Hello, This has already been discussed on the IRC, and I will simply mention the decision that has been taking in the live discussion.
On Tue, Sep 22, 2009 at 02:21:38AM +0200, olafbuddenha...@gmx.net wrote: > On Mon, Aug 17, 2009 at 06:07:20AM +0200, olafbuddenha...@gmx.net wrote: > > > settrans myroot unionfs / --mount overlay && chroot myroot > > > > (BTW, this is an obvious use case for union-mounting with something > > different than the underlying node -- why didn't I see this before? > > Actually there is a problem here: with transparent mounting, the > translator gets a port to the real underlying node. However, the real > underlying node is meaningless in this case, as we don't even include it > in the union. Effectively, we use the explicitely included directory as > underlying filesystem. > > I wonder whether passing, when there is no -u, a port to the first > "normal" directory in the union instead, would suffice to make things > behave as expected? Or would it cause even more confusion?... The adopted solution is to *always* give the mountee the port to the first normal directory, because in the usual use case (underlying filesystem + mountee) the first normal directory will be exactly the port to the underlying filesystem. I'm not sure as to the priority I should assign to this task among my tasks, however. Regards, scolobb
