On Thu, Apr 20, 2006 at 02:27:01PM +0200, u+codalist-p4pg_at_chalmers.se wrote:
> I meant in fact rename/unlink conflict, which is more of a conflict than
> unlink/unlink - the intent is less clear in this case.
> Nevertheless, I would treat "unlink a non-existent name" as no conflict
> independently of whether that was another unlink or a rename which conflicts.

I hadn't even considered the unlink/rename possibility. Interesting,
I'll have to try some combinations and see how resolution and
reintegration handle these cases.

> By the way, how is say local-global conflict unlink/rename solvable,
> when you are trying to rename a file which no longer exists?

Any type of conflict that involves unlink/rmdir is difficult. Typically
when the directory on the server still has the old name we would be able
to figure out how to show the conflict.

But I haven't checked the cases where we unlink a file that was renamed
or rename a file that was unlinked. As far as reintegration is concerned
we do correctly handle the case when a local rename unlinks the target
and that was already removed at the servers.

> Does it silently disappear (replacing so to say "rename" with "unlink")?
> There seems to be no way to recover it, anyway, if its contents is not in
> the local cache?

I'll give it a try.

Jan