gpu: Initial implementation of CommitOverlayPlanes

ui/gl/gl_surface_egl.cc

Index: ui/gl/gl_surface_egl.cc
diff --git a/ui/gl/gl_surface_egl.cc b/ui/gl/gl_surface_egl.cc
index 7c36984843a7eadbf12da2bf7c3ec969542a492e..0da838d0789aa1bcc285553890c4d01f9924fadd 100644
--- a/ui/gl/gl_surface_egl.cc
+++ b/ui/gl/gl_surface_egl.cc
@@ -731,6 +731,14 @@ gfx::SwapResult NativeViewGLSurfaceEGL::PostSubBuffer(int x,
   return gfx::SwapResult::SWAP_ACK;
 }
 
+bool NativeViewGLSurfaceEGL::SupportsCommitOverlayPlanes() {
+  return true;

[piman, 2015/11/25 06:49:28]

This is not right for GLSurfaceOzoneEGL, which inherits from
NativeViewGLSurfaceEGL.
Did you mean this only for Android?

[watk, 2015/11/25 21:10:13]

Oops, I didn't really mean to include this change. I removed it to keep this
change focused on the plumbing. I'll do the Android specific stuff in a follow
up. 

+}
+
+gfx::SwapResult NativeViewGLSurfaceEGL::CommitOverlayPlanes() {
+  return gfx::SwapResult::SWAP_ACK;

[piman, 2015/11/25 06:49:28]

It feels like the planes should be scheduled only at this point - i.e.
ScheduleOverlayPlanes would keep pending changes in a list, and only here commit
the list. Even if not atomic, at least it would follow the same scheduling logic
as the rest.

[watk, 2015/11/25 21:10:13]

Yeah, I can do that (removed for this CL). I think I'll pull
GLSurfaceOzoneSurfaceless::Overlay into a new file like
ui/gl/gl_surface_overlay.h and basically do what GLSurfaceOzoneSurfaceless where
it keeps a vector of overlays. It also looks like it's getting time to subclass
GLSurfaceEGL for android at that point?

+}
+
 VSyncProvider* NativeViewGLSurfaceEGL::GetVSyncProvider() {
   return vsync_provider_.get();
 }