No warning on unfulfilled #[expect]

[lnicola]

#[expect(non_snake_case)]
fn main() {
    let _foo = 0;
}

The other half of #18056.

[ChayimFriedman2]

I deliberately did not implement this because our lint handling isn't complete (e.g. I think it is disabled in proc macros for some reason), and also we will need a way to tell whether some lint is (fully) implemented.

[ChayimFriedman2]

This will also be quite hard to achieve: if we have an #[expect(...)] mod foo; declaration in a file, we need to check all lints in foo.rs to determine if we should warn, which means checking foo's children up to the entire crate. Since we don't want a diagnostics request to be that expensive, that means caching the diagnostics of a file, and the only way to cache between requests is in the DB. But if it's in the DB, it cannot be in crates/ide-diagnostics, and all handlers depend on IDE functionality...

[lnicola]

Yeah, it's fine to leave this for some later time which might never come. I filed it pre-emptively because I suspect people will complain anyway.

[Veykril]

Imo this is something we shouldn't even bother implementing. Any small divergence between r-a and rustc could make us emit a false diagnostic from this. "Having to" fallback to cargo for this to be checked seem absolutely fine in my eyes as this is purely a check that your allow is actually doing something.