Addition of flush to SkGrTexturePixelRef::deepCopy(...)

Addition of necessary flush to copyToTexturePixelRef(...).

Ganesh batches drawing operations, which means that the copy operation, which does not flush the pipeline, can capture stale contents.  By forcing a flush, the up-to-date version of the texture is copied.

TEST=none
BUG=124054
Committed: https://code.google.com/p/skia/source/detail?r=3785

[twiz1, 2012-04-26 22:54:21]

PTAL.

[brian2, 2012-04-26 23:42:22]

On 2012/04/26 22:54:21, twiz1 wrote:
> PTAL.

Jeff, nice catch! I think the flush should go into GrContext::copyTexture,
though. We should probably have a comment similar to the one in
resolveRenderTarget, just above copyTexture in GrContext.cpp, indicating that we
aren't tracking whether there are pending writes to the src texture.

[bsalomon, 2012-04-27 00:04:58]

On 2012/04/26 23:42:22, brian2 wrote:
> On 2012/04/26 22:54:21, twiz1 wrote:
> > PTAL.
> 
> Jeff, nice catch! I think the flush should go into GrContext::copyTexture,
> though. We should probably have a comment similar to the one in
> resolveRenderTarget, just above copyTexture in GrContext.cpp, indicating that
we
> aren't tracking whether there are pending writes to the src texture.

BTW this comment was from me (accidentally posted from personal account).

[twiz1, 2012-04-27 19:22:25]

On 2012/04/27 00:04:58, bsalomon wrote:
> On 2012/04/26 23:42:22, brian2 wrote:
> > On 2012/04/26 22:54:21, twiz1 wrote:
> > > PTAL.
> > 
> > Jeff, nice catch! I think the flush should go into GrContext::copyTexture,
> > though. We should probably have a comment similar to the one in
> > resolveRenderTarget, just above copyTexture in GrContext.cpp, indicating
that
> we
> > aren't tracking whether there are pending writes to the src texture.
> 
> BTW this comment was from me (accidentally posted from personal account).

Thanks for your review, Brian.  I made the changes you suggested.

It would be interesting if we could selectively flush to the given texture,
instead of flushing the whole pipeline.

[bsalomon, 2012-04-27 19:27:16]

On 2012/04/27 19:22:25, twiz1 wrote:
> On 2012/04/27 00:04:58, bsalomon wrote:
> > On 2012/04/26 23:42:22, brian2 wrote:
> > > On 2012/04/26 22:54:21, twiz1 wrote:
> > > > PTAL.
> > > 
> > > Jeff, nice catch! I think the flush should go into GrContext::copyTexture,
> > > though. We should probably have a comment similar to the one in
> > > resolveRenderTarget, just above copyTexture in GrContext.cpp, indicating
> that
> > we
> > > aren't tracking whether there are pending writes to the src texture.
> > 
> > BTW this comment was from me (accidentally posted from personal account).
> 
> Thanks for your review, Brian.  I made the changes you suggested.
> 
> It would be interesting if we could selectively flush to the given texture,
> instead of flushing the whole pipeline.

Agreed. We will have to do that at some point.

LGTM