When a help message is associated with a warning message, the two message you should be printed together and the source code printed only once.

[dzamlo]

For exemple, for the following code:

extern crate libc;

fn main() {
}

The output of rustc is:

bar.rs:1:1: 1:19 warning: use of unstable library feature 'libc'
bar.rs:1 extern crate libc;
         ^~~~~~~~~~~~~~~~~~
bar.rs:1:1: 1:19 help: add #![feature(libc)] to the crate attributes to silence this warning
bar.rs:1 extern crate libc;
         ^~~~~~~~~~~~~~~~~~

It should be:

bar.rs:1:1: 1:19 warning: use of unstable library feature 'libc'
bar.rs:1:1: 1:19 help: add #![feature(libc)] to the crate attributes to silence this warning
bar.rs:1 extern crate libc;
         ^~~~~~~~~~~~~~~~~~

I don't know if there is any help message associated with an error message, but if this the case, the same issue apply.

[caipre]

Looks like the referenced merge addresses error + help combos (albeit not in the order described above):

$ echo 'mod foo;' > test.rs
$ rustc ./test.rs
test.rs:1:5: 1:8 error: file not found for module `foo`
test.rs:1 mod foo;
              ^~~
test.rs:1:5: 1:8 help: name the file either foo.rs or foo/mod.rs inside the directory ""

But this still happens for error + warning combos:

use std::vec::Drain;

fn main () {
}

$ rustc ./test.rs
./test.rs:1:5: 1:20 error: use of unstable library feature 'collections': recently added as part of collections reform 2
./test.rs:1 use std::vec::Drain;
                ^~~~~~~~~~~~~~~
./test.rs:1:5: 1:20 warning: unused import, #[warn(unused_imports)] on by default
./test.rs:1 use std::vec::Drain;
                ^~~~~~~~~~~~~~~
error: aborting due to previous error

Though this may be deliberate since the error and warning are about different issues.

[carols10cents]

(albeit not in the order described above):

So since helps look like they're a subdiagnostic of errors, and errors get output before their subdiagnostics, the ordering looks intentional.

With the error+warning case, I think those are actually two separate diagnostics, that is, the warning is not the subdiagnostic of the error, so they shouldn't be combined.

Sooo I think this issue might be done? If I'm understanding the intentionality of the current error display?

[brson]

Warnings have undergone a lot of churn lately, and some of them have been grouped together in logical ways. @jonathandturner @nikomatsakis @nrc @pnkfelix is this issue still relevant?

[pnkfelix]

I think we can close this. The main redundancy has been addressed, and I think the current handling of help messages works at least for the case described

[nikomatsakis]

Agreed.