Monday, February 9, 2015
[ eae gave high level update on planning meeting; separate announcement on that to follow ] We added three new team members to this week which is very exciting! Welcome aboard, we're glad yo have you!
- James Maclean (wjmaclean) in WAT, new to project warden.
- Paul Meyer (paulmeyer) in WAT, new to blink and warden.
- Koji Ishii (kojii) in TOK, new to the team but has already been helping out with text rendering performance for awhile. Performance Tracking (benjhayden)
- Refined android profiling pipeline, similar to existing linux one.
- Results match that of linux with a few exceptions, more time spent in powf() and loading fonts. Almost 3% of total layout time is spent in
powf() translating color spaces for fonts.
- Started to look into FontCache on android which doesn't seem as effective as on other platforms. Scrolling (skobes) [crbug.com/417782]
- Smooth scrolling now works the same for overflow divs as for frames. win_blink flakiness (dsinclair)
- Worked to reduce win_blink flakiness by triaging flakiness and timeouts.
- Flakiness went from 40% -> 10% (!), it is now one of the faster bots.
- Much of remaining flakiness due to font antialiasing/subpixel rendering differences. Rename Rendering -> Layout [crbug.com/450612] (dsinclair, bsittler, eae, hartmanng)
- All non-prefixed files renamed. 188 prefixes files remaining. (dsinclair)
- Most of the linbox tree has been moved. (hartmanng) List markers (dsinclair)
- Trying to figure out how to deal with an ancestor being moved. Opera has some ideas and there might be multi-col logic that can be reused. Text (kojii, eae)
- Make RenderCombinedText less obstructive to the line breaker and paint code. Mostly donecrbug.com/433176, selection paint issue crbug.com/438852 and a bit more cleanup still remains. (kojii)
- Complex path perf issue that had been stalled due to the slow legal permission is finally clear now, hope to re-start with Dominik this week. (kojii)
- Folded line_break into TextBreakIteratod, removed unused logic and de-templetized a bunch of functions. (eae)
- Difference in preferred width between text and containing element caused regression on Facebook due to strict overflow handling. Has fix but writing layout test is difficult due to simplified text-rendering pipeline for tests. (szager)
- Fixed ZWJ handling on simple text path on Mac. (eae)