Add support for converting cc::FilterOperations into an SkImageFilter

cc/trees/layer_tree_host_common.cc

Index: cc/trees/layer_tree_host_common.cc
diff --git a/cc/trees/layer_tree_host_common.cc b/cc/trees/layer_tree_host_common.cc
index 309e910a510cd3bca3432da99bbdad73b1a49d07..26c6973cf2f7691dcdb34d405c27b9cbf75fd64b 100644
--- a/cc/trees/layer_tree_host_common.cc
+++ b/cc/trees/layer_tree_host_common.cc
@@ -555,8 +555,7 @@ static bool SubtreeShouldRenderToSeparateSurface(
   }
 
   // If the layer uses a CSS filter.
-  if (!layer->filters().IsEmpty() || !layer->background_filters().IsEmpty() ||
-      layer->filter()) {
+  if (!layer->filters().IsEmpty() || !layer->background_filters().IsEmpty()) {
     DCHECK(!is_root);
     return true;
   }
@@ -1548,7 +1547,8 @@ static void CalculateDrawPropertiesInternal(
 
     // TODO(senorblanco): make this smarter for the SkImageFilter case (check
     // for pixel-moving filters)
-    if (layer->filters().HasFilterThatMovesPixels() || layer->filter())
+    if (layer->filters().HasReferenceFilter() ||

[Stephen White, 2013/09/11 17:56:25]

It occurs to me that HasFilterThatMovesPixels() should probably return true
(conservatively) when it has a reference filter. Then we could just call that,
and skip the HasReferenceFilter() call here. Could be done in a followup.

That'll make it easier to clean this up once SkImageFilter gets the smarts.

[danakj, 2013/09/11 18:01:16]

That's true, but elsewhere we check that and then use the outsets to tell how
far they moved. And with a reference filter those outsets are wrong. So I'd
prefer to keep it a dcheck as is.

[Stephen White, 2013/09/11 18:04:19]

Ahh, ok. In that case, ignore me.

+        layer->filters().HasFilterThatMovesPixels())
       nearest_ancestor_surface_that_moves_pixels = render_surface;
 
     render_surface->SetNearestAncestorThatMovesPixels(