Scala 2.13.8 (was 2.13.6)

[SethTisue]

Not sure what the policy is — is this eligible for 3.1.1, or does it wait for 3.2?

Also not sure if 2.13.7 was skipped for a reason, or just nobody thought of it. Regardless, I've added opening a PR such as this one to the Scala 2 release steps, for next time.

[SethTisue]

Is the test failure something I should just add a checkfile update for?

[som-snytt]

No, the output is wrong:

Type argument C does not have the same kind as its bound [_$$1]

instead of underscore

-- Error: tests/neg/i8752.scala:3:39 ------------------------------------------------------------------------------------------------------------
3 |trait Arround1[C <:[X]=>>IterableOps[X,C,C[X]] ] // error // error
  |                                       ^
  |                                       Type argument C does not have the same kind as its bound [_]

Vulpix not running in partest --show-diff mode is really too bad, as it's not possible to run diff on CI output.

But it's great that the message uses its and not it's.

[SethTisue]

Is this conceivably related to the Scala 2 version bump?

[som-snytt]

As Wallace Shawn famously says in the movie, "Inconceivable!"

[som-snytt]

No, it wasn't caused by this Scala version bump, because it's already broken by bumping to 2.13.7!

[som-snytt]

Eerily, 2.13.7 includes "Fix 2.13.7 regression affecting wildcards and F-bounded types" and happens to break "F-bounds of type parameters may not be type lambdas/Type argument C does not have the same kind as its bound [_]" on the other compiler. This is why the compilers share the least amount of code possible.

[SethTisue]

fyi @lrytz @joroKr21, does this ring any bells?

[som-snytt]

It's due to the wildcard in the sig, but I didn't decide if the change is as intended. I saw the discussion with Lukas, which I'll revisit. I guess Scala 3 will also emit funny-looking temp names in messages? so who cares, just update the check file.

[SethTisue]

I saw the discussion with Lukas, which I'll revisit

link?

in the meantime, I've updated the checkfile.

[som-snytt]

After poking around where I don't yet understand, I guessed scala/scala#9806 or scala/scala#9815 changed the signature (of IterableOps?) seen by Scala 3 so it gets _$$1 instead of _. trait IterableOps[+A, +CC[_], +C]. But if it doesn't ring @joroKr21 's bell, probably I am off base.

[joroKr21]

RefChecks can't change the signature though - it's just checks. Maybe scala/bug#12488 but that applies only to existential types and this is a higher-kinded type right?

[smarter]

It'd be nice to also keep the stdlib213 submodule up-to-date.

[SethTisue]

It'd be nice to also keep the stdlib213 submodule up-to-date

ah, good call, that didn't occur to me.

I've added it to the Scala 2 release steps: scala/scala-dev@1602898

and I've opened #14322

fyi @lrytz

[SethTisue]

Is this being considered for 3.1.x, or will it wait for 3.2?

[smarter]

This PR is part of 3.1.2-RC1 (you can see the first tag a PR was part of at the top of the page on github).

[SethTisue]

duh, sorry. I also should have thought to check this:

Welcome to Scala 3.1.2-RC2 (17.0.2, Java OpenJDK 64-Bit Server VM).
Type in expressions for evaluation. Or try :help.
                                                                                                    
scala> util.Properties.versionString
val res0: String = version 2.13.8