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 Wow, three bug reports for the same problem in one day! We should
16 probably make fixing this a real priority :-).
18 Anyway, thanks for reporting.
20 The following patch will fix the problems with setting breakpoints in
21 dynamically loaded objects:
23 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00230.html
25 This patch isn't checked in yet (ping Michael/JimB), but I hope this
26 will be in the next GDB release.
28 There should really be a test in the testsuite for this problem, since
29 it keeps coming up :-(. Any volunteers?
35 x86 linux GDB and SIGALRM (???)
36 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00803.html
38 This problem has been fixed, but a regression test still needs to be
39 added to the testsuite:
40 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00309.html
44 [The test has been submitted for approval - cagney]
48 RFD: infrun.c: No bpstat_stop_status call after proceed over break?
49 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00665.html
51 GDB misses watchpoint triggers after proceeding over a breakpoint on
56 GDB 5.0 doesn't work on Linux/SPARC
58 There are two parts to this.
60 o GDB 5.0 doesn't work on GNU/Linux/SPARC32
62 o GDB 5.0 doesn't work on the new target
65 GDB does build on both these targets.
67 The first problem is the one that should be fixed.
71 GDB 5.1 - New features
72 ======================
74 The following new features should be included in 5.1.
78 Pascal (Pierre Muller, David Taylor)
80 Pierre Muller has contributed patches for adding Pascal Language
83 2 pascal language patches inserted in database
84 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00521.html
87 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00496.html
89 [I think this has been merged, need to confirm - cagney]
93 Java (Anthony Green, David Taylor)
95 Anthony Green has a number of Java patches that did not make it into
96 the 5.0 release. The first two are in cvs now, but the third needs
97 some fixing up before it can go in.
100 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00512.html
103 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00515.html
105 Patch: handle N_MAIN stab
106 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00527.html
112 A predicate to this is the multi-arching of SOFTWARE_SINGLE_STEP(). A
113 patch has been submitted.
120 The following code cleanups will hopefully be applied to GDB 5.1.
124 Fix copyright notices.
126 Turns out that ``1998-2000'' isn't considered valid :-(
128 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00467.html
132 GDB 5.1 - Known Problems
133 ========================
139 The z8k has suffered bit rot and is known to not build. The problem
140 was occuring in the opcodes directory.
146 The m88k has suffered bit rot and is known to not build.
150 Solaris 8 x86 CURSES_H problem
151 http://sources.redhat.com/ml/gdb/2000-07/msg00038.html
153 The original problem was worked around with:
155 2000-06-06 Michael Snyder <msnyder@cygnus.com>
157 * configure.in: Enable autoconf to find curses.h on Solaris 2.8.
158 * configure: Regenerate.
160 When building both GDB and SID using the same source tree the problem
161 will still occure. sid/component/configure.in mis-configures
162 <curses.h> and leaves wrong information in the config cache.
171 GDB 5.2 - New features
172 ======================
176 GCC 3.0 ABI support (but hopefully sooner...).
180 Objective C/C++ support (but hopefully sooner...).
184 Import of readline 4.2
191 The following cleanups have been identified as part of GDB 5.2.
195 Remove old code that does not use ui_out functions and all the related
196 "ifdef"s. This also allows the elimination of -DUI_OUT from
197 Makefile.in and configure.in.
203 Eliminate warnings for all targets on at least one host for one of the
204 -W flags. Flags up for debate include: -Wswitch -Wcomment -trigraphs
205 -Wtrigraphs -Wunused-function -Wunused-label -Wunused-variable
206 -Wunused-value -Wchar-subscripts -Wtraditional -Wshadow -Wcast-qual
207 -Wcast-align -Wwrite-strings -Wconversion -Wstrict-prototypes
208 -Wmissing-prototypes -Wmissing-declarations -Wredundant-decls
209 -Woverloaded-virtual -Winline
213 Deprecate, if not delete, the following:
219 Replaced by, on the target side
221 and on core-gdb side:
222 {read,write}_register_gen()
223 Remote.c will need to use something
224 other than REGISTER_BYTE() and
225 REGISTER_RAW_SIZE() when unpacking
228 STORE_PSEUDO_REGISTER
229 FETCH_PSEUDO_REGISTER
230 Now handed by the methods
231 gdbarch_{read,write}_register()
232 which sits between core GDB and
236 REGISTER_CONVERT_TO_RAW
237 REGISTER_CONVERT_TO_VIRTUAL
238 I think these three are redundant.
239 gdbarch_register_{read,write} can
240 do any conversion it likes.
242 REGISTER_VIRTUAL_SIZE
243 MAX_REGISTER_VIRTUAL_SIZE
244 REGISTER_VIRTUAL_TYPE
245 I think these can be replaced by
247 FRAME_REGISTER_TYPE(frame, regnum)
248 REGISTER_TYPE(regnum)
252 FRAME_REGISTER_INFO (frame, ...)
255 If nothing else rename this so that
256 how it relates to rawreg and the
260 The size of the cache can be computed
263 IS_TRAPPED_INTERNALVAR
264 The pseudo registers should eventually make
269 Obsolete the targets.
275 Obsolete the protocols:
279 ``As of version 5.3, WindRiver has removed the RDB server (RDB
280 protocol support is built into gdb).'' -- Till.
284 Restructure gdb directory tree so that it avoids any 8.3 and 14
289 Convert GDB build process to AUTOMAKE.
291 See also sub-directory configure below.
293 The current convention is (kind of) to use $(<header>_h) in all
294 dependency lists. It isn't done in a consistent way.
298 GDB 5.2 - Known Problems
299 ========================
303 Code Cleanups: General
304 ======================
306 The following are more general cleanups and fixes. They are not tied
307 to any specific release.
311 Investigate changing --target=a29k-amd-udi to a29k-*-coff* and
312 rationalize *.mt files. The got-ya is in remote-eb.c - it has its own
313 custom tty manipulation - it should be using the serial object.
317 Rename read_register{,_pid}() to read_unsigned_register{,_pid}().
321 Problem with weak functions
322 http://sourceware.cygnus.com/ml/gdb/2000-05/msg00060.html
324 Dan Nicolaescu writes:
325 > It seems that gdb-4.95.1 does not display correctly the function when
326 > stoping in weak functions.
328 > It stops in a function that is defined as weak, not in the function
329 > that is actually run...
333 Follow through `make check' with --enable-shared.
335 When the srcware tree is configured with --enable-shared, the `expect'
336 program won't run properly. Jim Wilson found out gdb has a local hack
337 to set LD_LIBRARY_PATH, but, AFAIK, no other project has been hacked
340 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00845.html
344 Delete macro TARGET_BYTE_ORDER_SELECTABLE.
346 Patches in the database.
350 printcmd.c (print_address_numeric):
352 NOTE: This assumes that the significant address information is kept in
353 the least significant bits of ADDR - the upper bits were either zero
354 or sign extended. Should ADDRESS_TO_POINTER() or some
355 ADDRESS_TO_PRINTABLE() be used to do the conversion?
359 The BFD directory requires bug-fixed AUTOMAKE et.al.
361 AUTOMAKE 1.4 incorrectly set the TEXINPUTS environment variable. It
362 contained the full path to texinfo.tex when it should have only
363 contained the directory. The bug has been fixed in the current
364 AUTOMAKE sources. Automake snapshots can be found in:
365 ftp://sourceware.cygnus.com/pub/gdb/snapshots
366 and ftp://sourceware.cygnus.com/pub/binutils
370 Find something better than DEFAULT_BFD_ARCH, DEFAULT_BFD_VEC to
371 determine the default isa/byte-order.
375 Rely on BFD_BIG_ENDIAN and BFD_LITTLE_ENDIAN instead of host dependent
376 BIG_ENDIAN and LITTLE_ENDIAN.
380 Eliminate more compiler warnings.
382 Of course there also needs to be the usual debate over which warnings
383 are valid and how to best go about this.
385 One method: choose a single option; get agreement that it is
386 reasonable; try it out to see if there isn't anything silly about it
387 (-Wunused-parameters is an example of that) then incrementally hack
390 The other method is to enable all warnings and eliminate them from one
395 Elimination of ``(catch_errors_ftype *) func''.
397 Like make_cleanup_func it isn't portable.
398 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00791.html
399 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00814.html
403 Nuke #define CONST_PTR.
411 [PATCH/5] src/intl/Makefile.in:distclean additions
412 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00363.html
414 Do not forget to merge the patch back into the trunk.
418 Rationalize the host-endian code (grep for HOST_BYTE_ORDER).
420 At present defs.h includes <endian.h> (which is linux specific) yet
421 almost nothing depends on it. Suggest "gdb_endian.h" which can also
422 handle <machine/endian.h> and only include that where it is really
427 Replace savestring() with something from libiberty.
429 An xstrldup()? but that would have different semantics.
433 Rationalize use of floatformat_unknown in GDB sources.
435 Instead of defaulting to floatformat_unknown, should hosts/targets
436 specify the value explicitly?
438 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00447.html
442 Add a ``name'' member to include/floatformat.h:struct floatformat.
443 Print that name in gdbarch.c.
447 Sort out the harris mess in include/floatformat.h (it hardwires two
448 different floating point formats).
452 See of the GDB local floatformat_do_doublest() and libiberty's
453 floatformat_to_double (which was once GDB's ...) can be merged some
458 Eliminate mmalloc(), mstrsave() et.al. from GDB.
460 Also eliminate it from defs.h.
464 Eliminate PTR. ISO-C allows ``void *''.
470 GDB should never abort. GDB should either throw ``error ()'' or
471 ``internal_error ()''. Better still GDB should naturally unwind with
476 GDB probably doesn't build on FreeBSD pre 2.2.x
477 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00378.html
479 Fixes to get FreeBSD working on 2.2.x, 3.x and 4.x caused the code to
484 Deprecate "fg". Apparently ``fg'' is actually continue.
486 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00417.html
490 Deprecate current use of ``floatformat_unknown''.
492 Require all targets to explicitly provide their float format instead
493 of defaulting to floatformat unknown. Doing the latter leads to nasty
496 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00447.html
500 Rationalize floatformat_to_double() vs floatformat_to_doublest().
502 Looks like GDB migrated floatformat_to_double() to libiberty but then
503 turned around and created a ..._to_doublest() the latter containing
506 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00472.html
510 Move floatformat_ia64_ext to libiberty/include floatformat.[ch].
512 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00466.html
516 The ``maintenance deprecate set endian big'' command doesn't notice
517 that it is deprecating ``set endian'' and not ``set endian big'' (big
518 is implemented using an enum). Is anyone going to notice this?
522 When tab expanding something like ``set arch<tab>'' ignore the
523 deprecated ``set archdebug'' and expand to ``set architecture''.
527 Eliminate ``arm_register_names[j] = (char *) regnames[j]'' and the
528 like from arm-tdep.c.
532 Fix uses of ->function.cfunc = set_function().
534 The command.c code calls sfunc() when a set command. Rather than
535 change it suggest fixing the callback function so that it is more
538 http://sourceware.cygnus.com/ml/gdb-patches/2000-06/msg00062.html
540 See also ``Fix implementation of ``target xxx''.'' below.
544 IRIX 3.x support is probably broken.
548 Delete sim/SIM_HAVE_BREAKPOINTS and gdb/SIM_HAS_BREAKPOINTS.
549 http://sourceware.cygnus.com/ml/gdb-patches/2000-07/msg00042.html
551 Apart from the d30v, are there any sim/common simulators that make use
554 A brief summary of what happened is that sim/common/sim-break.c was
555 created as a good idea. It turned out a better idea was to use
556 SIM_SIGBREAK and have GDB pass back sim_resume (..., SIGBREAK).
560 Move remote_remove_hw_breakpoint, remote_insert_hw_breakpoint,
561 remote_remove_watchpoint, remote_insert_watchpoint into target vector.
565 Eliminate ``extern'' from C files.
569 Replace ``STREQ()'' et.al. with ``strcmp() == 0'' et.al.
571 Extreme care is recommeded - perhaps only modify tests that are
572 exercised by the testsuite (as determined using some type of code
577 Replace the file gdb/CONTRIBUTE with a file that is generated from the
578 gdb/doc/*.texinfo directory.
582 Rewrite/break up sparcl-tdep.c so that it uses ser*.c as the mechanism
583 for accessing either the serial or UDP port.
587 New Features and Fixes
588 ======================
590 These are harder than cleanups but easier than work involving
591 fundamental architectural change.
595 Hardware watchpoint problems on x86 OSes, including Linux:
597 1. Delete/disable hardware watchpoints should free hardware debug
599 2. Watch for different values on a viariable with one hardware debug
602 According to Eli Zaretskii <eliz@delorie.com>:
604 These are not GDB/ia32 issues per se: the above features are all
605 implemented in the DJGPP port of GDB and work in v5.0. Every
606 x86-based target should be able to lift the relevant parts of
607 go32-nat.c and use them almost verbatim. You get debug register
608 sharing through reference counts, and the ability to watch large
609 regions (up to 16 bytes) using multiple registers. (The required
610 infrastructure in high-level GDB application code, mostly in
611 breakpoint.c, is also working since v5.0.)
615 Add built-by, build-date, tm, xm, nm and anything else into gdb binary
616 so that you can see how the GDB was created.
620 Add an "info bfd" command that displays supported object formats,
621 similarly to objdump -i.
623 Is there a command already?
627 Fix ``I'm sorry, Dave, I can't do that.'' from symfile.c.
629 This requires internationalization.
635 (gdb) p fwprintf(stdout,L"%S\n", f)
636 No symbol "L" in current context.
640 Cleanup configury support for optional sub-directories.
642 Check how GCC handles multiple front ends for an example of how things
643 could work. A tentative first step is to rationalize things so that
644 all sub directories are handled in a fashion similar to gdb/mi.
646 See also automake above.
650 Add a transcript mechanism to GDB.
652 Such a mechanism might log all gdb input and output to a file in a
653 form that would allow it to be replayed. It could involve ``gdb
654 --transcript=FILE'' or it could involve ``(gdb) transcript file''.
658 Can the xdep files be replaced by autoconf?
662 Document trace machinery
666 Document ui-out and ui-file.
668 http://sourceware.cygnus.com/ml/gdb/2000-04/msg00121.html
672 Update texinfo.tex to latest?
676 Incorporate agentexpr.texi into gdb.texinfo
678 agentexpr.texi mostly describes the details of the byte code used for
679 tracepoints, not the internals of the support for this in GDB. So it
680 looks like gdb.texinfo is a better place for this information.
682 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00566.html
686 Document overlay machinery.
690 ``(gdb) catch signal SIGNAL''
692 Overlaps with ``handle SIGNAL'' but the implied behavior is different.
693 You can attach commands to a catch but not a handle. A handle has a
694 limited number of hardwired actions.
700 o readline/*.h bitrot
702 The TUI isn't up-to-date with
703 respect to the readline currently
704 bundled with GDB. Importing a
705 new readline is on the 5.1 wish
706 list so this can only get worse.
708 Grep for things like term_cursor_move.
710 (To be honest, I don't see anyone
711 importing a new readline before 5.1 is
714 o tui.c:va_catch_errors() bitrot
716 This nasty piece of work used knowledge
717 of the internals of GDBs error functions :-(
718 Ever since those internals were cleaned
719 up this code has been broken. :-(
721 o tuiWin.c:c_makeVisibleWithNewHeight() broken
722 tuiLayout.c:_extractDisplayStartAddr() broken
724 Both these function call find_line_pc()
725 incorrectly (wrong args, wrong return value).
727 I suspect this bug has always been there!
728 It had been hidden because those files
729 didn't include the necessary header files
732 o tuiRegs() host dependant
734 Not suprisingly, this isn't a very portable
735 section of code. However, I'm sure people
736 could live with no regs in the short to
739 o defs.h: #include "tui.h" et.al.
741 I'm not sure where this came from.
742 It was a really bad idea.
744 To get things to compile I did a nasty
745 hack (Just declare what was needed and
746 replace any expressions like xx->y.z()
747 in GDB proper with function calls). I
748 could commit it slightly cleaned up if
751 Medium Term. the #ifdef TUI and TuiDo()
752 should be changed to hooks (like GDBTK).
753 The gdb-events.[hc] is there for that
756 o tui.c:_tuiReset() host dependant
758 tui.c contains a lump of termio[s]
759 I suspect an equivalent block of
760 code can be lifted from readline.
761 An equivalent readline function may
764 o curses.h vs ncurses.h.
766 Simple portability problem.
770 This function is a mystery - where is it?
772 o tui-file.[hc] cleanup
774 This can be significantly simplified.
776 o The code should be pacified. (-Werror -W...)
778 There are plenty of #includes,
779 duplicate #includes, missing function decls
782 Some of the problems I found were through
783 fixing a few of the warnings.
785 o The code should be GNUtified.
787 It would be very nice to have this code
788 look like the rest of GDB. That way people
789 would be more accepting of it as a true
792 Until it is GNUtified it is going to stick
793 out like a sore thumb to the programmer.
795 o The code should be clearly copyrighted
797 (FSF, with due credit to HP)
801 Add support for ``gdb --- PROGRAM ARGS ...''.
802 Add support for ``gdb -cmd=...''
804 Along with many variations. Check:
806 ????? for a full discussion.
812 Implement ``(gdb) !ls''.
814 Which is very different from ``(gdb) ! ls''. Implementing the latter
817 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00034.html
821 Change the (char *list[]) to (const char (*)[]) so that dynamic lists can
826 When tab expanding something like ``set arch<tab>'' ignore the
827 deprecated ``set archdebug'' and expand to ``set architecture''.
831 Replace the code that uses the host FPU with an emulator of the target
836 The "ocd reset" command needs to flush the dcache, which requires breaking
837 the abstraction layer between the target independent and target code. One
838 way to address this is provide a generic "reset" command and target vector.
840 http://sources.redhat.com/ml/gdb-patches/2000-10/msg00011.html
847 New languages come onto the scene all the time.
851 Re: Various C++ things
853 value_headof/value_from_vtable_info are worthless, and should be
854 removed. The one place in printcmd.c that uses it should use the RTTI
857 RTTI for g++ should be using the typeinfo functions rather than the
858 vtables. The typeinfo functions are always at offset 4 from the
859 beginning of the vtable, and are always right. The vtables will have
860 weird names like E::VB sometimes. The typeinfo function will always
861 be "E type_info function", or somesuch.
863 value_virtual_fn_field needs to be fixed so there are no failures for
864 virtual functions for C++ using g++.
866 Testsuite cases are the major priority right now for C++ support,
867 since i have to make a lot of changes that could potentially break
872 Add support for Modula3
874 Get DEC/Compaq to contribute their Modula-3 support.
878 Remote Protocol Support
879 =======================
883 Revised UDP support (was: Re: [Fwd: [patch] UDP transport support])
884 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00000.html
886 (Broken) support for GDB's remote protocol across UDP is to be
887 included in the follow-on release.
889 It should be noted that UDP can only work when the [Gg] packet fits in
892 There is also much debate over the merit of this.
896 Migrate qfThreadInfo packet -> qThreadInfo. (Andrew Cagney)
898 Add support for packet enable/disable commands with these thread
899 packets. General cleanup.
901 [PATCH] Document the ThreadInfo remote protocol queries
902 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00832.html
904 [PATCH] "info threads" queries for remote.c
905 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00831.html
909 Remote protocol doco feedback.
911 Too much feedback to mention needs to be merged in (901660). Search
912 for the word ``remote''.
915 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00023.html
916 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00056.html
917 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00382.html
921 GDB doesn't recover gracefully from remote protocol errors.
923 GDB wasn't checking for NAKs from the remote target. Instead a NAK is
924 ignored and a timeout is required before GDB retries. A pre-cursor to
925 fixing this this is making GDB's remote protocol packet more robust.
927 While downloading to a remote protocol target, gdb ignores packet
928 errors in so far as it will continue to download with chunk N+1 even
929 if chunk N was not correctly sent. This causes gdb.base/remote.exp to
930 take a painfully long time to run. As a PS that test needs to be
931 fixed so that it builds on 16 bit machines.
935 Fix the ``!'' packet.
937 JT reported that the existing targets do, in fact return ``OK'' so it
938 is possible to merge remote and extended-remote targets.
942 Drop ``<address>'' from the [SsCc] packets.
944 I don't think that GDB generates them so having it in the protocol is
949 Fix doco on the ``q'' packet.
951 It has evolved into a generic RPC. The notes should reflect this and,
952 perhaps, the ``Q'' packet can be deprecated.
954 The doco should mention that ``OK'' is a valid packet response.
956 The doco should explain why ``OK'' needs to be a valid packet
961 Add the cycle step command.
963 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00237.html
967 Resolve how to scale things to support very large packets.
971 Resolve how to handle a target that changes things like its endianess
972 on the fly - should it be returned in the ``T'' packet?
974 Underlying problem is that the register file is target endian. If the
975 target endianess changes gdb doesn't know.
977 Suggest encoding registers as NN!VALUE.
981 GDB should allow incomming packets to be larger than outgoing ones. A
982 fully loaded T packet (containing all registers) can be very large -
983 definitly larger than a corresponding Gg packet.
990 If / when GDB starts to support the debugging of multi-processor
991 (rather than multi-thread) applications the symtab code will need to
992 be updated a little so that several independent symbol tables are
993 active at a given time.
995 The other interesting change is a clarification of the exact meaning
996 of CORE_ADDR and that has had consequences for a few targets (that
997 were abusing that data type).
1001 Investiagate ways of reducing memory.
1005 Investigate ways of improving load time.
1009 Get the d10v to use POINTER_TO_ADDRESS and ADDRESS_TO_POINTER.
1011 Consequence of recent symtab clarification. No marks for figuring out
1012 who maintains the d10v.
1016 Get the MIPS to correctly sign extend all address <-> pointer
1019 Consequence of recent symtab clarification. No marks for figuring out
1020 who maintains the MIPS.
1024 GDB truncates 64 bit enums.
1026 http://sourceware.cygnus.com/ml/gdb-patches/2000-06/msg00290.html
1033 There are never to many testcases.
1037 Better thread testsuite.
1041 Better C++ testsuite.
1045 Look at adding a GDB specific testsuite directory so that white box
1046 tests of key internals can be added (eg ui_file).
1050 Separate out tests that involve the floating point (FP).
1052 (Something for people brining up new targets). FP and non-fp tests
1053 are combined. I think there should be set of basic tests that
1054 exercise pure integer support and then a more expanded set that
1055 exercise FP and FP/integer interactions.
1057 As an example, the MIPS, for n32 as problems with passing FP's and
1058 structs. Since most inferior call tests include FP it is difficult to
1059 determine of the integer tests are ok.
1063 Architectural Changes: General
1064 ==============================
1066 These are harder than simple cleanups / fixes and, consequently
1067 involve more work. Typically an Architectural Change will be broken
1068 down into a more digestible set of cleanups and fixes.
1072 Cleanup software single step.
1074 At present many targets implement software single step by directly
1075 blatting memory (see rs6000-tdep.c). Those targets should register
1076 the applicable breakpoints using the breakpoint framework. Perhaphs a
1077 new internal breakpoint class ``step'' is needed.
1081 Replace READ_FP() with FRAME_HANDLE().
1083 READ_FP() is a hangover from the days of the vax when the ABI really
1084 did have a frame pointer register. Modern architectures typically
1085 construct a virtual frame-handle from the stack pointer and various
1086 other bits of string.
1088 Unfortunately GDB still treats this synthetic FP register as though it
1089 is real. That in turn really confuses users (arm and ``print $fp'' VS
1090 ``info registers fp''). The synthetic FP should be separated out of
1091 the true register set presented to the user.
1095 Register Cache Cleanup (below from Andrew Cagney)
1097 I would depict the current register architecture as something like:
1099 High GDB --> Low GDB
1104 register + REGISTER_BYTE(reg_nr)
1107 -------------------------
1108 | extern register[] |
1109 -------------------------
1111 where neither the high (valops.c et.al.) or low gdb (*-tdep.c) are
1112 really clear on what mechanisms they should be using to manipulate that
1113 buffer. Further, much code assumes, dangerously, that registers are
1114 contigious. Having got mips-tdep.c to support multiple ABIs, believe
1115 me, that is a bad assumption. Finally, that register cache layout is
1116 determined by the current remote/local target and _not_ the less
1117 specific target ISA. In fact, in many cases it is determined by the
1118 somewhat arbitrary layout of the [gG] packets!
1121 How I would like the register file to work is more like:
1149 The main objectives being:
1151 o a clear separation between the low
1152 level target and the high level GDB
1154 o a mechanism that solves the general
1155 problem of register aliases, overlaps
1156 etc instead of treating them as optional
1157 extras that can be wedged in as an after
1158 thought (that is a reasonable description
1159 of the current code).
1161 Identify then solve the hard case and the
1162 rest just falls out. GDB solved the easy
1163 case and then tried to ignore the real
1166 o a removal of the assumption that the
1167 mapping between the register cache
1168 and virtual registers is largely static.
1169 If you flip the USR/SSR stack register
1170 select bit in the status-register then
1171 the corresponding stack registers should
1174 o a mechanism that clearly separates the
1175 gdb internal register cache from any
1176 target (not architecture) dependent
1177 specifics such as [gG] packets.
1179 Of course, like anything, it sounds good in theory. In reality, it
1180 would have to contend with many<->many relationships at both the
1181 virt<->cache and cache<->target level. For instance:
1184 Modifying an mmx register may involve
1185 scattering values across both FP and
1186 mmpx specific parts of a buffer
1189 When writing back a SP it may need to
1190 both be written to both SP and USP.
1195 Rather than let this like the last time it was discussed, just slip, I'm
1196 first going to add this e-mail (+ references) to TODO. I'd then like to
1197 sketch out a broad strategy I think could get us there.
1200 First thing I'd suggest is separating out the ``extern registers[]''
1201 code so that we can at least identify what is using it. At present
1202 things are scattered across many files. That way we can at least
1203 pretend that there is a cache instead of a global array :-)
1205 I'd then suggest someone putting up a proposal for the pseudo-reg /
1206 high-level side interface so that code can be adopted to it. For old
1207 code, initially a blanket rename of write_register_bytes() to
1208 deprecated_write_register_bytes() would help.
1210 Following that would, finaly be the corresponding changes to the target.
1214 Check that GDB can handle all BFD architectures (Andrew Cagney)
1216 There should be a test that checks that BFD/GDB are in sync with
1217 regard to architecture changes. Something like a test that first
1218 queries GDB for all supported architectures and then feeds each back
1219 to GDB.. Anyone interested in learning how to write tests? :-)
1223 Architectural Change: Multi-arch et al.
1224 =======================================
1226 The long term objective is to remove all assumptions that there is a
1227 single target with a single address space with a single instruction
1228 set architecture and single application binary interface.
1230 This is an ongoing effort. The first milestone is to enable
1231 ``multi-arch'' where by all architectural decisions are made at
1234 It should be noted that ``gdbarch'' is really ``gdbabi'' and
1235 ``gdbisa''. Once things are multi-arched breaking that down correctly
1236 will become much easier.
1240 GDBARCH cleanup (Andrew Cagney)
1242 The non-generated parts of gdbarch.{sh,h,c} should be separated out
1243 into arch-utils.[hc].
1245 Document that gdbarch_init_ftype could easily fail because it didn't
1246 identify an architecture.
1250 Fix BELIEVE_PPC_PROMOTION. Change it to BELIEVE_PPC_PROMOTION_P?
1252 At present there is still #ifdef BELIEVE_PPC_PROMOTION code in the
1257 Fix target_signal_from_host() etc.
1259 The name is wrong for starters. ``target_signal'' should probably be
1260 ``gdb_signal''. ``from_host'' should be ``from_target_signal''.
1261 After that it needs to be multi-arched and made independent of any
1262 host signal numbering.
1264 Once this is done, the signal enum can probably be moved to
1265 include/gdb so that it is available to embedded stubs.
1269 Update ALPHA so that it uses ``struct frame_extra_info'' instead of
1272 This is a barrier to replacing mips_extra_func_info with something
1273 that works with multi-arch.
1277 Multi-arch mips_extra_func_info.
1279 This first needs the alpha to be updated so that it uses ``struct
1284 Rationalize TARGET_SINGLE_FORMAT and TARGET_SINGLE_BIT et al.
1286 Surely one of them is redundant.
1290 Convert ALL architectures to MULTI-ARCH.
1294 Select the initial multi-arch ISA / ABI based on --target or similar.
1296 At present the default is based on what ever is first in the BFD
1297 archures table. It should be determined based on the ``--target=...''
1302 Make MIPS pure multi-arch.
1304 It is only at the multi-arch enabled stage.
1310 Enable the code to recognize --enable-targets=.... like BINUTILS does.
1312 Can the tm.h and nm.h files be eliminated by multi-arch.
1316 Architectural Change: MI, LIBGDB and scripting languages
1317 ========================================================
1319 See also architectural changes related to the event loop. LIBGDB
1320 can't be finished until there is a generic event loop being used by
1323 The long term objective is it to be possible to integrate GDB into
1324 scripting languages.
1328 Implement generic ``(gdb) commmand > file''
1330 Once everything is going through ui_file it should be come fairly
1333 http://sourceware.cygnus.com/ml/gdb/2000-04/msg00104.html
1337 Replace gdb_stdtarg with gdb_targout (and possibly gdb_targerr).
1339 gdb_stdtarg is easily confused with gdb_stdarg.
1343 Extra ui_file methods - dump.
1345 Very useful for whitebox testing.
1349 Eliminate error_begin().
1351 With ui_file, there is no need for the statefull error_begin ()
1356 Send normal output to gdb_stdout.
1357 Send error messages to gdb_stderror.
1358 Send debug and log output log gdb_stdlog.
1360 GDB still contains many cases where (f)printf or printf_filtered () is
1361 used when it should be sending the messages to gdb_stderror or
1362 gdb_stdlog. The thought of #defining printf to something has crossed
1367 Re-do GDB's output pager.
1369 GDB's output pager still relies on people correctly using *_filtered
1370 for gdb_stdout and *_unfiltered for gdb_stdlog / gdb_stderr.
1371 Hopefully, with all normal output going to gdb_stdout, the pager can
1372 just look at the ui_file that the output is on and then use that to
1373 decide what to do about paging. Sounds good in theory.
1377 Check/cleanup MI documentation.
1379 The list of commands specified in the documentation needs to be
1380 checked against the mi-cmds.c table in a mechanical way (so that they
1381 two can be kept up-to-date).
1385 Convert MI into libgdb
1387 MI provides a text interface into what should be many of the libgdb
1388 functions. The implementation of those functions should be separated
1389 into the MI interface and the functions proper. Those functions being
1390 moved to gdb/lib say.
1396 The first part can already be found in defs.h.
1400 MI's input does not use buffering.
1402 At present the MI interface reads raw characters of from an unbuffered
1403 FD. This is to avoid several nasty buffer/race conditions. That code
1404 should be changed so that it registers its self with the event loop
1405 (on the input FD) and then push commands up to MI as they arrive.
1407 The serial code already does this.
1411 Make MI interface accessible from existing CLI.
1415 Add a breakpoint-edit command to MI.
1417 It would be similar to MI's breakpoint create but would apply to an
1418 existing breakpoint. It saves the need to delete/create breakpoints
1419 when ever they are changed.
1423 Add directory path to MI breakpoint.
1425 That way the GUI's task of finding the file within which the
1426 breakpoint was set is simplified.
1430 Add a mechanism to reject certain expression classes to MI
1432 There are situtations where you don't want GDB's expression
1433 parser/evaluator to perform inferior function calls or variable
1434 assignments. A way of restricting the expression parser so that such
1435 operations are not accepted would be very helpful.
1439 Remove sideffects from libgdb breakpoint create function.
1441 The user can use the CLI to create a breakpoint with partial
1442 information - no file (gdb would use the file from the last
1445 The libgdb interface currently affects that environment which can lead
1446 to confusion when a user is setting breakpoints via both the MI and
1449 This is also a good example of how getting the CLI ``right'' will be
1454 Move gdb_lasterr to ui_out?
1456 The way GDB throws errors and records them needs a re-think. ui_out
1457 handles the correct output well. It doesn't resolve what to do with
1458 output / error-messages when things go wrong.
1462 do_setshow_command contains a 1024 byte buffer.
1464 The function assumes that there will never be any more than 1024 bytes
1465 of enum. It should use mem_file.
1469 Should struct cmd_list_element . completer take the command as an
1474 Should the bulk of top.c:line_completion_function() be moved to
1475 command.[hc]? complete_on_cmdlist() and complete_on_enums() could
1476 then be made private.
1480 top.c (execute_command): Should a command being valid when the target
1481 is running be made an attribute (predicate) to the command rather than
1482 an explicit set of tests.
1486 top.c (execute_command): Should the bulk of this function be moved
1487 into command.[hc] so that top.c doesn't grub around in the command
1492 Architectural Change: Async
1493 ===========================
1495 While GDB uses an event loop when prompting the user for input. That
1496 event loop is not exploited by targets when they allow the target
1497 program to continue. Typically targets still block in (target_wait())
1498 until the program again halts.
1500 The closest a target comes to supporting full asynchronous mode are
1501 the remote targets ``async'' and ``extended-async''.
1505 Asynchronous expression evaluator
1507 Inferior function calls hang GDB.
1511 Fix implementation of ``target xxx''.
1513 At present when the user specifies ``target xxxx'', the CLI maps that
1514 directly onto a target open method. It is then assumed that the
1515 target open method should do all sorts of complicated things as this
1516 is the only chance it has. Check how the various remote targets
1517 duplicate the target operations. Check also how the various targets
1518 behave differently for purely arbitrary reasons.
1520 What should happen is that ``target xxxx'' should call a generic
1521 ``target'' function and that should then co-ordinate the opening of
1522 ``xxxx''. This becomes especially important when you're trying to
1523 open an asynchronous target that may need to perform background tasks
1524 as part of the ``attach'' phase.
1526 Unfortunately, due to limitations in the old/creaking command.h
1527 interface, that isn't possible. The function being called isn't told
1528 of the ``xxx'' or any other context information.
1530 Consequently a precursor to fixing ``target xxxx'' is to clean up the
1531 CLI code so that it passes to the callback function (attatched to a
1532 command) useful information such as the actual command and a context
1533 for that command. Other changes such as making ``struct command''
1534 opaque may also help.
1537 http://sourceware.cygnus.com/ml/gdb-patches/2000-06/msg00062.html
1541 Make "target xxx" command interruptible.
1543 As things become async this becomes possible. A target would start
1544 the connect and then return control to the event loop. A cntrl-c
1545 would notify the target that the operation is to be abandoned and the
1546 target code could respond.
1550 Add a "suspend" subcommand of the "continue" command to suspend gdb
1551 while continuing execution of the subprocess. Useful when you are
1552 debugging servers and you want to dodge out and initiate a connection
1553 to a server running under gdb.
1562 Frequently requested but not approved requests.
1566 Eliminate unused argument warnings using ATTRIBUTE_UNUSED.
1568 The benefits on this one are thought to be marginal - GDBs design
1569 means that unused parameters are very common. GCC 3.0 will also
1570 include the option -Wno-unused-parameter which means that ``-Wall
1571 -Wno-unused-parameters -Werror'' can be specified.
1580 This list is not up to date, and opinions vary about the importance or
1581 even desirability of some of the items. If you do fix something, it
1582 always pays to check the below.
1586 @c This does not work (yet if ever). FIXME.
1587 @c @item --parse=@var{lang} @dots{}
1588 @c Configure the @value{GDBN} expression parser to parse the listed languages.
1589 @c @samp{all} configures @value{GDBN} for all supported languages. To get a
1590 @c list of all supported languages, omit the argument. Without this
1591 @c option, @value{GDBN} is configured to parse all supported languages.
1595 START_INFERIOR_TRAPS_EXPECTED need never be defined to 2, since that
1596 is its default value. Clean this up.
1600 It should be possible to use symbols from shared libraries before we know
1601 exactly where the libraries will be loaded. E.g. "b perror" before running
1602 the program. This could maybe be done as an extension of the "breakpoint
1603 re-evaluation" after new symbols are loaded.
1607 Make single_step() insert and remove breakpoints in one operation.
1609 [If this is talking about having single_step() insert the breakpoints,
1610 run the target then pull the breakpoints then it is wrong. The
1611 function has to return as control has to eventually be passed back to
1612 the main event loop.]
1616 Speed up single stepping by avoiding extraneous ptrace calls.
1620 Speed up single stepping by not inserting and removing breakpoints
1621 each time the inferior starts and stops.
1623 Breakpoints should not be inserted and deleted all the time. Only the
1624 one(s) there should be removed when we have to step over one. Support
1625 breakpoints that don't have to be removed to step over them.
1627 [this has resulted in numerous debates. The issue isn't clear cut]
1631 Provide "voodoo" debugging of core files. This creates a zombie
1632 process as a child of the debugger, and loads it up with the data,
1633 stack, and regs of the core file. This allows you to call functions
1634 in the executable, to manipulate the data in the core file.
1640 GDB reopens the source file on every line, as you "next" through it.
1642 [still true? I've a memory of this being fixed]
1646 Perhaps "i source" should take an argument like that of "list".
1650 Remove "at 0xnnnn" from the "b foo" response, if `print address off' and if
1651 it matches the source line indicated.
1655 The prompt at end of screen should accept space as well as CR.
1659 Backtrace should point out what the currently selected frame is, in
1660 its display, perhaps showing "@3 foo (bar, ...)" or ">3 foo (bar,
1661 ...)" rather than "#3 foo (bar, ...)".
1665 "i program" should work for core files, and display more info, like what
1666 actually caused it to die.
1670 "x/10i" should shorten the long name, if any, on subsequent lines.
1674 "next" over a function that longjumps, never stops until next time you happen
1675 to get to that spot by accident. E.g. "n" over execute_command which has
1680 "set zeroprint off", don't bother printing members of structs which
1681 are entirely zero. Useful for those big structs with few useful
1686 GDB does four ioctl's for every command, probably switching terminal modes
1687 to/from inferior or for readline or something.
1691 terminal_ours versus terminal_inferior: cache state. Switch should be a noop
1692 if the state is the same, too.
1696 "i frame" shows wrong "arglist at" location, doesn't show where the args
1697 should be found, only their actual values.
1701 There should be a way for "set" commands to validate the new setting
1702 before it takes effect.
1706 "ena d" is ambiguous, why? "ena delete" seems to think it is a command!
1710 i line VAR produces "Line number not known for symbol ``var''.". I
1711 thought we were stashing that info now!
1715 We should be able to write to random files at hex offsets like adb.
1719 [elena - delete this]
1721 Handle add_file with separate text, data, and bss addresses. Maybe
1722 handle separate addresses for each segment in the object file?
1726 [Jimb/Elena delete this one]
1728 Handle free_named_symtab to cope with multiply-loaded object files
1729 in a dynamic linking environment. Should remember the last copy loaded,
1730 but not get too snowed if it finds references to the older copy.
1734 [elena delete this also]
1736 Remove all references to:
1743 now that we have BFD. All remaining are in machine dependent files.
1747 Re-organize help categories into things that tend to fit on a screen
1752 Add in commands like ADB's for searching for patterns, etc. We should
1753 be able to examine and patch raw unsymboled binaries as well in gdb as
1754 we can in adb. (E.g. increase the timeout in /bin/login without source).
1756 [actually, add ADB interface :-]
1760 When doing "step" or "next", if a few lines of source are skipped between
1761 the previous line and the current one, print those lines, not just the
1762 last line of a multiline statement.
1766 Handling of "&" address-of operator needs some serious overhaul
1767 for ANSI C and consistency on arrays and functions.
1768 For "float point[15];":
1769 ptype &point[4] ==> Attempt to take address of non-lvalue.
1770 For "char *malloc();":
1771 ptype malloc ==> "char *()"; should be same as
1772 ptype &malloc ==> "char *(*)()"
1773 call printf ("%x\n", malloc) ==> weird value, should be same as
1774 call printf ("%x\n", &malloc) ==> correct value
1778 Fix dbxread.c symbol reading in the presence of interrupts. It
1779 currently leaves a cleanup to blow away the entire symbol table when a
1780 QUIT occurs. (What's wrong with that? -kingdon, 28 Oct 1993).
1782 [I suspect that the grype was that, on a slow system, you might want
1783 to cntrl-c and get just half the symbols and then load the rest later
1784 - scary to be honest]
1788 Mipsread.c reads include files depth-first, because the dependencies
1789 in the psymtabs are way too inclusive (it seems to me). Figure out what
1790 really depends on what, to avoid recursing 20 or 30 times while reading
1795 value_add() should be subtracting the lower bound of arrays, if known,
1796 and possibly checking against the upper bound for error reporting.
1800 When listing source lines, check for a preceding \n, to verify that
1801 the file hasn't changed out from under us.
1803 [fixed by some other means I think. That hack wouldn't actually work
1804 reliably - the file might move such that another \n appears. ]
1808 Get all the remote systems (where the protocol allows it) to be able to
1809 stop the remote system when the GDB user types ^C (like remote.c
1810 does). For ebmon, use ^Ak.
1814 Possible feature: A version of the "disassemble" command which shows
1815 both source and assembly code ("set symbol-filename on" is a partial
1818 [has this been done? It was certainly done for MI and GDBtk]
1822 investigate "x/s 0" (right now stops early) (I think maybe GDB is
1823 using a 0 address for bad purposes internally).
1827 Make "info path" and path_command work again (but independent of the
1828 environment either of gdb or that we'll pass to the inferior).
1832 Make GDB understand the GCC feature for putting octal constants in
1833 enums. Make it so overflow on an enum constant does not error_type
1834 the whole type. Allow arbitrarily large enums with type attributes.
1835 Put all this stuff in the testsuite.
1839 Make TYPE_CODE_ERROR with a non-zero TYPE_LENGTH more useful (print
1840 the value in hex; process type attributes). Add this to the
1841 testsuite. This way future compilers can add new types and old
1842 versions of GDB can do something halfway reasonable.
1846 Fix mdebugread.c:parse_type to do fundamental types right (see
1847 rs6000_builtin_type in stabsread.c for what "right" is--the point is
1848 that the debug format fixes the sizes of these things and it shouldn't
1849 depend on stuff like TARGET_PTR_BIT and so on. For mdebug, there seem
1850 to be separate bt* codes for 64 bit and 32 bit things, and GDB should
1851 be aware of that). Also use a switch statement for clarity and speed.
1855 Investigate adding symbols in target_load--some targets do, some
1860 Put dirname in psymtabs and change lookup*symtab to use dirname (so
1861 /foo/bar.c works whether compiled by cc /foo/bar.c, or cd /foo; cc
1866 Merge xcoffread.c and coffread.c. Use breakpoint_re_set instead of
1871 Make a watchpoint which contains a function call an error (it is
1872 broken now, making it work is probably not worth the effort).
1876 New test case based on weird.exp but in which type numbers are not
1877 renumbered (thus multiply defining a type). This currently causes an
1878 infinite loop on "p v_comb".
1882 [Hey! Hint Hint Delete Delete!!!]
1884 Fix 386 floating point so that floating point registers are real
1885 registers (but code can deal at run-time if they are missing, like
1886 mips and 68k). This would clean up "info float" and related stuff.
1890 gcc -g -c enummask.c then gdb enummask.o, then "p v". GDB complains
1891 about not being able to access memory location 0.
1893 -------------------- enummask.c
1914 If try to modify value in file with "set write off" should give
1915 appropriate error not "cannot access memory at address 0x65e0".
1919 Allow core file without exec file on RS/6000.
1923 Make sure "shell" with no arguments works right on DOS.
1927 Make gdb.ini (as well as .gdbinit) be checked on all platforms, so
1928 the same directory can be NFS-mounted on unix or DOS, and work the
1933 [Is this another delete???]
1935 Get SECT_OFF_TEXT stuff out of objfile_relocate (might be needed to
1936 get RS/6000 to work right, might not be immediately relevant).
1940 Work out some kind of way to allow running the inferior to be done as
1941 a sub-execution of, eg. breakpoint command lists. Currently running
1942 the inferior interupts any command list execution. This would require
1943 some rewriting of wait_for_inferior & friends, and hence should
1944 probably be done in concert with the above.
1948 Add function arguments to gdb user defined functions.
1952 Add convenience variables that refer to exec file, symbol file,
1953 selected frame source file, selected frame function, selected frame
1958 Modify the handling of symbols grouped through BINCL/EINCL stabs to
1959 allocate a partial symtab for each BINCL/EINCL grouping. This will
1960 seriously decrease the size of inter-psymtab dependencies and hence
1961 lessen the amount that needs to be read in when a new source file is
1966 Add a command for searching memory, a la adb. It specifies size,
1967 mask, value, start address. ADB searches until it finds it or hits
1968 an error (or is interrupted).
1972 Remove the range and type checking code and documentation, if not