805492 - The new decompiler is hitting an AssertNoGC()

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Terrence Cole [:terrence]]

Jason hit this failing stack: https://gist.github.com/3954084

Looking into it now.

Comment 1

[Terrence Cole [:terrence]]

Attachment: v0

This should either fix the problem or make the assert happen closer to the problem.

Comment 2

[:Ms2ger (he/him; ⌚ UTC+1/+2)]

Comment on attachment 675202 [details] [diff] [review]
v0

Review of attachment 675202 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/src/jsopcode.cpp
@@ +328,3 @@
>  {
> +    AssertCanGC();
> +    RootedScript script(cx, scriptArg);

This can just take a HandleScript now, no?

Comment 3

[Terrence Cole [:terrence]]

No, it's part of the API.

Comment 4

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

Comment on attachment 675202 [details] [diff] [review]
v0

Review of attachment 675202 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/src/jsopcode.cpp
@@ +328,3 @@
>  {
> +    AssertCanGC();
> +    RootedScript script(cx, scriptArg);

Apparently, the JSScript* is for calling from gdb, where constructing a Handle might be annoying. And given that this isn't exposed in any header file, that seems valid.

Comment 5

[Terrence Cole [:terrence]]

Green try at:
https://tbpl.mozilla.org/?tree=Try&rev=6c9f78601e21

Pushed at:
https://hg.mozilla.org/integration/mozilla-inbound/rev/a3709787bfca

Comment 6

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/a3709787bfca