CrashKill/2014-03-31
From MozillaWiki
Contents
General
Previous Action Items
- [DONE] juanb to file xtls.dll crash (release, beta) - bug 987531
- Naoki to try and contact Geeksphone on their 1.5 crashes
Socorro
- For some fun with crash comments, see Adrian's "dogeboard"
- Work Week last week
- Supersearch via API has landed, will ship later today or tomorrow morning.
Flash
Desktop
- BitGuard crash bug 985922 is #1 on 27 but hasn't been spiking as much as expected when we unthrottled 28 again.
- Spike across channels: bug 989992 mozilla::dom::XULDocument::OnScriptCompileComplete or js::frontend::TokenStream::getTokenInternal with jid1-vW9nopuIAJiRHw@jetpack add-on
Beta
- Total crash volume is up ~0.3-0.4 crashes / 100 ADI compared to 28.
- Did comparisons of Beta 28 and Beta 29 via explosiveness last week, those regressions show up:
- bug 988548 (Intel gfx driver via D2D and canvas2d) is pretty large, probably #1 in sum, half of the whole regression
- bug 970362 (F-Secure)
- bug 988961 InitFromBailout (bug 985228 is similar but about trunk tests)
- a few net::Seer crashes (either need new bug or bug 987248) - juanb to file a new bug for those.
- bug 988619 is expected side effect of security fixes (probably moving other crashes to that signature)
- bug 976536 (GC) is waiting for Till, who is back from vacation
- bug 856796 (Yarr) existed before but regressed significantly
- bug 976181 (nsContentUtils::IsCallerChrome) now assigned to bholley
- bug 987531 (xtls.dll) occurring on multiple branches, external cause
Aurora
Trunk
- bug 619558 has enabled GGC on trunk with the Nightly build of 2014-03-29, please expect and report new JS crashes, some signatures like those involving "Nursery" are a dead giveaway, but other areas can be affected.
- Two possible related bugs this morning: bug 990090 and bug 990082 (lizzard)
- sqlite update crashes seem to be fixed, fast work! bug 986577
Android
- Increased crash rate seems to be due to devices that had h264 blocklist that received OS upgrades so that the blocklist does not apply
- updated topcrash-android keyword on bugs
B2G
General
- a lot of crashing occurring on Geeksphone for 1.5, 1.4; tried to contact them no response yet.
- 1.5 has crashes but no symbols so we can't take action.
- need symbols for various devices still
- is there any way we can easily get dev to push their symbols? Do we want that?
- Flame Build tested : no white screen of death crashes.
1.5
- bug 970403 - crash in android_atomic_add
- happens in 1.4; driver issue. Devs need to update
1.4
- bug 989408 - crash in mozilla::layers::PLayerTransactionChild::SendPLayerConstructor(mozilla::layers::PLayerChild*)
- occurs in 1.3; need to see if it happens on 1.1
- bug 988105 - getUserMedia crashes on seccomp enabled builds on 1.3
- bug 986250 - nsLocalFile::Create() crashes B2G content process
1.3
- bug 945335 - crash in audioTrack_callBack_pullFromBuffQueue
- bug 986365 - Crash in supplicant because of closing supplicant while waiting for event
- Dev investigating
Notes
- KaiRo has a blog post (written in the first week of March) on long-term graph and efficiency of the stability program, but wants OK/review from PR and they ended up being busy first with GDC and releases and now with something different. Hopefully this will resolve soon. Dynamic long-term stability graphs also exist now but will point the group to them only once the blog post is out.
Upcoming PTO
Action Items
- Naoki to try and contact Geeksphone on their 1.5 crashes (waiting for reply from GP)
- juanb to file a new bug for the net::Seer regressions in 29