rustdoc: const trait bound modifiers lead to duplicate bounds being render (because it currently strips the modifiers)

[GoldsteinE]

For example, see documentation for the Ord trait from the standard library:

pub trait Ord: Eq + PartialOrd<Self> {
    fn cmp(&self, other: &Self) -> Ordering;

    fn max(self, other: Self) -> Self { ... }
    fn min(self, other: Self) -> Self { ... }
    fn clamp(self, min: Self, max: Self) -> Self
    where
        Self: PartialOrd<Self>,
        //    ^^^^^^^^^^^^^^^^ - this bound is already specified above
    { ... }
}

This is slightly confusing IMO.

Related to #91187

[fmease]

This happens for both locally and externally defined bindings (it's not a reexports issue).
@rustbot label T-rustdoc
@rustbot claim

[fmease]

If anyone would like to work on this, for the local case one solution is – in clean_maybe_renamed_item in branch ItemKind::Trait – to pass the generics & the bounds of the trait to clean_trait_item and to deduplicate the bounds in respect to the maybe-const generics of the respective associated function.

Actually, this is a can on worms: Rustdoc doesn't dedupe normal bounds either (for local / HIR definitions), e.g.

pub trait Tr { fn f() where Self: Eq, Self: Eq; } // `Self: Eq` not deduped
pub trait Super {}
pub trait Sub: Super { fn f() where Self: Super; } // `Self: Super` not deduped

Unassigning myself since I'm prioritizing a rustdoc PR I've been working on for some time.

[GuillaumeGomez]

I'm taking a look then.

[GoldsteinE]

AFAIU ~const doesn’t exist now, so this could be closed?

[GuillaumeGomez]

The underlying issue is still there so keeping it open (you can see it in #105392).