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 RFD: infrun.c: No bpstat_stop_status call after proceed over break?
16 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00665.html
18 GDB misses watchpoint triggers after proceeding over a breakpoint on
23 x86 linux GDB and SIGALRM (???)
24 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00803.html
26 This problem has been fixed, but a regression test still needs to be
27 added to the testsuite:
28 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00309.html
34 Can't build IRIX -> arm GDB.
35 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00356.html
38 > Now I'm building for an embedded arm target. If there is a way of turning
39 > remote-rdi off, I couldn't find it. It looks like it gets built by default
40 > in gdb/configure.tgt(line 58) Anyway, the build dies in
41 > gdb/rdi-share/unixcomm.c. SERPORT1 et. al. never get defined because we
42 > aren't one of the architectures supported.
46 Problem with weak functions
47 http://sourceware.cygnus.com/ml/gdb/2000-05/msg00060.html
49 Dan Nicolaescu writes:
50 > It seems that gdb-4.95.1 does not display correctly the function when
51 > stoping in weak functions.
53 > It stops in a function that is defined as weak, not in the function
54 > that is actually run...
58 GDB 5.0 doesn't work on Linux/SPARC
62 Thread support. Right now, as soon as a thread finishes and exits,
63 you're hosed. This problem is reported once a week or so.
67 Wow, three bug reports for the same problem in one day! We should
68 probably make fixing this a real priority :-).
70 Anyway, thanks for reporting.
72 The following patch will fix the problems with setting breakpoints in
73 dynamically loaded objects:
75 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00230.html
77 This patch isn't checked in yet (ping Michael/JimB), but I hope this
78 will be in the next GDB release.
80 There should really be a test in the testsuite for this problem, since
81 it keeps coming up :-(. Any volunteers?
88 http://sources.redhat.com/ml/gdb/2000-07/msg00038.html
90 Is the Solaris 8 x86 problem fixed? When you configure it, configure
91 incorrectly determines that I have no curses.h. This causes mucho
92 compilation errors later on.
94 Simply editing the config.h to define CURSES_H fixes the problem, and
95 then the build works fine.
97 The status for this problem:
99 Solaris 8 x86 (PIII-560)
102 I had the same problem with several of the snapshots shortly before
103 5.0 became official, and 5.0 has the same problem.
105 I sent some mail in about it long ago, and never saw a reply.
107 I haven't had time to figure it out myself, especially since I get all
108 confused trying to figure out what configure does, I was happy to find
115 GDB 5.1 - New features
116 ======================
118 The following new features should be included in 5.1.
122 Enable MI by default. Old code can be deleted after 5.1 is out.
126 Pascal (Pierre Muller, David Taylor)
128 Pierre Muller has contributed patches for adding Pascal Language
131 2 pascal language patches inserted in database
132 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00521.html
135 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00496.html
139 Java (Anthony Green, David Taylor)
141 Anthony Green has a number of Java patches that did not make it into
142 the 5.0 release. The first two are in cvs now, but the third needs
143 some fixing up before it can go in.
146 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00512.html
149 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00515.html
151 Patch: handle N_MAIN stab
152 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00527.html
158 Modify gdb to work correctly with Pascal.
162 Revised UDP support (was: Re: [Fwd: [patch] UDP transport support])
163 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00000.html
165 (Broken) support for GDB's remote protocol across UDP is to be
166 included in the follow-on release.
168 It should be noted that UDP can only work when the [Gg] packet fits in
171 There is also much debate over the merit of this.
178 The following code cleanups will hopefully be applied to GDB 5.1.
182 Change documentation to GFDL license.
184 ``It is time to make an effort to start using the GFDL more
185 thoroughly. Would all GNU maintainers please change the license to
186 the GFDL, for all manuals and other major documentation files?
188 The GFDL and some instructions for using it can be found in
189 http://www.gnu.org/copyleft/''
195 Delete macro TARGET_BYTE_ORDER_SELECTABLE.
197 Patches in the database.
201 Fix copyright notices.
203 Turns out that ``1998-2000'' isn't considered valid :-(
205 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00467.html
211 Eliminate all uses of PARAMS in GDB's source code.
215 printcmd.c (print_address_numeric):
217 NOTE: This assumes that the significant address information is kept in
218 the least significant bits of ADDR - the upper bits were either zero
219 or sign extended. Should ADDRESS_TO_POINTER() or some
220 ADDRESS_TO_PRINTABLE() be used to do the conversion?
226 Eliminate all warnings for at least one host/target for the flags:
227 -Wimplicit -Wreturn-type -Wcomment -Wtrigraphs -Wformat -Wparentheses
228 -Wpointer-arith -Wuninitialized
232 Follow through `make check' with --enable-shared.
234 When the srcware tree is configured with --enable-shared, the `expect'
235 program won't run properly. Jim Wilson found out gdb has a local hack
236 to set LD_LIBRARY_PATH, but, AFAIK, no other project has been hacked
239 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00845.html
243 GDB 5.1 Known Problems
244 ======================
250 The z8k has suffered bit rot and is known to not build. The problem
251 was occuring in the opcodes directory.
