Mobile/Testing/05 16 12
From MozillaWiki
Contents
Previous Action Items
- ctalbert to double check that the throbberstart/throbberstop stuff didn't land monday with the reliability fixes- this did land last monday and accounts for the 'win' in performance numbers.
- Armen & Joduinn to write a plan for what needs to happen so that one of us (ateam or mobile dev) can work on getting x86/armv6 builders in place (possibly using a loaner machine)- need to define list of tools and ensure our builders have all the tools, mozconfigs need to be checked into the tree.
- Callek and Jmaher to investigate hanging bcontroller.py processes on trobo test - bug 752222, work to cleanup bcontroller.py zombie processes.
- Ajuma to follow up this week with fixes to reftest harness in bug 748088 - patches exist, 67 failures exist.
- Wlach having issues getting sut agent to install apk's on ICS, but it seems we can fix it by adding LD_LIBRARY_PATH setting.
- FIXED: See bug 753484
Status reports
Dev team
- Trying to improve Robocop reliability
Rel Eng
- SUTAgent 1.08 deployed in the last week, some minor hiccups, but is deployed fine now.
- screen resolution changes on Friday
- Callek has some work coming up to stop buildbot after every run and verify before starting up again
A Team
- [total failure] rate is 15.89%. Lower than last week, and this includes a handful of red jobs due to the sutagent upgrade on Monday:)
- working on reftests with ajuma, 67 failures with browser.tabs.remote=false need attention
- sutagent 1.08 deployed, no resolution changes for crashtest and jsreftest now!
- robocop based talos tests troublesome, failing a lot, I saw local crashes and the minidump was going to the /data/data directory instead of staying in the profile, verified the environment variables are correct.
S1/S2 Automation
- Phones went down on Sunday's power outage.
- Updated the code to better recover from sdcard issues
- Restarted the phones yesterday
- Building a staging environment for autophone so that our production system isn't our staging system as well.
Eideticker
- Dashboard finally getting new numbers again! http://wrla.ch/eideticker/dashboard/
- Working on adjusting tests for different screen sizes on Galaxy Nexus, will create new dashboard for that platform after that's done (probably comparing fennec vs. stock at first, with Chrome coming shortly after)
- Priorities are still being tracked here: https://etherpad.mozilla.org/eideticker-priorities
- Does anything need adjusting now that the beta has been released? e.g. Should we drop testing of the XUL version, and set the beta as the new "old" version?
Panda
- [wlach] I filed bug 751347 about qualifying the graphics performance of the Pandaboard ES using Eideticker. Thoughts on the priority of this?
- New image built and available for download
- Used a new method to avoid the "button dance" during flash.
- Still have reboot issues and mac address issues with this build.
- But Linaro produced builds do not have the reboot issue. I think that we need to have a look at what kernel Linaro is basing their builds on and we should change to use that (the image above used the pre-compiled kernel from the android source checkout).
Round Table
- who is on point for bug 754837 and what are the next steps after the security review is complete?
- review will be scheduled today
Action Items
- ctalbert to file bug for nagios to monitor autophone and eideticker machines
- ted to ensure armv6 and x86 mozconfigs are checked into m-c for builds
- armen to schedule releng time for getting x86/armv6 builders going
- Callek and Jmaher to investigate hanging bcontroller.py processes on trobo test - bug 752222, work to cleanup bcontroller.py zombie processes.
- ajuma/jmaher - get the 67 failures commented on in the manifest so we can turn on reftests with browser.tabs.remote=false