Platform/2012-10-02

« previous week | index | next week »

Notices / Schedule

• FF15 beta 5 was released last Thursday
• FF16 beta 6 went to build earlier today, will be released in the next 1-2 days
• FF16 sign-offs will be held tomorrow. Still investigating:
  • bug 789075 – Video playback causes browser to stop responding (mostly) or crash in Cleopatra tool
  • Remaining stability issues

• CTP blocklisting (FF17) and the OS X 10.5 EOL hotfix (FF16) will be in external testing over the next couple of days

• FF16.0's final build will be Friday, shipped on Tuesday
  • ARMv6 support will remain on Beta (FF17) for further testing coverage
  • ESR10.0.8 will be shipped alongside 16.0

• Merge day is Monday, late morning PT. Make sure any changes you'd like uplifted are on mozilla-central (not inbound)

Firefox OS

• Feature freeze. Some features straggling in, being tracked heavily. Everyone else focusing on performance, stability and polish.
• Reconfiguring the ecosystem post-feature-freeze - pushing on new Gaia developer requirements: unit tests, reviews, automation, bugzilla.
• Coordinating with engineering managers who are donating resources, and coordinating with Developer Engagement outreach for external contributions.
• Prototype devices: If you're a developer working on Boot2Gecko, request an Unagi device on ServiceNow, explaining that it's for development not for dogfooding program.

Firefox Development

• Stub installer landed! Great work by Robert Strong and Brian Bondy in bug 322206
• Frank Yan and Stephen Horlander landed the front-end parts of our initial stab at HiDPI support (for Mac Retina displays): bug 781327
• Mark Hammond, Shane Caraveo, Jared Wein and Felipe Gomes continue to rock it on Social API fixes
• Margaret Leibovic landed telemetry for web page load time, as measured by the Firefox front-end: bug 790213
• Good number of front-end random orange bugs were fixed by Tim and Dao: bug 789709, bug 789003, bug 774009, bug 789102, bug 794436

Firefox Developer Tools

Work week in London last week. Progress was made on chrome debugging, b2g web console, source maps, the developer toolbox, and other small improvements.

Blog posts:

• http://antennasoft.net/robcee/2012/10/01/firefox-devtools-hax/
• http://blog.astithas.com/2012/10/debugging-firefox.html
• http://paulrouget.com/e/devtoolsMeetupUpdate/

Add-on SDK

• Main focuses for Q4 will be mobile support and landing in our APIs in Firefox
• Landing in Firefox plan

Performance

• This week's Snappy status

• Jeff has a prototype of bug 792199 that only decodes visible images immediately when switching to a tab
• Necko team landed bug 769764 gets rid of nasty proxy-related issues
• Brian's work to prevent cache lost on crash (bug 105843) shows the following expected improvements
  • Windows: 9.3% -> 2.88%
  • Linux: 22.34% -> 6.07%
  • Darwin: 11.85% -> 2.96%
• Robcee is working to integrate the SPS profiler into dev tools - defaulting to JS only (bug 795268) http://cl.ly/Jloc

GFX

• Images are now downscaled using a high-quality scaling algorithm (bug 486918).
  • This is disabled on mobile, where we worry about speed/multicore abilities.
  • It's also disabled on OS X, which has high-quality downscaling built in.
  • There have been some regressions (crashes mostly), but we hope to keep it on for Fx 18.
• Q4 GFX goals; important parts:
  • Deliver a D3D10 OMTC implementation for the Windows Metro product.
  • Work on OMTC for other platforms.
  • Work on the Azure project to improve drawing performance.
  • All of this is subject to B2G project needs.

JS

• Work week in Mountain View - October 22nd through October 26th.

Layout

• CSS3 Flexbox implementation landed: bug 666041
  • Currently disabled via a build flag by default, pending followups for correctness & conformance with later-game spec-changes. (tracked in bug 783409)
  • Soon to be controlled by a pref instead of build flag (bug 796212), so that nightly testers can turn it on.

Media

• Most remaining pieces of WebRTC are about to land (most are r+ already)
  • DataChannels and SCTP
  • PeerConnection DOM
  • media/mtransport
  • media/webrtc/signaling
• Initial landing will be preffed-off by default; we hope to pref-on for FF19

DOM

WebAPI

• majority of work landed for B2G feature freeze
• sms:, tel:, and mailto: URI schemes landed

Network

Identity

We've got a B2G implementation plan for ID (and Payments).

https://bugzilla.mozilla.org/show_bug.cgi?id=794552

Need some help on PopupManager (+ crashing B2G Linux desktop build).

