Choosing "Try Ubuntu" on daily-legacy image produced a crash

Bug #2036440 reported by Brian Murray
18
This bug affects 3 people
Affects Status Importance Assigned to Milestone
gnome-shell (Ubuntu)
Fix Released
Undecided
Unassigned

Bug Description

I was booting up a daily-legacy image from 20230918 and received a crash report after choosing "Try Ubuntu".

ProblemType: Crash
DistroRelease: Ubuntu 23.10
Package: gnome-shell 45~rc-0ubuntu3
Uname: Linux 6.5.0-5-generic x86_64
Architecture: amd64
CurrentDesktop: ubuntu:GNOME
Date: Mon Sep 18 15:20:58 2023
ExecutablePath: /usr/bin/gnome-shell
ExecutableTimestamp: 1694375959
ProcCmdline: gnome-shell --sm-disable --mode=ubiquity
ProcCwd: /home/ubuntu
ProcEnviron:
 LANG=C.UTF-8
 PATH=(custom, no user)
 XDG_RUNTIME_DIR=<set>
Signal: 6
SourcePackage: gnome-shell
UserGroups: adm cdrom dip lpadmin plugdev sambashare sudo users

Revision history for this message
Brian Murray (brian-murray) wrote :
tags: removed: need-amd64-retrace
tags: added: cuqa-manual-testing need-amd64-retrace
Revision history for this message
Apport retracing service (apport) wrote : Crash report cannot be processed

Thank you for your report!

However, processing it in order to get sufficient information for the
developers failed as the report has a core dump which is invalid. The
corruption may have happened on the system which the crash occurred or during
transit.

Thank you for your understanding, and sorry for the inconvenience!

Changed in gnome-shell (Ubuntu):
status: New → Invalid
tags: removed: need-amd64-retrace
Revision history for this message
Brian Murray (brian-murray) wrote :

Looking at the retracer log file regarding this crash something seems amiss.

09/18/23 15:39:06: fill_pool: retrace pool now: {2036440}
09/18/23 15:39:06: retracing LP: #2036440 (left in pool: 0)
09/18/23 15:39:13: retracing LP: #2036440 failed with status: 1
ERROR: Invalid core dump: "/tmp/apport_core_8b0m3n31" is not a core dump: file format not recognized

I'd have expected the core dump to part of the crash file.

Revision history for this message
Brian Murray (brian-murray) wrote :

I'm attaching the crash file which does have a CoreDump section.

information type: Private → Public
Revision history for this message
Brian Murray (brian-murray) wrote :

This bears some investigation into whether or not the core dump is actually a part of the crash file and if not why it isn't. We want to be able to provide crash reports from the live environment.

tags: added: rls-mm-incoming
Revision history for this message
Brian Murray (brian-murray) wrote :

I'm manually able to retrace the crash file from comment #4, although that fails, with the apport. Perhaps there is a period of time where the core dump is not attached to the crash file?

Revision history for this message
Benjamin Drung (bdrung) wrote :

Since this crash is about gnome-shell, maybe this is just another case of bug #2015857?

Revision history for this message
Daniel van Vugt (vanvugt) wrote :

Signal 6 is SIGABRT which usually means an assertion failed, which means the system log will contain an explanation. Next time it happens please run:

  journalctl -b0 > journal.txt

and attach the resulting text file here.

Changed in gnome-shell (Ubuntu):
status: Invalid → Incomplete
Revision history for this message
Brian Murray (brian-murray) wrote :

Couldn't we modify the apport hook for gnome-shell to gather the relevant log file when ever there is a 'Signal: 6' in the crash report?

Revision history for this message
Daniel van Vugt (vanvugt) wrote :

Indeed I've been meaning to learn about apport because I would like ALL bug reports to automatically come with:

  journalctl -b0 > journal.txt
  journalctl -b-1 > prevjournal.txt

Revision history for this message
Brian Murray (brian-murray) wrote :

gnome-shell already ships an apport package hook which is installed in /usr/share/apport/package-hooks/source_gnome-shell.py and that seems to be debian/source_gnome-shell.py from the source package. It should be relatively straight forward to modify that and an update to it is something the SRU team would likely (I say likely due to the install base of gnome-shell) as an stable release update to the package. It is something the Ubuntu Release team would definitely accept as an update post Feature Freeze.

