Allow creation of ImageBitmaps from WebGL canvases.

Source/core/page/ImageBitmap.cpp

Index: Source/core/page/ImageBitmap.cpp
diff --git a/Source/core/page/ImageBitmap.cpp b/Source/core/page/ImageBitmap.cpp
index 2f83dd31d15ef34c1697d544f39b783227b699fb..f748a3a08008e1a5e6fb23d889dbc74f9ad6438b 100644
--- a/Source/core/page/ImageBitmap.cpp
+++ b/Source/core/page/ImageBitmap.cpp
@@ -9,6 +9,7 @@
 #include "core/html/HTMLImageElement.h"
 #include "core/html/HTMLVideoElement.h"
 #include "core/html/ImageData.h"
+#include "core/html/canvas/CanvasRenderingContext.h"
 #include "core/platform/graphics/BitmapImage.h"
 #include "core/platform/graphics/GraphicsContext.h"
 #include "wtf/RefPtr.h"
@@ -71,6 +72,10 @@ ImageBitmap::ImageBitmap(HTMLCanvasElement* canvas, const IntRect& cropRect)
     IntRect srcRect = intersection(cropRect, IntRect(IntPoint(), canvasSize));
     IntRect dstRect(IntPoint(), srcRect.size());
 
+    CanvasRenderingContext* sourceContext = canvas->renderingContext();
+    if (!sourceContext || !sourceContext->isAccelerated() || !sourceContext->is2d())

[Justin Novosad, 2013/08/19 20:20:51]

Do we really need to call makeRenderingResultsAvailable() in all these cases? 
If it is just needed for WebGL, shouldn't the condition be:
if (sourceContext && !sourceContext->is2d())
If the other conditions are necessary, what use cases do they fix?  Do we have
tests for those use cases?

+        canvas->makeRenderingResultsAvailable();
+
     m_buffer = ImageBuffer::create(canvasSize);
     m_buffer->context()->drawImageBuffer(canvas->buffer(), dstRect, srcRect);
     m_bitmap = m_buffer->copyImage(DontCopyBackingStore);