drawVertices bug on gpu side

Fixed the bug in drawVertices when no colors or texture coordinates are provided by modifying the grprimitive to klines and adjusting the indices.

BUG=skia:2745
Committed: https://skia.googlesource.com/skia/+/612364de87cb2012239c4f6435e8e11760617fdc

Committed: https://skia.googlesource.com/skia/+/32a311bfc056169c9f3b1c299c57ff5139a0d8bc

[dandov, 2014-07-14 13:47:39]

I found this bug when I was trying to draw a patch using drawVertices, in the
GPU side it crashed when I didn't pass the color array to the function.

[jvanverth1, 2014-07-14 14:29:20]

https://codereview.chromium.org/387113002/diff/20001/src/gpu/SkGpuDevice.cpp
File src/gpu/SkGpuDevice.cpp (right):

https://codereview.chromium.org/387113002/diff/20001/src/gpu/SkGpuDevice.cpp#...
src/gpu/SkGpuDevice.cpp:1618: SkPaint copy(paint);
I don't think we'd want to copy the paint all the time if we don't have to.
Maybe move the declaration of grPaint here, create the copy only in the hairline
case, and put the  specific SkPaint2GrPaintNoShader calls inside the 'if' and
the 'else' blocks.

https://codereview.chromium.org/387113002/diff/20001/src/gpu/SkGpuDevice.cpp#...
src/gpu/SkGpuDevice.cpp:1688: if (copy.getAlpha() != 255) {
The paint color doesn't change in the hairline case, so you could use the
original paint here.

[dandov, 2014-07-14 15:31:34]

Moved declaration of GrPaint to the top and only copy the SkPaint when it is
necessary (when the colors and texture coordinates weren't provided).

https://codereview.chromium.org/387113002/diff/20001/src/gpu/SkGpuDevice.cpp
File src/gpu/SkGpuDevice.cpp (right):

https://codereview.chromium.org/387113002/diff/20001/src/gpu/SkGpuDevice.cpp#...
src/gpu/SkGpuDevice.cpp:1618: SkPaint copy(paint);
On 2014/07/14 14:29:19, jvanverth1 wrote:
> I don't think we'd want to copy the paint all the time if we don't have to.
> Maybe move the declaration of grPaint here, create the copy only in the
hairline
> case, and put the  specific SkPaint2GrPaintNoShader calls inside the 'if' and
> the 'else' blocks.

Done.

https://codereview.chromium.org/387113002/diff/20001/src/gpu/SkGpuDevice.cpp#...
src/gpu/SkGpuDevice.cpp:1688: if (copy.getAlpha() != 255) {
On 2014/07/14 14:29:20, jvanverth1 wrote:
> The paint color doesn't change in the hairline case, so you could use the
> original paint here.

Done.

[dandov, 2014-07-14 17:36:38]

Removed memcpy for the same indices that were passed and added auto deleter for
indices that are allocated in the call.

https://codereview.chromium.org/387113002/diff/70001/src/gpu/SkGpuDevice.cpp
File src/gpu/SkGpuDevice.cpp (right):

https://codereview.chromium.org/387113002/diff/70001/src/gpu/SkGpuDevice.cpp#...
src/gpu/SkGpuDevice.cpp:1652: idxDeleter.reset(idx);
added auto deleter to avoid memory leak

https://codereview.chromium.org/387113002/diff/70001/src/gpu/SkGpuDevice.cpp#...
src/gpu/SkGpuDevice.cpp:1665: idx = (uint16_t*)indices;
do not copy indices when not needed

[dandov, 2014-07-14 20:10:14]

Removed cast for const uint16_t*

https://codereview.chromium.org/387113002/diff/90001/src/gpu/SkGpuDevice.cpp
File src/gpu/SkGpuDevice.cpp (right):

https://codereview.chromium.org/387113002/diff/90001/src/gpu/SkGpuDevice.cpp#...
src/gpu/SkGpuDevice.cpp:1616: const uint16_t* outIndices;
now pointer to const to avoid cast from pointer to non-const

[jvanverth1, 2014-07-14 20:17:07]

lgtm

[dandov, 2014-07-14 21:13:28]

The CQ bit was checked by dandov@google.com

[commit-bot: I haz the power, 2014-07-14 21:14:16]

CQ is trying da patch. Follow status at
 https://skia-tree-status.appspot.com/cq/dandov@google.com/387113002/90001

[commit-bot: I haz the power, 2014-07-14 21:26:09]

Change committed as 612364de87cb2012239c4f6435e8e11760617fdc

[dandov, 2014-07-14 21:51:37]

A revert of this CL has been created in
https://codereview.chromium.org/395613002/ by dandov@google.com.

The reason for reverting is: compare gms failed on 6 images, I will check them
and determine if they need a rebaseline or if the output is incorrect. .

[dandov, 2014-07-15 17:16:56]

The CQ bit was checked by dandov@google.com

[commit-bot: I haz the power, 2014-07-15 17:18:04]

CQ is trying da patch. Follow status at
 https://skia-tree-status.appspot.com/cq/dandov@google.com/387113002/110001

[commit-bot: I haz the power, 2014-07-16 02:46:31]

Change committed as 32a311bfc056169c9f3b1c299c57ff5139a0d8bc