Simple gist to show problem when using companion object extending trait and sub-class using the same trait

trait-conflictor.scala

object MyApp extends App {
  val b = new B()
}

trait Conflictor {
	val log = "reduction"
}

class A {}

object A extends Conflictor {}

class B extends A with Conflictor {}

Btw : It fails in scala 2.9.2, it works in scala 2.10.0. Interesting.

Apparently, in Scala 2.9.2, A makes log final. See javap for A:

public class A extends java.lang.Object implements scala.ScalaObject{
    public static final void Conflictor$_setter_$log_$eq(java.lang.String);
    public static final java.lang.String log();
    public A();
}

Hence the problem for B.

This is an answer, but it may not be right :)

The object A is instantiated where defined. So it defines a val log on itself.

If I remember well, Scala compiles object A so that java code can call A.log (as a static value), so there should be a "log" member in class A calling log in A$ (am I right?)

So this could explain why B cannot extend A with Conflictor, because it tries to 'define' log a second time (there's already one in A)

yes but :

case class ahoy(s:String) {def apply(s:String) = ahoy(s + this.s)}

ahoy("a")("b") == ahoy("ba")

works fine. (there are a lot more collisions here)

So my idea is the compiler is avoiding collision in A, but place unsafe static accessor for log if it can.

This one works :

object MyApp extends App {
  val b = new B()
}

trait Conflictor {
    val log = "reduction"
}

class A extends Conflictor {}

object A extends Conflictor {}

class B extends A with Conflictor {}