`class extends null` is broken by ECMAScript

[shinenelson]

class extends null was reverted by ECMAScript in tc39/ecma262#781 ( tc39/ecma262@c57ef95 ) and is still broken.

The kata is abnormal since it doesn't have a solution in the test itself.
So, the question is whether we should have that test for class extends null?

However, I found a possible solution thanks to devdocs.io ( which had a previous version of the MDN Web Docs; revision ) :

class NullClass extends null {}
assert.equal(Object.getPrototypeOf(NullClass), Function.prototype);

[wolframkriesing]

if i read the spec (I am not sure if I read the wrong part) it looks like its is possible
https://tc39.es/ecma262/#sec-runtime-semantics-classdefinitionevaluation
in 5e it says "If superclass is null, then" and MDN also still says "The .prototype of the extension must be an Object or null."

Do I read those wrong?

[wolframkriesing]

though i dont get it work anywhere neither, here chrome (canary) console

> class A extends null {}
undefined
> new A
VM79:1 Uncaught TypeError: Super constructor null of A is not a constructor
    at new A (<anonymous>:1:1)
    at <anonymous>:1:1

and node repl

> class A extends null {}
undefined
> new A
Uncaught TypeError: Super constructor null of A is not a constructor
    at new A (repl:1:1)

[wolframkriesing]

yeah the only useful thing to learn is what you described. But not really of any practical value, is it?

So I would rather change this test to verify that a parent must be callable with super() and since null has no super it throws. Does that makes sense?

    it('a class extending `null` can not be instantiated, since calling `super()` on null fails', () => {
      class NullClass extends null {}
      assert.throws(() => new NullClass);
    });

like this

[shinenelson]

I'm not really sure. Wouldn't that be kind of misleading new developers to assume that a class may extend null?

We should either remove the whole test or explain the rationale behind the use-case. And also somehow probably document somewhere ( in the test itself ) that this is a gray-area test.

Do I read those wrong?

No, what you're reading is correct. The spec was first made to extend null, but then apparently many bad things happened and they reverted it ( see links in issue description )

I found this question on StackOverflow that questions this use-case altogether. The question has different errors of this use-case not working

PS : the issue title is probably misleading, what I meant was that it was broken by ECMAScript and that it wasn't our fault.

[wolframkriesing]

i agree, either 1) remove the test or 2) make it more explicit, maybe like that:

it('`extends null` is not supported', () => {
      class NullClass extends null {}
      assert.throws(() => new NullClass);
});

just to mention the edge case and its solution.
What do you think?

[shinenelson]

actually, it is supported by the spec, it's just not working as intended. So, maybe, the test should be extends null is broken or something along those lines maybe?

[shinenelson]

the MDN revision used the term 'not allowed', so, I guess 'not supported' is close enough. I'd favor 'not allowed' though.