Add convex polygon rendering effect and GM to test it.

Add convex polygon rendering effect and GM to test it.

BUG=skia:2051
Committed: http://code.google.com/p/skia/source/detail?r=13242

[bsalomon, 2014-01-29 20:18:29]

The idea is to use this for clipping. This change is just the effect and GM that
directly tests it.

https://codereview.chromium.org/149683004/diff/40001/gm/beziereffects.cpp
File gm/beziereffects.cpp (right):

https://codereview.chromium.org/149683004/diff/40001/gm/beziereffects.cpp#new...
gm/beziereffects.cpp:26: static const GrVertexAttrib kAttribs[] = {
I think this will break some compilers because it is used as a template arg. Try
runs will find out...

https://codereview.chromium.org/149683004/diff/40001/gm/convexpolyeffect.cpp
File gm/convexpolyeffect.cpp (right):

https://codereview.chromium.org/149683004/diff/40001/gm/convexpolyeffect.cpp#...
gm/convexpolyeffect.cpp:25: static const GrVertexAttrib kAttribs[] = {
ditto.

[robertphillips, 2014-01-29 20:57:55]

https://codereview.chromium.org/149683004/diff/40001/gm/beziereffects.cpp
File gm/beziereffects.cpp (right):

https://codereview.chromium.org/149683004/diff/40001/gm/beziereffects.cpp#new...
gm/beziereffects.cpp:26: static const GrVertexAttrib kAttribs[] = {
Yep - I seem to recall the extern was required.

https://codereview.chromium.org/149683004/diff/40001/gm/convexpolyeffect.cpp
File gm/convexpolyeffect.cpp (right):

https://codereview.chromium.org/149683004/diff/40001/gm/convexpolyeffect.cpp#...
gm/convexpolyeffect.cpp:66: static const SkScalar kRadius = 50.f;
You can ignore this hyper-nit but, add a space between 's' and '}'?

https://codereview.chromium.org/149683004/diff/40001/gm/convexpolyeffect.cpp#...
gm/convexpolyeffect.cpp:98: 
Do we need this struct?

https://codereview.chromium.org/149683004/diff/40001/gm/convexpolyeffect.cpp#...
gm/convexpolyeffect.cpp:137: SkRect bounds = p.getBounds();
Can we more conservatively outset? By 1?

https://codereview.chromium.org/149683004/diff/40001/src/gpu/effects/GrConvex...
File src/gpu/effects/GrConvexPolyEffect.cpp (right):

https://codereview.chromium.org/149683004/diff/40001/src/gpu/effects/GrConvex...
src/gpu/effects/GrConvexPolyEffect.cpp:150: SkASSERT(n <= kMaxEdges);
Should we just clamp n at kMaxEdges? We could document that if you pass us too
many we're just going to truncate the list.

https://codereview.chromium.org/149683004/diff/40001/src/gpu/effects/GrConvex...
src/gpu/effects/GrConvexPolyEffect.cpp:179: edges[i] = random->nextSScalar1();
Don't we need to normalize here?

https://codereview.chromium.org/149683004/diff/40001/src/gpu/effects/GrConvex...
File src/gpu/effects/GrConvexPolyEffect.h (right):

https://codereview.chromium.org/149683004/diff/40001/src/gpu/effects/GrConvex...
src/gpu/effects/GrConvexPolyEffect.h:7: 
Fix this name?

https://codereview.chromium.org/149683004/diff/40001/src/gpu/effects/GrConvex...
src/gpu/effects/GrConvexPolyEffect.h:39: /**
Do we want to state in the comment that n must be less then kMaxEdges?

https://codereview.chromium.org/149683004/diff/40001/src/gpu/effects/GrConvex...
src/gpu/effects/GrConvexPolyEffect.h:40: * edges is a set of n edge equations.
It contains 3*n values. The edges should form a convex
equations rather than equalizations?

[bsalomon, 2014-01-29 22:19:42]

https://codereview.chromium.org/149683004/diff/40001/gm/convexpolyeffect.cpp
File gm/convexpolyeffect.cpp (right):

https://codereview.chromium.org/149683004/diff/40001/gm/convexpolyeffect.cpp#...
gm/convexpolyeffect.cpp:66: static const SkScalar kRadius = 50.f;
On 2014/01/29 20:57:55, robertphillips wrote:
> You can ignore this hyper-nit but, add a space between 's' and '}'?

Done.

https://codereview.chromium.org/149683004/diff/40001/gm/convexpolyeffect.cpp#...
gm/convexpolyeffect.cpp:98: 
On 2014/01/29 20:57:55, robertphillips wrote:
> Do we need this struct?

Done.

https://codereview.chromium.org/149683004/diff/40001/gm/convexpolyeffect.cpp#...
gm/convexpolyeffect.cpp:137: SkRect bounds = p.getBounds();
On 2014/01/29 20:57:55, robertphillips wrote:
> Can we more conservatively outset? By 1?

I chose a larger outset so that there'd be some area around the path inside the
bounding geometry. That way if the effect produces artifacts at the corners or
edges they won't get clipped out by the geometry. I'll add a comment.

https://codereview.chromium.org/149683004/diff/40001/src/gpu/effects/GrConvex...
File src/gpu/effects/GrConvexPolyEffect.cpp (right):

https://codereview.chromium.org/149683004/diff/40001/src/gpu/effects/GrConvex...
src/gpu/effects/GrConvexPolyEffect.cpp:150: SkASSERT(n <= kMaxEdges);
On 2014/01/29 20:57:55, robertphillips wrote:
> Should we just clamp n at kMaxEdges? We could document that if you pass us too
> many we're just going to truncate the list.

The Create() function is responsible for handling this. Currently it returns
NULL. This assert is just here to ensure that all the Create() variants are
doing their job. It really should never fire even if a public caller gave a
large n. Comment added.

https://codereview.chromium.org/149683004/diff/40001/src/gpu/effects/GrConvex...
src/gpu/effects/GrConvexPolyEffect.cpp:179: edges[i] = random->nextSScalar1();
On 2014/01/29 20:57:55, robertphillips wrote:
> Don't we need to normalize here?

Done.

https://codereview.chromium.org/149683004/diff/40001/src/gpu/effects/GrConvex...
File src/gpu/effects/GrConvexPolyEffect.h (right):

https://codereview.chromium.org/149683004/diff/40001/src/gpu/effects/GrConvex...
src/gpu/effects/GrConvexPolyEffect.h:7: 
On 2014/01/29 20:57:55, robertphillips wrote:
> Fix this name?

nice catch, guess you can tell how I started this effect :)

https://codereview.chromium.org/149683004/diff/40001/src/gpu/effects/GrConvex...
src/gpu/effects/GrConvexPolyEffect.h:39: /**
On 2014/01/29 20:57:55, robertphillips wrote:
> Do we want to state in the comment that n must be less then kMaxEdges?

Done.

https://codereview.chromium.org/149683004/diff/40001/src/gpu/effects/GrConvex...
src/gpu/effects/GrConvexPolyEffect.h:40: * edges is a set of n edge equations.
It contains 3*n values. The edges should form a convex
On 2014/01/29 20:57:55, robertphillips wrote:
> equations rather than equalizations?

Done.

[robertphillips, 2014-01-29 22:22:14]

lgtm provided bots go green.

[bsalomon, 2014-01-29 22:33:53]

On 2014/01/29 22:22:14, robertphillips wrote:
> lgtm provided bots go green.

oops I typed static.. meant extern. will try again later.

[commit-bot: I haz the power, 2014-01-30 17:51:11]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/bsalomon@google.com/149683004/130001

[commit-bot: I haz the power, 2014-01-30 18:15:55]

Change committed as 13242

[commit-bot: I haz the power, 2014-01-30 18:15:55]

CQ bit was unchecked on CL. Ignoring.