Releases, locking and ldb

Andrew Bartlett abartlet at samba.org
Wed Jun 28 07:01:53 UTC 2017


On Wed, 2017-06-28 at 15:31 +1200, Andrew Bartlett wrote:
> On Wed, 2017-06-28 at 10:05 +1200, Andrew Bartlett via samba-technical
> wrote:
> > On Wed, 2017-06-28 at 00:00 +0200, Stefan Metzmacher wrote:
> > > 
> > > Can you prepare a branch on master with the required reverts?
> > > It should be possible to copy the lib/ldb lib/tdb directories
> > > to a 4.6 branch and let is pass autobuild.
> > 
> > Sure, I'll do that today.
> 
> It was a little more involved than that (showing in a way why these are
> not trivially made subprojects), but see:
>   
>  ldb-safe-locking-10
>   and 
>  samba-4-6-with-ldb-revert
> 
>  in my samba repo.
> 
> Andrew Bartlett

Sadly both this branch and v4-6-test fails autobuild in the Catalyst
cloud with the same error, so we cannot draw any conclusions as to the
effectiveness of this proposed solution:

[297(1142)/2043 at 32m16s] samba3.ntlm_auth.krb5 with old ccache(ktest:local)
UNEXPECTED(failure): samba3.ntlm_auth.krb5 with old ccache(ktest:local).ntlm_auth with krb5 gss-spnego-client and gss-spnego server(ktest:local)
REASON: Exception: Exception: could not obtain winbind domain name!
could not obtain winbind netbios name!
could not obtain winbind domain name!
could not obtain winbind separator!
Traceback (most recent call last):
  File "/home/ubuntu/autobuild/b29597/samba/source3/torture/test_ntlm_auth.py", line 329, in <module>
    main()
  File "/home/ubuntu/autobuild/b29597/samba/source3/torture/test_ntlm_auth.py", line 277, in main
    buf = readLine(client_in)
  File "/home/ubuntu/autobuild/b29597/samba/source3/torture/test_ntlm_auth.py", line 43, in readLine
    raise ReadChildError()
__main__.ReadChildError


In thinking about this, I don't see why attempting to provide an ldb
version that is backwards compatible to 4.6 is a task that has do be
done in the linear master repo, rather than in a branch unconstrained
by the release deadlines.  We should contain ourselves in the short
term to the patches that prevent building 4.6 with 1.2.0

The only good news is that I have confirmed that 4.6 fails with 1.1.31
for 'make test TESTS=fsmo', and that the proposed 1.1.32 fixes this
specific test. 

This took me the whole work day, so the lack of a clear resolution is a
little deflating, but I hope this helps.  

Please advise what the most useful next step towards a resolution here
is.

Thanks,

Andrew Bartlett

-- 
Andrew Bartlett                       http://samba.org/~abartlet/
Authentication Developer, Samba Team  http://samba.org
Samba Developer, Catalyst IT          http://catalyst.net.nz/services/samba




More information about the samba-technical mailing list