Fix panicking_unwrap FP on field access with implicit deref

[profetia]

Closes #16188

---

changelog: [panicking_unwrap] fix FP on field access with implicit deref

[rustbot]

r? @llogiq

rustbot has assigned @llogiq.
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

[github-actions]

Lintcheck changes for c905503

Lint	Added	Removed	Changed
clippy::unnecessary_unwrap	0	2	0

This comment will be updated if you push new changes

[llogiq]

Thank you!

View changes since this review

[ada4a]

Why does this work though?... It definitely no longer aligns with what the comment is describing

[ada4a]

I think what happened here is that I got my conditions wrong: when an non-field projections appears, we're supposed to return true, as it might be modifying something somehow -- but what the code was actually doing is returning false inside the .all(), meaning that the whole function would return false as well.

I came up with the following patch, which goes through each case a bit more thouroughly -- what do you think?

{
  if !is_potentially_local_place(*local_id, place) {
      // This can't be a local place -- no point checking any further
      return false;
  }
  
  let mut field_indices = field_indices.iter().rev().copied();
  for proj in &place.projections {
      let ProjectionKind::Field(f_idx, _) = proj.kind else {
          // This is a projection we didn't expect -- it _might_ be mutating something
          return true;
      };
  
      let Some(field_idx) = field_indices.next() else {
          // The field projections are going deeper than our `Local` -- a mutation in there
          // can't change whether the `Local` is the success or error variant
          return false;
      };
  
      if f_idx != field_idx {
          // The projection is going into an unrelated field -- it won't mutate our `Local`
          return false;
      }
  }
  
  // There are no more projections, and we haven't strayed off the field access chain that our `Local`
  // is located on, so a mutation of this place might mutate our `Local`
  true
}

[profetia]

I don't really get your point, but for the case of the issue, there is a deref before the field access

[ada4a]

What I'm trying to say is that just filtering out non-field projections might be dangerous, as they might end up modifying the Local... somehow. So I think we should instead early-return when encountering them. The patch above achieves that using the first let-else