#47341 breaks knockout.mapping tests on Definitely Typed

[sandersn]

Knockout.mapping tests rely on the old signature relation rules. The repro is fairly small:

interface Kow<T> {
    (value: T): void;
}
declare var src: Kow<true> | Kow<false>;
var tgt: Kow<boolean> = src;

Expected: src assignable to target.
Actual: Error, type 'boolean' is not assignable to type 'false'.

[ahejlsberg]

Hmm, the repro also fails with TS 4.5.4. Am I missing something?

[ahejlsberg]

Actually, here's a repro that only fails with #47341.

interface Kow<T> {
    (): void;
    (value: T): void;
}
declare var src: Kow<true> | Kow<false>;
var tgt: Kow<boolean> = src;

But really, that ought to fail, it just happened not to previously because we cross-related the signatures. I mean, src might reference a Kow<true> which shouldn't be callable with a boolean.

[ahejlsberg]

And, just to add to the bizarreness of our current behavior, we error if the type is written as an aliased type instead of an interface:

type Kow<T> = {
    (): void;
    (value: T): void;
}

The two obviously should behave the same.

[sandersn]

So it sounds like main is in a better state than 4.5.4, right? If that's the case then this issue should be closed.

[ahejlsberg]

Yes, I'd say main has better and more consistent behavior.