bug#39948: 28.0.50; crash in fchmodat

[Stephen Berman]

On Fri, 06 Mar 2020 17:26:54 +0100 Stephen Berman <stephen.berman@gmx.net> 
wrote:

> On Fri, 06 Mar 2020 16:31:50 +0100 Robert Pluim <rpluim@gmail.com> wrote:
>
>>>>>>> On Fri, 06 Mar 2020 15:16:43 +0100, Stephen Berman
>> <stephen.berman@gmx.net> said:
>>
>>     Stephen> I updated from master today and now Emacs is crashing
>>     Stephen> when I use Gnus.  The first time it happened I been
>>     Stephen> reading news groups for a while, then email arrived and
>>     Stephen> when I pulled it into Gnus, Emacs crashed.  Then I
>>     Stephen> restarted Emacs under GDB and now get the crash already
>>     Stephen> on starting Gnus (with my initializations; it doesn't
>>     Stephen> happen when I start an unconfigured Gnus in Emacs -Q).  I
>>     Stephen> tried to get a full backtrace, but the output of `bt
>>     Stephen> full' seemed to be in an endless loop; here's the start
>>     Stephen> of the backtrace:
>>
>> Wild Guess: does reverting 07da629926daf849aab248175c88cf53a5e21558
>> help? Or maybe 9d626dffc6ba62c0d7a1a5c712f576ed8684fd66 ?
>
> Reverting 07da629 does appear to prevent the crash.  Specifically, as a
> sanity check, before reverting, I ran my build from master under GDB
> again, started Gnus, but now Emacs didn't crash; however, then I
> composed a mail to myself in Gnus and upon sending it, Emacs did crash
> exactly as before in fchmodat.  This I did `git revert -n
> 07da629926daf849aab248175c88cf53a5e21558', rebuilt, started Emacs again,
> started Gnus, there was no crash, sent a mail to myself, again with no
> crash, and pulled the mail into Gnus also without the crash.  So it
> appears that that commit is indeed the cause of the crash.  Thanks for
> pinpointing it.

I also got the crash using Tramp, so I guess 9d626dff also needs to be
fixed (or reverted).

Steve Berman