On Fri, 2025-12-05 at 10:14 +1000, David Airlie wrote:
> > I've been noticing flakes in openQA tests caused by mysterious
> > segfaults. The first was that KDE live install tests started failing
> > because slitherer - the qt6-based runner for anaconda-webui which Neal
> > wrote and which is used on all lives except Workstation - would crash
> > and so the installer never ran. This is happening to maybe 1/4 of KDE
> > (and i3) live install tests. The crash is always a segfault in
> > libQt6WebEngineCore.so.6.10.1 , like this:
> >
> > Nov 23 22:58:46 localhost-live slitherer[3647]: Failed to register with
> > host portal QDBusError("org.freedesktop.portal.Error.Failed", "Could not
> > register app ID: App info not found for
> > 'org.fedoraproject.AnacondaInstaller'")
> > Nov 23 22:58:46 localhost-live liveinst[3647]: libva error:
> > /usr/lib64/dri/virtio_gpu_drv_video.so init failed
> > Nov 23 22:58:46 localhost-live audit[3647]: ANOM_ABEND auid=1000 uid=1000
> > gid=1000 ses=2 subj=unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023
> > pid=3647 comm="slitherer" exe="/usr/bin/slitherer" sig=11 res=1
> > Nov 23 22:58:46 localhost-live kernel: slitherer[3647]: segfault at 2dc ip
> > 00007fc97480ecab sp 00007ffd290db700 error 4 in
> > libQt6WebEngineCore.so.6.10.1[15dcab,7fc9746b1000+a32a000] likely on CPU 1
> > (core 1, socket 0)
> > Nov 23 22:58:46 localhost-live kernel: Code: 8d 7b 58 4c 89 e6 e8 04 15 c4
> > 00 c6 83 e8 03 00 00 01 48 8d 05 3e 95 7c 0b 48 8b 00 48 8b 80 28 02 00 00
> > 4c 8b a8 30 0b 00 00 <41> 8b 9d dc 02 00 00 85 db 0f 85 3e 01 00 00 4c 8d
> > b5 50 ff ff ff
> > Nov 23 22:58:46 localhost-live systemd-coredump[3771]: Process 3647
> > (slitherer) of user 1000 terminated abnormally with signal 11/SEGV,
> > processing...
> >
> > . I filed this as a slitherer bug -
> > https://bugzilla.redhat.com/show_bug.cgi?id=2417493 - but we couldn't
> > find a cause. We do think it starting happening right after the qt6
> > 6.10.1 update -
> > https://bodhi.fedoraproject.org/updates/FEDORA-2025-9417bbae5b .
> >
> > Then in the last week or so I've noticed Workstation tests sometimes
> > failing because gnome-shell crashes on login, with...you guessed it...a
> > segfault. It always looks more or less like this:
> >
> > Dec 04 08:11:03 fedora audit[1]: SERVICE_START pid=1 uid=0 auid=4294967295
> > ses=4294967295 subj=system_u:system_r:init_t:s0 msg='unit=sssd-kcm
> > comm="systemd" exe="/usr/lib/systemd/systemd" hostname=? addr=? terminal=?
