1230710 - Reenable arrow functions and eval inside derived class constructors

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Eric Faust [:efaust]]

Attachment: Patch

Thanks to Jan, this should just work. the patch was pretty straightforward.

Comment 1

[Jason Orendorff [:jorendorff]]

Comment on attachment 8696150 [details] [diff] [review]
Patch

Review of attachment 8696150 [details] [diff] [review]:
-----------------------------------------------------------------

nice.

::: js/src/frontend/BytecodeEmitter.cpp
@@ +3451,5 @@
> +    // Handle the eval case. Only accept the strict variant, as eval in a
> +    // derived class constructor must be strict.
> +    if (setOp == JSOP_STRICTSETNAME) {
> +        if (!emitAtomOp(name, JSOP_GETNAME))
> +            return false;

Do you know how we prevent `with ({".this": blah})` from capturing this GETNAME?

Test?

::: js/src/vm/Interpreter-inl.h
@@ +205,5 @@
>      }
>  
> +    // We do our own explicit checking for |this|
> +    if (name == cx->names().dotThis)
> +        return true;

The comment should say "The bytecode" rather than "We"

...I guess I kind of wish we did it the other way, got rid of JSOP_CHECKTHISREINIT, and just use JSOP_CHECKLEXICAL and CHECKALIASEDLEXICAL, and the CheckUninitializedLexical call below. Go ahead and take the benefit of `.this` being less special.

Your call though.

Comment 2

[Shu-yu Guo [:shu]]

Comment on attachment 8696150 [details] [diff] [review]
Patch

Review of attachment 8696150 [details] [diff] [review]:
-----------------------------------------------------------------

This looks fine to me. Just fine.

::: js/src/frontend/BytecodeEmitter.cpp
@@ +3451,5 @@
> +    // Handle the eval case. Only accept the strict variant, as eval in a
> +    // derived class constructor must be strict.
> +    if (setOp == JSOP_STRICTSETNAME) {
> +        if (!emitAtomOp(name, JSOP_GETNAME))
> +            return false;

In response to Jason's question, there's special magic in WithObjects to prevent .this from leaking into the scope.

::: js/src/vm/Interpreter-inl.h
@@ +205,5 @@
>      }
>  
> +    // We do our own explicit checking for |this|
> +    if (name == cx->names().dotThis)
> +        return true;

Also in response to Jason's comment. I don't care too much either way. More uniformity the better, sure, but better error messages for "this" TDZ errors already demand some specialty.

Comment 3

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/3c31d67c74d2

Comment 4

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/3c31d67c74d2