Why can't Rust infer this closure lifetime?

[crawdad]

I have some code that implements a trait for all Fn(&str) closures:

trait Target {}

impl<T: Fn(&str)> Target for T {}

fn takes(_: impl Target) {}

fn test() {
    takes(|_| {});
}

The function call fails to compile:

  = note: closure with signature `fn(&'2 str)` must implement `FnOnce<(&'1 str,)>`, for any lifetime `'1`...
  = note: ...but it actually implements `FnOnce<(&'2 str,)>`, for some specific lifetime `'2`

Rust can figure out the type, but it botches the lifetimes. But if I give the parameter a type myself, it works:

takes(|_: &str| {});

Why is that needed? Can I fix the definition somehow so Rust can infer the lifetime properly?

(Playground link)

[SkiFire13]

Type inference of closures with HRTBs involved has always been pretty junky. (The Fn(&str) is equivalent to for<'a> Fn(&'a str), and those for<'a> bounds are called HRTBs, Higher Ranked Trait Bounds)

See Confusing type error due to strange inferred type for a closure argument · Issue #41078 · rust-lang/rust · GitHub

[Yandros]

@crawdad

Can I fix the definition somehow so Rust can infer the lifetime properly?

The only thing you can do in that regard is forgo the Target trait and require an explicit impl Fn(&str) parameter.

[crawdad]

Thanks. I guess I'll have to avoid traits for this then.