[Chromium] WebGL active texture state corrupted upon WebGL contents present

DrawingBuffer::publishToPlatformLayer is corrupting the WebGL active texture state.  The bound texture must be restored after the texture copy.

Webkit issue tracking crbug.com/105045.

Created an attachment (id=116387) [details]
Patch

(From update of attachment 116387 [details])
View in context: https://bugs.webkit.org/attachment.cgi?id=116387&action=review

> Source/WebCore/platform/graphics/chromium/DrawingBufferChromium.cpp:66
> +    , m_activeTextureUnit(0)

The default value for m_activeTextureUnit should probably be GraphicsContext3D::GL_TEXTURE0, not 0, just in case the WebGL app doesn't call glActiveTexture() before DrawingBuffer::publishToPlatformLayer() gets called.

> Source/WebCore/platform/graphics/gpu/mac/DrawingBufferMac.mm:47
> +    , m_activeTextureUnit(0)

Same as above.

> Source/WebCore/platform/graphics/gpu/qt/DrawingBufferQt.cpp:42
> +    , m_activeTextureUnit(0)

Same as above.

> Source/WebCore/platform/graphics/gtk/DrawingBufferGtk.cpp:44
> +    , m_activeTextureUnit(0)

Same as above.

(From update of attachment 116387 [details])
Nice work tracking this down. Looks like it needs to be merged to top of tree, but r=me with that addressed.

BTW, possible to write a layout test for this?

Created an attachment (id=116400) [details]
Patch

(From update of attachment 116400 [details])
Attachment 116400 [details] did not pass qt-ews (qt):
Output: http://queues.webkit.org/results/10635041

(From update of attachment 116400 [details])
Attachment 116400 [details] did not pass gtk-ews (gtk):
Output: http://queues.webkit.org/results/10636045

(From update of attachment 116400 [details])
Attachment 116400 [details] did not pass chromium-ews (chromium-xvfb):
Output: http://queues.webkit.org/results/10632048

Created an attachment (id=116422) [details]
Patch

Created an attachment (id=116467) [details]
Patch

(From update of attachment 116467 [details])
View in context: https://bugs.webkit.org/attachment.cgi?id=116467&action=review

Looks good; nice job on the test case. A few minor nits in the test that can be cleaned up later. r=me

> LayoutTests/fast/canvas/webgl/webgl-texture-binding-preserved.html:22
> +var requestAnimationFrame;

nit: this variable is no longer needed.

> LayoutTests/fast/canvas/webgl/webgl-texture-binding-preserved.html:26
> +  // before reaching this point.  If the texture state was corruted, then 

nit: corruted -> corrupted

> LayoutTests/fast/canvas/webgl/webgl-texture-binding-preserved.html:42
> +  canvas.width = 50; canvas.height = 50;

nit: not needed

> LayoutTests/fast/canvas/webgl/webgl-texture-binding-preserved.html:63
> +  setTimeout(draw, 100);

nit: add comment that requestAnimationFrame was tried for this purpose and it didn't work. Also that the 100 ms timeout seemed to reproduce the problem 100% of the time even during fast page reloads.

(From update of attachment 116467 [details])
Clearing flags on attachment: 116467

Committed r101120: <http://trac.webkit.org/changeset/101120>

All reviewed patches have been landed.  Closing bug.

The test added by this patch is failing on Mac:
http://build.webkit.org/results/SnowLeopard%20Intel%20Release%20(Tests)/r101177%20(34952)/fast/canvas/webgl/webgl-texture-binding-preserved-pretty-diff.html

Created an attachment (id=116638) [details]
Patch

(In reply to comment #14)
> The test added by this patch is failing on Mac:
> http://build.webkit.org/results/SnowLeopard%20Intel%20Release%20(Tests)/r101177%20(34952)/fast/canvas/webgl/webgl-texture-binding-preserved-pretty-diff.html

Investigating the failure.  This test should pass on all browsers, and does not rely on Chrome-specific behaviour.

(From update of attachment 116638 [details])
OK.  r=me

(In reply to comment #14)
> The test added by this patch is failing on Mac:
> http://build.webkit.org/results/SnowLeopard%20Intel%20Release%20(Tests)/r101177%20(34952)/fast/canvas/webgl/webgl-texture-binding-preserved-pretty-diff.html

The test is failing because safari is failing to construct the shaders used by the test.  Added a suppression, and a new bug here:  https://bugs.webkit.org/show_bug.cgi?id=73148

Re-opening the bug, to allow the CQ to process pending patch #5.

(From update of attachment 116638 [details])
Clearing flags on attachment: 116638

Committed r101696: <http://trac.webkit.org/changeset/101696>

All reviewed patches have been landed.  Closing bug.