Added Z scale when X and Y scale to spot lights and point lights

Added Z scale when X and Y scale to spot lights and point lights

Z scale is set as the average of X scale and Y scale.

BUG=skia:

Committed: http://code.google.com/p/skia/source/detail?r=13798

[Stephen White, 2014-03-12 19:46:23]

Looks good, but needs tests. I suggest adding a test case or two to
imagefiltersscaled.

[sugoi1, 2014-03-12 20:27:48]

On 2014/03/12 19:46:23, Stephen White wrote:
> Looks good, but needs tests. I suggest adding a test case or two to
> imagefiltersscaled.

Tests added

[Stephen White, 2014-03-12 20:47:08]

LGTM

[Stephen White, 2014-03-12 20:51:00]

Hmm.. actually, I patched this in locally, and the spotlight versions don't look
correct. Granted, the scale is non-uniform, so it could be the Z approximation.
But it looks like the target X is incorrect.

[Stephen White, 2014-03-12 20:53:30]

On 2014/03/12 20:51:00, Stephen White wrote:
> Hmm.. actually, I patched this in locally, and the spotlight versions don't
look
> correct. Granted, the scale is non-uniform, so it could be the Z
approximation.
> But it looks like the target X is incorrect.

Checking your new GM against master, it looks like that problem predates this
patch.

[Stephen White, 2014-03-12 20:57:44]

On 2014/03/12 20:53:30, Stephen White wrote:
> On 2014/03/12 20:51:00, Stephen White wrote:
> > Hmm.. actually, I patched this in locally, and the spotlight versions don't
> look
> > correct. Granted, the scale is non-uniform, so it could be the Z
> approximation.
> > But it looks like the target X is incorrect.
> 
> Checking your new GM against master, it looks like that problem predates this
> patch.

Looks like it's because we compute the S vector between target & location in the
spotlight constructor, and don't adjust it for non-uniform scale in the matrix
later. 
We could either try to munge the S vector by the CTM at draw time, or keep the 
target around, transform that by the CTM and compute S at draw time. The latter
is 
probably easier.

Would you mind taking a stab at fixing that in this patch too, just so we don't
have to rebaseline the images twice?

[sugoi1, 2014-03-13 18:25:17]

On 2014/03/12 20:57:44, Stephen White wrote:
> On 2014/03/12 20:53:30, Stephen White wrote:
> > On 2014/03/12 20:51:00, Stephen White wrote:
> > > Hmm.. actually, I patched this in locally, and the spotlight versions
don't
> > look
> > > correct. Granted, the scale is non-uniform, so it could be the Z
> > approximation.
> > > But it looks like the target X is incorrect.
> > 
> > Checking your new GM against master, it looks like that problem predates
this
> > patch.
> 
> Looks like it's because we compute the S vector between target & location in
the
> spotlight constructor, and don't adjust it for non-uniform scale in the matrix
> later. 
> We could either try to munge the S vector by the CTM at draw time, or keep the

> target around, transform that by the CTM and compute S at draw time. The
latter
> is 
> probably easier.
> 
> Would you mind taking a stab at fixing that in this patch too, just so we
don't
> have to rebaseline the images twice?

Fixed

[Stephen White, 2014-03-13 18:47:46]

LGTM. Thanks for the extra effort.

https://codereview.chromium.org/198013002/diff/40001/expectations/gm/ignored-...
File expectations/gm/ignored-tests.txt (right):

https://codereview.chromium.org/198013002/diff/40001/expectations/gm/ignored-...
expectations/gm/ignored-tests.txt:58: colormatrix
I think you'll have to updated to ToT; this was removed.

[sugoi1, 2014-03-13 18:54:23]

The CQ bit was checked by sugoi@chromium.org

[commit-bot: I haz the power, 2014-03-13 18:54:25]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/sugoi@chromium.org/198013002/60001

[commit-bot: I haz the power, 2014-03-13 19:45:45]

Change committed as 13798