Don't shrink starting_rect in the first round.

Don't shrink starting_rect in the first round.

Sometimes a target_area smaller than the viewport (starting_rect) is
passed to ExpandRectEquallyToAreaBoundedBy(). If we shrink the
starting_rect to the target_area before considering the bounds, a
bound in the viewport but at the edge of viewport will be excluded
from prioritization.

In the first round only expand but not shrink starting_rect.

BUG=231370
TEST=PictureLayerTilingSet.ShrinkRectContainingBounds

[Xianzhu, 2013-04-16 21:50:09]

This may be related to https://codereview.chromium.org/13895005/ but perhaps
resolves different issues.

[danakj, 2013-04-16 21:53:04]

What do you think of this compared to https://codereview.chromium.org/13895005/
?

[enne (OOO), 2013-04-16 21:54:11]

Punting review of this to Dana.

[Xianzhu, 2013-04-16 22:12:08]

On 2013/04/16 21:53:04, danakj wrote:
> What do you think of this compared to
https://codereview.chromium.org/13895005/
> ?

I made this change before I knew yours which covers the case of this one.

This patch resolves the chrome-os issue when target_area is smaller than the
normal (not enormous) viewport size.

I'm still reading your patch. Will abandon this one if it can be fully covered
by yours.

[Xianzhu, 2013-04-16 22:44:50]

I'm sure this one is covered by 13895005. Abandoned.

[Xianzhu, 2013-04-17 00:14:38]

On 2013/04/16 22:44:50, Xianzhu wrote:
> I'm sure this one is covered by 13895005. Abandoned.

Reopened as we need a simple safe fix to be merged into m27. Consider this CL as
a candidate.