Implement getConstantColorComponents() for matrix convolution.

Implement getConstantColorComponents() for matrix convolution.

R=senorblanco@chromium.org
Committed: https://code.google.com/p/skia/source/detail?r=7146

[bsalomon, 2013-01-11 21:17:33]

I noticed just after landing the CL that added getConstantColorComponents() that
I overlooked this effect. It is inheriting GrSingleTextureEffect's version which
seems to be incorrect.

[Stephen White, 2013-01-11 21:22:48]

On 2013/01/11 21:17:33, bsalomon wrote:
> I noticed just after landing the CL that added getConstantColorComponents()
that
> I overlooked this effect. It is inheriting GrSingleTextureEffect's version
which
> seems to be incorrect.

Maybe the single texture effect should be conservative instead?

Do we have test coverage for this?

[bsalomon, 2013-01-11 21:30:19]

On 2013/01/11 21:22:48, Stephen White wrote:
> On 2013/01/11 21:17:33, bsalomon wrote:
> > I noticed just after landing the CL that added getConstantColorComponents()
> that
> > I overlooked this effect. It is inheriting GrSingleTextureEffect's version
> which
> > seems to be incorrect.
> 
> Maybe the single texture effect should be conservative instead?

I think single texture effect should be an abstract base class that doesn't
implement this function (forcing subclasses to do it). It's implementation of
getCCC() is correct for its GrGLEffect, but subclasses of GrSingleTextureEffect
have their own GrGLEffects.

> 
> Do we have test coverage for this?

In general there are lots of GMs that take different code paths if
srcAlphaIsOne() returns true (based on the result of GrEffect::getCCC()). But
this isn't necessarily true for every effect's implementation.

[Stephen White, 2013-01-11 21:34:07]

OK, LGTM.