Commit | Line | Data |
---|---|---|
a933dad1 | 1 | Debugging GNU Emacs |
bfd6d01a | 2 | Copyright (C) 1985, 2000, 2001, 2002, 2003, 2004, |
5b0d63bc | 3 | 2005, 2006 Free Software Foundation, Inc. |
a933dad1 DL |
4 | |
5 | Permission is granted to anyone to make or distribute verbatim copies | |
6 | of this document as received, in any medium, provided that the | |
7 | copyright notice and permission notice are preserved, | |
8 | and that the distributor grants the recipient permission | |
9 | for further redistribution as permitted by this notice. | |
10 | ||
11 | Permission is granted to distribute modified versions | |
12 | of this document, or of portions of it, | |
13 | under the above conditions, provided also that they | |
14 | carry prominent notices stating who last changed them. | |
15 | ||
437368fe EZ |
16 | [People who debug Emacs on Windows using native Windows debuggers |
17 | should read the Windows-specific section near the end of this | |
18 | document.] | |
19 | ||
42a3c627 | 20 | ** When you debug Emacs with GDB, you should start it in the directory |
7483daa1 RS |
21 | where the executable was made. That directory has a .gdbinit file |
22 | that defines various "user-defined" commands for debugging Emacs. | |
42a3c627 | 23 | |
70d9a9cd DK |
24 | ** When you are trying to analyze failed assertions, it will be |
25 | essential to compile Emacs either completely without optimizations or | |
26 | at least (when using GCC) with the -fno-crossjumping option. Failure | |
27 | to do so may make the compiler recycle the same abort call for all | |
28 | assertions in a given function, rendering the stack backtrace useless | |
29 | for identifying the specific failed assertion. | |
30 | ||
42a3c627 | 31 | ** It is a good idea to run Emacs under GDB (or some other suitable |
437368fe EZ |
32 | debugger) *all the time*. Then, when Emacs crashes, you will be able |
33 | to debug the live process, not just a core dump. (This is especially | |
34 | important on systems which don't support core files, and instead print | |
35 | just the registers and some stack addresses.) | |
36 | ||
42a3c627 | 37 | ** If Emacs hangs, or seems to be stuck in some infinite loop, typing |
437368fe EZ |
38 | "kill -TSTP PID", where PID is the Emacs process ID, will cause GDB to |
39 | kick in, provided that you run under GDB. | |
40 | ||
41 | ** Getting control to the debugger | |
a933dad1 | 42 | |
3102e429 | 43 | `Fsignal' is a very useful place to put a breakpoint in. |
a933dad1 DL |
44 | All Lisp errors go through there. |
45 | ||
3102e429 | 46 | It is useful, when debugging, to have a guaranteed way to return to |
eb55f651 | 47 | the debugger at any time. When using X, this is easy: type C-z at the |
3102e429 EZ |
48 | window where Emacs is running under GDB, and it will stop Emacs just |
49 | as it would stop any ordinary program. When Emacs is running in a | |
50 | terminal, things are not so easy. | |
51 | ||
52 | The src/.gdbinit file in the Emacs distribution arranges for SIGINT | |
53 | (C-g in Emacs) to be passed to Emacs and not give control back to GDB. | |
54 | On modern POSIX systems, you can override that with this command: | |
55 | ||
7718638c | 56 | handle SIGINT stop nopass |
3102e429 EZ |
57 | |
58 | After this `handle' command, SIGINT will return control to GDB. If | |
59 | you want the C-g to cause a QUIT within Emacs as well, omit the | |
60 | `nopass'. | |
61 | ||
62 | A technique that can work when `handle SIGINT' does not is to store | |
63 | the code for some character into the variable stop_character. Thus, | |
a933dad1 DL |
64 | |
65 | set stop_character = 29 | |
66 | ||
67 | makes Control-] (decimal code 29) the stop character. | |
68 | Typing Control-] will cause immediate stop. You cannot | |
69 | use the set command until the inferior process has been started. | |
70 | Put a breakpoint early in `main', or suspend the Emacs, | |
71 | to get an opportunity to do the set command. | |
72 | ||
8ef597fe NR |
73 | When Emacs is running in a terminal, it is useful to use a separate terminal |
74 | for the debug session. This can be done by starting Emacs as usual, then | |
75 | attaching to it from gdb with the `attach' command which is explained in the | |
76 | node "Attach" of the GDB manual. | |
77 | ||
a933dad1 DL |
78 | ** Examining Lisp object values. |
79 | ||
80 | When you have a live process to debug, and it has not encountered a | |
81 | fatal error, you can use the GDB command `pr'. First print the value | |
82 | in the ordinary way, with the `p' command. Then type `pr' with no | |
83 | arguments. This calls a subroutine which uses the Lisp printer. | |
84 | ||
1ce9f40a KS |
