bug#58334: 29.0.50; ASAN heap use after free in gui_produce_glyphs

[Eli Zaretskii]

> From: Gerd Möllmann <gerd.moellmann@gmail.com>
> Cc: Eli Zaretskii <eliz@gnu.org>,  58334@debbugs.gnu.org
> Date: Fri, 07 Oct 2022 10:54:34 +0200
> 
> Po Lu <luangruo@yahoo.com> writes:
> 
> > Gerd Möllmann <gerd.moellmann@gmail.com> writes:
> >
> >> And, while vaccuming, I also wondered what happens with the glyph
> >> matrices, and maybe other global state?
> >
> > Isn't input blocked wherever the glyph matrices are modified?
> > If not,
> 
> Partly.  Eli please correct me if this has changed.
> 
> Redisplay has two phases:
> 
> 1. Building desired matrices, that is, what should be on the screen
> after redisplay.
> 
> 2. Updating current matrices from desired matrices, and bringing that on
> the screen.
> 
> Only phase 2 has input blocked because the current matrices are
> modified, which are shared state with the GUI code.

I don't see input blocked in phase 2, either.  We interrupt phase 2 if
input is pending, but even that only in some cases.

The GUI code runs in the same thread as phase 2, so there's no reason
to synchronize anything here, AFAIU.  If you are thinking about
expose_frame and friends, then that was changed years ago not to run
from the signal handler, so it cannot cause any problems to code that
modifies the glyph matrices.