Have DOMWindow deal with references instead of pointers when possible

Have DOMWindow deal with references instead of pointers when possible

Have DOMWindow deal with references instead of pointers when possible. This
makes the code look safer and gets rid of unnecessary null checks.

This CL also updates the generated bindings so that they allow for the
implementation to return references instead of pointers in more cases.

R=adamk@chromium.org, nbarth@chromium.org, adamk

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=168422

[Nils Barth (inactive), 2014-03-04 02:50:39]

haraken: we're ok with using WTF::getPtr if it's useful, right?

LGTM
While the WTF::getPtr makes the CG a little more complicated,
the benefits to safety and Blink code simplicity make it worthwhile.
(I assume performance impact is minimal or zero?
 Do we need to worry about this?)

Also, since this and previous CL (removing [SetWrapperReferenceFromReference])
mean that we're using
WTF::getPtr (not eliminating it),
could you remove the two FIXME lines
in v8_interface.py and v8_callback_interface.py that read:
# FIXME: remove if can eliminate WTF::getPtr
...since it looks like we're not eliminating it anytime soon.

[Inactive, 2014-03-04 03:20:45]

On 2014/03/04 02:50:39, Nils Barth wrote:
> While the WTF::getPtr makes the CG a little more complicated,
> the benefits to safety and Blink code simplicity make it worthwhile.
> (I assume performance impact is minimal or zero?
>  Do we need to worry about this?)

There should be no impact on performance getPtr() is inlined and it returns p if
p is a pointer or &p if p is a reference.
 
> Also, since this and previous CL (removing [SetWrapperReferenceFromReference])
> mean that we're using
> WTF::getPtr (not eliminating it),
> could you remove the two FIXME lines
> in v8_interface.py and v8_callback_interface.py that read:
> # FIXME: remove if can eliminate WTF::getPtr
> ...since it looks like we're not eliminating it anytime soon.

I will remove them.

[haraken, 2014-03-04 03:27:23]

> haraken: we're ok with using WTF::getPtr if it's useful, right?

Yes. A few months ago I was thinking that we might be able to remove getPtr (for
refactoring purpose) but now I'm getting to understand that it's useful.

LGTM.

[Inactive, 2014-03-04 03:31:37]

On 2014/03/04 02:50:39, Nils Barth wrote:
> Also, since this and previous CL (removing [SetWrapperReferenceFromReference])
> mean that we're using
> WTF::getPtr (not eliminating it),
> could you remove the two FIXME lines
> in v8_interface.py and v8_callback_interface.py that read:
> # FIXME: remove if can eliminate WTF::getPtr
> ...since it looks like we're not eliminating it anytime soon.

Done.

[Inactive, 2014-03-04 03:32:06]

adamk / abarth: could you please take a look at the tiny change in web/ ?

[Nils Barth (inactive), 2014-03-04 03:51:14]

On 2014/03/04 03:31:37, Chris Dumez wrote:
> On 2014/03/04 02:50:39, Nils Barth wrote:
> > Also, since this and previous CL (removing
[SetWrapperReferenceFromReference])
> > mean that we're using
> > WTF::getPtr (not eliminating it),
> > could you remove the two FIXME lines
> > in v8_interface.py and v8_callback_interface.py that read:
> > # FIXME: remove if can eliminate WTF::getPtr
> > ...since it looks like we're not eliminating it anytime soon.
> 
> Done.

Thanks Chris!

[adamk, 2014-03-04 04:52:04]

rslgtm for web, but I wouldn't be surprised if you need some module owners (or
abarth)

[Inactive, 2014-03-04 05:00:11]

On 2014/03/04 04:52:04, adamk wrote:
> rslgtm for web, but I wouldn't be surprised if you need some module owners (or
> abarth)

It looks like haraken is a modules OWNER as well, which is why my Chrome plugin
indicated this folder as reviewed.

[Inactive, 2014-03-05 03:10:37]

Committed patchset #3 manually as r168422 (presubmit successful).