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 (The names in paren indicate people that posted the original problem.)
  15
  16 --
  17
  18 GDB requires GCC to build under IRIX
  19
  20 IRIX, being more pedantic than GCC reports as errors certain
  21 assignments that GCC treats as warnings.
  22
  23 This can be worked around by building GDB with the GCC compiler.
  24
  25 --
  26
  27 The BFD directory requires bug-fixed AUTOMAKE et.al.
  28
  29 AUTOMAKE 1.4 incorrectly set the TEXINPUTS environment variable.  It
  30 contained the full path to texinfo.tex when it should have only
  31 contained the directory.  The bug has been fixed in the current
  32 AUTOMAKE sources.  Automake snapshots can be found in:
  33         ftp://sourceware.cygnus.com/pub/gdb/snapshots
  34 and     ftp://sourceware.cygnus.com/pub/binutils
  35
  36 --
  37
  38 Possible regressions with some devel GCCs.
  39 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00475.html
  40
  41 gcc-2.95.2 outputs a line note *before* the prologue (and one for the
  42 closing brace after the epilogue, instead of before it, as it used to
  43 be). By disabling the RTL-style prologue generating mechanism
  44 (undocumented GCC option -mno-schedule-prologue), you get back the
  45 traditional behaviour.
  46 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00510.html
  47
  48 This should now be fixed.
  49
  50 --
  51
  52 RFD: infrun.c: No bpstat_stop_status call after proceed over break?
  53 (Peter Schauer)
  54 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00665.html
  55
  56 GDB misses watchpoint triggers after proceeding over a breakpoint on
  57 x86 targets.
  58
  59 --
  60
  61 x86 linux GDB and SIGALRM (???)
  62 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00803.html
  63
  64 I know there are problems with single stepping through signal
  65 handlers.  These problems were present in 4.18.  They were just masked
  66 because 4.18 failed to recognize signal handlers.  Fixing it is not
  67 easy, and will require changes to handle_inferior_event(), that I
  68 prefer not to make before the 5.0 release.
  69
  70 Mark
  71
  72 --
  73
  74 Revised UDP support (was: Re: [Fwd: [patch] UDP transport support])
  75 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00000.html
  76
  77 (Broken) support for GDB's remote protocol across UDP is to be
  78 included in the follow-on release.
  79
  80 --
  81
  82 Can't build IRIX -> arm GDB.
  83 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00356.html
  84
  85 David Whedon writes:
  86 > Now I'm building for an embedded arm target.  If there is a way of turning
  87 > remote-rdi off, I couldn't find it. It looks like it gets built by default
  88 > in gdb/configure.tgt(line 58) Anyway, the build dies in
  89 > gdb/rdi-share/unixcomm.c.  SERPORT1 et. al. never get defined because we
  90 > aren't one of the architectures supported.
  91
  92 --
  93
  94 Problem with weak functions
  95 http://sourceware.cygnus.com/ml/gdb/2000-05/msg00060.html
  96
  97 Dan Nicolaescu writes:
  98 > It seems that gdb-4.95.1  does not display correctly the function when
  99 > stoping in weak functions.
 100 >
 101 > It stops in a function that is defined as weak, not in the function
 102 > that is actualy run...
 103
 104 --
 105
 106                 Code Cleanups: Next Release
 107                 ===========================
 108
 109 The following are small cleanups that will hopefully be completed by
 110 the follow on to 5.0.
 111
 112 --
 113
 114 Delete macro TARGET_BYTE_ORDER_SELECTABLE.
 115
 116 Patches in the database.
 117
 118 --
 119
 120 Purge PARAMS
 121
 122 Eliminate all uses of PARAMS in GDB's source code.
 123
 124 --
 125
 126 Elimination of make_cleanup_func. (Andrew Cagney)
 127
 128 make_cleanup_func elimination
 129 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00791.html
 130 http://sourceware.cygnus.com/ml/gdb-patches/2000-q1/msg00814.html
 131
 132 --
 133
 134 Fix copyright notices.
 135
 136 Turns out that ``1998-2000'' isn't considered valid :-(
 137
 138 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00467.html
 139
 140 --
 141
 142                 Code Cleanups: General
 143                 ======================
 144
 145 The following are more general cleanups and fixes.  They are not tied
 146 to any specific release.
 147
 148 --
 149
 150 ATTRIBUTE_UNUSED
 151
 152 The need for this as almost been eliminated.  The next version of GCC
 153 (assuming cagney gets the relevant patch committed) will be able to
 154 supress unused parameter warnings.
 155
 156 --
 157
 158 Eliminate more compiler warnings.
 159
 160 Of course there also needs to be the usual debate over which warnings
 161 are valid and how to best go about this.
 162
 163 One method: choose a single option; get agreement that it is
 164 reasonable; try it out to see if there isn't anything silly about it
 165 (-Wunused-parameters is an example of that) then incrementally hack
 166 away.
 167
 168 The other method is to enable all warnings and eliminate them from one
 169 file at a time.
 170
 171 --
 172
 173 Elimination of ``(catch_errors_ftype *) func''.
 174
 175 Like make_cleanup_func it isn't portable.
 176
 177 --
 178
 179 Nuke USG define.
 180
 181 --
 182
 183 [PATCH/5] src/intl/Makefile.in:distclean additions
 184 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00363.html
 185
 186 Do not forget to merge the patch back into the trunk.
 187
 188 --
 189
 190 Rationalize the host-endian code (grep for HOST_BYTE_ORDER).
 191
 192 At present defs.h includes <endian.h> (which is linux specific) yet
 193 almost nothing depends on it.  Suggest "gdb_endian.h" which can also
 194 handle <machine/endian.h> and only include that where it is really
 195 needed.
