1 If you find inaccuracies in this list, please send mail to
2 gdb-patches@sourceware.cygnus.com. If you would like to work on any
3 of these, you should consider sending mail to the same address, to
4 find out whether anyone else is working on it.
10 Below is a list of problems identified during the GDB 5.0 release
11 cycle. People hope to have these problems fixed in 5.1.
15 Hardware watchpint problems on x86 OSes, including Linux:
17 1. Delete/disable hardware watchpoints should free hardware debug
19 2. Watch for different values on a viariable with one hardware debug
22 According to Eli Zaretskii <eliz@delorie.com>:
24 These are not GDB/ia32 issues per se: the above features are all
25 implemented in the DJGPP port of GDB and work in v5.0. Every
26 x86-based target should be able to lift the relevant parts of
27 go32-nat.c and use them almost verbatim. You get debug register
28 sharing through reference counts, and the ability to watch large
29 regions (up to 16 bytes) using multiple registers. (The required
30 infrastructure in high-level GDB application code, mostly in
31 breakpoint.c, is also working since v5.0.)
35 RFD: infrun.c: No bpstat_stop_status call after proceed over break?
36 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00665.html
38 GDB misses watchpoint triggers after proceeding over a breakpoint on
43 x86 linux GDB and SIGALRM (???)
44 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00803.html
46 This problem has been fixed, but a regression test still needs to be
47 added to the testsuite:
48 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00309.html
54 Can't build IRIX -> arm GDB.
55 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00356.html
58 > Now I'm building for an embedded arm target. If there is a way of turning
59 > remote-rdi off, I couldn't find it. It looks like it gets built by default
60 > in gdb/configure.tgt(line 58) Anyway, the build dies in
61 > gdb/rdi-share/unixcomm.c. SERPORT1 et. al. never get defined because we
62 > aren't one of the architectures supported.
66 Problem with weak functions
67 http://sourceware.cygnus.com/ml/gdb/2000-05/msg00060.html
69 Dan Nicolaescu writes:
70 > It seems that gdb-4.95.1 does not display correctly the function when
71 > stoping in weak functions.
73 > It stops in a function that is defined as weak, not in the function
74 > that is actually run...
78 GDB 5.0 doesn't work on Linux/SPARC
82 Thread support. Right now, as soon as a thread finishes and exits,
83 you're hosed. This problem is reported once a week or so.
87 Wow, three bug reports for the same problem in one day! We should
88 probably make fixing this a real priority :-).
90 Anyway, thanks for reporting.
92 The following patch will fix the problems with setting breakpoints in
93 dynamically loaded objects:
95 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00230.html
97 This patch isn't checked in yet (ping Michael/JimB), but I hope this
98 will be in the next GDB release.
100 There should really be a test in the testsuite for this problem, since
101 it keeps coming up :-(. Any volunteers?
108 http://sources.redhat.com/ml/gdb/2000-07/msg00038.html
110 Is the Solaris 8 x86 problem fixed? When you configure it, configure
111 incorrectly determines that I have no curses.h. This causes mucho
112 compilation errors later on.
114 Simply editing the config.h to define CURSES_H fixes the problem, and
115 then the build works fine.
117 The status for this problem:
119 Solaris 8 x86 (PIII-560)
122 I had the same problem with several of the snapshots shortly before
123 5.0 became official, and 5.0 has the same problem.
125 I sent some mail in about it long ago, and never saw a reply.
127 I haven't had time to figure it out myself, especially since I get all
128 confused trying to figure out what configure does, I was happy to find
135 GDB 5.1 - New features
136 ======================
138 The following new features should be included in 5.1.
142 Enable MI by default. Old code can be deleted after 5.1 is out.
146 Pascal (Pierre Muller, David Taylor)
148 Pierre Muller has contributed patches for adding Pascal Language
151 2 pascal language patches inserted in database
152 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00521.html
155 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00496.html
159 Java (Anthony Green, David Taylor)
161 Anthony Green has a number of Java patches that did not make it into
162 the 5.0 release. The first two are in cvs now, but the third needs
163 some fixing up before it can go in.
166 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00512.html
169 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00515.html
171 Patch: handle N_MAIN stab
172 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00527.html
178 Modify gdb to work correctly with Pascal.
182 Revised UDP support (was: Re: [Fwd: [patch] UDP transport support])
183 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00000.html
185 (Broken) support for GDB's remote protocol across UDP is to be
186 included in the follow-on release.
188 It should be noted that UDP can only work when the [Gg] packet fits in
191 There is also much debate over the merit of this.
198 The following code cleanups will hopefully be applied to GDB 5.1.
202 Delete macro TARGET_BYTE_ORDER_SELECTABLE.
204 Patches in the database.
208 Fix copyright notices.
210 Turns out that ``1998-2000'' isn't considered valid :-(
212 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00467.html
218 Eliminate all uses of PARAMS in GDB's source code.
222 printcmd.c (print_address_numeric):
