CrashKill/Firefox4
General Agenda
- To resurrect the Crashkill effort targeted specifically at the 2.0 release. We meet weekly to triage bugs and discuss issues.
- We would like to focus the developers' efforts on addressing the most serious crashers particularly regressions.
- We will also be monitoring crash trends and reporting on general stability.
- We will use the keywords (reproducible, needs-info, testcase-wanted) to provide more detail on what is needed to move forward with the bug.
- We will make sure the module owner is cc'd on the bug so they can be assigned appropriately.
- Communicating status to the development team and others on the stability of the app.
- Old Crashkill notes and info can be found at https://wiki.mozilla.org/CrashKill.
Latest Stability Data
- Beta9 - 2.35 crashes per 100 ADU with 1,200,757 users.
- Beta8 (comparable) - 2.26 crashes per 100 ADU with 1,211,936 users.
- Trunk - Low volume of users ~15K users. Looking at new regressions.
Status
(Updated Jan 19th, 2011)
We are now in the final stages of the release so monitoring for stability regressions is key. A few things we are tracking closely....
Crash Regressions From 3.6.13
19 of the top 30 browser crashes are new in 3.6.13. Here is a complete list along with current status.
- Bug 590373 - Top of the list for Beta9. Appears to be a graphics card issue. Lots of discussion in the bug. FIXED
- Bug 595351 - js::mjit::EnterMethodJIT crash. We didn't think we could do much about this but we now have additional data that might help us track down the problem. Someone is working on it. Currently a softblocker.
- Bug 626631 - Fixing this might fix Bug 595975. FIXED
- Bug 620984 - FIXED
- Bug 626343 - Hardblocker, Final+.
- Bug 500105 - Not a blocker since can't repro and don't know the cause.
- Bug 615098 - Softblocker, Final+.
- Bug 612103 - FIXED
- Bug 584169 - Unconfirmed
- Bug 612098 - Can be closed once Bug 615475 is fixed.
- Bug 602225 - @ js::gc::MarkObject. Not a blocker, still needs to be looked at.
- Bug 601102 - Not a blocker, still needs some review.
- Bug 595740 - Fixed
- Bug 594596 - Related to Bug 591599. Both are not blockers.
- Bug 626195 - Not a blocker, some comments.
- Bug 591513 - Crash in older Betas, moved up in Beta9.
- Bug 605290 - Softblocker targeted for .x
- Bug 614250 - There is a related Bug 598498. It's unclear if we can do anything with this.
Plugin Crashes
(Marcia to help fill in stuff we want to monitor with plugins)
- Bug 572134 - This is the biggy that got fixed. Waiting for Beta10 to see impact on users. This should noticably reduce the Flash crashes.
- Crash data reviews shows a correlation to Adblock Plus with these crashes, although it was reproducible without AP.
- Bug 585913 - This was a problem introduced with the Flash release on Nov 3rd. We are working with Adobe to get this fixed but no timeline on that.
Trunk Regressions
These were the ones appearing on the top 20 list.
- Bug 625330 - Win64
- Bug 613716 - Win64
- Bug 626599 - fixed in Tracemonkey
- Bug 625753 - Might depend on Bug 625315 which is not a blocker.
- Bug 625247 - Fixed
- Bug 626841 - Fixed
Bug Lists
A bunch of queries we like to refer to.
- Top crashes - (keyword=topcrash and blocking 2.0=beta, final)
- All top crashers
- Top crashers - nominated
- Top crashers - reproducible
- Top crashers - needs info
- Top crashers - regressions
- Top crashes - not modified since 2010-09-01
- Top crashers - fixed (so we can go back verify in the next beta).
- Unassigned (nobody@mozilla.org, general@js.bugs)
- All crashes - (keyword=crash, topcrash and blocking 2.0=beta, final)
Analysis
Meeting Logistics
- Dial-in for the meeting is x290
- The Meeting is now every! Tuesday at 12:30AM till 1PM Pacific
- 650-903-0800 or 650-215-1282 x92 Conf# 290 (US/INTL)
- 1-800-707-2533 (pin 369) Conf# 290 (US)
- MV - Meet in Zombocom Conference Room on Floor 3.
- Backchannel on irc.mozilla.org is #planning