DEBUG 42.8 KB
Newer Older
Dave Love's avatar
#  
Dave Love committed
1
Debugging GNU Emacs
2

Paul Eggert's avatar
Paul Eggert committed
3
Copyright (C) 1985, 2000-2019 Free Software Foundation, Inc.
4
See the end of the file for license conditions.
Dave Love's avatar
#  
Dave Love committed
5

6
** Preliminaries
Dave Love's avatar
#  
Dave Love committed
7

8 9 10
This section can be skipped if you are already familiar with building
Emacs with debug info, configuring and starting GDB, and simple GDB
debugging techniques.
11

12 13 14
*** Configuring Emacs for debugging

It is best to configure and build Emacs with special options that will
15
make the debugging easier.  Here are the configure-time options we
16 17 18
recommend (they are in addition to any other options you might need,
such as --prefix):

19 20
  ./configure --enable-checking='yes,glyphs' --enable-check-lisp-object-type \
    CFLAGS='-O0 -g3'
21

22 23 24 25 26
The -O0 flag is important, as debugging optimized code can be hard.
If the problem happens only with optimized code, you may need to
enable optimizations.  If that happens, try using -Og first instead of
-O2, as -Og disables some optimizations that make debugging some code
exceptionally hard.
27

28 29
Older versions of GCC may need more than just the -g3 flag.  For more,
search for "analyze failed assertions" below.
30 31 32 33 34 35 36 37 38 39 40 41 42

The 2 --enable-* switches are optional.  They don't have any effect on
debugging with GDB, but will compile additional code that might catch
the problem you are debugging much earlier, in the form of assertion
violation.  The --enable-checking option also enables additional
functionality useful for debugging display problems; see more about
this below under "Debugging Emacs redisplay problems".

Emacs needs not be installed to be debugged, you can debug the binary
created in the 'src' directory.

*** Configuring GDB

43 44 45
To start GDB to debug Emacs, you can simply type "gdb ./emacs RET" at
the shell prompt (assuming you do that from the directory of the Emacs
executable, usually the 'src' sub-directory of the Emacs tree).
Eli Zaretskii's avatar
Eli Zaretskii committed
46 47
However, we recommend starting GDB from Emacs, see below.

48
When you debug Emacs with GDB, you should start GDB in the directory
49 50
where the Emacs executable was made (the 'src' directory in the Emacs
source tree).  That directory has a .gdbinit file that defines various
51 52 53 54
"user-defined" commands for debugging Emacs.  (These commands are
described below under "Examining Lisp object values" and "Debugging
Emacs Redisplay problems".)

55 56
Starting the debugger from Emacs, via the "M-x gdb" command (described
below), when the current buffer visits one of the Emacs C source files
Eli Zaretskii's avatar
Eli Zaretskii committed
57 58 59 60
will automatically start GDB in the 'src' directory.  If you invoke
"M-x gdb" from a buffer whose default directory is different, such as
from the "*scratch*" buffer, you can change the default directory with
the "M-x cd" command before starting the debugger.
61

Eli Zaretskii's avatar
Eli Zaretskii committed
62 63 64
Recent GDB versions by default do not automatically load .gdbinit
files in the directory where you invoke GDB.  With those versions of
GDB, you will see a warning when GDB starts, like this:
65 66 67

  warning: File ".../src/.gdbinit" auto-loading has been declined by your `auto-load safe-path' set to "$debugdir:$datadir/auto-load".

68
The simplest way to fix this is to add the following line to your
Eli Zaretskii's avatar
Eli Zaretskii committed
69
~/.gdbinit file (creating such a file if it doesn't already exist):
70 71 72 73 74 75 76 77

  add-auto-load-safe-path /path/to/emacs/src/.gdbinit

There are other ways to overcome that difficulty, they are all
described in the node "Auto-loading safe path" in the GDB user manual.
If nothing else helps, type "source /path/to/.gdbinit RET" at the GDB
prompt, to unconditionally load the GDB init file.

Eli Zaretskii's avatar
Eli Zaretskii committed
78 79 80 81 82 83 84 85
Running GDB on macOS sometimes brings an error message like this:

  Unable to find Mach task port for process-id NNN: (os/kern) failure (0x5).

To overcome this, search the Internet for the phrase "Unable to find
Mach task port for process-id", and you will find detailed
instructions to follow.

86 87 88
*** Use the Emacs GDB UI front-end

We recommend using the GUI front-end for GDB provided by Emacs.  With
89
it, you can start GDB by typing "M-x gdb RET".  This will suggest the
90 91 92 93 94
file name of the default binary to debug; if the suggested default is
not the Emacs binary you want to debug, change the file name as
needed.  Alternatively, if you want to attach the debugger to an
already running Emacs process, change the GDB command shown in the
minibuffer to say this:
95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148

   gdb -i=mi -p PID

where PID is the numerical process ID of the running Emacs process,
displayed by system utilities such as 'top' or 'ps' on Posix hosts and
Task Manager on MS-Windows.

Once the debugger starts, open the additional windows provided by the
GDB UI, by typing "M-x gdb-many-windows RET".  (Alternatively, click
Gud->GDB-MI->Display Other Windows" from the menu bar.)  At this
point, make your frame large enough (or full-screen) such that the
windows you just opened have enough space to show the content without
horizontal scrolling.

You can later restore your window configuration with the companion
command "M-x gdb-restore-windows RET", or by deselecting "Display
Other Windows" from the menu bar.

*** Setting initial breakpoints

Before you let Emacs run, you should now set breakpoints in the code
which you want to debug, so that Emacs stops there and lets GDB take
control.  If the code which you want to debug is executed under some
rare conditions, or only when a certain Emacs command is manually
invoked, then just set your breakpoint there, let Emacs run, and
trigger the breakpoint by invoking that command or reproducing those
rare conditions.

