On Mon, Oct 29, 2007 at 11:12:57AM -0700, Ralf W. Grosse-Kunstleve wrote:
I'm testing Boost.Python compilation with the latest g++ 4.3.0
Index: boost/config/stdlib/libstdcpp3.hpp =================================================================== --- boost/config/stdlib/libstdcpp3.hpp (revision 40556) +++ boost/config/stdlib/libstdcpp3.hpp (working copy) @@ -68,6 +68,11 @@ # define BOOST_HAS_SLIST # define BOOST_HAS_HASH # define BOOST_SLIST_HEADER <ext/slist> +# if defined(__GLIBCXX__) && __GLIBCXX__ < 20071018 // GCC >= 4.3.0
A stupid question: Why do you use __GLIBCXX__ instead of __GNUC__ and __GNUC_MINOR__ as described on http://gcc.gnu.org/onlinedocs/cpp/Common-Predefined-Macros.html?
# define BOOST_HASH_SET_HEADER <ext/hash_set> # define BOOST_HASH_MAP_HEADER <ext/hash_map>
Is the location of hash_set a property of the glibc and not of the compiler or belong both together? I do not see how changing the path of a header file could cause a ABI bump of glibc ... Jens