On Apr 7, 10:52 am, Max Erickson <[email protected]> wrote: > Most of the remaining failing tests have unicode characters. > > I guess the problem is with file input in the mingw libraries, but I > haven't figured out where to look yet. Any tips there? >
The root of the issue is that std::isspace (in read_line in textual.cc) is eating the last byte of the Utf-8 Euro symbol. I think the function is implemented in terms of 0 to 255 and is then interpreting the byte as -84. Locale stuff on mingw seems to have pretty thin support. An appropriate fix certainly isn't obvious to me. Max
