Better handling of DocumentOrderedMap same-ID lookups during tree removals

third_party/WebKit/Source/core/dom/DocumentOrderedMap.h

Index: third_party/WebKit/Source/core/dom/DocumentOrderedMap.h
diff --git a/third_party/WebKit/Source/core/dom/DocumentOrderedMap.h b/third_party/WebKit/Source/core/dom/DocumentOrderedMap.h
index e4c311633288140eae0884d907f5798a293950a7..5a6f7dc92d0283b23606b2183fe4c0126c322bb7 100644
--- a/third_party/WebKit/Source/core/dom/DocumentOrderedMap.h
+++ b/third_party/WebKit/Source/core/dom/DocumentOrderedMap.h
@@ -62,7 +62,13 @@ public:
 
     DECLARE_TRACE();
 
+#if ENABLE(ASSERT)
+    void willBeRemovingId(const AtomicString*);

[esprehn, 2015/12/18 00:42:26]

willRemoveId()

no "Be"

[sof, 2015/12/18 07:31:53]

Renamed to willRemoveId().

+#endif
+
 private:
+    DocumentOrderedMap();

[esprehn, 2015/12/18 00:42:26]

we should declare the destructor too

[sof, 2015/12/18 07:31:53]

Done, but see below.

+
     template<bool keyMatches(const AtomicString&, const Element&)>
     Element* get(const AtomicString&, const TreeScope*) const;
 
@@ -84,6 +90,9 @@ private:
     using Map = WillBeHeapHashMap<AtomicString, OwnPtrWillBeMember<MapEntry>>;
 
     mutable Map m_map;
+#if ENABLE(ASSERT)
+    const AtomicString* m_removingId;

[tkent, 2015/12/18 00:30:57]

Can you make this |AtomicString| and check AtomicString::isNull() instead of
nullptr?

[esprehn, 2015/12/18 00:42:26]

+1

[sof, 2015/12/18 07:31:53]

Done that; slight downside with such a design is it leaves you with a
non-trivial finalizer on the Oilpan side.

+#endif
 };
 
 inline bool DocumentOrderedMap::contains(const AtomicString& id) const