260 Fix at least one thread bug.
264 GDB 5.2 - New features
265 ======================
269 GCC 3.0 ABI support (but hopefully sooner...).
273 Objective C/C++ support (but hopefully sooner...).
280 The following cleanups have been identified as part of GDB 5.2.
284 Remove old code that does not use ui_out functions and all the related
285 "ifdef"s. This also allows the elimination of -DUI_OUT from
286 Makefile.in and configure.in.
290 Eliminate one more compiler warnings.
294 Restructure gdb directory tree so that it avoids any 8.3 and 14
299 Convert GDB build process to AUTOMAKE.
301 See also sub-directory configure below.
303 The current convention is (kind of) to use $(<header>_h) in all
304 dependency lists. It isn't done in a consistent way.
308 Code Cleanups: General
309 ======================
311 The following are more general cleanups and fixes. They are not tied
312 to any specific release.
316 The BFD directory requires bug-fixed AUTOMAKE et.al.
318 AUTOMAKE 1.4 incorrectly set the TEXINPUTS environment variable. It
319 contained the full path to texinfo.tex when it should have only
320 contained the directory. The bug has been fixed in the current
321 AUTOMAKE sources. Automake snapshots can be found in:
322 ftp://sourceware.cygnus.com/pub/gdb/snapshots
323 and ftp://sourceware.cygnus.com/pub/binutils
327 Find something better than DEFAULT_BFD_ARCH, DEFAULT_BFD_VEC to
328 determine the default isa/byte-order.
332 Rely on BFD_BIG_ENDIAN and BFD_LITTLE_ENDIAN instead of host dependent
333 BIG_ENDIAN and LITTLE_ENDIAN.
337 Eliminate more compiler warnings.
339 Of course there also needs to be the usual debate over which warnings
340 are valid and how to best go about this.
342 One method: choose a single option; get agreement that it is
343 reasonable; try it out to see if there isn't anything silly about it
344 (-Wunused-parameters is an example of that) then incrementally hack
347 The other method is to enable all warnings and eliminate them from one
352 Elimination of ``(catch_errors_ftype *) func''.
354 Like make_cleanup_func it isn't portable.
355 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00791.html
356 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00814.html
360 Nuke #define CONST_PTR.
368 [PATCH/5] src/intl/Makefile.in:distclean additions
369 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00363.html
371 Do not forget to merge the patch back into the trunk.
375 Rationalize the host-endian code (grep for HOST_BYTE_ORDER).
377 At present defs.h includes <endian.h> (which is linux specific) yet
378 almost nothing depends on it. Suggest "gdb_endian.h" which can also
379 handle <machine/endian.h> and only include that where it is really
384 Replace savestring() with something from libiberty.
386 An xstrldup()? but that would have different semantics.
390 Rationalize use of floatformat_unknown in GDB sources.
392 Instead of defaulting to floatformat_unknown, should hosts/targets
393 specify the value explicitly?
395 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00447.html
399 Add a ``name'' member to include/floatformat.h:struct floatformat.
400 Print that name in gdbarch.c.
404 Sort out the harris mess in include/floatformat.h (it hardwires two
405 different floating point formats).
409 See of the GDB local floatformat_do_doublest() and libiberty's
410 floatformat_to_double (which was once GDB's ...) can be merged some
415 Eliminate mmalloc(), mstrsave() et.al. from GDB.
417 Also eliminate it from defs.h.
421 Eliminate PTR. ISO-C allows ``void *''.
427 GDB should never abort. GDB should either throw ``error ()'' or
428 ``internal_error ()''. Better still GDB should naturally unwind with
433 GDB probably doesn't build on FreeBSD pre 2.2.x
434 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00378.html
436 Fixes to get FreeBSD working on 2.2.x, 3.x and 4.x caused the code to
441 Deprecate "fg". Apparently ``fg'' is actually continue.
443 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00417.html
447 Deprecate current use of ``floatformat_unknown''.
449 Require all targets to explicitly provide their float format instead
450 of defaulting to floatformat unknown. Doing the latter leads to nasty
453 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00447.html
457 Rationalize floatformat_to_double() vs floatformat_to_doublest().
459 Looks like GDB migrated floatformat_to_double() to libiberty but then
460 turned around and created a ..._to_doublest() the latter containing
463 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00472.html
467 Move floatformat_ia64_ext to libiberty/include floatformat.[ch].
469 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00466.html
473 The ``maintenance deprecate set endian big'' command doesn't notice
474 that it is deprecating ``set endian'' and not ``set endian big'' (big
475 is implemented using an enum). Is anyone going to notice this?
479 When tab expanding something like ``set arch<tab>'' ignore the
480 deprecated ``set archdebug'' and expand to ``set architecture''.
484 Eliminate ``arm_register_names[j] = (char *) regnames[j]'' and the
485 like from arm-tdep.c.
489 Fix uses of ->function.cfunc = set_function().
491 The command.c code calls sfunc() when a set command. Rather than
492 change it suggest fixing the callback function so that it is more
495 http://sourceware.cygnus.com/ml/gdb-patches/2000-06/msg00062.html
497 See also ``Fix implementation of ``target xxx''.'' below.
501 IRIX 3.x support is probably broken.
505 Delete sim/SIM_HAVE_BREAKPOINTS and gdb/SIM_HAS_BREAKPOINTS.
506 http://sourceware.cygnus.com/ml/gdb-patches/2000-07/msg00042.html
508 Apart from the d30v, are there any sim/common simulators that make use
511 A brief summary of what happened is that sim/common/sim-break.c was
512 created as a good idea. It turned out a better idea was to use
513 SIM_SIGBREAK and have GDB pass back sim_resume (..., SIGBREAK).
517 Move remote_remove_hw_breakpoint, remote_insert_hw_breakpoint,
518 remote_remove_watchpoint, remote_insert_watchpoint into target vector.
