Java8 invokeinterface type-unification bug

TypeUnifailcation.java

package example;

import java.util.Optional;

class Foo {}
interface Bar {
  String barMethod();
}

public class TypeUnifailcation extends Foo implements Bar {
  @Override
  public String barMethod() {
    return "now you see me...";
  }

  public static <T extends Foo & Bar> void win(T t) {
    Optional.of(t)
             .map(x -> x.barMethod()) // invokestatic to anonymous synthetic lambda
             .ifPresent(System.out::println);
  }

  public static <T extends Foo & Bar> void alsoWin(T t) {
    Optional.<Bar>of(t)
             .map(Bar::barMethod) // invokeinterface with type specified by parametricity (Bar)
             .ifPresent(System.out::println);
  }

  public static <T extends Foo & Bar> void fail(T t) {
    // might as well go ahead and compile successfully
    // and just produce invalid byte code
    Optional.of(t)
             .map(Bar::barMethod) // invokeinterface with wrong type (Foo)
             .ifPresent(System.out::println);
  }

  public static void main(String... args) {
    TypeUnifailcation example = new TypeUnifailcation();
    win(example);
    alsoWin(example);
    fail(example);
  }
}

Exactly.