"consider adding an explicit lifetime bound `R: 'static" suggestion is misleading

[stepancheg]

Consider an (incorrect) code:

trait Foo {}

struct Bar<R>(R);

impl<R> Foo for Bar<R> {
}

fn bb<R>(r: R) -> Box<Foo> {
    Box::new(Bar(r))
}

fn main() {
    let a = 10;
    let _b = bb(&a);
}

Rustc suggests:

consider adding an explicit lifetime bound `R: 'static`

This suggestion R: 'static is not good enough, code still won't compile.

The problem is that suggestion is misleading, because proper fix should be

fn bb<'r, R : 'r>(r: R) -> Box<Foo + 'r> { ... }

People who don't understand lifetimes well (like me), could stuck after adding 'static bound.

So better suggestion could be something like:

consider adding an explicit lifetime bound `R: 'static`,
or lifetime parameter `<'r, R: 'r> ... -> ... Trait + 'r`

or do not mention 'static at all:

consider adding an explicit lifetime bound to `R`

And also, I think, E0310 could have an example with non-static lifetime bound.

[estebank]

CC #33652