new lint: Recursive `Display` impl

[oherrala]

It's possible to implement Display trait in such a way which recursively calls itself until stack is exhausted. There's probably more variations how to encounter this bug. I provide one example.

Consider the following code:

use std::fmt::{Display, Formatter, Result};

struct Foo;

impl Display for Foo {
    fn fmt(&self, f: &mut Formatter) -> Result {
        write!(f, "{}", self)
    }
}

fn main() {
    println!("{}", Foo);
}

Compiles perfectly fine, but there's stack trace at run time:

$ cargo run -q

thread 'main' has overflowed its stack
fatal runtime error: stack overflow
Abort trap: 6

The bug is in line write!(f, "{}", self) which recursively calls itself by trying to use Display trait implementation for self.

[clarfonthey]

Huh, it's funny that this doesn't trigger the unconditional recursion lint upstream. I'd consider this to be a compiler bug.

[Manishearth]

I suspect that's because the formatting machinery goes through some intermediate functions.

[clarfonthey]

Ah right, the lint only checks one layer of function calls, so fn a() { b() } fn b() { a() } doesn't lint.

[mimoo]

any update here :) ? This would be an awesome lint!

[flip1995]

There is the to_string_in_display lint, which already checks for one case. This lint could be enhanced to also check for things like using the formatting machinery inside the Display implementation.

[giraffate]

It looks like that the recursive_format_impl covers this, so I'm closing this.