224 NOTE: This assumes that the significant address information is kept in
225 the least significant bits of ADDR - the upper bits were either zero
226 or sign extended. Should ADDRESS_TO_POINTER() or some
227 ADDRESS_TO_PRINTABLE() be used to do the conversion?
233 Eliminate all warnings for at least one host/target for the flags:
234 -Wimplicit -Wreturn-type -Wcomment -Wtrigraphs -Wformat -Wparentheses
235 -Wpointer-arith -Wuninitialized
239 Follow through `make check' with --enable-shared.
241 When the srcware tree is configured with --enable-shared, the `expect'
242 program won't run properly. Jim Wilson found out gdb has a local hack
243 to set LD_LIBRARY_PATH, but, AFAIK, no other project has been hacked
246 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00845.html
255 Fix at least one thread bug.
259 GDB 5.2 - New features
260 ======================
264 Objective C/C++ Support. Bu hopefully sooner...
271 The following cleanups have been identified as part of GDB 5.2.
275 Eliminate more compiler warnings.
279 Restructure gdb directory tree so that it avoids any 8.3 and 14
284 Convert GDB build process to AUTOMAKE.
286 See also sub-directory configure below.
288 The current convention is (kind of) to use $(<header>_h) in all
289 dependency lists. It isn't done in a consistent way.
293 Code Cleanups: General
294 ======================
296 The following are more general cleanups and fixes. They are not tied
297 to any specific release.
301 The BFD directory requires bug-fixed AUTOMAKE et.al.
303 AUTOMAKE 1.4 incorrectly set the TEXINPUTS environment variable. It
304 contained the full path to texinfo.tex when it should have only
305 contained the directory. The bug has been fixed in the current
306 AUTOMAKE sources. Automake snapshots can be found in:
307 ftp://sourceware.cygnus.com/pub/gdb/snapshots
308 and ftp://sourceware.cygnus.com/pub/binutils
312 Find something better than DEFAULT_BFD_ARCH, DEFAULT_BFD_VEC to
313 determine the default isa/byte-order.
317 Rely on BFD_BIG_ENDIAN and BFD_LITTLE_ENDIAN instead of host dependent
318 BIG_ENDIAN and LITTLE_ENDIAN.
322 Eliminate more compiler warnings.
324 Of course there also needs to be the usual debate over which warnings
325 are valid and how to best go about this.
327 One method: choose a single option; get agreement that it is
328 reasonable; try it out to see if there isn't anything silly about it
329 (-Wunused-parameters is an example of that) then incrementally hack
332 The other method is to enable all warnings and eliminate them from one
337 Elimination of ``(catch_errors_ftype *) func''.
339 Like make_cleanup_func it isn't portable.
340 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00791.html
341 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00814.html
345 Nuke #define CONST_PTR.
353 [PATCH/5] src/intl/Makefile.in:distclean additions
354 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00363.html
356 Do not forget to merge the patch back into the trunk.
360 Rationalize the host-endian code (grep for HOST_BYTE_ORDER).
362 At present defs.h includes <endian.h> (which is linux specific) yet
363 almost nothing depends on it. Suggest "gdb_endian.h" which can also
364 handle <machine/endian.h> and only include that where it is really
369 Replace strsave() + mstrsave() with libiberty:xstrdup().
373 Replace savestring() with something from libiberty.
375 An xstrldup()? but that would have different semantics.
379 Rationalize use of floatformat_unknown in GDB sources.
381 Instead of defaulting to floatformat_unknown, should hosts/targets
382 specify the value explicitly?
384 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00447.html
388 Add a ``name'' member to include/floatformat.h:struct floatformat.
389 Print that name in gdbarch.c.
393 Sort out the harris mess in include/floatformat.h (it hardwires two
394 different floating point formats).
398 See of the GDB local floatformat_do_doublest() and libiberty's
399 floatformat_to_double (which was once GDB's ...) can be merged some
404 Eliminate mmalloc() from GDB.
406 Also eliminate it from defs.h.
410 Eliminate PTR. ISO-C allows ``void *''.
416 GDB should never abort. GDB should either throw ``error ()'' or
417 ``internal_error ()''. Better still GDB should naturally unwind with
422 Add __LINE__ and __FILE__ to internal_error().
426 GDB probably doesn't build on FreeBSD pre 2.2.x
427 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00378.html
429 Fixes to get FreeBSD working on 2.2.x, 3.x and 4.x caused the code to
434 Deprecate "fg". Apparently ``fg'' is actually continue.
436 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00417.html
440 Deprecate current use of ``floatformat_unknown''.
442 Require all targets to explicitly provide their float format instead
443 of defaulting to floatformat unknown. Doing the latter leads to nasty
446 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00447.html
450 Rationalize floatformat_to_double() vs floatformat_to_doublest().
452 Looks like GDB migrated floatformat_to_double() to libiberty but then
453 turned around and created a ..._to_doublest() the latter containing
456 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00472.html
460 Move floatformat_ia64_ext to libiberty/include floatformat.[ch].
462 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00466.html
466 The ``maintenance deprecate set endian big'' command doesn't notice
467 that it is deprecating ``set endian'' and not ``set endian big'' (big
468 is implemented using an enum). Is anyone going to notice this?
472 When tab expanding something like ``set arch<tab>'' ignore the
473 deprecated ``set archdebug'' and expand to ``set architecture''.
477 Eliminate ``arm_register_names[j] = (char *) regnames[j]'' and the
478 like from arm-tdep.c.
