EmitCreateIteratorResult loads map from function's context

EmitCreateIteratorResult loads map from function's context

Caching or serialization can cause full-codegen output to be shared
between contexts.  CreateIteratorResult, however, was doing the wrong
thing by creating results with the map that was current when the code
was generated.  Instead, we should chase pointers to load the right map
from the function's context.

R=verwaest@chromium.org
BUG=v8:3656
LOG=N

Committed: https://code.google.com/p/v8/source/detail?r=24987

[Toon Verwaest, 2014-10-29 16:02:20]

lgtm, thanks for fixing this.

[wingo, 2014-10-29 16:11:10]

Committed patchset #1 (id:1) manually as 24987 (presubmit successful).

[rossberg, 2014-10-29 17:25:04]

Note that this needs backmerging to the 3.28 branch. (Should we do 3.27 as
well?)

[rossberg, 2014-10-30 09:53:43]

On 2014/10/29 17:25:04, rossberg wrote:
> Note that this needs backmerging to the 3.28 branch. (Should we do 3.27 as
> well?)

Actually, I meant 3.29 (and 3.28?). Somehow I'm always off by one with Chrome
and V8 versions...

[wingo, 2014-10-30 13:49:41]

On 2014/10/30 09:53:43, rossberg wrote:
> On 2014/10/29 17:25:04, rossberg wrote:
> > Note that this needs backmerging to the 3.28 branch. (Should we do 3.27 as
> > well?)
> 
> Actually, I meant 3.29 (and 3.28?). Somehow I'm always off by one with Chrome
> and V8 versions...

I can do both if you like.  Would you like new CLs against the 3.29 and 3.28
branches?  Does "git cl dcommit" do the right thing in those cases?

[Yang, 2014-10-30 13:52:27]

On 2014/10/30 13:49:41, wingo wrote:
> On 2014/10/30 09:53:43, rossberg wrote:
> > On 2014/10/29 17:25:04, rossberg wrote:
> > > Note that this needs backmerging to the 3.28 branch. (Should we do 3.27 as
> > > well?)
> > 
> > Actually, I meant 3.29 (and 3.28?). Somehow I'm always off by one with
Chrome
> > and V8 versions...
> 
> I can do both if you like.  Would you like new CLs against the 3.29 and 3.28
> branches?  Does "git cl dcommit" do the right thing in those cases?

Let me do that by next monday. We got a set of scripts to do this.

[rossberg, 2014-10-30 14:09:53]

On 2014/10/30 13:52:27, Yang wrote:
> On 2014/10/30 13:49:41, wingo wrote:
> > On 2014/10/30 09:53:43, rossberg wrote:
> > > On 2014/10/29 17:25:04, rossberg wrote:
> > > > Note that this needs backmerging to the 3.28 branch. (Should we do 3.27
as
> > > > well?)
> > > 
> > > Actually, I meant 3.29 (and 3.28?). Somehow I'm always off by one with
> Chrome
> > > and V8 versions...
> > 
> > I can do both if you like.  Would you like new CLs against the 3.29 and 3.28
> > branches?  Does "git cl dcommit" do the right thing in those cases?
> 
> Let me do that by next monday. We got a set of scripts to do this.

Cool, thanks Yang.