522 Eliminate ``extern'' from C files.
526 Replace ``STREQ()'' et.al. with ``strcmp() == 0'' et.al.
528 Extreme care is recommeded - perhaps only modify tests that are
529 exercised by the testsuite (as determined using some type of code
534 Replace the file gdb/CONTRIBUTE with a file that is generated from the
535 gdb/doc/*.texinfo directory.
539 New Features and Fixes
540 ======================
542 These are harder than cleanups but easier than work involving
543 fundamental architectural change.
547 Hardware watchpoint problems on x86 OSes, including Linux:
549 1. Delete/disable hardware watchpoints should free hardware debug
551 2. Watch for different values on a viariable with one hardware debug
554 According to Eli Zaretskii <eliz@delorie.com>:
556 These are not GDB/ia32 issues per se: the above features are all
557 implemented in the DJGPP port of GDB and work in v5.0. Every
558 x86-based target should be able to lift the relevant parts of
559 go32-nat.c and use them almost verbatim. You get debug register
560 sharing through reference counts, and the ability to watch large
561 regions (up to 16 bytes) using multiple registers. (The required
562 infrastructure in high-level GDB application code, mostly in
563 breakpoint.c, is also working since v5.0.)
567 Add built-by, build-date, tm, xm, nm and anything else into gdb binary
568 so that you can see how the GDB was created.
572 Add an "info bfd" command that displays supported object formats,
573 similarly to objdump -i.
575 Is there a command already?
579 Fix ``I'm sorry, Dave, I can't do that.'' from symfile.c.
581 This requires internationalization.
587 (gdb) p fwprintf(stdout,L"%S\n", f)
588 No symbol "L" in current context.
592 Cleanup configury support for optional sub-directories.
594 Check how GCC handles multiple front ends for an example of how things
595 could work. A tentative first step is to rationalize things so that
596 all sub directories are handled in a fashion similar to gdb/mi.
598 See also automake above.
602 Add a transcript mechanism to GDB.
604 Such a mechanism might log all gdb input and output to a file in a
605 form that would allow it to be replayed. It could involve ``gdb
606 --transcript=FILE'' or it could involve ``(gdb) transcript file''.
610 Can the xdep files be replaced by autoconf?
614 Document trace machinery
618 Document ui-out and ui-file.
620 http://sourceware.cygnus.com/ml/gdb/2000-04/msg00121.html
624 Update texinfo.tex to latest?
628 Incorporate agentexpr.texi into gdb.texinfo
630 agentexpr.texi mostly describes the details of the byte code used for
631 tracepoints, not the internals of the support for this in GDB. So it
632 looks like gdb.texinfo is a better place for this information.
634 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00566.html
638 Document overlay machinery.
642 ``(gdb) catch signal SIGNAL''
644 Overlaps with ``handle SIGNAL'' but the implied behavior is different.
645 You can attach commands to a catch but not a handle. A handle has a
646 limited number of hardwired actions.
650 Get the TUI working on all platforms.
654 Add support for ``gdb --- PROGRAM ARGS ...''.
655 Add support for ``gdb -cmd=...''
657 Along with many variations. Check:
659 ????? for a full discussion.
665 Implement ``(gdb) !ls''.
667 Which is very different from ``(gdb) ! ls''. Implementing the latter
670 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00034.html
674 Change the (char *list[]) to (const char (*)[]) so that dynamic lists can
679 When tab expanding something like ``set arch<tab>'' ignore the
680 deprecated ``set archdebug'' and expand to ``set architecture''.
684 Replace the code that uses the host FPU with an emulator of the target
689 The "ocd reset" command needs to flush the dcache, which requires breaking
690 the abstraction layer between the target independent and target code. One
691 way to address this is provide a generic "reset" command and target vector.
693 http://sources.redhat.com/ml/gdb-patches/2000-10/msg00011.html
702 Generic: lin-thread cannot handle thread exit (Mark Kettenis, Michael
703 Snyder) http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00525.html
705 The thread_db assisted debugging code doesn't handle exiting threads
706 properly, at least in combination with glibc 2.1.3 (the framework is
707 there, just not the actual code). There are at least two problems
708 that prevent this from working.
710 As an additional reference point, the pre thread_db code did not work
715 GNU/Linux/x86 and random thread signals (and Solaris/SPARC but not
717 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00336.html
719 Christopher Blizzard writes:
721 So, I've done some more digging into this and it looks like Jim
722 Kingdon has reported this problem in the past:
724 http://sourceware.cygnus.com/ml/bug-gdb/1999-10/msg00058.html
726 I can reproduce this problem both with and without Tom's patch. Has
727 anyone seen this before? Maybe have a solution for it hanging around?
730 There's a test case for this documented at:
732 when debugging threaded applications you get extra SIGTRAPs
733 http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=9565
735 [There should be a GDB testcase - cagney]
739 GDB5 TOT on unixware 7
740 http://sourceware.cygnus.com/ml/gdb/2000-04/msg00119.html
743 > I just spun the top of tree of the GDB5 branch on UnixWare 7. As a
744 > practical matter, the current thread support is somewhat more annoying
745 > than when GDB was thread-unaware.
