cc::Fix computation of surface backfacing-ness when using property trees

cc::Fix computation of surface backfacing-ness when using property trees

Property trees currently use the owning layer's draw transform to
compute the render surface's backfacing-ness. This computation is
flawed as the owning layer's draw transform computed using property
trees has only the sublayer scale and not the transform to target. So,
we should use the surface's draw transform instead. This change also
fixes the unit test,
LayerTreeHostCommonTest.DoNotIncludeBackfaceInvisibleSurfaces when we
enable impl-thread property trees.

BUG=551629
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel

Committed: https://crrev.com/5aa88b8d81780575ce87eebe36ec3c11e103505c
Cr-Commit-Position: refs/heads/master@{#358736}

[jaydasika, 2015-11-09 22:56:45]

jaydasika@chromium.org changed reviewers:
+ ajuma@chromium.org

[jaydasika, 2015-11-09 22:56:45]

https://codereview.chromium.org/1430353002/diff/20001/cc/trees/layer_tree_hos...
File cc/trees/layer_tree_host_common.cc (right):

https://codereview.chromium.org/1430353002/diff/20001/cc/trees/layer_tree_hos...
cc/trees/layer_tree_host_common.cc:1769:
render_surface->SetDrawTransform(draw_transform);
CDP and PT render surface draw transform don't match without this change.

[ajuma, 2015-11-09 23:31:22]

https://codereview.chromium.org/1430353002/diff/20001/cc/trees/layer_tree_hos...
File cc/trees/layer_tree_host_common.cc (right):

https://codereview.chromium.org/1430353002/diff/20001/cc/trees/layer_tree_hos...
cc/trees/layer_tree_host_common.cc:1769:
render_surface->SetDrawTransform(draw_transform);
On 2015/11/09 22:56:45, jaydasika wrote:
> CDP and PT render surface draw transform don't match without this change.

Is this because the "skipped" surface is still winding up in the RSLL, and
having its draw transform compared? See the suggestion below for possibly fixing
this.

https://codereview.chromium.org/1430353002/diff/20001/cc/trees/layer_tree_hos...
cc/trees/layer_tree_host_common.cc:2508: layer->draw_properties().render_target
= nullptr;
Would calling RemoveSurfaceForEarlyExit here fix the CDP/PT draw transform
mismatch that you mention above (without the need for the code added earlier
that computes that draw transform of a skipped surface)?

[jaydasika, 2015-11-10 00:06:40]

https://codereview.chromium.org/1430353002/diff/20001/cc/trees/layer_tree_hos...
File cc/trees/layer_tree_host_common.cc (right):

https://codereview.chromium.org/1430353002/diff/20001/cc/trees/layer_tree_hos...
cc/trees/layer_tree_host_common.cc:2508: layer->draw_properties().render_target
= nullptr;
On 2015/11/09 23:31:21, ajuma (slow till Nov. 16) wrote:
> Would calling RemoveSurfaceForEarlyExit here fix the CDP/PT draw transform
> mismatch that you mention above (without the need for the code added earlier
> that computes that draw transform of a skipped surface)?

No that won't still fix the mismatch. We either need to add the code in CDP(at
previous comment) or call draw_transform of layer for !use_property_tree case
here.

[ajuma, 2015-11-10 00:17:15]

lgtm

https://codereview.chromium.org/1430353002/diff/20001/cc/trees/layer_tree_hos...
File cc/trees/layer_tree_host_common.cc (right):

https://codereview.chromium.org/1430353002/diff/20001/cc/trees/layer_tree_hos...
cc/trees/layer_tree_host_common.cc:1766: gfx::Transform draw_transform =
combined_transform;
Please add a comment that we need this logic since the surface-skipping code we
use later depends on the surface's draw transform.

https://codereview.chromium.org/1430353002/diff/20001/cc/trees/layer_tree_hos...
cc/trees/layer_tree_host_common.cc:2508: layer->draw_properties().render_target
= nullptr;
On 2015/11/10 00:06:40, jaydasika wrote:
> On 2015/11/09 23:31:21, ajuma (slow till Nov. 16) wrote:
> > Would calling RemoveSurfaceForEarlyExit here fix the CDP/PT draw transform
> > mismatch that you mention above (without the need for the code added earlier
> > that computes that draw transform of a skipped surface)?
> 
> No that won't still fix the mismatch. We either need to add the code in CDP(at
> previous comment) or call draw_transform of layer for !use_property_tree case
> here.

Oh, I think I understand: the old code relied on the fact that when we
early-return in the code at the previous comment, the owning layer's transform
doesn't get reparented by the surface.

[jaydasika, 2015-11-10 00:46:07]

The CQ bit was checked by jaydasika@chromium.org

[commit-bot: I haz the power, 2015-11-10 00:50:01]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1430353002/20001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1430353002/20001

[commit-bot: I haz the power, 2015-11-10 01:48:13]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-11-10 01:50:12]

Patchset 2 (id:??) landed as
https://crrev.com/5aa88b8d81780575ce87eebe36ec3c11e103505c
Cr-Commit-Position: refs/heads/master@{#358736}