CrashKill/2013-03-18

General

What's up with all the recent backfills?
- bug 850782 - combination of late ADU arrivals over the last couple of weeks, and a problem with crontabber not re-running the ADU job for 24 hours. Actively working on a solution.
Potential metrics maintenance window this weekend - will go to CAB tomorrow for approval. Need feedback about feasibility from this group.
Last week saw some DST issues, now resolved.
Hang reports from about:crashes are still redirecting to crash-analysis. When do we want to change this?
Email-on-crash is now working, occasional failures (bug 851511)
Planning Q2 goals: please come up with a list of what is wanted in Q2

The newest 11.7 beta fixed the most grave crashes of the initial one, a few regressions still remain, mostly bug 849622, bug 851876.

KaiRo did some investigation on increases of empty dump crashes, see bug 837835, looks like we regressed in 19 and again in 20.

Nothing notable, but crash rates remain on the high side (empty dumps play into that, see above).
No change in ongoing issues/concerns (D3D/D2D11-6.2/Win7 812683/819028, Dual-Intel/NVidia 806786, Crossrider 836263, Norton 840106).

Fix for bug 787158 (SPDY) seems to have worked.
bug 798274 (gfxContext::PushGroupAndCopyBackground) is now rising somewhat on beta.
bug 836263 (Crossrider) has changed signature as expected and seems to be a bit lower in b5 than previously.

bug 836951 (nsSecureBrowserUIImpl::MapInternalToExternalState) and bug 837370 (js::types::TypeCompartment::addPendingRecompile) seem to be somewhat declining even though no fix landed.

Several new topcrashes:

bug 850741 (js::AutoMaybeTouchDeadZones::AutoMaybeTouchDeadZones)
bug 851806 (js::ShapeTable::search)
bug 851807 (JSObject::setLastProperty)
Already fixed: bug 851851 (nsWindowSH::GlobalScopePolluterNewResolve)
- bug 752828 - android.database.sqlite.SQLiteDatabaseLockedException: database is locked at android.database.sqlite.SQLiteStatement.native_executeSql(Native Method) on ICS
  - happens in release, STR in the bug.

bug 846465 (libstagefright, HTC devices, JB) still main worry on beta.
bug 823236 (mozilla::layers::ShmemYCbCrImage::IsValid) and bug 827171 (nsPluginStreamListenerPeer::OnStartRequest) can be confirmed as fixed on 20.0b5
Overall crash rate on Beta 20 looks good, and startup crashes are notably low compared to the usual picture of Firefox for Android.
bug 845734 - crash in mozilla::MediaPluginHost::CreateDecoder @ libcutils.so@0x50.. on Sony Xperia T, V and Z running JB
- Dev blocked by needing device, QA trying to resolve by getting Dev access to device through device anywhere
bug 812881 - crash in nsMediaPluginHost::DestroyDecoder @ libsomxcore.so@0x1826 on Samsung Galaxy SIII with qcom/samsunggolden/espresso hw running JB
- need information from dev (see comment 33)

Uplifts:
- bug 803791 - maybe I'm not seeing this right... is this not in beta? Is beta locked down?

Top crashers in Nightly:
- bug 845867 - crash in mozilla::layers::LayerManagerOGL::WorldTransformRect @ libGLES_hgl.so@0x4.... on Samsung Galaxy Y and Ace with Broadcom VideoCore IV GPU running Gingerbread
- bug 851806 - crash in js::ObjectImpl::nativeLookup @ js::ShapeTable::search / bug 812319
- bug 834243 - crash in mozilla::layers::LayerManagerOGL::Initialize with abort message: "We need a context on Android"
  - shows in beta as well/wontfix for beta
Top crashers in Aurora :
- bug 847021 - android.database.sqlite.SQLiteException: Can't upgrade read-only database from version <n> to 14 at android.database.sqlite.SQLiteOpenHelper.getReadableDatabase(SQLiteOpenHelper.java) at org.mozilla.gecko.db.LocalBrowserDB.filterAllSites
- bug 852063 - java.lang.OutOfMemoryError: bitmap size exceeds VM budget at android.graphics.BitmapFactory.nativeDecodeByteArray(Native Method) at org.mozilla.gecko.AboutHomeContent.getThumbnailsFromCursor(AboutHomeContent.java)
Top crashers in Beta :
- bug 782223 - crash in libGLESv2_tegra.so@0x4ae.. on Asus Nexus 7 (NVIDIA Tegra 3) with JB
- bug 808378 - crash in nsMediaPluginHost::CreateDecoder @ libOMX_Core.so@0x1... on devices with rk30board HW and Mali-400 MP GPU running JB

Remember that we are - by design (mobile data costs) - only sending crash reports via wifi. QA people need to connect to wifi to send reports (queued up ones should all be sent in that case). We know we will not get many crash reports for users in Brazil where wifi is scarce.
Woohoo! Starting to see some report counts in 22:
- https://crash-stats.mozilla.com/topcrasher/byversion/B2G/22.0a1/7
- still have to rely on search : https://crash-stats.mozilla.com/query/query?product=B2G&version=ALL%3AALL&range_value=1&range_unit=weeks&date=03%2F18%2F2013+17%3A18%3A41&query_search=signature&query_type=contains&query=&reason=&build_id=&process_type=any&hang_type=any&do_query=1

bug 850845 - Camera - crash when trying to open a second camera instance
- MC crash fixed; QA needs to double check and confirm.
Socorro will need some UI changes for B2g to make analyzing it easier (ie different gecko / gaia / firmware )
main crashes we are seeing are decoding images, camera, sms, bluetooth and misc system crashes.