Typer regression in http4s/http4s

[WojciechMazur]

Based on the OpenCB failure in http4s/http4s - build logs

Compiler version

3.9.0-RC4
Works fine in 3.8.4

Bisect points to 7954ffc / #26161

Minimized code

trait QValue
object QValue:
  val One: QValue = new QValue {}

object LanguageTag:
  def apply(primaryTag: String, subTags: String*): LanguageTag =
    LanguageTag(primaryTag, QValue.One, subTags.toList)

final case class LanguageTag(
    primaryTag: String,
    q: QValue = QValue.One,
    subTags: List[String] = Nil
)

def test(pair: (String, List[String])): LanguageTag =
  pair match
    case (main: String, sub: collection.Seq[String]) =>
      LanguageTag(main, QValue.One, sub)

Output

-- [E134] Type Error: /Users/wmazur/projects/scala/community-build3/http4s.scala:19:6 
19 |      LanguageTag(main, QValue.One, sub)
   |      ^^^^^^^^^^^
   |None of the overloaded alternatives of method apply in object LanguageTag with types
   | (primaryTag: String, subTags: String*): LanguageTag
   | (primaryTag²: String, q: QValue, subTags²: List[String]): LanguageTag
   |match arguments ((main : String), (QValue.One : QValue), (sub : scala.collection.Seq[String]))
   |
   |where:    primaryTag  is a reference to a value parameter
   |          primaryTag² is a reference to a value parameter
   |          subTags     is a reference to a value parameter
   |          subTags²    is a reference to a value parameter

Expectation

Should most likely continue to compile

[SolalPirelli]

@WojciechMazur sorry if I'm being dense or more ill than I think I am, but isn't this failure expected? If you add .toList to sub in the last line it compiles, but without it how could a Seq be passed where a List is expected?

It seems reasonable that this accidentally passed before because we incorrectly typed sub as a List[String] since that's the type of pair._2, but now that we type it according to the case...