Possible issue with pgtk branch

Possible issue with pgtk branch

[chad]

[-- Attachment #1: Type: text/plain, Size: 1355 bytes --]

I've been trying out the feature/pgtk branch for a while, and I've run into
a mysterious shutdown (not crash) that I haven't been able to isolate yet.
I'm wondering if anyone else is seeing something similar, so I can gather
enough data to file a big report.

I'm running emacs in daemon mode, with at least 2 frames. The shutdown
happens in eww buffers, when I'm scrolling, and perhaps always (at least a
couple times) when I'm scrolling up with M-v. Perhaps relevant, I often use
eww to make web pages more readable, and this often involves me making the
buffer writable (C-x C-q) and removing (killing, ala C-k and C-w)
extraneous junk, which includes images and styled text in various sizes and
perhaps fonts.

This isn't reliable enough for me to catch it under gdb, but I'm trying.
Has anyone else seen anything like this, or perhaps the above behavior
description is enough to suggest something I might do to tickle the issue?

Thanks in advance,
~Chad

P.S. My environment is pretty weird; I'm running emacs inside Crostini, a
debian container running on a lightweight vm inside ChromeOS. This system
uses a wayland-based display engine with a chromeos-custom compositor with
X11 support, described in some detail here:
https://chromium.googlesource.com/chromiumos/platform2/+/413a646b5b7067a989bab2ef9fca0c3a4515cc22/vm_tools/sommelier/README.md

[-- Attachment #2: Type: text/html, Size: 1670 bytes --]

Re: Possible issue with pgtk branch

[Eli Zaretskii]

> From: chad <yandros@gmail.com>
> Date: Thu, 14 Jan 2021 16:55:45 -0800
> 
> I'm running emacs in daemon mode, with at least 2 frames. The shutdown happens in eww buffers, when I'm
> scrolling, and perhaps always (at least a couple times) when I'm scrolling up with M-v. Perhaps relevant, I
> often use eww to make web pages more readable, and this often involves me making the buffer writable
> (C-x C-q) and removing (killing, ala C-k and C-w) extraneous junk, which includes images and styled text in
> various sizes and perhaps fonts.
> 
> This isn't reliable enough for me to catch it under gdb, but I'm trying.

Why isn't it reliable enough to catch?  How did you try catching it?

Re: Possible issue with pgtk branch

[chad]

[-- Attachment #1: Type: text/plain, Size: 756 bytes --]

On Thu, Jan 14, 2021 at 11:08 PM Eli Zaretskii <eliz@gnu.org> wrote:

> > From: chad <yandros@gmail.com>
> > Date: Thu, 14 Jan 2021 16:55:45 -0800
> >
> > I'm running emacs in daemon mode, with at least 2 frames. The shutdown
> happens in eww buffers, when I'm
> > scrolling, and perhaps always (at least a couple times) when I'm
> scrolling up with M-v.
> >
> > This isn't reliable enough for me to catch it under gdb, but I'm trying.
>
> Why isn't it reliable enough to catch?  How did you try catching it?
>

I noticed that "scrolling" and "scrolling up" made it happen a couple
times, but that's not been enough to reproduce it so far. I'm still trying
to find a reliable repro, but I had seen it enough that it seemed worth
asking on the list.

~Chad

[-- Attachment #2: Type: text/html, Size: 1198 bytes --]

Re: Possible issue with pgtk branch

[Eli Zaretskii]

> From: chad <yandros@gmail.com>
> Date: Fri, 15 Jan 2021 09:49:44 -0800
> Cc: EMACS development team <emacs-devel@gnu.org>
> 
>  > I'm running emacs in daemon mode, with at least 2 frames. The shutdown happens in eww buffers,
>  when I'm
>  > scrolling, and perhaps always (at least a couple times) when I'm scrolling up with M-v. 
>  > 
>  > This isn't reliable enough for me to catch it under gdb, but I'm trying.
> 
>  Why isn't it reliable enough to catch?  How did you try catching it?
> 
> I noticed that "scrolling" and "scrolling up" made it happen a couple times, but that's not been enough to
> reproduce it so far. I'm still trying to find a reliable repro, but I had seen it enough that it seemed worth asking
> on the list.

My suggestion is to catch the shutdown itself.  That should go through
kill-emacs, so a breakpoint there should be enough.  If you want to be
more sure, set a breakpoint on 'exit' and 'abort' (2 library
functions).  Then run Emacs and use it as usual.  When these
breakpoints break, a backtrace should tell us what is going on.

Thanks.