- Wednesdays - 9:30am Pacific, 12:30pm Eastern, 16:30 UTC
- 650-903-0800 or 650-215-1282 x92 Conf# 8605 (US/INTL)
- 1-800-707-2533 (pin 369) Conf# 8605 (US)
- irc.mozilla.org #mobile for backchannel
- Beta 2
- Code freeze: Oct 19 (Yesterday)
Major Topics for This Week
- Primarily working on Beta 2 blockers
- Triaging the blocker list to move out non-critical bugs
- New Core (nee Android) theme is underway
- Working to get an acceptable level of theme in place
- Polish and tweaking will move to beta 3
- tracking theme work here (cleaned up today - should be easier to follow)
- using "polish" keyword in bugzilla for theme polish nitpicking
- some small theme patches going in today (?) which should take a few things off that list
- JPAKE/weave design work ongoing with beltzner, etc.
- opening some bugs (or adding testcases) to make themeing-useful CSS techniques faster
- new dlopen plus making sqlite not be stupid got our install size way down
- everyone thank mwu
- patch to make breakpad work with new dlopen waiting for review from mwu
- expect to land today
- file picker for chrome landed yesterday, content support landed today
- needs testing
- need to figure out l10n for java files
- added lots or polish to updater and notifications, pretty good experience now
- Android build bustage post mortem, by DougT
- Asked to push Tim's patch
- Applied it, and corrected small conflict
- Pushed to try.
- Android didn't compile.
- Got a new patch from Tim.
- Pushed to try
- For some reason, try didn't build Android this time.
- Locally imported patch to my android builder
- Built cleanly.
- Pushed to M-C, and left
- Failure Points:
- My mozconfig was for the emulator (disabled thumb2)
- Shouldn't have just left.
- High confidence on things would "just work"
- Try restricted builds for no good reason.
- I shaved.
- (blassey) I did too. Not sure why you didn't get a thumb2 build on try
- MacOS X - Content not rendering without user creating UI events
- Beta 2 Blocker
- Looks like mac isnt processing ipc messages it needs to
- Priority of child process (Alon can update everyone)
- Discussion in bug on dynamic verses constant child priority
- Constant priority makes the child slower (at least 3% on page load for one test).
- Dynamic priority is hard.
- I am bias.
- Content Sink pref change loaded (Bug 604762)
- Basically we make the child process less "responsive" and make it do more work before processing events
- Saw page load perf bump by 3%
- 3 tegras running 3 talos suites on m-c and tracemonkey only
- adding more suites over time
- scrounging up more tegras
- bugfixing as well -- please let us know if something's not working right
- currently tracking why the Talos results drop off of TBPL after test run is complete - bug 605856
- android multilocale -- untested poc manual build w/out packaged manifests
- 4.0b2 will be lsblakk
- aki pto Nov 20-Dec4; trying to hand over process before then
- bugfixing & documenting
- New Bugs over the week: filed
- Need a triage of bugs nominated over the past week
- Going through theme work
- Top issues from today:
- Seeing some good feedback coming in from Android nightly and Maemo beta users
- Results from nightly Smoketest runs
- DougT spent 2-3 days fixing talos.
- The Fennec CLH was registered using a prefix that loaded before talos CLHs.
- Alot of voodoo trying to get the right options, right config file, right files on disk.
- TPan doesn't work on desktop
- Tp4 doesn't produce the correct starting output on desktop
- Not all bad news, Ts does work on desktop
- I am pretty sure that these failures would also fail on the n900 given their subsystems (linux, apache) are basically the same.
- What I want is a way to locally test performance using the tools we currently have.
- (aki has considered rewriting talos in mozharness, which would hopefully make it easier to config, but this is not a small project and would require buy-in, also probably not as high priority as 4.0b2 releng blockers)