bug#34623: 25.3; infinite un-quittable loop somwhere under redisplay_window()

[Eli Zaretskii]

tags 34623 moreinfo
thanks

> From: "Greg A. Woods" <woods@robohack.ca>
> Date: Fri, 22 Feb 2019 14:26:04 -0800
> 
> So I was editing a message in a Wanderlust draft buffer, and I had done
> narrow-to-region, then a replace-regexp (to insert a tab at the
> beginning of every line in the region), then it hung with the CPU
> spinning and the only response being some redraws under cursor moves
> that triggered hi-lighting in other windows (i.e. keyboard-quit is
> ignored).  I don't think I managed to type "C-x n w" to widen the window
> before it hung (that is I probably typed it, but I think it was hung
> before it accepted my keystrokes, as I think that window had gone blank
> after the replace -- though I may be mistaken).
> 
> I tried finding the shortest backtrace by continuing and stopping
> randomly, and this isn't far from it, but due to compiler optimisation
> and bytecode compiled lisp, this still may not be very useful.
> 
> I tried to get more info with xbacktrace, but my gdb does not have
> Python support, and after fooling around too much I managed to crash the
> emacs process upon detach and ended up with a borked unusable core file.

Thanks, but the backtrace doesn't provide enough information to see
what caused this.  What I see is that redisplay called a fontification
function, that's all.  I cannot know what function it was nor why it
looped (if it looped).

I suggest to upgrade to Emacs 26.1 or even to a pretest of soon to be
released Emacs 26.2, as a few loops in redisplay in some tricky
situations were discovered and fixed since 25.3 was released.  And in
any case, if this ever happens again, please use the technique
described in etc/DEBUG, under "If the symptom of the bug is that Emacs
fails to respond", it explains how to find out the place of the
infloop, if there is one.