resolve: prohibit anon const non-static lifetimes

[davidtwco]

Fixes #75323, fixes #74447 and fixes #73375.

This PR prohibits non-static lifetimes in anonymous constants when only the min_const_generics feature is enabled. To do so, to_region_vid's bug! had to be changed into a delayed bug, which unfortunately required providing it a TyCtxt.

---

While I am happy with how the implementation of the error turned out in rustc_passes::check_const, emitting an error wasn't sufficient to avoid hitting the ICE later. I also tried implementing the error in rustc_mir::transform::check_consts::validation and that worked, but it didn't silence the ICE either. To silence the ICE, I changed it to a delayed bug which worked but was more invasive that I would have liked, and required I return an incorrect lifetime. It's possible that this check should be implemented earlier in the compiler to make the invasive changes unnecessary, but I wasn't sure where that would be and wanted to get some feedback first.
The approach taken by this PR has been changed to implement the error in name resolution, which ended up being much simpler.

cc @rust-lang/wg-const-eval
r? @lcnr

[lcnr]

Does this PR break the following snippet which already compiles on stable?
This restriction should only apply to anonymous constants, associated constants should not be affected here.

edit: ignore the above questions, didn't see the check on AnonConst, this already works correctly

#![feature(min_const_generics)]
struct Foo<'a>(&'a ());

impl<'a> Foo<'a> {
    const ASSOC: usize = {
        let x: &'a usize = &7;
        *x
    };
}

It might be easier to forbid these lifetimes during name resolution instead where we hopefully should be able to replace them with ty::ReStatic. I would expect this be a less invasive change which does not require us to change bug to a delay_span_bug.

[davidtwco]

Pushed a change that makes the test check both min_const_generics and const_generics.

Does this PR break the following snippet which already compiles on stable?
This restriction should only apply to anonymous constants, associated constants should not be affected here.

As you've noticed, it doesn't break that snippet and only applies to anonymous constants.

It might be easier to forbid these lifetimes during name resolution instead where we hopefully should be able to replace them with ty::ReStatic. I would expect this be a less invasive change which does not require us to change bug to a delay_span_bug.

I can give this approach a go if you'd prefer.

[davidtwco]

@lcnr I've updated the PR to implement this in name resolution (cc @petrochenkov), which did end up being much cleaner.

[lcnr]

I think this looks good now, but I am not too familiar with lifetime resolution, so maybe

r? @nikomatsakis for the final approval.

[hameerabbasi]

One request: Right now, it's hard to tell from the test which parts cause an ICE and which parts compile on full. It might be good to split the test up into multiple tests into src/test/ui/const-generics/issues, but this is only a suggestion.

[davidtwco]

I removed the example from this comment from the test because it gives a very similar but still different ICE on CI and that causes it to fail - I wasn't able to work out why that was, even after rebasing locally.

One request: Right now, it's hard to tell from the test which parts cause an ICE and which parts compile on full. It might be good to split the test up into multiple tests into src/test/ui/const-generics/issues, but this is only a suggestion.

Given the above, I've kept a single test since the ICE is only produced from the cases that emit an error under min_const_generics and each ICE is the same.

[varkor]

Thanks!

@bors r+ rollup

[bors]

📌 Commit eacfb2b has been approved by varkor