HandleKind::Interface now carries an ObjectImpl, because JS cares wha…

[mhammond]

…t the impl is

@bendk I think this is what we discussed - as you'll see in the patch, I think that either using ObjectImpl is wrong, or what this patch needs more work so that it always gets the precise variant in the places where today it doesn't know what it is.

Seems to work with js, although I've not actually tested it 😅

[bendk]

These changes look good to me, although I feel like some of my changes in the last commit might have eliminated the special cases that we needed these for in the first place.

In any case, more information doesn't seem bad, maybe some other bindings will want it. For that reason I like using ObjectImpl rather than only distinguishing interface vs trait.

[mhammond]

For that reason I like using ObjectImpl rather than only distinguishing interface vs trait.

But it seems like at least a couple of places in callback_interfaces.rs doesn't actually know whether it is a ::Trait or a CallbackTrait - so languages which care about the distinction might get it wrong?

(At least It think both these variants use vtables in the same way?)

[bendk]

But it seems like at least a couple of places in callback_interfaces.rs doesn't actually know whether it is a ::Trait or a CallbackTrait - so languages which care about the distinction might get it wrong?

Ahh, i didn't realize that. In that case, I'm for a boolean.

[mhammond]

Ahh, i didn't realize that. In that case, I'm for a boolean.

I ended up just adding a new variant - wdyt?

[bendk]

Looks good to me. I like the new variant.