Platform/2010-04-13

« previous week | index | next week »

Notices / Schedule

• Firefox 3.6.4 code freeze was last night @ 11:59 pm PST
• Had great convergence last night, thanks!
• bug 558390 was found late last night and is now the long pole. It has landed on m-c. We want to test some m-c builds with it in before putting on the branch
• bug 558226 (loc changes) is waiting for bug 558390. Once bug 558390 is in, we'll get bug 558226.
• We are still following Releases/Firefox_3.6.4, though depending on build times we may have today's build pushed back to tomorrow
• 3.5.10 is on the backburner for now

Blocker Report

• Two open blockers for 1.9.2.4
• Bug 558390 was found late last night and is now the long pole. It has landed on m-c. We want to test some m-c builds with it in before putting on the branch
• Bug 558226 (loc changes) is waiting for bug 558390. Once 558390 is in, we'll get bug 558226

Browser / Front End

As always, planet firefox for the details! In the meantime, some notables:

• Restartless add-ons
• Extension manager UI redesign
• Services.jsm, CSS Transitions, OOPP crash UI tweaks
• Notifications redesign
• UX Update

GFX Update

• Cairo update is still blocked on bugs.

Hardware Acceleration Update

• Direct2D and OpenGL have both gotten faster over the past week. Please do test and file bugs.

Layout Update

• dbaron has patches posted for css3-values "calc" bug 363249 (awaiting review)
• dholbert posting patches for <animateMotion> bug 436418 today or tomorrow

Content Update

• TestXUL runs with multi process Necko! (jduell)

JS

• been killing lots of crufty, bug-inducing detritus lately: __count__, a variety of obsolete getter/setter syntaxes that nobody uses, __parent__ coming soon too
• gal working on iteration speedups; seeing large perf wins on sunspider fasta test, both interpreter and on trace
• JM progressing too

Electrolysis

• Fennec with SysV shared memory and X shared memory (for speed at scrolling) is up and running.
• [Mac] Fixes to Core Graphics, added Core Animation support (all under review) bug 557534 bug 555281 bug 556052
• [Mac] Crash reporter fix landed today. bug 557011
• [Mac] Affected by mouse input degredation bug 558629

• Last-minute bugs found from 3.6.3plugin1 fixed
  • plugins suck: quakelive was implementing cdecl exports instead of stdcall, and we just happened to work because our calling function had a frame pointer. After refactoring, it didn't. Whoops. bug 558390

• Top crasher in OOPP in branch (3.6.3plugin1, 3.6.4pre) we need to look at:

https://bugzilla.mozilla.org/show_bug.cgi?id=558647

• test-ipc.xul now mostly-kinda works with remoted networking! Lots to do still, including redirect handling, lots of SSL info handling, etc.

Startup Performance

• View snapshot results for startup and all other tests.
• More details on the project page.

Estimated win	Bug #	Summary	Owner	Status	Notes
~10%	"Dirty Profile" Startup Reduction	Constant startup relative to profile size	Shawn Wilsher
significant	bug 558200	Extension profiler platform api	Steve Fink		There is an easy way to provide feedback to extension developers to help them figure out why their extension is misbehaving. We need to implement that ASAP, so typical firefox installs(with lots of extensions) can be fast

Breakpad

• No news.
• 23 patches awaiting review from Google to give Mac OS X CFI-based unwinding and non-broken 64-bit support.
• Jim will be taking a sick dog to the vet, and may be late to the meeting.
• Ted is also out.

Tree Management

• scheduled downtime Saturday 17th April: 2pm-8pm. (2-5pm for netapp; 5-8pm for spinning back up everything)
  • closing all trees, including taking hg down, for netapp upgrade bug 502151
• unittest-on-talos project
  • need help with last of Fedora orange tests bug 554934 bug 558521
  • ok to turn off unittest-on-builder for linux, osx?
  • still working on rollout of winXP, win7
• osx10.6 64bit and linux64 builds/test/talos see summary here
  • still working on adding to release-automation; will not be part of FF3.6.4

Roundtable

• bug 502301 - initializers using ints often need explicit casts to compile with c++0x, but that would allow us to use u"" literals on newer GCCs and eventually stop using -fshort-wchar