bug#47244: 28.0.50; SIGSEGV in long-runnning Emacs

[Eli Zaretskii]

> Cc: mwd@cert.org, mwd@md5i.com, schwab@linux-m68k.org, 47244@debbugs.gnu.org
> From: martin rudalics <rudalics@gmx.at>
> Date: Wed, 31 Mar 2021 08:41:52 +0200
> 
>  > IME, there's any number of "real problems" out there, and the hope to
>  > find them all is ... too optimistic.
> 
> I'd be already happy to find just the one we talk about here.
> 
>  > There's nothing wrong with being
>  > defensive against unknown factors.
> 
> Agreed.  What I meant was that if, as we suppose, the mischief happens
> at time T1 somewhere in redisplay and is detected (by chance) and
> corrected at time T2 in the main loop, any side effects that mischief
> causes between T1 and T2 still go undetected and may cause arbitrary
> harm.

I think it's the other way around: we set the dead buffer as the
current one in the main loop, and then redisplay tries to redisplay
that window.

> In either case whatever you want to do here - set the selected window's
> buffer or just the current buffer in the main loop - would not interfere
> with finding the real culprit.  So I certainly won't object to your fix.

I will wait for a few days in the hope that the culprit is found.