Trigger extra GraphicsLayerGeometryUpdate for squashing if we reset compositedBounds on squashed la…

Trigger extra GraphicsLayerGeometryUpdate for squashing if we reset compositedBounds on squashed layers.

Currently we're skipping a necessary update case in squashed layers due to an
early-out in CompositedLayerMapping::updateGraphicsLayerGeometry() because
m_needToUpdateGeometry is false. This explicitly sets m_needToUpdateGeometry
to true when we reset the squashed layers' compositedBounds so that they're
correctly recomputed before trying to paint.

BUG=349308

[hartmanng, 2014-03-18 18:58:46]

Ian, please take a look. The fix is trivial, but I'm still working on a layout
test (have actually been unable so far to reproduce this issue without
https://codereview.chromium.org/131543013/ applied). I can keep looking for a
repro, but it's possible that we might need to wait until after that CL lands to
get a proper layout test for this.

Alternatively, since this CL is so small, it might make sense to just stick it
in https://codereview.chromium.org/131543013/ and remove the for loops and
associated FIXME from
https://codereview.chromium.org/131543013/diff/220001/LayoutTests/compositing....

wdyt?

[Ian Vollick, 2014-03-18 23:48:57]

On 2014/03/18 18:58:46, hartmanng wrote:
> Ian, please take a look. The fix is trivial, but I'm still working on a layout
> test (have actually been unable so far to reproduce this issue without
> https://codereview.chromium.org/131543013/ applied). I can keep looking for a
> repro, but it's possible that we might need to wait until after that CL lands
to
> get a proper layout test for this.
> 
> Alternatively, since this CL is so small, it might make sense to just stick it
> in https://codereview.chromium.org/131543013/ and remove the for loops and
> associated FIXME from
>
https://codereview.chromium.org/131543013/diff/220001/LayoutTests/compositing....
> 
> wdyt?

Nice one! If it's true that the issue only happens with my patch, and only when
univ overflow scroll is on, I think you should keep this separate.

[chrishtr, 2014-03-19 21:51:41]

This CL is obsoleted by https://codereview.chromium.org/204983002/