bug#20808: 24.5; Emacs looping with 100% CPU at line 15287 (?) of xdisp.c

[N. Jackson]

At 11:08 -0300 on Sunday 2015-06-14, Eli Zaretskii wrote:

> Yes, the information you collected is useful, thanks.  If you could
> come up with a reproducible recipe for this, it would be even better.
> Failing that, I could give you instructions regarding GDB commands
> that will collect some more data, in the hope that the data will allow
> to come up with a fix.

Well, I don't know exactly what I was doing to trigger the problem. I
suspect it happenned when I clicked a link in an email in Gnus, which
opened a web page in Eww. (It may have been relevant that I then 1)
scrolled in the Eww window and 2) that I hit `&' to open the web page in
an external browser -- not necessarilly in that order.)

It might be relevant that the web page contained quite a lot of images,
mostly small, and that Gnus opens Eww in the short (eleven lines high)
Summary buffer window.

[The typical view in Gnus when reading mail has the frame divided
horizontally (a top and a bottom window), with the short Summary buffer
window on top displaying the list of mail in the current "group" (mail
box/folder/directory), and a taller window at the bottom (the Article
buffer) displaying the content of the mail that is currently selected in
the Summary buffer.]

I was going to say that I doubt that I can reproduce the problem but
while trying the steps above, it happenned again. It seemed (perhaps) to
happen after scrolling the Eww window with the scroll thumb. The
backtrace was the same.

I haven't been able to reproduce the problem in my own build of Emacs
24.5, but I don't know yet if that's because I haven't taken exactly the
same steps, or if it's because of different build settings.

[I haven't reproduced the problem on master either, but there things are
different so I don't expect to. Eww now no longer dispays in the short
window, but in the taller one, and the Eww page layout is simpler and
cleaner.]

I will work on finding a simple reproducible recipe for the Fedora 21
build (preferably removing Gnus from the equation), and will also see if
I can reproduce the problem in my own build.

Meanwhile, if you think it is worth trying to debug the problem in the
Fedora 21 optimised build, please send instructions for steps to take in
GDB. I still have GDB attached to it.

Thanks.