Mobile/Testing/01 09 13
From MozillaWiki
Contents
Previous Action Items
- [callek] bug 824764 - turn off reftest for panda
-
[kmoir] bug 825984 - turn on panda for try/inbound -
[kmoir] turn on 198 panda boards -
[jmaher] add x86 automation status to meeting notes -
[wlach] file a bug and investigate "push-to-eideticker"- DONE: See eideticker section below
- [jmaher] talk to socorro folks about reporting crashes from automation
- they suggest submitting reports as normal and we can work out any custom reports
- [jmaher] look into core dumps from android
Status reports
Dev team
- unit test support landed; see https://wiki.mozilla.org/Mobile/Fennec/Android#cppunittests
- bug 803158 if no crash report is generated by a tegra ... use ndk-stack to get a stack
- bug 810471 Crash on startup during various tests
- bug 811793 Get at least a minimal subset of Android XPCShell tests running on tbpl
- robocop failure investigations: bugs 770483, 813107, 817440, 824067
- bug 761987 (nssCertificate_Destroy crash) -- developing a logging / try based approach to getting more info.
Rel Eng
- Second set of pandas in production.
- tests enabled on try and m-i
IT
- Tracker bug - 800 production ready panda boards bug 799922
- Remaining open bug is a tracker for RelEng work bug 805040
- Working on a higher density chassis bug 777393 - This work has restarted.
- Building a staging instance of Mozpool to test new versions and stage upgrades
- Includes "fake" pandas for a realistic Mozpool environment without using physical pandas
- Preparing to upgrade to Mozpool-2.0.0, supporting SUT checks against devices, automated device self-testing, active monitoring of unused devices, fixed CORS implementation, and internal implementation improvements
A Team
- Tegra - [10.33%] total failure rate
- many of these were due to bug 827474 - org.mozilla.fennec_ hanging around
- Panda - [11.74%] total failure rate
- rck
- [1] bug 761987
- [2] bug 761987
- [3] bug 761987
- [4] (timeout)
- j1
- [1] bug 761987
- [2] (timeout)
- [3] bug 761987
- [4] (infra)
- j2
- [1] bug 761987
- [2] (timeout)
- [3] bug 761987
- j3
- [1] bug 761987
- [2] bug 761987
- [3] bug 761987
- [4] (automation)
- rck
- updated sutagent to work with firefox 19 on mozilla-beta
- waiting on approval in bug 828099
- updated talos to support mozcrash and latest breakpad binaries, iterating a few times- starting to see more crashstacks
- will start looking into posting to socorro (live) from automation to bring better visibility to the crashes we are seeing.
x86 automation
- haven't gotten started yet. [bc]
Autophone
- Ready to deploy unit tests with 'newparser' to phones in haxxor. Need to coordinate with ctalbert.
Eideticker
- Hardware in MV was flakey so there are some gaps in the data. Everything humming along again now though.
- New, faster, easier, better test + dashboard code landed (bug 820924). Load times should be way lower, plus adding new tests is WAY easier than it was before.
- Filed bug 826084 for push-to-eideticker.
- Question: would a "bisection" mode for eideticker (where it could automatically download and get results for intermediate builds on inbound and central) provide a similar benefit at a lower cost in terms of time?
Round Table
Action Items
- [callek] bug 824764 - turn off reftest for panda
- [jmaher & jchen] look into core dumps from android
- [gbrown & jmaher] investigate replacing libc and libdvm with custom compiled versions we have symbols for
- [bc] to file bug to track x86 automations
- bug 828568 - [tracking] Get unittests running on Android x86 [Hardware]
- bug 828571 - [tracking] Get unittests running on Android x86 [Emulator]