Add extra check to format_impl lint for self.fmt()

[jamesmcm]

As mentioned in this Reddit comment:
https://www.reddit.com/r/rust/comments/tq602q/my_first_clippy_lint_detecting_use_of_self_as/i2fhdy1/

Previously use of self.fmt within the Format trait wasn't detected

Note we don't check fully qualified paths atm, but that shouldn't happen by accident anyway.

changelog: format_impl lint now checks for recursive use of self.fmt() too.

[rust-highfive]

r? @giraffate

(rust-highfive has picked a reviewer for you, use r? to override)

[giraffate]

Would this occur regardless of Display or Debug trait? It appears to be simply a recursive call.

If so, it would be better to implement this as a new lint. It may be similar to #2624.

[jamesmcm]

That's a good point, and yeah that would catch it too and is a more general solution.

The only difficult part there is writing the lint so it works for general functions and trait methods.

[jamesmcm]

It'd be similar to #8422 but that doesn't catch the case of ever having identical recursion e.g. recursing with exactly the same arguments we take in, at least when they are edited in another conditional branch atm.

i.e. it didn't trigger on:

enum Foo {
    Foo(usize),
    Bar(usize),
}

impl std::fmt::Display for Foo {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        match self {
            Self::Foo(_) => self.fmt(f),
            Self::Bar(x) => x.fmt(f),
        }
    }
}

[giraffate]

Thanks!

It might be better to enhance the only_used_in_recursion lint to catch this case.

[bors]

☔ The latest upstream changes (presumably #8794) made this pull request unmergeable. Please resolve the merge conflicts.

[blyxyas]

@jamesmcm I'm going to close this for now (we're doing cleaning of old PRs ❤️) are you interested in continuing development as a fix to only_used_in_recursion?

If you're still interested in this PR, don't hesitate in re-open it and continue your work.