Mobile/Testing/11 13 13

From MozillaWiki
Jump to: navigation, search

Previous Action Items

  • bc to file bugs for skia gl
  • callek and bc to discuss sutagent 1.20 deployment
    • Result in bug 933842 comment 20
    • Specifically after an adjustment patch/review/landing Callek will get the new SUTAgent into staging
  • blassey to look at bc's inventory and give feedback about possible os upgrades: http://bclary.com/blog/2013/07/19/autophone-inventory/
    • likely better to obtain new devices rather than upgrading -- want a variety of devices and os levels

Status reports

Dev team

  • status of eideticker and autophone reporting to TBPL?
    • depends on TBPL2, stood up at the end of Q4, switch over at the end of Q1
  • status of debug tests?

Rel Eng

  • (pmoore) setting up releng infrastructure to support new HD chassis.
    • pete got some higher priority work elsewhere at the moment, but is still owning this process
  • (armenzg) continuing work on x86 automation, see corresponding section below.
  • (Callek) stopped months-long running buildbot processes on Friday, after edmorley discovered pandas not disabling when slavealloc state changed.
  • (Callek) Continuing to progress on meta bug 932231 - "many panda masters do not have pandas attached"
    • Fixed bug 936615 - "Pandas should start buildbot when mozpool thinks the device is ready"
      • This bug not being fixed is what caused the tree closure over the weekend as panda after panda was losing connection to buildbot, good summary of the exact cause/effect case here is in Bug 932231 c#5
  • Releng has a Team Week next week in Boston. Callek/Joduinn to be reached by cell in case of an emergency during the week.

IT

A Team

  • tegra failure rate: [1.52%]
  • panda failure rate: [11.38%]
    • excluding retries: 2.2
  • Can't attend the meeting this week (in Portland for a separate meeting)

x86 automation

  • All tests running hidden on Cedar, with some failures
  • Complete:
  • Working on:

Autophone

  • The gs2 and gs3 phones did not have their ipfw pipes setup. I have made sure that all pipes are set from the inventory.
  • Does using the ipfw network conditioning provide any value? When testing the regression ranges for bugs 819984, 825612, 916099 with a different host without using ipfw, I was able to detect the regressions. I would like to turn ipfw conditioning off as it slows the tests down and does not provide value.
  • The new arrangement with the phones on the metal shelves was causing issues connecting to the registration server and to the local phonedash instance. I have remove the shelf and placed all the phones together next to the router. It appears to be working more reliably again.
  • I ran the s1s2 tests using Mark's old phones for the regression ranges in the following bugs and have enough information to determine the changesets.
    • bug 819984 - Regression in "time to throbber start" - Dec 5
    • bug 825612 - Regression in "time to throbber start" and "time to throbber stop" - Dec 22
    • bug 916099 Regression in "time to throbber stop" on Aug 28 2013
  • I have setup the autophone Canvas SkiaGL tests and am making a preliminary set of runs using Mark's old phones to evaluate their condition and determine the phone partitioning. Will start posting to Autolog and filing bugs by Thursday 11/14.

Eideticker

  • Updated dashboard UI will be deployed shortly (should be more pleasant to look at and use)

Round Table

  • bug 916258 - canvasmark failures on panda boards. appear to be resetting most system services from what looks to be a OOM. This results in about 1 retry per push (32 retries Nov 11th on Inbound, imagine all the branches and we have about 75-100 retries/day).
  • Ctalbert met with Appurify They would like to set up a trial for us to see if we could use their service.
    • Probably for performance, given that we often need specific performance targets on specific phones
    • They have wired up web page test for mobile, which is also interesting.
    • They would be willing to add firefox to their set ups and they have already proved that they can get our tests running (they ran mochitests on samsung sII for instance).
    • Let's discuss this.
  • wlach working on adding inbound support to mozregression for desktop (bug 935183). Useful for Android too?

Action Items

Retrieved from "https://wiki.mozilla.org/index.php?title=Mobile/Testing/11_13_13&oldid=764740"