If you are less lucky, and the code in question is run very
frequently, you will have to find some way of avoiding triggering your
breakpoint when the conditions for the buggy behavior did not yet
happen.  There's no single recipe for this, you will have to be
creative and study the code to see what's appropriate.  Some useful
tricks for that:

  . Make your breakpoint conditional on certain buffer or string
    position.  For example:

      (gdb) break foo.c:1234 if PT >= 9876

  . Set a break point in some rarely called function, then create the
    conditions for the bug, call that rare function, and when GDB gets
    control, set the breakpoint in the buggy code, knowing that it
    will now be called when the bug happens.

  . If the bug manifests itself as an error message, set a breakpoint
    in Fsignal, and when it breaks, look at the backtrace to see what
    triggers the error.

Some additional techniques are described below under "Getting control
to the debugger".

You are now ready to start your debugging session.

Eli Zaretskii's avatar
Eli Zaretskii committed
149 150
*** Running Emacs from GDB

151 152
If you are starting a new Emacs session, type "run", followed by any
command-line arguments (e.g., "-Q") into the *gud-emacs* buffer and
Eli Zaretskii's avatar
Eli Zaretskii committed
153 154
press RET.  If you ran GDB outside of Emacs, type "run" followed by
the command-line arguments at the GDB prompt instead.
155 156 157 158 159

If you attached the debugger to a running Emacs, type "continue" into
the *gud-emacs* buffer and press RET.

Many variables you will encounter while debugging are Lisp objects.
160 161 162 163
These are normally displayed as opaque pointers or integers that are
hard to interpret, especially if they represent long lists.
(They are instead displayed as structures containing these opaque
values, if --enable-check-lisp-object-type is in effect.)  You can
164
use the 'pp' command to display them in their Lisp form.  That command
165 166
displays its output on the standard error stream, which you
can redirect to a file using "M-x redirect-debugging-output".
167 168 169 170 171 172 173
This means that if you attach GDB to a running Emacs that was invoked
from a desktop icon, chances are you will not see the output at all,
or it will wind up in an obscure place (check the documentation of
your desktop environment).

Additional information about displaying Lisp objects can be found
under "Examining Lisp object values" below.
174 175 176 177 178 179 180

The rest of this document describes specific useful techniques for
debugging Emacs; we suggest reading it in its entirety the first time
you are about to debug Emacs, then look up your specific issues
whenever you need.

Good luck!
181 182

** When you are trying to analyze failed assertions or backtraces, it
Paul Eggert's avatar
Paul Eggert committed
183
is essential to compile Emacs with flags suitable for debugging.
184 185 186 187
Although CFLAGS="-O0 -g3" often suffices with modern compilers,
you may benefit further by using CFLAGS="-O0 -g3 -gdwarf-4", replacing
"4" by the highest version of DWARF that your compiler supports;
this is especially important for GCC versions older than 4.8.
188 189 190 191 192
With GCC and higher optimization levels such as -O2, the
-fno-omit-frame-pointer and -fno-crossjumping options are often
essential.  The latter prevents GCC from using the same abort call for
all assertions in a given function, rendering the stack backtrace
useless for identifying the specific failed assertion.
193

194
** It is a good idea to run Emacs under GDB (or some other suitable
195 196 197 198 199
debugger) *all the time*.  Then, when Emacs crashes, you will be able
to debug the live process, not just a core dump.  (This is especially
important on systems which don't support core files, and instead print
just the registers and some stack addresses.)

200
** If Emacs hangs, or seems to be stuck in some infinite loop, typing
201 202 203 204
"kill -TSTP PID", where PID is the Emacs process ID, will cause GDB to
kick in, provided that you run under GDB.

** Getting control to the debugger
Dave Love's avatar
#  
Dave Love committed
205

206 207 208 209 210
Setting a breakpoint in a strategic place, after loading Emacs into
the debugger, but before running it, is the most efficient way of
making sure control will be returned to the debugger when you need
that.

211
'Fsignal' is a very useful place to put a breakpoint in.  All Lisp
212
errors go through there.  If you are only interested in errors that
213 214
would fire the Lisp debugger, breaking at 'maybe_call_debugger' is
useful.
Dave Love's avatar
#  
Dave Love committed
215

216
Another technique for getting control to the debugger is to put a
217 218 219
breakpoint in some rarely used function.  One such convenient function
is Fredraw_display, which you can invoke at will interactively with
"M-x redraw-display RET".
Dave Love's avatar
#  
Dave Love committed
220

221 222 223
It is also useful to have a guaranteed way to return to the debugger
at any arbitrary time.  When using X, this is easy: type C-z at the
window where you are interacting with GDB, and it will stop Emacs just
224 225 226 227 228 229 230 231
as it would stop any ordinary program.  (This doesn't work if GDB was
attached to a running Emacs process; in that case, you will need to
type C-z to the shell window from which Emacs was started, or use the
"kill -TSTP" method described below.)

When Emacs is displaying on a text terminal, things are not so easy,
so we describe the various alternatives below (however, those of them
that use signals only work on Posix systems).
232 233

The src/.gdbinit file in the Emacs distribution arranges for SIGINT
234 235 236
(C-g in Emacs on a text-mode frame) to be passed to Emacs and not give
control back to GDB.  On modern systems, you can override that with
this command:
237

238
   handle SIGINT stop nopass
239

240
After this 'handle' command, SIGINT will return control to GDB.  If
Paul Eggert's avatar
Paul Eggert committed
241
you want the C-g to cause a quit within Emacs as well, omit the 'nopass'.
242 243
See the GDB manual for more details about signal handling and the
'handle' command.
244

245
A technique that can work when 'handle SIGINT' does not is to store
246
the code for some character into the variable stop_character.  Thus,
Dave Love's avatar
#  
Dave Love committed
247 248 249 250 251

    set stop_character = 29

makes Control-] (decimal code 29) the stop character.
Typing Control-] will cause immediate stop.  You cannot
252 253 254
use the set command until the inferior process has been started, so
start Emacs with the 'start' command, to get an opportunity to do the
above 'set' command.
Dave Love's avatar
#  
Dave Love committed
255

256 257
On a Posix host, you can also send a signal using the 'kill' command
from a shell prompt, like this:
Eli Zaretskii's avatar
Eli Zaretskii committed
258

259
   kill -TSTP Emacs-PID
260

261 262 263 264 265 266 267 268 269 270 271 272
where Emacs-PID is the process ID of Emacs being debugged.  Other
useful signals to send are SIGUSR1 and SIGUSR2; see "Error Debugging"
in the ELisp manual for how to use those.

