1030938 - Stop copy-constructing frame iterators in SavedStacks, and some other cleanup

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Boris Zbarsky [:bzbarsky]]

See bug 1027305 comment 16.

Comment 1

[Boris Zbarsky [:bzbarsky]]

Attachment: Stop copy-constructing iterators in SavedStacks code, because that copy constructor is really slow. Also, hash atoms as pointers, not strings

Comment 2

[Shu-yu Guo [:shu]]

Comment on attachment 8446729 [details] [diff] [review]
Stop copy-constructing iterators in SavedStacks code, because that copy constructor is really slow.  Also, hash atoms as pointers, not strings

Review of attachment 8446729 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/src/vm/SavedStacks.cpp
@@ +101,5 @@
>      if (source != lookup.source)
>          return false;
>  
>      JSAtom *functionDisplayName = existing->getFunctionDisplayName();
> +    if (functionDisplayName != lookup.functionDisplayName)

Oh nice.

@@ +477,5 @@
>      // in js/src/jit-test/tests/saved-stacks/bug-1006876-too-much-recursion.js).
>      JS_CHECK_RECURSION_DONT_REPORT(cx, return false);
>  
> +    RootedScript script(cx, iter.script());
> +    jsbytecode* pc = iter.pc();

Style nit: I think SM style does jsbytecode *pc

@@ +480,5 @@
> +    RootedScript script(cx, iter.script());
> +    jsbytecode* pc = iter.pc();
> +    RootedFunction callee(cx, iter.maybeCallee());
> +    // script and callee should keep compartment alive.
> +    JSCompartment* compartment = iter.compartment();

Ditto.

Comment 3

[Boris Zbarsky [:bzbarsky]]

Fixed the nits and https://hg.mozilla.org/integration/mozilla-inbound/rev/937b2131d8e2

Comment 4

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/937b2131d8e2