482 Fix uses of ->function.cfunc = set_function().
484 The command.c code calls sfunc() when a set command. Rather than
485 change it suggest fixing the callback function so that it is more
488 http://sourceware.cygnus.com/ml/gdb-patches/2000-06/msg00062.html
490 See also ``Fix implementation of ``target xxx''.'' below.
494 IRIX 3.x support is probably broken.
498 Delete sim/SIM_HAVE_BREAKPOINTS and gdb/SIM_HAS_BREAKPOINTS.
499 http://sourceware.cygnus.com/ml/gdb-patches/2000-07/msg00042.html
501 Apart from the d30v, are there any sim/common simulators that make use
504 A brief summary of what happened is that sim/common/sim-break.c was
505 created as a good idea. It turned out a better idea was to use
506 SIM_SIGBREAK and have GDB pass back sim_resume (..., SIGBREAK).
510 Move remote_remove_hw_breakpoint, remote_insert_hw_breakpoint,
511 remote_remove_watchpoint, remote_insert_watchpoint into target vector.
515 Eliminate ``extern'' from C files.
519 Replace ``STREQ()'' et.al. with ``strcmp() == 0'' et.al.
521 Extreme care is recommeded - perhaps only modify tests that are
522 exercised by the testsuite (as determined using some type of code
527 New Features and Fixes
528 ======================
530 These are harder than cleanups but easier than work involving
531 fundamental architectural change.
535 Add built-by, build-date, tm, xm, nm and anything else into gdb binary
536 so that you can see how the GDB was created.
540 Add an "info bfd" command that displays supported object formats,
541 similarly to objdump -i.
543 Is there a command already?
547 Fix ``I'm sorry, Dave, I can't do that.'' from symfile.c.
549 This requires internationalization.
555 (gdb) p fwprintf(stdout,L"%S\n", f)
556 No symbol "L" in current context.
560 Cleanup configury support for optional sub-directories.
562 Check how GCC handles multiple front ends for an example of how things
563 could work. A tentative first step is to rationalize things so that
564 all sub directories are handled in a fashion similar to gdb/mi.
566 See also automake above.
570 Add a transcript mechanism to GDB.
572 Such a mechanism might log all gdb input and output to a file in a
573 form that would allow it to be replayed. It could involve ``gdb
574 --transcript=FILE'' or it could involve ``(gdb) transcript file''.
578 Can the xdep files be replaced by autoconf?
582 Document trace machinery
586 Document ui-out and ui-file.
588 http://sourceware.cygnus.com/ml/gdb/2000-04/msg00121.html
592 Update texinfo.tex to latest?
596 Incorporate agentexpr.texi into gdb.texinfo
598 agentexpr.texi mostly describes the details of the byte code used for
599 tracepoints, not the internals of the support for this in GDB. So it
600 looks like gdb.texinfo is a better place for this information.
602 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00566.html
606 Document overlay machinery.
610 ``(gdb) catch signal SIGNAL''
612 Overlaps with ``handle SIGNAL'' but the implied behavior is different.
613 You can attach commands to a catch but not a handle. A handle has a
614 limited number of hardwired actions.
618 Get the TUI working on all platforms.
622 Add support for ``gdb --- PROGRAM ARGS ...''.
623 Add support for ``gdb -cmd=...''
625 Along with many variations. Check:
627 ????? for a full discussion.
633 Implement ``(gdb) !ls''.
635 Which is very different from ``(gdb) ! ls''. Implementing the latter
638 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00034.html
642 Change the (char *list[]) to (const char (*)[]) so that dynamic lists can
647 When tab expanding something like ``set arch<tab>'' ignore the
648 deprecated ``set archdebug'' and expand to ``set architecture''.
652 Replace the code that uses the host FPU with an emulator of the target
657 The "ocd reset" command needs to flush the dcache, which requires breaking
658 the abstraction layer between the target independent and target code. One
659 way to address this is provide a generic "reset" command and target vector.
661 http://sources.redhat.com/ml/gdb-patches/2000-10/msg00011.html
670 Generic: lin-thread cannot handle thread exit (Mark Kettenis, Michael
671 Snyder) http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00525.html
673 The thread_db assisted debugging code doesn't handle exiting threads
674 properly, at least in combination with glibc 2.1.3 (the framework is
675 there, just not the actual code). There are at least two problems
676 that prevent this from working.
678 As an additional reference point, the pre thread_db code did not work
683 GNU/Linux/x86 and random thread signals (and Solaris/SPARC but not
685 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00336.html
687 Christopher Blizzard writes:
689 So, I've done some more digging into this and it looks like Jim
690 Kingdon has reported this problem in the past:
692 http://sourceware.cygnus.com/ml/bug-gdb/1999-10/msg00058.html
694 I can reproduce this problem both with and without Tom's patch. Has
695 anyone seen this before? Maybe have a solution for it hanging around?
698 There's a test case for this documented at:
700 when debugging threaded applications you get extra SIGTRAPs
701 http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=9565
703 [There should be a GDB testcase - cagney]
707 GDB5 TOT on unixware 7
708 http://sourceware.cygnus.com/ml/gdb/2000-04/msg00119.html
711 > I just spun the top of tree of the GDB5 branch on UnixWare 7. As a
712 > practical matter, the current thread support is somewhat more annoying
713 > than when GDB was thread-unaware.
717 Migrate qfThreadInfo packet -> qThreadInfo. (Andrew Cagney)
719 Add support for packet enable/disable commands with these thread
720 packets. General cleanup.