When Emacs is displaying on a text terminal, it is useful to have a
separate terminal for the debug session.  This can be done by starting
Emacs as usual, then attaching to it from gdb with the 'attach'
command which is explained in the node "Attach" of the GDB manual.

On MS-Windows, you can alternatively start Emacs from its own separate
console by setting the new-console option before running Emacs under
GDB:
Eli Zaretskii's avatar
Eli Zaretskii committed
273 274 275 276

  (gdb) set new-console 1
  (gdb) run

277 278 279 280 281 282
If you do this, then typing C-c or C-BREAK into the console window
through which you interact with GDB will stop Emacs and return control
to the debugger, no matter if Emacs displays GUI or text-mode frames.
This is the only reliable alternative on MS-Windows to get control to
the debugger, besides setting breakpoints in advance.

Dave Love's avatar
#  
Dave Love committed
283 284 285
** Examining Lisp object values.

When you have a live process to debug, and it has not encountered a
286 287
fatal error, you can use the GDB command 'pr'.  First print the value
in the ordinary way, with the 'p' command.  Then type 'pr' with no
Dave Love's avatar
#  
Dave Love committed
288 289
arguments.  This calls a subroutine which uses the Lisp printer.

290
You can also use 'pp value' to print the emacs value directly.
Kim F. Storm's avatar
Kim F. Storm committed
291

292
To see the current value of a Lisp Variable, use 'pv variable'.
293

294 295 296
These commands send their output to stderr; if that is closed or
redirected to some file you don't know, you won't see their output.
This is particularly so for Emacs invoked on MS-Windows from the
297 298
desktop shortcut.  You can use the command 'redirect-debugging-output'
to redirect stderr to a file.
299

300
Note: It is not a good idea to try 'pr', 'pp', or 'pv' if you know that Emacs
Kim F. Storm's avatar
Kim F. Storm committed
301
is in deep trouble: its stack smashed (e.g., if it encountered SIGSEGV
302 303
due to stack overflow), or crucial data structures, such as 'obarray',
corrupted, etc.  In such cases, the Emacs subroutine called by 'pr'
304 305 306
might make more damage, like overwrite some data that is important for
debugging the original problem.

307 308
Also, on some systems it is impossible to use 'pr' if you stopped
Emacs while it was inside 'select'.  This is in fact what happens if
309
you stop Emacs while it is waiting.  In such a situation, don't try to
310 311
use 'pr'.  Instead, use 's' to step out of the system call.  Then
Emacs will be between instructions and capable of handling 'pr'.
Dave Love's avatar
#  
Dave Love committed
312

313 314
If you can't use 'pr' command, for whatever reason, you can use the
'xpr' command to print out the data type and value of the last data
315 316 317 318 319 320
value, For example:

    p it->object
    xpr

You may also analyze data values using lower-level commands.  Use the
321
'xtype' command to print out the data type of the last data value.
322 323
Once you know the data type, use the command that corresponds to that
type.  Here are these commands:
Dave Love's avatar
#  
Dave Love committed
324 325 326 327

    xint xptr xwindow xmarker xoverlay xmiscfree xintfwd xboolfwd xobjfwd
    xbufobjfwd xkbobjfwd xbuflocal xbuffer xsymbol xstring xvector xframe
    xwinconfig xcompiled xcons xcar xcdr xsubr xprocess xfloat xscrollbar
Eli Zaretskii's avatar
Eli Zaretskii committed
328
    xchartable xsubchartable xboolvector xhashtable xlist xcoding
329
    xcharset xfontset xfont
Dave Love's avatar
#  
Dave Love committed
330 331 332 333 334 335 336 337 338

Each one of them applies to a certain type or class of types.
(Some of these types are not visible in Lisp, because they exist only
internally.)

Each x... command prints some information about the value, and
produces a GDB value (subsequently available in $) through which you
can get at the rest of the contents.

339
In general, most of the rest of the contents will be additional Lisp
Dave Love's avatar
#  
Dave Love committed
340 341
objects which you can examine in turn with the x... commands.

342 343 344
Even with a live process, these x...  commands are useful for
examining the fields in a buffer, window, process, frame or marker.
Here's an example using concepts explained in the node "Value History"
345 346
of the GDB manual to print values associated with the variable
called frame.  First, use these commands:
347

Eli Zaretskii's avatar
Eli Zaretskii committed
348 349 350 351
  cd src
  gdb emacs
  b set_frame_buffer_list
  r -q
352

353
Then Emacs hits the breakpoint:
354

Eli Zaretskii's avatar
Eli Zaretskii committed
355 356 357 358 359 360 361 362 363 364 365 366 367 368
  (gdb) p frame
  $1 = 139854428
  (gdb) xpr
  Lisp_Vectorlike
  PVEC_FRAME
  $2 = (struct frame *) 0x8560258
  "emacs@localhost"
  (gdb) p *$
  $3 = {
    size = 1073742931,
    next = 0x85dfe58,
    name = 140615219,
    [...]
  }
369

370
Now we can use 'pp' to print the frame parameters:
371

Eli Zaretskii's avatar
Eli Zaretskii committed
372 373
  (gdb) pp $->param_alist
  ((background-mode . light) (display-type . color) [...])
374 375 376

The Emacs C code heavily uses macros defined in lisp.h.  So suppose
we want the address of the l-value expression near the bottom of
377
'add_command_key' from keyboard.c:
378 379 380

  XVECTOR (this_command_keys)->contents[this_command_key_count++] = key;

