Robert Ramey writes:
I've made changes in the trunk to address this problem.
Please try to run this test again to verifiy that I've indeed fixed it.
I've just checked the head of the trunk and all my serialization tests now pass. Thanks to everyone involved for fixing this issue. Cheers, -- Robert p.s. As a (trivial?) asside would it be reasonable to apply the following patch to avoid compilation warnings on gcc? Index: boost/serialization/extended_type_info_typeid.hpp =================================================================== --- boost/serialization/extended_type_info_typeid.hpp (revision 49317) +++ boost/serialization/extended_type_info_typeid.hpp (working copy) @@ -112,7 +112,7 @@ return NULL; } } - void destroy(void const * const p) const { + void destroy(void const * const /* p */) const { // the only current usage of extended type info is in the // serialization library. The statement below requires // that destructor of type T be public and this creates