743 $ ./gdb -nx -q --data-directory=data-directory -ex "set non-stop" -ex "tar rem :1234"
80744 (gdb) set debug remote
80745 (gdb) set debug infrun
80748 [infrun] proceed: enter
80749 [infrun] scoped_disable_commit_resumed: reason=proceeding
80750 [remote] Sending packet: $vCont;c:p14388.14388#90
80751 [infrun] reset: reason=proceeding
80752 [infrun] maybe_set_commit_resumed_all_targets: enabling commit-resumed for target remote
80753 [infrun] maybe_call_commit_resumed_all_targets: calling commit_resumed for target remote
80754 [infrun] proceed: exit
80755 [infrun] proceed: enter
80756 [infrun] scoped_disable_commit_resumed: reason=proceeding
80757 [remote] Sending packet: $vCont;c:p14388.1438a#b9
80758 [infrun] reset: reason=proceeding
80759 [infrun] maybe_set_commit_resumed_all_targets: enabling commit-resumed for target remote
80760 [infrun] maybe_call_commit_resumed_all_targets: calling commit_resumed for target remote
80761 [infrun] proceed: exit
80762 ... and so on for each thread ...
80764 Notice how we send one vCont;c for each thread. With the patch applied, we
80765 send a single vCont;c at the end:
80767 [infrun] scoped_disable_commit_resumed: reason=continue all threads in non-stop
80768 [infrun] proceed: enter
80769 [infrun] scoped_disable_commit_resumed: reason=proceeding
80770 [infrun] reset: reason=proceeding
80771 [infrun] proceed: exit
80772 [infrun] clear_proceed_status_thread: Thread 85790.85792
80773 [infrun] proceed: enter
80774 [infrun] scoped_disable_commit_resumed: reason=proceeding
80775 [infrun] reset: reason=proceeding
80776 [infrun] proceed: exit
80777 ... proceeding threads individually ...
80778 [infrun] reset: reason=continue all threads in non-stop
80779 [infrun] maybe_set_commit_resumed_all_targets: enabling commit-resumed for target remote
80780 [infrun] maybe_call_commit_resumed_all_targets: calling commit_resumed for target remote
80781 [remote] Sending packet: $vCont;c#a8
80783 Change-Id: I331dd2473c5aa5114f89854196fed2a8fdd122bb
80785 2021-11-08 Simon Marchi <simon.marchi@polymtl.ca>
80787 gdb: make dwarf2_find_containing_comp_unit take a dwarf2_per_bfd
80788 While reading another patch, I saw that this function didn't need to
80789 take a dwarf2_per_objfile, but could take a dwarf2_per_bfd instead.
80790 It doesn't change the behavior, but doing this shows that this function
80791 is objfile-independent (can work with only the shared per-bfd data).
80793 Change-Id: I58f9c9cef6688902e95226480285da2d0005d77f
80795 2021-11-08 Simon Marchi <simon.marchi@polymtl.ca>
80797 gdb: remove bpstat typedef, rename bpstats to bpstat
80798 I don't find that the bpstat typedef, which hides a pointer, is
80799 particularly useful. In fact, it confused me many times, and I just see
80800 it as something to remember that adds cognitive load. Also, with C++,
80801 we might want to be able to pass bpstats objects by const-reference, not
80802 necessarily by pointer.
80804 So, remove the bpstat typedef and rename struct bpstats to bpstat (since
80805 it represents one bpstat, it makes sense that it is singular).
80807 Change-Id: I52e763b6e54ee666a9e045785f686d37b4f5f849
80809 2021-11-08 Nick Alcock <nick.alcock@oracle.com>
80811 libctf: add CTF format specification
80812 It's been a long time since most of this was written: it's long past
80813 time to put it in the binutils source tree. It's believed correct and
80814 complete insofar as it goes: it documents format v3 (the current
80815 version) but not the libctf API or any earlier versions. (The
80816 earlier versions can be read by libctf but not generated by it, and you
80817 are highly unlikely ever to see an example of any of them.)
80820 2021-11-08 Nick Alcock <nick.alcock@oracle.com>
80822 * doc/ctf-spec.texi: New file.
80823 * configure.ac (MAKEINFO): Add.
80824 (BUILD_INFO): Likewise.
80825 (AC_CONFIG_FILES) [doc/Makefile]: Add.
80826 * Makefile.am [BUILD_INFO] (SUBDIRS): Add doc/.
80827 * doc/Makefile.am: New file.
80828 * doc/Makefile.in: Likewise.
80829 * configure: Regenerated.
80830 * Makefile.in: Likewise.
80832 2021-11-08 GDB Administrator <gdbadmin@sourceware.org>
80834 Automatic date update in version.in
80836 2021-11-07 Alan Modra <amodra@gmail.com>
80838 Correct ld script wildcard matching description
80839 Goes with commit 68bbb9f788d0
80841 * ld.texi (Input Section Wildcards): Delete paragraph incorrectly
80842 saying '*' does not match '/'.
80844 2021-11-07 Mike Frysinger <vapier@gentoo.org>
80846 sim: sh: fix conversion of PC to an integer
80847 On LLP64 targets where sizeof(long) != sizeof(void*), this code fails:
80848 sim/sh/interp.c:704:24: error: cast from pointer to integer of different size -Werror=pointer-to-int-cast]
80849 704 | do { memstalls += ((((long) PC & 3) != 0) ? (n) : ((n) - 1)); } while (0)
80852 Since this code simply needs to check alignment, cast it using uintptr_t
80853 which is the right type for this.
80855 2021-11-07 Mike Frysinger <vapier@gentoo.org>
80857 sim: sh: clean up time(NULL) call
80858 Casting 0 to a pointer via (long *) doesn't work on LLP64 targets:
80859 error: cast from pointer to integer of different size [-Werror=pointer-to-int-cast]
80861 It's also unnecessary here. We can simply pass NULL like every other
80864 2021-11-07 Mike Frysinger <vapier@gentoo.org>
80866 sim: sh: break utime logic out of _WIN32 check
80867 Some _WIN32 targets provide utime (like mingw), so move the header
80868 include out from _WIN32 and under the specific HAVE_UTIME_H check.
80870 sim: sh: drop errno extern
80871 This isn't needed on any reasonable target nowadays, and no other
80872 source does this, and breaks with some mingw targets, so punt the
80875 sim: sh: fix isnan redefinition with mingw targets
80876 The code assumes that all _WIN32 targets are the same and can
80877 define isnan to _isnan. For mingw targets, they provide an isnan
80878 define already, so no need for the fallback here.
80880 sim: arm/bfin/rx: undefine page size from system headers
80881 Some targets (like cygwin) will export page size defines that clash
80882 with our local usage here. Undefine the system one to fix building
80885 sim: ppc: switch to libiberty environ.h
80886 Drop our compat code and assume environ exists to simplify.
80887 We did this for all other targets already, but ppc was missed.
80889 sim: sh: enable -Werror everywhere
80890 With most of the warnings fixed in interp.c, we can enable -Werror
80891 here too now. There are some -Wmaybe-uninitialized warnings still
80892 lurking that look legitimate, but we don't flag those are fatal,
80893 and I don't have the expertise to dive into each opcode to figure
80894 out the right way to clean them up.
80896 sim: sh: fix uninitialized variable usage with pdmsb
80897 This block of code relies on i to control which bits to test and how
80898 many times to run through the loop, but it never actually initialized
80899 it. There is another chunk of code that handles the pdmsb instruction
80900 that sets i to 16, so use that here too assuming it's correct. The
80901 programming manual suggests this is the right value too, but I am by
80902 no means a SuperH DSP expert. The tests are still passing though ...
80904 sim: sh: constify a few read-only lookup tables
80906 sim: sh: fix various parentheses warnings
80907 Add parentheses to a bunch of places where the compiler suggests we
80908 do to avoid confusion to most readers.
80910 sim: sh: fix unused-value warnings
80911 These macro expansions are deliberate in not using the computed value
80912 so that they trigger side-effects (possible invalid memory accesses)
80913 but while otherwise being noops. Add a (void) cast so the compiler
80914 knows these are intentional.
80916 sim: sh: rework register layout with anonymous unions & structs
80917 Now that we require C11, we can leverage anonymous unions & structs
80918 to fix a long standing issue with the SH register layout. The use
80919 of sregs.i for sh-dsp has generated a lot of compiler warnings about
80920 the access being out of bounds -- it only has 7 elements declared,
80921 but code goes beyond that to reach into the fregs that follow. But
80922 now that we have anonymous unions, we can reduce the nested names
80923 and have sregs cover all of these registers.
80925 2021-11-07 GDB Administrator <gdbadmin@sourceware.org>
80927 Automatic date update in version.in
80929 2021-11-06 Tiezhu Yang <yangtiezhu@loongson.cn>
80931 sim: mips: use sim_fpu_to{32,64}u to fix build warnings
80932 Since the first argument type is unsigned32 or unsigned64, just use
80933 sim_fpu_to{32,64}u instead of sim_fpu_to{32,64}i to fix the following
80937 .../sim/mips/cp1.c: In function 'convert':
80938 .../sim/mips/cp1.c:1425:32: warning: pointer targets in passing argument 1 of 'sim_fpu_to32i' differ in signedness [-Wpointer-sign]
80939 status |= sim_fpu_to32i (&result32, &wop, round);
80941 In file included from .../sim/mips/sim-main.h:67,
80942 from .../sim/mips/cp1.c:46:
80943 .../sim/mips/../common/sim-fpu.h:270:22: note: expected 'signed32 *' {aka 'int *'} but argument is of type 'unsigned32 *' {aka 'unsigned int *'}
80944 INLINE_SIM_FPU (int) sim_fpu_to32i (signed32 *i, const sim_fpu *f,
80946 .../sim/mips/cp1.c:1429:32: warning: pointer targets in passing argument 1 of 'sim_fpu_to64i' differ in signedness [-Wpointer-sign]
80947 status |= sim_fpu_to64i (&result64, &wop, round);
80949 In file included from .../sim/mips/sim-main.h:67,
80950 from .../sim/mips/cp1.c:46:
80951 .../sim/mips/../common/sim-fpu.h:274:22: note: expected 'signed64 *' {aka 'long int *'} but argument is of type 'unsigned64 *' {aka 'long unsigned int *'}
80952 INLINE_SIM_FPU (int) sim_fpu_to64i (signed64 *i, const sim_fpu *f,
80954 .../sim/mips/cp1.c: In function 'convert_ps':
80955 .../sim/mips/cp1.c:1528:34: warning: pointer targets in passing argument 1 of 'sim_fpu_to32i' differ in signedness [-Wpointer-sign]
80956 status_u |= sim_fpu_to32i (&res_u, &wop_u, round);
80958 In file included from .../sim/mips/sim-main.h:67,
80959 from .../sim/mips/cp1.c:46:
80960 .../sim/mips/../common/sim-fpu.h:270:22: note: expected 'signed32 *' {aka 'int *'} but argument is of type 'unsigned32 *' {aka 'unsigned int *'}
80961 INLINE_SIM_FPU (int) sim_fpu_to32i (signed32 *i, const sim_fpu *f,
80963 .../sim/mips/cp1.c:1529:34: warning: pointer targets in passing argument 1 of 'sim_fpu_to32i' differ in signedness [-Wpointer-sign]
80964 status_l |= sim_fpu_to32i (&res_l, &wop_l, round);
80966 In file included from .../sim/mips/sim-main.h:67,
80967 from .../sim/mips/cp1.c:46:
80968 .../sim/mips/../common/sim-fpu.h:270:22: note: expected 'signed32 *' {aka 'int *'} but argument is of type 'unsigned32 *' {aka 'unsigned int *'}
80969 INLINE_SIM_FPU (int) sim_fpu_to32i (signed32 *i, const sim_fpu *f,
80972 2021-11-06 Alan Modra <amodra@gmail.com>
80975 Newer versions of bison emit a prototype for yyerror
80976 void yyerror (const char *);
80977 This clashes with some of our old code that declares yyerror to return
80978 an int. Fix that in most cases by modernizing yyerror. bfin-parse.y
80979 uses the return value all over the place, so for there disable
80980 generation of the prototype as specified by posix.
80983 * arparse.y (yyerror): Return void.
80984 * dlltool.c (yyerror): Likewise.
80985 * dlltool.h (yyerror): Likewise.
80986 * sysinfo.y (yyerror): Likewise.
80987 * windmc.h (yyerror): Likewise.
80988 * mclex.c (mc_error): Extract from ..
80989 (yyerror): ..here, both now returning void.
80991 * config/bfin-parse.y (yyerror): Define.
80992 (yyerror): Make static.
80993 * itbl-parse.y (yyerror): Return void.
80995 * deffilep.y (def_error): Return void.
80997 2021-11-06 Alan Modra <amodra@gmail.com>
80999 ubsan: undefined shift in mach-o.c
81000 This one was logically wrong too. If file_ptr was 64 bits, then -1U
81001 is extended to 0x00000000ffffffff, probably not what was intended
81004 * mach-o.c (FILE_ALIGN): Correct expression.
81006 2021-11-06 Fangrui Song <maskray@google.com>
81008 readelf: Support RELR in -S and -d and output
81009 readelf -r dumping support is not added in this patch.
81012 * elf/common.h: Add SHT_RELR, DT_RELR{,SZ,ENT}
81014 * elf.c (_bfd_elf_print_private_bfd_data): Add DT_RELR{,SZ,ENT}.
81016 * readelf.c (get_dynamic_type): Add DT_RELR{,SZ,ENT}.
81017 (get_section_type_name): Add SHT_RELR.
81019 2021-11-06 Fangrui Song <maskray@google.com>
81021 readelf: Make DT_PREINIT_ARRAYSZ's output style match DT_INIT_ARRAYSZ
81022 The output now looks like:
81024 - 0x0000000000000021 (PREINIT_ARRAYSZ) 0x10
81025 + 0x0000000000000021 (PREINIT_ARRAYSZ) 16 (bytes)
81026 0x0000000000000019 (INIT_ARRAY) 0xbefc90
81027 0x000000000000001b (INIT_ARRAYSZ) 536 (bytes)
81029 * readelf.c (process_dynamic_section): Handle DT_PREINIT_ARRAYSZ.
81031 2021-11-06 Mike Frysinger <vapier@gentoo.org>
81033 sim: clarify license text via COPYING file
81034 The project has been using GPL v3 for a while now in the source files,
81035 and the arm & ppc ports have carried a copy of the COPYING file. Lets
81036 move those up to the top sim dir like other projects to make it clear.
81038 Also drop the ppc/COPYING.LIB as it's not really referenced by any
81039 source as everything is GPL v3.
81041 2021-11-06 GDB Administrator <gdbadmin@sourceware.org>
81043 Automatic date update in version.in
81045 2021-11-05 Tom Tromey <tom@tromey.com>
81047 Introduce make_unique_xstrndup
81048 This adds a new make_unique_xstrndup function, which is the "n"
81049 analogue of make_unique_xstrdup. It also updates a couple existing
81050 places to use this function.
81052 2021-11-05 Pedro Alves <pedro@palves.net>
81054 Avoid /proc/pid/mem races (PR 28065)
81055 PR 28065 (gdb.threads/access-mem-running-thread-exit.exp intermittent
81056 failure) shows that GDB can hit an unexpected scenario -- it can
81057 happen that the kernel manages to open a /proc/PID/task/LWP/mem file,
81058 but then reading from the file returns 0/EOF, even though the process
81059 hasn't exited or execed.
81061 "0" out of read/write is normally what you get when the address space
81062 of the process the file was open for is gone, because the process
81063 execed or exited. So when GDB gets the 0, it returns memory access
81064 failure. In the bad case in question, the process hasn't execed or
81065 exited, so GDB fails a memory access when the access should have
81068 GDB has code in place to gracefully handle the case of opening the
81069 /proc/PID/task/LWP/mem just while the LWP is exiting -- most often the
81070 open fails with EACCES or ENOENT. When it happens, GDB just tries
81071 opening the file for a different thread of the process. The testcase
81072 is written such that it stresses GDB's logic of closing/reopening the
81073 /proc/PID/task/LWP/mem file, by constantly spawning short lived
81076 However, there's a window where the kernel manages to find the thread,
81077 but the thread exits just after and clears its address space pointer.
81078 In this case, the kernel creates a file successfully, but the file
81079 ends up with no address space associated, so a subsequent read/write
81080 returns 0/EOF too, just like if the whole process had execed or
81081 exited. This is the case in question that GDB does not handle.
81083 Oleg Nesterov gave this suggestion as workaround for that race:
81085 gdb can open(/proc/pid/mem) and then read (say) /proc/pid/statm.
81086 If statm reports something non-zero, then open() was "successfull".
81088 I think that might work. However, I didn't try it, because I realized
81089 we have another nasty race that that wouldn't fix.
81091 The other race I realized is that because we close/reopen the
81092 /proc/PID/task/LWP/mem file when GDB switches to a different inferior,
81093 then it can happen that GDB reopens /proc/PID/task/LWP/mem just after
81094 a thread execs, and before GDB has seen the corresponding exec event.
81095 I.e., we can open a /proc/PID/task/LWP/mem file accessing the
81096 post-exec address space thinking we're accessing the pre-exec address
81099 A few months back, Simon, Oleg and I discussed a similar race:
81101 [Bug gdb/26754] Race condition when resuming threads and one does an exec
81102 https://sourceware.org/bugzilla/show_bug.cgi?id=26754
81104 The solution back then was to make the kernel fail any ptrace
81105 operation until the exec event is consumed, with this kernel commit:
81107 commit dbb5afad100a828c97e012c6106566d99f041db6
81108 Author: Oleg Nesterov <oleg@redhat.com>
81109 AuthorDate: Wed May 12 15:33:08 2021 +0200
81110 Commit: Linus Torvalds <torvalds@linux-foundation.org>
81111 CommitDate: Wed May 12 10:45:22 2021 -0700
81113 ptrace: make ptrace() fail if the tracee changed its pid unexpectedly
81115 This however, only applies to ptrace, not to the /proc/pid/mem file
81116 opening case. Also, even if it did apply to the file open case, we
81117 would want to support current kernels until such a fix is more wide
81120 So all in all, this commit gives up on the idea of only ever keeping
81121 one /proc/pid/mem file descriptor open. Instead, make GDB open a
81122 /proc/pid/mem per inferior, and keep it open until the inferior exits,
81123 is detached or execs. Make GDB open the file right after the inferior
81124 is created or is attached to or forks, at which point we know the
81125 inferior is stable and stopped and isn't thus going to exec, or have a
81126 thread exit, and so the file open won't fail (unless the whole process
81127 is SIGKILLed from outside GDB, at which point it doesn't matter
81128 whether we open the file).
81130 This way, we avoid both races described above, at the expense of using
81131 more file descriptors (one per inferior).
81133 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28065
81134 Change-Id: Iff943b95126d0f98a7973a07e989e4f020c29419
81136 2021-11-05 Andrew Burgess <aburgess@redhat.com>
81138 gdb/testsuite: use gdb_get_line_number
81139 Replaces a hard coded line number with a use of gdb_get_line_number.
81141 I suspect that the line number has, over time, come adrift from where
81142 it was supposed to be stopping. When the test was first added, line
81143 770 pointed at the final 'return 0' in function main. Over time, as
81144 things have been added, line 770 now points at some random location in
81145 the middle of main.
81147 So, I've marked the 'return 0' with a comment, and now the test will
81150 I also removed an old comment from 1997 talking about how these tests
81151 will only pass with the HP compiler, followed by an additional comment
81152 from 2000 saying that the tests now pass with GCC.
81154 I get the same results before and after this change.
81156 2021-11-05 Alan Modra <amodra@gmail.com>
81158 PR28541, unstable cie offset in the output of readelf
81159 Calculating "0 - pointer" can indeed result in seeming randomness as
81160 the pointer address varies.
81163 * dwarf.c (display_debug_frames): Don't print cie offset when
81164 invalid, print "invalid" instead. Remove now redundant warning.
81166 2021-11-05 Alan Modra <amodra@gmail.com>
81168 Missing va_end in aarch64-dis.c
81169 * aarch64-dis.c (extract_fields): Invoke va_end.
81171 2021-11-05 Alan Modra <amodra@gmail.com>
81173 PR28530, Hang in objdump on machine with 196GB RAM
81174 Investigating the PR28530 testcase, which has a fuzzed compression
81175 header with an enormous size, I noticed that decompress_contents is
81176 broken when the size doesn't fit in strm.avail_out. It wouldn't be
81177 too hard to support larger sizes (patches welcome!) but for now just
81178 stop decompress_contents from returning rubbish.
81181 * compress.c (decompress_contents): Fail when uncompressed_size
81183 (bfd_init_section_decompress_status): Likewise.
81185 2021-11-05 Alan Modra <amodra@gmail.com>
81187 asan: alpha-vms: objdump buffer overflows
81188 * vms-alpha.c (evax_bfd_print_desc): Sanity check buffer access.
81189 (evax_bfd_print_valspec, evax_bfd_print_typspec): Likewise.
81190 (evax_bfd_print_dst): Likewise.
81192 2021-11-05 GDB Administrator <gdbadmin@sourceware.org>
81194 Automatic date update in version.in
81196 2021-11-04 Simon Marchi <simon.marchi@polymtl.ca>
81198 gdb: introduce "set index-cache enabled", deprecate "set index-cache on/off"
81199 The "set index-cache" command is used at the same time as a prefix
81200 command (prefix for "set index-cache directory", for example), and a
81201 boolean setting for turning the index-cache on and off. Even though I
81202 did introduce that, I now don't think it's a good idea to do something
81203 non-standard like this.
81205 First, there's no dedicated CLI command to show whether the index-cache
81206 is enabled, so it has to be custom output in the "show index-cache
81207 handler". Also, it means there's no good way a MI frontend can find out
81208 if the index-cache is enabled. "-gdb-show index-cache" doesn't show it
81209 in the MI output record:
81211 (gdb) interpreter-exec mi "-gdb-show index-cache"
81213 ~"The index cache is currently disabled.\n"
81214 ^done,showlist={option={name="directory",value="/home/simark/.cache/gdb"}}
81216 Fix this by introducing "set/show index-cache enabled on/off", regular
81217 boolean setting commands. Keep commands "set index-cache on" and "set
81218 index-cache off" as deprecated aliases of "set index-cache enabled",
81219 with respectively the default arguments "on" and "off".
81221 Update tests using "set index-cache on/off" to use the new command.
81222 Update the regexps in gdb.base/maint.exp to figure out whether the
81223 index-cache is enabled or not. Update the doc to mention the new
81226 Change-Id: I7d5aaaf7fd22bf47bd03e0023ef4fbb4023b37b3
81228 2021-11-04 Simon Marchi <simon.marchi@polymtl.ca>
81230 gdb: pass/return setting setter/getter scalar values by value
81231 The getter and setter in struct setting always receive and return values
81232 by const reference. This is not necessary for scalar values (like bool
81233 and int), but more importantly it makes it a bit annoying to write a
81234 getter, you have to use a scratch static variable or something similar
81235 that you can refer to:
81241 value = function_returning_bool ();
81245 Change the getter and setter function signatures to receive and return
81246 value by value instead of by reference, when the underlying data type is
81247 scalar. This means that string-based settings will still use
81248 references, but all others will be by value. The getter above would
81249 then be re-written as:
81254 return function_returning_bool ();
81257 This is useful for a patch later in this series that defines a boolean
81258 setting with a getter and a setter.
81260 Change-Id: Ieca3a2419fcdb75a6f75948b2c920b548a0af0fd
81262 2021-11-04 Simon Marchi <simon.marchi@polymtl.ca>
81264 gdb: remove command_class enum class_deprecated
81265 The class_deprecated enumerator isn't assigned anywhere, so remove it.
81266 Commands that are deprecated have cmd_list_element::cmd_deprecated set
81269 Change-Id: Ib35e540915c52aa65f13bfe9b8e4e22e6007903c
81271 2021-11-04 Simon Marchi <simon.marchi@polymtl.ca>
81273 gdb: remove unnecessary cmd_list_element::aliases nullptr checks
81274 Remove two unnecessary nullptr checks. If aliases is nullptr, then the
81275 for loops will simply be skipped.
81277 Change-Id: I9132063bb17798391f8d019af305383fa8e0229f
81279 2021-11-04 Simon Marchi <simon.marchi@polymtl.ca>
81281 gdbserver: re-generate configure
81282 I get some diffs when running autoconf in gdbserver, probably leftovers
81283 from commit 5dfe4bfcb969 ("Fix format_pieces selftest on Windows").
81284 Re-generate configure in that directory.
81286 Change-Id: Icdc9906af95fbaf1047a579914b2983f8ec5db08
81288 2021-11-04 H.J. Lu <hjl.tools@gmail.com>
81290 Revert "bfd: Always check sections with the corrupt size"
81291 This reverts commit e0f7ea91436dd308a094c4c101fd4169e8245a91.
81293 2021-11-04 H.J. Lu <hjl.tools@gmail.com>
81295 bfd: Always check sections with the corrupt size
81296 Always check sections with the corrupt size for non-MMO files. Skip MMO
81297 files for compress_status == COMPRESS_SECTION_NONE since MMO has special
81298 handling for COMPRESS_SECTION_NONE.
81301 * compress.c (bfd_get_full_section_contents): Always check
81302 sections with the corrupt size.
81304 2021-11-04 Nelson Chu <nelson.chu@sifive.com>
81306 RISC-V: Clarify the behavior of .option rvc or norvc.
81307 Add/Remove the rvc extension to/from the riscv_subsets once the
81308 .option rvc/norvc is set. So that we don't need to always check
81309 the riscv_opts.rvc in the riscv_subset_supports, just call the
81310 riscv_lookup_subset to search the subset list is enough.
81312 Besides, we will need to dump the instructions according to the
81313 elf architecture attributes. That means the dis-assembler needs
81314 to parse the architecture string from the elf attribute before
81315 dumping any instructions, and also needs to recognized the
81316 INSN_CLASS* classes from riscv_opcodes. Therefore, I suppose
81317 some functions will need to be moved from gas/config/tc-riscv.c
81318 to bfd/elfxx-riscv.c, including riscv_multi_subset_supports and
81319 riscv_subset_supports. This is one of the reasons why we need
81322 This patch passes the gcc/binutils regressions of rv32emc-elf,
81323 rv32i-elf, rv64gc-elf and rv64gc-linux toolchains.
81326 * elfxx-riscv.c (riscv_remove_subset): Remove the extension
81327 from the subset list.
81328 (riscv_update_subset): Add/Remove an extension to/from the
81329 subset list. This is used for the .option rvc or norvc.
81330 * elfxx-riscv.h: Added the extern bool riscv_update_subset.
81332 * config/tc-riscv.c (riscv_set_options): Removed the unused
81334 (riscv_opts): Likewise.
81335 (riscv_set_rve): Removed.
81336 (riscv_subset_supports): Removed the riscv_opts.rvc check.
81337 (riscv_set_arch): Don't need to call riscv_set_rve.
81338 (reg_lookup_internal): Call riscv_subset_supports to check
81339 whether the rve is supported.
81340 (s_riscv_option): Add/Remove the rvc extension to/from the
81341 subset list once the .option rvc/norvc is set.
81343 2021-11-04 Mike Frysinger <vapier@gentoo.org>
81345 sim: mips: fix missing prototype in multi-run generation
81346 The multi-run logic for mips involves a bit of codegen and rewriting
81347 of files to include per-architecture prefixes. That can result in
81348 files with missing prototypes which cause compiler errors. In the
81349 case of mips-sde-elf targets, we have:
81350 $srcdir/m16run.c -> $builddir/m16mips64r2_run.c
81351 sim_engine_run -> m16mips64r2_engine_run
81352 $srcdir/micromipsrun.c -> micromipsmicromips_run.c
81353 sim_engine_run -> micromips64micromips_engine_run
81355 micromipsmicromips_run.c:80:1: error: no previous prototype for 'micromips64micromips_engine_run' [-Werror=missing-prototypes]
81356 80 | micromips64micromips_engine_run (SIM_DESC sd, int next_cpu_nr, int nr_cpus,
81358 We generate headers for those prototypes in the configure script,
81359 but only include them in the generated multi-run.c file. Update the
81360 rewrite logic to turn the sim-engine.h include into the relevant
81361 generated engine include so these files also have their prototypes.
81362 $srcdir/m16run.c -> $builddir/m16mips64r2_run.c
81363 sim-engine.h -> m16mips64r2_engine.h
81364 $srcdir/micromipsrun.c -> micromipsmicromips_run.c
81365 sim-engine.h -> micromips64micromips_engine.h
81367 2021-11-04 Alan Modra <amodra@gmail.com>
81369 PR28540, segmentation fault on NULL byte_get
81371 * objdump.c (dump_bfd): Don't attempt load_separate_debug_files
81372 when byte_get is NULL.
81374 2021-11-04 GDB Administrator <gdbadmin@sourceware.org>
81376 Automatic date update in version.in
81378 2021-11-03 Mike Frysinger <vapier@gentoo.org>
81380 sim: ppc: inline common sim-fpu.c logic
81381 We will never bother building w/out a ../common/ sim directory,
81382 so drop ancient logic supporting that method.
81384 2021-11-03 Mike Frysinger <vapier@gentoo.org>
81386 sim: ppc: switch to common builds for callback objects
81387 We don't need to build this anymore ourselves since the common build
81388 includes it and produces the same object code. We also need to pull
81389 in the split constant modules after the refactoring and pulling them
81390 out of nltvals.def & targ-map.o. This doesn't matter for the sim
81391 directly, but does for gdb and other users of libsim.
81393 We also delete some conditional source tree logic since we already
81394 require this be the "new" combined tree with a ../common/ dir. This
81395 has been the case for decades at this point.
81397 2021-11-03 Simon Marchi <simon.marchi@polymtl.ca>
81399 gdb: fix gnu-nat build
81400 When building gnu-nat.c, we get:
81403 gnu-nat.c: In member function 'virtual void gnu_nat_target::create_inferior(const char*, const string&, char**, int)':
81404 gnu-nat.c:2117:13: error: 'struct inf' has no member named 'target_is_pushed'
81405 2117 | if (!inf->target_is_pushed (this))
81407 gnu-nat.c:2118:10: error: 'struct inf' has no member named 'push_target'
81408 2118 | inf->push_target (this);
81411 This is because of a confusion between the generic `struct inferior`
81412 variable and the gnu-nat-specific `struct inf` variable. Fix by
81413 referring to `inferior`, not `inf`.
81415 Adjust the comment on top of `struct inf` to clarify the purpose of that
81418 Co-Authored-By: Andrea Monaco <andrea.monaco@autistici.org>
81419 Change-Id: I2fe2f7f6ef61a38d79860fd262b08835c963fc77
81421 2021-11-03 Simon Marchi <simon.marchi@polymtl.ca>
81423 gdb/testsuite: set ASAN_OPTIONS=detect_leaks=0 while running tests
81424 We see some additional failures when running the testsuite against a GDB
81425 compiled with ASan, compared to a GDB compiled without ASan. Some of
81426 them are caused by the memory leak report shown by the GDB process when
81427 it exits, and the fact that it makes it exit with a non-zero exit code.
81429 I generally try to remember to set ASAN_OPTIONS=detect_leaks=0 in my
81430 environment when running the tests, but I don't always do it. I think
81431 it would be nice if the testsuite did it. I don't see any use to have
81432 leak detection when running the tests. That is, unless we ever have a
81433 test that ensures GDB doesn't leak memory, which isn't going to happen
81436 Here are some tests I found that were affected by this:
81438 gdb.base/batch-exit-status.exp
81439 gdb.base/many-headers.exp
81441 gdb.base/with-mf.exp
81442 gdb.dwarf2/gdb-add-index.exp
81443 gdb.dwarf2/gdb-add-index-symlink.exp
81444 gdb.dwarf2/imported-unit-runto-main.exp
81446 Change-Id: I784c7df8a13979eb96587f735c1d33ba2cc6e0ca
81448 2021-11-03 Tom Tromey <tromey@adacore.com>
81450 Use section name in warnings in display_debug_loc
81451 While looking at an apparently malformed executable with
81452 "readelf --debug-dump=loc", I got this warning:
81454 readelf: ./main: Warning: There is a hole [0x89 - 0x95] in .debug_loc section.
81456 However, the executable only has a .debug_loclists section.
81458 This patch fixes the warning messages in display_debug_loc to use the
81459 name of the section that is being processed.
81462 2021-11-03 Tom Tromey <tromey@adacore.com>
81464 * dwarf.c (display_debug_loc): Use section name in warnings.
81466 2021-11-03 Luis Machado <luis.machado@linaro.org>
81468 [AArch64] Make gdbserver register set selection dynamic
81469 The current register set selection mechanism for AArch64 is static, based
81470 on a pre-populated array of register sets.
81472 This means that we might potentially probe register sets that are not
81473 available. This is OK if the kernel errors out during ptrace, but probing the
81474 tag_ctl register, for example, does not result in a ptrace error if the kernel
81475 supports the tagged address ABI but not MTE (PR 28355).
81477 Making the register set selection dynamic, based on feature checks, solves
81478 this and simplifies the code a bit. It allows us to list all of the register
81479 sets only once, and pick and choose based on HWCAP/HWCAP2 or other properties.
81481 I plan to backport this fix to GDB 11 as well.
81483 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28355
81485 2021-11-03 Jan Kratochvil <jan.kratochvil@redhat.com>
81487 Fix LD_PRELOAD=/usr/lib64/libasan.so.6 gdb
81488 Currently for a binary compiled normally (without -fsanitize=address) but with
81489 LD_PRELOAD of ASAN one gets:
81491 $ ASAN_OPTIONS=detect_leaks=0:alloc_dealloc_mismatch=1:abort_on_error=1:fast_unwind_on_malloc=0 LD_PRELOAD=/usr/lib64/libasan.so.6 gdb
81492 =================================================================
81493 ==1909567==ERROR: AddressSanitizer: alloc-dealloc-mismatch (malloc vs operator delete []) on 0x602000001570
81494 #0 0x7f1c98e5efa7 in operator delete[](void*) (/usr/lib64/libasan.so.6+0xb0fa7)
81496 0x602000001570 is located 0 bytes inside of 2-byte region [0x602000001570,0x602000001572)
81497 allocated by thread T0 here:
81498 #0 0x7f1c98e5cd1f in __interceptor_malloc (/usr/lib64/libasan.so.6+0xaed1f)
81499 #1 0x557ee4a42e81 in operator new(unsigned long) (/usr/libexec/gdb+0x74ce81)
81500 SUMMARY: AddressSanitizer: alloc-dealloc-mismatch (/usr/lib64/libasan.so.6+0xb0fa7) in operator delete[](void*)
81501 ==1909567==HINT: if you don't care about these errors you may set ASAN_OPTIONS=alloc_dealloc_mismatch=0
81502 ==1909567==ABORTING
81504 Despite the code called properly operator new[] and operator delete[].
81505 But GDB's new-op.cc provides its own operator new[] which gets translated into
81506 malloc() (which gets recogized as operatore new(size_t)) but as it does not
81507 translate also operators delete[] Address Sanitizer gets confused.
81509 The question is how many variants of the delete operator need to be provided.
81510 There could be 14 operators new but there are only 4, GDB uses 3 of them.
81511 There could be 16 operators delete but there are only 6, GDB uses 2 of them.
81512 It depends on libraries and compiler which of the operators will get used.
81513 Currently being used:
81514 U operator new[](unsigned long)
81515 U operator new(unsigned long)
81516 U operator new(unsigned long, std::nothrow_t const&)
81517 U operator delete[](void*)
81518 U operator delete(void*, unsigned long)
81520 Tested on x86_64-linux.
81522 2021-11-03 Alan Modra <amodra@gmail.com>
81524 asan: dlltool buffer overflow: embedded NUL in string
81525 yyleng gives the pattern length, xstrdup just copies up to the NUL.
81526 So it is quite possible writing at an index of yyleng-2 overflows
81527 the xstrdup allocated string buffer. xmemdup quite handily avoids
81528 this problem, even writing the terminating NUL over the trailing
81529 quote. Use it in ldlex.l too where we'd already had a report of this
81530 problem and fixed it by hand, and to implement xmemdup0 in gas.
81533 * deflex.l (single and double quote strings): Use xmemdup.
81535 * as.h (xmemdup0): Use xmemdup.
81538 * ldlex.l (double quote string): Use xmemdup.
81540 2021-11-03 Mike Frysinger <vapier@gentoo.org>
81542 sim: mloop: mark a few conditionally used funcs as unused
81543 These are marked inline, so building w/gcc at higher optimization
81544 levels will automatically discard them. But building with -O0 will
81545 trigger unused function warnings, so fix that.
81547 The common before/after cover functions in the common mloop generator
81548 are not used by all architecture ports. Doesn't seem to be a hard
81549 requirement, so marking them optional (i.e. unused) is fine.
81551 The cris execute function is conditionally used depending on the
81552 fast-build mode settings, so mark it unused too.
81554 2021-11-03 Alan Modra <amodra@gmail.com>
81556 asan: assert (addr_ranges) <= (start)
81557 That assert would be more obvious if it were reported as
81558 "addr_ranges <= end_ranges". Fix that by using the obvious variable
81559 in the final loop. Stop the assertion by using a signed comparison:
81560 It's possible for the rounding up of the arange pointer to exceed the
81561 end of the block when the block size is fuzzed.
81563 * dwarf.c (display_debug_aranges): Use "end_ranges" in loop
81564 displaying ranges rather that "start". Simplify rounding up
81565 to 2*address_size boundary. Use signed comparison in loop.
81567 2021-11-03 Mike Frysinger <vapier@gentoo.org>
81569 sim: hoist cgen mloop rules up to common builds
81570 These rules don't depend on the target compiler settings, so hoist
81571 the build logic up to the common builds for better parallelization.
81573 We have to extend the genmloop.sh logic a bit to allow outputting
81574 to a subdir since it always assumed cwd was the right place.
81576 We leave the cgen maintainer rules in the subdirs for now as they
81577 aren't normally run, and they rely on cgen logic that has not yet
81580 2021-11-03 Mike Frysinger <vapier@gentoo.org>
81582 sim: hoist mn10300 & v850 igen rules up to common builds
81583 These rules don't depend on the target compiler settings, so hoist
81584 the build logic up to the common builds for better parallelization.
81586 We leave the mips rules in place as they depend on complicated
81587 arch-specific configure logic that needs to be untangled first.
81589 2021-11-03 Mike Frysinger <vapier@gentoo.org>
81591 sim: hoist gencode & opc2c build rules up to common builds
81592 These rules don't depend on the target compiler settings, so hoist
81593 the build logic up to the common builds for better parallelization.
81595 2021-11-03 Mike Frysinger <vapier@gentoo.org>
81597 opcodes: d10v: simplify header includes
81598 This file doesn't use anything from bfd (sysdep.h), so drop that
81599 include. This avoids an implicit dependency on the generated
81600 config.h which can be problematic for build-time tools.
81602 Also swap stdio.h for stddef.h. This file isn't doing or using
81603 any I/O structures, but it does need NULL.
81605 2021-11-03 Alan Modra <amodra@gmail.com>
81607 PR28523, ld.bfd created undefined symbols on ppc64
81608 This patch removes any fake (linker created) function descriptor
81609 symbol if its code entry symbol isn't dynamic, to ensure bogus dynamic
81610 symbols are not created. The change to func_desc_adjust requires that
81611 it be run only once, which means ppc64_elf_tls_setup can't call it for
81612 just a few selected symbols.
81615 * elf64-ppc.c (func_desc_adjust): If a function entry sym is
81616 not dynamic and has no plt entry, hide any associated fake
81617 function descriptor symbol.
81618 (ppc64_elf_edit): Move func_desc_adjust iteration over syms to..
81619 (ppc64_elf_tls_setup): ..here.
81621 2021-11-03 GDB Administrator <gdbadmin@sourceware.org>
81623 Automatic date update in version.in
81625 2021-11-02 Tom de Vries <tdevries@suse.de>
81627 [gdb/tdep, rs6000] Don't skip system call in skip_prologue
81628 I ran into a case where a breakpoint on _exit never triggered, because it was
81629 set past the end of the _exit prologue, past the end of the exit_group system
81630 call (which does not return).
81632 More concretely, the breakpoint was set at the last insn show here:
81634 Dump of assembler code for function _exit:
81635 0x00007ffff7e42ea0 <+0>: 12 00 4c 3c addis r2,r12,18
81636 0x00007ffff7e42ea4 <+4>: 60 43 42 38 addi r2,r2,17248
81637 0x00007ffff7e42ea8 <+8>: 00 00 00 60 nop
81638 0x00007ffff7e42eac <+12>: f8 ff e1 fb std r31,-8(r1)
81639 0x00007ffff7e42eb0 <+16>: 78 1b 7f 7c mr r31,r3
81640 0x00007ffff7e42eb4 <+20>: f0 ff c1 fb std r30,-16(r1)
81641 0x00007ffff7e42eb8 <+24>: ea 00 00 38 li r0,234
81642 0x00007ffff7e42ebc <+28>: a0 8b 22 e9 ld r9,-29792(r2)
81643 0x00007ffff7e42ec0 <+32>: 78 fb e3 7f mr r3,r31
81644 0x00007ffff7e42ec4 <+36>: 14 6a c9 7f add r30,r9,r13
81645 0x00007ffff7e42ec8 <+40>: 02 00 00 44 sc
81646 0x00007ffff7e42ecc <+44>: 26 00 00 7c mfcr r0
81647 0x00007ffff7e42ed0 <+48>: 00 10 09 74 andis. r9,r0,4096
81650 Fix this by treating system calls the same as branches in skip_prologue:
81651 by default, don't skip, such that the breakpoint is set at 0x00007ffff7e42eb8
81654 Tested on ppc64le-linux, on a power 8 machine.
81656 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28527
81658 2021-11-02 Tom de Vries <tdevries@zinfandel-3.arch.suse.de>
81660 [gdb/testsuite] Handle SIGILL in two gdb.arch powerpc test-cases
81661 On powerpc64le-linux, with test-case gdb.arch/powerpc-addpcis.exp I run into
81664 (gdb) PASS: gdb.arch/powerpc-addpcis.exp: get hexadecimal valueof "$r3"
81667 Program terminated with signal SIGILL, Illegal instruction.^M
81668 The program no longer exists.^M
81669 (gdb) PASS: gdb.arch/powerpc-addpcis.exp: set r4
81671 because it's a power9 insn, and I'm running on a power8 machine.
81673 Fix this by handling the SIGILL. Likewise in gdb.arch/powerpc-lnia.exp.
81675 Tested on powerpc64le-linux.
81677 2021-11-02 Andrew Burgess <andrew.burgess@embecosm.com>
81679 gdb/sim: update my email address
81682 * MAINTAINERS (Global Maintainers): Update my address.
81683 (Responsible Maintainers): Likewise.
81684 (Write After Approval): Likewise.
81688 * MAINTAINERS (Global Maintainers): Update my address.
81690 2021-11-02 Tom de Vries <tdevries@suse.de>
81692 [gdb/testsuite] Fix stepi test-cases with unix/-m32/-fPIE/-pie
81693 When running test-case gdb.base/step-indirect-call-thunk.exp with target board
81694 unix/-m32/-fPIE/-pie, I run into:
81697 0x5655552e 22 { /* inc.1 */^M
81699 0x56555530 22 { /* inc.1 */^M
81701 0x565555f7 in __x86.get_pc_thunk.ax ()^M
81702 (gdb) FAIL: gdb.base/step-indirect-call-thunk.exp: stepi into return thunk
81705 In contrast, with unix/-m32 we have instead:
81708 0x08048407 22 { /* inc.1 */^M
81710 23 return x + 1; /* inc.2 */^M
81712 0x0804840c 23 return x + 1; /* inc.2 */^M
81716 0x08048410 24 } /* inc.3 */^M
81718 0x0804848f in __x86_return_thunk ()^M
81719 (gdb) PASS: gdb.base/step-indirect-call-thunk.exp: stepi into return thunk
81722 The test-case doesn't expect to run into __x86.get_pc_thunk.ax, which is a
81723 PIC helper function for x86_64-linux.
81725 Fix this by insn-stepping through it.
81727 Likewise in a few other test-cases.
81729 Tested on x86_64-linux.
81731 2021-11-02 GDB Administrator <gdbadmin@sourceware.org>
81733 Automatic date update in version.in
81735 2021-11-01 Alan Modra <amodra@gmail.com>
81737 ARM: match armeb output for unwind-pacbti-m test
81738 * testsuite/gas/arm/unwind-pacbti-m.d: Match armeb output.
81740 2021-11-01 Bruno Larsen <blarsen@redhat.com>
81742 [gdb/doc]: Updated manpages to be consistent with help
81743 Updated manpages to be consistent with help information provided by the
81744 binary. The main changes are:
81746 * Making all long-form options have '--', instead of a single '-';
81747 * added most of the missing options to the manpage;
81748 * removed the information about using '+' instead of '-', since it
81749 doesn't seem to be supported anymore.
81751 This also fixes 2 upstream bugs:
81752 * https://sourceware.org/bugzilla/show_bug.cgi?id=23965; by adding
81753 --args to the manpage
81754 * https://sourceware.org/bugzilla/show_bug.cgi?id=10619; by adding the
81757 2021-11-01 Alan Modra <amodra@gmail.com>
81759 macho-o archive sanity checks
81760 Anti-fuzzing checks.
81762 * mach-o.c (bfd_mach_o_fat_archive_p): Sanity check entry offset
81763 and size against file size.
81765 2021-11-01 Alan Modra <amodra@gmail.com>
81767 objcopy buffer overflow
81768 "tocopy" in this code was an int, which when the size to be copied was
81769 larger than MAXINT could result in tocopy being negative. A negative
81770 value of course is less than BUFSIZE, but when converted to
81771 bfd_size_type is extremely large.
81774 * objcopy.c (copy_unknown_object): Correct calculation of "tocopy".
81775 Use better variable types.
81777 2021-11-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
81779 arm: add armv9-a architecture to -march
81780 Update also include:
81781 + New value of Tag_CPU_arch EABI attribute (22) is added.
81782 + Updated missing Tag_CPU_arch EABI attributes.
81783 + Updated how we combine archs 'v4t_plus_v6_m' as this mechanism
81784 have to handle new Armv9 as well.
81786 Regression tested on `arm-none-eabi` cross Binutils and no issues.
81790 * archures.c: Define bfd_mach_arm_9.
81791 * bfd-in2.h (bfd_mach_arm_9): Define bfd_mach_arm_9.
81792 * cpu-arm.c: Add 'armv9-a' option to -march.
81793 * elf32-arm.c (using_thumb2_bl): Update assert check.
81794 (arch_has_arm_nop): Add TAG_CPU_ARCH_V9.
81795 (bfd_arm_get_mach_from_attributes): Add case for TAG_CPU_ARCH_V9.
81797 (tag_cpu_arch_combine): Updated table.
81802 * readelf.c (arm_attr_tag_CPU_arch): Update with
81806 * arm.h: Update TAG_CPU_ARCH_ enums with correct values.
81810 * NEWS: Update docs.
81811 * config/tc-arm.c (get_aeabi_cpu_arch_from_fset): Return Armv9-a
81813 (aeabi_set_public_attributes): Update assert.
81814 * doc/c-arm.texi: Update docs.
81815 * testsuite/gas/arm/armv9-a_arch.d: New test.
81816 * testsuite/gas/arm/attr-march-all.d: Update test with v9.
81820 * elf/arm.h Update TAG_CPU_ARCH_ defines with correct values.
81821 * opcode/arm.h (ARM_EXT3_V9A): New macro.
81822 (ARM_ARCH_NONE): Updated with arm_feature_set.core size.
81823 (FPU_NONE): Updated.
81824 (ARM_ANY): Updated.
81825 (ARM_ARCH_UNKNOWN): New macro.
81826 (ARM_FEATURE_LOW): Updated.
81827 (ARM_FEATURE_CORE): Updated.
81828 (ARM_FEATURE_CORE_LOW): Updated.
81829 (ARM_FEATURE_CORE_HIGH): Updated.
81830 (ARM_FEATURE_COPROC): Updated.
81831 (ARM_FEATURE): Updated.
81832 (ARM_FEATURE_ALL): New macro.
81836 * arm-dis.c (select_arm_features): Support bfd_mach_arm_9.
81837 Also Update bfd_mach_arm_unknown to use new macro ARM_ARCH_UNKNOWN.
81839 2021-11-01 Mike Frysinger <vapier@gentoo.org>
81841 sim: iq2000: reduce -Wno-error scope
81842 Clean up the warnings in sim-if, then reduce the -Werror disable to
81843 the files that still aren't clean that now that we require GNU make
81844 and can set variables on a per-object basis.
81846 sim: lm32: reduce -Wno-error scope
81847 Clean up some warnings in dv-lm32cpu, and all in sim-if, then reduce
81848 the -Werror disable to the files that still aren't clean that now that
81849 we require GNU make and can set variables on a per-object basis.
81851 sim: frv: reduce -Wno-error scope
81852 Only two files in here still generates warnings, so reduce the -Werror
81853 disable to that now that we require GNU make and can set variables on
81854 a per-object basis.
81856 sim: m32r: reduce -Wno-error scope
81857 Only two files in here still generates warnings, so reduce the -Werror
81858 disable to that now that we require GNU make and can set variables on
81859 a per-object basis.
81861 sim: mips: reduce -Wno-error scope
81862 Fix a few printf warnings in sim-main.c, and then we're left with only
81863 one file in here still generating warnings, so reduce the -Werror
81864 disable to that alone now that we require GNU make and can set variables
81865 on a per-object basis.
81867 sim: erc32: reduce -Wno-error scope
81868 Only one file in here still generates warnings, so reduce the -Werror
81869 disable to that alone now that we require GNU make and can set variables
81870 on a per-object basis.
81872 sim: cris: reduce -Wno-error scope
81873 Only two files in here still generates warnings, so reduce the -Werror
81874 disable to that now that we require GNU make and can set variables on
81875 a per-object basis.
81877 sim: sh: reduce -Wno-error scope
81878 Only one file in here still generates warnings, so reduce the -Werror
81879 disable to that alone now that we require GNU make and can set variables
81880 on a per-object basis.
81882 sim: or1k: build with -Werror
81883 The only warnings left in this port are a few maybe-uninitialized,
81884 but we don't abort the build for them, so turn on -Werror everywhere.
81886 sim: igen: minor build output alignment fix
81887 The custom echo was off by one space relative to all the others.
81889 sim: ppc: fix the printf fix for 32-bit systems
81890 The time delta is a 64-bit value too.
81892 sim: m68hc11: clean up pointer casts
81893 The void *data field is used to past arbitrary data between event
81894 handlers, and these are using it to pass an integer. Fix up the
81895 casts to avoid using (long) to cast to/from pointers since there
81896 is no guarantee that's the right size.
81898 sim: d10v: clean up pointer casts
81899 Use %p to print pointers instead of trying to cast them to longs.
81901 sim: bfin: cast pointers using uintptr_t
81902 We can't assume that sizeof(long) == sizeof(void*), so change all
81903 these casts over to uintptr_t.
81905 sim: ppc: clean up printf format handling
81906 Don't blindly cast every possible type to (long). Change to the right
81907 printf format specifier whether it be a 64-bit type or a pointer.
81909 sim: ppc: switch core types to stdint.h types
81910 There's no need to define these ourselves anymore, so switch to the
81911 stdint.h types. This will be important when we start using PRI*
81912 defines with printf formats.
81914 sim: mn10300: clean up pointer casts
81915 The void *data field is used to past arbitrary data between event
81916 handlers, and these are using it to pass an enum. Fix up the casts
81917 to avoid using (long) to cast to/from pointers since there is no
81918 guarantee that's the right size.
81920 sim: events: clean up trace casts
81921 Don't blindly cast every possible type to (long). Change to the right
81922 printf format specifier whether it be a 64-bit type or a pointer.
81924 sim: ppc: handle \r in igen inputs [PR sim/28476]
81925 Make sure we consume & ignore \r bytes in inputs in case the file
81926 encodings are from a non-LF systems (e.g. Windows).
81928 sim: ppc: constify strings in igen tooling
81930 2021-11-01 GDB Administrator <gdbadmin@sourceware.org>
81932 Automatic date update in version.in
81934 2021-10-31 Tom Tromey <tom@tromey.com>
81936 Fix latent bug in DWARF test case
81937 On my branch that replaces the DWARF psymtab reader,
81938 dw2-stack-boundary.exp started failing. However, when I look at the
81939 output in gdb.log, it is correct:
81941 file /home/tromey/gdb/build/gdb/testsuite/outputs/gdb.dwarf2/dw2-stack-boundary/dw2-stack-boundary
81942 Reading symbols from /home/tromey/gdb/build/gdb/testsuite/outputs/gdb.dwarf2/dw2-stack-boundary/dw2-stack-boundary...
81943 During symbol reading: location description stack overflow
81944 During symbol reading: location description stack underflow
81946 What happens to cause the failure is that the two branches in
81947 gdb_test_multiple appear in this order:
81949 -re "\r\nDuring symbol reading: location description stack underflow" {
81951 -re "\r\nDuring symbol reading: location description stack overflow" {
81953 The first one will match the above, without causing the second one to
81954 ever match -- leading to a spurious failure.
81956 Anchoring the regexps seems to fix the problem, and works for the
81957 current gdb as well.
81959 2021-10-31 Tom Tromey <tom@tromey.com>
81961 Fix unittest.exp failure due to 'set debuginfod' addition
81962 The 'set debuginfod' change caused a regression in unittest.exp:
81964 Running selftest help_doc_invariants.
81965 help doc broken invariant: command 'info set debuginfod' help doc first line is not terminated with a '.' character
81966 help doc broken invariant: command 'set debuginfod' help doc first line is not terminated with a '.' character
81967 help doc broken invariant: command 'show debuginfod' help doc first line is not terminated with a '.' character
81968 Self test failed: self-test failed at ../../binutils-gdb/gdb/unittests/command-def-selftests.c:100
81970 This patch fixes the problem. I'm checking it in.
81972 2021-10-31 Mike Frysinger <vapier@gentoo.org>
81974 sim: ppc: use silent build rules here too
81975 The ppc codebase is unique and doesn't leverage common/, so have to
81976 add silent rules to it specifically.
81978 sim: rl78: drop obsolete manual dependency rules
81979 We have GNU make generate these for us automatically now, so there's
81980 no need to manually specify any deps.
81982 sim: drop unused targ-vals.h includes
81983 This is used in a few places where it's not needed. Drop the include
81984 to avoid the build-time generated header file as we move to drop it.
81986 sim: unify callback.o building
81987 Now that the use of TARGET_xxx defines have been removed, we can move
81988 this to the common logic so we only build it once for multi-targets.
81990 sim: nltvals: pull target open flags out into a dedicated source file
81991 Like we just did for pulling out the errno & signal maps, pull out the
81992 open flag map into a dedicated common file. All newlib ports are using
81993 the same map which makes it easy.
81995 sim: nltvals: localize TARGET_<open> defines
81996 Code should not be using these directly, instead they should be
81997 resolving these dynamically via the open_map. Rework the common
81998 callback code that was using the defines to use symbolic names
81999 instead, and localize some of the defines in the ARM code (since
82000 it's a bit unclear how many different APIs it supports currently),
82001 then remove the defines out of the header so no new code can rely on
82004 sim: nltvals: pull target signal out into a dedicated source file
82005 Like we just did for pulling out the errno map, pull out the signal
82006 map into a dedicated common file. All newlib ports are using the
82007 same signal map which makes it easy.
82009 2021-10-31 Mike Frysinger <vapier@gentoo.org>
82011 sim: nltvals: pull target errno out into a dedicated source file
82012 The current system maintains a list of target errno constants in the
82013 nltvals.def file, then runs a build-time tool to turn that into a C
82014 file. This list of errno values is the same for all arches, so we
82015 don't need the arch-specific flexibility. Further, these are only
82016 for newlib/libgloss environments, which makes it confusing to support
82017 other userland runtimes (like Linux). Let's simplify to make this
82018 easier to understand & build. We don't namespace the variables yet,
82019 but sets up the framework for it.
82021 Create a new target-newlib-errno.c template file. The template file
82022 is hand written, but the inline map is still automatically generated.
82024 This allows us to move it to the common set of objects so it's only
82025 built once in a multi-target build.
82027 Now we can remove the output from the gentmap build-time tool since
82028 it's checked into the tree.
82030 Then we stop including the errno lists in nltvals.def since nothing
82033 2021-10-31 Mike Frysinger <vapier@gentoo.org>
82035 sim: erc32: use silent build rules with sis linkage
82037 2021-10-31 Mike Frysinger <vapier@gentoo.org>
82039 sim: erc32: fix a few more build warnings
82040 Tweak the if indentation & brace style to avoid ambiguous warnings.
82042 Add ATTRIBUTE_UNUSED to UART functions that aren't used when FAST_UART
82043 is defined (which is the default).
82045 2021-10-31 Orgad Shaneh <orgads@gmail.com>
82047 sim: erc32: fix signedness compatibility and redefinition warnings
82049 2021-10-31 Mike Frysinger <vapier@gentoo.org>
82051 sim: add arch-specific conditional logic
82052 This will make it easy to include arch-specific logic (build files)
82053 as we migrate ports to the common top level build.
82055 sim: v850: delete old gencode logic
82056 The v850 port used to have a gencode helper, but it was deleted long
82057 ago. Clean up the settings that no longer make sense w/out it.
82059 sim: common: merge multiple clean commands
82060 This provides a minor speedup when cleaning in a multi-target build.
82062 sim: m32c: tighten up opc2c build output
82063 Drop the single debugging line that repeats the command line option,
82064 and use the silent build helpers to tighten up output.
82066 sim: tighten up build regen rules
82067 Update the makefile & configure related rules to use the silent
82070 sim: tighten up gencode output
82071 Update the gencode rules to use the silent build helpers.
82073 sim: igen: tighten up build output
82074 Add a new stamp helper for quiet builds, and don't dump the command
82075 line options when it runs. That isn't standard tool behavior, and
82076 doesn't really seem necessary in any way.
82078 sim: tighten up stamp rules
82079 Add a new ECHO_STAMP helper and convert existing stamp code over
82080 to it. This is mostly common rules and cgen mloop rules.
82082 sim: silence stamp touch rules
82083 We pretty much never care about these stamp touches, so silence them.
82084 Also switch to using $@ when it makes sense.
82086 sim: standardize move-if-change rules
82087 Use the srcroot path and make them all silent.
82089 sim: mips/v850: remove redundant variable setup
82090 The common/Make-common.in fragment already provides these variables.
82092 2021-10-31 Orgad Shaneh <orgads@gmail.com>
82094 sim: fix compilation on mingw64 [PR sim/28476]
82095 ...by reordering includes.
82099 sim/mips/sim-main.h defines UserMode, while there is a struct in winnt.h
82100 which has UserMode as a member. So if sim-main.h is included before winnt.h,
82105 registers.h defines CR, which is used as a member in winnt.h.
82107 winsock2.h is included by sys/time.h, so sys/time.h has to be included
82108 before registers.h.
82110 Bug: https://sourceware.org/PR28476
82112 2021-10-31 Alan Modra <amodra@gmail.com>
82114 Don't include coff/pe.h in coff-x86_64.c
82115 This (and other) code from coffcode.h is broken for x86_64_coff_vec,
82116 and has been ever since support was added in 2006 commit 99ad839030c1
82117 Here, bfd_coff_aoutsz must match coff_swap_aouthdr_out otherwise we
82118 end up writing garbage.
82120 /* Note that peicode.h fills in a PEAOUTHDR, not an AOUTHDR.
82121 include/coff/pe.h sets AOUTSZ == sizeof (PEAOUTHDR)). */
82123 bfd_size_type amount = bfd_coff_aoutsz (abfd);
82125 buff = (char *) bfd_malloc (amount);
82129 coff_swap_aouthdr_out (abfd, & internal_a, buff);
82130 amount = bfd_bwrite (buff, amount, abfd);
82132 We have removed support for --target=x86_64-coff, likely because it
82133 never worked properly, but still produce coff-x86_64.o with
82134 --enable-targets=all. This means objcopy can recognize x86_64 COFF
82135 files but will write garbage to the output file, a fact found by
82136 fuzzers. I suspect x86_64 COFF is still broken after this fix, and
82137 mention of coff-x86_64.* should be removed from bfd/Makefile.am.
82139 * coff-x86_64.c: Don't include coff/pe.h.
82140 (COFF_WITH_pex64): Don't define here.
82141 * pe-x86_64.c: Include coff/pe.h and other headers.
82142 (PEI_HEADERS): Define.
82144 2021-10-31 Alan Modra <amodra@gmail.com>
82146 Re: PR28420, ecoff fuzzing failures
82147 sym_ptr_ptr NULL results in segfaults.
82150 * ecoff.c (ecoff_slurp_reloc_table): Don't leave sym_ptr_ptr NULL.
82152 2021-10-31 Alan Modra <amodra@gmail.com>
82154 ubsan: alpha-vms: undefined shift
82155 * vms-alpha.c (evax_bfd_print_image): Shift left 1u.
82157 PR28518: signed integer overflow & free on unmalloced address
82159 * vms-alpha.c (build_module_list): Don't lose malloc buffer address.
82160 Use unsigned variables.
82162 2021-10-31 GDB Administrator <gdbadmin@sourceware.org>
82164 Automatic date update in version.in
82166 2021-10-30 Simon Marchi <simon.marchi@polymtl.ca>
82168 gdb: fix gdb.gdb/unittest.exp with C++17 compiler
82169 On a machine with gcc 11, I get:
82171 FAIL: gdb.gdb/unittest.exp: test_completion: tab complete "maintenance selftest string_v" (second tab) (timeout)
82172 FAIL: gdb.gdb/unittest.exp: test_completion: tab complete "maintenance selftest string_vie" (timeout)
82174 That's because when compiling with C++ >= 17, we use the standard
82175 version of string_view, and don't have a selftest for it. So the list
82176 of selftests shown by the tab completion when completing "string_v"
82179 Change the test to use the copy_* tests instead.
82181 Change-Id: I85f6aa44ee5fc9652b9bd4451e0506b89773526b
82183 2021-10-30 Aaron Merey <amerey@redhat.com>
82185 gdb.texinfo: Expand documentation for debuginfod
82186 Add section describing GDB's usage of debuginfod.
82188 Refer to this new section in the description of the '--with-debuginfod'
82191 Mention debuginfod in the 'Separate Debug Files' section.
82193 2021-10-30 Aaron Merey <amerey@redhat.com>
82195 gdb: add set/show commands for managing debuginfod
82196 Add 'set debuginfod' command. Accepts 'on', 'off' or 'ask' as an
82197 argument. 'on' enables debuginfod for the current session. 'off'
82198 disables debuginfod for the current session. 'ask' will prompt
82199 the user to either enable or disable debuginfod when the next query
82200 is about to be performed:
82202 This GDB supports auto-downloading debuginfo from the following URLs:
82204 Enable debuginfod for this session? (y or [n]) y
82205 Debuginfod has been enabled.
82206 To make this setting permanent, add 'set debuginfod on' to .gdbinit.
82208 For interactive sessions, 'ask' is the default. For non-interactive
82209 sessions, 'off' is the default.
82211 Add 'show debuginfod status' command. Displays whether debuginfod
82212 is set to 'on', 'off' or 'ask'.
82214 Add 'set/show debuginfod urls' commands. Accepts a string of
82215 space-separated debuginfod server URLs to be queried. The default
82216 value is copied from the DEBUGINFOD_URLS environment variable.
82218 Finally add 'set/show debuginfod verbose' commands to control whether
82219 debuginfod-related output is displayed. Verbose output is enabled
82223 Starting program: /bin/sleep 5
82224 Download failed: No route to host. Continuing without debug info for /lib64/libc.so.6.
82226 If GDB is not built with debuginfod then these commands will just display
82228 Support for debuginfod is not compiled into GDB.
82230 2021-10-30 GDB Administrator <gdbadmin@sourceware.org>
82232 Automatic date update in version.in
82234 2021-10-29 Simon Marchi <simon.marchi@polymtl.ca>
82236 gdb: remove TYPE_FIELD_DWARF_BLOCK
82237 Remove TYPE_FIELD_DWARF_BLOCK, replace with type::field +
82238 field::loc_dwarf_block.
82240 Change-Id: I10af9410bb5f46d342b8358a7956998c7e804b64
82242 2021-10-29 Simon Marchi <simon.marchi@polymtl.ca>
82244 gdb: remove TYPE_FIELD_STATIC_PHYSADDR
82245 Remove TYPE_FIELD_STATIC_PHYSADDR replace with type::field +
82246 field::loc_physaddr.
82248 Change-Id: Ica9bc4a48f34750ec82ec86c298d3ecece81bcbd
82250 2021-10-29 Simon Marchi <simon.marchi@polymtl.ca>
82252 gdb: remove TYPE_FIELD_STATIC_PHYSNAME
82253 Remove TYPE_FIELD_STATIC_PHYSNAME, replace with type::field +
82254 field::loc_physname.
82256 Change-Id: Ie35d446b67dd1d02f39998b406001bdb7e6d5abb
82258 2021-10-29 Simon Marchi <simon.marchi@polymtl.ca>
82260 gdb: remove TYPE_FIELD_ENUMVAL
82261 Remove TYPE_FIELD_ENUMVAL, replace with type::field +
82262 field::loc_enumval.
82264 Change-Id: I2ada73e4635aad3363ce2eb22c1dc52698ee2072
82266 2021-10-29 Simon Marchi <simon.marchi@polymtl.ca>
82268 gdb: remove TYPE_FIELD_BITPOS
82269 Remove TYPE_FIELD_BITPOS, replace its uses with type::field +
82272 Change-Id: Iccd8d5a77e5352843a837babaa6bd284162e0320
82274 2021-10-29 Simon Marchi <simon.marchi@polymtl.ca>
82276 gdb: remove TYPE_FIELD_LOC_KIND
82277 Remove TYPE_FIELD_LOC_KIND, replace its uses with type::field +
82280 Change-Id: Ib124a26365df82ac1d23df7962d954192913bd90
82282 2021-10-29 Simon Marchi <simon.marchi@polymtl.ca>
82284 gdb: remove FIELD_DWARF_BLOCK macro
82285 Remove FIELD_DWARF_BLOCK, replace its uses with field::loc_dwarf_block.
82287 Change-Id: I66b7d6a960cb5e341e61e21bd3cc9a6ac26de6a8
82289 2021-10-29 Simon Marchi <simon.marchi@polymtl.ca>
82291 gdb: remove FIELD_STATIC_PHYSADDR macro
82292 Remove FIELD_LOC_KIND_PHYSADDR, replace its uses with
82293 field::loc_physaddr.
82295 Change-Id: Ifd8b2bdaad75f42bfb1404ef8c396ffe7e10ac55
82297 2021-10-29 Simon Marchi <simon.marchi@polymtl.ca>
82299 gdb: remove FIELD_STATIC_PHYSNAME macro
82300 Remove FIELD_STATIC_PHYSNAME, replace its uses with field::loc_physname.
82302 Change-Id: Iaa8952410403b4eb5bbd68411feea27e2405d657
82304 2021-10-29 Simon Marchi <simon.marchi@polymtl.ca>
82306 gdb: remove FIELD_ENUMVAL macro
82307 Remove FIELD_ENUMVAL, replace its uses with field::loc_enumval.
82309 Change-Id: Id4861cee91a8bb583a9836f1aa5da0a320fbf4d9
82311 2021-10-29 Simon Marchi <simon.marchi@polymtl.ca>
82313 gdb: remove FIELD_BITPOS macro
82314 Remove FIELD_BITPOD, replace its uses with field::loc_bitpos.
82316 Change-Id: Idb99297e0170661254276c206383a7e9bf1a935a
82318 2021-10-29 Simon Marchi <simon.marchi@polymtl.ca>
82320 gdb: remove FIELD_LOC_KIND macro
82321 Remove FIELD_LOC_KIND, replace its uses with field::loc_kind or
82322 call_site_target::loc_kind.
82324 Change-Id: I0368d8c3ea269d491bb215aa70e32edbdf55f389
82326 2021-10-29 Tom Tromey <tromey@adacore.com>
82328 Add gdb.Architecture.integer_type Python function
82329 This adds a new Python function, gdb.Architecture.integer_type, which
82330 can be used to look up an integer type of a given size and
82331 signed-ness. This is useful to avoid dependency on debuginfo when a
82332 particular integer type would be useful.
82334 v2 moves this to be a method on gdb.Architecture and addresses other
82337 2021-10-29 Tom Tromey <tromey@adacore.com>
82339 Remove ada_value_print_inner
82340 I noticed that the only caller of ada_value_print_inner is
82341 valprint.c:do_val_print (via ada_language::value_print_inner), meaning
82342 that the try/catch logic in this function is redundant. This patch
82343 removes the wrapper function.
82345 Regression tested on x86-64 Fedora 34.
82347 2021-10-29 Tom Tromey <tromey@adacore.com>
82349 Document resolve_dynamic_type oddity
82350 Today I re-learned that resolve_dynamic_type can return a type for
82351 which is_dynamic_type returns true. This can happen for an array
82352 whose elements have dynamic type -- the array is reported as dynamic,
82353 but resolving the elements would be incorrect, because each element
82354 might have a different type after resolution.
82356 You can see the special case in resolve_dynamic_array_or_string:
82358 if (ary_dim != NULL && ary_dim->code () == TYPE_CODE_ARRAY)
82363 I looked into having the TYPE_CODE_ARRAY case in
82364 is_dynamic_type_internal follow this same logic, but that breaks down
82365 on the gdb.fortran/dynamic-ptype-whatis.exp test case. In particular
82366 this code in fortran_undetermined::evaluate:
82368 value *callee = std::get<0> (m_storage)->evaluate (nullptr, exp, noside);
82369 if (noside == EVAL_AVOID_SIDE_EFFECTS
82370 && is_dynamic_type (value_type (callee)))
82371 callee = std::get<0> (m_storage)->evaluate (nullptr, exp, EVAL_NORMAL);
82373 ... relies on is_dynamic_type returning true for such an array.
82375 I wasn't really sure of the best way to fix this, so in the meantime I
82376 wrote this patch, which documents the oddity so that I might have a
82377 chance of remembering this in the future.
82379 2021-10-29 Tom Tromey <tromey@adacore.com>
82381 Avoid self-test failures on x86-linux
82382 The disassembly tests in "maint selftest" will fail on x86-linux.
82383 This happens because opcodes rejects an attempt to disassemble for an
82384 arch with a 64-bit address size when bfd_vma is 32-bit.
82386 This patch avoids this problem by avoiding the test in this case. I
82387 chose to do it this way because this seems to be the only situation
82388 where opcodes checks the size of bfd_vma.
82390 For v2 of this patch, I've also updated memory_error_test to do the
82391 same thing. This is needed due to the "improve error reporting from
82392 the disassembler" patch.
82394 2021-10-29 Tom de Vries <tdevries@suse.de>
82396 [gdb/build] Fix build with --disable-unit-tests
82397 A build with --disable-unit-tests currently run into:
82399 ld: maint.o: in function \
82400 `maintenance_selftest_completer(cmd_list_element*, completion_tracker&,
82401 char const*, char const*)':
82402 src/gdb/maint.c:1183: undefined reference to \
82403 `selftests::for_each_selftest(
82404 gdb::function_view<
82405 void (std::__cxx11::basic_string<char,std::char_traits<char>,
82406 std::allocator<char> > const&)>)'
82409 Fix this by guarding the call to selftests::for_each_selftest in
82410 maintenance_selftest_completer with GDB_SELF_TEST, such that the "-verbose"
82411 completion still works.
82413 Rebuild on x86_64-linux and ran gdb.gdb/unittest.exp.
82415 2021-10-29 Enze Li <lienze2010@hotmail.com>
82417 Document "memory-tag-violations".
82418 * gdb/doc/gdb.texinfo: (Data): Document '-memory-tag-violations'.
82419 (Command Options): Update the example.
82421 2021-10-29 Tejas Belagod <tejas.belagod@arm.com>
82423 Support for a new pacbti unwind opcode.
82424 This patch adds readelf support for decoding the exception table
82425 opcode for restoring the RA_AUTH_CODE pseudo register defined by the
82427 (https://github.com/ARM-software/abi-aa/releases/download/2021Q1/ehabi32.pdf
82430 * readelf.c (decode_arm_unwind_bytecode): Add support to decode
82431 restoring RA_AUTH_CODE pseudo register.
82433 2021-10-29 Alan Modra <amodra@gmail.com>
82435 Re: arm: add unwinder encoding support for PACBTI
82436 Move the gas testsuite files to where they belong.
82438 2021-10-29 Alan Modra <amodra@gmail.com>
82440 ELF core file size checks
82441 Catch fuzzed segments where p_offset + p_filesz wraps, and limit error
82444 * elfcore.h (elf_core_file_p): Rewrite segment checks using
82445 bfd_get_file_size. Set read_only on file size errors.
82446 * elfcode.h (elf_swap_shdr_in): Don't repeat error message.
82448 2021-10-29 Alan Modra <amodra@gmail.com>
82450 obcopy vs. files with silly section alignment
82451 We already ignore stupid segment alignment when rewriting headers,
82452 ignore section alignment too.
82454 * elf.c (rewrite_elf_program_header): Ignore section alignment
82455 power greater than 62.
82457 2021-10-29 GDB Administrator <gdbadmin@sourceware.org>
82459 Automatic date update in version.in
82461 2021-10-28 Stafford Horne <shorne@gmail.com>
82463 gdb: Add OpenRISC gdbserver and native config news
82464 The previous patches added gdbserver and native debugging support
82465 for OpenRISC targets. This patch documents that in the news.
82467 gdb: or1k: add single step for linux native debugging
82468 Needed for single stepping in Linux, this adds the or1k implementation
82469 of or1k_software_single_step. Most of the implementation is borrowed
82470 from the bare metal single step code from or1k_single_step_through_delay
82471 which has been extracted and shared in helper function
82474 2021-10-28 Stafford Horne <shorne@gmail.com>
82476 gdb: or1k: add native linux support
82477 This patch adds support for running gdb natively on OpenRISC linux.
82478 Debugging support is provided via the linux PTRACE interface which is
82479 mostly handled by GDB genric code. This patch provides the logic of how
82480 to read and write the ptrace registers between linux and GDB.
82482 Single stepping is privided in a separate patch.
82484 2021-10-28 Stafford Horne <shorne@gmail.com>
82486 gdb: or1k: add generated linux descriptor file
82488 gdb: or1k: fixup linux regcache comment
82489 The old comment was not properly updated from the RISC-V example used.
82490 Update it to match OpenRISC.
82492 2021-10-28 Stafford Horne <shorne@gmail.com>
82494 gdb: or1k: implement gdb server
82495 This patch adds gdbserver support for OpenRISC. This has been used for
82496 debugging the glibc port that in being worked on here:
82498 https://github.com/openrisc/or1k-glibc/tree/or1k-port-2
82500 Hence the comment about registers definitions being inline with glibc.
82502 2021-10-28 Christian Biesinger <cbiesinger@google.com>
82504 [sim] Include defs.h in ppc/hw_memory.c
82505 To fix this error (seen on cygwin):
82506 /../../sim/ppc/../common ../../../sim/ppc/hw_memory.c
82507 In file included from ../../gnulib/import/stdlib.h:100,
82508 from ../../../sim/ppc/hw_memory.c:28:
82509 ../../gnulib/import/unistd.h:663:3: error: #error "Please include config.h first."
82510 663 | #error "Please include config.h first."
82512 ../../gnulib/import/unistd.h:665:24: error: expected ';' before 'extern'
82513 665 | _GL_INLINE_HEADER_BEGIN
82516 ../../gnulib/import/unistd.h:2806:22: error: expected ';' before 'extern'
82517 2806 | _GL_INLINE_HEADER_END
82521 2021-10-28 Markus Klein <markus.klein@sma.de>
82523 ARM assembler: Allow up to 32 single precision registers in the VPUSH and VPOP instructions.
82525 * config/tc-arm.c (do_vfp_nsyn_push_pop_check): New function.
82526 (do_vfp_nsyn_pop): Use the new function.
82527 (do_vfp_nsyn_push): Use the new function.
82528 * testsuite/gas/arm/v8_1m-mve.s: Add new instructions.
82529 * testsuite/gas/arm/v8_1m-mve.d: Updated expected disassembly.
82531 2021-10-28 Simon Marchi <simon.marchi@efficios.com>
82533 gdb: use ptid_t::to_string in infrun debug messages
82534 In debug messages, I think it would be more helpful to print ptid using
82535 the simple "pid.lwp.tid" notation in infrun debug messages. I am
82536 currently debugging some fork issues, and find the pid_to_str output not
82537 so useful, as it doesn't tell which process a thread belongs to.
82539 It currently shows up like this:
82541 [infrun] resume_1: step=1, signal=GDB_SIGNAL_0, trap_expected=0, current thread [Thread 0x7ffff7d95740 (LWP 892942)] at 0x55555555521f
82543 With the patch, it shows up like this:
82545 [infrun] resume_1: step=1, signal=GDB_SIGNAL_0, trap_expected=1, current thread [894072.894077.0] at 0x5555555551d9
82547 Change-Id: I130796d7dfb0d8e763b8358d8a6002701d80c4ea
82549 2021-10-28 Simon Marchi <simon.marchi@polymtl.ca>
82551 gdb: add selftest name completion
82552 After the previous commit, it is easy to add completion for selftest
82553 names. Again, this is not particularly high value, but I rarely touched
82554 completion, so it served as a simple example to get some practice.
82556 Change the for_each_selftest_ftype parameter to gdb::function_view, so
82557 that we can pass a lambda that captures things.
82559 Change-Id: I87cac299ddca9ca7eb0ffab78342e850a98d954c
82561 2021-10-28 Tejas Belagod <Tejas.Belagod@arm.com>
82563 arm: add unwinder encoding support for PACBTI
82564 This patch adds support for encoding the Return Address Authentication pseudo
82565 register - '.save {ra_auth_code}' as defined by the DWARF ABI - in the
82566 exception tables where the opcode is defined by the EHABI
82570 * config/tc-arm.c (arm_reg_type): Add new type REG_TYPE_PSEUDO.
82571 (reg_expected_msgs): Add message for pseudo reg type.
82572 (reg_list_els): Add new reg list type REGLIST_PSEUDO.
82573 (parse_reg_list): Handle new REGLIST_PSEUDO type.
82574 (s_arm_unwind_save_pseudo): Encode pseudo reg list save in exception
82576 (s_arm_unwind_save): Handle new REG_TYPE_PSEUDO.
82577 (reg_names): Add ra_auth_code pseudo register.
82578 * testsuite/gas/arm/unwind-pacbti-m.s: New test.
82579 * testsuite/gas/arm/unwind-pacbti-m.d: New test.
82580 * testsuite/gas/arm/unwind-pacbti-m-readelf.d: New test.
82582 2021-10-28 Simon Marchi <simon.marchi@polymtl.ca>
82584 gdb: add "maint set/show selftest verbose" commands and use process_options
82585 I saw the new -verbose switch to "maint selftests" and thought it would
82586 be nice for it to use the option framework. For example, that makes
82587 having completion easy. It's not that high value, given this is a
82588 maintenance command, but I had never used the framework myself, so it
82589 was a good way to practice.
82591 This patch also adds the "maint set/show selftest verbose" setting. It
82592 would be possible to use option framework without adding the setting,
82593 but using the framework makes adding the option almost trivial, so I
82596 Change-Id: I6687faa0713ff3da60b398253211777100094144
82598 2021-10-28 Tom de Vries <tdevries@suse.de>
82600 [gdb/testsuite] Update some test-cases to GPLv3
82601 I noticed some files in the test-suite have GPLv2 notices.
82603 Update these to GPLv3.
82605 2021-10-28 Simon Marchi <simon.marchi@polymtl.ca>
82607 gdb: add add_setshow_prefix_cmd
82608 There's a common pattern to call add_basic_prefix_cmd and
82609 add_show_prefix_cmd to add matching set and show commands. Add the
82610 add_setshow_prefix_cmd function to factor that out and use it at a few
82613 Change-Id: I6e9e90a30e9efb7b255bf839cac27b85d7069cfd
82615 2021-10-28 Tom de Vries <tdevries@suse.de>
82617 [gdb/testsuite] Require python in gdb.server/server-kill-python.exp
82618 I came across this when running test-case gdb.server/server-kill-python.exp
82619 with a gdb configured without python:
82621 builtin_spawn gdb -nw -nx -data-directory data-directory -iex set height 0 \
82622 -iex set width 0 -quiet -iex set height 0 -iex set width 0 \
82623 -ex source outputs/gdb.server/server-kill-python/file1.py^M
82624 FAIL: gdb.server/server-kill-python.exp: ensure inferior is running
82625 Executing on target: kill -9 28535 (timeout = 300)
82626 builtin_spawn -ignore SIGHUP kill -9 28535^M
82627 file1.py:1: Error in sourced command file:^M
82628 Undefined command: "import". Try "help".^M
82631 Fix this by testing for python support in the test-case.
82633 Tested on aarch64-linux (with python support disabled) and x86_64-linux (with
82634 python support enabled).
82636 2021-10-28 Tom de Vries <tdevries@suse.de>
82638 [gdb/testsuite] Fix assembly comments in gdb.dwarf2/clang-debug-names.exp.tcl
82639 On openSUSE Leap 15.2 aarch64 I ran into:
82641 clang-debug-names-debug.S:72: \
82642 Error: junk at end of line, first unrecognized character is `#'
82646 71 .Ldebug_names_start:
82647 72 .short 5 # Header: version
82650 Fix this by using the /* ... */ comment style instead:
82652 $ sed -i 's% #\([^"]*\)%/*\1 */%' clang-debug-names.exp.tcl
82655 Tested on aarch64-linux and x86_64-linux.
82657 2021-10-28 Tom de Vries <tdevries@suse.de>
82659 [gdb/symtab] Handle DW_AT_string_length with location list
82660 Consider a fortran routine where a string variable s is modified:
82670 When compiling with optimization level -O1 and printing the type of
82673 $ gdb -q -batch outputs/gdb.opt/fortran-string/fortran-string \
82677 Breakpoint 1 at 0x4006f7: file fortran-string.f90, line 21.
82679 Breakpoint 1, f (s=..., _s=_s@entry=3) at fortran-string.f90:21
82683 while with -O0 we have instead:
82685 type = character (3)
82688 The problem is that the type of s is:
82690 <1><2d6>: Abbrev Number: 21 (DW_TAG_string_type)
82691 <2d7> DW_AT_string_length: 0xbf (location list)
82692 <2db> DW_AT_byte_size : 4
82694 where the DW_AT_string_length is a location list, a case that is not handled
82695 by attr_to_dynamic_prop.
82697 Fix this by handling attr->form_is_section_offset () in attr_to_dynamic_prop.
82699 Tested on x86_64-linux.
82701 The test-case is based on gdb.opt/fortran-string.exp from
82702 https://src.fedoraproject.org/rpms/gdb/raw/f32/f/gdb-archer-vla-tests.patch .
82703 I've updated the copyrights to stretch to 2021.
82705 [ I've tried to create a dwarf assembly test-case for this, but didn't
82708 Co-Authored-By: Jan Kratochvil <jan.kratochvil@redhat.com>
82710 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=26910
82712 2021-10-28 Kavitha Natarajan <kavitha.natarajan@amd.com>
82714 [gdb/testsuite] Initialize anonymous union in gdb.cp/koenig.cc
82715 GDB test fails while running the test case gdb.cp/koenig.exp using
82719 No symbol "p_union" in current context.
82720 (gdb) FAIL: gdb.cp/koenig.exp: p foo (p_union)
82723 In the testcase, "p_union" is an unused/uninitialized variable of
82724 anonymous union type. Clang does not emit symbol for unused anonymous
82725 union/struct variables at any optimization level. Since the compiler
82726 itself is not emitting the symbol for "p_union", debug info is also
82727 not emitted when built with debug option. If the anonymous union is
82728 initialized (or used), then clang emits the symbol "p_union" which
82729 enables emitting debug info for "p_union".
82732 Cannot resolve function foo to any overloaded instance
82733 (gdb) PASS: gdb.cp/koenig.exp: p foo (p_union)
82736 2021-10-28 Alan Modra <amodra@gmail.com>
82738 asan: mmo: NULL dereferenc in mmo_xore_32
82739 mmo_get_loc can return NULL. It's commented even, and that the caller
82740 then must handle a split field. mmo_xore_* don't handle split fields,
82741 instead just segfault. Stop that happening, and refuse to recognise
82742 fuzzed mmo files that trigger this problem.
82744 * mmo.c (mmo_get_loc): Don't declare inline.
82745 (mmo_xore_64, mmo_xore_32, mmo_xore_16): Remove forward decls.
82746 Return pointer, don't dereference NULL.
82747 (mmo_scan): Return error on mmo_get_loc returning NULL.
82749 2021-10-28 Alan Modra <amodra@gmail.com>
82751 bfd: remove use of INLINE
82752 No need to use anything fancy, plain inline works just as well.
82754 * bfd-in.h (INLINE): Don't define.
82755 * bfd-in2.h: Regenerate.
82756 * aoutx.h: Replace use of INLINE with inline.
82757 * elf-eh-frame.c: Likewise.
82758 * elf32-score7.c: Likewise.
82759 * elfxx-mips.c: Likewise.
82760 * ihex.c: Likewise.
82761 * mach-o.c: Likewise.
82764 2021-10-28 Alan Modra <amodra@gmail.com>
82766 ASSERT in empty output section with address
82767 * ldlang.c (lang_do_assignments_1): Correct "dot" inside ignored
82769 * testsuite/ld-scripts/empty-address-4.d,
82770 * testsuite/ld-scripts/empty-address-4.s,
82771 * testsuite/ld-scripts/empty-address-4.t: New test.
82772 * testsuite/ld-scripts/empty-address.exp: Run it.
82774 2021-10-28 GDB Administrator <gdbadmin@sourceware.org>
82776 Automatic date update in version.in
82778 2021-10-27 Alan Modra <amodra@gmail.com>
82780 asan: alpha-vms: buffer overflows
82781 Yet more anti-fuzzer sanity checking
82783 * vms-alpha.c (evax_bfd_print_egsd): Sanity check record and
82784 name lengths before access.
82785 (evax_bfd_print_etir_stc_ir, evax_bfd_print_etir): Likewise.
82787 2021-10-27 Alan Modra <amodra@gmail.com>
82789 ubsan: arm: undefined shift
82790 left shift of 2 by 31 places cannot be represented in type 'int'
82792 * arm-dis.c (print_insn_thumb16): Avoid undefined behaviour.
82794 2021-10-27 Tom Tromey <tromey@adacore.com>
82796 Fix watchpoints with multiple threads on Windows
82797 A recent internal change pointed out that watchpoints were not working
82798 on Windows when the inferior was multi-threaded. This happened
82799 because the debug registers were only updated for certain threads --
82800 in particular, those that were being resumed and that were not marked
82801 as suspended. In the case of single-stepping, the need to update the
82802 debug registers in other threads could also be "forgotten".
82804 This patch changes windows-nat.c to mark all threads needing a debug
82805 register update. This brings the code closer to what gdbserver does
82806 (though, unfortunately, it still seems more complicated than needed).
82808 2021-10-27 Tom de Vries <tdevries@suse.de>
82810 [gdb/testsuite] Fix port detection in gdb.debuginfod/fetch_src_and_symbols.exp
82811 On OBS I ran into this failure with test-case
82812 gdb.debuginfod/fetch_src_and_symbols.exp:
82814 Failed to listen for connections: Address already in use^M
82815 [Thu Oct 21 11:48:49 2021] (559/559): started http server on IPv6 port=8000^M
82817 FAIL: gdb.debuginfod/fetch_src_and_symbols.exp: local_url: find port timeout
82820 The test-case is trying to start debuginfod on a port to see if it's
82821 available, and it handles either this message:
82822 "started http server on IPv4 IPv6 port=$port"
82823 meaning success, or:
82824 "failed to bind to port"
82825 meaning failure, in which case the debuginfod instance is killed, and we try
82828 The test-case only uses the v4 address 127.0.0.1, so fix this by:
82829 - accepting "started http server on IPv4 port=$port"
82830 - rejecting "started http server on IPv6 port=$port"
82832 Tested on x86_64-linux.
82834 2021-10-27 Simon Marchi <simon.marchi@efficios.com>
82836 gdb: fix value.c build on 32-bits
82837 When building on ARM (32-bits), we errors like this:
82839 /home/smarchi/src/binutils-gdb/gdb/value.c: In function 'gdb::array_view<const unsigned char> value_contents_for_printing(value*)':
82840 /home/smarchi/src/binutils-gdb/gdb/value.c:1252:35: error: narrowing conversion of 'length' from 'ULONGEST' {aka 'long long unsigned int'} to 'size_t' {aka 'unsigned int'} [-Werror=narrowing]
82841 1252 | return {value->contents.get (), length};
82844 Fix that by using gdb::make_array_view, which does the appropriate
82847 Change-Id: I7d6f2e75d7440d248b8fb18f8272ee92954b404d
82849 2021-10-27 Nelson Chu <nelson.chu@sifive.com>
82851 RISC-V: Tidy riscv assembler and disassembler.
82852 Tidy the gas/config/tc-riscv.c and opcodes/riscv-dis.c, to prepare for
82853 moving the released extensions (including released vendor extensions)
82854 from integration branch back to mainline.
82856 * Added parts of missing comments.
82858 * Updated md_show_usage.
82860 * For validate_riscv_insn, riscv_ip and print_insn_args, unify the
82861 following pointer names,
82862 - oparg: pointed to the parsed operand defined in the riscv_opcodes.
82863 - asarg: pointed to the parsed operand from assembly.
82864 - opargStart: recorded the parsed operand name from riscv_opcodes.
82865 - asargStart: recorded the parsed operand name from assembly.
82868 * config/tc-riscv.c: Added parts of missind comments and updated
82870 (riscv_multi_subset_supports): Tidy codes.
82871 (validate_riscv_insn): Unify the pointer names, oparg, asarg,
82872 opargStart and asargStart, to prepare for moving the released
82873 extensions from integration branch back to mainline.
82874 (riscv_ip): Likewise.
82875 (macro_build): Added fmtStart, also used to prepare for moving
82876 released extensions.
82877 (md_show_usage): Added missing descriptions for new options.
82879 * riscv-dis.c (print_insn_args): Unify the pointer names,
82880 oparg and opargStart, to prepare for moving the released
82881 extensions from integration branch back to mainline.
82883 2021-10-27 Maciej W. Rozycki <macro@embecosm.com>
82885 opcodes: Fix RPATH not being set for dynamic libbfd dependency
82886 If built as a shared library, libopcodes has a load-time dependency on
82887 libbfd, which is recorded in the dynamic section, however without a
82888 corresponding RPATH entry for the directory to find libbfd in. This
82889 causes loading to fail whenever libbfd is only pulled by libopcodes
82890 indirectly and libbfd has been installed in a directory that is not in
82891 the dynamic loader's search path.
82893 It does not happen with the programs included with binutils or GDB,
82894 because they all also pull libbfd when using libopcodes, but it can
82895 happen with external software, e.g.:
82898 gdbserver: error while loading shared libraries: libbfd-[...].so: cannot open shared object file: No such file or directory
82901 (not our `gdbserver').
82903 Indirect dynamic dependencies are handled by libtool automatically by
82904 adding RPATH entries as required, however our setup for libopcodes
82905 prevents this from happening by linking in libbfd with an explicit file
82906 reference sneaked through to the linker directly behind libtool's back
82907 via the `-Wl' linker command-line option rather than via `-l' combined
82908 with a suitable library search path specified via `-L', as it would be
82909 usually the case, or just referring to the relevant .la file in a fully
82910 libtool-enabled configuration such as ours.
82912 According to an observation in the discussion back in 2007[1][2][3] that
82913 has led to the current arrangement it is to prevent libtool from picking
82914 up the wrong version of libbfd. It does not appear to be needed though,
82915 not at least with our current libtool incarnation, as directly referring
82916 `libbfd.la' does exactly what it should, as previously suggested[4], and
82917 with no link-time reference to the installation directory other than to
82918 set RPATH. Uninstalled version of libopcodes has libbfd's build-time
82919 location prepended to RPATH too, as also expected.
82921 Use a direct reference to `libbfd.la' then, making the load error quoted
82922 above go away. Alternatively `-L' and `-l' could be used to the same
82923 effect, but it seems an unnecessary complication and just another way to
82924 circumvent rather than making use of libtool.
82928 [1] "compile failure due to undefined symbol",
82929 <https://sourceware.org/ml/binutils/2007-08/msg00476.html>
82931 [2] same, <https://sourceware.org/ml/binutils/2007-09/msg00000.html>
82933 [3] same, <https://sourceware.org/ml/binutils/2007-10/msg00019.html>
82935 [4] same, <https://sourceware.org/ml/binutils/2007-10/msg00034.html>
82938 * Makefile.am: Remove obsolete comment.
82939 * configure.ac: Refer `libbfd.la' to link shared BFD library
82941 * Makefile.in: Regenerate.
82942 * configure: Regenerate.
82944 2021-10-27 GDB Administrator <gdbadmin@sourceware.org>
82946 Automatic date update in version.in
82948 2021-10-27 H.J. Lu <hjl.tools@gmail.com>
82950 gold: Place .note.gnu.property section before other note sections
82951 Place the .note.gnu.property section before all other note sections to
82952 avoid being placed between other note sections with different alignments.
82955 * layout.cc (Layout::create_note): Set order to ORDER_PROPERTY_NOTE
82956 for the .note.gnu.property section.
82957 * layout.h (Output_section_order): Add ORDER_PROPERTY_NOTE.
82959 2021-10-26 Tom de Vries <tdevries@suse.de>
82961 [gdb/doc] Fix print inferior-events default
82962 In the docs about print inferior-events we read:
82964 By default, these messages will not be printed.
82967 That used to be the case, but is no longer so since commit f67c0c91715 "Enable
82968 'set print inferior-events' and improve detach/fork/kill/exit messages".
82970 Fix this by updating the docs.
82972 2021-10-26 GDB Administrator <gdbadmin@sourceware.org>
82974 Automatic date update in version.in
82976 2021-10-25 Simon Marchi <simon.marchi@polymtl.ca>
82978 gdb: change functions returning value contents to use gdb::array_view
82979 The bug fixed by this [1] patch was caused by an out-of-bounds access to
82980 a value's content. The code gets the value's content (just a pointer)
82981 and then indexes it with a non-sensical index.
82983 This made me think of changing functions that return value contents to
82984 return array_views instead of a plain pointer. This has the advantage
82985 that when GDB is built with _GLIBCXX_DEBUG, accesses to the array_view
82986 are checked, making bugs more apparent / easier to find.
82988 This patch changes the return types of these functions, and updates
82989 callers to call .data() on the result, meaning it's not changing
82990 anything in practice. Additional work will be needed (which can be done
82991 little by little) to make callers propagate the use of array_view and
82994 [1] https://sourceware.org/pipermail/gdb-patches/2021-September/182306.html
82996 Change-Id: I5151f888f169e1c36abe2cbc57620110673816f3
82998 2021-10-25 Simon Marchi <simon.marchi@efficios.com>
83000 gdbsupport: add assertions in array_view
83001 Add assertions to ensure we don't access an array_view out of bounds.
83002 Enable these assertions only when _GLIBCXX_DEBUG is set, as we did for
83005 Change-Id: Iffaee38252405073735ed123c8e57fde6b2c6be3
83007 2021-10-25 Simon Marchi <simon.marchi@polymtl.ca>
83009 gdbserver: make target_pid_to_str return std::string
83010 I wanted to write a warning that included two target_pid_to_str calls,
83013 warning (_("Blabla %s, blabla %s"),
83014 target_pid_to_str (ptid1),
83015 target_pid_to_str (ptid2));
83017 This doesn't work, because target_pid_to_str stores its result in a
83018 static buffer, so my message would show twice the same ptid. Change
83019 target_pid_to_str to return an std::string to avoid this. I don't think
83020 we save much by using a static buffer, but it is more error-prone.
83022 Change-Id: Ie3f649627686b84930529cc5c7c691ccf5d36dc2
83024 2021-10-25 H.J. Lu <hjl.tools@gmail.com>
83026 x86: Also handle stores for -muse-unaligned-vector-move
83027 * config/tc-i386.c (encode_with_unaligned_vector_move): Also
83029 * testsuite/gas/i386/unaligned-vector-move.s: Add stores.
83030 * testsuite/gas/i386/unaligned-vector-move.d: Updated.
83031 * testsuite/gas/i386/x86-64-unaligned-vector-move.d: Likewise.
83033 2021-10-25 Tom de Vries <tdevries@suse.de>
83035 [gdb/testsuite] Fix duplicate in gdb.mi/mi-var-cp.exp
83036 With test-case gdb.mi/mi-var-cp.exp I run into this duplicate:
83038 PASS: gdb.mi/mi-var-cp.exp: run to mi-var-cp.cc:104 (set breakpoint)
83039 PASS: gdb.mi/mi-var-cp.exp: create varobj for s
83040 PASS: gdb.mi/mi-var-cp.exp: create varobj for s
83041 DUPLICATE: gdb.mi/mi-var-cp.exp: create varobj for s
83044 This is due to a duplicate test name here:
83046 $ cat -n gdb/testsuite/gdb.mi/mi-var-cp.cc
83048 100 int reference_to_struct ()
83050 102 /*: BEGIN: reference_to_struct :*/
83054 106 mi_create_varobj S s "create varobj for s"
83055 107 mi_create_varobj R r "create varobj for s"
83058 Fix this by using "create varobj for r" instead.
83060 Tested on x86_64-linux.
83062 2021-10-25 Nick Alcock <nick.alcock@oracle.com>
83064 libctf, ld: handle nonrepresentable types better
83065 ctf_type_visit (used, among other things, by the type dumping code) was
83066 aborting when it saw a nonrepresentable type anywhere: even a single
83067 structure member with a nonrepresentable type caused an abort with
83068 ECTF_NONREPRESENTABLE. This is not useful behaviour, given that the
83069 abort comes from a type-resolution we are only doing in order to
83070 determine whether the type is a structure or union. We know
83071 nonrepresentable types can't be either, so handle that case and
83072 pass the nonrepresentable type down.
83074 (The added test verifies that the dumper now handles this case and
83075 prints nonrepresentable structure members as it already does
83076 nonrepresentable top-level types, rather than skipping the whole
83077 structure -- or, without the previous commit, skipping the whole types
83081 2021-10-25 Nick Alcock <nick.alcock@oracle.com>
83083 * testsuite/ld-ctf/nonrepresentable-member.*: New test.
83086 2021-10-25 Nick Alcock <nick.alcock@oracle.com>
83088 * ctf-types.c (ctf_type_rvisit): Handle nonrepresentable types.
83090 2021-10-25 Nick Alcock <nick.alcock@oracle.com>
83092 libctf: dump: do not stop dumping types on error
83093 If dumping of a single type fails, we obviously can't dump it; but just
83094 as obviously this doesn't make the other types in the types section
83095 invalid or undumpable. So we should not propagate errors seen when
83096 type-dumping, but rather ignore them and carry on, so we dump as many
83097 types as we can (leaving out the ones we can't grok).
83100 2021-10-25 Nick Alcock <nick.alcock@oracle.com>
83102 * ctf-dump.c (ctf_dump_type): Do not abort on error.
83104 2021-10-25 Nick Alcock <nick.alcock@oracle.com>
83106 binutils, ld: make objdump --ctf's parameter optional
83107 ld by default (and always, unless adjusted with a hand-rolled linker
83108 script) emits deduplicated CTF into the .ctf section. But viewing
83109 it needs you to explicitly tell objdump this: it doesn't default
83110 its argument, even though what you always end up typing is
83113 This is annoying, so make the argument optional.
83116 2021-10-25 Nick Alcock <nick.alcock@oracle.com>
83118 * objdump.c (usage): --ctf now has an optional argument.
83119 (main): Adjust accordingly.
83120 (dump_ctf): Default it.
83121 * doc/ctf.options.texi: Adjust.
83124 2021-10-25 Nick Alcock <nick.alcock@oracle.com>
83126 * testsuite/ld-ctf/array.d: Change --ctf=.ctf to --ctf.
83127 * testsuite/ld-ctf/conflicting-cycle-1.B-1.d: Likewise.
83128 * testsuite/ld-ctf/conflicting-cycle-1.B-2.d: Likewise.
83129 * testsuite/ld-ctf/conflicting-cycle-1.parent.d: Likewise.
83130 * testsuite/ld-ctf/conflicting-cycle-2.A-1.d: Likewise.
83131 * testsuite/ld-ctf/conflicting-cycle-2.A-2.d: Likewise.
83132 * testsuite/ld-ctf/conflicting-cycle-2.parent.d: Likewise.
83133 * testsuite/ld-ctf/conflicting-cycle-3.C-1.d: Likewise.
83134 * testsuite/ld-ctf/conflicting-cycle-3.C-2.d: Likewise.
83135 * testsuite/ld-ctf/conflicting-cycle-3.parent.d: Likewise.
83136 * testsuite/ld-ctf/conflicting-enums.d: Likewise.
83137 * testsuite/ld-ctf/conflicting-typedefs.d: Likewise.
83138 * testsuite/ld-ctf/cross-tu-cyclic-conflicting.d: Likewise.
83139 * testsuite/ld-ctf/cross-tu-cyclic-nonconflicting.d: Likewise.
83140 * testsuite/ld-ctf/cross-tu-into-cycle.d: Likewise.
83141 * testsuite/ld-ctf/cross-tu-noncyclic.d: Likewise.
83142 * testsuite/ld-ctf/cycle-1.d: Likewise.
83143 * testsuite/ld-ctf/cycle-2.A.d: Likewise.
83144 * testsuite/ld-ctf/cycle-2.B.d: Likewise.
83145 * testsuite/ld-ctf/cycle-2.C.d: Likewise.
83146 * testsuite/ld-ctf/data-func-conflicted.d: Likewise.
83147 * testsuite/ld-ctf/diag-cttname-null.d: Likewise.
83148 * testsuite/ld-ctf/diag-cuname.d: Likewise.
83149 * testsuite/ld-ctf/diag-parlabel.d: Likewise.
83150 * testsuite/ld-ctf/enum-forward.d: Likewise.
83151 * testsuite/ld-ctf/enums.d: Likewise.
83152 * testsuite/ld-ctf/forward.d: Likewise.
83153 * testsuite/ld-ctf/function.d: Likewise.
83154 * testsuite/ld-ctf/nonrepresentable.d: Likewise.
83155 * testsuite/ld-ctf/slice.d: Likewise.
83156 * testsuite/ld-ctf/super-sub-cycles.d: Likewise.
83158 2021-10-25 Nick Alcock <nick.alcock@oracle.com>
83160 binutils: make objdump/readelf --ctf-parent actually useful
83161 This option has been present since the very early days of the
83162 development of libctf as part of binutils, and it shows. Back in the
83163 earliest days, I thought we might handle ambiguous types by introducing
83164 new ELF sections on the fly named things like .ctf.foo.c for ambiguous
83165 types found only in foo.c, etc. This turned out to be a terrible idea,
83166 so we moved to using a CTF archive in the .ctf section which contained
83167 all the CTF dictionaries -- but the --ctf-parent option in objdump and
83168 readelf was never adjusted, and lingered as a mechanism to specify CTF
83169 parent dictionaries in sections other than .ctf, even though the linker
83170 has no way to produce parent dictionaries in different sections from
83171 their children, libctf's ctf_open can't handle such split-up
83172 parent/child dicts, and they are never found in the wild, emitted by GNU
83173 ld or by any known third-party linking tool.
83175 Meanwhile, the actually-useful ctf_link feature (albeit not used by ld)
83176 which lets you remap the names of CTF archive members (so you can end up
83177 with a parent archive member named something other than ".ctf", still
83178 contained with all its children in a single .ctf section) had no support
83179 in objdump or readelf: there was no way to tell them that these members
83180 were parents, so all the types in the associated child dicts always
83181 appeared corrupted, referencing nonexistent types from a parent objdump
83184 So adjust --ctf-parent so that rather than taking a section name it
83185 takes a member name instead (if not specified, the name is ".ctf", which
83186 is what GNU ld emits). Because the option was always useless before
83187 now, this is expected to have no backward-compatibility implications.
83189 As part of this, we have to slightly adjust the code which skips the
83190 archive member name if redundant: right now it skips it if it's ".ctf",
83191 on the assumption that this name will almost always be at the start
83192 of the objdump output and thus we'll end up with a shared dump
83193 and then smaller, headed dumps for the per-TU child dicts; but if
83194 the parent name has been changed, that won't be true any more.
83196 So change the rules to "members named .ctf which appear first in the
83197 first have their member name skipped". Since we now need to count
83198 members, move from ctf_archive_iter (for which passing in extra
83199 parameters requires defining a new struct and is clumsy) to
83200 ctf_archive_next, allowing us to just *call* dump_ctf_archive_member and
83201 maintain a member count in the obvious way. In the process we fix a
83202 tiny difference between readelf and objdump: if a ctf_dump ever failed,
83203 readelf skipped every later member, while objdump tried to keep going as
83204 much as it could. For a dumping tool the former is clearly preferable.
83207 2021-10-25 Nick Alcock <nick.alcock@oracle.com>
83209 * objdump.c (usage): --ctf-parent now takes a name, not a section.
83210 (dump_ctf): Don't open a separate section; use the parent_name in
83211 ctf_dict_open instead. Use ctf_archive_next, not ctf_archive_iter,
83212 so we can pass down a member count.
83213 (dump_ctf_archive_member): Add the member count; don't return
83214 anything. Import parents into children no matter what the
83215 parent's name, while still avoiding displaying the header for the
83216 common parent name of ".ctf".
83217 * readelf.c (usage): Adjust similarly.
83218 (dump_section_as_ctf): Likewise.
83219 (dump_ctf_archive_member): Likewise. Never stop iterating over
83220 archive members, even if ctf_dump of one member fails.
83221 * doc/ctf.options.texi: Adjust.
83223 2021-10-25 Alan Modra <amodra@gmail.com>
83225 objdump doesn't accept -L option
83226 A followup to commit ca0e11aa4b.
83228 * objdump.c (main): Add 'L' to short options and sort them.
83230 2021-10-25 Alan Modra <amodra@gmail.com>
83232 bfd_nonfatal_message, localise va_start
83233 Nothing to see here, just a little tidier.
83235 * bucomm.c (bfd_nonfatal_message): Localise va_list args.
83237 2021-10-25 Alan Modra <amodra@gmail.com>
83239 ubsan: _bfd_xcoff64_swap_aux_in left shift of negative value
83240 * coff64-rs6000.c (_bfd_xcoff64_swap_aux_in): Use bfd_vma for h.
83242 asan: evax_bfd_print_image buffer overflow
83243 * vms-alpha.c (evax_bfd_print_image): Sanity check printing of
83244 "image activator fixup" section.
83245 (evax_bfd_print_relocation_records): Sanity check buffer offsets.
83246 (evax_bfd_print_address_fixups): Likewise.
83247 (evax_bfd_print_reference_fixups): Likewise.
83249 2021-10-25 GDB Administrator <gdbadmin@sourceware.org>
83251 Automatic date update in version.in
83253 2021-10-24 Alan Modra <amodra@gmail.com>
83255 asan: c4x, c54x coff_canonicalize_reloc buffer overflow
83256 Sometimes the investigation of a fuzzing bug report leads into areas
83257 you'd rather not go. In this instance by the time I'd figured out the
83258 real cause was a target variant that had never been properly supported
83259 in binutils, the time needed to fix it was less than the time needed
83262 * coffcode.h (coff_set_alignment_hook): Call bfd_coff_swap_reloc_in
83263 not coff_swap_reloc_in.
83264 (coff_slurp_reloc_table): Likewise. Don't use RELOC type.
83265 (ticoff0_swap_table): Use coff_swap_reloc_v0_out and
83266 coff_swap_reloc_v0_in.
83267 * coffswap.h (coff_swap_reloc_v0_in, coff_swap_reloc_v0_out): New.
83268 * coff-tic54x.c (tic54x_lookup_howto): Don't abort.
83269 * coffgen.c (coff_get_normalized_symtab): Use PTR_ADD.
83270 * bfd-in.h (PTR_ADD, NPTR_ADD): Avoid warnings when passing an
83272 * bfd-in2.h: Regenerate.
83274 2021-10-24 Alan Modra <amodra@gmail.com>
83276 asan: arm-darwin: buffer overflow
83278 * mach-o-arm.c (bfd_mach_o_arm_canonicalize_one_reloc): Sanity
83279 check PAIR reloc in other branch of condition as was done for
83280 PR21813. Formatting. Delete debug printf.
83282 asan: aout: heap buffer overflow
83283 * aoutx.h (aout_get_external_symbols): Sanity check before writing
83284 zero index entry. Remove outdated comment.
83285 * pdp11.c (aout_get_external_symbols): Likewise.
83287 2021-10-24 liuzhensong <liuzhensong@loongson.cn>
83289 LoongArch ld support
83290 2021-10-22 Chenghua Xu <xuchenghua@loongson.cn>
83291 Zhensong Liu <liuzhensong@loongson.cn>
83292 Weinan Liu <liuweinan@loongson.cn>
83293 Xiaolin Tang <tangxiaolin@loongson.cn>
83296 * Makefile.am: Add LoongArch.
83297 * NEWS: Mention LoongArch support.
83298 * configure.tgt: Add LoongArch.
83299 * emulparams/elf32loongarch-defs.sh: New.
83300 * emulparams/elf32loongarch.sh: Likewise.
83301 * emulparams/elf64loongarch-defs.sh: Likewise.
83302 * emulparams/elf64loongarch.sh: Likewise.
83303 * emultempl/loongarchelf.em: Likewise.
83304 * Makefile.in: Regenerate.
83305 * po/BLD-POTFILES.in: Regenerate.
83307 * ld-loongarch-elf/disas-jirl.d: New.
83308 * ld-loongarch-elf/disas-jirl.s: Likewise.
83309 * ld-loongarch-elf/jmp_op.d: Likewise.
83310 * ld-loongarch-elf/jmp_op.s: Likewise.
83311 * ld-loongarch-elf/ld-loongarch-elf.exp: Likewise.
83312 * ld-loongarch-elf/macro_op.d: Likewise.
83313 * ld-loongarch-elf/macro_op.s: Likewise.
83314 * ld-loongarch-elf/syscall-0.s: Likewise.
83315 * ld-loongarch-elf/syscall-1.s: Likewise.
83316 * ld-loongarch-elf/syscall.d: Likewise.
83317 * ld-srec/srec.exp: Add LoongArch.
83318 * ld-unique/pr21529.d: Likewise.
83320 2021-10-24 liuzhensong <liuzhensong@loongson.cn>
83322 LoongArch gas support
83323 2021-10-22 Chenghua Xu <xuchenghua@loongson.cn>
83324 Zhensong Liu <liuzhensong@loongson.cn>
83325 Weinan Liu <liuweinan@loongson.cn>
83326 Xiaolin Tang <tangxiaolin@loongson.cn>
83329 * Makefile.am: Add LoongArch.
83330 * NEWS: Mention LoongArch support.
83331 * config/loongarch-lex-wrapper.c: New.
83332 * config/loongarch-lex.h: New.
83333 * config/loongarch-lex.l: New.
83334 * config/loongarch-parse.y: New.
83335 * config/tc-loongarch.c: New.
83336 * config/tc-loongarch.h: New.
83337 * configure.ac: Add LoongArch.
83338 * configure.tgt: Likewise.
83339 * doc/as.texi: Likewise.
83340 * doc/c-loongarch.texi: Likewise.
83341 * Makefile.in: Regenerate.
83342 * configure: Regenerate.
83343 * po/POTFILES.in: Regenerate.
83345 * gas/all/gas.exp: Add LoongArch.
83346 * gas/elf/elf.exp: Likewise.
83347 * gas/loongarch/4opt_op.d: New.
83348 * gas/loongarch/4opt_op.s: Likewise.
83349 * gas/loongarch/fix_op.d: Likewise.
83350 * gas/loongarch/fix_op.s: Likewise.
83351 * gas/loongarch/float_op.d: Likewise.
83352 * gas/loongarch/float_op.s: Likewise.
83353 * gas/loongarch/imm_op.d: Likewise.
83354 * gas/loongarch/imm_op.s: Likewise.
83355 * gas/loongarch/jmp_op.d: Likewise.
83356 * gas/loongarch/jmp_op.s: Likewise.
83357 * gas/loongarch/load_store_op.d: Likewise.
83358 * gas/loongarch/load_store_op.s: Likewise.
83359 * gas/loongarch/loongarch.exp: Likewise.
83360 * gas/loongarch/macro_op.d: Likewise.
83361 * gas/loongarch/macro_op.s: Likewise.
83362 * gas/loongarch/nop.d: Likewise.
83363 * gas/loongarch/nop.s: Likewise.
83364 * gas/loongarch/privilege_op.d: Likewise.
83365 * gas/loongarch/privilege_op.s: Likewise.
83366 * gas/loongarch/syscall.d: Likewise.
83367 * gas/loongarch/syscall.s: Likewise.
83368 * lib/gas-defs.exp: Add LoongArch.
83370 2021-10-24 liuzhensong <liuzhensong@loongson.cn>
83372 LoongArch binutils support
83373 2021-10-22 Chenghua Xu <xuchenghua@loongson.cn>
83374 Zhensong Liu <liuzhensong@loongson.cn>
83375 Weinan Liu <liuweinan@loongson.cn>
83377 * NEWS: Mention LoongArch support.
83378 * readelf.c: Add LoongArch.
83379 * testsuite/binutils-all/objdump.exp: Add LoongArch.
83381 2021-10-24 liuzhensong <liuzhensong@loongson.cn>
83383 LoongArch opcodes support
83384 2021-10-22 Chenghua Xu <xuchenghua@loongson.cn>
83385 Zhensong Liu <liuzhensong@loongson.cn>
83386 Weinan Liu <liuweinan@loongson.cn>
83389 * opcode/loongarch.h: New.
83390 * dis-asm.h: Declare print_loongarch_disassembler_options.
83392 * Makefile.am: Add LoongArch.
83393 * configure.ac: Likewise.
83394 * disassemble.c: Likewise.
83395 * disassemble.h: Declare print_insn_loongarch.
83396 * loongarch-coder.c: New.
83397 * loongarch-dis.c: New.
83398 * loongarch-opc.c: New.
83399 * Makefile.in: Regenerate.
83400 * configure: Regenerate.
83401 * po/POTFILES.in: Regenerate.
