1.9.0 Release Tracking Issue

[jhance]

I will be cutting the release branch today (truly a January release!) and probably actually release sometime next week.

Please comment here if there are any outstanding PRs you wish to ensure are in the release branch. I will cherry-pick them if they are not merged before I cut the branch.

[AlexWaygood]

We haven't done a typeshed sync in a while; we might want to do one before the next release. The automated one two weeks ago failed due to merge conflicts with the cherry-picked commits: https://github.com/python/mypy/actions/runs/7522542054

[cdce8p]

I would really love to see this one included. It's a simple one line fix (+ tests), that just needs a review.

• Fix missing type store for overloads #16803

This here would be nice as well, although I need to investigate one small issue first

• Fix disallow-any errors for Instance types (PEP 696) #16832

[jhance]

#16844 for typeshed sync. Thanks for pointing that out.

[hauntsaninja]

Could be nice to have #15043 in 1.9

[jhance]

The release branch is up (delayed due to typeshed 3.7 issue). So far it looks pretty good - I've tested it on our internal codebase and all the errors seem like they are the fault of our code, not of mypy.

I don't think we are going to block on #15043 for this release, because while the PR itself is apparently ok, it doesn't actually fix the issue completely that its trying to address. Releases are monthly (and this one, like all the ones I do, is late) so we could probably get it + more fixes in for next release.

[hauntsaninja]

It might be worth cherry picking python/typeshed#11353

[AlexWaygood]

It might be worth cherry picking python/typeshed#11353

No need — it's already in the release-1.9 branch (we did two typeshed syncs in quick succession)

[cdce8p]

Not sure if it's worth mentioning yet. The release will include support for basic TypeVar defaults PEP 696. Would be great to have some more early testers. An example from the PEP

class Bot: ...

BotT = TypeVar("BotT", bound=Bot, default=Bot)

class Context(Generic[BotT]):
    bot: BotT

class MyBot(Bot): ...

reveal_type(Context().bot)         # type is Bot  # notice this is not Any which is what it would be currently
reveal_type(Context[MyBot]().bot)  # type is MyBot

Important

Not yet working / supported

• Defaults for TypeVarTuples
• Defaults which reference other TypeVars
• The new syntax but I think that is obvious as mypy doesn't support PEP 695 yet.

[cdce8p]

Opened two cherry-pick PRs for 1.9:

• [Release 1.9] Stubtest: ignore a new protocol dunder (#16895) #16899
• [Release 1.9] Unsupport targetting 3.7. (#16883) #16900

[cdce8p]

@jhance What's the current status? Anything we can help with? AFAICT it's just removing the +dev version suffix and the blog post. Would be awesome to get the release out soon 🔥🚀

• [Release 1.9] Remove +dev from version #16930

[jhance]

Unfortunately I was distracted by other problems at work, followed by on PTO, followed by being sick. But I am back now. And opening a PR for the changelog now. (The good news is that we actually updated our pin internally to upgrade for Dropbox a few weeks ago, so the harder part was already done).

[JelleZijlstra]

The release went out today, thanks!

#17001 reported an issue with the self-check, which was fixed on master (#16949), but not in the release branch. It's probably not worth a bugfix release on its own, but if we do make a 1.9.1, it would be good to include that fix.

[JelleZijlstra]

For similar reasons, #16928 might be worth backporting (see #17005).

[hauntsaninja]

#17003 fixes a minor regression in 1.9

[JelleZijlstra]

#17015 reports another regression.

[DanielNoord]

Could it be that the git tag was misnamed? All versions use vX.X.X whereas 1.9.0 doesn't include the v. I don't see anything in the changelog about this.
This probably breaks some workflows that use the naming scheme of the tags to build from source (this is how I found out as well).

[svalentin]

Closing this issue since the 1.9.0 release is done!

[kloczek]

Could it be that the git tag was misnamed? All versions use vX.X.X whereas 1.9.0 doesn't include the v. I don't see anything in the changelog about this. This probably breaks some workflows that use the naming scheme of the tags to build from source (this is how I found out as well).

Yeah it broke mine 😋
My automation relies on exact versioning pattern.
I'm not against tagging patter changes but it would be good to know is that new pattern or just typo on tagging 😃