Attached patch applied. Compare to the proposed one I had to: - Disable the new check for __gnu_debug::basic_string<>, it is supported following Standard words. To do so I had to slithly review how _GLIBCXX_DEBUG_PEDANTIC was managed. - Add check on forward_list::insert_after 2013-08-01 François Dumont PR libstdc++/57779 * include/debug/formatter.h (_Debug_msg_id): Add __msg_insert_itself_range entry. * include/debug/functions.h (_Insert_range_from_self_is_safe<>): New, indicate container types supporting self range insertion in GNU implementation. (__foreign_iterator): New, check if an iterator points to a given sequence. * include/debug/macros.h (__glibcxx_check_insert_range): Add check using __foreign_iterator. (__gibcxx_check_insert_range_after): Likewise. * include/debug/string (_Insert_range_from_self_is_safe<>): Partially specialized to mark __gnu_debug::basic_string<> as supporting self range insert. * include/debug/list (_Insert_range_from_self_is_safe<>): Partially specialized to mark std::list as supporting self range insert if _GLIBCXX_DEBUG_PEDANTIC is not defined. * include/debug/forward_list (_Insert_range_from_self_is_safe<>): Likewise. * src/c++11/debug.cc (_S_debug_messages): Add __msg_insert_itself_range_entry message. (_Error_formatter::_Parameter::_M_print_description): Display iterator sequence address rather than sequence address when the parameter type is an iterator. (_Error_formatter::_M_print_word): Enhance behavior when displaying a word with an appended '\n'. * testsuite/util/debug/checks.h (check_insert4<>): New. * testsuite/23_containers/deque/debug/insert5_neg.cc: New. * testsuite/23_containers/vector/debug/insert5_neg.cc: Likewise. * testsuite/23_containers/vector/debug/insert6_neg.cc: Likewise. * testsuite/23_containers/vector/debug/57779_neg.cc: Likewise. * testsuite/23_containers/list/debug/insert5_neg.cc: Likewise. * testsuite/23_containers/forward_list/debug/insert_after4_neg.cc: Likewise. François On 07/31/2013 11:50 PM, Paolo Carlini wrote: > On 07/31/2013 09:58 PM, François Dumont wrote: >> Here is another proposal using std::common_type to find out how to >> instantiate std::less and std::greater_equal. Much easier with C++11 >> features indeed. >> >> Tested under linux x86_64 debug mode. >> >> Ok to commit ? > Ok, thanks! > > Paolo. >