Fix rendering of const keyword for functions

[GuillaumeGomez]

Fixes #44187.

r? @QuietMisdreavus

[QuietMisdreavus]

So there's something weird i noticed when digging into this. I saw that you took out the import for UnstableFeatures, even though i remember associated fns still using some kind of check.

It turns out, associated fns use a different check altogether:

    // FIXME(#24111): remove when `const_fn` is stabilized
    let vis_constness = if is_nightly_build() {
        constness
    } else {
        hir::Constness::NotConst
    };

However, this doesn't stop constness from showing up in the stable std docs. I'm starting to think the check here to make sure we don't print constness on stable builds is superfluous. We try to compile the crate anyway; if it builds, we're on a nightly build (or a bootstrap build), by definition. If we can't hide constness from the stable std docs, then we should just leave it in and treat it like the way we treat other unstable things in the stable docs: give people a chance to look into it and shape it.

cc @rust-lang/docs Pulling in the rest of the team here; what do y'all think about this idea?

[GuillaumeGomez]

Any news in here?

[steveklabnik]

I'm starting to think the check here to make sure we don't print constness on stable builds is superfluous.

👍 from me; we also had previously printed const ness on stable builds.

[QuietMisdreavus]

We talked about this on IRC, and are generally in favor of printing constness all the time. To this end, i'd like it if you took out the check for bare functions, right here so it doesn't even try to mask it (this is the check that doesn't work and always prints it).

[QuietMisdreavus]

travis failure:

[00:32:53] error: unused import: `rustc::session::config::nightly_options::is_nightly_build`
[00:32:53]   --> /checkout/src/librustdoc/html/render.rs:62:5
[00:32:53]    |
[00:32:53] 62 | use rustc::session::config::nightly_options::is_nightly_build;
[00:32:53]    |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[00:32:53]    |

[GuillaumeGomez]

What failure? :p

[QuietMisdreavus]

Would you mind adding something to this test to make sure we're rendering "const" on associated functions too? Something like this is what i have in mind:

// @has foo/struct.Foo.html
// @has - '//*[@class="method"]' 'pub const fn new()'
pub struct Foo(usize);

impl Foo {
    pub const fn new() -> Foo { Foo(0) }
}

[GuillaumeGomez]

Sure.

[QuietMisdreavus]

r=me pending travis

[QuietMisdreavus]

@bors r+ rollup

[bors]

📌 Commit 742ff5a has been approved by QuietMisdreavus