Fix FlatTreeTraversal for a slot in a document tree

third_party/WebKit/Source/core/dom/shadow/FlatTreeTraversal.cpp

Index: third_party/WebKit/Source/core/dom/shadow/FlatTreeTraversal.cpp
diff --git a/third_party/WebKit/Source/core/dom/shadow/FlatTreeTraversal.cpp b/third_party/WebKit/Source/core/dom/shadow/FlatTreeTraversal.cpp
index de750fbb3a57b4ec1a9ca5ee39176e8f98de7165..7c81459e95f3ffcdedf39a13f42b66b0f4c1f5f8 100644
--- a/third_party/WebKit/Source/core/dom/shadow/FlatTreeTraversal.cpp
+++ b/third_party/WebKit/Source/core/dom/shadow/FlatTreeTraversal.cpp
@@ -68,6 +68,10 @@ Node* FlatTreeTraversal::resolveDistributionStartingAt(
                       : sibling->previousSibling())) {
     if (isHTMLSlotElement(*sibling)) {
       const HTMLSlotElement& slot = toHTMLSlotElement(*sibling);
+      if (slot.isInDocumentTree()) {
+        const_cast<HTMLSlotElement&>(slot)

[yosin_UTC9, 2016/10/14 01:44:40]

Can we integrate collecting distributed nodes into update distribution phase?

[hayato, 2016/10/14 03:05:35]

Yeah, that is one of TODO items. That could not be done easily.
You can find TODO description in this file. :)

+            .updateDistributedNodesInDocumentTree();

[yosin_UTC9, 2016/10/14 01:44:40]

It is better that using |collectDistributionNodesInDocumentTree()| or another
which implies populating |m_distributedNoes| instead of
|updateDistributedNodesInDocumentTree()|, since |updateXXX()| looks like
updating flat tree and term "update" is too generic(?).
Or, |updateDistributedNodeCache()| to make using |const_cast| to feel sane?

[hayato, 2016/10/14 03:05:35]

I think using *update* is consistent within this file.

+      }
       if (Node* found = (direction == TraversalDirectionForward
                              ? slot.firstDistributedNode()
                              : slot.lastDistributedNode()))