Fixing deferred canvas bug caused by SkTwoPointConicalGradient declaring itself as opaque.

Fixing deferred canvas bug caused by SkTwoPointConicalGradient declaring itself as opaque.

BUG=https://code.google.com/p/chromium/issues/detail?id=222140
TEST=DeferredCanvas unit test + DRT with --enable-deferred-2d-canvas

Committed: http://code.google.com/p/skia/source/detail?r=8247

[bsalomon, 2013-03-19 21:27:06]

On 2013/03/19 20:39:11, junov wrote:

lgtm

[commit-bot: I haz the power, 2013-03-20 00:49:53]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/junov@chromium.org/12879005/5002

[commit-bot: I haz the power, 2013-03-20 00:50:01]

Change committed as 8247

[reed1, 2013-03-20 13:27:36]

I think there are times that conical draws everywhere (e.g. when the first
circle is completely inside the 2nd), at least that's how our GMs look. Seems
like we shouldn't take that into account, and not always return false.

[Justin Novosad, 2013-03-20 14:18:39]

After working out the math in my head, I think it is actually when the *center*
of the smallest of the two circles is inside the larger circle. By "inside", I
mean excluding the edge.

R1 != R2 && (C1 - C2).length() < max(R1, R2)

I want to be 100% sure this is correct before submitting it. For Now I prefer to
be conservative.  I will log a bug right away.

[reed1, 2013-03-20 14:47:02]

On 2013/03/20 14:18:39, junov wrote:
> After working out the math in my head, I think it is actually when the
*center*
> of the smallest of the two circles is inside the larger circle. By "inside", I
> mean excluding the edge.
> 
> R1 != R2 && (C1 - C2).length() < max(R1, R2)
> 
> I want to be 100% sure this is correct before submitting it. For Now I prefer
to
> be conservative.  I will log a bug right away.

Sure, lets file a bug, and perhaps add this idea to the comment in the code
where we return false for now.