85 | You can also use `pp value' to print the emacs value directly. |
86 | ||
87 | Note: It is not a good idea to try `pr' or `pp' if you know that Emacs | |
88 | is in deep trouble: its stack smashed (e.g., if it encountered SIGSEGV | |
89 | due to stack overflow), or crucial data structures, such as `obarray', | |
437368fe EZ |
90 | corrupted, etc. In such cases, the Emacs subroutine called by `pr' |
91 | might make more damage, like overwrite some data that is important for | |
92 | debugging the original problem. | |
93 | ||
3102e429 EZ |
94 | Also, on some systems it is impossible to use `pr' if you stopped |
95 | Emacs while it was inside `select'. This is in fact what happens if | |
96 | you stop Emacs while it is waiting. In such a situation, don't try to | |
97 | use `pr'. Instead, use `s' to step out of the system call. Then | |
98 | Emacs will be between instructions and capable of handling `pr'. | |
a933dad1 | 99 | |
3102e429 EZ |
100 | If you can't use `pr' command, for whatever reason, you can fall back |
101 | on lower-level commands. Use the `xtype' command to print out the | |
102 | data type of the last data value. Once you know the data type, use | |
103 | the command that corresponds to that type. Here are these commands: | |
a933dad1 DL |
104 | |
105 | xint xptr xwindow xmarker xoverlay xmiscfree xintfwd xboolfwd xobjfwd | |
106 | xbufobjfwd xkbobjfwd xbuflocal xbuffer xsymbol xstring xvector xframe | |
107 | xwinconfig xcompiled xcons xcar xcdr xsubr xprocess xfloat xscrollbar | |
108 | ||
109 | Each one of them applies to a certain type or class of types. | |
110 | (Some of these types are not visible in Lisp, because they exist only | |
111 | internally.) | |
112 | ||
113 | Each x... command prints some information about the value, and | |
114 | produces a GDB value (subsequently available in $) through which you | |
115 | can get at the rest of the contents. | |
116 | ||
437368fe | 117 | In general, most of the rest of the contents will be additional Lisp |
a933dad1 DL |
118 | objects which you can examine in turn with the x... commands. |
119 | ||
437368fe EZ |
120 | Even with a live process, these x... commands are useful for |
121 | examining the fields in a buffer, window, process, frame or marker. | |
122 | Here's an example using concepts explained in the node "Value History" | |
aa1f38cd NR |
123 | of the GDB manual to print values associated with the variable |
124 | called frame. First, use these commands: | |
437368fe EZ |
125 | |
126 | cd src | |
127 | gdb emacs | |
aa1f38cd | 128 | b set_frame_buffer_list |
177c0ea7 | 129 | r -q |
437368fe | 130 | |
8ef597fe | 131 | Then Emacs hits the breakpoint: |
437368fe EZ |
132 | |
133 | (gdb) p frame | |
aa1f38cd | 134 | $1 = 139854428 |
437368fe EZ |
135 | (gdb) xtype |
136 | Lisp_Vectorlike | |
137 | PVEC_FRAME | |
138 | (gdb) xframe | |
aa1f38cd | 139 | $2 = (struct frame *) 0x8560258 |
437368fe EZ |
140 | (gdb) p *$ |
141 | $3 = { | |
aa1f38cd NR |
142 | size = 1073742931, |
143 | next = 0x85dfe58, | |
144 | name = 140615219, | |
437368fe EZ |
145 | [...] |
146 | } | |
147 | (gdb) p $3->name | |
aa1f38cd | 148 | $4 = 140615219 |
437368fe EZ |
149 | |
150 | Now we can use `pr' to print the name of the frame: | |
151 | ||
152 | (gdb) pr | |
153 | "emacs@steenrod.math.nwu.edu" | |
154 | ||
155 | The Emacs C code heavily uses macros defined in lisp.h. So suppose | |
156 | we want the address of the l-value expression near the bottom of | |
157 | `add_command_key' from keyboard.c: | |
158 | ||
159 | XVECTOR (this_command_keys)->contents[this_command_key_count++] = key; | |
160 | ||
f6e405a2 NR |
161 | XVECTOR is a macro, so GDB only knows about it if Emacs has been compiled with |
162 | preprocessor macro information. GCC provides this if you specify the options | |
163 | `-gdwarf-2' and `-g3'. In this case, GDB can evaluate expressions like | |
164 | "p XVECTOR (this_command_keys)". | |
437368fe | 165 | |
f6e405a2 NR |
166 | When this information isn't available, you can use the xvector command in GDB |
167 | to get the same result. Here is how: | |
437368fe EZ |
168 | |
169 | (gdb) p this_command_keys | |
170 | $1 = 1078005760 | |
171 | (gdb) xvector | |
172 | $2 = (struct Lisp_Vector *) 0x411000 | |
173 | 0 | |
174 | (gdb) p $->contents[this_command_key_count] | |
175 | $3 = 1077872640 | |
176 | (gdb) p &$ | |
177 | $4 = (int *) 0x411008 | |
178 | ||
179 | Here's a related example of macros and the GDB `define' command. | |
180 | There are many Lisp vectors such as `recent_keys', which contains the | |
181 | last 100 keystrokes. We can print this Lisp vector | |
182 | ||
183 | p recent_keys | |
184 | pr | |
185 | ||
