Here's a short summary of Mozilla-related work I've done in weeks 46/2011 (November 14 - 20, 2011):
- Mozilla work / crash-stats:
Kept following and taking part in discussions on a 8.0.1 chemspill release for crash issues in 8.
Did some testing of WebGL demos on my tablet and found a crasher there as well as possible crash reporting improvements.
Tested the upcoming Socorro release on stage and found that the signature summary still needs improvements.
Provided more feedback on ongoing Socorro work.
Filed a bug on dealing with development branches in Socorro - this will come too late for current "birch" work, but we should get it in the future.
Figured out why my gut feeling on crash rates in future build-based Nightly/Aurora reports was correct, now with mathematical proof.
As usual, I also watched new/rising crashes closely and filed bugs for a number of those.
Helped Edmund get commit access.
Reviewed two build system patches that nobody else wanted to review.
Uploaded 2.5 versions of EarlyBlue and LCARStrek to AMO, now waiting for reviews.
Started work on the 2.6 versions of those themes, which match the new betas of SeaMonkey and Firefox, and should have most of that work done already, so I hope that this time I can get reviewed themes out around the actual release of the browser versions.
- Various Discussions/Topics:
SeaMonkey machine work progressing, more aggressively blocking bad add-ons/plugins/libraries in the future, Fennec native UI project and its fallout, SQLite and places performance, default-compatible add-ons but not themes, WebAPI progress, MPL2 progress, putting tablet to use again, buying and setting up my new Nokia N9 phone, etc.
The story about the Firefox 8.0.1 update that we are sending to Mac users and users of Firefox 4-7 now was one that heavily involved our CrashKill team as it's all based on basically 3 crash problems, 2 of which we could fix on our side (working around an Apple Java problem and blocking old and crashy Roboform versions - we couldn't [yet] fix Windows System Restore to be reliable for Firefox). Therefore, we have been looking into those issues heavily and had tons of email and bug traffic on them with QA and release managers as well as affected developers in the last two weeks. We came away with a number of lessons and the notion that we need to do a lot more about crashes being caused by third-party software, as all of those problems were not actually in our own code. Also, roughly half of all our Firefox crashes right now happen because of third-party software, and we can't let the situation continue that those issues make us look bad. Of course, we value the extensibility that people can combine their offers with ours, but we need to react and pressure them to do updates when their stuff crashes, or else block their software loading in our products and crashing them, as that makes life bad for users.
On a positive note, though, it looks like we fixed something in late October on Nightly that caused Flash hangs and crashes, and right now both Aurora and Nightly show significantly fewer of those issues - will be interesting to see if that holds up when Firefox 10 goes to beta and later release.