Mypy 1.6 Release Planning

[koogoro]

I'm planning to make the 1.6 release by the end of August.

Please post any issues you'd like to see fixed in the release here, or any PRs you'd like to have merged. I'll make the release branch release-1.6.

[koogoro]

I've run into a regression: #15932.

[tyralla]

If this is the right place to bring attention to old, so far undiscussed pull requests: #14440

[AlexWaygood]

It would be good to get #15960 (or an equivalent fix) in before the release. It fixes some recent stubtest regressions that I introduced.

(EDIT: Now merged! :)

[ilevkivskyi]

@koogoro I have a suggestion on the blog post: Could you please mention there that --new-type-infernece will be on by default in 1.7?

[koogoro]

The branch has been created. The new functionality in the attrs plugin to cover fields might cause some errors to pop up for users of attrs but this isn't really a regression and I don't think the primer noticed anything, so I've marked this as a feature request: #15980.

[koogoro]

@AlexWaygood That commit didn't make it onto the branch, and it also doesn't cherry-pick cleanly. Does that mean that the change that introduced the issue didn't make it onto the branch, or would it be good to include multiple more commits on the release branch?

[cdce8p]

@koogoro Thanks for creating the release branch. I noticed that it branched of over a week ago. Was that intentional? It's missing a lot of the recent PRs especially from @ilevkivskyi. If you're not too far along with testing it, might I suggest hard reseting the branch to the current master commit? 010da0b That would also include the stubtest fix @AlexWaygood mentioned.

https://github.com/python/mypy/commits/master
https://github.com/python/mypy/commits/release-1.6

[koogoro]

@cdce8p The wheels build failed for the latest commit, so I'll use the one before it, d7b2451. Let me quickly check if there are any large problems before hard resetting it.

[AlexWaygood]

@AlexWaygood That commit didn't make it onto the branch, and it also doesn't cherry-pick cleanly. Does that mean that the change that introduced the issue didn't make it onto the branch, or would it be good to include multiple more commits on the release branch?

I was hoping 0b303b5 would make it into the release, and if it does, it's important d7b2451 does also, as it fixes some regressions introduced by 0b303b5. But if neither commit makes it in, that's also okay, it can wait :)

It would be pretty useful if 48835a3 could be included in the release, though — this one fixes a crash that's blocking some typeshed updates.

[koogoro]

I did find one regression in d7b2451: #15984. This regression's new in between that and the current release point, maybe due to some ParamSpec change between the two?

[hauntsaninja]

It would be good to cherrypick #16004 (fix for #14740 which was highly requested config option). You may also need #15992 for CI to work

[hauntsaninja]

We should cherrypick #16013 in case we don't have a release before Python 3.12 is out

[koogoro]

Cherry-picked 0ae0c75 as a fix for #15984 and 379b52f as a fix for #15990. I'll cherry-pick #16004 and #16013 once they've been merged.

[Hnasar]

Can we please get 010da0b for #15975 ? (This bug was surfaced by a change in 1.5 4d394c1)

[hauntsaninja]

Both #16004 and #16013 are merged and should be cherry picked.

Quick note about #16004: it's extremely commonly requested, so we do need to do it, but possibly it might be a little annoying to upgrade for projects that use type ignore with error codes and that use --warn-unused-ignore. If we're concerned about that, some options are #16004 (comment). If we want to figure it out in release 1.7, then we should maybe revert #16004, #15840, #14740 on the 1.6 branch.

[cdce8p]

Small change to better handle the --new-type-inference flag. It would make sense to cherry-pick that as well IMO.

• Clear cache when adding --new-type-inference #16059

[Dreamsorcerer]

it might be a little annoying to upgrade for projects that use type ignore with error codes and that use --warn-unused-ignore

As a maintainer of 1 of the projects in the mypy primer diff, I'm fine with making these updates manually. More precise ignores seems like a good thing. We often have to make updates to remove ignores etc. as mypy gets better anyway.

[mr-c]

Getting #16124 in for mypy 1.6 would fix a regression that prevents cwltool (59k downloads a month on PyPI, total 4M PyPI, 475k conda) from upgrading mypy since 1.4.0.

[hauntsaninja]

@koogoro let me know if there's something I can do to help the release!

[ilevkivskyi]

@JukkaL @koogoro What is the situation with the release? What is currently the blocker?

[JukkaL]

I will work on finishing the release. I hope to get it out by early next week. I'm planning to start working on the 1.7 release soon afterwards.

[JukkaL]

I worked on the release blog post today, and hope to complete it tomorrow. After the blog post is done, we should be good to publish the release. If anybody has any other things needed before the release, please leave a comment.

[ilevkivskyi]

Just one small reminder on the blog post:

I have a suggestion on the blog post: Could you please mention there that --new-type-inference will be on by default in 1.7?

[cdce8p]

I mentioned it earlier but AFAICT it hasn't been cherry-picked yet. #15919 (comment)

Especially for the --new-type-inference flag, cherry-picking this PR would make sense IMO.

• Clear cache when adding --new-type-inference #16059

[mr-c]

@JukkaL please also cherry pick #16124 as per #15919 (comment)

[ilevkivskyi]

I think these two mentioned by @hauntsaninja are important

Both #16004 and #16013 are merged and should be cherry picked.

In particular, we must fail gracefully on Python 3.12 that was just released.

[JukkaL]

Cherry-picked these PRs to the release branch:

• Clear cache when adding --new-type-inference #16059
• Fix mypyc regression with pretty #16124
• Match note error codes to import error codes #16004
• Make PEP 695 constructs give a reasonable error message #16013

[JukkaL]

And the release is out: https://mypy-lang.blogspot.com/2023/10/mypy-16-released.html

[JukkaL]

I will update the website tomorrow, since it's getting late here.

[AlexWaygood]

A new crash with mypy 1.6, that I hit when trying to bump typeshed's mypy pin:

• [mypy 1.6 regression] Crash involving ParamSpec, TypeAlias and overloads #16245

[intgr]

@ymyzk Can you please add mypy 1.6.0 to mypy-play.net?

[ymyzk]

@intgr Yes, thanks for the reminder. I added mypy 1.6.0 and updated mypy-play.net ymyzk/mypy-playground#488

[intgr]

I experienced two regressions with mypy 1.6.0, both involving ParamSpec:

• [regression] Files with mixed strict-optional forget None'able argument (with ParamSpec decorator) #16246
• [regression] Function wrapped with ParamSpec[Concatenate[]] is not compatible with Callable[..., Any] #16247

[ilevkivskyi]

An obvious candidate for 1.6.1 #16251 (fixes two closely related crash regressions).

[AlexWaygood]

Another new crash in mypy 1.6 (that still reproduces with mypy master and the latest version of the release-1.6 branch):

• INTERNAL ERROR when using typing.ParamSpec #16257

[AlexWaygood]

#16258 fixes a regression of sorts, so might be worth including in a patch release? But the regression was a new in mypy 1.5 rather than mypy 1.6.

[ilevkivskyi]

Also CP for 1.6.1 #16259

[intgr]

FYI mypy website https://mypy-lang.org/ does not display the 1.6 release in "What's new"

[hauntsaninja]

I cherrypicked these to release-1.6 branch and will cut the release later today:

6b6504eb8 Fix crash on ParamSpec unification (for real) (#16259)
eb81e63e9 Fix crash on ParamSpec unification (#16251)

[hauntsaninja]

https://pypi.org/project/mypy/1.6.1/ is now a thing

I also left some notes at https://github.com/python/mypy/wiki/Release-Process#making-a-patch-release so I don't have to grep through my shell history next time to see if I forgot something