Fix WritePixels test on ANGLE.

Fix WritePixels test on ANGLE.

Don't upload BGRA to RGBA if not supported (ES2 w/ EXT BGRA extension).

R=senorblanco@chromium.org
Committed: https://code.google.com/p/skia/source/detail?r=7622

[bsalomon, 2013-02-05 21:07:45]

The GpuBitmapCopyTest is now passing on ANGLE too (either due to this change or
Leon's change to the test).

[Stephen White, 2013-02-05 21:28:08]

https://codereview.appspot.com/7305046/diff/4001/src/gpu/GrContext.cpp
File src/gpu/GrContext.cpp (right):

https://codereview.appspot.com/7305046/diff/4001/src/gpu/GrContext.cpp#newcod...
src/gpu/GrContext.cpp:1335: GrPixelConfig preferredConfig = config;
Kind of a nit, but I find it a little confusing to be referring to
preferredConfig below, even when we don't have one.  Could we just do:

config = preferredConfig;

inside the if-clause, then use config to build the desc, rather than using
preferredConfig?  Do we need to remember what the unswapped config was?  Maybe
we do, for the scratch-failed case at the end of this function.  If so, could we
do:

GrPixelConfig readConfig = config, preferredConfig;
bool swapRAndB = false;
if (...) {
   readConfig = preferredConfig;
}

https://codereview.appspot.com/7305046/diff/4001/tests/GpuBitmapCopyTest.cpp
File tests/GpuBitmapCopyTest.cpp (right):

https://codereview.appspot.com/7305046/diff/4001/tests/GpuBitmapCopyTest.cpp#...
tests/GpuBitmapCopyTest.cpp:111: //        if (type ==
GrContextFactory::kANGLE_GLContextType) {
Please remove commented-out code, if necessary.

[bsalomon, 2013-02-05 21:45:57]

https://codereview.appspot.com/7305046/diff/4001/src/gpu/GrContext.cpp
File src/gpu/GrContext.cpp (right):

https://codereview.appspot.com/7305046/diff/4001/src/gpu/GrContext.cpp#newcod...
src/gpu/GrContext.cpp:1335: GrPixelConfig preferredConfig = config;
On 2013/02/05 21:28:08, Stephen White wrote:
> Kind of a nit, but I find it a little confusing to be referring to
> preferredConfig below, even when we don't have one.  Could we just do:
> 
> config = preferredConfig;
> 
> inside the if-clause, then use config to build the desc, rather than using
> preferredConfig?  Do we need to remember what the unswapped config was?  Maybe
> we do, for the scratch-failed case at the end of this function.  If so, could
we
> do:
> 
> GrPixelConfig readConfig = config, preferredConfig;
> bool swapRAndB = false;
> if (...) {
>    readConfig = preferredConfig;
> }

Maybe the new patch is clearer? I renamed the param and the preferred config.

https://codereview.appspot.com/7305046/diff/4001/tests/GpuBitmapCopyTest.cpp
File tests/GpuBitmapCopyTest.cpp (right):

https://codereview.appspot.com/7305046/diff/4001/tests/GpuBitmapCopyTest.cpp#...
tests/GpuBitmapCopyTest.cpp:111: //        if (type ==
GrContextFactory::kANGLE_GLContextType) {
On 2013/02/05 21:28:08, Stephen White wrote:
> Please remove commented-out code, if necessary.

Oops, forgot to save before upload

[bsalomon, 2013-02-05 22:26:17]

New patch keeps the old preferred config functions. We ignore their result
unless it is an r/b swap of the original config. In that case we do a double
swap in order to use the preferred config: once during the scratch read/write
and again when copying the RT to/from the scratch.

[Stephen White, 2013-02-05 22:31:53]

LGTM