cc : Remove parent for mask and replica layers in LayerImpl

cc/trees/damage_tracker_unittest.cc

Index: cc/trees/damage_tracker_unittest.cc
diff --git a/cc/trees/damage_tracker_unittest.cc b/cc/trees/damage_tracker_unittest.cc
index 743631f2bca1c67b190a74c7cc34cb1979a58048..65775beb9df19fecf97a150e20b4a6b540d75877 100644
--- a/cc/trees/damage_tracker_unittest.cc
+++ b/cc/trees/damage_tracker_unittest.cc
@@ -1392,6 +1392,7 @@ TEST_F(DamageTrackerTest, VerifyDamageForReplicaMaskWithTransformOrigin) {
     gfx::Transform reflection;
     reflection.Scale3d(-1.0, 1.0, 1.0);
     grand_child1_replica->SetTransform(reflection);
+    grand_child1_replica->SetParent(grand_child1);

[ajuma, 2016/03/30 21:56:05]

Why do tests still need to set the parent of mask and replica layers?

[jaydasika, 2016/03/30 21:59:32]

We don't need parent for masks but need it for replica's as we build property
trees for replicas. (I mean replicas are part of the recursion of building
property trees like other children).

[ajuma, 2016/03/30 22:02:52]

Oh, ok. Do we have other tests that construct replicas on the main thread rather
than on LayerImpl? (Just want to make sure we have test coverage for replica
LayerImpls that aren't children of any layer.)

[jaydasika, 2016/03/30 22:10:08]

There are many tests. One random example :
LayerTreeHostMasksPixelTest.MaskWithReplicaOfClippedLayer

[ajuma, 2016/03/30 23:00:09]

Thanks for checking. Please add a comment here (and in the other tests) that
this is only needed for property tree building.

[jaydasika, 2016/03/30 23:29:46]

Done.

     grand_child1->SetReplicaLayer(std::move(grand_child1_replica));
     grand_child1->SetForceRenderSurface(true);
   }