diff options
Diffstat (limited to 'contrib/libstdc++/config/locale/generic/codecvt_members.cc')
-rw-r--r-- | contrib/libstdc++/config/locale/generic/codecvt_members.cc | 179 |
1 files changed, 148 insertions, 31 deletions
diff --git a/contrib/libstdc++/config/locale/generic/codecvt_members.cc b/contrib/libstdc++/config/locale/generic/codecvt_members.cc index ff65dff..8bdc227 100644 --- a/contrib/libstdc++/config/locale/generic/codecvt_members.cc +++ b/contrib/libstdc++/config/locale/generic/codecvt_members.cc @@ -38,7 +38,7 @@ namespace std { // Specializations. -#ifdef _GLIBCPP_USE_WCHAR_T +#ifdef _GLIBCXX_USE_WCHAR_T codecvt_base::result codecvt<wchar_t, char, mbstate_t>:: do_out(state_type& __state, const intern_type* __from, @@ -46,25 +46,61 @@ namespace std extern_type* __to, extern_type* __to_end, extern_type*& __to_next) const { - result __ret = error; - size_t __len = min(__from_end - __from, __to_end - __to); - size_t __conv = wcsrtombs(__to, &__from, __len, &__state); + result __ret = ok; + // The conversion must be done using a temporary destination buffer + // since it is not possible to pass the size of the buffer to wcrtomb + state_type __tmp_state(__state); - if (__conv == __len) - { - __from_next = __from; - __to_next = __to + __conv; - __ret = ok; - } - else if (__conv > 0 && __conv < __len) + // The conversion must be done by calling wcrtomb in a loop rather + // than using wcsrtombs because wcsrtombs assumes that the input is + // zero-terminated. + + // Either we can upper bound the total number of external characters to + // something smaller than __to_end - __to or the conversion must be done + // using a temporary destination buffer since it is not possible to + // pass the size of the buffer to wcrtomb + if (MB_CUR_MAX * (__from_end - __from) - (__to_end - __to) <= 0) + while (__from < __from_end) + { + const size_t __conv = wcrtomb(__to, *__from, &__tmp_state); + if (__conv == static_cast<size_t>(-1)) + { + __ret = error; + break; + } + __state = __tmp_state; + __to += __conv; + __from++; + } + else { - __from_next = __from; - __to_next = __to + __conv; - __ret = partial; + extern_type __buf[MB_LEN_MAX]; + while (__from < __from_end && __to < __to_end) + { + const size_t __conv = wcrtomb(__buf, *__from, &__tmp_state); + if (__conv == static_cast<size_t>(-1)) + { + __ret = error; + break; + } + else if (__conv > static_cast<size_t>(__to_end - __to)) + { + __ret = partial; + break; + } + + memcpy(__to, __buf, __conv); + __state = __tmp_state; + __to += __conv; + __from++; + } } - else - __ret = error; - + + if (__ret == ok && __from < __from_end) + __ret = partial; + + __from_next = __from; + __to_next = __to; return __ret; } @@ -75,25 +111,106 @@ namespace std intern_type* __to, intern_type* __to_end, intern_type*& __to_next) const { - result __ret = error; - size_t __len = min(__from_end - __from, __to_end - __to); - size_t __conv = mbsrtowcs(__to, &__from, __len, &__state); + result __ret = ok; + // This temporary state object is neccessary so __state won't be modified + // if [__from, __from_end) is a partial multibyte character. + state_type __tmp_state(__state); - if (__conv == __len) + // Conversion must be done by calling mbrtowc in a loop rather than + // by calling mbsrtowcs because mbsrtowcs assumes that the input + // sequence is zero-terminated. + while (__from < __from_end && __to < __to_end) { - __from_next = __from; - __to_next = __to + __conv; - __ret = ok; + size_t __conv = mbrtowc(__to, __from, __from_end - __from, + &__tmp_state); + if (__conv == static_cast<size_t>(-1)) + { + __ret = error; + break; + } + else if (__conv == static_cast<size_t>(-2)) + { + // It is unclear what to return in this case (see DR 382). + __ret = partial; + break; + } + else if (__conv == 0) + { + // XXX Probably wrong for stateful encodings + __conv = 1; + *__to = L'\0'; + } + + __state = __tmp_state; + __to++; + __from += __conv; } - else if (__conv > 0 && __conv < __len) + + // It is not clear that __from < __from_end implies __ret != ok + // (see DR 382). + if (__ret == ok && __from < __from_end) + __ret = partial; + + __from_next = __from; + __to_next = __to; + return __ret; + } + + int + codecvt<wchar_t, char, mbstate_t>:: + do_encoding() const throw() + { + // XXX This implementation assumes that the encoding is + // stateless and is either single-byte or variable-width. + int __ret = 0; + if (MB_CUR_MAX == 1) + __ret = 1; + return __ret; + } + + int + codecvt<wchar_t, char, mbstate_t>:: + do_max_length() const throw() + { + // XXX Probably wrong for stateful encodings. + int __ret = MB_CUR_MAX; + return __ret; + } + + int + codecvt<wchar_t, char, mbstate_t>:: + do_length(state_type& __state, const extern_type* __from, + const extern_type* __end, size_t __max) const + { + int __ret = 0; + state_type __tmp_state(__state); + + while (__from < __end && __max) { - __from_next = __from; - __to_next = __to + __conv; - __ret = partial; + size_t __conv = mbrtowc(NULL, __from, __end - __from, &__tmp_state); + if (__conv == static_cast<size_t>(-1)) + { + // Invalid source character + break; + } + else if (__conv == static_cast<size_t>(-2)) + { + // Remainder of input does not form a complete destination + // character. + break; + } + else if (__conv == 0) + { + // XXX Probably wrong for stateful encodings + __conv = 1; + } + + __state = __tmp_state; + __from += __conv; + __ret += __conv; + __max--; } - else - __ret = error; - + return __ret; } #endif |