| 1 | |
| 2 | gdb bug list |
| 3 | John Gilmore, gnu@cygnus.com |
| 4 | |
| 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. |
| 7 | |
| 8 | |
| 9 | It should be possible to use symbols from shared libraries before we know |
| 10 | exactly where the libraries will be loaded. E.g. "b perror" before running |
| 11 | the program. This could maybe be done as an extension of the "breakpoint |
| 12 | re-evaluation" after new symbols are loaded. |
| 13 | |
| 14 | Make single_step() insert and remove breakpoints in one operation. |
| 15 | |
| 16 | Speed up single stepping by avoiding extraneous ptrace calls. |
| 17 | |
| 18 | Speed up single stepping by not inserting and removing breakpoints |
| 19 | each time the inferior starts and stops. |
| 20 | |
| 21 | Breakpoints should not be inserted and deleted all the time. Only the |
| 22 | one(s) there should be removed when we have to step over one. Support |
| 23 | breakpoints that don't have to be removed to step over them. |
| 24 | |
| 25 | Speed up watchpoints by not single-stepping them, but do something |
| 26 | faster like single-line execution. Speed them up tremendously on |
| 27 | machines that have watchpoint registers. |
| 28 | |
| 29 | Update gdbint.texinfo to include doc on the directory structure and |
| 30 | the various tricks of building gdb. |
| 31 | |
| 32 | Do a tutorial in gdb.texinfo on how to do simple things in gdb. |
| 33 | E.g. how to set a breakpoint that just prints something and continues. |
| 34 | How to break on aborts. Etc. |
| 35 | |
| 36 | Provide "voodoo" debugging of core files. This creates a zombie |
| 37 | process as a child of the debugger, and loads it up with the data, |
| 38 | stack, and regs of the core file. This allows you to call functions |
| 39 | in the executable, to manipulate the data in the core file. |
| 40 | |
| 41 | GDB reopens the source file on every line, as you "next" through it. |
| 42 | |
| 43 | Referencing the vtbl member of a struct doesn't work. It prints OK |
| 44 | if you print the struct, but it gets 0 if you try to deref it. |
| 45 | |
| 46 | Persistent command history: A feature where you could save off a list |
| 47 | of the commands you did, so you can edit it into something that will bring |
| 48 | the target to the same place every time you source it. |
| 49 | This would also be useful for automated fast watchpointing; if you go |
| 50 | past the place where it watchpoints, you just start it over again and |
| 51 | do it more carefully. |
| 52 | |
| 53 | Deal with the SunOS 4.0 and 4.1.1 ptrace bug that loses the registers if |
| 54 | the stack is paged out. |
| 55 | |
| 56 | Finish the C++ exception handling stub routines. Lint points them out |
| 57 | as unused statics functions. |
| 58 | |
| 59 | Perhaps "i source" should take an argument like that of "list". |
| 60 | |
| 61 | See if coredep.c's fetch_core_registers can be used on more machines. |
| 62 | E.g. MIPS (mips-xdep.c). |
| 63 | |
| 64 | unpack_double() does not handle IEEE float on the target unless the host |
| 65 | is also IEEE. Death on a vax. |
| 66 | |
| 67 | Test cross-debugging Unix-to-Unix. |
| 68 | |
| 69 | Integrate the RAPP remote debugging protocol. |
| 70 | `rapp' runs under unix and acts as a remote stub (like rem-multi.shar |
| 71 | distributed with GDB version 3). Currently it just works over UDP |
| 72 | (network), not over a serial line. To get it running |
| 73 | * Compile GDB on the host machine as usual |
| 74 | * Compile rapp on the target machine, giving for both host and target |
| 75 | the type of the target machine |
| 76 | * Install "gdb" in /etc/services on both machines. |
| 77 | |
| 78 | Set up interface between GDB and INFO so that you can hop into interactive |
