Confusing notes with per-method where-bounds when bounds are not satisfied

[nagisa]

The following code:

trait Fruit {
    type Error;
}

struct Peach;
impl Fruit for Peach {
    type Error = ();
}

struct Banana<F>(F);
impl<F> Banana<F> {
    fn peel(&self) -> Result<(), F::Error> where F: Fruit {
        unimplemented!()
    }
}

fn pineapple() -> Result<(), ()> {
    let banana: Banana<&Peach> = Banana(&Peach);
    banana.peel()?;
    Ok(())
}

playground

outputs the following error:

error[E0599]: no method named `peel` found for struct `Banana<&Peach>` in the current scope
  --> src/lib.rs:19:12
   |
10 | struct Banana<F>(F);
   | -------------------- method `peel` not found for this
...
19 |     banana.peel()?;
   |            ^^^^ method not found in `Banana<&Peach>`
   |
   = note: the method `peel` exists but the following trait bounds were not satisfied:
           `<&Peach as Fruit>::Error = _`

The note in this error is confusing as it only leads further away from the solution and is also difficult to read. In particular it is talking about trait bounds, but the <&Peach as Fruit>::Error = _ presented is not even a trait bound. The real problem here is that F: Fruit is not satisfied.

The error message improves significantly when the bound is moved to the impl generic instead as such:

impl<F: Fruit> Banana<F> {
    fn peel(&self) -> Result<(), F::Error> {
        unimplemented!()
    }
}

We should strive for a similar note when the where-bound is applied to method like in the original example.

[estebank]

error[E0599]: the method `peel` exists for struct `Banana<&Peach>`, but its trait bounds were not satisfied
  --> src/lib.rs:19:12
   |
10 | struct Banana<F>(F);
   | ---------------- method `peel` not found for this struct
...
19 |     banana.peel()?;
   |            ^^^^ method cannot be called on `Banana<&Peach>` due to unsatisfied trait bounds
   |
   = note: the following trait bounds were not satisfied:
           `&Peach: Fruit`