Inclusive range patterns are always output as ..., which is linted against.

[jebrosen]

At https://github.com/dtolnay/syn/blob/master/src/expr.rs#L3706, ... is output even if the input was ..= in range patterns. It looks like this was decided on in #275.

In rust-lang/rust#61342 the use of ... was upgraded to a warn-by-default lint.

Reproduction example:

#[proc_macro]
pub fn passthru(input: TokenStream) -> TokenStream {
    let parsed: syn::Item = syn::parse(input).unwrap();
    let tokens = quote! { #parsed };
    tokens.into()
}

passthru!{
    fn a() {
        match 1 {
            2..=3 => {},
            _ => {},
        }
    }
}

warning: `...` range patterns are deprecated
  --> consumer2018/src/main.rs:23:14
   |
23 |             2..=3 => {},
   |              ^^^ help: use `..=` for an inclusive range
   |

The error is especially confusing because the Span is preserved.

Discovered via rwf2/Rocket#1033.

[dtolnay]

I would accept a PR to emit ranges as ..= now that the new syntax is well supported.

[jebrosen]

I'd be more than happy to work on adding that.

What is syn's policy on version support for these cases? IOW, does it need to be done conditionally based on whether the running rustc supports ..= in patterns?

[dtolnay]

Ah, good thought! I would be okay with either way. It would also be fine to make the behavior conditional later only if the simple change turns out to be disruptive.