On Jun 8 14:53, Achim Gratz wrote: > Corinna Vinschen <corinna-cygwin <at> cygwin.com> writes: > > > > //server/some/directory /mnt/server-share none binary 0 0 > > > #bind mounts > > > /mnt/server-share/some/other/directory /mnt/task1 none binary,bind > > > > > > That works well until I try to add another such bind mount and activate it > > > via 'mount -a' without starting a new Cygwin session. Cygwin tries to > > > create a mount to the underlying local directory of /mnt/task1 instead of > > > binding, which doesn't work of course. > > > > How does your second bind mount look like? > > # new bind mount > /mnt/server-share/some/other/directory /mnt/task2 none binary,bind > > Which tries to mount to > > %CYGWIN_ROOT%/mnt/server-share/some/other/directory > > when I do a 'mount -a' in the same session. If I instead put in > > //server/some/directory/some/other/directory none binary 0 0 > > then everything works OK, AFAICS. This might be related to the fact that > some bind mounts appear before the physical mounts in the mount table (not > in fstab!), at least if the mounts are processed in the same order.
Easily reproducible, thank you. I think I found the culprit. mount(1) always converts the mnt_fsname field to backslash notation. That breaks converting the bind mount from the expected POSIX notation to a valid Win32 path. I fixed mount(1) in git. If you're set up to build Cygwin yourself, you want to give the patch a try. No new DLL, just call the new mount(1) tool. Otherwise I'll build a new snapshot today or tomorrow (I'm hacking on something else right now). Thanks, Corinna -- Corinna Vinschen Please, send mails regarding Cygwin to Cygwin Maintainer cygwin AT cygwin DOT com Red Hat
pgpLQKhPIB2Eg.pgp
Description: PGP signature
