Platform/2009-10-13

From MozillaWiki
Jump to: navigation, search

« previous week | index | next week »

Notices / Schedule

Firefox 3.0.15 / Firefox 3.5.4

  • Firefox 3.0.15: builds made, were targeting beta for Wednesday, but have to re-spin for one issue
  • Firefox 3.5.4: builds made, beta on Friday, caught two issues which require a re-spin

Firefox 3.0.16 / Firefox 3.5.5

  • no schedule yet
  • please work on your blockers though!

Firefox 3.6b1

  • had some problems with Aero Peek tab previews on Windows 7, still several functional problems, but not crashing anymore
  • currently going through QA

Firefox 3.6 RC

  • fix yer blockers

Blocker Report

See more Firefox 3.6 related blocker queries, or learn about the new status and blocker flags

Browser / Front End

(Progress reports every weekend on Planet Firefox)

GFX Update

  • 0 beta blockers; 3 1.9.2 blockers - all with patches.
  • Jeff got his 0x0 stack unwinding improvement, bug 519616, landed in upstream breakpad; we're waiting on that fix to be taken into our instance of breakpad, bug 521231.
  • Bas is starting a DirectWrite font backend for Cairo's GDI backend, which will make it possible (on Windows 7, and whenever it's backported, on Windows Vista) to have better-quality antialiasing on text.

Layout Update

  • 1.9.2 blockers
      • 17 unfixed in layout, 4 with patches
      • 3 in video, all with patches
  • CSS Transitions landed (dbaron): landed with support for animations of lengths, percentages, and colors, floats added later in the week, and more properties hopefully coming soon (input on which are important is accepted)
  • Added support for animating float values in transitions/SVG (dholbert)
  • mats landed support for drawing the caret (not blinking) in reftests so that we can test caret position
  • Mac timing weirdness on machines with no screen (mstange): when there's no screen attached to the machine and VNC is not connected, certain graphics calls cause pauses, which means our Mac talos numbers respond weirdly (and not the way real Macs do) to changes in the pattern of graphics calls

Content Update

  • Closed all our P1 blockers, many thanks to Steven, mrbkap, Andreas, Waldo, and bz.
  • Continuing to work on topcrash bugs.
  • HTML parser
      • Off main thread parsing works
      • Speculative image, script, and stylesheet loading works.
      • Starting to think about what's remaining to enable by default.

Platform-specific Support Update

  • tab previews landed on trunk/1.9.2.
  • jump lists landed on trunk.
  • progress indicator platform code landed on trunk/1.9.2.
  • Still waiting on progress indicator browser patch tests for landing, bug 474060.
  • perf regressions, merge problems, bug fixes have been addressed.
  • (smichaud) Patch landed to restore JEP on 1.9.2 branch bug 517335

JS

  • chasing recording perf issues, having some success
  • recursion regressions getting fixed
  • ES5 parser changes landing

Startup Performance

Highlights from the weekly update post, and work started this week:

  • Summarized table of progress - will update this week, since doesn't have some major landings from the last week.
  • Ryan landed bug 475289 to lazily initialize the engineMetadataService, bug 520342 to consolidate service initialization in the microsummary service, and pushed bug 499123, coalescing browser about: pages, to the 3.6 branch.
  • Taras landed a bunch of optimizations to the fastload system in bug 412796
  • Enabled PGO for Places, mozStorage and SQLite in bug 419893
  • Ben Hsieh got review on bug 511761, which optimizes the fastload cache invalidation, and consolidates some of the invalidation triggers, ready to land.
  • Alice continued work on bug 510587, getting us ever closer to automated testing of cold startup.
  • Jonathan Kew has a new patch on bug 519445 to improve font-loading on the Mac, with a project 10% improvement in warm startup time.

Join us on IRC in #startup.

Security

  • Fix security holes! Just because they aren't "blockers" for a specific release doesn't mean they aren't important.


feature review date who interested
Windows TSF integration (1.9.2) unscheduled Jim Chen, roc
DNS Prefetching unscheduled Patrick McManus Jesse, bz, reed(?), ctalbert
New system metrics (and media queries) unscheduled ? dbaron


Electrolysis

  • typeaheadfind works, in the testbed at least
  • work on windowed and windowless plugins continues
  • IPDL improvements including supports for array and shared memory types depends on refactoring the lowering pass: cjones on it
  • the async drawWindow is reviewed: needs to land, and the API bits need to land in mozilla-central probably
  • out-of-process plugins now working well enough on linux to play flash games and watch youtube without crashes (mostly ;) )
  • status reports available at http://benjamin.smedbergs.us/weekly-updates.fcgi/project/electrolysis (there's a feed too)
  • need to get memory-mapping working in order for realistic testing of fennec responsiveness... no owner yet
  • ETA for JPW wrapper (mrbkap)?

Bugs needed for plugin testing:

  • bug 519568 - Talos metrics for plugins
  • bug 519570 - leaktest+electrolysis; count leaks for each subprocess
  • bug 519572 - ensure test frameworks expect subprocesses

Tree Management

  • waiting for go on 3.0.15, 3.5.4 respins
  • waiting for go on 3.6b1
  • accidental loss of mobile perf data results. Now restoring.
    • details in bug 521895
    • will be missing data (from 02oct-05oct approx?)

Roundtable

  • Most of our European community relies on these weekly minutes & notes to catch up on the development priorities for the gecko project, but they find the notes fairly incomprehensible.
    • These notes should not just be an agenda, but also include conclusions and "what we need help with".
    • Consider using wiki transclusion to include other wiki pages here. Use <onlyinclude> tags to restrict transclusion to only certain sub-sections.
  • Suggestion to be more aggressive about backouts for regressions
    • if a patch causes a perf or functional regression, it needs to be dealt with immediately
    • if it can't be dealt with immediately (developer doesn't have time, or can't be reached) it should be backed out
    • joduinn to look into sending emails to committers upon completion of Talos run