overflow during drop-check on a recursive type

[wolfiestyle]

I get a pretty strange error when trying to compile this: (playground link)

enum Stuff<T> {
    Value(T),
    Nested(Box<Stuff<Stuff<T>>>),
}

fn main() {
    let a = Stuff::Value(1);
}

output:

error[E0320]: overflow while adding drop-check rules for Stuff<i32>
 --> src/main.rs:7:9
  |
7 |     let a = Stuff::Value(1);
  |         ^
  |
  = note: overflowed on Stuff<Stuff<Stuff<Stuff<Stuff<Stuff< ...

If the type isn't used, the program compiles fine. Don't know if the type is actually representable or not, the error is misleading.

[sfackler]

How is it misleading? Stuff::<i32>::Nested contains a Stuff<Stuff<i32>>, which contains a Stuff<Stuff<Stuff<i32>>>, which contains a....

[Ixrec]

I wouldn't call it misleading either, but "overflow while adding drop-check rules" certainly isn't as clear and to-the-point as "Stuff<T> is unrepresentable" would be. It also seems bad that we don't emit this error until the type gets used somewhere.

[pnkfelix]

(I am debating about whether the I-hang label applies here. In my head, stack overflow errors are usually just the result of hangs being caught via very crude means...)

[pnkfelix]

Interestingly, the overflow here doesn't trigger an ICE?

[oli-obk]

Diagnostics triage: P-low, as it's fairly clear what's going on, just not a great diagnostic message