83403 2021-10-24 liuzhensong <liuzhensong@loongson.cn>
83405 LoongArch bfd support
83406 2021-10-22 Chenghua Xu <xuchenghua@loongson.cn>
83407 Zhensong Liu <liuzhensong@loongson.cn>
83408 Weinan Liu <liuweinan@loongson.cn>
83410 * Makefile.am: Add LoongArch.
83411 * archures.c: Likewise.
83412 * config.bfd: Likewise.
83413 * configure.ac: Likewise.
83414 * cpu-loongarch.c: New.
83415 * elf-bfd.h: Add LoongArch.
83416 * elf.c: Add LoongArch elfcore_grok_xxx.
83417 * elfnn-loongarch.c: New.
83418 * elfxx-loongarch.c: New.
83419 * elfxx-loongarch.h: New.
83420 * reloc.c: Add LoongArch BFD RELOC ENUM.
83421 * targets.c: Add LoongArch target.
83422 * Makefile.in: Regenerate.
83423 * bfd-in2.h: Regenerate.
83424 * configure: Regenerate.
83425 * libbfd.h: Regenerate.
83426 * po/BLD-POTFILES.in: Regenerate.
83427 * po/SRC-POTFILES.in: Regenerate.
83430 * elf/common.h: Add NT_LARCH_{CPUCFG,CSR,LSX,LASX}.
83431 * elf/loongarch.h: New.
83433 2021-10-24 GDB Administrator <gdbadmin@sourceware.org>
83435 Automatic date update in version.in
83437 2021-10-23 GDB Administrator <gdbadmin@sourceware.org>
83439 Automatic date update in version.in
83441 2021-10-22 H.J. Lu <hjl.tools@gmail.com>
83443 x86: Add -muse-unaligned-vector-move to assembler
83444 Unaligned load/store instructions on aligned memory or register are as
83445 fast as aligned load/store instructions on modern Intel processors. Add
83446 a command-line option, -muse-unaligned-vector-move, to x86 assembler to
83447 encode encode aligned vector load/store instructions as unaligned
83448 vector load/store instructions.
83450 * NEWS: Mention -muse-unaligned-vector-move.
83451 * config/tc-i386.c (use_unaligned_vector_move): New.
83452 (encode_with_unaligned_vector_move): Likewise.
83453 (md_assemble): Call encode_with_unaligned_vector_move for
83454 -muse-unaligned-vector-move.
83455 (OPTION_MUSE_UNALIGNED_VECTOR_MOVE): New.
83456 (md_longopts): Add -muse-unaligned-vector-move.
83457 (md_parse_option): Handle -muse-unaligned-vector-move.
83458 (md_show_usage): Add -muse-unaligned-vector-move.
83459 * doc/c-i386.texi: Document -muse-unaligned-vector-move.
83460 * testsuite/gas/i386/i386.exp: Run unaligned-vector-move and
83461 x86-64-unaligned-vector-move.
83462 * testsuite/gas/i386/unaligned-vector-move.d: New file.
83463 * testsuite/gas/i386/unaligned-vector-move.s: Likewise.
83464 * testsuite/gas/i386/x86-64-unaligned-vector-move.d: Likewise.
83466 2021-10-22 Tom Tromey <tromey@adacore.com>
83468 Fix 'uninstall' target
83469 This adds some missing code to the 'uninstall' targets in gdb and
83470 gdbserver. It also changes gdb's uninstall target so that it no
83471 longer tries to remove any man page -- this is already done (and more
83472 correctly) by doc/Makefile.in.
83474 I tested this with 'make install' followed by 'make uninstall', then
83475 examining the install tree for regular files. Only the 'dir' file
83476 remains, but this appears to just be how 'install-info' is intended to
83479 2021-10-22 Tom Tromey <tromey@adacore.com>
83481 Remove unused variables from gdbserver's Makefile
83482 This removes a number of unused variables from gdbserver's Makefile.
83483 I found these while working on the subsequent patches, and figured it
83484 would be cleaner to have a separate patch for the deletions.
83486 2021-10-22 Tom de Vries <tdevries@suse.de>
83488 [gdb/testsuite] Fix gdb.threads/linux-dp.exp
83489 On openSUSE Tumbleweed with glibc-debuginfo installed I get:
83491 (gdb) PASS: gdb.threads/linux-dp.exp: continue to breakpoint: thread 5's print
83493 #0 print_philosopher (n=3, left=33 '!', right=33 '!') at linux-dp.c:105^M
83494 #1 0x0000000000401628 in philosopher (data=0x40537c) at linux-dp.c:148^M
83495 #2 0x00007ffff7d56b37 in start_thread (arg=<optimized out>) \
83496 at pthread_create.c:435^M
83497 #3 0x00007ffff7ddb640 in clone3 () \
83498 at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:81^M
83499 (gdb) PASS: gdb.threads/linux-dp.exp: first thread-specific breakpoint hit
83501 while without debuginfo installed I get instead:
83503 (gdb) PASS: gdb.threads/linux-dp.exp: continue to breakpoint: thread 5's print
83505 #0 print_philosopher (n=3, left=33 '!', right=33 '!') at linux-dp.c:105^M
83506 #1 0x0000000000401628 in philosopher (data=0x40537c) at linux-dp.c:148^M
83507 #2 0x00007ffff7d56b37 in start_thread () from /lib64/libc.so.6^M
83508 #3 0x00007ffff7ddb640 in clone3 () from /lib64/libc.so.6^M
83509 (gdb) FAIL: gdb.threads/linux-dp.exp: first thread-specific breakpoint hit
83512 The problem is that the regexp used:
83514 "\(from .*libpthread\|at pthread_create\|in pthread_create\)"
83516 expects the 'from' part to match libpthread, but in glibc 2.34 libpthread has
83517 been merged into libc.
83519 Fix this by updating the regexp.
83521 Tested on x86_64-linux.
83523 2021-10-22 Tom de Vries <tdevries@suse.de>
83525 [gdb/testsuite] Fix FAILs in gdb.mi/mi-breakpoint-changed.exp
83526 Since commit e36788d1354 "[gdb/testsuite] Fix handling of nr_args < 3 in
83527 mi_gdb_test" we run into:
83529 PASS: gdb.mi/mi-breakpoint-changed.exp: test_auto_disable: mi runto main
83530 Expecting: ^(-break-insert -f pendfunc1[^M
83531 ]+)?((&.*)*.*~"Breakpoint 2 at.*\\n".*=breakpoint-created,\
83532 bkpt=\{number="2",type="breakpoint".*\}.*\n\^done[^M
83535 -break-insert -f pendfunc1^M
83536 ^done,bkpt={number="2",type="breakpoint",disp="keep",enabled="y",\
83537 addr="0x00007ffff7bd559e",func="pendfunc1",\
83538 file="gdb/testsuite/gdb.mi/pendshr1.c",\
83539 fullname="gdb/testsuite/gdb.mi/pendshr1.c",line="21",thread-groups=["i1"],\
83540 times="0",original-location="pendfunc1"}^M
83542 FAIL: gdb.mi/mi-breakpoint-changed.exp: test_auto_disable: \
83543 -break-insert -f pendfunc1 (unexpected output)
83546 The regexp expects a breakpoint-created event, but that's actually suppressed
83549 DEF_MI_CMD_MI_1 ("break-insert", mi_cmd_break_insert,
83550 &mi_suppress_notification.breakpoint),
83553 Fix this by updating the regexp.
83555 Likewise for the following:
83557 PASS: gdb.mi/mi-breakpoint-changed.exp: test_auto_disable: \
83558 -break-insert -f pendfunc1
83559 Expecting: ^(-break-enable count 1 2[^M
83560 ]+)?(=breakpoint-modified,\
83561 bkpt=\{number="2",type="breakpoint",disp="dis",enabled="y".*\}.*\n\^done[^M
83564 -break-enable count 1 2^M
83567 FAIL: gdb.mi/mi-breakpoint-changed.exp: test_auto_disable: \
83568 -break-enable count 1 2 (unexpected out\
83572 Tested on x86_64-linux.
83574 2021-10-22 Andrew Burgess <andrew.burgess@embecosm.com>
83576 gdb/python: move gdb.Membuf support into a new file
83577 In a future commit I'm going to be creating gdb.Membuf objects from a
83578 new file within gdb/python/py*.c. Currently all gdb.Membuf objects
83579 are created directly within infpy_read_memory (as a result of calling
83580 gdb.Inferior.read_memory()).
83582 Initially I split out the Membuf creation code into a new function,
83583 and left the new function in gdb/python/py-inferior.c, however, it
83584 felt a little random that the Membuf creation code should live with
83585 the inferior handling code.
83587 So, then I moved all of the Membuf related code out into a new file,
83588 gdb/python/py-membuf.c, the interface is gdbpy_buffer_to_membuf, which
83589 wraps an array of bytes into a gdb.Membuf object.
83591 Most of the code is moved directly from py-inferior.c with only minor
83592 tweaks to layout and replacing NULL with nullptr, hence, I've left the
83593 copyright date on py-membuf.c as 2009-2021 to match py-inferior.c.
83595 Currently, the only user of this code is still py-inferior.c, but in
83596 later commits this will change.
83598 There should be no user visible changes after this commit.
83600 2021-10-22 Andrew Burgess <andrew.burgess@embecosm.com>
83602 gdb/python: new gdb.architecture_names function
83603 Add a new function to the Python API, gdb.architecture_names(). This
83604 function returns a list containing all of the supported architecture
83605 names within the current build of GDB.
83607 The values returned in this list are all of the possible values that
83608 can be returned from gdb.Architecture.name().
83610 2021-10-22 Andrew Burgess <andrew.burgess@embecosm.com>
83612 gdb: make disassembler fprintf callback a static member function
83613 The disassemble_info structure has four callbacks, we have three of
83614 them as static member functions within gdb_disassembler, the fourth is
83615 just a global static function.
83617 However, this fourth callback, is still only used from the
83618 disassemble_info struct, so there's no real reason for its special
83621 This commit makes fprintf_disasm a static method within
83624 There should be no user visible changes after this commit.
83626 2021-10-22 Lewis Revill <lewis.revill@embecosm.com>
83628 RISC-V: Added ld testcase for pcgp relaxation.
83629 Consider the the pcgp-relax-02 testcase,
83634 .L1: auipc a0, %pcrel_hi(data_a)
83635 .L2: auipc a1, %pcrel_hi(data_b)
83636 addi a0, a0, %pcrel_lo(.L1)
83637 addi a1, a1, %pcrel_lo(.L2)
83650 If the first auipc is deleted, but we are still building the pcgp
83651 table (connect the high and low pcrel relocations), then there is
83652 an aliasing issue that we need some way to disambiguate which of
83653 the two symbols we are targeting. Therefore, Palmer thought of a
83654 way to use R_RISCV_DELETE to split this into two phases, so we
83655 could resolve the addresses before creating the ambiguities.
83657 This patch just add the ld testcase for the above case, in case we
83658 have changed something but break this.
83661 * testsuite/ld-riscv-elf/ld-riscv-elf.exp: Renamed pcgp-relax
83662 to pcgp-relax-01, and added pcgp-relax-02.
83663 * testsuite/ld-riscv-elf/pcgp-relax-01.d: Renmaed from pcgp-relax.
83664 * testsuite/ld-riscv-elf/pcgp-relax-01.s: Likewise.
83665 * testsuite/ld-riscv-elf/pcgp-relax-02.d: New testcase.
83666 * testsuite/ld-riscv-elf/pcgp-relax-02.s: Likewise.
83668 2021-10-22 Lewis Revill <lewis.revill@embecosm.com>
83670 RISC-V: Don't separate pcgp relaxation to another relax pass.
83671 Commit abd20cb637008da9d32018b4b03973e119388a0a and
83672 ebdcad3fddf6ec21f6d4dcc702379a12718cf0c4 introduced additional
83673 complexity into the paths run by the RISC-V relaxation pass in order to
83674 resolve the issue of accurately keeping track of pcrel_hi and pcrel_lo
83675 pairs. The first commit split up relaxation of these relocs into a pass
83676 which occurred after other relaxations in order to prevent the situation
83677 where bytes were deleted in between a pcrel_lo/pcrel_hi pair, inhibiting
83678 our ability to find the corresponding pcrel_hi relocation from the
83679 address attached to the pcrel_lo.
83681 Since the relaxation was split into two passes the 'again' parameter
83682 could not be used to perform the entire relaxation process again and so
83683 the second commit added a way to restart ldelf_map_segments, thus
83684 starting the whole process again.
83686 Unfortunately this process could not account for the fact that we were
83687 not finished with the relaxation process so in some cases - such as the
83688 case where code would not fit in a memory region before the
83689 R_RISCV_ALIGN relocation was relaxed - sanity checks in generic code
83692 This patch fixes all three of these concerns by reverting back to a
83693 system of having only one target relax pass but updating entries in the
83694 table of pcrel_hi/pcrel_lo relocs every time any bytes are deleted. Thus
83695 we can keep track of the pairs accurately, and we can use the 'again'
83696 parameter to restart the entire target relax pass, behaving in the way
83697 that generic code expects. Unfortunately we must still have an
83698 additional pass to delay deleting AUIPC bytes to avoid ambiguity between
83699 pcrel_hi relocs stored in the table after deletion. This pass can only
83700 be run once so we may potentially miss out on relaxation opportunities
83701 but this is likely to be rare.
83703 https://sourceware.org/bugzilla/show_bug.cgi?id=28410
83706 * elfnn-riscv.c (riscv_elf_link_hash_table): Removed restart_relax.
83707 (riscv_elf_link_hash_table_create): Updated.
83708 (riscv_relax_delete_bytes): Moved after the riscv_update_pcgp_relocs.
83709 Update the pcgp_relocs table whenever bytes are deleted.
83710 (riscv_update_pcgp_relocs): Add function to update the section
83711 offset of pcrel_hi and pcrel_lo, and also update the symbol value
83713 (_bfd_riscv_relax_call): Need to update the pcgp_relocs table
83714 when deleting codes.
83715 (_bfd_riscv_relax_lui): Likewise.
83716 (_bfd_riscv_relax_tls_le): Likewise.
83717 (_bfd_riscv_relax_align): Once we've handled an R_RISCV_ALIGN,
83718 we can't relax anything else, so set the sec->sec_flg0 to true.
83719 Besides, we don't need to update the pcgp_relocs table at this
83720 stage, so just pass NULL pointer as the pcgp_relocs table for
83721 riscv_relax_delete_bytes.
83722 (_bfd_riscv_relax_section): Use only one pass for all target
83724 (_bfd_riscv_relax_delete): Likewise, we don't need to update
83725 the pcgp_relocs table at this stage, and don't need to set
83726 the `again' since restart_relax mechanism is abandoned.
83727 (bfd_elfNN_riscv_restart_relax_sections): Removed.
83728 (_bfd_riscv_relax_section): Updated.
83729 * elfxx-riscv.h (bfd_elf32_riscv_restart_relax_sections): Removed.
83730 (bfd_elf64_riscv_restart_relax_sections): Likewise.
83732 * emultempl/riscvelf.em: Revert restart_relax changes and set
83734 * testsuite/ld-riscv-elf/align-small-region.d: New testcase.
83735 * testsuite/ld-riscv-elf/align-small-region.ld: Likewise.
83736 * testsuite/ld-riscv-elf/align-small-region.s: Likewise.
83737 * testsuite/ld-riscv-elf/restart-relax.d: Removed sine the
83738 restart_relax mechanism is abandoned.
83739 * testsuite/ld-riscv-elf/restart-relax.s: Likewise.
83740 * testsuite/ld-riscv-elf/ld-riscv-elf.exp: Updated.
83742 2021-10-22 Simon Marchi <simon.marchi@polymtl.ca>
83744 gdb: fix remote-sim.c build
83745 Commit 183be222907a ("gdb, gdbserver: make target_waitstatus safe")
83746 broke the remote-sim.c build. In fact, it does some wrong changes,
83747 result of a bad sed invocation.
83749 Fix it by adjusting the code to the new target_waitstatus API.
83751 Change-Id: I3236ff7ef7681fc29215f68be210ff4263760e91
83753 2021-10-22 GDB Administrator <gdbadmin@sourceware.org>
83755 Automatic date update in version.in
83757 2021-10-21 Simon Marchi <simon.marchi@efficios.com>
83759 gdb, gdbserver: make target_waitstatus safe
83760 I stumbled on a bug caused by the fact that a code path read
83761 target_waitstatus::value::sig (expecting it to contain a gdb_signal
83762 value) while target_waitstatus::kind was TARGET_WAITKIND_FORKED. This
83763 meant that the active union field was in fact
83764 target_waitstatus::value::related_pid, and contained a ptid. The read
83765 signal value was therefore garbage, and that caused GDB to crash soon
83766 after. Or, since that GDB was built with ubsan, this nice error
83769 /home/simark/src/binutils-gdb/gdb/linux-nat.c:1271:12: runtime error: load of value 2686365, which is not a valid value for type 'gdb_signal'
83771 Despite being a large-ish change, I think it would be nice to make
83772 target_waitstatus safe against that kind of bug. As already done
83773 elsewhere (e.g. dynamic_prop), validate that the type of value read from
83774 the union matches what is supposed to be the active field.
83776 - Make the kind and value of target_waitstatus private.
83777 - Make the kind initialized to TARGET_WAITKIND_IGNORE on
83778 target_waitstatus construction. This is what most users appear to do
83780 - Add setters, one for each kind. Each setter takes as a parameter the
83781 data associated to that kind, if any. This makes it impossible to
83782 forget to attach the associated data.
83783 - Add getters, one for each associated data type. Each getter
83784 validates that the data type fetched by the user matches the wait
83786 - Change "integer" to "exit_status", "related_pid" to "child_ptid",
83787 just because that's more precise terminology.
83790 That last point is semi-mechanical. There are a lot of obvious changes,
83791 but some less obvious ones. For example, it's not possible to set the
83792 kind at some point and the associated data later, as some users did.
83793 But in any case, the intent of the code should not change in this patch.
83795 This was tested on x86-64 Linux (unix, native-gdbserver and
83796 native-extended-gdbserver boards). It was built-tested on x86-64
83797 FreeBSD, NetBSD, MinGW and macOS. The rest of the changes to native
83798 files was done as a best effort. If I forgot any place to update in
83799 these files, it should be easy to fix (unless the change happens to
83800 reveal an actual bug).
83802 Change-Id: I0ae967df1ff6e28de78abbe3ac9b4b2ff4ad03b7
83804 2021-10-21 Simon Marchi <simon.marchi@polymtl.ca>
83806 gdbserver: initialize the members of lwp_info in-class
83807 Add a constructor to initialize the waitstatus members. Initialize the
83808 others in the class directly.
83810 Change-Id: I10f885eb33adfae86e3c97b1e135335b540d7442
83812 2021-10-21 Simon Marchi <simon.marchi@polymtl.ca>
83814 gdbserver: make thread_info non-POD
83815 Add a constructor and a destructor. The constructor takes care of the
83816 initialization that happened in add_thread, while the destructor takes
83817 care of the freeing that happened in free_one_thread. This is needed to
83818 make target_waitstatus non-POD, as thread_info contains a member of that
83821 Change-Id: I1db321b4de9dd233ede0d5c101950f1d6f1d13b7
83823 2021-10-21 Andrew Pinski <apinski@marvell.com>
83825 Fix ARMv8.4 for hw watchpoint and breakpoint
83826 Just like my previoius patch for ARMv8.1 and v8.2 (49ecef2a7da2ee9df4),
83827 this adds ARMv8.4 debug arch as being compatible for hw watchpoint
83830 Refactor code slightly in nat/aarch64-linux-hw-point.c (aarch64_linux_get_debug_reg_capacity)
83831 Since the two locations which check the debug arch are the same code currently, it is
83832 a good idea to factor it out to a new function and just use that function from
83833 aarch64_linux_get_debug_reg_capacity. This is also the first step to support
83834 ARMv8.4 debug arch.
83836 2021-10-21 Carl Love <cel@us.ibm.com>
83838 Fixes for gdb.mi/mi-break.exp
83839 Update the expected pattern for two of the tests.
83841 Matching pattern \" doesn't work. Use .* to match the \* pattern.
83843 2021-10-21 Tom de Vries <tdevries@suse.de>
83845 [gdb/tui] Fix breakpoint display functionality
83846 In commit 81e6b8eb208 "Make tui-winsource not use breakpoint_chain", a loop
83847 body was transformed into a lambda function body:
83849 - for (bp = breakpoint_chain;
83852 + iterate_over_breakpoints ([&] (breakpoint *bp) -> bool
83855 - a continue was replaced by a return, and
83856 - a final return was added.
83858 Then in commit 240edef62f0 "gdb: remove iterate_over_breakpoints function", we
83859 transformed back to a loop body:
83861 - iterate_over_breakpoints ([&] (breakpoint *bp) -> bool
83862 + for (breakpoint *bp : all_breakpoints ())
83864 but without reverting the changes that introduced the two returns.
83866 Consequently, breakpoints no longer show up in the tui source window.
83868 Fix this by reverting the changes that introduced the two returns.
83870 Build on x86_64-linux, tested with all .exp test-cases that contain
83873 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28483
83875 2021-10-21 Carl Love <cel@us.ibm.com>
83877 Fix test step-and-next-inline.cc
83878 The test expect the runto_main to stop at the first line of the function.
83879 Depending on the optimization level, gdb may stop in the prolog or after
83880 the prolog at the first line. To ensure the test stops at the first line
83881 of main, have it explicitly stop at a break point on the first line of the
83884 On PowerPC, the test passes when compiled with no optimization but fails
83885 with all levels of optimization due to gdb stopping in the prolog.
83887 2021-10-21 Tom Tromey <tromey@adacore.com>
83889 Fix latent Ada bug when accessing field offsets
83890 The "add accessors for field (and call site) location" patch caused a
83891 gdb crash when running the internal AdaCore testsuite. This turned
83892 out to be a latent bug in ada-lang.c.
83894 The immediate cause of the bug is that find_struct_field
83895 unconditionally uses TYPE_FIELD_BITPOS. This causes an assert for a
83898 This patch fixes the problem by doing two things. First, it changes
83899 find_struct_field to use a dummy value for the field offset in the
83900 situation where the offset is not actually needed by the caller. This
83901 works because the offset isn't used in any other way -- only as a
83904 Second, this patch assures that calls to find_struct_field use a
83905 resolved type when the offset is needed. For
83906 value_tag_from_contents_and_address, this is done by resolving the
83907 type explicitly. In ada_value_struct_elt, this is done by passing
83908 nullptr for the out parameters when they are not needed (the second
83909 call in this function already uses a resolved type).
83911 Note that, while we believe the parent field probably can't occur at a
83912 variable offset, the patch still updates this code path, just in case.
83914 I've updated an existing test case to reproduce the crash.
83915 I'm checking this in.
83917 2021-10-21 Alan Modra <amodra@gmail.com>
83919 -Waddress warning in ldelf.c
83920 ldelf.c: In function 'ldelf_after_open':
83921 ldelf.c:1049:43: warning: the comparison will always evaluate as 'true' for the address of 'elf_header' will never be NULL [-Waddress]
83922 1049 | && elf_tdata (abfd)->elf_header != NULL
83924 In file included from ldelf.c:37:
83925 ../bfd/elf-bfd.h:1957:21: note: 'elf_header' declared here
83926 1957 | Elf_Internal_Ehdr elf_header[1]; /* Actual data, but ref like ptr */
83928 * ldelf.c (ldelf_after_open): Remove useless elf_header test.
83930 2021-10-21 Alan Modra <amodra@gmail.com>
83932 Avoid -Waddress warnings in readelf
83934 readelf.c: In function 'find_section':
83935 readelf.c:349:8: error: the comparison will always evaluate as 'true' for the pointer operand in 'filedata->section_headers + (sizetype)((long unsigned int)i * 80)' must not be NULL [-Werror=address]
83936 349 | ((X) != NULL \
83938 readelf.c:761:9: note: in expansion of macro 'SECTION_NAME_VALID'
83939 761 | if (SECTION_NAME_VALID (filedata->section_headers + i)
83940 | ^~~~~~~~~~~~~~~~~~
83942 This will likely be fixed in gcc, but inline functions are nicer than
83945 * readelf.c (SECTION_NAME, SECTION_NAME_VALID),
83946 (SECTION_NAME_PRINT, VALID_SYMBOL_NAME, VALID_DYNAMIC_NAME),
83947 (GET_DYNAMIC_NAME): Delete. Replace with..
83948 (section_name, section_name_valid, section_name_print),
83949 (valid_symbol_name, valid_dynamic_name, get_dynamic_name): ..these
83950 new inline functions. Update use throughout file.
83952 2021-10-21 GDB Administrator <gdbadmin@sourceware.org>
83954 Automatic date update in version.in
83956 2021-10-20 Alan Modra <amodra@gmail.com>
83958 PR28417, std::string no longer allows accepting nullptr_t
83960 * incremental.cc (Sized_relobj_incr::do_section_name): Avoid
83961 std:string undefined behaviour.
83962 * options.h (Search_directory::Search_directory): Likewise.
83964 2021-10-20 Alan Modra <amodra@gmail.com>
83966 Re: PR27625, powerpc64 gold __tls_get_addr calls
83967 My previous PR27625 patch had a problem or two. For one, the error
83968 "__tls_get_addr call lacks marker reloc" on processing some calls
83969 before hitting a call without markers typically isn't seen. Instead a
83970 gold assertion fails. Either way it would be a hard error, which
83971 triggers on a file contained in libphobos.a when running the gcc
83972 testsuite. A warning isn't even appropriate since the call involved
83973 is one built by hand without any of the arg setup relocations that
83974 might result in linker optimisation.
83976 So this patch reverts most of commit 0af4fcc25dd5, instead entirely
83977 ignoring the problem of mis-optimising old-style __tls_get_addr calls
83978 without marker relocs. We can't handle them gracefully without
83979 another pass over relocations before decisions are made about GOT
83980 entries in Scan::global or Scan::local. That seems too costly, just
83981 to link object files from 2009. What's more, there doesn't seem to be
83982 any way to allow the libphobos explicit __tls_get_addr call, but not
83983 old TLS sequences without marker relocs. Examining instructions
83984 before the __tls_get_addr call is out of the question: program flow
83985 might reach the call via a branch. Putting an R_PPC64_TLSGD marker
83986 with zero sym on the call might be a solution, but current linkers
83987 will then merrily optimise away the call!
83990 * powerpc.cc (Powerpc_relobj): Delete no_tls_marker_, tls_marker_,
83991 and tls_opt_error_ variables and accessors. Remove all uses.
83993 2021-10-20 Tom Tromey <tom@tromey.com>
83995 Use std::string in print_one_catch_syscall
83996 This changes print_one_catch_syscall to use std::string, removing a
83997 bit of manual memory management.
83999 Use unique_xmalloc_ptr in breakpoint
84000 This changes struct breakpoint to use unique_xmalloc_ptr in a couple
84001 of spots, removing a bit of manual memory management.
84003 Use unique_xmalloc_ptr in bp_location
84004 This changes struct bp_location to use a unique_xmalloc_ptr, removing
84005 a bit of manual memory management.
84007 Use unique_xmalloc_ptr in watchpoint
84008 This changes struct watchpoint to use unique_xmalloc_ptr in a couple
84009 of places, removing a bit of manual memory management.
84011 Use unique_xmalloc_ptr in exec_catchpoint
84012 This changes struct exec_catchpoint to use a unique_xmalloc_ptr,
84013 removing a bit of manual memory management.
84015 Use unique_xmalloc_ptr in solib_catchpoint
84016 This changes struct solib_catchpoint to use a unique_xmalloc_ptr,
84017 removing a bit of manual memory management.
84019 2021-10-20 Christian Biesinger <cbiesinger@google.com>
84021 Make c-exp.y work with Bison 3.8+
84022 When using Bison 3.8, we get this error:
84024 ../../gdb/c-exp.y:3455:1: error: 'void c_print_token(FILE*, int, YYSTYPE)' defined but not used [-Werror=unused-function]
84026 That's because bison 3.8 removed YYPRINT support:
84027 https://savannah.gnu.org/forum/forum.php?forum_id=10047
84029 Accordingly, this patch only defines that function for Bison < 3.8.
84031 Change-Id: I3cbf2f317630bb72810b00f2d9b2c4b99fa812ad
84033 2021-10-20 GDB Administrator <gdbadmin@sourceware.org>
84035 Automatic date update in version.in
84037 2021-10-19 Tom de Vries <tdevries@suse.de>
84039 [gdb/testsuite] Reimplement gdb.gdb/python-interrupts.exp as unittest
84040 The test-case gdb.gdb/python-interrupts.exp:
84041 - runs to captured_command_loop
84042 - sets a breakpoint at set_active_ext_lang
84043 - calls a python command
84044 - verifies the command triggers the breakpoint
84045 - sends a signal and verifies the result
84047 The test-case is fragile, because (f.i. with -flto) it cannot be guaranteed
84048 that captured_command_loop and set_active_ext_lang are available for setting
84051 Reimplement the test-case as unittest, using:
84052 - execute_command_to_string to capture the output
84053 - try/catch to catch the "Error while executing Python code" exception
84054 - a new hook selftests::hook_set_active_ext_lang to raise the signal
84056 Tested on x86_64-linux.
84058 2021-10-19 Tom Tromey <tromey@adacore.com>
84060 Check index in type::field
84061 This changes gdb to check the index that is passed to type::field.
84062 This caught one bug in the Ada code when running the test suite
84063 (actually I found the bug first, then realized that the check would
84064 have helped), so this patch fixes that as well.
84066 Regression tested on x86-64 Fedora 34.
84068 2021-10-19 Tom Tromey <tromey@adacore.com>
84070 Fix Rust lex selftest when using libiconv
84071 The Rust lex selftest fails on our Windows build. I tracked this down
84072 to a use of UTF-32 as a parameter to convert_between_encodings. Here,
84073 iconv_open succeeds, but the actual conversion of a tab character
84074 fails with EILSEQ. I suspect that "UTF-32" is being interpreted as
84075 big-endian, as changing the call to use "UTF-32LE" makes it work.
84076 This patch implements this fix.
84078 2021-10-19 Tom Tromey <tromey@adacore.com>
84080 Fix format_pieces selftest on Windows
84081 The format_pieces selftest currently fails on Windows hosts.
84083 The selftest doesn't handle the "%ll" -> "%I64" rewrite that the
84084 formatter may perform, but also gdbsupport was missing a configure
84085 check for PRINTF_HAS_LONG_LONG. This patch fixes both issues.
84087 2021-10-19 Tom Tromey <tromey@adacore.com>
84089 Fix bug in dynamic type resolution
84090 A customer-reported problem led us to a bug in dynamic type
84091 resolution. resolve_dynamic_struct will recursively call
84092 resolve_dynamic_type_internal, passing it the sub-object for the
84093 particular field being resolved. While it offsets the address here,
84094 it does not also offset the "valaddr" -- the array of bytes describing
84097 This patch fixes the bug, by offsetting both. A test case is included
84098 that can be used to reproduce the bug.
84100 2021-10-19 Tom Tromey <tromey@adacore.com>
84102 Always use std::function for self-tests
84103 Now that there is a register_test variant that accepts std::function,
84104 it seems to me that the 'selftest' struct and accompanying code is
84105 obsolete -- simply always using std::function is simpler. This patch
84106 implements this idea.
84108 2021-10-19 Daniel Black <daniel@mariadb.org>
84110 Fix PR gdb/17917 Lookup build-id in remote binaries
84111 GDB doesn't support loading debug files using build-id from remote
84112 target filesystems.
84114 This is the case when gdbserver attached to a process and a gdb target
84115 remote occurs over tcp.
84117 With this change we make build-id lookups possible:
84119 (gdb) show debug-file-directory
84120 The directory where separate debug symbols are searched for is "/usr/local/lib/debug".
84121 (gdb) set debug-file-directory /usr/lib/debug
84123 The current system root is "target:".
84124 (gdb) target extended-remote :46615
84125 Remote debugging using :46615
84126 warning: Can not parse XML target description; XML support was disabled at compile time
84127 Reading /usr/sbin/mariadbd from remote target...
84128 warning: File transfers from remote targets can be slow. Use "set sysroot" to access files locally instead.
84129 Reading /usr/sbin/mariadbd from remote target...
84130 Reading symbols from target:/usr/sbin/mariadbd...
84131 Reading /usr/lib/debug/.build-id/6e/0a874dca5a7ff831396ddc0785d939a192efe3.debug from remote target...
84132 Reading /usr/lib/debug/.build-id/6e/0a874dca5a7ff831396ddc0785d939a192efe3.debug from remote target...
84133 Reading symbols from target:/usr/lib/debug/.build-id/6e/0a874dca5a7ff831396ddc0785d939a192efe3.debug...
84134 Reading /lib/x86_64-linux-gnu/libpcre2-8.so.0 from remote target...
84137 Before this change, the lookups would have been (GNU gdb (GDB) Fedora 10.2-3.fc34):
84139 (gdb) target extended-remote :46615
84140 Remote debugging using :46615
84141 Reading /usr/sbin/mariadbd from remote target...
84142 warning: File transfers from remote targets can be slow. Use "set sysroot" to access files locally instead.
84143 Reading /usr/sbin/mariadbd from remote target...
84144 Reading symbols from target:/usr/sbin/mariadbd...
84145 Reading /usr/sbin/0a874dca5a7ff831396ddc0785d939a192efe3.debug from remote target...
84146 Reading /usr/sbin/.debug/0a874dca5a7ff831396ddc0785d939a192efe3.debug from remote target...
84147 Reading /usr/lib/debug//usr/sbin/0a874dca5a7ff831396ddc0785d939a192efe3.debug from remote target...
84148 Reading /usr/lib/debug/usr/sbin//0a874dca5a7ff831396ddc0785d939a192efe3.debug from remote target...
84149 Reading target:/usr/lib/debug/usr/sbin//0a874dca5a7ff831396ddc0785d939a192efe3.debug from remote target...
84150 Missing separate debuginfo for target:/usr/sbin/mariadbd
84151 Try: dnf --enablerepo='*debug*' install /usr/lib/debug/.build-id/6e/0a874dca5a7ff831396ddc0785d939a192efe3.debug
84152 (No debugging symbols found in target:/usr/sbin/mariadbd)
84154 Observe it didn't look for
84155 /usr/lib/debug/.build-id/6e/0a874dca5a7ff831396ddc0785d939a192efe3.debug
84156 on the remote target (where it is) and expected them to be installed
84159 As a minor optimization, this also changes the build-id lookup such that
84160 if sysroot is empty, no second lookup of the same location is performed.
84162 Change-Id: I5181696d271c325a25a0805a8defb8ab7f9b3f55
84163 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=17917
84165 2021-10-19 Nick Clifton <nickc@redhat.com>
84167 Fix a potential illegal memory access when testing for a special LTO symbol name.
84168 bfd * linker.c (_bfd_generic_link_add_one_symbol): Test for a NULL
84169 name before checking to see if the symbol is __gnu_lto_slim.
84170 * archive.c (_bfd_compute_and_write_armap): Likewise.
84172 * nm.c (filter_symbols): Test for a NULL name before checking to
84173 see if the symbol is __gnu_lto_slim.
84174 * objcopy.c (filter_symbols): Likewise.
84176 2021-10-19 GDB Administrator <gdbadmin@sourceware.org>
84178 Automatic date update in version.in
84180 2021-10-18 Weimin Pan <weimin.pan@oracle.com>
84182 CTF: incorrect underlying type setting for enumeration types
84183 A bug was filed against the incorrect underlying type setting for
84184 an enumeration type, which was caused by a copy and paste error.
84185 This patch fixes the problem by setting it by calling objfile_int_type,
84186 which was originally dwarf2_per_objfile::int_type, with ctf_type_size bits.
84187 Also add error checking on ctf_func_type_info call.
84189 2021-10-18 GDB Administrator <gdbadmin@sourceware.org>
84191 Automatic date update in version.in
84193 2021-10-17 Alan Modra <amodra@gmail.com>
84195 PR28459, readelf issues bogus warning
84196 I'd missed the fact that the .debug_rnglists dump doesn't exactly
84197 display the contents of the section. Instead readelf rummages through
84198 .debug_info looking for DW_AT_ranges entries, then displays the
84199 entries in .debug_rnglists pointed at, sorted. A simpler dump of the
84200 actual section contents might be more useful and robust, but it was
84201 likely done that way to detect overlap and holes.
84203 Anyway, the headers in .debug_rnglists besides the first are ignored,
84204 and limiting to the unit length of the first header fails if there is
84205 more than one unit.
84208 * dwarf.c (display_debug_ranges): Don't constrain data to length
84211 2021-10-17 GDB Administrator <gdbadmin@sourceware.org>
84213 Automatic date update in version.in
84215 2021-10-16 H.J. Lu <hjl.tools@gmail.com>
84217 ld: Adjust pr28158.rd for glibc 2.34
84218 Adjust pr28158.rd for glibc 2.34:
84220 $ readelf -W --dyn-syms tmpdir/pr28158
84222 Symbol table '.dynsym' contains 4 entries:
84223 Num: Value Size Type Bind Vis Ndx Name
84224 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND
84225 1: 0000000000000000 0 FUNC GLOBAL DEFAULT UND __libc_start_main@GLIBC_2.34 (2)
84226 2: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__
84227 3: 000000000040401c 4 OBJECT GLOBAL DEFAULT 23 foo@VERS_2.0 (3)
84232 $ readelf -W --dyn-syms tmpdir/pr28158
84234 Symbol table '.dynsym' contains 4 entries:
84235 Num: Value Size Type Bind Vis Ndx Name
84236 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND
84237 1: 0000000000000000 0 FUNC GLOBAL DEFAULT UND __libc_start_main@GLIBC_2.2.5 (3)
84238 2: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__
84239 3: 000000000040401c 4 OBJECT GLOBAL DEFAULT 23 foo@VERS_2.0 (2)
84243 * testsuite/ld-elf/pr28158.rd: Adjusted for glibc 2.34.
84245 2021-10-16 GDB Administrator <gdbadmin@sourceware.org>
84247 Automatic date update in version.in
84249 2021-10-15 GDB Administrator <gdbadmin@sourceware.org>
84251 Automatic date update in version.in
84253 2021-10-14 Carl Love <cel@us.ibm.com>
84255 Powerpc: Add support for openat and fstatat syscalls
84256 [gdb] update ppc-linux-tdep.c
84258 Add argument to ppc_canonicalize_syscall for the wordsize.
84259 Add syscall entries for the openat and fstatat system calls.
84261 2021-10-14 Tom de Vries <tdevries@suse.de>
84263 [gdb/testsuite] Add .debug_loc support in dwarf assembler
84264 Add .debug_loc support in the dwarf assembler, and use it in new test-case
84265 gdb.dwarf2/loc-sec-offset.exp (which is based on
84266 gdb.dwarf2/loclists-sec-offset.exp).
84268 Tested on x86_64-linux.
84270 2021-10-14 Alan Modra <amodra@gmail.com>
84272 [GOLD] Re: PowerPC64: Don't pretend to support multi-toc
84273 We can't get at section->address() until everything is laid out, so
84274 trying to generalise the offset calculation rather than using a value
84275 of 0x8000 (the old object->toc_base_offset()) was bound to fail.
84276 got->g_o_t() is a little better than a hard-coded 0x8000.
84278 * powerpc.cc (Target_powerpc::Scan::local, global): Don't use
84279 toc_pointer() here.
84281 2021-10-14 Alan Modra <amodra@gmail.com>
84283 [GOLD] Two GOT sections for PowerPC64
84284 Split .got into two piece, one with the header and entries for small
84285 model got entries, the other with entries for medium/large model got
84286 entries. The idea is to better support mixed pcrel/non-pcrel code
84287 where non-pcrel small-model .toc entries need to be within 32k of the
84290 * target.h (Target::tls_offset_for_local): Add got param.
84291 (Target::tls_offset_for_global): Likewise.
84292 (Target::do_tls_offset_for_local, do_tls_offset_for_global): Likewise.
84293 * output.h (Output_data_got::Got_entry::write): Add got param.
84294 * output.cc (Output_data_got::Got_entry::write): Likewise, pass to
84295 tls_offset_for_local/global calls.
84296 (Output_data_got::do_write): Adjust to suit.
84297 * s390.cc (Target_s390::do_tls_offset_for_local): Likewise.
84298 (Target_s390::do_tls_offset_for_global): Likewise.
84299 * powerpc.cc (enum Got_type): Extend with small types, move from
84300 class Target_powerpc.
84301 (Target_powerpc::biggot_): New.
84302 (Traget_powerpc::do_tls_offset_for_local, do_tls_offset_for_global,
84303 got_size, got_section, got_base_offset): Handle biggot_.
84304 (Target_powerpc::do_define_standard_symbols): Adjust.
84305 (Target_powerpc::make_plt_section, do_finalize_sections): Likewise.
84306 (Output_data_got_powerpc::Output_data_got_powerpc): Only make
84307 64-bit header for small got section.
84308 (Output_data_got_powerpc::g_o_t): Only return a result for small
84310 (Output_data_got_powerpc::write): Only write small got section
84312 (Target_powerpc::Scan::local, global): Select small/big Got_type
84313 and section to suit reloc.
84314 (Target_powerpc::Relocate::relocate): Similarly.
84315 (Sort_toc_sections): Rewrite.
84317 2021-10-14 Alan Modra <amodra@gmail.com>
84319 [GOLD] PowerPC64: Don't pretend to support multi-toc
84320 Code in powerpc.cc is pretending to support a per-object toc pointer
84321 value, but powerpc gold has no real support for multi-toc. This patch
84322 removes the pretense, tidying quite a lot in preparation for a
84323 followup patch. If multi-toc is ever to be supported, don't revert
84324 this patch but start by adding object parameter to toc_pointer() and
84325 an object to Branch_stub_key.
84327 * powerpc.cc (Powerpc_relobj::toc_base_offset): Delete.
84328 (Target_powerpc::toc_pointer): New function. Use throughout.
84329 (Target_powerpc::got_base_offset): New function. Use throughout..
84330 (Output_data_got_powerpc::got_base_offset): ..in place of
84332 (Output_data_got_powerpc::Output_data_got_powerpc): Init
84333 header_index_ to -1u for 64-bit, and make header here.
84334 (Output_data_got_powerpc::set_final_data_size, reserve_ent): Don't
84335 make 64-bit header here.
84336 (Output_data_got_powerpc::g_o_t): Return toc pointer offset in
84337 section for 64-bit. Use throughout.
84338 (Stub_table): Remove toc_base_off_ from Branch_stub_key, and
84339 object param on add_long_branch_entry and find_long_branch_entry.
84342 2021-10-14 Alan Modra <amodra@gmail.com>
84344 Re: s12z/disassembler: call memory_error_func when appropriate
84345 Adjust for commit ba7c18a48457.
84347 * testsuite/gas/s12z/truncated.d: Update expected output.
84349 2021-10-14 GDB Administrator <gdbadmin@sourceware.org>
84351 Automatic date update in version.in
84353 2021-10-13 Tom de Vries <tdevries@suse.de>
84355 [gdb/exp] Improve <error reading variable> message
84356 When printing a variable x in a subroutine foo:
84359 integer(4) :: x (*)
84363 where x is an array with unknown bounds, we get:
84365 $ gdb -q -batch outputs/gdb.fortran/array-no-bounds/array-no-bounds \
84369 Breakpoint 1 at 0x4005cf: file array-no-bounds.f90, line 18.
84371 Breakpoint 1, foo (x=...) at array-no-bounds.f90:18
84373 $1 = <error reading variable>
84376 Improve the error message by printing the details of the error, such that we
84379 $1 = <error reading variable: failed to get range bounds>
84382 This is a change in gdb/valprint.c, and grepping through the sources reveals
84383 that this is a common pattern.
84385 Tested on x86_64-linux.
84387 2021-10-13 Carl Love <cel@us.ibm.com>
84389 PPC fix for stfiwx instruction (and additional stores with primary opcode of 31)
84390 [gdb] Fix address being recorded in rs6000-tdep.c, ppc_process_record_op31.
84392 The GDB record function was recording the variable addr that was passed in
84393 rather than the calculated effective address (ea) by the
84394 ppc_process_record_op31 function.
84396 2021-10-13 Andrew Burgess <andrew.burgess@embecosm.com>
84398 gdb: improve error reporting from the disassembler
84399 If the libopcodes disassembler returns a negative value then this
84400 indicates that the disassembly failed for some reason. In disas.c, in
84401 the function gdb_disassembler::print_insn we can see how this is
84402 handled; when we get a negative value back, we call the memory_error
84403 function, which throws an exception.
84405 The problem here is that the address used in the memory_error call is
84406 gdb_disassembler::m_err_memaddr, which is set in
84407 gdb_disassembler::dis_asm_memory_error, which is called from within
84408 the libopcodes disassembler through the
84409 disassembler_info::memory_error_func callback.
84411 However, for this to work correctly, every time the libopcodes
84412 disassembler returns a negative value, the libopcodes disassembler
84413 must have first called the memory_error_func callback.
84415 My first plan was to make m_err_memaddr a gdb::optional, and assert
84416 that it always had a value prior to calling memory_error, however, a
84417 quick look in opcodes/*-dis.c shows that there _are_ cases where a
84418 negative value is returned without first calling the memory_error_func
84419 callback, for example in arc-dis.c and cris-dis.c.
84421 Now, I think that a good argument can be made that these disassemblers
84422 must therefore be broken, except for the case where we can't read
84423 memory, we should always be able to disassemble the memory contents to
84424 _something_, even if it's just '.word 0x....'. However, I certainly
84425 don't plan to go and fix all of the disassemblers.
84427 What I do propose to do then, is make m_err_memaddr a gdb::optional,
84428 but now, instead of always calling memory_error, I add a new path
84429 which just calls error complaining about an unknown error. This new
84430 path is only used if m_err_memaddr doesn't have a value (indicating
84431 that the memory_error_func callback was not called).
84433 To test this I just augmented one of the disassemblers to always
84434 return -1, before this patch I see this:
84436 Dump of assembler code for function main:
84437 0x000101aa <+0>: Cannot access memory at address 0x0
84439 And after this commit I now see:
84441 Dump of assembler code for function main:
84442 0x000101aa <+0>: unknown disassembler error (error = -1)
84444 This doesn't really help much, but that's because there's no way to
84445 report non memory errors out of the disasembler, because, it was not
84446 expected that the disassembler would ever report non memory errors.
84448 2021-10-13 Tom de Vries <tdevries@suse.de>
84450 [gdb/testsuite] Fix gdb.fortran/call-no-debug.exp with native-gdbserver
84451 When running test-case gdb.fortran/call-no-debug.exp with target board
84452 native-gdbserver, I run into:
84454 (gdb) PASS: gdb.fortran/call-no-debug.exp: print string_func_ (&'abcdefg', 3)
84455 call (integer) string_func_ (&'abcdefg', 3)^M
84457 (gdb) FAIL: gdb.fortran/call-no-debug.exp: call (integer) string_func_ (&'abcdefg', 3)
84460 The problem is that gdb_test is used to match inferior output.
84462 Fix this by using gdb_test_stdio.
84464 Tested on x86_64-linux.
84466 2021-10-13 Tom de Vries <tdevries@suse.de>
84468 [gdb/testsuite] Require use_gdb_stub == 0 where appropriate
84469 When running with target board native-gdbserver, we run into a number of FAILs
84470 due to use of the start command (and similar), which is not supported when
84474 - requiring use_gdb_stub == 0 for the entire test-case, or
84475 - guarding some tests in the test-case with use_gdb_stub == 0.
84477 Tested on x86_64-linux.
84479 2021-10-13 Tom de Vries <tdevries@suse.de>
84481 [gdb/testsuite] Fix test name in gdb.python/python.exp
84482 When running test-case gdb.python/python.exp, we have:
84484 PASS: gdb.python/python.exp: starti via gdb.execute, not from tty
84485 PASS: gdb.python/python.exp: starti via interactive input
84488 The two tests are instances of the same test, with different values for
84489 starti command argument from_tty, so it's strange that the test names are so
84492 This is due to using a gdb_test nested in a gdb_test_multiple, with the inner
84493 one using a different test name than the outer one. [ That could still make
84494 sense if both produced passes, but that's not the case here. ]
84496 Fix this by using $gdb_test_name, such that we have:
84498 PASS: gdb.python/python.exp: starti via gdb.execute, not from tty
84499 PASS: gdb.python/python.exp: starti via gdb.execute, from tty
84502 Also make this more readable by using variables.
84504 Tested on x86_64-linux.
84506 2021-10-13 Tom de Vries <tdevries@suse.de>
84508 [gdb/testsuite] Fix gdb.base/batch-exit-status.exp with native-gdbserver
84509 When running test-case gdb.base/batch-exit-status.exp with target board
84510 native-gdbserver, I run into (added missing double quotes for clarity):
84512 builtin_spawn $build/gdb/testsuite/../../gdb/gdb -nw -nx \
84513 -data-directory $build/gdb/testsuite/../data-directory \
84514 -iex "set height 0" -iex "set width 0" \
84515 -ex "set auto-connect-native-target off" \
84516 -iex "set sysroot" -batch ""^M
84517 : No such file or directory.^M
84518 PASS: gdb.base/batch-exit-status.exp: 1x: \
84519 No such file or directory: [lindex $result 2] == 0
84520 FAIL: gdb.base/batch-exit-status.exp: 1x: \
84521 No such file or directory: [lindex $result 3] == $expect_status
84524 As in commit a02a90c114c "[gdb/testsuite] Set sysroot earlier in
84525 local-board.exp", the problem is the use of -ex for
84526 "set auto-connect-native-target off", which makes that the last command to
84527 be executed, and consequently determines the return status.
84529 Fix this by using -iex instead.
84531 Tested on x86_64-linux.
84533 2021-10-13 Tom de Vries <tdevries@suse.de>
84535 [gdb/testsuite] Remove quit in gdb.arch/i386-mpx.exp
84536 When running test-case gdb.arch/i386-mpx.exp with target board
84537 native-gdbserver, I run into:
84539 (gdb) PASS: gdb.arch/i386-mpx.exp: verify size for bnd0
84540 Remote debugging from host ::1, port 42328^M
84542 A debugging session is active.^M
84544 Inferior 1 [process 19679] will be killed.^M
84546 Quit anyway? (y or n) monitor exit^M
84547 Please answer y or n.^M
84548 A debugging session is active.^M
84550 Inferior 1 [process 19679] will be killed.^M
84552 Quit anyway? (y or n) WARNING: Timed out waiting for EOF in server after monitor exit
84555 The problem is that the test-case sends a quit at the end (without verifying
84556 the result of this in any way):
84561 Fix this by removing the quit.
84563 Tested on x86_64-linux.
84565 2021-10-13 GDB Administrator <gdbadmin@sourceware.org>
84567 Automatic date update in version.in
84569 2021-10-12 GDB Administrator <gdbadmin@sourceware.org>
84571 Automatic date update in version.in
84573 2021-10-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
84575 [ARM] Add support for M-profile MVE extension
84576 This patch adds support for the M-profile MVE extension, which includes the
84579 - New M-profile XML feature m-profile-mve
84580 - MVE vector predication status and control register (VPR)
84581 - p0 pseudo register (contained in the VPR)
84582 - q0 ~ q7 pseudo vector registers
84584 - Documentation update
84586 Pseudo register p0 is the least significant bits of vpr and can be accessed
84587 as $p0 or displayed through $vpr. For more information about the register
84588 layout, please refer to [1].
84590 The q0 ~ q7 registers map back to the d0 ~ d15 registers, two d registers
84593 The register dump looks like this:
84609 sp 0x0 0x0 <__Vectors>
84611 pc 0xd0c 0xd0c <Reset_Handler>
84612 xpsr 0x1000000 16777216
84613 d0 0 (raw 0x0000000000000000)
84614 d1 0 (raw 0x0000000000000000)
84615 d2 0 (raw 0x0000000000000000)
84616 d3 0 (raw 0x0000000000000000)
84617 d4 0 (raw 0x0000000000000000)
84618 d5 0 (raw 0x0000000000000000)
84619 d6 0 (raw 0x0000000000000000)
84620 d7 0 (raw 0x0000000000000000)
84621 d8 0 (raw 0x0000000000000000)
84622 d9 0 (raw 0x0000000000000000)
84623 d10 0 (raw 0x0000000000000000)
84624 d11 0 (raw 0x0000000000000000)
84625 d12 0 (raw 0x0000000000000000)
84626 d13 0 (raw 0x0000000000000000)
84627 d14 0 (raw 0x0000000000000000)
84628 d15 0 (raw 0x0000000000000000)
84630 vpr 0x0 [ P0=0 MASK01=0 MASK23=0 ]
84631 s0 0 (raw 0x00000000)
84632 s1 0 (raw 0x00000000)
84633 s2 0 (raw 0x00000000)
84634 s3 0 (raw 0x00000000)
84635 s4 0 (raw 0x00000000)
84636 s5 0 (raw 0x00000000)
84637 s6 0 (raw 0x00000000)
84638 s7 0 (raw 0x00000000)
84639 s8 0 (raw 0x00000000)
84640 s9 0 (raw 0x00000000)
84641 s10 0 (raw 0x00000000)
84642 s11 0 (raw 0x00000000)
84643 s12 0 (raw 0x00000000)
84644 s13 0 (raw 0x00000000)
84645 s14 0 (raw 0x00000000)
84646 s15 0 (raw 0x00000000)
84647 s16 0 (raw 0x00000000)
84648 s17 0 (raw 0x00000000)
84649 s18 0 (raw 0x00000000)
84650 s19 0 (raw 0x00000000)
84651 s20 0 (raw 0x00000000)
84652 s21 0 (raw 0x00000000)
84653 s22 0 (raw 0x00000000)
84654 s23 0 (raw 0x00000000)
84655 s24 0 (raw 0x00000000)
84656 s25 0 (raw 0x00000000)
84657 s26 0 (raw 0x00000000)
84658 s27 0 (raw 0x00000000)
84659 s28 0 (raw 0x00000000)
84660 s29 0 (raw 0x00000000)
84661 s30 0 (raw 0x00000000)
84662 s31 0 (raw 0x00000000)
84663 q0 {u8 = {0x0 <repeats 16 times>}, u16 = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}, u32 = {0x0, 0x0, 0x0, 0x0}, u64 = {0x0, 0x0}, f32 = {0x0, 0x0, 0x0, 0x0}, f64 = {0x0, 0x0}}
84664 q1 {u8 = {0x0 <repeats 16 times>}, u16 = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}, u32 = {0x0, 0x0, 0x0, 0x0}, u64 = {0x0, 0x0}, f32 = {0x0, 0x0, 0x0, 0x0}, f64 = {0x0, 0x0}}
84665 q2 {u8 = {0x0 <repeats 16 times>}, u16 = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}, u32 = {0x0, 0x0, 0x0, 0x0}, u64 = {0x0, 0x0}, f32 = {0x0, 0x0, 0x0, 0x0}, f64 = {0x0, 0x0}}
84666 q3 {u8 = {0x0 <repeats 16 times>}, u16 = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}, u32 = {0x0, 0x0, 0x0, 0x0}, u64 = {0x0, 0x0}, f32 = {0x0, 0x0, 0x0, 0x0}, f64 = {0x0, 0x0}}
84667 q4 {u8 = {0x0 <repeats 16 times>}, u16 = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}, u32 = {0x0, 0x0, 0x0, 0x0}, u64 = {0x0, 0x0}, f32 = {0x0, 0x0, 0x0, 0x0}, f64 = {0x0, 0x0}}
84668 q5 {u8 = {0x0 <repeats 16 times>}, u16 = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}, u32 = {0x0, 0x0, 0x0, 0x0}, u64 = {0x0, 0x0}, f32 = {0x0, 0x0, 0x0, 0x0}, f64 = {0x0, 0x0}}
84669 q6 {u8 = {0x0 <repeats 16 times>}, u16 = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}, u32 = {0x0, 0x0, 0x0, 0x0}, u64 = {0x0, 0x0}, f32 = {0x0, 0x0, 0x0, 0x0}, f64 = {0x0, 0x0}}
84670 q7 {u8 = {0x0 <repeats 16 times>}, u16 = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}, u32 = {0x0, 0x0, 0x0, 0x0}, u64 = {0x0, 0x0}, f32 = {0x0, 0x0, 0x0, 0x0}, f64 = {0x0, 0x0}}
84673 Built and regtested with a simulator.
84675 [1] https://developer.arm.com/documentation/ddi0553/bn
84677 Co-Authored-By: Luis Machado <luis.machado@linaro.org>
84679 2021-10-11 Luis Machado <luis.machado@linaro.org>
84681 [ARM] Refactor pseudo register numbering
84682 The pseudo register handling for ARM uses some hardcoded constants to
84683 determine types and names. In preparation to the upcoming MVE support
84684 patch (that will add another pseudo register), this patch refactors and
84685 reorganizes things in order to simplify handling of future pseudo registers.
84687 We keep track of the first pseudo register number in a group and the number of
84688 pseudo registers in that group.
84690 Right now we only have the S and Q pseudo registers.
84692 2021-10-11 Luis Machado <luis.machado@linaro.org>
84694 [ARM] Small refactoring of arm gdbarch initialization
84695 This is in preparation to MVE support, where we will define another
84696 pseudo register. We need to define the pseudo register numbers *after*
84697 accounting for all the registers in the XML description, so move
84698 the call to tdesc_use_registers up.
84700 If we don't do it, GDB's register count won't consider registers contained
84701 in the XML but ignored by GDB, throwing the register numbering off.
84703 2021-10-11 Luis Machado <luis.machado@linaro.org>
84705 [ARM] Refactor some constants
84706 In preparation for the MVE extension patch, this one refactors some of
84707 the register-related constants we have for ARM.
84709 Basically I'm separating counting constants from numbering constants.
84711 For example, ARM_A1_REGNUM is a numbering constant, whereas ARM_NUM_ARG_REGS
84712 is a counting constant.
84714 2021-10-11 Tom de Vries <tdevries@suse.de>
84716 [gdb/testsuite] Fix FAIL in gdb.mi/mi-var-child-f.exp
84717 When running test-case gdb.mi/mi-var-child-f.exp on openSUSE Tumbleweed
84718 (with glibc 2.34) I run into:
84721 PASS: gdb.mi/mi-var-child-f.exp: mi runto prog_array
84722 Expecting: ^(-var-create array \* array[^M
84723 ]+)?(\^done,name="array",numchild="[0-9]+",value=".*",type=.*,has_more="0"[^M
84726 -var-create array * array^M
84727 &"Attempt to use a type name as an expression.\n"^M
84728 ^error,msg="-var-create: unable to create variable object"^M
84730 FAIL: gdb.mi/mi-var-child-f.exp: create local variable array (unexpected output)
84733 The problem is that the name array is used both:
84734 - as the name for a local variable
84735 - as the name of a type in glibc, in file malloc/dynarray-skeleton.c, as included
84736 by nss/nss_files/files-hosts.c.
84738 Fix this by ignoring the shared lib symbols.
84740 Likewise in a couple of other fortran tests.
84742 Tested on x86_64-linux.
84744 2021-10-11 Andrew Burgess <andrew.burgess@embecosm.com>
84746 z80/disassembler: call memory_error_func when appropriate
84747 If a call to the read_memory_func fails then we should call the
84748 memory_error_func to notify the user of the disassembler of the
84749 address that was a problem.
84751 Without this GDB will report all memory errors as being at address
84756 * z80-dis.c (fetch_data): Call memory_error_func if the
84757 read_memory_func call fails.
84759 2021-10-11 Andrew Burgess <andrew.burgess@embecosm.com>
84761 s12z/disassembler: call memory_error_func when appropriate
84762 If a call to the read_memory_func fails then we should call the
84763 memory_error_func to notify the user of the disassembler of the
84764 address that was a problem.
84766 Without this GDB will report all memory errors as being at address
84771 * s12z-disc.c (abstract_read_memory): Call memory_error_func if
84772 the read_memory_func call fails.
84774 2021-10-11 Tom de Vries <tdevries@suse.de>
84776 [gdb/testsuite] Fix double debug info in gdb.dwarf2/dw2-ref-missing-frame.exp
84777 A mistake slipped in in commit a5ea23036d8 "[gdb/testsuite] Use function_range
84778 in gdb.dwarf2/dw2-ref-missing-frame.exp".
84780 Before the commit the main file was compiled with debug info, and the two
84783 if {[prepare_for_testing_full "failed to prepare" \
84784 [list $testfile {} $srcfile {} $srcfuncfile {} \
84785 $srcmainfile debug]]} {
84788 After the commit, all were compiled with debug info, and consequently, there
84789 are two versions of debug info for $srcfuncfile. This shows up as a FAIL when
84790 running the test-case with target boards readnow and cc-with-debug-names.
84792 Fix this by using prepare_for_testing_full, as before.
84794 Tested on x86_64-linux.
84796 Fixes: a5ea23036d8 ("[gdb/testsuite] Use function_range in
84797 gdb.dwarf2/dw2-ref-missing-frame.exp")
84799 2021-10-11 Tom de Vries <tdevries@suse.de>
84801 [gdb/testsuite] Use require for ensure_gdb_index
84804 if { [ensure_gdb_index $binfile] == -1 } {
84810 require {ensure_gdb_index $binfile} != -1
84812 and consequently, add a missing UNTESTED message.
84814 Tested on x86_64-linux, both with native and target board readnow.
84816 2021-10-11 Tom de Vries <tdevries@suse.de>
84818 [gdb/testsuite] Handle readnow in ensure_gdb_index
84819 When running test-case gdb.base/with-mf.exp with target board readnow, I run
84822 FAIL: gdb.base/with-mf.exp: check if index present
84824 This is since commit 6010fb0c49e "[gdb/testsuite] Fix full buffer in
84825 gdb.rust/dwindex.exp".
84827 Before that commit, the proc ensure_gdb_index would treat the line:
84829 .gdb_index: faked for "readnow"^M
84831 as proof that an index is already present (which is incorrect).
84833 Now, instead it generates aforementioned FAIL and continues to generate an
84836 Fix this by explicitly handling the readnow case in proc ensure_gdb_index,
84837 such that we bail out instead.
84839 Tested on x86_64-linux.
84841 2021-10-11 Tom de Vries <tdevries@suse.de>
84843 [gdb/testsuite] Fix gdb.dwarf2/gdb-add-index-symlink.exp
84844 The test-case gdb.dwarf2/gdb-add-index-symlink.exp interpretes a failure to
84845 add an index as a failure to add an index for a symlink:
84847 if { [ensure_gdb_index $symlink] == -1 } {
84848 fail "Unable to call gdb-add-index with a symlink to a symfile"
84853 However, it's possible that the gdb-add-index also fails with a regular
84854 file. Add a check for that situation.
84856 Tested on x86_64-linux.
84858 2021-10-11 Tom de Vries <tdevries@suse.de>
84860 [gdb/testsuite] Add proc require in lib/gdb.exp
84861 Add a new proc require in lib/gdb.exp, and use it to shorten:
84863 if { [gdb_skip_xml_test] } {
84864 # Valgrind gdbserver requires gdb with xml support.
84865 untested "missing xml support"
84871 require gdb_skip_xml_test 0
84874 Tested on x86_64-linux, both with and without a trigger patch that forces
84875 gdb_skip_xml_test to return 1.
84877 2021-10-11 Michael Forney <mforney@mforney.org>
84879 bfd: Remove use of void pointer arithmetic
84880 This is not valid in ISO C. Instead, use a pointer to bfd_byte.
84882 * peicode.h (pe_bfd_object_p): Remove use of void pointer
84885 2021-10-11 GDB Administrator <gdbadmin@sourceware.org>
84887 Automatic date update in version.in
84889 2021-10-10 GDB Administrator <gdbadmin@sourceware.org>
84891 Automatic date update in version.in
84893 2021-10-09 Tom de Vries <tdevries@suse.de>
84895 [gdb] Make execute_command_to_string return string on throw
84896 The pattern for using execute_command_to_string is:
84898 std::string output;
84899 output = execute_fn_to_string (fn, term_out);
84902 This results in a problem when using it in a try/catch:
84906 output = execute_fn_to_string (fn, term_out)
84908 catch (const gdb_exception &e)
84914 If an expection was thrown during execute_fn_to_string, then the output
84915 remains unassigned, while it could be worthwhile to known what output was
84916 generated by gdb before the expection was thrown.
84918 Fix this by returning the string using a parameter instead:
84920 execute_fn_to_string (output, fn, term_out)
84923 Also add a variant without string parameter, to support places where the
84924 function is used while ignoring the result:
84926 execute_fn_to_string (fn, term_out)
84929 Tested on x86_64-linux.
84931 2021-10-09 Tom de Vries <tdevries@suse.de>
84933 [gdb/testsuite] Add check-readmore
84934 Consider the gdb output:
84936 27 return SYSCALL_CANCEL (nanosleep, requested_time, remaining);^M
84938 Thread 2 "run-attach-whil" stopped.^M
84941 When trying to match the gdb prompt using gdb_test which uses '$gdb_prompt $',
84942 it may pass or fail.
84944 This sort of thing needs to be fixed (see commit b0e2f96b56b), but there's
84945 currently no way to reliably find this type of FAILs.
84947 We have check-read1, but that one actually make the test pass reliably.
84949 We need something like the opposite of check-read1: something that makes
84950 expect read a bit slower, or more exhaustively.
84952 Add a new test target check-readmore that implements this.
84954 There are two methods of implementing this in read1.c:
84955 - the first method waits a bit before doing a read
84956 - the second method does a read and then decides whether to
84957 return or to wait a bit and do another read, and so on.
84959 The second method is potentially faster, has less risc of timeout and could
84960 potentially detect more problems. The first method has a simpler
84963 The second method is enabled by default. The default waiting period is 10
84966 The first method can be enabled using:
84968 $ export READMORE_METHOD=1
84970 and the waiting period can be specified in miliseconds using:
84972 $ export READMORE_SLEEP=9
84975 Also a log file can be specified using:
84977 $ export READMORE_LOG=$(pwd -P)/LOG
84980 Tested on x86_64-linux.
84982 Testing with check-readmore showed these regressions:
84984 FAIL: gdb.base/bp-cmds-continue-ctrl-c.exp: run: stop with control-c (continue)
84985 FAIL: gdb.base/bp-cmds-continue-ctrl-c.exp: attach: stop with control-c (continue)
84988 I have not been able to find a problem in the test-case, and I think it's the
84989 nature of both the test-case and readmore that makes it run longer. Make
84990 these pass by increasing the alarm timeout from 60 to 120 seconds.
84992 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=27957
84994 2021-10-09 Tom de Vries <tdevries@suse.de>
84996 [gdb/testsuite] Fix fortran module tests with stressed cpu
84997 When running these test-cases:
84998 - gdb.fortran/info-modules.exp
84999 - gdb.fortran/module.exp
85000 - gdb.mi/mi-fortran-modules.exp
85001 in conjunction with:
85003 $ stress -c $(($(cat /proc/cpuinfo | grep -c "^processor") + 1))
85005 I run into timeouts.
85008 - "set auto-solib-add off" to avoid symbols of shared libs
85009 (which doesn't work for libc, now that libpthread_name_p has been
85010 updated to match libc)
85011 - "nosharedlibrary" to avoid symbols of libc
85013 Tested on x86_64-linux.
85015 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28133
85017 2021-10-09 Guillermo E. Martinez <guillermo.e.martinez@oracle.com>
85019 PR28415, invalid read in xtensa_read_table_entries
85022 * elf32-xtensa.c (xtensa_read_table_entries): Handle error
85023 return from retrieve_contents.
85025 2021-10-09 GDB Administrator <gdbadmin@sourceware.org>
85027 Automatic date update in version.in
85029 2021-10-08 Tom de Vries <tdevries@suse.de>
85031 [gdb/testsuite] Fix gdb.base/info-types-c++.exp with stressed cpu
85032 When running test-case gdb.base/info-types-c++.exp in conjunction with:
85034 $ stress -c $(($(cat /proc/cpuinfo | grep -c "^processor") + 1))
85038 FAIL: gdb.base/info-types-c++.exp: info types (timeout)
85041 Fix this by setting auto-solib-add to off.
85043 Tested on x86_64-linux.
85045 2021-10-08 Tom de Vries <tdevries@suse.de>
85047 [gdb/testsuite] Fix gdb.base/info_sources_2.exp with check-read1
85048 When running test-case gdb.base/info_sources_2.exp with check-read1, I run
85051 FAIL: gdb.base/info_sources_2.exp: args: : info sources (timeout)
85054 Fix this by consuming a "$src1, $src2, ..., $srcn: line bit by bit rather than
85057 Also add the missing handling of "Objfile has no debug information".
85059 Tested on x86_64-linux.
85061 2021-10-08 Tom de Vries <tdevries@suse.de>
85063 [gdb/testsuite] Fix gdb.mi/gdb2549.exp with check-read1
85064 When running test-case gdb.mi/gdb2549.exp with check-read1, I run into:
85066 FAIL: gdb.mi/gdb2549.exp: register values x (timeout)
85069 Fix this by applying the same fix as for "register values t" in commit
85070 478e490a4df "[gdb/testsuite] Fix gdb.mi/gdb2549.exp with check-read1".
85072 Tested on x86_64-linux.
85074 2021-10-08 Tom de Vries <tdevries@suse.de>
85076 [gdb/testsuite] Fix gdb.base/bt-on-error-and-warning.exp with check-read1
85077 When running test-case gdb.base/bt-on-error-and-warning.exp with check-read1,
85080 (gdb) maint internal-error foobar^M
85081 src/gdb/maint.c:82: internal-error: foobar^M
85082 A problem internal to GDB has been detectedFAIL: \
85083 gdb.base/bt-on-error-and-warning.exp: problem=internal-error, mode=on: \
85084 scan for backtrace (GDB internal error)
85085 Resyncing due to internal error.
85089 The corresponding gdb_test_multiple in the test-case contains:
85091 -early -re "^A problem internal to GDB has been detected,\r\n" {
85096 but instead this one triggers in gdb_test_multiple:
85098 -re ".*A problem internal to GDB has been detected" {
85099 fail "$message (GDB internal error)"
85100 gdb_internal_error_resync
85105 Fix this by likewise shortening the regexp to before the comma.
85107 Tested on x86_64-linux.
85109 2021-10-08 Tom de Vries <tdevries@suse.de>
85111 [gdb/testsuite] Add nopie in two test-cases
85112 When running test-case gdb.dwarf2/dw2-restrict.exp on openSUSE Leap 15.2 with
85113 gcc-PIE installed (switching compiler default to -fPIE/-pie), I get:
85115 gdb compile failed, ld: outputs/gdb.dwarf2/dw2-restrict/dw2-restrict0.o: \
85116 warning: relocation in read-only section `.text'
85117 ld: warning: creating DT_TEXTREL in a PIE
85118 UNTESTED: gdb.dwarf2/dw2-restrict.exp: failed to prepare
85121 This is due to using a hardcoded .S file that was generated with -fno-PIE.
85123 Fix this by adding the missing nopie.
85125 Likewise in gdb.arch/amd64-tailcall-noret.exp.
85127 Tested on x86_64-linux.
85129 2021-10-08 GDB Administrator <gdbadmin@sourceware.org>
85131 Automatic date update in version.in
85133 2021-10-07 Tom de Vries <tdevries@suse.de>
85135 [gdb/testsuite] Fix gdb.threads/check-libthread-db.exp with glibc 2.34
85136 When running test-case gdb.threads/check-libthread-db.exp on openSUSE
85137 Tumbleweed (with glibc 2.34) I get:
85141 [Thread debugging using libthread_db enabled]^M
85142 Using host libthread_db library "/lib64/libthread_db.so.1".^M
85143 Stopped due to shared library event:^M
85144 Inferior loaded /lib64/libm.so.6^M
85146 (gdb) FAIL: gdb.threads/check-libthread-db.exp: user-initiated check: continue
85149 The check expect the inferior to load libpthread, but since glibc 2.34
85150 libpthread has been integrated into glibc, and consequently it's no longer
85153 $ ldd outputs/gdb.threads/check-libthread-db/check-libthread-db
85154 linux-vdso.so.1 (0x00007ffe4cae4000)
85155 libm.so.6 => /lib64/libm.so.6 (0x00007f167c77c000)
85156 libc.so.6 => /lib64/libc.so.6 (0x00007f167c572000)
85157 /lib64/ld-linux-x86-64.so.2 (0x00007f167c86e000)
85160 Fix this by updating the regexp to expect libpthread or libc.
85162 Tested on x86_64-linux.
85164 2021-10-07 Tom de Vries <tdevries@suse.de>
85166 [gdb/testsuite] Fix gdb.guile/scm-type.exp with gcc 4.8
85167 With gcc 7.5.0, I get:
85169 (gdb) guile (print (type-range (field-type (type-field (value-type \
85170 (value-dereference f)) "items"))))^M
85172 (gdb) PASS: gdb.guile/scm-type.exp: lang_cpp: test_range: \
85173 on flexible array member: $cmd
85175 but with gcc 4.8.5, I get instead:
85177 (gdb) guile (print (type-range (field-type (type-field (value-type \
85178 (value-dereference f)) "items"))))^M
85180 (gdb) FAIL: gdb.guile/scm-type.exp: lang_cpp: test_range: \
85181 on flexible array member: $cmd
85184 There's a difference in debug info. With gcc 4.8.5, we have:
85186 <2><224>: Abbrev Number: 15 (DW_TAG_member)
85187 <225> DW_AT_name : items
85188 <22b> DW_AT_type : <0x231>
85189 <1><231>: Abbrev Number: 4 (DW_TAG_array_type)
85190 <232> DW_AT_type : <0x105>
85191 <2><23a>: Abbrev Number: 16 (DW_TAG_subrange_type)
85192 <23b> DW_AT_type : <0x11a>
85193 <23f> DW_AT_upper_bound : 0xffffffffffffffff
85195 and with gcc 7.5.0, we have instead:
85197 <2><89f>: Abbrev Number: 12 (DW_TAG_member)
85198 <8a0> DW_AT_name : items
85199 <8a6> DW_AT_type : <0x8ac>
85200 <1><8ac>: Abbrev Number: 17 (DW_TAG_array_type)
85201 <8ad> DW_AT_type : <0x29d>
85202 <2><8b5>: Abbrev Number: 41 (DW_TAG_subrange_type)
85203 <2><8b6>: Abbrev Number: 0
85206 As mentioned in commit 858c8f2c1b9 "gdb/testsuite: adjust
85207 gdb.python/flexible-array-member.exp expected pattern":
85209 Ideally, GDB would present a consistent and documented value for an
85210 array member declared with size 0, regardless of how the debug info
85214 As in gdb.python/flexible-array-member.exp, change the test to accept the two
85217 Tested on x86_64-linux.
85219 2021-10-07 Simon Marchi <simon.marchi@polymtl.ca>
85221 gdb: add accessors for field (and call site) location
85222 Add accessors for the various location values in struct field. This
85223 lets us assert that when we get a location value of a certain kind (say,
85224 bitpos), the field's location indeed contains a value of that kind.
85226 Remove the SET_FIELD_* macros, instead use the new setters directly.
85227 Update the FIELD_* macros used to access field locations to go through
85228 the getters. They will be removed in a subsequent patch.
85230 There are places where the FIELD_* macros are used on call_site_target
85231 structures, because it contains members of the same name (loc_kind and
85232 loc). For now, I have replicated the getters/setters in
85233 call_site_target. But we could perhaps eventually factor them in a
85234 "location" structure that can be used at both places.
85236 Note that the field structure, being zero-initialized, defaults to a
85237 bitpos location with value 0. While writing this patch, I tried to make
85238 it default to an "unset" location, to catch places where we would miss
85239 setting a field's location. However, I found that some places relied on
85240 the default being "bitpos 0", so I left it as-is. This change could
85241 always be done as follow-up work, making these places explicitly set the
85242 "bitpos 0" location.
85244 I found two issues to fix:
85246 - I got some failures in the gdb.base/infcall-nested-structs-c++.exp
85247 test. They were caused by two functions in amd64-tdep.c using
85248 TYPE_FIELD_BITPOS before checking if the location is of the bitpos
85249 kind, which they do indirectly through `field_is_static`. Simply
85250 move getting the bitpos below the field_is_static call.
85252 - I got a failure in gdb.xml/tdesc-regs.exp. It turns out that in
85253 make_gdb_type_enum, we set enum field values using SET_FIELD_BITPOS,
85254 and later access them through FIELD_ENUMVAL. Fix that by using
85255 set_loc_enumval to set the value.
85257 Change-Id: I53d3734916c46457576ba11dd77df4049d2fc1e8
85259 2021-10-07 Philipp Tomsich <philipp.tomsich@vrull.eu>
85261 RISC-V: Support aliases for Zbs instructions
85262 Add aliases for the non-immediate mnemonics of b{set,clr,inv,ext} to
85263 yencode the respective immediate insn b{set,clr,inv,ext}i when the
85264 second source operand is an immediate.
85266 2021-01-11 Philipp Tomsich <philipp.tomsich@vrull.eu>
85269 * testsuite/gas/riscv/b-ext.d: Add tests.
85270 * testsuite/gas/riscv/b-ext.s: Likewise.
85271 * testsuite/gas/riscv/b-ext-64.d: Likewise.
85272 * testsuite/gas/riscv/b-ext-64.s: Likewise.
85274 * riscv-opc.c (riscv_opcodes): Add aliases for Zbs.
85276 Suggested-by: Jan Beulich <jbeulich@suse.com>
85278 2021-10-07 Philipp Tomsich <philipp.tomsich@vrull.eu>
85280 RISC-V: Add support for Zbs instructions
85281 This change adds the Zbs instructions from the Zbs 1.0.0 specification.
85283 https://github.com/riscv/riscv-bitmanip/releases/tag/1.0.0
85284 for the frozen specification.
85286 2021-01-09 Philipp Tomsich <philipp.tomsich@vrull.eu>
85289 * elfxx-riscv.c (riscv_supported_std_z_ext): Added zbs.
85291 * config/tc-riscv.c (riscv_multi_subset_supports): Handle INSN_CLASS_ZBS.
85292 * testsuite/gas/riscv/b-ext.d: Test Zbs instructions.
85293 * testsuite/gas/riscv/b-ext.s: Likewise.
85294 * testsuite/gas/riscv/b-ext-64.d: Likewise.
85295 * testsuite/gas/riscv/b-ext-64.s: Likewise.
85297 * opcode/riscv-opc.h: Added MASK/MATCH/DECLARE_INSN for Zbs.
85298 * opcode/riscv.h (riscv_insn_class): Added INSN_CLASS_ZBS.
85300 * riscv-opc.c (riscv_supported_std_z_ext): Add zbs.
85302 2021-10-07 Philipp Tomsich <philipp.tomsich@vrull.eu>
85304 RISC-V: Update extension version for Zb[abc] to 1.0.0
85305 2021-10-06 Philipp Tomsich <philipp.tomsich@vrull.eu>
85308 * elfxx-riscv.c (riscv_supported_std_z_ext): Update the version
85309 number for zba, zbb and zbc to 1.0.0
85313 - Updated version numbers for zba, zbb and zbc to 1.0.0
85315 2021-10-07 Philipp Tomsich <philipp.tomsich@vrull.eu>
85317 RISC-V: Split Zb[abc] into commented sections
85318 The Zb[abc] opcodes are bundled just below the Privileged opcodes in
85319 riscv_opcodes, possibly giving the appearance that they are part of
85320 the Privileged spec for an uninitiated reader. This separates them
85321 out and adds comments above each section to clearly identify them as
85322 Zba, Zbb or Zbc opcodes.
85324 2021-10-04 Philipp Tomsich <philipp.tomsich@vrull.eu>
85327 * riscv-opc.c: Split of Zb[abc] instructions and add comments.
85329 2021-10-07 Alan Modra <amodra@gmail.com>
85331 PR28423, use-after-free in objdump
85332 XCOFF archives use a bi-directional linked list for file members. So
85333 one member points to both the previous member and the next member.
85334 Members may not be sequentially ordered in the file. This of course
85335 is over-engineered nonsense and an attractive target for fuzzers.
85336 (There is even a free list of members!) The testcase in PR28423 is an
85337 XCOFF archive with one member pointing to itself, which results in
85338 lots of bad behaviour. For example, "ar t" never terminates.
85340 The use-after-free with "objdump -r" happens like this: The first
85341 archive element is opened, its symbols are read and "canonicalized"
85342 for objdump, then relocations are read and printed. Those relocations
85343 use the canonicalized symbols, and also happen to be cached by the
85344 coff bfd backend support. objdump frees the symbols. The next
85345 archive element is then opened. This must be done before the first
85346 element is closed, because finding the next element uses data held in
85347 the currect element. Unfortunately the next element happens to be the
85348 original, so we aren't opening, we're reopening a bfd which has cached
85349 data. When the relocations are printed they use the cached copy
85350 containing references to the freed canonical symbols.
85352 This patch adds a little sanity checking to the XCOFF "open next
85353 archive file" support, so that it rejects archive members pointing at
85354 themselves. That is sufficient to cure this problem. Anything more
85355 is overkill. If someone deliberately fuzzes an XCOFF archive with an
85356 element loop then reports an "ar" bug when it runs forever, they will
85357 find their bug report closed WONTFIX.
85360 * coff-rs6000.c (_bfd_xcoff_read_ar_hdr): Save size occupied
85361 by member name in areltdata.extra_size.
85362 (_bfd_xcoff_openr_next_archived_file): Sanity check nextoff.
85363 * coff64-rs6000.c (xcoff64_openr_next_archived_file): Call
85364 _bfd_xcoff_openr_next_archived_file.
85366 2021-10-07 Alan Modra <amodra@gmail.com>
85368 PR28422, build_id use-after-free
85369 This fixes a bug in commit 5d9bbb73c1df. All fields preserved from a
85370 bfd in struct bfd_preserve need to be cleared in bfd_reinit.
85373 * format.c (bfd_reinit): Clear build_id.
85375 2021-10-07 Alan Modra <amodra@gmail.com>
85377 Change ridiculous section size error
85378 Rather than reporting "memory exhausted", report "file truncated".
85379 You can hit this error on small fuzzed object files, or on files that
85380 are actually truncated. In either case sizes can be such that an out
85381 of memory error is a little confusing.
85383 * compress.c (bfd_get_full_section_contents): Set
85384 bfd_error_file_truncated rather than bfd_error_no_memory when
85385 section size exceeds file size.
85387 2021-10-07 Tom de Vries <tdevries@suse.de>
85389 [gdb/testsuite] Fix FAIL in gdb.base/annota1.exp
85390 On openSUSE tumbleweed I run into:
85392 FAIL: gdb.base/annota1.exp: run until main breakpoint (timeout)
85394 due to a message related to libthread_db:
85397 [Thread debugging using libthread_db enabled]^M
85398 Using host libthread_db library "/lib64/libthread_db.so.1".^M
85400 ^Z^Zframes-invalid^M
85402 which is not matched by the regexp.
85404 Fix this by updating the regexp.
85406 Tested on x86_64-linux.
85408 2021-10-07 Tom de Vries <tdevries@suse.de>
85410 [gdb/testsuite] Refactor regexp in gdb.base/annota1.exp
85411 Refactor regexp in gdb.base/annota1.exp to reduce indentation and repetition.
85413 Tested on x86_64-linux.
85415 2021-10-07 GDB Administrator <gdbadmin@sourceware.org>
85417 Automatic date update in version.in
85419 2021-10-06 Andrew Burgess <andrew.burgess@embecosm.com>
85421 gdb/doc: improve 'show print elements' description
85422 The documentation for 'show print elements' contains the line:
85424 If the number is 0, then the printing is unlimited.
85426 However, this line is now out of date as can be seen by this GDB
85429 (gdb) set print elements 0
85430 (gdb) show print elements
85431 Limit on string chars or array elements to print is unlimited.
85433 The value 0 does indeed mean unlimited, and this is described in the
85434 'set print elements' section, however, for 'show print elements' the
85435 user will never see the value 0, so lets just remove that bit from the
85438 2021-10-06 Tom de Vries <tdevries@suse.de>
85440 [gdb/testsuite] Fix FAIL in gdb.tui/corefile-run.exp
85441 When running test-case gdb.tui/corefile-run.exp on openSUSE Tumbleweed,
85444 PASS: gdb.tui/corefile-run.exp: load corefile
85445 FAIL: gdb.tui/corefile-run.exp: run until the end
85448 What's going on is easier to see when also doing dump_screen if
85449 check_contents passes, and inspecting state at the preceding PASS:
85451 +-------------------------------------------------------------------------+
85452 exec No process In: L?? PC: ??
85454 [Thread debugging using libthread_db enabled]
85455 Using host libthread_db library "/lib64/libthread_db.so.1".
85456 Core was generated by `/data/gdb_versions/devel/build/gdb/testsuite/output
85457 s/gdb.tui/corefile-run/corefi'.
85458 Program terminated with signal SIGTRAP, Trace/breakpoint trap.
85460 --Type <RET> for more, q to quit, c to continue without paging--
85463 The problem is that we're getting a pagination prompt, and the subsequent run
85464 command is interpreted as an answer to that prompt.
85467 - detecting the gdb prompt in response to "load corefile", such that
85468 we detect the failure earlier, and
85469 - doing a "set pagination off" in Term::clean_restart.
85471 Tested on x86_64-linux.
85473 2021-10-06 Alan Modra <amodra@gmail.com>
85475 PR28420, ecoff fuzzing failures
85477 * coff-mips.c (mips_adjust_reloc_in): Replace abort with error
85478 message and return.
85479 * ecoff.c (ecoff_slurp_reloc_table): Remove assertion and aborts,
85480 instead handle errors gracefully.
85482 2021-10-06 Alan Modra <amodra@gmail.com>
85484 PR28402, fail to allocate line number array
85485 This fixes a situation where the COFF code allocated memory for
85486 internal representaion arrays before reading the external file data.
85487 That meant the allocation didn't have any sanity check against file
85491 * coffcode.h (buy_and_read): Malloc rather than alloc memory.
85492 (coff_slurp_line_table): Read native line number info before
85493 allocating memory for internal line number array. Adjust error
85494 paths to suit. Remove now unnecessary line number count check.
85495 (coff_slurp_reloc_table): Adjust to suit buy_and_read change.
85497 2021-10-06 Alan Modra <amodra@gmail.com>
85499 PR28403, null pointer dereference in disassemble_bytes
85500 Indexing of symbol and howto arrays wasn't checked in aout targets.
85503 * aout-ns32k.c (MY (reloc_howto)): Sanity check howto_table index.
85504 Make r_index unsigned.
85505 (MY_swap_std_reloc_in): Make r_index unsigned.
85506 * aoutx.h (MOVE_ADDRESS): Sanity check symbol r_index.
85507 (aout_link_input_section_std): Make r_index unsigned.
85508 (aout_link_input_section_ext): Likewise.
85509 * i386lynx.c (MOVE_ADDRESS): Sanity check symbol r_index.
85510 (swap_ext_reloc_in, swap_std_reloc_in): Make r_index unsigned.
85511 * pdp11.c (MOVE_ADDRESS): Sanity check symbol r_index.
85513 2021-10-06 Alan Modra <amodra@gmail.com>
85515 PR28401, invalid section name lookup
85516 The PR28401 testcase has a section named "", ie. an empty string.
85517 This results in some silly behaviour in load_debug_section, and
85518 dump_dwarf_section. Fix that. Note that this patch doesn't correct
85519 the main complaint in PR28401, "failed to allocate", since malloc
85520 failures on sections having huge bogus sizes are to be expected. We
85521 can't safely catch all such cases by comparing with file size, for
85522 example, where sections contain compressed data.
85525 * objdump.c (load_debug_section): Don't attempt to retrieve
85526 empty name sections.
85527 (dump_dwarf_section): Likewise.
85529 2021-10-06 GDB Administrator <gdbadmin@sourceware.org>
85531 Automatic date update in version.in
85533 2021-10-06 Tom de Vries <tdevries@suse.de>
85535 [gdb/testsuite] Make tui testing less verbose
85536 Currently, tui testing is rather verbose. When using these RUNTESTFLAGS to
85537 pick up all tui tests (17 in total):
85539 rtf=$(echo $(cd src/gdb/testsuite/; find gdb.* -type f -name *.exp* \
85540 | xargs grep -l tuiterm_env) )
85548 Most of the output is related to controlling the tui screen, but that does
85549 not give a top-level sense of how the test-case progresses.
85551 Put differently: a lot of bandwith is used to describe how we arrive at a
85552 certain tui screen state. But we don't actually always show the state we
85553 arrive at, unless there's a FAIL.
85555 And if there's say, a PASS that should actually be FAILing, it's hard to
85559 - dropping the -log on the call to verbose in _log. We still can get the
85560 same info back using runtest -v.
85561 - dumping the screen or box that we're checking, also when the test passes.
85563 Brings down verbosity to something more reasonable:
85569 Tested on x86_64-linux.
85571 2021-10-06 Tom de Vries <tdevries@suse.de>
85573 [gdb/testsuite] Add Term::dump_box in lib/tuiterm.exp
85574 Factor out new proc Term::get_region and use it to implement a
85575 new proc Term::dump_box, similar to Term::dump_screen.
85577 Tested on x86_64-linux.
85579 2021-10-05 Lancelot SIX <lsix@lancelotsix.com>
85581 gdb: Remove deprecated assertion in setting::get
85582 The commit 702991711a91bd47b209289562843a11e7009396 (gdb: Have setter
85583 and getter callbacks for settings) makes it possible for a setting not
85584 to be backed by a memory buffer but use callback functions instead to
85585 retrieve or set the setting's value.
85587 An assertion was not properly updated to take into account that the
85588 m_var member (which points to a memory buffer, if used) might be nullptr
85589 if the setting uses callback functions. If the setting is backed by a
85590 memory buffer, the m_var has to be non nullptr, which is already checked
85591 before the pointer is dereferenced.
85593 This commit removes this assertion as it is not valid anymore.
85595 2021-10-05 Tom Tromey <tromey@adacore.com>
85597 Remove 'varsize-limit'
85598 This makes the Ada-specific "varsize-limit" a synonym for
85599 "max-value-size", and removes the Ada-specific checks of the limit.
85601 I am not certain of the history here, but it seems to me that this
85602 code is fully obsolete now. And, removing this makes it possible to
85603 index large Ada arrays without triggering an error. A new test case
85604 is included to demonstrate this.
85606 2021-10-05 Tom Tromey <tromey@adacore.com>
85608 Allow lazy 'zero' value
85609 This changes value_zero to create a lazy value. In many cases,
85610 value_zero is called in expression evaluation to wrap a type in a
85611 non-eval context. It seems senseless to allocate a buffer in these
85614 A new 'is_zero' flag is added so we can preserve the existing
85615 assertions in value_fetch_lazy.
85617 A subsequent patch will add a test where creating a zero value would
85618 fail, due to the variable size check. However, the contents of this
85619 value are never needed, and so creating a lazy value avoids the error
85622 2021-10-05 Tom Tromey <tromey@adacore.com>
85624 Add lval_funcs::is_optimized_out
85625 This adds an is_optimized_out function pointer to lval_funcs, and
85626 changes value_optimized_out to call it. This new function lets gdb
85627 determine if a value is optimized out without necessarily fetching the
85628 value. This is needed for a subsequent patch, where an attempt to
85629 access a lazy value would fail due to the value size limit -- however,
85630 the access was only needed to determine the optimized-out state.
85632 2021-10-05 Tom de Vries <tdevries@suse.de>
85634 [gdb/testsuite] Fix FAIL in gdb.mi/mi-nsmoribund.exp
85635 Since commit e36788d1354 "[gdb/testsuite] Fix handling of nr_args < 3 in
85636 mi_gdb_test" we run into:
85638 PASS: gdb.mi/mi-nsmoribund.exp: print done = 1
85642 \*running,thread-id="[0-9]+"^M
85643 \*running,thread-id="[0-9]+"^M
85644 \*running,thread-id="[0-9]+"^M
85645 \*running,thread-id="[0-9]+"^M
85646 \*running,thread-id="[0-9]+"^M
85647 \*running,thread-id="[0-9]+"^M
85648 \*running,thread-id="[0-9]+"^M
85649 \*running,thread-id="[0-9]+"^M
85650 \*running,thread-id="[0-9]+"^M
85651 \*running,thread-id="[0-9]+"[^M
85654 103-exec-continue --all^M
85655 =library-loaded,id="/lib64/libgcc_s.so.1",target-name="/lib64/libgcc_s.so.1",\
85656 host-name="/lib64/libgcc_s.so.1",symbols-loaded="0",thread-group="i1",\
85657 ranges=[{from="0x00007ffff22a5010",to="0x00007ffff22b6365"}]^M
85659 *running,thread-id="5"^M
85661 FAIL: gdb.mi/mi-nsmoribund.exp: 103-exec-continue --all (unexpected output)
85664 The regexp expect running messages for all threads, but we only get one for
85667 The test-case uses non-stop mode, and when the exec-continue --all command is
85668 issued, thread 5 is stopped and all other threads are running. Consequently,
85669 only thread 5 is resumed, and reported as running.
85671 Fix this by updating the regexp.
85673 Tested on x86_64-linux.
85675 2021-10-05 Andrew Burgess <andrew.burgess@embecosm.com>
85677 gdb/python: fix memory leak in python inferior code
85678 When a user creates a gdb.Inferior object for the first time a new
85679 Python object is created. This object is then cached within GDB's
85680 inferior object using the registry mechanism (see
85681 inferior_to_inferior_object in py-inferior.c, specifically the calls
85682 to inferior_data and set_inferior_data).
85684 The Python Reference to the gdb.Inferior object held within the real
85685 inferior object ensures that the reference count on the Python
85686 gdb.Inferior object never reaches zero while the GDB inferior object
85687 continues to exist.
85689 At the same time, the gdb.Inferior object maintains a C++ pointer back
85690 to GDB's real inferior object. We therefore end up with a system that
85696 .----------. | .--------------.
85697 | |------------------->| |
85698 | inferior | | gdb.Inferior |
85699 | |<-------------------| |
85700 '----------' | '--------------'
85705 When GDB's inferior object is deleted (say the inferior exits) then
85706 py_free_inferior is called (thanks to the registry system), this
85707 function looks up the Python gdb.Inferior object and sets the C++
85708 pointer to nullptr and finally reduces the reference count on the
85709 Python gdb.Inferior object.
85711 If at this point the user still holds a reference to the Python
85712 gdb.Inferior object then nothing happens. However, the gdb.Inferior
85713 object is now in the non-valid state (see infpy_is_valid in
85714 py-inferior.c), but otherwise, everything is fine.
85716 However, if there are no further references to the Python gdb.Inferior
85717 object, or, once the user has given up all their references to the
85718 gdb.Inferior object, then infpy_dealloc is called.
85720 This function currently checks to see if the inferior pointer within
85721 the gdb.Inferior object is nullptr or not. If the pointer is nullptr
85722 then infpy_dealloc immediately returns.
85724 Only when the inferior point in the gdb.Inferior is not nullptr do
85725 we (a) set the gdb.Inferior reference inside GDB's inferior to
85726 nullptr, and (b) call the underlying Python tp_free function.
85728 There are a number things wrong here:
85730 1. The Python gdb.Inferior reference within GDB's inferior object
85731 holds a reference count, thus, setting this reference to nullptr
85732 without first decrementing the reference count would leak a
85733 reference, however...
85735 2. As GDB's inferior holds a reference then infpy_dealloc will never
85736 be called until GDB's inferior object is deleted. Deleting a GDB
85737 inferior ohject calls py_free_inferior, and so gives up the
85738 reference. At this point there is no longer a need to call
85739 set_inferior_data to set the field back to NULL, that field must
85740 have been cleared in order to get the reference count to zero, which
85743 3. If we know that py_free_inferior must be called before
85744 infpy_dealloc, then we know that the inferior pointer in
85745 gdb.Inferior will always be nullptr when infpy_dealloc is called,
85746 this means that the call to the underlying tp_free function will
85747 always be skipped. Skipping this call will cause Python to leak the
85748 memory associated with the gdb.Inferior object, which is what we
85749 currently always do.
85751 Given all of the above, I assert that the C++ pointer within
85752 gdb.Inferior will always be nullptr when infpy_dealloc is called.
85753 That's what this patch does.
85755 I wrote a test for this issue making use of Pythons tracemalloc
85756 module, which allows us to spot this memory leak.
85758 2021-10-05 Bhuvanendra Kumar N <Bhuvanendra.KumarN@amd.com>
85760 [gdb/testsuite] Use function_range in gdb.dwarf2/dw2-ref-missing-frame.exp
85761 Following 2 test points are failing with clang compiler
85763 (gdb) FAIL: gdb.dwarf2/dw2-ref-missing-frame.exp: func_nofb print
85764 (gdb) FAIL: gdb.dwarf2/dw2-ref-missing-frame.exp: func_loopfb print
85766 As in commit f677852bbda "[gdb/testsuite] Use function_range in
85767 gdb.dwarf2/dw2-abs-hi-pc.exp", the problem is that the CU and functions
85768 have an empty address range, due to using asm labels in global scope,
85769 which is a known source of problems, as explained in the comment of proc
85770 function_range in gdb/testsuite/lib/dwarf.exp. Hence fix this also by
85771 using function_range.
85773 Tested on x86_64-linux with gcc and clang.
85775 2021-10-05 Andrew Burgess <andrew.burgess@embecosm.com>
85777 gdb/python: add a new gdb_exiting event
85778 Add a new event, gdb.events.gdb_exiting, which is called once GDB
85779 decides it is going to exit.
85781 This event is not triggered in the case that GDB performs a hard
85782 abort, for example, when handling an internal error and the user
85783 decides to quit the debug session, or if GDB hits an unexpected,
85786 This event is triggered if the user just types 'quit' at the command
85787 prompt, or if GDB is run with '-batch' and has processed all of the
85790 The new event type is gdb.GdbExitingEvent, and it has a single
85791 attribute exit_code, which is the value that GDB is about to exit
85794 The event is triggered before GDB starts dismantling any of its own
85795 internal state, so, my expectation is that most Python calls should
85796 work just fine at this point.
85798 When considering this functionality I wondered about using the
85799 'atexit' Python module. However, this is triggered when the Python
85800 environment is shut down, which is done from a final cleanup. At
85801 this point we don't know for sure what other GDB state has already
85804 2021-10-05 Andrew Burgess <andrew.burgess@embecosm.com>
85806 gdb/python: update events test to handle missing exit_code
85807 The test gdb.python/py-events.exp sets up a handler for the gdb.exited
85808 event. Unfortunately the handler is slightly broken, it assumes that
85809 the exit_code attribute will always be present. This is not always
85812 In a later commit I am going to add more tests to py-events.exp test
85813 script, and in so doing I expose the bug in our handling of gdb.exited
85816 Just to be clear, GDB itself is fine, it is the test that is not
85817 written correctly according to the Python Events API.
85819 So, in this commit I fix the Python code in the test, and extend the
85820 test case to exercise more paths through the Python code.
85822 Additionally, I noticed that the gdb.exited event is used as an
85823 example in the documentation for how to write an event handler.
85824 Unfortunately the same bug that we had in our test was also present in
85825 the example code in the manual.
85827 So I've fixed that too.
85829 After this commit there is no functional change to GDB.
85831 2021-10-05 GDB Administrator <gdbadmin@sourceware.org>
85833 Automatic date update in version.in
85835 2021-10-04 Tom Tromey <tromey@adacore.com>
85837 Use unique_xmalloc_ptr<char> when demangling
85838 I noticed that some methods in language_defn could use
85839 unique_xmalloc_ptr<char> rather than a plain 'char *'. This patch
85840 implements this change, fixing up the fallout and changing
85841 gdb_demangle to also return this type. In one spot, std::string is
85842 used to simplify some related code, and in another, an auto_obstack is
85843 used to avoid manual management.
85845 Regression tested on x86-64 Fedora 34.
85847 2021-10-04 Tom Tromey <tromey@adacore.com>
85849 Minor boolean fix in windows-nat.c
85850 I noticed a spot in windows-nat.c that used '1' rather than the more
85851 appropriate 'true'. This patch fixes it.
85853 2021-10-04 Tom de Vries <tdevries@suse.de>
85855 [gdb/build] Add CXX_DIALECT to CXX
85856 Say we use a gcc version that (while supporting c++11) does not support c++11
85857 by default, and needs an -std setting to enable it.
85859 If gdb would use the default AX_CXX_COMPILE_STDCXX from autoconf-archive, then
85862 CXX="g++ -std=gnu++11"
85865 That mechanism however has the following problem (quoting from commit
85868 the top level Makefile passes CXX down to subdirs, and that overrides whatever
85869 gdb/Makefile may set CXX to. The result would be that a make invocation from
85870 the build/gdb/ directory would use "g++ -std=gnu++11" as expected, while a
85871 make invocation at the top level would not.
85874 Commit 0bcda685399 fixes this by using a custom AX_CXX_COMPILE_STDCXX which
85878 CXX_DIALECT=-std=gnu++11
85881 The problem reported in PR28318 is that using the custom instead of the
85882 default AX_CXX_COMPILE_STDCXX makes the configure test for std::thread
85885 We could simply add $CXX_DIALECT to the test for std::thread support, but
85886 that would have to be repeated for each added c++ support test.
85888 Instead, fix this by doing:
85890 CXX="g++ -std=gnu++11"
85891 CXX_DIALECT=-std=gnu++11
85894 This is somewhat awkward, since it results in -std=gnu++11 occuring twice in
85897 $ touch src/gdb/dwarf2/read.c
85898 $ ( cd build/gdb; make V=1 dwarf2/read.o )
85899 g++-4.8 -std=gnu++11 -x c++ -std=gnu++11 ...
85902 However, both settings are needed:
85903 - the switch in CXX for the std::thread tests (and other tests)
85904 - the switch in CXX_DIALECT so it can be appended in Makefiles, to
85905 counteract the fact that the top-level Makefile overrides CXX
85907 The code added in gdb/ax_cxx_compile_stdcxx.m4 is copied from the default
85908 AX_CXX_COMPILE_STDCXX from autoconf-archive.
85910 Tested on x86_64-linux.
85912 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28318
85914 2021-10-04 Simon Marchi <simon.marchi@polymtl.ca>
85916 [gdb/symtab] Use unrelocated addresses in call_site
85917 Consider test-case gdb.trace/entry-values.exp with target board
85920 Using this command we have an abbreviated version, and can see the correct
85921 @entry values for foo:
85923 $ gdb -q -batch outputs/gdb.trace/entry-values/entry-values \
85926 -ex "set print entry-values both" \
85928 Temporary breakpoint 1 at 0x679
85930 Temporary breakpoint 1, 0x0000555555554679 in main ()
85931 Breakpoint 2 at 0x55555555463e
85933 Breakpoint 2, 0x000055555555463e in foo (i=0, i@entry=2, j=2, j@entry=3)
85936 Now, let's try the same again, but run directly to foo rather than stopping at
85939 $ gdb -q -batch outputs/gdb.trace/entry-values/entry-values \
85941 -ex "set print entry-values both" \
85943 Breakpoint 1 at 0x63e
85945 Breakpoint 1, 0x000055555555463e in foo (i=0, i@entry=<optimized out>, \
85946 j=2, j@entry=<optimized out>)
85949 So, what explains the difference? Noteworthy, this is a dwarf assembly
85950 test-case, with debug info for foo and bar, but not for main.
85954 - this does not trigger expanding debug info, because there's none for main
85955 - we set a breakpoint at foo
85956 - this triggers expanding debug info. Relocated addresses are used in
85957 call_site info (because the exec is started)
85958 - we continue to foo, and manage to find the call_site info
85960 In the second case:
85961 - we set a breakpoint at foo
85962 - this triggers expanding debug info. Unrelocated addresses are used in
85963 call_site info (because the exec is not started)
85965 - this triggers objfile_relocate1, but it doesn't update the call_site
85967 - we don't manage to find the call_site info
85969 We could fix this by adding the missing call_site relocation in
85972 This solution however is counter-trend in the sense that we're trying to
85973 work towards the situation where when starting two instances of an executable,
85974 we need only one instance of debug information, implying the use of
85975 unrelocated addresses.
85977 So, fix this instead by using unrelocated addresses in call_site info.
85979 Tested on x86_64-linux.
85981 This fixes all remaining unix/-fno-PIE/-no-pie vs unix/-fPIE/-pie
85982 regressions, like f.i. PR24892.
85984 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=24892
85986 Co-Authored-By: Tom de Vries <tdevries@suse.de>
85988 2021-10-04 Simon Marchi <simon.marchi@polymtl.ca>
85990 [gdb/symtab] C++-ify call_site
85992 - add member function call_site::pc ()
85994 Tested on x86_64-linux.
85996 Co-Authored-By: Tom de Vries <tdevries@suse.de>
85998 2021-10-04 Tom de Vries <tdevries@suse.de>
86000 [gdb/symtab] Add call_site_eq and call_site_hash
86001 In commit b4c919f7525 "[gdb/symtab] Fix htab_find_slot call in
86002 read_call_site_scope" , I removed the comment:
86004 It must be the first field as we overload core_addr_hash and core_addr_eq for
86007 for field pc of struct call_site.
86009 However, this was not tested, and when indeed moving field pc to the second
86010 location, we run into a testsuite failure in gdb.trace/entry-values.exp.
86012 This is caused by core_addr_eq (the eq_f function for the htab) being
86013 called with a pointer to the pc field (as passed into htab_find_slot) and a
86014 pointer to a hash table element. Now that pc is no longer the first field,
86015 the pointer to hash table element no longer points to the pc field.
86017 This could be fixed by simply reinstating the comment, but we're trying to
86018 get rid of this kind of tricks that make refactoring more difficult.
86020 Instead, fix this by:
86021 - reverting commit b4c919f7525, apart from the comment removal, such that
86022 we're passing a pointer to element to htab_find_slot
86023 - updating the htab_find_slot call in compunit_symtab::find_call_site
86024 in a similar manner
86025 - adding a call_site_eq and call_site_hash, and using these in the hash table
86026 instead of core_addr_eq and core_addr_hash.
86028 Tested on x86_64-linux, both with and without a trigger patch that moves pc to
86029 the second location in struct call_site.
86031 2021-10-04 Tom Tromey <tromey@adacore.com>
86033 Fix remote-sim.c compilation
86034 The change "make string-like set show commands use std::string
86035 variable" caused remote-sim.c to fail to build. The issue is that the
86038 const std::string &sysroot = gdb_sysroot;
86039 if (is_target_filename (sysroot))
86040 sysroot += strlen (TARGET_SYSROOT_PREFIX);
86042 ... which isn't valid.
86044 This patch changes this code to use a 'const char *' again, fixing the
86047 2021-10-04 Bruno Larsen <blarsen@redhat.com>
86049 [gdb/testsuite] update analyze-racy-logs.py to python3
86050 Since python 2 is no longer supported on most distributions, update the
86051 script to run under python while while still being runnable under
86054 2021-10-04 Andrew Burgess <andrew.burgess@embecosm.com>
86056 gdbsupport: remove attempt to define TARGET_WORD_SIZE
86057 In the gdbsupport configure.ac file, there is an attempt to define
86058 TARGET_WORD_SIZE. This is done by running grep on the file
86061 The problem with this is, the file bfd-in3.h is generated into the bfd
86062 build directory when bfd is configured, and there is no dependency
86063 between the gdbsupport module and the bfd module, so, for example, if
86067 $ make all-gdbsupport
86069 Then bfd will neither be configured, or built. In this case
86070 TARGET_WORD_SIZE ends up being defined, but with no value because the
86071 grep on bfd-in3.h fails.
86073 However, it turns out that this doesn't matter; we don't actually use
86074 TARGET_WORD_SIZE anywhere.
86076 My proposal in this commit is to just remove the definition of
86077 TARGET_WORD_SIZE, the alternative would be to add a dependency between
86078 configure-gdbsupport and configure-bfd into Makefile.def, but adding a
86079 dependency for something we don't need seems pretty pointless.
86081 2021-10-04 Mike Frysinger <vapier@gentoo.org>
86083 sim: add --info-target for listing supported BFD targets
86084 It can be difficult to guess the exact bfd name, so add an option to
86085 list all the targets that the current build supports. This aligns with
86086 other simulator options like --info-architecture.
86088 2021-10-04 GDB Administrator <gdbadmin@sourceware.org>
86090 Automatic date update in version.in
86092 2021-10-03 Lancelot SIX <lsix@lancelotsix.com>
86094 gdb: Setting setter return a bool to tell if the value changed
86095 GDB can notify observers when a parameter is changed.
86097 To do that, do_set_command (in gdb/cli/cli-setshow.c) compares the new
86098 value against the old one before updating it, and based on that notifies
86099 observers. This looks like something like:
86101 int valuechanged = 0;
86102 switch (cmd->var.type ())
86106 LONGEST new_val = parse_and_eval_long (arg)
86107 if (new_val != cmd->var.get<int> ())
86109 cmd->var.get<int> (new_val);
86114 case var_zuinteger:
86117 = parse_cli_var_uinteger (c->var->type (), &arg, true);
86118 if (c->var->get<unsigned int> () != val)
86120 c->var->set<unsigned int> (val);
86121 option_changed = true;
86125 /* And so on for all possible var_types. */
86128 This comparison is done for each possible var_type, which leads to
86129 unnecessary logic duplication.
86131 In this patch I propose to move all those checks in one place within the
86132 setting setter method. This limits the code duplication and simplifies
86133 the do_set_command implementation.
86135 This patch also changes slightly the way a value change is detected.
86136 Instead of comparing the user provided value against the current value
86137 of the setting, we compare the value of the setting before and after the
86138 set operation. This is meant to handle edge cases where trying to set
86139 an unrecognized value would be equivalent to a noop (the actual value
86140 remains unchanged). Doing this requires that the original value needs
86141 to be copied before the update, which can be non trivial for
86144 There should be no user visible change introduced by this commit.
86146 Tested on x86_64 GNU/Linux.
86148 [1] https://review.lttng.org/c/binutils-gdb/+/5831/41
86150 Change-Id: If064b9cede3eb56275aacd2b286f74eceb1aed11
86152 2021-10-03 Lancelot SIX <lsix@lancelotsix.com>
86153 Simon Marchi <simon.marchi@polymtl.ca>
86155 gdb: Have setter and getter callbacks for settings
86156 The main motivation behind this improvement is to help the
86157 implementation of a patch Simon Marchi is preparing to fix a bug when
86158 MI or Python try to access parameters that are inferior dependent (see
86161 This commit extends the previous ones, which introduces the setting
86162 object to represent a static variable whose value can be set or shown
86163 with the appropriate commands. This patch proposes that a setting can
86164 either contain a pointer to a static variable holding a setting, or
86165 pointers to a pair of setter and getter callback functions.
86167 The callbacks functions can be used to retrieve or change the value with
86168 custom logic. This is useful when the source of truth for a given
86169 setting is not contained in the variable pointed to by the setting
86172 Given that the callback function call is hidden within the setting
86173 abstraction introduced earlier, none of the sites accessing the setting
86174 needs to be updated. The registered getter or setter is used whatever
86175 the way to access it is (through MI, Python, Guile, the "with" command
86176 and the $_gdb_setting / $_gdb_setting_str convenience functions).
86178 All the add_setshow_*_cmd are given a new overload that will accept the
86179 pair of function pointers (set / get functions) instead of the pointer
86180 to a global variable.
86182 Tested on GNU/Linux x86_64 with no regression observed.
86184 Change-Id: Ieb81fef57550632ff66e6aa85f637372a226be8c
86186 2021-10-03 Simon Marchi <simon.marchi@efficios.com>
86187 Lancelot SIX <lsix@lancelotsix.com>
86189 gdb: make string-like set show commands use std::string variable
86190 String-like settings (var_string, var_filename, var_optional_filename,
86191 var_string_noescape) currently take a pointer to a `char *` storage
86192 variable (typically global) that holds the setting's value. I'd like to
86193 "mordernize" this by changing them to use an std::string for storage.
86195 An obvious reason is that string operations on std::string are often
86196 easier to write than with C strings. And they avoid having to do any
86197 manual memory management.
86199 Another interesting reason is that, with `char *`, nullptr and an empty
86200 string often both have the same meaning of "no value". String settings
86201 are initially nullptr (unless initialized otherwise). But when doing
86202 "set foo" (where `foo` is a string setting), the setting now points to
86203 an empty string. For example, solib_search_path is nullptr at startup,
86204 but points to an empty string after doing "set solib-search-path". This
86205 leads to some code that needs to check for both to check for "no value".
86206 Or some code that converts back and forth between NULL and "" when
86207 getting or setting the value. I find this very error-prone, because it
86208 is very easy to forget one or the other. With std::string, we at least
86209 know that the variable is not "NULL". There is only one way of
86210 representing an empty string setting, that is with an empty string.
86212 I was wondering whether the distinction between NULL and "" would be
86213 important for some setting, but it doesn't seem so. If that ever
86214 happens, it would be more C++-y and self-descriptive to use
86215 optional<string> anyway.
86217 Actually, there's one spot where this distinction mattered, it's in
86218 init_history, for the test gdb.base/gdbinit-history.exp. init_history
86219 sets the history filename to the default ".gdb_history" if it sees that
86220 the setting was never set - if history_filename is nullptr. If
86221 history_filename is an empty string, it means the setting was explicitly
86222 cleared, so it leaves it as-is. With the change to std::string, this
86223 distinction doesn't exist anymore. This can be fixed by moving the code
86224 that chooses a good default value for history_filename to
86225 _initialize_top. This is ran before -ex commands are processed, so an
86226 -ex command can then clear that value if needed (what
86227 gdb.base/gdbinit-history.exp tests).
86229 Another small improvement, in my opinion is that we can now easily
86230 give string parameters initial values, by simply initializing the global
86231 variables, instead of xstrdup-ing it in the _initialize function.
86233 In Python and Guile, when registering a string-like parameter, we
86234 allocate (with new) an std::string that is owned by the param_smob (in
86235 Guile) and the parmpy_object (in Python) objects.
86237 This patch started by changing all relevant add_setshow_* commands to
86238 take an `std::string *` instead of a `char **` and fixing everything
86239 that failed to build. That includes of course all string setting
86240 variable and their uses.
86242 string_option_def now uses an std::string also, because there's a
86243 connection between options and settings (see
86244 add_setshow_cmds_for_options).
86246 The add_path function in source.c is really complex and twisted, I'd
86247 rather not try to change it to work on an std::string right now.
86248 Instead, I added an overload that copies the std:string to a `char *`
86249 and back. This means more copying, but this is not used in a hot path
86250 at all, so I think it is acceptable.
86252 Change-Id: I92c50a1bdd8307141cdbacb388248e4e4fc08c93
86254 2021-10-03 Lancelot SIX <lsix@lancelotsix.com>
86255 Simon Marchi <simon.marchi@polymtl.ca>
86257 gdb: Introduce setting construct within cmd_list_element
86258 cmd_list_element can contain a pointer to data that can be set and / or
86259 shown. This is achieved with the void* VAR member which points to the
86260 data that can be accessed, while the VAR_TYPE member (of type enum
86261 var_types) indicates how to interpret the data pointed to.
