mangling of member template constructors

[Alain Miniussi]

Alex Samuel wrote:
> 
> Folks,
> 
> Here's a mangling corner case that needs special treatment and should
> be added to the spec.  If you have a member template constructor,
> currently we mangle its return type (since return types of all
> template functions are mangled).  Since it's a constructor, it should
> have a special exemption, and not have a return type encoded.
> 
> For instance
> 
>     class K {
>     public:
>         template<typename T> K (T* p) {}
>     };
> 
>     template K::K<int> (int* p);
> 
> should be mangled _ZN1KC2IiEEPT_ rather than _ZN1KC2IiEEvPT_ (or
> something else, depending on what you consider its `return type' to
> be).

What would be the benefit ? (I am just concerned by having to handle yet
another specific case if I don;t see any advantage to it).

Regards,

Alain