940718 - Mark AutoJSContext as not GCing (and make this true)

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Terrence Cole [:terrence]]

We still have a few (false positive) hazards over AutoJSContext. I think the best thing to do here is to run a first GetSafeJSContext in nsContentUtils::Init -- where it is safe to GC -- so that we don't have to worry about a potential GC later.

Lets see if this theory works in practice:
https://tbpl.mozilla.org/?tree=Try&rev=20aedee8483e

Comment 1

[Andrew McCreight [:mccr8]]

Bobby, does this sound like a reasonable plan?

Comment 2

[Bobby Holley (:bholley)]

Rather than doing any hacky fixes to the static analysis, I would prefer an explicit separate call to InitSafeJSContext, and have GetSafeJSContext MOZ_ASSERT that mSafeJSContext is non-null.

In a perfect would, InitSafeJSContext would happen during XPConnect startup. But this is what used to happen, and caused a bunch of startup ordering pain (which I fixed in bug 905364). So I think just doing it from nsContentUtils is fine, given that JSContexts are on the way out anyway.

Comment 3

[Terrence Cole [:terrence]]

Great idea, calling it InitSafeJSContext: I'll make that split tomorrow. Try run looks good, modulo existing bustage on tip.

Comment 4

[Terrence Cole [:terrence]]

Attachment: hazard_AutoSafeJSContext-v1.diff

Not bad at all.

Comment 5

[Terrence Cole [:terrence]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/2267eab2fde3

Comment 6

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/2267eab2fde3