Tracking bug for adding Account Manager to Firefox.

Requesting blocking for Firefox 4.

Final or are you hoping to get this into one of the betas Dan?

With such a large new feature like the Account Manager, it should probably go through at least one beta release before the RCs and final.

We are currently targeting the end of August for landing the final bits, though it will be in multiple patches so some features (e.g. sign in/out) should land earlier.

We can approve patches, but at this time I'm not comfortable blocking on the feature based on schedule. That should not be taken as a statement about not wanting the feature, just about keeping focus on issues that would actually block the release at this time. We can renominate when we have a better understanding of how close we are on the specific bugs being used to track the Account Manager feature.

Initial implementation that has UI/functionality to sign in for saved accounts and sign out.

Awesome to see this pulling together. Have you guys got a tryserver run that you can run through compare-talos to get us some early idea of performance impact?

Gavin: For the service files it may be useful to check in order: 1) helper classes 2) core service Then the ui bits for firefox integration and profile-specific: 3) ui bits 4) user/pass form profile Then various other patches can be somewhat individually looked at: 5) startup glue 6) header parser + generated 7) mochitest

I pushed several runs of Ts to maple with some baseline of the pre-merged changeset: (platform: 2 base runs, 4 merged runs, %overhead) linux: b456 b455 472 464 472 483 3.7% linux64: b429 b419 434 434 441 431 2.6% os x: b660 b663 660 672 672 658 0.6% os x64: b715 b787 793 806 801 799 1.6% winxp: b348 b341 359 361 360 359 4.4% win7: b454 b448 450 461 460 464 1.7%

Do we know what's consuming so much time on startup on XP and Linux? Or do we at least have an idea?

I don't have a specific answer, but we do have some juicy targets for Ts speedup. There are a couple of things we do right now that we shouldn't, and they possibly affect Linux and XP more than the rest - I don't know the platform differences well enough to say.

I reran the Ts tests (5 times) with the removal of the xpcom component and moved startup to the browser delayed startup: linux: 2.1% lin64: 2.4% macos: 0.5% mac64: 2.2% winxp: 2.9% win 7: 0.3%

We simplified the ui a bit to get rid of the inline svg/clip-path and reran Ts: From 2 base and 6 test runs (5 test runs for windows as it's still building..): linux: 1.7% lin64: 2.4% macos: 0.2% mac64: 2.0% winxp: 1.3% win 7: 0.5% That's an average of 1.35% across all 6 platforms, but we're probably hitting noise on some of the platforms, so that might not be totally accurate.

Comment on attachment 471259 [details] [diff] [review] ui f47972d05473 The popup notifications bit looks like it makes sense to me, although I'm not an expert :) However, for css rules using the popup notification id, you should be aware of bug 591026, which appends "-notification" to the id you supply when building the notification. Also, this patch may eventually cause a conflict with bug 567814 and bug 588309, which are moving save password behavior to popup notifications.

Comment on attachment 471263 [details] [diff] [review] test f47972d05473 I took a look at the sjs files and the tests in general. They all looked sane to me, but I don't know anything about how the account manager works, so this feedback should be taken with a grain of salt.

Comment on attachment 471261 [details] [diff] [review] makefile f47972d05473 Nit: license header copied w/o updating the info ;) >+DIRS = \ >+ browser >+ $(NULL) You're missing browser/Makefile.in. Looks good otherwise.

Comment on attachment 471259 [details] [diff] [review] ui f47972d05473 Oops, missed the feedback flag!

I talked with thunder today, and this is going to be changing. With that being said, whoever ends up actually implementing this should chat with me about approaches since I'm *very* likely to be on the hook for reviews.

We're changing direction and putting the previous Account Manager protocol on hold for a while. Closing these bugs to reflect reality.