86263 With this pattern, the user of the cmd_list_element needs to know what
86264 is the storage type associated with a given VAR_TYPES in order to do
86265 the proper casting. No automatic safeguard is available to prevent
86266 miss-use of the pointer. Client code typically looks something like:
86268 switch (c->var_type)
86270 case var_zuinteger:
86271 unsigned int v = *(unsigned int*) c->var;
86275 bool v = *(bool *) c->var;
86281 This patch proposes to add an abstraction around the var_types and void*
86282 pointer pair. The abstraction is meant to prevent the user from having
86283 to handle the cast and verify that the data is read or written as a type
86284 that is coherent with the setting's var_type. This is achieved by
86285 introducing the struct setting which exposes a set of templated get /
86286 set member functions. The template parameter is the type of the
86287 variable that holds the referred variable.
86289 Using those accessors allows runtime checks to be inserted in order to
86290 ensure that the data pointed to has the expected type. For example,
86291 instantiating the member functions with bool will yield something
86294 const bool &get<bool> () const
86296 gdb_assert (m_var_type == var_boolean);
86297 gdb_assert (m_var != nullptr);
86298 return *static_cast<bool *> (m_var);
86300 void set<bool> (const bool &var)
86302 gdb_assert (m_var_type == var_boolean);
86303 gdb_assert (m_var != nullptr);
86304 *static_cast<bool *> (m_var) = var;
86307 Using the new abstraction, our initial example becomes:
86309 switch (c->var_type)
86311 case var_zuinteger:
86312 unsigned int v = c->var->get<unsigned int> ();
86316 bool v = c->var->get<bool> ();
86322 While the call site is still similar, the introduction of runtime checks
86323 help ensure correct usage of the data.
86325 In order to avoid turning the bulk of add_setshow_cmd_full into a
86326 templated function, and following a suggestion from Pedro Alves, a
86327 setting can be constructed from a pre validated type erased reference to
86328 a variable. This is what setting::erased_args is used for.
86330 Introducing an opaque abstraction to describe a setting will also make
86331 it possible to use callbacks to retrieve or set the value of the setting
86332 on the fly instead of pointing to a static chunk of memory. This will
86333 be done added in a later commit.
86335 Given that a cmd_list_element may or may not reference a setting, the
86336 VAR and VAR_TYPES members of the struct are replaced with a
86337 gdb::optional<setting> named VAR.
86339 Few internal function signatures have been modified to take into account
86340 this new abstraction:
86342 -The functions value_from_setting, str_value_from_setting and
86343 get_setshow_command_value_string used to have a 'cmd_list_element *'
86344 parameter but only used it for the VAR and VAR_TYPE member. They now
86345 take a 'const setting &' parameter instead.
86346 - Similarly, the 'void *' and a 'enum var_types' parameters of
86347 pascm_param_value and gdbpy_parameter_value have been replaced with a
86348 'const setting &' parameter.
86350 No user visible change is expected after this patch.
86352 Tested on GNU/Linux x86_64, with no regression noticed.
86354 Change-Id: Ie1d08c3ceb8b30b3d7bf1efe036eb8acffcd2f34
86356 2021-10-03 Mike Frysinger <vapier@gentoo.org>
86358 sim: filter out SIGSTKSZ [PR sim/28302]
86359 We map target signals to host signals so we can propagate signals
86360 between the host & simulated worlds. That means we need to know
86361 the symbolic names & values of all signals that might be sent.
86363 The tools that generate that list use signal.h and include all
86364 symbols that start with "SIG" so as to automatically include any
86365 new symbols that the C library might add. Unfortunately, this
86366 also picks up "SIGSTKSZ" which is not actually a signal itself,
86367 but a signal related setting -- it's the size of the stack when
86368 a signal is handled.
86370 By itself this doesn't super matter as we will never see a signal
86371 with that same value (since the range of valid signals tend to be
86372 way less than 1024, and the size of the default signal stack will
86373 never be that small). But with recent glibc changes that make this
86374 into a dynamic value instead of a compile-time constant, some users
86375 see build failures when building the sim.
86377 As suggested by Adam Sampson, update our scripts to ignore this
86378 symbol to simplify everything and avoid the build failure.
86380 Bug: https://sourceware.org/PR28302
86382 2021-10-03 Mike Frysinger <vapier@gentoo.org>
86384 sim: ppc: fallback when ln is not available [PR sim/18864]
86385 Not all systems have easy access to hard links or symlinks, so add
86386 fallback logic to the run->psim build code to handle those.
86388 Bug: https://sourceware.org/PR18864
86390 2021-10-03 Lancelot SIX <lsix@lancelotsix.com>
86392 gdb: Fix comment in riscv_scan_prologue
86393 I found an inaccurate comment in riscv_scan_prologue. This commit fixes
86396 2021-10-03 Lancelot SIX <lsix@lancelotsix.com>
86398 gdb: Support the c.mv insn in the riscv prologue scanner.
86399 While working on other problems, I encountered situations where GDB
86400 fails to properly unwind the stack because some functions use the C.MV
86401 instruction in the prologue. The prologue scanner stops when it hits
86402 this instruction assuming its job is done at this point. Unfortunately
86403 the prologue is not necessarily finished yet, preventing GDB to properly
86406 This commit adds support for handling such instruction in
86407 riscv_scan_prologue.
86409 Note that C.MV is part of the compressed instruction set. The MV
86410 counterpart from the base ISA is a pseudo instruction that expands to
86411 'ADDI RD,RS1,0' which is already supported.
86413 Tested on riscv64-linux-gnu.
86415 All feedback are welcome.
86417 2021-10-03 GDB Administrator <gdbadmin@sourceware.org>
86419 Automatic date update in version.in
86421 2021-10-02 Simon Marchi <simon.marchi@polymtl.ca>
86423 [gdb/symtab] Remove COMPUNIT_CALL_SITE_HTAB
86424 Remove macro COMPUNIT_CALL_SITE_HTAB, and provide access to the htab using
86426 - compunit_symtab::find_call_site
86427 - compunit_symtab::set_call_site_htab
86429 Tested on x86_64-linux.
86431 Co-Authored-By: Tom de Vries <tdevries@suse.de>
86433 2021-10-02 Simon Marchi <simon.marchi@polymtl.ca>
86435 gdb/python: fix a few flake8 warnings
86436 Fix these rather obvious warnings reported by flake8:
86438 ./lib/gdb/FrameIterator.py:16:1: F401 'gdb' imported but unused
86439 ./lib/gdb/FrameIterator.py:17:1: F401 'itertools' imported but unused
86440 ./lib/gdb/command/prompt.py:55:26: E712 comparison to False should be 'if cond is False:' or 'if not cond:'
86441 ./lib/gdb/command/explore.py:526:9: F841 local variable 'has_explorable_fields' is assigned to but never used
86442 ./lib/gdb/command/explore.py:697:56: E712 comparison to False should be 'if cond is False:' or 'if not cond:'
86443 ./lib/gdb/command/explore.py:736:62: E712 comparison to False should be 'if cond is False:' or 'if not cond:'
86444 ./lib/gdb/command/explore.py:767:61: E712 comparison to False should be 'if cond is False:' or 'if not cond:'
86445 ./lib/gdb/command/frame_filters.py:21:1: F401 'copy' imported but unused
86446 ./lib/gdb/command/frame_filters.py:22:1: F401 'gdb.FrameIterator.FrameIterator' imported but unused
86447 ./lib/gdb/command/frame_filters.py:23:1: F401 'gdb.FrameDecorator.FrameDecorator' imported but unused
86448 ./lib/gdb/command/frame_filters.py:25:1: F401 'itertools' imported but unused
86449 ./lib/gdb/command/frame_filters.py:179:17: E712 comparison to True should be 'if cond is True:' or 'if cond:'
86451 Change-Id: I4f49c0cb430359ee872222600c61d9c5283b09ab
86453 2021-10-02 GDB Administrator <gdbadmin@sourceware.org>
86455 Automatic date update in version.in
86457 2021-10-01 Luis Machado <luis.machado@linaro.org>
86459 Fix build failure for 32-bit targets
86460 When building master GDB, I ran into the following:
86462 binutils-gdb/gdb/bt-utils.c: In function 'int libbacktrace_print(void*, uintptr_t, const char*, int, const char*)':
86463 binutils-gdb/gdb/bt-utils.c:93:44: error: format '%lx' expects argument of type 'long unsigned int', but argument 4 has type 'uintptr_t {aka unsigned int}' [-Werror=format=]
86464 snprintf (buf, sizeof (buf), "0x%lx ", pc);
86466 Fix this by using %PRIxPTR as opposed to %lx.
86468 2021-10-01 Nick Clifton <nickc@redhat.com>
86470 Fix mistake in RX assembler documentation (special section names)
86472 2021-10-01 Simon Marchi <simon.marchi@polymtl.ca>
86474 [gdb/symtab] Fix htab_find_slot call in read_call_site_scope
86475 In read_call_site_scope we have:
86477 call_site_local.pc = pc;
86478 slot = htab_find_slot (cu->call_site_htab, &call_site_local, INSERT);
86481 The call passes a call_site pointer as element. OTOH, the hashtab is created
86482 using hash_f == core_addr_hash and eq_f == core_addr_eq, so the element
86483 will be accessed through a CORE_ADDR pointer.
86485 This is not wrong (at least in C), given that pc is the first field in
86488 Nevertheless, as in call_site_for_pc, make the htab_find_slot call match the
86489 used hash_f and eq_f by using &pc instead:
86491 slot = htab_find_slot (cu->call_site_htab, &pc, INSERT);
86494 Tested on x86_64-linux.
86496 Co-Authored-By: Tom de Vries <tdevries@suse.de>
86498 2021-10-01 Andrea Corallo <andrea.corallo@arm.com>
86500 PATCH bfd: Fix linker warning for recently introduced arm attributes
86501 2021-09-27 Andrea Corallo <andrea.corallo@arm.com>
86503 * elf-bfd.h (NUM_KNOWN_OBJ_ATTRIBUTES): Update value to cover
86504 'Tag_BTI_use' and 'Tag_PACRET_use'.
86506 2021-10-01 Simon Marchi <simon.marchi@polymtl.ca>
86508 gdb/testsuite/dwarf: use options for rnglists/loclists procs
86509 Change how rnglists and loclists procs to align them with how procs for
86510 aranges (and other things in the DWARF assembler) work. Instead of
86511 using "args" (variable number of parameters in TCL) and command-line
86512 style option arguments, use one leading "option" parameters, used as a
86513 kind of key/value dictionary of options parsed using `parse_options`.
86515 Change-Id: I63e60d17ae16a020ce4d6de44baf3d152ea42a1a
86517 2021-10-01 Simon Marchi <simon.marchi@polymtl.ca>
86519 gdb/testsuite/dwarf: don't define nested procs for rnglists/loclists
86520 When I wrote support for rnglists and loclists in the testsuite's DWARF
86521 assembler, I made it with nested procs, for example proc "table" inside
86522 proc "rnglists". The intention was that this proc "table" could only be
86523 used by the user while inside proc "rnglists"'s body. I had chosen very
86524 simple names, thinking there was no chance of name clashes. I recently
86525 learned that this is not how TCL works. This ends up defining a proc
86526 "table" in the current namespace ("Dwarf" in this case).
86528 Things still work if you generate rnglists and loclists in the same
86529 file, as each redefines its own procedures when executing. But if a
86530 user of the assembler happened to define a convenience "table" or
86531 "start_end" procedure, for example, it would get overriden.
86533 I'd like to change how this works to reduce the chances of a name clash.
86535 - Move the procs out of each other, so they are not defined in a nested
86537 - Prefix them with "_rnglists_" or "_loclists_".
86538 - While calling $body in the various procs, temporarily make the procs
86539 available under their "short" name. For example, while in rngllists'
86540 body, make _rnglists_table available as just "table". This allows
86541 existing code to keep working and keeps it not too verbose.
86542 - Modify with_override to allow the overriden proc to not exist. In
86543 that case, the temporary proc is deleted on exit.
86545 Note the non-conforming indentation when calling with_override in
86546 _loclists_list. This is on purpose: as we implement more loclists (and
86547 rnglists) entry types, the indentation would otherwise get larger and
86548 larger without much value for readability. So I think it's reasonable
86549 here to put them on the same level.
86551 Change-Id: I7bb48d26fcb0dba1ae4dada05c0c837212424328
86553 2021-10-01 Simon Marchi <simon.marchi@polymtl.ca>
86555 gdb: remove TYPE_FIELD_NAME and FIELD_NAME macros
86556 Remove the `TYPE_FIELD_NAME` and `FIELD_NAME` macros, changing all the
86557 call sites to use field::name directly.
86559 Change-Id: I6900ae4e1ffab1396e24fb3298e94bf123826ca6
86561 2021-10-01 Simon Marchi <simon.marchi@polymtl.ca>
86563 gdb: add field::name / field::set_name
86564 Add the `name` and `set_name` methods on `struct field`, in order to
86565 remove `FIELD_NAME` and `TYPE_FIELD_NAME` macros. In this patch, the
86566 macros are changed to use `field::name`, so all the call sites that are
86567 used to set the field's name are changed to use `field::set_name`.
86568 The next patch will remove the macros completely.
86570 Note that because of the name clash between the existing field named
86571 `name` and the new method, I renamed the field `m_name`. It is not
86572 private per-se, because we can't make `struct field` a non-POD yet, but
86573 it should be considered private anyway (not accessed outside `struct
86576 Change-Id: If16ddbca4e0c39d0ff9da420bb5cdebe5b9b0896
86578 2021-10-01 GDB Administrator <gdbadmin@sourceware.org>
86580 Automatic date update in version.in
86582 2021-09-30 Sergio Durigan Junior <sergiodj@sergiodj.net>
86584 [PR gdb/28369] Use get_shell on gdb/ser-pipe.c
86585 PR gdb/28369 reports that gdb/ser-pipe.c has an 'execl' function call
86586 with a hard-coded "/bin/sh" as its argument. We've had 'get_shell'
86587 for a while now, which is conscious about the SHELL environment and a
86588 better alternative to always calling "/bin/sh".
86590 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28369
86592 2021-09-30 Tom de Vries <tdevries@suse.de>
86594 [gdb/testsuite] Add untested for missing xml support in gdb.base/valgrind*.exp
86595 Add untested in case missing xml support is detected in test-cases
86596 gdb.base/valgrind*.exp.
86598 Tested on x86_64-linux.
86600 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
86602 arm: enable Cortex-R52+ CPU
86603 Patch is adding Cortex-R52+ as 'cortex-r52plus' command line
86604 flag for -mcpu option.
86608 * cpu-arm.c: New Cortex-R52+ CPU.
86612 * NEWS: Update docs.
86613 * config/tc-arm.c: New Cortex-R52+ CPU.
86614 * doc/c-arm.texi: Update docs.
86615 * testsuite/gas/arm/cpu-cortex-r52plus.d: New test.
86617 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
86619 aarch64: Enable Cortex-X2 CPU
86620 This patch is adding support for Cortex-X2 CPU.
86624 * NEWS: Update docs.
86625 * config/tc-aarch64.c: Add Cortex-X2.
86626 * doc/c-aarch64.texi: Update docs.
86628 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
86630 aarch64: Enable Cortex-A710 CPU
86631 This patch is adding support for Cortex-A710 CPU.
86635 * NEWS: Update docs.
86636 * config/tc-aarch64.c: Add Cortex-A710.
86637 * doc/c-aarch64.texi: Update docs.
86639 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
86641 aarch64: Enable Cortex-A510 CPU
86642 This patch is adding support for Cortex-A510 CPU.
86646 * NEWS: Update docs.
86647 * config/tc-aarch64.c: Add Cortex-A510.
86648 * doc/c-aarch64.texi: Update docs.
86650 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
86652 aarch64: Update AArch64 features command line options docs 2/2
86653 Patch is only sorting by 'Extension` column 'Architecture Extension'
86658 * doc/c-aarch64.texi: Update docs.
86660 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
86662 aarch64: Update AArch64 features command line options docs 1/2
86663 Patch is improving entries in "Architecture extensions" table in GAS
86668 * doc/c-aarch64.texi: Update docs.
86670 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
86672 aarch64: add armv9-a architecture to -march
86673 Patch is adding new 'armv9-a` command line flag to -march for AArch64.
86677 * config/tc-aarch64.c: Add 'armv9-a' command line flag.
86678 * docs/c-aarch64.text: Update docs.
86679 * NEWS: Update docs.
86683 * opcode/aarch64.h (AARCH64_FEATURE_V9): New define.
86684 (AARCH64_ARCH_V9): New define.
86686 2021-09-30 Simon Marchi <simon.marchi@polymtl.ca>
86688 gdb/testsuite: make runto_main not pass no-message to runto
86689 As follow-up to this discussion:
86691 https://sourceware.org/pipermail/gdb-patches/2020-August/171385.html
86693 ... make runto_main not pass no-message to runto. This means that if we
86694 fail to run to main, for some reason, we'll emit a FAIL. This is the
86695 behavior we want the majority of (if not all) the time.
86697 Without this, we rely on tests logging a failure if runto_main fails,
86698 otherwise. They do so in a very inconsisteny mannet, sometimes using
86699 "fail", "unsupported" or "untested". The messages also vary widly.
86700 This patch removes all these messages as well.
86702 Also, remove a few "fail" where we call runto (and not runto_main). by
86703 default (without an explicit no-message argument), runto prints a
86704 failure already. In two places, gdb.multi/multi-re-run.exp and
86705 gdb.python/py-pp-registration.exp, remove "message" passed to runto.
86706 This removes a few PASSes that we don't care about (but FAILs will still
86707 be printed if we fail to run to where we want to). This aligns their
86708 behavior with the rest of the testsuite.
86710 Change-Id: Ib763c98c5f4fb6898886b635210d7c34bd4b9023
86712 2021-09-30 Simon Marchi <simon.marchi@polymtl.ca>
86714 gdbsupport: make gdb_mkostemp_cloexec return a scoped_fd
86715 This encourages the callers to use automatic file descriptor management.
86717 Change-Id: I137a81df6f3607b457e28c35aafde8ed6f3a3344
86719 2021-09-30 Simon Marchi <simon.marchi@polymtl.ca>
86721 gdbsupport: make gdb_open_cloexec return scoped_fd
86722 Make gdb_open_cloexec return a scoped_fd, to encourage using automatic
86723 management of the file descriptor closing. Except in the most trivial
86724 cases, I changed the callers to just release the fd, which retains their
86725 existing behavior. That will allow the transition to using scoped_fd
86726 more to go gradually, one caller at a time.
86728 Change-Id: Ife022b403f96e71d5ebb4f1056ef6251b30fe554
86730 2021-09-30 Simon Marchi <simon.marchi@polymtl.ca>
86732 gdbsupport: move gdb_file_up to its own file
86733 The following patches wants to change gdb_fopen_cloexec and
86734 gdb_mkostemp_cloexec to return a scoped_fd. Doing this causes a cyclic
86735 include between scoped_fd.h and filestuff.h, that both want to include
86736 each other. scoped_fd.h includes filestuff.h because of the
86737 scoped_fd::to_file method's return value. filestuff.h would then
86738 include scoped_fd.h for gdb_fopen_cloexec's and gdb_mkostemp_cloexec's
86741 To fix that, move gdb_file_up to its own file, gdb_file.h.
86743 Change-Id: Ic82a48914b2aacee8f14af535b7469245f88b93d
86745 2021-09-30 Dimitar Dimitrov <dimitar@dinux.eu>
86747 ld: pru: Fix resource_table output section alignment
86748 My commit 261980de18b added alignment for the resource table symbol.
86749 But it is wrong. The Linux remoteproc driver loads and interprets the
86750 contents of the .resource_table ELF section, not of a table symbol.
86752 Without this patch, if the linker happens to output padding for symbol
86753 alignment, then the resource table contents as viewed by the kernel
86754 loader would "shift" and look corrupted.
86758 * scripttempl/pru.sc (.resource_table): Align the output
86759 section, not the first symbol.
86761 2021-09-30 Tom Tromey <tromey@adacore.com>
86763 Fix Windows crash from stop_pc change
86764 The "make thread_suspend_state::stop_pc optional" patch caused a
86765 regression on Windows when using shared libraries. I tracked this
86766 down to an unguarded use of stop_pc() in the TARGET_WAITKIND_LOADED
86767 case of handle_inferior_event. This patch fixes the bug by ensuring
86768 that the stop PC is set at this point.
86770 2021-09-30 Tom de Vries <tdevries@suse.de>
86772 [gdb/testsuite] Use untested in gdb.debuginfod/fetch_src_and_symbols.exp
86773 With running test-case gdb.debuginfod/fetch_src_and_symbols.exp with target
86774 board unix/-bad, I get:
86776 gcc: error: unrecognized command line option '-bad'^M
86777 compiler exited with status 1
86778 gdb compile failed, gcc: error: unrecognized command line option '-bad'
86779 FAIL: gdb.debuginfod/fetch_src_and_symbols.exp: compile
86782 Replace the FAIL with the usual:
86784 UNTESTED: gdb.debuginfod/fetch_src_and_symbols.exp: failed to compile
86787 Tested on x86_64-linux.
86789 2021-09-30 Tom de Vries <tdevries@suse.de>
86791 [gdb/testsuite] Remove redundant FAIL in gdb.base/info-os.exp
86792 When running test-case gdb.base/info-os.exp with target board unix/-bad, I run
86795 gdb compile failed, gcc: error: unrecognized command line option '-bad'
86796 UNTESTED: gdb.base/info-os.exp: failed to prepare
86797 FAIL: gdb.base/info-os.exp: cannot compile test program
86800 Remove the redundant FAIL.
86802 Tested on x86_64-linux.
86804 2021-09-30 Tom de Vries <tdevries@suse.de>
86806 [gdb/testsuite] Fix DUPLICATE in gdb.base/info-os.exp
86807 When running test-case gdb.base/info-os.exp, I run into:
86809 PASS: gdb.base/info-os.exp: get threads
86810 PASS: gdb.base/info-os.exp: get threads
86811 DUPLICATE: gdb.base/info-os.exp: get threads
86814 Fix this not doing pass followed by exp_continue in gdb_test_multiple.
86816 Tested on x86_64-linux.
86818 2021-09-30 Tom de Vries <tdevries@suse.de>
86820 [gdb/testsuite] Check compilation result in gdb.dwarf2/dw2-opt-structptr.exp
86821 When running test-case gdb.dwarf2/dw2-opt-structptr.exp with target board
86824 gdb compile failed, gcc: error: unrecognized command line option '-bad'
86825 UNTESTED: gdb.dwarf2/dw2-opt-structptr.exp: dw2-opt-structptr.exp
86826 UNTESTED: gdb.dwarf2/dw2-opt-structptr.exp: failed to compile
86827 ERROR: (dw2-opt-structptr) No such file or directory
86828 UNRESOLVED: gdb.dwarf2/dw2-opt-structptr.exp: console: set print object on
86831 Merge the two UNTESTEDs.
86833 Fix the UNRESOLVED by checking result of compilation.
86835 Tested on x86_64-linux.
86837 2021-09-30 Tom de Vries <tdevries@suse.de>
86839 [gdb/testsuite] Check compilation result in gdb.base/structs.exp
86840 When running test-case gdb.base/structs.exp with target board unix/-bad, I
86843 gdb compile failed, gcc: error: unrecognized command line option '-bad'
86844 UNTESTED: gdb.base/structs.exp: failed to prepare
86845 ERROR: tcl error sourcing src/gdb/testsuite/gdb.base/structs.exp.
86846 ERROR: can't read "use_gdb_stub": no such variable
86849 Fix this by checking the compilation result.
86851 Fix the resulting DUPLICATEs using with_test_prefix.
86853 Tested on x86_64-linux.
86855 2021-09-30 Tom de Vries <tdevries@suse.de>
86857 [gdb/testsuite] Prepare nodebug exec in gdb.base/cvexpr.exp
86858 When running test-case gdb.base/cvexpr.exp with target board unix/-bad, I get:
86860 gdb compile failed, gcc: error: unrecognized command line option '-bad'
86861 ERROR: tcl error sourcing src/gdb/testsuite/gdb.base/cvexpr.exp.
86862 ERROR: can't read "use_gdb_stub": no such variable
86865 This is triggered in a part of the test that claims to require no debug
86866 information, but uses the exec containing either dwarf or ctf.
86868 Fix this by preparing another executable compiled with nodebug, and using
86871 Also use with_test_prefix to mark the nodebug part, such that we have:
86873 gdb compile failed, gcc: error: unrecognized command line option '-bad'
86874 UNTESTED: gdb.base/cvexpr.exp: dwarf: failed to prepare
86875 gdb compile failed, gcc: error: unrecognized command line option '-bad'
86876 UNTESTED: gdb.base/cvexpr.exp: nodebug: failed to prepare
86879 Tested on x86_64-linux.
86881 2021-09-30 Tom de Vries <tdevries@suse.de>
86883 [gdb/testsuite] Fix DUPLICATE in gdb.base/cvexpr.exp
86886 DUPLICATE: gdb.base/cvexpr.exp: ptype int * restrict
86888 using with_test_prefix.
86890 Tested on x86_64-linux.
86892 2021-09-30 Tom de Vries <tdevries@suse.de>
86894 [gdb/testsuite] Check compilation result in gdb.base/call-sc.exp
86895 When running test-case gdb.base/call-sc.exp with target board unix/-bad, I
86898 gdb compile failed, gcc: error: unrecognized command line option '-bad'
86899 UNTESTED: gdb.base/call-sc.exp: failed to prepare
86900 ERROR: tcl error sourcing src/gdb/testsuite/gdb.base/call-sc.exp.
86901 ERROR: can't read "use_gdb_stub": no such variable
86904 Fix this by checking the compilation result.
86906 Fix the resulting DUPLICATE:
86908 DUPLICATE: gdb.base/call-sc.exp: failed to prepare
86910 using with_test_prefix.
86912 Tested on x86_64-linux.
86914 2021-09-30 Tom de Vries <tdevries@suse.de>
86916 [gdb/testsuite] Fix untested messages in gdb.mi/*.exp
86921 in a gdb.x/y.exp is:
86923 UNTESTED: gdb.x/y.exp: y.exp
86925 which is a bit pointless.
86927 Replace these untested messages in gdb.mi/*.exp with the usual "failed to
86934 where the variable is undefined.
86936 Tested on x86_64-linux.
86938 2021-09-30 Nick Clifton <nickc@redhat.com>
86940 make objcopy fail if it is asked to redefine symbols in an object file containing LTO information.
86941 * objcopy.c (filter_symbols): Fail if attempting to dredefine
86942 symbols in an LTO object file.
86944 2021-09-30 Tom de Vries <tdevries@suse.de>
86946 [gdb/testsuite] Fix full buffer in gdb.rust/dwindex.exp
86947 On ubuntu 18.04.5, I run into:
86949 (gdb) mt print objfiles dwindex^M
86951 Object file build/gdb/testsuite/outputs/gdb.rust/dwindex/dwindex: \
86952 Objfile at 0x55dab0b87a50, bfd at 0x55dab0b0cfa0, 1095 minsyms^M
86955 vendor/compiler_builtins/src/int/specialized_div_rem/mod.rs at 0x55dab0db0720^M
86957 library/std/src/sys/unix/stdio.rs at 0x55dab0d96320^M
86958 ERROR: internal buffer is full.
86959 UNRESOLVED: gdb.rust/dwindex.exp: check if index present
86962 Fix this by using -lbl in proc ensure_gdb_index.
86964 Tested on x86_64-linux.
86966 2021-09-30 Libor Bukata <libor.bukata@oracle.com>
86968 Add Solaris specific ELF note processing
86969 Add elfcore_grok_solaris_note function that enables to
86970 obtain process status, register values, and program info
86971 from Solaris's core files.
86974 * elf.c (elfcore_grok_solaris_note): Solaris specific ELF
86975 note parser. Better GDB's coredump analysis on Solaris...
86976 (elfcore_grok_solaris_note_impl): New function.
86977 (elfcore_grok_solaris_prstatus): New function.
86978 (elfcore_grok_solaris_info): New function.
86979 (elfcore_grok_solaris_lwpstatus): New function.
86980 (elf_parse_notes): Added "CORE" groker element.
86982 * elf/common.h: Add note segment constants for core files on
86985 2021-09-30 Frederic Cambus <fred@statdns.com>
86987 Add support to readelf for reading OpenBSD ELF core notes.
86988 * readelf.c (get_openbsd_elfcore_note_type): New function.
86989 (process_note): Add support for OpenBSD core notes.
86991 2021-09-30 GDB Administrator <gdbadmin@sourceware.org>
86993 Automatic date update in version.in
86995 2021-09-29 Tom de Vries <tdevries@suse.de>
86997 [gdb/testsuite] Fix gdb.base/break-interp.exp for ld.so without debug
86998 When running test-case gdb.base/break-interp.exp on openSUSE Leap 42.3, I get:
87000 (gdb) info addr dl_main^M
87001 Symbol "dl_main" is at 0x1750 in a file compiled without debugging.^M
87002 (gdb) FAIL: gdb.base/break-interp.exp: info addr dl_main
87004 while the regexp expects "Symbol \"dl_main\" is a function at address $hex\\."
87006 Fix this by also accepting this variant.
87008 Tested on x86_64-linux.
87010 2021-09-29 H.J. Lu <hjl.tools@gmail.com>
87012 Add a testcase for PR binutils/27202
87014 * testsuite/gas/elf/dwarf-5-loc0.d: New file.
87015 * testsuite/gas/elf/dwarf-5-loc0.s: Likewise.
87016 * testsuite/gas/elf/elf.exp: Run dwarf-5-loc0.
87018 2021-09-29 Pedro Alves <pedro@palves.net>
87020 Fix gdb.multi/multi-term-settings.exp race
87021 The gdb.multi/multi-term-settings.exp testcase sometimes fails like so:
87023 Running /home/pedro/gdb/mygit/src/gdb/testsuite/gdb.multi/multi-term-settings.exp ...
87024 FAIL: gdb.multi/multi-term-settings.exp: inf1_how=attach: inf2_how=attach: stop with control-c (SIGINT)
87026 It's easier to reproduce if you stress the machine at the same time, like e.g.:
87030 Looking at gdb.log, we see:
87033 Attaching to program: build/gdb/testsuite/outputs/gdb.multi/multi-term-settings/multi-term-settings, process 60422
87034 [New Thread 60422.60422]
87035 Reading symbols from /lib/x86_64-linux-gnu/libc.so.6...
87036 Reading symbols from /usr/lib/debug//lib/x86_64-linux-gnu/libc-2.31.so...
87037 Reading symbols from /lib64/ld-linux-x86-64.so.2...
87038 (No debugging symbols found in /lib64/ld-linux-x86-64.so.2)
87039 0x00007f2fc2485334 in __GI___clock_nanosleep (clock_id=<optimized out>, clock_id@entry <mailto:clock_id@entry>=0, flags=flags@entry <mailto:flags@entry>=0, req=req@entry <mailto:req@entry>=0x7ffe23126940, rem=rem@entry <mailto:rem@entry>=0x0) at ../sysdeps/unix/sysv/linux/clock_nanosleep.c:78
87040 78 ../sysdeps/unix/sysv/linux/clock_nanosleep.c: No such file or directory.
87041 (gdb) PASS: gdb.multi/multi-term-settings.exp: inf1_how=attach: inf2_how=attach: inf2: attach
87042 set schedule-multiple on
87043 (gdb) PASS: gdb.multi/multi-term-settings.exp: inf1_how=attach: inf2_how=attach: set schedule-multiple on
87045 Num Description Connection Executable
87046 1 process 60404 1 (extended-remote localhost:2349) build/gdb/testsuite/outputs/gdb.multi/multi-term-settings/multi-term-settings
87047 * 2 process 60422 1 (extended-remote localhost:2349) build/gdb/testsuite/outputs/gdb.multi/multi-term-settings/multi-term-settings
87048 (gdb) PASS: gdb.multi/multi-term-settings.exp: inf1_how=attach: inf2_how=attach: info inferiors
87049 pid=60422, count=46
87050 pid=60422, count=47
87051 pid=60422, count=48
87052 pid=60422, count=49
87053 pid=60422, count=50
87054 pid=60422, count=51
87055 pid=60422, count=52
87056 pid=60422, count=53
87057 pid=60422, count=54
87058 pid=60422, count=55
87059 pid=60422, count=56
87060 pid=60422, count=57
87061 pid=60422, count=58
87062 pid=60422, count=59
87063 pid=60422, count=60
87064 pid=60422, count=61
87065 pid=60422, count=62
87066 pid=60422, count=63
87067 pid=60422, count=64
87068 pid=60422, count=65
87069 pid=60422, count=66
87070 pid=60422, count=67
87071 pid=60422, count=68
87072 pid=60422, count=69
87073 pid=60404, count=54
87074 pid=60404, count=55
87075 pid=60404, count=56
87076 pid=60404, count=57
87077 pid=60404, count=58
87078 PASS: gdb.multi/multi-term-settings.exp: inf1_how=attach: inf2_how=attach: continue
87080 (gdb) FAIL: gdb.multi/multi-term-settings.exp: inf1_how=attach: inf2_how=attach: stop with control-c (SIGINT)
87082 If you look at the testcase's sources, you'll see that the intention
87083 is to resumes the program with "continue", wait to see a few of those
87084 "pid=..., count=..." lines, and then interrupt the program with
87085 Ctrl-C. But somehow, that resulted in GDB printing "Quit", instead of
87086 the Ctrl-C stopping the program with SIGINT.
87088 Here's what is happening:
87090 #1 - those "pid=..., count=..." lines we see above weren't actually
87091 output by the inferior after it has been continued (see #1).
87092 Note that "inf1_how" and "inf2_how" are "attach". What happened
87093 is that those "pid=..., count=..." lines were output by the
87094 inferiors _before_ they were attached to. We see them at that
87095 point instead of earlier, because that's where the testcase
87096 reads from the inferiors' spawn_ids.
87098 #2 - The testcase mistakenly thinks those "pid=..., count=..." lines
87099 happened after the continue was processed by GDB, meaning it has
87100 waited enough, and so sends the Ctrl-C. GDB hasn't yet passed
87101 the terminal to the inferior, so the Ctrl-C results in that
87104 The fix here is twofold:
87106 #1 - flush inferior output right after attaching
87108 #2 - consume the "Continuing" printed by "continue", indicating the
87109 inferior has the terminal. This is the same as done throughout
87110 the testsuite to handle this exact problem of sending Ctrl-C too
87113 gdb/testsuite/ChangeLog:
87114 yyyy-mm-dd Pedro Alves <pedro@palves.net <mailto:pedro@palves.net>>
87116 * gdb.multi/multi-term-settings.exp (create_inferior): Flush
87118 (coretest): Use $gdb_test_name. After issuing "continue", wait
87121 Change-Id: Iba7671dfe1eee6b98d29cfdb05a1b9aa2f9defb9
87123 2021-09-29 Tom de Vries <tdevries@suse.de>
87125 [gdb/testsuite] Disable vgdb tests if xml not supported
87126 I build gdb without xml support using --without-expat, and ran into:
87128 (gdb) target remote | vgdb --wait=2 --max-invoke-ms=2500 --pid=22032^M
87129 Remote debugging using | vgdb --wait=2 --max-invoke-ms=2500 --pid=22032^M
87130 relaying data between gdb and process 22032^M
87131 warning: Can not parse XML target description; XML support was disabled at \
87134 (gdb) PASS: gdb.base/valgrind-infcall.exp: continue #1
87135 p gdb_test_infcall ()^M
87136 Remote 'g' packet reply is too long (expected 560 bytes, got 800 bytes): ...^M
87137 (gdb) FAIL: gdb.base/valgrind-infcall.exp: p gdb_test_infcall ()
87140 After googling the error message with context valgrind gdbserver, I found
87141 indications that the Remote 'g' packet reply error is due to missing xml
87144 And here ( https://www.valgrind.org/docs/manual/manual-core-adv.html ) I
87147 GDB version needed for ARM and PPC32/64.
87149 You must use a GDB version which is able to read XML target description sent
87150 by a gdbserver. This is the standard setup if GDB was configured and built
87151 with the "expat" library. If your GDB was not configured with XML support, it
87152 will report an error message when using the "target" command. Debugging will
87153 not work because GDB will then not be able to fetch the registers from the
87154 Valgrind gdbserver.
87157 So I guess I'm running into the same problem for x86_64.
87159 Fix this by skipping all gdb.base/valgrind-*.exp tests if xml support is not
87160 available. Although only the gdb.base/valgrind-infcall*.exp produce fails,
87161 the Remote 'g' packet reply error occurs in all tests, so it seems prudent to
87164 Tested on x86_64-linux.
87166 2021-09-29 Tom de Vries <tdevries@suse.de>
87168 [gdb/testsuite] Fix gdb.python/py-breakpoint.exp with python 2
87169 With a gdb build using python 2.7, I run into:
87172 gdb.events.breakpoint_modified.connect(lambda bp: print(bp.enabled))^M
87173 File "<string>", line 1^M
87174 gdb.events.breakpoint_modified.connect(lambda bp: print(bp.enabled))^M
87176 SyntaxError: invalid syntax^M
87177 Error while executing Python code.^M
87178 (gdb) FAIL: gdb.python/py-breakpoint.exp: test_bkpt_auto_disable: \
87179 trap breakpoint_modified event
87182 This is caused by the following:
87183 - a lambda function body needs to be an expression
87184 - in python 2, print is a statement, while in python 3 it's a function
87185 - a function call is an expression, and a statement is not.
87187 Fix this by defining a function print_bp_enabled:
87189 def print_bp_enabled (bp):
87193 and using that instead.
87195 Tested on x86_64-linux.
87197 2021-09-29 Tom de Vries <tdevries@suse.de>
87199 [gdb/testsuite] Fix breakpoint detection in gdb.gdb/python-helper.exp
87200 With a gdb configured to be somewhat minimal, while still supporting python:
87202 $ gdb --configuration
87203 This GDB was configured as follows:
87204 configure --host=x86_64-pc-linux-gnu --target=x86_64-pc-linux-gnu
87205 --with-auto-load-dir=$debugdir:$datadir/auto-load
87206 --with-auto-load-safe-path=$debugdir:$datadir/auto-load
87208 --with-gdb-datadir=$install/share/gdb (relocatable)
87209 --with-jit-reader-dir=$install/lib64/gdb (relocatable)
87210 --without-libunwind-ia64
87212 --without-babeltrace
87217 --with-python-libdir=/usr/lib
87220 --disable-source-highlight
87221 --with-separate-debug-dir=/usr/lib/debug
87222 --with-system-gdbinit=$devel/system-gdbinit
87224 and using gcc 4.8 to build gdb (causing std::thread not to be used due to
87225 PR28318) I ran into:
87227 (gdb) PASS: gdb.gdb/python-helper.exp: start inner gdb
87230 Breakpoint 2, value_print () at src/gdb/valprint.c:1174^M
87231 1174 scoped_value_mark free_values;^M
87232 (xgdb) FAIL: gdb.gdb/python-helper.exp: hit breakpoint in inner gdb (timeout)
87235 The problem is that the regexp expects "hit Breakpoint $decimal". The "hit"
87238 The "hit" is printed by maybe_print_thread_hit_breakpoint, when
87239 show_thread_that_caused_stop returns true:
87242 show_thread_that_caused_stop (void)
87244 return highest_thread_num > 1;
87247 Apparently, that's not the case.
87249 Fix this by removing "hit" from the regexp, making the regexp more similar to
87250 what is used in say, continue_to_breakpoint.
87252 Tested on x86_64-linux.
87254 2021-09-29 Andrew Burgess <andrew.burgess@embecosm.com>
87256 gdb: fix build when libbacktrace and execinfo backtrace are not available
87259 commit abbbd4a3e0ca51132e7fb31a43f896d29894dae0
87260 Date: Wed Aug 11 13:24:33 2021 +0100
87262 gdb: use libbacktrace to create a better backtrace for fatal signals
87264 The build of GDB was broken iff, the execinfo backtrace API is not
87265 available, and, libbacktrace is either disabled, or not usable. In
87266 this case you'll see build errors like this:
87269 /home/username/src/binutils-gdb/gdb/bt-utils.c: In function 'void gdb_internal_backtrace()':
87270 /home/username/src/binutils-gdb/gdb/bt-utils.c:165:5: error: 'gdb_internal_backtrace_1' was not declared in this scope
87271 gdb_internal_backtrace_1 ();
87272 ^~~~~~~~~~~~~~~~~~~~~~~~
87274 This commit fixes the issue by guarding the call to
87275 gdb_internal_backtrace_1 with '#ifdef GDB_PRINT_INTERNAL_BACKTRACE',
87276 which is only defined when one of the backtrace libraries are
87279 2021-09-29 Andrew Burgess <andrew.burgess@embecosm.com>
87281 gdb/doc: use 'standard error stream' instead of 'stderr' in some places
87284 commit 91f2597bd24d171c1337a4629f8237aa47c59082
87285 Date: Thu Aug 12 18:24:59 2021 +0100
87287 gdb: print backtrace for internal error/warning
87289 I included some references to 'stderr', which, it was pointed out,
87290 would be better written as 'standard error stream'. See:
87292 https://sourceware.org/pipermail/gdb-patches/2021-September/182225.html
87294 This commit replaces the two instances of 'stderr' that I introduced.
87296 2021-09-29 Andrew Burgess <andrew.burgess@embecosm.com>
87298 gdb: fix manor -> manner typo in some comments
87299 In a recent commit I used 'manor' in some comments rather than
87300 'manner'. This commit fixes those two mistakes.
87302 I also looked through the gdb/ tree and found one additional instance
87303 of this mistake that this commit also fixes.
87305 2021-09-29 Alan Modra <amodra@gmail.com>
87307 PR27202, readelf -wL doesn't work on ".loc 0"
87308 For DWARF revision 4 and earlier, display_debug_lines_decoded
87309 populates the file_table array with entries read from .debug_line
87310 after the directory table. file_table[0] contains the first entry.
87311 DWARF rev 4 line number programs index this entry as file number one.
87312 DWARF revision 5 changes .debug_line format quite extensively, and in
87313 particular gives file number zero a meaning.
87316 * dwarf.c (display_debug_lines_decoded): Correct indexing used
87319 2021-09-29 Simon Marchi <simon.marchi@polymtl.ca>
87321 gdb: enable target_async around stop_all_threads call in process_initial_stop_replies
87322 The following scenario hangs:
87324 - maint set target-non-stop on
87325 - `gdbserver --attach`
87326 - a multi-threaded program
87332 $ gnome-calculator&
87334 $ ../gdbserver/gdbserver --once --attach :1234 495731
87335 Attached; pid = 495731
87336 Listening on port 1234
87340 $ ./gdb -nx -q --data-directory=data-directory /usr/bin/gnome-calculator -ex "maint set target-non-stop on" -ex "tar rem :1234"
87341 Reading symbols from /usr/bin/gnome-calculator...
87342 (No debugging symbols found in /usr/bin/gnome-calculator)
87343 Remote debugging using :1234
87348 - The protocol between gdb and gdbserver is in non-stop mode, but the
87349 user-visible behavior is all-stop
87350 - On connect, gdbserver sends one stop reply for one thread that is
87351 stops, the others stay running
87352 - In process_initial_stop_replies, gdb calls stop_all_threads to stop
87353 these other threads, because we are using the all-stop user-visible
87355 - stop_all_threads sends a stop request for all the running threads and
87356 then waits for resulting events
87357 - At this point, the remote target is in target_async(0) mode, which
87358 makes stop_all_threads not consider it for events
87359 - stop_all_threads loops indefinitely (it does not even block
87360 indefinitely, it is in an infinite busy loop) because there are no
87361 event sources. wait_one_event returns a TARGET_WAITKIND_NO_RESUMED
87364 Fix that by making the remote target async around the stop_all_threads
87367 I haven't implemented it because I'm not sure how to do it, but I think
87368 it would be a good idea to have, in stop_all_threads / wait_one /
87369 handle_one, an assert to check that if we are expecting one or more
87370 event, then there are some targets that are in a state where they can
87371 supply some events. Otherwise, we'll necessarily be stuck in this
87372 infinite loop, and it's probably due to a bug in GDB. I'm not too sure
87373 where to put this or how to express it though. Perhaps in
87374 stop_all_threads, here:
87376 for (int i = 0; i < waits_needed; i++)
87378 wait_one_event event = wait_one ();
87380 if (handle_one (event))
87384 If at that point, the returned event is TARGET_WAITKIND_NO_RESUMED,
87385 there's a problem. We expect some event, because we've asked some
87386 threads to stop, but all targets are answering that they won't have any
87387 events for us. That's a contradiction, and a sign that something has
87388 gone wrong. It could perhaps event be:
87390 gdb_assert (event.ws.kind != TARGET_WAITKIND_NO_RESUMED);
87392 in handle_one, as the idea is the same in prepare_for_detach.
87394 A bit more sophisticated would be: we know which targets we are
87395 expecting waits from, since we know which threads we have asked to
87396 stop. So if any of these targets returns TARGET_WAITKIND_NO_RESUMED,
87397 something is fishy.
87399 Add a test that tests attaching with gdbserver's --attach flag to a
87400 multi-threaded program, and then connecting to it. Without the fix, the
87401 test reproduces the hang.
87403 Change-Id: If6f6690a4887ca66693ef1af64791dda4c65f24f
87405 2021-09-29 GDB Administrator <gdbadmin@sourceware.org>
87407 Automatic date update in version.in
87409 2021-09-29 Simon Marchi <simon.marchi@polymtl.ca>
87411 gdb: fix darwin-nat build (again)
87412 I made a mistake in the previous patch. Adjust the format string to
87413 match the arguments.
87415 Change-Id: I4d45e0e0adb78eb3b5a06ba1a5287155940056ba
87417 2021-09-29 Simon Marchi <simon.marchi@polymtl.ca>
87419 gdb: fix darwin-nat build
87420 There are two errors of this kind:
87423 /Users/smarchi/src/binutils-gdb/gdb/darwin-nat.c:1175:19: error: format specifies type 'unsigned long' but the argument has type 'ULONGEST' (aka 'unsigned long long') [-Werror,-Wformat]
87424 ptid.pid (), ptid.tid ());
87427 Fix them by using ptid_t's to_string method.
87429 Change-Id: I52087d5f7ee0fc01ac8b3f87d4db0217cb0d7cc7
87431 2021-09-29 Simon Marchi <simon.marchi@polymtl.ca>
87433 gdb.base/foll-fork.exp: accept "info breakpoints" output in any order
87434 The test currently requires the "inf 1" breakpoint to be before the "inf
87435 2" breakpoint. This is not always the case:
87438 Num Type Disp Enb Address What
87439 2 breakpoint keep y <MULTIPLE>
87440 2.1 y 0x0000555555554730 in callee at /home/simark/src/binutils-gdb/gdb/testsuite/gdb.base/foll-fork.c:9 inf 2
87441 2.2 y 0x0000555555554730 in callee at /home/simark/src/binutils-gdb/gdb/testsuite/gdb.base/foll-fork.c:9 inf 1
87442 (gdb) FAIL: gdb.base/foll-fork.exp: follow-fork-mode=parent: detach-on-fork=off: cmd=next 2: test_follow_fork: info breakpoints
87444 Since add_location_to_breakpoint uses only the address as a criterion to
87445 sort locations, the order of locations at the same address is not
87446 stable: it will depend on the insertion order. Here, the insertion
87447 order comes from the order of SALs when creating the breakpoint, which
87448 can vary from machine to machine. While it would be more user-friendly
87449 to have a more stable order for printed breakpoint locations, it doesn't
87450 really matter for this test, and it would be hard to define an order
87451 that will be the same everywhere, all the time.
87453 So, loosen the regexp to accept "inf 1" and "inf 2" in any order.
87455 Co-Authored-By: Pedro Alves <pedro@palves.net>
87456 Change-Id: I5ada2e0c6ad0669e0d161bfb6b767229c0970d16
87458 2021-09-28 Cooper Qu <cooper.qu@linux.alibaba.com>
87460 RISC-V: Fix wrong version number when arch contains 'p'.
87461 When specify a default version for p extension in
87462 riscv_supported_std_ext[](elfxx-riscv.c) and assembling with
87463 -march=rv32imacp, the c extension's version in attribute will become
87464 0p0, the expectation is 2p0.
87466 TODO: Remember to add testcase when we have supported standrad p in
87471 * elfxx-riscv.c (riscv_parsing_subset_version): Break if p
87472 represent the 'p' extension.
87474 Change-Id: Ia4e0cf26f3d7d07acaee8cefd86707ecac663a59
87476 2021-09-28 Nelson Chu <nelson.chu@sifive.com>
87478 RISC-V: Allow to add numbers in the prefixed extension names.
87479 We need to allow adding numbers in the prefixed extension names, since
87480 the zve<32,64><d,f,x> extensions are included in the forzen rvv v1.0 spec
87481 recently. But there are two restrictions as follows,
87483 * The extension name ends with <number>p is invalid, since this may
87484 be confused with extension with <number>.0 version. We report errors
87487 Invalid format: [z|h|s|zvm|x][0-9a-z]+[0-9]+p
87489 * The extension name ends with numbers is valid, but the numbers will
87490 be parsed as major version, so try to avoid naming extensions like this.
87493 * elfxx-riscv.c (riscv_recognized_prefixed_ext): Renamed from
87494 riscv_valid_prefixed_ext/
87495 (riscv_parsing_subset_version): The extensions end with <number>p
87496 is forbidden, we already report the detailed errors in the
87497 riscv_parse_prefixed_ext, so clean the code and unused parameters.
87498 (riscv_parse_std_ext): Updated.
87499 (riscv_parse_prefixed_ext): Rewrite the parser to allow numbers
87500 in the prefixed extension names.
87502 * testsuite/gas/riscv/march-fail-invalid-x-01.d: New testcases.
87503 * testsuite/gas/riscv/march-fail-invalid-x-02.d: Likewise.
87504 * testsuite/gas/riscv/march-fail-invalid-z-01.d: Likewise.
87505 * testsuite/gas/riscv/march-fail-invalid-z-02.d: Likewise.
87506 * testsuite/gas/riscv/march-fail-invalid.l: Likewise.
87507 * testsuite/gas/riscv/march-fail-version-x.d: Removed.
87508 * testsuite/gas/riscv/march-fail-version-z.d: Likewise.
87509 * testsuite/gas/riscv/march-fail-version.l: Likewise.
87511 2021-09-28 Andrew Burgess <andrew.burgess@embecosm.com>
87513 gdb: print backtrace for internal error/warning
87514 This commit builds on previous work to allow GDB to print a backtrace
87515 of itself when GDB encounters an internal-error or internal-warning.
87516 This fixes PR gdb/26377.
87518 There's not many places where we call internal_warning, and I guess in
87519 most cases the user would probably continue their debug session. And
87520 so, in order to avoid cluttering up the output, by default, printing
87521 of a backtrace is off for internal-warnings.
87523 In contrast, printing of a backtrace is on by default for
87524 internal-errors, as I figure that in most cases hitting an
87525 internal-error is going to be the end of the debug session.
87527 Whether a backtrace is printed or not can be controlled with the new
87530 maintenance set internal-error backtrace on|off
87531 maintenance show internal-error backtrace
87533 maintenance set internal-warning backtrace on|off
87534 maintenance show internal-warning backtrace
87536 Here is an example of what an internal-error now looks like with the
87537 backtrace included:
87539 (gdb) maintenance internal-error blah
87540 ../../src.dev-3/gdb/maint.c:82: internal-error: blah
87541 A problem internal to GDB has been detected,
87542 further debugging may prove unreliable.
87543 ----- Backtrace -----
87544 0x5c61ca gdb_internal_backtrace_1
87545 ../../src.dev-3/gdb/bt-utils.c:123
87546 0x5c626d _Z22gdb_internal_backtracev
87547 ../../src.dev-3/gdb/bt-utils.c:165
87548 0xe33237 internal_vproblem
87549 ../../src.dev-3/gdb/utils.c:393
87550 0xe33539 _Z15internal_verrorPKciS0_P13__va_list_tag
87551 ../../src.dev-3/gdb/utils.c:470
87552 0x1549652 _Z14internal_errorPKciS0_z
87553 ../../src.dev-3/gdbsupport/errors.cc:55
87554 0x9c7982 maintenance_internal_error
87555 ../../src.dev-3/gdb/maint.c:82
87556 0x636f57 do_simple_func
87557 ../../src.dev-3/gdb/cli/cli-decode.c:97
87558 .... snip, lots more backtrace lines ....
87559 ---------------------
87560 ../../src.dev-3/gdb/maint.c:82: internal-error: blah
87561 A problem internal to GDB has been detected,
87562 further debugging may prove unreliable.
87563 Quit this debugging session? (y or n) y
87565 This is a bug, please report it. For instructions, see:
87566 <https://www.gnu.org/software/gdb/bugs/>.
87568 ../../src.dev-3/gdb/maint.c:82: internal-error: blah
87569 A problem internal to GDB has been detected,
87570 further debugging may prove unreliable.
87571 Create a core file of GDB? (y or n) n
87573 My hope is that this backtrace might make it slightly easier to
87574 diagnose GDB issues if all that is provided is the console output, I
87575 find that we frequently get reports of an assert being hit that is
87576 located in pretty generic code (frame.c, value.c, etc) and it is not
87577 always obvious how we might have arrived at the assert.
87579 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=26377
87581 2021-09-28 Andrew Burgess <andrew.burgess@embecosm.com>
87583 gdb: use libbacktrace to create a better backtrace for fatal signals
87584 GDB recently gained the ability to print a backtrace when a fatal
87585 signal is encountered. This backtrace is produced using the backtrace
87586 and backtrace_symbols_fd API available in glibc.
87588 However, in order for this API to actually map addresses to symbol
87589 names it is required that the application (GDB) be compiled with
87590 -rdynamic, which GDB is not by default.
87592 As a result, the backtrace produced often looks like this:
87594 Fatal signal: Bus error
87595 ----- Backtrace -----
87596 ./gdb/gdb[0x80ec00]
87597 ./gdb/gdb[0x80ed56]
87598 /lib64/libc.so.6(+0x3c6b0)[0x7fc2ce1936b0]
87599 /lib64/libc.so.6(__poll+0x4f)[0x7fc2ce24da5f]
87600 ./gdb/gdb[0x15495ba]
87601 ./gdb/gdb[0x15489b8]
87602 ./gdb/gdb[0x9b794d]
87603 ./gdb/gdb[0x9b7a6d]
87604 ./gdb/gdb[0x9b943b]
87605 ./gdb/gdb[0x9b94a1]
87606 ./gdb/gdb[0x4175dd]
87607 /lib64/libc.so.6(__libc_start_main+0xf3)[0x7fc2ce17e1a3]
87608 ./gdb/gdb[0x4174de]
87609 ---------------------
87611 This is OK if you have access to the exact same build of GDB, you can
87612 manually map the addresses back to symbols, however, it is next to
87613 useless if all you have is a backtrace copied into a bug report.
87615 GCC uses libbacktrace for printing a backtrace when it encounters an
87616 error. In recent commits I added this library into the binutils-gdb
87617 repository, and in this commit I allow this library to be used by
87618 GDB. Now (when GDB is compiled with debug information) the backtrace
87621 ----- Backtrace -----
87622 0x80ee08 gdb_internal_backtrace
87623 ../../src/gdb/event-top.c:989
87624 0x80ef0b handle_fatal_signal
87625 ../../src/gdb/event-top.c:1036
87628 0x154976f gdb_wait_for_event
87629 ../../src/gdbsupport/event-loop.cc:613
87630 0x1548b6d _Z16gdb_do_one_eventv
87631 ../../src/gdbsupport/event-loop.cc:237
87632 0x9b7b02 start_event_loop
87633 ../../src/gdb/main.c:421
87634 0x9b7c22 captured_command_loop
87635 ../../src/gdb/main.c:481
87636 0x9b95f0 captured_main
87637 ../../src/gdb/main.c:1353
87638 0x9b9656 _Z8gdb_mainP18captured_main_args
87639 ../../src/gdb/main.c:1368
87641 ../../src/gdb/gdb.c:32
87642 ---------------------
87644 Which seems much more useful.
87646 Use of libbacktrace is optional. If GDB is configured with
87647 --disable-libbacktrace then the libbacktrace directory will not be
87648 built, and GDB will not try to use this library. In this case GDB
87649 would try to use the old backtrace and backtrace_symbols_fd API.
87651 All of the functions related to writing the backtrace of GDB itself
87652 have been moved into the new files gdb/by-utils.{c,h}.
87654 2021-09-28 Andrew Burgess <andrew.burgess@embecosm.com>
87656 src-release.sh: add libbacktrace to GDB_SUPPORT_DIRS
87657 After the previous commit that imported libbacktrace from gcc, this
87658 commit updates src-release.sh so that the libbacktrace directory is
87659 included in the gdb release tar file.
87663 * src-release.sh (GDB_SUPPPORT_DIRS): Add libbacktrace.
87665 2021-09-28 Andrew Burgess <andrew.burgess@embecosm.com>
87667 Copy in libbacktrace from gcc
87668 This copies in libbacktrace from the gcc repository as it was in the
87669 commit 62e420293a293608f383d9b9c7f2debd666e9fc9. GDB is going to
87670 start using this library soon.
87672 A dependency between GDB and libbacktrace has already been added to
87673 the top level Makefile, so, after this commit, when building GDB,
87674 libbacktrace will be built first. However, libbacktrace is not yet
87675 linked into GDB, or used by GDB in any way.
87677 It is possible to stop libbacktrace being built by configuring the
87678 tree with --disable-libbacktrace.
87680 This commit does NOT update src-release.sh, that will be done in the
87681 next commit, this commit ONLY imports libbacktrace from gcc. This
87682 means that if you try to make a release of GDB from exactly this
87683 commit then the release tar file will not include libbacktrace.
87684 However, as libbacktrace is an optional dependency this is fine.
87686 2021-09-28 Andrew Burgess <andrew.burgess@embecosm.com>
87688 gdb: Add a dependency between gdb and libbacktrace
87689 GDB is going to start using libbacktrace, so add a build dependency.
87693 * Makefile.def: Add all-gdb dependency on all-libbacktrace.
87694 * Makefile.in: Regenerate.
87696 2021-09-28 Andrew Burgess <andrew.burgess@embecosm.com>
87698 top-level configure: setup target_configdirs based on repository
87699 The top-level configure script is shared between the gcc repository
87700 and the binutils-gdb repository.
87702 The target_configdirs variable in the configure.ac script, defines
87703 sub-directories that contain components that should be built for the
87704 target using the target tools.
87706 Some components, e.g. zlib, are built as both host and target
87709 This causes problems for binutils-gdb. If we run 'make all' in the
87710 binutils-gdb repository we end up trying to build a target version of
87711 the zlib library, which requires the target compiler be available.
87712 Often the target compiler isn't immediately available, and so the
87715 The problem with zlib impacted a previous attempt to synchronise the
87716 top-level configure scripts from gcc to binutils-gdb, see this thread:
87718 https://sourceware.org/pipermail/binutils/2019-May/107094.html
87720 And I'm in the process of importing libbacktrace in to binutils-gdb,
87721 which is also a host and target library, and triggers the same issues.
87723 I believe that for binutils-gdb, at least at the moment, there are no
87724 target libraries that we need to build.
87726 In the configure script we build three lists of things we want to
87727 build, $configdirs, $build_configdirs, and $target_configdirs, we also
87728 build two lists of things we don't want to build, $skipdirs and
87729 $noconfigdirs. We then remove anything that is in the lists of things
87730 not to build, from the list of things that should be built.
87732 My proposal is to add everything in target_configdirs into skipdirs,
87733 if the source tree doesn't contain a gcc/ sub-directory. The result
87734 is that for binutils-gdb no target tools or libraries will be built,
87735 while for the gcc repository, nothing should change.
87737 If a user builds a unified source tree, then the target tools and
87738 libraries should still be built as the gcc/ directory will be present.
87740 I've tested a build of gcc on x86-64, and the same set of target
87741 libraries still seem to get built. On binutils-gdb this change
87742 resolves the issues with 'make all'.
87746 * configure: Regenerate.
87747 * configure.ac (skipdirs): Add the contents of target_configdirs if
87748 we are not building gcc.
87750 2021-09-28 Gleb Fotengauer-Malinovskiy <glebfm@altlinux.org>
87752 PR28391, strip/objcopy --preserve-dates *.a: cannot set time
87753 After commit 985e0264516 copy_archive function began to pass invalid
87754 values to the utimensat(2) function when it tries to preserve
87755 timestamps in ar archives. This happens because the bfd_stat_arch_elt
87756 implementation for ar archives fills only the st_mtim.tv_sec part of
87757 the st_mtim timespec structure, but leaves the st_mtim.tv_nsec part
87758 and the whole st_atim timespec untouched leaving them uninitialized
87761 * ar.c (extract_file): Clear buf for preserve_dates.
87762 * objcopy.c (copy_archive): Likewise.
87764 2021-09-28 Mike Frysinger <vapier@gentoo.org>
87766 sim: drop weak func attrs on module inits
87767 When I first wrote this, I was thinking we'd scan all source files
87768 that existed and generate a complete init list. That means for any
87769 particular build, we'd probably have a few functions that didn't
87770 exist, so weak attributes was necessary. What I ended up scanning
87771 though was only the source files that went into a particular build.
87773 There was another concern too: a source file might be included, but
87774 the build settings would cause all of its contents to be skipped
87775 (via CPP defines). So scanning via naive grep would pick up names
87776 not actually available. A check of the source tree shows that we
87777 never do this, and it's pretty easy to institute a policy that we
87778 don't start (by at the very least including a stub init func).
87780 The use of weak symbols ends up causing a problem in practice: for
87781 a few modules (like profiling), nothing else pulls it in, so the
87782 linker omits it entirely, which leads to the profiling module never
87783 being available. So drop the weak markings since we know all these
87784 funcs will be available.
87786 2021-09-28 Cui,Lili <lili.cui@intel.com>
87788 x86: Print {bad} on invalid broadcast in OP_E_memory
87789 Don't print broadcast for scalar_mode, and print {bad} for invalid broadcast.
87794 * testsuite/gas/i386/bad-bcast.s: Add a new testcase.
87795 * testsuite/gas/i386/bad-bcast.d: Likewise.
87796 * testsuite/gas/i386/bad-bcast-intel.d: New.
87801 * i386-dis.c (static struct): Add no_broadcast.
87802 (OP_E_memory): Mark invalid broadcast with no_broadcast=1 and Print "{bad}"for it.
87803 (intel_operand_size): mark invalid broadcast with no_broadcast=1.
87804 (OP_XMM): Mark scalar_mode with no_broadcast=1.
87806 2021-09-28 Simon Marchi <simon.marchi@polymtl.ca>
87808 gdb: use intrusive_list for linux-nat lwp_list
87809 Replace the manually maintained linked list of lwp_info objects with
87810 intrusive_list. Replace the ALL_LWPS macro with all_lwps, which returns
87811 a range. Add all_lwps_safe as well, for use in iterate_over_lwps, which
87812 currently iterates in a safe manner.
87814 Change-Id: I355313502510acc0103f5eaf2fbde80897d6376c
87816 2021-09-28 Simon Marchi <simon.marchi@polymtl.ca>
87818 gdb: add destructor to lwp_info
87819 Replace the lwp_free function with a destructor. Make lwp_info
87820 non-copyable, since there is now a destructor (we wouldn't want an
87821 lwp_info object getting copied and this->arch_private getting deleted
87824 Change-Id: I09fcbe967e362566d3a06fed2abca2a9955570fa
87826 2021-09-28 Simon Marchi <simon.marchi@polymtl.ca>
87828 gdb: make lwp_info non-POD
87829 Initialize all fields in the class declaration directly. This opens the
87830 door to using intrusive_list, done in the following patch.
87832 Change-Id: I38bb27410cd9ebf511d310bb86fe2ea1872c3b05
87834 2021-09-28 GDB Administrator <gdbadmin@sourceware.org>
87836 Automatic date update in version.in
87838 2021-09-27 Simon Marchi <simon.marchi@efficios.com>
87840 gdb: don't share aspace/pspace on fork with "detach-on-fork on" and "follow-fork-mode child"
87841 We found that when handling forks, two inferiors can unexpectedly share
87842 their program space and address space. To reproduce:
87844 1. Using a test program that forks...
87845 2. "set follow-fork-mode child"
87846 3. "set detach-on-fork on" (the default)
87847 4. run to a breakpoint somewhere after the fork
87849 Step 4 should have created a new inferior:
87851 (gdb) info inferiors
87852 Num Description Connection Executable
87853 1 <null> /home/smarchi/build/wt/amd/gdb/fork
87854 * 2 process 251425 1 (native) /home/smarchi/build/wt/amd/gdb/fork
87856 By inspecting the state of GDB, we can see that the two inferiors now
87857 share one program space and one address space:
87861 (top-gdb) p inferior_list.m_front.num
87863 (top-gdb) p inferior_list.m_front.aspace
87864 $3 = (struct address_space *) 0x5595e2520400
87865 (top-gdb) p inferior_list.m_front.pspace
87866 $4 = (struct program_space *) 0x5595e2520440
87870 (top-gdb) p inferior_list.m_front.next.num
87872 (top-gdb) p inferior_list.m_front.next.aspace
87873 $6 = (struct address_space *) 0x5595e2520400
87874 (top-gdb) p inferior_list.m_front.next.pspace
87875 $7 = (struct program_space *) 0x5595e2520440
87877 You can then run inferior 1 again and the two inferiors will still
87878 erroneously share their spaces, but already at this point this is wrong.
87880 The cause of the bad {a,p}space sharing is in follow_fork_inferior.
87881 When following the child and detaching from the parent, we just re-use
87882 the parent's spaces, rather than cloning them. When we switch back to
87883 inferior 1 and run again, we find ourselves with two unrelated inferiors
87886 Fix that by creating new spaces for the parent after having moved them
87887 to the child. My initial implementation created new spaces for the
87888 child instead. Doing this breaks doing "next" over fork(). When "next"
87889 start, we record the symtab of the starting location. When the program
87890 stops, we compare that symtab with the symtab the program has stopped
87891 at. If the symtab or the line number has changed, we conclude the
87892 "next" is done. If we create a new program space for the child and copy
87893 the parent's program space to it with clone_program_space, it creates
87894 new symtabs for the child as well. When the child stop, but still on
87895 the fork() line, GDB thinks the "next" is done because the symtab
87896 pointers no longer match. In reality they are two symtab instances that
87897 represent the same file. But moving the spaces to the child and
87898 creating new spaces for the parent, we avoid this problem.
87900 Note that the problem described above happens today with "detach-on-fork
87901 off" and "follow-fork-mode child", because we create new spaces for the
87902 child. This will have to be addressed later.
87904 Test-wise, improve gdb.base/foll-fork.exp to set a breakpoint that is
87905 expected to have a location in each inferiors. Without the fix, when
87906 the two inferiors erroneously share a program space, GDB reports a
87909 Change-Id: Ifea76e14f87b9f7321fc3a766217061190e71c6e
87911 2021-09-27 Simon Marchi <simon.marchi@efficios.com>
87913 gdb.base/foll-fork.exp: use foreach_with_prefix to handle prefixes
87914 No behavior change in the test expected, other than in the test names.
87916 Change-Id: I111137483858ab0f23138439f2930009779a2b3d
87918 2021-09-27 Simon Marchi <simon.marchi@efficios.com>
87920 gdb.base/foll-fork.exp: rename variables
87921 Rename the variables / parameters used to match the corresponding GDB
87922 setting name, I find that easier to follow.
87924 Change-Id: Idcbddbbb369279fcf1e808b11a8c478f21b2a946
87926 2021-09-27 Simon Marchi <simon.marchi@efficios.com>
87928 gdb.base/foll-fork.exp: refactor to restart GDB between each portion of the test
87929 This test is difficult to follow and modify because the state of GDB is
87930 preserved some tests. Add a setup proc, which starts a new GDB and runs
87931 to main, and use it in all test procs. Use proc_with_prefix to avoid
87934 The check_fork_catchpoints proc also seems used to check for follow-fork
87935 support by checking if catchpoints are supported. If they are not, it
87936 uses "return -code return", which makes its caller return. I find this
87937 unnecessary complex, versus just returning a boolean. Modify it to do
87940 Change-Id: I23e62b204286c5e9c5c86d2727f7d33fb126ed08
87942 2021-09-27 Simon Marchi <simon.marchi@efficios.com>
87944 gdb.base/foll-fork.exp: remove gating based on target triplet
87945 It looks like this test has some code to check at runtime the support of
87946 fork handling of the target (see check_fork_catchpoints). So, it seems
87947 to me that the check based on target triplet at the beginning of the
87948 test is not needed. This kind of gating is generally not desirable,
87949 because we wouldn't think of updating it when adding fork support to a
87950 target. For example, FreeBSD supports fork, but it wasn't listed here.
87952 Change-Id: I6b55f2298edae6b37c3681fb8633d8ea1b5aabee
87954 2021-09-27 Simon Marchi <simon.marchi@efficios.com>
87956 gdb.base/foll-fork.exp: remove DUPLICATEs
87957 Remove DUPLICATEs, and and at the same time replace two uses of
87958 gdb_test_multiple with gdb_test. I don't think using gdb_test_multiple
87961 Change-Id: I8dcf097c3364e92d4f0e11f0c0f05dbb88e86742
87963 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
87965 libctf, lookup: fix bounds of pptrtab lookup
87966 An off-by-one bug in the check for pptrtab lookup meant that we could
87967 access the pptrtab past its bounds (*well* past its bounds),
87968 particularly if we called ctf_lookup_by_name in a child dict with "*foo"
87969 where "foo" is a type that exists in the parent but not the child and no
87970 previous lookups by name have been carried out. (Note that "*foo" is
87971 not even a valid thing to call ctf_lookup_by_name with: foo * is.
87972 Nonetheless, users sometimes do call ctf_lookup_by_name with invalid
87973 content, and it should return ECTF_NOTYPE, not crash.)
87975 ctf_pptrtab_len, as its name suggests (and as other tests of it in
87976 ctf-lookup.c confirm), is one higher than the maximum valid permissible
87977 index, so the comparison is wrong.
87979 (Test added, which should fail pretty reliably in the presence of this
87980 bug on any machine with 4KiB pages.)
87983 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
87985 * ctf-lookup.c (ctf_lookup_by_name_internal): Fix pptrtab bounds.
87986 * testsuite/libctf-writable/pptrtab-writable-page-deep-lookup.*:
87989 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
87991 libctf, testsuite: fix various warnings in tests
87992 These warnings are all off by default, but if they do fire you get
87993 spurious ERRORs when running make check-libctf.
87996 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
87998 * testsuite/libctf-lookup/enum-symbol.c: Remove unused label.
87999 * testsuite/libctf-lookup/conflicting-type-syms.c: Remove unused
88001 * testsuite/libctf-regression/pptrtab.c: Likewise.
88002 * testsuite/libctf-regression/type-add-unnamed-struct.c: Likewise.
88003 * testsuite/libctf-writable/pptrtab.c: Likewise.
88004 * testsuite/libctf-writable/reserialize-strtab-corruption.c:
88006 * testsuite/libctf-regression/nonstatic-var-section-ld-r.c: Fix
88008 * testsuite/libctf-regression/nonstatic-var-section-ld.c:
88010 * testsuite/libctf-regression/nonstatic-var-section-ld.lk: Adjust.
88011 * testsuite/libctf-writable/symtypetab-nonlinker-writeout.c: Fix
88014 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88016 libctf: fix handling of CTF symtypetab sections emitted by older GCC
88017 Older (pre-upstreaming) GCC emits a function symtypetab section of a
88018 format never read by any extant libctf. We can detect such CTF dicts by
88019 the lack of the CTF_F_NEWFUNCINFO flag in their header, and we do so
88020 when reading in the symtypetab section -- but if the set of symbols with
88021 types is sufficiently sparse, even an older GCC will emit a function
88024 In NEWFUNCINFO-capable compilers, this section will always be the exact
88025 same length as the corresponding function section (each is an array of
88026 uint32_t, associated 1:1 with each other). But this is not true for the
88027 older compiler, for which the sections are different lengths. We check
88028 to see if the function symtypetab section and its index are the same
88029 length, but we fail to skip this check when this is not a NEWFUNCINFO
88030 dict, and emit a spurious corruption error for a CTF dict we could
88031 have perfectly well opened and used.
88033 Fix trivial: check the flag (and fix the terrible grammar of the error
88034 message at the same time).
88037 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88039 * ctf-open.c (ctf_bufopen_internal): Don't complain about corrupt
88040 function index symtypetab sections if this is an old-format
88041 function symtypetab section (which should be ignored in any case).
88044 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88046 configure: regenerate in all projects that use libtool.m4
88047 (including sim/, which has no changelog.)
88050 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88052 * configure: Regenerate.
88055 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88057 * configure: Regenerate.
88060 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88062 * configure: Regenerate.
88065 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88067 * configure: Regenerate.
88070 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88072 * configure: Regenerate.
88075 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88077 * configure: Regenerate.
88078 * Makefile.in: Regenerate.
88081 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88083 * configure: Regenerate.
88086 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88088 * configure: Regenerate.
88090 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88092 libctf: try several possibilities for linker versioning flags
88093 Checking for linker versioning by just grepping ld --help output for
88094 mentions of --version-script is inadequate now that Solaris 11.4
88095 implements a --version-script with different semantics. Try linking a
88096 test program with a small wildcard-using version script with each
88097 supported set of flags in turn, to make sure that linker versioning is
88098 not only advertised but actually works.
88100 The Solaris "GNU-compatible" linker versioning is not quite
88101 GNU-compatible enough, but we can work around the differences by
88102 generating a new version script that removes the comments from the
88103 original (Solaris ld requires #-style comments), and making another
88104 version script for libctf-nonbfd in particular which doesn't mention any
88105 of the symbols that appear in libctf.la, to avoid Solaris ld introducing
88106 corresponding new NOTYPE symbols to match the version script.
88109 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88112 * configure.ac (VERSION_FLAGS): Replace with...
88113 (ac_cv_libctf_version_script): ... this multiple test.
88114 (VERSION_FLAGS_NOBFD): Substitute this too.
88115 * Makefile.am (libctf_nobfd_la_LDFLAGS): Use it. Split out...
88116 (libctf_ldflags_nover): ... non-versioning flags here.
88117 (libctf_la_LDFLAGS): Use it.
88118 * libctf.ver: Give every symbol not in libctf-nobfd a comment on
88119 the same line noting as much.
88121 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88123 libtool.m4: fix nm BSD flag detection
88124 Libtool needs to get BSD-format (or MS-format) output out of the system
88125 nm, so that it can scan generated object files for symbol names for
88126 -export-symbols-regex support. Some nms need specific flags to turn on
88127 BSD-formatted output, so libtool checks for this in its AC_PATH_NM.
88128 Unfortunately the code to do this has a pair of interlocking flaws:
88130 - it runs the test by doing an nm of /dev/null. Some platforms
88131 reasonably refuse to do an nm on a device file, but before now this
88132 has only been worked around by assuming that the error message has a
88133 specific textual form emitted by Tru64 nm, and that getting this
88134 error means this is Tru64 nm and that nm -B would work to produce
88135 BSD-format output, even though the test never actually got anything
88136 but an error message out of nm -B. This is fixable by nm'ing *nm
88137 itself* (since we necessarily have a path to it).
88139 - the test is entirely skipped if NM is set in the environment, on the
88140 grounds that the user has overridden the test: but the user cannot
88141 reasonably be expected to know that libtool wants not only nm but
88142 also flags forcing BSD-format output. Worse yet, one such "user" is
88143 the top-level Cygnus configure script, which neither tests for
88144 nor specifies any BSD-format flags. So platforms needing BSD-format
88145 flags always fail to set them when run in a Cygnus tree, breaking
88146 -export-symbols-regex on such platforms. Libtool also needs to
88147 augment $LD on some platforms, but this is done unconditionally,
88148 augmenting whatever the user specified: the nm check should do the
88151 One wrinkle: if the user has overridden $NM, a path might have been
88152 provided: so we use the user-specified path if there was one, and
88153 otherwise do the path search as usual. (If the nm specified doesn't
88154 work, this might lead to a few extra pointless path searches -- but
88155 the test is going to fail anyway, so that's not a problem.)
88157 (Tested with NM unset, and set to nm, /usr/bin/nm, my-nm where my-nm is a
88158 symlink to /usr/bin/nm on the PATH, and /not-on-the-path/my-nm where
88159 *that* is a symlink to /usr/bin/nm.)
88162 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88165 * libtool.m4 (LT_PATH_NM): Try BSDization flags with a user-provided
88166 NM, if there is one. Run nm on itself, not on /dev/null, to avoid
88167 errors from nms that refuse to work on non-regular files. Remove
88168 other workarounds for this problem. Strip out blank lines from the
88171 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88173 libtool.m4: augment symcode for Solaris 11
88174 This reports common symbols like GNU nm, via a type code of 'C'.
88177 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88180 * libtool.m4 (lt_cv_sys_global_symbol_pipe): Augment symcode for
88183 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88185 libctf: link against libiberty before linking in libbfd or libctf-nobfd
88186 This ensures that the CTF_LIBADD, which always contains at least this
88187 when doing a shared link:
88189 -L`pwd`/../libiberty/pic -liberty
88191 appears in the link line before any requirements pulled in by libbfd.la,
88192 which include -liberty but because it is install-time do not include the
88193 -L`pwd`/../libiberty/pic portion (in an indirect dep like this, the path
88194 comes from the libbfd.la file, and is an install path). libiberty also
88195 appears after libbfd in the link line by virtue of libctf-nobfd.la,
88196 because libctf-nobfd has to follow libbfd in the link line, and that
88197 needs symbols from libiberty too.
88199 Without this, an installed liberty might well be pulled in by libbfd,
88200 and if --enable-install-libiberty is not specified this libiberty might
88201 be completely incompatible with what is being installed and break either
88202 or boht of libbfd and libctf. (The specific problem observed here is
88203 that bsearch_r was not present, but other problems might easily be
88204 observed in future too.)
88206 Because ld links against libctf, this has a tendency to break the system
88207 linker at install time too, if installing with --prefix=/usr. That's
88208 quite unpleasant to recover from.
88211 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
88214 * Makefile.am (libctf_la_LIBADD): Link against libiberty
88215 before pulling in libbfd.la or pulling in libctf-nobfd.la.
88216 * Makefile.in: Regenerate.
88218 2021-09-27 Tom de Vries <tdevries@suse.de>
88220 [gdb/build] Fix build with g++-4.8
88221 When building g++-4.8, we run into:
88223 src/gdb/dwarf2/read.c:919:5: error: multiple fields in union \
88224 'partial_die_info::<anonymous union>' initialized
88233 CORE_ADDR lowpc = 0;
88234 CORE_ADDR highpc = 0;
88236 ULONGEST ranges_offset;
88240 The error looks incorrect, given that only one union member is initialized,
88241 and does not reproduce with newer g++.
88243 Nevertheless, work around this by moving the initialization to a constructor.
88245 [ I considered just removing the initialization, with the idea that access
88246 should be guarded by has_pc_info, but I ran into one failure in the testsuite,
88247 for gdb.base/check-psymtab.exp due to add_partial_symbol using lowpc without
88248 checking has_pc_info. ]
88250 Tested on x86_64-linux.
88252 2021-09-27 Andrew Burgess <andrew.burgess@embecosm.com>
88254 gdb: add setting to disable reading source code files
88255 In some situations it is possible that a user might not want GDB to
88256 try and access source code files, for example, the source code might
88257 be stored on a slow to access network file system.
88259 It is almost certainly possible that using some combination of 'set
88260 directories' and/or 'set substitute-path' a user can trick GDB into
88261 being unable to find the source files, but this feels like a rather
88262 crude way to solve the problem.
88264 In this commit a new option is add that stops GDB from opening and
88265 reading the source files. A user can run with source code reading
88266 disabled if this is required, then re-enable later if they decide
88267 that they now want to view the source code.
88269 2021-09-27 Andrew Burgess <andrew.burgess@embecosm.com>
88271 gdb: remove duplicate cmd_list_element declarations
88272 For some reason we have two locations where cmd_list_elements are
88273 declared, cli/cli-cmds.h and gdbcmd.h. Worse still there is
88274 duplication between these two locations.
88276 In this commit I have moved all of the cmd_list_element declarations
88277 from gdbcmd.h into cli/cli-cmds.h and removed the duplicates.
88279 There should be no user visible changes after this commit.
88281 2021-09-27 Andrew Burgess <andrew.burgess@embecosm.com>
88283 gdb: prevent an assertion when computing the frame_id for an inline frame
88284 I ran into this assertion while GDB was trying to unwind the stack:
88286 gdb/inline-frame.c:173: internal-error: void inline_frame_this_id(frame_info*, void**, frame_id*): Assertion `frame_id_p (*this_id)' failed.
88288 That is, when building the frame_id for an inline frame, GDB asks for
88289 the frame_id of the previous frame. Unfortunately, no valid frame_id
88290 was returned for the previous frame, and so the assertion triggers.
88292 What is happening is this, I had a stack that looked something like
88293 this (the arrows '->' point from caller to callee):
88295 normal_frame -> inline_frame
88297 However, for whatever reason (e.g. broken debug information, or
88298 corrupted stack contents in the inferior), when GDB tries to unwind
88299 "normal_frame", it ends up getting back effectively the same frame,
88300 thus the call stack looks like this to GDB:
88302 .-> normal_frame -> inline_frame
88306 Given such a situation we would expect GDB to terminate the stack with
88307 an error like this:
88309 Backtrace stopped: previous frame identical to this frame (corrupt stack?)
88311 However, the inline_frame causes a problem, and here's why:
88313 When unwinding we start from the sentinel frame and call
88314 get_prev_frame. We eventually end up in get_prev_frame_if_no_cycle,
88315 in here we create a raw frame, and as this is frame #0 we immediately
88318 However, eventually we will try to unwind the stack further. When we
88319 do this we inevitably needing to know the frame_id for frame #0, and
88320 so, eventually, we end up in compute_frame_id.
88322 In compute_frame_id we first find the right unwinder for this frame,
88323 in our case (i.e. for inline_frame) the $pc is within the function
88324 normal_frame, but also within a block associated with the inlined
88325 function inline_frame, as such the inline frame unwinder claims this
88328 Back in compute_frame_id we next compute the frame_id, for our
88329 inline_frame this means a call to inline_frame_this_id.
88331 The ID of an inline frame is based on the id of the previous frame, so
88332 from inline_frame_this_id we call get_prev_frame_always, this
88333 eventually calls get_prev_frame_if_no_cycle again, which creates
88334 another raw frame and calls compute_frame_id (for frames other than
88335 frame 0 we immediately compute the frame_id).
88337 In compute_frame_id we again identify the correct unwinder for this
88338 frame. Our $pc is unchanged, however, the fact that the next frame is
88339 of type INLINE_FRAME prevents the inline frame unwinder from claiming
88340 this frame again, and so, the standard DWARF frame unwinder claims
88343 We return to compute_frame_id and call the standard DWARF function to
88344 build the frame_id for normal_frame.
88346 With the frame_id of normal_frame figured out we return to
88347 compute_frame_id, and then to get_prev_frame_if_no_cycle, where we add
88348 the ID for normal_frame into the frame_id cache, and return the frame
88349 back to inline_frame_this_id.
88351 From inline_frame_this_id we build a frame_id for inline_frame and
88352 return to compute_frame_id, and then to get_prev_frame_if_no_cycle,
88353 which adds the frame_id for inline_frame into the frame_id cache.
88357 However, as we are trying to unwind the complete stack, we eventually
88358 ask for the previous frame of normal_frame, remember, at this point
88359 GDB doesn't know the stack is corrupted (with a cycle), GDB still
88360 needs to figure that out.
88362 So, we eventually end up in get_prev_frame_if_no_cycle where we create
88363 a raw frame and call compute_frame_id, remember, this is for the frame
88364 before normal_frame.
88366 The first task for compute_frame_id is to find the unwinder for this
88367 frame, so all of the frame sniffers are tried in order, this includes
88368 the inline frame sniffer.
88370 The inline frame sniffer asks for the $pc, this request is sent up the
88371 stack to normal_frame, which, due to its cyclic behaviour, tells GDB
88372 that the $pc in the previous frame was the same as the $pc in
88375 GDB spots that this $pc corresponds to both the function normal_frame
88376 and also the inline function inline_frame. As the next frame is not
88377 an INLINE_FRAME then GDB figures that we have not yet built a frame to
88378 cover inline_frame, and so the inline sniffer claims this new frame.
88379 Our stack is now looking like this:
88381 inline_frame -> normal_frame -> inline_frame
88383 But, we have not yet computed the frame id for the outer most (on the
88384 left) inline_frame. After the frame sniffer has claimed the inline
88385 frame GDB returns to compute_frame_id and calls inline_frame_this_id.
88387 In here GDB calls get_prev_frame_always, which eventually ends up
88388 in get_prev_frame_if_no_cycle again, where we create a raw frame and
88389 call compute_frame_id.
88391 Just like before, compute_frame_id tries to find an unwinder for this
88392 new frame, it sees that the $pc is within both normal_frame and
88393 inline_frame, but the next frame is, again, an INLINE_FRAME, so, just
88394 like before the standard DWARF unwinder claims this frame. Back in
88395 compute_frame_id we again call the standard DWARF function to build
88396 the frame_id for this new copy of normal_frame.
88398 At this point the stack looks like this:
88400 normal_frame -> inline_frame -> normal_frame -> inline_frame
88402 After compute_frame_id we return to get_prev_frame_if_no_cycle, where
88403 we try to add the frame_id for the new normal_frame into the frame_id
88404 cache, however, unlike before, we fail to add this frame_id as it is
88405 a duplicate of the previous normal_frame frame_id. Having found a
88406 duplicate get_prev_frame_if_no_cycle unlinks the new frame from the
88407 stack, and returns nullptr, the stack now looks like this:
88409 inline_frame -> normal_frame -> inline_frame
88411 The nullptr result from get_prev_frame_if_no_cycle is fed back to
88412 inline_frame_this_id, which forwards this to get_frame_id, which
88413 immediately returns null_frame_id. As null_frame_id is not considered
88414 a valid frame_id, this is what triggers the assertion.
88418 - inline_frame_this_id currently assumes that as the inline frame
88419 exists, we will always get a valid frame back from
88420 get_prev_frame_always,
88422 - get_prev_frame_if_no_cycle currently assumes that it is safe to
88423 return nullptr when it sees a cycle.
88425 Notice that in frame.c:compute_frame_id, this code:
88427 fi->this_id.value = outer_frame_id;
88428 fi->unwind->this_id (fi, &fi->prologue_cache, &fi->this_id.value);
88429 gdb_assert (frame_id_p (fi->this_id.value));
88431 The assertion makes it clear that the this_id function must always
88432 return a valid frame_id (e.g. null_frame_id is not a valid return
88433 value), and similarly in inline_frame.c:inline_frame_this_id this
88436 *this_id = get_frame_id (get_prev_frame_always (this_frame));
88438 gdb_assert (frame_id_p (*this_id));
88440 Makes it clear that every inline frame expects to be able to get a
88441 previous frame, which will have a valid frame_id.
88443 As I have discussed above, these assumptions don't currently hold in
88446 One possibility would be to move the call to get_prev_frame_always
88447 forward from inline_frame_this_id to inline_frame_sniffer, however,
88448 this falls foul of (in frame.c:frame_cleanup_after_sniffer) this
88451 /* No sniffer should extend the frame chain; sniff based on what is
88452 already certain. */
88453 gdb_assert (!frame->prev_p);
88455 This assert prohibits any sniffer from trying to get the previous
88456 frame, as getting the previous frame is likely to depend on the next
88457 frame, I can understand why this assertion is a good thing, and I'm in
88458 no rush to alter this rule.
88460 The solution proposed here takes onboard feedback from both Pedro, and
88461 Simon (see the links below). The get_prev_frame_if_no_cycle function
88462 is renamed to get_prev_frame_maybe_check_cycle, and will now not do
88463 cycle detection for inline frames, even when we spot a duplicate frame
88464 it is still returned. This is fine, as, if the normal frame has a
88465 duplicate frame-id then the inline frame will also have a duplicate
88466 frame-id. And so, when we reject the inline frame, the duplicate
88467 normal frame, which is previous to the inline frame, will also be
88470 In inline-frame.c the call to get_prev_frame_always is no longer
88471 nested inside the call to get_frame_id. There are reasons why
88472 get_prev_frame_always can return nullptr, for example, if there is a
88473 memory error while trying to get the previous frame, if this should
88474 happen then we now give a more informative error message.
88478 Patch v2: https://sourceware.org/pipermail/gdb-patches/2021-June/180208.html
88479 Feedback: https://sourceware.org/pipermail/gdb-patches/2021-July/180651.html
88480 https://sourceware.org/pipermail/gdb-patches/2021-July/180663.html
88482 Patch v3: https://sourceware.org/pipermail/gdb-patches/2021-July/181029.html
88483 Feedback: https://sourceware.org/pipermail/gdb-patches/2021-July/181035.html
88485 Additional input: https://sourceware.org/pipermail/gdb-patches/2021-September/182040.html
88487 2021-09-27 Tom de Vries <tdevries@suse.de>
88489 [gdb/testsuite] Fix gdb.base/dcache-flush.exp
88490 When running test-case gdb.base/dcache-flush.exp on ubuntu 18.04.5, I run into:
88492 (gdb) PASS: gdb.base/dcache-flush.exp: p var2
88494 Dcache 4096 lines of 64 bytes each.^M
88495 Contains data for Thread 0x7ffff7fc6b80 (LWP 3551)^M
88496 Line 0: address 0x7fffffffd4c0 [47 hits]^M
88497 Line 1: address 0x7fffffffd500 [31 hits]^M
88498 Line 2: address 0x7fffffffd5c0 [7 hits]^M
88499 Cache state: 3 active lines, 85 hits^M
88500 (gdb) FAIL: gdb.base/dcache-flush.exp: check dcache before flushing
88502 The regexp expects "Contains data for process $decimal".
88504 This is another case of thread_db_target::pid_to_str being used.
88506 Fix this by updating the regexp.
88508 Tested on x86_64-linux.
88510 2021-09-27 Tom de Vries <tdevries@suse.de>
88512 [gdb/testsuite] Test sw watchpoint in gdb.threads/process-dies-while-detaching.exp
88513 The test-case gdb.threads/process-dies-while-detaching.exp takes about 20s
88514 when using hw watchpoints, but when forcing sw watchpoints (using the patch
88515 mentioned in PR28375#c0), the test-case takes instead 3m14s.
88517 Also, it show a FAIL:
88521 Cannot find user-level thread for LWP 10324: generic error^M
88522 (gdb) FAIL: gdb.threads/process-dies-while-detaching.exp: single-process:
88523 continue: watchpoint: continue
88525 for which PR28375 was filed.
88527 Modify the test-case to:
88528 - add the hw/sw axis to the watchpoint testing, to ensure that we
88529 observe the sw watchpoint behaviour also on can-use-hw-watchpoints
88531 - skip the hw breakpoint testing if not supported
88532 - set the sw watchpoint later to avoid making the test
88533 too slow. This still triggers the same PR, but now takes just 24s.
88535 This patch adds a KFAIL for PR28375.
88537 Tested on x86_64-linux.
88539 2021-09-27 Simon Marchi <simon.marchi@polymtl.ca>
88541 gdb: fix indentation in gdbtypes.c
88542 Change-Id: I7bfbb9d349a1f474256800c45e28fe3b1de08771
88544 2021-09-27 GDB Administrator <gdbadmin@sourceware.org>
88546 Automatic date update in version.in
88548 2021-09-26 GDB Administrator <gdbadmin@sourceware.org>
88550 Automatic date update in version.in
88552 2021-09-26 Peter Bergner <bergner@linux.ibm.com>
88554 PowerPC: Enable mfppr mfppr32, mtppr and mtppr32 extended mnemonics on POWER5
88555 SPR 896 and the mfppr mfppr32, mtppr and mtppr32 extended mnemonics were added
88556 in ISA 2.03, so enable them on POWER5 and later.
88559 * ppc-opc.c (powerpc_opcodes) <mfppr, mfppr32, mtppr, mtppr32>: Enable
88560 on POWER5 and later.
88563 * testsuite/gas/ppc/power5.s: New test.
88564 * testsuite/gas/ppc/power5.d: Likewise.
88565 * testsuite/gas/ppc/ppc.exp: Run it.
88566 * testsuite/gas/ppc/power7.s: Remove tests for mfppr, mfppr32, mtppr
88568 * testsuite/gas/ppc/power7.d: Likewise.
88570 2021-09-25 Tom de Vries <tdevries@suse.de>
88572 [gdb/testsuite] Minimize gdb restarts
88573 Minimize gdb restarts, applying the following rules:
88574 - don't use prepare_for_testing unless necessary
88575 - don't use clean_restart unless necessary
88577 Also, if possible, replace build_for_executable + clean_restart
88578 with prepare_for_testing for brevity.
88580 Touches 68 test-cases.
88582 Tested on x86_64-linux.
88584 2021-09-25 Alan Modra <amodra@gmail.com>
88586 PR28346, segfault attempting to disassemble raw binary
88587 Don't attempt to access elf_section_data for non-ELF sections.
88590 * elf32-xtensa.c (xtensa_read_table_entries): Return zero entries
88593 2021-09-25 GDB Administrator <gdbadmin@sourceware.org>
88595 Automatic date update in version.in
88597 2021-09-24 Hans-Peter Nilsson <hp@axis.com>
88599 gas/testsuite/ld-elf/dwarf2-21.d: Pass -W
88600 Required for the expected "CU:" to be emitted for long
88601 source-paths. See binutils/dwarf.c:
88603 if (do_wide || strlen (directory) < 76)
88604 printf (_("CU: %s/%s:\n"), directory, file_table[0].name);
88606 printf ("%s:\n", file_table[0].name);
88608 See also commit 5f410aa50ce2c, "testsuite/ld-elf/pr26936.d:
88612 * testsuite/ld-elf/dwarf2-21.d: Pass -W.
88614 2021-09-24 Simon Marchi <simon.marchi@polymtl.ca>
88616 gdb: change thread_info::name to unique_xmalloc_ptr, add helper function
88617 This started out as changing thread_info::name to a unique_xmalloc_ptr.
88618 That showed that almost all users of that field had the same logic to
88619 get a thread's name: use thread_info::name if non-nullptr, else ask the
88620 target. Factor out this logic in a new thread_name free function. Make
88621 the field private (rename to m_name) and add some accessors.
88623 Change-Id: Iebdd95f4cd21fbefc505249bd1d05befc466a2fc
88625 2021-09-24 Tom Tromey <tom@tromey.com>
88627 Move value_true to value.h
88628 I noticed that value_true is declared in language.h and defined in
88629 language.c. However, as part of the value API, I think it would be
88630 better in one of those files. And, because it is very short, I
88631 changed it to be an inline function in value.h. I've also removed a
88632 comment from the implementation, on the basis that it seems obsolete
88633 -- if the change it suggests was needed, it probably would have been
88634 done by now; and if it is needed in the future, odds are it would be
88635 done differently anyway.
88637 Finally, this patch also changes value_true and value_logical_not to
88638 return a bool, and updates some uses.
88640 2021-09-24 Pedro Alves <pedro@palves.net>
88642 Make dcache multi-target-safe
88643 By inspection, I noticed that this code in dcache.c is not
88644 multi-target-aware:
88646 /* If this is a different inferior from what we've recorded,
88647 flush the cache. */
88649 if (inferior_ptid != dcache->ptid)
88651 This doesn't take into account that threads of different targets may
88652 have the same ptid.
88654 Fixed by also storing/comparing the process_stratum_target.
88656 Tested on x86-64-linux-gnu, native and gdbserver.
88658 Change-Id: I4d9d74052c696b72d28cb1c77b697b911725c8d3
88660 2021-09-24 Pedro Alves <pedro@palves.net>
88662 Fix 'FAIL: gdb.perf/disassemble.exp: python Disassemble().run()'
88663 We currently have one FAIL while running "make check-perf":
88665 PerfTest::assemble, run ...
88666 python Disassemble().run()
88667 Traceback (most recent call last):
88668 File "<string>", line 1, in <module>
88669 File "/home/pedro/rocm/gdb/src/gdb/testsuite/gdb.perf/lib/perftest/perftest.py", line 64, in run
88671 File "<string>", line 25, in warm_up
88672 gdb.error: No symbol "ada_evaluate_subexp" in current context.
88673 Error while executing Python code.
88674 (gdb) FAIL: gdb.perf/disassemble.exp: python Disassemble().run()
88677 The gdb.perf/disassemble.exp testcase debugs GDB with itself, runs to
88678 main, and then disassembles a few GDB functions. The problem is that
88679 most(!) functions it is trying to disassemble are now gone...
88681 This commit fixes the issue by simply picking some other functions to
88684 It would perhaps be better to come up with some test program to
88685 disassemble, one that would stay the same throughout the years,
88686 instead of disassembling GDB itself. I don't know why that wasn't
88687 done to begin with. I'll have to leave that for another rainy day,
88691 yyyy-mm-dd Pedro Alves <pedro@palves.net>
88693 * gdb.perf/disassemble.py (Disassemble::warm_up): Disassemble
88694 evaluate_subexp_do_call instead of ada_evaluate_subexp.
88695 (Disassemble::warm_up): Disassemble "captured_main",
88696 "run_inferior_call" and "update_global_location_list" instead of
88697 "evaluate_subexp_standard" and "c_parse_internal".
88699 Change-Id: I89d1cca89ce2e495dea5096e439685739cc0d3df
88701 2021-09-24 Pedro Alves <pedro@palves.net>
88703 Fix all PATH problems in testsuite/gdb.perf/
88704 Currently "make check-perf" triggers ~40 PATH messages in gdb.sum:
88707 PATH: gdb.perf/backtrace.exp: python sys.path.insert(0, os.path.abspath("/home/pedro/rocm/gdb/build/gdb/../../src/gdb/testsuite/gdb.perf/lib"))
88708 PATH: gdb.perf/backtrace.exp: python exec (open ('/home/pedro/rocm/gdb/build/gdb/testsuite/outputs/gdb.perf/backtrace/backtrace.py').read ())
88711 This commit fixes them. E.g. before/after gdb.sum diff:
88713 -PASS: gdb.perf/backtrace.exp: python import os, sys
88714 -PASS: gdb.perf/backtrace.exp: python sys.path.insert(0, os.path.abspath("/home/pedro/rocm/gdb/build-master/gdb/../../src/gdb/testsuite/gdb.perf/lib"))
88715 -PATH: gdb.perf/backtrace.exp: python sys.path.insert(0, os.path.abspath("/home/pedro/rocm/gdb/build-master/gdb/../../src/gdb/testsuite/gdb.perf/lib"))
88716 -PASS: gdb.perf/backtrace.exp: python exec (open ('/home/pedro/rocm/gdb/build-master/gdb/testsuite/outputs/gdb.perf/backtrace/backtrace.py').read ())
88717 -PATH: gdb.perf/backtrace.exp: python exec (open ('/home/pedro/rocm/gdb/build-master/gdb/testsuite/outputs/gdb.perf/backtrace/backtrace.py').read ())
88718 +PASS: gdb.perf/backtrace.exp: setup perftest: python import os, sys
88719 +PASS: gdb.perf/backtrace.exp: setup perftest: python sys.path.insert(0, os.path.abspath("${srcdir}/gdb.perf/lib"))
88720 +PASS: gdb.perf/backtrace.exp: setup perftest: python exec (open ('${srcdir}/gdb.perf/backtrace.py').read ())
88723 yyyy-mm-dd Pedro Alves <pedro@palves.net>
88725 * lib/perftest.exp (PerfTest::_setup_perftest): Use
88726 with_test_prefix. Add explicit test names to python invocations,
88727 with "$srcdir" not expanded.
88729 Change-Id: I50a31b04b7abdea754139509e4a34ae9263118a4
88731 2021-09-24 Pedro Alves <pedro@palves.net>
88733 Fix all DUPLICATE problems in testsuite/gdb.perf/
88734 Currently running "make check-perf" shows:
88737 # of duplicate test names 6008
88740 All those duplicate test names come from gdb.perf/skip-command.exp.
88741 This commit fixes them, using with_test_prefix.
88744 yyyy-mm-dd Pedro Alves <pedro@palves.net>
88746 * gdb.perf/skip-command.exp (run_skip_bench): Wrap each for
88747 iteration in with_test_prefix.
88749 Change-Id: I38501cf70bc6b60306ee7228996ee7bcd858dc1b
88751 2021-09-24 Tom Tromey <tromey@adacore.com>
88753 Fix handling of DW_AT_data_bit_offset
88754 A newer version of GCC will now emit member locations using just
88755 DW_AT_data_bit_offset, like:
88757 <3><14fe>: Abbrev Number: 1 (DW_TAG_member)
88758 <14ff> DW_AT_name : (indirect string, offset: 0x215e): nb_bytes
88759 <1503> DW_AT_decl_file : 1
88760 <1503> DW_AT_decl_line : 10
88761 <1504> DW_AT_decl_column : 7
88762 <1505> DW_AT_type : <0x150b>
88763 <1509> DW_AT_bit_size : 31
88764 <150a> DW_AT_data_bit_offset: 64
88766 whereas earlier versions would emit something like:
88768 <3><164f>: Abbrev Number: 7 (DW_TAG_member)
88769 <1650> DW_AT_name : (indirect string, offset: 0x218d): nb_bytes
88770 <1654> DW_AT_decl_file : 1
88771 <1655> DW_AT_decl_line : 10
88772 <1656> DW_AT_decl_column : 7
88773 <1657> DW_AT_type : <0x165f>
88774 <165b> DW_AT_byte_size : 4
88775 <165c> DW_AT_bit_size : 31
88776 <165d> DW_AT_bit_offset : 1
88777 <165e> DW_AT_data_member_location: 8
88779 That is, DW_AT_data_member_location is not emitted any more. This is
88780 a change due to the switch to DWARF 5 by default.
88782 This change pointed out an existing bug in gdb, namely that the
88783 attr_to_dynamic_prop depends on the presence of
88784 DW_AT_data_member_location. This patch moves the handling of
88785 DW_AT_data_bit_offset into handle_data_member_location, and updates
88786 attr_to_dynamic_prop to handle this new case.
88788 A new test case is included. This test fails with GCC 11, but passes
88789 with an earlier version of GCC.
88791 2021-09-24 Tom de Vries <tdevries@suse.de>
88793 [gdb/testsuite] Don't leave gdb instance running after function_range
88794 A typical dwarf assembly test-case start like this:
88796 standard_testfile .c -debug.S
88798 set asm_file [standard_output_file $srcfile2]
88799 Dwarf::assemble $asm_file {
88803 if { [prepare_for_testing "failed to prepare" ${testfile} \
88804 [list $srcfile $asm_file] {nodebug}] } {
88809 When accidentally using build_for_executable instead of
88810 prepare_for_testing (or intentionally using it but forgetting to add
88811 clean_restart $binfile or some such) the mistake may not be caught, because
88812 another gdb instance is still running, and we may silently end up testing
88813 compiler-generated DWARF.
88815 This can be caused by something relatively obvious, like an earlier
88816 prepare_for_testing or clean_restart, but also by something more obscure like
88817 function_range, which may even be triggered by dwarf assembly like this:
88819 {MACRO_AT_func {main}}
88822 Fix this by calling gdb_exit at the end of function_range.
88824 Also fix the fallout of that in test-case gdb.dwarf2/dw2-bad-elf.exp, where a
88825 get_sizeof call used the gdb instance left lingering by function_range.
88827 [ A better and more complete fix would add a new proc get_exec_info, that would
88828 be called at the start of the dwarf assembly body:
88830 Dwarf::assemble $asm_file {
88831 get_exec_info {main foo} {int void*}
88834 - do a prepare_for_testing with $srcfile (roughtly equivalent to what
88835 MACRO_AT_func does,
88836 - call function_range for all functions main and foo, without starting a
88838 - set corresponding variables at the call-site: main_start, main_len,
88839 main_end, foo_start, foo_len, foo_end.
88840 - get size for types int and void*
88841 - set corresponding variables at the call-site: int_size, void_ptr_size.
88844 Tested on x86_64-linux.
88846 2021-09-24 Tom de Vries <tdevries@suse.de>
88848 [gdb/testsuite] Use pie instead of -fpie/-pie
88849 I noticed two test-cases where -fpie is used. Using the canonical pie option
88850 will usually get one -fPIE instead.
88852 That choice is justified here in gdb_compile:
88854 # For safety, use fPIE rather than fpie. On AArch64, m68k, PowerPC
88855 # and SPARC, fpie can cause compile errors due to the GOT exceeding
88856 # a maximum size. On other architectures the two flags are
88857 # identical (see the GCC manual). Note Debian9 and Ubuntu16.10
88858 # onwards default GCC to using fPIE. If you do require fpie, then
88859 # it can be set using the pie_flag.
88860 set flag "additional_flags=-fPIE"
88863 There is no indication that using -fpie rather than -fPIE is on purpose, so
88866 Tested on x86_64-linux.
88868 2021-09-24 Tom de Vries <tdevries@suse.de>
88870 [gdb/testsuite] Factor out dump_info in gdb.testsuite/dump-system-info.exp
88871 Factor out new proc dump_info in test-case gdb.testsuite/dump-system-info.exp,
88872 and in the process:
88874 - remove unnecessary "test -r /proc/cpuinfo"
88876 Tested on x86_64-linux.
88878 Co-Authored-By: Pedro Alves <pedro@palves.net>
88880 2021-09-24 Pedro Alves <pedro@palves.net>
88882 gdb/testsuite: Make it possible to use TCL variables in DWARF assembler loclists
88883 It is currently not possible to use variables in locations lists. For
88886 diff --git a/gdb/testsuite/gdb.dwarf2/loclists-multiple-cus.exp b/gdb/testsuite/gdb.dwarf2/loclists-multiple-cus.exp
88887 index 6b4f5c8cbb8..cdbf948619f 100644
88888 --- a/gdb/testsuite/gdb.dwarf2/loclists-multiple-cus.exp
88889 +++ b/gdb/testsuite/gdb.dwarf2/loclists-multiple-cus.exp
88890 @@ -30,6 +30,8 @@ if {![dwarf2_support]} {
88894 +set myconst 0x123456
88896 # Test with 32-bit and 64-bit DWARF.
88897 foreach_with_prefix is_64 {false true} {
88899 @@ -49,6 +51,7 @@ foreach_with_prefix is_64 {false true} {
88900 global func1_addr func1_len
88901 global func2_addr func2_len
88905 # The CU uses the DW_FORM_loclistx form to refer to the .debug_loclists
88907 @@ -107,7 +110,7 @@ foreach_with_prefix is_64 {false true} {
88910 start_length $func1_addr $func1_len {
88911 - DW_OP_constu 0x123456
88912 + DW_OP_constu $myconst
88918 $ make check TESTS="*/loclists-multiple-cus.exp"
88920 gdb compile failed, build/gdb/testsuite/outputs/gdb.dwarf2/loclists-multiple-cus/loclists-multiple-cus-dw32.S: Assembler messages:
88921 build/gdb/testsuite/outputs/gdb.dwarf2/loclists-multiple-cus/loclists-multiple-cus-dw32.S:78: Error: leb128 operand is an undefined symbol: $myconst
88924 That means $myconst was copied literally to the generated assembly
88927 This patch fixes it, by running subst on the location list body, in
88928 the context of the caller. The fix is applied to both
88929 Dwarf::loclists::table::list_::start_length and
88930 Dwarf::loclists::table::list_::start_end.
88932 Reported-by: Zoran Zaric <Zoran.Zaric@amd.com>
88934 Change-Id: I615a64431857242d9f477d5699e3732df1b31322
88936 2021-09-24 Tom de Vries <tdevries@suse.de>
88938 [gdb/testsuite] Fix DUPLICATEs in gdb.dwarf2/implptr-64bit.exp
88939 When running test-case gdb.dwarf2/implptr-64bit.exp with target board
88940 unix/-m32, I noticed:
88942 DUPLICATE: gdb.dwarf2/implptr-64bit.exp: failed to prepare
88945 Fix this by using with_test_prefix.
88947 Tested on x86_64-linux.
88949 2021-09-24 Tom de Vries <tdevries@suse.de>
88951 [gdb/testsuite] Fix DUPLICATEs gdb.dwarf2/dw2-is-stmt.exp
88952 Fix these DUPLICATEs by using with_test_prefix:
88954 DUPLICATE: gdb.dwarf2/dw2-is-stmt.exp: ensure we saw a valid line pattern, 1
88955 DUPLICATE: gdb.dwarf2/dw2-is-stmt.exp: ensure we saw a valid line pattern, 2
88958 Tested on x86_64-linux.
88960 2021-09-24 Tom de Vries <tdevries@suse.de>
88962 [gdb/testsuite] Fix set $var val in gdb.dwarf2/dw2-is-stmt.exp
88963 When doing a testrun with:
88965 $ make check RUNTESTFLAGS=$(cd $src/gdb/testsuite/; echo gdb.dwarf2/*.exp)
88969 ERROR: tcl error sourcing gdb.dwarf2/dw2-is-stmt.exp.
88970 ERROR: expected integer but got "dw2-abs-hi-pc-world.c"
88975 The variable i is set in gdb.dwarf2/dw2-abs-hi-pc.exp, and leaks to
88976 gdb.dwarf2/dw2-is-stmt.exp. It's not removed by gdb_cleanup_globals because i
88977 is set as global variable by runtest.exp, which does:
88979 for { set i 0 } { $i < $argc } { incr i } {
88981 at toplevel but forgets to unset the variable.
88983 Fix this by removing '$' in front of the variable name when doing set:
88989 Tested on x86_64-linux.
88991 2021-09-24 Tom de Vries <tdevries@suse.de>
88993 [gdb/testsuite] Fix DUPLICATE in gdb.base/load-command.exp
88994 Fix this duplicate:
88996 DUPLICATE: gdb.base/load-command.exp: check initial value of the_variable
88998 by using with_test_prefix.
89000 Tested on x86_64-linux.
89002 2021-09-24 Tom de Vries <tdevries@suse.de>
89004 [gdb/testsuite] Use pie/nopie instead of ldflags=-pie/-no-pie
89005 I noticed two test-case that use ldflags=-pie and ldflags-no-pie, instead of
89006 the canonical pie and nopie options, which would typically also add
89007 additional_flags=-fPIE respectively additional_flags=-fno-pie.
89009 There is no indication that this is on purpose, so replace these with pie and
89012 Tested on x86_64-linux.
89014 2021-09-24 Tom de Vries <tdevries@suse.de>
89016 [gdb/testsuite] Add gdb.testsuite/dump-system-info.exp
89017 When interpreting the testsuite results, it's often relevant what kind of
89018 machine the testsuite ran on. On a local machine one can just do
89019 /proc/cpuinfo, but in case of running tests using a remote system
89020 that distributes test runs to other remote systems that are not directly
89021 accessible, that's not possible.
89023 Fix this by dumping /proc/cpuinfo into the gdb.log, as well as lsb_release -a
89026 We could do this at the start of each test run, by putting it into unix.exp
89027 or some such. However, this might be too verbose, so we choose to put it into
89028 its own test-case, such that it get triggered in a full testrun, but not when
89029 running one or a subset of tests.
89031 We put the test-case into the gdb.testsuite directory, which is currently the
89032 only place in the testsuite where we do not test gdb. [ Though perhaps this
89033 could be put into a new gdb.info directory, since the test-case doesn't
89034 actually test the testsuite. ]
89036 Tested on x86_64-linux.
89038 2021-09-24 GDB Administrator <gdbadmin@sourceware.org>
89040 Automatic date update in version.in
89042 2021-09-23 Tom Tromey <tom@tromey.com>
89044 Change pointer_type to a method of struct type
89045 I noticed that pointer_type is declared in language.h and defined in
89046 language.c. However, it really has to do with types, so it should
89047 have been in gdbtypes.h all along.
89049 This patch changes it to be a method on struct type. And, I went
89050 through uses of TYPE_IS_REFERENCE and updated many spots to use the
89051 new method as well. (I didn't update ones that were in arch-specific
89052 code, as I couldn't readily test that.)
89054 2021-09-23 Tom de Vries <tdevries@suse.de>
89056 [gdb/testsuite] Support -fPIE/-fno-PIE/-pie/-no-pie in gdb_compile_rust
89057 When running gdb.rust/*.exp test-cases with target board unix/-fPIE/-pie, I
89060 builtin_spawn -ignore SIGHUP rustc --color never gdb.rust/watch.rs \
89061 -g -lm -fPIE -pie -o outputs/gdb.rust/watch/watch^M
89062 error: Unrecognized option: 'f'^M
89064 compiler exited with status 1
89067 The problem is that -fPIE and -fpie are gcc options, but for rust we use
89068 rustc, which has different compilation options.
89070 Fix this by translating the gcc options to rustc options in gdb_compile_rust,
89071 similar to how that is done for ada in target_compile_ada_from_dir.
89073 Likewise for unix/-fno-PIE/-no-pie.
89075 Tested on x86_64-linux, with:
89078 - unix/-fno-PIE/-no-pie
89079 specifically, on openSUSE Leap 15.2 both with package gcc-PIE:
89080 - installed (making gcc default to PIE)
89081 - uninstalled (making gcc default to non-PIE).
89084 2021-09-23 Tom de Vries <tdevries@suse.de>
89086 [gdb/testsuite] Use pie instead of -fPIE -pie
89087 Replace {additional_flags=-fPIE ldflags=-pie} with {pie}.
89089 This makes sure that the test-cases properly error out when using target board
89090 unix/-fno-PIE/-no-pie.
