Default DNS server for Samba 4.0

[Andrew Bartlett]

On Mon, 2012-09-10 at 00:03 +0200, Michael Adam wrote:
> On 2012-09-07 at 08:02 +1000, Andrew Bartlett wrote:
> > On Thu, 2012-09-06 at 15:09 +0200, Michael Adam wrote:
> > > I absolutely agree to Kai here: Requiring the code to be covered
> > > with tests for accepting to change a default is invalid and seems
> > > very unfair to me when the code from which the default status is
> > > to be taken has almost no automatic test coverage at all.
> > 
> > I think it is a perfectly reasonable requirement.  If you want to change
> > the default, then there is a higher standard than just adding another
> > optional feature.  If you care so much about this, perhaps you could
> > work with Kai to add the tests?
> > 
> > I've worked with, written and debugged way to much code that 'should
> > work' or was at one point manually tested, and I'm really quite
> > surprised to hear an argument that having code outside the automated
> > test regime is acceptable.  As the person doing the most work on the DC
> > side of Samba 4.0 at the moment, the single thing that continues to save
> > my backside on a daily basis is 'make test' and autobuild.  We don't
> > have the large development team we used to have on the AD DC, and that's
> > sad.  But that means even more that I want to ensure that critical parts
> > of our system are as tested as they can be, despite and during changes
> > to the underlying libraries, and that *new defaults* that are proposed
> > are even more tested than what they propose to obsolete.  
> 
> Yes, and for the internal DNS code, this is the case! There are
> more automatic tests for it than for the bind setup. And I will
> try and get some ressources to help extend and complete the test
> suite.
> 
> You don't really need to convince me, that test-covered code is
> a good thing and something toward which we should strive. But
> again, I don't by the veto to switch defaults from a mechanism
> that has caused so much grief in the past and is essentially not
> covered by tests to a new code that is also run in a couple of
> setups and covered with tests to some extent at least (with an
> upward trend).
> 
> 
> So where are we with this discussion?

At the moment, to my reading of the code, the ACL support is not
implemented correctly (wrong owner of created objects).  The tests I
described in my last mail would confirm if they are or are not, and
given we have had to go over that particular issue twice already, we
really, really should have an automated test for that. 

> Kai has proposed the change of default.
> All voices in this thread except yours, Andrew, seem to be supporting the change.

For all the huffing and puffing over this, I've not see any progress on
any of the testing requirements I've suggested.  I'm sorry that everyone
else seems to have promised Kai that he could switch this on by default
as soon as the code is complete, but without that testing, my position
remains unchanged at this point. 

Andrew Bartlett

-- 
Andrew Bartlett                                http://samba.org/~abartlet/
Authentication Developer, Samba Team           http://samba.org