Spurious is not a class type error when inheriting a class without call parentheses() through a type alias

[neko-kai]

Compiler version

3.3.1

Minimized code

https://scastie.scala-lang.org/1OfTRArhRuKz5Kq59dUcHg

final abstract class ForcedRecompilationToken[T]
object ForcedRecompilationToken {
  implicit def default: ForcedRecompilationToken["abc"] = null
}

class GoodNoParens[T](implicit ev: ForcedRecompilationToken[T])
type BadNoParens[T] = GoodNoParens[T]

// error
object A extends BadNoParens

// ok
object B extends BadNoParens()
object C extends GoodNoParens

object App extends App {
  println("compiled")
}

Output

GoodNoParens is not a class type

Expectation

Similar situation as in #15976, but this time merely using a type alias to refer to the class breaks the fix for the aforementioned issue and causes a spurious 'not a class type' error.

[odersky]

I think this is as it should be. BadNoParens is a higher kinded type, which is not a legal parent. BadNoParens() is a call to a constructor with inferred type parameters. GoodNoParens is a class type for which special rules apply (i.e. we infer parameters regardless).

I don't really see a strong reason for changing these rules.

[neko-kai]

@odersky BadNoParens is an alias to GoodNoParens, I believe that using it should not be materially different from using GoodNoParens directly, when at all possible. Otherwise some concepts, such as exports, stop working correctly. This bug also reproduces when using an export instead of a type alias - https://scastie.scala-lang.org/veWopnkESKO8aCdUPm352Q

final abstract class ForcedRecompilationToken[T]
object ForcedRecompilationToken {
  implicit def default: ForcedRecompilationToken["abc"] = null
}

object x {
class GoodNoParens[T](implicit ev: ForcedRecompilationToken[T])
}
export x.GoodNoParens as BadNoParens

// error
object A extends BadNoParens

// ok
object B extends BadNoParens()
object C extends x.GoodNoParens

object App extends App {
  println("compiled")
}

This worked correctly on Scala 2 and is currently blocking me from porting existing library code to Scala 3...[1]

[odersky]

OK, trying a fix...

[neko-kai]

@odersky Thank you! 🙏