Fix ICE when Self is used in enum discriminant of a generic enum

[GokhanKabar]

View all comments

Fixes #153756
Let discriminant AnonConst inherit parent generics via Node::Variant in generics_of, and emit a proper error instead of span_bug! for the TooGeneric case in wfcheck.

[rustbot]

r? @petrochenkov

rustbot has assigned @petrochenkov.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

• Owners of files modified in this PR: compiler, types
• compiler, types expanded to 69 candidates
• Random selection from 15 candidates

[petrochenkov]

There's probably a way to implement this better.
r? types

[BoxyUwU]

r? BoxyUwU

[BoxyUwU]

Yeah this isn't quite what we want 🤔 The general structure here is supposed to be that name resolution rejects uses of generic parameters, and the things it can't do we reject during HIR ty lowering.

Name resolution can't/doesn't reject uses of Self which alias a generic type, nor can it reject generic parameters introduce by type dependent name resolution (e.g. <Self>::Assoc turning into <u8 as Trait<T>>::Assoc). So we have special cases for those in HIR ty lowering.

Instead of adding a new place where we check for illegal generic parameters can you look into why the existing checks in HIR ty lowering aren't working correctly for discriminant exprs. I expect that what happened is that the refactoring in #150519 was overly fit to const generics rather than all anon consts and so stopped validating stuff that it should.

https://github.com/rust-lang/rust/pull/150519/changes#diff-ce4273e1e949bf3052de7a08466c4dad785890b24091122954541b615b0ba199R2161

[rustbot]

HIR ty lowering was modified

cc @fmease

[BoxyUwU]

looks good to me other than the above comment :)

[BoxyUwU]

I think the logic here is still slightly wrong (see #154281), we should be handling DefKind::InlineConst and DefKind::Closure here, not just DefKind::AnonConst. Would you be interested in looking into that issue too? I expect it shouldn't be too much additional work on top of what this PR already does.

I would be happy to merge this PR as is or for you to bundle a fix for that issue into this PR. Up to you. I assigned the issue to myself just to make sure I don't lose track of it but I would love for someone else to take it on if possible ^^

[theemathas]

This code currently compiles in stable rust. If this PR makes this code fail, then a crater run is needed for this PR.

#[repr(usize)]
enum What {
    X = size_of::<*mut Self>(),
}

fn main() {}

[GokhanKabar]

This code currently compiles in stable rust. If this PR makes this code fail, then a crater run is needed for this PR.

#[repr(usize)]
enum What {
    X = size_of::<*mut Self>(),
}

fn main() {}

Tested this it still compiles fine. In a non-generic enum, Self resolves to a concrete SelfTyAlias which gets lowered to the actual enum type before check_param_uses_if_mcg sees it, so the folder never flags it. No crater run needed for this specific case.

[rustbot]

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.