Commit | Line | Data |
---|---|---|
4a44dc1a JK |
1 | Contents |
2 | -------- | |
dd3b648e | 3 | |
4a44dc1a JK |
4 | If you find inaccuracies in this list, please send mail to |
5 | bug-gdb@prep.ai.mit.edu. | |
dd3b648e | 6 | |
4a44dc1a JK |
7 | * Things to do for Mach. |
8 | * General to do list. | |
dd3b648e | 9 | |
4a44dc1a JK |
10 | Things to do for Mach |
11 | --------------------- | |
12 | ||
13 | This section is up to date as of 28 Oct 1993. | |
14 | ||
15 | All my attempted compilation was on douglas.gnu.ai.mit.edu. | |
16 | ||
17 | 0. Get it to compile and run again, especially for non-threaded | |
18 | programs (some of the following are sub-tasks for this). | |
19 | ||
20 | 1. attach_command still contains a call to wait_for_inferior which is | |
21 | wrong for Mach. Need to figure out a way to push this functionality | |
22 | into target_attach (perhaps by having target_attach, for non-Mach | |
23 | targets, call a function which does what is now in attach_command). | |
24 | ||
25 | 2. jtv's port contains an #ifdef which skips the call to | |
26 | insert_step_breakpoint right after SOLIB_CREATE_INFERIOR_HOOK, but | |
27 | goes ahead and calls insert_breakpoints. I don't understand this--the | |
28 | comment would appear to apply to all breakpoints. Perhaps it is an | |
29 | artifact from a previous version of the Mach port? (BTW, the modern | |
30 | equivalent is the call to proceed from m3_create_inferior; proceed | |
31 | inserts breakpoints). | |
32 | ||
33 | 3. Get the thread stuff to use the new generic thread code (enhancing | |
34 | the generic thread code to include any missing features). This is | |
35 | necessary to make thread-specific breakpoints work again. If someone | |
36 | wants to try to patch up the old Mach threads code, need to deal with | |
37 | the hooks for PREPARE_TO_PROCEED and ATTACH_TO_THREAD, which I haven't | |
38 | merged--can these go in target_resume()? | |
39 | ||
40 | 4. BFD problem--"Undefined symbol _aout_32_swap_exec_header_in". | |
41 | ||
42 | 5. The linker complains about mfree and so on being multiply defined. | |
43 | As one version is in a library, this seems like a linker bug. But I | |
44 | think changing MMALLOC_LIB to MMALLOC in the .mh files is all that is | |
45 | necessary to fix it. | |
46 | ||
47 | 6. i386_mach3_float_info and register_addr were undefined in the | |
48 | link. I haven't investigated, but probably just another easy | |
49 | configuration thing or something. | |
50 | ||
51 | 7. I couldn't find mach_port_t in any of the headers in | |
52 | /usr/include/*.h or /usr/include/mach/*.h (I think those are the two | |
53 | places I tried). Typedeffing it to void * in nm-m3.h seemed to work, | |
54 | but of course that's hardly an elegant solution. | |
55 | ||
56 | 8. Implement the features which CMU gdb has which the main GDB does | |
57 | not. This could be done by getting paperwork from CMU and merging | |
58 | their changes, or by reimplementing them. | |
59 | ||
60 | General To Do List | |
61 | ------------------ | |
62 | ||
63 | This to do list is probably not up to date, and opinions may vary | |
64 | about the importance or even desirability of some of the items. | |
dd3b648e | 65 | |
dd3b648e RP |
66 | It should be possible to use symbols from shared libraries before we know |
67 | exactly where the libraries will be loaded. E.g. "b perror" before running | |
68 | the program. This could maybe be done as an extension of the "breakpoint | |
69 | re-evaluation" after new symbols are loaded. | |
70 | ||
71 | Make single_step() insert and remove breakpoints in one operation. | |
72 | ||
73 | Speed up single stepping by avoiding extraneous ptrace calls. | |
74 | ||
75 | Speed up single stepping by not inserting and removing breakpoints | |
76 | each time the inferior starts and stops. | |
77 | ||
adf2bb58 JG |
78 | Breakpoints should not be inserted and deleted all the time. Only the |
79 | one(s) there should be removed when we have to step over one. Support | |
80 | breakpoints that don't have to be removed to step over them. | |
81 | ||
e05fe481 JK |
82 | Speed up watchpoints by using debug registers, page table diddling (on |
83 | SunOS4, can call mprotect() in the inferior; on other machines can do | |
84 | something simpler), etc. | |
dd3b648e | 85 | |
b1dcd6ac | 86 | Update gdbint.texinfo to include doc on the directory structure and |
dd3b648e RP |
87 | the various tricks of building gdb. |
