Fix squashing to treat perspective and preserve3d elements as transform ancestors.

Fix squashing to treat perspective and preserve3d elements as transform ancestors.

This is necessary because they can affect the local transform of an element, so
elements with different perspecitive/transform3d ancestors cannot squash together.

BUG=404953, 403819
R=vollick@chromium.org

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=180799

[Ian Vollick, 2014-08-22 20:21:40]

lgtm

[chrishtr, 2014-08-22 20:41:25]

The CQ bit was checked by chrishtr@chromium.org

[commit-bot: I haz the power, 2014-08-22 20:42:23]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/chrishtr@chromium.org/497153003/20001

[commit-bot: I haz the power, 2014-08-22 21:44:14]

FYI, CQ is re-trying this CL (attempt #1).
The failing builders are:
  linux_blink_dbg on tryserver.blink
(http://build.chromium.org/p/tryserver.blink/builders/linux_blink_dbg/builds/2...)
  win_blink_rel on tryserver.blink
(http://build.chromium.org/p/tryserver.blink/builders/win_blink_rel/builds/24030)

[chrishtr, 2014-08-22 22:54:42]

Ian, I had to change the code in updateOverflowControlsHostLayerGeometry,
because
it's not true any more that the transform ancestor of the compositing container
is the same
as that of this layer. I also don't think it was always true before - there can
be non-transform
reasons for compositing the ancestor.

However, if there are no 3D transforms between us and the compositing stacking
ancestor, we can ignore any intermediate perspective or preserves3d and map
directly to it. Though I don't understand what happens if m_owningLayer itself
has a 3D transform on it. How is that supposed to work?

Meantime I've made the 2D update, and it passes all tests.

[chrishtr, 2014-08-22 23:19:40]

Committed patchset #4 manually as 180799 (presubmit successful).

[chrishtr, 2014-08-22 23:21:20]

I dcommitted to be able to merge stuff into M37. Happy to revise the changes on
Monday if my change is wrong.

[chrishtr, 2014-08-25 22:41:54]

Ian, could you take a look at the changes I made in the last revision? Not sure
it's correct.

[Ian Vollick, 2014-08-26 01:13:38]

On 2014/08/25 at 22:41:54, chrishtr wrote:
> Ian, could you take a look at the changes I made in the last revision? Not
sure it's correct.

Sure. Your changes in updateOverflowControlsHostLayerGeometry look correct to me
(and less roundabout).

[chrishtr, 2014-08-26 18:12:49]

On 2014/08/26 at 01:13:38, vollick wrote:
> On 2014/08/25 at 22:41:54, chrishtr wrote:
> > Ian, could you take a look at the changes I made in the last revision? Not
sure it's correct.
> 
> Sure. Your changes in updateOverflowControlsHostLayerGeometry look correct to
me (and less roundabout).

Hmm, ok. But what if m_owningLayer ha a 3D transform? Is that not possible in
this scenario?

[Ian Vollick, 2014-08-26 18:16:48]

On 2014/08/26 18:12:49, chrishtr wrote:
> On 2014/08/26 at 01:13:38, vollick wrote:
> > On 2014/08/25 at 22:41:54, chrishtr wrote:
> > > Ian, could you take a look at the changes I made in the last revision? Not
> sure it's correct.
> > 
> > Sure. Your changes in updateOverflowControlsHostLayerGeometry look correct
to
> me (and less roundabout).
> 
> Hmm, ok. But what if m_owningLayer ha a 3D transform? Is that not possible in
> this scenario?

If m_owningLayer has a 3D transform, then its overflow controls will never need
to be reparented as its owning layer will force a stacking context.

[chrishtr, 2014-08-26 18:25:21]

On 2014/08/26 at 18:16:48, vollick wrote:
> On 2014/08/26 18:12:49, chrishtr wrote:
> > On 2014/08/26 at 01:13:38, vollick wrote:
> > > On 2014/08/25 at 22:41:54, chrishtr wrote:
> > > > Ian, could you take a look at the changes I made in the last revision?
Not
> > sure it's correct.
> > > 
> > > Sure. Your changes in updateOverflowControlsHostLayerGeometry look correct
to
> > me (and less roundabout).
> > 
> > Hmm, ok. But what if m_owningLayer ha a 3D transform? Is that not possible
in
> > this scenario?
> 
> If m_owningLayer has a 3D transform, then its overflow controls will never
need to be reparented as its owning layer will force a stacking context.

Right ok. I'll add an assert.