Release 20.0

[pradyunsg]

New Year, New Release!

We've had >600 commits since the last release (yay?) so I'm not able to use the GitHub UI to "quickly" figure out a summary of the changes since the last tag. :)

As usual, we need to pick our release manager first, then hand off to them to take care of the release.

---

Notes for the release manager:

• The release process should be significantly easier now, thanks to some automation: https://pip.pypa.io/en/latest/development/release-process/#creating-a-new-release.
• If you're on Windows, please test-drive the prepare-release and build-release commands before the actual release. I don't have a Windows machine currently, to test things on. :)
• If something about the automation isn't working, mention @pradyunsg (or drop me a mail, because my GitHub notifications Inbox is a dumpster fire).

[pradyunsg]

I'm happy to be the RM for this release (and the next 2 or 3, as well but we'll get to that when the time comes), because I know I'll be available.

[pradyunsg]

Given that no one seems opposed to me being the RM (holler here), I'll self-assign myself as the RM for pip 20.0. 😉

Looking at my calendar, I'll be with @uranusjr this upcoming week and can make the release the week after that.

Let's go with 21st Jan as a "soft target" date for the release. :)

[brainwane]

@pradyunsg @uranusjr Per the conversation happening about the opt-in flag and #5727, I think it's likely we'll want to incorporate some part of this feature, enough to bootstrap our early testers, into the 20.0 release. Can we do that by the end of January?

[pradyunsg]

I'm not sure what's being requested here. Is this a request to delay the release? Is the motivation to add the opt-in flag, that's currently being discussed, as a part of this release? I don't think we gain much by exposing a no-op option -- is there a specific task that we know this would help with (maybe a user testing guide?)?

All that aside, if we do want to introduce the flag ASAP, I don't see any reason we can't have a 20.0.1 release that introduces such a flag a week after 20.0 (other than PyPI bandwidth cost reasons).

[brainwane]

I'm not sure what's being requested here. Is this a request to delay the release?

It's a request for information; could we add the relevant functionality by the end of January, if we wanted to?

Is the motivation to add the opt-in flag, that's currently being discussed, as a part of this release?

Yes.

I was under the impression that we were trying to avoid making a 20.0.1 release; if you (as 20.0 release manager) are fine with introducing the flag in a 20.0.x release then that solves the problem and you don't need to answer my request for information, and can ignore it. :-)

[pfmoore]

It's a request for information; could we add the relevant functionality by the end of January, if we wanted to?

I'm not the RM, but yes, we could do so (add a flag that said what resolver to use, which only allowed the "legacy" option for now, and so in effect did nothing). Doing that is pretty trivial. I'd be willing to write that PR if needed.

What I've done for now is open #7603 to discuss what such a flag would look like, and to thrash out details like whether there is any point having the flag in advance of having more than one resolver to choose. If we want to, we can make that feature a release blocker.

[pradyunsg]

could we add the relevant functionality by the end of January, if we wanted to?

Yep.

Honestly, I'm fine either way - if we're able to get everything into master, before EoD 20th IST, all good.

Otherwise, we'll make another release if we need to. I'm personally expecting that we'd have at least 1 bug fix release anyway (given how many changes are in master right now) and, yes, I'm fine with making a release that just adds the no-op flag.

[pradyunsg]

It's release day!

I'll be starting work to make the release in a few hours -- if anyone has any blocking concerns, please holler ASAP. :)

[webknjaz]

Yaaay!

[pradyunsg]

Alrighty. The release has been made.

• Uploaded files to PyPI.
• Updated get-pip.py.
• Announced on:
  • discuss.python.org
  • distutils-sig + pypa-dev

Looks like we already have our first user reports of problems popping up: #7217.

[messa]

Is it possible to roll back? See #7620 please.

Thank you for releasing 20.0.1 quickly :)

[k3dr1]

There is a problem with installing the packages (throws an Import error) when I try it like this "pip install tensorflow". So anyway, how do I roll back?

[pradyunsg]

In case someone wasn't looking for the past 2 hours, we had a major-ish outage. pip 20.0 was basically not-usable. pip 20.0.1 has been released with a hotfix.

#7624 improves our automation to prevent similar breakages in the future. :)

[pradyunsg]

I'll make a bugfix release that updates packaging with the fix for #7626 tomorrow. Yay Fridays.

[pradyunsg]

Okie then -- merged the fix for #7626 into master.

I'll go eat some snacks now, and then come back to do the make-a-release stuff, in about half an hour or so.

[vkataev]

I'll go eat some snacks now, and then come back to do the make-a-release stuff, in about half an hour or so.

Thank you for your great valuable work! Mistakes can happen and it's expected sometimes things may misbehave :) At least a workaround exists and hopefully 20.0.2 will resolve it.

[pradyunsg]

That took a bit longer than expected -- working on the release now. :)

---

Thanks for the kind words @vkataev!

[pradyunsg]

pip 20.0.2 has been released:

• Pushed commits and tags directly to pip's master branch.
• Uploaded files to PyPI.
• Updated get-pip.py.

[pradyunsg]

Coming back to this release cycle: looks like we're only waiting on a fix for #7629 (there's an open PR w/ changes requested for the past week: #7655).

That'd be the bugfix release 20.0.3, which would be the last release in the 20.0 cycle. :)

[pradyunsg]

If there's no progress/updates on #7629 in this week, I'm gonna defer it to pip 20.1 (April) and call this release cycle done-and-dusted.

[pradyunsg]

Alrighty! Looks like there's still a bunch more work needed for #7629; by updating / breaking up #7655 and no one has stepped up for doing it. I'm gonna say that it's missed the train for this release cycle, and defer it to 20.1 (April release).

[pradyunsg]

With that, it looks like that's all for this release cycle! Thanks to everyone who contributed / worked with us on this release! ^>^