89092 Tested on x86_64-linux.
89094 2021-09-23 Tom de Vries <tdevries@suse.de>
89096 [gdb/testsuite] Fix probe test in gdb.base/break-interp.exp
89097 When running test-case gdb.base/break-interp.exp on ubuntu 18.04.5, we have:
89100 #0 0x00007eff7ad5ae12 in ?? () from break-interp-LDprelinkNOdebugNO^M
89101 #1 0x00007eff7ad71f50 in ?? () from break-interp-LDprelinkNOdebugNO^M
89102 #2 0x00007eff7ad59128 in ?? () from break-interp-LDprelinkNOdebugNO^M
89103 #3 0x00007eff7ad58098 in ?? () from break-interp-LDprelinkNOdebugNO^M
89104 #4 0x0000000000000002 in ?? ()^M
89105 #5 0x00007fff505d7a32 in ?? ()^M
89106 #6 0x00007fff505d7a94 in ?? ()^M
89107 #7 0x0000000000000000 in ?? ()^M
89108 (gdb) FAIL: gdb.base/break-interp.exp: ldprelink=NO: ldsepdebug=NO: \
89109 first backtrace: dl bt
89112 Using the backtrace, the test-case tries to establish that we're stopped in
89115 However, the backtrace only shows an address, because:
89116 - the dynamic linker contains no minimal symbols and no debug info, and
89117 - gdb is build without --with-separate-debug-dir so it can't find the
89118 corresponding .debug file, which does contain the mimimal symbols and
89121 As in "[gdb/testsuite] Improve probe detection in gdb.base/break-probes.exp",
89122 fix this by doing info probes and grepping for the address.
89124 Tested on x86_64-linux.
89126 2021-09-23 Tom de Vries <tdevries@suse.de>
89128 [gdb/testsuite] Improve probe detection in gdb.base/break-probes.exp
89129 When running test-case gdb.base/break-probes.exp on ubuntu 18.04.5, we have:
89132 Starting program: break-probes^M
89133 Stopped due to shared library event (no libraries added or removed)^M
89135 #0 0x00007ffff7dd6e12 in ?? () from /lib64/ld-linux-x86-64.so.2^M
89136 #1 0x00007ffff7dedf50 in ?? () from /lib64/ld-linux-x86-64.so.2^M
89137 #2 0x00007ffff7dd5128 in ?? () from /lib64/ld-linux-x86-64.so.2^M
89138 #3 0x00007ffff7dd4098 in ?? () from /lib64/ld-linux-x86-64.so.2^M
89139 #4 0x0000000000000001 in ?? ()^M
89140 #5 0x00007fffffffdaac in ?? ()^M
89141 #6 0x0000000000000000 in ?? ()^M
89142 (gdb) UNSUPPORTED: gdb.base/break-probes.exp: probes not present on this system
89145 Using the backtrace, the test-case tries to establish that we're stopped in
89146 dl_main, which is used as proof that we're using probes.
89148 However, the backtrace only shows an address, because:
89149 - the dynamic linker contains no minimal symbols and no debug info, and
89150 - gdb is build without --with-separate-debug-dir so it can't find the
89151 corresponding .debug file, which does contain the mimimal symbols and
89154 Fix this by instead printing the pc and grepping for the value in the
89155 info probes output:
89158 $1 = 0x7ffff7dd6e12^M
89159 (gdb) info probes^M
89160 Type Provider Name Where Object ^M
89162 stap rtld init_start 0x00007ffff7dd6e12 /lib64/ld-linux-x86-64.so.2 ^M
89167 Tested on x86_64-linux.
89169 2021-09-23 Tom de Vries <tdevries@suse.de>
89171 [gdb/testsuite] Handle failing probe detection in gdb.base/break-probes.exp
89172 When running test-case gdb.base/break-probes.exp on ubuntu 18.04.5, we have:
89175 #0 0x00007ffff7dd6e12 in ?? () from /lib64/ld-linux-x86-64.so.2^M
89176 #1 0x00007ffff7dedf50 in ?? () from /lib64/ld-linux-x86-64.so.2^M
89177 #2 0x00007ffff7dd5128 in ?? () from /lib64/ld-linux-x86-64.so.2^M
89178 #3 0x00007ffff7dd4098 in ?? () from /lib64/ld-linux-x86-64.so.2^M
89179 #4 0x0000000000000001 in ?? ()^M
89180 #5 0x00007fffffffdaac in ?? ()^M
89181 #6 0x0000000000000000 in ?? ()^M
89182 (gdb) FAIL: gdb.base/break-probes.exp: ensure using probes
89185 The test-case intends to emit an UNTESTED in this case, but fails to do so
89186 because it tries to do it in a regexp clause in a gdb_test_multiple, which
89187 doesn't trigger. Instead, a default clause triggers which produces the FAIL.
89189 Also the use of UNTESTED is not appropriate, and we should use UNSUPPORTED
89192 Fix this by silencing the FAIL, and emitting an UNSUPPORTED after the
89195 if { ! $using_probes } {
89196 + unsupported "probes not present on this system"
89201 Tested on x86_64-linux.
89203 2021-09-23 Tom de Vries <tdevries@suse.de>
89205 [gdb/testsuite] Use early-out style in gdb.base/break-probes.exp
89206 Reduce indentation and improve readability in test-case
89207 gdb.base/break-probes.exp by replacing:
89221 Tested on x86_64-linux.
89223 2021-09-23 Pedro Alves <pedro@palves.net>
89225 Test that frame info/IDs are stable/consistent
89226 This adds a testcase that tests that the unwinder produces consistent
89227 frame info and frame IDs by making sure that "info frame" shows the
89228 same result when stopped at a function (level == 0), compared to when
89229 we find the same frame in the stack at a level > 0.
89231 E.g., on x86-64, right after running to main, we see:
89234 Stack level 0, frame at 0x7fffffffd340:
89235 rip = 0x555555555168 in main (gdb.base/backtrace.c:41); saved rip = 0x7ffff7dd90b3
89237 Arglist at 0x7fffffffd330, args:
89238 Locals at 0x7fffffffd330, Previous frame's sp is 0x7fffffffd340
89240 rbp at 0x7fffffffd330, rip at 0x7fffffffd338
89243 and then after continuing to a function called by main, and selecting
89244 the "main" frame again, we see:
89247 Stack level 3, frame at 0x7fffffffd340:
89248 rip = 0x555555555172 in main (gdb.base/backtrace.c:41); saved rip = 0x7ffff7dd90b3
89249 caller of frame at 0x7fffffffd330
89251 Arglist at 0x7fffffffd330, args:
89252 Locals at 0x7fffffffd330, Previous frame's sp is 0x7fffffffd340
89254 rbp at 0x7fffffffd330, rip at 0x7fffffffd338
89257 The only differences should be in the stack level, the 'rip = '
89258 address, and the presence of the "caller of frame at" info. All the
89259 rest should be the same. If it isn't, it probably means that the
89260 frame base, the frame ID, etc. aren't stable & consistent.
89262 The testcase exercises both the DWARF and the heuristic unwinders,
89263 using "maint set dwarf unwinder on/off".
89265 Tested on {x86-64 -m64, x86-64 -m32, Aarch64, Power8} GNU/Linux.
89267 Change-Id: I795001c82cc70d543d197415e3f80ce5dc7f3452
89269 2021-09-23 Tom Tromey <tromey@adacore.com>
89271 Change get_ada_task_ptid parameter type
89272 get_ada_task_ptid currently takes a 'long' as its 'thread' parameter
89273 type. However, on some platforms this is actually a pointer, and
89274 using 'long' can sometimes end up with the value being sign-extended.
89275 This sign extension can cause problems later, if the tid is then later
89276 used as an address again.
89278 This patch changes the parameter type to ULONGEST and updates all the
89279 uses. This approach preserves sign extension on the targets where it
89280 is apparently intended, while avoiding it on others.
89282 Co-Authored-By: John Baldwin <jhb@FreeBSD.org>
89284 2021-09-23 Tom Tromey <tromey@adacore.com>
89286 Change ptid_t::tid to ULONGEST
89287 The ptid_t 'tid' member is normally used as an address in gdb -- both
89288 bsd-uthread and ravenscar-thread use it this way. However, because
89289 the type is 'long', this can cause problems with sign extension.
89291 This patch changes the type to ULONGEST to ensure that sign extension
89294 2021-09-23 Tom Tromey <tromey@adacore.com>
89296 Remove defaulted 'tid' parameter to ptid_t constructor
89297 I wanted to find, and potentially modify, all the spots where the
89298 'tid' parameter to the ptid_t constructor was used. So, I temporarily
89299 removed this parameter and then rebuilt.
89301 In order to make it simpler to search through the "real" (nonzero)
89302 uses of this parameter, something I knew I'd have to do multiple
89303 times, I removed any ", 0" from constructor calls.
89305 Co-Authored-By: John Baldwin <jhb@FreeBSD.org>
89307 2021-09-23 Tom Tromey <tom@tromey.com>
89309 Style the "XXX" text in ptype/o
89310 This patch changes gdb to use the 'highlight' style on the "XXX" text
89311 in the output of ptype/o.
89313 2021-09-23 GDB Administrator <gdbadmin@sourceware.org>
89315 Automatic date update in version.in
89317 2021-09-22 Tom de Vries <tdevries@suse.de>
89319 [gdb/testsuite] Fix gdb.python/py-events.exp
89320 With test-case gdb.python/py-events.exp on ubuntu 18.04.5 we run into:
89322 (gdb) info threads^M
89323 Id Target Id Frame ^M
89324 * 1 Thread 0x7ffff7fc3740 (LWP 31467) "py-events" do_nothing () at \
89325 src/gdb/testsuite/gdb.python/py-events-shlib.c:19^M
89326 (gdb) FAIL: gdb.python/py-events.exp: get current thread
89329 The info thread commands uses "Thread" instead of "process" because
89330 libpthread is already loaded:
89332 new objfile name: /lib/x86_64-linux-gnu/libdl.so.2^M
89333 [Thread debugging using libthread_db enabled]^M
89334 Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".^M
89335 event type: new_objfile^M
89336 new objfile name: /lib/x86_64-linux-gnu/libpthread.so.0^M
89338 and consequently thread_db_target::pid_to_str is used.
89340 Fix this by parsing the "Thread" expression.
89342 Tested on x86_64-linux.
89344 2021-09-22 Tom de Vries <tdevries@suse.de>
89346 [gdb] Add maint selftest -verbose option
89347 The print_one_insn selftest in gdb/disasm-selftests.c contains:
89349 /* If you want to see the disassembled instruction printed to gdb_stdout,
89350 set verbose to true. */
89351 static const bool verbose = false;
89354 Make this parameter available in the maint selftest command using a new option
89355 -verbose, such that we can do:
89357 (gdb) maint selftest -verbose print_one_insn
89360 Tested on x86_64-linux.
89362 2021-09-22 Alan Modra <amodra@gmail.com>
89364 dwarf2 sub-section test
89365 This is a testcase for the bug fixed by commit 5b4846283c3d. When
89366 running the testcase on ia64 targets I found timeouts along with lots
89367 of memory being consumed, due to ia64 gas not tracking text
89368 sub-sections. Trying to add nops for ".nop 16" in ".text 1" resulting
89369 in them being added to subsegment 0, with no increase to subsegment 1
89370 size. This patch also fixes that problem.
89372 Note that the testcase fails on ft32-elf, mn10200-elf, score-elf,
89373 tic5x-elf, and xtensa-elf. The first two are relocation errors, the
89374 last three appear to be the .nop directive failing to emit the right
89375 number of nops. I didn't XFAIL any of them.
89377 * config/tc-ia64.c (md): Add last_text_subseg.
89378 (ia64_flush_insns, dot_endp): Use last_text_subseg.
89379 (ia64_frob_label, md_assemble): Set last_text_subseg.
89380 * testsuite/gas/elf/dwarf2-21.d,
89381 * testsuite/gas/elf/dwarf2-21.s: New test.
89382 * testsuite/gas/elf/elf.exp: Run it.
89384 2021-09-22 Alan Modra <amodra@gmail.com>
89386 Fix x86 "FAIL: TLS -fno-pic -shared"
89387 Fix a typo in commit 5d0869d9872a
89389 * testsuite/ld-i386/tlsnopic.rd: Typo fix.
89391 2021-09-22 GDB Administrator <gdbadmin@sourceware.org>
89393 Automatic date update in version.in
89395 2021-09-21 Nick Clifton <nickc@redhat.com>
89397 Change the linker's heuristic for computing the entry point for binaries so that shared libraries default to an entry point of 0.
89398 * ldlang.c (lang_end): When computing the entry point, only
89399 try the start address of the entry section when creating an
89401 * ld.texi (Entry point): Update description of heuristic used to
89402 choose the entry point.
89403 testsuite/ld-alpha/tlspic.rd: Update expected entry point address.
89404 testsuite/ld-arm/tls-gdesc-got.d: Likewise.
89405 testsuite/ld-i386/tlsnopic.rd: Likewise.
89406 testsuite/ld-ia64/tlspic.rd: Likewise.
89407 testsuite/ld-sparc/gotop32.rd: Likewise.
89408 testsuite/ld-sparc/gotop64.rd: Likewise.
89409 testsuite/ld-sparc/tlssunnopic32.rd: Likewise.
89410 testsuite/ld-sparc/tlssunnopic64.rd: Likewise.
89411 testsuite/ld-sparc/tlssunpic32.rd: Likewise.
89412 testsuite/ld-sparc/tlssunpic64.rd: Likewise.
89413 testsuite/ld-tic6x/shlib-1.rd: Likewise.
89414 testsuite/ld-tic6x/shlib-1b.rd: Likewise.
89415 testsuite/ld-tic6x/shlib-1r.rd: Likewise.
89416 testsuite/ld-tic6x/shlib-1rb.rd: Likewise.
89417 testsuite/ld-tic6x/shlib-noindex.rd: Likewise.
89418 testsuite/ld-x86-64/pr14207.d: Likewise.
89419 testsuite/ld-x86-64/tlsdesc.rd: Likewise.
89420 testsuite/ld-x86-64/tlspic.rd: Likewise.
89421 testsuite/ld-x86-64/tlspic2.rd: Likewise.
89423 2021-09-21 Tom de Vries <tdevries@suse.de>
89425 [gdb/testsuite] Handle supports_memtag in gdb.base/gdb-caching-proc.exp
89426 In test-case gdb.base/gdb-caching-proc.exp, we run all procs declared with
89427 gdb_caching_proc. Some of these require a gdb instance, some not.
89429 We could just do a clean_restart every time, but that would amount to 44 gdb
89430 restarts. We try to minimize this by doing this only for the few procs that
89431 need it, and hardcoding those in the test-case.
89433 For those procs, we do a clean_restart, execute the proc, and then do a
89434 gdb_exit, to make sure the gdb instance doesn't linger such that we detect
89435 procs that need a gdb instance but are not listed in the test-case.
89437 However, that doesn't work in the case of gnat_runtime_has_debug_info. This
89438 proc doesn't require a gdb instance because it starts its own. But it doesn't
89439 clean up the gdb instance, and since it's not listed, the test-case
89440 doesn't clean up the gdb instance eiter. Consequently, the proc
89441 supports_memtag (which should be listed, but isn't) uses the gdb instance
89442 started by gnat_runtime_has_debug_info rather than throwing an error. Well,
89443 unless gnat_runtime_has_debug_info fails before starting a gdb instance, in
89444 which case we do run into the error.
89447 - doing gdb_exit unconditionally
89448 - fixing the resulting error by adding supports_memtag in the test-case to
89449 the "needing gdb instance" list
89451 Tested on x86_64-linux.
89453 2021-09-21 Felix Willgerodt <felix.willgerodt@intel.com>
89455 gdb, doc: Add ieee_half and bfloat16 to list of predefined target types.
89456 For some reason these two weren't added to the list when they were orginally
89460 2021-09-21 Felix Willgerodt <felix.willgerodt@intel.com>
89462 * gdb.texinfo (Predefined Target Types): Mention ieee_half and bfloat16.
89464 2021-09-21 GDB Administrator <gdbadmin@sourceware.org>
89466 Automatic date update in version.in
89468 2021-09-20 Tom de Vries <tdevries@suse.de>
89470 [gdb/testsuite] Fix gdb.ada/interface.exp with gcc-9
89471 When running test-case gdb.ada/interface.exp with gcc-9, we run into:
89473 (gdb) info locals^M
89474 s = (x => 1, y => 2, w => 3, h => 4)^M
89475 r = (x => 1, y => 2, w => 3, h => 4)^M
89476 (gdb) FAIL: gdb.ada/interface.exp: info locals
89479 The failure is caused by the regexp expecting variable r followed by
89482 Fix this by allowing variable s followed by variable r as well.
89484 Tested on x86_64-linux.
89486 2021-09-20 Tom de Vries <tdevries@suse.de>
89488 [gdb/testsuite] Fix gdb.ada/mi_prot.exp
89489 When running test-case gdb.ada/mi_prot.exp with gcc 8.5.0, we run into:
89492 Expecting: ^(-stack-list-arguments --no-frame-filters 1[^M
89493 ]+)?(\^done,stack=.*[^M
89496 -stack-list-arguments --no-frame-filters 1^M
89497 ^done,stack-args=[frame={level="0",args=[{name="<_object>",value="(ceiling_priority =\
89498 > 97, local => 0)"},{name="v",value="5"},{name="<_objectO>",value="true"}]},frame={le\
89499 vel="1",args=[{name="v",value="5"},{name="<_objectO>",value="true"}]},frame={level="2\
89502 FAIL: gdb.ada/mi_prot.exp: -stack-list-arguments --no-frame-filters 1 (unexpected out\
89506 Fix this by updating the regexp to expect "^done,stack-args=" instead of
89509 Tested on x86_64-linux.
89511 2021-09-20 Tom de Vries <tdevries@suse.de>
89513 [gdb/testsuite] Register test for each arch separately in register_test_foreach_arch
89514 In gdb/disasm-selftests.c we have:
89516 selftests::register_test_foreach_arch ("print_one_insn",
89517 selftests::print_one_insn_test);
89521 $ gdb -q -batch -ex "maint selftest print_one_insn" 2>&1 \
89523 Running selftest print_one_insn.
89527 Change the semantics register_test_foreach_arch such that a version of
89528 print_one_insn is registered for each architecture, such that we have:
89530 $ gdb -q -batch -ex "maint selftest print_one_insn" 2>&1 \
89532 Running selftest print_one_insn::A6.
89533 Running selftest print_one_insn::A7.
89534 Running selftest print_one_insn::ARC600.
89539 This makes it f.i. possible to do:
89541 $ gdb -q -batch a.out -ex "maint selftest print_one_insn::armv8.1-m.main"
89542 Running selftest print_one_insn::armv8.1-m.main.
89543 Self test failed: self-test failed at src/gdb/disasm-selftests.c:165
89544 Ran 1 unit tests, 1 failed
89547 Tested on x86_64-linux with an --enable-targets=all build.
89549 2021-09-20 Tom de Vries <tdevries@suse.de>
89551 [gdb] Change register_test to use std::function arg
89552 Change register_test to use std::function arg, such that we can do:
89554 register_test (test_name, [=] () { SELF_CHECK (...); });
89557 Tested on x86_64-linux.
89559 2021-09-20 Tom de Vries <tdevries@suse.de>
89561 [gdb/testsuite] Fix gdb.ada/big_packed_array.exp xfail for -m32
89562 With test-case gdb.ada/big_packed_array.exp and target board unix/-m32 I run
89566 $2 = (0 => 0 <repeats 24 times>, 160)^M
89567 (gdb) FAIL: gdb.ada/big_packed_array.exp: scenario=minimal: print bad
89570 The problem is that while the variable is an array of 196 bits (== 24.5 bytes),
89571 the debug information describes it as 25 unsigned char. This is PR
89572 gcc/101643, and the test-case contains an xfail for this, which catches only:
89575 $2 = (0 => 0 <repeats 25 times>)^M
89578 Fix this by updating the xfail pattern.
89580 Tested on x86_64-linux.
89582 2021-09-20 Simon Marchi <simon.marchi@polymtl.ca>
89584 gdbsupport/gdb_proc_service.h: use decltype instead of typeof
89585 Bug 28341 shows that GDB fails to compile when built with -std=c++11.
89586 I don't know much about the use case, but according to the author of the
89589 I encountered the scenario where CXX is set to "g++ -std=c++11" when
89590 I try to compile binutils under GCC as part of the GCC 3-stage
89591 compilation, which is common for building a cross-compiler.
89593 The author of the bug suggests using __typeof__ instead of typeof. But
89594 since we're using C++, we might as well use decltype, which is standard.
89595 This is what this patch does.
89597 The failure (and fix) can be observed by configuring GDB with CXX="g++
89601 In file included from /home/simark/src/binutils-gdb/gdbserver/gdb_proc_service.h:22,
89602 from /home/simark/src/binutils-gdb/gdbserver/linux-low.h:27,
89603 from /home/simark/src/binutils-gdb/gdbserver/linux-low.cc:20:
89604 /home/simark/src/binutils-gdb/gdbserver/../gdbsupport/gdb_proc_service.h:177:50: error: expected constructor, destructor, or type conversion before (token
89605 177 | __attribute__((visibility ("default"))) typeof (SYM) SYM
89607 /home/simark/src/binutils-gdb/gdbserver/../gdbsupport/gdb_proc_service.h:179:1: note: in expansion of macro PS_EXPORT
89608 179 | PS_EXPORT (ps_get_thread_area);
89611 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28341
89612 Change-Id: I84fbaae938209d8d935ca08dec9b7e6a0dd1bda0
89614 2021-09-20 Andrew Burgess <andrew.burgess@embecosm.com>
89616 riscv: print .2byte or .4byte before an unknown instruction encoding
89617 When the RISC-V disassembler encounters an unknown instruction, it
89618 currently just prints the value of the bytes, like this:
89620 Dump of assembler code for function custom_insn:
89621 0x00010132 <+0>: addi sp,sp,-16
89622 0x00010134 <+2>: sw s0,12(sp)
89623 0x00010136 <+4>: addi s0,sp,16
89624 0x00010138 <+6>: 0x52018b
89625 0x0001013c <+10>: 0x9c45
89627 My proposal, in this patch, is to change the behaviour to this:
89629 Dump of assembler code for function custom_insn:
89630 0x00010132 <+0>: addi sp,sp,-16
89631 0x00010134 <+2>: sw s0,12(sp)
89632 0x00010136 <+4>: addi s0,sp,16
89633 0x00010138 <+6>: .4byte 0x52018b
89634 0x0001013c <+10>: .2byte 0x9c45
89636 Adding the .4byte and .2byte opcodes. The benefit that I see here is
89637 that in the patched version of the tools, the disassembler output can
89638 be fed back into the assembler and it should assemble to the same
89639 binary format. Before the patch, the disassembler output is invalid
89642 I've started a RISC-V specific test file under binutils so that I can
89643 add a test for this change.
89645 binutils/ChangeLog:
89647 * testsuite/binutils-all/riscv/riscv.exp: New file.
89648 * testsuite/binutils-all/riscv/unknown.d: New file.
89649 * testsuite/binutils-all/riscv/unknown.s: New file.
89653 * riscv-dis.c (riscv_disassemble_insn): Print a .%dbyte opcode
89654 before an unknown instruction, '%d' is replaced with the
89655 instruction length.
89657 2021-09-20 Alan Modra <amodra@gmail.com>
89659 Fix allocate_filenum last dir/file checks
89660 * dwarf2dbg.c (allocate_filenum) Correct use of last_used_dir_len.
89662 2021-09-20 Alan Modra <amodra@gmail.com>
89664 Re: PR28149, debug info with wrong file association
89665 Fixes segfaults when building aarch64-linux kernel, due to only doing
89666 part of the work necessary when allocating file numbers late. I'd
89667 missed looping over subsegments, which resulted in some u.filename
89668 entries left around and later interpreted as u.view.
89671 * dwarf2dbg.c (purge_generated_debug): Iterate over subsegs too.
89672 (dwarf2_finish): Call do_allocate_filenum for all subsegs too,
89673 in a separate loop before subsegs are chained.
89675 2021-09-20 Alan Modra <amodra@gmail.com>
89677 Move eelf_mipsel_haiki.c to ALL_64_EMULATION_SOURCES
89678 --enable-targets=all on a 32-bit host results in a link failure with
89679 undefined references due to elfxx-mips.c not being compiled. This
89680 patch fixes that by putting eelf_mipsel_haiki.c in the correct
89681 EMULATION_SOURCES Makefile variable. I've also added a bunch of
89682 missing file dependencies and sorted a few things so that it's easier
89683 to verify dependencies are present.
89685 * Makfile.am: Add missing haiku dependencies, sort.
89686 (ALL_EMULATION_SOURCES): Sort. Move eelf_mipsel_haiku.c to..
89687 (ALL_64_EMULATION_SOURCES): ..here. Sort.
89688 * Makfile.in: Regenerate.
89690 2021-09-20 GDB Administrator <gdbadmin@sourceware.org>
89692 Automatic date update in version.in
89694 2021-09-19 H.J. Lu <hjl.tools@gmail.com>
89696 elf: Don't set version info on unversioned symbols
89697 Don't set version info on unversioned symbols when seeing a hidden
89698 versioned symbol after an unversioned definition and the default
89704 * elflink.c (elf_link_add_object_symbols): Don't set version info
89705 on unversioned symbols.
89710 * testsuite/ld-elf/pr28348.rd: New file.
89711 * testsuite/ld-elf/pr28348.t: Likewise.
89712 * testsuite/ld-elf/pr28348a.c: Likewise.
89713 * testsuite/ld-elf/pr28348b.c: Likewise.
89714 * testsuite/ld-elf/pr28348c.c: Likewise.
89715 * testsuite/ld-elf/shared.exp: Run PR ld/28348 tests.
89717 2021-09-19 Mike Frysinger <vapier@gentoo.org>
89719 gdb: manual: update @inforef to @xref
89720 The @inforef command is deprecated, and @xref does the samething.
89721 Also had to update the text capitalization to match current manual.
89722 Verified that info & HTML links work.
89724 2021-09-19 Weimin Pan <weimin.pan@oracle.com>
89726 CTF: multi-CU and archive support
89727 Now gdb is capable of debugging executable, which consists of multiple
89728 compilation units (CUs) with the CTF debug info. An executable could
89729 potentially have one or more archives, which, in CTF context, contain
89732 all changes were made in ctfread.c in which elfctf_build_psymtabs was
89733 modified to handle archives, via the ctf archive iterator and its callback
89734 build_ctf_archive_member and scan_partial_symbols was modified to scan
89735 archives, which are treated as subfiles, to build the psymtabs.
89737 Also changes were made to handle CTF's data object section and function
89738 info section which now share the same format of their contents - an array
89739 of type IDs. New functions ctf_psymtab_add_stt_entries, which is called by
89740 ctf_psymtab_add_stt_obj and ctf_psymtab_add_stt_func, and add_stt_entries,
89741 which is called by add_stt_obj and add_stt_func when setting up psymtabs
89742 and full symtab, respectively.
89744 2021-09-19 GDB Administrator <gdbadmin@sourceware.org>
89746 Automatic date update in version.in
89748 2021-09-18 Tom de Vries <tdevries@suse.de>
89750 [gdb/testsuite] Fix gdb.server/server-kill.exp with -m32
89751 When running test-case gdb.server/server-kill.exp with target board unix/-m32,
89754 0xf7fd6b20 in _start () from /lib/ld-linux.so.2^M
89755 (gdb) Executing on target: kill -9 13082 (timeout = 300)
89756 builtin_spawn -ignore SIGHUP kill -9 13082^M
89758 (gdb) FAIL: gdb.server/server-kill.exp: kill_pid_of=server: test_unwind_syms: bt
89761 The test-case expects the backtrace command to trigger remote communication,
89762 which then should result in a "Remote connection closed" or similar.
89764 However, no remote communication is triggered, because we hit the "Check that
89765 this frame is unwindable" case in get_prev_frame_always_1.
89767 We don't hit this problem in the kill_pid_of=inferior case, because there we
89768 run to main before doing the backtrace.
89770 Fix this by doing the same in the kill_pid_of=server case.
89772 Tested on x86_64-linux.
89774 2021-09-18 Tom de Vries <tdevries@suse.de>
89776 [gdb/ada] Handle artificial local symbols
89777 With current master and gcc 7.5.0/8.5.0, we have this timeout:
89780 Multiple matches for s^M
89782 [1] s at src/gdb/testsuite/gdb.ada/interface/foo.adb:20^M
89783 [2] s at src/gdb/testsuite/gdb.ada/interface/foo.adb:?^M
89784 > FAIL: gdb.ada/interface.exp: print s (timeout)
89787 [ The FAIL doesn't reproduce with gcc 9.3.1. This difference in
89788 behaviour bisects to gcc commit d70ba0c10de.
89790 The FAIL with earlier gcc bisects to gdb commit ba8694b650b. ]
89792 The FAIL is caused by gcc generating this debug info describing a named
89793 artificial variable:
89795 <2><1204>: Abbrev Number: 31 (DW_TAG_variable)
89796 <1205> DW_AT_name : s.14
89797 <1209> DW_AT_type : <0x1213>
89798 <120d> DW_AT_artificial : 1
89799 <120d> DW_AT_location : 5 byte block: 91 e0 7d 23 18 \
89800 (DW_OP_fbreg: -288; DW_OP_plus_uconst: 24)
89803 An easy way to fix this would be to simply not put named artificial variables
89804 into the symbol table. However, that causes regressions for Ada. It relies
89805 on being able to get the value from such variables, using a named reference.
89807 Fix this instead by marking the symbol as artificial, and:
89808 - ignoring such symbols in ada_resolve_variable, which fixes the FAIL
89809 - ignoring such ada symbols in do_print_variable_and_value, which prevents
89810 them from showing up in "info locals"
89812 Note that a fix for the latter was submitted here (
89813 https://sourceware.org/pipermail/gdb-patches/2008-January/054994.html ), and
89814 this patch borrows from it.
89816 Tested on x86_64-linux.
89818 Co-Authored-By: Joel Brobecker <brobecker@adacore.com>
89820 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28180
89822 2021-09-18 GDB Administrator <gdbadmin@sourceware.org>
89824 Automatic date update in version.in
89826 2021-09-17 Alan Modra <amodra@gmail.com>
89828 PR28149 part 2, purge generated line info
89829 Mixing compiler generated line info with gas generated line info is
89830 generally just confusing. Also .loc directives with non-zero view
89831 fields might reference a previous .loc. It becomes a little more
89832 tricky to locate that previous .loc if there might be gas generated
89833 line info present too. Mind you, we turn off gas generation of line
89834 info on seeing compiler generated line info, so any reference back
89835 won't hit gas generated line info. At least, if the view info is
89836 sane. Unfortunately, gas needs to handle mangled source.
89839 * dwarf2dbg.c (purge_generated_debug): New function.
89840 (dwarf2_directive_filename): Call the above.
89841 (out_debug_line): Don't segfault after purging.
89842 * testsuite/gas/i386/dwarf2-line-4.d: Update expected output.
89843 * testsuite/gas/i386/dwarf4-line-1.d: Likewise.
89844 * testsuite/gas/i386/dwarf5-line-1.d: Likewise.
89845 * testsuite/gas/i386/dwarf5-line-2.d: Likewise.
89847 2021-09-17 Alan Modra <amodra@gmail.com>
89849 PR28149, debug info with wrong file association
89850 gcc-11 and gcc-12 pass -gdwarf-5 to gas, in order to prime gas for
89851 DWARF 5 level debug info. Unfortunately it seems there are cases
89852 where the compiler does not emit a .file or .loc dwarf debug directive
89853 before any machine instructions. (Note that the .file directive
89854 typically emitted as the first line of assembly output doesn't count as
89855 a dwarf debug directive. The dwarf .file has a file number before the
89858 This patch delays allocation of file numbers for gas generated line
89859 debug info until the end of assembly, thus avoiding any clashes with
89860 compiler generated file numbers. Two fixes for test case source are
89861 necessary; A .loc can't use a file number that hasn't already been
89862 specified with .file.
89864 A followup patch will remove all the gas generated line info on
89865 seeing a .file directive.
89868 * dwarf2dbg.c (num_of_auto_assigned): Delete.
89869 (current): Update initialisation.
89870 (set_or_check_view): Replace all accesses to view with u.view.
89871 (dwarf2_consume_line_info): Likewise.
89872 (dwarf2_directive_loc): Likewise. Assert that we aren't generating
89874 (dwarf2_gen_line_info_1): Don't call set_or_check_view on
89875 gas generated line entries.
89876 (dwarf2_gen_line_info): Set and track filenames for gas generated
89877 line entries. Simplify generation of labels.
89878 (get_directory_table_entry): Use filename_cmp when comparing dirs.
89879 (do_allocate_filenum): New function.
89880 (dwarf2_where): Set u.filename and filenum to -1 for gas generated
89882 (dwarf2_directive_filename): Remove num_of_auto_assigned handling.
89883 (process_entries): Update view field access. Call
89884 do_allocate_filenum.
89885 * dwarf2dbg.h (struct dwarf2_line_info): Add filename field in
89886 union aliasing view.
89887 * testsuite/gas/i386/dwarf2-line-3.s: Add .file directive.
89888 * testsuite/gas/i386/dwarf2-line-4.s: Likewise.
89889 * testsuite/gas/i386/dwarf2-line-4.d: Update expected output.
89890 * testsuite/gas/i386/dwarf4-line-1.d: Likewise.
89891 * testsuite/gas/i386/dwarf5-line-1.d: Likewise.
89892 * testsuite/gas/i386/dwarf5-line-2.d: Likewise.
89894 2021-09-17 Alan Modra <amodra@gmail.com>
89896 [GOLD] PowerPC64 support for sym+addend GOT entries
89897 Pass addends to all the GOT handling functions, plus remove some
89898 extraneous asserts.
89901 * powerpc.cc (Output_data_got_powerpc): Add addend parameter to
89902 all methods creating got entries.
89903 (Target_powerpc::Scan::local): Pass reloc addend to got handling
89904 functions, and when creating dynamic got relocations.
89905 (Target_powerpc::Scan::global): Likewise.
89906 (Target_powerpc::Relocate::relocate): Likewise. Remove extraneous
89909 2021-09-17 Alan Modra <amodra@gmail.com>
89911 [GOLD] Got_entry::write addends
89912 This takes care of writing out GOT entries with addends. The local
89913 symbol case was already largely handled, except for passing the addend
89914 to tls_offset_for_local which might need the addend in a
89915 local_got_offset call. That's needed also in tls_offset_for_global.
89917 I'm assuming here that GOT entries for function symbols won't ever
89918 have addends, and in particular that a GOT entry referencing PLT call
89919 stub code won't want an offset into the code.
89922 * output.cc (Output_data_got::Got_entry::write): Include addend
89923 in global symbol value. Pass addend to tls_offset_for_*.
89924 * powerpc.cc (Target_powerpc::do_tls_offset_for_local): Handle addend.
89925 (Target_powerpc::do_tls_offset_for_global): Likewise.
89926 * s390.cc (Target_s390::do_tls_offset_for_local): Likewise.
89927 (Target_s390::do_tls_offset_for_global): Likewise.
89928 * target.h (Target::tls_offset_for_local): Add addend param.
89929 (Target::tls_offset_for_global): Likewise.
89930 (Target::do_tls_offset_for_local): Likewise.
89931 (Target::do_tls_offset_for_global): Likewise.
89933 2021-09-17 Alan Modra <amodra@gmail.com>
89935 [GOLD] Output_data_got create entry method addends
89936 This patch makes all the Output_data_got methods that create new
89937 entries accept an optional addend.
89940 * output.h (Output_data_got::add_global): Add optional addend
89941 parameter. Update comment. Delete overload without addend.
89942 (Output_data_got::add_global_plt): Likewise.
89943 (Output_data_got::add_global_tls): Likewise.
89944 (Output_data_got::add_global_with_rel): Likewise.
89945 (Output_data_got::add_global_pair_with_rel): Likewise.
89946 (Output_data_got::add_local_plt): Likewise.
89947 (Output_data_got::add_local_tls): Likewise.
89948 (Output_data_got::add_local_tls_pair): Likewise.
89949 (Output_data_got::reserve_local): Likewise.
89950 (Output_data_got::reserve_global): Likewise.
89951 (Output_data_got::Got_entry): Include addend in global sym
89952 constructor. Delete local sym constructor without addend.
89953 * output.cc (Output_data_got::add_global): Add addend param,
89954 pass to got handling methods.
89955 (Output_data_got::add_global_plt): Likewise.
89956 (Output_data_got::add_global_with_rel): Likewise.
89957 (Output_data_got::add_global_pair_with_rel): Likewise.
89958 (Output_data_got::add_local_plt): Likewise.
89959 (Output_data_got::add_local_tls_pair): Likewise.
89960 (Output_data_got::reserve_local): Likewise.
89961 (Output_data_got::reserve_global): Likewise.
89963 2021-09-17 Alan Modra <amodra@gmail.com>
89965 [GOLD] Output_data_got tidy
89966 Some Output_data_got methods already have support for addends, but
89967 were implemented as separate methods. This removes unnecessary code
89970 Relobj::local_has_got_offset and others there get a similar treatment.
89971 Comments are removed since it should be obvious without a comment, and
89972 the existing comments are not precisely what the code does. For
89973 example, a local_has_got_offset call without an addend does not return
89974 whether the local symbol has *a* GOT offset of type GOT_TYPE, it
89975 returns whether there is a GOT entry of type GOT_TYPE for the symbol
89976 with addend of zero.
89979 * output.h (Output_data_got::add_local): Make addend optional.
89980 (Output_data_got::add_local_with_rel): Likewise.
89981 (Output_data_got::add_local_pair_with_rel): Likewise.
89982 * output.cc (Output_data_got::add_local): Delete overload
89984 (Output_data_got::add_local_with_rel): Likewise.
89985 (Output_data_got::add_local_pair_with_rel): Likewise.
89986 * object.h (Relobj::local_has_got_offset): Make addend optional.
89987 Delete overload without addend later. Update comment.
89988 (Relobj::local_got_offset): Likewise.
89989 (Relobj::set_local_got_offset): Likewise.
89991 2021-09-17 Alan Modra <amodra@gmail.com>
89993 [GOLD] Remove addend from Local_got_entry_key
89994 This patch removes the addend from Local_got_entry_key, which is
89995 unnecessary now that Got_offset_list has an addend. Note that it
89996 might be advantageous to keep the addend in Local_got_entry_key when
89997 linking objects containing a large number of section_sym+addend@got
89998 relocations. I opted to save some memory by removing the field but
89999 left the class there in case we might need to restore {sym,addend}
90000 lookup. That's also why this change is split out from the
90001 Got_offset_list change.
90004 * object.h (Local_got_entry_key): Delete addend_ field.
90005 Adjust constructor and methods to suit.
90006 * object.cc (Sized_relobj::do_for_all_local_got_entries):
90009 2021-09-17 Alan Modra <amodra@gmail.com>
90011 [GOLD] Got_offset_list: addend field
90012 This is the first in a series of patches aimed at supporting GOT
90013 entries against symbol plus addend generally for PowerPC64 rather than
90014 just section symbol plus addend as gold has currently.
90016 This patch adds an addend field to Got_offset_list, so that both local
90017 and global symbols can have GOT entries with addend.
90020 * object.h (Got_offset_list): Add addend_ field, init in both
90021 constructors. Adjust all accessors to suit.
90022 (Sized_relobj::do_local_has_got_offset): Adjust to suit.
90023 (Sized_relobj::do_local_got_offset): Likewise.
90024 (Sized_relobj::do_set_local_got_offset): Likewise.
90025 * symtab.h (Symbol::has_got_offset): Add optional addend param.
90026 (Symbol::got_offset, Symbol::set_got_offset): Likewise.
90027 * incremental.cc (Local_got_offset_visitor::visit): Add unused
90028 uint64_t parameter with FIXME.
90029 (Global_got_offset_visitor::visit): Add unused uint64_t parameter.
90031 2021-09-17 Henry Castro <hcvcastro@gmail.com>
90033 Fix segfault when running ia16-elf-gdb
90034 "A problem internal to GDB has been detected,
90035 further debugging may prove unreliable."
90039 2021-09-17 Nelson Chu <nelson.chu@sifive.com>
90041 RISC-V: Merged extension string tables and their version tables into one.
90042 There are two main reasons for this patch,
90044 * In the past we had two extension tables, one is used to record all
90045 supported extensions in bfd/elfxx-riscv.c, another is used to get the
90046 default extension versions in gas/config/tc-riscv.c. It is hard to
90047 maintain lots of tables in different files, but in fact we can merge
90048 them into just one table. Therefore, we now define many riscv_supported_std*
90049 tables, which record names and versions for all supported extensions.
90050 We not only use these tables to initialize the riscv_ext_order, but
90051 also use them to get the default versions of extensions, and decide if
90052 the extensions should be enbaled by default.
90054 * We add a new filed `default_enable' for the riscv_supported_std* tables,
90055 to decide if the extension should be enabled by default. For now if the
90056 `default_enable' field of the extension is set to EXT_DEFAULT, then we
90057 should enable the extension when the -march and elf architecture attributes
90058 are not set. In the future, I suppose the `default_enable' can be set
90059 to lots of EXT_<VENDOR>, each vendor can decide to open which extensions,
90060 when the target triple of vendor is chosen.
90062 The elf/linux regression tests of riscv-gnu-toolchain are passed.
90065 * elfnn-riscv.c (cpu-riscv.h): Removed sine it is included in
90067 (riscv_merge_std_ext): Updated since the field of rpe is changed.
90068 * elfxx-riscv.c (cpu-riscv.h): Removed.
90069 (riscv_implicit_subsets): Added implicit extensions for g.
90070 (struct riscv_supported_ext): Used to be riscv_ext_version. Moved
90071 from gas/config/tc-riscv.c, and added new field `default_enable' to
90072 decide if the extension should be enabled by default.
90073 (EXT_DEFAULT): Defined for `default_enable' field.
90074 (riscv_supported_std_ext): It used to return the supported standard
90075 architecture string, but now we move ext_version_table from
90076 gas/config/tc-riscv.c to here, and rename it to riscv_supported_std_ext.
90077 Currently we not only use the table to initialize riscv_ext_order, but
90078 also get the default versions of extensions, and decide if the extensions
90079 should be enbaled by default.
90080 (riscv_supported_std_z_ext): Likewise, but is used for z* extensions.
90081 (riscv_supported_std_s_ext): Likewise, but is used for s* extensions.
90082 (riscv_supported_std_h_ext): Likewise, but is used for h* extensions.
90083 (riscv_supported_std_zxm_ext): Likewise, but is used for zxm* extensions.
90084 (riscv_all_supported_ext): Includes all supported extension tables.
90085 (riscv_known_prefixed_ext): Updated.
90086 (riscv_valid_prefixed_ext): Updated.
90087 (riscv_init_ext_order): Init the riscv_ext_order table according to
90088 riscv_supported_std_ext.
90089 (riscv_get_default_ext_version): Moved from gas/config/tc-riscv.c.
90090 Get the versions of extensions from riscv_supported_std* tables.
90091 (riscv_parse_add_subset): Updated.
90092 (riscv_parse_std_ext): Updated.
90093 (riscv_set_default_arch): Set the default subset list according to
90094 the default_enable field of riscv_supported_*ext tables.
90095 (riscv_parse_subset): If the input ARCH is NULL, then we call
90096 riscv_set_default_arch to set the default subset list.
90097 * elfxx-riscv.h (cpu-riscv.h): Included.
90098 (riscv_parse_subset_t): Removed get_default_version field, and added
90099 isa_spec field to replace it.
90100 (extern riscv_supported_std_ext): Removed.
90102 * (bfd/cpu-riscv.h): Removed.
90103 (struct riscv_ext_version): Renamed and moved to bfd/elfxx-riscv.c.
90104 (ext_version_table): Likewise.
90105 (riscv_get_default_ext_version): Likewise.
90106 (ext_version_hash): Removed.
90107 (init_ext_version_hash): Removed.
90108 (riscv_set_arch): Updated since the field of rps is changed. Besides,
90109 report error when the architecture string is empty.
90110 (riscv_after_parse_args): Updated.
90112 2021-09-17 GDB Administrator <gdbadmin@sourceware.org>
90114 Automatic date update in version.in
90116 2021-09-16 Tom de Vries <tdevries@suse.de>
90118 [gdb/testsuite] Fix interrupted sleep in multi-threaded test-cases
90119 When running test-case gdb.threads/continue-pending-status.exp with native, I
90124 PASS: gdb.threads/continue-pending-status.exp: attempt 0: continue for ctrl-c
90126 Thread 1 "continue-pendin" received signal SIGINT, Interrupt.^M
90127 [Switching to Thread 0x7ffff7fc4740 (LWP 1276)]^M
90128 0x00007ffff758e4c0 in __GI___nanosleep () at nanosleep.c:27^M
90129 27 return SYSCALL_CANCEL (nanosleep, requested_time, remaining);^M
90130 (gdb) PASS: gdb.threads/continue-pending-status.exp: attempt 0: caught interrupt
90132 but with target board unix/-m32, I run into:
90136 PASS: gdb.threads/continue-pending-status.exp: attempt 0: continue for ctrl-c
90137 [Thread 0xf74aeb40 (LWP 31957) exited]^M
90138 [Thread 0xf7cafb40 (LWP 31956) exited]^M
90139 [Inferior 1 (process 31952) exited normally]^M
90143 The problem is that the sleep (300) call at the end of main is interrupted,
90144 which causes the inferior to exit before the ctrl-c can be send.
90146 This problem is described at "Interrupted System Calls" in the docs, and the
90147 suggested solution (using a sleep loop) indeed fixes the problem.
90149 Fix this instead using the more prevalent:
90153 while (1) sleep (1);
90155 which is roughly equivalent because the sleep is called at the end of main,
90156 but slightly better because it guards against hangs from the start rather than
90157 from the end of main.
90159 Likewise in gdb.base/watch_thread_num.exp.
90161 Likewise in gdb.btrace/enable-running.exp, but use the sleep loop there,
90162 because the sleep is not called at the end of main.
90164 Tested on x86_64-linux.
90166 2021-09-16 Mike Frysinger <vapier@gentoo.org>
90168 gdb: manual: fix werrors typo
90170 2021-09-16 GDB Administrator <gdbadmin@sourceware.org>
90172 Automatic date update in version.in
90174 2021-09-15 Tom de Vries <tdevries@suse.de>
90176 [gdb/testsuite] Use function_range in gdb.dwarf2/dw2-abs-hi-pc.exp
90177 When I run test-case gdb.dwarf2/dw2-abs-hi-pc.exp with gcc, we have:
90179 (gdb) break hello^M
90180 Breakpoint 1 at 0x4004c0: file dw2-abs-hi-pc-hello.c, line 24.^M
90181 (gdb) PASS: gdb.dwarf2/dw2-abs-hi-pc.exp: break hello
90183 but with clang, I run into:
90185 (gdb) break hello^M
90186 Breakpoint 1 at 0x4004e4^M
90187 (gdb) FAIL: gdb.dwarf2/dw2-abs-hi-pc.exp: break hello
90190 The problem is that the CU and function both have an empty address range:
90192 <0><d2>: Abbrev Number: 1 (DW_TAG_compile_unit)
90193 <108> DW_AT_name : dw2-abs-hi-pc-hello.c
90194 <123> DW_AT_low_pc : 0x4004e0
90195 <127> DW_AT_high_pc : 0x4004e0
90196 <1><12f>: Abbrev Number: 2 (DW_TAG_subprogram)
90197 <131> DW_AT_name : hello
90198 <13a> DW_AT_low_pc : 0x4004e0
90199 <13e> DW_AT_high_pc : 0x4004e0
90202 The address ranges are set like this in dw2-abs-hi-pc-hello-dbg.S:
90204 .4byte .hello_start /* DW_AT_low_pc */
90205 .4byte .hello_end /* DW_AT_high_pc */
90207 where the labels refer to dw2-abs-hi-pc-hello.c:
90211 asm (".hello_start: .globl .hello_start\n");
90215 asm (".hello0: .globl .hello0\n");
90217 asm (".hello1: .globl .hello1\n");
90219 asm (".hello_end: .globl .hello_end\n");
90222 Using asm labels in global scope is a known source of problems, as explained
90223 in the comment of proc function_range in gdb/testsuite/lib/dwarf.exp.
90225 Fix this by using function_range instead.
90227 Tested on x86_64-linux with gcc and clang-7 and clang-12.
90229 2021-09-15 Claudiu Zissulescu <claziss@synopsys.com>
90231 arc: Fix got-weak linker test
90232 Use regular expressions to fix the got-weak linker test.
90235 * testsuite/got-weak.d: Update test.
90237 2021-09-15 Andrew Burgess <andrew.burgess@embecosm.com>
90239 bfd: fix incorrect type used in sizeof
90240 Noticed in passing that we used 'sizeof (char **)' when calculating
90241 the size of a list of 'char *' pointers. Of course, this isn't really
90242 going to make a difference anywhere, but we may as well be correct.
90244 There should be no user visible changes after this commit.
90248 * archures.c (bfd_arch_list): Use 'char *' instead of 'char **'
90249 when calculating space for a string list.
90251 2021-09-15 Tom de Vries <tdevries@suse.de>
90253 [gdb/doc] Fix typo in maint selftest entry
90254 Fix typo "will by" -> "will be".
90256 2021-09-15 Tom de Vries <tdevries@suse.de>
90258 [bfd] Ensure unique printable names for bfd archs
90259 Remove duplicate entry in bfd_ft32_arch and bfd_rx_arch.
90261 Fix printable name for bfd_mach_n1: "nh1" -> "n1".
90264 * cpu-ft32.c (arch_info_struct): Remove "ft32" entry.
90265 * cpu-rx.c (arch_info_struct): Remove "rx" entry.
90266 * cpu-nds32.c (bfd_nds32_arch): Fix printable name for bfd_mach_n1
90269 2021-09-15 Alan Modra <amodra@gmail.com>
90271 PR28328, dlltool ice
90273 * archive.c (bfd_ar_hdr_from_filesystem): Don't use bfd_set_input_error
90274 here, our caller will do that.
90276 2021-09-15 GDB Administrator <gdbadmin@sourceware.org>
90278 Automatic date update in version.in
90280 2021-09-14 Tom de Vries <tdevries@suse.de>
90282 [gdb/testsuite] Fix gdb_load_no_complaints with gnu-debuglink
90283 When running test-case gdb.dwarf2/dw2-ranges-psym-warning.exp with target
90284 board gnu-debuglink I run into:
90286 (gdb) file dw2-ranges-psym-warning^M
90287 Reading symbols from dw2-ranges-psym-warning...^M
90288 Reading symbols from .debug/dw2-ranges-psym-warning.debug...^M
90289 (gdb) FAIL: gdb.dwarf2/dw2-ranges-psym-warning.exp: No complaints
90292 Fix this by updating the regexp in gdb_load_no_complaints.
90294 Tested on x86_64-linux.
90296 2021-09-14 Tom de Vries <tdevries@suse.de>
90298 [gdb/symtab] Fix function range handling in psymtabs
90299 Consider the test-case from this patch.
90303 (gdb) PASS: gdb.dwarf2/dw2-ranges-psym-warning.exp: continue
90305 warning: (Internal error: pc 0x4004b6 in read in psymtab, but not in symtab.)^M
90307 warning: (Internal error: pc 0x4004b6 in read in psymtab, but not in symtab.)^M
90309 warning: (Internal error: pc 0x4004b6 in read in psymtab, but not in symtab.)^M
90311 warning: (Internal error: pc 0x4004b6 in read in psymtab, but not in symtab.)^M
90313 warning: (Internal error: pc 0x4004b6 in read in psymtab, but not in symtab.)^M
90315 warning: (Internal error: pc 0x4004b6 in read in psymtab, but not in symtab.)^M
90317 read in psymtab, but not in symtab.)^M
90320 (gdb) FAIL: gdb.dwarf2/dw2-ranges-psym-warning.exp: bt
90323 This happens as follows.
90327 <1><31>: Abbrev Number: 4 (DW_TAG_subprogram)
90328 <33> DW_AT_name : foo
90329 <37> DW_AT_ranges : 0x0
90333 00000000 00000000004004c1 00000000004004d2
90334 00000000 00000000004004ae 00000000004004af
90335 00000000 <End of list>
90337 which have a hole at at [0x4004af,0x4004c1).
90339 However, the address map of the partial symtabs incorrectly maps addresses
90340 in the hole (such as 0x4004b6 in the backtrace) to the foo CU.
90342 The address map of the full symbol table of the foo CU however does not
90343 contain the addresses in the hole, which is what the warning / internal error
90346 Fix this by making sure that ranges of functions are read correctly.
90348 The patch adds a bit to struct partial_die_info, in this hole (shown for
90351 /* 11: 7 | 4 */ unsigned int canonical_name : 1;
90352 /* XXX 4-byte hole */
90353 /* 16 | 8 */ const char *raw_name;
90355 So there's no increase in size for 64-bit, but AFAIU there will be an increase
90358 Tested on x86_64-linux.
90362 2021-08-10 Tom de Vries <tdevries@suse.de>
90365 * dwarf2/read.c (struct partial_die_info): Add has_range_info and
90366 range_offset field.
90367 (add_partial_subprogram): Handle pdi->has_range_info.
90368 (partial_die_info::read): Set pdi->has_range_info.
90370 gdb/testsuite/ChangeLog:
90372 2021-08-10 Tom de Vries <tdevries@suse.de>
90375 * gdb.dwarf2/dw2-ranges-psym-warning-main.c: New test.
90376 * gdb.dwarf2/dw2-ranges-psym-warning.c: New test.
90377 * gdb.dwarf2/dw2-ranges-psym-warning.exp: New file.
90379 2021-09-14 Tom de Vries <tdevries@suse.de>
90381 [gdb/symtab] Fix CU list in .debug_names for dummy CUs
90382 With current trunk and target board cc-with-debug-names we have:
90384 (gdb) file dw2-ranges-psym^M
90385 Reading symbols from dw2-ranges-psym...^M
90386 warning: Section .debug_names in dw2-ranges-psym has abbreviation_table of \
90387 size 1 vs. written as 28, ignoring .debug_names.^M
90388 (gdb) set complaints 0^M
90389 (gdb) FAIL: gdb.dwarf2/dw2-ranges-psym.exp: No complaints
90392 The executable has 8 compilation units:
90394 $ readelf -wi dw2-ranges-psym | grep @
90395 Compilation Unit @ offset 0x0:
90396 Compilation Unit @ offset 0x2e:
90397 Compilation Unit @ offset 0xa5:
90398 Compilation Unit @ offset 0xc7:
90399 Compilation Unit @ offset 0xd2:
90400 Compilation Unit @ offset 0x145:
90401 Compilation Unit @ offset 0x150:
90402 Compilation Unit @ offset 0x308:
90404 of which the ones at 0xc7 and 0x145 are dummy CUs (that is, they do not
90405 contain a single DIE), which were added by recent commit 5ef670d81fd
90406 "[gdb/testsuite] Add dummy start and end CUs in dwarf assembly".
90408 The .debug_names section contains this CU table:
90419 The last two entries are incorrect, and the entries for the dummy CUs are
90422 The last two entries are incorrect because here in write_debug_names we write
90423 the dimension of the CU list as 8:
90425 /* comp_unit_count - The number of CUs in the CU list. */
90426 header.append_uint (4, dwarf5_byte_order,
90427 per_objfile->per_bfd->all_comp_units.size ()
90428 - per_objfile->per_bfd->tu_stats.nr_tus);
90430 while the actual dimension of the CU list is 6.
90432 The discrepancy is caused by this code which skips the dummy CUs:
90434 for (int i = 0; i < per_objfile->per_bfd->all_comp_units.size (); ++i)
90437 /* CU of a shared file from 'dwz -m' may be unused by this main
90438 file. It may be referenced from a local scope but in such
90439 case it does not need to be present in .debug_names. */
90440 if (psymtab == NULL)
90443 because they have a null partial symtab.
90445 We can fix this by writing the actual dimension of the CU list, but that still
90446 leaves the dummy CUs out of the CU list. The purpose of having these is to
90447 delimit the end of preceding CUs.
90450 - removing the code that skips the dummy CUs (note that the same change
90451 was done for .gdb_index in commit efba5c2319d '[gdb/symtab] Handle PU
90452 without import in "save gdb-index"'.
90453 - verifying that all units are represented in the CU/TU lists
90454 - using the actual CU list size when writing the dimension of the CU list
90455 (and likewise for the TU list).
90457 Tested on x86_64-linux with native and target board cc-with-debug-names.
90459 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28261
90461 2021-09-14 Tom de Vries <tdevries@suse.de>
90463 [gdb/testsuite] Generate .debug_aranges in gdb.dwarf2/locexpr-data-member-location.exp
90464 When running test-case gdb.dwarf2/locexpr-data-member-location.exp with target
90465 board cc-with-debug-names, all tests pass but we run into PR28261:
90468 Starting program: locexpr-data-member-location ^M
90469 warning: Section .debug_names in locexpr-data-member-location-lib.so has \
90470 abbreviation_table of size 1 vs. written as 37, ignoring .debug_names.^M
90473 Using a patch that fixes PR28261, the warning is gone, but we run into:
90475 FAIL: gdb.dwarf2/locexpr-data-member-location.exp: step into foo
90478 This is due a missing .debug_aranges contribution for the CU declared in
90479 gdb.dwarf2/locexpr-data-member-location.exp.
90481 Fix this by adding the missing .debug_aranges contribution.
90483 Tested on x86_64-linux.
90485 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
90487 arc: Fix potential invalid pointer access when fixing got symbols.
90488 When statically linking, it can arrive to an undefined weak symbol of
90489 which its value cannot be determined. However, we are having pieces of
90490 code which doesn't take this situation into account, leading to access
90491 a structure which may not be initialized. Fix this situation and add a
90495 xxxx-xx-xx Cupertino Miranda <cmiranda@synopsys.com>
90496 Claudiu Zissulescu <claziss@synopsys.com>
90498 * arc-got.h (arc_static_sym_data): New structure.
90499 (get_static_sym_data): New function.
90500 (relocate_fix_got_relocs_for_got_info): Move the computation fo
90501 symbol value and section to above introduced function, and use
90505 xxxx-xx-xx Claudiu Zissulescu <claziss@synopsys.com>
90507 * ld-arc/got-weak.d: New file.
90508 * ld-arc/got-weak.s: Likewise.
90513 2021-09-14 Mike Frysinger <vapier@gentoo.org>
90515 sim: bfin: add support for SDL2
90516 This probably should have been ported long ago, but better late than
90517 never. We keep support for both versions for now since both projects
90518 tend to have long lifetimes. Maybe consider dropping SDL1 in another
90521 2021-09-14 GDB Administrator <gdbadmin@sourceware.org>
90523 Automatic date update in version.in
90525 2021-09-14 Tom Tromey <tom@tromey.com>
90527 Remove use of __CYGNUSCLIB__
90528 I found a check of __CYGNUSCLIB__ in dbxread.c. I think this is dead
90529 code. This patch removes it.
90531 2021-09-13 Tom de Vries <tdevries@suse.de>
90533 [gdb/testsuite] Check for valid test name
90534 When running gdb.base/batch-exit-status.exp I noticed that the test name
90535 contains a newline:
90537 PASS: gdb.base/batch-exit-status.exp: : No such file or directory\.^M
90538 : No such file or directory\.: [lindex $result 2] == 0
90541 Check for this in ::CheckTestNames::check, such that we have a warning:
90543 PASS: gdb.base/batch-exit-status.exp: : No such file or directory\.^M
90544 : No such file or directory\.: [lindex $result 2] == 0
90545 WARNING: Newline in test name
90548 Tested on x86_64-linux.
90550 2021-09-13 Tom de Vries <tdevries@suse.de>
90552 [gdb/tdep] Fix exec check in gdb_print_insn_arm
90553 With a gdb build with --enable-targets=all we run into a KFAIL:
90555 KFAIL: gdb.gdb/unittest.exp: executable loaded: maintenance selftest, \
90556 failed none (PRMS: gdb/27891)
90560 Running selftest print_one_insn.^M
90561 Self test failed: arch armv8.1-m.main: self-test failed at \
90562 disasm-selftests.c:165^M
90565 The test fails because we expect disassembling of one arm insn to consume 4
90566 bytes and produce (using verbose = true in disasm-selftests.c):
90570 but instead the disassembler uses thumb mode and only consumes 2
90571 bytes and produces:
90576 The failure does not show up in the "no executable loaded" variant because
90577 this code in gdb_print_insn_arm isn't triggered:
90579 if (current_program_space->exec_bfd () != NULL)
90580 info->flags |= USER_SPECIFIED_MACHINE_TYPE;
90582 and consequently we do this in print_insn:
90584 if ((info->flags & USER_SPECIFIED_MACHINE_TYPE) == 0)
90585 info->mach = bfd_mach_arm_unknown;
90587 and don't set force_thumb to true in select_arm_features.
90589 The code in gdb_print_insn_arm makes the assumption that the disassembly
90590 architecture matches the exec architecture, which in this case is incorrect,
90591 because the exec architecture is x86_64, and the disassembly architecture is
90592 armv8.1-m.main. Fix that by explicitly checking it:
90594 if (current_program_space->exec_bfd () != NULL
90595 && (current_program_space->exec_bfd ()->arch_info
90596 == gdbarch_bfd_arch_info (gdbarch)))
90599 This fixes the print_one_insn failure, so remove the KFAIL.
90601 Tested on x86_64-linux.
90603 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=27891
90605 2021-09-13 Tom de Vries <tdevries@suse.de>
90607 [gdb/tdep] Reset force_thumb in parse_arm_disassembler_options
90608 With a gdb build with --enable-targets=all, we have 2 arch-specific failures
90609 in selftest print_one_insn:
90611 $ gdb -q -batch a.out -ex "maint selftest print_one_insn" 2>&1 \
90612 | grep "Self test failed: arch "
90613 Self test failed: arch armv8.1-m.main: self-test failed at \
90614 disasm-selftests.c:165
90615 Self test failed: arch arm_any: self-test failed at disasm-selftests.c:165
90619 During the first failed test, force_thumb is set to true, and remains so until
90620 and during the second test, which causes the second failure.
90622 Fix this by resetting force_thumb to false in parse_arm_disassembler_options,
90623 such that we get just one failure:
90625 $ gdb -q -batch a.out -ex "maint selftest print_one_insn" 2>&1 \
90626 | grep "Self test failed: arch "
90627 Self test failed: arch armv8.1-m.main: self-test failed at \
90628 disasm-selftests.c:165
90632 Tested on x86_64-linux.
90634 2021-09-13 Tom Tromey <tromey@adacore.com>
90636 Fix no-Python build
90637 A build without Python will currently fail, because
90638 selftests::test_python uses gdb_python_initialized, which is only
90639 conditionally defined.
90641 This patch fixes the build by making test_python also be conditionally
90642 defined. I chose this approach because the selftest will fail if
90643 Python is not enabled, so it didn't seem useful to leave it defined.
90645 2021-09-13 Nelson Chu <nelson.chu@sifive.com>
90647 RISC-V: Update the assembler insn testcase.
90648 Since the 0x57 is preserved for the vadd.vv instruction in the integration
90649 branch, remove it to make sure the testcase can work.
90652 * testsuite/gas/riscv/insn.d: Remove 0x57 since it is preserved
90653 for vadd.vv instruction.
90654 * testsuite/gas/riscv/insn.s: Likewise.
90656 2021-09-13 Jan Beulich <jbeulich@suse.com>
90658 MIPS: don't use get_symbol_name() for section parsing. With s_change_section() later calling obj_elf_section(), it seems better to pre-parse the section name by the same function that will be used there. This way no differences in what is accepted will result.
90659 gas * config/tc-mips.c (s_change_section): Use obj_elf_section_name to
90660 parse the section name.
90662 ia64: don't use get_symbol_name() for section parsing. With cross_section() later calling obj_elf_section(), it seems better to pre-parse the section name by the same function that will be used there. This way no differences in what is accepted will result.
90663 gas * config/tc-ia64.c (cross_section): Use obj_elf_section_name to
90664 parse the section name.
90666 2021-09-13 Tom de Vries <tdevries@suse.de>
90668 [gdb/testsuite] Fix gdb.gdb/selftest.exp
90669 With a gdb build with CFLAGS "-O2 -g -flto=auto", I run into:
90671 #7 gdb_main (args=0x7fffffffd220) at src/gdb/main.c:1368^M
90672 #8 main (argc=<optimized out>, argv=<optimized out>) at src/gdb/gdb.c:32^M
90673 (gdb) FAIL: gdb.gdb/selftest.exp: backtrace through signal handler
90675 which means that this regexp in proc test_with_self fails:
90677 -re "#0.*(read|poll).*in main \\(.*\\) at .*gdb\\.c.*$gdb_prompt $" {
90680 The problem is that gdb_main has been inlined into main, and consequently the
90687 #x <address> in <fn> ...
90690 Fix this by updating the regexp to not require "in" before " main".
90692 Tested on x86_64-linux.
90694 2021-09-13 Alan Modra <amodra@gmail.com>
90696 Re: Deprecate a.out support for NetBSD targets
90697 * config.bfd: Correct m68-*-*bsd* obsolete target match.
90699 2021-09-13 Tom de Vries <tdevries@suse.de>
90701 [gdb/testsuite] Fix test name in gdb.base/batch-exit-status.exp
90702 When running gdb.base/batch-exit-status.exp I noticed that the test name
90703 contains a newline:
90705 PASS: gdb.base/batch-exit-status.exp: : No such file or directory\.^M
90706 : No such file or directory\.: [lindex $result 2] == 0
90709 The mistake is that I passed an output regexp argument to a parameter
90710 interpreted as testname prefix. Fix this by passing a testname prefix
90713 Add support for checking output, to be able to handle the output regexp
90716 Tested on x86_64-linux.
90718 2021-09-13 GDB Administrator <gdbadmin@sourceware.org>
90720 Automatic date update in version.in
90722 2021-09-12 Tom de Vries <tdevries@suse.de>
90724 [gdb/testsuite] Set sysroot earlier in local-board.exp
90725 When running test-case gdb.base/batch-exit-status.exp for native, it passes.
90726 But with target board cc-with-debug-names, we run into (added missing double
90727 quotes for clarity):
90729 builtin_spawn $build/gdb/testsuite/../../gdb/gdb -nw -nx \
90730 -data-directory $build/gdb/testsuite/../data-directory \
90731 -iex "set height 0" -iex "set width 0" -ex "set sysroot" -batch ""^M
90732 : No such file or directory.^M
90733 PASS: gdb.base/batch-exit-status.exp: \
90734 : No such file or directory\.: [lindex $result 2] == 0
90735 FAIL: gdb.base/batch-exit-status.exp: \
90736 : No such file or directory\.: [lindex $result 3] == $expect_status
90739 The difference between the passing and failing case is that with native we
90740 have (leaving out set height/width for brevity):
90742 $ gdb -batch ""; echo $?
90743 : No such file or directory.
90746 and with target board cc-with-debug-names:
90748 $ gdb -ex "set sysroot" -batch ""; echo $?
90749 : No such file or directory.
90753 The difference is expected. GDB returns the exit status of the last executed
90754 command. In the former case that's 'file ""', which fails. In the latter case,
90755 that's 'set sysroot', which succeeds.
90757 Fix this by setting sysroot using -iex instead of -ex in local-board.exp, such
90758 that we have the expected:
90760 $ gdb -iex "set sysroot" -batch ""; echo $?
90761 : No such file or directory.
90765 Tested on x86_64-linux.
90767 2021-09-12 GDB Administrator <gdbadmin@sourceware.org>
90769 Automatic date update in version.in
90771 2021-09-11 Mike Frysinger <vapier@gentoo.org>
90773 sim: run: change help short option to -h
90774 It's unclear why -H was picked over the more standard -h, but since
90775 -h is still not used, just change -H to -h to match pretty much every
90776 other tool in the sourceware tree.
90778 2021-09-11 GDB Administrator <gdbadmin@sourceware.org>
90780 Automatic date update in version.in
90782 2021-09-10 Tom de Vries <tdevries@suse.de>
90784 [gdb/testsuite] Reimplement gdb.gdb/python-selftest.exp as unittest
90785 The test-case gdb.gdb/python-selftest.exp:
90786 - patches the gdb_python_initialized variable in gdb to 0
90787 - checks that the output of a python command is "Python not initialized"
90789 Reimplement gdb.gdb/python-selftest.exp as unittest, using:
90790 - execute_command_to_string to capture the output
90791 - try/catch to catch the "Python not initialized" exception.
90793 Tested on x86_64-linux.
90795 2021-09-10 Tom de Vries <tdevries@suse.de>
90797 [gdb/testsuite] Fix DUPLICATE in gdb.base/global-var-nested-by-dso.exp
90798 Fix DUPLICATE in gdb.base/global-var-nested-by-dso.exp by naming commands more
90801 2021-09-10 Tom de Vries <tdevries@suse.de>
90803 [gdb/testsuite] Fix DUPLICATE in gdb.base/skip-solib.exp
90804 Fix DUPLICATE in gdb.base/skip-solib.exp by using with_test_prefix.
90806 Also fix indentation style and long lines, remove outdated question/answer
90807 bits, and use multi_line.
90809 2021-09-10 Tom de Vries <tdevries@suse.de>
90811 [gdb/testsuite] Fix handling of nr_args < 3 in mi_gdb_test
90812 The documentation of mi_gdb_test states that the command, pattern and message
90813 arguments are mandatory:
90815 # mi_gdb_test COMMAND PATTERN MESSAGE [IPATTERN] -- send a command to gdb;
90819 However, this is not checked, and when mi_gdb_test is called with less than 3
90820 arguments, it passes or fails silently.
90822 Fix this by using the following semantics:
90823 - if there are 1 or 2 arguments, use the command as the message.
90824 - if there is 1 argument, use ".*" as the pattern.
90825 - if there are no or too much arguments, error out.
90827 Fix a PATH issue in gdb.mi/mi-logging.exp, introduced by using the command as
90828 message. Fix a few other trivial-looking FAILs.
90830 There are 11 less trivial-looking FAILs left in gdb.mi in test-cases:
90831 - mi-nsmoribund.exp
90832 - mi-breakpoint-changed.exp
90835 Tested on x86_64-linux.
90837 2021-09-10 Tom de Vries <tdevries@suse.de>
90839 [gdb/testsuite] Add string_list_to_regexp
90840 A regexp pattern with escapes like this is hard to read:
90842 set re "~\"\[$\]$decimal = 1\\\\n\"\r\n\\^done"
90845 We can make it more readable by spacing out parts (which allows us to also use
90846 the curly braces where that's convenient):
90848 set re [list "~" {"} {[$]} $decimal " = 1" "\\\\" "n" {"} "\r\n" "\\^" "done"]
90849 set re [join $re ""]
90851 or by using string_to_regexp:
90854 [string_to_regexp {~"$}] \
90856 [string_to_regexp " = 1\\n\"\r\n^done"]]
90857 set re [join $re ""]
90859 Note: we have to avoid applying string_to_list to decimal, which is already a
90862 Add a proc string_list_to_regexp to make it easy to do both:
90865 [string_list_to_regexp ~ {"} $] \
90867 [string_list_to_regexp " = 1" \\ n {"} \r\n ^ done]]
90870 Also add a test-case gdb.testsuite/string_to_regexp.exp.
90872 2021-09-10 Tom de Vries <tdevries@suse.de>
90874 [gdb/testsuite] Handle unrecognized command line option in gdb_compile_test
90875 When running the gdb testsuite with gnatmake-4.8, I get many fails of the
90878 gcc: error: unrecognized command line option '-fgnat-encodings=all'^M
90879 gnatmake: "gdb.ada/O2_float_param/foo.adb" compilation error^M
90880 compiler exited with status 1
90881 compilation failed: gcc ... gdb.ada/O2_float_param/foo.adb
90882 gcc: error: unrecognized command line option '-fgnat-encodings=all'
90883 gnatmake: "gdb.ada/O2_float_param/foo.adb" compilation error
90884 FAIL: gdb.ada/O2_float_param.exp: scenario=all: compilation foo.adb
90887 Fix this by marking the test unsupported instead, such that we have:
90889 UNSUPPORTED: gdb.ada/O2_float_param.exp: scenario=all: compilation foo.adb \
90890 (unsupported option '-fgnat-encodings=all')
90893 Tested on x86_64-linux.
90895 2021-09-10 Alan Modra <amodra@gmail.com>
90897 PowerPC, sanity check r_offset in relocate_section
90898 * elf32-ppc.c (offset_in_range): New function.
90899 (ppc_elf_vle_split16): Sanity check r_offset before accessing
90900 section contents. Return status.
90901 (ppc_elf_relocate_section): Sanity check r_offset before
90902 accessing section contents. Don't segfault on NULL howto.
90904 Re: gas: Use the directory name in .file 0
90906 * testsuite/gas/elf/dwarf-5-file0-2.s: Use %object rather than
90907 @object, .4byte instead of .long, and .asciz instead of .string.
90909 2021-09-10 Mike Frysinger <vapier@gentoo.org>
90911 etc: switch to automake
90912 There's no content in here currently, so switching to automake is
90913 pretty easy with a stub file.
90915 etc: rename configure.in to configure.ac
90916 The .in name has been deprecated for a long time in favor of .ac.
90918 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
90920 gas: Use the directory name in .file 0
90921 DWARF5 allows .file 0 to take an optional directory name. Set the entry
90922 0 of the directory table to the directory name in .file 0.
90925 * dwarf2dbg.c (get_directory_table_entry): Add an argument for
90926 the directory name in .file 0 and use it, instead of PWD.
90927 (allocate_filenum): Pass NULL to get_directory_table_entry.
90928 (allocate_filename_to_slot): Pass the incoming dirname to
90929 get_directory_table_entry.
90930 * testsuite/gas/elf/dwarf-5-file0-2.d: New file.
90931 * testsuite/gas/elf/dwarf-5-file0-2.s: Likewise.
90932 * testsuite/gas/elf/elf.exp: Run dwarf-5-file0-2.
90934 2021-09-10 GDB Administrator <gdbadmin@sourceware.org>
90936 Automatic date update in version.in
90938 2021-09-09 Yoshinori Sato <ysato@users.sourceforge.jp>
90940 gdb: Enable target rx-*-*linux.
90941 I added rx-*-linux in binutils few yaers ago.
90942 But missing this changes,
90944 2021-09-09 Tom de Vries <tdevries@suse.de>
90946 [gdb/testsuite] Fix gdb.base/coredump-filter-build-id.exp with older eu-unstrip
90947 On openSUSE Leap 42.3 with eu-unstrip 0.158, we run into:
90949 (gdb) PASS: gdb.base/coredump-filter-build-id.exp: save corefile
90950 First line of eu-unstrip: \
90951 0x400000+0x202000 f4ae8502bd6a14770182382316bc595e9dc6f08b@0x400284 - - [exe]
90952 FAIL: gdb.base/coredump-filter-build-id.exp: gcore dumped mapping with build-id
90955 The test expects an actual file name instead of '[exe]', but that only got
90956 introduced with eu-unstrip 0.161. Before it printed '[exe]' or '[pie]'.
90958 Fix this by updating the regexp.
90960 Tested on x86_64-linux.
90962 2021-09-09 Tom de Vries <tdevries@suse.de>
90964 [gdb/testsuite] Fix various issues in gdb.mi/mi-sym-info.exp
90965 I noticed this failure in gdb.mi/mi-sym-info.exp with gcc-4.8:
90967 FAIL: gdb.mi/mi-sym-info.exp: -symbol-info-functions --max-results 1 \
90968 (unexpected output)
90970 due to function f2 instead of f3 being listed.
90972 AFAICT, this is caused by a difference in debug info:
90974 $ readelf -wi outputs/gdb.mi/mi-sym-info/mi-sym-info1.o \
90975 | egrep "DW_AT_name.*: f[1-3]"
90976 <72> DW_AT_name : f1
90977 <a1> DW_AT_name : f2
90978 <d0> DW_AT_name : f3
90982 $ readelf -wi outputs/gdb.mi/mi-sym-info/mi-sym-info1.o \
90983 | egrep "DW_AT_name.*: f[1-3]"
90984 <f4> DW_AT_name : f3
90985 <123> DW_AT_name : f2
90986 <152> DW_AT_name : f1
90988 and the command documentation does not mention an imposed order, so fix this
90989 by allowing f2 as well.
90991 Doing this fix, it made sense to do a refactoring of adding f2_re and f3_re
90992 variables, in order to write (?:$f2_re|$f3_re), and I applied the same pattern
90995 Furthermore, I found a silent FAIL due to calling mi_gdb_proc with 2 args, fix
90996 by updating the regexp.
90998 Then I ran with clang and found another FAIL, fix by updating the regexp.
91000 Tested on x86_64-linux with gcc-4.8.5, gcc-7.5.0, gcc-11.2.1, clang-7.0.1 and
91003 2021-09-09 Tom de Vries <tdevries@suse.de>
91005 [gdb/testsuite] Reimplement gdb.gdb/complaints.exp as unittest
91006 When building gdb with "-Wall -O2 -g -flto=auto", I run into:
91008 (gdb) call clear_complaints()^M
91009 No symbol "clear_complaints" in current context.^M
91010 (gdb) FAIL: gdb.gdb/complaints.exp: clear complaints
91013 The problem is that lto has optimized away the clear_complaints function
91014 and consequently the selftest doesn't work.
91016 Fix this by reimplementing the selftest as a unit test.
91018 Factor out two new functions:
91020 execute_fn_to_ui_file (struct ui_file *file, std::function<void(void)> fn);
91022 execute_fn_to_string (std::function<void(void)> fn, bool term_out);
91023 and use the latter to capture the complaints output.
91025 Tested on x86_64-linux.
91027 2021-09-09 Andrew Burgess <andrew.burgess@embecosm.com>
91029 gdb/python: remove all uses of Py_TPFLAGS_HAVE_ITER
91030 Python 2 has a bit flag Py_TPFLAGS_HAVE_ITER which can be passed as
91031 part of the tp_flags field when defining a new object type. This flag
91032 is not defined in Python 3 and so we define it to 0 in
91033 python-internal.h (when IS_PY3K is defined).
91035 The meaning of this flag is that the object has the fields tp_iter and
91036 tp_iternext. Note the use of "has" here, the flag says nothing about
91037 the values in those fields, just that the type object has the fields.
91039 In early versions of Python 2 these fields were no part of the
91040 PyTypeObject struct, they were added in version 2.2 (see
91041 https://docs.python.org/release/2.3/api/type-structs.html). And so,
91042 there could be a some code compiled out there which has a PyTypeObject
91043 structure within it that doesn't even have the tp_iter and tp_iternext
91044 fields, attempting to access these fields would be undefined
91047 And so Python added the Py_TPFLAGS_HAVE_ITER flag. If the flag is
91048 present then Python is free to access the tp_iter and tp_iternext
91051 If we consider GDB then we always assume that the tp_iter and
91052 tp_iternext fields are part of PyTypeObject. If someone was crazy
91053 enough to try and compile GDB against Python 2.1 then we'd get lots of
91054 build errors saying that we were passing too many fields when
91055 initializing PyTypeObject structures. And so, I claim, we can be sure
91056 that GDB will always be compiled with a version of Python that has the
91057 tp_iter and tp_iternext fields in PyTypeObject.
91059 Next we can look at the Py_TPFLAGS_DEFAULT flag. In Python 2, each
91060 time additional fields are added to PyTypeObject a new Py_TPFLAGS_*
91061 flag would be defined to indicate whether those flags are present or
91062 not. And, those new flags would be added to Py_TPFLAGS_DEFAULT. And
91063 so, in the latest version of Python 2 the Py_TPFLAGS_DEFAULT flag
91064 includes Py_TPFLAGS_HAVE_ITER (see
91065 https://docs.python.org/2.7/c-api/typeobj.html).
91067 In GDB we pass Py_TPFLAGS_DEFAULT as part of the tp_flags for all
91070 And so, in this commit, I propose to remove all uses of
91071 Py_TPFLAGS_HAVE_ITER from GDB, it's simply not needed.
91073 There should be no user visible changes after this commit.
91075 2021-09-09 Mike Frysinger <vapier@gentoo.org>
91077 sim: accept -EB/-EL short options
91078 Many GNU tools accept -EB/-EL as short options for selecting big &
91079 little endian modes. While the sim has an -E option, it requires
91080 spelling out "big" and "little". Adding support for -EB & -EL is
91081 thus quite trivial, so lets round it out to be less annoying.
91083 sim: ppc: drop support for std-config.h overrides
91084 Only the ppc arch supports this kind of source file override logic.
91085 All the others expose knobs via configure flags, and for some of
91086 these, the ppc code does as well. For others, it doesn't make sense
91087 to ever change them. Since it's unlikely anyone is using this, drop
91088 it all to simplify the code (and to get us a little closer to the
91091 sim: ppc: enable use of gnulib
91092 All other sim arches are using this now, so finish up the logic in
91093 the ppc arch to enable gnulib usage here too.
91095 sim: drop old O_NDELAY & FNBLOCK support
91096 We use these older names inconsistently in the sim codebase, and time
91097 has moved on long ago, so drop support for these non-standard names.
91098 POSIX provides O_NONBLOCK for us, so use it everywhere.
91100 2021-09-09 Mike Frysinger <vapier@gentoo.org>
91102 sim: dv-sockser: enable for mingw targets too
91103 We have enough functionality from gnulib now to build sockser on
91106 Non-blocking I/O is supported when F_GETFL/F_SETFL are unavailable,
91107 but we can address that in a follow up commit. This mirrors what
91108 is done in other places in the sim already.
91110 2021-09-09 Mike Frysinger <vapier@gentoo.org>
91112 sim: cgen: workaround Windows VOID define
91113 The cgen framework provides a "VOID" type for code to use, but this
91114 defines ends up conflicting with the standard Windows VOID define.
91115 Since they actually define to the same thing ("void"), undef it here
91116 to fix the Windows build.
91118 We might want to reconsider the need for "VOID" in cgen, but that
91119 will take larger discussion & coordination with the cgen project.
91121 2021-09-09 Mike Frysinger <vapier@gentoo.org>
91123 sim: dv-sockser: move sim-main.h include after system includes
91124 The sim-main.h header is a bit of a dumping ground. Every arch can
91125 (and many do) define all sorts of weird & common names that end up
91126 conflicting with system headers. So including it before the system
91127 headers sets us up for pain. v850 is a good example of this -- when
91128 building for mingw, we see weird failures:
91130 $ i686-w64-mingw32-gcc ... -c -o dv-sockser.o ../../../../sim/v850/../common/dv-sockser.c
91131 In file included from ../../../../sim/v850/sim-main.h:11,
91132 from ../../../../sim/v850/../common/dv-sockser.c:24:
91133 ../../../../sim/v850/../common/sim-base.h:97:32: error: expected ')' before '->' token
91134 97 | # define STATE_CPU(sd, n) ((sd)->cpu[0])
91137 While gcc is unhelpful at first, running it through the preprocessor
91138 by hand shows more details:
91140 $ i686-w64-mingw32-gcc ... -E -dD -o dv-sockser.i ../../../../sim/v850/../common/dv-sockser.c
91141 $ i686-w64-mingw32-gcc -c dv-sockser.i
91142 In file included from /usr/i686-w64-mingw32/usr/include/minwindef.h:163,
91143 from /usr/i686-w64-mingw32/usr/include/windef.h:9,
91144 from /usr/i686-w64-mingw32/usr/include/windows.h:69,
91145 from /usr/i686-w64-mingw32/usr/include/winsock2.h:23,
91146 from ../../gnulib/import/sys/socket.h:684,
91147 from ../../gnulib/import/netinet/in.h:43,
91148 from ../../../../sim/v850/../common/dv-sockser.c:39:
91149 /usr/i686-w64-mingw32/usr/include/winnt.h:4803:25: error: expected ')' before '->' token
91150 4803 | DWORD State;
91154 This is because v850 sets up this common name:
91156 All of this needs cleaning up someday, but since the dv-sockser code
91157 definitely should be fixed in this way, lets do that now and unblock
91160 2021-09-09 Mike Frysinger <vapier@gentoo.org>
91162 sim: mips: delete unused PSIZE define
91163 It's unclear what this define is for as it appears to be unused, and
91164 has never been used in the history of the mips sim. Delete it to tidy
91165 up, and to fix build errors for Windows targets that have a standard
91166 "PSIZE" struct in their system headers. This doesn't show up yet as
91167 most sim files don't include many system headers, but enabling sockser
91168 code for mingw uncovers the conflict.
91170 Unfortunately the error produced by gcc is inscrutable, but running
91171 it through the preprocessor manually manages to provide a pointer to
91172 the underlying issue.
91174 $ i686-w64-mingw32-gcc ... -c -o dv-sockser.o ../../../../sim/mips/../common/dv-sockser.c
91175 <command-line>: error: expected identifier or '(' before numeric constant
91176 In file included from /usr/i686-w64-mingw32/usr/include/windows.h:71,
91177 from /usr/i686-w64-mingw32/usr/include/winsock2.h:23,
91178 from ../../gnulib/import/sys/socket.h:684,
91179 from ../../gnulib/import/netinet/in.h:43,
91180 from ../../../../sim/mips/../common/dv-sockser.c:39:
91181 /usr/i686-w64-mingw32/usr/include/wingdi.h:2934:59: error: unknown type name 'LPSIZE'; did you mean 'LPSIZEL'?
91182 2934 | WINGDIAPI WINBOOL WINAPI GetAspectRatioFilterEx(HDC hdc,LPSIZE lpsize);
91187 $ i686-w64-mingw32-gcc ... -E -dD -o dv-sockser.i ../../../../sim/mips/../common/dv-sockser.c
91188 $ i686-w64-mingw32-gcc -c dv-sockser.i
91189 In file included from /usr/i686-w64-mingw32/usr/include/windows.h:69,
91190 from /usr/i686-w64-mingw32/usr/include/winsock2.h:23,
91191 from ../../gnulib/import/sys/socket.h:684,
91192 from ../../gnulib/import/netinet/in.h:43,
91193 from ../../../../sim/mips/../common/dv-sockser.c:39:
91194 /usr/i686-w64-mingw32/usr/include/windef.h:104:9: error: expected identifier or '(' before numeric constant
91195 104 | } SIZE,*PSIZE,*LPSIZE;
91198 2021-09-09 Mike Frysinger <vapier@gentoo.org>
91200 sim: ppc: switch to common warning flags
91201 Now that the ppc code has been cleaned up enough to use the same set
91202 of warning flags as the common code, delete the ppc-specific configure
91203 logic so we can leverage what the common code already defined for us.
91205 2021-09-09 Tom de Vries <tdevries@suse.de>
91207 sim: ppc: enable -Wpointer-sign warnings
91208 When compiling with --enable-werror and CFLAGS="-O0 -g -Wall", we run into:
91210 src/sim/ppc/hw_memory.c: In function 'hw_memory_init_address':
91211 src/sim/ppc/hw_memory.c:204:7: error: pointer targets in passing argument 4 \
91212 of 'device_find_integer_array_property' differ in signedness \
91213 [-Werror=pointer-sign]
91218 Fix these by adding an explicit pointer cast. It's a bit ugly to use APIs
91219 based on signed integers to read out unsigned values, but in practice, this
91220 is par for the course in the ppc code.
91222 We already use signed APIs and assign the result to unsigned values a lot:
91223 see how device_find_integer_property returns a signed integer (cell), but
91224 then assign it to unsigned types. The array APIs are not used that often
91225 which is why we don't see many warnings, and we disable warnings when we
91226 assign signed integers to unsigned integers in general.
91228 The dtc/libfdt project (which is the standard in other projects) processes
91229 the fdt blob as a series of bytes without any type information. Typing is
91230 left to the caller. They have core APIs that read/write bytes, and a few
91231 helper functions to cast/convert those bytes to the right value (e.g. u32).
91232 In this ppc sim code, the core APIs use signed integers, and the callers
91233 convert to unsigned, usually implicitly.
91235 We could add some core APIs to the ppc sim that deal with raw bytes and then
91236 add some helpers to convert to the right type, but that seems like a lot of
91237 lifting for what boils down to a cast, and is effectively equivalent to all
91238 the implicit assignments we use elsewhere. Long term, a lot of the ppc code
91239 should either get converted to existing sim common code, or we should stand
91240 up proper APIs in the common code first, or use standard libraries to do all
91241 the processing (e.g. libfdt). Either way, this device.c code would all get
91242 deleted, and callers (like these hw_*.c files) would get converted. Which
91243 is also why we go with a cast rather new (but largely unused) APIs.
91245 2021-09-09 Mike Frysinger <vapier@gentoo.org>
91247 sim: ppc: enable -Wmissing-declarations & -Wmissing-prototypes
91248 This aligns with common code which already uses this flag. We have
91249 to add another local prototype to fix the failure, and add another
91250 local decl for the SIM_DESC type. Unwinding these will require a
91251 lot more work & conversions in the process, so going with the decl
91252 for now unblocks the warning unification.
91254 2021-09-09 Mike Frysinger <vapier@gentoo.org>
91256 sim: microblaze: replace custom basic types with common ones
91257 The basic "byte" type conflicts with Windows headers, and we already
91258 have common types that provide the right sizes. So replace these with
91259 the common ones to avoid issues.
91262 In file included from /usr/i686-w64-mingw32/usr/include/wtypes.h:8,
91263 from /usr/i686-w64-mingw32/usr/include/winscard.h:10,
91264 from /usr/i686-w64-mingw32/usr/include/windows.h:97,
91265 from /usr/i686-w64-mingw32/usr/include/winsock2.h:23,
91266 from ../../gnulib/import/sys/socket.h:684,
91267 from ../../gnulib/import/netinet/in.h:43,
91268 from .../build/sim/../../../sim/microblaze/../common/dv-sockser.c:39:
91269 /usr/i686-w64-mingw32/usr/include/rpcndr.h:63:25: error: conflicting types for 'byte'; have 'unsigned char'
91270 63 | typedef unsigned char byte;
91272 In file included from .../buildsim/../../../sim/microblaze/sim-main.h:21,
91273 from .../buildsim/../../../sim/microblaze/../common/dv-sockser.c:24:
91274 .../buildsim/../../../sim/microblaze/microblaze.h:94:25: note: previous declaration of 'byte' with type 'byte' {aka 'char'}
91275 94 | typedef char byte;
91277 make: *** [Makefile:513: dv-sockser.o] Error 1
91279 2021-09-09 Jim Wilson <jimw@sifive.com>
91281 RISC-V: Pretty print values formed with lui and addiw.
91282 The disassembler has support to pretty print values created by an lui/addi
91283 pair, but there is no support for addiw. There is also no support for
91284 c.addi and c.addiw. This patch extends the pretty printing support to
91285 handle these 3 instructions in addition to addi. Existing testcases serve
91286 as tests for the new feature.
91289 * riscv-dis.c (maybe_print_address): New arg wide. Sign extend when
91291 (print_insn_args): Fix calls to maybe_print_address. Add checks for
91292 c.addi, c.addiw, and addiw, and call maybe_print_address for them.
91295 * testsuite/gas/riscv/insn.d: Update for disassembler change.
91296 * testsuite/gas/li32.d, testsuite/gas/li64.d: Likwise.
91297 * testsuite/gas/lla64.d: Likewise.
91299 2021-09-09 Mike Frysinger <vapier@gentoo.org>
91301 sim: ppc: align format string settings with common code
91302 This copies logic used in the common sim warning configure code to fix
91303 build errors for mingw targets. Turning format warnings on triggers
91304 a failure in the debug.c file, so apply a minor fix at the same time.
91306 sim: ppc: drop unnecessary config includes
91307 This file is compiled for the --host & --build system which leads to
91308 including the configure generated config.h in both environments.
91309 This obviously doesn't work when the two targets don't look alike at
91310 all and can cause build failures here (e.g. a mingw host & a linux
91311 build). Since we don't actually need any config settings in this
91312 very simple file, drop the includes entirely.
91314 2021-09-09 GDB Administrator <gdbadmin@sourceware.org>
91316 Automatic date update in version.in
91318 2021-09-08 Mike Frysinger <vapier@gentoo.org>
91320 gnulib: import various network functions
91321 Some sim ports use these to provide networking functionality via the
91322 dv-sockser module or via direct emulation for a few ports.
91324 Gdb seems to build just fine still too.
91326 2021-09-08 Tom Tromey <tromey@adacore.com>
91328 Fix unit test build on Windows
91329 Like Tom de Vries' earlier patch to fix the no-CXX_STD_THREAD case in
91330 maint.c, this patch fixes a similar problem in
91331 parallel-for-selftests.c. This fixes a build failure on Windows.
91333 2021-09-08 Alan Modra <amodra@gmail.com>
91335 PowerPC64, sanity check r_offset in relocate_section
91336 This hardens the powerpc64 linker code transformations.
91338 * elf64-ppc.c (is_8byte_reloc, offset_in_range): New functions.
91339 (ppc64_elf_relocate_section): Sanity check r_offset before
91340 accessing section contents for various code transformations.
91342 2021-09-08 Alan Modra <amodra@gmail.com>
91344 PowerPC64: Avoid useless work on R_PPC64_TPREL34
91345 _bfd_elf_ppc_at_tprel_transform doesn't handle prefix instructions,
91346 and I'm not inclined to implement code editing for them.
91348 * elf64-ppc.c (ppc64_elf_relocate_section): Don't attempt tprel
91349 transform for R_PPC64_TPREL34.
91351 2021-09-08 Andrew Burgess <andrew.burgess@embecosm.com>
91353 gdb: make thread_suspend_state::stop_pc optional
91354 Currently the stop_pc field of thread_suspect_state is a CORE_ADDR and
91355 when we want to indicate that there is no stop_pc available we set
91356 this field back to a special value.
91358 There are actually two special values used, in post_create_inferior
91359 the stop_pc is set to 0. This is a little unfortunate, there are
91360 plenty of embedded targets where 0 is a valid pc value. The more
91361 common special value for stop_pc though, is set in
91362 thread_info::set_executing, where the value (~(CORE_ADDR) 0) is used.
91364 This commit changes things so that the stop_pc is instead a
91365 gdb::optional. We can now explicitly reset the field to an
91366 uninitialised state, we also have asserts that we don't read the
91367 stop_pc when its in an uninitialised state (both in
91368 gdbsupport/gdb_optional.h, when compiling with _GLIBCXX_DEBUG
91369 defined, and in thread_info::stop_pc).
91371 One situation where a thread will not have a stop_pc value is when the
91372 thread is stopped as a consequence of GDB being in all stop mode, and
91373 some other thread stopped at an interesting event. When GDB brings
91374 all the other threads to a stop those other threads will not have a
91375 stop_pc set (thus avoiding an unnecessary read of the pc register).
91377 Previously, when GDB passed through handle_one (in infrun.c) the
91378 threads executing flag was set to false and the stop_pc field was left
91379 unchanged, i.e. it would (previous) have been left as ~0.
91381 Now, handle_one leaves the stop_pc with no value.
91383 This caused a problem when we later try to set these threads running
91384 again, in proceed() we compare the current pc with the cached stop_pc.
91385 If the thread was stopped via handle_one then the stop_pc would have
91386 been left as ~0, and the compare (in proceed) would (likely) fail.
91387 Now however, this compare tries to read the stop_pc when it has no
91388 value and this would trigger an assert.
91390 To resolve this I've added thread_info::stop_pc_p() which returns true
91391 if the thread has a cached stop_pc. We should only ever call
91392 thread_info::stop_pc() if we know that there is a cached stop_pc,
91393 however, this doesn't mean that every call to thread_info::stop_pc()
91394 needs to be guarded with a call to thread_info::stop_pc_p(), in most
91395 cases we know that the thread we are looking at stopped due to some
91396 interesting event in that thread, and so, we know that the stop_pc is
91399 After running the testsuite I've seen no other situations where
91400 stop_pc is read uninitialised.
91402 There should be no user visible changes after this commit.
91404 2021-09-08 Tom de Vries <tdevries@suse.de>
91406 [gdb/build] Fix build with undefined CXX_STD_THREAD
91407 When building gdb on openSUSE Leap 42.3, we trigger the case that
91408 CXX_STD_THREAD is undefined, and run into:
91410 gdb/maint.c: In function 'void maintenance_show_worker_threads \
91411 (ui_file*, int, cmd_list_element*, const char*)':
91412 gdb/maint.c:877:14: error: 'gdb::thread_pool' has not been declared
91413 gdb::thread_pool::g_thread_pool->thread_count ());
91415 Makefile:1647: recipe for target 'maint.o' failed
91416 make[1]: *** [maint.o] Error 1
91419 Fix this by handling the undefined CXX_STD_THREAD case in
91420 maintenance_show_worker_threads, such that we get:
91422 $ gdb -q -batch -ex "maint show worker-thread"
91423 The number of worker threads GDB can use is 0.
91426 Tested on x86_64-linux.
91428 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28312
91430 2021-09-08 Mike Frysinger <vapier@gentoo.org>
91432 sim: update configure target list
91433 Fix sorting of the list, and update the globs to match the list used
91434 in gdb's configure script.
91436 gdb: cris: enable sim integration
91437 The sim side is already ready to go for cris, so wire it up.
91439 gdb: aarch64: enable sim integration
91440 The sim side is already ready to go for aarch64, so wire it up.
91442 gdb: sim: consolidate configure settings
91443 Moving all the sim settings to one section makes it easier to track,
91444 and makes it easier to keep it aligned with the sim target tests.
91445 The gdb logic was duplicating this when handling different OS targets
91446 instead of having a single cpu check. Now it's more obvious that the
91447 sim is tied to a cpu and not related to the OS.
91449 2021-09-08 GDB Administrator <gdbadmin@sourceware.org>
91451 Automatic date update in version.in
91453 2021-09-07 Tom Tromey <tromey@adacore.com>
91455 Remove unused declaration from gdbserver/win32-low.h
91456 I noticed that gdbserver/win32-low.h has an unused declaration. This
91457 code was changed a while ago, but this declaration slipped through.
91458 This patch removes it. Tested by rebuilding.
91460 2021-09-07 Andrew Burgess <andrew.burgess@embecosm.com>
91462 gdb: make use of std::string in utils.c
91463 Replace some of the manual string management (malloc/free) with
91464 std::string when creating commands in utils.c.
91466 Things are a little bit messy as, creating the prefix commands (using
91467 add_basic_prefix_cmd and add_show_prefix_cmd), doesn't copy the doc
91468 string, while creating the actual set/show commands (using
91469 add_setshow_enum_cmd) does copy the doc string.
91471 As a result, I have retained the use of xstrprintf when creating the
91472 prefix command doc strings, but switched to using std::string when
91473 creating the actual set/show commands.
91475 There should be no user visible changes after this commit.
91477 2021-09-07 Luis Machado <luis.machado@linaro.org>
91479 Revert: [AArch64] MTE corefile support
91480 bfd * elf.c (elfcore_make_memtag_note_section): New function.
91481 (elfcore_grok_note): Handle NT_MEMTAG note types.
91483 binutils* readelf.c (get_note_type): Handle NT_MEMTAG note types.
91485 include * elf/common.h (NT_MEMTAG): New constant.
91486 (NT_MEMTAG_TYPE_AARCH_MTE): New constant.
91488 2021-09-07 Andrew Burgess <andrew.burgess@embecosm.com>
91490 gdb: use bool instead of int in struct internal_problem
91491 Change struct internal_problem (gdb/utils.c) to use bool instead of
91492 int, update the 3 static instances of this structure that we create to
91493 use true/false instead of 1/0.
91495 I've also updated the comments on struct internal_problem as the
91496 existing comment doesn't seem to be referring to the structure, it
91497 talks about returning something, which doesn't make sense in this
91500 There should be no user visible changes after this commit.
91502 2021-09-07 Andrew Burgess <andrew.burgess@embecosm.com>
91504 gdb: make thread_info::executing private
91505 Rename thread_info::executing to thread_info::m_executing, and make it
91506 private. Add a new get/set member functions, and convert GDB to make
91509 The only real change of interest in this patch is in thread.c where I
91510 have deleted the helper function set_executing_thread, and now just
91511 use the new set function thread_info::set_executing. However, the old
91512 helper function set_executing_thread included some code to reset the
91513 thread's stop_pc, so I moved this code into the new function
91514 thread_info::set_executing. However, I don't believe there is
91515 anywhere that this results in a change of behaviour, previously the
91516 executing flag was always set true through a call to
91517 set_executing_thread anyway.
91519 2021-09-07 Nick Clifton <nickc@redhat.com>
91521 Fix an illegal memory access triggered by an atempt to disassemble a corrupt xtensa binary.
91523 * elf32-xtensa.c (elf_xtensa_do_reloc): Add check for put of range
91526 2021-09-07 Andrew Burgess <andrew.burgess@embecosm.com>
91528 gdb/python: new function to add values into GDB's history
91529 The guile API has (history-append! <value>) to add values into GDB's
91530 history list. There is currently no equivalent in the Python API.
91532 This commit adds gdb.add_history(<value>) to the Python API, this
91533 function takes <value> a gdb.Value (or anything that can be passed to
91534 the constructor of gdb.Value), and adds the value it represents to
91535 GDB's history list. The index of the newly added value is returned.
91537 2021-09-07 Nick Clifton <nickc@redhat.com>
91539 Fix illegal memory access triggered by an attempt to disassemble a corrupt RISC-V binary.
91541 * elfxx-riscv.c (riscv_elf_add_sub_reloc): Add check for out of
91544 2021-09-07 Tom de Vries <tdevries@suse.de>
91546 [gdb/testsuite] Handle internal-error in gdb_unload
91547 When reverting commit 5a20fadc841 and using gdb_unload instead of runto "bar"
91548 to trigger the internal-error in test-case
91549 gdb.dwarf2/locexpr-data-member-location.exp, we run into:
91551 ERROR: couldn't unload file in $gdb (timeout).
91553 and the test-case takes about 1 minute.
91555 Fix this by handling internal-error in gdb_unload, such that we have:
91557 ERROR: Couldn't unload file in $gdb (GDB internal error).
91558 ERROR: Could not resync from internal error (eof)
91562 Tested on x86_64-linux.
91564 2021-09-07 Alan Modra <amodra@gmail.com>
91566 PR28307, segfault in ppc64_elf_toc64_reloc
91567 Adds missing bfd_reloc_offset_in_range checks to various relocation
91571 * elf32-ppc.c (ppc_elf_addr16_ha_reloc): Range check reloc offset.
91572 * elf64-ppc.c (ppc64_elf_ha_reloc, ppc64_elf_brtaken_reloc): Likewise.
91573 (ppc64_elf_toc64_reloc, ppc64_elf_prefix_reloc): Likewise.
91575 2021-09-07 GDB Administrator <gdbadmin@sourceware.org>
91577 Automatic date update in version.in
91579 2021-09-07 Tom de Vries <tdevries@suse.de>
91581 [gdb/testsuite] Handle internal-error in gdb_run_cmd
91582 When reverting commit 5a20fadc841 the test-case
91583 gdb.dwarf2/locexpr-data-member-location.exp fails like this:
91585 FAIL: gdb.dwarf2/locexpr-data-member-location.exp: running to bar in runto \
91586 (GDB internal error)
91587 ERROR: Could not resync from internal error (eof)
91589 and takes 1 minute to run.
91591 The long running time is caused by running into a timeout in gdb_run_cmd, at
91595 The program being debugged has been started already.^M
91596 Start it from the beginning? (y or n) y^M
91597 /home/vries/gdb_versions/devel/src/gdb/gdbtypes.c:5583: internal-error: \
91598 Unexpected type field location kind: 4^M
91599 A problem internal to GDB has been detected,^M
91600 further debugging may prove unreliable.^M
91601 Quit this debugging session? (y or n)
91604 Fix this by detecting the internal-error in gdb_run_cmd. We don't handle it
91605 in gdb_run_cmd, but stash the gdb output back into the buffer using
91606 -notransfer, and let the caller proc runto deal with it.
91608 After the fix, the test-case just takes 2 seconds.
91610 Tested on x86_64-linux.
91612 2021-09-06 Lancelot SIX <lsix@lancelotsix.com>
91614 gdb: rename gdb/testsuite/gdb.arch/riscv64-unwind-prologue-with-ld-lw.c
91615 A previous commit added the
91616 gdb.arch/riscv64-unwind-prologue-with-ld-lw.exp testcase, but one of its
91617 associated file was named after a previous version of the test.
91619 This commit fixes this and makes sure that all the files linked to this
91620 testcase share the same prefix in the name.
91622 Tested on riscv64 GNU/Linux.
91624 2021-09-06 Tom de Vries <tdevries@suse.de>
91626 [gdb/testsuite] Handle eof in gdb_internal_error_resync
91627 Before commit 5a20fadc841 the test-case
91628 gdb.dwarf2/locexpr-data-member-location.exp fails like this:
91630 FAIL: gdb.dwarf2/locexpr-data-member-location.exp: running to bar in runto \
91631 (GDB internal error)
91632 ERROR: : spawn id exp9 not open
91635 -i exp9 -timeout 10
91636 -re "Quit this debugging session\\? \\(y or n\\) $" {
91637 send_gdb "n\n" answer
91641 ("uplevel" body line 1)
91642 invoked from within
91643 "uplevel $body" NONE : spawn id exp9 not open
91644 ERROR: Could not resync from internal error (timeout)
91649 ERROR: : spawn id exp9 not open
91651 by handling eof in gdb_internal_error_resync, such that we have instead:
91653 FAIL: gdb.dwarf2/locexpr-data-member-location.exp: running to bar in runto \
91654 (GDB internal error)
91655 ERROR: Could not resync from internal error (eof)
91658 Tested on x86_64-linux.
91660 2021-09-06 Tom Tromey <tom@tromey.com>
91662 Remove some complaints.h includes
91663 There are a few includes of complaints.h that aren't necessary. This
91664 patch removes them. Tested by rebuilding.
91666 2021-09-06 Nick Clifton <nickc@redhat.com>
91668 Fix an illegal memory access triggered by disassembling corrupt s390x binaries.
91670 * elfxx-score7.c (score_elf_gprel15_reloc): If there is no output bfd
91671 treat the reloc as undefined.
91673 Fix potential use on an uninitialised vairable in the MCore assembler.
91675 Fix potential uninitialised variable in microblaze assembler code.
91677 2021-09-06 Yinjun Zhang <yinjun.zhang@corigine.com>
91679 Add a sanity check to the init_nfp6000_mecsr_sec() function in the NFP disassembler.
91681 2021-09-06 Alexandra Hájková <ahajkova@redhat.com>
91683 gdbtypes.c: Add the case for FIELD_LOC_KIND_DWARF_BLOCK
91684 The case for FIELD_LOC_KIND_DWARF_BLOCK was missing for
91685 switch TYPE_FIELD_LOC_KIND. Thas caused an internal-error
91686 under some circumstances.
91690 2021-09-06 GDB Administrator <gdbadmin@sourceware.org>
91692 Automatic date update in version.in
91694 2021-09-05 GDB Administrator <gdbadmin@sourceware.org>
91696 Automatic date update in version.in
91698 2021-09-04 Tom de Vries <tdevries@suse.de>
91700 [gdb/testsuite] Check avx support in gdb.arch/amd64-disp-step-avx.exp
91701 On a machine on Open Build Service I'm running into a SIGILL for test-case
91702 gdb.arch/amd64-disp-step-avx.exp:
91704 Program received signal SIGILL, Illegal instruction.^M
91705 test_rip_vex2 () at gdb.arch/amd64-disp-step-avx.S:40^M
91706 40 vmovsd ro_var(%rip),%xmm0^M
91707 (gdb) FAIL: gdb.arch/amd64-disp-step-avx.exp: vex2: \
91708 continue to test_rip_vex2_end
91710 The SIGILL happens when trying to execute the first avx instruction in the
91713 I can't directly access the machine, but looking at the log for test-case
91714 gdb.arch/i386-avx.exp, it seems that there's no avx support:
91716 Breakpoint 1, main (argc=1, argv=0x7fffffffd6b8) at gdb.arch/i386-avx.c:68^M
91717 68 if (have_avx ())^M
91718 (gdb) print have_avx ()^M
91723 - adding a gdb_caching_proc have_avx, similar to have_mpx, using the have_avx
91724 function from gdb.arch/i386-avx.c
91725 - using proc have_avx in both gdb/testsuite/gdb.arch/amd64-disp-step-avx.exp
91726 and gdb/testsuite/gdb.arch/i386-avx.exp.
91728 Tested on my x86_64-linux laptop with avx support, where both test-cases pass.
91730 gdb/testsuite/ChangeLog:
91732 2021-09-04 Tom de Vries <tdevries@suse.de>
91735 * gdb/testsuite/gdb.arch/i386-avx.c (main): Remove call to have_avx.
91736 (have_avx): Move ...
91737 * gdb/testsuite/lib/gdb.exp (have_avx): ... here. New proc.
91738 * gdb/testsuite/gdb.arch/amd64-disp-step-avx.exp: Use have_avx.
91739 * gdb/testsuite/gdb.arch/i386-avx.exp: Same.
91741 2021-09-04 Mike Frysinger <vapier@gentoo.org>
91743 gnulib: import sys_wait
91744 A few sims use this to emulate process syscalls.
91745 Gdb builds seem to still be fine.
91747 2021-09-04 GDB Administrator <gdbadmin@sourceware.org>
91749 Automatic date update in version.in
91751 2021-09-03 Tom Tromey <tromey@adacore.com>
91753 Use CORE_ADDR as return type from x86_dr_low_get_addr
91754 On a Windows build locally, watchpoints started failing. I tracked
91755 this down to x86_dr_low_get_addr returning an 'unsigned long'... in
91756 this particular build, this is a 32-bit type, but the inferior is a
91759 This patch fixes the problem by changing the return type. No other
91760 change is required, because this matches the function pointer in
91761 struct x86_dr_low_type.
91763 2021-09-03 Kevin Buettner <kevinb@redhat.com>
91765 Test case reproducing PR28030 bug
91766 The original reproducer for PR28030 required use of a specific
91767 compiler version - gcc-c++-11.1.1-3.fc34 is mentioned in the PR,
91768 though it seems probable that other gcc versions might also be able to
91769 reproduce the bug as well. This commit introduces a test case which,
91770 using the DWARF assembler, provides a reproducer which is independent
91771 of the compiler version. (Well, it'll work with whatever compilers
91772 the DWARF assembler works with.)
91774 To the best of my knowledge, it's also the first test case which uses
91775 the DWARF assembler to provide debug info for a shared object. That
91776 being the case, I provided more than the usual commentary which should
91777 allow this case to be used as a template when a combo shared
91778 library / DWARF assembler test case is required in the future.
91780 I provide some details regarding the bug in a comment near the
91781 beginning of locexpr-dml.exp.
91783 This problem was difficult to reproduce; I found myself constantly
91784 referring to the backtrace while trying to figure out what (else) I
91785 might be missing while trying to create a reproducer. Below is a
91786 partial backtrace which I include for posterity.
91788 #0 internal_error (
91789 file=0xc50110 "/ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/gdbtypes.c", line=5575,
91790 fmt=0xc520c0 "Unexpected type field location kind: %d")
91791 at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdbsupport/errors.cc:51
91792 #1 0x00000000006ef0c5 in copy_type_recursive (objfile=0x1635930,
91793 type=0x274c260, copied_types=0x30bb290)
91794 at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/gdbtypes.c:5575
91795 #2 0x00000000006ef382 in copy_type_recursive (objfile=0x1635930,
91796 type=0x274ca10, copied_types=0x30bb290)
91797 at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/gdbtypes.c:5602
91798 #3 0x0000000000a7409a in preserve_one_value (value=0x24269f0,
91799 objfile=0x1635930, copied_types=0x30bb290)
91800 at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/value.c:2529
91801 #4 0x000000000072012a in gdbscm_preserve_values (
91802 extlang=0xc55720 <extension_language_guile>, objfile=0x1635930,
91803 copied_types=0x30bb290)
91804 at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/guile/scm-value.c:94
91805 #5 0x00000000006a3f82 in preserve_ext_lang_values (objfile=0x1635930,
91806 copied_types=0x30bb290)
91807 at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/extension.c:568
91808 #6 0x0000000000a7428d in preserve_values (objfile=0x1635930)
91809 at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/value.c:2579
91810 #7 0x000000000082d514 in objfile::~objfile (this=0x1635930,
91811 __in_chrg=<optimized out>)
91812 at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/objfiles.c:549
91813 #8 0x0000000000831cc8 in std::_Sp_counted_ptr<objfile*, (__gnu_cxx::_Lock_policy)2>::_M_dispose (this=0x1654580)
91814 at /usr/include/c++/11/bits/shared_ptr_base.h:348
91815 #9 0x00000000004e6617 in std::_Sp_counted_base<(__gnu_cxx::_Lock_policy)2>::_M_release (this=0x1654580) at /usr/include/c++/11/bits/shared_ptr_base.h:168
91816 #10 0x00000000004e1d2f in std::__shared_count<(__gnu_cxx::_Lock_policy)2>::~__shared_count (this=0x190bb88, __in_chrg=<optimized out>)
91817 at /usr/include/c++/11/bits/shared_ptr_base.h:705
91818 #11 0x000000000082feee in std::__shared_ptr<objfile, (__gnu_cxx::_Lock_policy)2>::~__shared_ptr (this=0x190bb80, __in_chrg=<optimized out>)
91819 at /usr/include/c++/11/bits/shared_ptr_base.h:1154
91820 #12 0x000000000082ff0a in std::shared_ptr<objfile>::~shared_ptr (
91821 this=0x190bb80, __in_chrg=<optimized out>)
91822 at /usr/include/c++/11/bits/shared_ptr.h:122
91823 #13 0x000000000085ed7e in __gnu_cxx::new_allocator<std::_List_node<std::shared_ptr<objfile> > >::destroy<std::shared_ptr<objfile> > (this=0x114bc00,
91824 __p=0x190bb80) at /usr/include/c++/11/ext/new_allocator.h:168
91825 #14 0x000000000085e88d in std::allocator_traits<std::allocator<std::_List_node<std::shared_ptr<objfile> > > >::destroy<std::shared_ptr<objfile> > (__a=...,
91826 __p=0x190bb80) at /usr/include/c++/11/bits/alloc_traits.h:531
91827 #15 0x000000000085e50c in std::__cxx11::list<std::shared_ptr<objfile>, std::allocator<std::shared_ptr<objfile> > >::_M_erase (this=0x114bc00, __position=
91828 std::shared_ptr<objfile> (expired, weak count 1) = {get() = 0x1635930})
91829 at /usr/include/c++/11/bits/stl_list.h:1925
91830 #16 0x000000000085df0e in std::__cxx11::list<std::shared_ptr<objfile>, std::allocator<std::shared_ptr<objfile> > >::erase (this=0x114bc00, __position=
91831 std::shared_ptr<objfile> (expired, weak count 1) = {get() = 0x1635930})
91832 at /usr/include/c++/11/bits/list.tcc:158
91833 #17 0x000000000085c748 in program_space::remove_objfile (this=0x114bbc0,
91835 at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/progspace.c:210
91836 #18 0x000000000082d3ae in objfile::unlink (this=0x1635930)
91837 at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/objfiles.c:487
91838 #19 0x000000000082e68c in objfile_purge_solibs ()
91839 at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/objfiles.c:875
91840 #20 0x000000000092dd37 in no_shared_libraries (ignored=0x0, from_tty=1)
91841 at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/solib.c:1236
91842 #21 0x00000000009a37fe in target_pre_inferior (from_tty=1)
91843 at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/target.c:2496
91844 #22 0x00000000007454d6 in run_command_1 (args=0x0, from_tty=1,
91845 run_how=RUN_NORMAL)
91846 at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/infcmd.c:437
91848 I'll note a few points regarding this backtrace:
91850 Frame #1 is where the internal error occurs. It's caused by an
91851 unhandled case for FIELD_LOC_KIND_DWARF_BLOCK. The fix for this bug
91852 adds support for this case.