381 382
XVECTOR is a macro, so GDB only knows about it if Emacs has been compiled with
preprocessor macro information.  GCC provides this if you specify the options
383
'-gdwarf-N' (where N is 2 or higher) and '-g3'.  In this case, GDB can
Eli Zaretskii's avatar
Eli Zaretskii committed
384
evaluate expressions like "p XVECTOR (this_command_keys)".
385

386 387
When this information isn't available, you can use the xvector command in GDB
to get the same result.  Here is how:
388

Eli Zaretskii's avatar
Eli Zaretskii committed
389 390 391 392 393 394 395 396 397
  (gdb) p this_command_keys
  $1 = 1078005760
  (gdb) xvector
  $2 = (struct Lisp_Vector *) 0x411000
  0
  (gdb) p $->contents[this_command_key_count]
  $3 = 1077872640
  (gdb) p &$
  $4 = (int *) 0x411008
398

399 400
Here's a related example of macros and the GDB 'define' command.
There are many Lisp vectors such as 'recent_keys', which contains the
401
last 300 keystrokes.  We can print this Lisp vector
402

Eli Zaretskii's avatar
Eli Zaretskii committed
403 404
  p recent_keys
  pr
405

406 407 408
But this may be inconvenient, since 'recent_keys' is much more verbose
than 'C-h l'.  We might want to print only the last 10 elements of
this vector.  'recent_keys' is updated in keyboard.c by the command
409 410 411

  XVECTOR (recent_keys)->contents[recent_keys_index] = c;

412
So we define a GDB command 'xvector-elts', so the last 10 keystrokes
413
are printed by
414

Eli Zaretskii's avatar
Eli Zaretskii committed
415
  xvector-elts recent_keys recent_keys_index 10
416 417 418

where you can define xvector-elts as follows:

Eli Zaretskii's avatar
Eli Zaretskii committed
419 420 421 422 423 424 425 426 427 428 429 430
  define xvector-elts
  set $i = 0
  p $arg0
  xvector
  set $foo = $
  while $i < $arg2
  p $foo->contents[$arg1-($i++)]
  pr
  end
  document xvector-elts
  Prints a range of elements of a Lisp vector.
  xvector-elts  v n i
431
  prints 'i' elements of the vector 'v' ending at the index 'n'.
Eli Zaretskii's avatar
Eli Zaretskii committed
432
  end
433 434 435

** Getting Lisp-level backtrace information within GDB

436
The most convenient way is to use the 'xbacktrace' command.  This
437
shows the names of the Lisp functions that are currently active.
438

439
If that doesn't work (e.g., because the 'backtrace_list' structure is
440 441 442 443 444 445 446 447 448 449 450 451 452
corrupted), type "bt" at the GDB prompt, to produce the C-level
backtrace, and look for stack frames that call Ffuncall.  Select them
one by one in GDB, by typing "up N", where N is the appropriate number
of frames to go up, and in each frame that calls Ffuncall type this:

   p *args
   pr

This will print the name of the Lisp function called by that level
of function calling.

By printing the remaining elements of args, you can see the argument
values.  Here's how to print the first argument:
453

454 455 456 457 458 459 460 461 462 463 464 465
   p args[1]
   pr

If you do not have a live process, you can use xtype and the other
x...  commands such as xsymbol to get such information, albeit less
conveniently.  For example:

   p *args
   xtype

and, assuming that "xtype" says that args[0] is a symbol:

466
   xsymbol
467

Eli Zaretskii's avatar
Eli Zaretskii committed
468 469 470 471 472 473 474 475 476 477 478 479
** Debugging Emacs redisplay problems

If you configured Emacs with --enable-checking='glyphs', you can use redisplay
tracing facilities from a running Emacs session.

The command "M-x trace-redisplay RET" will produce a trace of what redisplay
does on the standard error stream.  This is very useful for understanding the
code paths taken by the display engine under various conditions, especially if
some redisplay optimizations produce wrong results.  (You know that redisplay
optimizations might be involved if "M-x redraw-display RET", or even just
typing "M-x", causes Emacs to correct the bad display.)  Since the cursor
blinking feature triggers periodic redisplay cycles, we recommend disabling
480
'blink-cursor-mode' before invoking 'trace-redisplay', so that you have less
Eli Zaretskii's avatar
Eli Zaretskii committed
481
clutter in the trace.  You can also have up to 30 last trace messages dumped to
482
standard error by invoking the 'dump-redisplay-history' command.
Eli Zaretskii's avatar
Eli Zaretskii committed
483 484 485 486

To find the code paths which were taken by the display engine, search xdisp.c
for the trace messages you see.

487
The command 'dump-glyph-matrix' is useful for producing on standard error
Eli Zaretskii's avatar
Eli Zaretskii committed
488 489
stream a full dump of the selected window's glyph matrix.  See the function's
doc string for more details.  If you are debugging redisplay issues in
490
text-mode frames, you may find the command 'dump-frame-glyph-matrix' useful.
Eli Zaretskii's avatar
Eli Zaretskii committed
491

492 493
Other commands useful for debugging redisplay are 'dump-glyph-row' and
'dump-tool-bar-row'.
Eli Zaretskii's avatar
Eli Zaretskii committed
494 495 496 497

If you run Emacs under GDB, you can print the contents of any glyph matrix by
just calling that function with the matrix as its argument.  For example, the
following command will print the contents of the current matrix of the window
498
whose pointer is in 'w':
Eli Zaretskii's avatar
Eli Zaretskii committed
499 500 501 502 503 504 505 506 507 508 509

  (gdb) p dump_glyph_matrix (w->current_matrix, 2)

(The second argument 2 tells dump_glyph_matrix to print the glyphs in
a long form.)

The Emacs display code includes special debugging code, but it is normally
disabled.  Configuring Emacs with --enable-checking='yes,glyphs' enables it.

