On Thu, Jul 07, 2005 at 04:03:57PM -0400, Jan Harkes wrote:
> On Thu, Jul 07, 2005 at 01:47:09PM -0600, Patrick Walsh wrote:
> > 	And I've hit upon what I think must be the problem:
> > 
> > # cfs whereis /coda/director/snapin/pool_scm
> >   dir224  dir225  dir225

btw, the crash occurred because we were trying to create two directory
entries with the same (server) name, but pointing at different (volume
replica) objects for the server-server repair directory.

    dir224 -> 1000004.x.x
    dir225 -> 2000004.x.x
    dir225 -> 200000a.x.x

I wonder why it creates the expanded directory as soon as the conflict
is detected though. We haven't even called beginrepair yet. The code
probably always expands a conflict internally, but masks the fact that
it is already expanded by returning the dangling symlink attributes
until the volume has been marked as 'repair session active'.

Jan