Updated useSWOnlyPath method to use path chain instead of raw path renderer methods

This CL re-encapsulates the path renderers. I have verified that the useSWOnlyPath function still only enables the SW-only path on the complexclip_aa GM slide.

[robertphillips, 2012-05-23 13:55:02]

http://codereview.appspot.com/6211083/diff/1/src/gpu/GrClipMaskManager.cpp
File src/gpu/GrClipMaskManager.cpp (right):

http://codereview.appspot.com/6211083/diff/1/src/gpu/GrClipMaskManager.cpp#ne...
src/gpu/GrClipMaskManager.cpp:112: // block could disappear
What is your opinion on this Brian?

[bsalomon, 2012-05-23 15:01:20]

http://codereview.appspot.com/6211083/diff/1/src/gpu/GrClipMaskManager.cpp
File src/gpu/GrClipMaskManager.cpp (right):

http://codereview.appspot.com/6211083/diff/1/src/gpu/GrClipMaskManager.cpp#ne...
src/gpu/GrClipMaskManager.cpp:112: // block could disappear
On 2012/05/23 13:55:03, robertphillips wrote:
> What is your opinion on this Brian?

Hmm... Maybe for now we just do temp.setConvexity() before the path_needs_...
call? That is definitely safe.

It does feel wasteful to create a path here (and during the draw later). We have
some code in GrContext that draws AA rects. We could give ourselves a to-do to
encapsulate it in a helper class that could be invoked by the clip manager
without ever turning the rects into paths.

[robertphillips, 2012-05-23 15:34:45]

Wouldn't this basically just entail moving GrContext::fillAARect to
GrDrawTarget?

[bsalomon, 2012-05-23 15:37:25]

On 2012/05/23 15:34:45, robertphillips wrote:
> Wouldn't this basically just entail moving GrContext::fillAARect to
> GrDrawTarget?

Maybe a helper class that takes a draw target as a param?

[robertphillips, 2012-05-24 11:54:38]

http://codereview.appspot.com/6211083/diff/1/src/gpu/GrClipMaskManager.cpp
File src/gpu/GrClipMaskManager.cpp (right):

http://codereview.appspot.com/6211083/diff/1/src/gpu/GrClipMaskManager.cpp#ne...
src/gpu/GrClipMaskManager.cpp:112: // block could disappear
On 2012/05/23 15:01:20, bsalomon wrote:
> On 2012/05/23 13:55:03, robertphillips wrote:
> > What is your opinion on this Brian?
> 
> Hmm... Maybe for now we just do temp.setConvexity() before the path_needs_...
> call? That is definitely safe.
> 
> It does feel wasteful to create a path here (and during the draw later). We
have
> some code in GrContext that draws AA rects. We could give ourselves a to-do to
> encapsulate it in a helper class that could be invoked by the clip manager
> without ever turning the rects into paths.

Done.

[bsalomon, 2012-05-24 12:37:39]

On 2012/05/24 11:54:38, robertphillips wrote:
> http://codereview.appspot.com/6211083/diff/1/src/gpu/GrClipMaskManager.cpp
> File src/gpu/GrClipMaskManager.cpp (right):
> 
>
http://codereview.appspot.com/6211083/diff/1/src/gpu/GrClipMaskManager.cpp#ne...
> src/gpu/GrClipMaskManager.cpp:112: // block could disappear
> On 2012/05/23 15:01:20, bsalomon wrote:
> > On 2012/05/23 13:55:03, robertphillips wrote:
> > > What is your opinion on this Brian?
> > 
> > Hmm... Maybe for now we just do temp.setConvexity() before the
path_needs_...
> > call? That is definitely safe.
> > 
> > It does feel wasteful to create a path here (and during the draw later). We
> have
> > some code in GrContext that draws AA rects. We could give ourselves a to-do
to
> > encapsulate it in a helper class that could be invoked by the clip manager
> > without ever turning the rects into paths.
> 
> Done.

I take it back, I forgot we had the logic in SkPath to know rects are convex
without checking. LGTM on patchset 1.

[robertphillips, 2012-05-24 12:43:01]

Received verbal LGTM from Brian

committed as r4038