186 | But this may be inconvenient, since `recent_keys' is much more verbose | |
187 | than `C-h l'. We might want to print only the last 10 elements of | |
188 | this vector. `recent_keys' is updated in keyboard.c by the command | |
189 | ||
190 | XVECTOR (recent_keys)->contents[recent_keys_index] = c; | |
191 | ||
192 | So we define a GDB command `xvector-elts', so the last 10 keystrokes | |
177c0ea7 | 193 | are printed by |
437368fe EZ |
194 | |
195 | xvector-elts recent_keys recent_keys_index 10 | |
196 | ||
197 | where you can define xvector-elts as follows: | |
198 | ||
199 | define xvector-elts | |
200 | set $i = 0 | |
201 | p $arg0 | |
202 | xvector | |
203 | set $foo = $ | |
204 | while $i < $arg2 | |
177c0ea7 | 205 | p $foo->contents[$arg1-($i++)] |
437368fe EZ |
206 | pr |
207 | end | |
208 | document xvector-elts | |
209 | Prints a range of elements of a Lisp vector. | |
210 | xvector-elts v n i | |
211 | prints `i' elements of the vector `v' ending at the index `n'. | |
212 | end | |
213 | ||
214 | ** Getting Lisp-level backtrace information within GDB | |
215 | ||
3102e429 EZ |
216 | The most convenient way is to use the `xbacktrace' command. This |
217 | shows the names of the Lisp functions that are currently active. | |
437368fe EZ |
218 | |
219 | If that doesn't work (e.g., because the `backtrace_list' structure is | |
220 | corrupted), type "bt" at the GDB prompt, to produce the C-level | |
221 | backtrace, and look for stack frames that call Ffuncall. Select them | |
222 | one by one in GDB, by typing "up N", where N is the appropriate number | |
223 | of frames to go up, and in each frame that calls Ffuncall type this: | |
224 | ||
225 | p *args | |
226 | pr | |
227 | ||
228 | This will print the name of the Lisp function called by that level | |
229 | of function calling. | |
230 | ||
231 | By printing the remaining elements of args, you can see the argument | |
232 | values. Here's how to print the first argument: | |
177c0ea7 | 233 | |
437368fe EZ |
234 | p args[1] |
235 | pr | |
236 | ||
237 | If you do not have a live process, you can use xtype and the other | |
238 | x... commands such as xsymbol to get such information, albeit less | |
239 | conveniently. For example: | |
240 | ||
241 | p *args | |
242 | xtype | |
243 | ||
244 | and, assuming that "xtype" says that args[0] is a symbol: | |
245 | ||
177c0ea7 | 246 | xsymbol |
437368fe | 247 | |
c167547b NR |
248 | ** Using GDB in Emacs |
249 | ||
250 | Debugging with GDB in Emacs offers some advantages over the command line (See | |
251 | the GDB Graphical Interface node of the Emacs manual). There are also some | |
252 | features available just for debugging Emacs: | |
253 | ||
254 | 1) The command gud-pp isavailable on the tool bar (the `pp' icon) and allows | |
255 | the user to print the s-expression of the variable at point, in the GUD | |
256 | buffer. | |
257 | ||
258 | 2) Pressing `p' on a component of a watch expression that is a lisp object | |
259 | in the speedbar prints its s-expression in the GUD buffer. | |
260 | ||
261 | 3) The STOP button on the tool bar is adjusted so that it sends SIGTSTP | |
262 | instead of the usual SIGINT. | |
263 | ||
264 | 4) The command gud-pv has the global binding 'C-x C-a C-v' and prints the | |
265 | value of the lisp variable at point. | |
266 | ||
437368fe EZ |
267 | ** Debugging what happens while preloading and dumping Emacs |
268 | ||
269 | Type `gdb temacs' and start it with `r -batch -l loadup dump'. | |
270 | ||
271 | If temacs actually succeeds when running under GDB in this way, do not | |
272 | try to run the dumped Emacs, because it was dumped with the GDB | |
273 | breakpoints in it. | |
274 | ||
275 | ** Debugging `temacs' | |
276 | ||
277 | Debugging `temacs' is useful when you want to establish whether a | |
278 | problem happens in an undumped Emacs. To run `temacs' under a | |
279 | debugger, type "gdb temacs", then start it with `r -batch -l loadup'. | |
280 | ||
281 | ** If you encounter X protocol errors | |
282 | ||
283 | Try evaluating (x-synchronize t). That puts Emacs into synchronous | |
284 | mode, where each Xlib call checks for errors before it returns. This | |
285 | mode is much slower, but when you get an error, you will see exactly | |
286 | which call really caused the error. | |
287 | ||
ec4054f0 EZ |
288 | You can start Emacs in a synchronous mode by invoking it with the -xrm |
289 | option, like this: | |
290 | ||
9031cdf2 | 291 | emacs -xrm "emacs.synchronous: true" |
ec4054f0 EZ |
292 | |
293 | Setting a breakpoint in the function `x_error_quitter' and looking at | |
294 | the backtrace when Emacs stops inside that function will show what | |
295 | code causes the X protocol errors. | |
