ES6 classes: When a class extends B, super() invokes B.prototype.constructor() instead of B()

[Claude Pache]

Testcase:

```js
var called = null
class B { constructor() { called = 'B' } }
class C extends B {}
B.prototype.constructor = function F() { called = 'F' }

new C
called // should be 'B', is 'F'

```

It seems to me that JSC follows an old drafted semantics of `super()`, when it was equivalent to `super.constructor()`.


For reference, see the runtime semantics of `SuperCall : super Arguments` in the final spec:

    http://www.ecma-international.org/ecma-262/6.0/#sec-super-keyword-runtime-semantics-evaluation

and pay attention to the definition of GetSuperConstructor():

   http://www.ecma-international.org/ecma-262/6.0/#sec-getsuperconstructor

[Yusuke Suzuki]

Created attachment 262857 [details]
Patch

[Yusuke Suzuki]

Created attachment 262858 [details]
Patch

[Saam Barati]

Comment on attachment 262858 [details]
Patch

r=me
Does the latest spec change the "emitSuperBaseForCallee" static function in BytecodeGenerator?

[Yusuke Suzuki]

(In reply to comment #3)
> Comment on attachment 262858 [details]
> Patch
> 
> r=me
> Does the latest spec change the "emitSuperBaseForCallee" static function in
> BytecodeGenerator?

Yup. emitSuperBaseForCallee is correct in the latest spec :)
http://ecma-international.org/ecma-262/6.0/#sec-getsuperbase

[WebKit Commit Bot]

Comment on attachment 262858 [details]
Patch

Clearing flags on attachment: 262858

Committed r190847: <http://trac.webkit.org/changeset/190847>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.