cc: Move more functionality from host to LayerTreeImpl

cc/delegated_renderer_layer_impl_unittest.cc

Index: cc/delegated_renderer_layer_impl_unittest.cc
diff --git a/cc/delegated_renderer_layer_impl_unittest.cc b/cc/delegated_renderer_layer_impl_unittest.cc
index a125d8626cc8bdd5d5a080b2b80e8f4d36e37a5e..6c0b1683a025e44aab1a0dba0a4cc6c2d0f58b36 100644
--- a/cc/delegated_renderer_layer_impl_unittest.cc
+++ b/cc/delegated_renderer_layer_impl_unittest.cc
@@ -6,6 +6,7 @@
 
 #include "cc/append_quads_data.h"
 #include "cc/layer_tree_host_impl.h"
+#include "cc/layer_tree_impl.h"
 #include "cc/quad_sink.h"
 #include "cc/render_pass_draw_quad.h"
 #include "cc/scoped_ptr_vector.h"
@@ -106,7 +107,7 @@ public:
         // Get the RenderPass generated by layerBefore to come before the delegated RenderPasses.
         rootLayer->addChild(layerBefore.Pass());
 
-        m_hostImpl->setRootLayer(rootLayer.Pass());
+        m_hostImpl->activeTree()->SetRootLayer(rootLayer.Pass());

[danakj, 2012/12/31 15:18:38]

IIUC, in general the LTH would never set the root layer directly on the active
tree, right? It set it on the pending tree, then cause a tree swap when ready?


It seems like these tests that were written against the LTHImpl are getting
further from the truth. I should probably make these tests create main thread
objects instead of playing around with impl stuff directly. WDYT?

[enne (OOO), 2012/12/31 23:52:36]

Yeah, with impl-side painting, the LTH would only set the pending tree.  See the
way tree sync works with the syncTree variable in layer_tree_host.cc.

However, in tests, I don't think it's *that* problematic to just "skip right to
the active tree" where you're really just trying to construct a tree to test
with, rather than trying to test the commit flow.

Honestly, I find it more problematic that we need so much infrastructure to test
what passes a layer tree generates.  Having to add a pending tree in that mix
doesn't make that any better.  In other words, I understand where you're coming
from and I don't disagree, but I think I'd rather see us spend time on better
abstractions rather than making things more real by pulling in more boilerplate
to do simple things.

[danakj, 2013/01/01 00:39:22]

Yah okay, fair enough. Hmmm abstractions..

     }
 
 protected:
@@ -422,7 +423,7 @@ public:
 
         rootLayer->addChild(delegatedRendererLayer.PassAs<LayerImpl>());
 
-        m_hostImpl->setRootLayer(rootLayer.Pass());
+        m_hostImpl->activeTree()->SetRootLayer(rootLayer.Pass());
     }
 
 protected: