GNU bug report logs - #76473
31.0.50; gnutls_pproc leak

Previous Next

Full log

Message #30 received at 76473-done <at> debbugs.gnu.org (full text, mbox):

From: Pip Cet <pipcet <at> protonmail.com>
To: Eli Zaretskii <eliz <at> gnu.org>
Cc: 76473-done <at> debbugs.gnu.org, eller.helmut <at> gmail.com
Subject: Re: bug#76473: 31.0.50; gnutls_pproc leak
Date: Sat, 22 Feb 2025 14:34:27 +0000

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

>> Date: Fri, 21 Feb 2025 19:58:19 +0000
>> From: Pip Cet <pipcet <at> protonmail.com>
>> Cc: eller.helmut <at> gmail.com, 76473 <at> debbugs.gnu.org
>>
>> "Eli Zaretskii" <eliz <at> gnu.org> writes:
>>
>> >> Date: Fri, 21 Feb 2025 18:44:17 +0000
>> >> From: Pip Cet <pipcet <at> protonmail.com>
>> >> Cc: eller.helmut <at> gmail.com, 76473 <at> debbugs.gnu.org
>> >>
>> >> "Eli Zaretskii" <eliz <at> gnu.org> writes:
>> >>
>> >> >> Cc: 76473 <at> debbugs.gnu.org
>> >> >> Date: Fri, 21 Feb 2025 18:02:34 +0000
>> >> >> From:  Pip Cet via "Bug reports for GNU Emacs,
>> >> >>  the Swiss army knife of text editors" <bug-gnu-emacs <at> gnu.org>
>> >> >>
>> >> >> > This patch would fix that:
>> >> >>
>> >> >> I believe it would!  Is anyone testing the w32/w64 + gnutls combination?
>> >> >
>> >> > In general, or on the igc branch?
>> >>
>> >> The latter.  This bug is specific to feature/igc, I think, since that's
>> >> the only branch that has gnutls_pproc.
>> >
>> > If you or Helmut tell me what to try and what to notice, I can do
>> > that.
>>
>> Can you watch the PVEC_PROCESS line (in M-x igc-stats) or the
>> xzalloc-ambig line (in M-x igc-roots-stats) while creating and
>> destroying a number of gnutls processes?
>>
>> Without the patch, the numbers should go up but never down.  With the
>> patch, all should be fine :-)
>
> I indeed see that with the patch, igc-collect and/or deleting the
> processes (in the list-processes display) makes the number of
> PVEC_PROCESS objects go down, whereas without the patch it stays at
> the same value it had before (and goes up when I invoke eww for
> another HTTPS connection).  However, even with the patch, the number
> never goes back to zero, only to 1.  Don't know if it's important or
> not.

Thanks!  Since there is no process in M-x list-processes (such as a
server process), I think it's most likely that an "ambiguous" reference
to a destroyed process survives on the bytecode stack or the C stack.
So my guess it's that's a harmless side-effect of conservative garbage
collection.

> Anyway, the patch seems to be TRT, thanks.

Thanks, Helmut!  I've changed the subject line to reference the bug
number, but otherwise what I just pushed should be identical to what you
sent!

Closing this bug.

Pip

Previous Next

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