Allow using generic trait methods in const fn

[jonas-schievink]

Next step for #67792, this now also allows code like the following:

struct S;

impl const PartialEq for S {
    fn eq(&self, _: &S) -> bool {
        true
    }
}

const fn equals_self<T: PartialEq>(t: &T) -> bool {
    *t == *t
}

pub const EQ: bool = equals_self(&S);

This works by threading const-ness of trait predicates through trait selection, in particular through ParamCandidate, and exposing it in the resulting ImplSource.

Since this change makes two bounds T: Trait and T: ?const Trait that only differ in their const-ness be treated like different bounds, candidate winnowing has been changed to drop the ?const candidate in favor of the const candidate, to avoid ambiguities when both a const and a non-const bound is present.

[rust-highfive]

r? @oli-obk

(rust_highfive has picked a reviewer for you, use r? to override)

[oli-obk]

This is absolutely great! A few points for clarity:

• Please add tests for mixes of ?const Trait bounds and Trait bounds, where there would be ambiguities for figuring out which bound should be called.
• Please add a test for calling a generic method with bounds from withing a generic function with bounds. Ideally different tests for both with ?const and without on the method's bounds
  • or will this just hit the same issue as src/test/ui/rfc-2632-const-trait-impl/call-generic-method-nonconst.rs?
• Track the problem about the actual usage within Tracking issue for RFC 2632, impl const Trait for Ty and ~const (tilde const) syntax #67792

[jonas-schievink]

I've added more tests, but I'm not sure what you mean by this:

Track the problem about the actual usage within #67792

[oli-obk]

I added some text about the issue of using types that do not impl const Trait for const fn parameters that have Trait bounds and thus expect const Trait impls.

@bors r+

[bors]

📌 Commit cb40684 has been approved by oli-obk

[RalfJung]

Do we have tests to ensure that this only works when a feature flag is set?

[oli-obk]

We have https://github.com/rust-lang/rust/blob/6af388b25050bca26710be7e4030e17bf6d8d2f7/src/test/ui/rfc-2632-const-trait-impl/feature-gate.rs, but that only checks for impls, not actually using them.

[jonas-schievink]

There are no const impls in libcore/std, so using one currently also requires defining one. I'll add a note to the tracking issue.

[leonardo-m]

I think if you write the keywords in the wrong order "const impl PartialEq for S {" the compiler could give a bit better error message.

[leonardo-m]

See also: #79378

[usbalbin]

Is something like this supposed to work with this PR? Or is that something which might get implemented in a later stage? :)

impl<T, U> const Into<U> for T
where
    U: From<T>,
{
    fn into(self) -> U {
        U::from(self)
    }
}

I currently get this error

error[E0015]: calls in constant functions are limited to constant functions, tuple structs and tuple variants

[oli-obk]

This is unimplemented as of now, but being looked into since around a week

[usbalbin]

Oh, sorry. Thanks :)