852762 - Arrow functions are not automatically strict after all

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Jason Orendorff [:jorendorff]]

There is also the question of whether they have arguments, which I'll have to figure out from the meeting notes. This bug can address both.

<brendan> rs=me on arguments :-P

Comment 1

[Rick Waldron [:rwaldron]]

In case any future readers are interested, the resolution of record for this decision is here:

https://mail.mozilla.org/pipermail/es-discuss/2013-February/028632.html

(Formatted: https://github.com/rwldrn/tc39-notes/blob/master/es6/2013-01/jan-30.md)

Comment 2

[Jason Orendorff [:jorendorff]]

Attachment: v1

Comment 3

[Jason Orendorff [:jorendorff]]

Attachment: v2 - add another test or two

Comment 4

[Jason Orendorff [:jorendorff]]

Review ping! :)

Comment 5

[Eddy Bruel [:ejpbruel]]

Comment on attachment 730273 [details] [diff] [review]
v2 - add another test or two

Review of attachment 730273 [details] [diff] [review]:
-----------------------------------------------------------------

Looks mostly good to me. r+ with comments addressed.

::: js/src/jit-test/tests/arrow-functions/arguments-2.js
@@ +1,1 @@
> +// 'arguments' in arrow functions nested in other functions

Based on what this test does, this means that arrow functions get their own instance of arguments, right? That wasn't clear to me from this comment.

::: js/src/jit-test/tests/arrow-functions/arguments-4.js
@@ +10,3 @@
>  ];
>  
> +

Spurious newline here

::: js/src/jit-test/tests/arrow-functions/strict-1.js
@@ +10,3 @@
>  
> +f = (a = {x: 1, x: 2}) => b => { "use strict"; return a.x; };
> +assertEq(f()(0), 2);

I'm not sure what this tests for. Should the inner => be strict here? What about the outer =>? And how does this assertion confirm that?

::: js/src/jit-test/tests/arrow-functions/strict-2.js
@@ +2,5 @@
>  
>  load(libdir + "asserts.js");
>  
>  assertThrowsInstanceOf(
> +    () => Function("(a = function (obj) { with (obj) f(); }) => { 'use strict'; }"),

Why do we need the outer () => Function(...) here? I would assume this test throws even without that.

::: js/src/jit-test/tests/arrow-functions/syntax-errors.js
@@ +14,5 @@
>      "{p} => p",
>      "(...x => expr)",
>      "1 || a => a",
> +    "'use strict' => {}",
> +    "package => {'use strict';}",    // tricky: FutureReservedWord in strict mode code only

How do we make sure that this only throws a syntax error in strict mode? And if we don't, could you file a followup bug for this?

Comment 6

[David Bruant]

I added a note on MDN about this bug [1] Feel free to remove it when this bug is fixed.

[1] https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/arrow_functions#Upcoming_change

Comment 7

[Jason Orendorff [:jorendorff]]

(In reply to Eddy Bruel [:ejpbruel] from comment #5)
> > +// 'arguments' in arrow functions nested in other functions
> 
> Based on what this test does, this means that arrow functions get their own
> instance of arguments, right? That wasn't clear to me from this comment.

Yes, they do. I improved the comments in several of the tests.

> ::: js/src/jit-test/tests/arrow-functions/strict-1.js
> @@ +10,3 @@
> >  
> > +f = (a = {x: 1, x: 2}) => b => { "use strict"; return a.x; };
> > +assertEq(f()(0), 2);
> 
> I'm not sure what this tests for. Should the inner => be strict here? What
> about the outer =>? And how does this assertion confirm that?

Yes; no; the fact that it compiles and runs confirms that the outer arrow is non-strict. I only check the return value because why not.

> ::: js/src/jit-test/tests/arrow-functions/strict-2.js
> @@ +2,5 @@
> >  
> >  load(libdir + "asserts.js");
> >  
> >  assertThrowsInstanceOf(
> > +    () => Function("(a = function (obj) { with (obj) f(); }) => { 'use strict'; }"),
> 
> Why do we need the outer () => Function(...) here? I would assume this test
> throws even without that.

Another way to write this test would be:

    // |jit-test| error: SyntaxError
    (a = function (obj) { with (obj) f(); }) => { "use strict"; };

Using assertThrowsInstanceOf requires putting the test case in quotes and using ()=>Function(...) to parse it on demand. Otherwise the error would be thrown at parse time, before assertThrowsInstanceOf is ever called, so the test would fail.

> ::: js/src/jit-test/tests/arrow-functions/syntax-errors.js
> > +    "package => {'use strict';}",    // tricky: FutureReservedWord in strict mode code only
> 
> How do we make sure that this only throws a syntax error in strict mode? And
> if we don't, could you file a followup bug for this?

Added a test for this (in the same file).

Comment 8

[Jason Orendorff [:jorendorff]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/60d580712ec0

Comment 9

[Phil Ringnalda (:philor)]

https://hg.mozilla.org/mozilla-central/rev/60d580712ec0

Comment 10

[Tom S [:evilpie]]

Noted on:
https://developer.mozilla.org/en-US/docs/Mozilla/Firefox/Releases/24
https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/arrow_functions