Python 3.6 (only) for 4.10? (was: Re: Samba 4.9rc1 and blocker bugs)

[Alexander Bokovoy]

On ke, 04 heinä 2018, Andrew Bartlett via samba-technical wrote:
> On Tue, 2018-07-03 at 16:35 +0200, Andreas Schneider via samba-
> technical wrote:
> > Hello,
> > 
> > Samba 4.9rc1 will be released while I'm on vacation. I didn't find a release 
> > bug for Samba 4.9 but there is a blocker which I opened:
> > 
> > Support building Samba with Python3
> > https://bugzilla.samba.org/show_bug.cgi?id=13504
> > 
> > Alexander is working on it to get it ready in time and he just has a few 
> > things failing. Thanks for your hard work on that Alexander. There is a 
> > mailinglist threads which he updates with his progress.
> 
> I agree this is important, but we agreed as a team that we do not have
> 'blocker bugs'.   (That is also why there is no release bug).
I think you are mixing up Python3 and waf upgrade. We should have waf
upgrade as a requirement for 4.9. Python3 is a next target.

For waf upgrade I'm close to completing it:
https://gitlab.com/samba-team/devel/samba/pipelines/25041255

This is still using Python 2 but waf 2.0.8 gets us possibility to run
a build with Python 3.

> I'm not hopeful that the many remaining issues around Samba's Python3
> support can be addressed in the next few weeks.  Joe already found more
> issues in Noel's recent patch set (but got dragged onto customer work),
> and that is still just a WIP.  
> 
> Fundamental issues around LDB bytes and string handling and the
> behaviour of str() in practice remain to be sorted out.
> 
> As I say on the bug, the best 4.9 can hope for is that the --without-
> python build builds with python3. 
> 
> Finally, I think we do better to align for Samba 4.10.  Hopefully we
> will have a RHEL8 by then, with Python 3.6+ on the system.  
> 
> While the build can naturally continue to operate with whatever waf
> 2.0.8 is compatible with, I hope we can then set Python 3.6 as the
> minimum Python version for runtime use.  (Backports of Samba using
> Python will need to link to a backported Python3).
> 
> That in turn would allow us for 4.10, if we actually finish, to support
> just one, modern Python at runtime (specifically in the AD DC also the
> testsuite), which would help many, many things.
The goal is to get waf 2.0.8 for 4.9, otherwise we realistically
would be set back for about a year in any python3 effort due to need to
coordinate upgrades of waf code in libtdb/libldb/libtevent/libtalloc
etc.

-- 
/ Alexander Bokovoy