New lint: wildcard pattern in let

[Vlad-Shcherbina]

I think there should be a warning about the wildcard pattern in let.
I can't imagine any situation where this makes sense, and there are cases where it's a mistake caused by taking _ for identifier though it's not.

Wrong:

{
    let _ = my scope guard or whatever
    // dropped right away

    do stuff
}

Correct:

{
    let _g = my scope guard or whatever

    do stuff

    // dropped at the end of the scope
}

[flip1995]

One use case is to prevent warnings on #[must_use] functions.

let _ = returning_result_i_dont_want_to_deal_with()

But this could be implemented as a restriction lint.

[Manishearth]

We could specifically lint on known guard types

[Vlad-Shcherbina]

Many known guards (like std::cell::Ref, std::sync::MutexGuard) are meant to be accessed in scope, so this warning would be unlikely to trigger.
And I ran into this problem with a custom guard type in the first place.

Indeed I forgot that let _ = is idiomatically used to suppress must_use warnings.
I wish there was unused() function in the standard library (same behavior as std::mem::drop()) instead.

---

Two ideas that don't work:

1. Warn iff there is custom Drop impl. The problem is that it will cause false positives on functions returning strings.
2. Warn iff the expression isn't annotated with must_use. The problem is that properly implemented guards are likely must_use.

[lolbinarycat]

i feel like this makes a decent restriction lint. let _ is fine for most codebases, but some may wish to actually use all must_use types, and this enables that (along with forbid(unused_must_use))

[lolbinarycat]

@rustbot claim

[flip1995]

and this enables that (along with forbid(unused_must_use))

unused_must_use doesn't trigger on let _ = must_use();: Playground. So this lint would be a prior step to get this lint to trigger.

[lolbinarycat]

yes, that was my point. this lint (once implemented) would be mainly useful in conjunction with unused_must_use.

i have a partial implementation of this, and i successfully registered the lint.. but for some reason it never produces any output, not even debug prints?

[flip1995]

To add a new lint, using cargo dev new_lint is usually the easiest way. Now that you have down that manually, you still have to register the lint (bottom of the page)