Platform/2009-09-08

From MozillaWiki
Jump to: navigation, search

« previous week | index | next week »

Notices / Schedule

Firefox 3.0.14 / Firefox 3.5.3

  • on track to ship tomorrow

Firefox 3.0.15 / Firefox 3.5.4

Blocker Report

note: queries have changed to use new status1.9.2 flag (want to know more about our brand new flags?)

Browser / Front End

GFX Update

  • Decode-on-draw (bug 435296) very very close - needed to make some large-ish changes at the last minute to address roc's review comments.
  • WebGL - canvas3d refresh patch landed. New patch with latest spec tweaks soon.

Layout Update

  • Cocoa printing implemented (mstange)
  • W3C reftest framework (jwatt, fantasai)
  • Unfiled bugs? Plugin tearing, personas
  • 1.9.2 approval requests
  • cpearce experimenting with VMWare record+replay: stability much improved, some serious performance issues, new hardware may help

Content Update

  • Slow week, not much to report (vacations, personal leave, ...)
  • HTML5 parser update (hsivonen)

Platform-specific Support Update

  • [josh] Still making steady progress on 64-bit Mac OS X builds, more issues like critical sqlite library conflict coming up.
  • [josh] Will give summary about Java for Mac OS X in Firefox 3.6 and up.
    • Java2 plugin not considered ready, not being used by Safari
    • Firefox 3.6 requires the Java2 plugin, we can hack together a way for it to find it
    • Java2 plugin causes some small regressions, as well
    • other option is restoring OJI, which would be a significant effort
  • [rob] Aero peek widget patch has final reviews. Working on tests.
  • [sid] Progress taskbar patches being updated to new interfaces.
  • [jimm] Jump lists patches progress waiting on some DDE installer issues currently being worked on.
  • [jimm] Completed Win7 image builds, need to get these tested this week.

JS

  • Should move 1.9.2 branch to newer nanojit code. More secure (no Fragmento), better performance, merging getting really tough in the tracer.
  • ES5 update (tracking bug bug 445494):
    • Strict mode: jimb (tracking bug bug 482298), now with 19 sub-bugs --- track just the details you care about! Parser-based stuff will likely make the code freeze; runtime aspects will take longer.
    • Meta-object interface: waldo (tracking bug bug 430133)

Startup Performance

Weekly update notes

Key points

  • Mac: Progress on warm startup - 3.6 is 10% faster than 3.5. 3.7 is 20% faster than 3.5. No numbers for mac ccld startup yet.
  • Windows: Flat on warm startup. Cold startup, we have a *regression* in 3.7 that we're working on identifying.

Need help

  • Windows cold startup oddity: My numbers showed that Firefox 3.5 somehow starts up cold *much* faster than previous or subsequent versions. Any input appreciated.
  • Need input from Mac folks on feasibility of bug 513954
  • Need an XPConnect/component-loader person to take bug 512827

Security

Electrolysis

  • init code 98% done: Moving on to implementing HTTP Channel sub-protocol for simple HTTP requests.
  • Electrolysis+Jetpack plans posted: https://developer.mozilla.org/en/Multi-Process_Architecture/Jetpack
  • Shout out to Josh Matthews who has done great work starting to hook up TypeAheadFind... it can highlight all the words now, though highlighting the specific word is still not working, and chrome doesn't know when a match was found
  • Plugin work ongoing: still trying to set up plugin mochitests for IPC and not-IPC cases
  • Haven't been tracking the following projects closely (will do better!):
    • windowless plugins (jmuizelaar/karlt)
    • JPW (mrbkap)

Tree Management

  • new Talos suite: Dirty profile MAX coming in next downtime bug 514070, bug 510587
    • dirty profile small/medium already live
  • l10n nightly updates for en-US now timely again bug 511901
    • the workaround is en-US as higher priority, l10n done in smaller chunks
    • still working out "real" fix, doing updates concurrently in the pool

Roundtable

  • introducing Bob Moss
  • 30-50% of Linux Electrolysis builds are failing due to out-of-memory linking libxul... bug 498393. Can we get attention to this?