cc: The RenderSurface contentRect can have negative position.

cc/render_surface_impl.cc

Index: cc/render_surface_impl.cc
diff --git a/cc/render_surface_impl.cc b/cc/render_surface_impl.cc
index c7a346b41e3c143321ec37617229d0e8014f715d..4278a2dd593c1fa6bc4f9cdfe38f66cfb31279f2 100644
--- a/cc/render_surface_impl.cc
+++ b/cc/render_surface_impl.cc
@@ -209,7 +209,8 @@ void RenderSurfaceImpl::appendQuads(QuadSink& quadSink, AppendQuadsData& appendQ
             owningLayerDrawScale.y());
         // This assumes that the owning layer clips its subtree when a mask is
         // present.
-        DCHECK(gfx::RectF(unclippedSurfaceSize).Contains(contentRect()));
+        DCHECK(gfx::RectF(unclippedSurfaceSize).Contains(
+            gfx::Rect(contentRect().size())));

[shawnsingh, 2013/02/04 21:51:08]

Thanks for catching this.  I think it might be a stricter (better) check to
include positions in this DCHECK too.  Is there some reason we shouldnt?

i.e.   DCHECK(gfx::RectF(unclippedSurfacePos,
unclippedSurfaceSize).Contains(contentRect())));

[danakj, 2013/02/04 21:52:15]

I just didn't think we have the position of the unclipped surface here. Do we?

[shawnsingh, 2013/02/04 21:59:24]

Oh, I see, you're right.

Also, thinking a second time - your change looks fine, but isn't it true that we
pin the origin of the surface to the owning layer, ** in the case where the
layer does have a mask / masksToBounds ?

So what is the case where a contentRect on a surface with mask would have
actually triggered the original DCHECK incorrectly?

 
         float uvScaleX = contentRect().width() / unclippedSurfaceSize.width();
         float uvScaleY = contentRect().height() / unclippedSurfaceSize.height();