Re: [boost] is_incomplete

From: David Abrahams (dave_at_[hidden])
Date: 2002-10-23 11:27:58

Yitzhak Sapir <yitzhaks_at_[hidden]> writes:

> On Wed, 23 Oct 2002, Peter Dimov wrote:
> 
> > From: "Aleksey Gurtovoy" <agurtovoy_at_[hidden]>
> > > Peter Dimov wrote:
> > > > is_incomplete can never work. It violates ODR. ;-)
> > >
> > > You are right, of course. It should be something like this instead:
> >
> > [...]
> >
> > > Hmm, actually we can do even better:
> > >
> > >     #define BOOST_IS_INCOMPLETE(T) \
> > >         is_incomplete< T, boost::mpl::integral_c<long,__LINE__> > \
> > >     /**/
> > >
> > >     template< typename T >
> > >     struct something
> > >         : mpl::if_c<
> > >               BOOST_IS_INCOMPLETE(T)::value
> > >             , something_impl1<T>
> > >             , something_impl2<T>
> > >             >
> > >     {
> > >     };
> >
> > Now something<> violates ODR. ;-)
> 
> It could be used as BOOST_STATIC_ASSERT_COMPLETE(T).  

That's spelled 

        typedef int not_a_complete_type[sizeof(T)];

-- 
                    David Abrahams
dave_at_[hidden] * http://www.boost-consulting.com
Building C/C++ Extensions for Python: Dec 9-11, Austin, TX
http://www.enthought.com/training/building_extensions.html