722 [PATCH] Document the ThreadInfo remote protocol queries
723 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00832.html
725 [PATCH] "info threads" queries for remote.c
726 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00831.html
733 New languages come onto the scene all the time.
737 Re: Various C++ things
739 value_headof/value_from_vtable_info are worthless, and should be
740 removed. The one place in printcmd.c that uses it should use the RTTI
743 RTTI for g++ should be using the typeinfo functions rather than the
744 vtables. The typeinfo functions are always at offset 4 from the
745 beginning of the vtable, and are always right. The vtables will have
746 weird names like E::VB sometimes. The typeinfo function will always
747 be "E type_info function", or somesuch.
749 value_virtual_fn_field needs to be fixed so there are no failures for
750 virtual functions for C++ using g++.
752 Testsuite cases are the major priority right now for C++ support,
753 since i have to make a lot of changes that could potentially break
758 Add support for Modula3
760 Get DEC/Compaq to contribute their Modula-3 support.
764 Remote Protocol Support
765 =======================
769 Remote protocol doco feedback.
771 Too much feedback to mention needs to be merged in (901660). Search
772 for the word ``remote''.
775 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00023.html
776 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00056.html
777 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00382.html
781 GDB doesn't recover gracefully from remote protocol errors.
783 GDB wasn't checking for NAKs from the remote target. Instead a NAK is
784 ignored and a timeout is required before GDB retries. A pre-cursor to
785 fixing this this is making GDB's remote protocol packet more robust.
787 While downloading to a remote protocol target, gdb ignores packet
788 errors in so far as it will continue to download with chunk N+1 even
789 if chunk N was not correctly sent. This causes gdb.base/remote.exp to
790 take a painfully long time to run. As a PS that test needs to be
791 fixed so that it builds on 16 bit machines.
795 Add the cycle step command.
797 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00237.html
801 Resolve how to scale things to support very large packets.
805 Resolve how to handle a target that changes things like its endianess
806 on the fly - should it be returned in the ``T'' packet?
808 Underlying problem is that the register file is target endian. If the
809 target endianess changes gdb doesn't know.
813 Rename read_register{,_pid}() to read_unsigned_register{,_pid}().
820 If / when GDB starts to support the debugging of multi-processor
821 (rather than multi-thread) applications the symtab code will need to
822 be updated a little so that several independent symbol tables are
823 active at a given time.
825 The other interesting change is a clarification of the exact meaning
826 of CORE_ADDR and that has had consequences for a few targets (that
827 were abusing that data type).
831 Investiagate ways of reducing memory.
835 Investigate ways of improving load time.
839 Get the d10v to use POINTER_TO_ADDRESS and ADDRESS_TO_POINTER.
841 Consequence of recent symtab clarification. No marks for figuring out
842 who maintains the d10v.
846 Get the MIPS to correctly sign extend all address <-> pointer
849 Consequence of recent symtab clarification. No marks for figuring out
850 who maintains the MIPS.
854 GDB truncates 64 bit enums.
856 http://sourceware.cygnus.com/ml/gdb-patches/2000-06/msg00290.html
863 There are never to many testcases.
867 Better thread testsuite.
871 Better C++ testsuite.
875 Look at adding a GDB specific testsuite directory so that white box
876 tests of key internals can be added (eg ui_file).
880 Separate out tests that involve the floating point (FP).
882 (Something for people brining up new targets). FP and non-fp tests
883 are combined. I think there should be set of basic tests that
884 exercise pure integer support and then a more expanded set that
885 exercise FP and FP/integer interactions.
887 As an example, the MIPS, for n32 as problems with passing FP's and
888 structs. Since most inferior call tests include FP it is difficult to
889 determine of the integer tests are ok.
893 Architectural Changes: General
894 ==============================
896 These are harder than simple cleanups / fixes and, consequently
897 involve more work. Typically an Architectural Change will be broken
898 down into a more digestible set of cleanups and fixes.
902 Cleanup software single step.
904 At present many targets implement software single step by directly
905 blatting memory (see rs6000-tdep.c). Those targets should register
906 the applicable breakpoints using the breakpoint framework. Perhaphs a
907 new internal breakpoint class ``step'' is needed.
911 Replace READ_FP() with FRAME_HANDLE().
913 READ_FP() is a hangover from the days of the vax when the ABI really
914 did have a frame pointer register. Modern architectures typically
915 construct a virtual frame-handle from the stack pointer and various
916 other bits of string.
918 Unfortunately GDB still treats this synthetic FP register as though it
919 is real. That in turn really confuses users (arm and ``print $fp'' VS
920 ``info registers fp''). The synthetic FP should be separated out of
921 the true register set presented to the user.
925 Register Cache Cleanup (below from Andrew Cagney)
927 I would depict the current register architecture as something like:
934 register + REGISTER_BYTE(reg_nr)
937 -------------------------
938 | extern register[] |
939 -------------------------
941 where neither the high (valops.c et.al.) or low gdb (*-tdep.c) are
942 really clear on what mechanisms they should be using to manipulate that
943 buffer. Further, much code assumes, dangerously, that registers are
944 contigious. Having got mips-tdep.c to support multiple ABIs, believe
945 me, that is a bad assumption. Finally, that register cache layout is
946 determined by the current remote/local target and _not_ the less
947 specific target ISA. In fact, in many cases it is determined by the
948 somewhat arbitrary layout of the [gG] packets!
