fix(generic_const_exprs): Fix predicate inheritance for children of opaque types

[Neo-Zhixing]

Fixes #99705

We currently have a special case to perform predicate inheritance when the const item is in the generics. I think we're also going to need this for opaque return types. When evaluating the predicates applied to the associated item, it'll inherit from its parent, the opaque type, which will never have predicates applied. This PR bypass the opaque typed parent and inherit predicates directly from the function itself.

[rust-highfive]

Thanks for the pull request, and welcome! The Rust team is excited to review your changes, and you should hear from @compiler-errors (or someone else) soon.

Please see the contribution instructions for more information.

[Neo-Zhixing]

r? @BoxyUwU

[Neo-Zhixing]

@rustbot label +A-const-generics +F-generic_const_exprs

[BoxyUwU]

I suspect this probably doesnt work right for -> impl Foo<Assoc = impl Bar = Baz<{ T::ASSOC }>>. I am not that familiar with opaque type stuff so going to r? @oli-obk

[oli-obk]

I think this will get resolved by @spastorino's work which removes this parent generics behaviour of return position impl trait. This PR is just more proof that doing that is important. It feels like a band-aid that I would prefer not to merge

[Neo-Zhixing]

@oli-obk Sure, I'm OK with closing this for now and wait for the RPIT refactor to happen. Do you have a pointer to where that is happening & the timeline? Just want to get this bug properly documented here so that other people don't end up wasting time on this.

[oli-obk]

I don't know where @spastorino is tracking this work. I'll let him post a link in case there's a hackmd.

We could link #99705 from the generic_const_exprs tracking issue so it doesn't get lost.

[Neo-Zhixing]

Ok sounds good. Let's keep this PR open in the mean time so that those who need this now can use the branch as a custom toolchain.

[spastorino]

For some reason I've missed this one. Anyway, my changes are close to be completed. Right now I'm fixing some bugs on my branch.

[spastorino]

PR with RPIT changes is now up #101345, we still need to address some problems that crater have found.

[bors]

☔ The latest upstream changes (presumably #102306) made this pull request unmergeable. Please resolve the merge conflicts.

[pnkfelix]

discussed in T-compiler triage meeting.

reassigning ownership for design decisions here to T-types.

@rustbot label: -T-compiler +T-types

[oli-obk]

I think we should "just do this PR" and remove the changes when we'll actually change the opaque type parenting logic.

So... after much work that didn't get merged after all, @Neo-Zhixing can you rebase this PR, we can then merge it.

[Neo-Zhixing]

@oli-obk changes merged with master. Thanks!

[compiler-errors]

@Neo-Zhixing we don't actually allow merge conflicts in the rust repo: https://rustc-dev-guide.rust-lang.org/git.html#no-merge-policy

Can you rebase this commit instead?

[Neo-Zhixing]

My apologies, I thought GitHub would be able to squash those commits for you.
Rebased. Thanks @compiler-errors

[oli-obk]

@bors r+

[bors]

📌 Commit ccd9e07281b43567de28ad9f6601c49464e44bd6 has been approved by oli-obk

It is now in the queue for this repository.

[BoxyUwU]

did we test if -> impl Foo<Assoc = impl Bar<{ T::ASSOC }>> is fixed by this PR?

#![feature(generic_const_exprs)]

trait Foo {
    type Assoc;
}

trait Bar<const N: usize> {}

impl Foo for () {
    type Assoc = ();
}
impl<const N: usize> Bar<N> for () {}

trait Trait {
    const ASSOC: usize;
}

fn foo<T: Trait>() -> impl Foo<Assoc = impl Bar<{ T::ASSOC }>> {}

[oli-obk]

@bors r- let's add the test, but I think it should work

[Neo-Zhixing]

@rustbot label -S-waiting-on-author +S-waiting-on-review
@oli-obk @BoxyUwU Test added for the nested impl. Indeed it works! Thanks

[oli-obk]

@bors r+

[bors]

📌 Commit 744fa61 has been approved by oli-obk

It is now in the queue for this repository.

[BoxyUwU]

Is there no way to get nested opaque items in current rustc 🤔 that was what i was going for with -> impl Trait<Assoc = impl Other<{ T::ASSOC }>>. Judging by the fact that the code i posted compiles it kind of implies that is nested opaques so I am slightly confused as to how this PR fixes that ✨

[oli-obk]

Is there no way to get nested opaque items in current rustc

yes there is. Your example does indeed show that, but it's not an issue because the inner opaque item fetches the predicates of the outer one, which fetches it from the function.

[BoxyUwU]

I thought the bug was that opaque type items dont have all the predicates of the parent, so anon consts inheriting predicates from the opaque is incorrect since they should have "all" of them. Clearly i've misunderstood something here since if that was true the anon const having a parent of the outter opaque type would not be correct 😅

[bors]

⌛ Testing commit 744fa61 with merge 96787c4...