Last 3 autobuilds have failed...

[Alexander Bokovoy]

On pe, 12 tammi 2018, Andreas Schneider wrote:
> On Friday, 12 January 2018 11:02:50 CET Alexander Bokovoy wrote:
> > On pe, 12 tammi 2018, Alexander Bokovoy via samba-technical wrote:
> > > On pe, 12 tammi 2018, Andreas Schneider via samba-technical wrote:
> > > > On Friday, 12 January 2018 09:04:33 CET Björn JACKE via samba-technical 
> wrote:
> > > > > On 2018-01-11 at 08:38 -0800 Jeremy Allison via samba-technical sent 
> off:
> > > > > > [2199(16318)/2230 at 3h40m54s]
> > > > > > samba.blackbox.wbinfo(ad_member:local)(ad_member:local)
> > > > > > UNEXPECTED(failure): samba.blackbox.wbinfo(ad_member:local).confirm
> > > > > > output of wbinfo -i against ad_member(ad_member:local) REASON:
> > > > > > Exception:
> > > > > > Exception: expected
> > > > > > 'ADDOMAIN/administrator:*:100000:100004:Administrator:/home/ADDOMAIN
> > > > > > /Doma
> > > > > > in Users/administr +ator:/bin/false' or
> > > > > > 'ADDOMAIN/administrator:*:100000:100004::/home/ADDOMAIN/Domain
> > > > > > Users/administrator:/bin/false' got 'ADDOMAIN/
> > > > > > +administrator:*:100000:100004::/home/ADDOMAIN/domain
> > > > > > users/administrator:/bin/false'
> > > > > 
> > > > > I also got that error a couple of times and it looks a bit
> > > > > frightening. If
> > > > > the upper/lower case writing of of nss_winbind is eratic, this is
> > > > > calling
> > > > > for a lot of trouble. Does anybody know what the error might be here?
> > > > > It
> > > > > wouldn't be good if we ignore this error and it ends up a 4.8
> > > > > regression
> > > > > bug finally.
> > > > 
> > > > Uppercase lowercase?
> > > > 
> > > > There is "Administrator" missing after the forth colon! Could be related
> > > > to
> > > > the last substitute changes!
> > > 
> > > No, GECOS field can be empty -- the test actually says about it as it
> > > expects either existing GECOS or empty one. However, path is including
> > > lowercased 'domain users' group name but the test expects it intact.
> > 
> > Attached patch normalizes both tested and expected lines to the same
> > case before comparison. It still reports the original tested and
> > expected lines in case of an error.
> 
> but if the test is flapping it means the case changes in some runs. So either 
> it picks up a wrong config file or something else is wrong, isn't it?
'template homdir' is configured in setup_admember() of Samba3.pm:
       template homedir = /home/%D/%G/%U

%G substitution uses gidtoname() which is using getgrgid() call and
returns gr_name out of it. However, we do map 'domain users' SID to
domusers group, so technically this should be
/home/$DOMAIN/domusers/administrator if things are working properly,
isn't it?

-- 
/ Alexander Bokovoy