88 | ||
89 | Do a tutorial in gdb.texinfo on how to do simple things in gdb. | |
90 | E.g. how to set a breakpoint that just prints something and continues. | |
91 | How to break on aborts. Etc. | |
92 | ||
dd3b648e RP |
93 | Provide "voodoo" debugging of core files. This creates a zombie |
94 | process as a child of the debugger, and loads it up with the data, | |
95 | stack, and regs of the core file. This allows you to call functions | |
96 | in the executable, to manipulate the data in the core file. | |
97 | ||
98 | GDB reopens the source file on every line, as you "next" through it. | |
99 | ||
100 | Referencing the vtbl member of a struct doesn't work. It prints OK | |
101 | if you print the struct, but it gets 0 if you try to deref it. | |
102 | ||
103 | Persistent command history: A feature where you could save off a list | |
104 | of the commands you did, so you can edit it into something that will bring | |
adf2bb58 | 105 | the target to the same place every time you source it. |
dd3b648e RP |
106 | This would also be useful for automated fast watchpointing; if you go |
107 | past the place where it watchpoints, you just start it over again and | |
108 | do it more carefully. | |
109 | ||
adf2bb58 JG |
110 | Deal with the SunOS 4.0 and 4.1.1 ptrace bug that loses the registers if |
111 | the stack is paged out. | |
dd3b648e RP |
112 | |
113 | Finish the C++ exception handling stub routines. Lint points them out | |
114 | as unused statics functions. | |
115 | ||
adf2bb58 | 116 | Perhaps "i source" should take an argument like that of "list". |
dd3b648e RP |
117 | |
118 | See if coredep.c's fetch_core_registers can be used on more machines. | |
119 | E.g. MIPS (mips-xdep.c). | |
120 | ||
dd3b648e RP |
121 | unpack_double() does not handle IEEE float on the target unless the host |
122 | is also IEEE. Death on a vax. | |
123 | ||
dd3b648e RP |
124 | Set up interface between GDB and INFO so that you can hop into interactive |
125 | INFO and back out again. When running under Emacs, should use Emacs | |
126 | info, else fork the info program. Installation of GDB should install | |
127 | its texinfo files into the info tree automagically, including the readline | |
adf2bb58 | 128 | texinfo files. |
dd3b648e | 129 | |
adf2bb58 | 130 | "help address" ought to find the "help set print address" entry. |
dd3b648e RP |
131 | |
132 | Remove the VTBL internal guts from printouts of C++ structs, unless | |
133 | vtblprint is set. | |
134 | ||
b1dcd6ac | 135 | Remove "at 0xnnnn" from the "b foo" response, if `print address off' and if |
dd3b648e RP |
136 | it matches the source line indicated. |
137 | ||
138 | The prompt at end of screen should accept space as well as CR. | |
139 | ||
4a44dc1a JK |
140 | "List" should put you into a pseudo-"more" where you can hit space to |
141 | get more, forever to eof. (questionable--you can already hit return | |
142 | to get more, and modal user interfaces are evil -kingdon, 28 Oct | |
143 | 1993). | |
dd3b648e RP |
144 | |
145 | Check STORE_RETURN_VALUE on all architectures. Check near it in tm-sparc.h | |
146 | for other bogosities. | |
147 | ||
148 | Check for storage leaks in GDB, I'm sure there are a lot! | |
149 | ||
150 | vtblprint of a vtbl should demangle the names it's printing. | |
151 | ||
4a44dc1a JK |
152 | Backtrace should point out what the currently selected frame is, in |
153 | its display, perhaps showing "@3 foo (bar, ...)" or ">3 foo (bar, | |
154 | ...)" rather than "#3 foo (bar, ...)". | |
dd3b648e RP |
155 | |
156 | "i program" should work for core files, and display more info, like what | |
157 | actually caused it to die. | |
158 | ||
dd3b648e RP |
159 | "x/10i" should shorten the long name, if any, on subsequent lines. |
160 | ||
161 | Check through the code for FIXME comments and fix them. dbxread.c, | |
162 | blockframe.c, and plenty more. | |
163 | ||
164 | "next" over a function that longjumps, never stops until next time you happen | |
165 | to get to that spot by accident. E.g. "n" over execute_command which has | |
166 | an error. | |
167 | ||
adf2bb58 | 168 | Watchpoints seem not entirely reliable, though they haven't failed me recently. |
dd3b648e RP |
169 | |
170 | "set zeroprint off", don't bother printing members of structs which are entirely | |
171 | zero. Useful for those big structs with few useful members. | |
172 | ||
