Mypy 0.990 Release Planning

[svalentin]

I am planning to make a 0.990 release around late October.

Please post here any issues you'd like to see fixed in the release, or PRs you'd like to have merged. There is no release branch yet - I will update the issue once it exists.

[hauntsaninja]

Possible things to fix:

• Type Narrowing failure in 0.981 #13800
• no-redef when import-importing a module to satisfy a protocol #13803
• Inconsistent errors reported for protocol matching #13850
• Add a note on variable annotation within unchecked function #13851 (comment)
• Tests fail due to failing compilation of int_ops.c #13819
• mypy 0.982 run 15 minutes (previous version needs only one minute) #13821

[cdce8p]

It would be awesome if these PRs could be included. The first one is already approved

• Add error-code for truthy-function #13686
• Add error-code for truthy-iterable #13762

[ilevkivskyi]

@sobolevn IIUC you are working on LiteralString, right? Will you have time to finish it in time for 0.990?

[sobolevn]

Yes, I will :)

[svalentin]

How are we doing on the potential things to fix for next release list? Do all of them have someone working on them?
I'm thinking of cutting a release branch tomorrow. (We can of course add more stuff after)

[svalentin]

Sorry got carried away. Will make the release branch today!
Looking to release this around Tuesday unless there's any pressing issues.

[svalentin]

Created origin/release-0.990 at ec6d9d9

[hauntsaninja]

Cherry picked 2d70ac0 (useful for typeshed, no user visible changes)

[hauntsaninja]

Cherry picked 5319fa3 (likely addresses performance regression in #13821, no mypy_primer hits, maybe should be solved within mypy rather than in typeshed)

[hauntsaninja]

I think it would be very good (but not necessarily blocking) to get in #13892 cc @JukkaL @sobolevn

False negatives related to #13890 / #13472 show up pretty often (e.g. we just got another report in #13966, where things would silently not work)

Edit: I've now merged this in and cherry picked to release branch

[sobolevn]

Yes, reporing invalid ParamSpec usage seems quite important.

[hauntsaninja]

Okay, I've gone ahead and merged #13892 and cherry picked it to release branch. There are probably still some improvements we can make, but the basic checks should catch many user errors and it seems important we include it in the release.

[svalentin]

Unfortunately we'll have to postpone the release by 2-3 days. Jukka found that mypy crashes if except* is used in python 3.11 (https://peps.python.org/pep-0654/). The thinking is to make it at least not crash so we can announce the start of (basic) support for 3.11 in this release. This way maybe we can get some testing until 1.0.0 version.

[sobolevn]

Ok, I think I won't be able to finish LiteralString support in this release, there's a lot going on right now.

And I don't want to rush it.

Looks like we will have to move it to the next release. Sorry!

[hauntsaninja]

@svalentin I think treating except* exactly like except is pretty viable for "make it not crash". Opened #13991 for this

#12840 contains two other language changes that we should add tests for to make sure mypy can handle (but they're not headline features, so it's not as important and they likely won't crash mypy)

[ilevkivskyi]

FWIW I was going to work on proper except* support this and/or next weekend. But I don't want to postpone release for this, so please go ahead with a minimal no-crash support.

[svalentin]

v0.990 wheels should be on pypi now! Please test it out!
Will publish blog post later today or tomorrow morning if there's no problems.

[svalentin]

Blog post is out: https://mypy-lang.blogspot.com/2022/11/mypy-0990-released.html

[svalentin]

http://mypy-lang.org/ is updated with the link to the new version and blog post!

[sobolevn]

@svalentin thank you for your hard work!

[sobolevn]

And congrats on the release 🎉

[svalentin]

Thank you!
Though it looks like we might need a point release: #14027

[svalentin]

Reopening this for a point release to fix some of the reported issues!

[svalentin]

Will cherry pick fa22a49 for the 0.991 point release.
#14027 is probably something we want to CP as well once we have fix.
If there's anything else, please let me know!

[svalentin]

Not strictly related to this release, but we got to do it now too, with @JukkaL:
mypy blog site mypy-lang.blogspot.com will now redirect HTTP to HTTPS. No more HTTP! This should hopefully improve Google search ranking too.

[intgr]

If #14042 isn't just a fluke, it seems like a pretty important thing to fix.

UPDATE: Fix landed on master.

[intgr]

By the way, https://mypy-play.net still doesn't have mypy 0.990 (and not even 0.982).

[hauntsaninja]

Want to ask over at https://github.com/ymyzk/mypy-playground/issues ? (no one here has control over mypy-play.net)

[hauntsaninja]

Fixes to cherry pick:
#14060 (this one is bad enough that we should make a release sometime in the next couple of days regardless of state of other issues)
#14038

Things worth cherry picking once fixed:
#14027 (PR for this at #14064)
#14046 (PR for this at #14063 )
#14048 (docs, PR for this at #14065)

(feel free to edit this message)

[svalentin]

Yes! #14060 and #14038 are already CPed in the release-0.990 branch. Waiting for the other 3.

[hauntsaninja]

Made PRs for the other three issues!

[ilevkivskyi]

Merged two out of three, and left feedback for the third one.

[hauntsaninja]

Thanks! All are merged now, just cherry picked to release-0.990 branch

[svalentin]

Thanks! I'll go through the release formalities on Monday

[AlexWaygood]

Is there any chance of getting a patch release with #14077 included? #14030 was experienced as quite a bad regression for heavy users of the stdlib ast module (though typeshed is at least partly to blame for the regression 🙂)

[ilevkivskyi]

I don't like this idea. It is not a regression, some people may actually like semantics in 0.990. Point releases are for bad obvious bugs (and crashes).

[AlexWaygood]

I don't like this idea. It is not a regression, some people may actually like semantics in 0.990. Point releases are for bad obvious bugs (and crashes).

If we merged the changes to the ast stubs as suggested in python/typeshed#9130, would you be open to cherry-picking those changes into the patch release?

[ilevkivskyi]

Yes, this is a safer option, unless someone else is against.

[AlexWaygood]

Okay, I merged the typeshed change. It would be great if python/typeshed@f9cd5ee could be cherry-picked to the patch release :)

[svalentin]

I manually included python/typeshed@f9cd5ee into the release branch - the commit 6077d19

[AlexWaygood]

Thanks so much @svalentin!

[svalentin]

v0.991 wheels should be on pypi now!

[svalentin]

Marking this as Done!
Will reopen if we need another point release.