GrClip no longer translates its clips (to better mimic SkClipStack's behavior)

This seemed the cleanest way to get the translation capabilities required for the SkClipStack delivered apart from all the other changes (i.e., dumbing down GrClip a bit)

[bsalomon, 2012-07-27 18:24:43]

http://codereview.appspot.com/6445052/diff/1001/src/gpu/GrClipMaskManager.cpp
File src/gpu/GrClipMaskManager.cpp (right):

http://codereview.appspot.com/6445052/diff/1001/src/gpu/GrClipMaskManager.cpp...
src/gpu/GrClipMaskManager.cpp:768: GrMatrix m;
I think you can just say:

drawState->viewMatrix()->setTranslate(...);

and save a matrix copy.

http://codereview.appspot.com/6445052/diff/1001/src/gpu/GrClipMaskManager.cpp...
src/gpu/GrClipMaskManager.cpp:774: }
Would it be any better to fixup rtRect and bounds here rather than in the two
places below (and in process_initial_clip_elements)?

[robertphillips, 2012-07-30 12:01:32]

I think we should revisit the canvas-space/device-space issue after we've
implemented the logic to prevent clipping to the render target in some cases.
That feature will require a more mature handling of the coordinate spaces and
switching everything to canvas space after that might be less jarring.

http://codereview.appspot.com/6445052/diff/1001/src/gpu/GrClipMaskManager.cpp
File src/gpu/GrClipMaskManager.cpp (right):

http://codereview.appspot.com/6445052/diff/1001/src/gpu/GrClipMaskManager.cpp...
src/gpu/GrClipMaskManager.cpp:768: GrMatrix m;
On 2012/07/27 18:24:43, bsalomon wrote:
> I think you can just say:
> 
> drawState->viewMatrix()->setTranslate(...);
> 
> and save a matrix copy.

Done.

http://codereview.appspot.com/6445052/diff/1001/src/gpu/GrClipMaskManager.cpp...
src/gpu/GrClipMaskManager.cpp:774: }
I attempted to switch everything over to working in canvas space but it didn't
work out well. This method (createStencilClipMask) worked out pretty well but
the other two (createAlphaClipMask & createSoftwareClipMask) got a bit messy.
The main problem is/was that the transform to device coordinates is non-linear
(due to the clamp to the render target bounds) and in some places we really do
require the device coordinate bounds (clearStencilClip/flushScissor).

I have removed the creation of the "temp" Rect for the contains calls which does
streamline things a bit.

[bsalomon, 2012-07-30 12:46:12]

LGTM

http://codereview.appspot.com/6445052/diff/1001/src/gpu/GrClipMaskManager.cpp
File src/gpu/GrClipMaskManager.cpp (right):

http://codereview.appspot.com/6445052/diff/1001/src/gpu/GrClipMaskManager.cpp...
src/gpu/GrClipMaskManager.cpp:774: }
On 2012/07/30 12:01:32, robertphillips wrote:
> I attempted to switch everything over to working in canvas space but it didn't
> work out well. This method (createStencilClipMask) worked out pretty well but
> the other two (createAlphaClipMask & createSoftwareClipMask) got a bit messy.
> The main problem is/was that the transform to device coordinates is non-linear
> (due to the clamp to the render target bounds) and in some places we really do
> require the device coordinate bounds (clearStencilClip/flushScissor).
> 
> I have removed the creation of the "temp" Rect for the contains calls which
does
> streamline things a bit.

I'll take your word for it. I think we should consider some prefixes in this
file that indicate which space everything is in. And perhaps canvas_to_device
device_to_canvas inline functions for points/rects.

[robertphillips, 2012-07-31 12:18:48]

committed as r4848