Adrian, I did not remove them actually, Fabrice told that the library functions are causing problems with network drives: nt.c: /* stat has been fixed since MSVC 5.0. Oh, and do not encapsulater stat for non-MS compilers, too */ /* #### popineau(a)ese-metz.fr says they still might be broken. Oh well... Let's add that `1 ||' condition.... --kkm */ #if 1 || defined(_MSC_VER) && _MSC_VER < 1100 /* Since stat is encapsulated on Windows NT, we need to encapsulate the equally broken fstat as well. */ int _cdecl fstat (int handle, struct stat *buffer) { The problem must be elsewhere... I'll look at, thanks for the report anyway! Some time ago, Adrian Aichner wrote... Perhaps. Big K I just did the test, and got the correct results. Could it be that b31 went out with library versions of stat/fstat? I'll put a note into nt.c describing this problem. directory-sep-char:/ path:d:\tmp\ file-exists-p:t file-truename:d:/tmp/ expand-file-name:d:/tmp/ directory-file-name:d:\tmp find-file-name-handler:nil path:d:\tmp/ file-exists-p:t file-truename:d:/tmp/ expand-file-name:d:/tmp/ directory-file-name:d:\tmp find-file-name-handler:nil path:d:\tmp file-exists-p:t file-truename:d:/tmp expand-file-name:d:/tmp directory-file-name:d:\tmp find-file-name-handler:nil directory-sep-char:\ path:d:\tmp\ file-exists-p:t file-truename:d:\tmp\ expand-file-name:d:\tmp\ directory-file-name:d:\tmp find-file-name-handler:nil path:d:\tmp/ file-exists-p:t file-truename:d:\tmp\ expand-file-name:d:\tmp\ directory-file-name:d:\tmp find-file-name-handler:nil path:d:\tmp file-exists-p:t file-truename:d:\tmp expand-file-name:d:\tmp directory-file-name:d:\tmp find-file-name-handler:nil