Revision history for this message
Daniel van Vugt (vanvugt) wrote :

I meant all bug reports, not just gnome-shell. But gnome-shell is the best place to start.

Revision history for this message
Brian Murray (brian-murray) wrote :

Oh, in that case you'll want to look at /usr/share/apport/general-hooks/ubuntu.py which is provided by the apport package.

Revision history for this message
Daniel van Vugt (vanvugt) wrote :

I have more questions than answers now...

* It looks like source_gnome-shell.py already attaches journalctl output, so why didn't it here?

* data/general-hooks/generic.py also attaches journal errors, so why didn't it here?

* Do we really want all Ubuntu bug reports to have journal data that might be sensitive?

* Can we compress system logs before attaching them?

* Why is ubuntu.py source maintained in debian/patches/Add-general-hooks.patch instead of a standalone file?

* What's the criteria for going in ubuntu.py vs upstream https://github.com/canonical/apport ?

It all seems a little too hard right now. I have other things I need to be doing instead of learning apport and Python.

Revision history for this message
Brian Murray (brian-murray) wrote :

The gnome-shell crash is easily reproducible by choosing Try Ubuntu when using a mantic desktop legacy image.

Revision history for this message
Daniel van Vugt (vanvugt) wrote :

Unfortunately that log doesn't contain a gnome-shell crash.

Changed in apport (Ubuntu):
status: New → Invalid
Revision history for this message
Brian Murray (brian-murray) wrote :

We could create an /etc/apport/report-ignore/ file with the contents '/usr/bin/gnome-shell' so then at least the crash report dialog would not appear. However, after choosing try Ubuntu you presented with a black screen and some console output so its still not a great experience.

Changed in ubiquity (Ubuntu):
milestone: none → ubuntu-23.10
Revision history for this message
Brian Murray (brian-murray) wrote :

Could the gnome-shell crash be a consequence of how we are booting the Legacy image and starting the desktop environment? I noticed the following in the crash report "gnome-shell --sm-disable --mode=ubiquity".

Revision history for this message
Daniel van Vugt (vanvugt) wrote :

Oh sorry, I missed the "legacy" part. I'll try and reproduce it this week.

Revision history for this message
Daniel van Vugt (vanvugt) wrote (last edit ):

Bug confirmed but it does unfortunately result in bug 2015857, and no assertion mentioned in the system log.

I think the circumstances are consistent with bug 2036889 though (asserting on X11 session shutdown), so we should focus on fixing that first.

no longer affects: ubiquity (Ubuntu)
Changed in gnome-shell (Ubuntu):
status: Incomplete → Confirmed
no longer affects: apport (Ubuntu)
Revision history for this message
Daniel van Vugt (vanvugt) wrote :
Changed in gnome-shell (Ubuntu):
status: Confirmed → Fix Released
milestone: none → ubuntu-23.10
Revision history for this message
Brian Murray (brian-murray) wrote :

I confirm that the crash did not occur with the latest daily-legacy image, thanks!

ubuntu@ubuntu:~$ cat /media/cdrom/.disk/info
Ubuntu Legacy 23.10 "Mantic Minotaur" - Daily amd64 (20231004)ubuntu@ubuntu:~$
ubuntu@ubuntu:~$ apt-cache policy gnome-shell
gnome-shell:
  Installed: 45.0-1ubuntu2
  Candidate: 45.0-1ubuntu2
  Version table:
 *** 45.0-1ubuntu2 500
        500 http://archive.ubuntu.com/ubuntu mantic/main amd64 Packages
        100 /var/lib/dpkg/status
ubuntu@ubuntu:~$ ls /var/crash
ubuntu@ubuntu:~$

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.