fix #780 fix Object.assign when undefined value and inextensible

[tuchida]

ref. #780

Object.keys(Object.assign({a:undefined}, {b:undefined})).join()
// expected:<a[,b]> but was:<a[]>

I found another problem with it, so I fixed it while I was at it.

var obj = Object.freeze({});
Object.assign(obj, { a: 1 });
// An error to be thrown, but nothing was thrown.

[p-bakker]

Haven't looked at the changes, but haven't these fixes improved the test262 results?

[tuchida]

The undefined test does not seem to be in test262.
The Object.freeze test also does not seem to be in the one used by Rhino. The latest version has it, but it requires the implementation of Computed Property Names.

• Test Object.assign with non-extensible/sealed/frozen targets tc39/test262#2945

[p-bakker]

Right...

The latest version has it, but it requires the implementation of Computed Property Names.

Not just Computed Property Names though. Also at least a proper const impl and unicode regex

[p-bakker]

Abstract Object Operations as defined in the ecmascript spec ought to be placed in AbstractEcmaObjectOperations.

Please heck the guidelines at the top of the file for some instructions on how to

[tuchida]

ref. 204cd5f
I made to move to AbstractEcmaObjectOperations.

[tuchida]

I rebased it.

[gbrail]

Yes -- this looks good. Thanks!