Building Emacs like that activates many assertions which scrutinize
display code operation more than Emacs does normally.  (To see the
510
code which tests these assertions, look for calls to the 'eassert'
Eli Zaretskii's avatar
Eli Zaretskii committed
511 512 513
macros.)  Any assertion that is reported to fail should be investigated.

When you debug display problems running emacs under X, you can use
514
the 'ff' command to flush all pending display updates to the screen.
515 516 517 518

The src/.gdbinit file defines many useful commands for dumping redisplay
related data structures in a terse and user-friendly format:

519 520 521 522 523 524 525 526
 'ppt' prints value of PT, narrowing, and gap in current buffer.
 'pit' dumps the current display iterator 'it'.
 'pwin' dumps the current window 'win'.
 'prow' dumps the current glyph_row 'row'.
 'pg' dumps the current glyph 'glyph'.
 'pgi' dumps the next glyph.
 'pgrow' dumps all glyphs in current glyph_row 'row'.
 'pcursor' dumps current output_cursor.
527

528 529 530
The above commands also exist in a version with an 'x' suffix which takes an
object of the relevant type as argument.  For example, 'pgrowx' dumps all
glyphs in its argument, which must be of type 'struct glyph_row'.
Eli Zaretskii's avatar
Eli Zaretskii committed
531 532 533 534 535

Since redisplay is performed by Emacs very frequently, you need to place your
breakpoints cleverly to avoid hitting them all the time, when the issue you are
debugging did not (yet) happen.  Here are some useful techniques for that:

536
 . Put a breakpoint at 'Fredraw_display' before running Emacs.  Then do
Eli Zaretskii's avatar
Eli Zaretskii committed
537 538 539 540 541 542
   whatever is required to reproduce the bad display, and invoke "M-x
   redraw-display".  The debugger will kick in, and you can set or enable
   breakpoints in strategic places, knowing that the bad display will be
   redrawn from scratch.

 . For debugging incorrect cursor position, a good place to put a breakpoint is
543 544
   in 'set_cursor_from_row'.  The first time this function is called as part of
   'redraw-display', Emacs is redrawing the minibuffer window, which is usually
Eli Zaretskii's avatar
Eli Zaretskii committed
545
   not what you want; type "continue" to get to the call you want.  In general,
546
   always make sure 'set_cursor_from_row' is called for the right window and
Eli Zaretskii's avatar
Eli Zaretskii committed
547 548 549
   buffer by examining the value of w->contents: it should be the buffer whose
   display you are debugging.

550 551
 . 'set_cursor_from_row' is also a good place to look at the contents of a
   screen line (a.k.a. "glyph row"), by means of the 'pgrow' GDB command.  Of
Eli Zaretskii's avatar
Eli Zaretskii committed
552
   course, you need first to make sure the cursor is on the screen line which
553 554
   you want to investigate.  If you have set a breakpoint in 'Fredraw_display',
   as advised above, move cursor to that line before invoking 'redraw-display'.
Eli Zaretskii's avatar
Eli Zaretskii committed
555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571

 . If the problem happens only at some specific buffer position or for some
   specific rarely-used character, you can make your breakpoints conditional on
   those values.  The display engine maintains the buffer and string position
   it is processing in the it->current member; for example, the buffer
   character position is in it->current.pos.charpos.  Most redisplay functions
   accept a pointer to a 'struct it' object as their argument, so you can make
   conditional breakpoints in those functions, like this:

    (gdb) break x_produce_glyphs if it->current.pos.charpos == 1234

   For conditioning on the character being displayed, use it->c or
   it->char_to_display.

 . You can also make the breakpoints conditional on what object is being used
   for producing glyphs for display.  The it->method member has the value
   GET_FROM_BUFFER for displaying buffer contents, GET_FROM_STRING for
572 573
   displaying a Lisp string (e.g., a 'display' property or an overlay string),
   GET_FROM_IMAGE for displaying an image, etc.  See 'enum it_method' in
Eli Zaretskii's avatar
Eli Zaretskii committed
574
   dispextern.h for the full list of values.
575

576 577 578
** Following longjmp call.

Recent versions of glibc (2.4+?) encrypt stored values for setjmp/longjmp which
579
prevents GDB from being able to follow a longjmp call using 'next'.  To
580 581 582
disable this protection you need to set the environment variable
LD_POINTER_GUARD to 0.

583 584 585 586 587 588
** Using GDB in Emacs

Debugging with GDB in Emacs offers some advantages over the command line (See
the GDB Graphical Interface node of the Emacs manual).  There are also some
features available just for debugging Emacs:

589
1) The command gud-print is available on the tool bar (the 'p' icon) and
590 591
   allows the user to print the s-expression of the variable at point,
   in the GUD buffer.
592

593
2) Pressing 'p' on a component of a watch expression that is a lisp object
594 595
   in the speedbar prints its s-expression in the GUD buffer.

596 597 598
3) The STOP button on the tool bar and the Signals->STOP menu-bar menu
   item are adjusted so that they send SIGTSTP instead of the usual
   SIGINT.
599 600 601 602

4) The command gud-pv has the global binding 'C-x C-a C-v' and prints the
   value of the lisp variable at point.

603 604
** Debugging what happens while preloading and dumping Emacs

605 606 607
Debugging 'temacs' is useful when you want to establish whether a
problem happens in an undumped Emacs.  To run 'temacs' under a
debugger, type "gdb temacs", then start it with 'r -batch -l loadup'.
Eli Zaretskii's avatar
Eli Zaretskii committed
608

609
If you need to debug what happens during dumping, start it with 'r -batch -l
Eli Zaretskii's avatar
Eli Zaretskii committed
610 611
loadup dump' instead.  For debugging the bootstrap dumping, use "loadup
bootstrap" instead of "loadup dump".
612 613 614 615 616 617 618

If temacs actually succeeds when running under GDB in this way, do not
try to run the dumped Emacs, because it was dumped with the GDB
breakpoints in it.

** If you encounter X protocol errors

619 620 621 622 623 624 625 626 627
The X server normally reports protocol errors asynchronously,
so you find out about them long after the primitive which caused
the error has returned.

