gdb/TODO

   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.
   5
   6
   7                 Known problems in GDB 5.0
   8                 =========================
   9
  10 Below is a list of problems identified during the GDB 5.0 release
  11 cycle.  People hope to have these problems fixed in a follow-on
  12 release.
  13
  14 --
  15
  16 The BFD directory requires bug-fixed AUTOMAKE et.al.
  17
  18 AUTOMAKE 1.4 incorrectly set the TEXINPUTS environment variable.  It
  19 contained the full path to texinfo.tex when it should have only
  20 contained the directory.  The bug has been fixed in the current
  21 AUTOMAKE sources.  Automake snapshots can be found in:
  22         ftp://sourceware.cygnus.com/pub/gdb/snapshots
  23 and     ftp://sourceware.cygnus.com/pub/binutils
  24
  25 --
  26
  27 RFD: infrun.c: No bpstat_stop_status call after proceed over break?
  28 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00665.html
  29
  30 GDB misses watchpoint triggers after proceeding over a breakpoint on
  31 x86 targets.
  32
  33 --
  34
  35 x86 linux GDB and SIGALRM (???)
  36 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00803.html
  37
  38 I know there are problems with single stepping through signal
  39 handlers.  These problems were present in 4.18.  They were just masked
  40 because 4.18 failed to recognize signal handlers.  Fixing it is not
  41 easy, and will require changes to handle_inferior_event(), that I
  42 prefer not to make before the 5.0 release.
  43
  44 Mark
  45
  46 --
  47
  48 Revised UDP support (was: Re: [Fwd: [patch] UDP transport support])
  49 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00000.html
  50
  51 (Broken) support for GDB's remote protocol across UDP is to be
  52 included in the follow-on release.
  53
  54 --
  55
  56 Can't build IRIX -> arm GDB.
  57 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00356.html
  58
  59 David Whedon writes:
  60 > Now I'm building for an embedded arm target.  If there is a way of turning
  61 > remote-rdi off, I couldn't find it. It looks like it gets built by default
  62 > in gdb/configure.tgt(line 58) Anyway, the build dies in
  63 > gdb/rdi-share/unixcomm.c.  SERPORT1 et. al. never get defined because we
  64 > aren't one of the architectures supported.
  65
  66 --
  67
  68 Problem with weak functions
  69 http://sourceware.cygnus.com/ml/gdb/2000-05/msg00060.html
  70
  71 Dan Nicolaescu writes:
  72 > It seems that gdb-4.95.1  does not display correctly the function when
  73 > stoping in weak functions.
  74 >
  75 > It stops in a function that is defined as weak, not in the function
  76 > that is actualy run...
  77
  78 --
  79
  80 GDB 5.0 doesn't work on Linux/SPARC
  81
  82 --
  83
  84                 Code Cleanups: Next Release
  85                 ===========================
  86
  87 The following are small cleanups that will hopefully be completed by
  88 the follow on to 5.0.
  89
  90 --
  91
  92 Delete macro TARGET_BYTE_ORDER_SELECTABLE.
  93
  94 Patches in the database.
  95
  96 --
  97
  98 Purge PARAMS.
  99
 100 Eliminate all uses of PARAMS in GDB's source code.
 101
 102 --
 103
 104 Elimination of make_cleanup_func. (Andrew Cagney)
 105
 106 make_cleanup_func elimination
 107 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00791.html
 108 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00814.html
 109
 110 --
 111
 112 Fix copyright notices.
 113
 114 Turns out that ``1998-2000'' isn't considered valid :-(
 115
 116 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00467.html
 117
 118 --
 119
 120                 Code Cleanups: General
 121                 ======================
 122
 123 The following are more general cleanups and fixes.  They are not tied
 124 to any specific release.
 125
 126 --
 127
 128 Eliminate more compiler warnings.
 129
 130 Of course there also needs to be the usual debate over which warnings
 131 are valid and how to best go about this.
 132
 133 One method: choose a single option; get agreement that it is
 134 reasonable; try it out to see if there isn't anything silly about it
 135 (-Wunused-parameters is an example of that) then incrementally hack
 136 away.
 137
 138 The other method is to enable all warnings and eliminate them from one
 139 file at a time.
 140
 141 --
 142
 143 Elimination of ``(catch_errors_ftype *) func''.
 144
 145 Like make_cleanup_func it isn't portable.
 146
 147 --
 148
 149 Nuke USG define.
 150
 151 --
 152
 153 [PATCH/5] src/intl/Makefile.in:distclean additions
 154 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00363.html
 155
 156 Do not forget to merge the patch back into the trunk.
 157
 158 --
 159
 160 Rationalize the host-endian code (grep for HOST_BYTE_ORDER).
 161
 162 At present defs.h includes <endian.h> (which is linux specific) yet
 163 almost nothing depends on it.  Suggest "gdb_endian.h" which can also
 164 handle <machine/endian.h> and only include that where it is really
 165 needed.
 166
 167 --
 168
 169 Replace asprintf() calls with xasprintf() calls.
 170
 171 As with things like strdup() most calls to asprintf() don't check the
 172 return value.
 173
 174 --
 175
 176 Replace strsave() + mstrsave() with libiberty:xstrdup().
 177
 178 --
 179
 180 Replace savestring() with something from libiberty.
 181
 182 An xstrldup()? but that would have different semantics.
 183
 184 --
 185
 186 Rationalize use of floatformat_unknown in GDB sources.
 187
 188 Instead of defaulting to floatformat_unknown, should hosts/targets
 189 specify the value explicitly?
 190
 191 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00447.html
 192
 193 --
 194
 195 Add a ``name'' member to include/floatformat.h:struct floatformat.
 196 Print that name in gdbarch.c.
 197
 198 --
 199
 200 Sort out the harris mess in include/floatformat.h (it hardwires two
 201 different floating point formats).
 202
 203 --
 204
 205 See of the GDB local floatformat_do_doublest() and libiberty's
 206 floatformat_to_double (which was once GDB's ...) can be merged some
 207 how.
