Re: Boost.Format final submission

From: rogeeff (rogeeff_at_[hidden])
Date: 2002-02-23 10:25:41

--- In boost_at_y..., "David Abrahams" <david.abrahams_at_r...> wrote:
> 
> ----- Original Message -----
> From: "rogeeff" <rogeeff_at_m...>
> To: <boost_at_y...>
> Sent: Saturday, February 23, 2002 10:00 AM
> Subject: [boost] Re: Boost.Format final submission
> 
> 
> > --- In boost_at_y..., Samuel Krempp <krempp_at_c...> wrote:
> > > On Sat, 2002-02-23 at 04:25, Samuel Krempp wrote:
> > > > Okay, I've finally accessed a MSVC6 compiler, and fixed all
> > problems
> > > > format2 had with this compiler.
> > >
> > > argh, no, false alarm, there's still an overload MSVC can't
> > resolve..
> > >
> > > I'm still looking for a workaround..
> > >
> > > --
> > > Samuel
> >
> > First of all add boost/config.hpp into format.hpp
> > and make group1 constructor explicit. But you still end up with 
the
> > following ambiguity on MSVC:
> >
> > template<typename T>
> > struct A {
> > };
> >
> > template<typename T>
> > void foo( T const& ) {}
> >
> > template<typename T>
> > void foo( A<T> const& ) {}
> >
> > int main(int, char*[])
> > {
> >     A<int> a_int;
> >
> >     foo( a_int );
> >
> >     return 0;
> > }
> >
> > error C2667: 'foo' : none of 2 overload have a best conversion
> >
> > Anyone knows a fix?
> 
> It depends whether 'A' is supposed to represent a user-defined type 
or a
> template supplied by the library. If the latter, you can intrude on 
foo and
> detect that T is an A<T> using the sizeof() trick, then dispatch as
> appropriate. Otherwise, I'm afraid there's no way.
> 
> -Dave

It's latter. Could you clarify the trick, please?