posixacl test failure (flapping)

[Andrew Bartlett]

On Fri, 2017-02-10 at 19:04 -0600, Scott Lovenberg wrote:
> On Fri, Feb 10, 2017 at 6:41 PM, Andrew Bartlett <abartlet at samba.org>
> wrote:
> > One of the flapping tests we have is this one:
> > 
> > We get 11, not 15 ACE values in the posix ACL.  Does anybody have
> > any
> > idea how we sometimes come up short here?
> > 
> > (You can see the expected values at
> > /memdisk/abartlet/a/b626728/samba/bin/python/samba/tests/posixacl.p
> > y:70
> > 7)
> > 
> > Thanks,
> > 
> > Andrew Bartlett
> > 
> > UNEXPECTED(failure):
> > samba.tests.posixacl.samba.tests.posixacl.PosixAclMappingTests.test
> > _set
> > ntacl_policies_check_getposixacl(ad_dc_no_nss:local)
> > REASON: Exception: Exception: Traceback (most recent call last):
> >   File
> > "/memdisk/abartlet/a/b626728/samba/bin/python/samba/tests/posixacl.
> > py",
> > line 707, in test_setntacl_policies_check_getposixacl
> >     self.assertEquals(posix_acl.count, 15,
> > self.print_posix_acl(posix_acl))
> > AssertionError: a_type: 3
> > a_perm: 7
> > gid: 3000002
> > a_type: 1
> > a_perm: 6
> > uid: 1018
> > a_type: 5
> > a_perm: 0
> > a_type: 2
> > a_perm: 6
> > a_type: 1
> > a_perm: 7
> > uid: 3000002
> > a_type: 4
> > a_perm: 7
> > a_type: 3
> > a_perm: 5
> > gid: 3000003
> > a_type: 3
> > a_perm: 7
> > gid: 3000004
> > a_type: 3
> > a_perm: 5
> > gid: 3000005
> > a_type: 3
> > a_perm: 7
> > gid: 3000006
> > a_type: 6
> > a_perm: 7
> > --
> > Andrew Bartlett                       http://samba.org/~abartlet/
> > Authentication Developer, Samba Team  http://samba.org
> > Samba Developer, Catalyst IT          http://catalyst.net.nz/servic
> > es/samba
> > 
> > 
> > 
> 
> Wasn't there an artificial limit on the number of ACEs (specifically,
> the number of uids/gids that could have an entry in the list, not the
> size of the entry itself) on some platforms
> (Solaris/OpenSolaris/BSD?) at some point in the past or is my memory
> hazy on the subject?  I feel like it was something file system and
> kernel dependent due to metadata space, but I'm also hazy on those
> particulars.  If this isn't flapping on all platforms, that might be
> an avenue to verify.  Although I'm thinking it's doubtful since 11
> isn't exactly a magic number that's round in base 2 and my gut says
> that the max number of entries is very probably a round value in
> affected platforms.

Indeed.  However on this case it is on the same platform.  I've seen it
both on sn-devel and on Ubuntu 14.04 x86_64 VMs on the Catalyst Cloud.

That is, in this case the platform is constant, but this only collapses
from 15 to 11 entries sometimes.

Additionally, we don't actually store a real ACL, our code generates a
proposed posix ACL and puts it in a fake xattr in a tdb, so the problem
is probably in our generic code, perhaps in idmap somewhere.

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