296 | ||
1d3adc26 EZ |
297 | Some bugs related to the X protocol disappear when Emacs runs in a |
298 | synchronous mode. To track down those bugs, we suggest the following | |
299 | procedure: | |
300 | ||
301 | - Run Emacs under a debugger and put a breakpoint inside the | |
302 | primitive function which, when called from Lisp, triggers the X | |
303 | protocol errors. For example, if the errors happen when you | |
304 | delete a frame, put a breakpoint inside `Fdelete_frame'. | |
305 | ||
306 | - When the breakpoint breaks, step through the code, looking for | |
307 | calls to X functions (the ones whose names begin with "X" or | |
308 | "Xt" or "Xm"). | |
309 | ||
310 | - Insert calls to `XSync' before and after each call to the X | |
311 | functions, like this: | |
312 | ||
313 | XSync (f->output_data.x->display_info->display, 0); | |
314 | ||
315 | where `f' is the pointer to the `struct frame' of the selected | |
316 | frame, normally available via XFRAME (selected_frame). (Most | |
317 | functions which call X already have some variable that holds the | |
318 | pointer to the frame, perhaps called `f' or `sf', so you shouldn't | |
319 | need to compute it.) | |
320 | ||
321 | If your debugger can call functions in the program being debugged, | |
322 | you should be able to issue the calls to `XSync' without recompiling | |
323 | Emacs. For example, with GDB, just type: | |
324 | ||
325 | call XSync (f->output_data.x->display_info->display, 0) | |
326 | ||
327 | before and immediately after the suspect X calls. If your | |
328 | debugger does not support this, you will need to add these pairs | |
329 | of calls in the source and rebuild Emacs. | |
330 | ||
331 | Either way, systematically step through the code and issue these | |
332 | calls until you find the first X function called by Emacs after | |
333 | which a call to `XSync' winds up in the function | |
334 | `x_error_quitter'. The first X function call for which this | |
335 | happens is the one that generated the X protocol error. | |
336 | ||
337 | - You should now look around this offending X call and try to figure | |
338 | out what is wrong with it. | |
339 | ||
46def989 JD |
340 | ** If Emacs causes errors or memory leaks in your X server |
341 | ||
342 | You can trace the traffic between Emacs and your X server with a tool | |
343 | like xmon, available at ftp://ftp.x.org/contrib/devel_tools/. | |
344 | ||
345 | Xmon can be used to see exactly what Emacs sends when X protocol errors | |
346 | happen. If Emacs causes the X server memory usage to increase you can | |
347 | use xmon to see what items Emacs creates in the server (windows, | |
348 | graphical contexts, pixmaps) and what items Emacs delete. If there | |
349 | are consistently more creations than deletions, the type of item | |
350 | and the activity you do when the items get created can give a hint where | |
351 | to start debugging. | |
352 | ||
437368fe EZ |
353 | ** If the symptom of the bug is that Emacs fails to respond |
354 | ||
355 | Don't assume Emacs is `hung'--it may instead be in an infinite loop. | |
356 | To find out which, make the problem happen under GDB and stop Emacs | |
357 | once it is not responding. (If Emacs is using X Windows directly, you | |
358 | can stop Emacs by typing C-z at the GDB job.) Then try stepping with | |
359 | `step'. If Emacs is hung, the `step' command won't return. If it is | |
360 | looping, `step' will return. | |
361 | ||
362 | If this shows Emacs is hung in a system call, stop it again and | |
363 | examine the arguments of the call. If you report the bug, it is very | |
364 | important to state exactly where in the source the system call is, and | |
365 | what the arguments are. | |
366 | ||
367 | If Emacs is in an infinite loop, try to determine where the loop | |
368 | starts and ends. The easiest way to do this is to use the GDB command | |
369 | `finish'. Each time you use it, Emacs resumes execution until it | |
370 | exits one stack frame. Keep typing `finish' until it doesn't | |
371 | return--that means the infinite loop is in the stack frame which you | |
372 | just tried to finish. | |
373 | ||
374 | Stop Emacs again, and use `finish' repeatedly again until you get back | |
375 | to that frame. Then use `next' to step through that frame. By | |
376 | stepping, you will see where the loop starts and ends. Also, examine | |
377 | the data being used in the loop and try to determine why the loop does | |
378 | not exit when it should. | |
379 | ||
380 | ** If certain operations in Emacs are slower than they used to be, here | |
381 | is some advice for how to find out why. | |
382 | ||
