Fix ASSERT failure in RenderGeometryMap.

Fix ASSERT failure in RenderGeometryMap.

This ASSERT is too strong, and doesn't quite reflect the
intended behaviour. Invisible layers are skipped when
recomputing location in
RenderLayer::updateLayerPositionsAfterScroll(), but the
ASSERT assumes that even invisible layers should be correctly
positioned.

This patch changes the ASSERT to more accurately reflect
the correct intended behaviour, by skipping invisible
layers.

BUG=306100
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=159922

[hartmanng, 2013-10-15 13:53:19]

Ian, please take a look.

[Ian Vollick, 2013-10-15 14:38:06]

On 2013/10/15 13:53:19, hartmanng wrote:
> Ian, please take a look.

lgtm!

[Ian Vollick, 2013-10-15 14:40:18]

On 2013/10/15 14:38:06, Ian Vollick wrote:
> On 2013/10/15 13:53:19, hartmanng wrote:
> > Ian, please take a look.
> 
> lgtm!

Please add a layout test, though. If this regresses in future, it will be
helpful to catch on the dbg bots.

[hartmanng, 2013-10-17 20:38:38]

+ layout test

PTAL

[Ian Vollick, 2013-10-17 20:52:35]

On 2013/10/17 20:38:38, hartmanng wrote:
> + layout test
> 
> PTAL

Sweet. Thanks for the test :) slgtm

[commit-bot: I haz the power, 2013-10-17 20:58:10]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/hartmanng@chromium.org/27104002/47001

[shawnsingh, 2013-10-17 21:53:02]

On 2013/10/17 20:58:10, I haz the power (commit-bot) wrote:
> CQ is trying da patch. Follow status at
> https://chromium-status.appspot.com/cq/hartmanng%40chromium.org/27104002/47001

Hey Glenn, Do you think the MapToContainer( point ) overloaded version has the
same problem with its own ASSERT?  I'm running into that ASSERT which seems
bogus on https://codereview.chromium.org/26110004/ too...

[shawnsingh, 2013-10-17 21:55:23]

On 2013/10/17 21:53:02, shawnsingh wrote:
> On 2013/10/17 20:58:10, I haz the power (commit-bot) wrote:
> > CQ is trying da patch. Follow status at
> >
https://chromium-status.appspot.com/cq/hartmanng%2540chromium.org/27104002/47001
> 
> Hey Glenn, Do you think the MapToContainer( point ) overloaded version has the
> same problem with its own ASSERT?  I'm running into that ASSERT which seems
> bogus on https://codereview.chromium.org/26110004/ too...

actually, on second thought, the error i'm running into is that the last()
accessor returns nothing, bceause the vector is empty.  Do you know whether it's
reasonable to bail out of this ASSERT if size() == 0 before trying to assert the
condition?  I think the same applies to both overloaded implementations of
mapToContainer.

[commit-bot: I haz the power, 2013-10-18 01:49:56]

Change committed as 159922