Friday, 21 August 2015

Re: Python 3.5 transition - Update

Previously, I've described our plans for Python 3.5 in Wily and X, and I'd
like to give a quick status/update about where we are, now that we're in
feature freeze.

PEP 478 documents Python 3.5's release schedule upstream. We currently have
3.5 rc 1 in Wily, with two more candidates scheduled upstream, and a final
release in September. Barring unforeseen delays, I expect we'll see 3.5 final
in the final Wily release.

The change to python3-defaults which enables Python 3.5 as a supported version
cleared -proposed just before the freeze and should now be in Wily. There's
still a tail of build failures (not even counting autopkgtests), which are
most easily seen in the PPA I'm keeping up-to-date:

Any help you can provide for these would be greatly appreciated.

Since we're past feature freeze, it's not IMO appropriate to push for making
Python 3.5 the default Python 3 version for Wily. Fortunately we do not need
a mass rebuild to make this change so I plan to do it early in X, and we can
evaluate whether it makes sense to drop Python 3.4 for the X LTS (which would
require a mass rebuild).

The biggest set of build failures against 3.5 are for Django. I had hoped to
be able to back port enough of the Python 3.5 support into Django 1.7 but that
was unfeasible. Django 1.8.3 itself does build fine with Python 3.5, and for
it and all the related packages I've created a dependent PPA here:

There are quite a few build failures, and again, all help in fixing these
would be greatly appreciated. We'll need an FFE to get Django 1.8 into Wily,
but I think it's worth it if we can whip this PPA into shape soon enough.

Note that Debian is currently stalled on the Django 1.8 transition. It's in
experimental but hasn't migrated to unstable mostly due to lack of manpower.
If you care about the state of Django in Ubuntu and Debian, please help!