increasing the python requirement (was Re: [rfc] upgrade pqm chroot to lucid)

[Martin Pool]

> To the best of my knowledge, here is a rough survey of which python
> version is used by which OS (at least which one is used on Babune or to
> build the installers, mentioning available 2.x version between parens):
>
> osx 10.5: 2.5 (2.6)
>
> osx 10.6: 2.6
> windows:  2.6
> freebsd: 2.6
> gentoo: 2.6
> hardy: 2.5 (2.4)
> jaunty: 2.6 (2.5, 2.4 probably but not installed on babune)
> karmic: 2.6 (2.5, 2.4)
> lucid: 2.6 (2.5)
> maverick: 2.6 (2.7 status unclear, I had to uninstall it)
> natty: err, no, not under babune yet[1] ;)
>
>    > I'm not sure about the 2.3 series, but I would be perfectly fine if
>    > bzr2.4+ was only restricted to python2.5/(2.6?) compatibility, and we
>    > stopped worrying about python2.4.
>
> Indeed. I think we should decide on this before starting 2.4 so we can
> announce that 2.3 will be the last stable version to support 2.4 (2.5?
> 2.6 ?!?!).
>
> Given the overview above, and considering 2.3 will be released in
> February 2011, I wonder if we can drop 2.5 support (as well as 2.4)
> since hardy is becoming the last remaining platform requiring it.
>
> I don't think we can jump to 3.x but that's still something we should
> think about.

If it's true, and it seems to be true, that every platform important
to our developers is on at least python  2.5, perhaps it's useful to
drop the python 2.4 requirement from the moment we branch off bzr 2.4
in about February 2011.  (Unfortunate coincidence of numbers.)

According to <http://sivang.blogspot.com/2009/11/setting-up-python26-on-ubuntu-804-lts.html>
python2.6 is in a PPA for Hardy if you really need it.

>From memory 2.6 has relatively more relevant new features we'd like to
use than 2.5 does, so there is an incentive to bump our requirement
all the way to 2.6.

Before deciding when to go to 3.0 I'd like to take a spike on actually
doing the conversion, or getting bzr going under 2to3, so that we have
a better measure of how buggy or hard it's going to be.

-- 
Martin