bug#71556: GNOME Konsol crashes

bug#71556: GNOME Konsol crashes

[Simon Josefsson via Bug reports for GNU Guix]

[-- Attachment #1: Type: text/plain, Size: 1573 bytes --]

Hi.

GNOME Konsol has crashed for me, bringing down all 10+ windows I had
open (and plenty of ongoing work halted...).  I've now seen this twice,
so I suppose it counts as reproducible, even though I don't know how to
trigger it.  I recall that it happened when I used the mouse to maybe
resize windows or switch workspace somehow.

This is guix v1.3.0-60008-g7e00fb9f31 with a fairly normal system
configuration [1] on a fairly ordinary amd64 desktop (MSI Z790P).
Syslog has these two crashes:

Jun 12 17:53:49 localhost vmunix: [244491.668326] .kgx-real[2161]: segfault at 29 ip 00007fd22dc31d6d sp 00007ffed19f0630 error 6 in crocus_dri.so[7fd22da94000+165b000] likely on CPU 10 (core 20, socket 0)
Jun 12 17:53:49 localhost vmunix: [244491.668348] Code: c0 75 42 48 8b 93 08 1a 00 00 48 8d 8b 08 1a 00 00 49 8d 44 24 08 66 48 0f 6e c9 66 48 0f 6e c2 66 0f 6c c1 41 0f 11 44 24 08 <48> 89 42 08 48 89 83 08 1a 00 00 f0 83 6d 00 01 75 31 5b 5d 41 5c

Jun 13 20:45:36 localhost vmunix: [96351.516581] .kgx-real[2371]: segfault at 8 ip 00007f34fc831d6d sp 00007ffe27647ae0 error 6 in crocus_dri.so[7f34fc694000+165b000] likely on CPU 30 (core 46, socket 0)
Jun 13 20:45:36 localhost vmunix: [96351.516595] Code: c0 75 42 48 8b 93 08 1a 00 00 48 8d 8b 08 1a 00 00 49 8d 44 24 08 66 48 0f 6e c9 66 48 0f 6e c2 66 0f 6c c1 41 0f 11 44 24 08 <48> 89 42 08 48 89 83 08 1a 00 00 f0 83 6d 00 01 75 31 5b 5d 41 5c

Anyone else seeing this?  I know it is hard to debug based on this
information...

/Simon

[1] https://gitlab.com/jas/sjd-cosmos/-/blob/master/yeti.sjd.se/config.scm

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 255 bytes --]

bug#71556: GNOME Konsol crashes

[Liliana Marie Prikler]

Am Freitag, dem 14.06.2024 um 18:34 +0200 schrieb Simon Josefsson:
> Hi.
> 
> GNOME Konsol has crashed for me, bringing down all 10+ windows I had
> open (and plenty of ongoing work halted...).  I've now seen this
> twice, so I suppose it counts as reproducible, even though I don't
> know how to trigger it.  I recall that it happened when I used the
> mouse to maybe resize windows or switch workspace somehow.
> 
> This is guix v1.3.0-60008-g7e00fb9f31 with a fairly normal system
> configuration [1] on a fairly ordinary amd64 desktop (MSI Z790P).
> Syslog has these two crashes:
… and a fairly normal GPU?

The crash indicates some driver issue, probably in Mesa.  Since GNOME
is increasingly hardware accelerated, you tend to notice them in a lot
of places.  You can twiddle the GSK_RENDERER environment variable to
something less buggy, potentially at the cost of performance. 
Particularly the cairo one ought to do slow, but correct CPU rendering.


Cheers

bug#71556: GNOME Konsol crashes

[Simon Josefsson via Bug reports for GNU Guix]

> 21 juni 2024 kl. 06:32 skrev Liliana Marie Prikler <liliana.prikler@gmail.com>:
> 
> Am Freitag, dem 14.06.2024 um 18:34 +0200 schrieb Simon Josefsson:
>> Hi.
>> 
>> GNOME Konsol has crashed for me, bringing down all 10+ windows I had
>> open (and plenty of ongoing work halted...).  I've now seen this
>> twice, so I suppose it counts as reproducible, even though I don't
>> know how to trigger it.  I recall that it happened when I used the
>> mouse to maybe resize windows or switch workspace somehow.
>> 
>> This is guix v1.3.0-60008-g7e00fb9f31 with a fairly normal system
>> configuration [1] on a fairly ordinary amd64 desktop (MSI Z790P).
>> Syslog has these two crashes:
> … and a fairly normal GPU?

Built-in UHD 770 in the Intel 14900K CPU (no overclocking). I am using Dasharo system firmware if that makes a difference.

I’ve used this system extensively and think some recent update may have introduced it as I started seeing it only recently. Hard to tell though since it only happened after 4+ hours of interactive work each time.

> The crash indicates some driver issue, probably in Mesa.  Since GNOME
> is increasingly hardware accelerated, you tend to notice them in a lot
> of places.  You can twiddle the GSK_RENDERER environment variable to
> something less buggy, potentially at the cost of performance.
> Particularly the cairo one ought to do slow, but correct CPU rendering.

Ah this makes sense. I will try and see if I notice it again. It hasn’t happened for any other application.

/Simon