Change Size::IsEmpty() to be consistent with Rect::IsEmpty()

Change Size::IsEmpty() to be consistent with Rect::IsEmpty()
Change Size to not accept negative dimensions to be consistent with Rect.

BUG=10992
TEST=base_unittests.exe --gtest_filter=RectTest.IsEmpty


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=14567

[Erik does not do reviews, 2009-04-24 20:25:07]

I'm a bit nervous about a low level change like this.  The try server seem
happy, but if you guys have any other ideas, let me know.

[brettw, 2009-04-24 20:30:24]

+Ben who I think worked on this.

[Ben Goodger (Google), 2009-04-24 20:39:56]

On 2009/04/24 20:30:24, brettw wrote:
> +Ben who I think worked on this.

LGTM - though when measuring dimensions I find width_ == 0 more readable than
!width_. Just my opinion though.

[darin (slow to review), 2009-04-24 21:32:42]

WebCore's IntRect and IntSize actually do <= 0 comparisons for the isEmpty
check.  I think that might make sense for us too, but it certainly seems a bit
more risky.

[Erik does not do reviews, 2009-04-26 21:25:10]

On 2009/04/24 21:32:42, darin wrote:
> WebCore's IntRect and IntSize actually do <= 0 comparisons for the isEmpty
> check.  I think that might make sense for us too, but it certainly seems a bit
> more risky.

I made this change, but then noticed when I tried to unit test it that Rect
doesn't allow negative dimensions.  But Size does.  Whee!  So now I've changed
Size to not allow them in the same way.  How could anything go wrong?

[Erik does not do reviews, 2009-04-26 22:47:16]

I submitted this TBR since you had all basically bought in to the original
change, and since I could easily test and roll it back on the weekend if it
caused any grief.  It appears to have passed all of the bots.