another unhelpful backtrace from redisplay-internal

another unhelpful backtrace from redisplay-internal

[Madhu]

I have a sure recipe for a crash in recent emacs but the configuration
is complex and I'm afraid I don't have a simple test case. I'm posting
the initial backtrace here in case anyone can suggest things I should
look for, or some ways I could narrow it down.


Thread 1 "emacs" hit Breakpoint 1, terminate_due_to_signal
(sig=sig@entry=6,
    backtrace_limit=backtrace_limit@entry=40) at ../../src/emacs.c:371
371       signal (sig, SIG_DFL);
(gdb) back
#0  terminate_due_to_signal (sig=sig@entry=6,
    backtrace_limit=backtrace_limit@entry=40) at ../../src/emacs.c:371
#1  0x000000000041f740 in emacs_abort () at ../../src/sysdep.c:2448
#2  0x000000000041d051 in select_window (window=XIL(0x1165305),
    norecord=XIL(0x30), inhibit_point_swap=<optimized out>)
    at ../../src/lisp.h:780
#3  0x0000000000458f20 in gui_consider_frame_title (
    frame=frame@entry=XIL(0x1123255)) at ../../src/lisp.h:1048
#4  0x0000000000472394 in redisplay_window (window=XIL(0x1165305),
    just_this_one_p=<optimized out>) at ../../src/xdisp.c:18940
#5  0x0000000000476b9b in redisplay_window_0 (
    window=window@entry=XIL(0x1165305)) at ../../src/xdisp.c:16179
#6  0x000000000056c114 in internal_condition_case_1 (
    bfun=bfun@entry=0x476b70 <redisplay_window_0>,
    arg=arg@entry=XIL(0x1165305), handlers=<optimized out>,
    hfun=hfun@entry=0x43a0d0 <redisplay_window_error>) at
    ../../src/eval.c:1379
#7  0x0000000000443554 in redisplay_windows (window=XIL(0x1165305))
    at ../../src/xdisp.c:16159
#8  0x0000000000462e1d in redisplay_internal () at
../../src/xdisp.c:15627
#9  0x00000000005040f7 in read_char (commandflag=1, map=XIL(0x2bca3b3),
    prev_event=XIL(0), used_mouse_menu=0x7fffffffd9cb, end_time=0x0)
    at ../../src/keyboard.c:2493

#10 0x0000000000506a6e in read_key_sequence (keybuf=<optimized out>,
    prompt=XIL(0), dont_downcase_last=<optimized out>,
    can_return_switch_frame=true, fix_current_buffer=true,
    prevent_redisplay=<optimized out>) at ../../src/keyboard.c:9549
#11 0x00000000005080ee in command_loop_1 () at ../../src/lisp.h:1048
#12 0x000000000056c087 in internal_condition_case (
    bfun=bfun@entry=0x507f10 <command_loop_1>,
    handlers=handlers@entry=XIL(0x90), hfun=hfun@entry=0x4ff120
    <cmd_error>)
    at ../../src/eval.c:1355
#13 0x00000000004f9f2c in command_loop_2 (ignore=ignore@entry=XIL(0))
    at ../../src/lisp.h:1048
#14 0x000000000056bfe1 in internal_catch (tag=tag@entry=XIL(0xd4d0),
    func=func@entry=0x4f9f10 <command_loop_2>, arg=arg@entry=XIL(0))
    at ../../src/eval.c:1116
#15 0x00000000004f9ed4 in command_loop () at ../../src/lisp.h:1048
#16 0x00000000004fed36 in recursive_edit_1 () at
../../src/keyboard.c:714
#17 0x00000000004ff060 in Frecursive_edit () at ../../src/keyboard.c:786
#18 0x00000000004259c9 in main (argc=7, argv=<optimized out>)
    at ../../src/emacs.c:2059

Lisp Backtrace:
"redisplay_internal (C function)" (0x0)


The crash happens when quitting a sly-db-mode buffer in its own window
which has popped up in a frame.  The window being redisplayed is the
window which is selected after the sly-db window quits.

Emacs is compiled with -g3 -O2, and I'm not sure if all the info to
debug this is available.

Any suggestions please?
--Madhu


(gdb) up
#1  0x000000000041f740 in emacs_abort () at ../../src/sysdep.c:2448
2448      terminate_due_to_signal (SIGABRT, 40);
(gdb) up
#2  0x000000000041d051 in select_window (window=XIL(0x1165305),
    norecord=XIL(0x30), inhibit_point_swap=<optimized out>)
    at ../../src/lisp.h:780
780       lisp_h_CHECK_TYPE (ok, predicate, x);
(gdb) up
#3  0x0000000000458f20 in gui_consider_frame_title (
    frame=frame@entry=XIL(0x1123255)) at ../../src/lisp.h:1048
1048      return make_lisp_symbol (&lispsym(index));
(gdb) p &lispsym(index)
$1 = (struct Lisp_Symbol *) 0x17fffe29037f80
(gdb) pp &lispsym(index)
#<INVALID_LISP_OBJECT 0x17fffe29037f80>
(gdb) pp lispsym(index)
Cannot access memory at address 0x17fffe29037f80
(gdb) up
#4  0x0000000000472394 in redisplay_window (window=XIL(0x1165305),
    just_this_one_p=<optimized out>) at ../../src/xdisp.c:18940
18940         gui_consider_frame_title (w->frame);
(gdb) pp window
#<window 3 on 1.l>
(gdb) up
#5  0x0000000000476b9b in redisplay_window_0 (
    window=window@entry=XIL(0x1165305)) at ../../src/xdisp.c:16179
16179       redisplay_window (window, false);
(gdb) up
#6  0x000000000056c114 in internal_condition_case_1 (
    bfun=bfun@entry=0x476b70 <redisplay_window_0>,
    arg=arg@entry=XIL(0x1165305), handlers=<optimized out>,
    hfun=hfun@entry=0x43a0d0 <redisplay_window_error>) at
    ../../src/eval.c:1379
1379          Lisp_Object val = bfun (arg);
(gdb) up
#7  0x0000000000443554 in redisplay_windows (window=XIL(0x1165305))
    at ../../src/xdisp.c:16159
16159             internal_condition_case_1 (redisplay_window_0, window,
(gdb) up
#8  0x0000000000462e1d in redisplay_internal () at
../../src/xdisp.c:15627
15627                     redisplay_windows (FRAME_ROOT_WINDOW (f));

Re: another unhelpful backtrace from redisplay-internal

[Eli Zaretskii]

> From: Madhu <enometh@meer.net>
> Date: Sat, 07 Mar 2020 21:02:36 +0530
> 
> I have a sure recipe for a crash in recent emacs but the configuration
> is complex and I'm afraid I don't have a simple test case. I'm posting
> the initial backtrace here in case anyone can suggest things I should
> look for, or some ways I could narrow it down.

"Recent emacs" from which branch of the repository?  Also, how recent?

(Crashes should be reported with "M-x report-emacs-bug", as there's no
doubt they are bugs of some sort, and therefore no reason to discuss
them here.  Using report-emacs-bug would have included all the
information I'm asking about above, automatically.)

> Emacs is compiled with -g3 -O2, and I'm not sure if all the info to
> debug this is available.

The backtrace shows incorrect files and lines, and most values are
"optimized out" (and others are therefore suspect to not be the
truth).

> Any suggestions please?

Rebuild without optimizations and try reproducing the problem with
your recipe.  If successful, post the backtrace.

Thanks.

Re: another unhelpful backtrace from redisplay-internal

[Eli Zaretskii]

> Date: Sat, 07 Mar 2020 18:14:22 +0200
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: emacs-devel@gnu.org
> 
> Rebuild without optimizations and try reproducing the problem with
> your recipe.  If successful, post the backtrace.

P.S.  Your email server seems to reject email from gnu.org, so my
direct response isn't going to come through.

Re: another unhelpful backtrace from redisplay-internal

[Madhu]

* Eli Zaretskii <838skc9lod.fsf@gnu.org> :
Wrote on Sat, 07 Mar 2020 18:25:38 +0200:

>> Date: Sat, 07 Mar 2020 18:14:22 +0200
>> From: Eli Zaretskii <eliz@gnu.org>
>> Rebuild without optimizations and try reproducing the problem with
>> your recipe.  If successful, post the backtrace.

The backtrace is substantially the same. I'll post it once I'm past the
greylist

> P.S.  Your email server seems to reject email from gnu.org, so my
> direct response isn't going to come through.

[I apologize for that -- I'll drop a note to support and see if they can
whitelist it.  (I expect they will continue to reject it for the time
being)]

Re: another unhelpful backtrace from redisplay-internal

[Madhu]

* Eli Zaretskii <838skc9lod.fsf@gnu.org> :
Wrote on Sat, 07 Mar 2020 18:25:38 +0200:

>> Date: Sat, 07 Mar 2020 18:14:22 +0200
>> From: Eli Zaretskii <eliz@gnu.org>
>> Rebuild without optimizations and try reproducing the problem with
>> your recipe.  If successful, post the backtrace.

The backtrace is substantially the same. I'll post it once I'm past the
greylist

> P.S.  Your email server seems to reject email from gnu.org, so my
> direct response isn't going to come through.

[I apologize for that -- I'll drop a note to support and see if they can
whitelist it.  (I expect they will continue to reject it for the time
being)]

Re: another unhelpful backtrace from redisplay-internal

[Eli Zaretskii]

> From: Madhu <enometh@meer.net>
> Date: Sat, 07 Mar 2020 23:17:13 +0530
> 
> >> Rebuild without optimizations and try reproducing the problem with
> >> your recipe.  If successful, post the backtrace.
> 
> The backtrace is substantially the same.

I very much hope it isn't, because the original one was full of
incorrect information, such as that select_window is defined on lisp.h
etc.

Please also tell which branch is this and what is the last commit you
have on that branch.  It's important for us to be able to correctly
interpret the file name/line number info in your backtraces.