91854 Frame #22 - it's a partial backtrace - shows that GDB is attempting to
91855 (re)run the program. You can see the exact command sequence that was
91856 used for reproducing this problem in the PR (at
91857 https://sourceware.org/bugzilla/show_bug.cgi?id=28030), but in a
91858 nutshell, after starting the program and advancing to the appropriate
91859 source line, GDB was asked to step into libstdc++; a "finish" command
91860 was issued, returning a value. The fact that a value was returned is
91861 very important. GDB was then used to step back into libstdc++. A
91862 breakpoint was set on a source line in the library after which a "run"
91863 command was issued.
91865 Frame #19 shows a call to objfile_purge_solibs. It's aptly named.
91867 Frame #7 is a call to the destructor for one of the objfile solibs; it
91868 turned out to be the one for libstdc++.
91870 Frames #6 thru #3 show various value preservation frames. If you look
91871 at preserve_values() in gdb/value.c, the value history is preserved
91872 first, followed by internal variables, followed by values for the
91873 extension languages (python and guile).
91875 2021-09-03 Tom de Vries <tdevries@suse.de>
91877 [gdb/testsuite] Add untested case in gdb.gdb/complaints.exp
91878 When building gdb with "-Wall -O2 -g -flto=auto", I run into:
91880 (gdb) call clear_complaints()^M
91881 No symbol "clear_complaints" in current context.^M
91882 (gdb) FAIL: gdb.gdb/complaints.exp: clear complaints
91885 The problem is that lto has optimized away clear_complaints, and consequently
91886 the selftests cannot run.
91889 - using info function to detect presence of clear_complaints
91890 - handling the absence of clear_complaints by calling untested
91892 (gdb) UNTESTED: gdb.gdb/complaints.exp: \
91893 Cannot find clear_complaints, skipping test
91896 Tested on x86_64-linux.
91898 gdb/testsuite/ChangeLog:
91900 2021-09-03 Tom de Vries <tdevries@suse.de>
91902 * gdb.gdb/complaints.exp: Use untested if clear_complaints cannot
91905 2021-09-03 Felix Willgerodt <felix.willgerodt@intel.com>
91907 gdb: Enable finish command and inferior calls for _Float16 on amd64 and i386.
91908 Values of type _Float16 and _Float16 _Complex can now be used on CPUs with
91909 AVX512-FP16 support. Return values of those types are located in XMM0.
91910 Compiler support for gcc and clang is in progress, see e.g.:
91911 https://gcc.gnu.org/pipermail/gcc-patches/2021-July/574117.html
91914 2021-07-21 Felix Willgerodt <Felix.Willgerodt@intel.com>
91916 * amd64-tdep.c (amd64_classify): Classify _Float16 and
91917 _Float16 _Complex as AMD64_SSE.
91918 * i386-tdep.c (i386_extract_return_value): Read _Float16 and
91919 _Float16 _Complex from xmm0.
91921 gdb/testsuite/ChangeLog:
91922 2021-07-21 Felix Willgerodt <Felix.Willgerodt@intel.com>
91924 * gdb.arch/x86-avx512fp16-abi.c: New file.
91925 * gdb.arch/x86-avx512fp16-abi.exp: New file.
91927 2021-09-03 Felix Willgerodt <felix.willgerodt@intel.com>
91929 Add half support for AVX512 register view.
91930 This adds support for the half datatype, FP16, to the AVX512 register printing.
91933 2020-07-21 Felix Willgerodt <Felix.Willgerodt@intel.com>
91935 * i386-tdep.c (i386_zmm_type) <v32_half>: New field.
91936 (i386_ymm_type) <v16_half>: New field.
91937 (i386_gdbarch_init): Add set_gdbarch_half_format.
91938 * features/i386/64bit-avx512.xml: Add half type.
91939 * features/i386/64bit-avx512.c: Regenerated.
91940 * features/i386/64bit-sse.xml: Add half type.
91941 * features/i386/64bit-sse.c: Regenerated.
91943 gdb/testsuite/ChangeLog:
91944 2021-07-21 Felix Willgerodt <Felix.Willgerodt@intel.com>
91946 * gdb.arch/x86-avx512fp16.c: New file.
91947 * gdb.arch/x86-avx512fp16.exp: New file.
91948 * lib/gdb.exp (skip_avx512fp16_tests): New function.
91950 2021-09-03 Felix Willgerodt <felix.willgerodt@intel.com>
91952 gdb, i386: Enable AVX512-bfloat16 for i386 targets.
91953 Values of type bfloat16 can also be used on 32-bit targets, which was missed
91954 in the original enablement. This also adjusts the testcase to pass with
91955 "unix/-m32", where only the lower 8 AVX registers are available.
91958 2021-07-21 Felix Willgerodt <Felix.Willgerodt@intel.com>
91960 * features/i386/32bit-sse.xml: Add bfloat16 type.
91961 * features/i386/32bit-sse.c: Regenerated.
91963 gdb/testsuite/ChangeLog:
91964 2021-07-21 Felix Willgerodt <Felix.Willgerodt@intel.com>
91966 * gdb.arch/x86-avx512bf16.exp: Only use x/z/ymm 0-7.
91968 2021-09-03 Tom de Vries <tdevries@suse.de>
91970 [gdb/testsuite] Add untested case in selftest_setup
91971 When building gdb with "-Wall -O2 -g -flto=auto", I run into:
91973 FAIL: gdb.gdb/python-helper.exp: breakpoint in captured_main \
91974 (got interactive prompt)
91975 FAIL: gdb.gdb/python-helper.exp: run until breakpoint at captured_main
91976 WARNING: Couldn't test self
91978 and similar in gdb.gdb/selftest.exp.
91980 The first FAIL in more detail:
91982 (gdb) break captured_main^M
91983 Function "captured_main" not defined.^M
91984 Make breakpoint pending on future shared library load? (y or [n]) n^M
91985 (gdb) FAIL: gdb.gdb/python-helper.exp: breakpoint in captured_main \
91986 (got interactive prompt)
91989 The problem is that lto has optimized away the captured_main function
91990 and consequently the selftests dependent on that cannot run.
91993 - using gdb_breakpoint to detect failure to set the breakpoint
91994 - handling the failure to set a breakpoint by calling untested
91995 - not emitting the warning if we've already got untested
91998 (gdb) UNTESTED: gdb.gdb/python-helper.exp: Cannot set breakpoint at \
91999 captured_main, skipping testcase.
92002 gdb/testsuite/ChangeLog:
92004 2021-09-02 Tom de Vries <tdevries@suse.de>
92006 * lib/selftest-support.exp: Emit untested when not being able to set
92009 2021-09-03 Alan Modra <amodra@gmail.com>
92012 Fixes a few issues:
92013 1) If you use "-fsanitize=address,undefined" in CFLAGS, the Makefile
92014 attempt to trim off -fsanitize options left us with ",undefined".
92015 2) ld_compile adds CFLAGS_FOR_TARGET itself, no need to pass it.
92016 3) CFLAGS might be needed linking bootstrap test.
92018 * Makefile.am (CFLAGS_FOR_TARGET, CXXFLAGS_FOR_TARGET): Trim off
92020 * Makefile.in: Regenerate.
92021 * testsuite/ld-bootstrap/bootstrap.exp: Use CFLAGS when linking.
92022 * testsuite/ld-cdtest/cdtest.exp: Use CFLAGS_FOR_TARGET when
92024 * testsuite/ld-auto-import/auto-import.exp: Don't pass
92025 CFLAGS_FOR_TARGET to ld_compile.
92026 * testsuite/ld-cygwin/exe-export.exp: Likewise.
92027 * testsuite/ld-elfvers/vers.exp: Likewise.
92028 * testsuite/ld-elfvsb/elfvsb.exp: Likewise.
92029 * testsuite/ld-elfweak/elfweak.exp: Likewise.
92030 * testsuite/ld-gc/gc.exp: Likewise.
92031 * testsuite/ld-pe/pe-compile.exp: Likewise.
92032 * testsuite/ld-pe/pe-run.exp: Likewise.
92033 * testsuite/ld-pe/pe-run2.exp: Likewise.
92034 * testsuite/ld-plugin/plugin.exp: Likewise.
92035 * testsuite/ld-shared/shared.exp: Likewise.
92036 * testsuite/ld-elfcomm/elfcomm.exp: Likewise, and don't allow
92037 nios2 testing to trash CFLAGS_FOR_TARGET.
92038 * testsuite/ld-scripts/crossref.exp: Don't pass options in
92039 CC_FOR_TARGET, do so in CFLAGS_FOR_TARGET instead.
92040 * testsuite/ld-srec/srec.exp: Likewise, and for CXX.
92042 2021-09-03 Alan Modra <amodra@gmail.com>
92044 CC_FOR_TARGET et al
92045 The top level Makefile, the ld Makefile and others, define
92046 CC_FOR_TARGET to be a compiler for the binutils target machine. This
92047 is the compiler that should be used for almost all tests with C
92048 source. There are _FOR_TARGET versions of CFLAGS, CXX, and CXXFLAGS
92049 too. This was all supposed to work with the testsuite .exp files
92050 using CC for the target compiler, and CC_FOR_HOST for the host
92051 compiler, with the makefiles passing CC=$CC_FOR_TARGET and
92052 CC_FOR_HOST=$CC to the runtest invocation.
92054 One exception to the rule of using CC_FOR_TARGET is the native-only ld
92055 bootstrap test, which uses the newly built ld to link a copy of
92056 itself. Since the files being linked were created with the host
92057 compiler, the boostrap test should use CC and CFLAGS, in case some
92058 host compiler option provides needed libraries automatically.
92059 However, bootstrap.exp used CC where it should have used CC_FOR_HOST.
92060 I set about fixing that problem, then decided that playing games in
92061 the makefiles with CC was a bad idea. Not only is it confusing, but
92062 other dejagnu code knows about CC_FOR_TARGET. See dejagnu/target.exp.
92064 So this patch gets rid of the makefile variable renaming and changes
92065 all the .exp files to use the correct _FOR_TARGET variables.
92066 CC_FOR_HOST and CFLAGS_FOR_HOST disappear. A followup patch will
92067 correct bootstrap.exp to use CFLAGS, and a number of other things I
92071 * testsuite/lib/binutils-common.exp (run_dump_test): Use
92072 CC_FOR_TARGET and CFLAGS_FOR_TARGET rather than CC and CFLAGS.
92074 * Makefile.am (check-DEJAGNU): Don't set CC to CC_FOR_TARGET
92075 and similar. Pass variables with unchanged names. Don't set
92076 CC_FOR_HOST or CFLAGS_FOR_HOST.
92077 * Makefile.in: Regenerate.
92078 * testsuite/config/default.exp: Update default CC and similar.
92079 (compiler_supports, plug_opt): Use CC_FOR_TARGET.
92080 * testsuite/ld-cdtest/cdtest.exp: Replace all uses of CC with
92081 CC_FOR_TARGET, and similarly for CFLAGS, CXX and CXXFLAGS.
92082 * testsuite/ld-auto-import/auto-import.exp: Likewise.
92083 * testsuite/ld-cygwin/exe-export.exp: Likewise.
92084 * testsuite/ld-elf/dwarf.exp: Likewise.
92085 * testsuite/ld-elf/indirect.exp: Likewise.
92086 * testsuite/ld-elf/shared.exp: Likewise.
92087 * testsuite/ld-elfcomm/elfcomm.exp: Likewise.
92088 * testsuite/ld-elfvers/vers.exp: Likewise.
92089 * testsuite/ld-elfvsb/elfvsb.exp: Likewise.
92090 * testsuite/ld-elfweak/elfweak.exp: Likewise.
92091 * testsuite/ld-gc/gc.exp: Likewise.
92092 * testsuite/ld-ifunc/ifunc.exp: Likewise.
92093 * testsuite/ld-mn10300/mn10300.exp: Likewise.
92094 * testsuite/ld-pe/pe-compile.exp: Likewise.
92095 * testsuite/ld-pe/pe-run.exp: Likewise.
92096 * testsuite/ld-pe/pe-run2.exp: Likewise.
92097 * testsuite/ld-pie/pie.exp: Likewise.
92098 * testsuite/ld-plugin/lto.exp: Likewise.
92099 * testsuite/ld-plugin/plugin.exp: Likewise.
92100 * testsuite/ld-scripts/crossref.exp: Likewise.
92101 * testsuite/ld-selective/selective.exp: Likewise.
92102 * testsuite/ld-sh/sh.exp: Likewise.
92103 * testsuite/ld-shared/shared.exp: Likewise.
92104 * testsuite/ld-srec/srec.exp: Likewise.
92105 * testsuite/ld-undefined/undefined.exp: Likewise.
92106 * testsuite/ld-unique/unique.exp: Likewise.
92107 * testsuite/ld-x86-64/tls.exp: Likewise.
92108 * testsuite/lib/ld-lib.exp: Likewise.
92110 * Makefile.am (check-DEJAGNU): Don't set CC to CC_FOR_TARGET.
92111 Pass CC and CC_FOR_TARGET. Don't set CC_FOR_HOST.
92112 * Makefile.in: Regenerate.
92113 * testsuite/config/default.exp: Update default CC and similar.
92114 * testsuite/lib/ctf-lib.exp (run_native_host_cmd): Use CC rather
92116 (run_lookup_test): Use CC_FOR_TARGET and CFLAGS_FOR_TARGET.
92118 2021-09-03 Alan Modra <amodra@gmail.com>
92120 pj: asan: out of bounds, ubsan: left shift of negative
92121 * pj-dis.c: Include libiberty.h.
92122 (print_insn_pj): Don't index op->arg past array bound. Don't
92123 left shift negative int.
92125 ubsan: alpha: member access within null pointer
92126 * elf64-alpha.c (elf64_alpha_relax_with_lituse): Avoid UB.
92128 ubsan: libctf: applying zero offset to null pointer
92129 * ctf-open.c (init_symtab): Avoid ubsan error.
92131 2021-09-03 Alan Modra <amodra@gmail.com>
92134 --enable-maintainer-mode showed a number of files needing to be
92135 regenerated, and in the case of ld/Makefile.in that the file was
92136 regenerated by hand. Nothing to see here really.
92139 * Makefile.am (ALL_64_EMULATION_SOURCES): Sort haiku entry.
92140 * Makefile.in: Regenerate.
92141 * po/BLD-POTFILES.in: Regenerate.
92143 * configure: Regenerate.
92145 * configure: Regenerate.
92147 2021-09-03 Fangrui Song <maskray@google.com>
92149 gold: --export-dynamic-symbol: don't imply -u
92153 * archive.cc (Library_base::should_include_member): Don't handle
92154 --export-dynamic-symbol.
92155 * symtab.cc (Symbol_table::do_add_undefined_symbols_from_command_line):
92158 2021-09-03 GDB Administrator <gdbadmin@sourceware.org>
92160 Automatic date update in version.in
92162 2021-09-02 Alexander von Gluck IV <kallisti5@unixzen.com>
92164 Add support for the haiku operating system. These are the os support patches we have been grooming and maintaining for quite a few years over on git.haiku-os.org. All of these architectures are working and most have been stable for quite some time.
92166 2021-09-02 Nick Clifton <nickc@redhat.com>
92168 Fix the V850 assembler's generation of relocations for the st.b instruction.
92170 gas * config/tc-v850.c (handle_lo16): Also accept
92171 BFD_RELOC_V850_LO16_SPLIT_OFFSET.
92172 * testsuite/gas/v850/split-lo16.s: Add extra line.
92173 * testsuite/gas/v850/split-lo16.d: Update expected disassembly.
92175 opcodes * v850-opc.c (D16): Use BFD_RELOC_V850_LO16_SPLIT_OFFSET in place
92178 2021-09-02 Alan Modra <amodra@gmail.com>
92180 SHT_SYMTAB_SHNDX handling
92181 .symtab_shndx section contents is an array, one entry for each symbol
92182 in .symtab, present when the number of symbols exceeds a little less
92183 than 64k. Since the mapping is 1-1 with symbols there is no need to
92184 keep both dest_index and destshndx_index in elf_sym_strtab. Instead,
92185 just make sure that the shndx pointers to the swap functions are kept
92186 NULL when .symtab_shndx does not exist. Also, strtabcount in the
92187 linker's elf hash table is incremented in lock-step with the output
92188 symcount, so that can disappear too.
92190 2021-09-02 Alan Modra <amodra@gmail.com>
92192 PTR_ADD and NPTR_ADD for bfd.h
92193 This defines a couple of macros used to avoid ubsan complaints about
92194 calculations involving NULL pointers. PTR_ADD should be used in the
92195 case where it is known that the offset is always zero with a NULL
92196 pointer, and you'd like to know if a non-zero offset is ever used.
92197 NPTR_ADD should be rarely used, but is defined for cases where a
92198 non-zero offset is expected and should be ignored if the pointer is
92202 * bfd-in.h (PTR_ADD, NPTR_ADD): Define.
92203 * bfd-in2.h: Regenerate.
92204 * elf-eh-frame.c (adjust_eh_frame_local_symbols): Avoid NULL
92205 pointer calculations.
92206 * elflink.c (_bfd_elf_strip_zero_sized_dynamic_sections): Likewise.
92207 (bfd_elf_add_dt_needed_tag, elf_finalize_dynstr): Likewise.
92208 (elf_link_add_object_symbols, elf_link_input_bfd): Likewise.
92209 (bfd_elf_final_link, bfd_elf_gc_record_vtinherit): Likewise.
92211 * objdump.c (disassemble_section): Use PTR_ADD for rel_ppend.
92213 2021-09-02 Alan Modra <amodra@gmail.com>
92215 obstack.h __PTR_ALIGN vs. ubsan
92216 Current ubsan complains on every use of __PTR_ALIGN (when ptrdiff_t is
92217 as large as a pointer), due to making calculations relative to a NULL
92218 pointer. This patch avoids the problem by extracting out and
92219 simplifying __BPTR_ALIGN for the usual case. I've continued to use
92220 ptrdiff_t here, where it might be better to throw away __BPTR_ALIGN
92221 entirely and just assume uintptr_t exists.
92223 * obstack.h (__PTR_ALIGN): Expand and simplify __BPTR_ALIGN
92224 rather than calculating relative to a NULL pointer.
92226 2021-09-02 GDB Administrator <gdbadmin@sourceware.org>
92228 Automatic date update in version.in
92230 2021-09-01 Tom de Vries <tdevries@suse.de>
92232 [gdb/testsuite] Fix dwo path in fission-*.S
92233 [ Using $build for /home/vries/gdb_versions/devel/build to make things a bit
92236 When using make check// to run test-case gdb.dwarf2/fission-base.exp:
92238 ( cd $build/gdb; make check//unix RUNTESTFLAGS="fission-base.exp" )
92243 $build/gdb/testsuite.unix/outputs/gdb.dwarf2/fission-base/fission-base^M
92244 Reading symbols from \
92245 $build/gdb/testsuite.unix/outputs/gdb.dwarf2/fission-base/fission-base...^M
92246 warning: Could not find DWO CU \
92247 $build/gdb/testsuite.1/outputs/gdb.dwarf2/fission-base/fission-base.dwo \
92248 (0x807060504030201) referenced by CU at offset 0xc7 [in module \
92249 $build/gdb/testsuite.unix/outputs/gdb.dwarf2/fission-base/fission-base]^M
92252 The problem is that the executable refers to the dwo file using path name
92253 $build/gdb/testsuite.1/outputs/gdb.dwarf2/fission-base/fission-base.dwo,
92254 while the actual dwo file is at
92255 $build/gdb/testsuite.unix/outputs/gdb.dwarf2/fission-base/fission-base.dwo.
92257 This is caused by this trick in fission-base.S:
92259 #define XSTR(s) STR(s)
92262 .asciz XSTR(DWO) # DW_AT_GNU_dwo_name
92266 $ echo | gcc -E -dD - | grep "define unix"
92269 I used this trick to avoid doing additional_flags=-DDWO=\"$dwo\", since I was
92270 concerned that there could be quoting issues.
92272 However, I've found other uses of this pattern, f.i. in
92273 gdb/testsuite/gdb.base/corefile-buildid.exp:
92275 additional_flags=-DSHLIB_NAME=\"$dlopen_lib\"]
92279 - using additional_flags=-DDWO=\"$dwo\" and
92280 - using plain DWO instead of XSTR(DWO)
92282 Likewise in other gdb.dwarf2/fission*.exp test-cases.
92284 Tested on x86_64-linux, using make check//unix.
92286 gdb/testsuite/ChangeLog:
92288 2021-09-01 Tom de Vries <tdevries@suse.de>
92291 * gdb.dwarf2/fission-base.S: Use DWO instead of XSTR(DWO).
92292 * gdb.dwarf2/fission-loclists-pie.S: Same.
92293 * gdb.dwarf2/fission-loclists.S: Same.
92294 * gdb.dwarf2/fission-reread.S: Same.
92295 * gdb.dwarf2/fission-base.exp: Use additional_flags=-DDWO=\"$dwo\".
92296 * gdb.dwarf2/fission-loclists-pie.exp: Same.
92297 * gdb.dwarf2/fission-loclists.exp: Same.
92298 * gdb.dwarf2/fission-reread.exp: Same.
92300 2021-09-01 Tom de Vries <tdevries@suse.de>
92302 [gdb/testsuite] Fix gdb.fortran/call-no-debug.exp symbol search
92303 On openSUSE Tumbleweed I ran into:
92305 (gdb) ptype outstring_func.part^M
92306 No symbol "outstring_func" in current context.^M
92307 (gdb) FAIL: gdb.fortran/call-no-debug.exp: ptype outstring_func.part
92309 while on openSUSE Leap 15.2 I have instead:
92311 (gdb) ptype string_func_^M
92312 type = <unknown return type> ()^M
92313 (gdb) PASS: gdb.fortran/call-no-debug.exp: ptype string_func_
92316 The difference is caused by the result for "info function string_func", which
92317 is this for the latter:
92319 (gdb) info function string_func^M
92320 All functions matching regular expression "string_func":^M
92322 Non-debugging symbols:^M
92323 0x000000000040089c string_func_^M
92325 but this for the former:
92327 (gdb) info function string_func^M
92328 All functions matching regular expression "string_func":^M
92330 Non-debugging symbols:^M
92331 0x00000000004012bb string_func_^M
92332 0x00007ffff7bac5b0 outstring_func.part^M
92333 0x00007ffff7bb1a00 outstring_func.part^M
92336 The extra symbols are part of glibc:
92338 $ nm /lib64/libc.so.6 | grep string_func
92339 00000000000695b0 t outstring_func.part.0
92340 000000000006ea00 t outstring_func.part.0
92343 If glibc debug info is installed, we get instead:
92345 (gdb) info function string_func^M
92346 All functions matching regular expression "string_func":^M
92348 File /usr/src/debug/glibc-2.33-9.1.x86_64/stdio-common/vfprintf-internal.c:^M
92349 236: static int outstring_func(int, size_t, const unsigned int *, FILE *);^M
92351 File vfprintf-internal.c:^M
92352 236: static int outstring_func(int, size_t, const unsigned char *, FILE *);^M
92354 Non-debugging symbols:^M
92355 0x00000000004012bb string_func_^M
92357 and the FAIL doesn't trigger.
92359 Fix this by calling "info function string_func" before starting the exec, such
92360 that only symbols of the exec are taken into account.
92362 Tested on x86_64-linux.
92364 gdb/testsuite/ChangeLog:
92366 2021-09-01 Tom de Vries <tdevries@suse.de>
92368 * gdb.fortran/call-no-debug.exp: Avoid shared lib symbols for
92369 find_mangled_name calls.
92371 2021-09-01 Yinjun Zhang <yinjun.zhang@corigine.com>
92373 nfp: add validity check of island and me
92374 AddressSanitizer detects heap-buffer-overflow when running
92375 "objdump -D" for nfp .nffw files.
92378 * nfp-dis.c (_NFP_ISLAND_MAX, _NFP_ME_MAX): Define.
92379 (nfp_priv_data): ..and use here.
92380 (_print_instrs): Sanity check island and menum.
92382 2021-09-01 Alan Modra <amodra@gmail.com>
92384 PR28250, Null pointer dereference in debug_class_type_samep
92385 Typo fix, obviously should be m1->variants != NULL, not
92386 m1->variants == NULL.
92389 * debug.c (debug_class_type_samep): Correct m1->variants test.
92391 2021-09-01 GDB Administrator <gdbadmin@sourceware.org>
92393 Automatic date update in version.in
92395 2021-08-31 Simon Marchi <simon.marchi@polymtl.ca>
92397 gdb: remove breakpoint_find_if
92398 Remove breakpoint_find_if, replace its sole usage with using
92399 all_breakpoints directly instead. At the same time, change return
92402 Change-Id: I9ec392236b4804b362d16ab563330b9c07311106
92404 2021-08-31 Nick Clifton <nickc@redhat.com>
92406 Update the how-to-make-a-release document so that a check for empty manual pages is included. cf PR 28144
92408 2021-08-31 Nelson Chu <nelson.chu@sifive.com>
92410 RISC-V: Extend .insn directive to support hardcode encoding.
92411 The .insn directive can let users use their own instructions, or
92412 some new instruction, which haven't supported in the old binutils.
92413 For example, if users want to use sifive cache instruction, they
92414 cannot just write "cflush.d1.l1" in the assembly code, they should
92415 use ".insn i SYSTEM, 0, x0, x10, -0x40". But the .insn directive
92416 may not easy to use for some cases, and not so friendly to users.
92417 Therefore, I believe most of the users will use ".word 0xfc050073",
92418 to encode the instructions directly, rather than use .insn. But
92419 once we have supported the mapping symbols, the .word directives
92420 are marked as data, so disassembler won't dump them as instructions
92421 as usual. I have discussed this with Kito many times, we all think
92422 extend the .insn direcitve to support the hardcode encoding, is the
92423 easiest way to resolve the problem. Therefore, there are two more
92424 .insn formats are proposed as follows,
92426 (original) .insn <type>, <operand1>, <operand2>, ...
92427 .insn <insn-length>, <value>
92430 The <type> is string, and the <insn-length> and <value> are constants.
92433 * config/tc-riscv.c (riscv_ip_hardcode): Similar to riscv_ip,
92434 but assembles an instruction according to the hardcode values
92435 of .insn directive.
92436 * doc/c-riscv.texi: Document two new .insn formats.
92437 * testsuite/gas/riscv/insn-fail.d: New testcases.
92438 * testsuite/gas/riscv/insn-fail.l: Likewise.
92439 * testsuite/gas/riscv/insn-fail.s: Likewise.
92440 * testsuite/gas/riscv/insn.d: Updated.
92441 * testsuite/gas/riscv/insn.s: Likewise.
92443 2021-08-31 John Baldwin <jhb@FreeBSD.org>
92445 Use gdbfmt for vprintf_filtered.
92446 gdbfmt was already used for printf_filtered, so using it for
92447 vprintf_filtered is more consistent.
92449 As a result, all callers of vfprintf_maybe_filtered now use gdbfmt, so
92450 the function can be simplified to assume the gdbfmt case and remove
92451 the associated bool argument. Similary, vprintf_filtered is now a
92452 simple wrapper around vfprintf_filtered.
92454 2021-08-31 GDB Administrator <gdbadmin@sourceware.org>
92456 Automatic date update in version.in
92458 2021-08-30 John Baldwin <jhb@FreeBSD.org>
92460 fbsd-nat: Don't use '%jd' and '%ju' with printf_filtered.
92461 The handler for 'info proc status' for native processes on FreeBSD
92462 uses the 'j' size modifier along with uintmax_t / intmax_t casts to
92463 output integer values for types such as off_t that are not aliases of
92464 a basic C type such as 'int' or 'long'. printf_filtered does not
92465 support the 'j' modifer, so this resulted in runtime errors in
92468 (gdb) info proc stat
92472 Parent process: 8673
92473 Process group: 8674
92475 Unrecognized format specifier 'j' in printf
92477 Instead, use plongest and pulongest to generate the output strings of
92478 these integer values.
92480 2021-08-30 Simon Marchi <simon.marchi@polymtl.ca>
92482 gdb: fix build error in unittests/parallel-for-selftests.c
92483 We get this error when building GDB on some platforms. I get it using
92484 g++-10 on Ubuntu 20.04 (installed using the distro package). It was
92485 also reported by John Baldwin, using a clang that uses libc++.
92487 CXX unittests/parallel-for-selftests.o
92488 cc1plus: warning: command line option '-Wmissing-prototypes' is valid for C/ObjC but not for C++
92489 /home/smarchi/src/binutils-gdb/gdb/unittests/parallel-for-selftests.c: In function 'void selftests::parallel_for::test(int)':
92490 /home/smarchi/src/binutils-gdb/gdb/unittests/parallel-for-selftests.c:53:30: error: use of deleted function 'std::atomic<int>::atomic(const std::atomic<int>&)'
92491 53 | std::atomic<int> counter = 0;
92493 In file included from /usr/include/c++/9/future:42,
92494 from /home/smarchi/src/binutils-gdb/gdb/../gdbsupport/thread-pool.h:29,
92495 from /home/smarchi/src/binutils-gdb/gdb/../gdbsupport/parallel-for.h:26,
92496 from /home/smarchi/src/binutils-gdb/gdb/unittests/parallel-for-selftests.c:22:
92497 /usr/include/c++/9/atomic:755:7: note: declared here
92498 755 | atomic(const atomic&) = delete;
92500 /usr/include/c++/9/atomic:759:17: note: after user-defined conversion: 'constexpr std::atomic<int>::atomic(std::atomic<int>::__integral_type)'
92501 759 | constexpr atomic(__integral_type __i) noexcept : __base_type(__i) { }
92504 I haven't dug to know why it does not happen everywhere, but this patch
92505 fixes it by using the constructor to initialize the variable, rather
92506 than the assignment operator.
92508 Change-Id: I6b27958171bf6187f6a875657395fd10441db7e6
92510 2021-08-30 Nelson Chu <nelson.chu@sifive.com>
92512 RISC-V: PR28291, Fix the gdb fails that PR27916 caused.
92513 * According to PR28291, we get the following unexpected gdb behavior,
92515 (gdb) disassemble 0x0,+4
92516 Dump of assembler code from 0x0 to 0x4:
92517 0x0000000000000000:
92518 0x0000000000000001:
92519 0x0000000000000002:
92520 0x0000000000000003:
92521 End of assembler dump.
92523 * This patch should fix it to the right behavior,
92525 (gdb) disassemble 0x0,+4
92526 Dump of assembler code from 0x0 to 0x4:
92527 0x0000000000000000: Cannot access memory at address 0x0
92531 * riscv-dis.c (print_insn_riscv): Return STATUS if it is not zero.
92533 2021-08-30 Tom Tromey <tom@tromey.com>
92535 Add some parallel_for_each tests
92536 Tom de Vries noticed that a patch in the DWARF scanner rewrite series
92537 caused a regression in parallel_for_each -- it started crashing in the
92538 case where the number of threads is 0 (there was an unchecked use of
92539 "n-1" that was used to size an array).
92541 He also pointed out that there were no tests of parallel_for_each.
92542 This adds a few tests of parallel_for_each, primarily testing that
92543 different settings for the number of threads will work. This test
92544 catches the bug that he found in that series.
92546 2021-08-30 Tom Tromey <tom@tromey.com>
92548 Add a show function for "maint show worker-threads"
92549 I wanted to see how many threads gdb thought it was using, but
92550 "maint show worker-threads" only reported "unlimited". This patch
92551 adds a show function so that it will now report the number of threads
92554 Regression tested on x86-64 Fedora 34.
92556 2021-08-30 Tom de Vries <tdevries@suse.de>
92558 [gdb/cli] Don't assert on empty string for core-file
92559 With current gdb we run into:
92562 : No such file or directory.
92563 pathstuff.cc:132: internal-error: \
92564 gdb::unique_xmalloc_ptr<char> gdb_abspath(const char*): \
92565 Assertion `path != NULL && path[0] != '\0'' failed.
92568 Fix this by skipping the call to gdb_abspath in core_target_open in the
92569 empty-string case, such that we have instead:
92572 : No such file or directory.
92573 : No such file or directory.
92577 Tested on x86_64-linux.
92581 2021-08-30 Tom de Vries <tdevries@suse.de>
92584 * gdb/corelow.c (core_target_open): Skip call to gdb_abspath in the
92587 gdb/testsuite/ChangeLog:
92589 2021-08-30 Tom de Vries <tdevries@suse.de>
92592 * gdb.base/batch-exit-status.exp: Add gdb '' and gdb '' '' tests.
92594 2021-08-30 Nelson Chu <nelson.chu@sifive.com>
92596 RISC-V: PR27916, Support mapping symbols.
92597 Similar to ARM/AARCH64, we add mapping symbols in the symbol table,
92598 to mark the start addresses of data and instructions. The $d means
92599 data, and the $x means instruction. Then the disassembler uses these
92600 symbols to decide whether we should dump data or instruction.
92602 Consider the mapping-04 test case,
92615 .word 0x1 # No need to add mapping symbols.
92618 $ riscv64-unknown-elf-as tmp.s -o tmp.o
92619 $ riscv64-unknown-elf-objdump -d tmp.o
92621 Disassembly of section .text:
92623 0000000000000000 <.text>:
92624 0: 00001001 .word 0x00001001 # Marked $d, .fill directive.
92625 4: 00001001 .word 0x00001001
92626 8: 00000001 .word 0x00000001 # .byte + part of .word.
92627 c: 00 .byte 0x00 # remaining .word.
92628 d: 00 .byte 0x00 # Marked $d, odd byte of alignment.
92629 e: 0001 nop # Marked $x, nops for alignment.
92630 10: 00a50533 add a0,a0,a0
92631 14: 20022002 .word 0x20022002 # Marked $d, .fill directive.
92632 18: 20022002 .word 0x20022002
92633 1c: 2002 .short 0x2002
92634 1e: 00b585b3 add a1,a1,a1 # Marked $x.
92635 22: 0001 nop # Section tail alignment.
92638 * Use $d and $x to mark the distribution of data and instructions.
92639 Alignments of code are recognized as instructions, since we usually
92640 fill nops for them.
92642 * If the alignment have odd bytes, then we cannot just fill the nops
92643 into the spaces. We always fill an odd byte 0x00 at the start of
92644 the spaces. Therefore, add a $d mapping symbol for the odd byte,
92645 to tell disassembler that it isn't an instruction. The behavior
92646 is same as Arm and Aarch64.
92648 The elf/linux toolchain regressions all passed. Besides, I also
92649 disable the mapping symbols internally, but use the new objudmp, the
92650 regressions passed, too. Therefore, the new objudmp should dump
92651 the objects corretly, even if they don't have any mapping symbols.
92655 * cpu-riscv.c (riscv_elf_is_mapping_symbols): Define mapping symbols.
92656 * cpu-riscv.h: extern riscv_elf_is_mapping_symbols.
92657 * elfnn-riscv.c (riscv_maybe_function_sym): Do not choose mapping
92658 symbols as a function name.
92659 (riscv_elf_is_target_special_symbol): Add mapping symbols.
92662 * testsuite/binutils-all/readelf.s: Updated.
92663 * testsuite/binutils-all/readelf.s-64: Likewise.
92664 * testsuite/binutils-all/readelf.s-64-unused: Likewise.
92665 * testsuite/binutils-all/readelf.ss: Likewise.
92666 * testsuite/binutils-all/readelf.ss-64: Likewise.
92667 * testsuite/binutils-all/readelf.ss-64-unused: Likewise.
92670 * config/tc-riscv.c (make_mapping_symbol): Create a new mapping symbol.
92671 (riscv_mapping_state): Decide whether to create mapping symbol for
92672 frag_now. Only add the mapping symbols to text sections.
92673 (riscv_add_odd_padding_symbol): Add the mapping symbols for the
92674 riscv_handle_align, which have odd bytes spaces.
92675 (riscv_check_mapping_symbols): Remove any excess mapping symbols.
92676 (md_assemble): Marked as MAP_INSN.
92677 (riscv_frag_align_code): Marked as MAP_INSN.
92678 (riscv_init_frag): Add mapping symbols for frag, it usually called
92679 by frag_var. Marked as MAP_DATA for rs_align and rs_fill, and
92680 marked as MAP_INSN for rs_align_code.
92681 (s_riscv_insn): Marked as MAP_INSN.
92682 (riscv_adjust_symtab): Call riscv_check_mapping_symbols.
92683 * config/tc-riscv.h (md_cons_align): Defined to riscv_mapping_state
92685 (TC_SEGMENT_INFO_TYPE): Record mapping state for each segment.
92686 (TC_FRAG_TYPE): Record the first and last mapping symbols for the
92687 fragments. The first mapping symbol must be placed at the start
92689 (TC_FRAG_INIT): Defined to riscv_init_frag.
92690 * testsuite/gas/riscv/mapping-01.s: New testcase.
92691 * testsuite/gas/riscv/mapping-01a.d: Likewise.
92692 * testsuite/gas/riscv/mapping-01b.d: Likewise.
92693 * testsuite/gas/riscv/mapping-02.s: Likewise.
92694 * testsuite/gas/riscv/mapping-02a.d: Likewise.
92695 * testsuite/gas/riscv/mapping-02b.d: Likewise.
92696 * testsuite/gas/riscv/mapping-03.s: Likewise.
92697 * testsuite/gas/riscv/mapping-03a.d: Likewise.
92698 * testsuite/gas/riscv/mapping-03b.d: Likewise.
92699 * testsuite/gas/riscv/mapping-04.s: Likewise.
92700 * testsuite/gas/riscv/mapping-04a.d: Likewise.
92701 * testsuite/gas/riscv/mapping-04b.d: Likewise.
92702 * testsuite/gas/riscv/mapping-norelax-04a.d: Likewise.
92703 * testsuite/gas/riscv/mapping-norelax-04b.d: Likewise.
92704 * testsuite/gas/riscv/no-relax-align.d: Updated.
92705 * testsuite/gas/riscv/no-relax-align-2.d: Likewise.
92708 * opcode/riscv.h (enum riscv_seg_mstate): Added.
92712 * riscv-dis.c (last_map_symbol, last_stop_offset, last_map_state):
92713 Added to dump sections with mapping symbols.
92714 (riscv_get_map_state): Get the mapping state from the symbol.
92715 (riscv_search_mapping_symbol): Check the sorted symbol table, and
92716 then find the suitable mapping symbol.
92717 (riscv_data_length): Decide which data size we should print.
92718 (riscv_disassemble_data): Dump the data contents.
92719 (print_insn_riscv): Handle the mapping symbols.
92720 (riscv_symbol_is_valid): Marked mapping symbols as invalid.
92722 2021-08-30 Tom de Vries <tdevries@suse.de>
92724 [gdb/testsuite] Improve argument syntax of proc arange
92725 The current syntax of proc arange is:
92727 proc arange { arange_start arange_length {comment ""} {seg_sel ""} } {
92729 and a typical call looks like:
92734 This style is somewhat annoying because if you want to specify the last
92735 parameter, you need to give the default values of all the other optional ones
92738 arange $start $len "" $seg_sel
92741 Update the syntax to:
92743 proc arange { options arange_start arange_length } {
92749 such that a typical call looks like:
92751 arange {} $start $len
92753 and a call using seg_sel looks like:
92760 Also update proc aranges, which already has an options argument, to use the
92761 new proc parse_options.
92763 Tested on x86_64-linux.
92765 Co-Authored-By: Simon Marchi <simon.marchi@polymtl.ca>
92767 2021-08-30 GDB Administrator <gdbadmin@sourceware.org>
92769 Automatic date update in version.in
92771 2021-08-29 GDB Administrator <gdbadmin@sourceware.org>
92773 Automatic date update in version.in
92775 2021-08-28 H.J. Lu <hjl.tools@gmail.com>
92777 ld: Change indirect symbol from IR to undefined
92781 * elflink.c (_bfd_elf_merge_symbol): Change indirect symbol from
92787 * testsuite/ld-plugin/lto.exp: Run PR ld/28264 test.
92788 * testsuite/ld-plugin/pr28264-1.d: New file.
92789 * testsuite/ld-plugin/pr28264-2.d: Likewise.
92790 * testsuite/ld-plugin/pr28264-3.d: Likewise.
92791 * testsuite/ld-plugin/pr28264-4.d: Likewise.
92792 * testsuite/ld-plugin/pr28264.c: Likewise.
92793 * testsuite/ld-plugin/pr28264.ver: Likewise.
92795 2021-08-28 Alan Modra <amodra@gmail.com>
92797 PR28264, ld.bfd crash on linking efivar with LTO
92800 * linker.c (_bfd_generic_link_add_one_symbol <MIND>): Check
92801 that string is non-NULL.
92803 2021-08-28 GDB Administrator <gdbadmin@sourceware.org>
92805 Automatic date update in version.in
92807 2021-08-27 Tom de Vries <tdevries@suse.de>
92809 [gdb/symtab] Don't write .gdb_index symbol table with empty entries
92810 When comparing the sizes of the index files generated for shlib
92811 outputs/gdb.dwarf2/dw2-zero-range/shr1.sl, I noticed a large difference
92812 between .debug_names:
92814 $ gdb -q -batch $shlib -ex "save gdb-index -dwarf-5 ."
92815 $ du -b -h shr1.sl.debug_names shr1.sl.debug_str
92816 61 shr1.sl.debug_names
92817 0 shr1.sl.debug_str
92821 $ gdb -q -batch $shlib -ex "save gdb-index ."
92822 $ du -b -h shr1.sl.gdb-index
92823 8.2K shr1.sl.gdb-index
92826 The problem is that the .gdb_index contains a non-empty symbol table with only
92829 Fix this by making the symbol table empty, such that we have instead:
92831 $ du -b -h shr1.sl.gdb-index
92832 184 shr1.sl.gdb-index
92835 Tested on x86_64-linux.
92837 2021-08-27 Tom de Vries <tdevries@suse.de>
92839 [gdb/testsuite] Generate .debug_aranges in gdb.dlang/watch-loc.exp
92840 Before commit 5ef670d81fd "[gdb/testsuite] Add dummy start and end CUs in
92841 dwarf assembly" we had in exec outputs/gdb.dlang/watch-loc/watch-loc a D
92842 compilation unit at offset 0xc7:
92844 Compilation Unit @ offset 0xc7:
92845 Length: 0x4c (32-bit)
92847 Abbrev Offset: 0x64
92849 <0><d2>: Abbrev Number: 2 (DW_TAG_compile_unit)
92850 <d3> DW_AT_language : 19 (D)
92852 with a corresponding .debug_aranges entry:
92854 Offset into .debug_info: 0xc7
92863 After that commit we have a dummy CU at offset 0xc7 and the D compilation unit
92866 Compilation Unit @ offset 0xc7:
92867 Length: 0x7 (32-bit)
92869 Abbrev Offset: 0x64
92871 Compilation Unit @ offset 0xd2:
92872 Length: 0x4c (32-bit)
92874 Abbrev Offset: 0x65
92876 <0><dd>: Abbrev Number: 2 (DW_TAG_compile_unit)
92877 <de> DW_AT_language : 19 (D)
92879 while the .debug_aranges entry still points to 0xc7.
92881 The problem is that the test-case uses a hack (quoting from
92882 commit 75f06e9dc59):
92884 [ Note: this is a non-trivial test-case. The file watch-loc-dw.S contains a
92885 .debug_info section, but not an .debug_aranges section or any actual code.
92886 The file watch-loc.c contains code and a .debug_aranges section, but no other
92887 debug section. So, the intent for the .debug_aranges section in watch-loc.c
92888 is to refer to a compilation unit in the .debug_info section in
92891 and adding the dummy CU caused that hack to stop working.
92893 Fix this by moving the generation of .debug_aranges from watch-loc.c to
92894 watch-loc.exp, such that we have:
92896 Offset into .debug_info: 0xd2
92905 Tested on x86_64-linux.
92907 2021-08-27 Tom de Vries <tdevries@suse.de>
92909 [gdb/testsuite] Generate .debug_aranges entry for dummy CU
92910 A best practise for DWARF [1] is to generate .debug_aranges entries for CUs
92911 even if they have no address range.
92913 Generate .debug_arange entries for the dummy CUs added by the DWARF assembler.
92915 Tested on x86_64-linux.
92917 [1] http://wiki.dwarfstd.org/index.php?title=Best_Practices
92919 2021-08-27 Tom de Vries <tdevries@suse.de>
92921 [gdb/testsuite] Add .debug_aranges in more test-cases
92922 A couple of test-cases fail when run with target board cc-with-debug-names due
92923 to missing .debug_aranges entries for the CUs added by the dwarf assembler.
92925 Add a .debug_aranges entry for those CUs.
92927 Tested on x86_64-linux.
92929 2021-08-27 Tom de Vries <tdevries@suse.de>
92931 [gdb/testsuite] Support .debug_aranges in dwarf assembly
92932 Add a proc aranges such that we can generate .debug_aranges sections in dwarf
92935 cu { label cu_label } {
92939 aranges {} cu_label {
92940 arange $addr $len [<comment>] [$segment_selector]
92944 Tested on x86_64-linux.
92946 2021-08-27 Tom de Vries <tdevries@suse.de>
92948 [gdb/testsuite] Add label option to proc cu
92949 We can use current dwarf assembly infrastructure to declare a label that marks
92950 the start of the CU header:
92952 declare_labels header_start_cu_a
92953 _section ".debug_info"
92954 header_start_cu_a : cu {} {
92956 _section ".debug_info"
92957 header_start_cu_b : cu {} {
92960 on the condition that we switch to the .debug_info section before, which makes
92961 this style of use fragile.
92963 Another way to achieve the same is to use the label as generated by the cu
92969 set header_start_cu_a $_cu_label
92972 set header_start_cu_b $_cu_label
92974 but again that seems fragile given that adding a new CU inbetween will
92975 silently result in the wrong value for the label.
92977 Add a label option to proc cu such that we can simply do:
92979 cu { label header_start_cu_a } {
92981 cu { label header_start_cu_b } {
92985 Tested on x86_64-linux.
92987 2021-08-27 GDB Administrator <gdbadmin@sourceware.org>
92989 Automatic date update in version.in
92991 2021-08-26 Andrew Burgess <andrew.burgess@embecosm.com>
92993 gdb: remove some stray newlines in debug output
92994 I spotted a couple of stray newlines that were left at the end of
92995 debug message during conversion to the new debug output scheme. These
92996 messages are part of the 'set debug lin-lwp 1' output.
92998 2021-08-26 GDB Administrator <gdbadmin@sourceware.org>
93000 Automatic date update in version.in
93002 2021-08-25 GDB Administrator <gdbadmin@sourceware.org>
93004 Automatic date update in version.in
93006 2021-08-24 Tom Tromey <tom@tromey.com>
93008 Fix two regressions caused by CU / TU merging
93009 PR symtab/28160 and PR symtab/27893 concern GDB crashes in the test
93010 suite when using the "fission" target board. They are both caused by
93011 the patches that merge the list of CUs with the list of TUs (and to a
93012 lesser degree by the patches to share DWARF data across objfiles), and
93013 the underlying issue is the same: it turns out that reading a DWO can
93014 cause new type units to be created. This means that the list of
93015 dwarf2_per_cu_data objects depends on precisely which CUs have been
93016 expanded. However, because the type units can be created while
93017 expanding a CU means that the vector of CUs can expand while it is
93018 being iterated over -- a classic mistake. Also, because a TU can be
93019 added later, it means the resize_symtabs approach is incorrect.
93021 This patch fixes resize_symtabs by removing it, and having set_symtab
93022 resize the vector on demand. It fixes the iteration problem by
93023 introducing a safe (index-based) iterator and changing the relevant
93026 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28160
93027 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=27893
93029 2021-08-24 Alan Modra <amodra@gmail.com>
93031 Real programmers don't configure gcc using --with-ld
93032 * testsuite/lib/ld-lib.exp (run_host_cmd): Give a clue as to why
93033 gcc -B doesn't pick up the ld under test.
93035 2021-08-24 Alan Modra <amodra@gmail.com>
93037 objdump -S test fail on mingw
93039 FAIL: objdump --source-comment
93040 is seen on mingw for the simple reason that gcc adds a .exe suffix on
93041 the output file if not already present. Fix that, and tidy some objcopy
93044 * testsuite/lib/binutils-common.exp (exeext): New proc.
93045 * testsuite/binutils-all/objcopy.exp (exe, test_prog): Use it here.
93046 (objcopy_remove_relocations_from_executable): Catch objcopy errors.
93047 Only run on ELF targets.
93048 * testsuite/binutils-all/objdump.exp (exe): Set variable.
93049 (test_build_id_debuglink, test_objdump_S): Use exe file suffix.
93051 2021-08-24 James Bowman (FTDI-UK) <james.bowman@ftdichip.com>
93053 FT32: Remove recursion in ft32_opcode
93054 The function ft32_opcode used recursion. This could cause a stack
93055 overflow. Replaced with a pair of non-recursive functions.
93058 * ft32-dis.c: Formatting.
93059 (ft32_opcode1): Split out from..
93060 (ft32_opcode): ..here.
93062 2021-08-24 GDB Administrator <gdbadmin@sourceware.org>
93064 Automatic date update in version.in
93066 2021-08-23 Tom Tromey <tom@tromey.com>
93068 Fix a latent bug in dw2-ranges-overlap.exp
93069 dw2-ranges-overlap.exp creates a program where a psymtab has two
93070 address ranges, and a function without debug info whose address is
93071 between these two ranges. Then it sets a breakpoint on this function
93072 and runs to it, expecting that the language should remain "auto; c"
93075 However, this test case also has a "main" function described (briefly)
93076 in the DWARF, and this function is given language C++. Also, a
93077 breakpoint stop sets the current language to the language that was
93078 used when setting the breakpoint.
93080 My new DWARF scanner decides that this "main" is the main program and
93081 sets the current language to C++ at startup, causing this test to
93084 This patch fixes the test in a simple way, by introducing a new
93085 function that takes the place of "main" in the DWARF. I think this
93086 still exercises the original problem, but also avoids problems with my
93089 It seemed safe to me to submit this separately.
93091 2021-08-23 Tom de Vries <tdevries@suse.de>
93093 [gdb] Fix 'not in executable format' error message
93094 With trying to load a non-executable file into gdb, we run into PR26880:
93096 $ gdb -q -batch test.c
93097 "0x7ffc87bfc8d0s": not in executable format: \
93098 file format not recognized
93101 The problem is caused by using %ps in combination with the error function
93102 (note that confusingly, it does work in combination with the warning
93105 Fix this by using plain "%s" instead.
93107 Tested on x86_64-linux.
93111 2021-08-22 Tom de Vries <tdevries@suse.de>
93114 * gdb/exec.c (exec_file_attach): Use %s instead of %ps in call to
93117 gdb/testsuite/ChangeLog:
93119 2021-08-22 Tom de Vries <tdevries@suse.de>
93122 * gdb.base/non-executable.exp: New file.
93124 2021-08-23 Tom de Vries <tdevries@suse.de>
93126 [gdb/testsuite] Use compiler-generated instead of gas-generated stabs
93127 The test-case gdb.dwarf2/dw2-ranges.exp is the only one in the gdb testsuite
93128 that uses gas-generated stabs.
93130 While the use seems natural alongside the use of gas-generated dwarf in the
93131 same test-case, there are a few known issues, filed on the gdb side as:
93132 - PR symtab/12497 - "stabs: PIE relocation does not work"
93133 - PR symtab/28221 - "[readnow, stabs] FAIL: gdb.dwarf2/dw2-ranges.exp: \
93135 and on the gas side as:
93136 - PR gas/28233 - "[gas, --gstabs] Generate stabs more similar to gcc"
93138 The test-case contains a KFAIL for PR12497, but it's outdated and fails to
93141 The intention of the test-case is to test gas-generated dwarf, and using
93142 gcc-generated stabs instead of gas-generated stabs works fine.
93144 Supporting compiler-generated stabs is already a corner-case for gdb, and
93145 there's no current commitment/incentive to support/workaround gas-generated
93146 stabs, which can be considered a corner-case of a corner-case.
93148 Work around these problem by using compiler-generated stabs in the test-case.
93150 Tested on x86_64-linux.
93152 gdb/testsuite/ChangeLog:
93154 2021-08-22 Tom de Vries <tdevries@suse.de>
93156 * gdb.dwarf2/dw2-ranges.exp: Use compiler-generated stabs.
93158 2021-08-23 Tom de Vries <tdevries@suse.de>
93160 [gdb/testsuite] Add dummy start and end CUs in dwarf assembly
93161 Say one compiles a hello.c:
93166 On openSUSE Leap 15.2 and Tumbleweed, the CU for hello.c is typically not the
93167 first in .debug_info, nor the last, due to presence of debug information in
93168 objects for sources like:
93169 - ../sysdeps/x86_64/start.S
93171 - ../sysdeps/x86_64/crti.S
93173 - ../sysdeps/x86_64/crtn.S.
93175 On other systems, say ubuntu 18.04.5, the CU for hello.c is typically the
93176 first and the last in .debug_info.
93178 This difference has caused me to find some errors in the dwarf assembly
93179 using openSUSE, that didn't show up on other platforms.
93181 Force the same situation on other platforms by adding a dummy start
93184 Tested on x86_64-linux.
93186 gdb/testsuite/ChangeLog:
93188 2021-08-22 Tom de Vries <tdevries@suse.de>
93191 * lib/dwarf.exp (Dwarf::dummy_cu): New proc.
93192 (Dwarf::assemble): Add dummy start and end CU.
93194 2021-08-23 Tom de Vries <tdevries@suse.de>
93196 [gdb/testsuite] Fix dw2-ranges-psym.exp with -readnow
93197 When running test-case gdb.dwarf2/dw2-ranges-psym.exp with target board
93198 -readnow, I run into:
93200 (gdb) file dw2-ranges-psym^M
93201 Reading symbols from dw2-ranges-psym...^M
93202 Expanding full symbols from dw2-ranges-psym...^M
93203 (gdb) set complaints 0^M
93204 (gdb) FAIL: gdb.dwarf2/dw2-ranges-psym.exp: No complaints
93207 The problem is that the regexp expects a gdb prompt immediately after the
93208 "Reading symbols" line.
93210 Fix this by updating the regexp.
93212 Tested on x86_64-linux.
93214 gdb/testsuite/ChangeLog:
93216 2021-08-22 Tom de Vries <tdevries@suse.de>
93218 * lib/gdb.exp (gdb_load_no_complaints): Update regexp to allow
93219 "Expanding full symbols" Line.
93221 2021-08-23 GDB Administrator <gdbadmin@sourceware.org>
93223 Automatic date update in version.in
93225 2021-08-22 Mike Frysinger <vapier@gentoo.org>
93227 sim: m32r: add __linux__ hack for non-Linux hosts
93228 The m32r Linux syscall emulation logic assumes the host environment
93229 directly matches -- it's being run on 32-bit little endian Linux.
93230 This breaks building for non-Linux systems, so put all the code in
93231 __linux__ ifdef checks. This code needs a lot of love to make it
93232 work everywhere, but let's at least unbreak it for non-Linux hosts.
93234 2021-08-22 GDB Administrator <gdbadmin@sourceware.org>
93236 Automatic date update in version.in
93238 2021-08-21 GDB Administrator <gdbadmin@sourceware.org>
93240 Automatic date update in version.in
93242 2021-08-20 Mike Frysinger <vapier@gentoo.org>
93244 sim: nltvals: switch output mode to a directory
93245 In preparation for this script generating more files, change the output
93246 argument to specify a directory. This drops the stdout behavior, but
93247 since no one really runs this tool directly, it's not a big deal.
93249 2021-08-20 GDB Administrator <gdbadmin@sourceware.org>
93251 Automatic date update in version.in
93253 2021-08-19 Simon Marchi <simon.marchi@polymtl.ca>
93255 gdb: use bool in notify_command_param_changed_p and do_set_command
93256 Trivial patch to use bool instead of int.
93258 Change-Id: I9e5f8ee4305272a6671cbaaaf2f0484eff0d1ea5
93260 2021-08-19 H.J. Lu <hjl.tools@gmail.com>
93262 x86: Put back 3 aborts in OP_E_memory
93263 Put back 3 aborts where invalid lengths should have been filtered out.
93268 * testsuite/gas/i386/bad-bcast.s: Add a comment.
93273 * * i386-dis.c (OP_E_memory): Put back 3 aborts.
93275 2021-08-19 H.J. Lu <hjl.tools@gmail.com>
93277 x86: Avoid abort on invalid broadcast
93278 Print "{bad}" on invalid broadcast instead of abort.
93283 * testsuite/gas/i386/bad-bcast.d: New file.
93284 * testsuite/gas/i386/bad-bcast.s: Likewise.
93285 * testsuite/gas/i386/i386.exp: Run bad-bcast.
93290 * i386-dis.c (OP_E_memory): Print "{bad}" on invalid broadcast
93293 2021-08-19 Aaron Merey <amerey@redhat.com>
93295 gdb/solib: Refactor scan_dyntag
93296 scan_dyntag is unnecessarily duplicated in solib-svr4.c and solib-dsbt.c.
93298 Move this function to solib.c and rename it to gdb_bfd_scan_elf_dyntag.
93299 Also add it to solib.h so it is included in both solib-svr4 and solib-dsbt.
93301 2021-08-19 GDB Administrator <gdbadmin@sourceware.org>
93303 Automatic date update in version.in
93305 2021-08-18 Will Schmidt <will_schmidt@vnet.ibm.com>
93307 [gdb] [rs6000] Add ppc64_linux_gcc_target_options method.
93308 Add a method to set the gcc target options for the ppc64 targets.
93309 This change sets an empty value, which allows the gcc
93310 default values (-mcmodel=medium) be used, instead of -mcmodel=large
93311 which is set by the default_gcc_target_options hook.
93313 [gdb] [rs6000] Add ppc64*_gnu_triplet_regexp methods.
93314 Add methods to set the target triplet so we can
93315 find the proper gcc when our gcc is named of
93316 the form powerpc64{le}-<foo>-gcc or ppc64{le}-<foo>-gcc.
93318 2021-08-18 Alan Modra <amodra@gmail.com>
93320 Re: as: Replace the removed symbol with the versioned symbol
93321 Some targets, typically embedded without shared libraries, replace the
93322 relocation symbol with a section symbol (see tc_fix_adjustable).
93323 Allow the test to pass for such targets. Fixes the following.
93325 avr-elf +FAIL: symver symver16
93326 d10v-elf +FAIL: symver symver16
93327 dlx-elf +FAIL: symver symver16
93328 ip2k-elf +FAIL: symver symver16
93329 m68k-elf +FAIL: symver symver16
93330 mcore-elf +FAIL: symver symver16
93331 pj-elf +FAIL: symver symver16
93332 s12z-elf +FAIL: symver symver16
93333 visium-elf +FAIL: symver symver16
93334 z80-elf +FAIL: symver symver16
93337 * testsuite/gas/symver/symver16.d: Relax reloc match.
93339 2021-08-18 Alan Modra <amodra@gmail.com>
93341 [GOLD] PowerPC64 relocation overflow for -Os register save/restore funcs
93342 Fixes a silly mistake in calculating the address of -Os out-of-line
93343 register save/restore function copies. Copies of these linker defined
93344 functions are added to stub sections when the original (in
93345 target->savres_section) can't be reached.
93347 * powerpc.cc (Target_powerpc::Relocate::relocate): Correct address
93348 calculation of out-of-line save/restore function copies.
93350 2021-08-18 Alan Modra <amodra@gmail.com>
93352 Another ld script backtrack
93353 * ldgram.y (length_spec): Throw away look-ahead NAME.
93355 2021-08-18 Mike Frysinger <vapier@gentoo.org>
93357 gdb: fix spacing on CCLD silent rules
93359 sim: nltvals: localize TARGET_<ERRNO> defines
93360 Code should not be using these directly, instead they should be
93361 resolving these dynamically via cb_host_to_target_errno maps.
93362 Fix the Blackfin code and remove the defines out of the header
93363 so no new code can rely on them.
93365 2021-08-18 Mike Frysinger <vapier@gentoo.org>
93367 sim: rename ChangeLog files to ChangeLog-2021
93368 Now that ChangeLog entries are no longer used for sim patches,
93369 this commit renames all relevant sim ChangeLog to ChangeLog-2021,
93370 similar to what we would do in the context of the "Start of New
93373 The purpose of this change is to avoid people merging ChangeLog
93374 entries by mistake when applying existing commits that they are
93375 currently working on.
93377 Also throw in a .gitignore entry to keep people from adding new
93378 ChangeLog files anywhere in the sim tree.
93380 2021-08-18 GDB Administrator <gdbadmin@sourceware.org>
93382 Automatic date update in version.in
93384 2021-08-17 Simon Marchi <simon.marchi@polymtl.ca>
93386 gdb: fix thread_step_over_chain_length
93387 If I debug a single-thread program and look at the infrun debug logs, I
93390 [infrun] start_step_over: stealing global queue of threads to step, length = 2
93392 That makes no sense... turns out there's a buglet in
93393 thread_step_over_chain_length, "num" should be initialized to 0. I
93394 think this bug is a leftover from an earlier version of the code (not
93395 merged upstream) that manually walked the list, where the first item was
93396 implicitly counted (hence the 1).
93398 Change-Id: I0af03aa93509aed36528be5076894dc156a0b5ce
93400 2021-08-17 Shahab Vahedi <shahab@synopsys.com>
93402 opcodes: Fix the auxiliary register numbers for ARC HS
93403 The numbers for the auxiliary registers "tlbindex" and
93404 "tlbcommand" of ARCv2HS are incorrect. This patch makes
93405 the following changes to correct that error.
93407 ,------------.-----------------.---------------.
93408 | aux. reg. | old (incorrect) | new (correct) |
93409 |------------+-----------------+---------------|
93410 | tlbindex | 0x463 | 0x464 |
93411 | tlbcommand | 0x464 | 0x465 |
93412 `------------^-----------------^---------------'
93415 2021-08-17 Shahab Vahedi <shahab@synopsys.com>
93417 * arc-regs.h (DEF): Fix the register numbers.
93419 2021-08-17 H.J. Lu <hjl.tools@gmail.com>
93421 gdb: Don't assume r_ldsomap when r_version > 1 on Linux
93422 The r_ldsomap field is specific to Solaris (part of librtld_db), and
93423 should never be accessed for Linux. glibc is planning to add a field
93424 to support multiple namespaces. But there will be no r_ldsomap when
93425 r_version is bumped to 2. Add linux_[ilp32|lp64]_fetch_link_map_offsets
93426 to set r_ldsomap_offset to -1 and use them for Linux targets.
93428 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28236
93430 2021-08-17 H.J. Lu <hjl.tools@gmail.com>
93432 gdbserver: Check r_version < 1 for Linux debugger interface
93433 Update gdbserver to check r_version < 1 instead of r_version != 1 so
93434 that r_version can be bumped for a new field in the glibc debugger
93435 interface to support multiple namespaces. Since so far, the gdbserver
93436 only reads fields defined for r_version == 1, it is compatible with
93439 All future glibc debugger interface changes will be backward compatible.
93440 If there is ever the need for backward incompatible change to the glibc
93441 debugger interface, a new DT_XXX element will be provided to access the
93442 new incompatible interface.
93444 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=11839
93446 2021-08-17 Andrea Corallo <andrea.corallo@arm.com>
93448 PATCH [4/4] arm: Add Tag_PACRET_use build attribute
93450 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93452 * elf32-arm.c (elf32_arm_merge_eabi_attributes): Add
93453 'Tag_PACRET_use' case.
93456 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93458 * readelf.c (arm_attr_tag_PAC_extension): Declare.
93459 (arm_attr_public_tags): Add 'PAC_extension' lookup.
93462 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93464 * arm.h: Define 'Tag_PACRET_use' enum.
93467 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93469 * config/tc-arm.c (arm_convert_symbolic_attribute): Add
93470 'Tag_PACRET_use' to the attribute_table.
93473 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93475 * elf/arm.h (elf_arm_reloc_type): Add 'Tag_PACRET_use'.
93477 2021-08-17 Andrea Corallo <andrea.corallo@arm.com>
93479 PATCH [3/4] arm: Add Tag_BTI_use build attribute
93481 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93483 * elf32-arm.c (elf32_arm_merge_eabi_attributes): Add
93484 'Tag_BTI_use' case.
93487 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93489 * readelf.c (arm_attr_tag_PAC_extension): Declare.
93490 (arm_attr_public_tags): Add 'PAC_extension' lookup.
93493 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93495 * arm.h: Define 'Tag_BTI_use' enum.
93498 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93500 * config/tc-arm.c (arm_convert_symbolic_attribute): Add
93501 'Tag_BTI_use' to the attribute_table.
93504 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93506 * elf/arm.h (elf_arm_reloc_type): Add 'Tag_BTI_use'.
93508 2021-08-17 Andrea Corallo <andrea.corallo@arm.com>
93510 PATCH [2/4] arm: Add Tag_BTI_extension build attribute
93512 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93514 * elf32-arm.c (elf32_arm_merge_eabi_attributes): Add
93515 'Tag_BTI_extension' case.
93518 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93520 * readelf.c (arm_attr_tag_PAC_extension): Declare.
93521 (arm_attr_public_tags): Add 'PAC_extension' lookup.
93524 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93526 * arm.h: Define 'Tag_BTI_extension' enum.
93529 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93531 * config/tc-arm.c (arm_convert_symbolic_attribute): Add
93532 'Tag_BTI_extension' to the attribute_table.
93535 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93537 * elf/arm.h (elf_arm_reloc_type): Add 'Tag_BTI_extension'.
93539 2021-08-17 Andrea Corallo <andrea.corallo@arm.com>
93541 PATCH [1/4] arm: Add Tag_PAC_extension build attribute
93543 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93545 * elf32-arm.c (elf32_arm_merge_eabi_attributes): Add
93546 'Tag_PAC_extension' case.
93549 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93551 * readelf.c (arm_attr_tag_PAC_extension): Declare.
93552 (arm_attr_public_tags): Add 'PAC_extension' lookup.
93555 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93557 * arm.h: Define 'Tag_PAC_extension' enum.
93560 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93562 * config/tc-arm.c (arm_convert_symbolic_attribute): Add
93563 'Tag_PAC_extension' to the attribute_table.
93566 2021-07-06 Andrea Corallo <andrea.corallo@arm.com>
93568 * elf/arm.h (elf_arm_reloc_type): Add 'Tag_PAC_extension'.
93570 2021-08-17 H.J. Lu <hjl.tools@gmail.com>
93572 x86: Always run fp tests
93573 Always run fp tests since the size of .tfloat, .ds.x, .dc.x and .dcb.x
93574 directive outputs is always 10 bytes. There is no need for fp-elf32 nor
93578 * testsuite/gas/i386/fp-elf32.d: Removed.
93579 * testsuite/gas/i386/fp-elf64.d: Likewise.
93580 * testsuite/gas/i386/fp.s: Remove NO_TFLOAT_PADDING codes.
93581 * testsuite/gas/i386/i386.exp: Don't run fp-elf32 nor fp-elf64.
93584 2021-08-17 GDB Administrator <gdbadmin@sourceware.org>
93586 Automatic date update in version.in
93588 2021-08-16 H.J. Lu <hjl.tools@gmail.com>
93590 x86: Don't pad .tfloat directive output
93591 .tfloat output should always be 10 bytes without padding, independent
93592 of psABIs. In glibc, x86 assembly codes expect 10-byte .tfloat output.
93593 This also reduces .ds.x output and .tfloat output with hex input from
93594 12 bytes to 10 bytes to match .tfloat output.
93597 * NEWS: Mention changes of .ds.x output and .tfloat output with
93599 * config/tc-i386.c (x86_tfloat_pad): Removed.
93600 * config/tc-i386.h (X_PRECISION_PAD): Changed to 0.
93601 (x86_tfloat_pad): Removed.
93602 * testsuite/gas/i386/fp.s: If NO_TFLOAT_PADDING isn't defined,
93603 add explicit paddings after .tfloat, .ds.x, .dc.x and .dcb.x
93605 * testsuite/gas/i386/i386.exp (ASFLAGS): Append
93606 "--defsym NO_TFLOAT_PADDING=1" when running the fp test.
93608 2021-08-16 Tom Tromey <tromey@adacore.com>
93610 Fix register regression in DWARF evaluator
93611 On an internal test case, using an arm-elf target, commit ba5bc3e5a92
93612 ("Make DWARF evaluator return a single struct value") causes a
93613 regression. (It doesn't happen for any of the other cross targets
93614 that I test when importing upstream gdb.)
93616 I don't know if there's an upstream gdb test case showing the same
93617 problem... I can only really run native tests with dejagnu AFAIK.
93619 The failure manifests like this:
93621 Breakpoint 1, file_1.export_1 (param_1=<error reading variable: Unable to access DWARF register number 64>, str=...) at [...]/file_1.adb:5
93623 Whereas when it works it looks like:
93625 Breakpoint 1, file_1.export_1 (param_1=99.0, str=...) at [...]/file_1.adb:5
93627 The difference is that the new code uses the passed-in gdbarch,
93628 whereas the old code used the frame's gdbarch, when handling
93629 DWARF_VALUE_REGISTER.
93631 This patch restores the use of the frame's arch.
93633 2021-08-16 Tom Tromey <tromey@adacore.com>
93635 Fix Ada regression due to DWARF expression series
93636 Commit 0579205aec4 ("Simplify dwarf_expr_context class interface")
93637 caused a regression in the internal AdaCore test suite. I didn't try
93638 to reproduce this with the GDB test suite, but the test is identical
93639 to gdb.dwarf2/dynarr-ptr.exp.
93641 The problem is that this change:
93643 case DW_OP_push_object_address:
93644 /* Return the address of the object we are currently observing. */
93645 - if (this->data_view.data () == nullptr
93646 - && this->obj_address == 0)
93647 + if (this->m_addr_info == nullptr)
93649 ... slightly changes the logic here. In particular, it's possible for
93650 the caller to pass in a non-NULL m_addr_info, but one that looks like:
93652 (top) p *this.m_addr_info
93663 In this case, an additional check is needed. With the current code,
93664 what happens instead is that the computation computes an incorrect
93665 address -- but one that does not fail in read_memory, due to the
93666 precise memory map of the embedded target in question.
93668 This patch restores the old logic.
93670 2021-08-16 Patrick Monnerat <patrick@monnerat.net>
93672 Notify observer of breakpoint auto-disabling
93673 As breakpoint_modified observer is currently notified upon breakpoint stop
93674 before handling auto-disabling when enable count is reached, the observer
93675 is never notified of the disabling.
93677 The problem affects:
93678 - The MI interpreter enabled= value when reporting =breakpoint-modified
93679 - A Python event handler for breakpoint_modified using the "enabled"
93680 member of its parameter
93681 - insight: breakpoint GUI window is not properly updated upon auto-disable
93683 This patch moves the observer notification after the auto-disabling
93684 code and implements corresponding tests for the MI and Python cases.
93686 Fixes https://sourceware.org/bugzilla/show_bug.cgi?id=23336
93688 Change-Id: I0c50df4789334071e5390cb46b3ca0d4a7f83c61
93690 2021-08-16 H.J. Lu <hjl.tools@gmail.com>
93692 as: Replace the removed symbol with the versioned symbol
93693 When a symbol removed by .symver is used in relocation and there is one
93694 and only one versioned symbol, don't remove the symbol. Instead, mark
93695 it to be removed and replace the removed symbol used in relocation with
93696 the versioned symbol before generating relocation.
93699 * symbols.c (symbol_flags): Add removed.
93700 (symbol_entry_find): Updated.
93701 (symbol_mark_removed): New function.
93702 (symbol_removed_p): Likewise.
93703 * symbols.h (symbol_mark_removed): New prototype.
93704 (symbol_removed_p): Likewise.
93705 * write.c (write_relocs): Call obj_fixup_removed_symbol on
93706 removed fixp->fx_addsy and fixp->fx_subsy if defined.
93707 (set_symtab): Don't add a symbol if symbol_removed_p returns true.
93708 * config/obj-elf.c (elf_frob_symbol): Don't remove the symbol
93709 if it is used on relocation. Instead, mark it as to be removed
93710 and issue an error if the symbol has more than one versioned name.
93711 (elf_fixup_removed_symbol): New function.
93712 * config/obj-elf.h (elf_fixup_removed_symbol): New prototype.
93713 (obj_fixup_removed_symbol): New.
93714 * testsuite/gas/symver/symver11.d: Updated expected error
93716 * testsuite/gas/symver/symver16.d: New file.
93717 * testsuite/gas/symver/symver16.s: Likewise.
93719 2021-08-16 GDB Administrator <gdbadmin@sourceware.org>
93721 Automatic date update in version.in
93723 2021-08-15 GDB Administrator <gdbadmin@sourceware.org>
93725 Automatic date update in version.in
93727 2021-08-14 Alan Modra <amodra@gmail.com>
93729 ld script fill pattern expression
93730 It turns out we do need to backtrack when parsing after all. The
93731 fill_opt component in the section rule swiches to EXPRESSION and back
93732 to SCRIPT, and to find the end of an expression it is necessary to
93733 look ahead one token.
93735 * ldgram.y (section): Throw away lookahead NAME token.
93736 (overlay_section): Likewise.
93737 * testsuite/ld-elf/overlay.t: Add fill pattern on overlays.
93738 Test fill pattern before stupidly named normal sections too,
93739 and before /DISCARD/.
93741 2021-08-14 GDB Administrator <gdbadmin@sourceware.org>
93743 Automatic date update in version.in
93745 2021-08-13 Alan Modra <amodra@gmail.com>
93747 ld lexer tidy, possibly break the world
93748 This tidies the states in which ld lexer rules are enabled.
93749 This change will quite likely trip over issues similar to those
93750 mentioned in the new ldlex.l comments, so please test it out.
93752 * ldgram.y (wildcard_name): Remove now unnecessary components.
93753 * ldlex.l: Restrict many rules' states. Remove -l expression
93754 state rule. Comment on lookahead state madness and need for
93755 /DISCARD/ in expression state.
93757 2021-08-13 Alan Modra <amodra@gmail.com>
93759 ld script lower-case absolute and sizeof_headers
93760 I think these happened by accident, so let's see what breaks if they
93763 * ldlex.l: Remove lower case "absolute" and "sizeof_headers"
93765 * ld.texi: Remove sizeof_headers index.
93766 * testsuite/ld-mmix/mmohdr1.ld: Use SIZEOF_HEADERS.
93768 2021-08-13 Alan Modra <amodra@gmail.com>
93770 tidy mri script extern
93771 MRI mode generally doesn't flip lexer states, so let's make MRI mode
93772 "extern" not do so either.
93774 * ldgram.y (extern_name_list): Don't change lex state here.
93775 (ifile_p1): Change state here on EXTERN instead.
93777 2021-08-13 Alan Modra <amodra@gmail.com>
93779 Re: PR28217, Syntax error when memory region contains a hyphen
93780 I discovered some more errors when tightening up the lexer rules.
93781 Just because we INCLUDE a file doesn't mean we've switched states.
93784 * ldgram.y (statement): Don't switch lexer state on INCLUDE.
93785 (mri_script_command, ifile_p1, memory_spec, section): Likewise.
93787 2021-08-13 Lifang Xia <lifang_xia@c-sky.com>
93789 PR28168: [CSKY] Fix stack overflow in disassembler
93791 Stack overflow with a large float. %f is not a goot choice for this.
93792 %f should be replaced with %.7g.
93795 * testsuite/gas/csky/pr28168.d: New testcase for PR 28168.
93796 * testsuite/gas/csky/pr28168.s: Likewise.
93797 * testsuite/gas/csky/v2_float_part2.d: Following the new format.
93798 * opcodes/csky-dis.c (csky_output_operand): %.7g replaces %f.
93800 2021-08-13 Alan Modra <amodra@gmail.com>
93802 PR28217, Syntax error when memory region contains a hyphen
93803 The saga of commit 40726f16a8d7 continues. This attacks the problem
93804 of switching between SCRIPT and EXPRESSION state lexing by removing
93805 the need to do so for phdrs like ":text". Instead {WILDCHAR}*
93806 matching, the reason why ":text" lexed as one token, is restricted to
93807 within the braces of a section or overlay statement. The new WILD
93808 lexer state is switched at the non-optional brace tokens, so
93809 ldlex_backup is no longer needed. I've also removed the BOTH state,
93810 which doesn't seem to be needed any more. Besides rules involving
93811 error reporting, there was just one place where SCRIPT appeared
93812 without BOTH, the {WILDCHAR}* rule, three where BOTH appears without
93813 SCRIPT for tokens that only need EXPRESSION state, and two where BOTH
93814 appears alongside INPUT_LIST. (Since I'm editing the wild and
93815 filename rules, removing BOTH and adding WILD can also be seen as
93816 renaming the old BOTH state to SCRIPT and renaming the old SCRIPT
93817 state to WILD with a reduced scope.)
93819 As a followup, I'll look at removing EXPRESSION state from some lexer
93820 rules that no longer need it due to this cleanup.
93823 * ldgram.y (exp <ORIGIN, LENGTH>): Use paren_script_name.
93824 (section): Parse within braces of section in wild mode, and
93825 after brace back in script mode. Remove ldlex_backup call.
93826 Similarly for OVERLAY.
93827 (overlay_section): Similarly.
93828 (script_file): Replace ldlex_both with ldlex_script.
93829 * ldlex.h (ldlex_wild): Declare.
93830 (ldlex_both): Delete.
93831 * ldlex.l (BOTH): Delete. Remove state from all rules.
93832 (WILD): New state. Enable many tokens in this state.
93833 Enable filename match in SCRIPT mode. Enable WILDCHAR match
93834 in WILD state, disable in SCRIPT mode.
93835 (ldlex_wild): New function.
93836 * ldfile.c (ldfile_try_open_bfd): Replace ldlex_both call with
93839 2021-08-13 Alan Modra <amodra@gmail.com>
93842 Since ns32k-netbsd is as yet not removed, just marked obsolete,
93843 the target should still be accepted with --enable-obsolete.
93845 I also enabled ns32k-openbsd in ld since there doesn't seem to be a
93846 good reason why that target is not supported there but is elsewhere.
93849 * config.bfd: Allow ns32k-netbsd.
93851 * configure.tgt: Allow ns32k-openbsd.
93853 2021-08-13 GDB Administrator <gdbadmin@sourceware.org>
93855 Automatic date update in version.in
93857 2021-08-13 Lancelot SIX <lsix@lancelotsix.com>
93859 gdb: riscv_scan_prologue: handle LD and LW instructions
93860 While working on the testsuite, I ended up noticing that GDB fails to
93861 produce a full backtrace from a thread waiting in pthread_join. When
93862 selecting the waiting thread and using the 'bt' command, the following
93863 result can be observed:
93866 #0 0x0000003ff7fccd20 in __futex_abstimed_wait_common64 () from /lib/riscv64-linux-gnu/libpthread.so.0
93867 #1 0x0000003ff7fc43da in __pthread_clockjoin_ex () from /lib/riscv64-linux-gnu/libpthread.so.0
93868 Backtrace stopped: frame did not save the PC
93870 On my platform, I do not have debug symbols for glibc, so I need to rely
93871 on prologue analysis in order to unwind stack.
93873 Here is what the function prologue looks like:
93875 (gdb) disassemble __pthread_clockjoin_ex
93876 Dump of assembler code for function __pthread_clockjoin_ex:
93877 0x0000003ff7fc42de <+0>: addi sp,sp,-144
93878 0x0000003ff7fc42e0 <+2>: sd s5,88(sp)
93879 0x0000003ff7fc42e2 <+4>: auipc s5,0xd
93880 0x0000003ff7fc42e6 <+8>: ld s5,-2(s5) # 0x3ff7fd12e0
93881 0x0000003ff7fc42ea <+12>: ld a5,0(s5)
93882 0x0000003ff7fc42ee <+16>: sd ra,136(sp)
93883 0x0000003ff7fc42f0 <+18>: sd s0,128(sp)
93884 0x0000003ff7fc42f2 <+20>: sd s1,120(sp)
93885 0x0000003ff7fc42f4 <+22>: sd s2,112(sp)
93886 0x0000003ff7fc42f6 <+24>: sd s3,104(sp)
93887 0x0000003ff7fc42f8 <+26>: sd s4,96(sp)
93888 0x0000003ff7fc42fa <+28>: sd s6,80(sp)
93889 0x0000003ff7fc42fc <+30>: sd s7,72(sp)
93890 0x0000003ff7fc42fe <+32>: sd s8,64(sp)
93891 0x0000003ff7fc4300 <+34>: sd s9,56(sp)
93892 0x0000003ff7fc4302 <+36>: sd a5,40(sp)
93894 As far as prologue analysis is concerned, the most interesting part is
93895 done at address 0x0000003ff7fc42ee (<+16>): 'sd ra,136(sp)'. This stores
93896 the RA (return address) register on the stack, which is the information
93897 we are looking for in order to identify the caller.
93899 In the current implementation of the prologue scanner, GDB stops when
93900 hitting 0x0000003ff7fc42e6 (<+8>) because it does not know what to do
93901 with the 'ld' instruction. GDB thinks it reached the end of the
93902 prologue but have not yet reached the important part, which explain
93903 GDB's inability to unwind past this point.
93905 The section of the prologue starting at <+4> until <+12> is used to load
93906 the stack canary[1], which will then be placed on the stack at <+36> at
93907 the end of the prologue.
93909 In order to have the prologue properly handled, this commit proposes to
93910 add support for the ld instruction in the RISC-V prologue scanner.
93911 I guess riscv32 would use lw in such situation so this patch also adds
93912 support for this instruction.
93914 With this patch applied, gdb is now able to unwind past pthread_join:
93917 #0 0x0000003ff7fccd20 in __futex_abstimed_wait_common64 () from /lib/riscv64-linux-gnu/libpthread.so.0
93918 #1 0x0000003ff7fc43da in __pthread_clockjoin_ex () from /lib/riscv64-linux-gnu/libpthread.so.0
93919 #2 0x0000002aaaaaa88e in bar() ()
93920 #3 0x0000002aaaaaa8c4 in foo() ()
93921 #4 0x0000002aaaaaa8da in main ()
93923 I have had a look to see if I could reproduce this easily, but in my
93924 simple testcases using '-fstack-protector-all', the canary is loaded
93925 after the RA register is saved. I do not have a reliable way of
93926 generating a prologue similar to the problematic one so I forged one
93929 The testsuite have been run on riscv64 ubuntu 21.01 with no regression
93932 [1] https://en.wikipedia.org/wiki/Buffer_overflow_protection#Canaries
93934 2021-08-12 Tom Tromey <tromey@adacore.com>
93936 Update documentation to mention Pygments
93937 Philippe Blain pointed out that the gdb documentation does not mention
93938 that Pygments may be used for source highlighting. This patch updates
93939 the docs to reflect how highlighting is actually done.
93941 2021-08-12 Simon Marchi <simon.marchi@polymtl.ca>
93943 gdb: make gdbarch_printable_names return a vector
93944 I noticed that gdbarch_selftest::operator() leaked the value returned by
93945 gdbarch_printable_names. Make gdbarch_printable_names return an
93946 std::vector and update callers. That makes it easier for everyone
93947 involved, less manual memory management.
93949 Change-Id: Ia8fc028bdb91f787410cca34f10bf3c5a6da1498
93951 2021-08-12 Carl Love <cel@us.ibm.com>
93953 Improve forward progress test in python.exp
93954 The test steps into func2 and than does an up to get back to the previous
93955 frame. The test checks that the line number you are at after the up command
93956 is greater than the line where the function was called from. The
93957 assembly/codegen for the powerpc target includes a NOP after the
93960 func2 (); /* Break at func2 call site. /
93961 10000694: 59 00 00 48 bl 100006ec
93962 10000698: 00 00 00 60 nop
93963 return 0; / Break to end. */
93964 1000069c: 00 00 20 39 li r9,0
93966 The PC at the instruction following the branch-link is 0x10000698 which
93967 GDB.find_pc_line() maps to the same line number as the bl instruction.
93968 GDB did move past the branch-link location thus making forward progress.
93970 The following proposed fix adds an additional PC check to see if forward
93971 progress was made. The line test is changed from greater than to greater
93974 2021-08-12 Jiangshuai Li <jiangshuai_li@c-sky.com>
93976 gdb:csky rm tdesc_has_registers in csky_register_name
93977 As CSKY arch has not parsed target-description.xml in csky_gdbarch_init,
93978 when a remote server, like csky-qemu or gdbserver, send a target-description.xml
93979 to gdb, tdesc_has_registers will return ture, but tdesc_register_name (gdbarch, 0)
93980 will return NULL, so a cmd "info registers r0" will not work.
93982 Function of parsing target-description.xml will be add later for CSKY arch,
93983 now it is temporarily removed to allow me to do other supported tests.
93985 2021-07-15 Jiangshuai Li <jiangshuai_li@c-sky.com>
93987 * csky-tdep.c : not using tdesc funtions in csky_register_name
93989 2021-08-12 Alan Modra <amodra@gmail.com>
93991 Re: gas: support NaN flavors
93992 Fixes tic4x-coff FAIL: simple FP constants
93994 * testsuite/gas/all/float.s: Make NaN tests conditional on hasnan.
93995 * testsuite/gas/all/gas.exp: Define hasnan.
93997 2021-08-12 GDB Administrator <gdbadmin@sourceware.org>
93999 Automatic date update in version.in
94001 2021-08-11 H.J. Lu <hjl.tools@gmail.com>
94003 ld: Update the pass and fail strings of PR ld/28138 test
94005 * testsuite/ld-plugin/lto.exp: Update the pass and fail strings
94006 of PR ld/28138 test to indicate which part of the test passed
94009 2021-08-11 Darius Galis <darius.galis@cyberthorstudios.com>
94011 Fix a typo in the RX asse,bler. The Double-precision floating-point exception handling control register name is DECNT not DCENT.
94012 * config/rx-parse.y (DECNT): Fixed typo.
94013 * testsuite/gas/rx/dpopm.sm (DECNT): Fixed typo.
94014 * testsuite/gas/rx/dpushm.sm (DECNT): Fixed typo.
94015 * testsuite/gas/rx/macros.inc (DECNT): Fixed typo.
94017 2021-08-11 Nick Clifton <nickc@redhat.com>
94019 Fix an internal error in the CSKY assembler when asked to resolve an overlarge constant.
94021 * config/tc-csky.c (md_apply_fix): Correctly handle a fixup that
94022 involves an overlarge constant.
94024 2021-08-11 Luis Machado <luis.machado@linaro.org>
94026 Add 3 new PAC-related ARM note types
94027 The following patch synchronizes includes/objdump/readelf with the Linux
94028 Kernel in terms of ARM regset notes.
94030 We're currently missing 3 of them:
94034 NT_ARM_PAC_ENABLED_KEYS
94036 We don't need GDB to bother with this at the moment, so this doesn't update
94037 bfd/elf.c. If needed, we can do it in the future.
94041 * readelf.c (get_note_type): Handle new ARM PAC notes.
94045 * common.h (NT_ARM_PACA_KEYS, NT_ARM_PACG_KEYS)
94046 (NT_ARM_PAC_ENABLED_KEYS): New constants.
94048 2021-08-11 Nick Clifton <nickc@redhat.com>
94050 Updated Portuguese translation for the binutils sub-directory.
94052 2021-08-11 John Ericson <git@JohnEricson.me>
94054 Deprecate a.out support for NetBSD targets.
94055 As discussed previously, a.out support is now quite deprecated, and in
94056 some cases removed, in both Binutils itself and NetBSD, so this legacy
94057 default makes little sense. `netbsdelf*` and `netbsdaout*` still work
94058 allowing the user to be explicit about there choice. Additionally, the
94059 configure script warns about the change as Nick Clifton requested.
94061 One possible concern was the status of NetBSD on NS32K, where only a.out
94062 was supported. But per [1] NetBSD has removed support, and if it were to
94063 come back, it would be with ELF. The binutils implementation is
94064 therefore marked obsolete, per the instructions in the last message.
94066 With that patch and this one applied, I have confirmed the following:
94068 --target=i686-unknown-netbsd
94069 --target=i686-unknown-netbsdelf
94072 --target=i686-unknown-netbsdaout
94073 properly fails because target is deprecated.
94075 --target=vax-unknown-netbsdaout builds completely except for gas, where
94076 the target is deprecated.
94078 [1]: https://mail-index.netbsd.org/tech-toolchain/2021/07/19/msg004025.html
94080 bfd/config.bfd | 43 +++++++++++++--------
94081 bfd/configure.ac | 5 +--
94082 binutils/testsuite/binutils-all/nm.exp | 2 +-
94083 binutils/testsuite/lib/binutils-common.exp | 7 +---
94084 config/picflag.m4 | 4 +-
94085 gas/configure.tgt | 9 +++--
94086 gas/testsuite/gas/arm/blx-bl-convert.d | 2 +-
94087 gas/testsuite/gas/arm/blx-local-thumb.d | 2 +-
94088 gas/testsuite/gas/sh/basic.exp | 2 +-
94089 gdb/configure.host | 34 +++++++----------
94090 gdb/configure.tgt | 2 +-
94091 gdb/testsuite/gdb.asm/asm-source.exp | 6 +--
94092 intl/configure | 2 +-
94093 ld/configure.tgt | 44 +++++++++++-----------
94094 ld/testsuite/ld-arm/arm-elf.exp | 4 +-
94095 ld/testsuite/ld-elf/elf.exp | 2 +-
94096 ld/testsuite/ld-elf/shared.exp | 4 +-
94097 libiberty/configure | 4 +-
94099 2021-08-11 Andrew Burgess <andrew.burgess@embecosm.com>
94101 gdb: don't print backtrace when dumping core after an internal error
94102 Currently, when GDB hits an internal error, and the user selects to
94103 dump core, the recently added feature to write a backtrace to the
94104 console will kick in, and print a backtrace as well as dumping the
94107 This was certainly not my intention when adding the backtrace on fatal
94108 signal functionality, this feature was intended to produce a backtrace
94109 when GDB crashes due to some fatal signal, internal errors should have
94110 continued to behave as they did before, unchanged.
94112 In this commit I set the signal disposition of SIGABRT back to SIG_DFL
94113 just prior to the call to abort() that GDB uses to trigger the core
94114 dump, this prevents GDB reaching the code that writes the backtrace to
94117 I've also added a test that checks we don't see a backtrace on the
94118 console after an internal error.
94120 2021-08-11 Andrew Burgess <andrew.burgess@embecosm.com>
94122 gdb: register SIGBUS, SIGFPE, and SIGABRT handlers
94123 Register handlers for SIGBUS, SIGFPE, and SIGABRT. All of these
94124 signals are setup as fatal signals that will cause GDB to terminate.
94125 However, by passing these signals through the handle_fatal_signal
94126 function, a user can arrange to see a backtrace when GDB
94127 terminates (see maint set backtrace-on-fatal-signal).
94129 In normal use of GDB there should be no user visible changes after
94130 this commit. Only if GDB terminates with one of the above signals
94131 will GDB change slightly, potentially printing a backtrace before
94134 I've added new tests for SIGFPE, SIGBUS, and SIGABRT.
94136 2021-08-11 Andrew Burgess <andrew.burgess@embecosm.com>
94138 gdb: print backtrace on fatal SIGSEGV
94139 This commit adds a new maintenance feature, the ability to print
94140 a (limited) backtrace if GDB dies due to a fatal signal.
94142 The backtrace is produced using the backtrace and backtrace_symbols_fd
94143 functions which are declared in the execinfo.h header, and both of
94144 which are async signal safe. A configure check has been added to
94145 check for these features, if they are not available then the new code
94146 is not compiled into GDB and the backtrace will not be printed.
94148 The motivation for this new feature is to aid in debugging GDB in
94149 situations where GDB has crashed at a users site, but the user is
94150 reluctant to share core files, possibly due to concerns about what
94151 might be in the memory image within the core file. Such a user might
94152 be happy to share a simple backtrace that was written to stderr.
94154 The production of the backtrace is on by default, but can switched off
94155 using the new commands:
94157 maintenance set backtrace-on-fatal-signal on|off
94158 maintenance show backtrace-on-fatal-signal
94160 Right now, I have hooked this feature in to GDB's existing handling of
94161 SIGSEGV only, but this will be extended to more signals in a later
94164 One additional change I have made in this commit is that, when we
94165 decide GDB should terminate due to the fatal signal, we now
94166 raise the same fatal signal rather than raising SIGABRT.
94168 Currently, this is only effecting our handling of SIGSEGV. So,
94169 previously, if GDB hit a SEGV then we would terminate GDB with a
94170 SIGABRT. After this commit we will terminate GDB with a SIGSEGV.
94172 This feels like an improvement to me, we should still get a core dump,
94173 but in many shells, the user will see a more specific message once GDB
94174 exits, in bash for example "Segmentation fault" rather than "Aborted".
94176 Finally then, here is an example of the output a user would see if GDB
94177 should hit an internal SIGSEGV:
94179 Fatal signal: Segmentation fault
94180 ----- Backtrace -----
94181 ./gdb/gdb[0x8078e6]
94182 ./gdb/gdb[0x807b20]
94183 /lib64/libpthread.so.0(+0x14b20)[0x7f6648c92b20]
94184 /lib64/libc.so.6(__poll+0x4f)[0x7f66484d3a5f]
94185 ./gdb/gdb[0x1540f4c]
94186 ./gdb/gdb[0x154034a]
94187 ./gdb/gdb[0x9b002d]
94188 ./gdb/gdb[0x9b014d]
94189 ./gdb/gdb[0x9b1aa6]
94190 ./gdb/gdb[0x9b1b0c]
94191 ./gdb/gdb[0x41756d]
94192 /lib64/libc.so.6(__libc_start_main+0xf3)[0x7f66484041a3]
94193 ./gdb/gdb[0x41746e]
94194 ---------------------
94195 A fatal error internal to GDB has been detected, further
94196 debugging is not possible. GDB will now terminate.
94198 This is a bug, please report it. For instructions, see:
94199 <https://www.gnu.org/software/gdb/bugs/>.
94201 Segmentation fault (core dumped)
94203 It is disappointing that backtrace_symbols_fd does not actually map
94204 the addresses back to symbols, this appears, in part, to be due to GDB
94205 not being built with -rdynamic as the manual page for
94206 backtrace_symbols_fd suggests, however, even when I do add -rdynamic
94207 to the build of GDB I only see symbols for some addresses.
94209 We could potentially look at alternative libraries to provide the
94210 backtrace (e.g. libunwind) however, the solution presented here, which
94211 is available as part of glibc is probably a good baseline from which
94212 we might improve things in future.
94214 2021-08-11 Andrew Burgess <andrew.burgess@embecosm.com>
94216 gdb: rename async_init_signals to gdb_init_signals
94217 The async_init_signals has, for some time, dealt with async and sync
94218 signals, so removing the async prefix makes sense I think.
94220 Additionally, as pointed out by Pedro:
94224 The comments relating to SIGTRAP and SIGQUIT within this function are
94227 The comments for SIGTRAP talk about the signal disposition (SIG_IGN)
94228 being passed to the inferior, meaning the signal disposition being
94229 inherited by GDB's fork children. However, we now call
94230 restore_original_signals_state prior to forking, so the comment on
94231 SIGTRAP is redundant.
94233 The comments for SIGQUIT are similarly out of date, further, the
94234 comment on SIGQUIT talks about problems with BSD4.3 and vfork,
94235 however, we have not supported BSD4.3 for several years now.
94237 Given the above, it seems that changing the disposition of SIGTRAP is
94238 no longer needed, so I've deleted the signal() call for SIGTRAP.
94240 Finally, the header comment on the function now called
94241 gdb_init_signals was getting quite out of date, so I've updated it
94242 to (hopefully) better reflect reality.
94244 There should be no user visible change after this commit.
94246 2021-08-11 Andrew Burgess <andrew.burgess@embecosm.com>
94248 gdb: register signal handler after setting up event token
94249 This commit fixes the smallest of small possible bug related to signal
94250 handling. If we look in async_init_signals we see code like this:
94252 signal (SIGQUIT, handle_sigquit);
94254 create_async_signal_handler (async_do_nothing, NULL, "sigquit");
94256 Then if we look in handle_sigquit we see code like this:
94258 mark_async_signal_handler (sigquit_token);
94259 signal (sig, handle_sigquit);
94261 Finally, in mark_async_signal_handler we have:
94263 async_handler_ptr->ready = 1;
94265 Where async_handler_ptr will be sigquit_token.
94267 What this means is that if a SIGQUIT arrive in async_init_signals
94268 after handle_sigquit has been registered, but before sigquit_token has
94269 been initialised, then GDB will most likely crash.
94271 The chance of this happening is tiny, but fixing this is trivial, just
94272 ensure we call create_async_signal_handler before calling signal, so
94275 There are no tests for this. Trying to land a signal in the right
94276 spot is pretty hit and miss. I did try changing the current HEAD GDB
94279 signal (SIGQUIT, handle_sigquit);
94282 create_async_signal_handler (async_do_nothing, NULL, "sigquit");
94284 And confirmed that this did result in a crash, after my change I tried
94288 create_async_signal_handler (async_do_nothing, NULL, "sigquit");
94289 signal (SIGQUIT, handle_sigquit);
94292 And GDB now starts up just fine.
94296 * event-top.c (async_init_signals): For each signal, call signal
94297 only after calling create_async_signal_handler.
94299 2021-08-11 Andrew Burgess <andrew.burgess@embecosm.com>
94301 gdb: terminate upon receipt of SIGFPE
94302 GDB's SIGFPE handling is broken, this is PR gdb/16505 and
94305 We currently try to use an async event token to process SIGFPE. So,
94306 when a SIGFPE arrives the signal handler calls
94307 mark_async_signal_handler then returns, effectively ignoring the
94310 The intention is that later the event loop will see that the async
94311 token associated with SIGFPE has been marked and will call the async
94312 handler, which just throws an error.
94314 The problem is that SIGFPE is not safe to ignore. Ignoring a
94315 SIGFPE (unless it is generated artificially, e.g. by raise()) is
94316 undefined behaviour, after ignoring the signal on many targets we
94317 return to the instruction that caused the SIGFPE to be raised, which
94318 immediately causes another SIGFPE to be raised, we get stuck in an
94319 infinite loop. The behaviour is certainly true on x86-64.
94321 To view this behaviour I simply added some dummy code to GDB that
94322 performed an integer divide by zero, compiled this on x86-64
94323 GNU/Linux, ran GDB and saw GDB hang.
94325 In this commit, I propose to remove all special handling of SIGFPE and
94326 instead just let GDB make use of the default SIGFPE action, that is,
94327 to terminate the process.
94329 The only user visible change here should be:
94331 - If a user sends a SIGFPE to GDB using something like kill,
94332 previously GDB would just print an error and remain alive, now GDB
94333 will terminate. This is inline with what happens if the user
94334 sends GDB a SIGSEGV from kill though, so I don't see this as an
94337 - If a bug in GDB causes a real SIGFPE, previously the users GDB
94338 session would hang. Now the GDB session will terminate. Again,
94339 this is inline with what happens if GDB receives a SIGSEGV due to
94342 In bug gdb/16505 there is mention that it would be nice if GDB did
94343 more than just terminate when receiving a fatal signal. I haven't
94344 done that in this commit, but later commits will move in that
94347 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=16505
94348 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=17891
94350 2021-08-11 Alan Modra <amodra@gmail.com>
94352 PR28198, Support # as linker script comment marker
94354 * ldlex.l: Combine rules for handling newline, whitespace and
94355 comments. Extend # comment handling to all states.
94357 2021-08-11 Alan Modra <amodra@gmail.com>
94360 I've been tripped up before thinking the "end" rule was the "END"
94361 token. Let's use a better name. The formatting changes are for
94362 consistency within rules, and making it a little easier to visually
94363 separate tokens from mid-rule actions.
94365 * ldgram.y (separator): Rename from "end". Update uses.
94366 (statement): Formatting. Move ';' match to beginning.
94367 (paren_script_name): Formatting. Simplify.
94368 (must_be_exp, section): Formatting.
94370 2021-08-11 Alan Modra <amodra@gmail.com>
94372 Mention whitespace in script expressions
94373 Inside an output section statement, ld's parser can't tell whether a
94376 is an assignment to dot or a file named ".+=4".
94378 * ld.texi (expressions): Mention need for whitespace.
94380 2021-08-11 Matt Jacobson <mhjacobson@me.com>
94382 Add a -mno-dollar-line-separator command line option to the AVR assembler.
94383 Some frontends, like the gcc Objective-C frontend, emit symbols with $
94384 characters in them. The AVR target code in gas treats $ as a line separator,
94385 so the code doesn?t assemble correctly.
94387 Provide a machine-specific option to disable treating $ as a line separator.
94389 * config/tc-avr.c (enum options): Add option flag.
94390 (struct option): Add option -mno-dollar-line-separator.
94391 (md_parse_option): Adjust treatment of $ when option is present.
94392 * config/tc-avr.h: Use avr_line_separator_chars.
94394 2021-08-11 Nick Clifton <nickc@redhat.com>
94396 Fix typo in previous delta
94398 2021-08-11 Jan Beulich <jbeulich@suse.com>
94400 gas: fold IEEE encoding of -Inf with that of +Inf
94401 The respective results differ only by the sign bits - there's no need to
94402 have basically identical (partially even arch-specific) logic twice.
94403 Simply set the sign bit at the end of encoding the various formats.
94405 2021-08-11 Jan Beulich <jbeulich@suse.com>
94407 gas: support NaN flavors
94408 Like for infinity, there isn't just a single NaN. The sign bit may be
94409 of interest and, going beyond infinity, whether the value is quiet or
94410 signalling may be even more relevant to be able to encode.
94412 Note that an anomaly with x86'es double extended precision NaN values
94413 gets taken care of at the same time: For all other formats a positive
94414 value with all mantissa bits set was used, while here a negative value
94415 with all non-significant mantissa bits clear was chose for an unknown
94418 For m68k, since I don't know their X_PRECISION floating point value
94419 layout, a warning gets issued if any of the new flavors was attempted
94420 to be encoded that way. However likely it may be that, given that the
94421 code lives in a source file supposedly implementing IEEE-compliant
94422 formats, the bit patterns of the individual words match x86'es, I didn't
94423 want to guess so. And my very, very old paper doc doesn't even mention
94424 floating point formats other than single and double.
94426 2021-08-11 Jan Beulich <jbeulich@suse.com>
94428 Arm64: leave .bfloat16 processing to common code
94429 With x86 support having been implemented by extending atof-ieee.c, avoid
94430 unnecessary code duplication in md_atof(). This will then also allow to
94431 take advantage of adjustments made there without needing to mirror them
94434 Arm32: leave more .bfloat16 processing to common code
94435 With x86 support having been implemented by extending atof-ieee.c, avoid
94436 unnecessary code duplication in md_atof(). This will then also allow to
94437 take advantage of adjustments made there without needing to mirror them
94440 gas: make 2nd argument of .dcb.* consistently optional
94441 Unlike the forms consuming/producing integer data, the floating point
94442 ones so far required the 2nd argument to be present, contrary to
94443 documentation. To avoid code duplication, split float_length() out of
94444 hex_float() (taking the opportunity to adjust error message wording).
94446 2021-08-11 Jan Beulich <jbeulich@suse.com>
94448 x86: introduce .bfloat16 directive
94449 This is to be able to generate data acted upon by AVX512-BF16 and
94450 AMX-BF16 insns. While not part of the IEEE standard, the format is
94451 sufficiently standardized to warrant handling in config/atof-ieee.c.
94452 Arm, where custom handling was implemented, may want to leverage this as
94453 well. To be able to also use the hex forms supported for other floating
94454 point formats, a small addition to the generic hex_float() is needed.
94456 Extend existing x86 testcases.
94458 2021-08-11 Jan Beulich <jbeulich@suse.com>
94460 x86: introduce .hfloat directive
94461 This is to be able to generate data passed to {,V}CVTPH2PS and acted
94462 upon by AVX512-FP16 insns. To be able to also use the hex forms
94463 supported for other floating point formats, a small addition to the
94464 generic hex_float() is needed.
94466 Extend existing x86 testcases.
94468 2021-08-11 Jan Beulich <jbeulich@suse.com>
94470 x86/ELF: fix .tfloat output with hex input
94471 The ELF psABI-s are quite clear here: On 32-bit the data type is 12
94472 bytes long (with 2 bytes of trailing padding), while on 64-bit it is 16
94473 bytes long (with 6 bytes of padding). Make hex_float() capable of
94474 handling such padding.
94476 Note that this brings the emitted data size of .dc.x / .dcb.x in line
94477 also for non-ELF targets; so far they were different depending on input
94478 format (dec vs hex).
94480 Extend the existing x86 testcases.
94482 2021-08-11 Jan Beulich <jbeulich@suse.com>
94484 x86/ELF: fix .ds.x output
94485 The ELF psABI-s are quite clear here: On 32-bit the underlying data type
94486 is 12 bytes long (with 2 bytes of trailing padding), while on 64-bit it
94487 is 16 bytes long (with 6 bytes of padding). Make s_space() capable of
94488 handling 'x' (and 'p') type floating point being other than 12 bytes
94489 wide (also adjusting documentation). This requires duplicating the
94490 definition of X_PRECISION in the target speciifc header; the compiler
94491 would complain if this was out of sync with config/atof-ieee.c.
94493 Note that for now padding space doesn't get separated from actual
94494 storage, which means that things will work correctly only for little-
94495 endian cases, and which also means that by specifying large enough
94496 numbers padding space can be set to non-zero. Since the logic is needed
94497 for a single little-endian architecture only for now, I'm hoping that
94498 this might be acceptable for the time being; otherwise the change will
94499 become more intrusive.
94501 Note also that this brings the emitted data size of .ds.x vs .tfloat in
94502 line for non-ELF targets as well; the issue will be even more obvious
94503 when further taking into account a subsequent patch fixing .dc.x/.dcb.x
94504 (where output sizes currently differ depending on input format).
94506 Extend existing x86 testcases.
94508 2021-08-11 Jan Beulich <jbeulich@suse.com>
94510 x86/ELF: fix .tfloat output
94511 The ELF psABI-s are quite clear here: On 32-bit the data type is 12
94512 bytes long (with 2 bytes of trailing padding), while on 64-bit it is 16
94513 bytes long (with 6 bytes of padding). Make ieee_md_atof() capable of
94514 handling such padding, and specify the needed padding for x86 (leaving
94515 non-ELF targets alone for now). Split the existing x86 testcase.
94517 x86: have non-PE/COFF BEOS be recognized as ELF
94518 BEOS, unless explicitly requesting *-*-beospe* targets, uses standard
94519 ELF. None of the newly enabled tests in the testsuite fail for me.
94521 2021-08-11 Alan Modra <amodra@gmail.com>
94523 PR28163, Segment fault in function rl78_special_reloc
94524 Relocation offset checks were completely missing in the rl78 backend,
94525 allowing a relocation to write over memory anywhere. This was true
94526 for rl78_special_reloc, a function primarily used when applying debug
94527 relocations, and in rl78_elf_relocate_section used by the linker.
94529 This patch fixes those problems by correcting inaccuracies in the
94530 relocation howtos, then uses those howtos to sanity check relocation
94531 offsets before applying relocations. In addition, the patch
94532 implements overflow checking using the howto information rather than
94533 the ad-hoc scheme implemented in relocate_section. I implemented the
94534 overflow checking in rl78_special_reloc too.
94536 * elf32-rl78.c (RL78REL, RL78_OP_REL): Add mask parameter.
94537 (rl78_elf_howto_table): Set destination masks. Correct size and
94538 bitsize of DIR32_REV. Correct complain_on_overflow for many relocs
94539 as per tests in relocate_section. Add RH_SFR. Correct bitsize
94540 for RH_SADDR. Set size to 3 and bitsize to 0 for all OP relocs.
94541 (check_overflow): New function.
94542 (rl78_special_reloc): Check that reloc address is within section.
94543 Apply relocations using reloc howto. Check for overflow.
94545 (rl78_elf_relocate_section): Sanity check r_offset. Perform
94546 overflow checking using reloc howto.
94548 2021-08-11 GDB Administrator <gdbadmin@sourceware.org>
94550 Automatic date update in version.in
94552 2021-08-10 Tom Tromey <tom@tromey.com>
94554 Ignore .debug_types when reading .debug_aranges
94555 I noticed that the fission-reread.exp test case can cause a complaint
94556 when run with --target_board=cc-with-debug-names:
94558 warning: Section .debug_aranges in [...]/fission-reread has duplicate debug_info_offset 0x0, ignoring .debug_aranges.
94560 The bug here is that this executable has both .debug_info and
94561 .debug_types, and both have a CU at offset 0x0. This triggers the
94564 Because .debug_types doesn't provide any address ranges, these CUs can
94565 be ignored. That is, this bug turns out to be another regression from
94566 the info/types merger patch.
94568 This patch fixes the problem by having this loop igore type units.
94569 fission-reread.exp is updated to test for the bug.
94571 2021-08-10 Tom Tromey <tom@tromey.com>
94573 Generalize addrmap dumping
94574 While debugging another patch series, I wanted to dump an addrmap. I
94575 came up with this patch, which generalizes the addrmap-dumping code
94576 from psymtab.c and moves it to addrmap.c. psymtab.c is changed to use
94579 2021-08-10 Simon Marchi <simon.marchi@polymtl.ca>
94581 gdb: iterate only on vfork parent threads in handle_vfork_child_exec_or_exit
94582 I spotted what I think is a buglet in proceed_after_vfork_done. After a
94583 vfork child exits or execs, we resume all the threads of the parent. To
94584 do so, we iterate on all threads using iterate_over_threads with the
94585 proceed_after_vfork_done callback. Each thread is resumed if the
94586 following condition is true:
94588 if (thread->ptid.pid () == pid
94589 && thread->state == THREAD_RUNNING
94590 && !thread->executing
94591 && !thread->stop_requested
94592 && thread->stop_signal () == GDB_SIGNAL_0)
94594 where `pid` is the pid of the vfork parent. This is not multi-target
94595 aware: since it only filters on pid, if there is an inferior with the
94596 same pid in another target, we could end up resuming a thread of that
94597 other inferior. The chances of the stars aligning for this to happen
94598 are tiny, but still.
94600 Fix that by iterating only on the vfork parent's threads, instead of on
94601 all threads. This is more efficient, as we iterate on just the required
94602 threads (inferiors have their own thread list), and we can drop the pid
94603 check. The resulting code is also more straightforward in my opinion,
94606 Change-Id: I14647da72e2bf65592e82fbe6efb77a413a4be3a
94608 2021-08-10 Nick Clifton <nickc@redhat.com>
94610 Updated Serbian and Russian translations for various sub-directories
94612 2021-08-10 George Barrett <bob@bob131.so>
94614 guile: fix smob exports
94615 Before Guile v2.1 [1], calls to `scm_make_smob_type' implicitly added
94616 the created class to the exports list of (oop goops); v2.1+ does not
94617 implicitly create bindings in any modules. This means that the GDB
94618 manual subsection documenting exported types is not quite right when GDB
94619 is linked against Guile <v2.1 (types are exported from (oop goops))
94620 instead of (gdb)) and incorrect when linked against Guile v2.1+ (types
94621 are not bound to any variables at all!).
94623 There is a range of cases in which it's necessary or convenient to be
94624 able to refer to a GDB smob type, for instance:
94626 - Pattern matching based on the type of a value.
94627 - Defining GOOPS methods handling values from GDB (GOOPS methods
94628 typically use dynamic dispatch based on the types of the arguments).
94629 - Type-checking assertions when applying some defensive programming on
94631 - Generally any other situation one might encounter in a dynamically
94632 typed language that might need some introspection.
94634 If you're more familiar with Python, it would be quite similar to being
94635 unable to refer to the classes exported from the GDB module (which is to
94636 say: not crippling for the most part, but makes certain tasks more
94637 difficult than necessary).
94639 This commit makes a small change to GDB's smob registration machinery
94640 to make sure registered smobs get exported from the current
94641 module. This will likely cause warnings to the user about conflicting
94642 exports if they load both (gdb) and (oop goops) from a GDB linked
94643 against Guile v2.0, but it shouldn't impact functionality (and seemed
94644 preferable to trying to un-export bindings from (oop goops) if v2.0
94647 [1]: This changed with Guile commit
94648 28d0871b553a3959a6c59e2e4caec1c1509f8595
94652 2021-06-07 George Barrett <bob@bob131.so>
94654 * guile/scm-gsmob.c (gdbscm_make_smob_type): Export registered
94655 smob type from the current module.
94657 gdb/testsuite/ChangeLog:
94659 2021-06-07 George Barrett <bob@bob131.so>
94661 * gdb.guile/scm-gsmob.exp (test exports): Add tests to make
94662 sure the smob types currently listed in the GDB manual get
94663 exported from the (gdb) module.
94665 Change-Id: I7dcd791276b48dfc9edb64fc71170bbb42a6f6e7
94667 2021-08-10 GDB Administrator <gdbadmin@sourceware.org>
94669 Automatic date update in version.in
94671 2021-08-09 Nick Clifton <nickc@redhat.com>
94673 GAS: DWARF-5: Ensure that the 0'th entry in the directory table contains the current working directory.
94674 * dwarf2dbg.c (get_directory_table_entry): Ensure that dir[0]
94675 contains current working directory.
94676 (out_dir_and_file_list): Likewise.
94677 * testsuite/gas/elf/dwarf-5-dir0.s: New test source file.
94678 * testsuite/gas/elf/dwarf-5-dir0.d: New test driver.
94679 * testsuite/gas/elf/elf.exp: Run the new test.
94680 * testsuite/gas/elf/dwarf-5-file0.d: Adjust expected output.
94681 * testsuite/gas/i386/dwarf5-line-1.d: Likewise.
94682 * testsuite/gas/i386/dwarf5-line-2.d: Likewise.
94684 2021-08-09 GDB Administrator <gdbadmin@sourceware.org>
94686 Automatic date update in version.in
94688 2021-08-08 Tom Tromey <tom@tromey.com>
94690 Include objfiles.h in a few .c files
94691 I found a few .c files that rely on objfiles.h, but that only include
94692 it indirectly, via dwarf2/read.h -> psympriv.h. If that include is
94693 removed (something my new DWARF indexer series does), then the build
94696 It seemed harmless and correct to add these includes now, making the
94697 eventual series a little smaller.
