Platform/2009-07-07

« previous week | index | next week »

aiming for builds early next week
blocking1.9.1.1 and approval1.9.1.1 flags should be used
goal of this release should be a quick-turnaround that:
- topcrash fixes, security/stability patches
- narrow scope, small changes
- contrary to some reports on the Internet, this is the usual process for Firefox and software releases; the 3.5 release was strong, stable and solid, and feedback has been extremely positive. Near the end of the release we become extremely conservative about patches to accept; the 3.5.1 release is a quick update to fold in some patches that came up late in the 3.5 release cycle.

Firefox 3.6a1

code freeze 21-July?

Blocker Report

Gecko 1.9.1.1 / Firefox 3.5.1

Browser / Front End

posted the Q3 team goals for Firefox on the wiki, as well as the evaluation of last quarter's goals
Firefox.next
- moving on to TSnap bugs which aim to make Firefox observably faster
- some of us are working on WinCE compatibility issues
- also returning to polish bugs and exploratory sprints
- Reminder to review the Firefox.Next plan and add your comments/feedback on the discussion page or in the dev-apps-firefox thread

Polish update: Firefox is 59% shiny (-1% change)
- 43 remaining easy polish bugs (whiteboard [polish-easy])
- 39 remaining hard polish bugs(whiteboard [polish-hard])

pub?key=pMZGKUlD9NOPg4oJGs1CUMw&oid=1&output=image&foo=.png

GFX Update

bug 753 work continues apace, with several people building on it.
- If you are doing work that depends in any way on imgIContainer, gfxIImageFrame, nsIImage, you should base them on bug 753.
Decode-on-draw work is happening in bug 435296.
Jeff's working on an event profiler that has shown some bugs in imagelib; once those are resolved there will undoubtedly be others.
#4 topcrasher bug 470487 fix should now be possible that nsWindow refactoring has landed. There have been very few, quite easy to fix regressions from that refactoring.

Layout Update

Nothing major to report since last week
SVG/MathML in HTML5 parser rocks
HTML5 parser creates giant text nodes, should raise priority of work on performance of layout of very large text nodes

Content Update

HTML5 parser landed! (preffed off -- about:config pref is "html5.enable".)
peterv's slimwrappers landed, moving on to further optimizations that can be done now that the main work enabled.
Ben Newman is working with Henri Sivonen on figuring out what still needs to be done with the generated code in the HTML5 parser (cleanup of the generated code etc, so work on the generator itself).
Josh is almost (completely?) done with XPCOM plugin interface removal and interface consolidation in the plugin module.
Starting to work out what the security model for JetPack's should look like.
LiveConnect R.I.P.

Mac OS X Update

Steven Michaud fixed major OnStopFrame crasher, bug bug 499600.
Markus Stange is very close to having a completed patch for fullscreen in 1.9.2, bug 420491.
64-bit Mac OS X port blocked on gfx bug 493280.
Cocoa NPAPI event model coming along, part of it landed, maybe 70% complete at this point.

JS

landed https://bugzilla.mozilla.org/show_bug.cgi?id=200505. big perf win on some benchmarks.

Static Analysis

Dan Witte will be working on static analysis for code cleanup, performance, and correctness improvements
for instance, bug 502775, bug 172937, and more seriously JS GC safety - bug 421934
suggestions for analysis passes we could use are welcome

Security

Fuzz-testers Martijn Wargers and Paul Nickerson are visiting Mountain View this week. We'll be scheduling meetings with lots of groups, such as the people working on the HTML5 parser.

There might be areas that need fuzzing that we're not aware of. If you need something fuzzed, find us near the QA area this week!

Electrolysis

Phase I completed. Repository is in place and builds on Windows/Linux.

cjones still work on minor protocol issues and bsmedberg hooking up automated ipdl generation.

bent working on xpcshell extensions that will allow scripting the remote process for testing and prototyping.

bsmedberg doing local buildbots until RelEng can hook up a full project... reporting to tinderbox.

Tree Management

tp4 enabled on mozilla-central last week
- tp4 == 100pages, c15mins to complete; tp3 == 400pages, c60mins to complete
- still to enable on other branches, nom'd 1.9.1.1 bug
- once tp4 rollout completed, want to power off fast talos machines, and EOL tp3
new graph server (graphs.m.o), old is now called graphs-old.m.o
- new dashboard at [1]
working with IT to power up more machines in colo; needed for places, electrolysis, wait times.

Roundtable

Everyone should read other teams' goals.
keeping trunk stable for the next few months (beltzner)
- landings that affect broad areas of the code should be discussed here and in dev-planning, first
- landings with many regressions should be backed out instead of waiting for follow-on fixes
- goal is to stabilize for 3.6 alphas/betas upon which Fennec will likely ship
major update from 1.9.1-nightly to trunk-nightly? (beltzner)
triaging blocking nominations earlier in the cycle? (dbaron)
- getting regressions fixed when people still remember the code that caused them
- having a smaller pile at the end of the cycle
- keeping mozilla-central in more-shippable state
Can we use a multi-state flag for blocking-1.9.2? Proposed states (bsmedberg):
- ?
- -
- next alpha
- last alpha
- beta
- final