749 Migrate qfThreadInfo packet -> qThreadInfo. (Andrew Cagney)
751 Add support for packet enable/disable commands with these thread
752 packets. General cleanup.
754 [PATCH] Document the ThreadInfo remote protocol queries
755 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00832.html
757 [PATCH] "info threads" queries for remote.c
758 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00831.html
765 New languages come onto the scene all the time.
769 Re: Various C++ things
771 value_headof/value_from_vtable_info are worthless, and should be
772 removed. The one place in printcmd.c that uses it should use the RTTI
775 RTTI for g++ should be using the typeinfo functions rather than the
776 vtables. The typeinfo functions are always at offset 4 from the
777 beginning of the vtable, and are always right. The vtables will have
778 weird names like E::VB sometimes. The typeinfo function will always
779 be "E type_info function", or somesuch.
781 value_virtual_fn_field needs to be fixed so there are no failures for
782 virtual functions for C++ using g++.
784 Testsuite cases are the major priority right now for C++ support,
785 since i have to make a lot of changes that could potentially break
790 Add support for Modula3
792 Get DEC/Compaq to contribute their Modula-3 support.
796 Remote Protocol Support
797 =======================
801 Remote protocol doco feedback.
803 Too much feedback to mention needs to be merged in (901660). Search
804 for the word ``remote''.
807 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00023.html
808 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00056.html
809 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00382.html
813 GDB doesn't recover gracefully from remote protocol errors.
815 GDB wasn't checking for NAKs from the remote target. Instead a NAK is
816 ignored and a timeout is required before GDB retries. A pre-cursor to
817 fixing this this is making GDB's remote protocol packet more robust.
819 While downloading to a remote protocol target, gdb ignores packet
820 errors in so far as it will continue to download with chunk N+1 even
821 if chunk N was not correctly sent. This causes gdb.base/remote.exp to
822 take a painfully long time to run. As a PS that test needs to be
823 fixed so that it builds on 16 bit machines.
827 Add the cycle step command.
829 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00237.html
833 Resolve how to scale things to support very large packets.
837 Resolve how to handle a target that changes things like its endianess
838 on the fly - should it be returned in the ``T'' packet?
840 Underlying problem is that the register file is target endian. If the
841 target endianess changes gdb doesn't know.
845 Rename read_register{,_pid}() to read_unsigned_register{,_pid}().
852 If / when GDB starts to support the debugging of multi-processor
853 (rather than multi-thread) applications the symtab code will need to
854 be updated a little so that several independent symbol tables are
855 active at a given time.
857 The other interesting change is a clarification of the exact meaning
858 of CORE_ADDR and that has had consequences for a few targets (that
859 were abusing that data type).
863 Investiagate ways of reducing memory.
867 Investigate ways of improving load time.
871 Get the d10v to use POINTER_TO_ADDRESS and ADDRESS_TO_POINTER.
873 Consequence of recent symtab clarification. No marks for figuring out
874 who maintains the d10v.
878 Get the MIPS to correctly sign extend all address <-> pointer
881 Consequence of recent symtab clarification. No marks for figuring out
882 who maintains the MIPS.
886 GDB truncates 64 bit enums.
888 http://sourceware.cygnus.com/ml/gdb-patches/2000-06/msg00290.html
895 There are never to many testcases.
899 Better thread testsuite.
903 Better C++ testsuite.
907 Look at adding a GDB specific testsuite directory so that white box
908 tests of key internals can be added (eg ui_file).
912 Separate out tests that involve the floating point (FP).
914 (Something for people brining up new targets). FP and non-fp tests
915 are combined. I think there should be set of basic tests that
916 exercise pure integer support and then a more expanded set that
917 exercise FP and FP/integer interactions.
919 As an example, the MIPS, for n32 as problems with passing FP's and
920 structs. Since most inferior call tests include FP it is difficult to
921 determine of the integer tests are ok.
925 Architectural Changes: General
926 ==============================
928 These are harder than simple cleanups / fixes and, consequently
929 involve more work. Typically an Architectural Change will be broken
930 down into a more digestible set of cleanups and fixes.
934 Cleanup software single step.
936 At present many targets implement software single step by directly
937 blatting memory (see rs6000-tdep.c). Those targets should register
938 the applicable breakpoints using the breakpoint framework. Perhaphs a
939 new internal breakpoint class ``step'' is needed.
943 Replace READ_FP() with FRAME_HANDLE().
945 READ_FP() is a hangover from the days of the vax when the ABI really
946 did have a frame pointer register. Modern architectures typically
947 construct a virtual frame-handle from the stack pointer and various
948 other bits of string.
950 Unfortunately GDB still treats this synthetic FP register as though it
951 is real. That in turn really confuses users (arm and ``print $fp'' VS
952 ``info registers fp''). The synthetic FP should be separated out of
953 the true register set presented to the user.
957 Register Cache Cleanup (below from Andrew Cagney)
959 I would depict the current register architecture as something like:
966 register + REGISTER_BYTE(reg_nr)
969 -------------------------
970 | extern register[] |
971 -------------------------
973 where neither the high (valops.c et.al.) or low gdb (*-tdep.c) are
974 really clear on what mechanisms they should be using to manipulate that
975 buffer. Further, much code assumes, dangerously, that registers are
976 contigious. Having got mips-tdep.c to support multiple ABIs, believe
977 me, that is a bad assumption. Finally, that register cache layout is
978 determined by the current remote/local target and _not_ the less
979 specific target ISA. In fact, in many cases it is determined by the
980 somewhat arbitrary layout of the [gG] packets!
