[SIP-71] `into[A]` type wrapper lost in mirror generation

[prolativ]

Compiler version

3.7.3-RC1-bin-20250619-b1da8fc-NIGHTLY

Minimized code

import scala.language.experimental.into
import Conversion.into

import scala.deriving.Mirror

object Opaques:
  opaque into type MyInto[+A] >: A = A

import Opaques.MyInto

case class Foo(x: Int)
case class Bar(foo: into[Foo])
case class Baz(foo: MyInto[Foo])

given Conversion[Int, Foo] = Foo(_)

def test = 
  val barMirror = summon[Mirror.Of[Bar]]
  summon[barMirror.MirroredElemTypes =:= (into[Foo] *: EmptyTuple.type)] // Doesn't compile unexpectedly
  summon[barMirror.MirroredElemTypes =:= (Foo *: EmptyTuple.type)] // Compiles unexpectedly

  val bazMirror = summon[Mirror.Of[Baz]]
  summon[bazMirror.MirroredElemTypes =:= (MyInto[Foo] *: EmptyTuple.type)] // Compiles as expected
  summon[bazMirror.MirroredElemTypes =:= (Foo *: EmptyTuple.type)] // Doesn't compile as expected

Expectation

into type should be preserved in mirror generation

[odersky]

I think it's better not to preserve it for mirror generation. into is strictly for the external parameter interface. We erase it for the other info of methods and classes. So I argue this is as it should be. I'll add a test that checks that the behavior stays as reported.

[prolativ]

Is there any other possible situation when there's a mismatch between the type of a parameter of a case class and the corresponding type in the generated mirror? IMO this is very unexpected. The presence or absence of into in a constructor might be valuable information for typeclass derivation and I think it should be preserved.

[odersky]

I believe the contrary. Typeclass derivation should not care about this.

[prolativ]

One example of such mismatch that came to my mind now is varargs.
E.g. for

case class Foo(xs: Int*)

we get

type MirroredElemTypes = Seq[Int] @Repeated *: EmptyTuple.type

[odersky]

I will not have time to deal with this any further. If you think there should be a change make a PR against both spec and implementation and we can discuss it. But I believe we are wasting time here.

[prolativ]

OK, let's keep the behaviour as it is given that it's well documented