[Command buffer] workaround the framebuffer completeness bug for Intel Mac

gpu/command_buffer/service/framebuffer_manager.h

Index: gpu/command_buffer/service/framebuffer_manager.h
diff --git a/gpu/command_buffer/service/framebuffer_manager.h b/gpu/command_buffer/service/framebuffer_manager.h
index eab5809d63145562acb0564d46d59c61a454ad40..2bf59f019d15c4c19dd79f58e8b5395552449712 100644
--- a/gpu/command_buffer/service/framebuffer_manager.h
+++ b/gpu/command_buffer/service/framebuffer_manager.h
@@ -147,6 +147,10 @@ class GPU_EXPORT Framebuffer : public base::RefCounted<Framebuffer> {
     return has_been_bound_ && !IsDeleted();
   }
 
+  bool ReadBufferHasNoImage() const;
+  bool DrawBuffersHaveNoImage(GLsizei& num, GLenum* buf) const;

[Zhenyao Mo, 2016/11/14 19:42:24]

1) I think DrawBuffersHaveMissingImage is a better name.  Otherwise it's
confusing because it could be read as "no images are attached to any draw
buffers".

2) "&num": this is against chromium coding style.  Use GLsizei* num instead.

3) Why do we need num? I think we can just get rid of it.

[yunchao, 2016/11/15 16:21:59]

Agreed.
Done.
Mac Intel driver still think it is framebuffer incomplete if 1) we reset
COLOR_ATTACHMENTi to GL_NONE for readBuffer or drawBuffers those are missing
image, 2)After reset, there is not valid read/draw buffer(all of them are
GL_NONE after reset). 
As a result, use parameter num to record the valid drawbuffers number after
reset. Generate INVALID_OPERATION （read） or return directly (draw) in Chromium
if num is 0. Don't go to driver to check framebuffer status for this situation.

+  bool ColorBuffersHaveImage() const;
+
   bool HasColorAttachment(int index) const;
   bool HasDepthAttachment() const;
   bool HasStencilAttachment() const;