Plans for pdb_ads and auth_netlogond?

[Matthieu Patou]

On 06/28/2012 03:44 PM, Andrew Bartlett wrote:
> On Mon, 2012-06-25 at 08:30 -0700, Jeremy Allison wrote:
>> On Fri, Jun 22, 2012 at 10:54:01PM -0700, Matthieu Patou wrote:
>>> If not let's keep them for 4.0 to give this code a chance (mostly
>>> the same thing that we did with ntvfs with the noticeable exception
>>> that ntvfs allowed a decent AD experience although not perfect).
>>> Can we agree that:
>>> * for 4.1 we (== people interested in keeping this code) would have
>>> introduced some unit tests
>>> * for 4.2 development in (or related) to those module a restarted
>>> significantly
>>> Failing to comply with this would mean that we should seriously
>>> consider the removal of this modules ?
>>>
>>> Comments ?
>> +1 from me. Very reasonable approach - thanks !
> If there is genuine interest in continuing this approach, then certainly
> these modules could be restored by those interested in maintaining them.
> But I think that they should be restored with appropriate unit or
> integration tests, so that other developers making changes across the
> source tree can do so with confidence that they are not breaking these
> modules.  (This is particularly important given that these modules are
> desired to form part of the very foundation of our AD DC).
Well I think that Volker, Simo and Micheal expressed interest for them 
with the closing window for 4.0 coming closer and the work that some 
have with SMB3 I personally don't think that it's fair to require unit 
tests, the only risk without it is that they might not work at all 
anymore very soon and that nobody could be blamed for this. Knowing this 
if someone decide to contribute integration tests before the proposed 
"date" of 4.1 that's good because it should limit the roting effect but 
otherwise I really like to keep comprehensive tests in 4.1 as the 
referee for removal decision.
> I say this because while there has been an number of changes to these
> modules in the past 12 months, but they all have been this kind of
> consequential change, but without tests it seems unlikely that any
> subtle issues accidentally introduced would have been picked up.
>
> Once in, I think that one release cycle would be a reasonable timeframe
> to demonstrate a restart of significant development.
Yeah that's the idea of my plan and I propose to stick to this one if 
you don't mind.

Matthieu.

-- 
Matthieu Patou
Samba Team
http://samba.org