Re-land: cc: Make Overlay unit tests work when sync query extension is available.

cc/output/gl_renderer.cc

Index: cc/output/gl_renderer.cc
diff --git a/cc/output/gl_renderer.cc b/cc/output/gl_renderer.cc
index a302f7e783f1027647364793efbb580468b6473c..2b298e19e258cfae0edd9ebdb3d85962a2b411e4 100644
--- a/cc/output/gl_renderer.cc
+++ b/cc/output/gl_renderer.cc
@@ -466,9 +466,6 @@ static ResourceProvider::ResourceId WaitOnResourceSyncPoints(
 }
 
 void GLRenderer::BeginDrawingFrame(DrawingFrame* frame) {
-  if (frame->device_viewport_rect.IsEmpty())
-    return;
-
   TRACE_EVENT0("cc", "GLRenderer::BeginDrawingFrame");
 
   scoped_refptr<ResourceProvider::Fence> read_lock_fence;
@@ -499,6 +496,9 @@ void GLRenderer::BeginDrawingFrame(DrawingFrame* frame) {
   }
   resource_provider_->SetReadLockFence(read_lock_fence.get());
 
+  if (frame->device_viewport_rect.IsEmpty())

[danakj, 2014/10/29 15:18:26]

If there's an empty viewport can you not just make FinishDrawingFrame early out
and not do a bunch of work here?

[reveman, 2014/10/29 16:14:52]

Maybe. It would assume that we're not performing any operations that require a
read lock fence in these cases. It's not clear from this code if that is safe.

Alternatively, we could just remove this check and leave it up to the caller.

[danakj, 2014/10/29 17:04:38]

ie make DirectRenderer check this? That sounds good.

+    return;
+
   // Insert WaitSyncPointCHROMIUM on quad resources prior to drawing the frame,
   // so that drawing can proceed without GL context switching interruptions.
   DrawQuad::ResourceIteratorCallback wait_on_resource_syncpoints_callback =