move texture flush and param update to separate function

[bsalomon, 2012-06-01 20:03:30]

The GL state flush was broken into GrGpuGL's portion and the subclass's portion.
This is the first change to undo that.

[TomH, 2012-06-01 21:00:25]

LGTM

http://codereview.appspot.com/6266044/diff/1/src/gpu/gl/GrGpuGL_program.cpp
File src/gpu/gl/GrGpuGL_program.cpp (right):

http://codereview.appspot.com/6266044/diff/1/src/gpu/gl/GrGpuGL_program.cpp#n...
src/gpu/gl/GrGpuGL_program.cpp:508: this->flushRenderTarget(rect);
How did we get by without this before?

[bsalomon, 2012-06-01 21:04:28]

http://codereview.appspot.com/6266044/diff/1/src/gpu/gl/GrGpuGL_program.cpp
File src/gpu/gl/GrGpuGL_program.cpp (right):

http://codereview.appspot.com/6266044/diff/1/src/gpu/gl/GrGpuGL_program.cpp#n...
src/gpu/gl/GrGpuGL_program.cpp:508: this->flushRenderTarget(rect);
On 2012/06/01 21:00:25, TomH wrote:
> How did we get by without this before?

They were done in the correct order before as well. Because I moved
flushBoundTextureAndParam into the loop in the above I had to move the RT flush
below the loop.

[bsalomon, 2012-06-04 13:17:22]

Closed, r4134.

[bsalomon, 2012-06-04 17:58:13]

When this was landed the first time it broke stuff. The problem was that we
relied on flushRenderTarget() to invalidate the programs' cached view matrices
when the view port changes. However, I had moved the call the
flushRenderTarget() below flushViewMatrix(). I decoupled these by having the
each program remember the last viewport size it saw.

[TomH, 2012-06-04 19:00:37]

ReLGTM

[bsalomon, 2012-06-04 19:28:08]

Relanded as r4142