With an UTF-8 locale, when using codecvt<wchar_t, char, mbstate_t> with the following input {0xc3, 0xa9, 0x20, 0xf0, 0xaa, 0x98} the function claims to have consumed all six bytes, albeit it has only returned relevant data for three bytes, the result being {0xe9, 0x20}.
-- Summary: codecvt doesn't update how much input it has consumed correctly Product: gcc Version: 4.4.3 Status: UNCONFIRMED Severity: normal Priority: P3 Component: libstdc++ AssignedTo: unassigned at gcc dot gnu dot org ReportedBy: loufoque at gmail dot com http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45156