"Unused imports" warning has false-positives in the presence of syntax extensions

[catamorphism]

For example, code with deriving(IterBytes) seems to need a use core; in scope. however, this causes a warning: unused import. Removing the import causes a resolve error, though.

Or maybe the answer is for syntax extensions to not require a certain module to be in scope... I'm not sure.

[catamorphism]

Nominating for milestone 5, production-ready.

[alexcrichton]

The emission of unused imports changed a fair amount in #6093, so this may be fixed by that. Is there a simple case that I can test one? I tried:

#[deriving(IterBytes)]
enum A{
  B, C, D
}

fn main(){}

and I didn't get any errors. This makes me think that it may only pertain to things with #[no_core] ?

[catamorphism]

Hmm, maybe you do need #[no_core]. I'll try to make a test case.

[catamorphism]

Yeah, @alexcrichton seems to be right. This only happens inside rustc. If you want to test it, try removing the use core; directive from librustc/middle/ty.rs.

rustc doesn't have #[no_core], though. I still don't know what's going on...

[alexcrichton]

This doesn't look to be an error, just an unfortunate stage0 thing. If you add #[cfg(stage0)], the warning goes away. The reason for this is that the #[deriving(IterBytes)] attribute used to generated code of the form:

#[deriving(IterBytes)]
enum A { B, }
#[doc = "Automatically derived."]
pub impl ::core::to_bytes::IterBytes for A {
    pub fn iter_bytes(&self, __lsb0: bool, __f: core::to_bytes::Cb) {
        match *self { B => { 0u.iter_bytes(__lsb0, __f); } }; }
    }

whereas now it spits out

#[deriving(IterBytes)]
enum A { B, }
#[doc = "Automatically derived."]
pub impl ::core::to_bytes::IterBytes for A {
    pub fn iter_bytes(&self, __arg_0: ::bool, __arg_1: ::core::to_bytes::Cb)
     -> ::bool {
        match *self { B => 0u.iter_bytes(__arg_0, __arg_1) }
    }
}

Which if you notice on the second argument, the path was changed from a relative path to an absolute one, which would explain why in later stages the use core isn't needed, but in stage0 it is.

[catamorphism]

Ah, ok. If it's a "this'll go away with the next snapshot" thing, then I'll be happy :-)