On Thu, Dec 16, 2004 at 07:59:05AM -0700, Mark Paulus wrote: >Other than the way I proposed, I'm not sure how to fix this, since the >issue seems to be that conv_to_win32_path() needs to get the trailing >dot in it's input argument, and check() is stripping it out. The only >way I can see to fix this behaviour is to leave the trailing dots in >the string. Maybe conv_to_win32_path needs to deal/strip with the >trailing dots, depending upon whether it's a managed filesystem or not?
Yes, *of course* there either has to be special accommodations for managed mounts or you have to show that your change will not affect normal cygwin operation for non-managed mounts. Your patch just essentially nuked most of cygwin's handling of trailing dots. Unless you can prove that the previous code was misguided in the general case of non-managed mounts, doing this is obviously the wrong solution. cgf