NoSuchMethodError when creating abstract override val

[Atry]

Welcome to Scala 2.12.2 (Java HotSpot(TM) 64-Bit Server VM, Java 1.8.0_102).
Type in expressions for evaluation. Or try :help.

scala> :paste
// Entering paste mode (ctrl-D to finish)

class A(override val f: Float) extends B

trait B {
 def f: Float = 1.0f
}

trait C extends B {
 abstract override val f = super.f + 100.0f
}

trait D extends B {
 abstract override val f = super.f + 1000.0f
}

new A(10.0f) with C with D {}

// Exiting paste mode, now interpreting.

java.lang.NoSuchMethodError: C.f$(LC;)F
  at $anon$1.$line3$$read$D$$super$f(<console>:29)
  at D.$init$(<console>:24)
  ... 30 elided

[Atry]

The same code doesn't crash 2.11.11

[lrytz]

This goes deep into the corner cases of mixin and fields.

Note that in 2.11.11, (new A(10.0f) with C with D {}).f is 1000f. Not sure if this is a bug, it's a result of the wiring of the C$/D$_super_$f and C$/D$super$f accessors generated in the anonymous class.

// from 2.11
public final class Test..anon extends A implements C, D {
    private final float f;

    public float f() {
        return this.f;
    }

    public /* synthetic */ float D$$super$f() {
        return this.f();
    }

    public void D$_setter_$f_$eq(float x$1) {
        this.f = x$1;
    }

    public /* synthetic */ float C$$super$f() {
        return super.f();
    }

    public void C$_setter_$f_$eq(float x$1) { }

    public Test..anon() {
        super(10.0f);        // sets the field in A
        C$class.$init$(this);  // gets `10` through C$$super$f(), adds 100, call C$_setter_$f_$eq, which discards it
        D$class.$init$(this); // gets `0` through D$$super$f, adds 1000, calls D$_setter_$f_$eq, stores in Test$anon.f
    }
}

In 2.12, the C$$super$f method attempts to invoke super[D].f, which is translated in the backend to INVOKESTATIC D.f$, but this method doesn't exist because the getter D.f is abstract. We didn't see any super calls to abstract methods before, so this is not special-cased. But also, the getter symbol D.f doesn't have the deferred flag.

[Atry]

You are right. It would make more sense if the result is 1110.0f.

[lrytz]

@adriaanm might be able to tell what is the expected/spec'd behavior

[SethTisue]

#notfixedindotty