bug#69431: 30.0.50; Strange fontificaion behavior

[Andrea Corallo <acorallo@gnu.org>]

Björn Bidar <bjorn.bidar@thaodan.de> writes:

> Eli Zaretskii <eliz@gnu.org> writes:
>
>>> From: Björn Bidar <bjorn.bidar@thaodan.de>
>>> Cc: acorallo@gnu.org,  yantar92@posteo.net,  monnier@iro.umontreal.ca,
>>>   69431@debbugs.gnu.org,  hirofumi@mail.parknet.co.jp
>>> Date: Sun, 07 Apr 2024 18:00:26 +0300
>>>
>>> >> Is there already a bug for this issue or should I open a new one?
>>>
>>> > It depends, I think.  What is the recipe for what you see?
>>>
>>> I can only trigger the bug when I load enough packages to trigger the bug
>>> First I thought the bug is triggered when I modus themes but it also
>>> happens without, I just have to load enough packages.
>>>
>>>
>>> I attach my init.el (with private information removed) and
>>> the logs I had from  earlier and now.
>>
>> It is some kind of recursive call:
>>
>>   face-attribute->face-attribute-merged-with->face_attribute->...
>>
>> Looks like a separate problem to me.  It is important to understand
>> what face causes this, and what is that face's spec.
>>
>> Also, please make a point of loading the src/.gdbinit file from the
>> Emacs source tree before running Emacs under GDB, so that the
>> backtrace includes also the Lisp backtrace, and shows Lisp objects in
>> human-readable format, for easier reading.
>
> The backtrace was with src/.gdbinit loaded however the lisp wasn't
> included the backtace I suspect it is because:
> Cannot access memory at address 0x7fffff66ff7f
>
>>> I tried to find which package exactly triggers the bug but I did not
>>> manage to do that. I only know that it is triggered through the use of
>>> faces.
>>
>> Both backtraces are almost identical, and in both the problem seems to
>> be triggered by loading and enabling a theme.  So I don't think I
>> understand what you say here about "enough packages" -- I don't
>> suppose all of the packages you load are themes, are they?  IOW, can
>> you show a backtrace from a crash that is not caused by loading a
>> theme?  E.g., what happens if you remove all theme loading and related
>> stuff from your init files?
>
> What do you mean by related?
> If I don't load a theme Emacs doesn't crash but if I start Emacs with -Q
> and load the theme it's not enough to crash Emacs that's why I mean by
> enough packages essentially.
>
> Eventually I isolated the exact problem what caused the bug:
> 1. (setq max-lisp-eval-depth 32768)
> 2. load-theme modus-vivendi (any modus theme works as a trigger)

I tried the following recipe on latest master but can't reproduce.

BTW this is different bug from the one being treated here, ideally would
be better to have a dedicated bug.

  Andrea