Fixed leak of render target associated with ColorBuffer.

Fixed leak of render target associated with ColorBuffer.

Following r853, I am assuming that any render target returned by FrameBuffer::getRenderTarget has to be released. Updating the usage of getRenderTarget to reflect that.

This is not reproducible in Chrome at the moment but it will be once I start rendering to pbuffers.

I ran the WebGL conformance tests. No regressions or crashes.
Committed: http://code.google.com/p/angleproject/source/detail?r=903

[apatrick1, 2011-12-07 03:27:15]

Hi Daniel,

I think I found some leaks from r853.

Thanks,

Al

[nicolas, 2011-12-07 16:59:54]

Nice catch. This may actually explain a failure to reset the device that I've
observed recently...

The patch looks good to me. Tiny nitpick: there's no need to test for "if
(renderTarget)" in readPixels.