Scala 2.12：对于普遍量化的SAM特征,Java 8方法参考的等价物是什

在Java中,以下程序返回true：

import java.util.function.Function;
import java.util.function.Supplier;

@FunctionalInterface
public interface Maybe<A> {

  <X> X fold(Supplier<X> empty,Function<A,X> just);

  static <A,X> X empty0(Supplier<X> empty,X> just) {
    return empty.get();
  }

  static <A> Maybe<A> empty() {
      return Maybe::empty0;
  }

  static void main(String[] args) {
    Maybe<?> emptyString = Maybe.<String>empty();
    Maybe<?> emptyInt = Maybe.<Integer>empty();

    System.out.println(emptyString == emptyInt); // print "true".
  }
}

我试图将此编码移植到scala 2.12但它不编译：

@FunctionalInterface
trait Maybe[A] {
  def fold[X](empty: => X,just: A => X): X
}

object Maybe {
  def empty0[A,X](empty: => X,just: A => X): X = empty

  def empty[A]: Maybe[A] = empty0(_,_) // does not compile

  def main(args: Array[String]): Unit = {
    val emptyString: Maybe[String] = Maybe.empty
    val emptyInt: Maybe[Integer] = Maybe.empty

    print(emptyString eq emptyInt) // how to make this print "true"???
  }
}

我得到的错误是：
扩展函数缺少参数类型((x $1：<错误>,x $2：<错误>)=> empty0(x $1,x $2))

我的目标是让scalac触发Javac完成的相同优化,使java程序打印为“true”.只要它不使用asInstanceOf和Nothing / variance注释,我对满足scalac所需的任何东西都是开放的.

编辑：由于目前不支持,我为此开了一个feature request on the scala issue tracker(请投票给它！;-).