Tox: With setuptools dropping python2.6, do we finally have enough precedent to drop it ourselves?

[asottile]

http://setuptools.readthedocs.io/en/latest/history.html#v37-0-0

[nicoddemus]

I'm 👍 on the idea. We can also consider dropping python3.3.

[nicoddemus]

Let's make sure to use the python_requires option in setup.py. Ref: pytest-dev/pluggy#106.

[RonnyPfannschmidt]

please note that tox can drop support for running on said python versions without dropping support for running venvs/tests in them - due to more versatile usage - even in orgs i wonder if a reasonably representative survey can be done

[sigmavirus24]

Frankly, if companies/organizations are so stuck on Python 2.6 that they can't get to Python 2.7 they're probably also paying some company (e.g., RedHat) truck-sized payments to support their long-dead version of Python. I don't think, in that case, the community should be doing this maintenance work for free. And distributors with models similar to RedHat will do backports for customers without the community taking on this burden. I'd be happier to see tox make a clean break with insecure versions of Python that aren't supported by the community.

[RonnyPfannschmidt]

@sigmavirus24 i'm all for dropping support for running on python2.6 - but i wonder how expensive it is to keep running of python 2.6 either in core or in a plugin

[sigmavirus24]

You mean running virtualenv -p python2.6? That's probably not expensive for tox developers but that adds costs to virtualenv deprecating support for 2.6 and let's be real, people testing on 2.6 are already pinning setuptools, pip, wheel and lots of other projects to keep their environment working. They should already be pinning known working versions of everything else, in which case, why do you need to support that case for running tests against py26? Eventually that will have added cost. I'm just proposing you avoid that ever happening 😄

[RonnyPfannschmidt]

@sigmavirus24 true, good point, thanks :)

[The-Compiler]

FWIW the Travis jobs for 2.6 and 3.3 started failing because of pytest dropping support for those.

[gaborbernat]

yeah, most new PRs we get fail on CI. I was thinking that given this we might as well drop support ourselves, and remove from CI. I'll go ahead with this.

[nicoddemus]

Playing devil's advocate, it should be possible to run pytest in 2.6/3.3 if you use pytest-3.2.*.

Having said that, I still think dropping support altogether makes more sense.

[gaborbernat]

Python 2.6 reached its EOL in 2013, four years later I think it's about time we drop support for it.

[sigmavirus24]

@nicoddemus No one really needs a devil's advocate. It's not really constructive in most conversations. And in this conversation, it's equivalent to advocating for team burnout.

[nicoddemus]

No one really needs a devil's advocate. It's not really constructive in most conversations. And in this conversation, it's equivalent to advocating for team burnout.

I certainly didn't mean that, only wanted to mention that pytest-3.2 still works in 2.6/3.3. I apologize if it came across that way.