Possible reason for f5a0ccc228ce1ef failure

[Michael Adam]

On 2016-03-22 at 16:29 +0100, Volker Lendecke wrote:
> Hi!
> 
> There is one possible reason for the failure mentioned in the commit
> message of f5a0ccc228ce1ef: I found that when the member environments
> join a domain, the winbind socket environment variable points at the
> DC's winbind, where the member's winbind does not run yet. This means
> that the join creates BUILTIN\Administrators with a rid allocated from
> the DC's idmap.tdb. The HWM in the member's idmap.tdb is not even
> started yet, so later assignments will get us clashes. If I now move
> the idmap range for members out of the DC's range space, I get an
> unexpected success for the nss tests. Would it be okay to unmark the
> knownfails that you entered with f5a0ccc228ce1ef?

Oh, if this phenomenon is fixed by such a patch of yours,
sure it would be ok! It would even be required. :-)
And thanks!

I also observed that with my yet-unpushed branch to tear down
and re-setup each enviroment before starting a testcase/suite
against it, these patches also succeed.

If you could show the patch, I'd happily review it in.

Speaking of the fix: Why is the member winbind socket
pointing at the DC's winbind? Is that a bug in our
selftest environment? After all, the socket of the
DC's winbind socket should be under (e.g.) st/ad_dc/ while
the member's winbind should be under st/ad_member/ ?
*scratching my head*... :-)

Thanks again for finding this!

Michael
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: not available
URL: <http://lists.samba.org/pipermail/samba-technical/attachments/20160322/0b20a85a/signature.sig>