CrashKill/2014-09-29

General

Previous Action Items

• [KaiRo] follow up with relman and gfx on what we can do about 33 crash rates
• [KaiRo] discuss with lonnen and bsmedberg what we can do to make it valuable and possible for them to make the CrashKill meeting (scheduling? agenda?)
• (from last week) [bsmedberg] Talk to e10s devs to prioritize e10s stability issues seen on Nightly 35
• (from last week) [bsmedberg] Talk to GFX team about beta graphics issues
• [KaiRo] Continue discussion with tchung on who will cover FxOS while Naoki is away.
  • [nhirata] follow up to MTBF test; purposeful crash should be obvious crash signature + channel change : add -test-<testsuite> ie -test-MTBF.
• [KaiRo] finish up discussion with relman on prerequisites in approval questionnaire

Socorro

• Advanced Search is targeted to be killed. If you use it for anything that is more complicated or otherwise difficult to do with Super Search, please make sure there's a blocker to bug 1052761 filed on that.
• no changes to production last week

Desktop

Beta

• OMTC on Windows:
  • 33.0b7 with OMTC off comes down to almost the same level we had in late 32 beta in overall crash rate.
  • Bas has a patch that should fix a lot of OOM|small cases that are not really OOM, see bug 1072404
  • For b8, that patch will be taken as well as turning OMTC back on in bug 1074045
  • A few signatures (gfxUtils::EncodeSourceSurface, gfxASurface::SetSubpixelAntialiasingEnabled, CDevice::ID3D10Device1_CopySubresourceRegion_) were rising up with b7, but probably just non-OTMC issues that will be solved with enabling it again in b8.

Aurora

• Main issue (#2 topcrash after OOM|small) is still bug 1064107 crash in RtlEnterCriticalSection | LockImpl::Lock()
• bug 1074196 sending crash rates up for 34 (startup crash)

Trunk

• bug 1074196 is a new URL classifier crash that spiked yesterday and is #1 for that day on both Nightly and Aurora.
• OOM|small numbers are down significantly after bug 1072404 landed.

Android

Anthony Hughes

• bug 1072332 crash in java.lang.NullPointerException: at org.mozilla.gecko.db.PerProfileDatabases.getDatabaseHelperForProfile(PerProfileDatabases.java)
  • #2 in Nightly -- bandaid landed on 9'26 but crashes are still appearing in the latest Nightly builds
• bug 1069369 Android crash in mozilla::image::RasterImage::DecodeSomeData
  • #1 in Nightly but dropping significantly since the fix landed on 9'26 (down ~6%)
  • Overall crash rate is down to ~25c/100i vs ~40c/100i last week vs ~10c/100i two weeks ago

Notes

• Proposals for sessions in coincident work week:
  • What would the ideal Socorro look like (for your/our purposes)?
  • QA CrashKill work: What are we doing? What should we do? What are we missing or overdoing?
  • QA: How can we get our next Scoobidiver?

Action Items

• [bsmedberg] get traction on URL cassifier bug 1074196