383 | Stop Emacs repeatedly during the slow operation, and make a backtrace | |
384 | each time. Compare the backtraces looking for a pattern--a specific | |
385 | function that shows up more often than you'd expect. | |
386 | ||
387 | If you don't see a pattern in the C backtraces, get some Lisp | |
388 | backtrace information by typing "xbacktrace" or by looking at Ffuncall | |
389 | frames (see above), and again look for a pattern. | |
390 | ||
391 | When using X, you can stop Emacs at any time by typing C-z at GDB. | |
392 | When not using X, you can do this with C-g. On non-Unix platforms, | |
393 | such as MS-DOS, you might need to press C-BREAK instead. | |
394 | ||
a933dad1 DL |
395 | ** If GDB does not run and your debuggers can't load Emacs. |
396 | ||
397 | On some systems, no debugger can load Emacs with a symbol table, | |
398 | perhaps because they all have fixed limits on the number of symbols | |
399 | and Emacs exceeds the limits. Here is a method that can be used | |
400 | in such an extremity. Do | |
401 | ||
402 | nm -n temacs > nmout | |
403 | strip temacs | |
404 | adb temacs | |
405 | 0xd:i | |
406 | 0xe:i | |
407 | 14:i | |
408 | 17:i | |
409 | :r -l loadup (or whatever) | |
410 | ||
411 | It is necessary to refer to the file `nmout' to convert | |
412 | numeric addresses into symbols and vice versa. | |
413 | ||
414 | It is useful to be running under a window system. | |
415 | Then, if Emacs becomes hopelessly wedged, you can create | |
416 | another window to do kill -9 in. kill -ILL is often | |
417 | useful too, since that may make Emacs dump core or return | |
418 | to adb. | |
419 | ||
420 | ||
421 | ** Debugging incorrect screen updating. | |
422 | ||
423 | To debug Emacs problems that update the screen wrong, it is useful | |
424 | to have a record of what input you typed and what Emacs sent to the | |
425 | screen. To make these records, do | |
426 | ||
427 | (open-dribble-file "~/.dribble") | |
428 | (open-termscript "~/.termscript") | |
429 | ||
430 | The dribble file contains all characters read by Emacs from the | |
431 | terminal, and the termscript file contains all characters it sent to | |
432 | the terminal. The use of the directory `~/' prevents interference | |
433 | with any other user. | |
434 | ||
435 | If you have irreproducible display problems, put those two expressions | |
436 | in your ~/.emacs file. When the problem happens, exit the Emacs that | |
437 | you were running, kill it, and rename the two files. Then you can start | |
438 | another Emacs without clobbering those files, and use it to examine them. | |
125f929e MB |
439 | |
440 | An easy way to see if too much text is being redrawn on a terminal is to | |
441 | evaluate `(setq inverse-video t)' before you try the operation you think | |
442 | will cause too much redrawing. This doesn't refresh the screen, so only | |
443 | newly drawn text is in inverse video. | |
437368fe | 444 | |
3f715e77 EZ |
445 | The Emacs display code includes special debugging code, but it is |
446 | normally disabled. You can enable it by building Emacs with the | |
447 | pre-processing symbol GLYPH_DEBUG defined. Here's one easy way, | |
448 | suitable for Unix and GNU systems, to build such a debugging version: | |
449 | ||
450 | MYCPPFLAGS='-DGLYPH_DEBUG=1' make | |
451 | ||
452 | Building Emacs like that activates many assertions which scrutinize | |
453 | display code operation more than Emacs does normally. (To see the | |
454 | code which tests these assertions, look for calls to the `xassert' | |
455 | macros.) Any assertion that is reported to fail should be | |
456 | investigated. | |
457 | ||
458 | Building with GLYPH_DEBUG defined also defines several helper | |
459 | functions which can help debugging display code. One such function is | |
460 | `dump_glyph_matrix'. If you run Emacs under GDB, you can print the | |
461 | contents of any glyph matrix by just calling that function with the | |
462 | matrix as its argument. For example, the following command will print | |
463 | the contents of the current matrix of the window whose pointer is in | |
464 | `w': | |
465 | ||
466 | (gdb) p dump_glyph_matrix (w->current_matrix, 2) | |
467 | ||
468 | (The second argument 2 tells dump_glyph_matrix to print the glyphs in | |
469 | a long form.) You can dump the selected window's current glyph matrix | |
470 | interactively with "M-x dump-glyph-matrix RET"; see the documentation | |
471 | of this function for more details. | |
472 | ||
473 | Several more functions for debugging display code are available in | |
71d6b459 EZ |
474 | Emacs compiled with GLYPH_DEBUG defined; type "C-h f dump- TAB" and |
475 | "C-h f trace- TAB" to see the full list. | |
