[PATCH] Refuse to commit a faulty reindex

[Andrew Bartlett]

On Wed, 2018-04-04 at 16:45 +0200, Stefan Metzmacher wrote:
> Hi Andrew,
> 
> Am 04.04.2018 um 01:25 schrieb Andrew Bartlett via samba-technical:
> > +        if hasattr(self, 'IDXGUID'):
> > +            self.assertEquals(len(res), 3)
> > +        else:
> > +            # We should not return this entry twice, but sadly
> > +            # we have not yet fixed
> > +            # https://bugzilla.samba.org/show_bug.cgi?id=13361
> > +            self.assertEquals(len(res), 4)
> > +
> 
> Isn't 5c1504b94d1417894176811f18c5d450de22cfd2 the fix for bug #13361 ?

Sadly only for the GUID index mode.  

There is much more to fix here, but I had to step back and get patches
that work for the re-index case rather than fix all the bugs as it was
just becoming a rat-hole.

Not only do we have this duplicate return for multi-value case-wise
equal attributes, we also have major issues where the index gets
written to while we do a modify or add, but not rolled back when that
modify or add fails (eg due to constraint violation).  

I had to step back from all that and fix the pressing issue or else I
wasn't going to get any fixes to our users. 

Ouch!

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