Release 0.540 planning

[JukkaL]

We are going to switch to more frequent mypy releases. Our new target is ~3 weeks between releases. (Vacations, travel and other distractions will cause the actual time between releases vary.)

Timeline for the next release:

• I'll cut the release branch on Friday, October 13.
• The tentative release date is Wednesday, October 18.

Please add to this issue PRs and fixes that you think need to be included in the release, but please be reasonable. If there's no PR yet we likely won't have time to finish the review by Friday -- but on a more positive note, the next release after 0.540 won't be too far away!

[ilevkivskyi]

Among my PRs:

• Allow subscripted aliases at function scope #4000 (simplifies/unifies rules for type aliases)
• Fix --incremental crash on Type[...] #4038 (fixes metaclass attribute lookup and corresponding --incremental crash)
• Check code after inferred UninhabitedType #4059 (check code after inferred UninhabitedType, this fixes several potential false negatives)
• Allow assignments to multiple targets from union types #4067 (multiassign from union, an oldest high priority issue plus a bonus crash)
• Push correct enclosing class while deferring a method #4073 (a recent regression)

Plus (strict-)optionally:

• Last strict optional fixes #4070 (depends on Narrow types after 'in' operator #4072) it fixes no issues, so probably is not urgent, but nice to have.

[ilevkivskyi]

On a second thought last two are probably too tight for Friday, let's focus on other five.

[JelleZijlstra]

It would be nice to have some form of #4076 in, since it fixes a nasty bug in config file parsing (even mypy's own config was affected).

[gvanrossum]

I am hoping to come up with a decision about the lxml stubs -- can we fix them, should we remove them (they can be revived as a separate project once @ethanhs's PEP 561 is implemented). @JelleZijlstra do you have a strong opinion here?

[JelleZijlstra]

Don't feel strongly about it. Maybe waiting for PEP 561 makes sense.

[elazarg]

#4016 (correct subtype check for self argument in an attribute/property access)

[gvanrossum]

Could a fix for #4097 be included? It seems to be a regression that appeared in 0.530.

[ilevkivskyi]

@gvanrossum

Could a fix for #4097 be included? It seems to be a regression that appeared in 0.530.

I am working on it now. If everything goes well (i.e. if I will have more time today), then I will make a PR by the end of the day.

[JukkaL]

Could a fix for #4097 be included? It seems to be a regression that appeared in 0.530.

Yes, unless it's particularly risky. Fixes to regressions may be cherry picked even after the release branch has been cut.

[gvanrossum]

I've cut the release branch: https://github.com/python/mypy/tree/release-0.540. We can cherry-pick the fix for #4097 if it's ready and deemed safe.

Of the suggested PRs, these didn't make it in and I don't expect them to be cp'ed:

• Check code after inferred UninhabitedType #4059
• Last strict optional fixes #4070
• remove all lxml stubs for now typeshed#1664 (removing lxml stubs)

There will be another release about three weeks later.

[ilevkivskyi]

Whether #4070 gets in the release or not, I would prefer it reviewed and merged in master soon, since it is prone to merge conflicts.

[gvanrossum]

I just talked to Jukka and we're going to do the release tomorrow from what's currently in the release-0.540 branch (also current master).

[JukkaL]

The new release date is Friday instead of Wednesday to give it more time to be tested internally at Dropbox. I'm not expecting any additional changes.