Array Iterator next should check for own property

Array Iterator next should check for own property

Since we are using private symbols for the internal slots we need to
check for a local property.

BUG=None
LOG=Y
R=rossberg@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=21399

[arv (Not doing code reviews), 2014-05-07 16:26:53]

Follow up CL from https://codereview.chromium.org/258793005/

[Toon Verwaest, 2014-05-07 16:42:13]

Lg semantically, but andreas had some performance concern with
%HasLocalProperty. I'll defer to him, given that I'm on holidays for the next 2
weeks and the tree is closed at least until friday for the branch.

[rossberg, 2014-05-08 08:59:20]

On 2014/05/07 16:42:13, Toon Verwaest wrote:
> Lg semantically, but andreas had some performance concern with
> %HasLocalProperty. I'll defer to him, given that I'm on holidays for the next
2
> weeks and the tree is closed at least until friday for the branch.

Yes, Toon and I discussed this very issue the other day. IMO, the "right" fix
would be to define GET_PRIVATE with %GetOwnProperty. However, this would be far
more expensive than the current definition, for all reads -- see
https://code.google.com/p/v8/issues/detail?id=3320, which I just opened
yesterday. Since addressing the performance issue would be a bigger chunk of
work, we decided to punt on it for the time being.

But as middle ground I think it would make sense to at least redefine
HAS_PRIVATE to use %HasLocalProperty. Then you can implement the correct
behaviour without violating the abstraction provided by the PRIVATE macros.

[rossberg, 2014-05-20 09:14:28]

On 2014/05/08 08:59:20, rossberg wrote:
> On 2014/05/07 16:42:13, Toon Verwaest wrote:
> > Lg semantically, but andreas had some performance concern with
> > %HasLocalProperty. I'll defer to him, given that I'm on holidays for the
next
> 2
> > weeks and the tree is closed at least until friday for the branch.
> 
> Yes, Toon and I discussed this very issue the other day. IMO, the "right" fix
> would be to define GET_PRIVATE with %GetOwnProperty. However, this would be
far
> more expensive than the current definition, for all reads -- see
> https://code.google.com/p/v8/issues/detail?id=3320, which I just opened
> yesterday. Since addressing the performance issue would be a bigger chunk of
> work, we decided to punt on it for the time being.
> 
> But as middle ground I think it would make sense to at least redefine
> HAS_PRIVATE to use %HasLocalProperty. Then you can implement the correct
> behaviour without violating the abstraction provided by the PRIVATE macros.

I landed https://codereview.chromium.org/290373002/ so you can use HAS_PRIVATE
now.

[arv (Not doing code reviews), 2014-05-20 21:12:46]

Now using HAS_PRIVATE.

[rossberg, 2014-05-21 06:59:39]

lgtm

[adamk, 2014-05-21 08:05:22]

Committed patchset #2 manually as r21399 (presubmit successful).