3f715e77 | 476 | |
1ce9f40a KS |
477 | When you debug display problems running emacs under X, you can use |
478 | the `ff' command to flush all pending display updates to the screen. | |
479 | ||
437368fe EZ |
480 | |
481 | ** Debugging LessTif | |
482 | ||
483 | If you encounter bugs whereby Emacs built with LessTif grabs all mouse | |
484 | and keyboard events, or LessTif menus behave weirdly, it might be | |
485 | helpful to set the `DEBUGSOURCES' and `DEBUG_FILE' environment | |
486 | variables, so that one can see what LessTif was doing at this point. | |
487 | For instance | |
177c0ea7 | 488 | |
6806e867 | 489 | export DEBUGSOURCES="RowColumn.c:MenuShell.c:MenuUtil.c" |
437368fe | 490 | export DEBUG_FILE=/usr/tmp/LESSTIF_TRACE |
2aa25884 | 491 | emacs & |
437368fe EZ |
492 | |
493 | causes LessTif to print traces from the three named source files to a | |
2aa25884 EZ |
494 | file in `/usr/tmp' (that file can get pretty large). The above should |
495 | be typed at the shell prompt before invoking Emacs, as shown by the | |
496 | last line above. | |
437368fe EZ |
497 | |
498 | Running GDB from another terminal could also help with such problems. | |
499 | You can arrange for GDB to run on one machine, with the Emacs display | |
500 | appearing on another. Then, when the bug happens, you can go back to | |
501 | the machine where you started GDB and use the debugger from there. | |
502 | ||
503 | ||
437368fe EZ |
504 | ** Debugging problems which happen in GC |
505 | ||
f46cc673 EZ |
506 | The array `last_marked' (defined on alloc.c) can be used to display up |
507 | to 500 last objects marked by the garbage collection process. | |
62578de5 | 508 | Whenever the garbage collector marks a Lisp object, it records the |
1a527e27 EZ |
509 | pointer to that object in the `last_marked' array, which is maintained |
510 | as a circular buffer. The variable `last_marked_index' holds the | |
511 | index into the `last_marked' array one place beyond where the pointer | |
512 | to the very last marked object is stored. | |
437368fe EZ |
513 | |
514 | The single most important goal in debugging GC problems is to find the | |
515 | Lisp data structure that got corrupted. This is not easy since GC | |
516 | changes the tag bits and relocates strings which make it hard to look | |
517 | at Lisp objects with commands such as `pr'. It is sometimes necessary | |
518 | to convert Lisp_Object variables into pointers to C struct's manually. | |
437368fe | 519 | |
1a527e27 EZ |
520 | Use the `last_marked' array and the source to reconstruct the sequence |
521 | that objects were marked. In general, you need to correlate the | |
522 | values recorded in the `last_marked' array with the corresponding | |
523 | stack frames in the backtrace, beginning with the innermost frame. | |
524 | Some subroutines of `mark_object' are invoked recursively, others loop | |
525 | over portions of the data structure and mark them as they go. By | |
526 | looking at the code of those routines and comparing the frames in the | |
527 | backtrace with the values in `last_marked', you will be able to find | |
528 | connections between the values in `last_marked'. E.g., when GC finds | |
529 | a cons cell, it recursively marks its car and its cdr. Similar things | |
530 | happen with properties of symbols, elements of vectors, etc. Use | |
531 | these connections to reconstruct the data structure that was being | |
532 | marked, paying special attention to the strings and names of symbols | |
533 | that you encounter: these strings and symbol names can be used to grep | |
534 | the sources to find out what high-level symbols and global variables | |
535 | are involved in the crash. | |
536 | ||
537 | Once you discover the corrupted Lisp object or data structure, grep | |
538 | the sources for its uses and try to figure out what could cause the | |
539 | corruption. If looking at the sources doesn;t help, you could try | |
540 | setting a watchpoint on the corrupted data, and see what code modifies | |
541 | it in some invalid way. (Obviously, this technique is only useful for | |
542 | data that is modified only very rarely.) | |
543 | ||
544 | It is also useful to look at the corrupted object or data structure in | |
545 | a fresh Emacs session and compare its contents with a session that you | |
546 | are debugging. | |
437368fe | 547 | |
93699019 EZ |
548 | ** Debugging problems with non-ASCII characters |
549 | ||
550 | If you experience problems which seem to be related to non-ASCII | |
551 | characters, such as \201 characters appearing in the buffer or in your | |
552 | files, set the variable byte-debug-flag to t. This causes Emacs to do | |
