bug#6278: Segmentation fault at exit if server-mode

bug#6278: Segmentation fault at exit if server-mode

[Lennart Borgman]

I get a segmentation fault on w32 at Emacs exit if server-mode is
enabled and with -Q. This is with a trunk checkout from 2010-05-22.



*** Here is a backtrace and the last messages from gdb:


warning: BTMMHOOK 27.05.2010 05:01:44 Thread<0E68> Hook DLL loaded

[New Thread 3168.0xf98]
warning: BTMMHOOK 27.05.2010 05:01:52 Thread<0A24> Hook DLL unloaded


Program received signal SIGSEGV, Segmentation fault.
0x7c809eda in KERNEL32!IsBadReadPtr () from C:\WINDOWS\system32\kernel32.dll
(gdb) bt
#0  0x7c809eda in KERNEL32!IsBadReadPtr ()
   from C:\WINDOWS\system32\kernel32.dll
#1  0x5f180000 in ?? ()
#2  0x03105cce in ?? ()
   from C:\Program Files\Tall Emu\Online Armor\oawatch.dll
#3  0x030f33d7 in ?? ()
   from C:\Program Files\Tall Emu\Online Armor\oawatch.dll
#4  0x03108476 in ?? ()
   from C:\Program Files\Tall Emu\Online Armor\oawatch.dll
#5  0x0310949e in ?? ()
   from C:\Program Files\Tall Emu\Online Armor\oawatch.dll
#6  0x030f3ea0 in ?? ()
   from C:\Program Files\Tall Emu\Online Armor\oawatch.dll
#7  0x030f41d6 in ?? ()
   from C:\Program Files\Tall Emu\Online Armor\oawatch.dll
#8  0x7c90118a in ntdll!LdrSetAppCompatDllRedirectionCallback ()
   from C:\WINDOWS\system32\ntdll.dll
#9  0x030f0000 in ?? ()
#10 0x7c923ada in ntdll!RtlDestroyQueryDebugBuffer ()
   from C:\WINDOWS\system32\ntdll.dll
#11 0x7c81caae in KERNEL32!IsValidLocale ()
   from C:\WINDOWS\system32\kernel32.dll
#12 0x7c81cb26 in KERNEL32!ExitProcess ()
   from C:\WINDOWS\system32\kernel32.dll
#13 0x77c39d45 in strerror () from C:\WINDOWS\system32\msvcrt.dll
#14 0x77c39e78 in msvcrt!_initterm () from C:\WINDOWS\system32\msvcrt.dll
#15 0x77c39e90 in msvcrt!exit () from C:\WINDOWS\system32\msvcrt.dll
#16 0x01002ed6 in Fkill_emacs (arg=46135322) at emacs.c:2093
#17 0x01022d0f in Ffuncall (nargs=1, args=0x82f450) at eval.c:3073
#18 0x0116e455 in Fbyte_code (bytestr=19032417, vector=19032437, maxdepth=20)
    at bytecode.c:680
#19 0x0102344d in funcall_lambda (fun=19032389, nargs=1, arg_vector=0x82f694)
    at eval.c:3260
#20 0x01022f2c in Ffuncall (nargs=2, args=0x82f690) at eval.c:3119
#21 0x0116e455 in Fbyte_code (bytestr=19032649, vector=19032669, maxdepth=12)
    at bytecode.c:680
#22 0x0102344d in funcall_lambda (fun=19032621, nargs=1, arg_vector=0x82f914)
    at eval.c:3260
#23 0x01022f2c in Ffuncall (nargs=2, args=0x82f910) at eval.c:3119
#24 0x0116d94f in Fcall_interactively (function=47311594,
    record_flag=46135322, keys=46156549) at callint.c:869
#25 0x01022d63 in Ffuncall (nargs=4, args=0x82fb90) at eval.c:3079
#26 0x010228a3 in call3 (fn=46281234, arg1=47311594, arg2=46135322,
    arg3=46135322) at eval.c:2901
#27 0x01014299 in Fcommand_execute (cmd=47311594, record_flag=46135322,
    keys=46135322, special=46135322) at keyboard.c:10397
#28 0x010069d5 in command_loop_1 () at keyboard.c:1755
#29 0x010206bf in internal_condition_case (bfun=0x1006315 <command_loop_1>,
    handlers=46192882, hfun=0x1005d06 <cmd_error>) at eval.c:1510
#30 0x0100607a in command_loop_2 () at keyboard.c:1356
#31 0x010201b0 in internal_catch (tag=46191050,
    func=0x1006057 <command_loop_2>, arg=46135322) at eval.c:1246
#32 0x01006030 in command_loop () at keyboard.c:1335
#33 0x01005922 in recursive_edit_1 () at keyboard.c:950
#34 0x01005a86 in Frecursive_edit () at keyboard.c:1012
#35 0x0100284d in main (argc=2, argv=0xa94348) at emacs.c:1782

