Consider the following situation:
trait M { } // uses N
object M { def apply = new M { } }
trait N { self: L => }
trait L { val m: B }
type B
M needs new trees definitions that are created in N. I don't want to pass around these new trees because the class hierarchy is in fact very long and would result in modifying a lot the code. Instead I was proposed to do:
trait M { self: N with L => } // uses N
object M { def apply = new M with N with L { } }
trait N { self: L => }
trait L { val m: B }
type B
but then I get error:
object creation impossible, since value m in trait L of type B is not defined
this is coming from the definition of object M. I have tried different combinations and none works.
How can I solve this error?
Discussion
I learn that when writing new M {} I get an anonymous class extending (object??) M, so I tried modifying to new M with N with L {} but then I get the error of unimplemented members from L.
References
You may find more details on the problem on this commit. Namely, type M is MeasureAnnotation, N is StructuralSize and L is SolverProvider.
If you want to build the program you should git clone and then run sbt followed by universal:stage. You need to have installed z3 or cvc4.
I think your code snippet does not accurately reflect your actual commit. In your commit you used a self type on the object, too, which does not make sense. A self-type on a trait describes a requirement that concrete instances mixing in this trait must satisfy. However, an object is already a concrete type, so there is no point in defining further requirements.
In short: do not repeat the self-type from a trait on its companion object.