 208
 209 --
 210
 211 Eliminate mmalloc() from GDB.
 212
 213 Also eliminate it from defs.h.
 214
 215 --
 216
 217 Eliminate PTR.  ISO-C allows ``void *''.
 218
 219 --
 220
 221 Eliminate abort ().
 222
 223 GDB should never abort.  GDB should either throw ``error ()'' or
 224 ``internal_error ()''.  Better still GDB should naturally unwind with
 225 an error status.
 226
 227 --
 228
 229 GDB probably doesn't build on FreeBSD pre 2.2.x
 230 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00378.html
 231
 232 Fixes to get FreeBSD working on 2.2.x, 3.x and 4.x caused the code to
 233 suffer bit rot.
 234
 235 --
 236
 237 Updated readline
 238
 239 Readline 4.? is out.  A merge wouldn't hurt.  Patches are in:
 240
 241 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00436.html
 242
 243 --
 244
 245 Deprecate "fg".  Apparently ``fg'' is actually continue.
 246
 247 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00417.html
 248
 249 --
 250
 251 Deprecate current use of ``floatformat_unknown''.
 252
 253 Require all targets to explicitly provide their float format instead
 254 of defaulting to floatformat unknown.  Doing the latter leads to nasty
 255 bugs.
 256
 257 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00447.html
 258
 259 --
 260
 261 Rationalize floatformat_to_double() vs floatformat_to_doublest().
 262
 263 Looks like GDB migrated floatformat_to_double() to libiberty but then
 264 turned around and created a ..._to_doublest() the latter containing
 265 several bug fixes.
 266
 267 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00472.html
 268
 269 --
 270
 271 Move floatformat_ia64_ext to libiberty/include floatformat.[ch].
 272
 273 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00466.html
 274
 275 --
 276
 277 Always build ser-tcp.c.
 278
 279 The patch as submitted was just going to add ser-tcp.c to the Alpha's
 280 makefile.  A better patch is to instead add ser-tcp.c to SER_HARDWARE
 281 and make it a standard part of all debuggers.
 282
 283 If problems occure then configure.in can sort them out.
 284
 285 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00544.html
 286
 287 --
 288
 289 Follow through `make check' with --enable-shared.
 290
 291 When the srcware tree is configured with --enable-shared, the `expect'
 292 program won't run properly.  Jim Wilson found out gdb has a local hack
 293 to set LD_LIBRARY_PATH, but, AFAIK, no other project has been hacked
 294 similarly.
 295
 296 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00845.html
 297
 298 --
 299
 300 Change the parameter ``char *list[]'' (etc) to ``const char (*)[]'' so
 301 that dynamic lists from things like gdbarch_printable_names() can be
 302 passed.
 303
 304 --
 305
 306 The ``maintenance deprecate set endian big'' command doesn't notice
 307 that it is deprecating ``set endian'' and not ``set endian big'' (big
 308 is implemented using an enum).  Is anyone going to notice this?
 309
 310 --
 311
 312 When tab expanding something like ``set arch<tab>'' ignore the
 313 deprecated ``set archdebug'' and expand to ``set architecture''.
 314
 315 --
 316
 317
 318                         New Features and Fixes
 319                         ======================
 320
 321 These are harder than cleanups but easier than work involving
 322 fundamental architectural change.
 323
 324 --
 325
 326 Add built-by, build-date, tm, xm, nm and anything else into gdb binary
 327 so that you can see how the GDB was created.
 328
 329 Some of these (*m.h) would be added to the generated config.h.  That
 330 in turn would fix a long standing bug where by the build process many
 331 not notice a changed tm.h file.  Since everything depends on config.h,
 332 a change to *m.h forces a change to config.h and, consequently forces
 333 a rebuild.
 334
 335 --
 336
 337 Add an "info bfd" command that displays supported object formats,
 338 similarly to objdump -i.
 339
 340 Is there a command already?
 341
 342 --
 343
 344 Fix ``I'm sorry, Dave, I can't do that.'' from symfile.c.
 345
 346 This requires internationalization.
 347
 348 --
 349
 350 Convert GDB build process to AUTOMAKE.
 351
 352 See also sub-directory configure below.
 353
 354 The current convention is (kind of) to use $(<header>_h) in all
 355 dependency lists.  It isn't done in a consistent way.
 356
 357 --
 358
 359 Cleanup configury support for optional sub-directories.
 360
 361 Check how GCC handles multiple front ends for an example of how things
 362 could work.  A tentative first step is to rationalize things so that
 363 all sub directories are handled in a fashion similar to gdb/mi.
 364
 365 See also automake above.
 366
 367 --
 368
 369 Restructure gdb directory tree so that it avoids any 8.3 and 14
 370 filename problems.
 371
 372 --
 373
 374 Add a transcript mechanism to GDB.
 375
 376 Such a mechanism might log all gdb input and output to a file in a
 377 form that would allow it to be replayed.  It could involve ``gdb
 378 --transcript=FILE'' or it could involve ``(gdb) transcript file''.
 379
 380 --
 381
 382 Can the xdep files be replaced by autoconf?
 383
 384 --
 385
 386 Document trace machinery
 387
 388 --
 389
 390 Document ui-out and ui-file.
 391
 392 http://sourceware.cygnus.com/ml/gdb/2000-04/msg00121.html
 393
 394 --
 395
 396 Update texinfo.tex to latest?
 397
 398
 399
 400 --
 401
 402 Incorporate agentexpr.texi into gdb.texinfo
 403
 404 agentexpr.texi mostly describes the details of the byte code used for
 405 tracepoints, not the internals of the support for this in GDB.  So it
 406 looks like gdb.texinfo is a better place for this information.
 407
 408 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00566.html
 409
 410 --
 411
 412 Document overlay machinery.
 413
 414 --
 415
 416 ``(gdb) catch signal SIGNAL''
 417
 418 Overlaps with ``handle SIGNAL'' but the implied behavour is different.
 419 You can attach commands to a catch but not a handle.  A handle has a
 420 limited number of hardwired actions.
