🤖 User test baselines have changed for fix-leaked-independent-type-params

[typescript-bot]

This test run was triggerd by a request on microsoft#42211
Please review the diff and merge if no changes are unexpected.
You can view the build log here.

cc @weswigham

[weswigham]

@millsp you wouldn't happen to know offhand if this can be worked around trivially in ts-toolbelt or not, would you? I can look into a workaround for my change to avoid triggering these, but this probably indicates that if you caused these types to get their variance measured before, they'd issue this error.

[millsp]

I think (I assume) that it comes from:

https://github.com/millsp/ts-toolbelt/blob/51c9df18bdf0cdeec016ad5715620023f7434897/src/Number/Minus.ts#L20

and

https://github.com/millsp/ts-toolbelt/blob/51c9df18bdf0cdeec016ad5715620023f7434897/src/Number/Minus.ts#L43

which are mistakes. Just for testing, if you can put number instead of N2 on both lines, then (i believe) it should work. Let me know, because there might be other optimizations that I had to work on :) In any case, TypeScript has to move forward, and I will adapt for future releases.

[weswigham]

Yeah, you can definitely trigger these errors in currently shipping typescript (or at least the current nightly) - just force variance calculation of the type by using it in a comparison of some kind (doesn't even have to make much sense):

type Box<X> = {f: MinusNegative<Cast<X, Iteration>, Cast<X, Iteration>, Map>};

const x = null as any as Box<number>;
const y: Box<12> = x;

y will trigger all these Excessive stack depth errors. :(

So if you can fix it, you probably should - it might already be causing headaches for your consumers.

[weswigham]

Changing those two lines to number as you suggested doesn't seem to fix the issue, however.

[millsp]

Will look into it