On Tue, 2005-09-27 at 12:16 +0800, Nicholas Hemsley wrote: > I wanto to setup UnionFS to write any changes made to either > filesystem to a stacked filesystem. > i.e. > > /var/trunk/ contains the read-only filesystem > /var/developers/<developer> contains a filesystem stacked ontop. This > will initially be empty, & cleaned when the user does a SVN checkin, > on a per-file basis (if this is indeed possible with a script hook in > SVN). When the user saves a file, it get's written here. > > I assume this is possible with unionfs. However: what happens when a > file gets deleted? I dont want it deleted from the trunk, but it > should appear to the developer that it has dissapeared. When they do a > svn checkin, then the file can get physicall deleted from trunk (THis > has nothing to do with unionFS however). Does unionFS handle this > case, or will the developer be unable to delete from their FS. You'll end up with N unionfs, one per developer along the lines of: /mnt/union/<devname> /var/developers/<devname>=rw:/var/trunk=ro
If you delete /mnt/union/<devname>/foo, /var/developers/<devname>/.wh.foo will get created, and /var/trunk/foo will be untouched. I don't know if SVN actually uses separate files for each versioned file though (I know it stores things in BDB, but not the precise semantics). This may not yield the results you expect if many things are jammed into the same file and you have several copy-on-write versions of it (you might get quickly out of sync). You'll have to ask someone more familiar with subversion. Charles _______________________________________________ unionfs mailing list [email protected] http://www.fsl.cs.sunysb.edu/mailman/listinfo/unionfs
