Generic inference from LHS of assignment when destructuring can lead to unexpected `any`s

[jkillian]

Bug Report

🔎 Search Terms

Generic inference, destructuring, binding patterns

🕗 Version & Regression Information

Exists in 4.3.5

• This is the behavior in every version I tried, and I reviewed the FAQ for entries about generics

⏯ Playground Link

Playground link with relevant code

💻 Code

declare function genericFunc<T>(): T;
const { bar } = genericFunc();
// `bar` is `any` here.

🙁 Actual behavior

bar is any and T is inferred to be { bar: any }.

🙂 Expected behavior

The inferred type of T should be unknown and there should be an error from trying to destructure it.
(Or, T could still be inferred as { bar: any } but then noImplicitAny should trigger an error. This seems at a glance less desirable to me than the former, but I don't know enough how generic inference works to make a more informed recommendation. Anyways, some sort of error no matter how it happens is what would be ideal!)

Notes: see #43371 (comment) for the original context.

[andrewbranch]

The inference of { bar: any } may or may not be avoidable without breaking a lot of other inferences, but it should at least error under noImplicitAny.

[andrewbranch]

Related: #43605

[jkillian]

Related to the discussion in #46012 I believe

[andrewbranch]

Yeah. Probably not going to fix the noImplicitAny error for 4.5 because I want to investigate making this whole pattern an error in the 4.6 timeframe, as discussed in that design meeting.