NTSTATUS trick for NTTIME?

[Andrew Bartlett]

On Wed, 2011-03-09 at 08:36 +0100, Volker Lendecke wrote:
> On Wed, Mar 09, 2011 at 06:12:33PM +1100, Andrew Bartlett wrote:
> > I'm not as confident as you that this is a small change however, and
> > would like to see the full extent of the patches (or an analysis of it)
> > before we agree to the change.  
> 
> If it was a small change, I would have just put it in. I am
> well aware that this is intrusive, that's why I posted the
> question on samba-technical first.

Thank you. 

> It was meant as an incremental approach. I will not be able
> to provide a full conversion patchset before this is being
> accepted. You might have noticed that I used the typedef to
> nt_time_t in my patch proposal to enable co-existence and
> gradual conversion.
> 
> That seems unacceptable, so lets better stick with the
> current, flaky approach.

What I'm asking for is an idea of the eventual scale. If you are willing
to work on the patch in parts, then that's great, but as the full
benefit is only available on full implementation (when the type changes
and NTTIME no longer exists) I'm cautious and would like to see some
metrics on the scale of that.

> > I strongly disagree on this point.  We must never again introduce Samba3
> > or Samba4 only types, or special coding rules.  Any rules we choose to
> 
> We have this, and I don't see it go away. See for example
> talloc_tos(), something that is not acceptable in Samba4.

I should clarify my position on this, so you may make proposals here
under an easier light.  I would be very happy to consider patches to
Samba4 that introduce the correct hooks for talloc_tos() in the event
loop and libraries, including at the library interfaces that OpenChange
uses.  I've not had the time or interest to take on this challenge at
the moment, and so some small instances of talloc_tos() use have been
re-factored out of code that is made it common, as that has been the
smaller task at the time. 

I regret that I have not made my changed position on this matter as
clear as I could have.  

> See for example nested event loops, something which is not
> acceptable in Samba3. We just have different coding styles
> and need to find ways to live with it.

I think we need to, and in the interests of our users must, do better
than simple co-existence. 

Andrew Bartlett

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