October 2021 Release

[uranusjr]

Tracking issue for pip 21.3.

Anyone volunteers to be the release manager @pypa/pip-committers?

[sbidoul]

I can not be the release manager for this one: October being the month of the annual Odoo release I'll be too busy in the Odoo community. I may take January or April.

I'll be around to help, and at least handle any issues related to editables, though.

[pfmoore]

I might be able to make time for this, but only if it's a relatively routine release. I definitely won't be able to support multiple followup bugfix releases, so if we aren't OK with taking a hard line with "it'll be sorted in 22.1 and not before", then I'd rather not take this on. IMO, we've got into a habit of releasing too many bugfix releases, and it's not practical to sustain given the level of resource we have.

Are we comfortable with a release on a weekend early in Oct, shipped from whatever is on main at that point, with no commitment to bugfix releases for anything beyond complete showstoppers?

[pradyunsg]

I'm happy to pick up the baton for this time around, given that everyone else has limited availability. :)

I expect there to be a bunch of activity and fallout in this release, since we're going to ship in-tree-build-by-default, will likely try to ship editables support, and have a bunch of removals (there's a few gone_in="21.3" deprecations) as well.

[pfmoore]

OK, cool. I still think we should as a matter of principle try to cut back on the number of bugfix releases we do. Either that, or we may start to find that a 3-month release cycle isn't sustainable (we've barely finished doing 21.2 bugfix releases and we're starting on 21.3).

Maybe we need to have a separate issue to review our release policy?

[pradyunsg]

Maybe we need to have a separate issue to review our release policy?

I'm ambivalent. This works, so does a separate issue. :)

I still think we should as a matter of principle try to cut back on the number of bugfix releases we do. Either that, or we may start to find that a 3-month release cycle isn't sustainable

I tend to agree.

FWIW, I'm also fine with deferring that decision to whomever is cutting the releases. I personally prefer to keeping the issue open as short as possible, actually getting the first release on a week day in the first half of the month and so on -- all of which was tricky with the resolver rollout. And I was full-time working on pip in that period. :)

Anyway, for 21.2, we had a whole bunch of follow ups on the sysconfig <-> distutils migration and the resolver -- but the issue stayed open for quite a while after the final release "just in case", which... meh, it's fine but we can also get away with not doing that.

[uranusjr]

I think the sysconfig thing needs to stay open for a while since there are still some obscure issues popping up once in a while. Or alternatively, since we’ve already committed to the switch in 3.10 and those cases are obscure anyway, we can completely silence the warning in 21.3 and see what happens after 3.10.

[pradyunsg]

but the issue stayed open for quite a while after the final release "just in case"

I was referring to the release issue, not the sysconfig migration-related reporting location. :)

And, definitely agree that we should keep the warnings around for a while. Ambivalent on what to do for 3.10+ -- I'm inclined to keep them as-is, until we drop distutils entirely from 3.12. 🤷🏽