Move computation of RenderLayer::isUnclippedDescendant into CompositingPropertyUpdater

Move computation of RenderLayer::isUnclippedDescendant into CompositingPropertyUpdater

Prior to this CL, we computed RenderLayer::isUnclippedDescendant in a separate
O(n^2) walk of the out-of-flow RenderLayers. On workloads like famo.us, this
walk is quite expensive because everything is out-of-flow positioned, and
O(n^2) is large.

After this CL, we compute this information as we walk down the tree in the
CompositingPropertyUpdater, which means we can compute the information in
linear time and we can entirely remove the separate tree walk. Moving this
computation also lets us rip out a large amount of machinery that existed to
prune the extra tree walk that this CL removes.

Now that we compute this information with the CompositingPropertyUpdater,
we compute it even when composited overflow scrolling is off.  This confused
some tests that used Internals to poke this state.  I've added a workaround to
keep theses tests working identically.  I'll remove the work around and update
the tests in the next CL.

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=175225

[Ian Vollick, 2014-06-01 19:25:35]

I love it. LGTM.

https://codereview.chromium.org/309743002/diff/20001/Source/core/rendering/Re...
File Source/core/rendering/RenderLayer.cpp (left):

https://codereview.chromium.org/309743002/diff/20001/Source/core/rendering/Re...
Source/core/rendering/RenderLayer.cpp:114: , m_isUnclippedDescendant(false)
Amazing!

https://codereview.chromium.org/309743002/diff/20001/Source/core/rendering/co...
File Source/core/rendering/compositing/RenderLayerCompositor.h (left):

https://codereview.chromium.org/309743002/diff/20001/Source/core/rendering/co...
Source/core/rendering/compositing/RenderLayerCompositor.h:105: void
updateCompositingRequirementsState();
Awesome!

https://codereview.chromium.org/309743002/diff/40001/Source/core/rendering/co...
File Source/core/rendering/compositing/CompositingReasonFinder.cpp (right):

https://codereview.chromium.org/309743002/diff/40001/Source/core/rendering/co...
Source/core/rendering/compositing/CompositingReasonFinder.cpp:147: // this value
isn't stale.
What would you think of using the disabler here?

[abarth-chromium, 2014-06-01 19:49:36]

https://codereview.chromium.org/309743002/diff/40001/Source/core/rendering/co...
File Source/core/rendering/compositing/CompositingReasonFinder.cpp (right):

https://codereview.chromium.org/309743002/diff/40001/Source/core/rendering/co...
Source/core/rendering/compositing/CompositingReasonFinder.cpp:147: // this value
isn't stale.
On 2014/06/01 19:25:36, Ian Vollick wrote:
> What would you think of using the disabler here?

There isn't a disabler for this ASSERT because it's guarding the inputs to the
computing update rather than the outputs.  Hopefully there won't be any more
violations...

This would be clearer if we renamed ancestorDependentProperties() to
compositingInputs()    :)

[abarth-chromium, 2014-06-01 19:49:41]

The CQ bit was checked by abarth@chromium.org

[commit-bot: I haz the power, 2014-06-01 19:49:49]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/abarth@chromium.org/309743002/40001

[commit-bot: I haz the power, 2014-06-01 21:07:32]

FYI, CQ is re-trying this CL (attempt #1).
The failing builders are:
  blink_presubmit on tryserver.blink
(http://build.chromium.org/p/tryserver.blink/builders/blink_presubmit/builds/6614)

[commit-bot: I haz the power, 2014-06-01 21:21:40]

Change committed as 175225

[ojan, 2014-06-01 23:29:20]

https://codereview.chromium.org/309743002/diff/40001/Source/core/rendering/co...
File Source/core/rendering/compositing/RenderLayerCompositor.cpp (left):

https://codereview.chromium.org/309743002/diff/40001/Source/core/rendering/co...
Source/core/rendering/compositing/RenderLayerCompositor.cpp:357: // before
moving this function after checking the dirty bits.
This comment doesn't seem accurate anymore. I mean...it kind of is. FWIW, if we
wanted to, I think we could move the enableCompositingModeIfNeeded check to be
after the dirty bit checking now that we don't early return on !m_compositing
anymore.

https://codereview.chromium.org/309743002/diff/40001/Source/core/rendering/co...
Source/core/rendering/compositing/RenderLayerCompositor.cpp:359:
updateCompositingRequirementsState();
Amazing.