ntlm_auth autobuild failures on the Catalyst Cloud after XFILE

Jeremy Allison jra at samba.org
Fri Jan 6 18:03:58 UTC 2017


On Fri, Jan 06, 2017 at 09:11:39PM +1300, Andrew Bartlett wrote:
> I'm sorry to add to this debate around flapping tests, but at the same
> time I don't want to miss out a possibly important data point.
> 
> Since around the time of the XFILE -> FILE conversion (13 Dec), the
> builds that my team at Catalyst have been doing on the Catalyst cloud
> have been failing with the attached.  (We also see the ctdb errors that
> martin has been fixing).
> [1809(11135)/2025 at 2h28m47s] samba.ntlm_auth.(ad_dc:local)
> UNEXPECTED(failure): samba.ntlm_auth.(ad_dc:local).ntlm_auth with
> NTLMSSP gss-spnego-client and gss-spnego server against winbind with
> require-membership-of(ad_dc:local)
> REASON: Exception: Exception: Could not determine hostname for target
> computer, cannot use kerberos
> Traceback (most recent call last):
>   File
> "/home/ubuntu/autobuild/b15247/samba/source3/torture/test_ntlm_auth.py"
> , line 329, in <module>
>     main()
>   File
> "/home/ubuntu/autobuild/b15247/samba/source3/torture/test_ntlm_auth.py"
> , line 277, in main
>     buf = readLine(client_in)
>   File
> "/home/ubuntu/autobuild/b15247/samba/source3/torture/test_ntlm_auth.py"
> , line 43, in readLine
>     raise ReadChildError()
> __main__.ReadChildError
> 
> FAILED (1 failures, 0 errors and 0 unexpected successes in 0
> testsuites)

Just FYI - this is almost *certainly* another subtle hiding
bug in using printf/fprintf when emitting an error condition.

Using the changes to test_ntlm_auth.py I posted will help
track this down - specifically the addition of:

@@ -40,6 +58,8 @@ def readLine(pipe):
     buf = os.read(pipe, 2047)
     newline = buf.find('\n')
     if newline == -1:
+        dbg("JRA: Missing newline")
+        dbg("JRA: in string |%s|" % buf)
         raise ReadChildError()
     return buf[:newline]

probably means it's not finding a newline in the
buffer out. That debug statement will help you
find out exactly where in the ntlm_auth code
it's going wrong.



More information about the samba-technical mailing list