To get clear information about the cause of an error, try evaluating
(x-synchronize t).  That puts Emacs into synchronous mode, where each
Xlib call checks for errors before it returns.  This mode is much
slower, but when you get an error, you will see exactly which call
really caused the error.
628

629 630 631
You can start Emacs in a synchronous mode by invoking it with the -xrm
option, like this:

632
    emacs -xrm "emacs.synchronous: true"
633

634
Setting a breakpoint in the function 'x_error_quitter' and looking at
635 636 637
the backtrace when Emacs stops inside that function will show what
code causes the X protocol errors.

638 639 640 641 642
Note that the -xrm option may have no effect when you make an Emacs
process invoked with the -nw option a server and want to trace X
protocol errors from subsequent invocations of emacsclient in a GUI
frame.  In that case calling the initial Emacs via

643
  emacs -nw --eval '(setq x-command-line-resources "emacs.synchronous: true")'
644 645 646

should give more reliable results.

647 648 649 650 651 652 653 654
For X protocol errors related to displaying unusual characters or to
font-related customizations, try invoking Emacs like this:

  XFT_DEBUG=16 emacs -xrm "emacs.synchronous: true"

This should produce information from the libXft library which could
give useful hints regarding font-related problems in that library.

655 656 657 658 659 660 661
Some bugs related to the X protocol disappear when Emacs runs in a
synchronous mode.  To track down those bugs, we suggest the following
procedure:

  - Run Emacs under a debugger and put a breakpoint inside the
    primitive function which, when called from Lisp, triggers the X
    protocol errors.  For example, if the errors happen when you
662
    delete a frame, put a breakpoint inside 'Fdelete_frame'.
663 664 665 666 667

  - When the breakpoint breaks, step through the code, looking for
    calls to X functions (the ones whose names begin with "X" or
    "Xt" or "Xm").

668
  - Insert calls to 'XSync' before and after each call to the X
669 670 671 672
    functions, like this:

       XSync (f->output_data.x->display_info->display, 0);

673
    where 'f' is the pointer to the 'struct frame' of the selected
674 675
    frame, normally available via XFRAME (selected_frame).  (Most
    functions which call X already have some variable that holds the
676
    pointer to the frame, perhaps called 'f' or 'sf', so you shouldn't
677 678 679
    need to compute it.)

    If your debugger can call functions in the program being debugged,
680
    you should be able to issue the calls to 'XSync' without recompiling
681 682 683 684 685 686 687 688 689 690
    Emacs.  For example, with GDB, just type:

       call XSync (f->output_data.x->display_info->display, 0)

    before and immediately after the suspect X calls.  If your
    debugger does not support this, you will need to add these pairs
    of calls in the source and rebuild Emacs.

    Either way, systematically step through the code and issue these
    calls until you find the first X function called by Emacs after
691 692
    which a call to 'XSync' winds up in the function
    'x_error_quitter'.  The first X function call for which this
693 694 695 696 697
    happens is the one that generated the X protocol error.

  - You should now look around this offending X call and try to figure
    out what is wrong with it.

Jan Djärv's avatar
Jan Djärv committed
698 699 700
** If Emacs causes errors or memory leaks in your X server

You can trace the traffic between Emacs and your X server with a tool
701
like xmon.
Jan Djärv's avatar
Jan Djärv committed
702 703 704 705 706 707 708 709 710

Xmon can be used to see exactly what Emacs sends when X protocol errors
happen.  If Emacs causes the X server memory usage to increase you can
use xmon to see what items Emacs creates in the server (windows,
graphical contexts, pixmaps) and what items Emacs delete.  If there
are consistently more creations than deletions, the type of item
and the activity you do when the items get created can give a hint where
to start debugging.

711 712
** If the symptom of the bug is that Emacs fails to respond

713
Don't assume Emacs is 'hung'--it may instead be in an infinite loop.
714 715
To find out which, make the problem happen under GDB and stop Emacs
once it is not responding.  (If Emacs is using X Windows directly, you
716 717 718 719 720
can stop Emacs by typing C-z at the GDB job.  On MS-Windows, run Emacs
as usual, and then attach GDB to it -- that will usually interrupt
whatever Emacs is doing and let you perform the steps described
below.)

721 722
Then try stepping with 'step'.  If Emacs is hung, the 'step' command
won't return.  If it is looping, 'step' will return.
723 724 725 726 727 728 729 730

If this shows Emacs is hung in a system call, stop it again and
examine the arguments of the call.  If you report the bug, it is very
important to state exactly where in the source the system call is, and
what the arguments are.

If Emacs is in an infinite loop, try to determine where the loop
starts and ends.  The easiest way to do this is to use the GDB command
731 732
'finish'.  Each time you use it, Emacs resumes execution until it
exits one stack frame.  Keep typing 'finish' until it doesn't
733 734 735
return--that means the infinite loop is in the stack frame which you
just tried to finish.

736 737
Stop Emacs again, and use 'finish' repeatedly again until you get back
to that frame.  Then use 'next' to step through that frame.  By
738 739 740 741
stepping, you will see where the loop starts and ends.  Also, examine
the data being used in the loop and try to determine why the loop does
not exit when it should.

742
On GNU and Unix systems, you can also trying sending Emacs SIGUSR2,
743
which, if 'debug-on-event' has its default value, will cause Emacs to
744
attempt to break it out of its current loop and into the Lisp
745 746 747
debugger.  (See the node "Debugging" in the ELisp manual for the
details about the Lisp debugger.)  This feature is useful when a
C-level debugger is not conveniently available.
Daniel Colascione's avatar
Daniel Colascione committed
748

749 750 751 752 753 754 755 756 757 758 759 760 761 762 763
** If certain operations in Emacs are slower than they used to be, here
is some advice for how to find out why.

Stop Emacs repeatedly during the slow operation, and make a backtrace
each time.  Compare the backtraces looking for a pattern--a specific
function that shows up more often than you'd expect.