 196
 197 --
 198
 199 Replace asprintf() calls with xasprintf() calls.
 200
 201 As with things like strdup() most calls to asprintf() don't check the
 202 return value.
 203
 204 --
 205
 206 Replace strsave() + mstrsave() with libiberty:xstrdup().
 207
 208 --
 209
 210 Replace savestring() with something from libiberty.
 211
 212 An xstrldup()? but that would have different semantics.
 213
 214 --
 215
 216 Rationalize use of floatformat_unknown in GDB sources.
 217
 218 Instead of defaulting to floatformat_unknown, should hosts/targets
 219 specify the value explicitly?
 220
 221 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00447.html
 222
 223 --
 224
 225 Add a ``name'' member to include/floatformat.h:struct floatformat.
 226 Print that name in gdbarch.c.
 227
 228 --
 229
 230 Sort out the harris mess in include/floatformat.h (it hardwires two
 231 different floating point formats).
 232
 233 --
 234
 235 See of the GDB local floatformat_do_doublest() and libiberty's
 236 floatformat_to_double (which was once GDB's ...) can be merged some
 237 how.
 238
 239 --
 240
 241 Eliminate mmalloc() from GDB.
 242
 243 Also eliminate it from defs.h.
 244
 245 --
 246
 247 Eliminate PTR.  ISO-C allows ``void *''.
 248
 249 --
 250
 251 Eliminate abort ().
 252
 253 GDB should never abort.  GDB should either throw ``error ()'' or
 254 ``internal_error ()''.  Better still GDB should naturally unwind with
 255 an error status.
 256
 257 --
 258
 259 GDB probably doesn't build on FreeBSD pre 2.2.x
 260 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00378.html
 261
 262 Fixes to get FreeBSD working on 2.2.x, 3.x and 4.x caused the code to
 263 suffer bit rot.
 264
 265 --
 266
 267 Updated readline
 268
 269 Readline 4.? is out.  A merge wouldn't hurt.  Patches are in:
 270
 271 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00436.html
 272
 273 --
 274
 275 Deprecate "fg".  Apparently ``fg'' is actually continue.
 276
 277 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00417.html
 278
 279 --
 280
 281 Deprecate current use of ``floatformat_unknown''.
 282
 283 Require all targets to explicitly provide their float format instead
 284 of defaulting to floatformat unknown.  Doing the latter leads to nasty
 285 bugs.
 286
 287 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00447.html
 288
 289 --
 290
 291 Rationalize floatformat_to_double() vs floatformat_to_doublest().
 292
 293 Looks like GDB migrated floatformat_to_double() to libiberty but then
 294 turned around and created a ..._to_doublest() the latter containing
 295 several bug fixes.
 296
 297 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00472.html
 298
 299 --
 300
 301 Move floatformat_ia64_ext to libiberty/include floatformat.[ch].
 302
 303 http://sourceware.cygnus.com/ml/gdb-patches/2000-05/msg00466.html
 304
 305 --
 306
 307 Always build ser-tcp.c.
 308
 309 The patch as submitted was just going to add ser-tcp.c to the Alpha's
 310 makefile.  A better patch is to instead add ser-tcp.c to SER_HARDWARE
 311 and make it a standard part of all debuggers.
 312
 313 If problems occure then configure.in can sort them out.
 314
 315 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00544.html
 316
 317 --
 318
 319 Follow through `make check' with --enable-shared.
 320
 321 When the srcware tree is configured with --enable-shared, the `expect'
 322 program won't run properly.  Jim Wilson found out gdb has a local hack
 323 to set LD_LIBRARY_PATH, but, AFAIK, no other project has been hacked
 324 similarly.
 325
 326 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00845.html
 327
 328 --
 329
 330
 331                         New Features and Fixes
 332                         ======================
 333
 334 These are harder than cleanups but easier than work involving
 335 fundamental architectural change.
 336
 337 --
 338
 339 Add built-by, build-date, tm, xm, nm and anything else into gdb binary
 340 so that you can see how the GDB was created.
 341
 342 Some of these (*m.h) would be added to the generated config.h.  That
 343 in turn would fix a long standing bug where by the build process many
 344 not notice a changed tm.h file.  Since everything depends on config.h,
 345 a change to *m.h forces a change to config.h and, consequently forces
 346 a rebuild.
 347
 348 --
 349
 350 Add an "info bfd" command that displays supported object formats,
 351 similarly to objdump -i.
 352
 353 Is there a command already?
 354
 355 --
 356
 357 Fix ``I'm sorry, Dave, I can't do that.'' from symfile.c.
 358
 359 This requires internationalization.
 360
 361 --
 362
 363 Convert GDB build process to AUTOMAKE.
 364
 365 See also sub-directory configure below.
 366
 367 --
 368
 369 Cleanup configury support for optional sub-directories.
 370
 371 Check how GCC handles multiple front ends for an example of how things
 372 could work.  A tentative first step is to rationalize things so that
 373 all sub directories are handled in a fashion similar to gdb/mi.
 374
 375 See also automake above.
 376
 377 --
 378
 379 Restructure gdb directory tree so that it avoids any 8.3 and 14
 380 filename problems.
