I'm trying the example on the "Character Set Conversions" page of the Boost.Locale documentation: http://www.boost.org/doc/libs/1_49_0/libs/locale/doc/html/charset_handling.h... (the full code I'm running is at the end of this message). When compiling with gcc: g++ -o test -lboost_locale test.cpp It run correctly - it prints "שלום" to standard output. However, when compiling with clang: clang++ -o test -lboost_locale test.cpp It produces, as far as I can tell, gibberish - it prints "של×" to standard output. The hex values of the output are: 0xc3 0x97 0xc2 0xa9 0xc3 0x97 0xc2 0x9c 0xc3 0x97 0xc2 0x95 0xc3 0x97 0xc2 0x9d -------- What I'd like is to have a stream that accepts wide strings in the current locale and produces narrow strings in UTF-8 - am I completely off the track? If so, what should I do instead, and if not, what am I doing wrong? Thanks! - Jesse Beder ------- (code follows) #include #include #include #include #include <iostream> namespace io = boost::iostreams; class consumer { public: typedef char char_type; typedef io::sink_tag category; std::streamsize write(const char* s, std::streamsize n) { std::cout.write(s,n); return n; } }; int main() { typedef io::code_converter<consumer> converter_device; typedef io::stream converter_stream; consumer cons; converter_device dev; boost::locale::generator gen; dev.imbue(gen("en_US.UTF-8")); dev.open(cons); converter_stream stream; stream.open(dev); stream << L"שלום"; return 0; }