Possible 4.3 regression with generic function and rest in object destructuring

[OliverJAsh]

Bug Report

🔎 Search Terms

🕗 Version & Regression Information

• This changed between versions 4.2 and 4.3 beta

⏯ Playground Link

Playground link with relevant code

💻 Code

type Params = {
    foo: string;
} & ({ tag: 'a'; type: number } | { tag: 'b'; type: string });

const getType = <P extends Params>(params: P) => {
    const {
        // Omit
        foo,

        ...rest
    } = params;

    return rest;
};

declare const params: Params;

switch (params.tag) {
    case 'a': {
        // TS 4.2: number
        // TS 4.3: string | number
        const result = getType(params).type;

        break;
    }
    case 'b': {
        // TS 4.2: string
        // TS 4.3: string | number
        const result = getType(params).type;

        break;
    }
}

🙁 Actual behavior

4.3 behaviour should match 4.2 behaviour. See comments in code.

🙂 Expected behavior

See comments in code.

[OliverJAsh]

This is marked as fixed but the behaviour in 4.3.2 still differs from 4.2 as per my comment above. Is that expected?

[OliverJAsh]

@weswigham Not sure this issue has been fixed, should we re-open it?

[weswigham]

It definitely looks fixed in the nightly playground... and in the associated PR's tests...

[OliverJAsh]

I had a look at the tests but I couldn't understand what behaviour they were trying to verify. Do we expect my example to behave the same in 4.3 as it did in 4.2, or is this an intended change? In my testing the behaviour in 4.3 and nightly differs from that of 4.2, as per my code comments.

[weswigham]

I see the behavior match what you noted as the 4.2 behavior in the nightly playground?

[OliverJAsh]

Yes you're right, sorry!