http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52476
--- Comment #2 from Daniel Krügler <daniel.kruegler at googlemail dot com> 2012-03-04 16:10:19 UTC --- (In reply to comment #1) > (we should double check that by avoiding any use of > fancy C++11 features like initializer lists in the testcase) I rewrote the testcase in C++03 form and based on TR1 unordered map: //------ #include <tr1/unordered_map> #include <iostream> #include <utility> typedef std::tr1::unordered_multimap<int, int> map_type; void printHashTable(const map_type& map) { for (unsigned i = 0; i < map.bucket_count(); ++i) { std::cout << "b[" << i << "]:" << std::endl; for (map_type::const_local_iterator it = map.begin(i); it != map.end(i); ++it) { std::cout << " " << map.hash_function()(it->first) << " [" << it->first << "," << it->second << "]" << std::endl; } } std::cout << "----------------------" << std::endl; } int main() { typedef std::pair<int, int> P; const P input1[] = { P(0,0), P(1,0), P(2,0), P(3,0), P(4,0), P(1,1) }; map_type dict(input1, input1 + sizeof(input1)/sizeof(input1[0])); printHashTable(dict); const P input2[] = { P(3,1), P(3,2), P(5,0) }; dict.insert(input2, input2 + sizeof(input2)/sizeof(input2[0])); printHashTable(dict); dict.max_load_factor(0.5); printHashTable(dict); } //------ The incorrect runtime behaviour is unchanged compared to the original form. This means, one could actually remove the [C++11] tag from the bug title.