[libcxx] [test] Follow LWG 2520 "N4089 broke initializing unique_ptr<T[]> from a nullptr".

[STL_MSFT]

Summary

Follow LWG 2520 "N4089 broke initializing unique_ptr<T[]> from a nullptr".

This library issue permits nullptr, but not 0 or NULL (which can be 0). In null_ctor.pass.cpp, replace 0 with nullptr. In pointer_deleter01.pass.cpp, which is testing both NULL and nullptr, mark the NULL test as libcxx-specific.

[EricWF]

I see how the wording doesn't require supporting the literal '0', but if you provide a concrete 'std::nullptr_t' parameter your going to allow the conversion from '0'. I think this should probably be clarified in the standard. However I'm fine taking this for now.

[EricWF]

r276587.

test/std/utilities/memory/unique.ptr/unique.ptr.runtime/null_ctor.pass.cpp

public:
     void operator()(void*) {++state_;}
 };
 
 int main()
 {
     Deleter d;
     assert(d.state() == 0);
     {
-    std::unique_ptr<int[], Deleter&> p(0, d);
+    std::unique_ptr<int[], Deleter&> p(nullptr, d);
     assert(p.get() == 0);
     assert(&p.get_deleter() == &d);
     }
     assert(d.state() == 0);
 }

test/std/utilities/memory/unique.ptr/unique.ptr.runtime/unique.ptr.runtime.ctor/pointer_deleter01.pass.cpp

int main()
     assert(A::count == 3);
     std::unique_ptr<A[], Deleter<A[]> > s(p, Deleter<A[]>());
     assert(s.get() == p);
     assert(s.get_deleter().state() == 0);
     }
     assert(A::count == 0);
 
     { // LWG#2520 says that nullptr is a valid input as well as null
+#ifdef _LIBCPP_VERSION
     std::unique_ptr<A[], Deleter<A[]> > s1(NULL, Deleter<A[]>());
+#endif
     std::unique_ptr<A[], Deleter<A[]> > s2(nullptr, Deleter<A[]>());
     }
     assert(A::count == 0);
 }