Plugins

• Update on click-to-play eng status (dkeeler)

Mobile

• progressive tile by tile drawing landed, pref'd off (pref: "layers.progressive-paint")
• safe browsing landed and pref'd off

Accessibility

• No Audible.
• Helping FF OS priory work: implementing crash reporting (hub)
• Good progress on mobile a11y.
• Fixing oranges.

Tree Management

• armv6 and armv7 testing is enabled per checkin on all branches; armv7-no-ion is per checkin on m-c, m-i and try only
• new record # of test jobs run on September 26: 41,598 test jobs in one day.

Security

Bugs needing info for security action

No results. 0 Total; 0 Open (0%); 0 Resolved (0%); 0 Verified (0%);

Stability Report

Socorro

Desktop

18.0a1 Nightly

• bug 795594: Extremely huge (startup) crash spike with Windows 64bit, probably MS compiler issue with PGO, bug 795748 turned off PGO for Win64 for the moment to fix for now.
  • Made us aware that running no tests at all for Win64 means that we don't even check if builds start up, despite us having quite a few Nightly users on Win64.
  • Discussion brought up that while we're running tests for tier-1 by-checkin builds (and implicitly testing startup through them), we're not gating pushing Nightlies (or daily Aurora builds) and their updates on running any tests on those builds and therefore can push updates that don't even launch.
• Unassigned new crashers: bug 795728 (GFX, probably from DLBI landing), bug 771407 (layout, recurring DLBI regression), bug 795819 (layout, probably from DLBI landing)
• bug 795940 is a new crash that has a patch up for review, probably also fixes bug 795942.
• bug 793175 is a high-rank topcrash (#2/#3 depending on the view) with a reviewed patch up, but not landed and no comment for a week. bug 758531 is probably a duplicate.
• bug 795829 crash in mozilla::layers::Layer::CalculateScissorRect
• bug 795826 crash in mozilla::`anonymous namespace'::ContainerState::Finish
• bug 795819 crash in mozilla::FrameLayerBuilder::BuildContainerLayerFor
• bug 795749 crash in mozilla::FrameLayerBuilder::DrawThebesLayer
• bug 795728 crash in mozilla::layers::ContainerLayer::DefaultComputeEffectiveTransforms

Beta

• bug 795683 crash in mozilla::plugins::PStreamNotify::Transition with abort messages: "__delete__()d actor" or "corrupted actor state" mainly with FlashGot
• bug 795675 crash in JSC::Yarr::interpret on JB
• bug 795962 crash in F412627940
• hangs dominate the landscape of Top 20 crashes in Beta 5

Aurora

• bug 795104 crash in _VEC_memcpy | js_NewStringCopyN

Flash

• New labs version of Flash was released on 9/26 - version 11.5.x

Mobile

Nightly :

• 9 out of top 10 are startup crashes
  • mostly due to bug 790139 has better symbols; freetype font issue.

Aurora:

• Aurora crash rate seem to have spiked up a little
• bug 777460, bug 791441 are on the rise in Aurora
• lots of libxul crashes. Not sure what to do with them as of yet.

Beta:

• Crashes are almost as good as Release crashes. Because the ADU are lower comparatively to Release, Naoki believe that the crash rate may be lower in the release channel once beta is pushed to release.

Release :

• bug 756140 something we've talked about.

QA:

• a lot more libxul crashes are shown at the bottom of nightly and other channels which is disconcerting.
• bug 752828 looking at the patch from bug 741224, it might be helpful to get the API levels for the crashers?
  • Need STR from QA.
• bug 750217 is not fully fixed.

Monitoring:

• bug 745512 #6 on release, not shown in beta and above.
• bug 774622 has already been pushed about 10 days ago to aurora and beta.
• js::gc::MarkObjectRange, ion::Cannon and java.lang.NullPointerException: at org.mozilla.gecko.BrowserApp.onPrepareOptions are nightly only crashes that have been fixed.

See Mobile Notes for Mobile specific Socorro notes

• 15.0.1 Release top crashes
• 16.0b4 Beta top crashes
• Aurora top crashes
• Nightly top crashes

B2G

• B2G crashes are starting to show in socorro!
  • We still need to get symbols for these crashes to make them useful (which requires builds coming from releng infrastructure).
• Discussions on UX flow nearing final decision, will need implementation soon.
• Work is ongoing on getting content crashes to work and getting annotations up - see dependencies of bug b2g-crash-reporting.

Roundtable