@0x57e11a So my best guess as to what's going on there (although I'm not a Rust expert by any stretch of the imagination) is that the closure can be interpreted/typed in multiple ways, and in the first case, it gets a type hint from the specified return signature so it knows the type to interpret it as, but in the second case it doesn't have that information (as I don't think the type hint propagates backwards through `clos`?) and so it guesses the wrong type.

That doesn't really explain the error message, which I think is just wrong/vague and probably should be filed as a bug (error clarity bug at least). Maybe you can get more insight into what it expects by looking at the supposed type that the language server thinks `clos` has, and see how it differs from the return value signature?

@0x57e11a (Also you could probably force it into the correct shape by providing some sort of explicit type declaration for `clos` but that's... not great, and I'm not sure what declaration that would be exactly)

@joepie91 unfortunately, because the return type is `impl ...`, which is knowns as RPIT (Return Position Impl Trait), it's not something this one could name in a local, as RPIT is an opaque (but concrete) type that implements that trait

ultimately the workaround is to just allocate and have owned parameters, which isn't much when it's ~5x per file

@0x57e11a Yeah, I'm out of suggestions here unfortunately 😔 Pretty much every time I try to work with impl return types I have different but similar issues

@joepie91 probably also doesn't help that there *was* a `#![feature(impl_trait_in_bindings)]` nightly feature for a while, but it got removed due to causing ICEs