553 | some extra checks, such as look for broken relations between byte and | |
554 | character positions in buffers and strings; the resulting diagnostics | |
555 | might pinpoint the cause of the problem. | |
556 | ||
036cb5a2 EZ |
557 | ** Debugging the TTY (non-windowed) version |
558 | ||
559 | The most convenient method of debugging the character-terminal display | |
560 | is to do that on a window system such as X. Begin by starting an | |
561 | xterm window, then type these commands inside that window: | |
562 | ||
563 | $ tty | |
564 | $ echo $TERM | |
565 | ||
566 | Let's say these commands print "/dev/ttyp4" and "xterm", respectively. | |
567 | ||
568 | Now start Emacs (the normal, windowed-display session, i.e. without | |
569 | the `-nw' option), and invoke "M-x gdb RET emacs RET" from there. Now | |
570 | type these commands at GDB's prompt: | |
571 | ||
572 | (gdb) set args -nw -t /dev/ttyp4 | |
573 | (gdb) set environment TERM xterm | |
574 | (gdb) run | |
575 | ||
576 | The debugged Emacs should now start in no-window mode with its display | |
577 | directed to the xterm window you opened above. | |
578 | ||
e039053d EZ |
579 | Similar arrangement is possible on a character terminal by using the |
580 | `screen' package. | |
581 | ||
19cf8f36 EZ |
582 | ** Running Emacs built with malloc debugging packages |
583 | ||
584 | If Emacs exhibits bugs that seem to be related to use of memory | |
585 | allocated off the heap, it might be useful to link Emacs with a | |
586 | special debugging library, such as Electric Fence (a.k.a. efence) or | |
587 | GNU Checker, which helps find such problems. | |
588 | ||
589 | Emacs compiled with such packages might not run without some hacking, | |
590 | because Emacs replaces the system's memory allocation functions with | |
591 | its own versions, and because the dumping process might be | |
592 | incompatible with the way these packages use to track allocated | |
593 | memory. Here are some of the changes you might find necessary | |
594 | (SYSTEM-NAME and MACHINE-NAME are the names of your OS- and | |
595 | CPU-specific headers in the subdirectories of `src'): | |
596 | ||
597 | - In src/s/SYSTEM-NAME.h add "#define SYSTEM_MALLOC". | |
598 | ||
599 | - In src/m/MACHINE-NAME.h add "#define CANNOT_DUMP" and | |
600 | "#define CANNOT_UNEXEC". | |
601 | ||
602 | - Configure with a different --prefix= option. If you use GCC, | |
603 | version 2.7.2 is preferred, as some malloc debugging packages | |
604 | work a lot better with it than with 2.95 or later versions. | |
605 | ||
606 | - Type "make" then "make -k install". | |
607 | ||
608 | - If required, invoke the package-specific command to prepare | |
609 | src/temacs for execution. | |
610 | ||
611 | - cd ..; src/temacs | |
612 | ||
613 | (Note that this runs `temacs' instead of the usual `emacs' executable. | |
614 | This avoids problems with dumping Emacs mentioned above.) | |
615 | ||
616 | Some malloc debugging libraries might print lots of false alarms for | |
617 | bitfields used by Emacs in some data structures. If you want to get | |
618 | rid of the false alarms, you will have to hack the definitions of | |
619 | these data structures on the respective headers to remove the `:N' | |
620 | bitfield definitions (which will cause each such field to use a full | |
621 | int). | |
622 | ||
270bf00e EZ |
623 | ** How to recover buffer contents from an Emacs core dump file |
624 | ||
625 | The file etc/emacs-buffer.gdb defines a set of GDB commands for | |
626 | recovering the contents of Emacs buffers from a core dump file. You | |
627 | might also find those commands useful for displaying the list of | |
628 | buffers in human-readable format from within the debugger. | |
629 | ||
437368fe EZ |
630 | ** Some suggestions for debugging on MS Windows: |
631 | ||
632 | (written by Marc Fleischeuers, Geoff Voelker and Andrew Innes) | |
633 | ||
3102e429 | 634 | To debug Emacs with Microsoft Visual C++, you either start emacs from |
961e2394 EZ |
635 | the debugger or attach the debugger to a running emacs process. |
636 | ||
637 | To start emacs from the debugger, you can use the file bin/debug.bat. | |
638 | The Microsoft Developer studio will start and under Project, Settings, | |
3102e429 | 639 | Debug, General you can set the command-line arguments and Emacs's |
437368fe EZ |
640 | startup directory. Set breakpoints (Edit, Breakpoints) at Fsignal and |
641 | other functions that you want to examine. Run the program (Build, | |
642 | Start debug). Emacs will start and the debugger will take control as | |
643 | soon as a breakpoint is hit. | |
644 | ||
