Re: Regular crashes with abort() in mark_maybe_pointer with trunk

[Jim Meyering <jim@meyering.net>]

Julien Danjou wrote:
> Since at least yesterday, latest trunk version crashes regularly.
>
> I've grabbed 2 backtraces so far, attached.
>
> I'm sure that the change responsible has been done between the commits:
>
> commit 1e211d886c96420904ad197cfee7fce7ae7b9d83
> Author: Stefan Monnier <monnier@iro.umontreal.ca>
> Date:   Fri Jun 22 09:42:38 2012 -0400
>
>     Provide generalized variables in core Elisp.
>
>
> commit b17778189e1df0a56e48a5f6f0b52f7bebbdfdd7
> Author: Julien Danjou <julien@danjou.info>
> Date:   Thu Jun 28 12:40:24 2012 +0200
>
>     term: rework color cutomizing
>
>
> I know this because emacs-snapshot for Debian that I package worked
> correctly for the 20120622 version, and stopped working with this last
> commit. Something in between changed something causing this.
>
> If nobody has an idea, I can try to bisect but that will require
> sometimes since the crash is still random to me (not sure how to
> reproduce for sure).

I've been seeing intermittent crashes and bootstrap failures (stack
traces always point to GC-related code) on the trunk for months now.
The failure comes and goes, seemingly at random, when unrelated changes
are committed.  Bisection has not been useful to me, perhaps because
the underlying problem is merely triggered by the presence of new code
inducing some sort of change in the bootstrapped temacs.

For example, between June 19 and some time in June 25-27, I failed
to bootstrap from the trunk because of this sort of problem.
However, since June 28th bootstrap succeeded.
Here are date/times of my recent successful trunk bootstraps:

  emacs-2012-06-01.07h20/
  emacs-2012-06-02.06h19/
  emacs-2012-06-03.12h59/
  emacs-2012-06-04.15h56/
  emacs-2012-06-06.20h46/
  emacs-2012-06-08.05h44/
  emacs-2012-06-09.15h54/
  emacs-2012-06-16.06h27/
  emacs-2012-06-18.07h52/
  emacs-2012-06-19.05h12/
  emacs-2012-06-28.13h43/
  emacs-2012-06-29.05h29/
  emacs-2012-06-30.09h07/