evaluation error hides "impl item missing" error

[lcnr]

trait Trait {
    const N: usize;
}

impl Trait for i32 {}

fn f()
where
    [(); <i32 as Trait>::N]:,
{}

fn main() {}

results in

error[[E0080]](https://doc.rust-lang.org/nightly/error-index.html#E0080): evaluation of constant value failed
 --> src/main.rs:9:10
  |
9 |     [(); <i32 as Trait>::N]:,
  |          ^^^^^^^^^^^^^^^^^ referenced constant has errors

without showing that N is missing in the trait impl.

Taken from the comment by @carado in #74713 (comment)

[Rageking8]

@rustbot label T-compiler

[Rageking8]

Possible regression from stable to nightly. (Tested in Rust Playground)

Both Stable 1.61.0 and 1.62.0-beta.7 (2022-06-26 747075d) result in this error :

error[E0046]: not all trait items implemented, missing: `N`
 --> src/main.rs:5:1
  |
2 |     const N: usize;
  |     --------------- `N` from trait
...
5 | impl Trait for i32 {}
  | ^^^^^^^^^^^^^^^^^^ missing `N` in implementation

For more information about this error, try `rustc --explain E0046`.

However on Nightly version: 1.64.0-nightly (2022-06-27 2f3ddd9)

This is the error :

error[E0080]: evaluation of constant value failed
 --> src/main.rs:9:10
  |
9 |     [(); <i32 as Trait>::N]:,
  |          ^^^^^^^^^^^^^^^^^ referenced constant has errors

For more information about this error, try `rustc --explain E0080`.

[Rageking8]

@rustbot label regression-from-stable-to-nightly

[apiraino]

if my bisection is correct, commit 1f34da9 seems to trigger this. In #96591 cc @b-naber

searched nightlies: from nightly-2022-04-01 to nightly-2022-06-28
regressed nightly: nightly-2022-06-15
searched commits: from ca122c7 to 1f34da9
regressed commit: 1f34da9

bisected with cargo-bisect-rustc v0.6.0

Host triple: x86_64-unknown-linux-gnu
Reproduce with:

cargo bisect-rustc --start=2022-04-01 --script=./test.sh 

$ cat test.sh 
#!/bin/bash
cargo check 2>&1 | grep "from trait"

[apiraino]

WG-prioritization assigning priority (Zulip discussion).

The missing diagnostic message in this case can possibly make fixing the code quite confusing. Thanks @TaKO8Ki for offering some help here :)

@rustbot label -I-prioritize +P-high

[compiler-errors]

Hi @TaKO8Ki, are you working on this issue? I've been keeping tabs on this one for a bit and I decided to look into it since it's been a few weeks.

I think I may have found a bug in the way that we resolve const items and I would like to put up a PR, unless you've already found a solution. 😸

Also, marking this as a regression from stable to beta, since beta cutoff has happened since this issue was filed, I think.