Oilpan: Remove guardRef and guardDeref from TreeScope.

Source/core/html/shadow/DateTimeEditElement.cpp

Index: Source/core/html/shadow/DateTimeEditElement.cpp
diff --git a/Source/core/html/shadow/DateTimeEditElement.cpp b/Source/core/html/shadow/DateTimeEditElement.cpp
index 28a7b63ddb66441ddc4267a14d170ebd49c729d0..11dde15fc88d6923f1566ee06496c48eb029b959 100644
--- a/Source/core/html/shadow/DateTimeEditElement.cpp
+++ b/Source/core/html/shadow/DateTimeEditElement.cpp
@@ -448,8 +448,10 @@ DateTimeEditElement::DateTimeEditElement(Document& document, EditControlOwner& e
 
 DateTimeEditElement::~DateTimeEditElement()
 {
+#if !ENABLE(OILPAN)
     for (size_t fieldIndex = 0; fieldIndex < m_fields.size(); ++fieldIndex)
         m_fields[fieldIndex]->removeEventHandler();

[haraken, 2014/04/25 05:21:42]

Just help me understand: Even though we have not yet made Node::m_parent a
Member, why is it safe to remove this code? Since Node::m_parent is still a raw
pointer, it's possible that the DateTimeEditElement dies earlier the
DateTiemFieldElements. Then the back pointer from the DateTiemFieldElements to
the DateTimeEditElement will become stale pointers.

[Mads Ager (chromium), 2014/04/25 10:58:25]

This is in a shadow tree and is not accessible from the outside. Therefore, it
would seem to me that this is all dead when the shadow tree owner is dead. That
at least will be the case when we trace in all directions. Therefore, I would
prefer to not spent more time convincing myself that this is true right now.
Instead, I will focus on implementing the full tracing for the TreeShared
hierarchies which will ensure that this is correct.

+#endif
 }
 
 inline Element* DateTimeEditElement::fieldsWrapperElement() const