Re: [libstdc++ patch] TR1 result_of implementation

[Paolo Carlini <pcarlini at suse dot de>]

Douglas Gregor wrote:

> This patch adds support for result_of (section 3.4 of TR1). This is the 
> library-only solution that does not make sure of typeof. If we get a 
> reference-preserving typeof (or can fake it), we should consider replacing 
> this implementation with one that uses the reference-preserving typeof 
> directly.
>  
Yes, this issue - reference-preserving typeof - already surfaced a bit 
in the past, but I'm a little confused about the expected developments: 
is there someone actually working on it?!? Gaby should know better...

> All function object tests pass on i686-pc-linux-gnu; full test suite is 
> running right. Okay to commit if everything passes?
>  
You have my personal ok (of course wait 1-2 days for the other 
maintainers), modulo a curiosity:

> +   template<typename _Tp>
> +     class _Has_result_type_helper
> +     {
> +       template<typename _Up>
> +       struct _Wrap_type
> +       { };
> + 
> +       template<typename _Up>
> +         static __sfinae_types::__one
> +         __test(_Wrap_type<typename _Up::result_type>*);
> + 
> +       template<typename _Up>
> +         static __sfinae_types::__two __test(...);
> + 
> +     public:
> +       static const bool __value = sizeof(__test<_Tp>(0)) == 1;
> +     };
>  
Is there something wrong with deriving from __sfinae_types, as I'm 
usually doing?!?

Also, watch-out for overlong lines, maybe you have got a couple...

Paolo.