Targeting the NULL EGLImage causes slowness on the Geeksphone Peak,
and assumedly, other "Adreno (TM) 205" devices. Achieve the same
effect by deleting the GL texture instead.
ShadowImageLayerOGL::RenderLayer used to call gl()->GetSharedHandleDetails()
directly and uses that result.
This makes us do something similar. As a concequence we also drop the
mTextureTransform member because we are not caching it anymore.
I do wonder if it would make more sense for the content thread to send
the transform along with the update to make sure that it always matches.
This would also make it so that we have no chance of causing a GC
in the compositor to get the transform by calling through java.
ShadowImageLayerOGL::RenderLayer used to call gl()->GetSharedHandleDetails()
directly and uses that result.
This makes us do something similar. As a concequence we also drop the
mTextureTransform member because we are not caching it anymore.
I do wonder if it would make more sense for the content thread to send
the transform along with the update to make sure that it always matches.
This would also make it so that we have no chance of causing a GC
in the compositor to get the transform by calling through java.
There were a number of issues with the current implementation. It couldn't deal
with NULL mSharedHandles. Additionally it was binding things to the wrong
samplers. Finally it was not properly setting the texture transform (which
defaults to all 0). Finally in a debug build an assert would be hit upon client
shutdown due to it not clearing mDescriptor.
This patch fixes all of those issues which makes flash work again.
