More switching to standard type traits

third_party/WebKit/Source/platform/heap/ThreadingTraits.h

Index: third_party/WebKit/Source/platform/heap/ThreadingTraits.h
diff --git a/third_party/WebKit/Source/platform/heap/ThreadingTraits.h b/third_party/WebKit/Source/platform/heap/ThreadingTraits.h
index fef13ef6931856b060354f4f61ca2515afcb6d23..2657fefed7eed980200b5321c9ea8dda44f92103 100644
--- a/third_party/WebKit/Source/platform/heap/ThreadingTraits.h
+++ b/third_party/WebKit/Source/platform/heap/ThreadingTraits.h
@@ -35,8 +35,8 @@ class Node;
 class NodeList;
 
 template<typename T,
-    bool mainThreadOnly = WTF::IsSubclass<typename std::remove_const<T>::type, Node>::value
-        || WTF::IsSubclass<typename std::remove_const<T>::type, NodeList>::value> struct DefaultThreadingTrait;
+    bool mainThreadOnly = std::is_convertible<typename std::remove_const<T>::type*, Node*>::value
+        || std::is_convertible<typename std::remove_const<T>::type*, NodeList*>::value> struct DefaultThreadingTrait;

[Nico, 2015/12/18 14:14:14]

Now that I think about it, this is a risky change, right? It's a good thing that
is_base_of requires the types to be complete (unless they're the same).
Consider:

#include <type_traits>
class A; class B;
static_assert(std::is_base_of<A, B>::value, "");

This doesn't build, it complains that A and B aren't complete.

Now consider:
#include <type_traits>
class A; class B;
static_assert(std::is_convertible<B*, A*>::value, "");


This builds, and the static_assert fails. Now we add the definitions:

#include <type_traits>
class A; class B;
class A {}; class B : public A {};
static_assert(std::is_convertible<B*, A*>::value, "");

This builds, and the static_assert doesn't fail! This means is_convertible will
have different results based on if classes are defined or not (which is
unpredictable). This is actually fairly worrying. Maybe we shouldn't be using
std::is_convertible much...

 
 template<typename T>
 struct DefaultThreadingTrait<T, false> {