 421
 422 --
 423
 424 Get the TUI working on all platforms.
 425
 426 --
 427
 428 Add support for ``gdb --- PROGRAM ARGS ...''.
 429 Add support for ``gdb -cmd=...''
 430
 431 Along with many variations.  Check:
 432
 433 ????? for a full discussion.
 434
 435 for a discussion.
 436
 437 --
 438
 439 Implement ``(gdb) !ls''.
 440
 441 Which is very different from ``(gdb) ! ls''.  Implementing the latter
 442 is trivial.
 443
 444 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00034.html
 445
 446 --
 447
 448 Change the (char *list[]) to (const char (*)[]) so that dynamic lists can
 449 be passed.
 450
 451 --
 452
 453 When tab expanding something like ``set arch<tab>'' ignore the
 454 deprecated ``set archdebug'' and expand to ``set architecture''.
 455
 456 --
 457
 458 Replace the code that uses the host FPU with an emulator of the target
 459 FPU.
 460
 461 --
 462
 463                         Thread Support
 464                         ==============
 465
 466 --
 467
 468 Generic: lin-thread cannot handle thread exit (Mark Kettenis, Michael
 469 Snyder) http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00525.html
 470
 471 The thread_db assisted debugging code doesn't handle exiting threads
 472 properly, at least in combination with glibc 2.1.3 (the framework is
 473 there, just not the actual code).  There are at least two problems
 474 that prevent this from working.
 475
 476 As an additional reference point, the pre thread_db code did not work
 477 either.
 478
 479 --
 480
 481 GNU/Linux/x86 and random thread signals (and Solaris/SPARC but not
 482 Solaris/x86).
 483 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00336.html
 484
 485 Christopher Blizzard writes:
 486
 487 So, I've done some more digging into this and it looks like Jim
 488 Kingdon has reported this problem in the past:
 489
 490 http://sourceware.cygnus.com/ml/bug-gdb/1999-10/msg00058.html
 491
 492 I can reproduce this problem both with and without Tom's patch.  Has
 493 anyone seen this before?  Maybe have a solution for it hanging around?
 494 :)
 495
 496 There's a test case for this documented at:
 497
 498 when debugging threaded applications you get extra SIGTRAPs
 499 http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=9565
 500
 501 [There should be a GDB testcase - cagney]
 502
 503 --
 504
 505 GDB5 TOT on unixware 7
 506 http://sourceware.cygnus.com/ml/gdb/2000-04/msg00119.html
 507
 508 Robert Lipe writes:
 509 > I just spun the top of tree of the GDB5 branch on UnixWare 7.  As a
 510 > practical matter, the current thread support is somewhat more annoying
 511 > than when GDB was thread-unaware.
 512
 513 --
 514
 515 Migrate qfThreadInfo packet -> qThreadInfo. (Andrew Cagney)
 516
 517 Add support for packet enable/disable commands with these thread
 518 packets.  General cleanup.
 519
 520 [PATCH] Document the ThreadInfo remote protocol queries
 521 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00832.html
 522
 523 [PATCH] "info threads" queries for remote.c
 524 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00831.html
 525
 526 --
 527
 528                         Language Support
 529                         ================
 530
 531 New languages come onto the scene all the time.
 532
 533 --
 534
 535 Pascal (Pierre Muller, David Taylor)
 536
 537 Pierre Muller has contributed patches for adding Pascal Language
 538 support to GDB.
 539
 540 2 pascal language patches inserted in database
 541 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00521.html
 542
 543 Indent -gnu ?
 544 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00496.html
 545
 546 --
 547
 548 Java (Anthony Green, David Taylor)
 549
 550 Anthony Green has a number of Java patches that did not make it into
 551 the 5.0 release.
 552
 553 Patch: java tests
 554 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00512.html
 555
 556 Patch: java booleans
 557 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00515.html
 558
 559 Patch: handle N_MAIN stab
 560 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00527.html
 561
 562 --
 563
 564 [Comming...]
 565
 566 Modify gdb to work correctly with Pascal.
 567
 568 --
 569
 570 Re: Various C++ things
 571
 572 value_headof/value_from_vtable_info are worthless, and should be
 573 removed.  The one place in printcmd.c that uses it should use the RTTI
 574 functions.
 575
 576 RTTI for g++ should be using the typeinfo functions rather than the
 577 vtables.  The typeinfo functions are always at offset 4 from the
 578 beginning of the vtable, and are always right. The vtables will have
 579 weird names like E::VB sometimes.  The typeinfo function will always
 580 be "E type_info function", or somesuch.
 581
 582 value_virtual_fn_field needs to be fixed so there are no failures for
 583 virtual functions for C++ using g++.
 584
 585 Testsuite cases are the major priority right now for C++ support,
 586 since i have to make a lot of changes that could potentially break
 587 each other.
 588
 589 --
 590
 591 Add support for Modula3
 592
 593 Get DEC/Compaq to contribute their Modula-3 support.
 594
 595 --
 596
 597                 Remote Protocol Support
 598                 =======================
 599
 600 --
 601
 602 set/show remote X-packet ...
 603
 604 ``(gdb) help set remote X-packet'' doesn't list the applicable
 605 responses.  The help message needs to be expanded.
 606
 607 --
 608
 609 Remote protocol doco feedback.
 610
 611 Too much feedback to mention needs to be merged in (901660).  Search
 612 for the word ``remote''.
 613
 614
 615 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00023.html
 616 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00056.html
 617 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00382.html
 618
 619 --
 620
 621 GDB doesn't recover gracefully from remote protocol errors.
 622
 623 GDB wasn't checking for NAKs from the remote target.  Instead a NAK is
 624 ignored and a timeout is required before GDB retries.  A pre-cursor to
 625 fixing this this is making GDB's remote protocol packet more robust.
 626
 627 While downloading to a remote protocol target, gdb ignores packet
 628 errors in so far as it will continue to edownload with chunk N+1 even
 629 if chunk N was not correctly sent.  This causes gdb.base/remote.exp to
 630 take a painfully long time to run.  As a PS that test needs to be
 631 fixed so that it builds on 16 bit machines.
 632
 633 --
 634
 635 Add the cycle step command.
 636
 637 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00237.html
 638
 639 --
 640
 641 Resolve how to scale things to support very large packets.
 642
 643 --
 644
 645 Resolve how to handle a target that changes things like its endianess
 646 on the fly - should it be returned in the ``T'' packet?
 647
 648 Underlying problem is that the register file is target endian.  If the
 649 target endianess changes gdb doesn't know.
 650
 651 --
 652
 653                         Symbol Support
 654                         ==============
 655
 656 If / when GDB starts to support the debugging of multi-processor
 657 (rather than multi-thread) applications the symtab code will need to
 658 be updated a little so that several independant symbol tables are
 659 active at a given time.
 660
 661 The other interesting change is a clarification of the exact meaning
 662 of CORE_ADDR and that has had consequences for a few targets (that
 663 were abusing that data type).
 664
 665 --
 666
 667 Investiagate ways of reducing memory.
 668
 669 --
 670
 671 Investigate ways of improving load time.
 672
 673 --
 674
 675 Get the d10v to use POINTER_TO_ADDRESS and ADDRESS_TO_POINTER.
 676
 677 Consequence of recent symtab clarification.  No marks for figuring out
 678 who maintains the d10v.
 679
 680 --
 681
 682 Get the MIPS to correctly sign extend all address <-> pointer
 683 conversions.
 684
 685 Consequence of recent symtab clarification.  No marks for figuring out
 686 who maintains the MIPS.
 687
 688 --
 689
 690                         Testsuite Support
 691                         =================
 692
 693 There are never to many testcases.
 694
 695 --
 696
 697 Better thread testsuite.
 698
 699 --
 700
 701 Better C++ testsuite.
 702
 703 --
 704
 705 Look at adding a GDB specific testsuite directory so that white box
 706 tests of key internals can be added (eg ui_file).
 707
 708 --
 709
 710 Separate out tests that involve the floating point (FP).
 711
 712 (Something for people brining up new targets).  FP and non-fp tests
 713 are combined.  I think there should be set of basic tests that
 714 exercise pure integer support and then a more expanded set that
 715 exercise FP and FP/integer interactions.
 716
 717 As an example, the MIPS, for n32 as problems with passing FP's and
 718 structs.  Since most inferior call tests include FP it is difficult to
 719 determine of the integer tests are ok.
 720
 721 --
 722
 723                 Architectural Changes: General
 724                 ==============================
 725
 726 These are harder than simple cleanups / fixes and, consequently
 727 involve more work.  Typically an Architectural Change will be broken
 728 down into a more digestible set of cleanups and fixes.
 729
 730 --
 731
 732 Cleanup software single step.
 733
 734 At present many targets implement software single step by directly
 735 blatting memory (see rs6000-tdep.c).  Those targets should register
 736 the applicable breakpoints using the breakpoint framework.  Perhaphs a
 737 new internal breakpoint class ``step'' is needed.
 738
 739 --
 740
 741 Replace READ_FP() with FRAME_HANDLE().
 742
 743 READ_FP() is a hangover from the days of the vax when the ABI really
 744 did have a frame pointer register.  Modern architectures typically
 745 construct a virtual frame-handle from the stack pointer and various
 746 other bits of string.
 747
 748 Unfortunatly GDB still treats this synthetic FP register as though it
 749 is real.  That in turn really confuses users (arm and ``print $fp'' VS
 750 ``info registers fp'').  The synthetic FP should be separated out of
 751 the true register set presented to the user.
 752
 753 --
 754
 755 Register Cache Cleanup (below from Andrew Cagney)
 756
 757 I would depict the current register architecture as something like:
 758
 759         High GDB  -->   Low GDB
 760             |              |
 761            \|/            \|/
 762             --- REG NR -----
 763                    |
 764     register + REGISTER_BYTE(reg_nr)
 765                    |
 766                   \|/
 767         -------------------------
 768         |  extern register[]    |
 769         -------------------------
 770
 771 where neither the high (valops.c et.al.) or low gdb (*-tdep.c) are
 772 really clear on what mechanisms they should be using to manipulate that
 773 buffer.   Further, much code assumes, dangerously, that registers are
 774 contigious.  Having got mips-tdep.c to support multiple ABIs, believe
 775 me, that is a bad assumption.  Finally, that register cache layout is
 776 determined by the current remote/local target and _not_ the less
 777 specific target ISA.  In fact, in many cases it is determined by the
 778 somewhat arbitrary layout of the [gG] packets!
 779
 780
 781 How I would like the register file to work is more like:
 782
 783
 784           High GDB
 785              |
 786             \|/
 787        pseudo reg-nr
 788              |
 789         map pseudo <->
 790         random cache
 791         bytes
 792              |
 793             \|/
 794         ------------
 795         | register |
 796         |  cache   |
 797         ------------
 798             /|\
 799              |
 800         map random cache
 801         bytes to target
 802         dependant i-face
 803             /|\
 804              |
 805        target dependant
 806         such as [gG] packet
 807         or ptrace buffer
 808
 809 The main objectives being:
 810
 811         o       a clear separation between the low
 812                 level target and the high level GDB
 813
 814         o       a mechanism that solves the general
 815                 problem of register aliases, overlaps
 816                 etc instead of treating them as optional
 817                 extras that can be wedged in as an after
 818                 thought (that is a reasonable description
 819                 of the current code).
 820
 821                 Identify then solve the hard case and the
 822                 rest just falls out.  GDB solved the easy
 823                 case and then tried to ignore the real
 824                 world :-)
 825
 826         o       a removal of the assumption that the
 827                 mapping between the register cache
 828                 and virtual registers is largely static.
 829                 If you flip the USR/SSR stack register
 830                 select bit in the status-register then
 831                 the corresponding stack registers should
 832                 reflect the change.
 833
 834         o       a mechanism that clearly separates the
 835                 gdb internal register cache from any
 836                 target (not architecture) dependant
 837                 specifics such as [gG] packets.
 838
 839 Of course, like anything, it sounds good in theory.  In reality, it
 840 would have to contend with many<->many relationships at both the
 841 virt<->cache and cache<->target level.  For instance:
 842
 843         virt<->cache
 844                 Modifying an mmx register may involve
 845                 scattering values across both FP and
 846                 mmpx specific parts of a buffer
 847
 848         cache<->target
 849                 When writing back a SP it may need to
 850                 both be written to both SP and USP.
 851
 852
 853 Hmm,
 854
 855 Rather than let this like the last time it was discussed, just slip, I'm
 856 first going to add this e-mail (+ references) to TODO.  I'd then like to
 857 sketch out a broad strategy I think could get us there.
 858
 859
 860 First thing I'd suggest is separating out the ``extern registers[]''
 861 code so that we can at least identify what is using it.  At present
 862 things are scattered across many files.  That way we can at least
 863 pretend that there is a cache instead of a global array :-)
 864
 865 I'd then suggest someone putting up a proposal for the pseudo-reg /
 866 high-level side interface so that code can be adopted to it.  For old
 867 code, initially a blanket rename of write_register_bytes() to
 868 deprecated_write_register_bytes() would help.
 869
 870 Following that would, finaly be the corresponding changes to the target.
 871
 872 --
 873
 874 Check that GDB can handle all BFD architectures (Andrew Cagney)
 875
 876 There should be a test that checks that BFD/GDB are in sync with
 877 regard to architecture changes.  Something like a test that first
 878 queries GDB for all supported architectures and then feeds each back
 879 to GDB..  Anyone interested in learning how to write tests?  :-)
 880
 881 --
 882
 883                 Architectural Change: Multi-arch et al.
 884                 =======================================
 885
 886 The long term objective is to remove all assumptions that there is a
 887 single target with a single address space with a single instruction
 888 set architecture and single application binary interface.
 889
 890 This is an ongoing effort.  The first milestone is to enable
 891 ``multi-arch'' where by all architectural decisions are made at
 892 runtime.
 893
 894 It should be noted that ``gdbarch'' is really ``gdbabi'' and
 895 ``gdbisa''.  Once things are multi-arched breaking that down correctly
 896 will become much easier.
 897
 898 --
 899
 900 GDBARCH cleanup (Andrew Cagney)
 901
 902 The non-generated parts of gdbarch.{sh,h,c} should be separated out
 903 into arch-utils.[hc].
 904
 905 Document that gdbarch_init_ftype could easily fail because it didn't
 906 identify an architecture.
 907
 908 --
 909
 910 Fix BELIEVE_PPC_PROMOTION.  Change it to BELIEVE_PPC_PROMOTION_P?
 911
 912 At present there is still #ifdef BELIEVE_PPC_PROMOTION code in the
 913 symtab file.
 914
 915 --
 916
 917 Fix target_signal_from_host() etc.
 918
 919 The name is wrong for starters.  ``target_signal'' should probably be
 920 ``gdb_signal''.  ``from_host'' should be ``from_target_signal''.
 921 After that it needs to be multi-arched and made independant of any
 922 host signal numbering.
 923
 924 --
 925
 926 Update ALPHA so that it uses ``struct frame_extra_info'' instead of
 927 EXTRA_FRAME_INFO.
 928
 929 This is a barrier to replacing mips_extra_func_info with something
 930 that works with multi-arch.
 931
 932 --
 933
 934 Multi-arch  mips_extra_func_info.
 935
 936 This first needs the alpha to be updated so that it uses ``struct
 937 frame_extra_info''.
 938
 939 --
 940
 941 Rationalize TARGET_SINGLE_FORMAT and TARGET_SINGLE_BIT et al.
 942
 943 Surely one of them is redundant.
 944
 945 --
 946
 947 Convert ALL architectures to MULTI-ARCH.
 948
 949 --
 950
 951 Select the initial multi-arch ISA / ABI based on --target or similar.
 952
 953 At present the default is based on what ever is first in the BFD
 954 archures table.  It should be determined based on the ``--target=...''
 955 name.
 956
 957 --
 958
 959 Truly multi-arch.
 960
 961 Enable the code to recognize --enable-targets=.... like BINUTILS does.
 962
 963 Can the tm.h and nm.h files be eliminated by multi-arch.
 964
 965 --
 966
 967         Architectural Change: MI, LIBGDB and scripting languages
 968         ========================================================
 969
 970 See also architectural changes related to the event loop.  LIBGDB
 971 can't be finished until there is a generic event loop being used by
 972 all targets.
 973
 974 The long term objective is it to be possible to integrate GDB into
 975 scripting languages.
 976
 977 --
 978
 979 Implement generic ``(gdb) commmand > file''
 980
 981 Once everything is going through ui_file it should be come fairly
 982 easy.
 983
 984 http://sourceware.cygnus.com/ml/gdb/2000-04/msg00104.html
 985
 986 --
 987
 988 Replace gdb_stdtarg with gdb_targout (and possibly gdb_targerr).
 989
 990 gdb_stdtarg is easily confused with gdb_stdarg.
 991
 992 --
 993
 994 Extra ui_file methods - dump.
 995
 996 Very useful for whitebox testing.
 997
 998 --
 999
1000 Eliminate error_begin().
1001
1002 With ui_file, there is no need for the statefull error_begin ()
1003 function.
1004
1005 --
1006
1007 Send normal output to gdb_stdout.
1008 Send error messages to gdb_stderror.
1009 Send debug and log output log gdb_stdlog.
1010
1011 GDB still contains many cases where (f)printf or printf_filtered () is
1012 used when it should be sending the messages to gdb_stderror or
1013 gdb_stdlog.  The thought of #defining printf to something has crossed
1014 peoples minds ;-)
1015
1016 --
1017
1018 Re-do GDB's output pager.
1019
1020 GDB's output pager still relies on people correctly using *_filtered
1021 for gdb_stdout and *_unfiltered for gdb_stdlog / gdb_stderr.
1022 Hopefully, with all normal output going to gdb_stdout, the pager can
1023 just look at the ui_file that the output is on and then use that to
1024 decide what to do about paging.  Sounds good in theory.
1025
1026 --
1027
1028 Check/cleanup MI documentation.
1029
1030 The list of commands specified in the documentation needs to be
1031 checked against the mi-cmds.c table in a mechanical way (so that they
1032 two can be kept up-to-date).
1033
1034 --
1035
1036 Convert MI into libgdb
1037
1038 MI provides a text interface into what should be many of the libgdb
1039 functions.  The implementation of those functions should be separated
1040 into the MI interface and the functions proper.  Those functions being
1041 moved to gdb/lib say.
1042
1043 --
1044
1045 Create libgdb.h
1046
1047 The first part can already be found in defs.h.
1048
1049 --
1050
1051 MI's input does not use buffering.
1052
1053 At present the MI interface reads raw characters of from an unbuffered
1054 FD.  This is to avoid several nasty buffer/race conditions.  That code
1055 should be changed so that it registers its self with the event loop
1056 (on the input FD) and then push commands up to MI as they arrive.
1057
1058 The serial code already does this.
1059
1060 --
1061
1062 Make MI interface accessable from existing CLI.
1063
1064 --
1065
1066 Add a breakpoint-edit command to MI.
1067
1068 It would be similar to MI's breakpoint create but would apply to an
1069 existing breakpoint.  It saves the need to delete/create breakpoints
1070 when ever they are changed.
1071
1072 --
1073
1074 Add directory path to MI breakpoint.
1075
1076 That way the GUI's task of finding the file within which the
1077 breakpoint was set is simplified.
1078
1079 --
1080
1081 Add a mechanism to reject certain expression classes to MI
1082
1083 There are situtations where you don't want GDB's expression
1084 parser/evaluator to perform inferior function calls or variable
1085 assignments.  A way of restricting the expression parser so that such
1086 operations are not accepted would be very helpful.
1087
1088 --
1089
1090 Remove sideffects from libgdb breakpoint create function.
1091
1092 The user can use the CLI to create a breakpoint with partial
1093 information - no file (gdb would use the file from the last
1094 breakpoint).
1095
1096 The libgdb interface currently affects that environment which can lead
1097 to confusion when a user is setting breakpoints via both the MI and
1098 the CLI.
1099
1100 This is also a good example of how getting the CLI ``right'' will be
1101 hard.
1102
1103 --
1104
1105 Move gdb_lasterr to ui_out?
1106
1107 The way GDB throws errors and records them needs a re-think.  ui_out
1108 handles the correct output well.  It doesn't resolve what to do with
1109 output / error-messages when things go wrong.
1110
1111 --
1112
1113                 Architectural Change: Async
1114                 ===========================
1115
1116 While GDB uses an event loop when prompting the user for input.  That
1117 event loop is not exploited by targets when they allow the target
1118 program to continue.  Typically targets still block in (target_wait())
1119 until the program again halts.
1120
1121 The closest a target comes to supporting full asynchronous mode are
1122 the remote targets ``async'' and ``extended-async''.
1123
1124 --
1125
1126 Asynchronous expression evaluator
1127
1128 Inferior function calls hang GDB.
1129
1130 --
1131
1132 Fix implementation of  ``target xxx''.
1133
1134 At present when the user specifies ``target xxxx'', the CLI maps that
1135 directly onto a target open method.  It is then assumed that the
1136 target open method should do all sorts of complicated things as this
1137 is the only chance it has.  Check how the various remote targets
1138 duplicate the target operations.  Check also how the various targets
1139 behave differently for purely arbitrary reasons.
1140
1141 What should happen is that ``target xxxx'' should call a generic
1142 ``target'' function and that should then co-ordinate the opening of
1143 ``xxxx''.  This becomes especially important when you're trying to
1144 open an asynchronous target that may need to perform background tasks
1145 as part of the ``attach'' phase.
1146
1147 Unfortunatly, due to limitations in the old/creaking command.h
1148 interface, that isn't possible.  The function being called isn't told
1149 of the ``xxx'' or any other context information.
1150
1151 Consequently a precursor to fixing ``target xxxx'' is to clean up the
1152 CLI code so that it passes to the callback function (attatched to a
1153 command) useful information such as the actual command and a context
1154 for that command.  Other changes such as making ``struct command''
1155 opaque may also help.
1156
1157 --
1158
1159 Make "target xxx" command interruptible.
1160
1161 As things become async this becomes possible.  A target would start
1162 the connect and then return control to the event loop.  A cntrl-c
1163 would notify the target that the operation is to be abandoned and the
1164 target code could respond.
1165
1166 --
1167
1168 Add a "suspend" subcommand of the "continue" command to suspend gdb
1169 while continuing  execution of the subprocess.  Useful when you are
1170 debugging servers and you want to dodge out and initiate a connection
1171 to a server running under gdb.
1172
1173 [hey async!!]
1174
1175 --
1176
1177                         TODO FAQ
1178                         ========
1179
1180 Frequently requested but not approved requests.
1181
1182 --
1183
1184 Eliminate unused argument warnings using ATTRIBUTE_UNUSED.
1185
1186 The benefits on this one are thought to be marginal - GDBs design
1187 means that unused parameters are very common.  GCC 3.0 will also
1188 include the option -Wno-unused-parameter which means that ``-Wall
1189 -Wno-unused-parameters -Werror'' can be specified.
1190
1191 --
1192
1193
1194
1195                         Legacy Wish List
1196                         ================
1197
1198 This list is not up to date, and opinions vary about the importance or
1199 even desirability of some of the items.  If you do fix something, it
1200 always pays to check the below.
1201
1202 --
1203
1204 @c This does not work (yet if ever).  FIXME.
1205 @c @item --parse=@var{lang} @dots{}
1206 @c Configure the @value{GDBN} expression parser to parse the listed languages.
1207 @c @samp{all} configures @value{GDBN} for all supported languages.  To get a
1208 @c list of all supported languages, omit the argument.  Without this
1209 @c option, @value{GDBN} is configured to parse all supported languages.
