T&& in templated function and class

I've just stumbled upon that T&& in class and function means different things.
In function:

template<class T> void f(T&& t){};   // t is R or L-value
...
int i=0; 
f(i);   // t in f is lvalue
f(42);  // t in f is rvalue  

In class:

template<class T> 
struct S { 
       S(T&& t){}  // t is only R-value? 
};
...
int i;
S<int> ss(i);   // compile error - cannot bind lvalue to ‘int&&’

Does this means that if we haveT&& t in class, than t will be only rvalue?
Can some one point me where I can get more info about this?
Does it means that I need to write two method overloads for L and R-values?

ANSWER
As Alf's example shows, t in function and class can be Lvalue or Rvalue.

In your function T is deduced from the actual argument. The main use for that particular combination is perfect forwarding. In the class template T is not deduced, it must be specified.

E.g., this compiles nicely with both g++ and msvc:

template<class T> 
struct S { 
       S(T&& t){}
};

int main()
{
    int i;
    S< int& > ss(i);
}