If you don't see a pattern in the C backtraces, get some Lisp
backtrace information by typing "xbacktrace" or by looking at Ffuncall
frames (see above), and again look for a pattern.

When using X, you can stop Emacs at any time by typing C-z at GDB.
When not using X, you can do this with C-g.  On non-Unix platforms,
such as MS-DOS, you might need to press C-BREAK instead.

Dave Love's avatar
#  
Dave Love committed
764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779
** If GDB does not run and your debuggers can't load Emacs.

On some systems, no debugger can load Emacs with a symbol table,
perhaps because they all have fixed limits on the number of symbols
and Emacs exceeds the limits.  Here is a method that can be used
in such an extremity.  Do

    nm -n temacs > nmout
    strip temacs
    adb temacs
    0xd:i
    0xe:i
    14:i
    17:i
    :r -l loadup   (or whatever)

780
It is necessary to refer to the file 'nmout' to convert
Dave Love's avatar
#  
Dave Love committed
781 782 783
numeric addresses into symbols and vice versa.

It is useful to be running under a window system.
784 785 786
Then, if Emacs becomes hopelessly wedged, you can create another
window to do kill -9 in.  kill -ILL is often useful too, since that
may make Emacs dump core or return to adb.
Dave Love's avatar
#  
Dave Love committed
787

Eli Zaretskii's avatar
Eli Zaretskii committed
788
** Debugging incorrect screen updating on a text terminal.
Dave Love's avatar
#  
Dave Love committed
789 790 791 792 793 794 795 796 797 798

To debug Emacs problems that update the screen wrong, it is useful
to have a record of what input you typed and what Emacs sent to the
screen.  To make these records, do

(open-dribble-file "~/.dribble")
(open-termscript "~/.termscript")

The dribble file contains all characters read by Emacs from the
terminal, and the termscript file contains all characters it sent to
799
the terminal.  The use of the directory '~/' prevents interference
Dave Love's avatar
#  
Dave Love committed
800 801 802 803 804 805
with any other user.

If you have irreproducible display problems, put those two expressions
in your ~/.emacs file.  When the problem happens, exit the Emacs that
you were running, kill it, and rename the two files.  Then you can start
another Emacs without clobbering those files, and use it to examine them.
806 807

An easy way to see if too much text is being redrawn on a terminal is to
808
evaluate '(setq inverse-video t)' before you try the operation you think
809 810
will cause too much redrawing.  This doesn't refresh the screen, so only
newly drawn text is in inverse video.
811 812 813 814 815

** Debugging LessTif

If you encounter bugs whereby Emacs built with LessTif grabs all mouse
and keyboard events, or LessTif menus behave weirdly, it might be
816
helpful to set the 'DEBUGSOURCES' and 'DEBUG_FILE' environment
817 818
variables, so that one can see what LessTif was doing at this point.
For instance
819

820
  export DEBUGSOURCES="RowColumn.c:MenuShell.c:MenuUtil.c"
821
  export DEBUG_FILE=/usr/tmp/LESSTIF_TRACE
822
  emacs &
823 824

causes LessTif to print traces from the three named source files to a
825
file in '/usr/tmp' (that file can get pretty large).  The above should
826 827
be typed at the shell prompt before invoking Emacs, as shown by the
last line above.
828 829 830 831 832 833 834 835

Running GDB from another terminal could also help with such problems.
You can arrange for GDB to run on one machine, with the Emacs display
appearing on another.  Then, when the bug happens, you can go back to
the machine where you started GDB and use the debugger from there.

** Debugging problems which happen in GC

836
The array 'last_marked' (defined on alloc.c) can be used to display up
837
to the 512 most-recent objects marked by the garbage collection process.
838
Whenever the garbage collector marks a Lisp object, it records the
839 840 841
pointer to that object in the 'last_marked' array, which is maintained
as a circular buffer.  The variable 'last_marked_index' holds the
index into the 'last_marked' array one place beyond where the pointer
842
to the very last marked object is stored.
843 844 845 846

The single most important goal in debugging GC problems is to find the
Lisp data structure that got corrupted.  This is not easy since GC
changes the tag bits and relocates strings which make it hard to look
847
at Lisp objects with commands such as 'pr'.  It is sometimes necessary
848 849
to convert Lisp_Object variables into pointers to C struct's manually.

850
Use the 'last_marked' array and the source to reconstruct the sequence
851
that objects were marked.  In general, you need to correlate the
852
values recorded in the 'last_marked' array with the corresponding
853
stack frames in the backtrace, beginning with the innermost frame.
854
Some subroutines of 'mark_object' are invoked recursively, others loop
855 856
over portions of the data structure and mark them as they go.  By
looking at the code of those routines and comparing the frames in the
857 858
backtrace with the values in 'last_marked', you will be able to find
connections between the values in 'last_marked'.  E.g., when GC finds
859 860 861 862 863 864 865 866 867 868
a cons cell, it recursively marks its car and its cdr.  Similar things
happen with properties of symbols, elements of vectors, etc.  Use
these connections to reconstruct the data structure that was being
marked, paying special attention to the strings and names of symbols
that you encounter: these strings and symbol names can be used to grep
the sources to find out what high-level symbols and global variables
are involved in the crash.

Once you discover the corrupted Lisp object or data structure, grep
the sources for its uses and try to figure out what could cause the
Glenn Morris's avatar
Glenn Morris committed
869
corruption.  If looking at the sources doesn't help, you could try
870 871 872 873 874 875 876
setting a watchpoint on the corrupted data, and see what code modifies
it in some invalid way.  (Obviously, this technique is only useful for
data that is modified only very rarely.)

It is also useful to look at the corrupted object or data structure in
a fresh Emacs session and compare its contents with a session that you
are debugging.
877

Eli Zaretskii's avatar
Eli Zaretskii committed
878 879 880 881 882 883 884 885 886
** Debugging problems with non-ASCII characters

If you experience problems which seem to be related to non-ASCII
characters, such as \201 characters appearing in the buffer or in your
files, set the variable byte-debug-flag to t.  This causes Emacs to do
some extra checks, such as look for broken relations between byte and
character positions in buffers and strings; the resulting diagnostics
might pinpoint the cause of the problem.

