3 John Gilmore, gnu@cygnus.com
5 This bug list is probably not up to date or accurate, but it reflects
6 some known bugs in gdb, if you are into bug-hunting.
9 Update the TODO list with all the lists of gdb bugs lying around on paper.
11 It should be possible to use symbols from shared libraries before we know
12 exactly where the libraries will be loaded. E.g. "b perror" before running
13 the program. This could maybe be done as an extension of the "breakpoint
14 re-evaluation" after new symbols are loaded.
16 Make single_step() insert and remove breakpoints in one operation.
18 Speed up single stepping by avoiding extraneous ptrace calls.
20 Speed up single stepping by not inserting and removing breakpoints
21 each time the inferior starts and stops.
23 Breakpoints should not be inserted and deleted all the time. Only the
24 one(s) there should be removed when we have to step over one. Support
25 breakpoints that don't have to be removed to step over them.
27 Speed up watchpoints by not single-stepping them, but do something
28 faster like single-line execution. Speed them up tremendously on
29 machines that have watchpoint registers.
31 Update gdb-int.texinfo to include doc on the directory structure and
32 the various tricks of building gdb.
34 Do a tutorial in gdb.texinfo on how to do simple things in gdb.
35 E.g. how to set a breakpoint that just prints something and continues.
36 How to break on aborts. Etc.
38 Provide "voodoo" debugging of core files. This creates a zombie
39 process as a child of the debugger, and loads it up with the data,
40 stack, and regs of the core file. This allows you to call functions
41 in the executable, to manipulate the data in the core file.
43 GDB reopens the source file on every line, as you "next" through it.
45 Referencing the vtbl member of a struct doesn't work. It prints OK
46 if you print the struct, but it gets 0 if you try to deref it.
48 Persistent command history: A feature where you could save off a list
49 of the commands you did, so you can edit it into something that will bring
50 the target to the same place every time you source it.
51 This would also be useful for automated fast watchpointing; if you go
52 past the place where it watchpoints, you just start it over again and
55 Deal with the SunOS 4.0 and 4.1.1 ptrace bug that loses the registers if
56 the stack is paged out.
58 Finish the C++ exception handling stub routines. Lint points them out
59 as unused statics functions.
61 Perhaps "i source" should take an argument like that of "list".
63 See if coredep.c's fetch_core_registers can be used on more machines.
64 E.g. MIPS (mips-xdep.c).
66 unpack_double() does not handle IEEE float on the target unless the host
67 is also IEEE. Death on a vax.
69 Test cross-debugging Unix-to-Unix.
71 Check the RAPP remote protocol. What is it? It's in Makefile.in
72 and one ChangeLog entry.
74 Set up interface between GDB and INFO so that you can hop into interactive
75 INFO and back out again. When running under Emacs, should use Emacs
76 info, else fork the info program. Installation of GDB should install
77 its texinfo files into the info tree automagically, including the readline
80 "help address" ought to find the "help set print address" entry.
82 Remove the VTBL internal guts from printouts of C++ structs, unless
85 Remove "at 0xnnnn" from the "b foo" response, if !addressprint and if
86 it matches the source line indicated.
88 The prompt at end of screen should accept space as well as CR.
90 "List" should put you into a pseudo-"more" where you can hit space
91 to get more, forever to eof.
93 Check STORE_RETURN_VALUE on all architectures. Check near it in tm-sparc.h
96 Check for storage leaks in GDB, I'm sure there are a lot!
98 vtblprint of a vtbl should demangle the names it's printing.
100 Backtrace should point out what the currently selected frame is, in its
101 display, perhaps showing "@3 foo (bar, ...)" rather than "#3 foo (bar, ...)".
103 "i program" should work for core files, and display more info, like what
104 actually caused it to die.
106 Hitting ^Z to an inferior doesn't work right, it takes several continues
107 to make it actually go.
109 "x/10i" should shorten the long name, if any, on subsequent lines.
111 Check through the code for FIXME comments and fix them. dbxread.c,
112 blockframe.c, and plenty more.
114 "next" over a function that longjumps, never stops until next time you happen
115 to get to that spot by accident. E.g. "n" over execute_command which has
118 Watchpoints seem not entirely reliable, though they haven't failed me recently.
120 "set zeroprint off", don't bother printing members of structs which are entirely
121 zero. Useful for those big structs with few useful members.
123 GDB does four ioctl's for every command, probably switching terminal modes
124 to/from inferior or for readline or something.
126 terminal_ours versus terminal_inferior: cache state. Switch should be a noop
127 if the state is the same, too.
131 Clean up invalid_float handling so gdb doesn't coredump when it tries to
132 access a NaN. While this might work on SPARC, other machines are not
135 "b value_at ; commands ; continue ; end" stops EVERY OTHER TIME!
136 Then once you enter a command, it does the command, runs two more
137 times, and then stops again! Bizarre... (This behaviour has been
138 modified, but it is not yet 100% predictable when e.g. the commands
139 call functions in the child, and while there, the child is interrupted
140 with a signal, or hits a breakpoint.)
142 Symbol completion with TAB does not unmangle names!
144 help completion, help history should work.
146 Symbol completion doesn't handle e.g. W::f. (symtab.c,
147 make_symbol_completion_list).
149 Check that we can handle stack trace through varargs AND alloca in same
152 wait_for_inferior loops forever if wait() gives it an error.
154 "i frame" shows wrong "arglist at" location, doesn't show where the args
155 should be found, only their actual values.
157 Symbolic display of addrs, (& disassembly prefixes), don't show static
158 fns, e.g. enable_command in gdb.
160 'ptype yylval' ==> "union YYSTYPE { ..... }". However, it is not a
161 union YYSTYPE, but is simply a YYSTYPE, which is a typedef for an
164 There should be a way for "set" commands to validate the new setting
165 before it takes effect.
167 The "display" command should become the "always" command, e.g.
172 "always x/i $pc", etc.
174 A mess of floating point opcodes are missing from sparc-opcode.h.
175 Also, a little program should test the table for bits that are
176 overspecified or underspecified. E.g. if the must-be-ones bits
177 and the must-be-zeroes bits leave some fields unexamined, and the format
178 string leaves them unprinted, then point this out. If multiple
179 non-alias patterns match, point this out too. Finally, there should
180 be a sparc-optest.s file that tries each pattern out. This file
181 should end up coming back the same (modulo transformation comments)
182 if fed to "gas" then the .o is fed to gdb for disassembly.
184 Merge the xxx-opcode.h files with gas again...
186 Eliminate all the core_file_command's in all the xdep files.
187 Eliminate separate declarations of registers[] everywhere.
189 "ena d" is ambiguous, why? "ena delete" seems to think it is a command!
191 Line numbers are off in some spots. In proceed() at 1st "oneproc = 1",
192 it seems to run that statement, but it doesn't actually.
194 Perhaps move the tdep and xdep files, and the tm and xm files, into a config
195 subdirectory. If not, at least straighten out their names so that
196 they all start with the machine name.
198 inferior_status should include stop_print_frame. It won't need to be
199 reset in wait_for_inferior after bpstat_stop_status call, then.
201 i line VAR produces "Line number not known for symbol ``var''.". I
202 thought we were stashing that info now!
204 Make sure we can handle executables with no symbol info, e.g. /bin/csh.
206 We should be able to write to executables that aren't running.
208 We should be able to write to random files at hex offsets like adb.
210 Make "target xxx" command interruptible.
212 Handle add_file with separate text, data, and bss addresses. Maybe
213 handle separate addresses for each segment in the object file?
215 Handle free_named_symtab to cope with multiply-loaded object files
216 in a dynamic linking environment. Should remember the last copy loaded,
217 but not get too snowed if it finds references to the older copy.
219 The original BFD core dump reading routine would itself coredump when fed
220 a garbage file as a core file. Does the current one?
222 Generalize and Standardize the RPC interface to a target program,
223 improve it beyond the "ptrace" interface, and see if it can become a
224 standard for remote debugging.
226 Remove all references to:
233 now that we have BFD. All remaining are in machine dependent files.
235 When quitting with a running program, if a core file was previously
236 examined, you get "Couldn't read float regs from core file"...if
237 indeed it can't. generic_mourn_inferior...
239 Check signal argument to remote proceed's and error if set.
241 Sort help and info output.
243 Re-organize help categories into things that tend to fit on a screen
246 renote-nindy.c handles interrupts poorly; it error()s out of badly
247 chosen places, e.g. leaving current_frame zero, which causes core dumps
250 Add in commands like ADB's for searching for patterns, etc. We should
251 be able to examine and patch raw unsymboled binaries as well in gdb as
252 we can in adb. (E.g. increase the timeout in /bin/login without source).
254 Those xdep files that call register_addr without defining it are
255 probably simply broken. When reconfiguring this part of gdb, I could
256 only make guesses about how to redo some of those files, and I
257 probably guessed wrong, or left them "for later" when I have a
258 machine that can attempt to build them.
260 Use the complain() mechanism for handling all the error() calls in dbxread.c,
261 and in similar situations in coffread.c and mipsread.c.
263 When doing "step" or "next", if a few lines of source are skipped between
264 the previous line and the current one, print those lines, not just the
265 last line of a multiline statement.
267 When searching for C++ superclasses in value_cast in valops.c, we must
268 not search the "fields", only the "superclasses". There might be a
269 struct with a field name that matches the superclass name. This can
270 happen when the struct was defined before the superclass (before the
271 name became a typedef).
273 Handling of "&" address-of operator needs some serious overhaul
274 for ANSI C and consistency on arrays and functions.
275 For "float point[15];":
276 ptype &point[4] ==> Attempt to take address of non-lvalue.
277 p &point[4] ==> Dereferences point[4] rather than giving you point+4.
278 For "char *malloc();":
279 ptype malloc ==> "char *()"; should be same as
280 ptype &malloc ==> "char *(*)()"
281 call printf ("%x\n", malloc) ==> wierd value, should be same as
282 call printf ("%x\n", &malloc) ==> correct value
284 Fix dbxread.c symbol reading in the presence of interrupts. It currently
285 leaves a cleanup to blow away the entire symbol table when a QUIT occurs.
287 Mipsread.c reads include files depth-first, because the dependencies
288 in the psymtabs are way too inclusive (it seems to me). Figure out what
289 really depends on what, to avoid recursing 20 or 30 times while reading
292 value_add() should be subtracting the lower bound of arrays, if known,
293 and possibly checking against the upper bound for error reporting.
295 mipsread.c symbol table allocation and deallocation should be checked.
296 My suspicion is that it's full of memory leaks.
298 SunOS should have a target_lookup_symbol() for common'd things allocated
299 by the shared library linker ld.so.
301 When listing source lines, check for a preceding \n, to verify that
302 the file hasn't changed out from under us.
304 When listing source lines, eat leading whitespace corresponding to the
305 line-number prefix we print. This avoids long lines wrapping.
307 mipsread.c needs to check for old symtabs and psymtabs for the same
308 files, the way it happens for dbxread.c and coffread.c, for VxWorks
309 incremental symbol table reloading.
311 When attached to a non-child process, ^C or other signals are not
312 propagated to the child. Do this in the GDB signal handler, using
313 target_kill(). AMD version: ^C should do ^Ak to stop ebmon.