DragonFly commits List (threaded) for 2009-03
Re: DragonFly-184.108.40.2064.g574066 master sbin/hammer cmd_cleanup.c
Ok, I added support for the @@-1:%05d mode. It does make 'df' nicer,
at least if you do not have any slave PFSs.
I'm going to note something here to try to avoid some future confusion.
The actual softlinks in the filesystem are "@@PFS%05d". However, the
HAMMER VFS presents them in the longer form when you stat/ls/readlink.
So it is actually the HAMMER VFS which must present the shorter form.
The reason for doing this is primarily to support slave PFSs and
master->slave and (eventually) slave->master transitions. On a slave
PFS the 'current' transaction id is the most recently synced mirroring
operation, meaning that it can change from moment to moment, so we
can't lay down a static softlink.
By generating a dynamic softlink DragonFly's namecache does not get
confused whenever slave PFSs are updated by a mirroring operation.
A person CD'd into a previously synced slave TID will still see that
snapshot while a re-CD into the mount will see the latest synced