951 How I would like the register file to work is more like:
979 The main objectives being:
981 o a clear separation between the low
982 level target and the high level GDB
984 o a mechanism that solves the general
985 problem of register aliases, overlaps
986 etc instead of treating them as optional
987 extras that can be wedged in as an after
988 thought (that is a reasonable description
989 of the current code).
991 Identify then solve the hard case and the
992 rest just falls out. GDB solved the easy
993 case and then tried to ignore the real
996 o a removal of the assumption that the
997 mapping between the register cache
998 and virtual registers is largely static.
999 If you flip the USR/SSR stack register
1000 select bit in the status-register then
1001 the corresponding stack registers should
1004 o a mechanism that clearly separates the
1005 gdb internal register cache from any
1006 target (not architecture) dependent
1007 specifics such as [gG] packets.
1009 Of course, like anything, it sounds good in theory. In reality, it
1010 would have to contend with many<->many relationships at both the
1011 virt<->cache and cache<->target level. For instance:
1014 Modifying an mmx register may involve
1015 scattering values across both FP and
1016 mmpx specific parts of a buffer
1019 When writing back a SP it may need to
1020 both be written to both SP and USP.
1025 Rather than let this like the last time it was discussed, just slip, I'm
1026 first going to add this e-mail (+ references) to TODO. I'd then like to
1027 sketch out a broad strategy I think could get us there.
1030 First thing I'd suggest is separating out the ``extern registers[]''
1031 code so that we can at least identify what is using it. At present
1032 things are scattered across many files. That way we can at least
1033 pretend that there is a cache instead of a global array :-)
1035 I'd then suggest someone putting up a proposal for the pseudo-reg /
1036 high-level side interface so that code can be adopted to it. For old
1037 code, initially a blanket rename of write_register_bytes() to
1038 deprecated_write_register_bytes() would help.
1040 Following that would, finaly be the corresponding changes to the target.
1044 Check that GDB can handle all BFD architectures (Andrew Cagney)
1046 There should be a test that checks that BFD/GDB are in sync with
1047 regard to architecture changes. Something like a test that first
1048 queries GDB for all supported architectures and then feeds each back
1049 to GDB.. Anyone interested in learning how to write tests? :-)
1053 Architectural Change: Multi-arch et al.
1054 =======================================
1056 The long term objective is to remove all assumptions that there is a
1057 single target with a single address space with a single instruction
1058 set architecture and single application binary interface.
1060 This is an ongoing effort. The first milestone is to enable
1061 ``multi-arch'' where by all architectural decisions are made at
1064 It should be noted that ``gdbarch'' is really ``gdbabi'' and
1065 ``gdbisa''. Once things are multi-arched breaking that down correctly
1066 will become much easier.
1070 GDBARCH cleanup (Andrew Cagney)
1072 The non-generated parts of gdbarch.{sh,h,c} should be separated out
1073 into arch-utils.[hc].
1075 Document that gdbarch_init_ftype could easily fail because it didn't
1076 identify an architecture.
1080 Fix BELIEVE_PPC_PROMOTION. Change it to BELIEVE_PPC_PROMOTION_P?
1082 At present there is still #ifdef BELIEVE_PPC_PROMOTION code in the
1087 Fix target_signal_from_host() etc.
1089 The name is wrong for starters. ``target_signal'' should probably be
1090 ``gdb_signal''. ``from_host'' should be ``from_target_signal''.
1091 After that it needs to be multi-arched and made independent of any
1092 host signal numbering.
1096 Update ALPHA so that it uses ``struct frame_extra_info'' instead of
1099 This is a barrier to replacing mips_extra_func_info with something
1100 that works with multi-arch.
1104 Multi-arch mips_extra_func_info.
1106 This first needs the alpha to be updated so that it uses ``struct
1111 Rationalize TARGET_SINGLE_FORMAT and TARGET_SINGLE_BIT et al.
1113 Surely one of them is redundant.
1117 Convert ALL architectures to MULTI-ARCH.
1121 Select the initial multi-arch ISA / ABI based on --target or similar.
1123 At present the default is based on what ever is first in the BFD
1124 archures table. It should be determined based on the ``--target=...''
1129 Make MIPS pure multi-arch.
1131 It is only at the multi-arch enabled stage.
1137 Enable the code to recognize --enable-targets=.... like BINUTILS does.
1139 Can the tm.h and nm.h files be eliminated by multi-arch.
1143 Architectural Change: MI, LIBGDB and scripting languages
1144 ========================================================
1146 See also architectural changes related to the event loop. LIBGDB
1147 can't be finished until there is a generic event loop being used by
1150 The long term objective is it to be possible to integrate GDB into
1151 scripting languages.
1155 Implement generic ``(gdb) commmand > file''
1157 Once everything is going through ui_file it should be come fairly
1160 http://sourceware.cygnus.com/ml/gdb/2000-04/msg00104.html
1164 Replace gdb_stdtarg with gdb_targout (and possibly gdb_targerr).
1166 gdb_stdtarg is easily confused with gdb_stdarg.
