ui: Remove implicit flooring in skia rect conversion methods.

ui: Remove implicit flooring in skia rect conversion methods.

The current SkRectToRect method implicitly floors all components of thre rect
and this is bad. Replace this method with SkRectToRectF which keeps everything
as floating point.

We also add a conversion method from RectF to SkRect.

Tests:
ui_unittests:RectTest.SkiaRectConversions

R=sky
BUG=147395
Moved From: https://codereview.chromium.org/11275089/

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=167620

[danakj, 2012-11-13 00:21:26]

Hi Sky,

I'm picking up https://codereview.chromium.org/11275089/ here.

I used ToEnclosingRect in most of the places. And just avoided converting to an
int at all in a couple.

[danakj, 2012-11-13 00:24:52]

+jamesr for cc/ since you reviewed the previous incarnation

[jamesr, 2012-11-13 00:30:23]

cc lgtm

[sky, 2012-11-13 01:18:18]

LGTM

[danakj, 2012-11-13 02:09:13]

Fixed tests.

aura_tests:WindowTest.Transform will pass when we switch to using doubles in
matrices. Rotation by 90 degrees is too lossy with floats.

[sky, 2012-11-13 17:18:18]

SLGTM

[commit-bot: I haz the power, 2012-11-13 23:15:32]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/danakj@chromium.org/11359172/9003

[danakj, 2012-11-14 02:30:30]

Hi Sky. We need one more round for this. (No pun intended..)

The root_window.cc case seems to be flaky depending on the machine and it makes
me uneasy. In its case, it is truly expecting the resulting RectF to be an
integer rect still, with some floating point error.

So I added ToNearestRect() for this case, with some comments and DCHECKs
accordingly.

What do you think of that?

[sky, 2012-11-14 02:45:41]

I don't think WindowTest.Transform is really valid. Its rotating the rootwindow,
which we have no need to do.

https://codereview.chromium.org/11359172/diff/7020/ui/gfx/rect_conversions.h
File ui/gfx/rect_conversions.h (right):

https://codereview.chromium.org/11359172/diff/7020/ui/gfx/rect_conversions.h#...
ui/gfx/rect_conversions.h:21: // interger rect with floating poing error. If it
is an arbitrary RectF, then
interger -> integer
poing -> point

[sky, 2012-11-14 02:45:49]

LGTM

[danakj, 2012-11-14 02:50:58]

On 2012/11/14 02:45:49, sky wrote:
> LGTM

Thanks.

[commit-bot: I haz the power, 2012-11-14 05:24:18]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/danakj@chromium.org/11359172/12024

[commit-bot: I haz the power, 2012-11-14 07:28:14]

Change committed as 167620