| 79 | INFO and back out again. When running under Emacs, should use Emacs |
| 80 | info, else fork the info program. Installation of GDB should install |
| 81 | its texinfo files into the info tree automagically, including the readline |
| 82 | texinfo files. |
| 83 | |
| 84 | "help address" ought to find the "help set print address" entry. |
| 85 | |
| 86 | Remove the VTBL internal guts from printouts of C++ structs, unless |
| 87 | vtblprint is set. |
| 88 | |
| 89 | Remove "at 0xnnnn" from the "b foo" response, if `print address off' and if |
| 90 | it matches the source line indicated. |
| 91 | |
| 92 | The prompt at end of screen should accept space as well as CR. |
| 93 | |
| 94 | "List" should put you into a pseudo-"more" where you can hit space |
| 95 | to get more, forever to eof. |
| 96 | |
| 97 | Check STORE_RETURN_VALUE on all architectures. Check near it in tm-sparc.h |
| 98 | for other bogosities. |
| 99 | |
| 100 | Check for storage leaks in GDB, I'm sure there are a lot! |
| 101 | |
| 102 | vtblprint of a vtbl should demangle the names it's printing. |
| 103 | |
| 104 | Backtrace should point out what the currently selected frame is, in its |
| 105 | display, perhaps showing "@3 foo (bar, ...)" rather than "#3 foo (bar, ...)". |
| 106 | |
| 107 | "i program" should work for core files, and display more info, like what |
| 108 | actually caused it to die. |
| 109 | |
| 110 | Hitting ^Z to an inferior doesn't work right, it takes several continues |
| 111 | to make it actually go. |
| 112 | |
| 113 | "x/10i" should shorten the long name, if any, on subsequent lines. |
| 114 | |
| 115 | Check through the code for FIXME comments and fix them. dbxread.c, |
| 116 | blockframe.c, and plenty more. |
| 117 | |
| 118 | "next" over a function that longjumps, never stops until next time you happen |
| 119 | to get to that spot by accident. E.g. "n" over execute_command which has |
| 120 | an error. |
| 121 | |
| 122 | Watchpoints seem not entirely reliable, though they haven't failed me recently. |
| 123 | |
| 124 | "set zeroprint off", don't bother printing members of structs which are entirely |
| 125 | zero. Useful for those big structs with few useful members. |
| 126 | |
| 127 | GDB does four ioctl's for every command, probably switching terminal modes |
| 128 | to/from inferior or for readline or something. |
| 129 | |
| 130 | terminal_ours versus terminal_inferior: cache state. Switch should be a noop |
| 131 | if the state is the same, too. |
| 132 | |
| 133 | ptype $i6 = void??! |
| 134 | |
| 135 | Clean up invalid_float handling so gdb doesn't coredump when it tries to |
| 136 | access a NaN. While this might work on SPARC, other machines are not |
| 137 | configured right. |
| 138 | |
| 139 | "b value_at ; commands ; continue ; end" stops EVERY OTHER TIME! |
| 140 | Then once you enter a command, it does the command, runs two more |
| 141 | times, and then stops again! Bizarre... (This behaviour has been |
| 142 | modified, but it is not yet 100% predictable when e.g. the commands |
| 143 | call functions in the child, and while there, the child is interrupted |
| 144 | with a signal, or hits a breakpoint.) |
| 145 | |
| 146 | help completion, help history should work. |
| 147 | |
| 148 | Check that we can handle stack trace through varargs AND alloca in same |
| 149 | function, on 29K. |
| 150 | |
| 151 | wait_for_inferior loops forever if wait() gives it an error. |
| 152 | |
| 153 | "i frame" shows wrong "arglist at" location, doesn't show where the args |
| 154 | should be found, only their actual values. |
| 155 | |
| 156 | Symbolic display of addrs, (& disassembly prefixes), don't show static |
| 157 | fns, e.g. enable_command in gdb. |
| 158 | |
| 159 | 'ptype yylval' ==> "union YYSTYPE { ..... }". However, it is not a |