94699 2021-08-08 GDB Administrator <gdbadmin@sourceware.org>
94701 Automatic date update in version.in
94703 2021-08-07 Alan Modra <amodra@gmail.com>
94705 PR28186, SEGV elf.c:7991:30 in _bfd_elf_fixup_group_sections
94707 * elf.c (_bfd_elf_fixup_group_sections): Don't segfault on
94708 objcopy/strip with NULL output_section.
94710 2021-08-07 Alan Modra <amodra@gmail.com>
94712 PR28176, rl78 complex reloc divide by zero
94713 This is a bit more than just preventing the divide by zero. Most of
94714 the patch is tidying up error reporting, so that for example, linking
94715 an object file with a reloc stack underflow produces a linker error
94716 rather than just displaying a message that might be ignored.
94719 * elf32-rl78.c (RL78_STACK_PUSH, RL78_STACK_POP): Delete.
94720 (rl78_stack_push, rl78_stack_pop): New inline functions.
94721 (rl78_compute_complex_reloc): Add status and error message params.
94722 Use new inline stack handling functions. Report stack overflow
94723 or underflow, and divide by zero.
94724 (rl78_special_reloc): Return status and error message from
94725 rl78_compute_complex_reloc.
94726 (rl78_elf_relocate_section): Similarly. Modernise reloc error
94727 reporting. Delete unused bfd_reloc_other case. Don't assume
94728 DIR24S_PCREL overflow is due to undefined function.
94729 (rl78_offset_for_reloc): Adjust to suit rl78_compute_complex_reloc.
94731 2021-08-07 GDB Administrator <gdbadmin@sourceware.org>
94733 Automatic date update in version.in
94735 2021-08-06 Tom de Vries <tdevries@suse.de>
94737 [gdb/symtab] Recognize .gdb_index symbol table with empty entries as empty
94738 When reading a .gdb_index that contains a non-empty symbol table with only
94739 empty entries, gdb doesn't recognize it as empty.
94741 Fix this by recognizing that the constant pool is empty, and then setting the
94742 symbol table to empty.
94744 Tested on x86_64-linux.
94748 2021-08-01 Tom de Vries <tdevries@suse.de>
94751 * dwarf2/read.c (read_gdb_index_from_buffer): Handle symbol table
94752 filled with empty entries.
94754 gdb/testsuite/ChangeLog:
94756 2021-08-01 Tom de Vries <tdevries@suse.de>
94759 * gdb.dwarf2/dw2-zero-range.exp: Remove kfail.
94761 2021-08-06 Tom Tromey <tromey@adacore.com>
94763 Unconditionally define _initialize_addrmap
94764 The way that init.c is generated does not allow for an initialization
94765 function to be conditionally defined -- doing so will result in a link
94768 This patch fixes a build problem that arises from such a conditional
94769 definition. It can be reproduce with --disable-unit-tests.
94771 2021-08-06 Tom de Vries <tdevries@suse.de>
94773 [gdb/symtab] Fix zero address complaint for shlib
94774 In PR28004 the following warning / Internal error is reported:
94777 -iex "set sysroot $(pwd -P)/repro" \
94782 Program terminated with signal SIGABRT, Aborted.
94783 #0 0x00007ff8fe8e5d22 in raise () from repro/usr/lib/libc.so.6
94784 [Current thread is 1 (LWP 1762498)]
94785 #1 0x00007ff8fe8cf862 in abort () from repro/usr/lib/libc.so.6
94786 warning: (Internal error: pc 0x7ff8feb2c21d in read in psymtab, \
94787 but not in symtab.)
94788 warning: (Internal error: pc 0x7ff8feb2c218 in read in psymtab, \
94789 but not in symtab.)
94791 #2 0x00007ff8feb2c21e in __gnu_debug::_Error_formatter::_M_error() const \
94792 [clone .cold] (warning: (Internal error: pc 0x7ff8feb2c21d in read in \
94793 psymtab, but not in symtab.)
94795 ) from repro/usr/lib/libstdc++.so.6
94798 The warning is about the following:
94799 - in find_pc_sect_compunit_symtab we try to find the address
94800 (0x7ff8feb2c218 / 0x7ff8feb2c21d) in the symtabs.
94801 - that fails, so we try again in the partial symtabs.
94802 - we find a matching partial symtab
94803 - however, the partial symtab has a full symtab, so
94804 we should have found a matching symtab in the first step.
94808 (gdb) info sym 0x7ff8feb2c218
94809 __gnu_debug::_Error_formatter::_M_error() const [clone .cold] in \
94810 section .text of repro/usr/lib/libstdc++.so.6
94811 (gdb) info sym 0x7ff8feb2c21d
94812 __gnu_debug::_Error_formatter::_M_error() const [clone .cold] + 5 in \
94813 section .text of repro/usr/lib/libstdc++.so.6
94815 which correspond to unrelocated addresses 0x9c218 and 0x9c21d:
94817 $ nm -C repro/usr/lib/libstdc++.so.6.0.29 | grep 000000000009c218
94818 000000000009c218 t __gnu_debug::_Error_formatter::_M_error() const \
94821 which belong to function __gnu_debug::_Error_formatter::_M_error() in
94822 /build/gcc/src/gcc/libstdc++-v3/src/c++11/debug.cc.
94824 The partial symtab that is found for the addresses is instead the one for
94825 /build/gcc/src/gcc/libstdc++-v3/src/c++98/bitmap_allocator.cc, which is
94828 This happens as follows.
94830 The bitmap_allocator.cc CU has DW_AT_ranges at .debug_rnglist offset 0x4b50:
94832 00004b50 0000000000000000 0000000000000056
94833 00004b5a 00000000000a4790 00000000000a479c
94834 00004b64 00000000000a47a0 00000000000a47ac
94837 When reading the first range 0x0..0x56, it doesn't trigger the "start address
94838 of zero" complaint here:
94840 /* A not-uncommon case of bad debug info.
94841 Don't pollute the addrmap with bad data. */
94842 if (range_beginning + baseaddr == 0
94843 && !per_objfile->per_bfd->has_section_at_zero)
94845 complaint (_(".debug_rnglists entry has start address of zero"
94846 " [in module %s]"), objfile_name (objfile));
94850 because baseaddr != 0, which seems incorrect given that when loading the
94851 shared library individually in gdb (and consequently baseaddr == 0), we do see
94854 Consequently, we run into this case in dwarf2_get_pc_bounds:
94856 if (low == 0 && !per_objfile->per_bfd->has_section_at_zero)
94857 return PC_BOUNDS_INVALID;
94859 which then results in this code in process_psymtab_comp_unit_reader being
94860 called with cu_bounds_kind == PC_BOUNDS_INVALID, which sets the set_addrmap
94863 scan_partial_symbols (first_die, &lowpc, &highpc,
94864 cu_bounds_kind <= PC_BOUNDS_INVALID, cu);
94866 and consequently, the CU addrmap gets build using address info from the
94869 During that process, addrmap_set_empty is called with a range that includes
94870 0x9c218 and 0x9c21d:
94874 (gdb) p /x end_inclusive
94877 but it's called for a function at DIE 0x54153 with DW_AT_ranges at 0x40ae:
94879 000040ae 00000000000b1ee0 00000000000b200e
94880 000040b9 000000000009989c 00000000000998c4
94881 000040c3 <End of list>
94883 and neither range includes 0x9c218 and 0x9c21d.
94885 This is caused by this code in partial_die_info::read:
94887 if (dwarf2_ranges_read (ranges_offset, &lowpc, &highpc, cu,
94891 which pretends that the function is located at addresses 0x9989c..0xb200d,
94892 which is indeed not the case.
94894 This patch fixes the first problem encountered: fix the "start address of
94895 zero" complaint warning by removing the baseaddr part from the condition.
94896 Same for dwarf2_ranges_process.
94898 The effect is that:
94899 - the complaint is triggered, and
94900 - the warning / Internal error is no longer triggered.
94902 This does not fix the observed problem in partial_die_info::read, which is
94905 Tested on x86_64-linux.
94907 Co-Authored-By: Simon Marchi <simon.marchi@polymtl.ca>
94911 2021-07-29 Simon Marchi <simon.marchi@polymtl.ca>
94912 Tom de Vries <tdevries@suse.de>
94915 * gdb/dwarf2/read.c (dwarf2_rnglists_process, dwarf2_ranges_process):
94916 Fix zero address complaint.
94917 * gdb/testsuite/gdb.dwarf2/dw2-zero-range-shlib.c: New test.
94918 * gdb/testsuite/gdb.dwarf2/dw2-zero-range.c: New test.
94919 * gdb/testsuite/gdb.dwarf2/dw2-zero-range.exp: New file.
94921 2021-08-06 Alan Modra <amodra@gmail.com>
94923 Re: Add tests for Intel AVX512_FP16 instructions
94924 * testsuite/gas/i386/x86-64-avx512_fp16_pseudo_ops.d: Pass with
94925 mingw section padding.
94927 2021-08-06 Alan Modra <amodra@gmail.com>
94930 It matters not at all if pc is incremented from its initial NULL
94931 value, but avoid this silly runtime ubsan error.
94933 * doc/chew.c (perform): Avoid incrementing NULL pc.
94935 2021-08-06 Alan Modra <amodra@gmail.com>
94937 bfd_reloc_offset_in_range overflow
94938 This patch is more about the style of bounds checking we ought to use,
94939 rather than a real problem. An overflow of "octet + reloc_size" can
94940 only happen with huge sections which would certainly cause out of
94943 * reloc.c (bfd_reloc_offset_in_range): Avoid possible overflow.
94945 2021-08-06 Alan Modra <amodra@gmail.com>
94947 PR28175, Segment fault in coff-tic30.c reloc_processing
94948 The obj_convert table shouldn't be accessed without first checking the
94949 index against the table size.
94952 * coff-tic30.c (reloc_processing): Sanity check reloc symbol index.
94953 * coff-z80.c (reloc_processing): Likewise.
94954 * coff-z8k.c (reloc_processing): Likewise.
94956 2021-08-06 Alan Modra <amodra@gmail.com>
94958 PR28173, nds32_elf_howto_table index out of bounds
94959 Indexing the howto table was seriously broken by a missing entry, and
94960 use of assertions about user input rather than testing the input.
94963 * elf32-nds32.c (nds32_elf_howto_table): Add missing empty howto.
94964 (bfd_elf32_bfd_reloc_type_table_lookup): Replace assertions with
94965 range checks. Return NULL if unsupported reloc type. Remove
94966 dead code. Take an unsigned int param.
94967 (nds32_info_to_howto_rel): Test for NULL howto or howto name
94968 return from lookup. Remove assertion.
94969 (nds32_info_to_howto): Remove unnecessary ATTRIBUTE_UNUSED.
94970 Test for NULL howto or howto name return from lookup.
94972 2021-08-06 Alan Modra <amodra@gmail.com>
94974 PR28172, bfin_pcrel24_reloc heap-buffer-overflow
94975 bfin pcrel24 relocs are weird, they apply to the reloc address minus
94976 two. That means reloc addresses of 0 and 1 are invalid. Check that,
94977 and fix other reloc range checking.
94980 * elf32-bfin.c (bfin_pcrel24_reloc): Correct reloc range check.
94981 (bfin_imm16_reloc, bfin_byte4_reloc, bfin_bfd_reloc): Likewise.
94982 (bfin_final_link_relocate): Likewise.
94984 2021-08-06 GDB Administrator <gdbadmin@sourceware.org>
94986 Automatic date update in version.in
94988 2021-08-05 Will Schmidt <will_schmidt@vnet.ibm.com>
94990 [PATCH] GDB Testsuite, update compile-cplus.exp
94991 [PATCH] GDB Testsuite, update compile-cplus.exp
94993 Update the gdb.compile/compile-cplus.exp test to
94994 handle errors generated when passing bad arguments
94995 into the gdb-compile command.
94996 This matches changes made to gdb.compile/compile.exp
94997 in the past as part of
94998 "Migrate rest of compile commands to new options framework"
94999 e6ed716cd5514c08b9d7c469d185b1aa177dbc22
95001 2021-08-05 Will Schmidt <will_schmidt@vnet.ibm.com>
95003 [gdb] Handle .TOC. sections during gdb-compile for rs6000 target.
95004 [gdb] Handle .TOC. sections during gdb-compile for rs6000 target.
95006 When we encounter a .TOC. symbol in the object we are loading,
95007 we need to associate this with the .toc section in order to
95008 properly resolve other symbols in the object. IF a .toc section
95009 is not found, iterate the sections until we find one with the
95010 SEC_ALLOC flag. If that also fails, fall back to using
95011 the *ABS* section, pointed to by bfd_abs_section_ptr.
95013 2021-08-05 Simon Marchi <simon.marchi@polymtl.ca>
95015 gdb/testsuite: gdb.base/attach.exp: expose bug when testing with native-extended-gdbserver
95016 In gdb.base/attach.exp, proc do_attach_failure_tests, we attach to a
95017 process. When then try to attach to the same process in another
95018 inferior, expecting it to fail. We then come back to the first inferior
95019 and try to kill it, to clean up the test. When using the
95020 native-extended-gdbserver board, this "kill" test passes, even though it
95021 didn't actually work:
95025 Added inferior 2 on connection 1 (extended-remote localhost:2347)
95026 (gdb) PASS: gdb.base/attach.exp: do_attach_failure_tests: add empty inferior 2
95028 [Switching to inferior 2 [<null>] (<noexec>)]
95029 (gdb) PASS: gdb.base/attach.exp: do_attach_failure_tests: switch to inferior 2
95031 Attaching to process 817032
95032 Attaching to process 817032 failed
95033 (gdb) PASS: gdb.base/attach.exp: do_attach_failure_tests: fail to attach again
95035 [Switching to inferior 1 [process 817032] (/home/simark/build/binutils-gdb/gdb/testsuite/outputs/gdb.base/attach/attach)]
95036 [Switching to thread 1.1 (Thread 817032.817032)]
95037 #0 main () at /home/simark/src/binutils-gdb/gdb/testsuite/gdb.base/attach.c:19
95038 19 while (! should_exit)
95039 (gdb) PASS: gdb.base/attach.exp: do_attach_failure_tests: switch to inferior 1
95041 Kill the program being debugged? (y or n) y
95042 Remote connection closed <==== That's unexpected
95043 (gdb) PASS: gdb.base/attach.exp: do_attach_failure_tests: exit after attach failures
95045 When the second attach fails, gdbserver seems to break the connection
95046 (it hangs up on the existing remote target) and start listening again
95047 for incoming connections. This is documented in PR 19558 [1].
95049 Make the expected output regexp for the kill command tighter (it
95050 currently accepts anything). Use "set confirm off" so we don't have to
95051 deal with the confirmation. And to be really sure the extended-remote
95052 target still works, try to run the inferior again after killing. The
95053 now tests are kfail'ed when the target is gdbserver.
95055 [1] https://sourceware.org/bugzilla/show_bug.cgi?id=19558
95057 gdb/testsuite/ChangeLog:
95059 * gdb.base/attach.exp (do_attach_failure_tests): Make kill
95060 regexp tighter, run inferior after killing it. Kfail when
95061 target is gdbserver.
95063 Change-Id: I99c5cd3968ce2ec962ace35b016f842a243b7a0d
95065 2021-08-05 Simon Marchi <simon.marchi@polymtl.ca>
95067 gdb/testsuite: gdb.base/attach.exp: fix support check in test_command_line_attach_run
95068 When running this test with the native-extended-gdbserver, we get:
95070 main () at /home/simark/src/binutils-gdb/gdb/testsuite/gdb.base/attach.c:19
95071 19 while (! should_exit)
95072 The program being debugged has been started already.
95073 Start it from the beginning? (y or n) PASS: gdb.base/attach.exp: cmdline attach run: run to prompt
95075 Don't know how to run. Try "help target".
95076 (gdb) FAIL: gdb.base/attach.exp: cmdline attach run: run to main
95078 This test tests using both "-p <pid>" and "-ex start" on the command line,
95079 making sure that we first attach and then run.
95081 Normally, after that "y", we should see the program running again.
95082 However, a particuliarity of the native-extended-gdbserver is that it
95083 uses "set auto-connect-native-target off" on the command line. The full
95084 GDB command line is:
95086 ./gdb -nw -nx -data-directory /home/simark/build/binutils-gdb/gdb/testsuite/../data-directory \
95087 -iex set height 0 -iex set width 0 -ex set auto-connect-native-target off \
95088 -ex set sysroot -quiet -iex set height 0 -iex set width 0 --pid=536609 -ex start
95090 The attach succeeds. I guess it is done before "set
95091 auto-connect-native-target off", or it somehow bypasses it. When the
95092 "start" is executed, the native target is unpushed, while killing the
95093 existing process, but not re-pushed, due to "set
95094 auto-connect-native-target off". So we get that "Don't know how to run"
95097 Really, I think it's a case of the test doing things incompatible with
95098 the board, I think it should just be skipped. And as we can see with
95099 the current code, there were some attempts at doing this, just using the
95102 - isnative: this is a dejagnu proc which checks if the target board has
95103 the same triplet as the build machine. In the case of
95104 native-extended-gdbserver, it does.
95105 - is_remote target: this checks whether the target board is remote, as
95106 in executing on a different machin. native-extended-gdbserver is not
95109 Since the --pid option specifically attaches to a process using the
95110 native target, change the test to use gdb_is_target_native instead.
95112 gdb/testsuite/ChangeLog:
95114 * gdb.base/attach.exp (test_command_line_attach_run): Use
95115 gdb_is_target_native to check if test is supported.
95117 Change-Id: I762e127f39623889999dc9ed2185540a0951bfb0
95119 2021-08-05 Simon Marchi <simon.marchi@polymtl.ca>
95121 gdb: target_waitstatus_to_string: print extra info for FORKED, VFORKED, EXECD
95122 Print the extra information contained in target_waitstatus for these
95123 events. For TARGET_WAITKIND_{FORKED,VFORKED}, the extra information is
95124 contained in related_pid, and is the ptid of the new process. For
95125 TARGET_WAITKIND_EXECD, it,s the exec'd path name in execd_pathname.
95126 Print it using the same format used for TARGET_WAITKIND_STOPPED and
95129 Here are sample outputs for all three events:
95131 [infrun] print_target_wait_results: target_wait (-1.0.0 [process -1], status) =
95132 [infrun] print_target_wait_results: 726890.726890.0 [process 726890],
95133 [infrun] print_target_wait_results: status->kind = vforked, related_pid = 726894.726894.0
95135 [infrun] print_target_wait_results: target_wait (-1.0.0 [process -1], status) =
95136 [infrun] print_target_wait_results: 727045.727045.0 [process 727045],
95137 [infrun] print_target_wait_results: status->kind = forked, related_pid = 727049.727049.0
95139 [infrun] print_target_wait_results: target_wait (-1.0.0 [process -1], status) =
95140 [infrun] print_target_wait_results: 727119.727119.0 [process 727119],
95141 [infrun] print_target_wait_results: status->kind = execd, execd_pathname = /usr/bin/ls
95143 Change-Id: I4416a74e3bf792a625a68bf26c51689e170f2184
95145 2021-08-05 Simon Marchi <simon.marchi@polymtl.ca>
95147 gdb: use ptid_t::to_string in print_target_wait_results
95148 The ptid_t::to_string method was introduced recently, to format a ptid_t
95149 for debug purposes. It formats the ptid exactly as is done in
95150 print_target_wait_results, so make print_target_wait_results use it.
95152 Change-Id: I0a81c8040d3e1858fb304cb28366b34d94eefe4d
95154 2021-08-05 Zoran Zaric <Zoran.Zaric@amd.com>
95156 Add as_lval argument to expression evaluator
95157 There are cases where the result of the expression evaluation is
95158 expected to be in a form of a value and not location description.
95160 One place that has this requirement is dwarf_entry_parameter_to_value
95161 function, but more are expected in the future. Until now, this
95162 requirement was fulfilled by extending the evaluated expression with
95163 a DW_OP_stack_value operation at the end.
95165 New implementation, introduces a new evaluation argument instead.
95167 * dwarf2/expr.c (dwarf_expr_context::fetch_result): Add as_lval
95169 (dwarf_expr_context::eval_exp): Add as_lval argument.
95170 * dwarf2/expr.h (struct dwarf_expr_context): Add as_lval
95171 argument to fetch_result and eval_exp methods.
95172 * dwarf2/frame.c (execute_stack_op): Add as_lval argument.
95173 * dwarf2/loc.c (dwarf_entry_parameter_to_value): Remove
95174 DWARF expression extension.
95175 (dwarf2_evaluate_loc_desc_full): Add as_lval argument support.
95176 (dwarf2_evaluate_loc_desc): Add as_lval argument support.
95177 (dwarf2_locexpr_baton_eval): Add as_lval argument support.
95179 2021-08-05 Zoran Zaric <zoran.zaric@amd.com>
95181 Simplify dwarf_expr_context class interface
95182 Idea of this patch is to get a clean and simple public interface for
95183 the dwarf_expr_context class, looking like:
95187 - push_address method and
95190 Where constructor should only ever require a target architecture
95191 information. This information is held in per object file
95192 (dwarf2_per_objfile) structure, so it makes sense to keep that
95193 structure as a constructor argument. It also makes sense to get the
95194 address size from that structure, but unfortunately that interface
95195 doesn't exist at the moment, so the dwarf_expr_context class user
95196 needs to provide that information.
95198 The push_address method is used to push a CORE_ADDR as a value on
95199 top of the DWARF stack before the evaluation. This method can be
95200 later changed to push any struct value object on the stack.
95202 The evaluate method is the method that evaluates a DWARF expression
95203 and provides the evaluation result, in a form of a single struct
95204 value object that describes a location. To do this, the method requires
95205 a context of the evaluation, as well as expected result type
95206 information. If the type information is not provided, the DWARF generic
95207 type will be used instead.
95209 To avoid storing the gdbarch information in the evaluator object, that
95210 information is now always acquired from the per_objfile object.
95212 All data members are now private and only visible to the evaluator
95213 class, so a m_ prefix was added to all of their names to reflect that.
95214 To make this distinction clear, they are also accessed through objects
95215 this pointer, wherever that was not the case before.
95219 * dwarf2/expr.c (dwarf_expr_context::dwarf_expr_context): Add
95220 address size argument.
95221 (dwarf_expr_context::read_mem): Change to use property_addr_info
95223 (dwarf_expr_context::evaluate): New function.
95224 (dwarf_expr_context::execute_stack_op): Change to use
95225 property_addr_info structure.
95226 * dwarf2/expr.h (struct dwarf_expr_context): New evaluate
95227 declaration. Change eval and fetch_result method to private.
95228 (dwarf_expr_context::gdbarch): Remove member.
95229 (dwarf_expr_context::stack): Make private and add m_ prefix.
95230 (dwarf_expr_context::addr_size): Make private and add
95232 (dwarf_expr_context::recursion_depth): Make private and add
95234 (dwarf_expr_context::max_recursion_depth): Make private and
95236 (dwarf_expr_context::len): Make private and add m_ prefix.
95237 (dwarf_expr_context::data): Make private and add m_ prefix.
95238 (dwarf_expr_context::initialized): Make private and add
95240 (dwarf_expr_context::pieces): Make private and add m_ prefix.
95241 (dwarf_expr_context::per_objfile): Make private and add
95243 (dwarf_expr_context::frame): Make private and add m_ prefix.
95244 (dwarf_expr_context::per_cu): Make private and add m_ prefix.
95245 (dwarf_expr_context::addr_info): Make private and add
95247 * dwarf2/frame.c (execute_stack_op): Change to call evaluate
95249 * dwarf2/loc.c (dwarf2_evaluate_loc_desc_full): Change to call
95251 (dwarf2_locexpr_baton_eval): Change to call evaluate method.
95253 2021-08-05 Zoran Zaric <zoran.zaric@amd.com>
95255 Make DWARF evaluator return a single struct value
95256 The patch is addressing the issue of class users writing and reading
95257 the internal data of the dwarf_expr_context class.
95259 At this point, all conditions are met for the DWARF evaluator to return
95260 an evaluation result in a form of a single struct value object.
95264 * dwarf2/expr.c (pieced_value_funcs): Chenge to static
95266 (allocate_piece_closure): Change to static function.
95267 (dwarf_expr_context::fetch_result): New function.
95268 * dwarf2/expr.h (struct piece_closure): Remove declaration.
95269 (struct dwarf_expr_context): fetch_result new declaration.
95270 fetch, fetch_address and fetch_in_stack_memory members move
95272 (allocate_piece_closure): Remove.
95273 * dwarf2/frame.c (execute_stack_op): Change to use
95275 * dwarf2/loc.c (dwarf2_evaluate_loc_desc_full): Change to use
95277 (dwarf2_locexpr_baton_eval): Change to use fetch_result.
95278 * dwarf2/loc.h (invalid_synthetic_pointer): Expose function.
95280 2021-08-05 Zoran Zaric <Zoran.Zaric@amd.com>
95282 Make value_copy also copy the stack data member
95283 Fixing a bug where the value_copy function did not copy the stack data
95284 and initialized members of the struct value. This is needed for the
95285 next patch where the DWARF expression evaluator is changed to return a
95286 single struct value object.
95288 * value.c (value_copy): Change to also copy the stack data
95289 and initialized members.
95291 2021-08-05 Zoran Zaric <zoran.zaric@amd.com>
95293 Move piece_closure and its support to expr.c
95294 Following 5 patches series is trying to clean up the interface of the
95295 DWARF expression evaluator class (dwarf_expr_context).
95297 After merging all expression evaluators into one class, the next
95298 logical step is to make a clean user interface for that class. To do
95299 that, we first need to address the issue of class users writing and
95300 reading the internal data of the class directly.
95302 Fixing the case of writing is simple, it makes sense for an evaluator
95303 instance to be per architecture basis. Currently, the best separation
95304 seems to be per object file, so having that data (dwarf2_per_objfile)
95305 as a constructor argument makes sense. It also makes sense to get the
95306 address size from that object file, but unfortunately that interface
95307 does not exist at the moment.
95309 Luckily, address size information is already available to the users
95310 through other means. As a result, the address size also needs to be a
95311 class constructor argument, at least until a better interface for
95312 acquiring that information from an object file is implemented.
95314 The rest of the user written data comes down to a context of an
95315 evaluated expression (compilation unit context, frame context and
95316 passed in buffer context) and a source type information that a result
95317 of evaluating expression is representing. So, it makes sense for all of
95318 these to be arguments of an evaluation method.
95320 To address the problem of reading the dwarf_expr_context class
95321 internal data, we first need to understand why it is implemented that
95324 This is actualy a question of which existing class can be used to
95325 represent both values and a location descriptions and why it is not
95328 The answer is in a struct value class/structure, but the problem is
95329 that before the evaluators were merged, only one evaluator had an
95330 infrastructure to resolve composite and implicit pointer location
95333 After the merge, we are now able to use the struct value to represent
95334 any result of the expression evaluation. It also makes sense to move
95335 all infrastructure for those location descriptions to the expr.c file
95336 considering that that is the only place using that infrastructure.
95338 What we are left with in the end is a clean public interface of the
95339 dwarf_expr_context class containing:
95343 - push_address method and
95346 The idea with this particular patch is to move piece_closure structure
95347 and the interface that handles it (lval_funcs) to expr.c file.
95349 While implicit pointer location descriptions are still not useful in
95350 the CFI context (of the AMD's DWARF standard extensions), the composite
95351 location descriptions are certainly necessary to describe a results of
95352 specific compiler optimizations.
95354 Considering that a piece_closure structure is used to represent both,
95355 there was no benefit in splitting them.
95359 * dwarf2/expr.c (struct piece_closure): Add from loc.c.
95360 (allocate_piece_closure): Add from loc.c.
95361 (bits_to_bytes): Add from loc.c.
95362 (rw_pieced_value): Add from loc.c.
95363 (read_pieced_value): Add from loc.c.
95364 (write_pieced_value): Add from loc.c.
95365 (check_pieced_synthetic_pointer): Add from loc.c.
95366 (indirect_pieced_value): Add from loc.c.
95367 (coerce_pieced_ref): Add from loc.c.
95368 (copy_pieced_value_closure): Add from loc.c.
95369 (free_pieced_value_closure): Add from loc.c.
95370 (sect_variable_value): Add from loc.c.
95371 * dwarf2/loc.c (sect_variable_value): Move to expr.c.
95372 (struct piece_closure): Move to expr.c.
95373 (allocate_piece_closure): Move to expr.c.
95374 (bits_to_bytes): Move to expr.c.
95375 (rw_pieced_value): Move to expr.c.
95376 (read_pieced_value): Move to expr.c.
95377 (write_pieced_value): Move to expr.c.
95378 (check_pieced_synthetic_pointer): Move to expr.c.
95379 (indirect_pieced_value): Move to expr.c.
95380 (coerce_pieced_ref): Move to expr.c.
95381 (copy_pieced_value_closure): Move to expr.c.
95382 (free_pieced_value_closure): Move to expr.c.
95384 2021-08-05 Zoran Zaric <zoran.zaric@amd.com>
95386 Merge evaluate_for_locexpr_baton evaluator
95387 The evaluate_for_locexpr_baton is the last derived class from the
95388 dwarf_expr_context class. It's purpose is to support the passed in
95389 buffer functionality.
95391 Although, it is not really necessary to merge this class with it's
95392 base class, doing that simplifies new expression evaluator design.
95394 Considering that this functionality is going around the DWARF standard,
95395 it is also reasonable to expect that with a new evaluator design and
95396 extending the push object address functionality to accept any location
95397 description, there will be no need to support passed in buffers.
95399 Alternatively, it would also makes sense to abstract the interaction
95400 between the evaluator and a given resource in the near future. The
95401 passed in buffer would then be a specialization of that abstraction.
95405 * dwarf2/expr.c (dwarf_expr_context::read_mem): Merge with
95406 evaluate_for_locexpr_baton implementation.
95407 * dwarf2/loc.c (class evaluate_for_locexpr_baton): Remove
95409 (evaluate_for_locexpr_baton::read_mem): Move to
95410 dwarf_expr_context.
95411 (dwarf2_locexpr_baton_eval): Instantiate dwarf_expr_context
95412 instead of evaluate_for_locexpr_baton class.
95414 2021-08-05 Zoran Zaric <zoran.zaric@amd.com>
95416 Remove empty frame and full evaluators
95417 There are no virtual methods that require different specialization in
95418 dwarf_expr_context class. This means that derived classes
95419 dwarf_expr_executor and dwarf_evaluate_loc_desc are not needed any
95422 As a result of this, the evaluate_for_locexpr_baton class base class
95423 is now the dwarf_expr_context class.
95425 There might be a need for a better class hierarchy when we know more
95426 about the direction of the future DWARF versions and gdb extensions,
95427 but that is out of the scope of this patch series.
95431 * dwarf2/frame.c (class dwarf_expr_executor): Remove class.
95432 (execute_stack_op): Instantiate dwarf_expr_context instead of
95433 dwarf_evaluate_loc_desc class.
95434 * dwarf2/loc.c (class dwarf_evaluate_loc_desc): Remove class.
95435 (dwarf2_evaluate_loc_desc_full): Instantiate dwarf_expr_context
95436 instead of dwarf_evaluate_loc_desc class.
95437 (struct evaluate_for_locexpr_baton): Derive from
95438 dwarf_expr_context.
95440 2021-08-05 Zoran Zaric <Zoran.Zaric@amd.com>
95442 Inline get_reg_value method of dwarf_expr_context
95443 The get_reg_value method is a small function that is only called once,
95444 so it can be inlined to simplify the dwarf_expr_context class.
95448 * dwarf2/expr.c (dwarf_expr_context::get_reg_value): Remove
95450 (dwarf_expr_context::execute_stack_op): Inline get_reg_value
95452 * dwarf2/expr.h (dwarf_expr_context::get_reg_value): Remove
95455 2021-08-05 Zoran Zaric <zoran.zaric@amd.com>
95457 Move push_dwarf_reg_entry_value to expr.c
95458 Following the idea of merging the evaluators, the
95459 push_dwarf_reg_entry_value method can be moved from
95460 dwarf_expr_executor and dwarf_evaluate_loc_desc classes
95461 to their base class dwarf_expr_context.
95466 (dwarf_expr_context::push_dwarf_reg_entry_value): Move from
95467 dwarf_evaluate_loc_desc.
95469 (dwarf_expr_executor::push_dwarf_reg_entry_value): Remove
95471 * dwarf2/loc.c (dwarf_expr_reg_to_entry_parameter): Expose
95473 (dwarf_evaluate_loc_desc::push_dwarf_reg_entry_value): Move to
95474 dwarf_expr_context.
95475 * dwarf2/loc.h (dwarf_expr_reg_to_entry_parameter): Expose
95478 2021-08-05 Zoran Zaric <zoran.zaric@amd.com>
95480 Move read_mem to dwarf_expr_context
95481 Following the idea of merging the evaluators, the read_mem method can
95482 be moved from dwarf_expr_executor and dwarf_evaluate_loc_desc classes
95483 to their base class dwarf_expr_context.
95487 * dwarf2/expr.c (dwarf_expr_context::read_mem): Move from
95488 dwarf_evaluate_loc_desc.
95489 * dwarf2/frame.c (dwarf_expr_executor::read_mem): Remove
95491 * dwarf2/loc.c (dwarf_evaluate_loc_desc::read_mem): Move to
95492 dwarf_expr_context.
95494 2021-08-05 Zoran Zaric <Zoran.Zaric@amd.com>
95496 Move get_object_address to dwarf_expr_context
95497 Following the idea of merging the evaluators, the get_object_address
95498 and can be moved from dwarf_expr_executor and dwarf_evaluate_loc_desc
95499 classes to their base class dwarf_expr_context.
95503 * dwarf2/expr.c (dwarf_expr_context::get_object_address): Move
95504 from dwarf_evaluate_loc_desc.
95505 (class dwarf_expr_context): Add object address member to
95506 dwarf_expr_context.
95507 * dwarf2/expr.h (dwarf_expr_context::get_frame_pc): Remove
95509 * dwarf2/frame.c (dwarf_expr_executor::get_object_address):
95511 * dwarf2/loc.c (dwarf_evaluate_loc_desc::get_object_address):
95512 move to dwarf_expr_context.
95513 (class dwarf_evaluate_loc_desc): Move object address member to
95514 dwarf_expr_context.
95516 2021-08-05 Zoran Zaric <zoran.zaric@amd.com>
95518 Move dwarf_call to dwarf_expr_context
95519 Following the idea of merging the evaluators, the dwarf_call and
95520 get_frame_pc method can be moved from dwarf_expr_executor and
95521 dwarf_evaluate_loc_desc classes to their base class dwarf_expr_context.
95522 Once this is done, the get_frame_pc can be replace with lambda
95527 * dwarf2/expr.c (dwarf_expr_context::dwarf_call): Move from
95528 dwarf_evaluate_loc_desc.
95529 (dwarf_expr_context::get_frame_pc): Replace with lambda.
95530 * dwarf2/expr.h (dwarf_expr_context::get_frame_pc): Remove
95532 * dwarf2/frame.c (dwarf_expr_executor::dwarf_call): Remove
95534 (dwarf_expr_executor::get_frame_pc): Remove method.
95535 * dwarf2/loc.c (dwarf_evaluate_loc_desc::get_frame_pc): Remove
95537 (dwarf_evaluate_loc_desc::dwarf_call): Move to
95538 dwarf_expr_context.
95539 (per_cu_dwarf_call): Inline function.
95541 2021-08-05 Zoran Zaric <zoran.zaric@amd.com>
95543 Move compilation unit info to dwarf_expr_context
95544 This patch moves the compilation unit context information and support
95545 from dwarf_expr_executor and dwarf_evaluate_loc_desc to
95546 dwarf_expr_context evaluator. The idea is to report an error when a
95547 given operation requires a compilation unit information to be resolved,
95548 which is not available.
95550 With this change, it also makes sense to always acquire ref_addr_size
95551 information from the compilation unit context, considering that all
95552 DWARF operations that refer to that information require a compilation
95553 unit context to be present during their evaluation.
95557 * dwarf2/expr.c (ensure_have_per_cu): New function.
95558 (dwarf_expr_context::dwarf_expr_context): Add compilation unit
95559 context information.
95560 (dwarf_expr_context::get_base_type): Move from
95561 dwarf_evaluate_loc_desc.
95562 (dwarf_expr_context::get_addr_index): Remove method.
95563 (dwarf_expr_context::dwarf_variable_value): Remove method.
95564 (dwarf_expr_context::execute_stack_op): Call compilation unit
95565 context info check. Inline get_addr_index and
95566 dwarf_variable_value methods.
95567 * dwarf2/expr.h (struct dwarf_expr_context): Add compilation
95569 (dwarf_expr_context::get_addr_index): Remove method.
95570 (dwarf_expr_context::dwarf_variable_value): Remove method.
95571 (dwarf_expr_context::ref_addr_size): Remove member.
95572 * dwarf2/frame.c (dwarf_expr_executor::get_addr_index): Remove
95574 (dwarf_expr_executor::dwarf_variable_value): Remove method.
95575 * dwarf2/loc.c (sect_variable_value): Expose function.
95576 (dwarf_evaluate_loc_desc::get_addr_index): Remove method.
95577 (dwarf_evaluate_loc_desc::dwarf_variable_value): Remove method.
95578 (class dwarf_evaluate_loc_desc): Move compilation unit context
95579 information to dwarf_expr_context class.
95580 * dwarf2/loc.h (sect_variable_value): Expose function.
95582 2021-08-05 Zoran Zaric <Zoran.Zaric@amd.com>
95584 Remove get_frame_cfa from dwarf_expr_context
95585 Following the idea of merging the evaluators, the get_frame_cfa method
95586 can be moved from dwarf_expr_executor and dwarf_evaluate_loc_desc
95587 classes to their base class dwarf_expr_context. Once this is done,
95588 it becomes apparent that the method is only called once and it can be
95591 It is also necessary to check if the frame context information was
95592 provided before the DW_OP_call_frame_cfa operation is executed.
95596 * dwarf2/expr.c (dwarf_expr_context::get_frame_cfa): Remove
95598 (dwarf_expr_context::execute_stack_op): Call frame context info
95599 check for DW_OP_call_frame_cfa. Remove use of get_frame_cfa.
95600 * dwarf2/expr.h (dwarf_expr_context::get_frame_cfa): Remove
95602 * dwarf2/frame.c (dwarf_expr_context::get_frame_cfa): Remove
95604 * dwarf2/loc.c (dwarf_expr_context::get_frame_cfa): Remove
95607 2021-08-05 Zoran Zaric <zoran.zaric@amd.com>
95609 Move frame context info to dwarf_expr_context
95610 Following 15 patches in this patch series is cleaning up the design of
95611 the DWARF expression evaluator (dwarf_expr_context) to make future
95612 extensions of that evaluator easier and cleaner to implement.
95614 There are three subclasses of the dwarf_expr_context class
95615 (dwarf_expr_executor, dwarf_evaluate_loc_desc and
95616 evaluate_for_locexpr_baton). Here is a short description of each class:
95618 - dwarf_expr_executor is evaluating a DWARF expression in a context
95619 of a Call Frame Information. The overridden methods of this subclass
95620 report an error if a specific DWARF operation, represented by that
95621 method, is not allowed in a CFI context. The source code of this
95622 subclass lacks the support for composite as well as implicit pointer
95623 location description.
95625 - dwarf_evaluate_loc_desc can evaluate any expression with no
95626 restrictions. All of the methods that this subclass overrides are
95627 actually doing what they are intended to do. This subclass contains
95628 a full support for all location description types.
95630 - evaluate_for_locexpr_baton subclass is a specialization of the
95631 dwarf_evaluate_loc_desc subclass and it's function is to add
95632 support for passed in buffers. This seems to be a way to go around
95633 the fact that DWARF standard lacks a bit offset support for memory
95634 location descriptions as well as using any location description for
95635 the push object address functionality.
95637 It all comes down to this question: what is a function of a DWARF
95638 expression evaluator?
95640 Is it to evaluate the expression in a given context or to check the
95641 correctness of that expression in that context?
95643 Currently, the only reason why there is a dwarf_expr_executor subclass
95644 is to report an invalid DWARF expression in a context of a CFI, but is
95645 that what the evaluator is supposed to do considering that the evaluator
95646 is not tied to a given DWARF version?
95648 There are more and more vendor and GNU extensions that are not part of
95649 the DWARF standard, so is it that impossible to expect that some of the
95650 extensions could actually lift the previously imposed restrictions of
95651 the CFI context? Not to mention that every new DWARF version is lifting
95652 some restrictions anyway.
95654 The thing that makes more sense for an evaluator to do, is to take the
95655 context of an evaluation and checks the requirements of every operation
95656 evaluated against that context. With this approach, the evaluator would
95657 report an error only if parts of the context, necessary for the
95658 evaluation, are missing.
95660 If this approach is taken, then the unification of the
95661 dwarf_evaluate_loc_desc, dwarf_expr_executor and dwarf_expr_context
95662 is the next logical step. This makes a design of the DWARF expression
95663 evaluator cleaner and allows more flexibility when supporting future
95664 vendor and GNU extensions.
95666 Additional benefit here is that now all evaluators have access to all
95667 location description types, which means that a vendor extended CFI
95668 rules could support composite location description as well. This also
95669 means that a new evaluator interface can be changed to return a single
95670 struct value (that describes the result of the evaluation) instead of
95671 a caller poking around the dwarf_expr_context internal data for answers
95672 (like it is done currently).
95674 This patch starts the merging process by moving the frame context
95675 information and support from dwarf_expr_executor and
95676 dwarf_evaluate_loc_desc to dwarf_expr_context evaluator. The idea
95677 is to report an error when a given operation requires a frame
95678 information to be resolved, if that information is not present.
95682 * dwarf2/expr.c (ensure_have_frame): New function.
95683 (read_addr_from_reg): Add from frame.c.
95684 (dwarf_expr_context::dwarf_expr_context): Add frame info to
95685 dwarf_expr_context.
95686 (dwarf_expr_context::read_addr_from_reg): Remove.
95687 (dwarf_expr_context::get_reg_value): Move from
95688 dwarf_evaluate_loc_desc.
95689 (dwarf_expr_context::get_frame_base): Move from
95690 dwarf_evaluate_loc_desc.
95691 (dwarf_expr_context::execute_stack_op): Call frame context info
95692 check. Remove use of read_addr_from_reg method.
95693 * dwarf2/expr.h (struct dwarf_expr_context): Add frame info
95694 member, read_addr_from_reg, get_reg_value and get_frame_base
95696 (read_addr_from_reg): Move to expr.c.
95697 * dwarf2/frame.c (read_addr_from_reg): Move to
95698 dwarf_expr_context.
95699 (dwarf_expr_executor::read_addr_from_reg): Remove.
95700 (dwarf_expr_executor::get_frame_base): Remove.
95701 (dwarf_expr_executor::get_reg_value): Remove.
95702 (execute_stack_op): Use read_addr_from_reg function instead of
95703 read_addr_from_reg method.
95704 * dwarf2/loc.c (dwarf_evaluate_loc_desc::get_frame_base): Move
95705 to dwarf_expr_context.
95706 (dwarf_evaluate_loc_desc::get_reg_value): Move to
95707 dwarf_expr_context.
95708 (dwarf_evaluate_loc_desc::read_addr_from_reg): Remove.
95709 (dwarf2_locexpr_baton_eval):Use read_addr_from_reg function
95710 instead of read_addr_from_reg method.
95712 2021-08-05 Zoran Zaric <Zoran.Zaric@amd.com>
95714 Cleanup of the dwarf_expr_context constructor
95715 Move the initial values for dwarf_expr_context class data members
95716 to the class declaration in expr.h.
95720 * dwarf2/expr.c (dwarf_expr_context::dwarf_expr_context):
95721 Remove initial data members values.
95722 * dwarf2/expr.h (dwarf_expr_context): Add initial values
95723 to the class data members.
95725 2021-08-05 Zoran Zaric <Zoran.Zaric@amd.com>
95727 Replace the symbol needs evaluator with a parser
95728 This patch addresses a design problem with the symbol_needs_eval_context
95729 class. It exposes the problem by introducing two new testsuite test
95732 To explain the issue, I first need to explain the dwarf_expr_context
95733 class that the symbol_needs_eval_context class derives from.
95735 The intention behind the dwarf_expr_context class is to commonize the
95736 DWARF expression evaluation mechanism for different evaluation
95737 contexts. Currently in gdb, the evaluation context can contain some or
95738 all of the following information: architecture, object file, frame and
95741 Depending on the information needed to evaluate a given expression,
95742 there are currently three distinct DWARF expression evaluators:
95744 - Frame: designed to evaluate an expression in the context of a call
95745 frame information (dwarf_expr_executor class). This evaluator doesn't
95746 need a compilation unit information.
95748 - Location description: designed to evaluate an expression in the
95749 context of a source level information (dwarf_evaluate_loc_desc
95750 class). This evaluator expects all information needed for the
95751 evaluation of the given expression to be present.
95753 - Symbol needs: designed to answer a question about the parts of the
95754 context information required to evaluate a DWARF expression behind a
95755 given symbol (symbol_needs_eval_context class). This evaluator
95756 doesn't need a frame information.
95758 The functional difference between the symbol needs evaluator and the
95759 others is that this evaluator is not meant to interact with the actual
95760 target. Instead, it is supposed to check which parts of the context
95761 information are needed for the given DWARF expression to be evaluated by
95762 the location description evaluator.
95764 The idea is to take advantage of the existing dwarf_expr_context
95765 evaluation mechanism and to fake all required interactions with the
95766 actual target, by returning back dummy values. The evaluation result is
95767 returned as one of three possible values, based on operations found in a
95770 - SYMBOL_NEEDS_NONE,
95771 - SYMBOL_NEEDS_REGISTERS and
95772 - SYMBOL_NEEDS_FRAME.
95774 The problem here is that faking results of target interactions can yield
95775 an incorrect evaluation result.
95777 For example, if we have a conditional DWARF expression, where the
95778 condition depends on a value read from an actual target, and the true
95779 branch of the condition requires a frame information to be evaluated,
95780 while the false branch doesn't, fake target reads could conclude that a
95781 frame information is not needed, where in fact it is. This wrong
95782 information would then cause the expression to be actually evaluated (by
95783 the location description evaluator) with a missing frame information.
95784 This would then crash the debugger.
95786 The gdb.dwarf2/symbol_needs_eval_fail.exp test introduces this
95787 scenario, with the following DWARF expression:
95789 DW_OP_addr $some_variable
95792 # conditional jump to DW_OP_bregx
95796 # jump to DW_OP_stack_value
95798 DW_OP_bregx $dwarf_regnum 0
95801 This expression describes a case where some variable dictates the
95802 location of another variable. Depending on a value of some_variable, the
95803 variable whose location is described by this expression is either read
95804 from a register or it is defined as a constant value 0. In both cases,
95805 the value will be returned as an implicit location description on the
95808 Currently, when the symbol needs evaluator fakes a memory read from the
95809 address behind the some_variable variable, the constant value 0 is used
95810 as the value of the variable A, and the check returns the
95811 SYMBOL_NEEDS_NONE result.
95813 This is clearly a wrong result and it causes the debugger to crash.
95815 The scenario might sound strange to some people, but it comes from a
95816 SIMD/SIMT architecture where $some_variable is an execution mask. In
95817 any case, it is a valid DWARF expression, and GDB shouldn't crash while
95818 evaluating it. Also, a similar example could be made based on a
95819 condition of the frame base value, where if that value is concluded to
95820 be 0, the variable location could be defaulted to a TLS based memory
95823 The gdb.dwarf2/symbol_needs_eval_timeout.exp test introduces a second
95824 scenario. This scenario is a bit more abstract due to the DWARF
95825 assembler lacking the CFI support, but it exposes a different
95826 manifestation of the same problem. Like in the previous scenario, the
95827 DWARF expression used in the test is valid:
95830 DW_OP_addr $some_variable
95833 # jump to DW_OP_fbreg
95841 # conditional jump to DW_OP_drop
95845 Similarly to the previous scenario, the location of a variable A is an
95846 implicit location description with a constant value that depends on a
95847 value held by a global variable. The difference from the previous case
95848 is that DWARF expression contains a loop instead of just one branch. The
95849 end condition of that loop depends on the expectation that a frame base
95850 value is never zero. Currently, the act of faking the target reads will
95851 cause the symbol needs evaluator to get stuck in an infinite loop.
95853 Somebody could argue that we could change the fake reads to return
95854 something else, but that would only hide the real problem.
95856 The general impression seems to be that the desired design is to have
95857 one class that deals with parsing of the DWARF expression, while there
95858 are virtual methods that deal with specifics of some operations.
95860 Using an evaluator mechanism here doesn't seem to be correct, because
95861 the act of evaluation relies on accessing the data from the actual
95862 target with the possibility of skipping the evaluation of some parts of
95865 To better explain the proposed solution for the issue, I first need to
95866 explain a couple more details behind the current design:
95868 There are multiple places in gdb that handle DWARF expression parsing
95869 for different purposes. Some are in charge of converting the expression
95870 to some other internal representation (decode_location_expression,
95871 disassemble_dwarf_expression and dwarf2_compile_expr_to_ax), some are
95872 analysing the expression for specific information
95873 (compute_stack_depth_worker) and some are in charge of evaluating the
95874 expression in a given context (dwarf_expr_context::execute_stack_op
95875 and decode_locdesc).
95877 The problem is that all those functions have a similar (large) switch
95878 statement that handles each DWARF expression operation. The result of
95879 this is a code duplication and harder maintenance.
95881 As a step into the right direction to solve this problem (at least for
95882 the purpose of a DWARF expression evaluation) the expression parsing was
95883 commonized inside of an evaluator base class (dwarf_expr_context). This
95884 makes sense for all derived classes, except for the symbol needs
95885 evaluator (symbol_needs_eval_context) class.
95887 As described previously the problem with this evaluator is that if the
95888 evaluator is not allowed to access the actual target, it is not really
95891 Instead, the desired function of a symbol needs evaluator seems to fall
95892 more into expression analysis category. This means that a more natural
95893 fit for this evaluator is to be a symbol needs analysis, similar to the
95894 existing compute_stack_depth_worker analysis.
95896 Another problem is that using a heavyweight mechanism of an evaluator
95897 to do an expression analysis seems to be an unneeded overhead. It also
95898 requires a more complicated design of the parent class to support fake
95901 The reality is that the whole symbol_needs_eval_context class can be
95902 replaced with a lightweight recursive analysis function, that will give
95903 more correct result without compromising the design of the
95904 dwarf_expr_context class. The analysis treats the expression byte
95905 stream as a DWARF operation graph, where each graph node can be
95906 visited only once and each operation can decide if the frame context
95907 is needed for their evaluation.
95909 The downside of this approach is adding of one more similar switch
95910 statement, but at least this way the new symbol needs analysis will be
95911 a lightweight mechnism and it will provide a correct result for any
95912 given DWARF expression.
95914 A more desired long term design would be to have one class that deals
95915 with parsing of the DWARF expression, while there would be a virtual
95916 methods that deal with specifics of some DWARF operations. Then that
95917 class would be used as a base for all DWARF expression parsing mentioned
95920 This however, requires a far bigger changes that are out of the scope
95921 of this patch series.
95923 The new analysis requires the DWARF location description for the
95924 argc argument of the main function to change in the assembly file
95925 gdb.python/amd64-py-framefilter-invalidarg.S. Originally, expression
95926 ended with a 0 value byte, which was never reached by the symbol needs
95927 evaluator, because it was detecting a stack underflow when evaluating
95928 the operation before. The new approach does not simulate a DWARF
95929 stack anymore, so the 0 value byte needs to be removed because it
95930 makes the DWARF expression invalid.
95934 * dwarf2/loc.c (class symbol_needs_eval_context): Remove.
95935 (dwarf2_get_symbol_read_needs): New function.
95936 (dwarf2_loc_desc_get_symbol_read_needs): Remove.
95937 (locexpr_get_symbol_read_needs): Use
95938 dwarf2_get_symbol_read_needs.
95940 gdb/testsuite/ChangeLog:
95942 * gdb.python/amd64-py-framefilter-invalidarg.S : Update argc
95943 DWARF location expression.
95944 * lib/dwarf.exp (_location): Handle DW_OP_fbreg.
95945 * gdb.dwarf2/symbol_needs_eval.c: New file.
95946 * gdb.dwarf2/symbol_needs_eval_fail.exp: New file.
95947 * gdb.dwarf2/symbol_needs_eval_timeout.exp: New file.
95949 2021-08-05 Cui,Lili <lili.cui@intel.com>
95951 [PATCH 2/2] Add tests for Intel AVX512_FP16 instructions
95952 Intel AVX512 FP16 instructions use maps 3, 5 and 6. Maps 5 and 6 use 3 bits
95953 in the EVEX.mmm field (0b101, 0b110). Map 5 is for instructions that were FP32
95954 in map 1 (0Fxx). Map 6 is for instructions that were FP32 in map 2 (0F38xx).
95955 There are some exceptions to this rule. Some things in map 1 (0Fxx) with imm8
95956 operands predated our current conventions; those instructions moved to map 3.
95957 FP32 things in map 3 (0F3Axx) found new opcodes in map3 for FP16 because map3
95958 is very sparsely populated. Most of the FP16 instructions share opcodes and
95959 prefix (EVEX.pp) bits with the related FP32 operations.
95961 Intel AVX512 FP16 instructions has new displacements scaling rules, please refer
95962 to the public software developer manual for detail information.
95966 2021-08-05 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
95967 H.J. Lu <hongjiu.lu@intel.com>
95968 Wei Xiao <wei3.xiao@intel.com>
95969 Lili Cui <lili.cui@intel.com>
95971 * testsuite/gas/i386/i386.exp: Run FP16 tests.
95972 * testsuite/gas/i386/avx512_fp16-intel.d: New test.
95973 * testsuite/gas/i386/avx512_fp16-inval-bcast.l: Ditto.
95974 * testsuite/gas/i386/avx512_fp16-inval-bcast.s: Ditto.
95975 * testsuite/gas/i386/avx512_fp16.d: Ditto.
95976 * testsuite/gas/i386/avx512_fp16.s: Ditto.
95977 * testsuite/gas/i386/avx512_fp16_pseudo_ops.d: Ditto.
95978 * testsuite/gas/i386/avx512_fp16_pseudo_ops.s: Ditto.
95979 * testsuite/gas/i386/avx512_fp16_vl-intel.d: Ditto.
95980 * testsuite/gas/i386/avx512_fp16_vl.d: Ditto.
95981 * testsuite/gas/i386/avx512_fp16_vl.s: Ditto.
95982 * testsuite/gas/i386/x86-64-avx512_fp16-intel.d: Ditto.
95983 * testsuite/gas/i386/x86-64-avx512_fp16-inval-bcast.l: Ditto.
95984 * testsuite/gas/i386/x86-64-avx512_fp16-inval-bcast.s: Ditto.
95985 * testsuite/gas/i386/x86-64-avx512_fp16.d: Ditto.
95986 * testsuite/gas/i386/x86-64-avx512_fp16.s: Ditto.
95987 * testsuite/gas/i386/x86-64-avx512_fp16_pseudo_ops.d: Ditto.
95988 * testsuite/gas/i386/x86-64-avx512_fp16_pseudo_ops.s: Ditto.
95989 * testsuite/gas/i386/x86-64-avx512_fp16_vl-intel.d: Ditto.
95990 * testsuite/gas/i386/x86-64-avx512_fp16_vl.d: Ditto.
95991 * testsuite/gas/i386/x86-64-avx512_fp16_vl.s: Ditto.
95992 * testsuite/gas/i386/x86-64-avx512_fp16-inval-register.l: Ditto.
95993 * testsuite/gas/i386/x86-64-avx512_fp16-inval-register.s: Ditto.
95994 * testsuite/gas/i386/x86-64-avx512_fp16-bad.d: Ditto.
95995 * testsuite/gas/i386/x86-64-avx512_fp16-bad.s: Ditto.
95996 * testsuite/gas/i386/x86-64-default-suffix-avx.d: Add new testcase.
95997 * testsuite/gas/i386/x86-64-default-suffix.d: Ditto.
95998 * testsuite/gas/i386/x86-64-default-suffix.s: Ditto.
95999 * testsuite/gas/i386/xmmword.l: Ditto.
96000 * testsuite/gas/i386/xmmword.s: Ditto.
96002 2021-08-05 Cui,Lili <lili.cui@intel.com>
96004 [PATCH 1/2] Enable Intel AVX512_FP16 instructions
96005 Intel AVX512 FP16 instructions use maps 3, 5 and 6. Maps 5 and 6 use 3 bits
96006 in the EVEX.mmm field (0b101, 0b110). Map 5 is for instructions that were FP32
96007 in map 1 (0Fxx). Map 6 is for instructions that were FP32 in map 2 (0F38xx).
96008 There are some exceptions to this rule. Some things in map 1 (0Fxx) with imm8
96009 operands predated our current conventions; those instructions moved to map 3.
96010 FP32 things in map 3 (0F3Axx) found new opcodes in map3 for FP16 because map3
96011 is very sparsely populated. Most of the FP16 instructions share opcodes and
96012 prefix (EVEX.pp) bits with the related FP32 operations.
96014 Intel AVX512 FP16 instructions has new displacements scaling rules, please refer
96015 to the public software developer manual for detail information.
96019 2021-08-05 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
96020 H.J. Lu <hongjiu.lu@intel.com>
96021 Wei Xiao <wei3.xiao@intel.com>
96022 Lili Cui <lili.cui@intel.com>
96024 * config/tc-i386.c (struct Broadcast_Operation): Adjust comment.
96025 (cpu_arch): Add .avx512_fp16.
96026 (cpu_noarch): Add noavx512_fp16.
96027 (pte): Add evexmap5 and evexmap6.
96028 (build_evex_prefix): Handle EVEXMAP5 and EVEXMAP6.
96029 (check_VecOperations): Handle {1to32}.
96030 (check_VecOperands): Handle CheckRegNumb.
96031 (check_word_reg): Handle Toqword.
96032 (i386_error): Add invalid_dest_and_src_register_set.
96033 (match_template): Handle invalid_dest_and_src_register_set.
96034 * doc/c-i386.texi: Document avx512_fp16, noavx512_fp16.
96038 2021-08-05 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
96039 H.J. Lu <hongjiu.lu@intel.com>
96040 Wei Xiao <wei3.xiao@intel.com>
96041 Lili Cui <lili.cui@intel.com>
96043 * i386-dis.c (EXwScalarS): New.
96048 (EXEvexXwb): Ditto.
96049 (DistinctDest_Fixup): Ditto.
96050 (enum): Add xh_mode, evex_half_bcst_xmmqh_mode, evex_half_bcst_xmmqdh_mode
96052 (enum): Add PREFIX_EVEX_0F3A08_W_0, PREFIX_EVEX_0F3A0A_W_0,
96053 PREFIX_EVEX_0F3A26, PREFIX_EVEX_0F3A27, PREFIX_EVEX_0F3A56,
96054 PREFIX_EVEX_0F3A57, PREFIX_EVEX_0F3A66, PREFIX_EVEX_0F3A67,
96055 PREFIX_EVEX_0F3AC2, PREFIX_EVEX_MAP5_10, PREFIX_EVEX_MAP5_11,
96056 PREFIX_EVEX_MAP5_1D, PREFIX_EVEX_MAP5_2A, PREFIX_EVEX_MAP5_2C,
96057 PREFIX_EVEX_MAP5_2D, PREFIX_EVEX_MAP5_2E, PREFIX_EVEX_MAP5_2F,
96058 PREFIX_EVEX_MAP5_51, PREFIX_EVEX_MAP5_58, PREFIX_EVEX_MAP5_59,
96059 PREFIX_EVEX_MAP5_5A_W_0, PREFIX_EVEX_MAP5_5A_W_1,
96060 PREFIX_EVEX_MAP5_5B_W_0, PREFIX_EVEX_MAP5_5B_W_1,
96061 PREFIX_EVEX_MAP5_5C, PREFIX_EVEX_MAP5_5D, PREFIX_EVEX_MAP5_5E,
96062 PREFIX_EVEX_MAP5_5F, PREFIX_EVEX_MAP5_78, PREFIX_EVEX_MAP5_79,
96063 PREFIX_EVEX_MAP5_7A, PREFIX_EVEX_MAP5_7B, PREFIX_EVEX_MAP5_7C,
96064 PREFIX_EVEX_MAP5_7D_W_0, PREFIX_EVEX_MAP6_13, PREFIX_EVEX_MAP6_56,
96065 PREFIX_EVEX_MAP6_57, PREFIX_EVEX_MAP6_D6, PREFIX_EVEX_MAP6_D7
96066 (enum): Add EVEX_MAP5 and EVEX_MAP6.
96067 (enum): Add EVEX_W_MAP5_5A, EVEX_W_MAP5_5B,
96068 EVEX_W_MAP5_78_P_0, EVEX_W_MAP5_78_P_2, EVEX_W_MAP5_79_P_0,
96069 EVEX_W_MAP5_79_P_2, EVEX_W_MAP5_7A_P_2, EVEX_W_MAP5_7A_P_3,
96070 EVEX_W_MAP5_7B_P_2, EVEX_W_MAP5_7C_P_0, EVEX_W_MAP5_7C_P_2,
96071 EVEX_W_MAP5_7D, EVEX_W_MAP6_13_P_0, EVEX_W_MAP6_13_P_2,
96072 (get_valid_dis386): Properly handle new instructions.
96073 (intel_operand_size): Handle new modes.
96074 (OP_E_memory): Ditto.
96076 * i386-dis-evex.h: Updated for AVX512_FP16.
96077 * i386-dis-evex-mod.h: Updated for AVX512_FP16.
96078 * i386-dis-evex-prefix.h: Updated for AVX512_FP16.
96079 * i386-dis-evex-reg.h : Updated for AVX512_FP16.
96080 * i386-dis-evex-w.h : Updated for AVX512_FP16.
96081 * i386-gen.c (cpu_flag_init): Add CPU_AVX512_FP16_FLAGS,
96082 and CPU_ANY_AVX512_FP16_FLAGS. Update CPU_ANY_AVX512F_FLAGS
96083 and CPU_ANY_AVX512BW_FLAGS.
96084 (cpu_flags): Add CpuAVX512_FP16.
96085 (opcode_modifiers): Add DistinctDest.
96086 * i386-opc.h (enum): (AVX512_FP16): New.
96087 (i386_opcode_modifier): Add reqdistinctreg.
96088 (i386_cpu_flags): Add cpuavx512_fp16.
96089 (EVEXMAP5): Defined as a macro.
96091 * i386-opc.tbl: Add Intel AVX512_FP16 instructions.
96092 * i386-init.h: Regenerated.
96093 * i386-tbl.h: Ditto.
96095 2021-08-05 Alan Modra <amodra@gmail.com>
96097 PR28167, vms-alpha build_module_list
96099 * vms-alpha.c (build_module_list): Malloc and free section contents.
96100 Don't read past end of section.
96102 2021-08-05 Alan Modra <amodra@gmail.com>
96104 PR28166, _bfd_elf_mips_get_relocated_section_contents
96105 Some of the code paths unpacking mips relocs left arelent->sym_ptr_ptr
96109 * elf64-mips.c (mips_elf64_slurp_one_reloc_table): Don't leave
96110 sym_ptr_ptr uninitialised.
96112 2021-08-05 Alan Modra <amodra@gmail.com>
96114 PR28165, buffer overflow in elf32-rx.c:rx_info_to_howto_rela
96116 * elf32-rx.c (rx_elf_howto_table): Add missing empty entries.
96117 (rx_info_to_howto_rela): Assert rx_elf_howto_table is correct size.
96118 Use actual size when sanity checking r_type.
96120 2021-08-05 Alan Modra <amodra@gmail.com>
96122 Re: elf: Treat undefined version as hidden
96123 Fix fallout in cris testsuite
96126 * ld-cris/libdso-1c.d: Update for version display change.
96127 * ld-cris/libdso-15b.d: Likewise.
96129 2021-08-05 Andrew Burgess <andrew.burgess@embecosm.com>
96131 gdb/testsuite: update test gdb.base/step-over-syscall.exp
96132 I was looking at PR gdb/19675 and the related test
96133 gdb.base/step-over-syscall.exp. This test includes a call to kfail
96134 when we are testing a displaced step over a clone syscall.
96136 While looking at the test I removed the call to kfail and ran the
96137 test, and was surprised that the test passed.
96139 I ran the test a few times and it does sometimes fail, but mostly it
96142 PR gdb/19675 describes how, when we displaced step over a clone, the
96143 new thread is created with a $pc in the displaced step buffer. GDB
96144 then fails to "fix" this $pc (for the new thread), and the thread will
96145 be set running with its current $pc value. This means that the new
96146 thread will just start executing from whatever happens to be after the
96147 displaced stepping buffer.
96149 In the original PR gdb/19675 bug report Yao Qi was seeing the new
96150 thread cause a segfault, the problem is, what actually happens is
96153 On my machine, I'm seeing the new thread reenter main, it then starts
96154 trying to run the test again (in the new thread). This just happens
96155 to be safe enough (in this simple test) that most of the time the
96156 inferior doesn't crash.
96158 In this commit I try to make the test slightly more likely to fail by
96159 doing a couple of things.
96161 First, I added a static variable to main, this is set true when the
96162 first thread enters main, if a second thread ever enters main then I
96165 Second, when the test is finishing I want to ensure that the new
96166 threads have had a chance to do "something bad" if they are going to.
96167 So I added a global counter, as each thread starts successfully it
96168 decrements the counter. The main thread does not proceed to the final
96169 marker function (where GDB has placed a breakpoint) until all threads
96170 have started successfully. This means that if the newly created
96171 thread doesn't successfully enter clone_fn then the counter will never
96172 reach zero and the test will timeout.
96174 With these two changes my hope is that the test should fail more
96175 reliably, and so, I have also changed the test to call setup_kfail
96176 before the specific steps that we expect to misbehave instead of just
96177 calling kfail and skipping parts of the test completely. The benefit
96178 of this is that if/when we fix GDB this test will start to KPASS and
96179 we'll know to update this test to remove the setup_kfail call.
96181 2021-08-05 GDB Administrator <gdbadmin@sourceware.org>
96183 Automatic date update in version.in
96185 2021-08-05 Lancelot SIX <lsix@lancelotsix.com>
96187 gdb: Use unwinder name in frame_info::to_string
96188 While working on a stack unwinding issue using 'set debug frame on', I
96189 noticed the frame_info::to_string method could be slightly improved.
96191 Unwinders have been given a name in
96192 a154d838a70e96d888620c072e2d6ea8bdf044ca. Before this patch, frame_info
96193 debug output prints the host address of the used unwinder, which is not
96194 easy to interpret. This patch proposes to use the unwinder name
96195 instead since we now have it.
96199 {level=1,type=NORMAL_FRAME,unwind=0x2ac1763ec0,pc=0x3ff7fc3460,id={stack=0x3ff7ea79b0,code=0x0000003ff7fc33ac,!special},func=0x3ff7fc33ac}
96203 {level=1,type=NORMAL_FRAME,unwinder="riscv prologue",pc=0x3ff7fc3460,id={stack=0x3ff7ea79b0,code=0x0000003ff7fc33ac,!special},func=0x3ff7fc33ac}
96205 Tested on riscv64-linux-gnu.
96207 2021-08-04 Simon Marchi <simon.marchi@polymtl.ca>
96209 gdb/testsuite: fix gdb.base/info-macros.exp with clang
96210 The test gdb.base/info-macros.exp says that it doesn't pass the "debug"
96211 option to prepare_for_testing because that would cause -g to appear
96212 after -g3 on the command line, and that would cause some gcc versions to
96213 not include macro info. I don't know what gcc versions this refers to.
96214 I tested with gcc 4.8, and that works fine with -g after -g3.
96216 The current state is problematic when testing with CC_FOR_TARGET=clang,
96217 because then only -fdebug-macro is included. No -g switch if included,
96218 meaning we get a binary without any debug info, and the test fails.
96220 One way to fix it would be to add "debug" to the options when the
96223 However, the solution I chose was to specify "debug" in any case, even
96224 for gcc. Other macro tests such as gdb.base/macscp.exp do perfectly
96225 fine with it. Also, this lets the test use the debug flag specified by
96226 the board file. For example, we can test with GCC and DWARF 5, with:
96228 $ make check RUNTESTFLAGS="--target_board unix/gdb:debug_flags=-gdwarf-5" TESTS="gdb.base/info-macros.exp"
96230 With the hard-coded -g3, this wouldn't actually test with DWARF 5.
96232 Change-Id: I33fa92ee545007d3ae9c52c4bb2d5be6b5b698f1
96234 2021-08-04 Simon Marchi <simon.marchi@polymtl.ca>
96236 gdb: avoid dereferencing empty str_offsets_base optional in dwarf_decode_macros
96237 Since 4d7188abfdf2 ("gdbsupport: add debug assertions in
96238 gdb::optional::get"), some macro-related tests fail on Ubuntu 20.04 with
96239 the system gcc 9.3.0 compiler when building with _GLIBCXX_DEBUG. For
96240 example, gdb.base/info-macros.exp results in:
96242 (gdb) break -qualified main
96243 /home/smarchi/src/binutils-gdb/gdb/../gdbsupport/gdb_optional.h:206: internal-error: T& gdb::optional<T>::get() [with T = long unsigned int]: Assertion `this->has_value ()' failed.
96245 The binary contains DWARF 4 debug info and includes a pre-standard
96246 (pre-DWARF 5) .debug_macro section. The CU doesn't have a
96247 DW_AT_str_offsets_base attribute (which doesn't exist in DWARF 4). The
96248 field dwarf2_cu::str_offsets_base is therefore empty. At
96249 dwarf2/read.c:24138, we unconditionally read the value in the optional,
96250 which triggers the assertion shown above.
96252 The same thing happens when building the test program with DWARF 5 with
96253 the same gcc compiler, as that version of gcc doesn't use indirect
96254 string forms, even with DWARF 5. So it still doesn't add a
96255 DW_AT_str_offsets_base attribute on the CU.
96257 Fix that by propagating down a gdb::optional<ULONGEST> for the str
96258 offsets base instead of ULONGEST. That value is only used in
96259 dwarf_decode_macro_bytes, when encountering an "strx" macro operation
96260 (DW_MACRO_define_strx or DW_MACRO_undef_strx). Add a check there that
96261 we indeed have a value in the optional before reading it. This is
96262 unlikely to happen, but could happen in theory with an erroneous file
96263 that uses DW_MACRO_define_strx but does not provide a
96264 DW_AT_str_offsets_base (in practice, some things would probably have
96265 failed before and stopped processing of debug info). I tested the
96266 complaint by inverting the condition and using a clang-compiled binary,
96267 which uses the strx operators. This is the result:
96269 During symbol reading: use of DW_MACRO_define_strx with unknown string offsets base [in module /home/simark/build/binutils-gdb/gdb/testsuite/outputs/gdb.base/info-macros/info-macros]
96271 The test now passes cleanly with the setup mentioned above, and the
96272 testsuite looks on par with how it was before 4d7188abfdf2.
96274 Change-Id: I7ebd2724beb7b9b4178872374c2a177aea696e77
96276 2021-08-04 Simon Marchi <simon.marchi@polymtl.ca>
96278 gdb: fix typo in complaint in dwarf2/macro.c
96279 I saw this complaint when my code had some bug, and spotted the typo.
96280 Fix it, and while at it mention DW_MACRO as well (it would be confusing
96281 to only see DW_MACINFO with a file that uses a DWARF 5 .debug_macro
96282 section). I contemplated the idea of passing the knowledge of whether
96283 we are dealing with a .debug_macro section or .debug_macinfo section, to
96284 print only the right one. But in the end, I don't think that trouble is
96285 necessary for a complaint nobody is going to see.
96287 Change-Id: I276ce8da65c3eac5304f64a1e246358ed29cdbbc
96289 2021-08-04 Simon Marchi <simon.marchi@polymtl.ca>
96291 gdb: fix warnings in bsd-kvm.c
96292 Building on OpenBSD, I get warnings like:
96295 /home/simark/src/binutils-gdb/gdb/bsd-kvm.c:241:18: error: ISO C++11 does not allow conversion from string literal to 'char *' [-Werror,-Wwritable-strings]
96296 nl[0].n_name = "_dumppcb";
96299 Silence those by adding casts.
96301 Change-Id: I2bef4eebcc306762a4e3e5b5c52f67ecf2820503
96303 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
96305 IBM Z: Remove lpswey parameter
96307 * s390-opc.c (INSTR_SIY_RD): New instruction format.
96308 (MASK_SIY_RD): New instruction mask.
96309 * s390-opc.txt: Change instruction format of lpswey to SIY_RD.
96312 * testsuite/gas/s390/zarch-arch14.d: Remove last operand of
96314 * testsuite/gas/s390/zarch-arch14.s: Likewise.
96316 2021-08-04 Alan Modra <amodra@gmail.com>
96318 PR28162, segment fault in mips_elf_assign_gp
96319 For the testcase in the PR, _bfd_mips_elf32_gprel16_reloc is passed a
96320 NULL output_bfd. As expected for reloc special functions if called by
96321 objdump or when final linking. The function attempts to find the
96323 output_bfd = symbol->section->output_section->owner;
96324 That makes some sense, since when handling a gp-relative reloc we need
96325 the relevant gp to which the symbol is relative. Possibly the gp
96326 value can be one for a shared library? But that doesn't seem useful
96327 or supported by the various abi docs and won't work as written.
96328 Symbols defined in shared libraries have section->output_section
96329 NULL, and what's more the code in mips_elf_assign_gp isn't set up to
96330 look at shared library symbols.
96332 Also, if the symbol is a SHN_ABS one the owner of *ABS* section is
96333 NULL, which will result in the testcase segfault. The only gp to
96334 which an absolute symbol can be relative is the linker output bfd when
96335 linking, or the input bfd when not. This patch arranges to do that
96336 for all gp-relative reloc symbols.
96338 * elf32-mips.c (_bfd_mips_elf32_gprel16_reloc): Don't use the
96339 section symbol to find the output bfd, use input_section.
96340 (mips_elf_gprel32_reloc, mips16_gprel_reloc): Likewise.
96341 * elf64-mips.c (mips_elf64_gprel16_reloc): Likewise.
96342 (mips_elf64_literal_reloc, mips_elf64_gprel32_reloc): Likewise.
96343 (mips16_gprel_reloc): Likewise.
96345 2021-08-04 Tom de Vries <tdevries@suse.de>
96347 [gdb/symtab] Use lambda function instead of addrmap_foreach_check
96348 Use a lambda function instead of addrmap_foreach_check,
96349 which removes the need for static variables.
96351 Also remove unnecessary static on local var temp_obstack in test_addrmap.
96355 2021-08-04 Tom de Vries <tdevries@suse.de>
96357 * addrmap.c (addrmap_foreach_check): Remove.
96358 (array, val1, val2): Move ...
96359 (test_addrmap): ... here. Remove static on temp_obstack. Use lambda
96360 function instead of addrmap_foreach_check.
96362 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
96364 elf: Treat undefined version as hidden
96365 Since undefined version can't be used to resolve any references without
96366 the original definition, treat it as hidden.
96371 * elf.c (_bfd_elf_get_symbol_version_string): Treat undefined
96377 * testsuite/ld-elf/linux-x86.exp: Run PR binutils/28158 tests.
96378 * testsuite/ld-elf/pr28158-1.c: New file.
96379 * testsuite/ld-elf/pr28158-2.S: Likewise.
96380 * testsuite/ld-elf/pr28158.nd: Likewise.
96381 * testsuite/ld-elf/pr28158.rd: Likewise.
96382 * testsuite/ld-elf/pr28158.t: Likewise.
96383 * testsuite/ld-elfvers/vers2.dsym: Updated.
96384 * testsuite/ld-elfvers/vers3.dsym: Likewise.
96385 * testsuite/ld-elfvers/vers6.dsym: Likewise.
96386 * testsuite/ld-elfvers/vers19.dsym: Likewise.
96387 * testsuite/ld-elfvers/vers22.dsym: Likewise.
96388 * testsuite/ld-elfvers/vers23.dsym: Likewise.
96389 * testsuite/ld-elfvers/vers23d.dsym: Likewise.
96390 * testsuite/ld-elfvers/vers27d4.dsym: Likewise.
96391 * testsuite/ld-elfvers/vers28c.dsym: Likewise.
96393 2021-08-04 Tom de Vries <tdevries@suse.de>
96395 [gdb/symtab] Implement addrmap_mutable_find
96396 Currently addrmap_mutable_find is not implemented:
96399 addrmap_mutable_find (struct addrmap *self, CORE_ADDR addr)
96401 /* Not needed yet. */
96402 internal_error (__FILE__, __LINE__,
96403 _("addrmap_find is not implemented yet "
96404 "for mutable addrmaps"));
96408 I implemented this because I needed it during debugging, to be able to do:
96410 (gdb) p ((dwarf2_psymtab *)addrmap_find (map, addr))->filename
96412 before and after a call to addrmap_set_empty.
96414 Since this is not used otherwise, added addrmap unit test.
96416 Build on x86_64-linux, tested by doing:
96418 $ gdb -q -batch -ex "maint selftest addrmap"
96419 Running selftest addrmap.
96420 Ran 1 unit tests, 0 failed
96425 2021-08-03 Tom de Vries <tdevries@suse.de>
96427 * gdb/addrmap.c (addrmap_mutable_find): Implement
96428 [GDB_SELF_TESTS] (CHECK_ADDRMAP_FIND): New macro.
96429 [GDB_SELF_TESTS] (core_addr, addrmap_foreach_check, test_addrmap)
96430 (_initialize_addrmap): New function.
96432 2021-08-04 Clément Chigot <clement.chigot@atos.net>
96434 gas: correctly output XCOFF tbss symbols with XTY_CM type.
96435 Global tbss symbols weren't correctly handled and were generating
96436 a symbol with XTY_SD instead of XTY_CM as expected.
96439 * config/tc-ppc.c (ppc_frog_symbol): Generate a XTY_CM when
96440 a symbol has a storage class of XMC_UL.
96442 2021-08-04 Clément Chigot <clement.chigot@atos.net>
96444 gas: always add dummy symbols when creating XCOFF sections.
96445 Most of the algorithms for XCOFF in tc-ppc.c assume that
96446 the csects field of a ppc_xcoff_section isn't NULL.
96447 This was already made for most of the sections with the creation
96449 This patch simply mades it default when creating a xcoff_section.
96452 * config/tc-ppc.c (ppc_init_xcoff_section): Always create
96454 (md_begin): Adjust ppc_init_xcoff_section call.
96455 (ppc_comm): Likewise.
96456 (ppc_change_csect): Likewise.
96458 2021-08-04 Alan Modra <amodra@gmail.com>
96460 PR28156, rename.c doesn't compile with MinGW
96461 Guard against lack of struct timespec definition.
96464 * rename.c (get_stat_atime, get_stat_mtime): Don't compile
96465 unless HAVE_UTIMENSAT is defined.
96467 2021-08-04 Alan Modra <amodra@gmail.com>
96469 PR28155, Superfluous "the" in the man page
96471 * ld.texi (Options <runtime library name>): Correct grammar.
96473 revise PE IMAGE_SCN_LNK_NRELOC_OVFL test
96474 * coffcode.h (coff_set_alignment_hook): Test that the resulting
96475 reloc count is not less than 0xffff.
96477 2021-08-04 Simon Marchi <simon.marchi@polymtl.ca>
96479 gdb: follow-fork: push target and add thread in target_follow_fork
96480 In the context of ROCm-gdb [1], the ROCm target sits on top of the
96481 linux-nat target. when a process forks, it needs to carry over some
96482 data from the forking inferior to the fork child inferior. Ideally, the
96483 ROCm target would implement the follow_fork target_ops method, but there
96484 are some small problems. This patch fixes these, which helps the ROCm
96485 target, but also makes things more consistent and a bit nicer in
96486 general, I believe.
96488 The main problem is: when follow-fork-mode is "parent",
96489 target_follow_fork is called with the parent as the current inferior.
96490 When it's "child", target_follow_fork is called with the child as the
96491 current inferior. This means that target_follow_fork is sometimes
96492 called on the parent's target stack and sometimes on the child's target
96495 The parent's target stack may contain targets above the process target,
96496 such as the ROCm target. So if follow-fork-child is "parent", the ROCm
96497 target would get notified of the fork and do whatever is needed. But
96498 the child's target stack, at that moment, only contains the exec and
96499 process target copied over from the parent. The child's target stack is
96500 set up by follow_fork_inferior, before calling target_follow_fork. In
96501 that case, the ROCm target wouldn't get notified of the fork.
96503 For consistency, I think it would be good to always call
96504 target_follow_fork on the parent inferior's target stack. I think it
96505 makes sense as a way to indicate "this inferior has called fork, do
96506 whatever is needed". The desired outcome of the fork (whether an
96507 inferior is created for the child, do we need to detach from the child)
96508 can be indicated by passed parameter.
96510 I therefore propose these changes:
96512 - make follow_fork_inferior always call target_follow_fork with the
96513 parent as the current inferior. That lets all targets present on the
96514 parent's target stack do some fork-related handling and push
96515 themselves on the fork child's target stack if needed.
96517 For this purpose, pass the child inferior down to target_follow_fork
96518 and follow_fork implementations. This is nullptr if no inferior is
96519 created for the child, because we want to detach from it.
96521 - as a result, in follow_fork_inferior, detach from the parent inferior
96522 (if needed) only after the target_follow_fork call. This is needed
96523 because we want to call target_follow_fork before the parent's
96524 target stack is torn down.
96526 - hand over to the targets in the parent's target stack (including the
96527 process target) the responsibility to push themselves, if needed, to
96528 the child's target stack. Also hand over the responsibility to the
96529 process target, at the same time, to create the child's initial
96530 thread (just like we do for follow_exec).
96532 - pass the child inferior to exec_on_vfork, so we don't need to swap
96533 the current inferior between parent and child. Nothing in
96534 exec_on_vfork depends on the current inferior, after this change.
96536 Although this could perhaps be replaced with just having the exec
96537 target implement follow_fork and push itself in the child's target
96538 stack, like the process target does... We would just need to make
96539 sure the process target calls beneath()->follow_fork(...). I'm not
96540 sure about this one.
96544 * target.h (struct target_ops) <follow_fork>: Add inferior*
96546 (target_follow_fork): Likewise.
96547 * target.c (default_follow_fork): Likewise.
96548 (target_follow_fork): Likewise.
96549 * fbsd-nat.h (class fbsd_nat_target) <follow_fork>: Likewise.
96550 (fbsd_nat_target::follow_fork): Likewise, and call
96551 inf_ptrace_target::follow_fork.
96552 * linux-nat.h (class linux_nat_target) <follow_fork>: Likewise.
96553 * linux-nat.c (linux_nat_target::follow_fork): Likewise, and
96554 call inf_ptrace_target::follow_fork.
96555 * obsd-nat.h (obsd_nat_target) <follow_fork>: Likewise.
96556 * obsd-nat.c (obsd_nat_target::follow_fork): Likewise, and call
96557 inf_ptrace_target::follow_fork.
96558 * remote.c (class remote_target) <follow_fork>: Likewise.
96559 (remote_target::follow_fork): Likewise, and call
96560 process_stratum_target::follow_fork.
96561 * process-stratum-target.h (class process_stratum_target)
96562 <follow_fork>: New.
96563 * process-stratum-target.c
96564 (process_stratum_target::follow_fork): New.
96565 * target-delegates.c: Re-generate.
96567 [1] https://github.com/ROCm-Developer-Tools/ROCgdb
96569 Change-Id: I460bd0af850f0485e8aed4b24c6d8262a4c69929
96571 2021-08-04 GDB Administrator <gdbadmin@sourceware.org>
96573 Automatic date update in version.in
96575 2021-08-03 Carl Love <cel@us.ibm.com>
96577 Fixes for mi-fortran-modules.exp fixes
96578 Output has additional information for a given filename.
96580 gdb/testsuite/ChangeLog
96581 * gdb.mi/mi-fortran-modules.exp (system_modules_pattern,
96582 system_module_symbols_pattern): Add check for additional symbols
96585 2021-08-03 Simon Marchi <simon.marchi@polymtl.ca>
96587 gdbsupport: add debug assertions in gdb::optional::get
96588 The libstdc++ version of optional contains some runtime checks enabled
96589 when _GLIBCXX_DEBUG is defined. I think it would be useful if our
96590 version contained similar checks.
96592 Add checks in the two `get` methods, also conditional on _GLIBCXX_DEBUG.
96593 I think it's simpler to use that macro rather than introducing a new
96594 GDB-specific one, as I think that if somebody is interested in enabling
96595 these runtime checks, they'll also be interested in enabling the
96596 libstdc++ runtime checks (and vice-versa).
96598 I implemented these checks using gdb_assert. Note that gdb_assert
96599 throws (after querying the user), and we are in noexcept methods. That
96600 means that std::terminate / abort will immediately be called. I think
96601 this is ok, since if those were "real" _GLIBCXX_DEBUG checks, abort
96602 would be called straight away.
96604 If I add a dummy failure, it looks like so:
96606 $ ./gdb -q -nx --data-directory=data-directory
96607 /home/simark/src/binutils-gdb/gdb/../gdbsupport/gdb_optional.h:206: internal-error: T& gdb::optional<T>::get() [with T = int]: Assertion `this->has_value ()' failed.
96608 A problem internal to GDB has been detected,
96609 further debugging may prove unreliable.
96610 Quit this debugging session? (y or n) n
96611 [1] 658767 abort (core dumped) ./gdb -q -nx --data-directory=data-directory
96613 Change-Id: Iadfdcd131425bd2ca6a2de30d7b22e9b3cc67793
96615 2021-08-03 Alok Kumar Sharma <AlokKumar.Sharma@amd.com>
96617 [gdb/testsuite] templates.exp to accept clang++ output
96618 Please consider below testcase with intended error.
96620 constexpr const char cstring[] = "Eta";
96621 template <const char*, typename T> class Column {};
96622 using quick = Column<cstring,double>; // cstring without '&'
96629 It produces below error.
96631 no member named 'ls' in 'Column<&cstring, double>'.
96633 Please note that error message contains '&' for cstring, which is absent
96635 Clang++ does not generate & in such cases and this should also be
96636 accepted as correct output.
96638 gdb/testsuite/ChangeLog:
96640 * gdb.cp/templates.exp: Accept different but correct output
96641 from the Clang++ compiled binary also.
96643 2021-08-03 GDB Administrator <gdbadmin@sourceware.org>
96645 Automatic date update in version.in
96647 2021-08-02 Tom Tromey <tromey@adacore.com>
96649 Handle compiler-generated suffixes in Ada names
96650 The compiler may add a suffix to a mangled name. A typical example
96651 would be splitting a function and creating a ".cold" variant.
96653 This patch changes Ada decoding (aka demangling) to handle these
96654 suffixes. It also changes the encoding process to handle them as
96657 A symbol like "function.cold" will now be displayed to the user as
96658 "function[cold]". The "." is not simply preserved because that is
96659 already used in Ada.
96661 2021-08-02 Tom Tromey <tromey@adacore.com>
96663 Remove uses of fprintf_symbol_filtered
96664 I believe that many calls to fprintf_symbol_filtered are incorrect.
96665 In particular, there are some that pass a symbol's print name, like:
96667 fprintf_symbol_filtered (gdb_stdout, sym->print_name (),
96668 current_language->la_language, DMGL_ANSI);
96670 fprintf_symbol_filtered uses the "demangle" global to decide whether
96671 or not to demangle -- but print_name does this as well. This can lead
96672 to double-demangling. Normally this could be innocuous, except I also
96673 plan to change Ada demangling in a way that causes this to fail.
96675 2021-08-02 Tom Tromey <tromey@adacore.com>
96677 Handle type qualifier for enumeration name
96678 Pierre-Marie noticed that the Ada expression "TYPE'(NAME)" resolved
96679 incorrectly when "TYPE" was an enumeration type. Here, "NAME" should
96682 This patch fixes this problem. Note that the patch is not perfect --
96683 it does not give an error if TYPE is an enumeration type but NAME is
96684 not an enumerator but does have some other meaning in scope. Fixing
96685 this proved difficult, and so I've left it out.
96687 2021-08-02 Tom Tromey <tromey@adacore.com>
96689 Remove the type_qualifier global
96690 The type_qualifier global is no longer needed in the Ada expression
96691 parser, so this removes it.
96693 2021-08-02 Tom Tromey <tromey@adacore.com>
96695 Defer Ada character literal resolution
96696 In Ada, an enumeration type can use a character literal as one of the
96697 enumerators. The Ada expression parser handles the appropriate
96700 It turns out, though, that this conversion was handled incorrectly.
96701 For an expression like TYPE'(EXP), the conversion would be done for
96702 any such literal appearing in EXP -- but only the outermost such
96703 expression should really be affected.
96705 This patch defers the conversion until the resolution phase, fixing
96708 2021-08-02 Tom Tromey <tromey@adacore.com>
96710 Refactor Ada resolution
96711 In a subsequent patch, it will be convenient if an Ada expression
96712 operation can supply its own replacement object. This patch refactors
96713 Ada expression resolution to make this possible.
96715 Remove add_symbols_from_enclosing_procs
96716 I noticed that add_symbols_from_enclosing_procs is empty, and can be
96717 removed. The one caller, ada_add_local_symbols, can also be
96718 simplified, removing some code that, I think, was an incorrect attempt
96719 to handle nested functions.
96721 2021-08-02 Tom Tromey <tromey@adacore.com>
96723 Avoid crash in varobj deletion
96724 PR varobj/28131 points out a crash in the varobj deletion code. It
96725 took a while to reproduce this, but essentially what happens is that a
96726 top-level varobj deletes its root object, then deletes the "dynamic"
96727 object. However, deletion of the dynamic object may cause
96728 ~py_varobj_iter to run, which in turn uses gdbpy_enter_varobj:
96730 gdbpy_enter_varobj::gdbpy_enter_varobj (const struct varobj *var)
96731 : gdbpy_enter (var->root->exp->gdbarch, var->root->exp->language_defn)
96735 However, because var->root has already been destroyed, this is
96738 I've added a new test case. This doesn't reliably crash, but the
96739 problem can easily be seen under valgrind (and, I presume, with ASAN,
96740 though I did not try this).
96742 Tested on x86-64 Fedora 32. I also propose putting this on the GDB 11
96743 branch, with a suitable ChangeLog entry of course.
96745 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28131
96747 2021-08-02 Tom de Vries <tdevries@suse.de>
96749 [gdb/testsuite] Fix gdb.dwarf2/dw2-using-debug-str.exp with cc-with-dwz-m
96750 When running with target board cc-with-dwz-m, we run into:
96752 (gdb) file dw2-using-debug-str-no-debug-str^M
96753 Reading symbols from dw2-using-debug-str-no-debug-str...^M
96754 (gdb) FAIL: gdb.dwarf2/dw2-using-debug-str.exp: file dw2-using-debug-str
96757 With native, the .debug_str section is present in the
96758 dw2-using-debug-str executable, and removed from the
96759 dw2-using-debug-str-no-debug-str executable. When loading the latter, a dwarf
96760 error is triggered.
96762 With cc-with-dwz-m, the .debug_str section is not present in the
96763 dw2-using-debug-str executable, because it's already moved to
96764 .tmp/dw2-using-debug-str.dwz. Consequently, the removal has no effect, and no
96765 dwarf error is triggered, which causes the FAIL.
96767 The same problem arises with target board cc-with-gnu-debuglink.
96769 Fix this by detecting whether the .debug_str section is missing, and skipping
96770 the remainder of the test-case.
96772 Tested on x86_64-linux.
96774 gdb/testsuite/ChangeLog:
96776 2021-08-02 Tom de Vries <tdevries@suse.de>
96778 * gdb.dwarf2/dw2-using-debug-str.exp: Handle missing .debug_str
96779 section in dw2-using-debug-str.
96781 2021-08-02 Tom de Vries <tdevries@suse.de>
96783 [gdb/testsuite] Fix gdb.dwarf2/dw2-using-debug-str.exp with cc-with-gdb-index
96784 When running with target board cc-with-gdb-index, we run into:
96786 (gdb) file dw2-using-debug-str-no-debug-str^M
96787 Reading symbols from dw2-using-debug-str-no-debug-str...^M
96788 Dwarf Error: DW_FORM_strp used without required section^M
96789 (gdb) FAIL: gdb.dwarf2/dw2-using-debug-str.exp: file dw2-using-debug-str
96792 The test expects the dwarf error, but has no matching pattern for the entire
96795 Fix this by updating the regexp.
96797 Tested on x86_64-linux.
96799 gdb/testsuite/ChangeLog:
96801 2021-08-02 Tom de Vries <tdevries@suse.de>
96803 * gdb.dwarf2/dw2-using-debug-str.exp: Update regexp to match
96804 cc-with-gdb-index output.
96806 2021-08-02 Tom de Vries <tdevries@suse.de>
96808 [gdb/testsuite] Fix gdb.dwarf2/per-bfd-sharing.exp with cc-with-gdb-index
96809 When running with target board cc-with-gdb-index, we run into:
96811 rm: cannot remove '/tmp/tmp.JmYTeiuFjj/*.gdb-index': \
96812 No such file or directory^M
96813 FAIL: gdb.dwarf2/per-bfd-sharing.exp: \
96814 couldn't remove files in temporary cache dir
96817 Fix this, as in gdb.base/index-cache.exp, by only FAILing when
96818 $expecting_index_cache_use.
96820 Tested on x86_64-linux.
96822 gdb/testsuite/ChangeLog:
96824 2021-08-02 Tom de Vries <tdevries@suse.de>
96826 * gdb.dwarf2/per-bfd-sharing.exp: Only expect index-cache files
96827 when $expecting_index_cache_use.
96829 2021-08-02 Tom de Vries <tdevries@suse.de>
96831 [gdb/testsuite] Fix gdb.dwarf2/gdb-index-nodebug.exp with cc-with-gdb-index
96832 When running with target board cc-with-gdb-index, we run into:
96834 (gdb) save gdb-index .^M
96835 Error while writing index for `gdb-index-nodebug': \
96836 Cannot use an index to create the index^M
96837 (gdb) FAIL: gdb.dwarf2/gdb-index-nodebug.exp: try to save gdb index
96840 Fix this by detecting an already present index, and marking the test
96843 Tested on x86_64-linux.
96845 gdb/testsuite/ChangeLog:
96847 2021-08-02 Tom de Vries <tdevries@suse.de>
96849 * gdb.dwarf2/gdb-index-nodebug.exp: Mark unsupported when index
96852 2021-08-02 Tom de Vries <tdevries@suse.de>
96854 [gdb/testsuite] Fix gdb.dwarf2/fission-relative-dwo.exp with cc-with-gdb-index
96855 When running with target board cc-with-gdb-index, we run into:
96857 gdb compile failed, warning: Could not find DWO CU \
96858 fission-relative-dwo.dwo(0x1234) referenced by CU at offset 0xc7 \
96859 [in module outputs/gdb.dwarf2/fission-relative-dwo/.tmp/fission-relative-dwo]
96860 UNTESTED: gdb.dwarf2/fission-relative-dwo.exp: fission-relative-dwo.exp
96861 ERROR: failed to compile fission-relative-dwo
96864 The problem is that:
96865 - the .dwo file is found relative to the executable, and
96866 - cc-with-tweaks.sh moves the executable to a temp dir, but not
96869 Fix this by copying the .dwo file alongside the executable in the temp dir.
96871 Verified changes using shellcheck.
96873 Tested on x86_64-linux.
96877 2021-08-02 Tom de Vries <tdevries@suse.de>
96879 * contrib/cc-with-tweaks.sh: Copy .dwo files alongside executable.
96881 2021-08-02 Shahab Vahedi <shahab@synopsys.com>
96883 gdb: Make the builtin "boolean" type an unsigned type
96884 When printing the fields of a register that is of a custom struct type,
96885 the "unpack_bits_as_long ()" function is used:
96888 cp_print_value_fields (...)
96889 value_field_bitfield (...)
96890 unpack_value_bitfield (...)
96891 unpack_bits_as_long (...)
96893 This function may sign-extend the extracted field while returning it:
96899 valmask = (((ULONGEST) 1) << bitsize) - 1;
96901 if (!field_type->is_unsigned ())
96902 if (val & (valmask ^ (valmask >> 1)))
96908 lsbcount: Number of lower bits to get rid of.
96909 bitsize: The bit length of the field to be extracted.
96910 val: The register value.
96911 field_type: The type of field that is being handled.
96913 While the logic here is correct, there is a problem when it is
96914 handling "field_type"s of "boolean". Those types are NOT marked
96915 as "unsigned" and therefore they end up being sign extended.
96916 Although this is not a problem for "false" (0), it definitely
96917 causes trouble for "true".
96919 This patch constructs the builtin boolean type as such that it is
96920 marked as an "unsigned" entity.
96922 The issue tackled here was first encountered for arc-elf32 target
96923 running on an x86_64 machine. The unit-test introduced in this change
96924 has passed for all the targets (--enable-targets=all) running on the
96927 Fixes: https://sourceware.org/PR28104
96929 2021-08-02 GDB Administrator <gdbadmin@sourceware.org>
96931 Automatic date update in version.in
96933 2021-08-01 Tom de Vries <tdevries@suse.de>
96935 [gdb/testsuite] Fix gdb.base/maint.exp with cc-with-gdb-index
96936 With target board cc-with-gdb-index we run into:
96938 FAIL: gdb.base/maint.exp: maint print statistics
96941 The output that is checked is:
96943 Statistics for 'maint':^M
96944 Number of "minimal" symbols read: 53^M
96945 Number of "full" symbols read: 40^M
96946 Number of "types" defined: 60^M
96947 Number of symbol tables: 7^M
96948 Number of symbol tables with line tables: 2^M
96949 Number of symbol tables with blockvectors: 2^M
96950 Number of read CUs: 2^M
96951 Number of unread CUs: 5^M
96952 Total memory used for objfile obstack: 20320^M
96953 Total memory used for BFD obstack: 4064^M
96954 Total memory used for string cache: 4064^M
96956 and the regexp doesn't match because it expects the "Number of read/unread
96957 CUs" lines in a different place.
96959 Fix this by updating the regexp.
96961 Tested on x86_64-linux.
96963 gdb/testsuite/ChangeLog:
96965 2021-08-01 Tom de Vries <tdevries@suse.de>
96967 * gdb.base/maint.exp: Update "maint print statistics" to match
96968 output with target board cc-with-gdb-index.
96970 2021-08-01 Tom de Vries <tdevries@suse.de>
96972 [gdb/testsuite] Fix gdb.base/index-cache.exp with cc-with-gdb-index
96973 With target board cc-with-gdb-index we run into:
96975 FAIL: gdb.base/index-cache.exp: couldn't remove files in temporary cache dir
96978 The problem is that there are no files to remove, because the index cache
96979 isn't used, as indicated by $expecting_index_cache_use.
96981 Fix this by only FAILing when $expecting_index_cache_use.
96983 Tested on x86_64-linux.
96985 gdb/testsuite/ChangeLog:
96987 2021-08-01 Tom de Vries <tdevries@suse.de>
96989 * gdb.base/index-cache.exp:
96991 2021-08-01 GDB Administrator <gdbadmin@sourceware.org>
96993 Automatic date update in version.in
96995 2021-07-31 GDB Administrator <gdbadmin@sourceware.org>
96997 Automatic date update in version.in
96999 2021-07-30 Tom Tromey <tom@tromey.com>
97001 Use iterator_range in more places
97002 This changes a couple of spots to replace custom iterator range
97003 classes with a specialization of iterator_range.
97005 Regression tested on x86-64 Fedora 34.
97007 2021-07-30 Tom Tromey <tom@tromey.com>
97009 Replace exception_print_same with operator!=
97010 I noticed that exception_print_same is only used in a single spot, and
97011 it seemed to be better as an operator!= method attached to
97014 Regression tested on x86-64 Fedora 34.
97016 2021-07-30 Tom de Vries <tdevries@suse.de>
97018 [gdb/build] Disable attribute nonnull
97019 With trunk gcc (12.0) we're running into a -Werror=nonnull-compare build
97020 breaker in gdb, which caused a broader review of the usage of the nonnull
97023 The current conclusion is that it's best to disable this. This is explained
97024 at length in the gdbsupport/common-defs.h comment.
97026 Tested by building with trunk gcc.
97030 2021-07-29 Tom de Vries <tdevries@suse.de>
97032 * gdbsupport/common-defs.h (ATTRIBUTE_NONNULL): Disable.
97034 2021-07-30 Clément Chigot <clement.chigot@atos.net>
97036 gas: ensure XCOFF DWARF subsection are initialized to 0
97037 debug_abbrev doesn't use end_exp to compute its size. However, it must
97038 be NULL. Otherwise, ppc_xcoff_end might try to access uninitialized
97042 * config/tc-ppc.c (ppc_dwsect): Use XCNEW instead of XNEW when creating
97045 2021-07-30 Clément Chigot <clement.chigot@atos.net>
97047 bfd: ensure that symbols targeted by DWARF relocations are kept in XCOFF
97048 This patch improves XCOFF garbage collector pass, in order to keep
97049 symbols being referenced only by special sections like DWARF sections.
97052 * xcofflink.c (xcoff_mark): Replace SEC_MARK by gc_mark.
97053 Look through relocations even if xcoff_section_data is NULL.
97054 (xcoff_sweep): Check if any sections of a file is kept before
97055 adding its special sections.
97056 Call xcoff_mark for special sessions being kept instead of just
97059 (xcoff_mark_symbol): Replace SEC_MARK by gc_mark.
97060 (xcoff_keep_symbol_p): Likewise.
97061 (bfd_xcoff_size_dynamic_sections): Likewise.
97062 (xcoff_find_tc0): Likewise.
97064 2021-07-30 Clément Chigot <clement.chigot@atos.net>
97066 bfd: avoid a crash when debug_section isn't created in XCOFF
97068 * xcofflink.c (bfd_xcoff_size_dynamic_sections):
97069 Add check to know if debug_section is initialized.
97071 2021-07-30 Alan Modra <amodra@gmail.com>
97073 readelf: catch archive_file_size of -1
97074 Fuzzers might put -1 in arhdr.ar_size. If the size is rounded up to
97075 and even number of bytes we get zero.
97077 * readelf.c (process_archive): Don't round up archive_file_size.
97078 Do round up next_arhdr_offset calculation.
97080 2021-07-30 Alan Modra <amodra@gmail.com>
97082 reloc_upper_bound size calculations
97083 Section reloc_count is an unsigned int. Adding one for a NULL
97084 terminator to an array of arelent pointers can wrap the count to
97085 zero. Avoid that by doing the addition as longs.
97087 * coffgen.c (coff_get_reloc_upper_bound): Don't overflow unsigned
97089 * elf.c (_bfd_elf_get_reloc_upper_bound): Likewise.
97090 * elf64-sparc.c (elf64_sparc_get_reloc_upper_bound): Likewise.
97091 * mach-o.c (bfd_mach_o_get_reloc_upper_bound): Likewise.
97092 * vms-alpha.c (alpha_vms_get_reloc_upper_bound): Likewise.
97094 2021-07-30 Alan Modra <amodra@gmail.com>
97096 Sanity check _bfd_coff_read_string_table
97097 * coffgen.c (_bfd_coff_read_string_table): Catch overflows
97098 when calculating string table file location.
97100 2021-07-30 Alan Modra <amodra@gmail.com>
97102 IMAGE_SCN_LNK_NRELOC_OVFL
97103 From microsoft docs: It is an error if IMAGE_SCN_LNK_NRELOC_OVFL is
97104 set and there are fewer than 0xffff relocations in the section.
97106 * coffcode.h (coff_set_alignment_hook): Sanity check overflow
97109 2021-07-30 Simon Marchi <simon.marchi@polymtl.ca>
97111 gdb: fix nr_bits gdb_assert in append_flags_type_field
97114 gdb_assert (nr_bits >= 1 && nr_bits <= type_bitsize);
97116 is not correct. Well, it's correct in that we do want the number of
97117 bits to be in the range [1, type_bitsize]. But we don't check anywhere
97118 that the end of the specified flag is within the containing type.
97120 The following code should generate a failed assertion, as the flag goes
97121 past the 32 bits of the underlying type, but it's currently not caught:
97124 test_print_flag (gdbarch *arch)
97126 type *flags_type = arch_flags_type (arch, "test_type", 32);
97127 type *field_type = builtin_type (arch)->builtin_uint32;
97128 append_flags_type_field (flags_type, 31, 2, field_type, "invalid");
97131 (You can test this by registering it as a selftest using
97132 selftests::register_test_foreach_arc and running.)
97134 Change the assertion to verify that the end bit is within the range of
97135 the underlying type. This implicitly verifies that nr_bits is not
97136 too big as well, so we don't need a separate assertion for that.
97138 Change-Id: I9be79e5fd7a5917bf25b03b598727e6274c892e8
97139 Co-Authored-By: Tony Tye <Tony.Tye@amd.com>
97141 2021-07-30 GDB Administrator <gdbadmin@sourceware.org>
97143 Automatic date update in version.in
97145 2021-07-29 John Baldwin <jhb@FreeBSD.org>
97147 obsd-nat: Report both thread and PID in ::pid_to_str.
97148 This improves the output of info threads when debugging multiple
97149 inferiors (e.g. after a fork with detach_on_fork disabled).
97151 2021-07-29 John Baldwin <jhb@FreeBSD.org>
97153 obsd-nat: Various fixes for fork following.
97154 - Don't use #ifdef's on ptrace ops. obsd-nat.h didn't include
97155 <sys/ptrace.h>, so the virtual methods weren't always overridden
97156 causing the fork following to not work. In addition, the thread and
97157 fork code is intertwined in ::wait and and the lack of #ifdef's
97158 there already assumed both were present. Finally, both of these
97159 ptrace ops have been present in OpenBSD for at least 10 years.
97161 - Move duplicated code to enable PTRACE_FORK event reporting to a
97162 single function and invoke it on new child processes reported via
97165 - Don't return early from PTRACE_FORK handling, but instead reset
97166 wptid to the correct ptid if the child reports its event before the
97167 parent. This allows the ptid fixup code to add thread IDs if the
97168 first event for a process is a PTRACE_FORK event. This also
97169 properly returns ptid's with thread IDs when reporting PTRACE_FORK
97172 - Handle detach_fork by skipping the PT_DETACH.
97174 2021-07-29 John Baldwin <jhb@FreeBSD.org>
97176 obsd-nat: Various fixes to obsd_nat_target::wait.
97177 - Call inf_ptrace_target::wait instead of duplicating the code.
97178 Replace a check for WIFSTOPPED on the returned status from waitpid
97179 by checking for TARGET_WAITKIND_STOPPED in the parsed status as is
97180 done in fbsd_nat_target::wait.
97182 - Don't use inferior_ptid when deciding if a new process is a child vs
97183 parent of the fork. Instead, use find_inferior_pid and assume that
97184 if an inferior already exists, the pid in question is the parent;
97185 otherwise, the pid is the child.
97187 - Don't use inferior_ptid when deciding if the ptid of the process
97188 needs to be updated with an LWP ID, or if this is a new thread.
97189 Instead, use the approach from fbsd-nat which is to check if a ptid
97190 without an LWP exists and if so update the ptid of that thread
97191 instead of adding a new thread.
97193 2021-07-29 John Baldwin <jhb@FreeBSD.org>
97195 x86-bsd-nat: Only define gdb_ptrace when using debug registers.
97196 This fixes an unused function warning on OpenBSD which does not
97197 support PT_GETDBREGS.
97199 2021-07-29 John Baldwin <jhb@FreeBSD.org>
97201 Don't compile x86 debug register support on OpenBSD.
97202 Simon Marchi tried gdb on OpenBSD, and it immediately segfaults when
97203 running a program. Simon tracked down the problem to x86_dr_low.get_status
97204 being nullptr at this point:
97206 (lldb) print x86_dr_low.get_status
97207 (unsigned long (*)()) $0 = 0x0000000000000000
97209 * thread #1, stop reason = step over
97210 * frame #0: 0x0000033b64b764aa gdb`x86_dr_stopped_data_address(state=0x0000033d7162a310, addr_p=0x00007f7ffffc5688) at x86-dregs.c:645:12
97211 frame #1: 0x0000033b64b766de gdb`x86_dr_stopped_by_watchpoint(state=0x0000033d7162a310) at x86-dregs.c:687:10
97212 frame #2: 0x0000033b64ea5f72 gdb`x86_stopped_by_watchpoint() at x86-nat.c:206:10
97213 frame #3: 0x0000033b64637fbb gdb`x86_nat_target<obsd_nat_target>::stopped_by_watchpoint(this=0x0000033b65252820) at x86-nat.h:100:12
97214 frame #4: 0x0000033b64d3ff11 gdb`target_stopped_by_watchpoint() at target.c:468:46
97215 frame #5: 0x0000033b6469b001 gdb`watchpoints_triggered(ws=0x00007f7ffffc61c8) at breakpoint.c:4790:32
97216 frame #6: 0x0000033b64a8bb8b gdb`handle_signal_stop(ecs=0x00007f7ffffc61a0) at infrun.c:6072:29
97217 frame #7: 0x0000033b64a7e3a7 gdb`handle_inferior_event(ecs=0x00007f7ffffc61a0) at infrun.c:5694:7
97218 frame #8: 0x0000033b64a7c1a0 gdb`fetch_inferior_event() at infrun.c:4090:5
97219 frame #9: 0x0000033b64a51921 gdb`inferior_event_handler(event_type=INF_REG_EVENT) at inf-loop.c:41:7
97220 frame #10: 0x0000033b64a827c9 gdb`infrun_async_inferior_event_handler(data=0x0000000000000000) at infrun.c:9384:3
97221 frame #11: 0x0000033b6465bd4f gdb`check_async_event_handlers() at async-event.c:335:4
97222 frame #12: 0x0000033b65070917 gdb`gdb_do_one_event() at event-loop.cc:216:10
97223 frame #13: 0x0000033b64af0db1 gdb`start_event_loop() at main.c:421:13
97224 frame #14: 0x0000033b64aefe9a gdb`captured_command_loop() at main.c:481:3
97225 frame #15: 0x0000033b64aed5c2 gdb`captured_main(data=0x00007f7ffffc6470) at main.c:1353:4
97226 frame #16: 0x0000033b64aed4f2 gdb`gdb_main(args=0x00007f7ffffc6470) at main.c:1368:7
97227 frame #17: 0x0000033b6459d787 gdb`main(argc=5, argv=0x00007f7ffffc6518) at gdb.c:32:10
97228 frame #18: 0x0000033b6459d521 gdb`___start + 321
97230 On BSDs, get_status is set in _initialize_x86_bsd_nat, but only if
97231 HAVE_PT_GETDBREGS is defined. PT_GETDBREGS doesn't exist on OpenBSD, so
97232 get_status (and the other fields of x86_dr_low) are left as nullptr.
97234 OpenBSD doesn't support getting or setting the x86 debug registers, so
97235 fix by omitting debug register support entirely on OpenBSD:
97237 - Change x86bsd_nat_target to only inherit from x86_nat_target if
97238 PT_GETDBREGS is supported.
97240 - Don't include x86-nat.o and nat/x86-dregs.o for OpenBSD/amd64. They
97241 were already omitted for OpenBSD/i386.
97243 2021-07-29 Carl Love <cel@us.ibm.com>
97245 Fix for gdb.tui/tui-layout-asm.exp
97246 The width of the window is too narrow to display the entire assembly line.
97247 The width of the columns in the window changes as the test walks thru the
97248 terminal window output. The column change results in the first and second
97249 reads of the same line to differ thus causing the test to fail. Increasing
97250 the width of the window keeps the column width consistent thru the test.
97252 If the test fails, the added check prints an message to the log file if
97253 the failure may be due to the window being too narrow.
97255 gdb/testsuite/ChangeLog
97257 * gdb.tui/tui-layout-asm.exp: Replace window width of 80 with the
97258 tui_asm_window_width variable for the width. Add if
97259 count_whitespace check.
97260 (count_whitespace): New proc
97262 2021-07-29 George Barrett <bob@bob131.so>
97264 guile/scm-math: indentation fixes
97265 Changes the indenting of a few expressions in
97266 vlscm_convert_typed_number to be better in line with the prevailing
97271 2021-07-30 George Barrett <bob@bob131.so>
97273 * guile/scm-math.c (vlscm_convert_typed_number): Fix the
97274 indentation of calls to gdbscm_make_out_of_range_error.
97276 Change-Id: I7463998b77c17a00e88058e89b52fa029ee40e03
97278 2021-07-29 George Barrett <bob@bob131.so>
97280 guile: fix make-value with pointer type
97281 Calling the `make-value' procedure with an integer value and a pointer
97282 type for the #:type argument triggers a failed assertion in
97283 `get_unsigned_type_max', as that function doesn't consider pointers to
97284 be an unsigned type. This commit fixes the issue by adding a separate
97285 code path for pointers.
97287 As previously suggested, range checking is done using a new helper
97288 function in gdbtypes.
97292 2021-07-30 George Barrett <bob@bob131.so>
97294 * gdbtypes.h (get_pointer_type_max): Add declaration.
97295 * gdbtypes.c (get_pointer_type_max): Add definition for new
97297 * guile/scm-math.c (vlscm_convert_typed_number): Add code path
97298 for handling conversions to pointer types without failing an
97301 gdb/testsuite/ChangeLog:
97303 2021-07-30 George Barrett <bob@bob131.so>
97305 * gdb.guile/scm-math.exp (test_value_numeric_ops): Add test
97306 for creating pointers with make-value.
97307 (test_make_pointer_value, test_pointer_numeric_range): Add
97308 test procedures containing checks for integer-to-pointer
97311 Change-Id: I9994dd1c848840a3d995f745e6d72867732049f0
97313 2021-07-29 George Barrett <bob@bob131.so>
97315 gdbtypes: return value from get_unsigned_type_max
97316 Changes the signature of get_unsigned_type_max to return the computed
97317 value rather than returning void and writing the value into a pointer
97318 passed by the caller.
97322 2021-07-30 George Barrett <bob@bob131.so>
97324 * gdbtypes.h (get_unsigned_type_max): Change signature to
97325 return the result instead of accepting a pointer argument in
97326 which to store the result.
97327 * gdbtypes.c (get_unsigned_type_max): Likewise.
97328 * guile/scm-math.c (vlscm_convert_typed_number): Update caller
97329 of get_unsigned_type_max.
97330 (vlscm_integer_fits_p): Likewise.
