Round in the range [-0.5, 0.5) when computing offsetFromRenderer and subpixelAccumulation

Round in the range [-0.5, 0.5) when computing offsetFromRenderer and subpixelAccumulation
for squashed layers.

Also remove the assert that the subpixel accumulation matches
the subpixel accumulation if the layer was separately
composited. This doesn't make sense, since the squashing
layer is positioned relative to the compositing ancestor
of m_owningLayer, not the squashing layer, which can have
different subpixel accumulations since subpixel
accumulation is relative to the compositing container of
the graphics layer.

BUG=369526
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=176441

[chrishtr, 2014-06-18 16:15:48]

I'm still struggling to find a test that reproduces this.
I have confirmed so far that it is actually impossible for
offsetFromSquashLayerOrigin to be negative (which makes sense
since it's an offset of a squashed layer to the origin of
its own space of squashed layers).

Therefore the effect of this CL must be that the rounding
behavior to the [0.5, 0.5) range somehow causes compositing
to match the behavior of non-composited pixel snapping. Still
looking.

[chrishtr, 2014-06-18 17:15:19]

Success! I now have a test.

The case in which it fails is when subpixel accumulation was
previously > 0.5. This causes the code in
CLM::doPaintTask to create a different interesection rect
between dirtyRect and the composited bounds of the layer,
since the logic is:

dirtyRect.intersect(pixelSnappedIntRect(compositedBounds +
subpixelAccumulation))

Notice that offsetFromRenderer does not come into play here.
Therefore it's important that subpixelAccumulation have the
right behavior or the intersected rect will be wrong.

[leviw_travelin_and_unemployed, 2014-06-18 17:33:35]

https://codereview.chromium.org/343543005/diff/40001/Source/core/rendering/co...
File Source/core/rendering/compositing/CompositedLayerMapping.cpp (left):

https://codereview.chromium.org/343543005/diff/40001/Source/core/rendering/co...
Source/core/rendering/compositing/CompositedLayerMapping.cpp:592:
toLayoutSize(computeOffsetFromCompositedAncestor(layers[i].renderLayer,
layers[i].renderLayer->ancestorCompositingLayer())).fraction());
I'm sad to see the assert leave, but I'll just cross my fingers on our test
coverage.

LGTM.

[chrishtr, 2014-06-18 18:32:28]

The CQ bit was checked by chrishtr@chromium.org

[commit-bot: I haz the power, 2014-06-18 18:33:04]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/chrishtr@chromium.org/343543005/40001

[commit-bot: I haz the power, 2014-06-18 18:37:23]

Change committed as 176441