3102e429 | 645 | You can also attach the debugger to an already running Emacs process. |
437368fe EZ |
646 | To do this, start up the Microsoft Developer studio and select Build, |
647 | Start debug, Attach to process. Choose the Emacs process from the | |
648 | list. Send a break to the running process (Debug, Break) and you will | |
649 | find that execution is halted somewhere in user32.dll. Open the stack | |
650 | trace window and go up the stack to w32_msg_pump. Now you can set | |
651 | breakpoints in Emacs (Edit, Breakpoints). Continue the running Emacs | |
652 | process (Debug, Step out) and control will return to Emacs, until a | |
653 | breakpoint is hit. | |
654 | ||
3102e429 | 655 | To examine the contents of a Lisp variable, you can use the function |
437368fe EZ |
656 | 'debug_print'. Right-click on a variable, select QuickWatch (it has |
657 | an eyeglass symbol on its button in the toolbar), and in the text | |
658 | field at the top of the window, place 'debug_print(' and ')' around | |
659 | the expression. Press 'Recalculate' and the output is sent to stderr, | |
660 | and to the debugger via the OutputDebugString routine. The output | |
661 | sent to stderr should be displayed in the console window that was | |
662 | opened when the emacs.exe executable was started. The output sent to | |
663 | the debugger should be displayed in the 'Debug' pane in the Output | |
664 | window. If Emacs was started from the debugger, a console window was | |
665 | opened at Emacs' startup; this console window also shows the output of | |
666 | 'debug_print'. | |
667 | ||
668 | For example, start and run Emacs in the debugger until it is waiting | |
669 | for user input. Then click on the `Break' button in the debugger to | |
670 | halt execution. Emacs should halt in `ZwUserGetMessage' waiting for | |
671 | an input event. Use the `Call Stack' window to select the procedure | |
672 | `w32_msp_pump' up the call stack (see below for why you have to do | |
673 | this). Open the QuickWatch window and enter | |
674 | "debug_print(Vexec_path)". Evaluating this expression will then print | |
3102e429 | 675 | out the contents of the Lisp variable `exec-path'. |
437368fe EZ |
676 | |
677 | If QuickWatch reports that the symbol is unknown, then check the call | |
678 | stack in the `Call Stack' window. If the selected frame in the call | |
679 | stack is not an Emacs procedure, then the debugger won't recognize | |
680 | Emacs symbols. Instead, select a frame that is inside an Emacs | |
681 | procedure and try using `debug_print' again. | |
682 | ||
683 | If QuickWatch invokes debug_print but nothing happens, then check the | |
684 | thread that is selected in the debugger. If the selected thread is | |
685 | not the last thread to run (the "current" thread), then it cannot be | |
686 | used to execute debug_print. Use the Debug menu to select the current | |
687 | thread and try using debug_print again. Note that the debugger halts | |
688 | execution (e.g., due to a breakpoint) in the context of the current | |
689 | thread, so this should only be a problem if you've explicitly switched | |
690 | threads. | |
691 | ||
3102e429 | 692 | It is also possible to keep appropriately masked and typecast Lisp |
437368fe EZ |
693 | symbols in the Watch window, this is more convenient when steeping |
694 | though the code. For instance, on entering apply_lambda, you can | |
695 | watch (struct Lisp_Symbol *) (0xfffffff & args[0]). | |
961e2394 EZ |
696 | |
697 | Optimizations often confuse the MS debugger. For example, the | |
698 | debugger will sometimes report wrong line numbers, e.g., when it | |
699 | prints the backtrace for a crash. It is usually best to look at the | |
700 | disassembly to determine exactly what code is being run--the | |
701 | disassembly will probably show several source lines followed by a | |
702 | block of assembler for those lines. The actual point where Emacs | |
703 | crashes will be one of those source lines, but not neccesarily the one | |
704 | that the debugger reports. | |
705 | ||
706 | Another problematic area with the MS debugger is with variables that | |
707 | are stored in registers: it will sometimes display wrong values for | |
708 | those variables. Usually you will not be able to see any value for a | |
709 | register variable, but if it is only being stored in a register | |
710 | temporarily, you will see an old value for it. Again, you need to | |
711 | look at the disassembly to determine which registers are being used, | |
712 | and look at those registers directly, to see the actual current values | |
713 | of these variables. | |
ab5796a9 MB |
714 | |
715 | ;;; arch-tag: fbf32980-e35d-481f-8e4c-a2eca2586e6b |