Good, glad that's taken care of now.
> I'm still curious as to why there is no read cacheing any more when
> sharing from WfW. This was working a few revisions ago, and I'm wondering
> if it can be fixed.
I had hoped this was an artifact of the broken timestamps, but if not
we'll have to track it down. Could you enable the "invalidating cache"
message (in inode.c, smb_revalidate_inode, when the mtime changes) and
see if this gets triggered when running bonnie? Normally there's no
invalidation when we go from writing the file to reading, so the reading
should come out of the cache.
Regards,
Bill