Boost logo

Boost Testing :

From: Markus Schöpflin (markus.schoepflin_at_[hidden])
Date: 2004-10-27 04:18:59

Robert Ramey wrote:

> I noticed the compiler warnings every time serializaton is invoked by the
> tests. I would be curious to know what we have to do to correct/suppress
> these. With other compilers I had to carefully adjust code which passes
> references in order to avoid some non-obvious C++ pitfalls in this area. So
> it wouldn't suprise me that we need one more tiny tweak for this compiler.
> Of course, I can't do that from here.

I tried to analyse we the compiler issues the warning. The last two items
of the instantiation stack look like this:

cxx: Warning: /vol1/tmp/boost/boost/boost/archive/detail/oserializer.hpp,
line 506: #461-D
           initial value of reference to non-const must be an lvalue
           detected during:
             instantiation of "void
                       T>::invoke(Archive &, const T &) [with
                       Archive=boost::archive::binary_oarchive, T=const A
                       at line 551

template<class Archive, class T>
inline void save(Archive & ar, const T &t){
         BOOST_DEDUCED_TYPENAME mpl::eval_if<is_pointer<T>,
             mpl::identity<detail::save_pointer_type<Archive, T> >,
         BOOST_DEDUCED_TYPENAME mpl::eval_if<is_enum<T>,
             mpl::identity<detail::save_enum_type<Archive, T> >,
         BOOST_DEDUCED_TYPENAME mpl::eval_if<is_array<T>,
             mpl::identity<detail::save_array_type<Archive, T> >,
             mpl::identity<detail::save_non_pointer_type<Archive, T> >
>::type typex;
     typex::invoke(ar, t); // <- *** 551 ***

             instantiation of "void boost::archive::save(Archive &, const T &)
                       [with Archive=boost::archive::binary_oarchive, T=const A
                       [10]]" at line 68 of

     void save_override(const T & t, BOOST_PFTO int)
         archive::save(* this->This(), t); // <- *** 68 ***

The only explanation I can come up with is that "*this->This()" creates a
temporary which is then passed to "invoke(ar, t)". Does that make sense?


Boost-testing list run by mbergal at