PartialOrd docs for generic tuples show the where clauses in a seemingly random order

[jruderman]

On https://doc.rust-lang.org/std/cmp/trait.PartialOrd.html:

The docs for PartialOrd on tuples show the where clauses in a weird order, obscuring the fact that the last element of the tuple is allowed to be unsized.

impl<A, B, C, D, E> PartialOrd<(A, B, C, D, E)> for (A, B, C, D, E) where
    C: PartialOrd<C> + PartialEq<C>,
    E: PartialOrd<E> + PartialEq<E> + ?Sized,
    B: PartialOrd<B> + PartialEq<B>,
    A: PartialOrd<A> + PartialEq<A>,
    D: PartialOrd<D> + PartialEq<D>, 

[GuillaumeGomez]

Found the issue: it's because we use a BTreeMap using Symbol as key. Problem being that Symbol is making integer based comparison and not string based, hence the current result. Trying to write a test now.

[camelid]

Note that this bug is specific to cross-crate re-exports; the same documentation in core renders correctly. There also seems to be another "bug" related to cross-crate re-exports and where clauses: In the core docs, the impl looks like this (which is how it's actually written in the source):

But in the std docs, the impl looks like this, with all of the bounds expressed as where clauses:

I think the re-exported version is actually better, because it's easier to read, but they should at least be consistent.