173 | GDB does four ioctl's for every command, probably switching terminal modes | |
174 | to/from inferior or for readline or something. | |
175 | ||
176 | terminal_ours versus terminal_inferior: cache state. Switch should be a noop | |
177 | if the state is the same, too. | |
178 | ||
179 | ptype $i6 = void??! | |
180 | ||
181 | Clean up invalid_float handling so gdb doesn't coredump when it tries to | |
182 | access a NaN. While this might work on SPARC, other machines are not | |
183 | configured right. | |
184 | ||
185 | "b value_at ; commands ; continue ; end" stops EVERY OTHER TIME! | |
186 | Then once you enter a command, it does the command, runs two more | |
187 | times, and then stops again! Bizarre... (This behaviour has been | |
188 | modified, but it is not yet 100% predictable when e.g. the commands | |
189 | call functions in the child, and while there, the child is interrupted | |
190 | with a signal, or hits a breakpoint.) | |
191 | ||
dd3b648e RP |
192 | help completion, help history should work. |
193 | ||
dd3b648e RP |
194 | Check that we can handle stack trace through varargs AND alloca in same |
195 | function, on 29K. | |
196 | ||
197 | wait_for_inferior loops forever if wait() gives it an error. | |
198 | ||
dd3b648e RP |
199 | "i frame" shows wrong "arglist at" location, doesn't show where the args |
200 | should be found, only their actual values. | |
201 | ||
dd3b648e RP |
202 | There should be a way for "set" commands to validate the new setting |
203 | before it takes effect. | |
204 | ||
dd3b648e RP |
205 | A mess of floating point opcodes are missing from sparc-opcode.h. |
206 | Also, a little program should test the table for bits that are | |
207 | overspecified or underspecified. E.g. if the must-be-ones bits | |
208 | and the must-be-zeroes bits leave some fields unexamined, and the format | |
209 | string leaves them unprinted, then point this out. If multiple | |
210 | non-alias patterns match, point this out too. Finally, there should | |
211 | be a sparc-optest.s file that tries each pattern out. This file | |
212 | should end up coming back the same (modulo transformation comments) | |
213 | if fed to "gas" then the .o is fed to gdb for disassembly. | |
214 | ||
dd3b648e RP |
215 | Eliminate all the core_file_command's in all the xdep files. |
216 | Eliminate separate declarations of registers[] everywhere. | |
217 | ||
218 | "ena d" is ambiguous, why? "ena delete" seems to think it is a command! | |
219 | ||
220 | Line numbers are off in some spots. In proceed() at 1st "oneproc = 1", | |
221 | it seems to run that statement, but it doesn't actually. | |
222 | ||
77eb2e8f JK |
223 | Perhaps move the tdep, xdep, and nat files, into the config |
224 | subdirectories. If not, at least straighten out their names so that | |
dd3b648e RP |
225 | they all start with the machine name. |
226 | ||
227 | inferior_status should include stop_print_frame. It won't need to be | |
228 | reset in wait_for_inferior after bpstat_stop_status call, then. | |
229 | ||
230 | i line VAR produces "Line number not known for symbol ``var''.". I | |
231 | thought we were stashing that info now! | |
232 | ||
dd3b648e RP |
233 | We should be able to write to random files at hex offsets like adb. |
234 | ||
dd3b648e RP |
235 | Make "target xxx" command interruptible. |
236 | ||
237 | Handle add_file with separate text, data, and bss addresses. Maybe | |
238 | handle separate addresses for each segment in the object file? | |
239 | ||
240 | Handle free_named_symtab to cope with multiply-loaded object files | |
241 | in a dynamic linking environment. Should remember the last copy loaded, | |
242 | but not get too snowed if it finds references to the older copy. | |
243 | ||
dd3b648e RP |
244 | The original BFD core dump reading routine would itself coredump when fed |
245 | a garbage file as a core file. Does the current one? | |
246 | ||
dd3b648e | 247 | Generalize and Standardize the RPC interface to a target program, |
adf2bb58 JG |
248 | improve it beyond the "ptrace" interface, and see if it can become a |
249 | standard for remote debugging. | |
dd3b648e RP |
250 | |
251 | Remove all references to: | |
252 | text_offset | |
253 | data_offset | |
254 | text_data_start | |
255 | text_end | |
256 | exec_data_offset | |
257 | ... | |
258 | now that we have BFD. All remaining are in machine dependent files. | |
259 | ||
260 | When quitting with a running program, if a core file was previously | |
261 | examined, you get "Couldn't read float regs from core file"...if | |
262 | indeed it can't. generic_mourn_inferior... | |
263 | ||
4a44dc1a JK |
264 | Have remote targets give a warning on a signal argument to |
265 | target_resume. Or better yet, extend the protocols so that it works | |
266 | like it does on the Unix-like systems. | |
dd3b648e | 267 | |
dd3b648e RP |
268 | Sort help and info output. |
269 | ||
270 | Re-organize help categories into things that tend to fit on a screen | |
271 | and hang together. | |
272 | ||
dd3b648e RP |
273 | renote-nindy.c handles interrupts poorly; it error()s out of badly |
274 | chosen places, e.g. leaving current_frame zero, which causes core dumps | |
275 | on the next command. | |
276 | ||
277 | Add in commands like ADB's for searching for patterns, etc. We should | |
278 | be able to examine and patch raw unsymboled binaries as well in gdb as | |
279 | we can in adb. (E.g. increase the timeout in /bin/login without source). | |
280 | ||
281 | Those xdep files that call register_addr without defining it are | |
282 | probably simply broken. When reconfiguring this part of gdb, I could | |
283 | only make guesses about how to redo some of those files, and I | |
284 | probably guessed wrong, or left them "for later" when I have a | |
285 | machine that can attempt to build them. | |
286 | ||
dd3b648e RP |
287 | When doing "step" or "next", if a few lines of source are skipped between |
288 | the previous line and the current one, print those lines, not just the | |
289 | last line of a multiline statement. | |
290 | ||
291 | When searching for C++ superclasses in value_cast in valops.c, we must | |
292 | not search the "fields", only the "superclasses". There might be a | |
293 | struct with a field name that matches the superclass name. This can | |
294 | happen when the struct was defined before the superclass (before the | |
295 | name became a typedef). | |
296 | ||
36b9d39c JG |
297 | Handling of "&" address-of operator needs some serious overhaul |
298 | for ANSI C and consistency on arrays and functions. | |
299 | For "float point[15];": | |
dd3b648e | 300 | ptype &point[4] ==> Attempt to take address of non-lvalue. |
36b9d39c JG |
301 | For "char *malloc();": |
302 | ptype malloc ==> "char *()"; should be same as | |
303 | ptype &malloc ==> "char *(*)()" | |
304 | call printf ("%x\n", malloc) ==> wierd value, should be same as | |
305 | call printf ("%x\n", &malloc) ==> correct value | |
dd3b648e | 306 | |
4a44dc1a JK |
307 | Fix dbxread.c symbol reading in the presence of interrupts. It |
308 | currently leaves a cleanup to blow away the entire symbol table when a | |
309 | QUIT occurs. (What's wrong with that? -kingdon, 28 Oct 1993). | |
ef98d5ac JG |
310 | |
311 | Mipsread.c reads include files depth-first, because the dependencies | |
312 | in the psymtabs are way too inclusive (it seems to me). Figure out what | |
313 | really depends on what, to avoid recursing 20 or 30 times while reading | |
314 | real symtabs. | |
315 | ||
316 | value_add() should be subtracting the lower bound of arrays, if known, | |
317 | and possibly checking against the upper bound for error reporting. | |
318 | ||
319 | mipsread.c symbol table allocation and deallocation should be checked. | |
320 | My suspicion is that it's full of memory leaks. | |
321 | ||
322 | SunOS should have a target_lookup_symbol() for common'd things allocated | |
323 | by the shared library linker ld.so. | |
324 | ||
325 | When listing source lines, check for a preceding \n, to verify that | |
326 | the file hasn't changed out from under us. | |
327 | ||
328 | When listing source lines, eat leading whitespace corresponding to the | |
329 | line-number prefix we print. This avoids long lines wrapping. | |
330 | ||
331 | mipsread.c needs to check for old symtabs and psymtabs for the same | |
332 | files, the way it happens for dbxread.c and coffread.c, for VxWorks | |
333 | incremental symbol table reloading. | |
334 | ||
4a44dc1a JK |
335 | Get all the remote systems (where the protocol allows it) to be able to |
336 | stop the remote system when the GDB user types ^C (like remote.c | |
337 | does). For ebmon, use ^Ak. | |
ce31a796 JK |
338 | |
339 | Possible feature: A version of the "disassemble" command which shows | |
2af4851f JK |
340 | both source and assembly code ("set symbol-filename on" is a partial |
341 | solution). |