Fix for GL_CHROMIUM_flipy in TexSubImage2D

Fix for GL_CHROMIUM_flipy in TexSubImage2D and unit test

TEST=none
BUG=none
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=110923

[bsalomon, 2011-11-17 19:14:14]

Hey guys, I'm not sure if this patch is correct or not. My impression of
GL_CHROMIUM_flipy is that when GL_UNPACK_FLIP_Y is set the only modification to
the behavior of the glTex(Sub)Image functions is that the rows in the ptr param
are unpacked in descending y-order rather than ascending. (xoffset,yoffset) is
still be the lower-left corner of the rectangle written. Is this incorrect? I'm
wondering if GL_UNPACK_FLIP_Y is also supposed to affect the way the rectangle
params to the SubImage functions are interpreted.

[greggman, 2011-11-17 20:11:07]

I'm not sure either. Are you the only user?

You'll need to change the docs in
gpu/GLES2/extensions/CHROMIUM/CHROMIUM_flipy.txt

There's a test for this in
src/gpu/command_buffer/client/gles2_implementation_unittest.cc in
GLESImplementationTest::TexImage2D2Writes

Was the test wrong?

[greggman, 2011-11-17 20:13:34]

Oh, the intent is that you pass the exact same parameters you'd pass to
TexSubImage2D or TexImage2D. The only change is the data is copied flipped. the
meaning of xoffset and yoffset, width, height do not change.

[bsalomon, 2011-11-18 16:51:32]

It doesn't appear that anyone else is using it. I think our understandings of
how it should work agree and there is just a bug in the implementation.

The existing tests didn't catch it because the two that set GL_UNPACK_FLIP_Y
only do a glTexImage2D not glTexSubImage2D. I added a new test in the latest
patch. It passes with the change but fails without.

[greggman, 2011-11-18 17:17:43]

lgtm