Release 0.630 planning

[ilevkivskyi]

I would like to release mypy 0.630 on Friday September 7 September 14. The release branch is going to be was cut on Tuesday September 4 Friday September 5. Please mention here any issues and PRs that you would like to include in the release.

Here is the list of release blockers:

• Importing packages can poison the mypy cache. #5534
• Mypy fails on Union[float, int] + Union[float, int] #2128
• Use package_data to ship typeshed and xml files #5517
• --disallow-incomplete-defs does not print errors when run without this argument previously #4913
• Add a customize_class_mro plugin hook #4567
• Perf regression caused by #5474 #5514 (last remaining blocker)
• Regression since 0.620: Argument 1 to "__div__" of "int" has incompatible type "float"; expected "int" #5557
• Remove the 'rop will not be called' error message #5571
• Regression caused by #5565 #5573
• -2 and --python-version 2.7 arguments differ in behaviour #5576 (Is this really important?)
• mypy error trying to check pymc3 #5560 (we don't have time for this and this is not a regression)
• False positive for overriding decorated method in an import cycle #5548 (ditto)

[ilevkivskyi]

I think these two should be fixed before the release: #5534 and #2128

[emmatyping]

I would like to get #5517 in, as it makes finding typeshed a lot simpler and more robust.

[Michael0x2a]

I was just about to mention #2128, actually -- I'm currently in the process of writing some code to resolve that issue and hope to submit a PR either later today or sometime tomorrow.

[srittau]

Is there any chance to have #4913 fixed in the release? Guido's comment seems to indicate it is an easy fix, although I can't judge this.

[gvanrossum]

Maybe #4567? It's a plugin hook for MRO calculation that some user would like to see. I think it's unobjectionable, though it currently has a non-trivial conflict (you edited the same code in your most recent fix or MRO-related issues).

[ilevkivskyi]

I think it also makes sense to fix #5514

[JelleZijlstra]

In general, it would be good to sync typeshed right before creating the branch, so that people's typeshed fixes make it in. Maybe that could be part of the standard release procedure.

[ilevkivskyi]

Maybe that could be part of the standard release procedure.

This is in our internal team playbook. I will make a sync on Monday probably.

[gvanrossum]

A regression: #5557.

[Michael0x2a]

This is cutting it close to the deadline (sorry!), but it'd be nice if #5571 could also be merged in. (It partially addresses some of the feedback regarding error messages in #5557).

[ilevkivskyi]

#5573 should be fixed before the release.

[gvanrossum]

I suspect #5576 may be worth fixing?

At this point we should probably consider postponing the release by a few business days.

I'd also like to see a consolidated list of all issues and PRs that are blocking the release.

[ilevkivskyi]

I'd also like to see a consolidated list of all issues and PRs that are blocking the release.

I updated the description with a list.

I suspect #5576 may be worth fixing?

Why is this important? This release is already late, I would limit ourselves to only critical things.

At this point we should probably consider postponing the release by a few business days.

Yes, I think we can do it on Monday-Tuesday. However, if there are no additional things I would cut release branch tomorrow.

[gvanrossum]

OK, I see #5576 is not that important -- it's not a regression AFAICT.

[gvanrossum]

Then it looks like there are no blockers left. Can you cut the branch tonight?

[ilevkivskyi]

Then it looks like there are no blockers left. Can you cut the branch tonight?

Hm, indeed. OK, I will do this now.

[ilevkivskyi]

OK, release branch is cut.

[ilevkivskyi]

After some thinking we decided that we should first fix #5514 and postpone the release by few days again -- new target date is Friday, September 14.

(I updated the issue description.)

[srittau]

Please have a look at python/typeshed#2456 as well. (Although I am not sure whether the changes causing this are in the release branch.)

[ilevkivskyi]

@srittau The PR that caused this didn't get into the release branch, but we anyway decided to revert it.

[Michael0x2a]

I'm not sure if we've started the release process yet or not, but if there's still time, could we perhaps sneak in #5595? It's almost entirely a docs change -- the only actual change in the code is a small tweak to the text we print out when running mypy --help.

If this is too late/too much trouble, that's totally fine -- this is a pretty low-priority thing. I'm bringing this up mostly because I think it would be a little cleaner if we could include all of the command line/config file stuff in one release instead of across two. (So this PR would be released along with #5333 and #5334.)

[ilevkivskyi]

We are kind of deep into the release process, in the sense that everything is ready, but the actual release was already postponed twice for some reasons. If everything will be OK, I will upload packages to PyPI on Monday morning. There is no big need for doc updates, since the website is anyway built from master. I mean you can already see your changes on https://mypy.readthedocs.io.

[Michael0x2a]

Only if you visit https://mypy.readthedocs.io/en/latest/ -- if you visit https://mypy.readthedocs.io/en/stable/ (which is what mypy-lang.org links to), you won't see the changes.

But anyways, no worries -- I figured I might as well try asking.

[ilevkivskyi]

The release is out!

[gvanrossum]

Congrats! Thank you for all your work on this.