941796 - (ExactRootingB2G) [meta] Exactly root B2G builds of the browser

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Andrew McCreight [:mccr8]]

This will be needed for compacting and generational garbage collection on B2G.  This will affect C++ code that handles JS pointers that isn't built as part of the desktop browser.

Comment 1

[Steve Fink [:sfink] [:s:]]

Attachment: Enable exact rooting by default on b2g

This patch enables exact rooting for B2G. r? glandium for the mechanism that I'm using to turn it on. r? nbp for deciding the timing of this stuff.

We could go straight to ggc, or we can do exact rooting for a week or so, then add in ggc. I've been told nbp is probably the best person to coordinate this with. What do you think?

Comment 2

[Steve Fink [:sfink] [:s:]]

There is one possible true hazard in b2g, fixed by bug 1017419. I'll land that before this one.

I am still working on automating the analysis for b2g and publishing it on tbpl (bug 898554). For now, I can continue to run it locally to detect regressions, but we're seeing very few hazards currently, and it doesn't seem like the relevant code is changing much.

Comment 3

[Steve Fink [:sfink] [:s:]]

Kicked off a try push at https://tbpl.mozilla.org/?tree=Try&rev=4267fa5c6a7d

Comment 4

[Mike Hommey [:glandium]]

Comment on attachment 8430957 [details] [diff] [review]
Enable exact rooting by default on b2g

Review of attachment 8430957 [details] [diff] [review]:
-----------------------------------------------------------------

I don't think my review is relevant here.

Comment 5

[Nicolas B. Pierron [:nbp]]

Comment on attachment 8430957 [details] [diff] [review]
Enable exact rooting by default on b2g

Review of attachment 8430957 [details] [diff] [review]:
-----------------------------------------------------------------

I think landing exact rooting is a good idea.  Still I think the landing of GGC should be conditioned by the addition of the exact rooting analysis (Bug 898554) as par of all-trees on tbpl (at least Gecko 32 trees).

Also we should pay attention to aurora's fork, as we are currently working on Gecko 32, which would be used as a base for B2G 2.0.

Comment 6

[Steve Fink [:sfink] [:s:]]

https://hg.mozilla.org/integration/b2g-inbound/rev/0f5c1cd2c79a

Comment 7

[Steve Fink [:sfink] [:s:]]

Sorry, should have put this here first: https://tbpl.mozilla.org/?tree=Try&rev=4267fa5c6a7d

Comment 8

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/0f5c1cd2c79a