param accessors should drop @ Repeated from the type.

[retronym]

cat sandbox/test.scala
class C(val f: Any*) {
  def g = f
}

class Test {
  def test1(c: C): Nothing = c.f
  def test2(c: C): Nothing = c.g
}

% scalac sandbox/test.scala
sandbox/test.scala:6: error: type mismatch;
 found   : c.f.type (with underlying type Any*)
 required: Nothing
  def test1(c: C): Nothing = c.f
                               ^
sandbox/test.scala:7: error: type mismatch;
 found   : Seq[Any]
 required: Nothing
  def test2(c: C): Nothing = c.g
                               ^
two errors found

% ./bin/dotc sandbox/test.scala
sandbox/test.scala:6: error: type mismatch:
 found   : scala.collection.Seq[Any] @Repeated(c.f)
 required: Nothing
  def test1(c: C): Nothing = c.f
                               ^
sandbox/test.scala:7: error: type mismatch:
 found   : scala.collection.Seq[Any] @Repeated
 required: Nothing
  def test2(c: C): Nothing = c.g
                               ^
two errors found

I believe that both dotc and scalac are wrong here to give the param accessor the type Any*/ Any @Repeated. Instead, the repeated-ness of the type should be dropped, in the same way that it is when computing the type of the explicity defined accessor g.

Followup from #617, /cc @smarter

[retronym]

Note: By name constructor parameters are not allowed to to have the val modifier (or, equivalently, be used as case class params), otherwise we would need to test that out as well.

[odersky]

I don't see why. @Repeated is an annotation. The type for @Repeated method parameters gets transformed into an ExprType. But I see nothing wrong with leaving the annotation on the Seq type.

[retronym]

Fair enough. I can't really remember raising this ticket :).

What's the meaning of the Repeated annotation? It isn't documented in the dotty codebase AFAICS. Does it have an analog in scalac?

[odersky]

#1080 adds some documentation to Repeated and other classes.