983 How I would like the register file to work is more like:
1011 The main objectives being:
1013 o a clear separation between the low
1014 level target and the high level GDB
1016 o a mechanism that solves the general
1017 problem of register aliases, overlaps
1018 etc instead of treating them as optional
1019 extras that can be wedged in as an after
1020 thought (that is a reasonable description
1021 of the current code).
1023 Identify then solve the hard case and the
1024 rest just falls out. GDB solved the easy
1025 case and then tried to ignore the real
1028 o a removal of the assumption that the
1029 mapping between the register cache
1030 and virtual registers is largely static.
1031 If you flip the USR/SSR stack register
1032 select bit in the status-register then
1033 the corresponding stack registers should
1036 o a mechanism that clearly separates the
1037 gdb internal register cache from any
1038 target (not architecture) dependent
1039 specifics such as [gG] packets.
1041 Of course, like anything, it sounds good in theory. In reality, it
1042 would have to contend with many<->many relationships at both the
1043 virt<->cache and cache<->target level. For instance:
1046 Modifying an mmx register may involve
1047 scattering values across both FP and
1048 mmpx specific parts of a buffer
1051 When writing back a SP it may need to
1052 both be written to both SP and USP.
1057 Rather than let this like the last time it was discussed, just slip, I'm
1058 first going to add this e-mail (+ references) to TODO. I'd then like to
1059 sketch out a broad strategy I think could get us there.
1062 First thing I'd suggest is separating out the ``extern registers[]''
1063 code so that we can at least identify what is using it. At present
1064 things are scattered across many files. That way we can at least
1065 pretend that there is a cache instead of a global array :-)
1067 I'd then suggest someone putting up a proposal for the pseudo-reg /
1068 high-level side interface so that code can be adopted to it. For old
1069 code, initially a blanket rename of write_register_bytes() to
1070 deprecated_write_register_bytes() would help.
1072 Following that would, finaly be the corresponding changes to the target.
1076 Check that GDB can handle all BFD architectures (Andrew Cagney)
1078 There should be a test that checks that BFD/GDB are in sync with
1079 regard to architecture changes. Something like a test that first
1080 queries GDB for all supported architectures and then feeds each back
1081 to GDB.. Anyone interested in learning how to write tests? :-)
1085 Architectural Change: Multi-arch et al.
1086 =======================================
1088 The long term objective is to remove all assumptions that there is a
1089 single target with a single address space with a single instruction
1090 set architecture and single application binary interface.
1092 This is an ongoing effort. The first milestone is to enable
1093 ``multi-arch'' where by all architectural decisions are made at
1096 It should be noted that ``gdbarch'' is really ``gdbabi'' and
1097 ``gdbisa''. Once things are multi-arched breaking that down correctly
1098 will become much easier.
1102 GDBARCH cleanup (Andrew Cagney)
1104 The non-generated parts of gdbarch.{sh,h,c} should be separated out
1105 into arch-utils.[hc].
1107 Document that gdbarch_init_ftype could easily fail because it didn't
1108 identify an architecture.
1112 Fix BELIEVE_PPC_PROMOTION. Change it to BELIEVE_PPC_PROMOTION_P?
1114 At present there is still #ifdef BELIEVE_PPC_PROMOTION code in the
1119 Fix target_signal_from_host() etc.
1121 The name is wrong for starters. ``target_signal'' should probably be
1122 ``gdb_signal''. ``from_host'' should be ``from_target_signal''.
1123 After that it needs to be multi-arched and made independent of any
1124 host signal numbering.
1128 Update ALPHA so that it uses ``struct frame_extra_info'' instead of
1131 This is a barrier to replacing mips_extra_func_info with something
1132 that works with multi-arch.
1136 Multi-arch mips_extra_func_info.
1138 This first needs the alpha to be updated so that it uses ``struct
1143 Rationalize TARGET_SINGLE_FORMAT and TARGET_SINGLE_BIT et al.
1145 Surely one of them is redundant.
1149 Convert ALL architectures to MULTI-ARCH.
1153 Select the initial multi-arch ISA / ABI based on --target or similar.
1155 At present the default is based on what ever is first in the BFD
1156 archures table. It should be determined based on the ``--target=...''
1161 Make MIPS pure multi-arch.
1163 It is only at the multi-arch enabled stage.
1169 Enable the code to recognize --enable-targets=.... like BINUTILS does.
1171 Can the tm.h and nm.h files be eliminated by multi-arch.
1175 Architectural Change: MI, LIBGDB and scripting languages
1176 ========================================================
1178 See also architectural changes related to the event loop. LIBGDB
1179 can't be finished until there is a generic event loop being used by
1182 The long term objective is it to be possible to integrate GDB into
1183 scripting languages.
1187 Implement generic ``(gdb) commmand > file''
1189 Once everything is going through ui_file it should be come fairly
1192 http://sourceware.cygnus.com/ml/gdb/2000-04/msg00104.html
1196 Replace gdb_stdtarg with gdb_targout (and possibly gdb_targerr).
1198 gdb_stdtarg is easily confused with gdb_stdarg.
1202 Extra ui_file methods - dump.
1204 Very useful for whitebox testing.
1208 Eliminate error_begin().
1210 With ui_file, there is no need for the statefull error_begin ()
1215 Send normal output to gdb_stdout.
1216 Send error messages to gdb_stderror.
1217 Send debug and log output log gdb_stdlog.
1219 GDB still contains many cases where (f)printf or printf_filtered () is
1220 used when it should be sending the messages to gdb_stderror or
1221 gdb_stdlog. The thought of #defining printf to something has crossed
1226 Re-do GDB's output pager.
1228 GDB's output pager still relies on people correctly using *_filtered
1229 for gdb_stdout and *_unfiltered for gdb_stdlog / gdb_stderr.
1230 Hopefully, with all normal output going to gdb_stdout, the pager can
1231 just look at the ui_file that the output is on and then use that to
1232 decide what to do about paging. Sounds good in theory.
1236 Check/cleanup MI documentation.
1238 The list of commands specified in the documentation needs to be
1239 checked against the mi-cmds.c table in a mechanical way (so that they
1240 two can be kept up-to-date).
1244 Convert MI into libgdb
1246 MI provides a text interface into what should be many of the libgdb
1247 functions. The implementation of those functions should be separated
1248 into the MI interface and the functions proper. Those functions being
1249 moved to gdb/lib say.
1255 The first part can already be found in defs.h.
1259 MI's input does not use buffering.
1261 At present the MI interface reads raw characters of from an unbuffered
1262 FD. This is to avoid several nasty buffer/race conditions. That code
1263 should be changed so that it registers its self with the event loop
1264 (on the input FD) and then push commands up to MI as they arrive.
1266 The serial code already does this.
1270 Make MI interface accessible from existing CLI.
1274 Add a breakpoint-edit command to MI.
1276 It would be similar to MI's breakpoint create but would apply to an
1277 existing breakpoint. It saves the need to delete/create breakpoints
1278 when ever they are changed.
1282 Add directory path to MI breakpoint.
1284 That way the GUI's task of finding the file within which the
1285 breakpoint was set is simplified.
1289 Add a mechanism to reject certain expression classes to MI
1291 There are situtations where you don't want GDB's expression
1292 parser/evaluator to perform inferior function calls or variable
1293 assignments. A way of restricting the expression parser so that such
1294 operations are not accepted would be very helpful.
1298 Remove sideffects from libgdb breakpoint create function.
1300 The user can use the CLI to create a breakpoint with partial
1301 information - no file (gdb would use the file from the last
1304 The libgdb interface currently affects that environment which can lead
1305 to confusion when a user is setting breakpoints via both the MI and
1308 This is also a good example of how getting the CLI ``right'' will be
1313 Move gdb_lasterr to ui_out?
1315 The way GDB throws errors and records them needs a re-think. ui_out
1316 handles the correct output well. It doesn't resolve what to do with
1317 output / error-messages when things go wrong.
1321 do_setshow_command contains a 1024 byte buffer.
1323 The function assumes that there will never be any more than 1024 bytes
1324 of enum. It should use mem_file.
1328 Should struct cmd_list_element . completer take the command as an
1333 Should the bulk of top.c:line_completion_function() be moved to
1334 command.[hc]? complete_on_cmdlist() and complete_on_enums() could
1335 then be made private.
1339 top.c (execute_command): Should a command being valid when the target
1340 is running be made an attribute (predicate) to the command rather than
1341 an explicit set of tests.
1345 top.c (execute_command): Should the bulk of this function be moved
1346 into command.[hc] so that top.c doesn't grub around in the command
1351 Architectural Change: Async
1352 ===========================
1354 While GDB uses an event loop when prompting the user for input. That
1355 event loop is not exploited by targets when they allow the target
1356 program to continue. Typically targets still block in (target_wait())
1357 until the program again halts.
1359 The closest a target comes to supporting full asynchronous mode are
1360 the remote targets ``async'' and ``extended-async''.
1364 Asynchronous expression evaluator
1366 Inferior function calls hang GDB.
1370 Fix implementation of ``target xxx''.
1372 At present when the user specifies ``target xxxx'', the CLI maps that
1373 directly onto a target open method. It is then assumed that the
1374 target open method should do all sorts of complicated things as this
1375 is the only chance it has. Check how the various remote targets
1376 duplicate the target operations. Check also how the various targets
1377 behave differently for purely arbitrary reasons.
1379 What should happen is that ``target xxxx'' should call a generic
1380 ``target'' function and that should then co-ordinate the opening of
1381 ``xxxx''. This becomes especially important when you're trying to
1382 open an asynchronous target that may need to perform background tasks
1383 as part of the ``attach'' phase.
1385 Unfortunately, due to limitations in the old/creaking command.h
1386 interface, that isn't possible. The function being called isn't told
1387 of the ``xxx'' or any other context information.
1389 Consequently a precursor to fixing ``target xxxx'' is to clean up the
1390 CLI code so that it passes to the callback function (attatched to a
1391 command) useful information such as the actual command and a context
1392 for that command. Other changes such as making ``struct command''
1393 opaque may also help.
1396 http://sourceware.cygnus.com/ml/gdb-patches/2000-06/msg00062.html
1400 Make "target xxx" command interruptible.
1402 As things become async this becomes possible. A target would start
1403 the connect and then return control to the event loop. A cntrl-c
1404 would notify the target that the operation is to be abandoned and the
1405 target code could respond.
1409 Add a "suspend" subcommand of the "continue" command to suspend gdb
1410 while continuing execution of the subprocess. Useful when you are
1411 debugging servers and you want to dodge out and initiate a connection
1412 to a server running under gdb.
1421 Frequently requested but not approved requests.
1425 Eliminate unused argument warnings using ATTRIBUTE_UNUSED.
1427 The benefits on this one are thought to be marginal - GDBs design
1428 means that unused parameters are very common. GCC 3.0 will also
1429 include the option -Wno-unused-parameter which means that ``-Wall
1430 -Wno-unused-parameters -Werror'' can be specified.
1439 This list is not up to date, and opinions vary about the importance or
1440 even desirability of some of the items. If you do fix something, it
1441 always pays to check the below.
1445 @c This does not work (yet if ever). FIXME.
1446 @c @item --parse=@var{lang} @dots{}
1447 @c Configure the @value{GDBN} expression parser to parse the listed languages.
1448 @c @samp{all} configures @value{GDBN} for all supported languages. To get a
1449 @c list of all supported languages, omit the argument. Without this
1450 @c option, @value{GDBN} is configured to parse all supported languages.
1454 START_INFERIOR_TRAPS_EXPECTED need never be defined to 2, since that
1455 is its default value. Clean this up.
1459 It should be possible to use symbols from shared libraries before we know
1460 exactly where the libraries will be loaded. E.g. "b perror" before running
1461 the program. This could maybe be done as an extension of the "breakpoint
1462 re-evaluation" after new symbols are loaded.
1466 Make single_step() insert and remove breakpoints in one operation.
1468 [If this is talking about having single_step() insert the breakpoints,
1469 run the target then pull the breakpoints then it is wrong. The
1470 function has to return as control has to eventually be passed back to
1471 the main event loop.]
1475 Speed up single stepping by avoiding extraneous ptrace calls.
1479 Speed up single stepping by not inserting and removing breakpoints
1480 each time the inferior starts and stops.
1482 Breakpoints should not be inserted and deleted all the time. Only the
1483 one(s) there should be removed when we have to step over one. Support
1484 breakpoints that don't have to be removed to step over them.
1486 [this has resulted in numerous debates. The issue isn't clear cut]
1490 Provide "voodoo" debugging of core files. This creates a zombie
1491 process as a child of the debugger, and loads it up with the data,
1492 stack, and regs of the core file. This allows you to call functions
1493 in the executable, to manipulate the data in the core file.
1499 GDB reopens the source file on every line, as you "next" through it.
1501 [still true? I've a memory of this being fixed]
1505 Perhaps "i source" should take an argument like that of "list".
1509 Remove "at 0xnnnn" from the "b foo" response, if `print address off' and if
1510 it matches the source line indicated.
1514 The prompt at end of screen should accept space as well as CR.
1518 Backtrace should point out what the currently selected frame is, in
1519 its display, perhaps showing "@3 foo (bar, ...)" or ">3 foo (bar,
1520 ...)" rather than "#3 foo (bar, ...)".
1524 "i program" should work for core files, and display more info, like what
1525 actually caused it to die.
1529 "x/10i" should shorten the long name, if any, on subsequent lines.
1533 "next" over a function that longjumps, never stops until next time you happen
1534 to get to that spot by accident. E.g. "n" over execute_command which has
1539 "set zeroprint off", don't bother printing members of structs which
1540 are entirely zero. Useful for those big structs with few useful
1545 GDB does four ioctl's for every command, probably switching terminal modes
1546 to/from inferior or for readline or something.
1550 terminal_ours versus terminal_inferior: cache state. Switch should be a noop
1551 if the state is the same, too.
1555 "i frame" shows wrong "arglist at" location, doesn't show where the args
1556 should be found, only their actual values.
1560 There should be a way for "set" commands to validate the new setting
1561 before it takes effect.
1565 "ena d" is ambiguous, why? "ena delete" seems to think it is a command!
1569 i line VAR produces "Line number not known for symbol ``var''.". I
1570 thought we were stashing that info now!
1574 We should be able to write to random files at hex offsets like adb.
1578 [elena - delete this]
1580 Handle add_file with separate text, data, and bss addresses. Maybe
1581 handle separate addresses for each segment in the object file?
1585 [Jimb/Elena delete this one]
1587 Handle free_named_symtab to cope with multiply-loaded object files
1588 in a dynamic linking environment. Should remember the last copy loaded,
1589 but not get too snowed if it finds references to the older copy.
1593 [elena delete this also]
1595 Remove all references to:
1602 now that we have BFD. All remaining are in machine dependent files.
1606 Re-organize help categories into things that tend to fit on a screen
1611 Add in commands like ADB's for searching for patterns, etc. We should
1612 be able to examine and patch raw unsymboled binaries as well in gdb as
1613 we can in adb. (E.g. increase the timeout in /bin/login without source).
1615 [actually, add ADB interface :-]
1619 When doing "step" or "next", if a few lines of source are skipped between
1620 the previous line and the current one, print those lines, not just the
1621 last line of a multiline statement.
1625 Handling of "&" address-of operator needs some serious overhaul
1626 for ANSI C and consistency on arrays and functions.
1627 For "float point[15];":
1628 ptype &point[4] ==> Attempt to take address of non-lvalue.
1629 For "char *malloc();":
1630 ptype malloc ==> "char *()"; should be same as
1631 ptype &malloc ==> "char *(*)()"
1632 call printf ("%x\n", malloc) ==> weird value, should be same as
1633 call printf ("%x\n", &malloc) ==> correct value
1637 Fix dbxread.c symbol reading in the presence of interrupts. It
1638 currently leaves a cleanup to blow away the entire symbol table when a
1639 QUIT occurs. (What's wrong with that? -kingdon, 28 Oct 1993).
1641 [I suspect that the grype was that, on a slow system, you might want
1642 to cntrl-c and get just half the symbols and then load the rest later
1643 - scary to be honest]
1647 Mipsread.c reads include files depth-first, because the dependencies
1648 in the psymtabs are way too inclusive (it seems to me). Figure out what
1649 really depends on what, to avoid recursing 20 or 30 times while reading
1654 value_add() should be subtracting the lower bound of arrays, if known,
1655 and possibly checking against the upper bound for error reporting.
1659 When listing source lines, check for a preceding \n, to verify that
1660 the file hasn't changed out from under us.
1662 [fixed by some other means I think. That hack wouldn't actually work
1663 reliably - the file might move such that another \n appears. ]
1667 Get all the remote systems (where the protocol allows it) to be able to
1668 stop the remote system when the GDB user types ^C (like remote.c
1669 does). For ebmon, use ^Ak.
1673 Possible feature: A version of the "disassemble" command which shows
1674 both source and assembly code ("set symbol-filename on" is a partial
1677 [has this been done? It was certainly done for MI and GDBtk]
1681 investigate "x/s 0" (right now stops early) (I think maybe GDB is
1682 using a 0 address for bad purposes internally).
1686 Make "info path" and path_command work again (but independent of the
1687 environment either of gdb or that we'll pass to the inferior).
1691 Make GDB understand the GCC feature for putting octal constants in
1692 enums. Make it so overflow on an enum constant does not error_type
1693 the whole type. Allow arbitrarily large enums with type attributes.
1694 Put all this stuff in the testsuite.
1698 Make TYPE_CODE_ERROR with a non-zero TYPE_LENGTH more useful (print
1699 the value in hex; process type attributes). Add this to the
1700 testsuite. This way future compilers can add new types and old
1701 versions of GDB can do something halfway reasonable.
1705 Fix mdebugread.c:parse_type to do fundamental types right (see
1706 rs6000_builtin_type in stabsread.c for what "right" is--the point is
1707 that the debug format fixes the sizes of these things and it shouldn't
1708 depend on stuff like TARGET_PTR_BIT and so on. For mdebug, there seem
1709 to be separate bt* codes for 64 bit and 32 bit things, and GDB should
1710 be aware of that). Also use a switch statement for clarity and speed.
1714 Investigate adding symbols in target_load--some targets do, some
1719 Put dirname in psymtabs and change lookup*symtab to use dirname (so
1720 /foo/bar.c works whether compiled by cc /foo/bar.c, or cd /foo; cc
1725 Merge xcoffread.c and coffread.c. Use breakpoint_re_set instead of
1730 Make a watchpoint which contains a function call an error (it is
1731 broken now, making it work is probably not worth the effort).
1735 New test case based on weird.exp but in which type numbers are not
1736 renumbered (thus multiply defining a type). This currently causes an
1737 infinite loop on "p v_comb".
1741 [Hey! Hint Hint Delete Delete!!!]
1743 Fix 386 floating point so that floating point registers are real
1744 registers (but code can deal at run-time if they are missing, like
1745 mips and 68k). This would clean up "info float" and related stuff.
1749 gcc -g -c enummask.c then gdb enummask.o, then "p v". GDB complains
1750 about not being able to access memory location 0.
1752 -------------------- enummask.c
1773 If try to modify value in file with "set write off" should give
1774 appropriate error not "cannot access memory at address 0x65e0".
1778 Allow core file without exec file on RS/6000.
1782 Make sure "shell" with no arguments works right on DOS.
1786 Make gdb.ini (as well as .gdbinit) be checked on all platforms, so
1787 the same directory can be NFS-mounted on unix or DOS, and work the
1792 [Is this another delete???]
1794 Get SECT_OFF_TEXT stuff out of objfile_relocate (might be needed to
1795 get RS/6000 to work right, might not be immediately relevant).
1799 Work out some kind of way to allow running the inferior to be done as
1800 a sub-execution of, eg. breakpoint command lists. Currently running
1801 the inferior interupts any command list execution. This would require
1802 some rewriting of wait_for_inferior & friends, and hence should
1803 probably be done in concert with the above.
1807 Add function arguments to gdb user defined functions.
1811 Add convenience variables that refer to exec file, symbol file,
1812 selected frame source file, selected frame function, selected frame
1817 Modify the handling of symbols grouped through BINCL/EINCL stabs to
1818 allocate a partial symtab for each BINCL/EINCL grouping. This will
1819 seriously decrease the size of inter-psymtab dependencies and hence
1820 lessen the amount that needs to be read in when a new source file is
1825 Add a command for searching memory, a la adb. It specifies size,
1826 mask, value, start address. ADB searches until it finds it or hits
1827 an error (or is interrupted).
1831 Remove the range and type checking code and documentation, if not