844112 - GC: Fix misc build and test failures in exactly rooted shell

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Comment 1

[Jon Coppeard (:jonco)]

Attachment: Proposed fix

I tried building and testing the shell with exact rooting enabled, and made this patch to fix a few small problems I found:

 - build failure in MarkExactStackRoots

 - fix initialization of JSContext::thingGCRooters

 - don't try and Mark Proxy::LazyProto magic object pointer

 - NULL out rooted Shape pointer in interpreter when it's set to SHAPE_REMOVED (and another place where it's contents are no longer needed)

Comment 2

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

Comment on attachment 717171 [details] [diff] [review]
Proposed fix

Review of attachment 717171 [details] [diff] [review]:
-----------------------------------------------------------------

Sorry, I could've given you the r+ before you left if the LazyProto thing hadn't scared me into looking harder.

::: js/src/gc/RootMarking.cpp
@@ +50,5 @@
>          return;
>  
> +    if (kind == THING_ROOT_OBJECT && *addr == Proxy::LazyProto)
> +        return;
> +

Ugh. You made me go look up what Proxy::LazyProto is. That's awful. So... if anyone else decides to use a special pointer value, it might collide with LazyProto? Bleh. We need some more general mechanism for this.

Which is not your fault and totally out of scope here. Ignore me; I'm just complaining.

Comment 3

[Jon Coppeard (:jonco)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/f12876112a28

Comment 4

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/f12876112a28