Platform/2010-06-15
From MozillaWiki
< Platform
« previous week | index | next week »
Contents
Notices / Schedule
- Firefox 3.6.4
- The release -> beta channel update was successful
- Release channel went from 325k -> 124k in 3 days, build #6 users specifically 211k -> 30k
- Built build #7 for bug 562198 on Friday
- Released build #7 to beta users yesterday afternoon
- Plan is to release build #7 as Firefox 3.6.4 final on Thursday, 2010-06-17
- Will have small uptake, but build #6 baked for a long time and the changes on top are fairly small
- this date may change again pending the feedback from Socorro team on recent crash stat confusion
- laura to report to the Product Planning team at tomorrow's meeting
- The release -> beta channel update was successful
- Firefox 3.5.10
- Tied to 3.6.4
- Firefox 3.6.6
- Announced the 1.9.2.5 -> 1.9.2.6 change on planet and planning
- Will be coming up with some scenarios and schedules this week for review
- Mozilla Developer Preview 5 (aka 1.9.3a5)
- Released yesterday (6/14/2010).
Blocker Report
- Firefox 3.6.4
- Nothing blocking, but watching this query to make sure nothing is missed
Browser / Front End
Working towards code freeze this week, lots of theme change, run those nightlies!
- Do we know status on Retained Layers, when we think it'll be available? We need it for the extension bar + app tabs UI. The UI team would like to do something like this
- jdrew thinks this is bug 130078 which is a b1 blocker
- roc says that he thinks that will work, but not sure if requirements have changed
- limi and faaborg to get in touch with roc and jdrew and follow up offline and work through requirements
GFX Update
- Harfbuzz landed! bug 449292
- What's Harfbuzz? I'm glad you asked! I talked about it last week!
- It's turned off by default. To turn it on:
- Set gfx.font_rendering.harfbuzz.level to 1; harfbuzz will be used for "simple" scripts (Latin, Cyrillic, CJK) that do not require specific complex shaping.
- You can set it to 2; Harfbuzz will then be used for all scripts (even though the results will currently be broken).
Hardware Acceleration Update
- OpenGL and OpenGL ES acceleration won't make it for Q2.
- To talk more about this, please participate in the dev.planning thread.
- To help out - and we would LIKE YOU TO HELP - please look at bug 565833.
Layout Update
- WebM landed and shipped in 1.9.3 alpha.
- Need to work on performance, especially non-accelerated scaling; will probably use more of the Chrome code [doublec]
- Working on 'buffered' TimeRanges; implemented for Ogg and Wave, working on WebM [cpearce, kinetik]
- Harfbuzz landed, currently off by default [jkew, jdaggett]
- Retained layers basically ready for review, just wrangling a few test failures [roc]
- Optimized scrolling of background-attachment:fixed in hard cases (faster than Safari)
- will be put up for review today, no need to delay!
- Reorganization of frame coordinate systems for zooming just about ready for review [tnikkel]
Content Update
- Necko e10s changes getting close to landing in mozilla-central (jduell)
- Remote XUL and XBL is being disabled for beta1, more communication about this around the corner.
- Websockets test infrastructure for non SSL websockets (ws://) is close - tracking down two failures, one on debug linux and one on opt mac, might need a content expert on hand to help with debugging but for now, it's in hand (see jgriffin for details). Still planning to land early next week.
- beltzner: do we want to hold off landing before the test infrastructure is working?
- jst wants to think about it
- jonas points out that manual tests can be run, clint pointed out that test runs are passing, just got some infra issues on getting them on all systems
Platform-specific Support Update
JS
- About to turn on conservative stack scanner.
- No longer necessary to explicitly root stack values.
- JM compiler with fat values has caught the speed of the old one, without as many optimizations applied.
- Compartmentalized heaps coming. Should make "GC pauses" much shorter.
Electrolysis
- Two issues getting F10S running on mochitest - bug 572149 and bug 572152.
- bug 572149 - JS crash in MessageManager.loadFrameScript() we need a hand debugging and fixing this. See jmaher for details.
- bug 572152 - Starting last weekend, we could suddenly write to an nsIFile object from content in E10S mochitest builds. This seems wrong. Is this expected behavior?
- Fennec trunk (mobile-browser) is currently being ported to use message manager and run on either mozilla-central and electrolysis branches. Work list
- Nightly blocked by bug 571438
Startup Performance
Security
Breakpad
Tree Management
- New mozilla-1.9.3 / mozilla-2.0.0 branch work blocked on figuring out which version # to use
- beltzner: use mozilla-1.9.3 for now, please!
- Please use TryServer wisely
- LOTS of people using bigger, shinier, TryServer (normal "15-35 per day"; 91 on thursday, 137 on friday)
- Please do NOT just repush to TryServer if you need to re-try one test suite. File a bug! Details here
Roundtable
- Beta 1 progress (beltzner)
- Are we happy not including the XPCOM changes (and any related Gecko version bump) in beta 1? (Mossop)
- Some feedback from Linuxtag (Pike):
- Getting rid of restarts, updates not in startup is a perfect match with people
- Hard to tell how much of perf comparison with chrome is apples-to-apples, both in platform integration and extension ecosystem/capabilities. Recent blog and forum posts helped.
- Having plans for shoveling tasks to other processes and GPU is good, should deliver something noticeable in Fx4
- Adblock+ is something like a Firebug, in that people evaluate "the browser that run adblock+". Recent perf blog by Wladimir helped a bunch, but people don't mind that much who's at fault. Can we expose better APIs for him?