Annoying Overloading ambiguity

Question

If the following code is compiled with Studio 12u1 or express 0610, the following compilation error is reported:
CC /tmp/bug.cc
"/tmp/bug.cc", line 32: Error: Overloading ambiguity between "built-in operator[](int, const char*)" and "Object::ValueProxy::operator[](const std::string &)".
gcc is happy with the code and I think it is valid (what is "built-in operator[](int, const char*)"?).
Removing the template operator T() gets rid of the error.
#include <string>
struct Value
struct BaseHolder {};
template <typename T> struct Holder : BaseHolder {};
template <typename T> operator const T() const; // <-- this line
Value operator[]( unsigned n );
Value( const Value& ) {}
struct Object
struct ValueProxy : Value
operator Value();
ValueProxy operator[](const std::string& );
ValueProxy operator[](const std::string& s);
int main()
Object o;
o["a"]["b"][0];
}

Answer 1

1. Please always put code examples, scripts, and compiler messages in "code" brackets, to ensure they are displayed correctly. In particular, some punctuation characters are otherwise taken as formatting directives and not displayed.
2. Class Value has a user-defined copy constructor, so the compiler will not generate a the default no-argument constructor.
The fragment struct ValueProxy : Value { ... }is not valid because no constructor for Value is available. (CC does not catch this error, but gcc does.) I removed the copy constructor to avoid this error.
3. The operator Value() conversion function in class ValueProxy would convert a ValueProxy to its base class, which is an implicit conversion. The conversion function will never be called for an explicit or implicit conversion. Perhaps this code does not work as you expect. (You could call the function explicitly by name, but then you might wind up with two different results for a conversion.)
4. I think the last line really is ambiguous, as indicated by CC. I think gcc is wrong to accept the code.