cgf wrote: > Is it correct to assume that only fhandler_base::open cares about >trailing dots?
Good point. This bring back memories. The initial motivation was to fix problems introduced by the use of NtCreateFile http://www.cygwin.com/ml/cygwin/2004-04/msg01250.html and there were successive changes 2004-04-30 Corinna Vinschen <[EMAIL PROTECTED]> * path.cc (normalize_posix_path): Remove trailing dots and spaces. http://cygwin.com/ml/cygwin-patches/2004-q2/msg00053.html 2004-05-06 Pierre Humblet <[EMAIL PROTECTED]> * path.cc (path_conv::check): Strip trailing dots and spaces and return error if the final component had only dots and spaces. (normalize_posix_path): Revert 2004-04-30. However, as a side effect, checking the tail in :check also cleanly fixed longstanding dormant issues with the path hash and with chroot (at least). So checking the tail in fhandler_base::open() is too late. It should be done before exiting :check(), perhaps only in the case where the path refers to a disk file, preferably with little processing overhead. Although it wasn't done before 2004/04, we should also make sure (I have no free time for the moment) that nothing goes wrong inside :check() while we lookup symbolic links. Pierre