1170 Extra ui_file methods - dump.
1172 Very useful for whitebox testing.
1176 Eliminate error_begin().
1178 With ui_file, there is no need for the statefull error_begin ()
1183 Send normal output to gdb_stdout.
1184 Send error messages to gdb_stderror.
1185 Send debug and log output log gdb_stdlog.
1187 GDB still contains many cases where (f)printf or printf_filtered () is
1188 used when it should be sending the messages to gdb_stderror or
1189 gdb_stdlog. The thought of #defining printf to something has crossed
1194 Re-do GDB's output pager.
1196 GDB's output pager still relies on people correctly using *_filtered
1197 for gdb_stdout and *_unfiltered for gdb_stdlog / gdb_stderr.
1198 Hopefully, with all normal output going to gdb_stdout, the pager can
1199 just look at the ui_file that the output is on and then use that to
1200 decide what to do about paging. Sounds good in theory.
1204 Check/cleanup MI documentation.
1206 The list of commands specified in the documentation needs to be
1207 checked against the mi-cmds.c table in a mechanical way (so that they
1208 two can be kept up-to-date).
1212 Convert MI into libgdb
1214 MI provides a text interface into what should be many of the libgdb
1215 functions. The implementation of those functions should be separated
1216 into the MI interface and the functions proper. Those functions being
1217 moved to gdb/lib say.
1223 The first part can already be found in defs.h.
1227 MI's input does not use buffering.
1229 At present the MI interface reads raw characters of from an unbuffered
1230 FD. This is to avoid several nasty buffer/race conditions. That code
1231 should be changed so that it registers its self with the event loop
1232 (on the input FD) and then push commands up to MI as they arrive.
1234 The serial code already does this.
1238 Make MI interface accessible from existing CLI.
1242 Add a breakpoint-edit command to MI.
1244 It would be similar to MI's breakpoint create but would apply to an
1245 existing breakpoint. It saves the need to delete/create breakpoints
1246 when ever they are changed.
1250 Add directory path to MI breakpoint.
1252 That way the GUI's task of finding the file within which the
1253 breakpoint was set is simplified.
1257 Add a mechanism to reject certain expression classes to MI
1259 There are situtations where you don't want GDB's expression
1260 parser/evaluator to perform inferior function calls or variable
1261 assignments. A way of restricting the expression parser so that such
1262 operations are not accepted would be very helpful.
1266 Remove sideffects from libgdb breakpoint create function.
1268 The user can use the CLI to create a breakpoint with partial
1269 information - no file (gdb would use the file from the last
1272 The libgdb interface currently affects that environment which can lead
1273 to confusion when a user is setting breakpoints via both the MI and
1276 This is also a good example of how getting the CLI ``right'' will be
1281 Move gdb_lasterr to ui_out?
1283 The way GDB throws errors and records them needs a re-think. ui_out
1284 handles the correct output well. It doesn't resolve what to do with
1285 output / error-messages when things go wrong.
1289 do_setshow_command contains a 1024 byte buffer.
1291 The function assumes that there will never be any more than 1024 bytes
1292 of enum. It should use mem_file.
1296 Should struct cmd_list_element . completer take the command as an
1301 Should the bulk of top.c:line_completion_function() be moved to
1302 command.[hc]? complete_on_cmdlist() and complete_on_enums() could
1303 then be made private.
1307 top.c (execute_command): Should a command being valid when the target
1308 is running be made an attribute (predicate) to the command rather than
1309 an explicit set of tests.
1313 top.c (execute_command): Should the bulk of this function be moved
1314 into command.[hc] so that top.c doesn't grub around in the command
1319 Architectural Change: Async
1320 ===========================
1322 While GDB uses an event loop when prompting the user for input. That
1323 event loop is not exploited by targets when they allow the target
1324 program to continue. Typically targets still block in (target_wait())
1325 until the program again halts.
1327 The closest a target comes to supporting full asynchronous mode are
1328 the remote targets ``async'' and ``extended-async''.
1332 Asynchronous expression evaluator
1334 Inferior function calls hang GDB.
1338 Fix implementation of ``target xxx''.
1340 At present when the user specifies ``target xxxx'', the CLI maps that
1341 directly onto a target open method. It is then assumed that the
1342 target open method should do all sorts of complicated things as this
1343 is the only chance it has. Check how the various remote targets
1344 duplicate the target operations. Check also how the various targets
1345 behave differently for purely arbitrary reasons.
1347 What should happen is that ``target xxxx'' should call a generic
1348 ``target'' function and that should then co-ordinate the opening of
1349 ``xxxx''. This becomes especially important when you're trying to
1350 open an asynchronous target that may need to perform background tasks
1351 as part of the ``attach'' phase.
1353 Unfortunately, due to limitations in the old/creaking command.h
1354 interface, that isn't possible. The function being called isn't told
1355 of the ``xxx'' or any other context information.
1357 Consequently a precursor to fixing ``target xxxx'' is to clean up the
1358 CLI code so that it passes to the callback function (attatched to a
1359 command) useful information such as the actual command and a context
1360 for that command. Other changes such as making ``struct command''
1361 opaque may also help.
1364 http://sourceware.cygnus.com/ml/gdb-patches/2000-06/msg00062.html
1368 Make "target xxx" command interruptible.
1370 As things become async this becomes possible. A target would start
1371 the connect and then return control to the event loop. A cntrl-c
1372 would notify the target that the operation is to be abandoned and the
1373 target code could respond.
