On Sun, Dec 22, 2002 at 01:36:34AM -0500, Oleg wrote: > On Saturday 21 December 2002 08:21 pm, Lars Ellenberg wrote: > > On Sat, Dec 21, 2002 at 02:51:29PM -0500, Oleg wrote: > > > That is cmkdir gives an error, but creates a directory, while cattach > > > fails altogether. Does anyone know why? And isn't CFS supposed to be file > > > system agnostic?
cfs doesn't work on all filesystems, see http://bugs.debian.org/136411 and /usr/share/doc/cfs/README.Debian. This is another case though. > cmkdir encrypted_directory > writes to file "encrypted_directory/...". Apparently, this can not be done > with FAT32. Had CFS used another file, such as "..a" instead of "...", it > would be able to work with FAT32, I believe. The linux vfat implementation doesn't support filenames ending with a dot, see linux/fs/vfat/namei.c:int vfat_add_entry(). I've tested linux-2.4.17: # mount |grep mnt; cd /mnt /tmp/fatfs on /mnt type vfat (rw,loop=/dev/loop0) # touch foo. # cp foo. foo cp: `foo.' and `foo' are the same file # cp foo. foo.. cp: `foo.' and `foo..' are the same file # Dots at the end of the filename are stripped, so creating a file '...', which cfs does, fails. I'm not sure if this is a bug in vfat, I didn't find it documented anywhere; google gave me this: http://www.netbsd.org/cgi-bin/query-pr-single.pl?number=14942 If this isn't a bug in vfat or a bug that cannot be fixed, it should be possible to work around this. If so, please file a wishlist bug against cfs. Regards, Gerrit.
