Please lint on `a.eq(b.into())` if `a.eq(b)` would work

[joshtriplett]

In the course of investigating crater breakage, I discovered many instances of a pattern that reduces to this:

fn main() {
    let s = String::from("hello");
    assert!(s.eq("hello".into()));
}

This compiles, presumably because "hello".into() has only one possible impl that would work. It breaks if another From impl is brought into scope. For instance, cargo add bstr and add use bstr as _; to the top of that example, and it breaks.

Could we lint against this?

cc @estebank; this is in the same spirit as #129249 .

[matthiaskrgr]

Doesn't https://rust-lang.github.io/rust-clippy/master/index.html#useless_conversion already cover this?

[joshtriplett]

@matthiaskrgr It does, but that's a clippy lint. I'm suggesting a rustc one, so that crates in the ecosystem are much more likely to encounter it and fix it.

That said, #129249 will probably cover this. @estebank, would your lint implementation catch this?

[estebank]

@joshtriplett I believe it could, I need to try and get it working again to confirm.

Edit: can confirm that it would be picked up by that lint:

error: this conversion is useless `&str` to `&str`
 --> bar.rs:3:18
  |
3 |     assert!(s.eq("hello".into()));
  |                  ^^^^^^^^^^^^^^
  |
  = note: `#[deny(self_type_conversion)]` on by default