Quote:Original post by Aiwendil
Thanks for the replies, that makes sense now.
One last thing. With
T::A *aObj
and
typename T::A* a6;
Would there be an error if T did not contain an A?
Of course.
Their example about using class instead of typename is one of the more interesting cases, and one that can cause confusion.
The first example says to take the value of A within the scope T, and multiply it by aObj. It is a subtle error, and generally not what the programmer meant. The code will work if you meant to do this:
class T {
public:
const int A = 4;
};
If A is not a value, the code should not compile.
The second says to create a pointer of T::A, which must be a type.
class T {
public:
struct A { ... };
};
If A is not a type, the code should not compile.