1377 Add a "suspend" subcommand of the "continue" command to suspend gdb
1378 while continuing execution of the subprocess. Useful when you are
1379 debugging servers and you want to dodge out and initiate a connection
1380 to a server running under gdb.
1389 Frequently requested but not approved requests.
1393 Eliminate unused argument warnings using ATTRIBUTE_UNUSED.
1395 The benefits on this one are thought to be marginal - GDBs design
1396 means that unused parameters are very common. GCC 3.0 will also
1397 include the option -Wno-unused-parameter which means that ``-Wall
1398 -Wno-unused-parameters -Werror'' can be specified.
1407 This list is not up to date, and opinions vary about the importance or
1408 even desirability of some of the items. If you do fix something, it
1409 always pays to check the below.
1413 @c This does not work (yet if ever). FIXME.
1414 @c @item --parse=@var{lang} @dots{}
1415 @c Configure the @value{GDBN} expression parser to parse the listed languages.
1416 @c @samp{all} configures @value{GDBN} for all supported languages. To get a
1417 @c list of all supported languages, omit the argument. Without this
1418 @c option, @value{GDBN} is configured to parse all supported languages.
1422 START_INFERIOR_TRAPS_EXPECTED need never be defined to 2, since that
1423 is its default value. Clean this up.
1427 It should be possible to use symbols from shared libraries before we know
1428 exactly where the libraries will be loaded. E.g. "b perror" before running
1429 the program. This could maybe be done as an extension of the "breakpoint
1430 re-evaluation" after new symbols are loaded.
1434 Make single_step() insert and remove breakpoints in one operation.
1436 [If this is talking about having single_step() insert the breakpoints,
1437 run the target then pull the breakpoints then it is wrong. The
1438 function has to return as control has to eventually be passed back to
1439 the main event loop.]
1443 Speed up single stepping by avoiding extraneous ptrace calls.
1447 Speed up single stepping by not inserting and removing breakpoints
1448 each time the inferior starts and stops.
1450 Breakpoints should not be inserted and deleted all the time. Only the
1451 one(s) there should be removed when we have to step over one. Support
1452 breakpoints that don't have to be removed to step over them.
1454 [this has resulted in numerous debates. The issue isn't clear cut]
1458 Provide "voodoo" debugging of core files. This creates a zombie
1459 process as a child of the debugger, and loads it up with the data,
1460 stack, and regs of the core file. This allows you to call functions
1461 in the executable, to manipulate the data in the core file.
1467 GDB reopens the source file on every line, as you "next" through it.
1469 [still true? I've a memory of this being fixed]
1473 Perhaps "i source" should take an argument like that of "list".
1477 Remove "at 0xnnnn" from the "b foo" response, if `print address off' and if
1478 it matches the source line indicated.
1482 The prompt at end of screen should accept space as well as CR.
1486 Backtrace should point out what the currently selected frame is, in
1487 its display, perhaps showing "@3 foo (bar, ...)" or ">3 foo (bar,
1488 ...)" rather than "#3 foo (bar, ...)".
1492 "i program" should work for core files, and display more info, like what
1493 actually caused it to die.
1497 "x/10i" should shorten the long name, if any, on subsequent lines.
1501 "next" over a function that longjumps, never stops until next time you happen
1502 to get to that spot by accident. E.g. "n" over execute_command which has
1507 "set zeroprint off", don't bother printing members of structs which
1508 are entirely zero. Useful for those big structs with few useful
1513 GDB does four ioctl's for every command, probably switching terminal modes
1514 to/from inferior or for readline or something.
1518 terminal_ours versus terminal_inferior: cache state. Switch should be a noop
1519 if the state is the same, too.
1523 "i frame" shows wrong "arglist at" location, doesn't show where the args
1524 should be found, only their actual values.
1528 There should be a way for "set" commands to validate the new setting
1529 before it takes effect.
1533 "ena d" is ambiguous, why? "ena delete" seems to think it is a command!
1537 i line VAR produces "Line number not known for symbol ``var''.". I
1538 thought we were stashing that info now!
1542 We should be able to write to random files at hex offsets like adb.
1546 [elena - delete this]
1548 Handle add_file with separate text, data, and bss addresses. Maybe
1549 handle separate addresses for each segment in the object file?
1553 [Jimb/Elena delete this one]
1555 Handle free_named_symtab to cope with multiply-loaded object files
1556 in a dynamic linking environment. Should remember the last copy loaded,
1557 but not get too snowed if it finds references to the older copy.
1561 [elena delete this also]
1563 Remove all references to:
1570 now that we have BFD. All remaining are in machine dependent files.
1574 Re-organize help categories into things that tend to fit on a screen
1579 Add in commands like ADB's for searching for patterns, etc. We should
1580 be able to examine and patch raw unsymboled binaries as well in gdb as
1581 we can in adb. (E.g. increase the timeout in /bin/login without source).
1583 [actually, add ADB interface :-]
1587 When doing "step" or "next", if a few lines of source are skipped between
1588 the previous line and the current one, print those lines, not just the
1589 last line of a multiline statement.
1593 Handling of "&" address-of operator needs some serious overhaul
1594 for ANSI C and consistency on arrays and functions.
1595 For "float point[15];":
1596 ptype &point[4] ==> Attempt to take address of non-lvalue.
1597 For "char *malloc();":
1598 ptype malloc ==> "char *()"; should be same as
1599 ptype &malloc ==> "char *(*)()"
1600 call printf ("%x\n", malloc) ==> weird value, should be same as
1601 call printf ("%x\n", &malloc) ==> correct value
1605 Fix dbxread.c symbol reading in the presence of interrupts. It
1606 currently leaves a cleanup to blow away the entire symbol table when a
1607 QUIT occurs. (What's wrong with that? -kingdon, 28 Oct 1993).
1609 [I suspect that the grype was that, on a slow system, you might want
1610 to cntrl-c and get just half the symbols and then load the rest later
1611 - scary to be honest]
1615 Mipsread.c reads include files depth-first, because the dependencies
1616 in the psymtabs are way too inclusive (it seems to me). Figure out what
1617 really depends on what, to avoid recursing 20 or 30 times while reading
1622 value_add() should be subtracting the lower bound of arrays, if known,
1623 and possibly checking against the upper bound for error reporting.
1627 When listing source lines, check for a preceding \n, to verify that
1628 the file hasn't changed out from under us.
1630 [fixed by some other means I think. That hack wouldn't actually work
1631 reliably - the file might move such that another \n appears. ]
1635 Get all the remote systems (where the protocol allows it) to be able to
1636 stop the remote system when the GDB user types ^C (like remote.c
1637 does). For ebmon, use ^Ak.
1641 Possible feature: A version of the "disassemble" command which shows
1642 both source and assembly code ("set symbol-filename on" is a partial
1645 [has this been done? It was certainly done for MI and GDBtk]
1649 investigate "x/s 0" (right now stops early) (I think maybe GDB is
1650 using a 0 address for bad purposes internally).
1654 Make "info path" and path_command work again (but independent of the
1655 environment either of gdb or that we'll pass to the inferior).
1659 Make GDB understand the GCC feature for putting octal constants in
1660 enums. Make it so overflow on an enum constant does not error_type
1661 the whole type. Allow arbitrarily large enums with type attributes.
1662 Put all this stuff in the testsuite.
1666 Make TYPE_CODE_ERROR with a non-zero TYPE_LENGTH more useful (print
1667 the value in hex; process type attributes). Add this to the
1668 testsuite. This way future compilers can add new types and old
1669 versions of GDB can do something halfway reasonable.
1673 Fix mdebugread.c:parse_type to do fundamental types right (see
1674 rs6000_builtin_type in stabsread.c for what "right" is--the point is
1675 that the debug format fixes the sizes of these things and it shouldn't
1676 depend on stuff like TARGET_PTR_BIT and so on. For mdebug, there seem
1677 to be separate bt* codes for 64 bit and 32 bit things, and GDB should
1678 be aware of that). Also use a switch statement for clarity and speed.
1682 Investigate adding symbols in target_load--some targets do, some
1687 Put dirname in psymtabs and change lookup*symtab to use dirname (so
1688 /foo/bar.c works whether compiled by cc /foo/bar.c, or cd /foo; cc
1693 Merge xcoffread.c and coffread.c. Use breakpoint_re_set instead of
1698 Make a watchpoint which contains a function call an error (it is
1699 broken now, making it work is probably not worth the effort).
1703 New test case based on weird.exp but in which type numbers are not
1704 renumbered (thus multiply defining a type). This currently causes an
1705 infinite loop on "p v_comb".
1709 [Hey! Hint Hint Delete Delete!!!]
1711 Fix 386 floating point so that floating point registers are real
1712 registers (but code can deal at run-time if they are missing, like
1713 mips and 68k). This would clean up "info float" and related stuff.
1717 gcc -g -c enummask.c then gdb enummask.o, then "p v". GDB complains
1718 about not being able to access memory location 0.
1720 -------------------- enummask.c
1741 If try to modify value in file with "set write off" should give
1742 appropriate error not "cannot access memory at address 0x65e0".
1746 Allow core file without exec file on RS/6000.
1750 Make sure "shell" with no arguments works right on DOS.
1754 Make gdb.ini (as well as .gdbinit) be checked on all platforms, so
1755 the same directory can be NFS-mounted on unix or DOS, and work the
1760 [Is this another delete???]
1762 Get SECT_OFF_TEXT stuff out of objfile_relocate (might be needed to
1763 get RS/6000 to work right, might not be immediately relevant).
1767 Work out some kind of way to allow running the inferior to be done as
1768 a sub-execution of, eg. breakpoint command lists. Currently running
1769 the inferior interupts any command list execution. This would require
1770 some rewriting of wait_for_inferior & friends, and hence should
1771 probably be done in concert with the above.
1775 Add function arguments to gdb user defined functions.
1779 Add convenience variables that refer to exec file, symbol file,
1780 selected frame source file, selected frame function, selected frame
1785 Modify the handling of symbols grouped through BINCL/EINCL stabs to
1786 allocate a partial symtab for each BINCL/EINCL grouping. This will
1787 seriously decrease the size of inter-psymtab dependencies and hence
1788 lessen the amount that needs to be read in when a new source file is
1793 Add a command for searching memory, a la adb. It specifies size,
1794 mask, value, start address. ADB searches until it finds it or hits
1795 an error (or is interrupted).
1799 Remove the range and type checking code and documentation, if not