Lisp Backtrace:
"kill-emacs" (0x82f454)
"save-buffers-kill-emacs" (0x82f694)
"save-buffers-kill-terminal" (0x82f914)
"call-interactively" (0x82fb94)
(gdb)

bug#6278: Segmentation fault at exit if server-mode

[Eli Zaretskii]

> From: Lennart Borgman <lennart.borgman@gmail.com>
> Date: Thu, 27 May 2010 05:08:42 +0200
> Cc: 
> 
> I get a segmentation fault on w32 at Emacs exit if server-mode is
> enabled and with -Q. This is with a trunk checkout from 2010-05-22.

Do you mean to say that a recipe for reproducing this is

  emacs -Q
  M-: (server-start) RET
  C-x C-c

?  And that this happens consistently, every time you do the above
steps?

bug#6278: Segmentation fault at exit if server-mode

[Lennart Borgman]

On Thu, May 27, 2010 at 7:42 PM, Eli Zaretskii <eliz@gnu.org> wrote:
>> From: Lennart Borgman <lennart.borgman@gmail.com>
>> Date: Thu, 27 May 2010 05:08:42 +0200
>> Cc:
>>
>> I get a segmentation fault on w32 at Emacs exit if server-mode is
>> enabled and with -Q. This is with a trunk checkout from 2010-05-22.
>
> Do you mean to say that a recipe for reproducing this is
>
>  emacs -Q
>  M-: (server-start) RET
>  C-x C-c
>
> ?  And that this happens consistently, every time you do the above
> steps?

Yes, but there is no visible crash. It is just the debugger that says this.

bug#6278: Segmentation fault at exit if server-mode

[Lennart Borgman]

>> Do you mean to say that a recipe for reproducing this is
>>
>>  emacs -Q
>>  M-: (server-start) RET
>>  C-x C-c
>>
>> ?  And that this happens consistently, every time you do the above
>> steps?
>
> Yes, but there is no visible crash. It is just the debugger that says this.


It looks like this was the same problem as bug 6284. I do not see this
problem now that I am using the patch for 6284.

bug#6278: Segmentation fault at exit if server-mode

[Lennart Borgman]

On Fri, May 28, 2010 at 2:14 PM, Lennart Borgman
<lennart.borgman@gmail.com> wrote:
>
> It looks like this was the same problem as bug 6284. I do not see this
> problem now that I am using the patch for 6284.

Too optimistic ;-)

I just saw it again.

bug#6278: Segmentation fault at exit if server-mode

[Lennart Borgman]

On Sat, May 29, 2010 at 2:04 AM, Lennart Borgman
<lennart.borgman@gmail.com> wrote:
> On Fri, May 28, 2010 at 2:14 PM, Lennart Borgman
> <lennart.borgman@gmail.com> wrote:
>>
>> It looks like this was the same problem as bug 6284. I do not see this
>> problem now that I am using the patch for 6284.
>
> Too optimistic ;-)
>
> I just saw it again.

I just saw this in dbg. It might be related but I am not sure since I
believe I got an error from make-frame in this instance.

This error might of course have to do with the current bidi work. My
checkout is from 2010-05-22 (and patched).

#24 0x011d3277 in w32_abort () at w32fns.c:8183
#25 0x010b87d9 in check_glyph_memory () at dispnew.c:2610
#26 0x01002f8d in shut_down_emacs (sig=0, no_x=0, stuff=45467674)
    at emacs.c:2198
#27 0x01002eb4 in Fkill_emacs (arg=45467674) at emacs.c:2104

bug#6278: Segmentation fault at exit if server-mode

[Eli Zaretskii]

> From: Lennart Borgman <lennart.borgman@gmail.com>
> Date: Sat, 29 May 2010 03:46:02 +0200
> Cc: 6278@debbugs.gnu.org
> 
> I just saw this in dbg. It might be related but I am not sure since I
> believe I got an error from make-frame in this instance.
> 
> This error might of course have to do with the current bidi work.

It's hard to do anything about such a crash without having a clear
recipe for reproducing it, starting with "emacs -Q".  Whatever problem
caused it, happened before you invoked kill-emacs, and in a different
place.

> My checkout is from 2010-05-22 (and patched).
> 
> #24 0x011d3277 in w32_abort () at w32fns.c:8183
> #25 0x010b87d9 in check_glyph_memory () at dispnew.c:2610
> #26 0x01002f8d in shut_down_emacs (sig=0, no_x=0, stuff=45467674)
>     at emacs.c:2198
> #27 0x01002eb4 in Fkill_emacs (arg=45467674) at emacs.c:2104

Please show the function check_glyph_memory in your sources, with line
numbers.  check_glyph_memory can abort in two different places, and
your sources are different from the current trunk, so it's hard to
know which one actually aborted.

Also, is this an optimized build or non-optimized one?  If the former,
the line numbers shown by the debugger cannot be trusted.

Finally, what are the values of glyph_matrix_count and
glyph_pool_count in frame #25?

bug#6278: Segmentation fault at exit if server-mode

[Lennart Borgman]

On Sat, May 29, 2010 at 8:32 AM, Eli Zaretskii <eliz@gnu.org> wrote:
>> From: Lennart Borgman <lennart.borgman@gmail.com>
>> Date: Sat, 29 May 2010 03:46:02 +0200
>> Cc: 6278@debbugs.gnu.org
>>
>> I just saw this in dbg. It might be related but I am not sure since I
>> believe I got an error from make-frame in this instance.
>>
>> This error might of course have to do with the current bidi work.
>
> It's hard to do anything about such a crash without having a clear
> recipe for reproducing it, starting with "emacs -Q".  Whatever problem
> caused it, happened before you invoked kill-emacs, and in a different
> place.

Yes, of course, but I have no better information.

>> My checkout is from 2010-05-22 (and patched).
>>
>> #24 0x011d3277 in w32_abort () at w32fns.c:8183
>> #25 0x010b87d9 in check_glyph_memory () at dispnew.c:2610
>> #26 0x01002f8d in shut_down_emacs (sig=0, no_x=0, stuff=45467674)
>>     at emacs.c:2198
>> #27 0x01002eb4 in Fkill_emacs (arg=45467674) at emacs.c:2104
>
> Please show the function check_glyph_memory in your sources, with line
> numbers.  check_glyph_memory can abort in two different places, and
> your sources are different from the current trunk, so it's hard to
> know which one actually aborted.

It aborts here (line 2610):

  if (glyph_matrix_count)
    abort ();

> Also, is this an optimized build or non-optimized one?

non-optimized.

> Finally, what are the values of glyph_matrix_count and
> glyph_pool_count in frame #25?

Sorry, I had to reboot. If this happens again I will look at it.

bug#6278: Segmentation fault at exit if server-mode

[Chong Yidong]

Since this bug report is for a modified Emacs, and no one else has
reported anything similar, and no additional information has been
forthcoming, I am closing it.