cc: Make impl-thread property trees handle fixed-position layers

cc: Make impl-thread property trees handle fixed-position layers

This makes impl-thread property trees correctly handle fixed-position
layers, as long as there's no bounds_delta (handling bounds_delta is
left for another CL). This CL makes the offset between a fixed-position
layer and its container unchangeable on the compositor thread, so that
it remains unaffected by impl scrolling of intermediate layers. This
also changes the transform node used as the parent of a fixed-position
node when the fixed-position container is scrollable: since scrolls of
this container aren't supposed to affect contained fixed-position
layers, yet do contribute to this node's local transform, this node's
parent is used as the transform tree parent of contained fixed-position
nodes.

This makes the portions of LayerPostionConstraintTests that do not
involve bounds_delta pass with verification enabled.

BUG=481585
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel

Committed: https://crrev.com/5e47dec44b5f6211afd795da37d8b4a257ae133c
Cr-Commit-Position: refs/heads/master@{#332055}

[ajuma, 2015-05-29 17:46:09]

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

[ajuma, 2015-05-29 17:47:08]

https://codereview.chromium.org/1144993004/diff/40001/cc/layers/layer_positio...
File cc/layers/layer_position_constraint_unittest.cc (right):

https://codereview.chromium.org/1144993004/diff/40001/cc/layers/layer_positio...
cc/layers/layer_position_constraint_unittest.cc:978:
great_grand_child_->SetScrollClipLayerId(root_->id());
This test scrolls great_grand_child_, so it should be scrollable (so that a
transform node gets created).

https://codereview.chromium.org/1144993004/diff/40001/cc/trees/property_tree_...
File cc/trees/property_tree_builder.cc (right):

https://codereview.chromium.org/1144993004/diff/40001/cc/trees/property_tree_...
cc/trees/property_tree_builder.cc:395: return;
This change causes about 16 unit tests to fail, since they rely on property
trees being rebuilt every frame. I'm not sure what the best approach is here.
One option would be to change those tests so that they explicitly set
needs_rebuild. Alternatively, we could add a way for tests that don't want to
rebuild (like LayerPositionConstraintTests) to opt out. Making the default be
"don't rebuild" seems better in the sense that non-test impl property trees
aren't rebuilt, so tests that want to rebuild anyway should be making a
conscious choice.

[enne (OOO), 2015-05-29 20:07:54]

lgtm

https://codereview.chromium.org/1144993004/diff/40001/cc/layers/layer_positio...
File cc/layers/layer_position_constraint_unittest.cc (right):

https://codereview.chromium.org/1144993004/diff/40001/cc/layers/layer_positio...
cc/layers/layer_position_constraint_unittest.cc:978:
great_grand_child_->SetScrollClipLayerId(root_->id());
On 2015/05/29 at 17:47:08, ajuma wrote:
> This test scrolls great_grand_child_, so it should be scrollable (so that a
transform node gets created).

Is that something that can be DCHECK'd?

https://codereview.chromium.org/1144993004/diff/40001/cc/trees/property_tree.h
File cc/trees/property_tree.h (right):

https://codereview.chromium.org/1144993004/diff/40001/cc/trees/property_tree....
cc/trees/property_tree.h:219: // that fixed-position layers are positioned by
Blink wrt to their layer tree
I will be so glad when we get to the point that we can just do this the right
way.

https://codereview.chromium.org/1144993004/diff/40001/cc/trees/property_tree_...
File cc/trees/property_tree_builder.cc (right):

https://codereview.chromium.org/1144993004/diff/40001/cc/trees/property_tree_...
cc/trees/property_tree_builder.cc:395: return;
On 2015/05/29 at 17:47:08, ajuma wrote:
> This change causes about 16 unit tests to fail, since they rely on property
trees being rebuilt every frame. I'm not sure what the best approach is here.
One option would be to change those tests so that they explicitly set
needs_rebuild. Alternatively, we could add a way for tests that don't want to
rebuild (like LayerPositionConstraintTests) to opt out. Making the default be
"don't rebuild" seems better in the sense that non-test impl property trees
aren't rebuilt, so tests that want to rebuild anyway should be making a
conscious choice.

Agreed that the default should be don't rebuild unless the flag is set.  Tests
that are doing weird things should mark rebuild explicitly if they really need
it.

[ajuma, 2015-05-29 21:41:07]

https://codereview.chromium.org/1144993004/diff/40001/cc/layers/layer_positio...
File cc/layers/layer_position_constraint_unittest.cc (right):

https://codereview.chromium.org/1144993004/diff/40001/cc/layers/layer_positio...
cc/layers/layer_position_constraint_unittest.cc:978:
great_grand_child_->SetScrollClipLayerId(root_->id());
On 2015/05/29 20:07:53, enne wrote:
> On 2015/05/29 at 17:47:08, ajuma wrote:
> > This test scrolls great_grand_child_, so it should be scrollable (so that a
> transform node gets created).
> 
> Is that something that can be DCHECK'd?

Added a DCHECK to LayerImpl::SetScrollDelta.

[ajuma, 2015-05-29 21:41:12]

The CQ bit was checked by ajuma@chromium.org

[ajuma, 2015-05-29 21:41:13]

The patchset sent to the CQ was uploaded after l-g-t-m from enne@chromium.org
Link to the patchset: https://codereview.chromium.org/1144993004/#ps60001
(title: "Add DCHECK")

[commit-bot: I haz the power, 2015-05-29 22:02:01]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1144993004/60001

[commit-bot: I haz the power, 2015-05-29 22:10:32]

Committed patchset #4 (id:60001)

[commit-bot: I haz the power, 2015-05-29 22:11:17]

Patchset 4 (id:??) landed as
https://crrev.com/5e47dec44b5f6211afd795da37d8b4a257ae133c
Cr-Commit-Position: refs/heads/master@{#332055}