Removing tombstone reanimation

[Andrew Bartlett]

On Sat, 2015-07-11 at 08:27 +1200, Andrew Bartlett wrote:
> On Fri, 2015-07-10 at 23:18 +0300, Kamen Mazdrashki wrote:
> > Hi Andrew,
> > 
> > On Fri, Jul 10, 2015 at 11:13 PM, Andrew Bartlett <abartlet at samba.org> wrote:
> > > On Sun, 2015-05-31 at 02:12 +0300, Kamen Mazdrashki wrote:
> > >> Hi Andrew,
> > >>
> > >> Could you please take a look at attached patch - it disables
> > >> tombstone_reanimation module.
> > >> It is still compiled, but skips registration and I have also removed it
> > >> from the list of dsdb modules.
> > >> Running tombstone_reanimation test fails nicely :)
> > >
> > > Kamen,
> > >
> > > I've reviewed this and am pushing it now.  We next need to try and
> > > re-add the dbcheck tests as not flapping, and remove
> > > tombstone_reanimation as falpping so when we put this back it is a real
> > > test.
> > >
> > > I did this now because we need dbcheck tests back for 4.3, and we need
> > > to ensure tombstone_reanimation isn't an accidental feature of 4.3 until
> > > we fix it.
> > >
> > > Do you have any plan for how to resolve this?  I feel really bad about
> > > removing the hard work you did here, it would really be great to have
> > > this back!
> > 
> > Frankly said, I don't understand what is the relation with dbcheck tests.
> > I thought it is tombstone_reanimation and deleted_objects flapping test's fault
> > (sorry for remaining silent for so long).
> > I will be happy to help bringing tombsone_reanimation back again, I just
> > need to know what to chase - dbcheck failures or tombstone_reanimation
> > failing occasionally.
> 
> It is probably all inter-related, because they test the same thing -
> that the DN is the expected DN.  That is:
>  - dbcheck detects that the DN and name don't end up correctly matching
>  - the deleted_objects and tombstone reanimations tests confim that the
> DN is as we expect after conflict resolution
> 
> The bug may well be in the repl_meta_data deleted object conflict
> resolution handling. 
> 
> I strongly suspect that once you fix the runtime tests, that the dbcheck
> test will be OK, or in the inverse, that the dbcheck test failure data
> might be enough to understand what is going wrong at runtime. 
> 
> I hope this helps,
> 
> Andrew Bartlett

Annoyingly, even removing tombstone_reanimation isn't so simple.  I now
get:
baseDN: DC=samba,DC=example,DC=com
UNEXPECTED(failure): samba4.ldap.acl.python(ad_dc_ntvfs).__main__.AclUndeleteTests.test_undelete(ad_dc_ntvfs)
REASON: Exception: Exception: Traceback (most recent call last):
  File "bin/python/samba/tests/__init__.py", line 141, in run
    testMethod()
  File "/memdisk/abartlet/a/b6733/samba/source4/dsdb/tests/python/acl.py", line 1730, in test_undelete
    self.assertEquals(num, ERR_INSUFFICIENT_ACCESS_RIGHTS)
AssertionError: 19 != 50

We really need this sorted before we branch for 4.3 - with the fapping
test issues, this should never have got into master (for which I'm as
much responsible, as I reviewed it) :-(

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