From: Ross Bencina (rossb_at_[hidden])
Date: 2001-06-29 16:04:46
Douglas Gregor <gregod_at_[hidden]> wrote:
> Borland C++ seems to have a bug regarding template argument deduction when
> dealing with function pointers. Consider this:
>
> template<typename T> void set(const T& t);
> int foo(int, int) {};
> // ...
> set(&foo);
I don't have enough context to be sure the following is the problem, but in
this case:
struct x{
int foo(int,int) {};
void z(){ set( &foo ); }
};
with Borland extensions on, Borland C++ treats &foo as a bound "closure__"
to this->foo. You need to say set( &x::foo ) to get a straight member
function pointer. I've never been sure how ANSI compliant that was..
De-lurkingly yours,
Ross.