const validation: properly ignore zero-sized UnsafeCell

[RalfJung]

Fixes #142948
r? @oli-obk

[rustbot]

Some changes occurred to the CTFE / Miri interpreter

cc @rust-lang/miri

Some changes occurred to the CTFE machinery

cc @RalfJung, @oli-obk, @lcnr

[lcnr]

this seems very footgunny, why do is_zst and size_and_align_of(&val.meta(), &val.layout).is_some_and(|(s, _a)| s.bytes() == 0) differ. When is using is_zst actually correct?

Okay, this is because we're not checking whether the type itself is a ZST, but that the value is a zero-sized. So we don't use is_zst to also handle slices and trait objects. Can you add this as a comment to the code in visit_value?

after that r=me

[RalfJung]

When is using is_zst actually correct?

Basically this is must_be_zst, but it can miss types that are only dynamically zero-sized.

[rustbot]

The Miri subtree was changed

cc @rust-lang/miri

Some changes occurred to MIR optimizations

cc @rust-lang/wg-mir-opt

[RalfJung]

I also added a 2nd commit making the size_and_align_of APIs in the interpreter a bit nicer.

[oli-obk]

@bors r=lcnr,oli-obk

[bors]

📌 Commit 7de39f5 has been approved by lcnr,oli-obk

It is now in the queue for this repository.