887 888 889 890 891 892 893 894 895 896 897 898
** Debugging the TTY (non-windowed) version

The most convenient method of debugging the character-terminal display
is to do that on a window system such as X.  Begin by starting an
xterm window, then type these commands inside that window:

  $ tty
  $ echo $TERM

Let's say these commands print "/dev/ttyp4" and "xterm", respectively.

Now start Emacs (the normal, windowed-display session, i.e. without
899
the '-nw' option), and invoke "M-x gdb RET emacs RET" from there.  Now
900 901 902 903 904 905 906 907 908
type these commands at GDB's prompt:

  (gdb) set args -nw -t /dev/ttyp4
  (gdb) set environment TERM xterm
  (gdb) run

The debugged Emacs should now start in no-window mode with its display
directed to the xterm window you opened above.

909
Similar arrangement is possible on a character terminal by using the
910
'screen' package.
911

912
On MS-Windows, you can start Emacs in its own separate console by
Eli Zaretskii's avatar
Eli Zaretskii committed
913 914 915 916 917
setting the new-console option before running Emacs under GDB:

  (gdb) set new-console 1
  (gdb) run

918
** Running Emacs with undefined-behavior sanitization
919

920 921
Building Emacs with undefined-behavior sanitization can help find
several kinds of low-level problems in C code, including:
922

923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939
  * Out-of-bounds access of many (but not all) arrays.
  * Signed integer overflow, e.g., (INT_MAX + 1).
  * Integer shifts by a negative or wider-than-word value.
  * Misaligned pointers and pointer overflow.
  * Loading a bool or enum value that is out of range for its type.
  * Passing NULL to or returning NULL from a function requiring nonnull.
  * Passing a size larger than the corresponding array to memcmp etc.
  * Passing invalid values to some builtin functions, e.g., __builtin_clz (0).
  * Reaching __builtin_unreachable calls (in Emacs, 'eassume' failure).

To use UndefinedBehaviorSanitizer with GCC and similar compilers,
append '-fsanitize=undefined' to CFLAGS, either when running
'configure' or running 'make'.  When supported, you can also specify
'bound-strict' and 'float-cast-overflow'.  For example:

  ./configure \
    CFLAGS='-O0 -g3 -fsanitize=undefined,bounds-strict,float-cast-overflow'
940

941 942
You may need to append '-static-libubsan' to CFLAGS if your version of
GCC is installed in an unusual location.
943

944 945
When using GDB to debug an executable with undefined-behavior
sanitization, the GDB command:
946

947
  (gdb) rbreak ^__ubsan_handle_
948

949 950 951
will let you gain control when an error is detected and before
UndefinedBehaviorSanitizer outputs to stderr or terminates the
program.
952

953
** Running Emacs with address sanitization
954

955
Building Emacs with address sanitization can help debug memory-use
956 957 958 959
problems, such as freeing the same object twice.  To use
AddressSanitizer with GCC and similar compilers, append
'-fsanitize=address' to CFLAGS, either when running 'configure' or
running 'make'.  Configure, build and run Emacs with
960 961
ASAN_OPTIONS='detect_leaks=0' in the environment to suppress
diagnostics of minor memory leaks in Emacs.  For example:
962

963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008
  export ASAN_OPTIONS='detect_leaks=0'
  ./configure CFLAGS='-O0 -g3 -fsanitize=address'
  make
  src/emacs

You may need to append '-static-libasan' to CFLAGS if your version of
GCC is installed in an unusual location.

When using GDB to debug an executable with address sanitization, the
GDB command:

  (gdb) rbreak ^__asan_report_

will let you gain control when an error is detected and before
AddressSanitizer outputs to stderr or terminates the program.

Address sanitization is incompatible with undefined-behavior
sanitization, unfortunately.  Address sanitization is also
incompatible with the --with-dumping=unexec option of 'configure'.

** Running Emacs under Valgrind

Valgrind <http://valgrind.org/> is free software that can be useful
when debugging low-level Emacs problems.  Unlike GCC sanitizers,
Valgrind does not need you to compile Emacs with special debugging
flags, so it can be helpful in investigating problems that vanish when
Emacs is recompiled with debugging enabled.  However, by default
Valgrind generates many false alarms with Emacs, and you will need to
maintain a suppressions file to suppress these false alarms and use
Valgrind effectively.  For example, you might invoke Valgrind this
way:

   valgrind --suppressions=valgrind.supp ./emacs

where valgrind.supp contains groups of lines like the following, which
suppresses some Valgrind false alarms during Emacs garbage collection:

   {
     Fgarbage_collect Cond - conservative garbage collection
     Memcheck:Cond
     ...
     fun:Fgarbage_collect
   }

Unfortunately Valgrind suppression files tend to be system-dependent,
so you will need to keep one around that matches your system.
1009

Eli Zaretskii's avatar
Eli Zaretskii committed
1010 1011 1012 1013 1014 1015 1016
** How to recover buffer contents from an Emacs core dump file

The file etc/emacs-buffer.gdb defines a set of GDB commands for
recovering the contents of Emacs buffers from a core dump file.  You
might also find those commands useful for displaying the list of
buffers in human-readable format from within the debugger.

1017 1018 1019

This file is part of GNU Emacs.

1020
GNU Emacs is free software: you can redistribute it and/or modify
1021
it under the terms of the GNU General Public License as published by
1022 1023
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
1024 1025 1026 1027 1028 1029 1030

GNU Emacs is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
1031
along with GNU Emacs.  If not, see <https://www.gnu.org/licenses/>.
1032

Nick Roberts's avatar
Nick Roberts committed
1033 1034 1035 1036 1037

Local variables:
mode: outline
paragraph-separate: "[ 	]*$"
end: