Obtain conic weights for GrPath cache keys from SkPathRef

Obtain conic weights for GrPath cache keys from SkPathRef

Obtain conic data for GrPath cache kSimpleVolatilePathDomain keys
directly from SkPathRef.

Makes GrPath a friend of SkPath to for accessing SkPathRef.

BUG=skia:4580

[Kimmo Kinnunen, 2015-11-26 09:54:20]

kkinnunen@nvidia.com changed reviewers:
+ fmalita@chromium.org, reed@google.com

[Kimmo Kinnunen, 2015-11-26 09:54:20]

fmalita@ for the implementation
reed@ for the SkPath API change

[reed1, 2015-11-30 15:36:39]

https://chromiumcodereview.appspot.com/1473973003/diff/1/src/gpu/GrPath.cpp
File src/gpu/GrPath.cpp (right):

https://chromiumcodereview.appspot.com/1473973003/diff/1/src/gpu/GrPath.cpp#n...
src/gpu/GrPath.cpp:77: if (pointCnt < 0) {
Same question here: why this check? (tho I know its not part of your change).

https://chromiumcodereview.appspot.com/1473973003/diff/1/src/gpu/GrPath.cpp#n...
src/gpu/GrPath.cpp:83: if (conicWeightCnt < 0) {
Why this check?

[Kimmo Kinnunen, 2015-12-01 06:35:14]

https://chromiumcodereview.appspot.com/1473973003/diff/1/src/gpu/GrPath.cpp
File src/gpu/GrPath.cpp (right):

https://chromiumcodereview.appspot.com/1473973003/diff/1/src/gpu/GrPath.cpp#n...
src/gpu/GrPath.cpp:83: if (conicWeightCnt < 0) {
On 2015/11/30 15:36:39, reed1 wrote:
> Why this check?

Because it's an int and used as a thing below that would overflow a buffer if
the int would be negative?

[reed1, 2015-12-01 14:37:21]

reed@google.com changed reviewers:
+ bsalomon@google.com

[reed1, 2015-12-01 14:37:21]

https://chromiumcodereview.appspot.com/1473973003/diff/1/src/gpu/GrPath.cpp
File src/gpu/GrPath.cpp (right):

https://chromiumcodereview.appspot.com/1473973003/diff/1/src/gpu/GrPath.cpp#n...
src/gpu/GrPath.cpp:83: if (conicWeightCnt < 0) {
On 2015/12/01 06:35:14, Kimmo Kinnunen wrote:
> On 2015/11/30 15:36:39, reed1 wrote:
> > Why this check?
> 
> Because it's an int and used as a thing below that would overflow a buffer if
> the int would be negative?

Yes, but can it ever be negative? This is an internal value, and typically we
don't spend runtime checking internal structures for self-consistency. Is there
a scenario where this can happen?

[Kimmo Kinnunen, 2015-12-01 16:42:07]

On 2015/12/01 14:37:21, reed1 wrote:
> https://chromiumcodereview.appspot.com/1473973003/diff/1/src/gpu/GrPath.cpp
> File src/gpu/GrPath.cpp (right):
> 
>
https://chromiumcodereview.appspot.com/1473973003/diff/1/src/gpu/GrPath.cpp#n...
> src/gpu/GrPath.cpp:83: if (conicWeightCnt < 0) {
> On 2015/12/01 06:35:14, Kimmo Kinnunen wrote:
> > On 2015/11/30 15:36:39, reed1 wrote:
> > > Why this check?
> > 
> > Because it's an int and used as a thing below that would overflow a buffer
if
> > the int would be negative?
> 
> Yes, but can it ever be negative? This is an internal value, and typically we
> don't spend runtime checking internal structures for self-consistency.

Sure, I removed the check. How about now?

> Is there a scenario where this can happen?

(I thought the use of signedness was due to reserving the possibility for such a
scenario.)