function template specialization ambiguity

[Rafal Dabrowa]

Consider the following code:
template <class T>
void f(T, int) {}
template <class T>
void f(int, T) {}
template<>
void f(int, int) {} // ambiguous ?

My compiler complains that the last definition is an "ambiguous template
specialization `f<>' for `void f(int, int)' ". I have tested this on a
few compilers and all they complain there. What is wrong in this
specialization ? Consider similar code:
template <class T, class C>
void f(int, T, C) {}
template <class T, class C>
void f(T, int, C) {}
template<class C>
void f(int, int, C) {}

Compiler does not complain that 3rd function is ambiguous. Why ? What is
the difference ?


Rafal

 

[Rob Williscroft]

Rafal Dabrowa wrote in in comp.lang.c++:

Consider the following code:
template <class T>
void f(T, int) {}
template <class T>
void f(int, T) {}
template<>
void f(int, int) {} // ambiguous ?

My compiler complains that the last definition is an "ambiguous
template
specialization `f<>' for `void f(int, int)' ". I have tested this on a
few compilers and all they complain there. What is wrong in this
specialization ? Consider similar code:
template <class T, class C>
void f(int, T, C) {}
template <class T, class C>
void f(T, int, C) {}
template<class C>
void f(int, int, C) {}

Compiler does not complain that 3rd function is ambiguous. Why ? What
is the difference ?

The diffrence is that in the second example you aren't
specializing anything, you are just providing an overload
that happens to be a template.

In the first example the explicit specialization f<>(int, int)
could be a specialization of either of the two overloads that
preceded it, hence the ambiguity.

HTH.



Rob.