> > res=success'
> > Dec 04 08:11:03 fedora gnome-shell[2088]: An object is already exported for
> > the interface org.freedesktop.IBus.Service at /org/freedesktop/IBus/Panel
> > Dec 04 08:11:03 fedora kernel: gnome-shell[2088]: segfault at 1012000e ip
> > 00007f09a8f39145 sp 00007ffd4ac89620 error 4 in
> > libgobject-2.0.so.0.8700.0[2f145,7f09a8f0a000+38000] likely on CPU 1 (core
> > 1, socket 0)
> > Dec 04 08:11:03 fedora gnome-shell[2088]: g_object_get_qdata: assertion
> > 'G_IS_OBJECT (object)' failed
> > Dec 04 08:11:03 fedora kernel: Code: 0f 84 8f 00 00 00 48 8b 07 48 85 c0 0f
> > 84 a3 00 00 00 48 8b 10 48 81 fa fc 03 00 00 76 3f 48 89 d6 b8 01 00 00 00
> > 48 83 e6 fc <f6> 46 0e 10 75 27 8b 7e 20 e8 8d 22 fd ff 48 89 c1 31 c0 48
> > 8d 15
> > Dec 04 08:11:03 fedora gnome-shell[2088]: g_object_is_floating: assertion
> > 'G_IS_OBJECT (object)' failed
> > Dec 04 08:11:03 fedora gnome-shell[2088]: g_object_set_qdata_full:
> > assertion 'G_IS_OBJECT (object)' failed
> > Dec 04 08:11:03 fedora gnome-shell[2088]: g_object_get_qdata: assertion
> > 'G_IS_OBJECT (object)' failed
> > Dec 04 08:11:03 fedora gnome-shell[2088]: g_object_weak_ref: assertion
> > 'G_IS_OBJECT (object)' failed
> > Dec 04 08:11:03 fedora gnome-shell[2088]: g_object_add_toggle_ref:
> > assertion 'G_IS_OBJECT (object)' failed
> > Dec 04 08:11:03 fedora gnome-shell[2088]: g_object_unref: assertion
> > 'G_IS_OBJECT (object)' failed
> > Dec 04 08:11:03 fedora gnome-shell[2088]: instance with invalid (NULL)
> > class pointer
> > Dec 04 08:11:03 fedora gnome-shell[2088]: g_signal_connect_closure_by_id:
> > assertion 'G_TYPE_CHECK_INSTANCE (instance)' failed
> > Dec 04 08:11:03 fedora gnome-shell[2088]: instance with invalid (NULL)
> > class pointer
> > Dec 04 08:11:03 fedora gnome-shell[2088]: g_signal_connect_closure_by_id:
> > assertion 'G_TYPE_CHECK_INSTANCE (instance)' failed
> > Dec 04 08:11:03 fedora gnome-shell[2088]: instance with invalid (NULL)
> > class pointer
> >
> > sometimes the segfault message appears slightly later, but it's always
> > around there. Again, there's no obvious reason gnome-shell would
> > suddenly have started segfaulting like this.
> >
> > As best as I can tell this started happening early on 2025-11-28. That
> > makes it seem like the most obvious possible trigger is gnome-shell-
> > 49.1-3.fc44 , which was tagged late on 2025-11-27. But that update
> > changed no code - it only bumped a dependency version:
> >
> > https://src.fedoraproject.org/rpms/gnome-shell/c/8e6d9a3cbaa5ee8b8c0e4e7977e5d9b12d437a0a?branch=rawhide
> >
> > so...I'm starting to suspect a toolchain issue, likely gcc or glibc.
> > I've updated the bug report that was initially for slitherer with info
> > on the gnome-shell crashes, and my tentative theorizing about triggers.
> >
> > Has anyone else noticed mysterious segfaults in stuff built in Rawhide
> > relatively recently (like, during November)? Anyone have any idea
> > what's going on?
> >
>
> There was an upstream kernel instabiltiy towards the end of the
> 6.18-rc cycle, this should be fixed now, I think rc6/rc7 were
> problematic, just mentioning in case it was relevant.
I don't *think* it's that, as it's still happening now, and the crashes
are always in the same place - it's not a "oh the system just randomly
fell over" kind of thing.
However, my neat theory above doesn't quite hold any more, since I
noticed the slitherer case happens on F43 as well as F44. So that
rather suggests it really is caused by qt 6.10.1 somehow - not just a
miscompilation of qt 6.10.1, or something - because we have an older
and stabler toolchain on F43, but the slitherer crashes started on both
F43 and F44 immediately after qt 6.10.1 went stable, for both releases.
I don't have a clue what's going on with the GNOME case yet though. :(
--
Adam Williamson (he/him/his)
Fedora QA
Fedora Chat: @adamwill:fedora.im | Mastodon: @[email protected]
https://www.happyassassin.net
--
_______________________________________________
devel mailing list -- [email protected]
To unsubscribe send an email to [email protected]
Fedora Code of Conduct:
https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives:
https://lists.fedoraproject.org/archives/list/[email protected]
Do not reply to spam, report it:
https://pagure.io/fedora-infrastructure/new_issue
