Mir

CI test failures in GLMark2Test

Bug #1515660 reported by Alexandros Frantzis
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Mir
Fix Released
Critical
Alexandros Frantzis

Bug Description

We have recently started seeing:

[ RUN ] GLMark2Test.benchmark_fullscreen_default
[1447168375.534318] mirplatform: Found graphics driver: android
[1447168375.535051] mirplatform: Found graphics driver: mesa-kms
[1447168375.535204] mirplatform: Found graphics driver: mesa-x11
[1447168375.535876] mirplatform: Found graphics driver: dummy
ERROR: Throw location unknown (consider using BOOST_THROW_EXCEPTION)
Dynamic exception type: N5boost16exception_detail10clone_implINS0_19error_info_injectorINS_6system12system_errorEEEEE
std::exception::what: bind: No such file or directory

unknown file: Failure
C++ exception with description "Failed to start server thread" thrown in SetUp().
[ FAILED ] GLMark2Test.benchmark_fullscreen_default (10052 ms)

https://jenkins.qa.ubuntu.com/job/mir-mediumtests-runner-touch/7388/consoleFull
https://jenkins.qa.ubuntu.com/job/mir-mediumtests-runner-touch/7396/consoleFull
https://jenkins.qa.ubuntu.com/job/mir-mediumtests-runner-touch/7399/consoleFull

Tags: testsfail
Revision history for this message
Alexandros Frantzis (afrantzis) wrote :

This has become a CI blocker, so raising priority to critical.

Changed in mir:
importance: High → Critical
assignee: nobody → Alexandros Frantzis (afrantzis)
Revision history for this message
Alexandros Frantzis (afrantzis) wrote :

The problem is that we fail to stop lightdm/unity8 before running the test servers:

I: Turning on the display
initctl: unable to determine sessions
initctl: unable to determine sessions
I: stopping unity8
initctl: unable to determine sessions
initctl: unable to determine sessions
initctl: Unknown job: unity8
I: stopping lightdm
initctl: unable to determine sessions
initctl: unable to determine sessions
initctl: Unknown instance:

This is something we have occasionally seen before, but it never was as consistently reproducible as it is now.

Revision history for this message
Alan Griffiths (alan-griffiths) wrote :

    echo "I: stopping unity8"
    exec_with_adb_user "/sbin/initctl stop unity8"
    echo "I: stopping lightdm"
    exec_with_adb "/sbin/initctl stop lightdm"

Hmm. When working on the phone I tend to use just "sudo stop lightdm" alone - and that works fine in practice. Maybe the above should just be:

    echo "I: stopping lightdm"
    exec_with_adb "stop lightdm"

Revision history for this message
Alexandros Frantzis (afrantzis) wrote :

New hypothesis: In our scripts we are rebooting the phone a second time after we have installed the CI packages. Since the CI packages may introduce protocol/ABI/API incompatibilities with USC and Unity8, the second reboot can fail, in the sense that we boot into a system with a messed up state (init jobs that have failed etc).

In the following branch I have moved the second reboot before the installation of packages:

https://code.launchpad.net/~afrantzis/+junk/mir-medium-test-runner-for-jenkins-wip

I have triggered three mir-mediumtests-runner-touch jobs with the with the update test runner and they have all succeeded.

Revision history for this message
Alexandros Frantzis (afrantzis) wrote :

I have released the changes in https://code.launchpad.net/~afrantzis/+junk/mir-medium-test-runner-for-jenkins-wip to our main branch lp:~mir-team/+junk/mir-medium-test-runner-for-jenkins

Changed in mir:
status: New → Fix Released
milestone: none → 0.18.0
Daniel van Vugt (vanvugt)
Changed in mir (Ubuntu):
importance: Undecided → Medium
status: New → Triaged
no longer affects: mir (Ubuntu)
Changed in mir:
milestone: 0.18.0 → none
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.
Subscribing...

Other bug subscribers

Remote bug watches

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