Revert of Fix bugs in clip parent calculations

Revert of Fix bugs in clip parent calculations (https://codereview.chromium.org/402283003/)

Reason for revert:
non-reparented-overlay-scrollbars.html fails on the following bots:

http://build.chromium.org/p/chromium.webkit/builders/WebKit%20Mac10.7/builds/29607
http://build.chromium.org/p/chromium.webkit/builders/WebKit%20Mac10.8/builds/27919
http://build.chromium.org/p/chromium.webkit/builders/WebKit%20Mac10.8%20%28retina%29/builds/16557

Original issue's description:
> Fix bugs in clip parent calculations
> 
> I noticed two bugs in the clip parent code - when walking up the layers
> in the stacking tree looking for clipping layers, I'd forgotten to
> consider the layers themselves (I'd only considered their clipping
> containers). I'd also incorrectly gated promotion for having a clip
> parent on having an ancestor scrolling layer that needed composited
> scrolling. This condition was bogus: we often need a clip parent
> precisely because we have no ancestor scrolling layer (eg, we're fixed
> positioned and have 'escaped' our ancestor scrollers in tree order).
> 
> BUG=None
> 
> Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=178649

TBR=abarth@chromium.org,vollick@chromium.org
NOTREECHECKS=true
NOTRY=true
BUG=None

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=178665

[johnme, 2014-07-22 15:06:03]

Created Revert of Fix bugs in clip parent calculations

[commit-bot: I haz the power, 2014-07-22 15:06:52]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/johnme@chromium.org/410683002/1

[commit-bot: I haz the power, 2014-07-22 15:08:01]

Change committed as 178665

[abarth-chromium, 2014-07-22 16:53:26]

Was this a real failure or just an instance of
https://code.google.com/p/chromium/issues/detail?id=394072 ?

[Ian Vollick, 2014-07-22 17:07:53]

On 2014/07/22 16:53:26, abarth wrote:
> Was this a real failure or just an instance of
> https://code.google.com/p/chromium/issues/detail?id=394072 ?

I'm looking into it, but I don't think so. With these failures, we're missing
the horizontal scrollbars in the 'actual' test results on mac. Those scrollbars
ought to be there with or without my patch, so something fishy is happening.
I'll post when I know more.

[johnme, 2014-07-22 17:26:02]

On 2014/07/22 16:53:26, abarth wrote:
> Was this a real failure or just an instance of
> https://code.google.com/p/chromium/issues/detail?id=394072 ?

This wasn't issue 394072 - I checked and the builder and tester were using the
exact same revisions. Also, it turned the tree red for 4 builds in a row.

[Ian Vollick, 2014-07-22 18:10:52]

On 2014/07/22 17:26:02, johnme wrote:
> On 2014/07/22 16:53:26, abarth wrote:
> > Was this a real failure or just an instance of
> > https://code.google.com/p/chromium/issues/detail?id=394072 ?
> 
> This wasn't issue 394072 - I checked and the builder and tester were using the
> exact same revisions. Also, it turned the tree red for 4 builds in a row.

I've found the root cause. The problem is unrelated to this patch. There's a
div-by-zero in cc's thumb geometry computation for the horizontal scrollbar (the
zero is the amount you can scroll on that axis). This only gets hit when the
scroller is composited, obv (otherwise the cc code wouldn't run). I'm not sure
why this only happens on these mac platforms, though. Maybe they handle the NaN
+ casting differently? (shrug) In any case, the fix seems trivial. Will have
something uploaded soon.

[Ian Vollick, 2014-07-22 19:02:10]

On 2014/07/22 18:10:52, Ian Vollick wrote:
> On 2014/07/22 17:26:02, johnme wrote:
> > On 2014/07/22 16:53:26, abarth wrote:
> > > Was this a real failure or just an instance of
> > > https://code.google.com/p/chromium/issues/detail?id=394072 ?
> > 
> > This wasn't issue 394072 - I checked and the builder and tester were using
the
> > exact same revisions. Also, it turned the tree red for 4 builds in a row.
> 
> I've found the root cause. The problem is unrelated to this patch. There's a
> div-by-zero in cc's thumb geometry computation for the horizontal scrollbar
(the
> zero is the amount you can scroll on that axis). This only gets hit when the
> scroller is composited, obv (otherwise the cc code wouldn't run). I'm not sure
> why this only happens on these mac platforms, though. Maybe they handle the
NaN
> + casting differently? (shrug) In any case, the fix seems trivial. Will have
> something uploaded soon.

Fix is in the cq, if you're interested:
https://codereview.chromium.org/409193002/

[abarth-chromium, 2014-07-22 19:31:40]

Thanks!