scope escape checking for qualified private

[scabug]

I can't figure out for sure from code or spec what is supposed to happen here, but I feel there must be a bug when "object X" is used as the accessible fallback to "X.type".

package privates {
  object A { 
    val b = B
    val c = C
  }
  
  // givn this way it correctly fails
  // private object B
  //
  // privates.scala:4: error: private object B escapes its defining scope as part of type object privates.B
  //     val b = B
  //         ^
  // one error found
   
  // Should this type be inferred for a public val in A?
  // I would think not.  But then the fallback should be AnyRef, or ScalaObject.
  private[privates] object B
  object C
}

package other {
  object C {
    val b = privates.A.b
    val c = privates.A.c
  }
}

This compiles, and look at the types of b and c as inferred in other.C.

// Decides privates.B.type is off limits, so infers "object B"
scala> other.C.b _
res0: () => object privates.B = <function0>

// Doesn't mind inferring privates.C.type
scala> other.C.c _
res1: () => privates.C.type = <function0>

[scabug]

Imported From: https://issues.scala-lang.org/browse/SI-4323?orig=1
Reporter: @paulp

[scabug]

@ingoem said:
This one goes to meeting for your personal entertainment.

[SethTisue]

I can't figure out for sure from code or spec what is supposed to happen here

someone can comment and/or reopen if they have an opinion on this

[som-snytt]

Use case

package p {
  object A {
    val b = B
    val c = C
  }

  private[p] object B
  object C

  class D {
    def f(x: B.type) = x
  }
}

package other {
  object C {
    val b = p.A.b
    val c = p.A.c
    def f = new p.D().d.f(b)
  }
}