97332 Change-Id: Ibb1bf0c0fa181fac7853147dfde082a7d1ae2323
97334 2021-07-29 Clément Chigot <clement.chigot@atos.net>
97336 gas: improve C_BSTAT and C_STSYM symbols handling on XCOFF
97337 A C_BSTAT debug symbol specifies the beginning of a static block.
97338 Its n_value is the index of the csect containing static symbols.
97339 A C_STSYM debug symbol represents the stabstring of a statically
97340 allocated symbol. Its n_value is the offset in the csect pointed
97341 by the containing C_BSTAT.
97343 These two special n_value were not correctly handled both when
97344 generating object files with gas or when reading them with objdump.
97345 This patch tries to improve that and, above all, to allow gas-generated
97346 object files with such symbols to be accepted by AIX ld.
97349 * coff-bfd.c (bfd_coff_get_syment): Adjust n_value of symbols
97350 having fix_value = 1 in order to be an index and not a memory
97352 * coffgen.c (coff_get_symbol_info): Likewize.
97353 (coff_print_symbol): Likewize.
97356 * config/tc-ppc.c (ppc_frob_label): Don't change within if
97358 (ppc_stabx): Remove workaround changing exp.X_add_symbol's
97360 * config/tc-ppc.h (struct ppc_tc_sy): Update comments.
97361 * symbols.c (resolve_symbol_value): Remove symbol update
97362 when final_val is 0 and it's an AIX debug symbol.
97363 * testsuite/gas/ppc/aix.exp: Add new tests.
97364 * testsuite/gas/ppc/xcoff-stsym-32.d: New test.
97365 * testsuite/gas/ppc/xcoff-stsym-64.d: New test.
97366 * testsuite/gas/ppc/xcoff-stsym.s: New test.
97368 2021-07-29 George Barrett <bob@bob131.so>
97370 Guile: temporary breakpoints
97371 Adds API to the Guile bindings for creating temporary breakpoints and
97372 querying whether an existing breakpoint object is temporary. This is
97373 effectively a transliteration of the Python implementation.
97375 It's worth noting that the added `is_temporary' flag is ignored in the
97376 watchpoint registration path. This replicates the behaviour of the
97377 Python implementation, but might be a bit surprising for users.
97381 2021-06-09 George Barrett <bob@bob131.so>
97383 * guile/scm-breakpoint.c (gdbscm_breakpoint_object::spec): Add
97384 is_temporary field.
97385 (temporary_keyword): Add keyword object for make-breakpoint
97387 (gdbscm_make_breakpoint): Accept #:temporary keyword argument
97388 and store the value in the allocated object's
97390 (gdbscm_register_breakpoint_x): Pass the breakpoint's
97391 spec.is_temporary value to create_breakpoint.
97392 (gdbscm_breakpoint_temporary): Add breakpoint-temporary?
97393 procedure implementation.
97394 (breakpoint_functions::make-breakpoint): Update documentation
97395 string and fix a typo.
97396 (breakpoint_functions::breakpoint-temporary?): Add
97397 breakpoint-temporary? procedure.
97398 (gdbscm_initialize_breakpoints): Initialise temporary_keyword
97400 NEWS (Guile API): Mention new temporary breakpoints API.
97404 2021-06-09 George Barrett <bob@bob131.so>
97406 * guile.texi (Breakpoints In Guile): Update make-breakpoint
97407 documentation to reflect new #:temporary argument.
97408 Add documentation for new breakpoint-temporary? procedure.
97410 gdb/testsuite/ChangeLog:
97412 2021-06-09 George Barrett <bob@bob131.so>
97414 * gdb.guile/scm-breakpoint.exp: Add additional tests for
97415 temporary breakpoints.
97417 Change-Id: I2de332ee7c256f5591d7141ab3ad50d31b871d17
97419 2021-07-29 GDB Administrator <gdbadmin@sourceware.org>
97421 Automatic date update in version.in
97423 2021-07-28 Simon Marchi <simon.marchi@polymtl.ca>
97425 gdb: clean up some things in features/Makefile
97426 Clean up some things I noticed:
97428 - we generate a regformats/microblaze-with-stack-protect.dat file. I
97429 don't think this is used. It could be used by a GDBserver built for
97430 Microblaze, but GDBserver isn't ported to Microblaze. So I don't
97431 think that's used at all. Remove the entry in features/Makefile and
97434 - There are a bunch of *-expedite values in features/Makefile for
97435 architectures for which we don't generate dat files. AFAIK, these
97436 *-expedite values are only used when generating dat files. Remove
97437 those that are not necessary.
97439 - 32bit-segments.xml is not listed in the Makfile, but it's used. This
97440 means that it wouldn't get re-generated if we were to change how C
97441 files are generated from the XML. It looks like it was simply
97444 Change-Id: I112d00db317102270e1df924473c37122ccb6c3a
97446 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
97448 x86: Simplify check for distinct TMM register operands
97449 If any pair of operands in AMX instructions with 3 TMM register operands
97450 are the same, the instruction will UD. Don't call register_number to
97451 check for distinct TMM register operands since all TMM register operands
97452 have the same size.
97454 * config/tc-i386.c (check_VecOperands): Remove register_number
97455 call when checking for distinct TMM register operands.
97457 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
97459 ld: Run tmpdir/pr28138 only for native build
97461 * testsuite/ld-plugin/lto.exp: Run tmpdir/pr28138 only for
97464 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
97466 bfd: Close the file descriptor if there is no archive fd
97467 Close the file descriptor if there is no archive plugin file descriptor
97468 to avoid running out of file descriptors on thin archives with many
97474 * plugin.c (bfd_plugin_close_file_descriptor): Close the file
97475 descriptor there is no archive plugin file descriptor.
97480 * testsuite/ld-plugin/lto.exp: Run ld/28138 tests.
97481 * testsuite/ld-plugin/pr28138.c: New file.
97482 * testsuite/ld-plugin/pr28138-1.c: Likewise.
97483 * testsuite/ld-plugin/pr28138-2.c: Likewise.
97484 * testsuite/ld-plugin/pr28138-3.c: Likewise.
97485 * testsuite/ld-plugin/pr28138-4.c: Likewise.
97486 * testsuite/ld-plugin/pr28138-5.c: Likewise.
97487 * testsuite/ld-plugin/pr28138-6.c: Likewise.
97488 * testsuite/ld-plugin/pr28138-7.c: Likewise.
97490 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
97492 ld: Report error reason when a library cannot be found
97493 With "ulimit -n 20", report:
97495 ld: cannot find -lgcc: Too many open files
97499 ld: cannot find -lgcc
97501 * ldfile.c (ldfile_open_file): Rport error reason when a library
97504 2021-07-28 Sergei Trofimovich <siarheit@google.com>
97506 texi2pod.pl: add no-op --no-split option support [PR28144]
97507 Change 2faf902da ("generate single html manual page by default")
97508 added use of --no-split option to makeinfo. binutils reuses
97509 makeinfo options for texi2pod.pl wrapper. Unsupported option
97510 led to silent manpage truncation.
97512 The change adds no-op option support.
97516 * texi2pod.pl: Handle no-op --no-split option.
97518 2021-07-28 Andrew Burgess <andrew.burgess@embecosm.com>
97520 gdb: fix missing space in some info variables output
97521 Fixes PR gdb/28121. When a user declares an array like this:
97523 int * const foo_1[3];
97525 And in GDB the user does this:
97527 (gdb) info variables foo
97528 All variables matching regular expression "foo":
97531 1: int * constfoo_1[3];
97533 Notice the missing space between 'const' and 'foo_1'. This is fixed
97534 in c_type_print_varspec_prefix (c-typeprint.c) by passing through the
97535 flag that indicates if a trailing space is needed, rather than hard
97536 coding the flag to false as we currently do.
97538 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28121
97540 2021-07-28 Tom de Vries <tdevries@suse.de>
97542 [gdb/symtab] Fix unhandled dwarf expression opcode with gcc-11 -gdwarf-5
97543 [ I've confused things by forgetting to add -gdwarf-4 in $subject of
97544 commit 0057a7ee0d9 "[gdb/testsuite] Add KFAILs for gdb.ada FAILs with
97545 gcc-11". So I'm adding here -gdwarf-5 in $subject, even though -gdwarf-5 is
97546 the default for gcc-11. I keep getting confused because of working with a
97547 system gcc-11 compiler that was patched to switch the default back to
97550 When running test-case gdb.ada/arrayptr.exp with gcc-11 (and default
97551 -gdwarf-5), I run into:
97553 (gdb) print pa_ptr.all^M
97554 Unhandled dwarf expression opcode 0xff^M
97555 (gdb) FAIL: gdb.ada/arrayptr.exp: scenario=all: print pa_ptr.all
97558 What happens is that pa_ptr:
97560 <2><1523>: Abbrev Number: 3 (DW_TAG_variable)
97561 <1524> DW_AT_name : pa_ptr
97562 <1529> DW_AT_type : <0x14fa>
97566 <2><14fa>: Abbrev Number: 2 (DW_TAG_typedef)
97567 <14fb> DW_AT_name : foo__packed_array_ptr
97568 <1500> DW_AT_type : <0x1504>
97569 <2><1504>: Abbrev Number: 4 (DW_TAG_pointer_type)
97570 <1505> DW_AT_byte_size : 8
97571 <1505> DW_AT_type : <0x1509>
97573 which is a pointer to a subrange:
97575 <2><1509>: Abbrev Number: 12 (DW_TAG_subrange_type)
97576 <150a> DW_AT_lower_bound : 0
97577 <150b> DW_AT_upper_bound : 0x3fffffffffffffffff
97578 <151b> DW_AT_name : foo__packed_array
97579 <151f> DW_AT_type : <0x15cc>
97580 <1523> DW_AT_artificial : 1
97581 <1><15cc>: Abbrev Number: 5 (DW_TAG_base_type)
97582 <15cd> DW_AT_byte_size : 16
97583 <15ce> DW_AT_encoding : 7 (unsigned)
97584 <15cf> DW_AT_name : long_long_long_unsigned
97585 <15d3> DW_AT_artificial : 1
97587 with upper bound of form DW_FORM_data16.
97589 In gdb/dwarf/attribute.h we have:
97591 /* Return non-zero if ATTR's value falls in the 'constant' class, or
97592 zero otherwise. When this function returns true, you can apply
97593 the constant_value method to it.
97595 DW_FORM_data16 is not considered as constant_value cannot handle
97597 bool form_is_constant () const;
97599 so instead we have attribute::form_is_block (DW_FORM_data16) == true.
97601 Then in attr_to_dynamic_prop for the upper bound, we get a PROC_LOCEXPR
97602 instead of a PROP_CONST and end up trying to evaluate the constant
97603 0x3fffffffffffffffff as if it were a locexpr, which causes the
97604 "Unhandled dwarf expression opcode 0xff".
97606 In contrast, with -gdwarf-4 we have:
97608 <164c> DW_AT_upper_bound : 18 byte block: \
97609 9e 10 ff ff ff ff ff ff ff ff 3f 0 0 0 0 0 0 0 \
97610 (DW_OP_implicit_value 16 byte block: \
97611 ff ff ff ff ff ff ff ff 3f 0 0 0 0 0 0 0 )
97614 Fix the dwarf error by translating the DW_FORM_data16 constant into a
97615 PROC_LOCEXPR, effectively by prepending 0x9e 0x10, such that we have same
97616 result as with -gdwarf-4:
97618 (gdb) print pa_ptr.all^M
97619 That operation is not available on integers of more than 8 bytes.^M
97620 (gdb) KFAIL: gdb.ada/arrayptr.exp: scenario=all: print pa_ptr.all \
97624 Tested on x86_64-linux, with gcc-11 and target board
97625 unix/gdb:debug_flags=-gdwarf-5.
97629 2021-07-25 Tom de Vries <tdevries@suse.de>
97631 * dwarf2/read.c (attr_to_dynamic_prop): Handle DW_FORM_data16.
97633 2021-07-28 will schmidt <will_schmidt@vnet.ibm.com>
97635 Externalize the _bfd_set_gp_value function
97636 This change adds an external-visible wrapper for the _bfd_set_gp_value
97637 function. This is a prerequisite for some gdb patches that better
97638 handle powerpc64le relocations against ".TOC.".
97640 * bfd.c (bfd_set_gp_value): New externally visible wrapper
97641 for _bfd_set_gp_value.
97642 * bfd-in2.h: Regenerate.
97644 2021-07-28 Alan Modra <amodra@gmail.com>
97646 PowerPC: ignore sticky options for .machine
97647 PowerPC gas and objdump for a long time have allowed certain -m/-M
97648 options that extend a base cpu with extra functional units to be
97649 specified before the base cpu. For example, "-maltivec -mpower4" is
97650 the same as "-mpower4 -maltivec". See
97651 https://sourceware.org/pipermail/binutils/2008-January/054935.html
97653 It doesn't make as much sense that .machine keep any of these
97654 "sticky" flags when handling a new base cpu. See gcc PR101393. I
97655 think that instead .machine ought to override the command line.
97656 That's what this patch does. It is still possible to extend cpu
97657 functionality with .machine. For example the following can be
97658 assembled when selecting a basic -mppc on the command line:
97663 Here, ".machine altivec" extends the ".machine power5" so that both
97664 the power5 "frin" instruction and the altivec "lvsr" instruction are
97665 enabled. Swapping the two ".machine" directives would result in
97666 failure to assemble "lvsr".
97668 This change will expose some assembly errors, such as the one in
97669 glibc/sysdeps/powerpc/powerpc64/tst-ucontext-ppc64-vscr.c, a file
97670 compiled with -maltivec but containing
97671 asm volatile (".machine push;\n"
97672 ".machine \"power5\";\n"
97674 "vspltisb %1,-1;\n"
97675 "vpkuwus %0,%0,%1;\n"
97679 : "=v" (v0), "=v" (v1)
97682 It's just wrong to choose power5 for a bunch of altivec instructions
97683 and in fact all of those .machine directives are unnecessary.
97685 * config/tc-ppc.c (ppc_machine): Don't use command line
97688 2021-07-28 GDB Administrator <gdbadmin@sourceware.org>
97690 Automatic date update in version.in
97692 2021-07-27 Tom de Vries <tdevries@suse.de>
97694 [gdb/testsuite] Add xfail for PR gcc/101643
97695 With gcc 8.5.0 I run into:
97698 $2 = (0 => 0 <repeats 25 times>)^M
97699 (gdb) FAIL: gdb.ada/big_packed_array.exp: scenario=minimal: print bad
97701 while with gcc 9.3.1 we have instead:
97704 $2 = (false <repeats 196 times>)^M
97705 (gdb) PASS: gdb.ada/big_packed_array.exp: scenario=minimal: print bad
97708 This is caused by gcc PR, which I've filed at
97709 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101643 "[debug, ada] packed array
97710 not described as packed".
97712 Fix by marking this as XFAIL.
97714 Tested on x86_64-linux.
97718 2021-07-27 Tom de Vries <tdevries@suse.de>
97721 * gdb/testsuite/gdb.ada/big_packed_array.exp: Add xfail.
97723 2021-07-27 Tom de Vries <tdevries@suse.de>
97725 [gdb/testsuite] Add xfail for PR gcc/101633
97726 With gcc 7.5.0, I run into:
97728 (gdb) print objects^M
97729 $1 = ((tag => object, values => ()), (tag => unused))^M
97730 (gdb) FAIL: gdb.ada/array_of_variant.exp: scenario=minimal: print entire array
97732 while with gcc 8.5.0 we have:
97734 (gdb) print objects^M
97735 $1 = ((tag => object, values => (2, 2, 2, 2, 2)), (tag => unused))^M
97736 (gdb) PASS: gdb.ada/array_of_variant.exp: scenario=minimal: print entire array
97739 This is due to a gcc PR, which I've filed at
97740 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101633 "Bug 101633 - [debug]
97741 DW_TAG_subrange_type missing DW_AT_upper_bound".
97743 Fix by marking this and related FAILs as XFAIL.
97745 Tested on x86_64-linux.
97749 2021-07-27 Tom de Vries <tdevries@suse.de>
97752 * gdb/testsuite/gdb.ada/array_of_variant.exp: Add xfails.
97754 2021-07-27 Andrew Burgess <andrew.burgess@embecosm.com>
97756 gdb: remove VALUE_FRAME_ID and fix another frame debug issue
97757 This commit was originally part of this patch series:
97759 (v1): https://sourceware.org/pipermail/gdb-patches/2021-May/179357.html
97760 (v2): https://sourceware.org/pipermail/gdb-patches/2021-June/180208.html
97761 (v3): https://sourceware.org/pipermail/gdb-patches/2021-July/181028.html
97763 However, that series is being held up in review, so I wanted to break
97764 out some of the non-related fixes in order to get these merged.
97766 This commit addresses two semi-related issues, both of which are
97767 problems exposed by using 'set debug frame on'.
97769 The first issue is in frame.c in get_prev_frame_always_1, and was
97770 introduced by this commit:
97772 commit a05a883fbaba69d0f80806e46a9457727fcbe74c
97773 Date: Tue Jun 29 12:03:50 2021 -0400
97775 gdb: introduce frame_debug_printf
97777 This commit replaced fprint_frame with frame_info::to_string.
97778 However, the former could handle taking a nullptr while the later, a
97779 member function, obviously requires a non-nullptr in order to make the
97780 function call. In one place we are not-guaranteed to have a
97781 non-nullptr, and so, there is the possibility of triggering undefined
97784 The second issue addressed in this commit has existed for a while in
97785 GDB, and would cause this assertion:
97787 gdb/frame.c:622: internal-error: frame_id get_frame_id(frame_info*): Assertion `fi->this_id.p != frame_id_status::COMPUTING' failed.
97789 We attempt to get the frame_id for a frame while we are computing the
97790 frame_id for that same frame.
97792 What happens is that when GDB stops we create a frame_info object for
97793 the sentinel frame (frame #-1) and then we attempt to unwind this
97794 frame to create a frame_info object for frame #0.
97796 In the test case used here to expose the issue we have created a
97797 Python frame unwinder. In the Python unwinder we attemt to read the
97798 program counter register.
97800 Reading this register will initially create a lazy register value.
97801 The frame-id stored in the lazy register value will be for the
97802 sentinel frame (lazy register values hold the frame-id for the frame
97803 from which the register will be unwound).
97805 However, the Python unwinder does actually want to examine the value
97806 of the program counter, and so the lazy register value is resolved
97807 into a non-lazy value. This sends GDB into value_fetch_lazy_register
97810 Now, inside this function, if 'set debug frame on' is in effect, then
97811 we want to print something like:
97813 frame=%d, regnum=%d(%s), ....
97815 Where 'frame=%d' will be the relative frame level of the frame for
97816 which the register is being fetched, so, in this case we would expect
97817 to see 'frame=0', i.e. we are reading a register as it would be in
97818 frame #0. But, remember, the lazy register value actually holds the
97819 frame-id for frame #-1 (the sentinel frame).
97821 So, to get the frame_info for frame #0 we used to call:
97823 frame = frame_find_by_id (VALUE_FRAME_ID (val));
97825 Where VALUE_FRAME_ID is:
97827 #define VALUE_FRAME_ID(val) (get_prev_frame_id_by_id (VALUE_NEXT_FRAME_ID (val)))
97829 That is, we start with the frame-id for the next frame as obtained by
97830 VALUE_NEXT_FRAME_ID, then call get_prev_frame_id_by_id to get the
97831 frame-id of the previous frame.
97833 The get_prev_frame_id_by_id function finds the frame_info for the
97834 given frame-id (in this case frame #-1), calls get_prev_frame to get
97835 the previous frame, and then calls get_frame_id.
97837 The problem here is that calling get_frame_id requires that we know
97838 the frame unwinder, so then have to try each frame unwinder in turn,
97839 which would include the Python unwinder.... which is where we started,
97840 and thus we have a loop!
97842 To prevent this loop GDB has an assertion in place, which is what
97845 Solving the assertion failure is pretty easy, if we consider the code
97846 in value_fetch_lazy_register and get_prev_frame_id_by_id then what we
97849 1. Start with a frame_id taken from a value,
97850 2. Lookup the corresponding frame,
97851 3. Find the previous frame,
97852 4. Get the frame_id for that frame, and
97853 5. Lookup the corresponding frame
97854 6. Print the frame's level
97856 Notice that steps 3 and 5 give us the exact same result, step 4 is
97857 just wasted effort. We could shorten this process such that we drop
97858 steps 4 and 5, thus:
97860 1. Start with a frame_id taken from a value,
97861 2. Lookup the corresponding frame,
97862 3. Find the previous frame,
97863 6. Print the frame's level
97865 This will give the exact same frame as a result, and this is what I
97866 have done in this patch by removing the use of VALUE_FRAME_ID from
97867 value_fetch_lazy_register.
97869 Out of curiosity I looked to see how widely VALUE_FRAME_ID was used,
97870 and saw it was only used in one other place in valops.c:value_assign,
97871 where, once again, we take the result of VALUE_FRAME_ID and pass it to
97872 frame_find_by_id, thus introducing a redundant frame_id lookup.
97874 I don't think the value_assign case risks triggering the assertion
97875 though, as we are unlikely to call value_assign while computing the
97876 frame_id for a frame, however, we could make value_assign slightly
97877 more efficient, with no real additional complexity, by removing the
97878 use of VALUE_FRAME_ID.
97880 So, in this commit, I completely remove VALUE_FRAME_ID, and replace it
97881 with a use of VALUE_NEXT_FRAME_ID, followed by a direct call to
97882 get_prev_frame_always, this should make no difference in either case,
97883 and resolves the assertion issue from value.c.
97885 As I said, this patch was originally part of another series, the
97886 original test relied on the fixes in that original series. However, I
97887 was able to create an alternative test for this issue by enabling
97888 frame debug within an existing test script.
97890 This commit probably fixes bug PR gdb/27938, though the bug doesn't
97891 have a reproducer attached so it is not possible to know for sure.
97893 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=27938
97895 2021-07-27 Chenghua Xu <xuchenghua@loongson.cn>
97897 Correct gs264e bfd_mach in mips_arch_choices.
97899 * mips-dis.c (mips_arch_choices): Correct gs264e bfd_mach.
97901 2021-07-27 Roland McGrath <mcgrathr@google.com>
97903 Fix ld test case that assumes --enable-textrel-check
97905 * testsuite/ld-x86-64/x86-64.exp (Build textrel-1): Use --warn-textrel.
97907 2021-07-27 GDB Administrator <gdbadmin@sourceware.org>
97909 Automatic date update in version.in
97911 2021-07-27 H.J. Lu <hjl.tools@gmail.com>
97913 bfd: Set error to bfd_error_malformed_archive only if unset
97914 When reading an archive member, set error to bfd_error_malformed_archive
97915 on open_nested_file failure only if the error is unset.
97918 * archive.c (_bfd_get_elt_at_filepos): Don't set error to
97919 bfd_error_malformed_archive if it has been set.
97921 2021-07-26 Carl Love <cel@us.ibm.com>
97923 Fix for mi-reverse.exp
97924 This test fails on PPC64 because PPC64 prints the value of 3.5 with
97925 more significant digits than on Intel. The patch updates the regular
97926 expression to allow for more significant digits on the constant.
97928 gdb/testsuite/ChangeLog
97930 * gdb.mi/mi-reverse.exp: mi_execute_to exec-step reverse add check
97931 for additional digits.
97933 2021-07-26 Tom Tromey <tromey@adacore.com>
97935 Fix the Windows build
97936 The gdb build was broken on Windows after the patch to change
97937 get_inferior_cwd. This patch fixes the build.
97939 2021-07-26 Shahab Vahedi <shahab@synopsys.com>
97941 gdb: Fix numerical field extraction for target description "flags"
97942 The "val_print_type_code_flags ()" function is responsible for
97943 extraction of fields for "flags" data type. These data types are
97944 used when describing a custom register type in a target description
97945 XML. The logic used for the extraction though is not sound:
97947 unsigned field_len = TYPE_FIELD_BITSIZE (type, field);
97949 = val >> (TYPE_FIELD_BITPOS (type, field) - field_len + 1);
97951 TYPE_FIELD_BITSIZE: The bit length of the field to be extracted.
97952 TYPE_FIELD_BITPOS: The starting position of the field; 0 is LSB.
97953 val: The register value.
97955 Imagine you have a field that starts at position 1 and its length
97956 is 4 bits. According to the third line of the code snippet the
97957 shifting right would become "val >> -2", or "val >> 0xfff...fe"
97958 to be precise. That will result in a "field_val" of 0.
97960 The correct extraction should be:
97962 ULONGEST field_val = val >> TYPE_FIELD_BITPOS (type, field);
97964 The rest of the algorithm that masks out the higher bits is OK.
97966 Co-Authored-By: Simon Marchi <simon.marchi@efficios.com>
97968 2021-07-26 Andrea Corallo <andrea.corallo@arm.com>
97970 PATCH [10/10] arm: Alias 'ra_auth_code' to r12 for pacbti.
97972 2021-06-11 Andrea Corallo <andrea.corallo@arm.com>
97974 * config/tc-arm.c (reg_names): Alias 'ra_auth_code' to r12.
97976 2021-07-26 Andrea Corallo <andrea.corallo@arm.com>
97978 PATCH [9/10] arm: add 'pacg' instruction for Armv8.1-M pacbti extension
97980 2021-06-11 Andrea Corallo <andrea.corallo@arm.com>
97982 * config/tc-arm.c (T16_32_TAB): Add '_pacg'.
97983 (do_t_pacbti_pacg): New function.
97984 (insns): Define 'pacg' insn.
97985 * testsuite/gas/arm/armv8_1-m-pacbti.d: Add 'pacg' test.
97986 * testsuite/gas/arm/armv8_1-m-pacbti.s: Likewise.
97989 2021-06-11 Andrea Corallo <andrea.corallo@arm.com>
97991 * arm-dis.c (thumb32_opcodes): Add 'pacg'.
97993 2021-07-26 Andrea Corallo <andrea.corallo@arm.com>
97995 PATCH [8/10] arm: add 'autg' instruction for Armv8.1-M pacbti extension
97997 2021-06-11 Andrea Corallo <andrea.corallo@arm.com>
97999 * config/tc-arm.c (T16_32_TAB): Add '_autg'.
98000 (insns): Define 'autg' insn.
98001 * testsuite/gas/arm/armv8_1-m-pacbti.d: Add autg test.
98002 * testsuite/gas/arm/armv8_1-m-pacbti.s: Likewise.
98005 2021-06-11 Andrea Corallo <andrea.corallo@arm.com>
98007 * arm-dis.c (thumb32_opcodes): Add 'autg'.
98009 2021-07-26 Andrea Corallo <andrea.corallo@arm.com>
98011 PATCH [7/10] arm: add 'bxaut' instruction for Armv8.1-M pacbti extension
98013 2021-06-11 Andrea Corallo <andrea.corallo@arm.com>
98015 * config/tc-arm.c (T16_32_TAB): Add '_bxaut'.
98016 (do_t_pacbti_nonop): New function.
98017 (insns): Define 'bxaut' insn.
98018 * testsuite/gas/arm/armv8_1-m-pacbti.d: Add 'bxaut' test.
98019 * testsuite/gas/arm/armv8_1-m-pacbti.s: Likewise.
98022 2021-06-11 Andrea Corallo <andrea.corallo@arm.com>
98024 * arm-dis.c (thumb32_opcodes): Add 'bxaut'.
98026 2021-07-26 Andrea Corallo <andrea.corallo@arm.com>
98028 PATCH [6/10] arm: Add -march=armv8.1-m.main+pacbti flag
98030 2021-06-11 Andrea Corallo <andrea.corallo@arm.com>
98032 * config/tc-arm.c (pacbti_ext): Define.
98033 (BAD_PACBTI): New macro.
98034 (armv8_1m_main_ext_table): Add 'pacbti' extension.
98037 2021-06-11 Andrea Corallo <andrea.corallo@arm.com>
98039 * opcode/arm.h (ARM_EXT3_PACBTI, ARM_AEXT3_V8_1M_MAIN_PACBTI): New
98042 2021-07-26 Andrea Corallo <andrea.corallo@arm.com>
98044 PATCH [5/10] arm: Extend again arm_feature_set struct to provide more bits
98046 2021-06-11 Andrea Corallo <andrea.corallo@arm.com>
98048 * opcode/arm.h (arm_feature_set): Extend 'core' field.
98049 (ARM_CPU_HAS_FEATURE, ARM_FSET_CPU_SUBSET, ARM_CPU_IS_ANY)
98050 (ARM_MERGE_FEATURE_SETS, ARM_CLEAR_FEATURE, ARM_FEATURE_EQUAL)
98051 (ARM_FEATURE_ZERO, ARM_FEATURE_CORE_EQUAL): Account for
98053 (ARM_FEATURE_CORE_HIGH_HIGH): New macro.
98055 2021-07-26 Andrea Corallo <andrea.corallo@arm.com>
98057 PATCH [4/10] arm: add 'pac' instruction for Armv8.1-M pacbti extension
98059 2021-06-11 Andrea Corallo <andrea.corallo@arm.com>
98061 * config/tc-arm.c (T16_32_TAB): Add '_pac'.
98062 (insns): Add 'pac' insn.
98063 * testsuite/gas/arm/armv8_1-m-pacbti-bad.l: Add pac tests.
98064 * testsuite/gas/arm/armv8_1-m-pacbti-bad.s: Likewise.
98065 * testsuite/gas/arm/armv8_1-m-pacbti.d: Likewise.
98066 * testsuite/gas/arm/armv8_1-m-pacbti.s: Likewise.
98069 2021-06-11 Andrea Corallo <andrea.corallo@arm.com>
98071 * arm-dis.c (thumb32_opcodes): Add 'pac'.
98073 2021-07-26 Andrea Corallo <andrea.corallo@arm.com>
98075 PATCH [3/10] arm: add 'aut' instruction for Armv8.1-M pacbti extension
98077 2021-06-11 Andrea Corallo <andrea.corallo@arm.com>
98079 * config/tc-arm.c (insns): Add 'aut.'
98080 (T16_32_TAB): Add '_aut'.
98081 * testsuite/gas/arm/armv8_1-m-pacbti-bad.l: Add 'aut' tests.
98082 * testsuite/gas/arm/armv8_1-m-pacbti-bad.s: Likewise.
98083 * testsuite/gas/arm/armv8_1-m-pacbti.d: Likewise.
98084 * testsuite/gas/arm/armv8_1-m-pacbti.s: Likewise.
98087 2021-06-11 Andrea Corallo <andrea.corallo@arm.com>
98089 * arm-dis.c (thumb32_opcodes): Add 'aut'.
98091 2021-07-26 Andrea Corallo <andrea.corallo@arm.com>
98093 PATCH [2/10] arm: add 'pacbti' instruction for Armv8.1-M pacbti extension
98095 2021-06-11 Andrea Corallo <andrea.corallo@arm.com>
98098 (enum operand_parse_code): Add OP_SP and OP_R12.
98099 (parse_operands): Add switch cases for OP_SP and OP_R12.
98100 (T16_32_TAB): Add '_pacbti'.
98101 (do_t_pacbti): New function.
98102 (insns): Add 'pacbti'.
98103 * testsuite/gas/arm/armv8_1-m-pacbti-bad.d: New file.
98104 * testsuite/gas/arm/armv8_1-m-pacbti-bad.l: Likewise.
98105 * testsuite/gas/arm/armv8_1-m-pacbti-bad.s: Likewise.
98106 * testsuite/gas/arm/armv8_1-m-pacbti.d: Add 'pacbti' to testcase.
98107 * testsuite/gas/arm/armv8_1-m-pacbti.s: Likewise.
98110 2021-06-11 Andrea Corallo <andrea.corallo@arm.com>
98112 * arm-dis.c (thumb32_opcodes): Add 'pacbti' instruction.
98114 2021-07-26 Andrea Corallo <andrea.corallo@arm.com>
98116 PATCH [1/10] arm: add 'bti' instruction for Armv8.1-M pacbti extension
98118 2021-06-11 Andrea Corallo <andrea.corallo@arm.com>
98120 * config/tc-arm.c (insns): Add 'bti' insn.
98121 * testsuite/gas/arm/armv8_1-m-pacbti.d: New file.
98122 * testsuite/gas/arm/armv8_1-m-pacbti.s: Likewise.
98125 2021-06-11 Andrea Corallo <andrea.corallo@arm.com>
98127 * arm-dis.c (thumb32_opcodes): Add bti instruction.
98129 2021-07-26 Andrew Burgess <andrew.burgess@embecosm.com>
98131 gdb: move remaining ChangeLogs to legacy files
98134 commit f069ea46a03ae868581d1c852da28e979ea1245a
98135 Date: Sat Jul 3 16:29:08 2021 -0700
98137 Rename gdb/ChangeLog to gdb/ChangeLog-2021
98139 The gdb/ChangeLog file was renamed, but all of the other ChangeLog
98140 files relating to gdb were left in place.
98142 As I understand things, the no ChangeLogs policy applies to all the
98143 GDB related directories, so this commit renames all of the remaining
98144 GDB related ChangeLog files.
98146 As with the original commit, the intention behind this commit is to
98147 hopefully stop people merging ChangeLog entries by mistake.
98149 The renames carried out in this commit are:
98151 gdb/doc/ChangeLog -> gdb/doc/ChangeLog-1991-2021
98152 gdb/stubs/ChangeLog -> gdb/stubs/ChangeLog-2012-2020
98153 gdb/testsuite/ChangeLog -> gdb/testsuite/ChangeLog-2014-2021
98154 gdbserver/ChangeLog -> gdbserver/ChangeLog-2002-2021
98155 gdbsupport/ChangeLog -> gdbsupport/ChangeLog-2020-2021
98157 2021-07-26 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
98159 gdb/mi: handle no condition argument case for -break-condition
98160 As reported in PR gdb/28076 [1], passing no condition argument to the
98161 -break-condition command (e.g.: "-break-condition 2") should clear the
98162 condition for breakpoint 2, just like CLI's "condition 2", but instead
98163 an error message is returned:
98165 ^error,msg="-break-condition: Missing the <number> and/or <expr> argument"
98167 The current implementation of the -break-condition command's argument
98168 handling (79aabb7308c "gdb/mi: add a '--force' flag to the
98169 '-break-condition' command") was done according to the documentation,
98170 where the condition argument seemed mandatory. However, the
98171 -break-condition command originally (i.e. before the 79aabb7308c
98172 patch) used the CLI's "cond" command, and back then not passing a
98173 condition argument was clearing out the condition. So, this is a
98174 regression in terms of the behavior.
98176 Fix the argument handling of the -break-condition command to allow not
98177 having a condition argument, and also update the document to make the
98178 behavior clear. Also add test cases to test the scenarios which were
98179 previously not covered.
98181 [1] https://sourceware.org/bugzilla/show_bug.cgi?id=28076
98183 2021-07-26 GDB Administrator <gdbadmin@sourceware.org>
98185 Automatic date update in version.in
98187 2021-07-25 GDB Administrator <gdbadmin@sourceware.org>
98189 Automatic date update in version.in
98191 2021-07-24 Alan Modra <amodra@gmail.com>
98193 Revert: PowerPC: Don't generate unused section symbols
98194 Blindly following x86 broke linux kernel builds.
98197 * elf32-ppc.c (TARGET_KEEP_UNUSED_SECTION_SYMBOLS): Define as true.
98198 * elf64-ppc.c (TARGET_KEEP_UNUSED_SECTION_SYMBOLS): Likewise.
98200 * testsuite/gas/ppc/power4.d: Adjust for section sym change.
98201 * testsuite/gas/ppc/test1elf32.d: Likewise.
98202 * testsuite/gas/ppc/test1elf64.d: Likewise.
98204 * testsuite/ld-powerpc/tlsexe.r: Adjust for section sym change.
98205 * testsuite/ld-powerpc/tlsexe32.r: Likewise.
98206 * testsuite/ld-powerpc/tlsexe32no.r: Likewise.
98207 * testsuite/ld-powerpc/tlsexeno.r: Likewise.
98208 * testsuite/ld-powerpc/tlsexenors.r: Likewise.
98209 * testsuite/ld-powerpc/tlsexers.r: Likewise.
98210 * testsuite/ld-powerpc/tlsexetoc.r: Likewise.
98211 * testsuite/ld-powerpc/tlsexetocrs.r: Likewise.
98212 * testsuite/ld-powerpc/tlsget.d: Likewise.
98213 * testsuite/ld-powerpc/tlsget.wf: Likewise.
98214 * testsuite/ld-powerpc/tlsget2.d: Likewise.
98215 * testsuite/ld-powerpc/tlsget2.wf: Likewise.
98216 * testsuite/ld-powerpc/tlsso.r: Likewise.
98217 * testsuite/ld-powerpc/tlsso32.r: Likewise.
98218 * testsuite/ld-powerpc/tlstocso.r: Likewise.
98220 2021-07-24 Alan Modra <amodra@gmail.com>
98222 Re: ld script expression parsing
98223 Commit 40726f16a8d7 broke references to sections within ADDR(), and
98224 overlays with weird section names.
98226 * ldgram.y (paren_script_name): New rule.
98227 (exp): Use it for ALIGNOF, SIZEOF, ADDR, and LOADADDR. Similarly
98228 ensure script mode parsing for section name in SEGMENT_START.
98229 (overlay_section): Delete unnecessary ldlex_script call. Backup
98230 on a lookahead NAME parsed in expression mode.
98231 * testsuite/ld-elf/overlay.s: Add more sections.
98232 * testsuite/ld-elf/overlay.t: Test '-' in section names.
98234 2021-07-24 Frederic Cambus <fred@statdns.com>
98236 Update the NetBSD system call table to match NetBSD-current.
98237 Generated from sys/sys/syscall.h revision 1.319.
98239 We can safely remove the _lwp_gettid syscall, which was never exposed
98240 in libc and never made it into a release.
98244 2021-07-23 Frederic Cambus <fred@statdns.com>
98246 * syscalls/netbsd.xml: Regenerate.
98248 2021-07-24 GDB Administrator <gdbadmin@sourceware.org>
98250 Automatic date update in version.in
98252 2021-07-23 Simon Marchi <simon.marchi@polymtl.ca>
98254 gdb/testsuite: test get/set value of unregistered Guile parameter
98255 When creating a parameter in Guile, you have to create it using
98256 make-parameter and then register it with GDB with register-parameter!.
98257 In between, it's still possible (though not documented) to set the
98258 parameter's value. I broke this use case by mistake while writing this
98259 series, so thought it would be good to have a test for it.
98261 I suppose that people could use this "feature" to give their parameter
98262 an initial value, even though make-parameter has an initial-value
98263 parameter for this. Nevertheless, changing this behavior could break
98264 some scripts, which is why I think it's important for it to be tested.
98266 Change-Id: I5b2103e3cec0cfdcccf7ffb00eb05fed8626e66d
98268 2021-07-23 Simon Marchi <simon.marchi@polymtl.ca>
98270 gdb: remove cmd_list_element::function::sfunc
98271 I don't understand what the sfunc function type in
98272 cmd_list_element::function is for. Compared to cmd_simple_func_ftype,
98273 it has an extra cmd_list_element parameter, giving the callback access
98274 to the cmd_list_element for the command being invoked. This allows
98275 registering the same callback with many commands, and alter the behavior
98276 using the cmd_list_element's context.
98278 From the comment in cmd_list_element, it sounds like at some point it
98279 was the callback function type for set and show functions, hence the
98280 "s". But nowadays, it's used for many more commands that need to access
98281 the cmd_list_element object (see add_catch_command for example).
98283 I don't really see the point of having sfunc at all, since do_sfunc is
98284 just a trivial shim that changes the order of the arguments. All
98285 commands using sfunc could just as well set cmd_list_element::func to
98286 their callback directly.
98288 Therefore, remove the sfunc field in cmd_list_element and everything
98289 that goes with it. Rename cmd_const_sfunc_ftype to cmd_func_ftype and
98290 use it for cmd_list_element::func, as well as for the add_setshow
98293 Change-Id: I1eb96326c9b511c293c76996cea0ebc51c70fac0
98295 2021-07-23 Simon Marchi <simon.marchi@polymtl.ca>
98297 gdb: rename cfunc to simple_func
98298 After browsing the CLI code for quite a while and trying really hard, I
98299 reached the conclusion that I can't give a meaningful explanation of
98300 what "sfunc" and "cfunc" functions are, in cmd_list_element. I don't
98301 see a logic at all. That makes it very difficult to do any kind of
98302 change. Unless somebody can make sense out of all that, I'd like to try
98303 to retro-fit some logic in the cmd_list_element callback function code
98304 so that we can understand what is going on, do some cleanups and add new
98307 The first change is about "cfunc". I can't figure out what the "c" in
98308 cfunc means. It's not const, because there's already "const" in
98309 "cmd_const_cfunc_ftype", and the previous "cmd_cfunc_ftype" had nothing
98310 const.. It's not "cmd" or "command", because there's already "cmd" in
98311 "cmd_const_cfunc_ftype".
98313 The "main" command callback, cmd_list_element::func, has three
98314 parameters, whereas cfunc has two. It is missing the cmd_list_element
98315 parameter. So the only reason I see for cfunc to exist is to be a shim
98316 between the three and two parameter versions. Most commands don't need
98317 to receive the cmd_list_element object, so adding it everywhere would be
98318 long and would just add more unnecessary boilerplate. So since this is
98319 the "simple" version of the callback, compared to the "full", I suggest
98320 renaming cmd_const_cfunc_ftype into cmd_simple_func_ftype, as well as
98321 everything (like the utility functions) that goes with it.
98323 Change-Id: I4e46cacfd77a66bc1cbf683f6a362072504b7868
98325 2021-07-23 Simon Marchi <simon.marchi@polymtl.ca>
98327 gdb: make inferior::m_terminal an std::string
98328 Same idea as the previous patch, but for m_terminal.
98330 Change-Id: If9367d5db8c976a4336680adca4ea5bc31ab64d2
98332 2021-07-23 Simon Marchi <simon.marchi@polymtl.ca>
98334 gdb: make inferior::m_cwd an std::string
98335 Same idea as the previous patch, but for m_cwd.
98337 To keep things consistent across the board, change get_inferior_cwd as
98338 well, which is shared with GDBserver. So update the related GDBserver
98341 Change-Id: Ia2c047fda738d45f3d18bc999eb67ceb8400ce4e
98343 2021-07-23 Simon Marchi <simon.marchi@polymtl.ca>
98345 gdb: make inferior::m_args an std::string
98346 With the current code, both a NULL pointer and an empty string can mean
98347 "no arguments". We don't need this distinction. Changing to a string
98348 has the advantage that there is now a single state for that (an empty
98349 string), which makes the code a bit simpler in my opinion.
98351 Change-Id: Icdc622820f7869478791dbaa84b4a1c7fec21ced
98353 2021-07-23 Simon Marchi <simon.marchi@polymtl.ca>
98355 gdb: add setter/getter for inferior cwd
98356 Add cwd/set_cwd to the inferior class, remove set_inferior_args.
98357 Keep get_inferior_args, because it is used from fork_inferior, in shared
98358 code. The cwd could eventually be passed as a parameter eventually,
98359 though, I think that would be cleaner.
98361 Change-Id: Ifb72ea865d7e6f9a491308f0d5c1595579d8427e
98363 2021-07-23 Simon Marchi <simon.marchi@polymtl.ca>
98365 gdb: add setter/getter for inferior arguments
98366 Add args/set_args to the inferior class, remove the set_inferior_args
98367 and get_inferior_args functions, that would just be wrappers around
98370 Change-Id: If87d52f3402ce08be26c32897ae8915d9f6d1ea3
98372 2021-07-23 Simon Marchi <simon.marchi@polymtl.ca>
98374 gdb: remove inferior::{argc,argv}
98375 There are currently two states that the inferior args can be stored.
98376 The main one is the `args` field, where they are stored as a single
98377 string. The other one is the `argc`/`argv` fields.
98379 This last one is only used for arguments passed in GDB's
98380 command line. And the only outcome is that when get_inferior_args is
98381 called, `argc`/`argv` are serialized into `args`. So really,
98382 `argc`/`argv` is just a staging area before moving the arguments in
98385 Simplify this by only keeping the `args` field. Change
98386 set_inferior_args_vector to immediately serialize the arguments into
98387 `args`, work that would be done in get_inferior_args later anyway.
98389 The only time where this work would be "wasted" is when the user passes
98390 some arguments on the command line, but does not end up running the
98391 program. But that just seems unlikely. And it's not that much work.
98393 Change-Id: Ica0b9859397c095f6530350c8fb3c36905f2044a
98395 2021-07-23 Simon Marchi <simon.marchi@polymtl.ca>
98397 gdb: un-share set_inferior_cwd declaration
98398 The declaration of set_inferior_cwd is currently shared between gdb and
98399 gdbserver, in gdbsupport/common-inferior.h. It doesn't need to be, as
98400 set_inferior_cwd is not called from common code. Only get_inferior_cwd
98403 The motivation for this is that a future patch will change the prototype
98404 of set_inferior_cwd in gdb, and I don't want to change it for gdbserver
98405 unnecessarily. I see this as a good cleanup in any case, to reduce to
98406 just the essential what is shared between GDB and GDBserver.
98408 Change-Id: I3127d27d078f0503ebf5ccc6fddf14f212426a73
98410 2021-07-23 Simon Marchi <simon.marchi@polymtl.ca>
98412 gdb.base/setshow.exp: fix duplicate test name
98415 DUPLICATE: gdb.base/setshow.exp: test_setshow_args: show args
98417 by giving some explicit test names.
98419 Change-Id: I2a738d3d3675ab9b45929e71f5aee0ea6bf92072
98421 2021-07-23 Simon Marchi <simon.marchi@polymtl.ca>
98423 gdb.base/setshow.exp: split in procs
98424 Split in multiple procs, one per topic, and start with a fresh GDB in
98425 each. I find it easier to work on a test with multiple smaller
98426 independent test procedures. For example, it's possible to comment all
98427 but one when working on one. It's also easier to add things without
98428 having to think about the impact on existing tests, and vice-versa.
98430 Change-Id: I19691eed8f9bcb975b2eeff7577cac66251bcbe2
98432 2021-07-23 Simon Marchi <simon.marchi@polymtl.ca>
98434 gdb.base/setshow.exp: use save_vars to save/restore gdb_prompt
98435 Using save_vars is a bit better than what we have now, as it ensures the
98436 variable gets restored if the code within it throws an error.
98438 Change-Id: I3bd6836e5b7efb61b078acadff1a1c8182c19a27
98440 2021-07-23 Simon Marchi <simon.marchi@polymtl.ca>
98442 gdb/testsuite: split gdb.python/py-parameter.exp in procs
98443 Split the file into multiple independent test procs, where each proc
98444 starts with a fresh GDB. I find it easier to understand what a test is
98445 doing when each part of the test is isolated and self-contained. It
98446 makes it easier to comment out some parts of the test while working /
98447 debugging a specific part. It also makes it easier to add new things
98448 (which a subsequent patch will do) without fear of impacting another part
98451 Change-Id: I8b4d52ac82b1492d79b679e13914ed177d8a836d
98453 2021-07-23 Carl Love <cel@us.ibm.com>
98455 Fix for gdb.python/py-breakpoint.exp
98456 Not all systems have hardware breakpoint support. Add a check
98457 to see if the system supports hardware breakpoints.
98459 gdb/testsuite/ChangeLog
98461 * gdb.python/py-breakpoint.exp (test_hardware_breakpoints): Add
98462 check for hardware breakpoint support.
98464 2021-07-23 Andrew Burgess <andrew.burgess@embecosm.com>
98466 gdb/testsuite: don't error when trying to unset last_spawn_tty_name
98467 In spawn_capture_tty_name (lib/gdb.exp) we either set or unset
98468 last_spawn_tty_name depending on whether spawn_out(slave,name) exists
98471 One situation that might cause spawn_out(slave,name) to not exists is
98472 if the spawn function is called with the argument -leaveopen, which is
98473 how it is called when processes are created as part of a pipeline, the
98474 created process has no tty, instead its output is written to a file
98477 If a pipeline is created consisting of multiple processes then there
98478 will be multiple sequential calls to spawn, all using -leaveopen. The
98479 first of these calls is fine, spawn_out(slave,name) is not set, and so
98480 in spawn_capture_tty_name we unset last_spawn_tty_name. However, on
98481 the second call to spawn, spawn_out(slave,name) is still not set and
98482 so in spawn_capture_tty_name we again try to unset
98483 last_spawn_tty_name, this now throws an error (as last_spawn_tty_name
98486 Fix this issue by using -nocomplain with the call to unset in
98487 spawn_capture_tty_name.
98489 Before this commit I was seeing gdb.base/gnu-debugdata.exp report 1
98490 pass, and 1 unsupported test. After this commit I now see 16 passes
98491 from this test script.
98493 I have also improved the code that used to do this:
98495 if { [info exists spawn_out] } {
98496 set ::last_spawn_tty_name $spawn_out(slave,name)
98501 The problem here is that we check for the existence of spawn_out, and
98502 then unconditionally read spawn_out(slave,name). A situation could
98503 arise where some other element of spawn_out is set,
98504 e.g. spawn_out(foo), in which case we would enter the if block and try
98505 to read a non-existent variable. After this commit we now check
98506 specifically for spawn_out(slave,name).
98508 Finally, it is worth noting that before this issue was fixed runtest
98509 itself, or rather the expect process behind runtest, would segfault
98510 while exiting. I haven't looked at all into what the problem is here
98511 that caused expect to crash, as fixing the bug in GDB's testing
98512 scripts made the segfault go away.
98514 2021-07-23 Jan Beulich <jbeulich@suse.com>
98516 x86: express unduly set rounding control bits in disassembly
98517 While EVEX.L'L are indeed ignored when EVEX.b stands for just SAE,
98518 EVEX.b itself is not ignored when an insn permits neither rounding
98521 While changing this aspect of EVEX.b handling, also alter unduly set
98522 embedded broadcast: Don't call BadOp(), screwing up subsequent
98523 disassembly, but emit "{bad}" instead.
98525 2021-07-23 GDB Administrator <gdbadmin@sourceware.org>
98527 Automatic date update in version.in
98529 2021-07-22 Tom de Vries <tdevries@suse.de>
98531 [gdb/testsuite] Fix FAILs due to PR gcc/101575
98532 When running test-case gdb.ada/formatted_ref.exp with gcc-11 and target board
98533 unix/gdb:debug_flags=-gdwarf-4 we run into:
98536 No definition of "s" in current context.^M
98537 (gdb) FAIL: gdb.ada/formatted_ref.exp: print/x s
98539 which is caused by "runto defs.adb:20" taking us to defs__struct1IP:
98541 (gdb) break defs.adb:20^M
98542 Breakpoint 1 at 0x402cfd: defs.adb:20. (2 locations)^M
98544 Starting program: formatted_ref ^M
98546 Breakpoint 1, defs__struct1IP () at defs.adb:20^M
98547 20 return s.x; -- Set breakpoint marker here.^M
98548 (gdb) print s1'access^M
98550 instead of the expected defs.f1:
98552 (gdb) break defs.adb:20^M
98553 Breakpoint 1 at 0x402d0e: file defs.adb, line 20.^M
98555 Starting program: formatted_ref ^M
98557 Breakpoint 1, defs.f1 (s=...) at defs.adb:20^M
98558 20 return s.x; -- Set breakpoint marker here.^M
98561 This is caused by incorrect line info due to gcc PR 101575 - "[gcc-11,
98562 -gdwarf-4] Missing .file <n> directive causes invalid line info".
98564 Fix this by when landing in defs__struct1IP:
98565 - xfailing the runto, and
98566 - issuing a continue to land in defs.f1.
98568 Likewise in a few other test-cases.
98570 Tested on x86_64-linux, with:
98572 - gcc-11 and target boards unix/gdb:debug_flags=-gdwarf-4 and
98573 unix/gdb:debug_flags=-gdwarf-5.
98575 gdb/testsuite/ChangeLog:
98577 2021-07-22 Tom de Vries <tdevries@suse.de>
98579 * gdb.ada/formatted_ref.exp: Add xfail for PR gcc/101575.
98580 * gdb.ada/iwide.exp: Same.
98581 * gdb.ada/pkd_arr_elem.exp: Same.
98583 2021-07-22 Jan Beulich <jbeulich@suse.com>
98585 x86: drop dq{b,d}_mode
98586 Their sole use is for {,V}EXTRACTPS / {,V}P{EXT,INS}RB respectively; for
98587 consistency also limit use of dqw_mode to Jdqw. 64-bit disassembly
98588 reflecting REX.W / VEX.W is not in line with the assembler's opcode
98589 table having NoRex64 / VexWIG in all respective templates, i.e. assembly
98590 input isn't being honored there either. Obviously the 0FC5 encodings of
98591 {,V}PEXTRW then also need adjustment for consistency reasons.
98593 x86: drop vex_scalar_w_dq_mode
98594 It has only a single use and can easily be represented by dq_mode
98595 instead. Plus its handling in intel_operand_size() was duplicating
98596 that of vex_vsib_{d,q}_w_dq_mode anyway.
98598 x86: drop xmm_m{b,w,d,q}_mode
98599 They're effectively redundant with {b,w,d,q}_mode.
98601 x86: fold duplicate vector register printing code
98602 The bulk of OP_XMM() can be easily reused also for OP_EX(). Break the
98603 shared logic out of the function, and invoke the new helper from both
98606 x86: drop vex_mode and vex_scalar_mode
98607 These are fully redundant with, respectively, x_mode and scalar_mode.
98609 x86: correct EVEX.V' handling outside of 64-bit mode
98610 Unlike the high bit of VEX.vvvv / EVEX.vvvv, EVEX.V' is not ignored
98611 outside of 64-bit mode. Oddly enough there already are tests for these
98612 cases, but their expectations were wrong. (This may have been based on
98613 an old SDM version, where the restriction wasn't properly spelled out.)
98615 x86: fold duplicate code in MOVSXD_Fixup()
98616 There's no need to have two paths printing the "xd" mnemonic suffix.
98618 x86: fold duplicate register printing code
98619 What so far was OP_E_register() can be easily reused also for OP_G().
98620 Add suitable parameters to the function and move the invocation of
98621 swap_operand() to OP_E(). Adjust MOVSXD's first operand: There never was
98622 a need to use movsxd_mode there, and its use gets in the way of the code
98625 x86-64: properly bounds-check %bnd<N> in OP_G()
98626 The restriction to %bnd0-%bnd3 requires to also check REX.R is clear,
98627 just like OP_E_Register() also includes REX.B in its check.
98629 x86-64: generalize OP_G()'s EVEX.R' handling
98630 EVEX.R' is invalid to be clear not only for mask registers, but also for
98631 GPRs - IOW everything handled in this function.
98633 2021-07-22 Jan Beulich <jbeulich@suse.com>
98635 x86: correct VCVT{,U}SI2SD rounding mode handling
98636 With EVEX.W clear the instruction doesn't ignore the rounding mode, but
98637 (like for other insns without rounding semantics) EVEX.b set causes #UD.
98638 Hence the handling of EVEX.W needs to be done when processing
98639 evex_rounding_64_mode, not at the decode stages.
98641 Derive a new 64-bit testcase from the 32-bit one to cover the different
98642 EVEX.W treatment in both cases.
98644 2021-07-22 Jan Beulich <jbeulich@suse.com>
98646 x86: drop OP_Mask()
98647 By moving its vex.r check there it becomes fully redundant with OP_G().
98649 2021-07-22 GDB Administrator <gdbadmin@sourceware.org>
98651 Automatic date update in version.in
98653 2021-07-21 Tom de Vries <tdevries@suse.de>
98655 [gdb/testsuite] Fix gdb.cp/step-and-next-inline.exp with gcc-11
98656 When running test-case gdb.cp/step-and-next-inline.exp with gcc-11, I run
98659 KPASS: gdb.cp/step-and-next-inline.exp: no_header: next step 1 \
98660 (PRMS symtab/25507)
98661 FAIL: gdb.cp/step-and-next-inline.exp: no_header: next step 2
98662 KPASS: gdb.cp/step-and-next-inline.exp: no_header: next step 3 \
98663 (PRMS symtab/25507)
98666 [ Note that I get the same result with gcc-11 and target board
98667 unix/gdb:debug_flags=-gdwarf-4, so this is not a dwarf 4 vs 5 issue. ]
98669 With gcc-10, I have this trace:
98671 64 get_alias_set (&xx);
98672 get_alias_set (t=0x601038 <xx>) at step-and-next-inline.cc:51
98675 52 && TREE_TYPE (t).z != 1
98677 53 && TREE_TYPE (t).z != 2
98679 54 && TREE_TYPE (t).z != 3)
98681 main () at step-and-next-inline.cc:65
98684 and with gcc-11, I have instead:
98686 64 get_alias_set (&xx);
98687 get_alias_set (t=0x601038 <xx>) at step-and-next-inline.cc:51
98689 52 && TREE_TYPE (t).z != 1
98691 53 && TREE_TYPE (t).z != 2
98693 54 && TREE_TYPE (t).z != 3)
98695 main () at step-and-next-inline.cc:65
98698 and with clang-10, I have instead:
98700 64 get_alias_set (&xx);
98701 get_alias_set (t=0x601034 <xx>) at step-and-next-inline.cc:51
98703 52 && TREE_TYPE (t).z != 1
98704 53 && TREE_TYPE (t).z != 2
98705 54 && TREE_TYPE (t).z != 3)
98708 main () at step-and-next-inline.cc:65
98712 The test-case tries to verify that we don't step into inlined function
98713 tree_check (lines 40-43) (so, with the clang trace we get that right).
98715 The test-case then tries to kfail the problems when using gcc, but this is
98716 done in such a way that the testing still gets out of sync after a failure.
98717 That is: the "next step 2" check that is supposed to match
98718 "TREE_TYPE (t).z != 2" is actually matching "TREE_TYPE (t).z != 1":
98721 52 && TREE_TYPE (t).z != 1^M
98722 (gdb) PASS: gdb.cp/step-and-next-inline.exp: no_header: next step 2
98725 Fix this by issuing extra nexts to arrive at the required lines.
98727 Tested on x86_64-linux, with gcc-8, gcc-9, gcc-10, gcc-11, clang-8, clang-10
98730 gdb/testsuite/ChangeLog:
98732 2021-07-20 Tom de Vries <tdevries@suse.de>
98734 * gdb.cp/step-and-next-inline.cc (tree_check, get_alias_set, main):
98735 Tag closing brace with comment.
98736 * gdb.cp/step-and-next-inline.h: Update to keep identical with
98737 step-and-next-inline.cc.
98738 * gdb.cp/step-and-next-inline.exp: Issue extra next when required.
98740 2021-07-21 Nick Clifton <nickc@redhat.com>
98742 Updated Russian translation for the bfd library
98744 2021-07-21 Luca Boccassi <luca.boccassi@microsoft.com>
98746 Allows linker scripts to set the SEC_READONLY flag.
98747 * ld.texi: Document new output section type.
98748 * ldgram.y: Add new token.
98749 * ldlang.c: Handle the new flag.
98750 * ldlang.h: Add readonly_section to list of section types.
98751 * ldlex.l: Add a new identifier.
98752 * testsuite/ld-scripts/output-section-types.t: New example linker script.
98753 * testsuite/ld-scripts/output-section-types.d: Test driver.
98754 * testsyute/ld-scripts/script.exp: Run the new test.
98756 2021-07-21 Tom de Vries <tdevries@suse.de>
98758 [gdb/testsuite] Fix FAILs due to PR gcc/101452
98759 When running test-case gdb.base/ptype-offsets.exp with gcc-11 (with -gdwarf-5
98760 default) or gcc-10 with target board unix/gdb:debug_flags=-gdwarf-5 we run
98761 into this regression:
98763 (gdb) ptype/o static_member^M
98764 /* offset | size */ type = struct static_member {^M
98765 - static static_member Empty;^M
98766 /* 0 | 4 */ int abc;^M
98768 /* total size (bytes): 4 */^M
98770 -(gdb) PASS: gdb.base/ptype-offsets.exp: ptype/o static_member
98771 +(gdb) FAIL: gdb.base/ptype-offsets.exp: ptype/o static_member
98774 This is caused by missing debug info, which I filed as gcc PR101452 - "[debug,
98775 dwarf-5] undefined static member removed by
98776 -feliminate-unused-debug-symbols".
98778 It's not clear yet whether this is a bug or a feature, but work around this in
98780 - defining the static member
98781 - adding additional_flags=-fno-eliminate-unused-debug-types.
98783 Tested on x86_64-linux.
98785 gdb/testsuite/ChangeLog:
98787 2021-07-20 Tom de Vries <tdevries@suse.de>
98789 * lib/gdb.exp (gcc_major_version): New proc.
98790 * gdb.base/ptype-offsets.cc: Define static member static_member::Empty.
98791 * gdb.cp/templates.exp: Define static member using -DGCC_BUG.
98792 * gdb.cp/m-static.exp: Add
98793 additional_flags=-fno-eliminate-unused-debug-types.
98794 * gdb.cp/pr-574.exp: Same.
98795 * gdb.cp/pr9167.exp: Same.
98797 2021-07-21 Tom de Vries <tdevries@suse.de>
98799 [gdb/testsuite] Add KFAILs for gdb.ada FAILs with gcc-11
98800 With gcc-11 we run into:
98802 (gdb) print pa_ptr.all^M
98803 That operation is not available on integers of more than 8 bytes.^M
98804 (gdb) KFAIL: gdb.ada/arrayptr.exp: scenario=all: print pa_ptr.all (PRMS: gdb/20991)
98807 This is due to PR exp/20991 - "__int128 type support". Mark this and similar
98810 Also mark this FAIL:
98812 (gdb) print pa_ptr(3)^M
98813 cannot subscript or call something of type `foo__packed_array_ptr'^M
98814 (gdb) FAIL: gdb.ada/arrayptr.exp: scenario=minimal: print pa_ptr(3)
98816 as a KFAIL for PR ada/28115 - "Support packed array encoded as
98817 DW_TAG_subrange_type".
98819 Tested on x86_64-linux, with gcc-10 and gcc-11.
98821 gdb/testsuite/ChangeLog:
98823 2021-07-21 Tom de Vries <tdevries@suse.de>
98825 * gdb.ada/arrayptr.exp: Add KFAILs for PR20991 and PR28115.
98826 * gdb.ada/exprs.exp: Add KFAILs for PR20991.
98827 * gdb.ada/packed_array_assign.exp: Same.
98829 2021-07-21 Alan Modra <amodra@gmail.com>
98832 Many places report errors of the nature "can't resolve a - b".
98833 This provides a utility function to report such errors consistently.
98834 I removed the section reporting and quotes around symbol names while I
98836 ifunc-2.s:4: Error: can't resolve `bar1' {.text.1 section} - `foo1' {.text.1 section}
98838 ifunc-2.s:4: Error: can't resolve bar1 - foo1
98840 In many cases the section names don't help the user very much in
98841 figuring out what went wrong, and the quotes if present arguably ought
98842 to be placed around the entire expression:
98843 can't resolve `bar1 - foo1'
98845 The patch also tidies some tc_get_reloc functions that leak memory on
98848 * write.h (as_bad_subtract): Declare.
98849 * write.c (as_bad_subtract): New function.
98850 (fixup_segment): Use as_bad_subtract.
98851 * config/tc-arc.c (md_apply_fix): Likewise.
98852 * config/tc-avr.c (md_apply_fix, tc_gen_reloc): Likewise.
98853 * config/tc-cris.c (md_apply_fix): Likewise.
98854 * config/tc-d10v.c (md_apply_fix): Likewise.
98855 * config/tc-d30v.c (md_apply_fix): Likewise.
98856 * config/tc-ft32.c (md_apply_fix): Likewise.
98857 * config/tc-h8300.c (tc_gen_reloc): Likewise.
98858 * config/tc-m68hc11.c (md_apply_fix): Likewise.
98859 * config/tc-mmix.c (mmix_frob_file): Likewise.
98860 * config/tc-mn10200.c (tc_gen_reloc): Likewise.
98861 * config/tc-nds32.c (nds32_apply_fix): Likewise.
98862 * config/tc-pru.c (md_apply_fix): Likewise.
98863 * config/tc-riscv.c (md_apply_fix): Likewise.
98864 * config/tc-s12z.c (md_apply_fix): Likewise.
98865 * config/tc-s390.c (md_apply_fix): Likewise.
98866 * config/tc-tilegx.c (md_apply_fix): Likewise.
98867 * config/tc-tilepro.c (md_apply_fix): Likewise.
98868 * config/tc-v850.c (md_apply_fix): Likewise.
98869 * config/tc-vax.c (md_apply_fix): Likewise.
98870 * config/tc-xc16x.c (tc_gen_reloc): Likewise.
98871 * config/tc-xgate.c (md_apply_fix): Likewise.
98872 * config/tc-xstormy16.c (xstormy16_md_apply_fix): Likewise.
98873 * config/tc-xtensa.c (md_apply_fix): Likewise.
98874 * config/tc-z80.c (tc_gen_reloc): Likewise.
98875 * config/tc-spu.c (md_apply_fix): Likewise.
98876 (tc_gen_reloc): Delete dead code. Free memory on error.
98877 * config/tc-cr16.c (tc_gen_reloc): Use as_bad_subtract. Free
98879 * config/tc-crx.c (tc_gen_reloc): Likewise.
98880 * config/tc-ppc.c (tc_gen_reloc): Likewise.
98881 * testsuite/gas/i386/ifunc-2.l: Adjust to suit changed error message.
98882 * testsuite/gas/mips/lui-2.l: Likewise.
98883 * testsuite/gas/tic6x/reloc-bad-1.l: Likewise.
98885 2021-07-21 John Ericson <git@JohnEricson.me>
98887 Remove `netbsdpe` support
98888 netbsdpe was deprecated in c2ce831330e10dab4703094491f80b6b9a5c2289.
98889 Since then, a release has passed (2.37), and it was marked obselete in
98890 5c9cbf07f3f972ecffe13d858010b3179df17b32. Unless I am mistaken, that
98891 means we can now remove support altogether.
98893 All branches in the "active" code are remove, and the target is
98894 additionally marked as obsolete next to the other removed ones for
98897 Per [1] from the NetBSD toolchain list, PE/COFF support was removed a
98898 decade ago. Furthermore, the sole mention of this target in the binutils
98899 commit history was in 2002. Together, I'm led to believe this target
98900 hasn't seen much attention in quite a while.
98902 [1]: https://mail-index.netbsd.org/tech-toolchain/2021/06/16/msg003996.html
98905 * config.bfd: Remove netbsdpe entry.
98907 * configure.ac: Remove netbsdpe entry.
98908 * testsuite/lib/binutils-common.exp (is_pecoff_format): Likewise.
98909 * configure: Regenerate.
98911 * configure.tgt: Remove netbsdpe entry.
98913 * configure.tgt: Add netbsdpe to removed targets.
98915 * configure.tgt: Remove netbsdpe entry.
98916 * testsuite/ld-bootstrap/bootstrap.exp: Likewise.
98918 2021-07-21 GDB Administrator <gdbadmin@sourceware.org>
98920 Automatic date update in version.in
98922 2021-07-20 Alan Modra <amodra@gmail.com>
98924 PR28106, build of 2.37 fails on FreeBSD and Clang
98925 https://en.cppreference.com/w/cpp/types/NULL says NULL might be
98926 defined as nullptr.
98927 https://en.cppreference.com/w/cpp/language/reinterpret_cast says
98928 reinterpret_cast can't be used on nullptr.
98932 * gc.h (gc_process_relocs): Use static_cast in Section_id constructor.
98934 2021-07-20 Luis Machado <luis.machado@linaro.org>
98936 Fix printing of non-address types when memory tagging is enabled
98937 When the architecture supports memory tagging, we handle
98938 pointer/reference types in a special way, so we can validate tags and
98941 Unfortunately, the currently implementation errors out when the user
98942 prints non-address values: composite types, floats, references, member
98943 functions and other things.
98948 Value can't be converted to integer.
98950 Non-existent internal variables:
98953 Value can't be converted to integer.
98955 The same happens for complex types and printing struct/union types.
98957 There are a few problems here.
98959 The first one is that after print_command_1 evaluates the expression
98960 to print, the tag validation code call value_as_address
98961 unconditionally, without making sure we have have a suitable type
98962 where it makes to sense to call it. That results in value_as_address
98963 (if it isn't given a pointer-like type) trying to treat the value as
98964 an integer and convert it to an address, which #1 - doesn't make sense
98965 (i.e., no sense in validating tags after "print 1"), and throws for
98966 non-integer-convertible types. We fix this by making sure we have a
98967 pointer or reference type first, and only if so then proceed to check
98968 if the address-like value has tags.
98970 The second is that we're calling value_as_address even if we have an
98971 optimized out or unavailable value, which throws, because the value's
98972 contents aren't fully accessible/readable. This error currently
98973 escapes out and aborts the print. This case is fixed by checking for
98974 optimized out / unavailable explicitly.
98976 Third, the tag checking process does not gracefully handle exceptions.
98977 If any exception is thrown from the tag validation code, we abort the
98978 print. E.g., the target may fail to access tags via a running thread.
98979 Or the needed /proc files aren't available. Or some other untold
98980 reason. This is a bit too rigid. This commit changes print_command_1
98981 to catch errors, print them, and still continue with the normal
98982 expression printing path instead of erroring out and printing nothing
98985 With this patch, printing works correctly again:
98988 $1 = {d = {f = {2.0546950501119882e-81, 2.0546950501119882e-81}, u = {3399988123389603631, 3399988123389603631}, s = {
98989 3399988123389603631, 3399988123389603631}}, s = {f = {1.59329203e-10, 1.59329203e-10, 1.59329203e-10, 1.59329203e-10}, u = {
98990 791621423, 791621423, 791621423, 791621423}, s = {791621423, 791621423, 791621423, 791621423}}, h = {bf = {1.592e-10,
98991 1.592e-10, 1.592e-10, 1.592e-10, 1.592e-10, 1.592e-10, 1.592e-10, 1.592e-10}, f = {0.11224, 0.11224, 0.11224, 0.11224, 0.11224,
98992 0.11224, 0.11224, 0.11224}, u = {12079, 12079, 12079, 12079, 12079, 12079, 12079, 12079}, s = {12079, 12079, 12079, 12079,
98993 12079, 12079, 12079, 12079}}, b = {u = {47 <repeats 16 times>}, s = {47 <repeats 16 times>}}, q = {u = {
98994 62718710765820030520700417840365121327}, s = {62718710765820030520700417840365121327}}}
99000 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28110
99002 2021-07-20 Nelson Chu <nelson.chu@sifive.com>
99004 RISC-V: Minor updates for architecture parser.
99005 * Two add subset functions is redundant. Keep the riscv_add_implicit_subset,
99006 and renamed it to riscv_add_subset. Besides, if the subset is added in order,
99007 then we just add it at the tail of the subset list.
99009 * Removed the "-march:" prefix from the error messages. Since not only the
99010 -march= option will use the parser, but also the architecture elf attributes,
99011 the default architecture setting and linker will use the same parser.
99013 * Use a function, riscv_parse_check_conflicts, to check the conflicts
99014 of extensions, including the rv64e and rv32q.
99016 The rv32emc-elf/rv32i-elf/rv32gc-linux/rv64gc-elf/rv64gc-linux regressions
99017 are tested and passed.
99020 * elfxx-riscv.c (riscv_lookup_subset): Check the subset tail list
99021 first. If the subset is added in order, then we can just add it to
99022 the tail without searching the whole list.
99023 (riscv_add_subset): Replaced by riscv_add_implicit_subset.
99024 (riscv_add_implicit_subset): Renamed to riscv_add_subset.
99025 (riscv_parse_add_subset): Updated.
99026 (riscv_parsing_subset_version): Removed the "-march:" prefix from
99028 (riscv_parse_prefixed_ext): Likewise.
99029 (riscv_parse_std_ext): Likewise. And move the rv<xlen>e check
99030 to riscv_parse_check_conflicts.
99031 (riscv_parse_check_conflicts): New function used to check conflicts.
99032 (riscv_parse_subset): Updated.
99034 * testsuite/gas/riscv/march-fail-base-02.l: Updated.
99035 * testsuite/gas/riscv/march-fail-unknown-std.l: Likewise.
99037 2021-07-20 GDB Administrator <gdbadmin@sourceware.org>
99039 Automatic date update in version.in
99041 2021-07-19 Simon Marchi <simon.marchi@polymtl.ca>
99043 gdb: set current thread in btrace_compute_ftrace_{bts,pt}
99044 As documented in bug 28086, test gdb.btrace/enable-new-thread.exp
99045 started failing with commit 0618ae414979 ("gdb: optimize
99046 all_matching_threads_iterator"):
99048 (gdb) record btrace^M
99049 (gdb) PASS: gdb.btrace/enable-new-thread.exp: record btrace
99051 Breakpoint 2 at 0x555555555175: file /home/smarchi/src/binutils-gdb/gdb/testsuite/gdb.btrace/enable-new-thread.c, line 24.^M
99054 /home/smarchi/src/binutils-gdb/gdb/inferior.c:303: internal-error: inferior* find_inferior_pid(process_stratum_target*, int): Assertion `pid != 0' failed.^M
99055 A problem internal to GDB has been detected,^M
99056 further debugging may prove unreliable.^M
99057 Quit this debugging session? (y or n) FAIL: gdb.btrace/enable-new-thread.exp: continue to breakpoint: cont to bp.1 (GDB internal error)
99059 Note that I only see the failure if GDB is compiled without libipt
99060 support. This is because GDB then makes use BTS instead of PT, so
99061 exercises different code paths.
99063 I think that the commit above just exposed an existing problem. The
99064 stack trace of the internal error is:
99066 #8 0x0000561cb81e404e in internal_error (file=0x561cb83aa2f8 "/home/smarchi/src/binutils-gdb/gdb/inferior.c", line=303, fmt=0x561cb83aa099 "%s: Assertion `%s' failed.") at /home/smarchi/src/binutils-gdb/gdbsupport/errors.cc:55
99067 #9 0x0000561cb7b5c031 in find_inferior_pid (targ=0x561cb8aafb60 <the_amd64_linux_nat_target>, pid=0) at /home/smarchi/src/binutils-gdb/gdb/inferior.c:303
99068 #10 0x0000561cb7b5c102 in find_inferior_ptid (targ=0x561cb8aafb60 <the_amd64_linux_nat_target>, ptid=...) at /home/smarchi/src/binutils-gdb/gdb/inferior.c:317
99069 #11 0x0000561cb7f1d1c3 in find_thread_ptid (targ=0x561cb8aafb60 <the_amd64_linux_nat_target>, ptid=...) at /home/smarchi/src/binutils-gdb/gdb/thread.c:487
99070 #12 0x0000561cb7f1b921 in all_matching_threads_iterator::all_matching_threads_iterator (this=0x7ffc4ee34678, filter_target=0x561cb8aafb60 <the_amd64_linux_nat_target>, filter_ptid=...) at /home/smarchi/src/binutils-gdb/gdb/thread-iter.c:125
99071 #13 0x0000561cb77bc462 in filtered_iterator<all_matching_threads_iterator, non_exited_thread_filter>::filtered_iterator<process_stratum_target* const&, ptid_t const&> (this=0x7ffc4ee34670) at /home/smarchi/src/binutils-gdb/gdb/../gdbsupport/filtered-iterator.h:42
99072 #14 0x0000561cb77b97cb in all_non_exited_threads_range::begin (this=0x7ffc4ee34650) at /home/smarchi/src/binutils-gdb/gdb/thread-iter.h:243
99073 #15 0x0000561cb7d8ba30 in record_btrace_target::record_is_replaying (this=0x561cb8aa6250 <record_btrace_ops>, ptid=...) at /home/smarchi/src/binutils-gdb/gdb/record-btrace.c:1411
99074 #16 0x0000561cb7d8bb83 in record_btrace_target::xfer_partial (this=0x561cb8aa6250 <record_btrace_ops>, object=TARGET_OBJECT_MEMORY, annex=0x0, readbuf=0x7ffc4ee34c58 "\260g\343N\374\177", writebuf=0x0, offset=140737352774277, len=1, xfered_len=0x7ffc4ee34ad8) at /home/smarchi/src/binutils-gdb/gdb/record-btrace.c:1437
99075 #17 0x0000561cb7ef73a9 in raw_memory_xfer_partial (ops=0x561cb8aa6250 <record_btrace_ops>, readbuf=0x7ffc4ee34c58 "\260g\343N\374\177", writebuf=0x0, memaddr=140737352774277, len=1, xfered_len=0x7ffc4ee34ad8) at /home/smarchi/src/binutils-gdb/gdb/target.c:1504
99076 #18 0x0000561cb7ef77da in memory_xfer_partial_1 (ops=0x561cb8aa6250 <record_btrace_ops>, object=TARGET_OBJECT_CODE_MEMORY, readbuf=0x7ffc4ee34c58 "\260g\343N\374\177", writebuf=0x0, memaddr=140737352774277, len=1, xfered_len=0x7ffc4ee34ad8) at /home/smarchi/src/binutils-gdb/gdb/target.c:1635
99077 #19 0x0000561cb7ef78b5 in memory_xfer_partial (ops=0x561cb8aa6250 <record_btrace_ops>, object=TARGET_OBJECT_CODE_MEMORY, readbuf=0x7ffc4ee34c58 "\260g\343N\374\177", writebuf=0x0, memaddr=140737352774277, len=1, xfered_len=0x7ffc4ee34ad8) at /home/smarchi/src/binutils-gdb/gdb/target.c:1664
99078 #20 0x0000561cb7ef7ba4 in target_xfer_partial (ops=0x561cb8aa6250 <record_btrace_ops>, object=TARGET_OBJECT_CODE_MEMORY, annex=0x0, readbuf=0x7ffc4ee34c58 "\260g\343N\374\177", writebuf=0x0, offset=140737352774277, len=1, xfered_len=0x7ffc4ee34ad8) at /home/smarchi/src/binutils-gdb/gdb/target.c:1721
99079 #21 0x0000561cb7ef8503 in target_read_partial (ops=0x561cb8aa6250 <record_btrace_ops>, object=TARGET_OBJECT_CODE_MEMORY, annex=0x0, buf=0x7ffc4ee34c58 "\260g\343N\374\177", offset=140737352774277, len=1, xfered_len=0x7ffc4ee34ad8) at /home/smarchi/src/binutils-gdb/gdb/target.c:1974
99080 #22 0x0000561cb7ef861f in target_read (ops=0x561cb8aa6250 <record_btrace_ops>, object=TARGET_OBJECT_CODE_MEMORY, annex=0x0, buf=0x7ffc4ee34c58 "\260g\343N\374\177", offset=140737352774277, len=1) at /home/smarchi/src/binutils-gdb/gdb/target.c:2014
99081 #23 0x0000561cb7ef809f in target_read_code (memaddr=140737352774277, myaddr=0x7ffc4ee34c58 "\260g\343N\374\177", len=1) at /home/smarchi/src/binutils-gdb/gdb/target.c:1869
99082 #24 0x0000561cb7937f4d in gdb_disassembler::dis_asm_read_memory (memaddr=140737352774277, myaddr=0x7ffc4ee34c58 "\260g\343N\374\177", len=1, info=0x7ffc4ee34e88) at /home/smarchi/src/binutils-gdb/gdb/disasm.c:139
99083 #25 0x0000561cb80ab66d in fetch_data (info=0x7ffc4ee34e88, addr=0x7ffc4ee34c59 "g\343N\374\177") at /home/smarchi/src/binutils-gdb/opcodes/i386-dis.c:194
99084 #26 0x0000561cb80ab7e2 in ckprefix () at /home/smarchi/src/binutils-gdb/opcodes/i386-dis.c:8628
99085 #27 0x0000561cb80adbd8 in print_insn (pc=140737352774277, info=0x7ffc4ee34e88) at /home/smarchi/src/binutils-gdb/opcodes/i386-dis.c:9587
99086 #28 0x0000561cb80abe4f in print_insn_i386 (pc=140737352774277, info=0x7ffc4ee34e88) at /home/smarchi/src/binutils-gdb/opcodes/i386-dis.c:8894
99087 #29 0x0000561cb7744a19 in default_print_insn (memaddr=140737352774277, info=0x7ffc4ee34e88) at /home/smarchi/src/binutils-gdb/gdb/arch-utils.c:1029
99088 #30 0x0000561cb7b33067 in i386_print_insn (pc=140737352774277, info=0x7ffc4ee34e88) at /home/smarchi/src/binutils-gdb/gdb/i386-tdep.c:4013
99089 #31 0x0000561cb7acd8f4 in gdbarch_print_insn (gdbarch=0x561cbae2fb60, vma=140737352774277, info=0x7ffc4ee34e88) at /home/smarchi/src/binutils-gdb/gdb/gdbarch.c:3478
99090 #32 0x0000561cb793a32d in gdb_disassembler::print_insn (this=0x7ffc4ee34e80, memaddr=140737352774277, branch_delay_insns=0x0) at /home/smarchi/src/binutils-gdb/gdb/disasm.c:795
99091 #33 0x0000561cb793a5b0 in gdb_print_insn (gdbarch=0x561cbae2fb60, memaddr=140737352774277, stream=0x561cb8ac99f8 <null_stream>, branch_delay_insns=0x0) at /home/smarchi/src/binutils-gdb/gdb/disasm.c:850
99092 #34 0x0000561cb793a631 in gdb_insn_length (gdbarch=0x561cbae2fb60, addr=140737352774277) at /home/smarchi/src/binutils-gdb/gdb/disasm.c:859
99093 #35 0x0000561cb77f53f4 in btrace_compute_ftrace_bts (tp=0x561cbba11210, btrace=0x7ffc4ee35188, gaps=...) at /home/smarchi/src/binutils-gdb/gdb/btrace.c:1107
99094 #36 0x0000561cb77f55f5 in btrace_compute_ftrace_1 (tp=0x561cbba11210, btrace=0x7ffc4ee35180, cpu=0x0, gaps=...) at /home/smarchi/src/binutils-gdb/gdb/btrace.c:1527
99095 #37 0x0000561cb77f5705 in btrace_compute_ftrace (tp=0x561cbba11210, btrace=0x7ffc4ee35180, cpu=0x0) at /home/smarchi/src/binutils-gdb/gdb/btrace.c:1560
99096 #38 0x0000561cb77f583b in btrace_add_pc (tp=0x561cbba11210) at /home/smarchi/src/binutils-gdb/gdb/btrace.c:1589
99097 #39 0x0000561cb77f5a86 in btrace_enable (tp=0x561cbba11210, conf=0x561cb8ac6878 <record_btrace_conf>) at /home/smarchi/src/binutils-gdb/gdb/btrace.c:1629
99098 #40 0x0000561cb7d88d26 in record_btrace_enable_warn (tp=0x561cbba11210) at /home/smarchi/src/binutils-gdb/gdb/record-btrace.c:294
99099 #41 0x0000561cb7c603dc in std::__invoke_impl<void, void (*&)(thread_info*), thread_info*> (__f=@0x561cbb6c4878: 0x561cb7d88cdc <record_btrace_enable_warn(thread_info*)>) at /usr/include/c++/10/bits/invoke.h:60
99100 #42 0x0000561cb7c5e5a6 in std::__invoke_r<void, void (*&)(thread_info*), thread_info*> (__fn=@0x561cbb6c4878: 0x561cb7d88cdc <record_btrace_enable_warn(thread_info*)>) at /usr/include/c++/10/bits/invoke.h:153
99101 #43 0x0000561cb7c5dc92 in std::_Function_handler<void (thread_info*), void (*)(thread_info*)>::_M_invoke(std::_Any_data const&, thread_info*&&) (__functor=..., __args#0=@0x7ffc4ee35310: 0x561cbba11210) at /usr/include/c++/10/bits/std_function.h:291
99102 #44 0x0000561cb7f2600f in std::function<void (thread_info*)>::operator()(thread_info*) const (this=0x561cbb6c4878, __args#0=0x561cbba11210) at /usr/include/c++/10/bits/std_function.h:622
99103 #45 0x0000561cb7f23dc8 in gdb::observers::observable<thread_info*>::notify (this=0x561cb8ac5aa0 <gdb::observers::new_thread>, args#0=0x561cbba11210) at /home/smarchi/src/binutils-gdb/gdb/../gdbsupport/observable.h:150
99104 #46 0x0000561cb7f1c436 in add_thread_silent (targ=0x561cb8aafb60 <the_amd64_linux_nat_target>, ptid=...) at /home/smarchi/src/binutils-gdb/gdb/thread.c:263
99105 #47 0x0000561cb7f1c479 in add_thread_with_info (targ=0x561cb8aafb60 <the_amd64_linux_nat_target>, ptid=..., priv=0x561cbb3f7ab0) at /home/smarchi/src/binutils-gdb/gdb/thread.c:272
99106 #48 0x0000561cb7bfa1d0 in record_thread (info=0x561cbb0413a0, tp=0x0, ptid=..., th_p=0x7ffc4ee35610, ti_p=0x7ffc4ee35620) at /home/smarchi/src/binutils-gdb/gdb/linux-thread-db.c:1380
99107 #49 0x0000561cb7bf7a2a in thread_from_lwp (stopped=0x561cba81db20, ptid=...) at /home/smarchi/src/binutils-gdb/gdb/linux-thread-db.c:429
99108 #50 0x0000561cb7bf7ac5 in thread_db_notice_clone (parent=..., child=...) at /home/smarchi/src/binutils-gdb/gdb/linux-thread-db.c:447
99109 #51 0x0000561cb7bdc9a2 in linux_handle_extended_wait (lp=0x561cbae25720, status=4991) at /home/smarchi/src/binutils-gdb/gdb/linux-nat.c:1981
99110 #52 0x0000561cb7bdf0f3 in linux_nat_filter_event (lwpid=435403, status=198015) at /home/smarchi/src/binutils-gdb/gdb/linux-nat.c:2920
99111 #53 0x0000561cb7bdfed6 in linux_nat_wait_1 (ptid=..., ourstatus=0x7ffc4ee36398, target_options=...) at /home/smarchi/src/binutils-gdb/gdb/linux-nat.c:3202
99112 #54 0x0000561cb7be0b68 in linux_nat_target::wait (this=0x561cb8aafb60 <the_amd64_linux_nat_target>, ptid=..., ourstatus=0x7ffc4ee36398, target_options=...) at /home/smarchi/src/binutils-gdb/gdb/linux-nat.c:3440
99113 #55 0x0000561cb7bfa2fc in thread_db_target::wait (this=0x561cb8a9acd0 <the_thread_db_target>, ptid=..., ourstatus=0x7ffc4ee36398, options=...) at /home/smarchi/src/binutils-gdb/gdb/linux-thread-db.c:1412
99114 #56 0x0000561cb7d8e356 in record_btrace_target::wait (this=0x561cb8aa6250 <record_btrace_ops>, ptid=..., status=0x7ffc4ee36398, options=...) at /home/smarchi/src/binutils-gdb/gdb/record-btrace.c:2547
99115 #57 0x0000561cb7ef996d in target_wait (ptid=..., status=0x7ffc4ee36398, options=...) at /home/smarchi/src/binutils-gdb/gdb/target.c:2608
99116 #58 0x0000561cb7b6d297 in do_target_wait_1 (inf=0x561cba6d8780, ptid=..., status=0x7ffc4ee36398, options=...) at /home/smarchi/src/binutils-gdb/gdb/infrun.c:3640
99117 #59 0x0000561cb7b6d43e in operator() (__closure=0x7ffc4ee36190, inf=0x561cba6d8780) at /home/smarchi/src/binutils-gdb/gdb/infrun.c:3701
99118 #60 0x0000561cb7b6d7b2 in do_target_wait (ecs=0x7ffc4ee36370, options=...) at /home/smarchi/src/binutils-gdb/gdb/infrun.c:3720
99119 #61 0x0000561cb7b6e67d in fetch_inferior_event () at /home/smarchi/src/binutils-gdb/gdb/infrun.c:4069
99120 #62 0x0000561cb7b4659b in inferior_event_handler (event_type=INF_REG_EVENT) at /home/smarchi/src/binutils-gdb/gdb/inf-loop.c:41
99121 #63 0x0000561cb7be25f7 in handle_target_event (error=0, client_data=0x0) at /home/smarchi/src/binutils-gdb/gdb/linux-nat.c:4227
99122 #64 0x0000561cb81e4ee2 in handle_file_event (file_ptr=0x561cbae24e10, ready_mask=1) at /home/smarchi/src/binutils-gdb/gdbsupport/event-loop.cc:575
99123 #65 0x0000561cb81e5490 in gdb_wait_for_event (block=0) at /home/smarchi/src/binutils-gdb/gdbsupport/event-loop.cc:701
99124 #66 0x0000561cb81e41be in gdb_do_one_event () at /home/smarchi/src/binutils-gdb/gdbsupport/event-loop.cc:212
99125 #67 0x0000561cb7c18096 in start_event_loop () at /home/smarchi/src/binutils-gdb/gdb/main.c:421
99126 #68 0x0000561cb7c181e0 in captured_command_loop () at /home/smarchi/src/binutils-gdb/gdb/main.c:481
99127 #69 0x0000561cb7c19d7e in captured_main (data=0x7ffc4ee366a0) at /home/smarchi/src/binutils-gdb/gdb/main.c:1353
99128 #70 0x0000561cb7c19df0 in gdb_main (args=0x7ffc4ee366a0) at /home/smarchi/src/binutils-gdb/gdb/main.c:1368
99129 #71 0x0000561cb7693186 in main (argc=11, argv=0x7ffc4ee367b8) at /home/smarchi/src/binutils-gdb/gdb/gdb.c:32
99131 At frame 45, the new_thread observable is fired. At this moment, the
99132 new thread isn't the current thread, inferior_ptid is null_ptid. I
99133 think this is ok: the new_thread observable doesn't give any guarantee
99134 on the global context when observers are invoked. Frame 35,
99135 btrace_compute_ftrace_bts, calls gdb_insn_length. gdb_insn_length
99136 doesn't have a thread_info or other parameter what could indicate where
99137 to read memory from, it implicitly uses the global context
99140 So we reach the all_non_exited_threads_range in
99141 record_btrace_target::record_is_replaying with a null inferior_ptid.
99142 The previous implemention of all_non_exited_threads_range didn't care,
99143 but the new one does. The problem of calling gdb_insn_length and
99144 ultimately trying to read memory with a null inferior_ptid already
99145 existed, but the commit mentioned above made it visible.
99147 Something between frames 40 (record_btrace_enable_warn) and 35
99148 (btrace_compute_ftrace_bts) needs to be switching the global context to
99149 make TP the current thread. Since btrace_compute_ftrace_bts takes the
99150 thread_info to work with as a parameter, that typically means that it
99151 doesn't require its caller to also set the global current context
99152 (current thread) when calling. If it needs to call other functions
99153 that do require the global current thread to be set, then it needs to
99154 temporarily change the current thread while calling these other
99155 functions. Therefore, switch and restore the current thread in
99156 btrace_compute_ftrace_bts.
99158 By inspection, it looks like btrace_compute_ftrace_pt may also call
99159 functions sensitive to the global context: it installs the
99160 btrace_pt_readmem_callback callback in the PT instruction decoder. When
99161 this function gets called, inferior_ptid must be set appropriately. Add
99162 a switch and restore in there too.
99164 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28086
99165 Change-Id: I407fbfe41aab990068bd102491aa3709b0a034b3
99167 2021-07-19 GDB Administrator <gdbadmin@sourceware.org>
99169 Automatic date update in version.in
99171 2021-07-18 Nick Clifton <nickc@redhat.com>
99173 Move pending-obsolesence targets onto the obsolete list.
99174 * config.bfd: Move pending obsoletion targets to obsolete list.
99176 Update how-to-make-a-release checklist with latest changes from 2.37 release
99178 2021-07-18 Michael Krasnyk <mkrasnyk@argo.ai>
99180 PR28098 Skip R_*_NONE relocation entries with zero r_sym without counting
99182 * reloc.cc (Track_relocs::advance): Skip R_*_NONE relocation entries
99183 with r_sym of zero without counting in advance method.
99185 2021-07-18 Simon Marchi <simon.marchi@polymtl.ca>
99187 gdb: convert nat/x86-dregs.c macros to functions
99188 I'm debugging why GDB crashes on OpenBSD/amd64, turns out it's because
99189 x86_dr_low.get_status is nullptr. It would have been useful to be able
99190 to break on x86_dr_low_get_status, so I thought it would be a good
99191 reason to convert these function-like macros into functions.
99193 Change-Id: Ic200b50ef8455b4697bc518da0fa2bb704cf4721
99195 2021-07-18 GDB Administrator <gdbadmin@sourceware.org>
99197 Automatic date update in version.in
99199 2021-07-17 Tom Tromey <tom@tromey.com>
99201 Fix file-name handling regression with DWARF index
99202 When run with the gdb-index or debug-names target boards, dup-psym.exp
99203 fails. This came up for me because my new DWARF scanner reuses this
99204 part of the existing index code, and so it registers as a regression.
99205 This is PR symtab/25834.
99207 Looking into this, I found that the DWARF index code here is fairly
99208 different from the psymtab code. I don't think there's a deep reason
99209 for this, and in fact, it seemed to me that the index code could
99210 simply mimic what the psymtab code already does.
99212 That is what this patch implements. The DW_AT_name and DW_AT_comp_dir
99213 are now stored in the quick file names table. This may require
99214 allocating a quick file names table even when DW_AT_stmt_list does not
99215 exist. Then, the functions that work with this data are changed to
99216 use find_source_or_rewrite, just as the psymbol code does. Finally,
99217 line_header::file_full_name is removed, as it is no longer needed.
99219 Currently, the index maintains a hash table of "quick file names".
99220 The hash table uses a deletion function to free the "real name"
99221 components when necessary. There's also a second such function to
99222 implement the forget_cached_source_info method.
99224 This bug fix patch will create a quick file name object even when
99225 there is no DW_AT_stmt_list, meaning that the object won't be entered
99226 in the hash table. So, this patch changes the memory management
99227 approach so that the entries are cleared when the per-BFD object is
99228 destroyed. (A dwarf2_per_cu_data destructor is not introduced,
99229 because we have been avoiding adding a vtable to that class.)
99231 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=25834
99233 2021-07-17 Tom Tromey <tom@tromey.com>
99235 Check for debug-types in map_symbol_filenames
99236 map_symbol_filenames can skip type units -- in fact I think it has to,
99237 due to the assertion at the top of dw2_get_file_names. This may be a
99238 regression due to the TU/CU unification patch, I did not check.
99240 Simplify DWARF file name caching
99241 The DWARF index file name caching code only records when a line table
99242 has been read and the reading failed. However, the code would be
99243 simpler if it recorded any attempt, which is what this patch
99246 Introduce find_source_or_rewrite
99247 The final bug fix in this series would duplicate the logic in
99248 psymtab_to_fullname, so this patch extracts the body of this function
99249 into a new function.
99251 Simplify file_and_directory storage management
99252 file_and_directory carries a std::string in case the compilation
99253 directory is computed, but a subsequent patch wants to preserve this
99254 string without also having to maintain the storage for it. So, this
99255 patch arranges for the compilation directory string to be stored in
99256 the per-BFD string bcache instead.
99258 Pass file_and_directory through DWARF line-decoding code
99259 This patch removes the redundant "comp_unit" parameter from
99260 compute_include_file_name, and arranges to pass a file_and_directory
99261 object from the readers down to this function. It also changes the
99262 partial symtab reader to use find_file_and_directory, rather than
99263 reimplement this functionality by hand.
99265 Rename and refactor psymtab_include_file_name
99266 In order to fix an index-related regression, I want to use
99267 psymtab_include_file_name in the DWARF index file-handling code. This
99268 patch renames this function and changes it to no longer require a
99269 partial symtab to be passed in. A subsequent patch will further
99270 refactor this code to remove the redundant parameter (which was always
99271 there but is now more obvious).
99273 2021-07-17 Sergey Belyashov <Sergey.Belyashov@gmail.com>
99275 Add basic Z80 CPU support
99277 - Z80 (all undocumented instructions)
99279 - eZ80 (Z80 mode only)
99282 Z80: https://www.zilog.com/manage_directlink.php?filepath=docs/z80/um0080&extn=.pdf
99283 Z180: https://www.zilog.com/manage_directlink.php?filepath=docs/z180/ps0140&extn=.pdf
99284 eZ80: http://www.zilog.com/force_download.php?filepath=YUhSMGNEb3ZMM2QzZHk1NmFXeHZaeTVqYjIwdlpHOWpjeTlWVFRBd056Y3VjR1Jt
99286 To debug Z80 programs using GDB you must configure and embed
99287 z80-stub.c to your program (SDCC compiler is required). Or
99288 you may use some simulator with GDB support.
99292 * Makefile.in (ALL_TARGET_OBS): Add z80-tdep.c.
99293 * NEWS: Mention z80 support.
99294 * configure.tgt: Handle z80*.
99295 * features/Makefile (XMLTOC): Add z80.xml.
99296 * features/z80-cpu.xml: New.
99297 * features/z80.c: Generate.
99298 * features/z80.xml: New.
99299 * z80-tdep.c: New file.
99300 * z80-tdep.h: New file.
99302 gdb/stubs/ChangeLog:
99304 * z80-stub.c: New file.
99306 Change-Id: Id0b7a6e210c3f93c6853c5e3031b7bcee47d0db9
99308 2021-07-17 Simon Marchi <simon.marchi@polymtl.ca>
99310 gdb: make all_inferiors_safe actually work
99311 The test gdb.threads/fork-plus-threads.exp fails since 08bdefb58b78
99312 ("gdb: make inferior_list use intrusive_list"):
99314 FAIL: gdb.threads/fork-plus-threads.exp: detach-on-fork=off: only inferior 1 left
99316 Looking at the log, we see that we are left with a bunch of inferiors in
99317 the detach-on-fork=off case:
99320 Num Description Connection Executable ^M
99321 * 1 <null> <snip>/fork-plus-threads ^M
99322 2 <null> <snip>/fork-plus-threads ^M
99323 3 <null> <snip>/fork-plus-threads ^M
99324 4 <null> <snip>/fork-plus-threads ^M
99325 5 <null> <snip>/fork-plus-threads ^M
99326 6 <null> <snip>/fork-plus-threads ^M
99327 7 <null> <snip>/fork-plus-threads ^M
99328 8 <null> <snip>/fork-plus-threads ^M
99329 9 <null> <snip>/fork-plus-threads ^M
99330 10 <null> <snip>/fork-plus-threads ^M
99331 11 <null> <snip>/fork-plus-threads ^M
99332 (gdb) FAIL: gdb.threads/fork-plus-threads.exp: detach-on-fork=off: only inferior 1 left
99334 when we expect to have just one. The problem is prune_inferiors not
99335 pruning inferiors. And this is caused by all_inferiors_safe not
99336 actually iterating on inferiors. The current implementation:
99338 inline all_inferiors_safe_range
99339 all_inferiors_safe ()
99344 default-constructs an all_inferiors_safe_range, which default-constructs
99345 an all_inferiors_safe_iterator as its m_begin field, which
99346 default-constructs a all_inferiors_iterator. A default-constructed
99347 all_inferiors_iterator is an end iterator, which means we have
99348 constructed an (end,end) all_inferiors_safe_range.
99350 We actually need to pass down the list on which we want to iterator
99351 (that is the inferior_list global), so that all_inferiors_iterator's
99352 first constructor is chosen. We also pass nullptr as the proc_target
99353 filter. In this case, we don't do any filtering, but if in the future
99354 all_inferiors_safe needed to allow filtering on process target (like
99355 all_inferiors does), we could pass down a process target pointer.
99357 basic_safe_iterator's constructor needs to be changed to allow
99358 constructing the wrapped iterator with multiple arguments, not just one.
99360 With this, gdb.threads/fork-plus-threads.exp is passing once again for
99363 Change-Id: I650552ede596e3590c4b7606ce403690a0278a01
99365 2021-07-17 GDB Administrator <gdbadmin@sourceware.org>
99367 Automatic date update in version.in
99369 2021-07-16 Lancelot SIX <lsix@lancelotsix.com>
99371 gdb: Support stepping out from signal handler on riscv*-linux
99372 Currently, gdb cannot step outside of a signal handler on RISC-V
99373 platforms. This causes multiple failures in gdb.base/sigstep.exp:
99375 FAIL: gdb.base/sigstep.exp: continue to handler, nothing in handler, step from handler: leave handler (timeout)
99376 FAIL: gdb.base/sigstep.exp: continue to handler, si+advance in handler, step from handler: leave handler (timeout)
99377 FAIL: gdb.base/sigstep.exp: continue to handler, nothing in handler, next from handler: leave handler (timeout)
99378 FAIL: gdb.base/sigstep.exp: continue to handler, si+advance in handler, next from handler: leave handler (timeout)
99379 FAIL: gdb.base/sigstep.exp: stepi from handleri: leave signal trampoline
99380 FAIL: gdb.base/sigstep.exp: nexti from handleri: leave signal trampoline
99382 === gdb Summary ===
99384 # of expected passes 587
99385 # of unexpected failures 6
99387 This patch adds support for stepping outside of a signal handler on
99390 Implementation is heavily inspired from mips_linux_syscall_next_pc and
99391 surroundings as advised by Pedro Alves.
99393 After this patch, all tests in gdb.base/sigstep.exp pass.
99395 Build and tested on riscv64-linux-gnu.
99397 2021-07-16 Lancelot SIX <lsix@lancelotsix.com>
99399 gdb/testsuite: Declare that riscv*-*-linux* cannot hardware_single_step
99400 Many tests fail in gdb/testsuite/gdb.base/sigstep.exp on
99401 riscv64-linux-gnu. Those tests check that when stepping, if the
99402 debuggee received a signal it should step inside the signal handler.
99404 This feature requires hardware support for single stepping (or at least
99405 kernel support), but none are available on riscv*-linux-gnu hosts, at
99406 the moment at least.
99408 This patch adds RISC-V to the list of configurations that does not
99409 have hardware single step capability, disabling tests relying on such
99412 Tested on riscv64-linux-gnu.
99414 2021-07-16 Tom Tromey <tom@tromey.com>
99416 Document quick_symbol_functions::expand_symtabs_matching invariant
99417 While working on my series to replace the DWARF psymbol reader, I
99418 noticed that the expand_symtabs_matching has an undocumented
99419 invariant. I think that, if this invariant is not followed, then GDB
99420 will crash. So, this patch documents this in the relevant spots and
99421 introduces some asserts to make it clear.
99423 Regression tested on x86-64 Fedora 32.
99425 2021-07-16 Tom Tromey <tromey@adacore.com>
99427 Fix array stride bug
99428 Investigation of using the Python API with an Ada program showed that
99429 an array of dynamic types was not being handled properly. I tracked
99430 this down to an oddity of how array strides are handled.
99432 In gdb, an array stride can be attached to the range type, via the
99433 range_bounds object. However, the stride can also be put into the
99434 array's first field. From create_range_type_with_stride:
99436 else if (bit_stride > 0)
99437 TYPE_FIELD_BITSIZE (result_type, 0) = bit_stride;
99439 It's hard to be sure why this is done, but I would guess a combination
99440 of historical reasons plus a desire (mentioned in a comment somewhere)
99441 to avoid modifying the range type.
99443 This patch fixes the problem by changing type::bit_stride to
99444 understand this convention. It also fixes one spot that reproduces
99447 Regression tested on x86-64 Fedora 32.
99449 2021-07-16 Giulio Benetti <giulio.benetti@benettiengineering.com>
99451 or1k: fix pc-relative relocation against dynamic on PC relative 26 bit relocation.
99452 bfd * elf32-or1k.c (or1k_elf_relocate_section): Use a separate entry
99453 in switch case R_OR1K_INSN_REL_26 where we need to check for
99454 !SYMBOL_CALLS_LOCAL() instead of !SYMBOL_REFERENCES_LOCAL().
99456 2021-07-16 Nick Clifton <nickc@redhat.com>
99458 Updated Swedish translation for the binutils sub-directory
99460 2021-07-16 GDB Administrator <gdbadmin@sourceware.org>
99462 Automatic date update in version.in
99464 2021-07-15 Tom Tromey <tromey@adacore.com>
99466 Avoid expression parsing crash with unknown language
99467 PR gdb/28093 points out that gdb crashes when language is set to
99468 "unknown" and expression parsing is attempted. At first I thought
99469 this was a regression due to the expression rewrite, but it turns out
99470 that older versions crash as well.
99472 This patch avoids the crash by changing the default expression parser
99473 to throw an exception. I think this is preferable -- the current
99474 behavior of silently doing nothing does not really make sense.
99476 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28093
99478 2021-07-15 Simon Marchi <simon.marchi@polymtl.ca>
99480 gdb: pass child_ptid and fork kind to target_ops::follow_fork
99481 This is a small cleanup I think would be nice, that I spotted while
99482 doing the following patch.
99486 * target.h (struct target_ops) <follow_fork>: Add ptid and
99487 target_waitkind parameters.
99488 (target_follow_fork): Likewise.
99489 * target.c (default_follow_fork): Likewise.
99490 (target_follow_fork): Likewise.
99491 * fbsd-nat.h (class fbsd_nat_target) <follow_fork>: Likewise.
99492 * fbsd-nat.c (fbsd_nat_target::follow_fork): Likewise.
99493 * linux-nat.h (class linux_nat_target) <follow_fork>: Likewise.
99494 * linux-nat.c (linux_nat_target::follow_fork): Likewise.
99495 * obsd-nat.h (class obsd_nat_target) <follow_fork>: Likewise.
99496 * obsd-nat.c (obsd_nat_target::follow_fork): Likewise.
99497 * remote.c (class remote_target) <follow_fork>: Likewise.
99498 * target-debug.h (target_debug_print_target_waitkind): New.
99499 * target-delegates.c: Re-generate.
99501 Change-Id: I5421a542f2e19100a22b74cc333d2b235d0de3c8
99503 2021-07-15 Simon Marchi <simon.marchi@polymtl.ca>
99505 gdb: call post_create_inferior at end of follow_fork_inferior
99506 GDB doesn't handle well the case of an inferior using the JIT interface
99507 to register JIT-ed objfiles and forking. If an inferior registers a
99508 code object using the JIT interface and then forks, the child process
99509 conceptually has the same code object loaded, so GDB should look it up
99510 and learn about it (it currently doesn't).
99512 To achieve this, I think it would make sense to have the
99513 inferior_created observable called when an inferior is created due to a
99514 fork in follow_fork_inferior. The inferior_created observable is
99515 currently called both after starting a new inferior and after attaching
99516 to an inferior, allowing various sub-components to learn about that new
99517 executing inferior. We can see handling a fork child just like
99518 attaching to it, so any work done when attaching should also be done in
99519 the case of a fork child.
99521 Instead of just calling the inferior_created observable, this patch
99522 makes follow_fork_inferior call the whole post_create_inferior function.
99523 This way, the attach and follow-fork code code paths are more alike.
99525 Given that post_create_inferior calls solib_create_inferior_hook,
99526 follow_fork_inferior doesn't need to do it itself, so those calls to
99527 solib_create_inferior_hook are removed.
99529 One question you may have: why not just call post_create_inferior at the
99530 places where solib_create_inferior_hook is currently called, instead of
99531 after target_follow_fork?
99533 - there's something fishy for the second solib_create_inferior_hook
99534 call site: at this point we have switched the current program space
99535 to the child's, but not the current inferior nor the current thread.
99536 So solib_create_inferior_hook (and everything under, including
99537 check_for_thread_db, for example) is called with inferior 1 as the
99538 current inferior and inferior 2's program space as the current
99539 program space. I think that's wrong, because at this point we are
99540 setting up inferior 2, and all that code relies on the current
99541 inferior. We could just add a switch_to_thread call before it to
99542 make inferior 2 the current one, but there are other problems (see
99545 - solib_create_inferior_hook is currently not called on the
99546 `follow_child && detach_fork` path. I think we need to call it,
99547 because we still get a new inferior in that case (even though we
99548 detach the parent). If we only call post_create_inferior where
99549 solib_create_inferior_hook used to be called, then the JIT
99550 subcomponent doesn't get informed about the new inferior, and that
99551 introduces a failure in the new gdb.base/jit-elf-fork.exp test.
99553 - if we try to put the post_create_inferior just after the
99554 switch_to_thread that was originally at line 662, or just before the
99555 call to target_follow_fork, we introduce a subtle failure in
99556 gdb.threads/fork-thread-pending.exp. What happens then is that
99557 libthread_db gets loaded (somewhere under post_create_inferior)
99558 before the linux-nat target learns about the LWPs (which happens in
99559 linux_nat_target::follow_fork). As a result, the ALL_LWPS loop in
99560 try_thread_db_load_1 doesn't see the child LWP, and the thread-db
99561 target doesn't have the chance to fill in thread_info::priv. A bit
99562 later, when the test does "info threads", and
99563 thread_db_target::pid_to_str is called, the thread-db target doesn't
99564 recognize the thread as one of its own, and delegates the request to
99565 the target below. Because the pid_to_str output is not the expected
99566 one, the test fails.
99568 This tells me that we need to call the process target's follow_fork
99569 first, to make the process target create the necessary LWP and thread
99570 structures. Then, we can call post_create_inferior to let the other
99571 components of GDB do their thing.
99573 But then you may ask: check_for_thread_db is already called today,
99574 somewhere under solib_create_inferior_hook, and that is before
99575 target_follow_fork, why don't we see this ordering problem!? Well,
99576 because of the first bullet point: when check_for_thread_db /
99577 thread_db_load are called, the current inferior is (erroneously)
99578 inferior 1, the parent. Because libthread_db is already loaded for
99579 the parent, thread_db_load early returns. check_for_thread_db later
99580 gets called by linux_nat_target::follow_fork. At this point, the
99581 current inferior is the correct one and the child's LWP exists, so
99584 Since we now call post_create_inferior after target_follow_fork, which
99585 calls the inferior_created observable, which calls check_for_thread_db,
99586 I don't think linux_nat_target needs to explicitly call
99587 check_for_thread_db itself, so that is removed.
99589 In terms of testing, this patch adds a new gdb.base/jit-elf-fork.exp
99590 test. It makes an inferior register a JIT code object and then fork.
99591 It then verifies that whatever the detach-on-fork and follow-fork-child
99592 parameters are, GDB knows about the JIT code object in all the inferiors
99593 that survive the fork. It verifies that the inferiors can unload that
99596 There isn't currently a way to get visibility into GDB's idea of the JIT
99597 code objects for each inferior. For the purpose of this test, add the
99598 "maintenance info jit" command. There isn't much we can print about the
99599 JIT code objects except their load address. So the output looks a bit
99600 bare, but it's good enough for the test.
99604 * NEWS: Mention "maint info jit" command.
99605 * infrun.c (follow_fork_inferior): Don't call
99606 solib_create_inferior_hook, call post_create_inferior if a new
99607 inferior was created.
99608 * jit.c (maint_info_jit_cmd): New.
99609 (_initialize_jit): Register new command.
99610 * linux-nat.c (linux_nat_target::follow_fork): Don't call
99611 check_for_thread_db.
99612 * linux-nat.h (check_for_thread_db): Remove declaration.
99613 * linux-thread-db.c (check_thread_signals): Make static.
99617 * gdb.texinfo (Maintenance Commands): Mention "maint info jit".
99619 gdb/testsuite/ChangeLog:
99621 * gdb.base/jit-elf-fork-main.c: New test.
99622 * gdb.base/jit-elf-fork-solib.c: New test.
99623 * gdb.base/jit-elf-fork.exp: New test.
99625 Change-Id: I9a192e55b8a451c00e88100669283fc9ca60de5c
99627 2021-07-15 Libor Bukata <libor.bukata@oracle.com>
99629 [gdb/procfs.c] Fix build failure in find_stop_signal
99630 It fixes a regression caused by commit
99631 1edb66d856c82c389edfd7610143236a68c76846 where thread_info::suspend was
99634 The public thread_info API has to be used to get stop signal and avoid
99639 2021-07-14 Libor Bukata <libor.bukata@oracle.com>
99641 * gdb/procfs.c (find_stop_signal): Use thread_info API.
99643 Change-Id: I53bc57a05cd0eca5f28ef0726d6faeeb306e7904
99645 2021-07-15 GDB Administrator <gdbadmin@sourceware.org>
99647 Automatic date update in version.in
99649 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
99651 x86: Add int1 as one byte opcode 0xf1
99652 Also change the x86 disassembler to disassemble 0xf1 as int1, instead of
99658 * testsuite/gas/i386/opcode.s: Add int1.
99659 * testsuite/gas/i386/x86-64-opcode.s: Add int1, int3 and int.
99660 * testsuite/gas/i386/opcode-intel.d: Updated.
99661 * testsuite/gas/i386/opcode-suffix.d: Likewise.
99662 * testsuite/gas/i386/opcode.d: Likewise.
99663 * testsuite/gas/i386/x86-64-opcode.d: Likewise.
99668 * i386-dis.c (dis386): Replace icebp with int1.
99669 * i386-opc.tbl: Add int1.
99670 * i386-tbl.h: Regenerate.
99672 2021-07-14 Alan Modra <amodra@gmail.com>
99674 gas: default TC_VALIDATE_FIX_SUB to 0
99675 gas/write.c provides a fallback TC_VALIDATE_FIX_SUB define that can be
99676 a problem for some targets, the problem being that a non-zero
99677 definition of TC_VALIDATE_FIX_SUB says that some uses of fx_subsy are
99678 OK, in effect that the target will handle fx_subsy in md_apply_fix
99679 and/or tc_gen_reloc. A lot of targets don't have the necessary
99680 md_apply_fix and tc_gen_reloc support. So a safer default is to
99681 disallow fx_subsy by default.
99683 I've had a good look over target usage of fx_subsy, and think I've
99684 caught all the cases where targets need TC_VALIDATE_FIX_SUB. Possible
99685 failures would be limited to alpha, microblaze, ppc and s390 (the
99686 targets that define UNDEFINED_DIFFERENCE_OK), or targets that generate
99687 fixups with BFD_RELOC_GPREL32/16 and use a syntax explicitly showing
99688 a difference expression.
99690 * write.c (TC_VALIDATE_FIX_SUB): Default to 0.
99691 * config/tc-hppa.h (TC_VALIDATE_FIX_SUB): Define.
99692 * config/tc-microblaze.h (TC_VALIDATE_FIX_SUB): Define.
99693 * config/tc-alpha.h (TC_VALIDATE_FIX_SUB): Define for ECOFF.
99694 * config/tc-ppc.h (TC_VALIDATE_FIX_SUB): Don't define for ELF.
99695 Do define for XCOFF.
99697 2021-07-14 Clément Chigot <clement.chigot@atos.net>
99699 objdump: add DWARF support for AIX
99700 DWARF sections have special names on AIX which need be handled
99701 by objdump in order to correctly print them.
99702 This patch also adds the correlation in bfd for future uses.
99705 * libxcoff.h (struct xcoff_dwsect_name): Add DWARF name.
99706 * coff-rs6000.c (xcoff_dwsect_names): Update.
