« previous week | index | next week »

Notices / Schedule

• Firefox 3.6.4 schedule is now TBD (see the blocker section)
  • Will likely firm it up tomorrow as issues start landing / new issues slow down
• Firefox 3.5.10 is still tied to Firefox 3.6.4
• Investigating alternatives if Firefox 3.6.4 issues keep appearing (turning OOPP off for certain plugins, etc).

Blocker Report

Firefox 3.6.4

8 issues we are currently tracking as blockers:

• Bug 558986
  • Crash on touchscreen / tablets
  • Needs more investigation

• Bug 564260
  • Has a fix, waiting on review
  • Initial review mentions the fix is scary

• Bug 564861
  • Reported by Microsoft
  • Has a fix which has landed on m-c
  • Need MS to confirm it fixes this and bug 559760

• Bug 563377
  • Has a fix which has landed on m-c
  • Blizzard verified it fixed his issue

• Bug 563847
  • Topcrash for 3.6.4
  • Has a fix that has been reviewed, needs to land

• Bug 559760
  • Originally thought it was a Microsoft issue
  • Likely a dupe/the same as bug 564861

• Bug 561818
  • Follow-on to a bug we fixed in build 3
  • Needs more investigation/STR

• Bug 564389
  • Not a Firefox problem, but tracking until resolution
  • Said they are whitelisting the files early this week
  • Did an ALE request as well
  • Need to do this again if we rebuild

Firefox 3.5.10

• None currently

Browser / Front End

Planet Firefox loves you!

• Addons manager relanded!
• Work started in earnest on Home Tab/AppTabs
• Startup timeline thread on dev.platform
• Firefox 4 Product Plan

GFX Update

• Benoit Jacob (bjacob) started in the Toronto office this week! Initially, he will be working on WebGL.

Hardware Acceleration Update

• A D2D correctness fix (large image support) landed last night, and along with it (strangely) came a lot of performance wins too.
• Mac layers (in the form of full-screen video) should be in soon, with a couple of known bugs we're working on.
• EGL layers (currently on Maemo) are also coming along.

Layout Update

• Need feedback from XUL authors on www-style as Tab Atkins refines his flexbox proposal
• dholbert working on SVG images
• Retained layers plus scrolling working pretty well on Mac (non-GL). A little extra overhead for simple pages, clear performance wins for complex cases, should really fly with GL. Patches trickling out for review. Need to work on other platforms, fix issues, get on trunk ASAP. Lets us test scrolling with reftests,l e.g. using mozReftestInvalidate and setting scrollLeft/scrollTop [roc]

Content Update

• Implemented most of multi-process Jetpack API (bnewman)
• Various performance work under way (bug domperf).
• MessageManager being ported from e10s to trunk (smaug)

Platform-specific Support Update

JS

JSD plans: https://wiki.mozilla.org/Platform/JSDebugv2 (now with more concrete estimates)

JM plans: https://wiki.mozilla.org/JaegerMonkey#Major_Optimizations current status: http://blog.mozilla.com/dmandelin/2010/05/10/jm-halfway/

GC plans: https://wiki.mozilla.org/JavaScript:SpiderMonkey:GC_Futures

Electrolysis

• Have diagnosed bug 558986. Have no clue how to fix it. It's a race/deadlock between COM/OLE cross-process RPC and Mozilla RPC. Helpwanted!

• bug 564415 says there are many more hardware interrupts running Flash OOPP than IPP. Is there some hardware/profiling expert who can help diagnose/explain this?

Startup Performance

Taras' Update

Security

Breakpad

Tree Management

• Networks Are Fun! Updates on MPT <-> Castro woes:
  • VPN implementation was changed on Thursday to fix the 8hr disconnect problem.
    • New setup was causing disconnects
    • Friday night started to revert back to old implementation.
    • Saturday morning finished reversion. STILL seeing disconnects.
    • To reduce WAN traffic, RelEng tried two changes:
      • Set up new master in Castro so that machines wouldn't have to traverse VPN for most test jobs. Migrating machines to new master took until Monday morning.
      • Disconnected fast build machines in Castro from masters in MPT. No way to move the masters to Castro as there's a mix of build machines in MPT and Castro. Disconnecting these machines also significantly reduces build capacity. This now reverted.
    • Random disconnects gone, but were seeing new DNS failures
  • Meeting with IT Monday morning; Derek changed VPN implementation again.
    • Seems to be much more stable now.
  • Not finished yet; now we're seeing cache corruption that we didn't see before.

• Fedora oranges
  • status update:
    • 2/10 test suites left to greenize (mochitest-other and reftest)
    • mochitest-other debug has been enabled
    • latest status update if wanted to read
  • Machines have been provided for devs to test their patches.
    • If Armen had plenty of time he would love to help/learn with devs how to test patches but I can't anymore (Windows unittests is now on my plate).

• More minis arrived mid-last week
  • this will improve wait times for new 64bit OS. 55 online already bug 557294

Roundtable

• Firefox 4 Product Plan (beltzner)
  • Read all about it! if you are surprised, email me (beltzner@mozilla) or send anonymous feedback
  • will be setting up mechanisms to track against this plan and tie it to blocking bugs, etc
  • what else would help make it real?
• TabCandy (aza & beltzner)
  • quick overview of what it is
  • needs faster image scaling - what's the cost there?
  • who can we talk to about this sort of stuff?
  • video!