ParamForwarding: do not require param accessors to be private[this]

[smarter]

Also mark the forwarder as Stable otherwise we get a RefChecks error.

This fixes #608.

Review by @odersky, @retronym .

[smarter]

Note that we do less parameter forwarding than scalac. See for example D and Y in paramForwarding.scala which don't get their own local fields in scalac but do in dotty.

[odersky]

LGTM, thanks!

[retronym]

I had a quick play with this and couldn't find any big holes.

Suggested test variations would include varying repeated / by-name modifiers in the sub/super class, and access modifiers.

I noticed that dotty's tree printer seems not to print the qualifier in qualified protected/private, I was testing that to see if the flags of the constructor param were correctly propagated to the accessor.

I managed to crash the compiler with:

class Outer {
  class A(protected[Outer] val theValue: Any*) {
    val theValueInA = theValue

    def getTheValue = theValue
  }
  class C(override protected[Outer] val theValue: Nothing) extends A(theValue)
}

I also noticed that the @Repeated annotation was propagated to the param accessor, I think this should be removed:

cat sandbox/test.scala && ./bin/dotc sandbox/test.scala
class C(val a: Int*)

class Test {
  def test(c: C): Nothing = c.a
}

sandbox/test.scala:4: error: type mismatch:
 found   : scala.collection.Seq[Int] @Repeated(c.a)
 required: Nothing
  def test(c: C): Nothing = c.a
                              ^
one error found

[smarter]

@retronym : Thanks for looking at this! Would you mind opening issues for the problems you found?

[retronym]

@smarter done

[odersky]

Qualified private is currently not a high priority, because we will most likely get rid of it. In that case it will be a task for the migration tool to find a rewrite.

[samuelgruetter]

So you would rewrite private[C] to just protected[C]?