Re: python 2.x

[Greg Troxel <gdt%lexort.com@localhost>]

coypu%sdf.org@localhost writes:

> Another pkgsrc post-branch, another discussion.
> python2 will become EOL upstream in a few months.

Over 8 months!  The EOL date is January 1, 2020:
  https://www.python.org/dev/peps/pep-0373/

So from the having-maintained-software-as-default viewpoint, together
with our norms about making potentially destabilizing changes early in
the branch cycle, it is necessary to change the default to some 3.x
choice by October 15.  Other options are nowish and July 15.

> Let's get a 3.x version as the default.

Separately from the deadline, it's a fair question to ask if the time is
now for the switch.

Are you proposing 3.7?  Or some other version, perhaps 3.6?  What's the
rationale for choosing among 3.x?  (Not complaining; I am just not clear
on what's the best choice.)

Do we have an idea of what the consequences would be?  Do we have any
platforms where 2.7 builds and 3.x does not?   While pyNN-foo gets built
multiversion, some things pick one, and hence depend on that version.
Do we know how much would break?

The end of 2.7 has been coming for a really long time.  Maybe this
really is it.  I am getting the impression that substantially all
maintained software (python modules) supports 3, and that larger
programs in python are getting there.  For example, weewx is not there
yet, but I don't see pkgsrc defaulting to 37 causing trouble for using
weewx on 27.

I expect fairly rapid progress in the upstream world making things work
with 3.x over the rest of the year, and therefore decreasing pain from
us switching the longer we wait.  But the pain may already be quite low.

Is anybody up for doing a bulk build with the default changed?