| 160 | union YYSTYPE, but is simply a YYSTYPE, which is a typedef for an |
| 161 | unnamed union. |
| 162 | |
| 163 | There should be a way for "set" commands to validate the new setting |
| 164 | before it takes effect. |
| 165 | |
| 166 | The "display" command should become the "always" command, e.g. |
| 167 | "always print XXX" |
| 168 | "always p/xxx XXX" |
| 169 | "always echo foo" |
| 170 | "always call XXX" |
| 171 | "always x/i $pc", etc. |
| 172 | |
| 173 | A mess of floating point opcodes are missing from sparc-opcode.h. |
| 174 | Also, a little program should test the table for bits that are |
| 175 | overspecified or underspecified. E.g. if the must-be-ones bits |
| 176 | and the must-be-zeroes bits leave some fields unexamined, and the format |
| 177 | string leaves them unprinted, then point this out. If multiple |
| 178 | non-alias patterns match, point this out too. Finally, there should |
| 179 | be a sparc-optest.s file that tries each pattern out. This file |
| 180 | should end up coming back the same (modulo transformation comments) |
| 181 | if fed to "gas" then the .o is fed to gdb for disassembly. |
| 182 | |
| 183 | Eliminate all the core_file_command's in all the xdep files. |
| 184 | Eliminate separate declarations of registers[] everywhere. |
| 185 | |
| 186 | "ena d" is ambiguous, why? "ena delete" seems to think it is a command! |
| 187 | |
| 188 | Line numbers are off in some spots. In proceed() at 1st "oneproc = 1", |
| 189 | it seems to run that statement, but it doesn't actually. |
| 190 | |
| 191 | Perhaps move the tdep and xdep files, and the tm and xm files, into a config |
| 192 | subdirectory. If not, at least straighten out their names so that |
| 193 | they all start with the machine name. |
| 194 | |
| 195 | inferior_status should include stop_print_frame. It won't need to be |
| 196 | reset in wait_for_inferior after bpstat_stop_status call, then. |
| 197 | |
| 198 | i line VAR produces "Line number not known for symbol ``var''.". I |
| 199 | thought we were stashing that info now! |
| 200 | |
| 201 | Make sure we can handle executables with no symbol info, e.g. /bin/csh. |
| 202 | |
| 203 | We should be able to write to random files at hex offsets like adb. |
| 204 | |
| 205 | Make "target xxx" command interruptible. |
| 206 | |
| 207 | Handle add_file with separate text, data, and bss addresses. Maybe |
| 208 | handle separate addresses for each segment in the object file? |
| 209 | |
| 210 | Handle free_named_symtab to cope with multiply-loaded object files |
| 211 | in a dynamic linking environment. Should remember the last copy loaded, |
| 212 | but not get too snowed if it finds references to the older copy. |
| 213 | |
| 214 | The original BFD core dump reading routine would itself coredump when fed |
| 215 | a garbage file as a core file. Does the current one? |
| 216 | |
| 217 | Generalize and Standardize the RPC interface to a target program, |
| 218 | improve it beyond the "ptrace" interface, and see if it can become a |
| 219 | standard for remote debugging. |
| 220 | |
| 221 | Remove all references to: |
| 222 | text_offset |
| 223 | data_offset |
| 224 | text_data_start |
| 225 | text_end |
| 226 | exec_data_offset |
| 227 | ... |
| 228 | now that we have BFD. All remaining are in machine dependent files. |
| 229 | |
| 230 | When quitting with a running program, if a core file was previously |
| 231 | examined, you get "Couldn't read float regs from core file"...if |
| 232 | indeed it can't. generic_mourn_inferior... |
| 233 | |
| 234 | Check signal argument to remote proceed's and error if set. |
| 235 | |
| 236 | Sort help and info output. |
| 237 | |
| 238 | Re-organize help categories into things that tend to fit on a screen |