 381
 382 --
 383
 384 Add a transcript mechanism to GDB.
 385
 386 Such a mechanism might log all gdb input and output to a file in a
 387 form that would allow it to be replayed.  It could involve ``gdb
 388 --transcript=FILE'' or it could involve ``(gdb) transcript file''.
 389
 390 --
 391
 392 Can the xdep files be replaced by autoconf?
 393
 394 --
 395
 396 Document trace machinery
 397
 398 --
 399
 400 Document ui-out and ui-file.
 401
 402 http://sourceware.cygnus.com/ml/gdb/2000-04/msg00121.html
 403
 404 --
 405
 406 Update texinfo.tex to latest?
 407
 408
 409
 410 --
 411
 412 Incorporate agentexpr.texi into gdb.texinfo
 413
 414 agentexpr.texi mostly describes the details of the byte code used for
 415 tracepoints, not the internals of the support for this in GDB.  So it
 416 looks like gdb.texinfo is a better place for this information.
 417
 418 http://sourceware.cygnus.com/ml/gdb-patches/2000-04/msg00566.html
 419
 420 --
 421
 422 Document overlay machinery.
 423
 424 --
 425
 426 ``(gdb) catch signal SIGNAL''
 427
 428 Overlaps with ``handle SIGNAL'' but the implied behavour is different.
 429 You can attach commands to a catch but not a handle.  A handle has a
 430 limited number of hardwired actions.
 431
 432 --
 433
 434 Get the TUI working on all platforms.
 435
 436 --
 437
 438 Add support for ``gdb --- PROGRAM ARGS ...''.
 439 Add support for ``gdb -cmd=...''
 440
 441 Along with many variations.  Check:
 442
 443 ????? for a full discussion.
 444
 445 for a discussion.
 446
 447 --
 448
 449 Implement ``(gdb) !ls''.
 450
 451 Which is very different from ``(gdb) ! ls''.  Implementing the latter
 452 is trivial.
 453
 454 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00034.html
 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 removed.
 573 The one place in printcmd.c that uses it should use the RTTI functions.
 574
 575 RTTI for g++ should be using the typeinfo functions rather than the vtables.
 576 The typeinfo functions are always at offset 4 from the beginning of the vtable,
 577 and are always right. The vtables will have weird names like E::VB sometimes.
 578 The typeinfo function will always be "E type_info function", or somesuch.
 579
 580 value_virtual_fn_field needs to be fixed so there are no failures for virtual
 581 functions for C++ using g++.
 582
 583 Testsuite cases are the major priority right now for C++ support, since i have
 584 to make a lot of changes that could potentially break each other.
 585
 586 --
 587
 588 Add support for Modula3
 589
 590 Get DEC/Compaq to contribute their Modula-3 support.
 591
 592 --
 593
 594                 Remote Protocol Support
 595                 =======================
 596
 597 --
 598
 599 set/show remote X-packet ...
 600
 601 ``(gdb) help set remote X-packet'' doesn't list the applicable
 602 responses.  The help message needs to be expanded.
 603
 604 --
 605
 606 Remote protocol doco feedback.
 607
 608 Too much feedback to mention needs to be merged in (901660).  Search
 609 for the word ``remote''.
 610
 611
 612 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00023.html
 613 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00056.html
 614 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00382.html
 615
 616 --
 617
 618 GDB doesn't recover gracefully from remote protocol errors.
 619
 620 GDB wasn't checking for NAKs from the remote target.  Instead a NAK is
 621 ignored and a timeout is required before GDB retries.  A pre-cursor to
 622 fixing this this is making GDB's remote protocol packet more robust.
 623
 624 While downloading to a remote protocol target, gdb ignores packet
 625 errors in so far as it will continue to edownload with chunk N+1 even
 626 if chunk N was not correctly sent.  This causes gdb.base/remote.exp to
 627 take a painfully long time to run.  As a PS that test needs to be
 628 fixed so that it builds on 16 bit machines.
 629
 630 --
 631
 632 Add the cycle step command.
 633
 634 http://sourceware.cygnus.com/ml/gdb/2000-q1/msg00237.html
 635
 636 --
 637
 638                         Symbol Support
 639                         ==============
 640
 641 If / when GDB starts to support the debugging of multi-processor
 642 (rather than multi-thread) applications the symtab code will need to
 643 be updated a little so that several independant symbol tables are
 644 active at a given time.
 645
 646 The other interesting change is a clarification of the exact meaning
 647 of CORE_ADDR and that has had consequences for a few targets (that
 648 were abusing that data type).
 649
 650 --
 651
 652 Investiagate ways of reducing memory.
 653
 654 --
 655
 656 Investigate ways of improving load time.
 657
 658 --
 659
 660 Get the d10v to use POINTER_TO_ADDRESS and ADDRESS_TO_POINTER.
 661
 662 Consequence of recent symtab clarification.  No marks for figuring out
 663 who maintains the d10v.
 664
 665 --
 666
 667 Get the MIPS to correctly sign extend all address <-> pointer
 668 conversions.
 669
 670 Consequence of recent symtab clarification.  No marks for figuring out
 671 who maintains the MIPS.
 672
 673 --
 674
 675                 Architectural Changes: General
 676                 ==============================
 677
 678 These are harder than simple cleanups / fixes and, consequently
 679 involve more work.  Typically an Architectural Change will be broken
 680 down into a more digestible set of cleanups and fixes.