1210
1211 --
1212
1213 START_INFERIOR_TRAPS_EXPECTED need never be defined to 2, since that
1214 is its default value.  Clean this up.
1215
1216 --
1217
1218 It should be possible to use symbols from shared libraries before we know
1219 exactly where the libraries will be loaded.  E.g. "b perror" before running
1220 the program.  This could maybe be done as an extension of the "breakpoint
1221 re-evaluation" after new symbols are loaded.
1222
1223 --
1224
1225 Make single_step() insert and remove breakpoints in one operation.
1226
1227 [If this is talking about having single_step() insert the breakpoints,
1228 run the target then pull the breakpoints then it is wrong.  The
1229 function has to return as control has to eventually be passed back to
1230 the main event loop.]
1231
1232 --
1233
1234 Speed up single stepping by avoiding extraneous ptrace calls.
1235
1236 --
1237
1238 Speed up single stepping by not inserting and removing breakpoints
1239 each time the inferior starts and stops.
1240
1241 Breakpoints should not be inserted and deleted all the time.  Only the
1242 one(s) there should be removed when we have to step over one.  Support
1243 breakpoints that don't have to be removed to step over them.
1244
1245 [this has resulted in numerous debates.  The issue isn't clear cut]
1246
1247 --
1248
1249 Provide "voodoo" debugging of core files.  This creates a zombie
1250 process as a child of the debugger, and loads it up with the data,
1251 stack, and regs of the core file.  This allows you to call functions
1252 in the executable, to manipulate the data in the core file.
1253
1254 [you wish]
1255
1256 --
1257
1258 GDB reopens the source file on every line, as you "next" through it.
1259
1260 [still true? I've a memory of this being fixed]
1261
1262 --
1263
1264 Perhaps "i source" should take an argument like that of "list".
1265
1266 --
1267
1268 Remove "at 0xnnnn" from the "b foo" response, if `print address off' and if
1269 it matches the source line indicated.
1270
1271 --
1272
1273 The prompt at end of screen should accept space as well as CR.
1274
1275 --
1276
1277 Backtrace should point out what the currently selected frame is, in
1278 its display, perhaps showing "@3 foo (bar, ...)" or ">3 foo (bar,
1279 ...)" rather than "#3 foo (bar, ...)".
1280
1281 --
1282
1283 "i program" should work for core files, and display more info, like what
1284 actually caused it to die.
1285
1286 --
1287
1288 "x/10i" should shorten the long name, if any, on subsequent lines.
1289
1290 --
1291
1292 "next" over a function that longjumps, never stops until next time you happen
1293 to get to that spot by accident.  E.g. "n" over execute_command which has
1294 an error.
1295
1296 --
1297
1298 "set zeroprint off", don't bother printing members of structs which
1299 are entirely zero.  Useful for those big structs with few useful
1300 members.
1301
1302 --
1303
1304 GDB does four ioctl's for every command, probably switching terminal modes
1305 to/from inferior or for readline or something.
1306
1307 --
1308
1309 terminal_ours versus terminal_inferior: cache state.  Switch should be a noop
1310 if the state is the same, too.
1311
1312 --
1313
1314 "i frame" shows wrong "arglist at" location, doesn't show where the args
1315 should be found, only their actual values.
1316
1317 --
1318
1319 There should be a way for "set" commands to validate the new setting
1320 before it takes effect.
1321
1322 --
1323
1324 "ena d" is ambiguous, why?  "ena delete" seems to think it is a command!
1325
1326 --
1327
1328 i line VAR produces "Line number not known for symbol ``var''.".  I
1329 thought we were stashing that info now!
1330
1331 --
1332
1333 We should be able to write to random files at hex offsets like adb.
1334
1335 --
1336
1337 [elena - delete this]
1338
1339 Handle add_file with separate text, data, and bss addresses.  Maybe
1340 handle separate addresses for each segment in the object file?
1341
1342 --
1343
1344 [Jimb/Elena delete this one]
1345
1346 Handle free_named_symtab to cope with multiply-loaded object files
1347 in a dynamic linking environment.  Should remember the last copy loaded,
1348 but not get too snowed if it finds references to the older copy.
1349
1350 --
1351
1352 [elena delete this also]
1353
1354 Remove all references to:
1355         text_offset
1356         data_offset
1357         text_data_start
1358         text_end
1359         exec_data_offset
1360         ...
1361 now that we have BFD.  All remaining are in machine dependent files.
1362
1363 --
1364
1365 Re-organize help categories into things that tend to fit on a screen
1366 and hang together.
1367
1368 --
1369
1370 Add in commands like ADB's for searching for patterns, etc.  We should
1371 be able to examine and patch raw unsymboled binaries as well in gdb as
1372 we can in adb.  (E.g. increase the timeout in /bin/login without source).
1373
1374 [actually, add ADB interface :-]
1375
1376 --
1377
1378 When doing "step" or "next", if a few lines of source are skipped between
1379 the previous line and the current one, print those lines, not just the
1380 last line of a multiline statement.
1381
1382 --
1383
1384 Handling of "&" address-of operator needs some serious overhaul
1385 for ANSI C and consistency on arrays and functions.
1386   For "float point[15];":
1387 ptype &point[4]   ==> Attempt to take address of non-lvalue.
1388   For "char *malloc();":
1389 ptype malloc      ==> "char *()";  should be same as
1390 ptype &malloc     ==> "char *(*)()"
1391 call printf ("%x\n", malloc)   ==> weird value, should be same as
1392 call printf ("%x\n", &malloc)  ==> correct value
1393
1394 --
1395
1396 Fix dbxread.c symbol reading in the presence of interrupts.  It
1397 currently leaves a cleanup to blow away the entire symbol table when a
1398 QUIT occurs.  (What's wrong with that? -kingdon, 28 Oct 1993).
1399
1400 [I suspect that the grype was that, on a slow system, you might want
1401 to cntrl-c and get just half the symbols and then load the rest later
1402 - scary to be honest]
1403
1404 --
1405
1406 Mipsread.c reads include files depth-first, because the dependencies
1407 in the psymtabs are way too inclusive (it seems to me).  Figure out what
1408 really depends on what, to avoid recursing 20 or 30 times while reading
1409 real symtabs.