99707 * coffcode.h (sec_to_styp_flags): Likewise.
99708 (coff_new_section_hook): Likewise.
99710 * dwarf.h (struct dwarf_section): Add XCOFF name.
99711 * dwarf.c (struct dwarf_section_display): Update.
99712 * objdump.c (load_debug_section): Add XCOFF name handler.
99713 (dump_dwarf_section): Likewise.
99715 * config/tc-ppc.c (ppc_change_debug_section): Update to
99716 match new name's field.
99718 2021-07-14 Tom de Vries <tdevries@suse.de>
99720 [gdb/testsuite] Fix gdb.base/gold-gdb-index.exp
99721 When running test-case gdb.base/gold-gdb-index.exp on openSUSE Tumbleweed,
99724 FAIL: gdb.base/gold-gdb-index.exp: maint info symtabs
99727 This is due to a dummy .gdb_index:
99729 Contents of the .gdb_index section:
99742 The dummy .gdb_index is ignored when loading the symbols, and instead partial
99743 symbols are used. Consequently, we get the same result as if we'd removed
99744 -Wl,--gdb-index from the compilation.
99746 Presumably, gold fails to generate a proper .gdb_index because it lacks
99749 Anyway, without a proper .gdb_index we can't test the gdb behaviour we're
99750 trying to excercise. Fix this by detecting whether we actually used a
99751 .gdb_index for symbol loading.
99753 Tested on x86_64-linux.
99755 gdb/testsuite/ChangeLog:
99757 2021-07-14 Tom de Vries <tdevries@suse.de>
99759 * lib/gdb.exp (have_index): New proc.
99760 * gdb.base/gold-gdb-index.exp: Use have_index.
99762 2021-07-14 Tom de Vries <tdevries@suse.de>
99764 [gdb/testsuite] Add missing skip_tui_tests
99765 When building gdb with --disable-tui, we run into:
99767 (gdb) frame apply all -- -^M
99768 Undefined command: "-". Try "help".^M
99769 (gdb) ERROR: Undefined command "frame apply all -- -".
99770 UNRESOLVED: gdb.base/options.exp: test-frame-apply: frame apply all -- -
99773 Fix this by detecting whether tui is supported, and skipping the tui-related
99774 tests otherwise. Same in some gdb.tui test-cases.
99776 Tested on x86_64-linux.
99778 gdb/testsuite/ChangeLog:
99780 2021-07-13 Tom de Vries <tdevries@suse.de>
99782 * gdb.base/options.exp: Skip tui-related tests when tui is not
99784 * gdb.python/tui-window-disabled.exp: Same.
99785 * gdb.python/tui-window.exp: Same.
99787 2021-07-14 GDB Administrator <gdbadmin@sourceware.org>
99789 Automatic date update in version.in
99791 2021-07-13 Lancelot SIX <lsix@lancelotsix.com>
99793 Use /bin/sh as shebang in gdb/make-init-c
99794 While testing the NixOS[1] packaging for gdb-11.0.90.tar.xz, I got the
99801 /nix/store/26a78ync552m8j4sbjavhvkmnqir8c9y-bash-4.4-p23/bin/bash: ./make-init-c: /usr/bin/env: bad interpreter: No such file or directory
99802 make[2]: *** [Makefile:1866: stamp-init] Error 126
99803 make[2]: *** Waiting for unfinished jobs....
99804 make[2]: Leaving directory '/build/gdb-11.0.90/gdb'
99805 make[1]: *** [Makefile:9814: all-gdb] Error 2
99806 make[1]: Leaving directory '/build/gdb-11.0.90'
99807 make: *** [Makefile:903: all] Error 2
99808 builder for '/nix/store/xs8my3rrc3l4kdlbpx0azh6q0v0jxphr-gdb-gdb-11.0.90.drv' failed with exit code 2
99809 error: build of '/nix/store/xs8my3rrc3l4kdlbpx0azh6q0v0jxphr-gdb-gdb-11.0.90.drv' failed
99811 In the nix build environment, /usr/bin/env is not present, only /bin/sh
99812 is. This patch makes sure that gdb/make-init-c uses '/bin/sh' as
99813 interpreter as this is the only one available on this platform.
99815 I do not think this change will cause regressions on any other
99818 [1] https://nixos.org/
99822 * make-init-c: Use /bin/sh as shebang.
99824 2021-07-13 John Baldwin <jhb@FreeBSD.org>
99826 arm-fbsd-nat: Use fetch_register_set and store_register_set.
99828 aarch64-fbsd-nat: Use fetch_register_set and store_register_set.
99830 riscv-fbsd-nat: Use fetch_register_set and store_register_set.
99832 fbsd-nat: Add helper functions to fetch and store register sets.
99833 In particular, this supports register sets described by a regcache_map
99834 which are fetched and stored with dedicated ptrace operations. These
99835 functions are intended to be used in architecture-specific
99836 fetch_registers and store_registers target methods.
99838 Add regcache_map_supplies helper routine.
99839 This helper can be used in the fetch_registers and store_registers
99840 target methods to determine if a register set includes a specific
99843 2021-07-13 Pedro Alves <pedro@palves.net>
99845 Avoid letting exceptions escape gdb_bfd_iovec_fileio_close (PR gdb/28080)
99846 Before PR gdb/28080 was fixed by the previous patch, GDB was crashing
99850 Detaching from program: target:/any/program, process 3671843
99851 Detaching from process 3671843
99852 Ending remote debugging.
99853 [Inferior 1 (process 3671843) detached]
99855 terminate called after throwing an instance of 'gdb_exception_error'
99856 Aborted (core dumped)
99858 Here's the exception above being thrown:
99861 #0 throw_error (error=TARGET_CLOSE_ERROR, fmt=0x555556035588 "Remote connection closed") at src/gdbsupport/common-exceptions.cc:222
99862 #1 0x0000555555bbaa46 in remote_target::readchar (this=0x555556a11040, timeout=10000) at src/gdb/remote.c:9440
99863 #2 0x0000555555bbb9e5 in remote_target::getpkt_or_notif_sane_1 (this=0x555556a11040, buf=0x555556a11058, forever=0, expecting_notif=0, is_notif=0x0) at src/gdb/remote.c:9928
99864 #3 0x0000555555bbbda9 in remote_target::getpkt_sane (this=0x555556a11040, buf=0x555556a11058, forever=0) at src/gdb/remote.c:10030
99865 #4 0x0000555555bc0e75 in remote_target::remote_hostio_send_command (this=0x555556a11040, command_bytes=13, which_packet=14, remote_errno=0x7fffffffcfd0, attachment=0x0, attachment_len=0x0) at src/gdb/remote.c:12137
99866 #5 0x0000555555bc1b6c in remote_target::remote_hostio_close (this=0x555556a11040, fd=8, remote_errno=0x7fffffffcfd0) at src/gdb/remote.c:12455
99867 #6 0x0000555555bc1bb4 in remote_target::fileio_close (During symbol reading: .debug_line address at offset 0x64f417 is 0 [in module build/gdb/gdb]
99868 this=0x555556a11040, fd=8, remote_errno=0x7fffffffcfd0) at src/gdb/remote.c:12462
99869 #7 0x0000555555c9274c in target_fileio_close (fd=3, target_errno=0x7fffffffcfd0) at src/gdb/target.c:3365
99870 #8 0x000055555595a19d in gdb_bfd_iovec_fileio_close (abfd=0x555556b9f8a0, stream=0x555556b11530) at src/gdb/gdb_bfd.c:439
99871 #9 0x0000555555e09e3f in opncls_bclose (abfd=0x555556b9f8a0) at src/bfd/opncls.c:599
99872 #10 0x0000555555e0a2c7 in bfd_close_all_done (abfd=0x555556b9f8a0) at src/bfd/opncls.c:847
99873 #11 0x0000555555e0a27a in bfd_close (abfd=0x555556b9f8a0) at src/bfd/opncls.c:814
99874 #12 0x000055555595a9d3 in gdb_bfd_close_or_warn (abfd=0x555556b9f8a0) at src/gdb/gdb_bfd.c:626
99875 #13 0x000055555595ad29 in gdb_bfd_unref (abfd=0x555556b9f8a0) at src/gdb/gdb_bfd.c:715
99876 #14 0x0000555555ae4730 in objfile::~objfile (this=0x555556515540, __in_chrg=<optimized out>) at src/gdb/objfiles.c:573
99877 #15 0x0000555555ae955a in std::_Sp_counted_ptr<objfile*, (__gnu_cxx::_Lock_policy)2>::_M_dispose (this=0x555556c20db0) at /usr/include/c++/9/bits/shared_ptr_base.h:377
99878 #16 0x000055555572b7c8 in std::_Sp_counted_base<(__gnu_cxx::_Lock_policy)2>::_M_release (this=0x555556c20db0) at /usr/include/c++/9/bits/shared_ptr_base.h:155
99879 #17 0x00005555557263c3 in std::__shared_count<(__gnu_cxx::_Lock_policy)2>::~__shared_count (this=0x555556bf0588, __in_chrg=<optimized out>) at /usr/include/c++/9/bits/shared_ptr_base.h:730
99880 #18 0x0000555555ae745e in std::__shared_ptr<objfile, (__gnu_cxx::_Lock_policy)2>::~__shared_ptr (this=0x555556bf0580, __in_chrg=<optimized out>) at /usr/include/c++/9/bits/shared_ptr_base.h:1169
99881 #19 0x0000555555ae747e in std::shared_ptr<objfile>::~shared_ptr (this=0x555556bf0580, __in_chrg=<optimized out>) at /usr/include/c++/9/bits/shared_ptr.h:103
99882 #20 0x0000555555b1c1dc in __gnu_cxx::new_allocator<std::_List_node<std::shared_ptr<objfile> > >::destroy<std::shared_ptr<objfile> > (this=0x5555564cdd60, __p=0x555556bf0580) at /usr/include/c++/9/ext/new_allocator.h:153
99883 #21 0x0000555555b1bb1d in std::allocator_traits<std::allocator<std::_List_node<std::shared_ptr<objfile> > > >::destroy<std::shared_ptr<objfile> > (__a=..., __p=0x555556bf0580) at /usr/include/c++/9/bits/alloc_traits.h:497
99884 #22 0x0000555555b1b73e in std::__cxx11::list<std::shared_ptr<objfile>, std::allocator<std::shared_ptr<objfile> > >::_M_erase (this=0x5555564cdd60, __position=std::shared_ptr<objfile> (expired, weak count 1) = {get() = 0x555556515540}) at /usr/include/c++/9/bits/stl_list.h:1921
99885 #23 0x0000555555b1afeb in std::__cxx11::list<std::shared_ptr<objfile>, std::allocator<std::shared_ptr<objfile> > >::erase (this=0x5555564cdd60, __position=std::shared_ptr<objfile> (expired, weak count 1) = {get() = 0x555556515540}) at /usr/include/c++/9/bits/list.tcc:158
99886 #24 0x0000555555b19576 in program_space::remove_objfile (this=0x5555564cdd20, objfile=0x555556515540) at src/gdb/progspace.c:210
99887 #25 0x0000555555ae4502 in objfile::unlink (this=0x555556515540) at src/gdb/objfiles.c:487
99888 #26 0x0000555555ae5a12 in objfile_purge_solibs () at src/gdb/objfiles.c:875
99889 #27 0x0000555555c09686 in no_shared_libraries (ignored=0x0, from_tty=1) at src/gdb/solib.c:1236
99890 #28 0x00005555559e3f5f in detach_command (args=0x0, from_tty=1) at src/gdb/infcmd.c:2769
99894 #14 0x0000555555ae4730 in objfile::~objfile (this=0x555556515540, __in_chrg=<optimized out>) at src/gdb/objfiles.c:573
99896 That's a dtor, thus noexcept. That's the reason for the
99899 The previous patch fixed things such that the exception above isn't
99900 thrown anymore. However, it's possible that e.g., the remote
99901 connection drops just while a user types "nosharedlibrary", or some
99902 other reason that leads to objfile::~objfile, and then we end up the
99903 same std::terminate problem.
99905 Also notice that frames #9-#11 are BFD frames:
99907 #9 0x0000555555e09e3f in opncls_bclose (abfd=0x555556bc27e0) at src/bfd/opncls.c:599
99908 #10 0x0000555555e0a2c7 in bfd_close_all_done (abfd=0x555556bc27e0) at src/bfd/opncls.c:847
99909 #11 0x0000555555e0a27a in bfd_close (abfd=0x555556bc27e0) at src/bfd/opncls.c:814
99911 BFD is written in C and thus throwing exceptions over such frames may
99912 either not clean up properly, or, may abort if bfd is not compiled
99913 with -fasynchronous-unwind-tables (x86-64 defaults that on, but not
99916 Thus frame #8 seems like a good place to swallow exceptions. More so
99917 since in this spot we already ignore target_fileio_close return
99918 errors. That's what this commit does. Without the previous fix, we'd
99922 Detaching from program: target:/any/program, process 2197701
99923 Ending remote debugging.
99924 [Inferior 1 (process 2197701) detached]
99925 warning: cannot close "target:/lib64/ld-linux-x86-64.so.2": Remote connection closed
99927 Note it prints a warning, which would still be a regression compared
99928 to GDB 10, if it weren't for the previous fix.
99931 yyyy-mm-dd Pedro Alves <pedro@palves.net>
99934 * gdb_bfd.c (gdb_bfd_close_warning): New.
99935 (gdb_bfd_iovec_fileio_close): Wrap target_fileio_close in
99936 try/catch and print warning on exception.
99937 (gdb_bfd_close_or_warn): Use gdb_bfd_close_warning.
99939 Change-Id: Ic7a26ddba0a4444e3377b0e7c1c89934a84545d7
99941 2021-07-13 Pedro Alves <pedro@palves.net>
99943 Fix detach with target remote (PR gdb/28080)
99944 Commit 408f66864a1a823591b26420410c982174c239a2 ("detach in all-stop
99945 with threads running") regressed "detach" with "target remote":
99948 Detaching from program: target:/any/program, process 3671843
99949 Detaching from process 3671843
99950 Ending remote debugging.
99951 [Inferior 1 (process 3671843) detached]
99953 terminate called after throwing an instance of 'gdb_exception_error'
99954 Aborted (core dumped)
99956 Here's the exception above being thrown:
99959 #0 throw_error (error=TARGET_CLOSE_ERROR, fmt=0x555556035588 "Remote connection closed") at src/gdbsupport/common-exceptions.cc:222
99960 #1 0x0000555555bbaa46 in remote_target::readchar (this=0x555556a11040, timeout=10000) at src/gdb/remote.c:9440
99961 #2 0x0000555555bbb9e5 in remote_target::getpkt_or_notif_sane_1 (this=0x555556a11040, buf=0x555556a11058, forever=0, expecting_notif=0, is_notif=0x0) at src/gdb/remote.c:9928
99962 #3 0x0000555555bbbda9 in remote_target::getpkt_sane (this=0x555556a11040, buf=0x555556a11058, forever=0) at src/gdb/remote.c:10030
99963 #4 0x0000555555bc0e75 in remote_target::remote_hostio_send_command (this=0x555556a11040, command_bytes=13, which_packet=14, remote_errno=0x7fffffffcfd0, attachment=0x0, attachment_len=0x0) at src/gdb/remote.c:12137
99964 #5 0x0000555555bc1b6c in remote_target::remote_hostio_close (this=0x555556a11040, fd=8, remote_errno=0x7fffffffcfd0) at src/gdb/remote.c:12455
99965 #6 0x0000555555bc1bb4 in remote_target::fileio_close (During symbol reading: .debug_line address at offset 0x64f417 is 0 [in module build/gdb/gdb]
99966 this=0x555556a11040, fd=8, remote_errno=0x7fffffffcfd0) at src/gdb/remote.c:12462
99967 #7 0x0000555555c9274c in target_fileio_close (fd=3, target_errno=0x7fffffffcfd0) at src/gdb/target.c:3365
99968 #8 0x000055555595a19d in gdb_bfd_iovec_fileio_close (abfd=0x555556b9f8a0, stream=0x555556b11530) at src/gdb/gdb_bfd.c:439
99969 #9 0x0000555555e09e3f in opncls_bclose (abfd=0x555556b9f8a0) at src/bfd/opncls.c:599
99970 #10 0x0000555555e0a2c7 in bfd_close_all_done (abfd=0x555556b9f8a0) at src/bfd/opncls.c:847
99971 #11 0x0000555555e0a27a in bfd_close (abfd=0x555556b9f8a0) at src/bfd/opncls.c:814
99972 #12 0x000055555595a9d3 in gdb_bfd_close_or_warn (abfd=0x555556b9f8a0) at src/gdb/gdb_bfd.c:626
99973 #13 0x000055555595ad29 in gdb_bfd_unref (abfd=0x555556b9f8a0) at src/gdb/gdb_bfd.c:715
99974 #14 0x0000555555ae4730 in objfile::~objfile (this=0x555556515540, __in_chrg=<optimized out>) at src/gdb/objfiles.c:573
99975 #15 0x0000555555ae955a in std::_Sp_counted_ptr<objfile*, (__gnu_cxx::_Lock_policy)2>::_M_dispose (this=0x555556c20db0) at /usr/include/c++/9/bits/shared_ptr_base.h:377
99976 #16 0x000055555572b7c8 in std::_Sp_counted_base<(__gnu_cxx::_Lock_policy)2>::_M_release (this=0x555556c20db0) at /usr/include/c++/9/bits/shared_ptr_base.h:155
99977 #17 0x00005555557263c3 in std::__shared_count<(__gnu_cxx::_Lock_policy)2>::~__shared_count (this=0x555556bf0588, __in_chrg=<optimized out>) at /usr/include/c++/9/bits/shared_ptr_base.h:730
99978 #18 0x0000555555ae745e in std::__shared_ptr<objfile, (__gnu_cxx::_Lock_policy)2>::~__shared_ptr (this=0x555556bf0580, __in_chrg=<optimized out>) at /usr/include/c++/9/bits/shared_ptr_base.h:1169
99979 #19 0x0000555555ae747e in std::shared_ptr<objfile>::~shared_ptr (this=0x555556bf0580, __in_chrg=<optimized out>) at /usr/include/c++/9/bits/shared_ptr.h:103
99980 #20 0x0000555555b1c1dc in __gnu_cxx::new_allocator<std::_List_node<std::shared_ptr<objfile> > >::destroy<std::shared_ptr<objfile> > (this=0x5555564cdd60, __p=0x555556bf0580) at /usr/include/c++/9/ext/new_allocator.h:153
99981 #21 0x0000555555b1bb1d in std::allocator_traits<std::allocator<std::_List_node<std::shared_ptr<objfile> > > >::destroy<std::shared_ptr<objfile> > (__a=..., __p=0x555556bf0580) at /usr/include/c++/9/bits/alloc_traits.h:497
99982 #22 0x0000555555b1b73e in std::__cxx11::list<std::shared_ptr<objfile>, std::allocator<std::shared_ptr<objfile> > >::_M_erase (this=0x5555564cdd60, __position=std::shared_ptr<objfile> (expired, weak count 1) = {get() = 0x555556515540}) at /usr/include/c++/9/bits/stl_list.h:1921
99983 #23 0x0000555555b1afeb in std::__cxx11::list<std::shared_ptr<objfile>, std::allocator<std::shared_ptr<objfile> > >::erase (this=0x5555564cdd60, __position=std::shared_ptr<objfile> (expired, weak count 1) = {get() = 0x555556515540}) at /usr/include/c++/9/bits/list.tcc:158
99984 #24 0x0000555555b19576 in program_space::remove_objfile (this=0x5555564cdd20, objfile=0x555556515540) at src/gdb/progspace.c:210
99985 #25 0x0000555555ae4502 in objfile::unlink (this=0x555556515540) at src/gdb/objfiles.c:487
99986 #26 0x0000555555ae5a12 in objfile_purge_solibs () at src/gdb/objfiles.c:875
99987 #27 0x0000555555c09686 in no_shared_libraries (ignored=0x0, from_tty=1) at src/gdb/solib.c:1236
99988 #28 0x00005555559e3f5f in detach_command (args=0x0, from_tty=1) at src/gdb/infcmd.c:2769
99990 So frame #28 already detached the remote process, and then we're
99991 purging the shared libraries. GDB had opened remote shared libraries
99992 via the target: sysroot, so it tries closing them. GDBserver is
99993 tearing down already, so remote communication breaks down and we close
99994 the remote target and throw TARGET_CLOSE_ERROR.
99998 #14 0x0000555555ae4730 in objfile::~objfile (this=0x555556515540, __in_chrg=<optimized out>) at src/gdb/objfiles.c:573
100000 That's a dtor, thus noexcept. That's the reason for the
100003 Stepping back a bit, why do we still have open remote files if we've
100004 managed to detach already, and, we're debugging with "target remote"?
100005 The reason is that commit 408f66864a1a823591b26420410c982174c239a2
100006 makes detach_command hold a reference to the target, so the remote
100007 target won't be finally closed until frame #28 returns. It's closing
100008 the target that invalidates target file I/O handles.
100010 This commit fixes the issue by not relying on target_close to
100011 invalidate the target file I/O handles, instead invalidate them
100012 immediately in remote_unpush_target. So when GDB purges the solibs,
100013 and we end up in target_fileio_close (frame #7 above), there's nothing
100014 to do, and we don't try to talk with the remote target anymore.
100016 The regression isn't seen when testing with
100017 --target_board=native-gdbserver, because that does "set sysroot" to
100018 disable the "target:" sysroot, for test run speed reasons. So this
100019 commit adds a testcase that explicitly tests detach with "set sysroot
100023 yyyy-mm-dd Pedro Alves <pedro@palves.net>
100026 * remote.c (remote_unpush_target): Invalidate file I/O target
100028 * target.c (fileio_handles_invalidate_target): Make extern.
100029 * target.h (fileio_handles_invalidate_target): Declare.
100031 gdb/testsuite/ChangeLog:
100032 yyyy-mm-dd Pedro Alves <pedro@palves.net>
100035 * gdb.base/detach-sysroot-target.exp: New.
100036 * gdb.base/detach-sysroot-target.c: New.
100038 Reported-By: Jonah Graham <jonah@kichwacoders.com>
100040 Change-Id: I851234910172f42a1b30e731161376c344d2727d
100042 2021-07-13 Tom de Vries <tdevries@suse.de>
100044 [gdb/testsuite] Fix check-libthread-db.exp FAILs with glibc 2.33
100045 When running test-case gdb.threads/check-libthread-db.exp on openSUSE
100046 Tumbleweed with glibc 2.33, I get:
100048 (gdb) maint check libthread-db^M
100049 Running libthread_db integrity checks:^M
100050 Got thread 0x7ffff7c79b80 => 9354 => 0x7ffff7c79b80; errno = 0 ... OK^M
100051 libthread_db integrity checks passed.^M
100052 (gdb) FAIL: gdb.threads/check-libthread-db.exp: user-initiated check: \
100053 libpthread.so not initialized (pattern 2)
100056 The test-case expects instead:
100058 Got thread 0x0 => 9354 => 0x0 ... OK^M
100060 which is what I get on openSUSE Leap 15.2 with glibc 2.26, and what is
100061 described in the test-case like this:
100063 # libthread_db should fake a single thread with th_unique == NULL.
100066 Using a breakpoint on check_thread_db_callback we can compare the two
100067 scenarios, and find that in the latter case we hit this code in glibc function
100068 iterate_thread_list in nptl_db/td_ta_thr_iter.c:
100070 if (next == 0 && fake_empty)
100072 /* __pthread_initialize_minimal has not run. There is just the main
100073 thread to return. We cannot rely on its thread register. They
100074 sometimes contain garbage that would confuse us, left by the
100075 kernel at exec. So if it looks like initialization is incomplete,
100076 we only fake a special descriptor for the initial thread. */
100077 td_thrhandle_t th = { ta, 0 };
100078 return callback (&th, cbdata_p) != 0 ? TD_DBERR : TD_OK;
100081 while in the former case we don't because this preceding statement doesn't
100084 err = DB_GET_FIELD (next, ta, head, list_t, next, 0);
100087 Note that the comment mentions __pthread_initialize_minimal, but in both cases
100088 it has already run before we hit the callback, so it's possible the comment is
100091 The change in behaviour bisect to glibc commit 1daccf403b "nptl: Move stack
100092 list variables into _rtld_global", which moves the initialization of stack
100093 list variables such as __stack_user to an earlier moment, which explains well
100094 enough the observed difference.
100096 Fix this by updating the regexp patterns to agree with what libthread-db is
100099 Tested on x86_64-linux, both with glibc 2.33 and 2.26.
100101 gdb/testsuite/ChangeLog:
100103 2021-07-07 Tom de Vries <tdevries@suse.de>
100106 * gdb.threads/check-libthread-db.exp: Update patterns for glibc 2.33.
100108 2021-07-13 Felix Willgerodt <felix.willgerodt@intel.com>
100110 gdb, dwarf: Don't follow the parent of a subprogram to get a prefix.
100111 During prefix resolution, if the parent is a subprogram, there is no need
100112 to go to the parent of the subprogram. The DIE will be local.
100133 The class F2 should not be seen as a member of F1.
100154 2021-06-23 Felix Willgerodt <felix.willgerodt@intel.com>
100156 * dwarf2/read.c (determine_prefix): Return an empty prefix if the
100157 parent is a subprogram.
100159 gdb/testsuite/ChangeLog:
100160 2021-06-23 Felix Willgerodt <felix.willgerodt@intel.com>
100162 * gdb.cp/nested-class-func-class.cc: New file.
100163 * gdb.cp/nested-class-func-class.exp: New file.
100165 2021-07-13 Simon Marchi <simon.marchi@polymtl.ca>
100167 gdb: disable commit-resumed on -exec-interrupt --thread-group
100168 As reported in PR gdb/28077, we hit an internal error when using
100169 -exec-interrupt with --thread-group:
100173 ~" Id Target Id Frame \n"
100174 ~"* 1 process 403312 \"loop\" (running)\n"
100177 -exec-interrupt --thread-group i1
100178 ~"/home/simark/src/binutils-gdb/gdb/target.c:3768: internal-error: void target_stop(ptid_t): Assertion `!proc_target->commit_resumed_state' failed.\nA problem internal to GDB has been detected,\nfurther debugging may prove unreliable.\nQuit this debugging session? (y or n) "
100180 This is because this code path never disables commit-resumed (a
100181 requirement for calling target_stop, as documented in
100182 process_stratum_target::»commit_resumed_state) before calling
100185 The other 3 code paths in mi_cmd_exec_interrupt use interrupt_target_1,
100186 which does it. But the --thread-group code path uses its own thing
100187 which doesn't do it. Fix this by adding a scoped_disable_commit_resumed
100190 Calling -exec-interrupt with --thread-group is apparently not tested at
100191 the moment (which is why this bug could creep in). Add a new test for
100192 that. The test runs two inferiors and tries to interrupt them with
100193 "-exec-interrupt --thread-group X".
100195 This will need to be merged in the gdb-11-branch, so here are ChangeLog
100200 * mi/mi-main.c (mi_cmd_exec_interrupt): Use
100201 scoped_disable_commit_resumed in the --thread-group case.
100203 gdb/testsuite/ChangeLog:
100205 * gdb.mi/interrupt-thread-group.c: New.
100206 * gdb.mi/interrupt-thread-group.exp: New.
100208 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28077
100209 Change-Id: I615efefcbcaf2c15d47caf5e4b9d82854b2a2fcb
100211 2021-07-13 Nelson Chu <nelson.chu@sifive.com>
100213 RISC-V: Enable elf attributes when default configure option isn't set.
100214 Since gcc commit, 3c70b3ca1ef58f302bf8c16d9e7c7bb8626408bf, we now enable
100215 elf attributes for all riscv targets by default in gcc. Therefore, I
100216 think binutils should have the same behavior, in case users are writing
100217 assembly files. If --enable-default-riscv-attribute isn't set, then we
100218 enable the elf attributes for all riscv targets by default.
100224 * testsuite/binutils-all/readelf.s: Add comments for riscv.
100225 * testsuite/binutils-all/readelf.s-64: Likewise.
100226 * testsuite/binutils-all/readelf.s-64-unused: Likewise.
100227 * testsuite/binutils-all/readelf.ss: Likewise.
100228 * testsuite/binutils-all/readelf.ss-64: Likewise.
100229 * testsuite/binutils-all/readelf.ss-64-unused: Likewise.
100233 * configure.ac: If --enable-default-riscv-attribute isn't set,
100234 then we enable the elf attributes for all riscv targets by
100236 * configure: Regenerated.
100238 2021-07-13 John Ericson <git@JohnEricson.me>
100240 Fix some dangling references to `netbsd-tdep`
100241 These files were renamed in 1b71cfcfdc3e13a655fefa6566b5564cec044c10,
100242 but evidentially a few dangling references were left behind. This causes
100245 $ ./configure --target i686-netbsdelf
100247 make: *** No rule to make target 'nbsd-tdep.c', needed by 'nbsd-tdep.o'. Stop.
100249 2021-07-13 Simon Marchi <simon.marchi@polymtl.ca>
100251 gdb: optimize all_matching_threads_iterator
100252 all_matching_threads_iterator is used extensively in some pretty fast
100253 paths, often under the all_non_exited_threads function.
100255 If a filter target and thread-specific ptid are given, it iterates on
100256 all threads of all inferiors of that target, to ultimately yield exactly
100257 on thread. And this happens quite often, which means we unnecessarily
100258 spend time iterating on threads to find the one we are looking for. The
100259 same thing happens if an inferior-specific ptid is given, although there
100260 the iterator yields all the threads of that inferior.
100262 In those cases, the callers of all_non_exited_threads could have
100263 different behaviors depending on the kind of ptid, to avoid this
100264 inefficiency, but that would be very tedious. Using
100265 all_non_exited_threads has the advantage that one simple implementation
100266 can work seamlessly on multiple threads or on one specific thread, just
100267 by playing with the ptid.
100269 Instead, optimize all_matching_threads_iterator directly to detect these
100270 different cases and limiting what we iterate on to just what we need.
100272 - if filter_ptid is minus_one_ptid, do as we do now: filter inferiors
100273 based on filter_target, iterate on all of the matching inferiors'
100275 - if filter_ptid is a pid-only ptid (then a filter_target must
100276 necessarily be given), look up that inferior and iterate on all its
100278 - otherwise, filter_ptid is a thread-specific ptid, so look up that
100279 specific thread and "iterate" only on it
100281 For the last case, what was an iteration on all threads of the filter
100282 target now becomes a call to find_thread_ptid, which is quite efficient
100283 now thanks to inferior::ptid_thread_map.
100287 * thread-iter.h (class all_matching_threads_iterator)
100288 <all_matching_threads_iterator>: Use default.
100289 <enum class mode>: New.
100290 <m_inf, m_thr>: Initialize.
100291 <m_filter_ptid>: Remove.
100292 * thread-iter.c (all_matching_threads_iterator::m_inf_matches):
100293 Don't filter on m_filter_ptid.
100294 (all_matching_threads_iterator::all_matching_threads_iterator):
100295 Choose path based on filter_ptid (all threads, all threads of
100296 inferior, single thread).
100297 (all_matching_threads_iterator::advance): Likewise.
100299 Change-Id: Ic6a19845f5f760fa1b8eac8145793c0ff431bbc9
100301 2021-07-13 Simon Marchi <simon.marchi@polymtl.ca>
100303 gdb: maintain ptid -> thread map, optimize find_thread_ptid
100304 When debugging a large number of threads (thousands), looking up a
100305 thread by ptid_t using the inferior::thread_list linked list can add up.
100307 Add inferior::thread_map, an std::unordered_map indexed by ptid_t, and
100308 change the find_thread_ptid function to look up a thread using
100309 std::unordered_map::find, instead of iterating on all of the
100310 inferior's threads. This should make it faster to look up a thread
100313 Change-Id: I3a8da0a839e18dee5bb98b8b7dbeb7f3dfa8ae1c
100314 Co-Authored-By: Pedro Alves <pedro@palves.net>
100316 2021-07-13 Simon Marchi <simon.marchi@polymtl.ca>
100318 gdb: optimize selection of resumed thread with pending event
100319 Consider a case where many threads (thousands) keep hitting a breakpoint
100320 whose condition evaluates to false. random_pending_event_thread is
100321 responsible for selecting a thread from an inferior among all that are
100322 resumed with a pending wait status. It is currently implemented by
100323 walking the inferior's thread list twice: once to count the number of
100324 candidates and once to select a random one.
100326 Since we now maintain a per target list of resumed threads with pending
100327 event, we can implement this more efficiently by walking that list and
100328 selecting the first thread that matches the criteria
100329 (random_pending_event_thread looks for an thread from a specific
100330 inferior, and possibly a filter ptid). It will be faster especially in
100331 the common case where there isn't any resumed thread with pending
100332 event. Currently, we have to iterate the thread list to figure this
100333 out. With this patch, the list of resumed threads with pending event
100334 will be empty, so it's quick to figure out.
100336 The random selection is kept, but is moved to
100337 process_stratum_target::random_resumed_with_pending_wait_status. The
100338 same technique is used: do a first pass to count the number of
100339 candidates, and do a second pass to select a random one. But given that
100340 the list of resumed threads with pending wait statuses will generally be
100341 short, or at least shorter than the full thread list, it should be
100344 Note that this isn't completely true, in case there are multiple
100345 inferiors on the same target. Imagine that inferior A has 10k resumed
100346 threads with pending wait statuses, and random_pending_event_thread is
100347 called with inferior B. We'll need to go through the list that contains
100348 inferior A's threads to realize that inferior B has no resumed threads
100349 with pending wait status. But I think that this is a corner /
100350 pathological case. And a possible fix for this situation would be to
100351 make random_pending_event_thread work per-process-target, rather than
100354 Change-Id: I1b71d01beaa500a148b5b9797745103e13917325
100356 2021-07-13 Simon Marchi <simon.marchi@polymtl.ca>
100358 gdb: optimize check for resumed threads with pending wait status in maybe_set_commit_resumed_all_targets
100359 Consider a test case where many threads (thousands) keep hitting a
100360 breakpoint whose condition evaluates to false.
100361 maybe_set_commit_resumed_all_targets is called at each handled event,
100362 when the scoped_disable_commit_resumed object in fetch_inferior_event is
100363 reset_and_commit-ed. One particularly expensive check in there is
100364 whether the target has at least one resumed thread with a pending wait
100365 status (in which case, we don't want to commit the resumed threads, as
100366 we want to consume this status first). It is currently implemented as
100367 walking all threads of the target.
100369 Since we now maintain a per-target list of resumed threads with pending
100370 status, we can do this check efficiently, by checking whether that list
100373 Add the process_stratum_target::has_resumed_with_pending_wait_status
100374 method for this, and use it in maybe_set_commit_resumed_all_targets.
100376 Change-Id: Ia1595baa1b358338f94fc3cb3af7f27092dad5b6
100378 2021-07-13 Simon Marchi <simon.marchi@polymtl.ca>
100380 gdb: maintain per-process-target list of resumed threads with pending wait status
100381 Looking up threads that are both resumed and have a pending wait
100382 status to report is something that we do quite often in the fast path
100383 and is expensive if there are many threads, since it currently requires
100384 walking whole thread lists.
100386 The first instance is in maybe_set_commit_resumed_all_targets. This is
100387 called after handling each event in fetch_inferior_event, to see if we
100388 should ask targets to commit their resumed threads or not. If at least
100389 one thread is resumed but has a pending wait status, we don't ask the
100390 targets to commit their resumed threads, because we want to consume and
100391 handle the pending wait status first.
100393 The second instance is in random_pending_event_thread, where we want to
100394 select a random thread among all those that are resumed and have a
100395 pending wait status. This is called every time we try to consume
100396 events, to see if there are any pending events that we we want to
100397 consume, before asking the targets for more events.
100399 To allow optimizing these cases, maintain a per-process-target list of
100400 threads that are resumed and have a pending wait status.
100402 In maybe_set_commit_resumed_all_targets, we'll be able to check in O(1)
100403 if there are any such threads simply by checking whether the list is
100406 In random_pending_event_thread, we'll be able to use that list, which
100407 will be quicker than iterating the list of threads, especially when
100408 there are no resumed with pending wait status threads.
100410 About implementation details: using the new setters on class
100411 thread_info, it's relatively easy to maintain that list. Any time the
100412 "resumed" or "pending wait status" property is changed, we check whether
100413 that should cause the thread to be added or removed from the list.
100415 In set_thread_exited, we try to remove the thread from the list, because
100416 keeping an exited thread in that list would make no sense (especially if
100417 the thread is freed). My first implementation assumed that a process
100418 stratum target was always present when set_thread_exited is called.
100419 That's however, not the case: in some cases, targets unpush themselves
100420 from an inferior and then call "exit_inferior", which exits all the
100421 threads. If the target is unpushed before set_thread_exited is called
100422 on the threads, it means we could mistakenly leave some threads in the
100423 list. I tried to see how hard it would be to make it such that targets
100424 have to exit all threads before unpushing themselves from the inferior
100425 (that would seem logical to me, we don't want threads belonging to an
100426 inferior that has no process target). That seemed quite difficult and
100427 not worth the time at the moment. Instead, I changed
100428 inferior::unpush_target to remove all threads of that inferior from the
100431 As of this patch, the list is not used, this is done in the subsequent
100434 The debug messages in process-stratum-target.c need to print some ptids.
100435 However, they can't use target_pid_to_str to print them without
100436 introducing a dependency on the current inferior (the current inferior
100437 is used to get the current target stack). For debug messages, I find it
100438 clearer to print the spelled out ptid anyway (the pid, lwp and tid
100439 values). Add a ptid_t::to_string method that returns a string
100440 representation of the ptid that is meant for debug messages, a bit like
100441 we already have frame_id::to_string.
100443 Change-Id: Iad8f93db2d13984dd5aa5867db940ed1169dbb67
100445 2021-07-13 Simon Marchi <simon.marchi@polymtl.ca>
100447 gdb: make thread_info::suspend private, add getters / setters
100448 A following patch will want to take some action when a pending wait
100449 status is set on or removed from a thread. Add a getter and a setter on
100450 thread_info for the pending waitstatus, so that we can add some code in
100453 The thing is, the pending wait status field is in the
100454 thread_suspend_state, along with other fields that we need to backup
100455 before and restore after the thread does an inferior function call.
100456 Therefore, make the thread_suspend_state member private
100457 (thread_info::suspend becomes thread_info::m_suspend), and add getters /
100458 setters for all of its fields:
100465 For the pending wait status, add the additional has_pending_waitstatus
100466 and clear_pending_waitstatus methods.
100468 I think this makes the thread_info interface a bit nicer, because we
100469 now access the fields as:
100475 thread->suspend.stop_pc
100477 The stop_pc field being in the `suspend` structure is an implementation
100478 detail of thread_info that callers don't need to be aware of.
100480 For the backup / restore of the thread_suspend_state structure, add
100481 save_suspend_to and restore_suspend_from methods. You might wonder why
100482 `save_suspend_to`, as opposed to a simple getter like
100484 thread_suspend_state &suspend ();
100486 I want to make it clear that this is to be used only for backing up and
100487 restoring the suspend state, _not_ to access fields like:
100489 thread->suspend ()->stop_pc
100491 Adding some getters / setters allows adding some assertions. I find
100492 that this helps understand how things are supposed to work. Add:
100494 - When getting the pending status (pending_waitstatus method), ensure
100495 that there is a pending status.
100496 - When setting a pending status (set_pending_waitstatus method), ensure
100497 there is no pending status.
100499 There is one case I found where this wasn't true - in
100500 remote_target::process_initial_stop_replies - which needed adjustments
100501 to respect that contract. I think it's because
100502 process_initial_stop_replies is kind of (ab)using the
100503 thread_info::suspend::waitstatus to store some statuses temporarily, for
100504 its internal use (statuses it doesn't intent on leaving pending).
100506 process_initial_stop_replies pulls out stop replies received during the
100507 initial connection using target_wait. It always stores the received
100508 event in `evthread->suspend.waitstatus`. But it only sets
100509 waitstatus_pending_p, if it deems the event interesting enough to leave
100510 pending, to be reported to the core:
100512 if (ws.kind != TARGET_WAITKIND_STOPPED
100513 || ws.value.sig != GDB_SIGNAL_0)
100514 evthread->suspend.waitstatus_pending_p = 1;
100516 It later uses this flag a bit below, to choose which thread to make the
100520 && thread->suspend.waitstatus_pending_p)
100523 And ultimately that's used if the user-visible mode is all-stop, so that
100524 we print the stop for that interesting thread:
100526 /* In all-stop, we only print the status of one thread, and leave
100527 others with their status pending. */
100530 thread_info *thread = selected;
100532 thread = lowest_stopped;
100536 print_one_stopped_thread (thread);
100539 But in any case (all-stop or non-stop), print_one_stopped_thread needs
100540 to access the waitstatus value of these threads that don't have a
100541 pending waitstatus (those that had TARGET_WAITKIND_STOPPED +
100542 GDB_SIGNAL_0). This doesn't work with the assertions I've
100545 So, change the code to only set the thread's wait status if it is an
100546 interesting one that we are going to leave pending. If the thread
100547 stopped due to a non-interesting event (TARGET_WAITKIND_STOPPED +
100548 GDB_SIGNAL_0), don't store it. Adjust print_one_stopped_thread to
100549 understand that if a thread has no pending waitstatus, it's because it
100550 stopped with TARGET_WAITKIND_STOPPED + GDB_SIGNAL_0.
100552 The call to set_last_target_status also uses the pending waitstatus.
100553 However, given that the pending waitstatus for the thread may have been
100554 cleared in print_one_stopped_thread (and that there might not even be a
100555 pending waitstatus in the first place, as explained above), it is no
100556 longer possible to do it at this point. To fix that, move the call to
100557 set_last_target_status in print_one_stopped_thread. I think this will
100558 preserve the existing behavior, because set_last_target_status is
100559 currently using the current thread's wait status. And the current
100560 thread is the last one for which print_one_stopped_thread is called. So
100561 by calling set_last_target_status in print_one_stopped_thread, we'll get
100562 the same result. set_last_target_status will possibly be called
100563 multiple times, but only the last call will matter. It just means
100564 possibly more calls to set_last_target_status, but those are cheap.
100566 Change-Id: Iedab9653238eaf8231abcf0baa20145acc8b77a7
100568 2021-07-13 Simon Marchi <simon.marchi@polymtl.ca>
100570 gdb: add setter / getter for thread_info resumed state
100571 A following patch will want to do things when a thread's resumed state
100572 changes. Make the `resumed` field private (renamed to `m_resumed`) and
100573 add a getter and a setter for it. The following patch in question will
100574 therefore be able to add some code to the setter.
100576 Change-Id: I360c48cc55a036503174313261ce4e757d795319
100578 2021-07-13 Simon Marchi <simon.marchi@polymtl.ca>
100580 gdb: use intrusive list for step-over chain
100581 The threads that need a step-over are currently linked using an
100582 hand-written intrusive doubly-linked list, so that seems a very good
100583 candidate for intrusive_list, convert it.
100585 For this, we have a use case of appending a list to another one (in
100586 start_step_over). Based on the std::list and Boost APIs, add a splice
100587 method. However, only support splicing the other list at the end of the
100588 `this` list, since that's all we need.
100590 Add explicit default assignment operators to
100591 reference_to_pointer_iterator, which are otherwise implicitly deleted.
100592 This is needed because to define thread_step_over_list_safe_iterator, we
100593 wrap reference_to_pointer_iterator inside a basic_safe_iterator, and
100594 basic_safe_iterator needs to be able to copy-assign the wrapped
100595 iterator. The move-assignment operator is therefore not needed, only
100596 the copy-assignment operator is. But for completeness, add both.
100598 Change-Id: I31b2ff67c7b78251314646b31887ef1dfebe510c
100600 2021-07-13 Pedro Alves <pedro@palves.net>
100602 gdb: make inferior_list use intrusive_list
100603 Change inferior_list, the global list of inferiors, to use
100604 intrusive_list. I think most other changes are somewhat obvious
100605 fallouts from this change.
100607 There is a small change in behavior in scoped_mock_context. Before this
100608 patch, constructing a scoped_mock_context would replace the whole
100609 inferior list with only the new mock inferior. Tests using two
100610 scoped_mock_contexts therefore needed to manually link the two inferiors
100611 together, as the second scoped_mock_context would bump the first mock
100612 inferior from the thread list. With this patch, a scoped_mock_context
100613 adds its mock inferior to the inferior list on construction, and removes
100614 it on destruction. This means that tests run with mock inferiors in the
100615 inferior list in addition to any pre-existing inferiors (there is always
100616 at least one). There is no possible pid clash problem, since each
100617 scoped mock inferior uses its own process target, and pids are per
100620 Co-Authored-By: Simon Marchi <simon.marchi@efficios.com>
100621 Change-Id: I7eb6a8f867d4dcf8b8cd2dcffd118f7270756018
100623 2021-07-13 Pedro Alves <pedro@palves.net>
100625 gdb: introduce intrusive_list, make thread_info use it
100626 GDB currently has several objects that are put in a singly linked list,
100627 by having the object's type have a "next" pointer directly. For
100628 example, struct thread_info and struct inferior. Because these are
100629 simply-linked lists, and we don't keep track of a "tail" pointer, when
100630 we want to append a new element on the list, we need to walk the whole
100631 list to find the current tail. It would be nice to get rid of that
100632 walk. Removing elements from such lists also requires a walk, to find
100633 the "previous" position relative to the element being removed. To
100634 eliminate the need for that walk, we could make those lists
100635 doubly-linked, by adding a "prev" pointer alongside "next". It would be
100636 nice to avoid the boilerplate associated with maintaining such a list
100637 manually, though. That is what the new intrusive_list type addresses.
100639 With an intrusive list, it's also possible to move items out of the
100640 list without destroying them, which is interesting in our case for
100641 example for threads, when we exit them, but can't destroy them
100642 immediately. We currently keep exited threads on the thread list, but
100643 we could change that which would simplify some things.
100645 Note that with std::list, element removal is O(N). I.e., with
100646 std::list, we need to walk the list to find the iterator pointing to
100647 the position to remove. However, we could store a list iterator
100648 inside the object as soon as we put the object in the list, to address
100649 it, because std::list iterators are not invalidated when other
100650 elements are added/removed. However, if you need to put the same
100651 object in more than one list, then std::list<object> doesn't work.
100652 You need to instead use std::list<object *>, which is less efficient
100653 for requiring extra memory allocations. For an example of an object
100654 in multiple lists, see the step_over_next/step_over_prev fields in
100657 /* Step-over chain. A thread is in the step-over queue if these are
100658 non-NULL. If only a single thread is in the chain, then these
100659 fields point to self. */
100660 struct thread_info *step_over_prev = NULL;
100661 struct thread_info *step_over_next = NULL;
100663 The new intrusive_list type gives us the advantages of an intrusive
100664 linked list, while avoiding the boilerplate associated with manually
100667 intrusive_list's API follows the standard container interface, and thus
100668 std::list's interface. It is based the API of Boost's intrusive list,
100671 https://www.boost.org/doc/libs/1_73_0/doc/html/boost/intrusive/list.html
100673 Our implementation is relatively simple, while Boost's is complicated
100674 and intertwined due to a lot of customization options, which our version
100677 The easiest way to use an intrusive_list is to make the list's element
100678 type inherit from intrusive_node. This adds a prev/next pointers to
100679 the element type. However, to support putting the same object in more
100680 than one list, intrusive_list supports putting the "node" info as a
100681 field member, so you can have more than one such nodes, one per list.
100683 As a first guinea pig, this patch makes the per-inferior thread list use
100684 intrusive_list using the base class method.
100686 Unlike Boost's implementation, ours is not a circular list. An earlier
100687 version of the patch was circular: the intrusive_list type included an
100688 intrusive_list_node "head". In this design, a node contained pointers
100689 to the previous and next nodes, not the previous and next elements.
100690 This wasn't great for when debugging GDB with GDB, as it was difficult
100691 to get from a pointer to the node to a pointer to the element. With the
100692 design proposed in this patch, nodes contain pointers to the previous
100693 and next elements, making it easy to traverse the list by hand and
100696 The intrusive_list object contains pointers to the first and last
100697 elements of the list. They are nullptr if the list is empty.
100698 Each element's node contains a pointer to the previous and next
100699 elements. The first element's previous pointer is nullptr and the last
100700 element's next pointer is nullptr. Therefore, if there's a single
100701 element in the list, both its previous and next pointers are nullptr.
100702 To differentiate such an element from an element that is not linked into
100703 a list, the previous and next pointers contain a special value (-1) when
100704 the node is not linked. This is necessary to be able to reliably tell
100705 if a given node is currently linked or not.
100707 A begin() iterator points to the first item in the list. An end()
100708 iterator contains nullptr. This makes iteration until end naturally
100709 work, as advancing past the last element will make the iterator contain
100710 nullptr, making it equal to the end iterator. If the list is empty,
100711 a begin() iterator will contain nullptr from the start, and therefore be
100712 immediately equal to the end.
100714 Iterating on an intrusive_list yields references to objects (e.g.
100715 `thread_info&`). The rest of GDB currently expects iterators and ranges
100716 to yield pointers (e.g. `thread_info*`). To bridge the gap, add the
100717 reference_to_pointer_iterator type. It is used to define
100720 Add a Python pretty-printer, to help inspecting intrusive lists when
100721 debugging GDB with GDB. Here's an example of the output:
100723 (top-gdb) p current_inferior_.m_obj.thread_list
100724 $1 = intrusive list of thread_info = {0x61700002c000, 0x617000069080, 0x617000069400, 0x61700006d680, 0x61700006eb80}
100726 It's not possible with current master, but with this patch [1] that I
100727 hope will be merged eventually, it's possible to index the list and
100728 access the pretty-printed value's children:
100730 (top-gdb) p current_inferior_.m_obj.thread_list[1]
100731 $2 = (thread_info *) 0x617000069080
100732 (top-gdb) p current_inferior_.m_obj.thread_list[1].ptid
100739 Even though iterating the list in C++ yields references, the Python
100740 pretty-printer yields pointers. The reason for this is that the output
100741 of printing the thread list above would be unreadable, IMO, if each
100742 thread_info object was printed in-line, since they contain so much
100743 information. I think it's more useful to print pointers, and let the
100744 user drill down as needed.
100746 [1] https://sourceware.org/pipermail/gdb-patches/2021-April/178050.html
100748 Co-Authored-By: Simon Marchi <simon.marchi@efficios.com>
100749 Change-Id: I3412a14dc77f25876d742dab8f44e0ba7c7586c0
100751 2021-07-13 GDB Administrator <gdbadmin@sourceware.org>
100753 Automatic date update in version.in
100755 2021-07-12 Tucker <tuckkern@sourceware@gmail.com>
100757 Add the SEC_ELF_OCTETS flag to debug sections created by the assembler.
100759 gas * config/obj-elf.c (obj_elf_change_section): Set the
100760 SEF_ELF_OCTETS flag on debug sections.
100762 2021-07-12 Tom de Vries <tdevries@suse.de>
100764 [gdb/testsuite] Fix gdb.btrace/tsx.exp on system with tsx disabled in microcode
100765 Recently I started to see this fail with trunk:
100767 (gdb) record instruction-history^M
100768 1 0x00000000004004ab <main+4>: call 0x4004b7 <test>^M
100769 2 0x00000000004004c6 <test+15>: mov $0x1,%eax^M
100770 3 0x00000000004004cb <test+20>: ret ^M
100771 (gdb) FAIL: gdb.btrace/tsx.exp: speculation indication
100774 This is due to an intel microcode update (1) that disables Intel TSX by default.
100776 Fix this by updating the pattern.
100778 Tested on x86_64-linux, with both gcc 7.5.0 and clang 12.0.1.
100780 [1] https://www.intel.com/content/www/us/en/support/articles/000059422/processors.html
100782 gdb/testsuite/ChangeLog:
100784 2021-07-12 Tom de Vries <tdevries@suse.de>
100787 * gdb.btrace/tsx.exp: Add pattern for system with tsx disabled in
100790 2021-07-12 Nick Clifton <nickc@redhat.com>
100792 Updated French translation for the binutils sub-directory
100794 Fix a translation problem for the text generated by readelf at the start of a dump of a dynamic section.
100796 binutils * readelf.c (process_dynamic_section): Use ngettext to help with translation of header text.
100798 2021-07-12 Tom de Vries <tdevries@suse.de>
100800 [gdb/testsuite] Fix gdb.mi/mi-info-sources.exp for extra debug info
100801 When running test-case gdb.mi/mi-info-sources.exp, I run into:
100803 Running src/gdb/testsuite/gdb.mi/mi-info-sources.exp ...
100804 ERROR: internal buffer is full.
100806 due to extra debug info from the shared libraries.
100808 Fix this by using "nosharedlibrary".
100810 Then I run into these FAILs:
100812 FAIL: gdb.mi/mi-info-sources.exp: debug_read=false: \
100813 -file-list-exec-source-files (unexpected output)
100814 FAIL: gdb.mi/mi-info-sources.exp: debug_read=true: \
100815 -file-list-exec-source-files (unexpected output)
100816 FAIL: gdb.mi/mi-info-sources.exp: debug_read=true: \
100817 -file-list-exec-source-files --group-by-objfile, look for \
100818 mi-info-sources.c (unexpected output)
100819 FAIL: gdb.mi/mi-info-sources.exp: debug_read=true: \
100820 -file-list-exec-source-files --group-by-objfile, look for \
100821 mi-info-sources-base.c (unexpected output)
100823 due to openSUSE executables which have debug info for objects from sources
100824 like sysdeps/x86_64/crtn.S.
100826 Fix these by updating the patterns, and adding "maint expand-symtabs" to
100827 reliably get fully-read objfiles.
100829 Then I run into FAILs when using the readnow target board. Fix these by
100830 skipping the relevant tests.
100832 Then I run into FAILs when using the cc-with-gnu-debuglink board. Fix these
100833 by updating the patterns.
100835 Tested on x86_64-linux, with native, check-read1, readnow, cc-with-gdb-index,
100836 cc-with-debug-names, cc-with-gnu-debuglink, cc-with-dwz, cc-with-dwz-m.
100838 gdb/testsuite/ChangeLog:
100840 2021-07-05 Tom de Vries <tdevries@suse.de>
100842 * lib/mi-support.exp (mi_readnow): New proc.
100843 * gdb.mi/mi-info-sources.exp: Use nosharedlibrary. Update patterns.
100844 Skip tests for readnow. Use "maint expand-symtabs".
100846 2021-07-12 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
100848 testsuite: fix whitespace problems in gdb.mi/mi-break.exp
100849 Replace leading 8-spaces with tab and remove trailing space in
100852 2021-07-12 GDB Administrator <gdbadmin@sourceware.org>
100854 Automatic date update in version.in
100856 2021-07-11 GDB Administrator <gdbadmin@sourceware.org>
100858 Automatic date update in version.in
100860 2021-07-10 Alan Modra <amodra@gmail.com>
100862 Tidy commit 49910fd88dcd
100863 Pointer range checking is UB if the values compared are outside the
100864 underlying array elements (plus one).
100866 * dwarf2.c (read_address): Remove accidental commit.
100867 (read_ranges): Compare offset rather than pointers.
100869 2021-07-10 Alan Modra <amodra@gmail.com>
100871 PR28069, assertion fail in dwarf.c:display_discr_list
100872 We shouldn't be asserting on anything to do with leb128 values, or
100873 reporting file and line numbers when something unexpected happens.
100874 leb128 data is of indeterminate length, perfect for fuzzer mayhem.
100875 It would only make sense to assert or report dwarf.c/readelf.c source
100876 lines if the code had already sized and sanity checked the leb128
100879 After removing the assertions, the testcase then gave:
100881 <37> DW_AT_discr_list : 5 byte block: 0 0 0 0 0 (label 0, label 0, label 0, label 0, <corrupt>
100882 readelf: Warning: corrupt discr_list - unrecognized discriminant byte 0x5
100884 <3d> DW_AT_encoding : 0 (void)
100885 <3e> DW_AT_identifier_case: 0 (case_sensitive)
100886 <3f> DW_AT_virtuality : 0 (none)
100887 <40> DW_AT_decimal_sign: 5 (trailing separate)
100889 So the DW_AT_discr_list was showing more data than just the 5 byte
100890 block. That happened due to "end" pointing a long way past the end of
100891 block, and uvalue decrementing past zero on one of the leb128 bytes.
100894 * dwarf.c (display_discr_list): Remove assertions. Delete "end"
100895 parameter, use initial "data" pointer as the end. Formatting.
100896 Don't count down bytes as they are read.
100897 (read_and_display_attr_value): Adjust display_discr_list call.
100898 (read_and_print_leb128): Don't pass __FILE__ and __LINE__ to
100900 * dwarf.h (report_leb_status): Don't report file and line
100901 numbers. Delete file and lnum parameters,
100902 (READ_ULEB, READ_SLEB): Adjust.
100904 2021-07-10 GDB Administrator <gdbadmin@sourceware.org>
100906 Automatic date update in version.in
100908 2021-07-09 H.J. Lu <hjl.tools@gmail.com>
100910 ld/NEWS: Clarify -z [no]indirect-extern-access
100911 -z [no]indirect-extern-access are only for x86 ELF linker.
100913 2021-07-09 H.J. Lu <hjl.tools@gmail.com>
100915 elf: Limits 2 GNU_PROPERTY_1_NEEDED tests to Linux/x86
100916 Run property-1_needed-1b.d and property-1_needed-1c.d, which pass
100917 -z [no]indirect-extern-access to linker, only run for Linux/x86 targets.
100919 * testsuite/ld-elf/property-1_needed-1b.d: Only run for
100921 * testsuite/ld-elf/property-1_needed-1c.d: Likewise.
100923 2021-07-09 H.J. Lu <hjl.tools@gmail.com>
100925 elf: Add GNU_PROPERTY_1_NEEDED check
100926 If GNU_PROPERTY_1_NEEDED_INDIRECT_EXTERN_ACCESS is set on any input
100929 1. Don't generate copy relocations.
100930 2. Turn off extern_protected_data since it implies
100931 GNU_PROPERTY_NO_COPY_ON_PROTECTED.
100932 3. Treate reference to protected symbols with indirect external access
100934 4. Set GNU_PROPERTY_1_NEEDED_INDIRECT_EXTERN_ACCESS on output.
100935 5. When generating executable, clear this bit when there are non-GOT or
100936 non-PLT relocations in input relocatable files without the bit set.
100937 6. Add -z [no]indirect-extern-access to control indirect external access.
100941 * elf-bfd (elf_obj_tdata): Add has_indirect_extern_access.
100942 (elf_has_indirect_extern_access): New.
100943 * elf-properties.c (_bfd_elf_parse_gnu_properties): Set
100944 elf_has_indirect_extern_access and elf_has_no_copy_on_protected
100945 when seeing GNU_PROPERTY_1_NEEDED_INDIRECT_EXTERN_ACCESS.
100946 (elf_write_gnu_propertie): Add an argument to pass link_info.
100947 Set needed_1_p for GNU_PROPERTY_1_NEEDED in memory.
100948 (_bfd_elf_link_setup_gnu_properties): Handle
100949 GNU_PROPERTY_1_NEEDED_INDIRECT_EXTERN_ACCESS for
100950 -z indirect-extern-access. Set nocopyreloc to true and
100951 extern_protected_data to false for indirect external access.
100952 (_bfd_elf_convert_gnu_properties): Updated.
100953 * elf32-i386.c (elf_i386_check_relocs): Set
100954 non_got_ref_without_indirect_extern_access on legacy non-GOT or
100956 * elf64-x86-64.c (elf_x86_64_check_relocs): Likewise.
100957 * elflink.c (_bfd_elf_symbol_refs_local_p): Return true for
100958 STV_PROTECTED symbols with indirect external access.
100959 * elfxx-x86.c (_bfd_x86_elf_adjust_dynamic_symbol): Clear
100960 indirect_extern_access for legacy non-GOT/non-PLT references.
100961 * elfxx-x86.h (elf_x86_link_hash_entry): Add
100962 non_got_ref_without_indirect_extern_access.
100966 * bfdlink.h (bfd_link_info): Add indirect_extern_access and
100967 needed_1_p. Change nocopyreloc to int.
100971 * NEWS: Mention -z [no]indirect-extern-access
100972 * ld.texi: Document -z [no]indirect-extern-access
100973 * ldmain.c (main): Initialize link_info.indirect_extern_access
100975 * emulparams/extern_protected_data.sh: Support
100976 -z [no]indirect-extern-access.
100977 * testsuite/ld-elf/indirect-extern-access-1.rd: New file
100978 * testsuite/ld-elf/indirect-extern-access-1a.c: Likewise.
100979 * testsuite/ld-elf/indirect-extern-access-1b.c: Likewise.
100980 * testsuite/ld-elf/indirect-extern-access-2.rd: Likewise.
100981 * testsuite/ld-elf/indirect-extern-access-2a.c: Likewise.
100982 * testsuite/ld-elf/indirect-extern-access-2b.c: Likewise.
100983 * testsuite/ld-elf/indirect-extern-access-3.rd: Likewise.
100984 * testsuite/ld-elf/indirect-extern-access.S: Likewise.
100985 * testsuite/ld-elf/property-1_needed-1b.d: Likewise.
100986 * testsuite/ld-elf/property-1_needed-1c.d: Likewise.
100987 * testsuite/ld-x86-64/indirect-extern-access.rd: Likewise.
100988 * testsuite/ld-x86-64/protected-data-1.h: Likewise.
100989 * testsuite/ld-x86-64/protected-data-1a.c: Likewise.
100990 * testsuite/ld-x86-64/protected-data-1b.c: Likewise.
100991 * testsuite/ld-x86-64/protected-data-2a.S: Likewise.
100992 * testsuite/ld-x86-64/protected-data-2b.S: Likewise.
100993 * testsuite/ld-x86-64/protected-func-2a.S: Likewise.
100994 * testsuite/ld-x86-64/protected-func-2b.S: Likewise.
100995 * testsuite/ld-x86-64/protected-func-2c.c: Likewise.
100996 * testsuite/ld-elf/linux-x86.exp: Run test with
100997 GNU_PROPERTY_1_NEEDED_INDIRECT_EXTERN_ACCESS.
100998 * testsuite/ld-x86-64/x86-64.exp: Run tests for protected
100999 function and data with indirect external access.
101001 2021-07-09 H.J. Lu <hjl.tools@gmail.com>
101003 elf: Add GNU_PROPERTY_1_NEEDED
101004 Add GNU_PROPERTY_1_NEEDED:
101006 #define GNU_PROPERTY_1_NEEDED GNU_PROPERTY_UINT32_OR_LO
101008 to indicate the needed properties by the object file.
101010 Add GNU_PROPERTY_1_NEEDED_INDIRECT_EXTERN_ACCESS:
101012 #define GNU_PROPERTY_1_NEEDED_INDIRECT_EXTERN_ACCESS (1U << 0)
101014 to indicate that the object file requires canonical function pointers and
101015 cannot be used with copy relocation.
101019 * readelf.c (decode_1_needed): New.
101020 (print_gnu_property_note): Handle GNU_PROPERTY_1_NEEDED.
101024 * elf/common.h (GNU_PROPERTY_1_NEEDED): New.
101025 (GNU_PROPERTY_1_NEEDED_INDIRECT_EXTERN_ACCESS): Likewise.
101029 * testsuite/ld-elf/property-1_needed-1a.d: New file.
101030 * testsuite/ld-elf/property-1_needed-1.s: Likewise.
101032 2021-07-09 GDB Administrator <gdbadmin@sourceware.org>
101034 Automatic date update in version.in
101036 2021-07-09 Lancelot SIX <lsix@lancelotsix.com>
101038 Remove unused parameter in maybe_software_singlestep
101039 While working around, I noticed that the last parameter of
101040 maybe_software_singlestep is never used. This path removes
101043 Built on x86_64-linux-gnu and riscv64-linux-gnu.
101047 * infrun.c (maybe_software_singlestep): Remove unused PC
101049 (resume_1): Update calls to maybe_software_singlestep.
101051 2021-07-08 H.J. Lu <hjl.tools@gmail.com>
101053 x86-64: Disallow PC reloc against weak undefined symbols in PIE
101054 Disallow PC relocations against weak undefined symbols in PIE since they
101055 can lead to non-zero address at run-time.
101060 * elf64-x86-64.c (elf_x86_64_relocate_section): Disallow PC
101061 relocations against weak undefined symbols in PIE.
101066 * testsuite/ld-x86-64/pie3.d: Expect linker error.
101068 2021-07-08 H.J. Lu <hjl.tools@gmail.com>
101070 ld: Limit cache size and add --max-cache-size=SIZE
101071 When link_info.keep_memory is true, linker caches the relocation
101072 information and symbol tables of input files in memory. When there
101073 are many input files with many relocations, we may run out of memory.
101074 Add --max-cache-size=SIZE to set the maximum cache size.
101079 * bfd.c (bfd): Add alloc_size.
101080 * elf-bfd.h (_bfd_elf_link_info_read_relocs): New.
101081 * elf32-i386.c (elf_i386_check_relocs): Use _bfd_link_keep_memory.
101083 * elf64-x86-64.c (elf_x86_64_check_relocs): Likewise.
101084 * elflink.c (_bfd_elf_link_read_relocs): Renamed to ...
101085 (_bfd_elf_link_info_read_relocs): This. Update cache_size.
101086 (_bfd_elf_link_read_relocs): New.
101087 (_bfd_elf_link_check_relocs): Call _bfd_elf_link_info_read_relocs
101088 instead of _bfd_elf_link_read_relocs.
101089 (elf_link_add_object_symbols): Likewise.
101090 (elf_link_input_bfd): Likewise.
101091 (init_reloc_cookie_rels): Likewise.
101092 (init_reloc_cookie): Update cache_size. Call
101093 _bfd_elf_link_info_read_relocs instead of
101094 _bfd_elf_link_read_relocs.
101096 (elf_gc_smash_unused_vtentry_relocs): Updated. Call
101097 _bfd_elf_link_info_read_relocs instead of
101098 _bfd_elf_link_read_relocs.
101099 (bfd_elf_gc_sections): Use link_info_ok. Pass &link_info_ok
101100 to elf_gc_smash_unused_vtentry_relocs.
101101 * libbfd-in.h (_bfd_link_keep_memory): New.
101102 * linker.c (_bfd_link_keep_memory): New.
101103 * opncls.c (bfd_alloc): Update alloc_size.
101104 * bfd-in2.h: Regenerated.
101110 * bfdlink.h (bfd_link_info): Add cache_size and max_cache_size.
101115 * NEWS: Mention --max-cache-size=SIZE.
101116 * ld.texi: Document --max-cache-size=SIZE.
101117 * ldlex.h (option_values): Add OPTION_MAX_CACHE_SIZE.
101118 * ldmain.c: (main): Set link_info.max_cache_size to -1.
101119 * lexsup.c (ld_options): Add --max-cache-size=SIZE.
101120 (parse_args): Support OPTION_MAX_CACHE_SIZE.
101121 * testsuite/ld-bootstrap/bootstrap.exp: Add test for
101124 2021-07-08 Simon Marchi <simon.marchi@polymtl.ca>
101126 gdb: don't set Linux-specific displaced stepping methods in s390_gdbarch_init
101127 According to bug 28056, running an s390x binary gives:
101130 Starting program: /usr/bin/ls
101131 /home/ubuntu/tmp/gdb-11.0.90.20210705/gdb/linux-tdep.c:2550: internal-error: displaced_step_prepare_status linux_displaced_step_prepare(gdbarch*, thread_info*, CORE_ADDR&): Assertion `gdbarch_data->num_disp_step_buffers > 0' failed.
101133 This is because the s390 architecture registers some Linux-specific
101134 displaced stepping callbacks in the OS-agnostic s390_gdbarch_init:
101136 set_gdbarch_displaced_step_prepare (gdbarch, linux_displaced_step_prepare);
101137 set_gdbarch_displaced_step_finish (gdbarch, linux_displaced_step_finish);
101138 set_gdbarch_displaced_step_restore_all_in_ptid
101139 (gdbarch, linux_displaced_step_restore_all_in_ptid);
101141 But then the Linux-specific s390_linux_init_abi_any passes
101142 num_disp_step_buffers=0 to linux_init_abi:
101144 linux_init_abi (info, gdbarch, 0);
101146 The problem happens when linux_displaced_step_prepare is called for the
101147 first time. It tries to allocate the displaced stepping buffers, but
101148 sees that the number of displaced stepping buffers for that architecture
101149 is 0, which is unexpected / invalid.
101151 s390_gdbarch_init should not register the linux_* callbacks, that is
101152 expected to be done by linux_init_abi. If debugging a bare-metal s390
101153 program, or an s390 program on another OS GDB doesn't know about, we
101154 wouldn't want to use them. We would either register no callbacks, if
101155 displaced stepping isn't supported, or register a different set of
101156 callbacks if we wanted to support displaced stepping in those cases.
101158 The commit that refactored the displaced stepping machinery and
101159 introduced these set_gdbarch_displaced_step_* calls is 187b041e2514
101160 ("gdb: move displaced stepping logic to gdbarch, allow starting
101161 concurrent displaced steps"). However, even before that,
101164 set_gdbarch_displaced_step_location (gdbarch, linux_displaced_step_location);
101166 ... which already seemed wrong. The Linux-specific callback was used
101167 even for non-Linux system. Maybe that was on purpose, because it would
101168 also happen to work in some other non-Linux case, or maybe it was simply
101169 a mistake. I'll assume that this was a small mistake when
101170 s390-tdep.{h,c} where factored out of s390-linux-tdep.c, in d6e589456475
101171 ("s390: Split up s390-linux-tdep.c into two files").
101173 Fix this by removing the setting of these displaced step callbacks from
101174 s390_gdbarch_init. Instead, pass num_disp_step_buffers=1 to
101175 linux_init_abi, in s390_linux_init_abi_any. Doing so will cause
101176 linux_init_abi to register these same callbacks. It will also mean that
101177 when debugging a bare-metal s390 executable or an executable on another
101178 OS that GDB doesn't know about, gdbarch_displaced_step_prepare won't be
101179 set, so displaced stepping won't be used.
101181 This patch will need to be merged in the gdb-11-branch, since this is a
101182 GDB 11 regression, so here's the ChangeLog entry:
101186 * s390-linux-tdep.c (s390_linux_init_abi_any): Pass 1 (number
101187 of displaced stepping buffers to linux_init_abi.
101188 * s390-tdep.c (s390_gdbarch_init): Don't set the Linux-specific
101189 displaced-stepping gdbarch callbacks.
101191 Change-Id: Ieab2f8990c78fde845ce7378d6fd4ee2833800d5
101192 Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28056
101194 2021-07-08 Simon Marchi <simon.marchi@polymtl.ca>
101196 gdb/Makefile.in: remove testsuite from SUBDIRS
101197 When distclean-ing a configured / built gdb directory, like so:
101199 $ ./configure && make all-gdb && make distclean
101201 The distclean operation fails with:
101203 Missing testsuite/Makefile
101205 If we look at the SUBDIRS variable in the generated gdb/Makefile,
101206 testsuite is there twice:
101208 SUBDIRS = doc testsuite data-directory testsuite
101210 So we try distclean-ing the testsuite directory twice. The second time,
101211 gdb/testsuite/Makefile doesn't exist, so it fails.
101213 The first "testsuite" comes from the @subdirs@ replacement, because of
101214 the `AC_CONFIG_SUBDIRS` macro in gdb/configure.ac. The second one is
101215 hard-coded in gdb/Makefile.in:
101217 SUBDIRS = doc @subdirs@ data-directory testsuite
101219 The hard-coded was added by:
101221 bdbbcd577460 ("Always build 'all' in gdb/testsuite")
101223 which came after `testsuite` was removed from @subdirs@ by:
101225 f99d1d37496f ("Remove gdb/testsuite/configure")
101227 My commit a100a94530eb ("gdb/testsuite: restore configure script")
101228 should have removed the hard-coded `testsuite`, since it added it back
101229 as a "subdir", but I missed it because I only looked f99d1d37496f to
101232 Fix this by removing the hard-coded one.
101234 This patch should be pushed to both master and gdb-11-branch, hence the
101239 * Makefile.in (SUBDIRS): Remove testsuite.
101241 Change-Id: I63e5590b1a08673c646510b3ecc74600eae9f92d
101243 2021-07-08 Nick Clifton <nickc@redhat.com>
101245 Updated Portuguese translation for the BFD sub-directory
101247 2021-07-08 Tom de Vries <tdevries@suse.de>
101249 [gdb/testsuite] Fix gdb.guile/scm-breakpoint.exp with guile 3.0
101250 When running test-case gdb.guile/scm-breakpoint.exp on openSUSE Tumbleweed
101251 with guile 3.0, I run into:
101253 (gdb) guile (define cp (make-breakpoint "syscall" #:type BP_CATCHPOINT))^M
101254 ERROR: In procedure make-breakpoint:^M
101255 In procedure gdbscm_make_breakpoint: unsupported breakpoint type in \
101256 position 3: "BP_CATCHPOINT"^M
101257 Error while executing Scheme code.^M
101258 (gdb) FAIL: gdb.guile/scm-breakpoint.exp: test_catchpoints: \
101259 create a catchpoint via the api
101262 The same test passes on openSUSE Leap 15.2 with guile 2.0, where the second
101263 line of the error message starts with the same prefix as the first:
101265 ERROR: In procedure gdbscm_make_breakpoint: unsupported breakpoint type in \
101266 position 3: "BP_CATCHPOINT"^M
101269 I observe the same difference in many other tests, f.i.:
101271 (gdb) gu (print (value-add i '()))^M
101272 ERROR: In procedure value-add:^M
101273 In procedure gdbscm_value_add: Wrong type argument in position 2: ()^M
101274 Error while executing Scheme code.^M
101275 (gdb) PASS: gdb.guile/scm-math.exp: catch error in guile type conversion
101277 but it doesn't cause FAILs anywhere else.
101279 Fix this by updating the regexp to make the "ERROR: " prefix optional.
101281 Tested on x86_64-linux, with both guile 2.0 and 3.0.
101283 gdb/testsuite/ChangeLog:
101285 2021-07-07 Tom de Vries <tdevries@suse.de>
101287 * gdb.guile/scm-breakpoint.exp: Make additional "ERROR: " prefix in
101288 exception printing optional.
101290 2021-07-08 Mike Frysinger <vapier@gentoo.org>
101292 sim: erc32: use libsim.a for common objects
101293 We're starting to move more objects to the common build that sis did
101294 not need before, so linking them is causing problems (when common
101295 objects end up needing symbols from non-common objects). Switch it
101296 to the libsim.a archive which will allow the link to pull out only
101299 2021-07-08 GDB Administrator <gdbadmin@sourceware.org>
101301 Automatic date update in version.in
101303 2021-07-07 Nick Clifton <nickc@redhat.com>
101305 Remove an accidental change to elfcode.h included as part of commit 6e0dfbf420.
101307 * elfcode.h (elf_swap_symbol_out): Revert accidental change that
101308 removed an abort if the shndx pointer is NULL.
101310 2021-07-07 H.J. Lu <hjl.tools@gmail.com>
101312 ld: Check archive only for archive member
101313 Since plugin_maybe_claim calls bfd_close on the original input BFD if it
101314 isn't an archive member, pass NULL to bfd_plugin_close_file_descriptor
101315 to indicate that the BFD isn't an archive member.
101320 * plugin.c (bfd_plugin_close_file_descriptor): Check archive
101322 (try_claim): Pass NULL to bfd_plugin_close_file_descriptor if
101323 it isn't an archive member.
101328 * plugin.c (plugin_input_file): Add comments for abfd and ibfd.
101329 (plugin_object_p): Set input->ibfd to NULL if it isn't an
101332 2021-07-07 Andreas Krebbel <krebbel@linux.ibm.com>
101334 Add changelog entries for last commit
101336 2021-07-07 Andreas Krebbel <krebbel@linux.ibm.com>
101338 IBM Z: Add another arch14 instruction
101341 * opcodes/s390-opc.txt: Add qpaci.
101345 * testsuite/gas/s390/zarch-arch14.d: Add qpaci.
101346 * testsuite/gas/s390/zarch-arch14.s: Add qpaci.
101348 2021-07-07 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
101350 Fix Solaris gprof build with --disable-nls
101351 gprof fails to compile on Solaris 10 and 11.3 with --disable-nls:
101353 In file included from /vol/src/gnu/binutils/hg/binutils-2.37-branch/git/gprof/gprof.h:33,
101354 from /vol/src/gnu/binutils/hg/binutils-2.37-branch/git/gprof/basic_blocks.c:24:
101355 /usr/include/libintl.h:45:14: error: expected identifier or '(' before 'const'
101356 45 | extern char *dcgettext(const char *, const char *, const int);
101358 /usr/include/libintl.h:46:14: error: expected identifier or '(' before 'const'
101359 46 | extern char *dgettext(const char *, const char *);
101361 /usr/include/libintl.h:47:14: error: expected identifier or '(' before 'const'
101362 47 | extern char *gettext(const char *);
101364 /vol/src/gnu/binutils/hg/binutils-2.37-branch/git/gprof/../bfd/sysdep.h:165:33:
101365 error: expected identifier or '(' before 'do'
101366 165 | # define textdomain(Domainname) do {} while (0)
101368 /vol/src/gnu/binutils/hg/binutils-2.37-branch/git/gprof/../bfd/sysdep.h:165:39:
101369 error: expected identifier or '(' before 'while'
101370 165 | # define textdomain(Domainname) do {} while (0)
101372 /vol/src/gnu/binutils/hg/binutils-2.37-branch/git/gprof/../bfd/sysdep.h:166:46:
101373 error: expected identifier or '(' before 'do'
101374 166 | # define bindtextdomain(Domainname, Dirname) do {} while (0)
101376 /vol/src/gnu/binutils/hg/binutils-2.37-branch/git/gprof/../bfd/sysdep.h:166:52:
101377 error: expected identifier or '(' before 'while'
101378 166 | # define bindtextdomain(Domainname, Dirname) do {} while (0)
101380 /usr/include/libintl.h:55:14: error: expected identifier or '(' before 'unsigned'
101381 55 | extern char *dcngettext(const char *, const char *,
101383 /usr/include/libintl.h:57:14: error: expected identifier or '(' before 'unsigned'
101384 57 | extern char *dngettext(const char *, const char *,
101386 /usr/include/libintl.h:59:14: error: expected identifier or '(' before 'unsigned'
101387 59 | extern char *ngettext(const char *, const char *, unsigned long int);
101390 This is a known issue already partially fixed in binutils/sysdep.h. For
101391 gprof, the same fix needs to be applied in bfd/sysdep.h, as the
101392 following patch does. Tested on i386-pc-solaris2.10 and
101395 2021-07-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
101398 * sysdep.h [!ENABLE_NLS]: Prevent inclusion of <libintl.h> on
101401 2021-07-07 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
101403 Check for strnlen declaration to fix Solaris 10 build
101404 binutils currently fails to compile on Solaris 10:
101406 /vol/src/gnu/binutils/hg/binutils-2.37-branch/git/bfd/opncls.c: In function 'bfd_get_debug_link_info_1':
101407 /vol/src/gnu/binutils/hg/binutils-2.37-branch/git/bfd/opncls.c:1231:16: error: implicit declaration of function 'strnlen' [-Werror=implicit-function-declaration]
101408 1231 | crc_offset = strnlen (name, size) + 1;
101410 /vol/src/gnu/binutils/hg/binutils-2.37-branch/git/bfd/opncls.c:1231:16: error: incompatible implicit declaration of built-in function 'strnlen' [-Werror]
101411 /vol/src/gnu/binutils/hg/binutils-2.37-branch/git/bfd/opncls.c: In function 'bfd_get_alt_debug_link_info':
101412 /vol/src/gnu/binutils/hg/binutils-2.37-branch/git/bfd/opncls.c:1319:20: error: incompatible implicit declaration of built-in function 'strnlen' [-Werror]
101413 1319 | buildid_offset = strnlen (name, size) + 1;
101416 and in a couple of other places. The platform lacks strnlen, and while
101417 libiberty.h can provide a fallback declaration, the necessary configure
101420 Fixed with the following patch. Tested on i386-pc-solaris2.10.
101422 2021-07-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
101425 * configure.ac: Check for strnlen declaration.
101426 * configure, config.in: Regenerate.
101429 * configure.ac: Check for strnlen declaration.
101430 * configure, config.in: Regenerate.
101432 2021-07-07 Nick Clifton <nickc@redhat.com>
101434 Fix problems translating messages when a percentage sign appears at the end of a string.
101436 gas * config/tc-i386.c (offset_in_range): Reformat error messages in
101437 order to fix problems when translating.
101438 (md_assemble): Likewise.
101439 * messages.c (as_internal_value_out_of_range): Likewise.
101440 * read.c (emit_expr_with_reloc): Likewise.
101441 * testsuite/gas/all/overflow.l Change expected output format.
101442 * po/gas.pot: Regenerate.
101444 bfd * coff-rs6000.c (xcoff_reloc_type_tls): Reformat error messages in
101445 order to fix problems when translating.
101446 * cofflink.c (_bfd_coff_write_global_sym): Likewise.
101447 * elfnn-aarch64.c (_bfd_aarch64_erratum_843419_branch_to_stub):
101449 * po/bfd.pot: Regenerate.
101451 2021-07-07 GDB Administrator <gdbadmin@sourceware.org>
101453 Automatic date update in version.in
101455 2021-07-06 Simon Marchi <simon.marchi@polymtl.ca>
101457 gdb: introduce iterator_range, remove next_adapter
101458 I was always a bit confused by next_adapter, because it kind of mixes
101459 the element type and the iterator type. In reality, it is not much more
101460 than a class that wraps two iterators (begin and end). However, it
101463 - you can construct the begin iterator by passing a pointer to the
101464 first element of the iterable
101465 - you can default-construct iterator to make the end iterator
101467 I think that by generalizing it a little bit, we can re-use it at more
101470 Rename it to "iterator_range". I think it describes a bit better: it's
101471 a range made by wrapping a begin and end iterator. Move it to its own
101472 file, since it's not related to next_iterator anymore.
101474 iterator_range has two constructors. The variadic one, where arguments
101475 are forwarded to construct the underlying begin iterator. The end
101476 iterator is constructed through default construction. This is a
101477 generalization of what we have today.
101479 There is another constructor which receives already constructed begin
101480 and end iterators, useful if the end iterator can't be obtained by
101481 default-construction. Or, if you wanted to make a range that does not
101482 end at the end of the container, you could pass any iterator as the
101485 This generalization allows removing some "range" classes, like
101486 all_inferiors_range. These classes existed only to pass some arguments
101487 when constructing the begin iterator. With iterator_range, those same
101488 arguments are passed to the iterator_range constructed and then
101489 forwarded to the constructed begin iterator.
101491 There is a small functional difference in how iterator_range works
101492 compared to next_adapter. next_adapter stored the pointer it received
101493 as argument and constructeur an iterator in the `begin` method.
101494 iterator_range constructs the begin iterator and stores it as a member.
101495 Its `begin` method returns a copy of that iterator.
101497 With just iterator_range, uses of next_adapter<foo> would be replaced
101500 using foo_iterator = next_iterator<foo>;
101501 using foo_range = iterator_range<foo_iterator>;
101503 However, I added a `next_range` wrapper as a direct replacement for
101504 next_adapter<foo>. IMO, next_range is a slightly better name than
101507 The rest of the changes are applications of this new class.
101511 * next-iterator.h (class next_adapter): Remove.
101512 * iterator-range.h: New.
101516 * breakpoint.h (bp_locations_range): Remove.
101517 (bp_location_range): New.
101518 (struct breakpoint) <locations>: Adjust type.
101519 (breakpoint_range): Use iterator_range.
101520 (tracepoint_range): Use iterator_range.
101521 * breakpoint.c (breakpoint::locations): Adjust return type.
101522 * gdb_bfd.h (gdb_bfd_section_range): Use iterator_range.
101523 * gdbthread.h (all_threads_safe): Pass argument to
101524 all_threads_safe_range.
101525 * inferior-iter.h (all_inferiors_range): Use iterator_range.
101526 (all_inferiors_safe_range): Use iterator_range.
101527 (all_non_exited_inferiors_range): Use iterator_range.
101528 * inferior.h (all_inferiors, all_non_exited_inferiors): Pass
101529 inferior_list as argument.
101530 * objfiles.h (struct objfile) <compunits_range>: Remove.
101531 <compunits>: Return compunit_symtab_range.
101532 * progspace.h (unwrapping_objfile_iterator)
101533 <unwrapping_objfile_iterator>: Take parameter by value.
101534 (unwrapping_objfile_range): Use iterator_range.
101535 (struct program_space) <objfiles_range>: Define with "using".
101537 <objfiles_safe_range>: Define with "using".
101538 <objfiles_safe>: Adjust.
101539 <solibs>: Return so_list_range, define here.
101540 * progspace.c (program_space::solibs): Remove.
101541 * psymtab.h (class psymtab_storage) <partial_symtab_iterator>:
101543 <partial_symtab_range>: Use iterator_range.
101544 * solist.h (so_list_range): New.
101545 * symtab.h (compunit_symtab_range):
101548 (compunit_filetabs): Change to a function.
101549 * thread-iter.h (inf_threads_range,
101550 inf_non_exited_threads_range, safe_inf_threads_range,
101551 all_threads_safe_range): Use iterator_range.
101552 * top.h (ui_range): New.
101553 (all_uis): Use ui_range.
101555 Change-Id: Ib7a9d2a3547f45f01aa1c6b24536ba159db9b854
101557 2021-07-06 Simon Marchi <simon.marchi@polymtl.ca>
101559 gdb/testsuite: restore configure script
101560 Commit f99d1d37496f ("Remove gdb/testsuite/configure") removed
101561 gdb/testsuite/configure, as anything gdb/testsuite/configure did could
101562 be done by gdb/configure.
101564 There is however one use case that popped up when this changed
101565 propagated to downstream consumers, to run the testsuite on an already
101566 built GDB. In the workflow of ROCm-GDB at AMD, a GDB package is built
101567 in a CI job. This GDB package is then tested on different machines /
101568 hardware configurations as part of other CI jobs. To achieve this,
101569 those CI jobs only configure the testsuite directory and run "make
101570 check" with an appropriate board file.
101572 In light of this use case, the way I see it is that gdb/testsuite could
101573 be considered its own project. It could be stored in a completely
101574 different repo if we want to, it just happens to be stored inside gdb/.
101576 Since the only downside of having gdb/testsuite/configure is that it
101577 takes a few more seconds to run, but on the other hand it's quite useful
101578 for some people, I propose re-adding it.
101580 In a sense, this is revert of f99d1d37496f, but it's not a direct
101581 git-revert, as some things have changed since.
101585 * configure.ac: Remove things that were moved from
101586 testsuite/configure.ac.
101587 * configure: Re-generate.
101589 gdb/testsuite/ChangeLog:
101591 * configure.ac: Restore.
101592 * configure: Re-generate.
101593 * aclocal.m4: Re-generate.
101594 * Makefile.in (distclean): Add config.status.
101595 (Makefile): Adjust paths.
101596 (lib/pdtrace): Adjust paths.
101599 Change-Id: Ic38c79485e1835712d9c99649c9dfb59667254f1
101601 2021-07-06 Joel Brobecker <brobecker@adacore.com>
101603 Rename gdb/ChangeLog to gdb/ChangeLog-2021
101604 Now that ChangeLog entries are no longer used for GDB patches,
101605 this commit renames the file gdb/ChangeLog to gdb/ChangeLog-2021,
101606 similar to what we would do in the context of the "Start of New
101609 The purpose of this change is to avoid people merging ChangeLog
101610 entries by mistake when applying existing commits that they are
101613 2021-07-06 Dan Streetman <ddstreet@canonical.com>
101615 sim: ppc: add missing empty targets
101616 These are copied from sim/common/Make-common.in.
101618 On ppc the build fails without at least the 'info' target, e.g.:
101621 make[4]: Entering directory '/<<BUILDDIR>>/gdb-10.2.2974.g5b45e89f56d+21.10.20210510155809/build/default/sim/ppc'
101622 make[4]: *** No rule to make target 'info'. Stop.
101624 2021-07-06 Yuri Chornoivan <yurchor@ukr.net>
101626 PR 28053: Fix spelling mistakes: usupported -> unsupported and relocatation -> relocation.
101628 2021-07-06 Michael Matz <matz@suse.de>
101630 elf/riscv: Fix relaxation with aliases [PR28021]
101631 the fix for PR22756 only changed behaviour for hidden aliases,
101632 but the same situation exists for non-hidden aliases: sym_hashes[]
101633 can contain multiple entries pointing to the same symbol structure
101634 leading to relaxation adjustment to be applied twice.
101636 Fix this by testing for duplicates for everything that looks like it
101642 * elfnn-riscv.c (riscv_relax_delete_bytes): Check for any
101646 * testsuite/ld-riscv-elf/relax-twice.ver: New.
101647 * testsuite/ld-riscv-elf/relax-twice-1.s: New.
101648 * testsuite/ld-riscv-elf/relax-twice-2.s: New.
101649 * testsuite/ld-riscv-elf/ld-riscv-elf.exp
101650 (run_relax_twice_test): New, and call it.
101652 2021-07-06 Pedro Alves <pedro@palves.net>
101653 Qingchuan Shi <qingchuan.shi@amd.com>
101655 Update gdb performance testsuite to be compatible with Python 3.8
101656 Running "make check-perf" on a system with Python 3.8 (e.g., Ubuntu
101657 20.04) runs into this Python problem:
101659 Traceback (most recent call last):
101660 File "<string>", line 1, in <module>
101661 File "/home/pedro/rocm/gdb/src/gdb/testsuite/gdb.perf/lib/perftest/perftest.py", line 65, in run
101663 File "<string>", line 35, in execute_test
101664 File "/home/pedro/rocm/gdb/src/gdb/testsuite/gdb.perf/lib/perftest/measure.py", line 45, in measure
101666 File "/home/pedro/rocm/gdb/src/gdb/testsuite/gdb.perf/lib/perftest/measure.py", line 102, in start
101667 self.start_time = time.clock()
101668 AttributeError: module 'time' has no attribute 'clock'
101669 Error while executing Python code.
101670 (gdb) FAIL: gdb.perf/single-step.exp: python SingleStep(1000).run()
101674 The problem is that the testsuite is using time.clock(), deprecated in
101675 Python 3.3 and finaly removed in Python 3.8. The guidelines say to
101676 use time.perf_counter() or time.process_time() instead depending on
101677 requirements. Looking at the current description of those functions,
101680 https://docs.python.org/3.10/library/time.html
101684 time.perf_counter() -> float
101686 Return the value (in fractional seconds) of a performance
101687 counter, i.e. a clock with the highest available resolution to
101688 measure a short duration. It does include time elapsed during
101689 sleep and is system-wide. (...)
101691 time.process_time() -> float
101693 Return the value (in fractional seconds) of the sum of the
101694 system and user CPU time of the current process. It does not
101695 include time elapsed during sleep. It is process-wide by
101698 I'm thinking that it's just best to record both instead of picking
101699 one. So this patch replaces the MeasurementCpuTime measurement class
101700 with two new classes -- MeasurementPerfCounter and
101701 MeasurementProcessTime. Correspondingly, this changes the reports in
101702 testsuite/perftest.log -- we have two new "perf_counter" and
101703 "process_time" measurements and the "cpu_time" measurement is gone. I
101704 don't suppose breaking backward compatibility here is a big problem.
101705 I suspect no one is really tracking long term performance using the
101706 perf testsuite today. And if they are, it shouldn't be hard to adjust.
101708 For backward compatility, with Python < 3.3, both perf_counter and
101709 process_time use the old time.clock.
101711 gdb/testsuite/ChangeLog:
101712 yyyy-mm-dd Qingchuan Shi <qingchuan.shi@amd.com>
101713 Pedro Alves <pedro@palves.net>
101715 * gdb.perf/lib/perftest/perftest.py: Import sys.
101716 (time.perf_counter, time.process_time): Map to time.clock on
101718 (MeasurementCpuTime): Delete, replaced by...
101719 (MeasurementPerfCounter, MeasurementProcessTime): .. these two new
101721 * gdb.perf/lib/perftest/perftest.py: Import MeasurementPerfCounter
101722 and MeasurementProcessTime instead of MeasurementCpuTime.
101723 (TestCaseWithBasicMeasurements): Use MeasurementPerfCounter and
101724 MeasurementProcessTime instead of MeasurementCpuTime.
101727 Change-Id: Ia850c05d5ce57d2dada70ba5b0061f566444aa2b
101729 2021-07-06 Pedro Alves <pedro@palves.net>
101731 gdb.perf/: FAIL on Python errors, avoid "ERROR: internal buffer is full"
101732 Currently, if you run make check-perf on a system with Python 3.8,
101733 tests seen to PASS, but they actually test a lot less than intended,
101736 PerfTest::assemble, run ...
101737 python BackTrace(64).run()
101738 Traceback (most recent call last):
101739 File "<string>", line 1, in <module>
101740 File "/home/pedro/rocm/gdb/src/gdb/testsuite/gdb.perf/lib/perftest/perftest.py", line 65, in run
101742 File "<string>", line 49, in execute_test
101743 File "/home/pedro/rocm/gdb/src/gdb/testsuite/gdb.perf/lib/perftest/measure.py", line 45, in measure
101745 File "/home/pedro/rocm/gdb/src/gdb/testsuite/gdb.perf/lib/perftest/measure.py", line 102, in start
101746 self.start_time = time.clock()
101747 AttributeError: module 'time' has no attribute 'clock'
101748 Error while executing Python code.
101749 (gdb) PASS: gdb.perf/backtrace.exp: python BackTrace(64).run()
101751 And then, after fixing the above Python compatibility issues (which
101752 will be a separate patch), I get 86 instances of overflowing expect's
101755 ERROR: internal buffer is full.
101756 UNRESOLVED: gdb.perf/single-step.exp: python SingleStep(1000).run()
101758 This patch fixes both problems by adding & using a gdb_test_python_run
101761 - checks for Python errors
101762 - consumes output line by line
101764 gdb/testsuite/ChangeLog:
101765 yyyy-mm-dd Pedro Alves <pedro@palves.net>
101767 * gdb.perf/backtrace.exp: Use gdb_test_python_run.
101768 * gdb.perf/disassemble.exp: Use gdb_test_python_run.
101769 * gdb.perf/single-step.exp: Use gdb_test_python_run.
101770 * gdb.perf/skip-command.exp: Use gdb_test_python_run.
101771 * gdb.perf/skip-prologue.exp: Use gdb_test_python_run.
101772 * gdb.perf/solib.exp: Use gdb_test_python_run.
101773 * gdb.perf/template-breakpoints.exp: Use gdb_test_python_run.
101774 * lib/perftest.exp (gdb_test_python_run): New.
101776 Change-Id: I007af36f164b3f4cda41033616eaaa4e268dfd2f
101778 2021-07-06 Tom de Vries <tdevries@suse.de>
101780 [gdb/testsuite] Remove read1 timeout factor from gdb.base/info-macros.exp
101781 At the moment some check-read1 timeouts are handled like this in
101782 gdb.base/info-macros.exp:
101784 gdb_test_multiple_with_read1_timeout_factor 10 "$test" $testname {
101788 -re ".*#define TWO.*\r\n$gdb_prompt" {
101791 -re ".*#define THREE.*\r\n$gdb_prompt" {
101794 -re ".*#define FOUR.*\r\n$gdb_prompt" {
101801 We could use gdb_test_lines, but it currently doesn't support verifying
101802 the absence of regexps, which is done using the clauses above calling fail.
101804 Fix this by using gdb_test_lines and adding a -re-not syntax to
101805 gdb_test_lines, such that we can do:
101807 gdb_test_lines $test $testname $r1.*$r2 \
101808 -re-not "#define TWO" \
101809 -re-not "#define THREE" \
101813 Tested on x86_64-linux, whith make targets check and check-read1.
101815 Also observed that check-read1 execution time is reduced from 6m35s to 13s.
101817 gdb/testsuite/ChangeLog:
101819 2021-07-06 Tom de Vries <tdevries@suse.de>
101821 * gdb.base/info-macros.exp: Replace use of
101822 gdb_test_multiple_with_read1_timeout_factor with gdb_test_lines.
101823 (gdb_test_multiple_with_read1_timeout_factor): Remove.
101824 * lib/gdb.exp (gdb_test_lines): Add handling or -re-not <regexp>.
101826 2021-07-06 Nelson Chu <nelson.chu@sifive.com>
101828 RISC-V: Fix the build broken with -Werror.
101833 * elfnn-riscv.c(riscv_elf_additional_program_headers): Removed the
101835 (riscv_elf_modify_segment_map): Added ATTRIBUTE_UNUSED for the
101838 2021-07-06 Tom de Vries <tdevries@suse.de>
101840 [gdb/symtab] Fix skipping of import of C++ CU
101841 Tom Tromey observed that when changing the language in
101842 gdb.dwarf2/imported-unit-bp.exp from c to c++, the test failed.
101844 This is due to this code in process_imported_unit_die:
101846 /* We're importing a C++ compilation unit with tag DW_TAG_compile_unit
101847 into another compilation unit, at root level. Regard this as a hint,
101849 if (die->parent && die->parent->parent == NULL
101850 && per_cu->unit_type == DW_UT_compile
101851 && per_cu->lang == language_cplus)
101854 which should have a partial symtabs counterpart.
101856 Add the missing counterpart in process_psymtab_comp_unit.
101858 Tested on x86_64-linux (openSUSE Leap 15.2), no regressions for config:
101859 - using default gcc version 7.5.0
101860 (with 5 unexpected FAILs)
101861 - gcc 10.3.0 and target board
101862 unix/-flto/-O0/-flto-partition=none/-ffat-lto-objects
101863 (with 1000 unexpected FAILs)
101867 2021-07-06 Tom de Vries <tdevries@suse.de>
101869 * dwarf2/read.c (scan_partial_symbols): Skip top-level imports of
101871 * testsuite/gdb.dwarf2/imported-unit-bp.exp: Moved to ...
101872 * testsuite/gdb.dwarf2/imported-unit-bp.exp.tcl: ... here.
101873 * testsuite/gdb.dwarf2/imported-unit-bp-c++.exp: New test.
101874 * testsuite/gdb.dwarf2/imported-unit-bp-c.exp: New test.
101875 * testsuite/gdb.dwarf2/imported-unit.exp: Update.
101877 2021-07-06 Kito Cheng <kito.cheng@sifive.com>
101879 RISC-V: Add PT_RISCV_ATTRIBUTES and add it to PHDR.
101880 We added PT_RISCV_ATTRIBUTES to program header to make
101881 .riscv.attribute easier to find in dynamic loader or kernel.
101884 https://github.com/riscv/riscv-elf-psabi-doc/pull/71
101890 * elfnn-riscv.c(RISCV_ATTRIBUTES_SECTION_NAME): New.
101891 (riscv_elf_additional_program_headers): Ditto.
101892 (riscv_elf_modify_segment_map): Ditto.
101893 (elf_backend_additional_program_headers): Ditto.
101894 (elf_backend_modify_segment_map): Ditto.
101895 (elf_backend_obj_attrs_section): Use RISCV_ATTRIBUTES_SECTION_NAME
101896 rather than string literal.
101900 * readelf.c(get_riscv_segment_type): New.
101901 (get_segment_type): Handle EM_RISCV.
101905 * elf/riscv.h (PT_RISCV_ATTRIBUTES): New.
101906 * testsuite/ld-elf/orphan-region.ld: Discard .riscv.attributes
101907 section for simplify testcase.
101908 * testsuite/ld-riscv-elf/attr-phdr.d: New.
101909 * testsuite/ld-riscv-elf/attr-phdr.s: Ditto.
101910 * testsuite/ld-riscv-elf/ld-riscv-elf.exp: Add attr-phdr to
101913 2021-07-06 Alan Modra <amodra@gmail.com>
101915 Re: PR28055, segfault in bpf special reloc function
101917 * elf64-bpf.c (bpf_elf_generic_reloc): Add missing ATTRIBUTE_UNUSED.
101919 2021-07-06 GDB Administrator <gdbadmin@sourceware.org>
101921 Automatic date update in version.in
101923 2021-07-05 Tom Tromey <tom@tromey.com>
101925 Simplify debug_names index writing
101926 This changes the .debug_names writer to find the TU indices in the
101927 main loop over all CUs and TUs. (An earlier patch applied this same
101928 treatment to the .gdb_index writer.)
101930 Simplify gdb_index writing
101931 write_gdbindex writes the CUs first, then walks the signatured type
101932 hash table to write out the TUs. However, now that CUs and TUs are
101933 unified in the DWARF reader, it's simpler to handle both of these in
101936 Minor cleanup to addrmap_index_data::previous_valid
101937 This changes addrmap_index_data::previous_valid to a bool, and
101940 2021-07-05 Tom Tromey <tom@tromey.com>
101942 Fix oddity in write_gdbindex
101943 My recent patch to unify CUs and TUs introduced an oddity in
101944 write_gdbindex. Here, we pass 'i' to recursively_write_psymbols, but
101945 we must instead pass 'counter', to handle the situation where a TU is
101948 I am not sure a test case for this is possible. I think it can only
101949 happen when using DWARF 5, where a TU appears in .debug_info.
101950 However, this situation is already not handled correctly by
101951 .gdb_index. I filed a bug about this.
101953 2021-07-05 Tom Tromey <tom@tromey.com>
101955 Fix warning in symtab.c
101956 The compiler gives this warning when building symtab.c:
101958 ../../binutils-gdb/gdb/symtab.c:4247:28: warning: 'to_match' may be used uninitialized in this function [-Wmaybe-uninitialized]
101960 This patch fixes the warning by adding a gdb_assert_not_reached.
101962 2021-07-05 H.J. Lu <hjl.tools@gmail.com>
101964 ld: Cache and reuse the IR archive file descriptor
101965 Linker plugin_object_p opens the IR archive for each IR archive member.
101966 For GCC plugin, plugin_object_p closes the archive file descriptor. But
101967 for LLVM plugin, the archive file descriptor remains open. If there are
101968 3000 IR archive members, there are 3000 file descriptors for them. We
101969 can run out of file descriptors petty easily.
101971 1. Add archive_plugin_fd and archive_plugin_fd_open_count to bfd so that
101972 we can cache and reuse the IR archive file descriptor for all IR archive
101973 members in the archive.
101974 2. Add bfd_plugin_close_file_descriptor to properly close the IR archive
101980 * archive.c (_bfd_archive_close_and_cleanup): Close the archive
101981 plugin file descriptor if needed.
101982 * bfd.c (bfd): Add archive_plugin_fd and
101983 archive_plugin_fd_open_count.
101984 * opncls.c (_bfd_new_bfd): Initialize to -1.
101985 * plugin.c (bfd_plugin_open_input): Cache and reuse the archive
101986 plugin file descriptor.
101987 (bfd_plugin_close_file_descriptor): New function.
101988 (try_claim): Call bfd_plugin_close_file_descriptor.
101989 * plugin.h (bfd_plugin_close_file_descriptor): New.
101990 * bfd-in2.h: Regenerated.
101995 * plugin.c (plugin_input_file): Add ibfd.
101996 (release_plugin_file_descriptor): New function.
101997 (release_input_file): Call release_plugin_file_descriptor to
101999 (plugin_object_p): Call release_plugin_file_descriptor to close
102000 input->fd. Also call release_plugin_file_descriptor if not
102002 * testsuite/config/default.exp (RANLIB): New.
102003 * testsuite/ld-plugin/lto.exp: Run ranlib test.
102005 2021-07-05 Nick Clifton <nickc@redhat.com>
102007 Restore the libiberty component of commit 50ad1254d5030d0804cbf89c758359ae202e8d55.
102008 This commit has not yet been applied to the master sources in the gcc repository.
102009 It was submitted here: https://gcc.gnu.org/pipermail/gcc-patches/2021-July/574405.html
102010 The commit allows options to be set for the AR and RANLIB programs used when building libiberty, which in turn allows building with LTO enabled.
102012 Updated translations (mainly Ukranian and French) triggered by creation of 2.37 branch.
102014 2021-07-05 Tom de Vries <tdevries@suse.de>
102016 [gdb/testsuite] Fix fail in gdb.fortran/ptype-on-functions.exp with gcc-7
102017 Since commit 05b85772061 "gdb/fortran: Add type info of formal parameter for
102020 (gdb) ptype say_string^M
102021 type = void (character*(*), integer(kind=4))^M
102022 (gdb) FAIL: gdb.fortran/ptype-on-functions.exp: ptype say_string
102025 The part of the commit causing the fail is:
102027 gdb_test "ptype say_string" \
102028 - "type = void \\(character\\*\\(\\*\\), integer\\(kind=\\d+\\)\\)"
102029 + "type = void \\(character\[^,\]+, $integer8\\)"
102031 which fails to take into account that for gcc-7 and before, the type for
102032 string length of a string argument is int, not size_t.
102034 Fix this by allowing both $integer8 and $integer4.
102036 Tested on x86_64-linux, with gcc-7 and gcc-10.
102038 gdb/testsuite/ChangeLog:
102040 2021-07-05 Tom de Vries <tdevries@suse.de>
102042 * gdb.fortran/ptype-on-functions.exp: Allow both $integer8 and
102043 $integer4 for size of string length.
102045 2021-07-05 Simon Marchi <simon.marchi@polymtl.ca>
102047 gdb: fall back on sigpending + sigwait if sigtimedwait is not available
102048 The macOS platform does not provide sigtimedwait, so we get:
102051 In file included from /Users/smarchi/src/binutils-gdb/gdb/compile/compile.c:46:
102052 /Users/smarchi/src/binutils-gdb/gdb/../gdbsupport/scoped_ignore_signal.h:69:4: error: use of undeclared identifier 'sigtimedwait'
102053 sigtimedwait (&set, nullptr, &zero_timeout);
102056 An alternative to sigtimedwait with a timeout of 0 is to use sigpending,
102057 to first check which signals are pending, and then sigwait, to consume
102058 them. Since that's slightly more expensive (2 syscalls instead of 1),
102059 keep using sigtimedwait for the platforms that provide it, and fall back
102060 to sigpending + sigwait for the others.
102064 * scoped_ignore_signal.h (struct scoped_ignore_signal)
102065 <~scoped_ignore_signal>: Use sigtimedwait if HAVE_SIGTIMEDWAIT
102066 is defined, else use sigpending + sigwait.
102068 Change-Id: I2a72798337e81dd1bbd21214736a139dd350af87
102069 Co-Authored-By: John Baldwin <jhb@FreeBSD.org>
102071 2021-07-05 Simon Marchi <simon.marchi@polymtl.ca>
102073 gdbsupport/common.m4: check for sigtimedwait
102074 The next patch will make the use of sigtimedwait conditional to whether
102075 the platform provides it. Start by adding a configure check for it.
102079 * common.m4 (GDB_AC_COMMON): Check for sigtimedwait.
102080 * config.in, configure: Re-generate.
102084 * config.in, configure: Re-generate.
102088 * config.in, configure: Re-generate.
102090 Change-Id: Ic7613fe14521b966b4d991bbcd0933ab14629c05
102092 2021-07-05 Alan Modra <amodra@gmail.com>
102094 Re: opcodes: constify & local meps macros
102095 Commit f375d32b35ce changed a generated file. Edit the source instead.
102097 * mep.opc (macros): Make static and const.
102098 (lookup_macro): Return and use const pointer.
102099 (expand_macro): Make mac param const.
102100 (expand_string): Make pmacro const.
102102 2021-07-05 Alan Modra <amodra@gmail.com>
102104 PR28055, segfault in bpf special reloc function
102105 The testcase in this PR tickled two bugs fixed here. output_bfd is
102106 NULL when a reloc special_function is called for final linking and
102107 when called from bfd_generic_get_relocated_section_contents. Clearly
102108 using output_bfd is wrong as it results in segfaults. Not only that,
102109 the endianness of the reloc field really should be that of the input.
102110 The second bug was not checking that the entire reloc field was
102111 contained in the section contents.
102114 * elf64-bpf.c (bpf_elf_generic_reloc): Use correct bfd for bfd_put
102115 and bfd_put_32 calls. Correct section limit checks.
102117 2021-07-05 Alan Modra <amodra@gmail.com>
102119 PR28047, readelf crash due to assertion failure
102120 DW_FORM_ref1, DW_FORM_ref2, DW_FORM_ref4, DW_FORM_ref1, and
102121 DW_FORM_ref_udata are all supposed to be within the containing unit.
102124 * dwarf.c (get_type_abbrev_from_form): Add cu_end parameter.
102125 Check DW_FORM_ref1 etc. arg against cu_end rather than end of
102126 section. Adjust all callers.
102128 2021-07-05 GDB Administrator <gdbadmin@sourceware.org>
102130 Automatic date update in version.in
102132 2021-07-04 Simon Marchi <simon.marchi@polymtl.ca>
102134 gdb: return early if no execution in darwin_solib_create_inferior_hook
102135 When loading a file using the file command on macOS, we get:
102137 $ ./gdb -nx --data-directory=data-directory -q -ex "file ./test"
102138 Reading symbols from ./test...
102139 Reading symbols from /Users/smarchi/build/binutils-gdb/gdb/test.dSYM/Contents/Resources/DWARF/test...
102140 /Users/smarchi/src/binutils-gdb/gdb/thread.c:72: internal-error: struct thread_info *inferior_thread(): Assertion `current_thread_ != nullptr' failed.
102141 A problem internal to GDB has been detected,
102142 further debugging may prove unreliable.
102143 Quit this debugging session? (y or n)
102147 * frame #0: 0x0000000101fcb826 gdb`internal_error(file="/Users/smarchi/src/binutils-gdb/gdb/thread.c", line=72, fmt="%s: Assertion `%s' failed.") at errors.cc:52:3
102148 frame #1: 0x00000001018a2584 gdb`inferior_thread() at thread.c:72:3
102149 frame #2: 0x0000000101469c09 gdb`get_current_regcache() at regcache.c:421:31
102150 frame #3: 0x00000001015f9812 gdb`darwin_solib_get_all_image_info_addr_at_init(info=0x0000603000006d00) at solib-darwin.c:464:34
102151 frame #4: 0x00000001015f7a04 gdb`darwin_solib_create_inferior_hook(from_tty=1) at solib-darwin.c:515:5
102152 frame #5: 0x000000010161205e gdb`solib_create_inferior_hook(from_tty=1) at solib.c:1200:3
102153 frame #6: 0x00000001016d8f76 gdb`symbol_file_command(args="./test", from_tty=1) at symfile.c:1650:7
102154 frame #7: 0x0000000100abab17 gdb`file_command(arg="./test", from_tty=1) at exec.c:555:3
102155 frame #8: 0x00000001004dc799 gdb`do_const_cfunc(c=0x000061100000c340, args="./test", from_tty=1) at cli-decode.c:102:3
102156 frame #9: 0x00000001004ea042 gdb`cmd_func(cmd=0x000061100000c340, args="./test", from_tty=1) at cli-decode.c:2160:7
102157 frame #10: 0x00000001018d4f59 gdb`execute_command(p="t", from_tty=1) at top.c:674:2
102158 frame #11: 0x0000000100eee430 gdb`catch_command_errors(command=(gdb`execute_command(char const*, int) at top.c:561), arg="file ./test", from_tty=1, do_bp_actions=true)(char const*, int), char const*, int, bool) at main.c:523:7
102159 frame #12: 0x0000000100eee902 gdb`execute_cmdargs(cmdarg_vec=0x00007ffeefbfeba0 size=1, file_type=CMDARG_FILE, cmd_type=CMDARG_COMMAND, ret=0x00007ffeefbfec20) at main.c:618:9
102160 frame #13: 0x0000000100eed3a4 gdb`captured_main_1(context=0x00007ffeefbff780) at main.c:1322:3
102161 frame #14: 0x0000000100ee810d gdb`captured_main(data=0x00007ffeefbff780) at main.c:1343:3
102162 frame #15: 0x0000000100ee8025 gdb`gdb_main(args=0x00007ffeefbff780) at main.c:1368:7
102163 frame #16: 0x00000001000044f1 gdb`main(argc=6, argv=0x00007ffeefbff8a0) at gdb.c:32:10
102164 frame #17: 0x00007fff20558f5d libdyld.dylib`start + 1
102166 The solib_create_inferior_hook call in symbol_file_command was added by
102167 commit ea142fbfc9c1 ("Fix breakpoints on file reloads for PIE
102168 binaries"). It causes solib_create_inferior_hook to be called while
102169 the inferior is not running, which darwin_solib_create_inferior_hook
102170 does not expect. darwin_solib_get_all_image_info_addr_at_init, in
102171 particular, assumes that there is a current thread, as it tries to get
102172 the current thread's regcache.
102174 Fix it by adding a target_has_execution check and returning early. Note
102175 that there is a similar check in svr4_solib_create_inferior_hook.
102179 * solib-darwin.c (darwin_solib_create_inferior_hook): Return
102182 Change-Id: Ia11dd983a1e29786e5ce663d0fcaa6846dc611bb
102184 2021-07-04 GDB Administrator <gdbadmin@sourceware.org>
102186 Automatic date update in version.in
102188 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
102190 gprof: Regenerate configure
102191 * configure: Regenerated.
102193 2021-07-03 Joel Brobecker <brobecker@adacore.com>
102195 Update NEWS post GDB 11 branch creation.
102198 * NEWS: Create a new section for the next release branch.
102199 Rename the section of the current branch, now that it has
102202 2021-07-03 Joel Brobecker <brobecker@adacore.com>
102204 Bump version to 12.0.50.DATE-git.
102205 Now that the GDB 11 branch has been created, we can
102206 bump the version number.
102210 GDB 11 branch created (4b51505e33441c6165e7789fa2b6d21930242927):
102211 * version.in: Bump version to 12.0.50.DATE-git.
102213 gdb/testsuite/ChangeLog:
102215 * gdb.base/default.exp: Change $_gdb_major to 12.
102217 2021-07-03 Tom Tromey <tom@tromey.com>
102219 Use 'bool' more idiomatically in dwarf_decode_lines
102220 I noticed a couple of spots related to dwarf_decode_lines where the
102221 'include_p' field was not being used idiomatically -- it is of type
102222 bool now, so treat it as such.
102225 2021-07-03 Tom Tromey <tom@tromey.com>
102227 * dwarf2/read.c (lnp_state_machine::record_line): Use 'true'.
102228 (dwarf_decode_lines): Remove '=='.
102230 2021-07-03 Nick Clifton <nickc@redhat.com>
102232 More minor updates to the how-to-make-a-release documentation
102234 Update version number and regenerate files
102236 Add markers for 2.37 branch
projects / platform / upstream / binutils.git / blob