Traits can be inherited multiple times without rustc throwing an error

[brendanzab]

The code below compiles and executes fine:

use cmp::Eq;

trait Hahaha: Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq,
              Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq,
              Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq,
              Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq,
              Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq,
              Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq, Eq {}

enum Lol = int;

pub impl Lol: Hahaha {
    pure fn eq(other: &Lol) -> bool { *self != **other }
    pure fn ne(other: &Lol) -> bool { *self == **other }
}

fn main() {
    if Lol(2) == Lol(4) {
        io::println("2 == 4");
    } else {
        io::println("2 != 4");
    }
}

Surely rustc should throw an error...?

[catamorphism]

I have a fix, just testing it.

[nikomatsakis]

I am not sure that this is a bug. Depending on the rules we settle on, it might be reasonable to allow (for example) trait ByWordOrByChar : Iterable<~str>, Iterable<char>. I had originally thought to forbid such things but @pcwalton made a good case that they are ok. Also, you can easily get multiple inheritance of the same trait (e.g., Eq) like so:

trait A : Eq {}
trait B : Eq {}
trait C : A, B {}

and trait semantics are designed to ensure that this has a reasonable meaning (as indeed does C : Eq, Eq).

[catamorphism]

My feeling is that while C: A, B as in your example is reasonable, C: Eq, Eq would never be intentional. I'm not sure if there is some situation where automatically generated code would look like this. Until we have a formal spec of the trait system, though, my inclination is to keep the patch. If someone else runs into it trying to do something that should be allowed, they'll make noise :-)