 681
 682 --
 683
 684 Cleanup software single step.
 685
 686 At present many targets implement software single step by directly
 687 blatting memory (see rs6000-tdep.c).  Those targets should register
 688 the applicable breakpoints using the breakpoint framework.  Perhaphs a
 689 new internal breakpoint class ``step'' is needed.
 690
 691 --
 692
 693 Replace READ_FP() with FRAME_HANDLE().
 694
 695 READ_FP() is a hangover from the days of the vax when the ABI really
 696 did have a frame pointer register.  Modern architectures typically
 697 construct a virtual frame-handle from the stack pointer and various
 698 other bits of string.
 699
 700 Unfortunatly GDB still treats this synthetic FP register as though it
 701 is real.  That in turn really confuses users (arm and ``print $fp'' VS
 702 ``info registers fp'').  The synthetic FP should be separated out of
 703 the true register set presented to the user.
 704
 705 --
 706
 707 Register Cache Cleanup (below from Andrew Cagney)
 708
 709 I would depict the current register architecture as something like:
 710
 711         High GDB  -->   Low GDB
 712             |              |
 713            \|/            \|/
 714             --- REG NR -----
 715                    |
 716     register + REGISTER_BYTE(reg_nr)
 717                    |
 718                   \|/
 719         -------------------------
 720         |  extern register[]    |
 721         -------------------------
 722
 723 where neither the high (valops.c et.al.) or low gdb (*-tdep.c) are
 724 really clear on what mechanisms they should be using to manipulate that
 725 buffer.   Further, much code assumes, dangerously, that registers are
 726 contigious.  Having got mips-tdep.c to support multiple ABIs, believe
 727 me, that is a bad assumption.  Finally, that register cache layout is
 728 determined by the current remote/local target and _not_ the less
 729 specific target ISA.  In fact, in many cases it is determined by the
 730 somewhat arbitrary layout of the [gG] packets!
 731
 732
 733 How I would like the register file to work is more like:
 734
 735
 736           High GDB
 737              |
 738             \|/
 739        pseudo reg-nr
 740              |
 741         map pseudo <->
 742         random cache
 743         bytes
 744              |
 745             \|/
 746         ------------
 747         | register |
 748         |  cache   |
 749         ------------
 750             /|\
 751              |
 752         map random cache
 753         bytes to target
 754         dependant i-face
 755             /|\
 756              |
 757        target dependant
 758         such as [gG] packet
 759         or ptrace buffer
 760
 761 The main objectives being:
 762
 763         o       a clear separation between the low
 764                 level target and the high level GDB
 765
 766         o       a mechanism that solves the general
 767                 problem of register aliases, overlaps
 768                 etc instead of treating them as optional
 769                 extras that can be wedged in as an after
 770                 thought (that is a reasonable description
 771                 of the current code).
 772
 773                 Identify then solve the hard case and the
 774                 rest just falls out.  GDB solved the easy
 775                 case and then tried to ignore the real
 776                 world :-)
 777
 778         o       a removal of the assumption that the
 779                 mapping between the register cache
 780                 and virtual registers is largely static.
 781                 If you flip the USR/SSR stack register
 782                 select bit in the status-register then
 783                 the corresponding stack registers should
 784                 reflect the change.
 785
 786         o       a mechanism that clearly separates the
 787                 gdb internal register cache from any
 788                 target (not architecture) dependant
 789                 specifics such as [gG] packets.
 790
 791 Of course, like anything, it sounds good in theory.  In reality, it
 792 would have to contend with many<->many relationships at both the
 793 virt<->cache and cache<->target level.  For instance:
 794
 795         virt<->cache
 796                 Modifying an mmx register may involve
 797                 scattering values across both FP and
 798                 mmpx specific parts of a buffer
 799
 800         cache<->target
 801                 When writing back a SP it may need to
 802                 both be written to both SP and USP.
 803
 804
 805 Hmm,
 806
 807 Rather than let this like the last time it was discussed, just slip, I'm
 808 first going to add this e-mail (+ references) to TODO.  I'd then like to
 809 sketch out a broad strategy I think could get us there.
 810
 811
 812 First thing I'd suggest is separating out the ``extern registers[]''
 813 code so that we can at least identify what is using it.  At present
 814 things are scattered across many files.  That way we can at least
 815 pretend that there is a cache instead of a global array :-)
 816
 817 I'd then suggest someone putting up a proposal for the pseudo-reg /
 818 high-level side interface so that code can be adopted to it.  For old
 819 code, initially a blanket rename of write_register_bytes() to
 820 deprecated_write_register_bytes() would help.
 821
 822 Following that would, finaly be the corresponding changes to the target.
 823
 824 --
 825
 826 Check that GDB can handle all BFD architectures (Andrew Cagney)
 827
 828 There should be a test that checks that BFD/GDB are in sync with
 829 regard to architecture changes.  Something like a test that first
 830 queries GDB for all supported architectures and then feeds each back
 831 to GDB..  Anyone interested in learning how to write tests?  :-)
 832
 833 --
 834
 835                 Architectural Change: Multi-arch et al.
 836                 =======================================
 837
 838 The long term objective is to remove all assumptions that there is a
 839 single target with a single address space with a single instruction
 840 set architecture and single application binary interface.
 841
 842 This is an ongoing effort.  The first milestone is to enable
 843 ``multi-arch'' where by all architectural decisions are made at
 844 runtime.
 845
 846 It should be noted that ``gdbarch'' is really ``gdbabi'' and
 847 ``gdbisa''.  Once things are multi-arched breaking that down correctly
 848 will become much easier.
 849
 850 --
 851
 852 GDBARCH cleanup (Andrew Cagney)
 853
 854 The non-generated parts of gdbarch.{sh,h,c} should be separated out
 855 into arch-utils.[hc].
 856
 857 Document that gdbarch_init_ftype could easily fail because it didn't
 858 identify an architecture.
 859
 860 --
 861
 862 Fix BELIEVE_PPC_PROMOTION.  Change it to BELIEVE_PPC_PROMOTION_P?
 863
 864 At present there is still #ifdef BELIEVE_PPC_PROMOTION code in the
 865 symtab file.
 866
 867 --
 868
 869 Fix ``set architecture <tab>''
 870
 871 This command should expand to a list of all supported architectures.
 872 At present ``info architecture'' needs to be used.  That is simply
 873 wrong.  It involves the use of add_set_enum_cmd().
 874
 875 --
 876
 877 Update ALPHA so that it uses ``struct frame_extra_info'' instead of
 878 EXTRA_FRAME_INFO.
 879
 880 This is a barrier to replacing mips_extra_func_info with something
 881 that works with multi-arch.
 882
 883 --
 884
 885 Multi-arch  mips_extra_func_info.
 886
 887 This first needs the alpha to be updated so that it uses ``struct
 888 frame_extra_info''.
 889
 890 --
 891
 892 Rationalize TARGET_SINGLE_FORMAT and TARGET_SINGLE_BIT et al.
 893
 894 Surely one of them is redundant.
 895
 896 --
 897
 898 Convert ALL architectures to MULTI-ARCH.
 899
 900 --
 901
 902 Select the initial multi-arch ISA / ABI based on --target or similar.
 903
 904 At present the default is based on what ever is first in the BFD
 905 archures table.  It should be determined based on the ``--target=...''
 906 name.
 907
 908 --
 909
 910 Truly multi-arch.
 911
 912 Enable the code to recognize --enable-targets=.... like BINUTILS does.
 913
 914 Can the tm.h and nm.h files be eliminated by multi-arch.
 915
 916 --
 917
 918         Architectural Change: MI, LIBGDB and scripting languages
 919         ========================================================
 920
 921 See also architectural changes related to the event loop.  LIBGDB
 922 can't be finished until there is a generic event loop being used by
 923 all targets.
 924
 925 The long term objective is it to be possible to integrate GDB into
 926 scripting languages.
 927
 928 --
 929
 930 Implement generic ``(gdb) commmand > file''
 931
 932 Once everything is going through ui_file it should be come fairly
 933 easy.
 934
 935 http://sourceware.cygnus.com/ml/gdb/2000-04/msg00104.html
 936
 937 --
 938
 939 Replace gdb_stdtarg with gdb_targout (and possibly gdb_targerr).
 940
 941 gdb_stdtarg is easily confused with gdb_stdarg.
 942
 943 --
 944
 945 Extra ui_file methods - dump.
 946
 947 These are for debugging / testing.  An aside is to set up a whitebox
 948 testsuite for key internals such as ui_file.
 949
 950 --
 951
 952 Eliminate error_begin().
 953
 954 With ui_file, there is no need for the statefull error_begin ()
 955 function.
 956
 957 --
 958
 959 Send normal output to gdb_stdout.
 960 Send error messages to gdb_stderror.
 961 Send debug and log output log gdb_stdlog.
 962
 963 GDB still contains many cases where (f)printf or printf_filtered () is
 964 used when it should be sending the messages to gdb_stderror or
 965 gdb_stdlog.  The thought of #defining printf to something has crossed
 966 peoples minds ;-)
 967
 968 --
 969
 970 Re-do GDB's output pager.
 971
 972 GDB's output pager still relies on people correctly using *_filtered
 973 for gdb_stdout and *_unfiltered for gdb_stdlog / gdb_stderr.
 974 Hopefully, with all normal output going to gdb_stdout, the pager can
 975 just look at the ui_file that the output is on and then use that to
 976 decide what to do about paging.  Sounds good in theory.
 977
 978 --
 979
 980 Check/cleanup MI documentation.
 981
 982 The list of commands specified in the documentation needs to be
 983 checked against the mi-cmds.c table in a mechanical way (so that they
 984 two can be kept up-to-date).
 985
 986 --
 987
 988 Convert MI into libgdb
 989
 990 MI provides a text interface into what should be many of the libgdb
 991 functions.  The implementation of those functions should be separated
 992 into the MI interface and the functions proper.  Those functions being
 993 moved to gdb/lib say.
 994
 995 --
 996
 997 Create libgdb.h
 998
 999 The first part can already be found in defs.h.
1000
1001 --
1002
1003 MI's input does not use buffering.
1004
1005 At present the MI interface reads raw characters of from an unbuffered
1006 FD.  This is to avoid several nasty buffer/race conditions.  That code
1007 should be changed so that it registers its self with the event loop
1008 (on the input FD) and then push commands up to MI as they arrive.
1009
1010 The serial code already does this.
1011
1012 --
1013
1014 Make MI interface accessable from existing CLI.
1015
1016 --
1017
1018 Add a breakpoint-edit command to MI.
1019
1020 It would be similar to MI's breakpoint create but would apply to an
1021 existing breakpoint.  It saves the need to delete/create breakpoints
1022 when ever they are changed.
1023
1024 --
1025
1026 Add directory path to MI breakpoint.
1027
1028 That way the GUI's task of finding the file within which the
1029 breakpoint was set is simplified.
1030
1031 --
1032
1033 Add a mechanism to reject certain expression classes to MI
1034
1035 There are situtations where you don't want GDB's expression
1036 parser/evaluator to perform inferior function calls or variable
1037 assignments.  A way of restricting the expression parser so that such
1038 operations are not accepted would be very helpful.
1039
1040 --
1041
1042 Remove sideffects from libgdb breakpoint create function.
1043
1044 The user can use the CLI to create a breakpoint with partial
1045 information - no file (gdb would use the file from the last
1046 breakpoint).
1047
1048 The libgdb interface currently affects that environment which can lead
1049 to confusion when a user is setting breakpoints via both the MI and
1050 the CLI.
1051
1052 This is also a good example of how getting the CLI ``right'' will be
1053 hard.
1054
1055 --
1056
1057 Move gdb_lasterr to ui_out?
1058
1059 The way GDB throws errors and records them needs a re-think.  ui_out
1060 handles the correct output well.  It doesn't resolve what to do with
1061 output / error-messages when things go wrong.
1062
1063 --
1064
1065                 Architectural Change: Async
1066                 ===========================
1067
1068 While GDB uses an event loop when prompting the user for input.  That
1069 event loop is not exploited by targets when they allow the target
1070 program to continue.  Typically targets still block in (target_wait())
1071 until the program again halts.
1072
1073 The closest a target comes to supporting full asynchronous mode are
1074 the remote targets ``async'' and ``extended-async''.
1075
1076 --
1077
1078 Asynchronous expression evaluator
1079
1080 Inferior function calls hang GDB.
1081
1082 --
1083
1084 Fix implementation of  ``target xxx''.
1085
1086 At present when the user specifies ``target xxxx'', the CLI maps that
1087 directly onto a target open method.  It is then assumed that the
1088 target open method should do all sorts of complicated things as this
1089 is the only chance it has.  Check how the various remote targets
1090 duplicate the target operations.  Check also how the various targets
1091 behave differently for purely arbitrary reasons.
1092
1093 What should happen is that ``target xxxx'' should call a generic
1094 ``target'' function and that should then co-ordinate the opening of
1095 ``xxxx''.  This becomes especially important when you're trying to
1096 open an asynchronous target that may need to perform background tasks
1097 as part of the ``attach'' phase.
1098
1099 Unfortunatly, due to limitations in the old/creaking command.h
1100 interface, that isn't possible.  The function being called isn't told
1101 of the ``xxx'' or any other context information.
1102
1103 Consequently a precursor to fixing ``target xxxx'' is to clean up the
1104 CLI code so that it passes to the callback function (attatched to a
1105 command) useful information such as the actual command and a context
1106 for that command.  Other changes such as making ``struct command''
1107 opaque may also help.
1108
1109 --
1110
1111 Make "target xxx" command interruptible.
1112
1113 As things become async this becomes possible.  A target would start
1114 the connect and then return control to the event loop.  A cntrl-c
1115 would notify the target that the operation is to be abandoned and the
1116 target code could respond.
1117
1118 --
1119
1120 Add a "suspend" subcommand of the "continue" command to suspend gdb
1121 while continuing  execution of the subprocess.  Useful when you are
1122 debugging servers and you want to dodge out and initiate a connection
1123 to a server running under gdb.
1124
1125 [hey async!!]
1126
1127 --
1128
1129                         Legacy Wish List
1130                         ================
1131
1132 This list is not up to date, and opinions vary about the importance or
1133 even desirability of some of the items.  If you do fix something, it
1134 always pays to check the below.
1135
1136 --
1137
1138 @c This does not work (yet if ever).  FIXME.
1139 @c @item --parse=@var{lang} @dots{}
1140 @c Configure the @value{GDBN} expression parser to parse the listed languages.
1141 @c @samp{all} configures @value{GDBN} for all supported languages.  To get a
1142 @c list of all supported languages, omit the argument.  Without this
1143 @c option, @value{GDBN} is configured to parse all supported languages.
1144
1145 --
1146
1147 START_INFERIOR_TRAPS_EXPECTED need never be defined to 2, since that
1148 is its default value.  Clean this up.
1149
1150 --
1151
1152 It should be possible to use symbols from shared libraries before we know
1153 exactly where the libraries will be loaded.  E.g. "b perror" before running
1154 the program.  This could maybe be done as an extension of the "breakpoint
1155 re-evaluation" after new symbols are loaded.
1156
1157 --
1158
1159 Make single_step() insert and remove breakpoints in one operation.
1160
1161 --
1162
1163 Speed up single stepping by avoiding extraneous ptrace calls.
1164
1165 --
1166
1167 Speed up single stepping by not inserting and removing breakpoints
1168 each time the inferior starts and stops.
1169
1170 Breakpoints should not be inserted and deleted all the time.  Only the
1171 one(s) there should be removed when we have to step over one.  Support
1172 breakpoints that don't have to be removed to step over them.
1173
1174 [this has resulted in numerous debates.  The issue isn't clear cut]
1175
1176 --
1177
1178 Provide "voodoo" debugging of core files.  This creates a zombie
1179 process as a child of the debugger, and loads it up with the data,
1180 stack, and regs of the core file.  This allows you to call functions
1181 in the executable, to manipulate the data in the core file.
1182
1183 [you wish]
1184
1185 --
1186
1187 GDB reopens the source file on every line, as you "next" through it.
1188
1189 [still true? I've a memory of this being fixed]
1190
1191 --
1192
1193 Perhaps "i source" should take an argument like that of "list".
1194
1195 --
1196
1197 Remove "at 0xnnnn" from the "b foo" response, if `print address off' and if
1198 it matches the source line indicated.
1199
1200 --
1201
1202 The prompt at end of screen should accept space as well as CR.
1203
1204 --
1205
1206 Backtrace should point out what the currently selected frame is, in
1207 its display, perhaps showing "@3 foo (bar, ...)" or ">3 foo (bar,
1208 ...)" rather than "#3 foo (bar, ...)".
1209
1210 --
1211
1212 "i program" should work for core files, and display more info, like what
1213 actually caused it to die.
1214
1215 --
1216
1217 "x/10i" should shorten the long name, if any, on subsequent lines.
1218
1219 --
1220
1221 "next" over a function that longjumps, never stops until next time you happen
1222 to get to that spot by accident.  E.g. "n" over execute_command which has
1223 an error.
1224
1225 --
1226
1227 "set zeroprint off", don't bother printing members of structs which
1228 are entirely zero.  Useful for those big structs with few useful
1229 members.
1230
1231 --
1232
1233 GDB does four ioctl's for every command, probably switching terminal modes
1234 to/from inferior or for readline or something.
1235
1236 --
1237
1238 terminal_ours versus terminal_inferior: cache state.  Switch should be a noop
1239 if the state is the same, too.
1240
1241 --
1242
1243 "i frame" shows wrong "arglist at" location, doesn't show where the args
1244 should be found, only their actual values.
1245
1246 --
1247
1248 There should be a way for "set" commands to validate the new setting
1249 before it takes effect.
1250
1251 --
1252
1253 "ena d" is ambiguous, why?  "ena delete" seems to think it is a command!
1254
1255 --
1256
1257 i line VAR produces "Line number not known for symbol ``var''.".  I
1258 thought we were stashing that info now!
1259
1260 --
1261
1262 We should be able to write to random files at hex offsets like adb.
1263
1264 --
1265
1266 [elena - delete this]
1267
1268 Handle add_file with separate text, data, and bss addresses.  Maybe
1269 handle separate addresses for each segment in the object file?
1270
1271 --
1272
1273 [Jimb/Elena delete this one]
1274
1275 Handle free_named_symtab to cope with multiply-loaded object files
1276 in a dynamic linking environment.  Should remember the last copy loaded,
1277 but not get too snowed if it finds references to the older copy.
1278
1279 --
1280
1281 [elena delete this also]
1282
1283 Remove all references to:
1284         text_offset
1285         data_offset
1286         text_data_start
1287         text_end
1288         exec_data_offset
1289         ...
1290 now that we have BFD.  All remaining are in machine dependent files.
1291
1292 --
1293
1294 Re-organize help categories into things that tend to fit on a screen
1295 and hang together.
1296
1297 --
1298
1299 Add in commands like ADB's for searching for patterns, etc.  We should
1300 be able to examine and patch raw unsymboled binaries as well in gdb as
1301 we can in adb.  (E.g. increase the timeout in /bin/login without source).
1302
1303 [actually, add ADB interface :-]
1304
1305 --
1306
1307 When doing "step" or "next", if a few lines of source are skipped between
1308 the previous line and the current one, print those lines, not just the
1309 last line of a multiline statement.
1310
1311 --
1312
1313 Handling of "&" address-of operator needs some serious overhaul
1314 for ANSI C and consistency on arrays and functions.
1315   For "float point[15];":
1316 ptype &point[4]   ==> Attempt to take address of non-lvalue.
1317   For "char *malloc();":
1318 ptype malloc      ==> "char *()";  should be same as
1319 ptype &malloc     ==> "char *(*)()"
1320 call printf ("%x\n", malloc)   ==> weird value, should be same as
1321 call printf ("%x\n", &malloc)  ==> correct value
1322
1323 --
1324
1325 Fix dbxread.c symbol reading in the presence of interrupts.  It
1326 currently leaves a cleanup to blow away the entire symbol table when a
1327 QUIT occurs.  (What's wrong with that? -kingdon, 28 Oct 1993).
1328
1329 [I suspect that the grype was that, on a slow system, you might want
1330 to cntrl-c and get just half the symbols and then load the rest later
1331 - scary to be honest]
1332
1333 --
1334
1335 Mipsread.c reads include files depth-first, because the dependencies
1336 in the psymtabs are way too inclusive (it seems to me).  Figure out what
1337 really depends on what, to avoid recursing 20 or 30 times while reading
1338 real symtabs.
1339
1340 --
1341
1342 value_add() should be subtracting the lower bound of arrays, if known,
1343 and possibly checking against the upper bound for error reporting.
1344
1345 --
1346
1347 When listing source lines, check for a preceding \n, to verify that
1348 the file hasn't changed out from under us.
1349
1350 [fixed by some other means I think.  That hack wouldn't actually work
1351 reliably - the file might move such that another \n appears. ]
1352
1353 --
1354
1355 Get all the remote systems (where the protocol allows it) to be able to
1356 stop the remote system when the GDB user types ^C (like remote.c
1357 does).  For ebmon, use ^Ak.
1358
1359 --
1360
1361 Possible feature: A version of the "disassemble" command which shows
1362 both source and assembly code ("set symbol-filename on" is a partial
1363 solution).
1364
1365 [has this been done?  It was certainly done for MI and GDBtk]
1366
1367 --
1368
1369 investigate "x/s 0" (right now stops early) (I think maybe GDB is
1370 using a 0 address for bad purposes internally).
1371
1372 --
1373
1374 Make "info path" and path_command work again (but independent of the
1375 environment either of gdb or that we'll pass to the inferior).
1376
1377 --
1378
1379 Make GDB understand the GCC feature for putting octal constants in
1380 enums.  Make it so overflow on an enum constant does not error_type
1381 the whole type.  Allow arbitrarily large enums with type attributes.
1382 Put all this stuff in the testsuite.
1383
1384 --
1385
1386 Make TYPE_CODE_ERROR with a non-zero TYPE_LENGTH more useful (print
1387 the value in hex; process type attributes).  Add this to the
1388 testsuite.  This way future compilers can add new types and old
1389 versions of GDB can do something halfway reasonable.
1390
1391 --
1392
1393 Fix mdebugread.c:parse_type to do fundamental types right (see
1394 rs6000_builtin_type in stabsread.c for what "right" is--the point is
1395 that the debug format fixes the sizes of these things and it shouldn't
1396 depend on stuff like TARGET_PTR_BIT and so on.  For mdebug, there seem
1397 to be separate bt* codes for 64 bit and 32 bit things, and GDB should
1398 be aware of that).  Also use a switch statement for clarity and speed.
1399
1400 --
1401
1402 Investigate adding symbols in target_load--some targets do, some
1403 don't.
1404
1405 --
1406
1407 Put dirname in psymtabs and change lookup*symtab to use dirname (so
1408 /foo/bar.c works whether compiled by cc /foo/bar.c, or cd /foo; cc
1409 bar.c).
1410
1411 --
1412
1413 Merge xcoffread.c and coffread.c.  Use breakpoint_re_set instead of
1414 fixup_breakpoints.
1415
1416 --
1417
1418 Make a watchpoint which contains a function call an error (it is
1419 broken now, making it work is probably not worth the effort).
1420
1421 --
1422
1423 New test case based on weird.exp but in which type numbers are not
1424 renumbered (thus multiply defining a type).  This currently causes an
1425 infinite loop on "p v_comb".
1426
1427 --
1428
1429 [Hey! Hint Hint Delete Delete!!!]
1430
1431 Fix 386 floating point so that floating point registers are real
1432 registers (but code can deal at run-time if they are missing, like
1433 mips and 68k).  This would clean up "info float" and related stuff.
1434
1435 --
1436
1437 gcc -g -c enummask.c then gdb enummask.o, then "p v".  GDB complains
1438 about not being able to access memory location 0.
1439
1440 -------------------- enummask.c
1441 enum mask
1442 {
1443   ANIMAL = 0,
1444   VEGETABLE = 1,
1445   MINERAL = 2,
1446   BASIC_CATEGORY = 3,
1447
1448   WHITE = 0,
1449   BLUE = 4,
1450   GREEN = 8,
1451   BLACK = 0xc,
1452   COLOR = 0xc,
1453
1454   ALIVE = 0x10,
1455
1456   LARGE = 0x20
1457 } v;
1458
1459 --
1460
1461 If try to modify value in file with "set write off" should give
1462 appropriate error not "cannot access memory at address 0x65e0".
1463
1464 --
1465
1466 Allow core file without exec file on RS/6000.
1467
1468 --
1469
1470 Make sure "shell" with no arguments works right on DOS.
1471
1472 --
1473
1474 Make gdb.ini (as well as .gdbinit) be checked on all platforms, so
1475 the same directory can be NFS-mounted on unix or DOS, and work the
1476 same way.
1477
1478 --
1479
1480 [Is this another delete???]
1481
1482 Get SECT_OFF_TEXT stuff out of objfile_relocate (might be needed to
1483 get RS/6000 to work right, might not be immediately relevant).
1484
1485 --
1486
1487 Work out some kind of way to allow running the inferior to be done as
1488 a sub-execution of, eg. breakpoint command lists.  Currently running
1489 the inferior interupts any command list execution.  This would require
1490 some rewriting of wait_for_inferior & friends, and hence should
1491 probably be done in concert with the above.
1492
1493 --
1494
1495 Add function arguments to gdb user defined functions.
1496
1497 --
1498
1499 Add convenience variables that refer to exec file, symbol file,
1500 selected frame source file, selected frame function, selected frame
1501 line number, etc.
1502
1503 --
1504
1505 Modify the handling of symbols grouped through BINCL/EINCL stabs to
1506 allocate a partial symtab for each BINCL/EINCL grouping.  This will
1507 seriously decrease the size of inter-psymtab dependencies and hence
1508 lessen the amount that needs to be read in when a new source file is
1509 accessed.
1510
1511 --
1512
1513 Add a command for searching memory, a la adb.  It specifies size,
1514 mask, value, start address.  ADB searches until it finds it or hits
1515 an error (or is interrupted).
1516
1517 --
1518
1519 Remove the range and type checking code and documentation, if not
1520 going to implement.
1521
1522 # Local Variables:
1523 # mode: text
1524 # End: