https://tc39.github.io/ecma262/#sec-ordinaryset

This combines steps 3.c.i (overwriting desc with a new descriptor) and the steps under 4. Looks fine.

Hm... looks like this is not fine.

receiver[property] = value will execute setters defined in the receiver's prototype, even though they should not be executed (the setter from super should be executed instead).

Reading spec is hard. So I just add another test case for accessors. This test case pass in V8, WebKit and Firefox so I assume it's correct.

I am confused about this, though, but hopefully this is not broken as it is here. Step 4.d.iv does a [[DefineOwnProperty]] call, but this just directly updates the .value in the descriptor without calling Object.defineProperty.

var o = { foo: 'bar' }; Object.getOwnPropertyDescriptor(o, 'foo').value = 'baz'

is not sufficient to change the value of o.foo.

The helper also just silently does nothing if not writable (needs to throw in strict mode), but solving that is a harder problem.

I think the right way to solve it is also to replace it with basically receiver[property] = value. This will also execute the OrdinarySet algorithm. However, there needs to be a check for 'value' in Object.getOwnPropertyDescriptor(receiver, property) (if there is a pre-existing descriptor). It needs to fail if the receiver[property] is a setter.

Yes, the silent fail in sloopy mode or throw in strict mode for writeable: false and no setter is a hard problem, that's why I do not want to deal with it in this PR. Let's just do one thing in one time.

IMHO DefineOwnProperty is not the same with receiver[property] = value.
Think this example.

class A {
}

class B extends A {
    constructor() {
        super()
        super.x = 3
        console.log(this.x)
        console.log(super.x)
    }

    set x(v) {
        console.log('i am invoked!') // will not invoked
        this._x = v
    }
    get x() {
        return this._x
    }
}
new B()

Good catch!

This needs writable: true, enumerable: true, configurable: true?