Notices / Schedule
- code freeze tomorrow at 11:59pm
- only a few blockers left though
- released last week
- contains lots of great stuff
- well received so far!
- code freeze for RC in late May required to hit June release window
- while we will aim to make RC1 perfect, previous Firefox releases have needed up to 3 RCs before we're ship-ready
- goal is to code freeze late in the week of the 18th
- we'll re-evaluate then based on remaining work to do
Since 1.9.1 branch ...
Past 2 weeks ...
- slow week due to all hands
- need to keep pressure and pace up!
- 6 blockers marked fixed1.9.1 over last week
- 21 bugs nominated for blocking over last week
- 14 patches approved for 1.9.1 over last week
- 77 blockers are FIXED but not yet landed on 1.9.1
- 38 bugs with approval1.9.1+ that are not yet landed on 1.9.1
Browser / Front End
- Blockers: 11 remaining & 5 nominations
- two tricky things (undo close window fallout, tab drag and drop)
- big props to Johnath and Dietrich for getting bug 486236 handled (again!)
- Polish update: Firefox is 55% shiny (+2% over two weeks ago)
GFX 1.9.1 Update
- 5 blocking bugs
- 4 of them are image rendering related, either perf or artifacts when zooming
- 1 is the nsWindow destroy bug, which our bandaid attempt didn't fix
- Roc is not available today
- 5 blockers on trunk (3 with patches)
- 10 untriaged noms (3 fixed)
- 0 blockers
- 19 blockers (6 with patches) (could be pruned, should not delay our schedule)
Mac OS X Update
23 blockers, around 4 with patches in hand. 5 noms.
These are bugs that fall outside of components covered by the Gfx, Content, Layout and JS groups:
Mobile 1.9.1 Update
- Initial docs with phases and SWAG schedule available.
- Initial focus will be responsiveness/stability, not security sandboxing
- We're looking for places where chrome JS currently touches content DOM directly... see mozilla.dev.platform post for more details.
jst: "we're going to need to remote arbitrary JS across processes to satisfy out of process plugins, so we may have the necessary machinery in place already"
- Decisions need to be made about how to approach the network stack (should we switch to chromium's network stack wholesale?), but I don't know the right way to have that conversation/make that decision (it's hard to know whether switching would reduce or extend the total time of the multiple-process work).
- why? "because the chromium network stack and IPC stack already work well together"
- but mapping that to existing necko API surface may be difficult
cjones: why do we want to centralize networking in the chrome process? Wouldn't it be better to have the content processes do their own networking? answers include: to avoid the weight of SSL/NSS in each process: to avoid complicating sharing activities for disk cache, cookies, and other shared state. we should evaluate this ourselves.
bent: we probably are going to have to make significant modifications to the Chromium IPC stack, because they use RTTI/exceptions ([pkasting] As a Chromium contributor I can tell you authoritatively that this is untrue; we don't use or enable RTTI/exceptions anywhere in the codebase.)
cjones: We should consider an IPC language, instead of a library, which we can use for typechecking and type safety, including protocol safety (e.g., read() only happens after open() and before close()).
- Fixed intermittent orange on talos, Tshutdown tests enabled in talos last week, details here
- infrastructure load for april
- more machines coming to help deal with wait times
- long downtime next week for firmware update on equallogic (held off from last week, and from before FF3.5b4).