Deriving can't be used within macros because it generates duplicate impls

[bstrie]

This:

macro_rules! define_vec (
    () => (
        mod foo {
            #[deriving(Eq)]
            pub struct bar;
        }
    )
)

define_vec!()

fn main() {}

...doesn't compile:

moon.rs:4:23: 4:25 error: conflicting implementations for a trait
moon.rs:4             #[deriving(Eq)]
                                 ^~
moon.rs:4:23: 4:25 note: note conflicting implementation here
moon.rs:4             #[deriving(Eq)]
                                 ^~
moon.rs:4:23: 4:25 error: conflicting implementations for a trait
moon.rs:4             #[deriving(Eq)]
                                 ^~
moon.rs:4:23: 4:25 note: note conflicting implementation here
moon.rs:4             #[deriving(Eq)]
                                 ^~
error: aborting due to 2 previous errors

Taking a look at the expanded code:

mod foo {
    #[deriving(Eq)]
    pub struct bar;
    #[doc = "Automatically derived."]
    pub impl ::std::cmp::Eq for bar {
        pub fn eq(&self, __arg_0: &bar) -> ::bool {
            match *__arg_0 { bar => match *self { bar => true } }
        }
        pub fn ne(&self, __arg_0: &bar) -> ::bool {
            match *__arg_0 { bar => match *self { bar => false } }
        }
    }
    #[doc = "Automatically derived."]
    pub impl ::std::cmp::Eq for bar {
        pub fn eq(&self, __arg_0: &bar) -> ::bool {
            match *__arg_0 { bar => match *self { bar => true } }
        }
        pub fn ne(&self, __arg_0: &bar) -> ::bool {
            match *__arg_0 { bar => match *self { bar => false } }
        }
    }
}


fn main() { }

Notice that it's generating a second identical impl for the given trait. I've tested that this happens with both Eq and ToStr, so I presume it applies to the deriving code in general.

[huonw]

IIRC, this actually happens with any syntax extension, but someone should double check. (I.e. when #[auto_encode] was around, it had the same behaviour.)

[emberian]

Still relevant

[huonw]

Triage: cool! This appears to have fixed itself. @cmr, could you bisect it? I don't remember anything that would've affected this.

(Tagging as needstest.)

[emberian]

@huonw Something somewhere in the let hygiene fixed it.

[huonw]

Ah, cool; makes sense.