Implement SkPictureImageFilter bounds computation.

Implement SkPictureImageFilter bounds computation.

SkPictureImageFilter::onFilterBounds() was unimplemented, causing
incorrect results for SVG filters and impl-side painting (see Chrome
bug https://code.google.com/p/chromium/issues/detail?id=375162).

BUG=skia:
R=reed@google.com

Committed: https://code.google.com/p/skia/source/detail?r=15003

[Stephen White, 2014-05-30 19:54:54]

reed@: PTAL. Thanks!

[reed1, 2014-05-30 20:10:19]

https://codereview.chromium.org/306733003/diff/60001/src/effects/SkPictureIma...
File src/effects/SkPictureImageFilter.cpp (right):

https://codereview.chromium.org/306733003/diff/60001/src/effects/SkPictureIma...
src/effects/SkPictureImageFilter.cpp:98: *dst = src;
Why are we not applying the CTM?

[Stephen White, 2014-05-30 20:21:24]

On 2014/05/30 20:10:19, reed1 wrote:
>
https://codereview.chromium.org/306733003/diff/60001/src/effects/SkPictureIma...
> File src/effects/SkPictureImageFilter.cpp (right):
> 
>
https://codereview.chromium.org/306733003/diff/60001/src/effects/SkPictureIma...
> src/effects/SkPictureImageFilter.cpp:98: *dst = src;
> Why are we not applying the CTM?

src is in device space already (hence the SkIRect).

[reed1, 2014-05-30 20:25:30]

On 2014/05/30 20:21:24, Stephen White wrote:
> On 2014/05/30 20:10:19, reed1 wrote:
> >
>
https://codereview.chromium.org/306733003/diff/60001/src/effects/SkPictureIma...
> > File src/effects/SkPictureImageFilter.cpp (right):
> > 
> >
>
https://codereview.chromium.org/306733003/diff/60001/src/effects/SkPictureIma...
> > src/effects/SkPictureImageFilter.cpp:98: *dst = src;
> > Why are we not applying the CTM?
> 
> src is in device space already (hence the SkIRect).

Then I don't understand what the method is for. Other filters seem to transform
their own bounds/vectors, and/or the src rect, by the CTM and then compute dst.
This one ignores its bounds and the ctm parameter. I must be misunderstanding
the dox for this method.

[reed1, 2014-05-30 20:31:54]

Perhaps calling the two parameters src and dst is what throws me off. I think I
see now how this CL is compatible with other effects.

lgtm

[Stephen White, 2014-05-30 20:35:10]

On 2014/05/30 20:31:54, reed1 wrote:
> Perhaps calling the two parameters src and dst is what throws me off. I think
I
> see now how this CL is compatible with other effects.
> 
> lgtm

Yeah it is a bit confusing; both src and dst are in device space. The CTM is
used to transform the filter's parameters, not the rects.

[Stephen White, 2014-05-30 20:36:18]

Committed patchset #4 manually as r15003 (presubmit successful).