super in class extending null should disallow super call in constructor

[bterlson]

It presently keys off of ClassHeritage being present but if it's present but null it should also be an early error. /cc @bmeck

[bterlson]

I'm also not entirely sure why we didn't make super in base class methods an early error (rather than a TypeError by way of MakeSuperPropertyReference). @allenwb thoughts?

[bterlson]

I'm also not entirely sure why we didn't make super in base class methods an early error (rather than a TypeError by way of MakeSuperPropertyReference). @allenwb thoughts?

[bakkot]

@bterlson, given that it works in object literals, it would be at least a little surprising if didn't work in base classes.

e.g.

''+(new class {
  toString(){
    return 'my ' + super.toString();
  }
}); // "my [object Object]"

[bakkot]

@bterlson, given that it works in object literals, it would be at least a little surprising if didn't work in base classes.

e.g.

''+(new class {
  toString(){
    return 'my ' + super.toString();
  }
}); // "my [object Object]"

[bterlson]

@bakkot right, thanks!

So presumably we could make super in methods for class extends null an early error?

[bterlson]

@bakkot right, thanks!

So presumably we could make super in methods for class extends null an early error?

[bterlson]

@ajklein points out that [[HomeObject]]'s prototype is determined dynamically, so the prototype could change.

[bterlson]

@ajklein points out that [[HomeObject]]'s prototype is determined dynamically, so the prototype could change.

[bakkot]

Are we talking about the literal null, vs e.g. class extends (()=>null)() {}? Because as I understood it ClassHeritage could evaluate to null at runtime - but we can't have early errors for that.

[bakkot]

Are we talking about the literal null, vs e.g. class extends (()=>null)() {}? Because as I understood it ClassHeritage could evaluate to null at runtime - but we can't have early errors for that.

[ljharb]

Thanks to @ajklein my understanding is now that if [[HomeObject]] ever became changeable, which was previously possible, that extends X could always change from null to non-null, so super access can't be an early error.

[ljharb]

Thanks to @ajklein my understanding is now that if [[HomeObject]] ever became changeable, which was previously possible, that extends X could always change from null to non-null, so super access can't be an early error.

[bterlson]

@bakkot yeah you're right, early error is only possible with the null literal (unless we throw early-ish during CDE, but I wouldn't argue for that I guess).

[bterlson]

@bakkot yeah you're right, early error is only possible with the null literal (unless we throw early-ish during CDE, but I wouldn't argue for that I guess).