| 239 | and hang together. |
| 240 | |
| 241 | renote-nindy.c handles interrupts poorly; it error()s out of badly |
| 242 | chosen places, e.g. leaving current_frame zero, which causes core dumps |
| 243 | on the next command. |
| 244 | |
| 245 | Add in commands like ADB's for searching for patterns, etc. We should |
| 246 | be able to examine and patch raw unsymboled binaries as well in gdb as |
| 247 | we can in adb. (E.g. increase the timeout in /bin/login without source). |
| 248 | |
| 249 | Those xdep files that call register_addr without defining it are |
| 250 | probably simply broken. When reconfiguring this part of gdb, I could |
| 251 | only make guesses about how to redo some of those files, and I |
| 252 | probably guessed wrong, or left them "for later" when I have a |
| 253 | machine that can attempt to build them. |
| 254 | |
| 255 | Use the complain() mechanism for handling all the error() calls in dbxread.c, |
| 256 | and in similar situations in coffread.c and mipsread.c. |
| 257 | |
| 258 | When doing "step" or "next", if a few lines of source are skipped between |
| 259 | the previous line and the current one, print those lines, not just the |
| 260 | last line of a multiline statement. |
| 261 | |
| 262 | When searching for C++ superclasses in value_cast in valops.c, we must |
| 263 | not search the "fields", only the "superclasses". There might be a |
| 264 | struct with a field name that matches the superclass name. This can |
| 265 | happen when the struct was defined before the superclass (before the |
| 266 | name became a typedef). |
| 267 | |
| 268 | Handling of "&" address-of operator needs some serious overhaul |
| 269 | for ANSI C and consistency on arrays and functions. |
| 270 | For "float point[15];": |
| 271 | ptype &point[4] ==> Attempt to take address of non-lvalue. |
| 272 | For "char *malloc();": |
| 273 | ptype malloc ==> "char *()"; should be same as |
| 274 | ptype &malloc ==> "char *(*)()" |
| 275 | call printf ("%x\n", malloc) ==> wierd value, should be same as |
| 276 | call printf ("%x\n", &malloc) ==> correct value |
| 277 | |
| 278 | Fix dbxread.c symbol reading in the presence of interrupts. It currently |
| 279 | leaves a cleanup to blow away the entire symbol table when a QUIT occurs. |
| 280 | |
| 281 | Mipsread.c reads include files depth-first, because the dependencies |
| 282 | in the psymtabs are way too inclusive (it seems to me). Figure out what |
| 283 | really depends on what, to avoid recursing 20 or 30 times while reading |
| 284 | real symtabs. |
| 285 | |
| 286 | value_add() should be subtracting the lower bound of arrays, if known, |
| 287 | and possibly checking against the upper bound for error reporting. |
| 288 | |
| 289 | mipsread.c symbol table allocation and deallocation should be checked. |
| 290 | My suspicion is that it's full of memory leaks. |
| 291 | |
| 292 | SunOS should have a target_lookup_symbol() for common'd things allocated |
| 293 | by the shared library linker ld.so. |
| 294 | |
| 295 | When listing source lines, check for a preceding \n, to verify that |
| 296 | the file hasn't changed out from under us. |
| 297 | |
| 298 | When listing source lines, eat leading whitespace corresponding to the |
| 299 | line-number prefix we print. This avoids long lines wrapping. |
| 300 | |
| 301 | mipsread.c needs to check for old symtabs and psymtabs for the same |
| 302 | files, the way it happens for dbxread.c and coffread.c, for VxWorks |
| 303 | incremental symbol table reloading. |
| 304 | |
| 305 | When attached to a non-child process, ^C or other signals are not |
| 306 | propagated to the child. Do this in the GDB signal handler, using |
| 307 | target_kill(). AMD version: ^C should do ^Ak to stop ebmon. |
| 308 | |