1410
1411 --
1412
1413 value_add() should be subtracting the lower bound of arrays, if known,
1414 and possibly checking against the upper bound for error reporting.
1415
1416 --
1417
1418 When listing source lines, check for a preceding \n, to verify that
1419 the file hasn't changed out from under us.
1420
1421 [fixed by some other means I think.  That hack wouldn't actually work
1422 reliably - the file might move such that another \n appears. ]
1423
1424 --
1425
1426 Get all the remote systems (where the protocol allows it) to be able to
1427 stop the remote system when the GDB user types ^C (like remote.c
1428 does).  For ebmon, use ^Ak.
1429
1430 --
1431
1432 Possible feature: A version of the "disassemble" command which shows
1433 both source and assembly code ("set symbol-filename on" is a partial
1434 solution).
1435
1436 [has this been done?  It was certainly done for MI and GDBtk]
1437
1438 --
1439
1440 investigate "x/s 0" (right now stops early) (I think maybe GDB is
1441 using a 0 address for bad purposes internally).
1442
1443 --
1444
1445 Make "info path" and path_command work again (but independent of the
1446 environment either of gdb or that we'll pass to the inferior).
1447
1448 --
1449
1450 Make GDB understand the GCC feature for putting octal constants in
1451 enums.  Make it so overflow on an enum constant does not error_type
1452 the whole type.  Allow arbitrarily large enums with type attributes.
1453 Put all this stuff in the testsuite.
1454
1455 --
1456
1457 Make TYPE_CODE_ERROR with a non-zero TYPE_LENGTH more useful (print
1458 the value in hex; process type attributes).  Add this to the
1459 testsuite.  This way future compilers can add new types and old
1460 versions of GDB can do something halfway reasonable.
1461
1462 --
1463
1464 Fix mdebugread.c:parse_type to do fundamental types right (see
1465 rs6000_builtin_type in stabsread.c for what "right" is--the point is
1466 that the debug format fixes the sizes of these things and it shouldn't
1467 depend on stuff like TARGET_PTR_BIT and so on.  For mdebug, there seem
1468 to be separate bt* codes for 64 bit and 32 bit things, and GDB should
1469 be aware of that).  Also use a switch statement for clarity and speed.
1470
1471 --
1472
1473 Investigate adding symbols in target_load--some targets do, some
1474 don't.
1475
1476 --
1477
1478 Put dirname in psymtabs and change lookup*symtab to use dirname (so
1479 /foo/bar.c works whether compiled by cc /foo/bar.c, or cd /foo; cc
1480 bar.c).
1481
1482 --
1483
1484 Merge xcoffread.c and coffread.c.  Use breakpoint_re_set instead of
1485 fixup_breakpoints.
1486
1487 --
1488
1489 Make a watchpoint which contains a function call an error (it is
1490 broken now, making it work is probably not worth the effort).
1491
1492 --
1493
1494 New test case based on weird.exp but in which type numbers are not
1495 renumbered (thus multiply defining a type).  This currently causes an
1496 infinite loop on "p v_comb".
1497
1498 --
1499
1500 [Hey! Hint Hint Delete Delete!!!]
1501
1502 Fix 386 floating point so that floating point registers are real
1503 registers (but code can deal at run-time if they are missing, like
1504 mips and 68k).  This would clean up "info float" and related stuff.
1505
1506 --
1507
1508 gcc -g -c enummask.c then gdb enummask.o, then "p v".  GDB complains
1509 about not being able to access memory location 0.
1510
1511 -------------------- enummask.c
1512 enum mask
1513 {
1514   ANIMAL = 0,
1515   VEGETABLE = 1,
1516   MINERAL = 2,
1517   BASIC_CATEGORY = 3,
1518
1519   WHITE = 0,
1520   BLUE = 4,
1521   GREEN = 8,
1522   BLACK = 0xc,
1523   COLOR = 0xc,
1524
1525   ALIVE = 0x10,
1526
1527   LARGE = 0x20
1528 } v;
1529
1530 --
1531
1532 If try to modify value in file with "set write off" should give
1533 appropriate error not "cannot access memory at address 0x65e0".
1534
1535 --
1536
1537 Allow core file without exec file on RS/6000.
1538
1539 --
1540
1541 Make sure "shell" with no arguments works right on DOS.
1542
1543 --
1544
1545 Make gdb.ini (as well as .gdbinit) be checked on all platforms, so
1546 the same directory can be NFS-mounted on unix or DOS, and work the
1547 same way.
1548
1549 --
1550
1551 [Is this another delete???]
1552
1553 Get SECT_OFF_TEXT stuff out of objfile_relocate (might be needed to
1554 get RS/6000 to work right, might not be immediately relevant).
1555
1556 --
1557
1558 Work out some kind of way to allow running the inferior to be done as
1559 a sub-execution of, eg. breakpoint command lists.  Currently running
1560 the inferior interupts any command list execution.  This would require
1561 some rewriting of wait_for_inferior & friends, and hence should
1562 probably be done in concert with the above.
1563
1564 --
1565
1566 Add function arguments to gdb user defined functions.
1567
1568 --
1569
1570 Add convenience variables that refer to exec file, symbol file,
1571 selected frame source file, selected frame function, selected frame
1572 line number, etc.
1573
1574 --
1575
1576 Modify the handling of symbols grouped through BINCL/EINCL stabs to
1577 allocate a partial symtab for each BINCL/EINCL grouping.  This will
1578 seriously decrease the size of inter-psymtab dependencies and hence
1579 lessen the amount that needs to be read in when a new source file is
1580 accessed.
1581
1582 --
1583
1584 Add a command for searching memory, a la adb.  It specifies size,
1585 mask, value, start address.  ADB searches until it finds it or hits
1586 an error (or is interrupted).
1587
1588 --
1589
1590 Remove the range and type checking code and documentation, if not
1591 going to implement.
1592
1593 # Local Variables:
1594 # mode: text
1595 # End: