GNU bug report logs - #79367
31.0.50; magit-commit sometimes doesn't work if diff-hl-update-async is t

Previous Next

Full log

View this message in rfc822 format

From: Zhengyi Fu <i <at> fuzy.me>
To: Eli Zaretskii <eliz <at> gnu.org>
Cc: sbaugh <at> janestreet.com, 79367 <at> debbugs.gnu.org, dmitry <at> gutov.dev
Subject: bug#79367: 31.0.50; magit-commit sometimes doesn't work if diff-hl-update-async is t
Date: Sat, 06 Sep 2025 16:28:40 +0800

Eli Zaretskii <eliz <at> gnu.org> writes:

>> From: Zhengyi Fu <i <at> fuzy.me>
>> Cc: sbaugh <at> janestreet.com,  dmitry <at> gutov.dev,  79367 <at> debbugs.gnu.org
>> Date: Sat, 06 Sep 2025 13:35:49 +0800
>> 
>> Eli Zaretskii <eliz <at> gnu.org> writes:
>> 
>> >> From: Zhengyi Fu <i <at> fuzy.me>
>> >> Cc: sbaugh <at> janestreet.com,  dmitry <at> gutov.dev,  79367 <at> debbugs.gnu.org
>> >> Date: Fri, 05 Sep 2025 21:36:16 +0800
>> >> 
>> >> Eli Zaretskii <eliz <at> gnu.org> writes:
>> >> 
>> >> > Thanks, then please show the value of the offending descriptor which
>> >> > causes the assertion violation.  For that, please run Emacs udner GDB,
>> >> > perform the recipe that reproduces the problem, and then type these
>> >> > GDB commands:
>> >> >
>> >> >  (gdb) frame 1
>> >> >  (gdb) print p->infd
>> >> >
>> >> > and tell what that produces.
>> >> 
>> >> The p->infd value is 8.
>> >
>> > OK, please repeat this several times to make sure it's always 8.
>> >
>> > If it is, then let's try to see which code leaves the .thread field
>> > uncleared.  Like this (please run from the src directory where you
>> > have the 'emacs' executable you have built):
>> >
>> >   $ gdb ./emacs
>> >   ...
>> >   (gdb) break process.c:8731
>> >   (gdb) run
>> >
>> > Emacs will start and stop in init_process_emacs, on line 8731 of
>> > process.c.  Then:
>> >
>> >   (gdb) watch fd_callback_info[8].thread
>> >   (gdb) commands
>> >    > bt 5
>> >    > continue
>> >    > end
>> >   (gdb) continue
>> >
>> > The above sets a watchpoint which will stop Emacs each time we assign
>> > some value to fd_callback_info[8].thread, and GDB will then show the
>> > first 5 frames of the call-stack, then continue.  Now run your recipe,
>> > until it gets SIGABRT, and post everything GDB produced until that
>> > point.  I hope that will tell us which code fails to clear the .thread
>> > member.
>> >
>> > Thanks.
>> 
>> The fd is 8 when running Emacs in terminal.
>> The fd is always 16 when running Emacs in X11.
>
> Thanks, please try the patch below, instead of the incorrect one I
> asked you to try before:
>
> diff --git a/src/process.c b/src/process.c
> index fa003c2..736098f 100644
> --- a/src/process.c
> +++ b/src/process.c
> @@ -4831,7 +4831,11 @@ deactivate_process (Lisp_Object proc)
>    /* Beware SIGCHLD hereabouts.  */
>  
>    for (i = 0; i < PROCESS_OPEN_FDS; i++)
> -    close_process_fd (&p->open_fd[i]);
> +    {
> +      fd_callback_info[p->open_fd[i]].thread = NULL;
> +      fd_callback_info[p->open_fd[i]].waiting_thread = NULL;
> +      close_process_fd (&p->open_fd[i]);
> +    }
>  
>    inchannel = p->infd;
>    eassert (inchannel < FD_SETSIZE);

This resolves the problem. The assertion failure no longer occurs when testing with the patch.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.