Point at unit structs on foreign crates in type errors when they are the pattern of a binding

[estebank]

Consts and unit structs in patterns can be confusing if they are mistaken for new bindings. We already provide some context for unit structs and consts that come from the current crate, we now also point at those from foreign crates, and we properly skip cases where the pattern has type parameters which can't be confused with a new binding. So not suggest making a new binding when other suggestions are already emitted, as the likelihood of the other suggestions being what the user intended is higher. Make new binding suggestion verbose.

Fix #129792.

error[E0308]: mismatched types
 --> fi.rs:8:9
  |
1 | struct percentage;
  | ----------------- unit struct defined here
...
8 |     let percentage = 4i32;
  |         ^^^^^^^^^^   ---- this expression has type `i32`
  |         |
  |         expected `i32`, found `percentage`
  |         `percentage` is interpreted as a unit struct, not a new binding
help: introduce a new binding instead
  |
8 -     let percentage = 4i32;
8 +     let other_percentage = 4i32;
  |

[rustbot]

r? @nnethercote

rustbot has assigned @nnethercote.
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
• compiler expanded to 69 candidates
• Random selection from 15 candidates

[nnethercote]

The example from #129792 seems reasonable:

error[E0308]: mismatched types
 --> fi.rs:8:9
  |
1 | struct percentage;
  | ----------------- unit struct defined here
...
8 |     let percentage = 4i32;
  |         ^^^^^^^^^^   ---- this expression has type `i32`
  |         |
  |         expected `i32`, found `percentage`
  |         `percentage` is interpreted as a unit struct, not a new binding
  |         help: introduce a new binding instead: `other_percentage`

The use of a lower-case p on percentage leads to the type/binding name confusion. It's easy to imagine somebody thinking that percentage is a binding.

The example from this PR's description seems much less likely. Do we really think the problem in that code is that the user intended None to be a new binding? To me that's clearly just a type error caused by misunderstanding the Box, the extra information about the binding just confuses things, and suggesting other_none is not helpful.

[estebank]

Agreed. The second commit uses the prior existence of suggestions as a signal that we shouldn't talk about introducing the new binding. The only case left in the test suite (other than the one in the ticket) is

error[E0308]: mismatched types
  --> $DIR/name-clash-nullary.rs:2:7
   |
LL |   let None: isize = 42;
   |       ^^^^  ----- expected due to this
   |       |
   |       expected `isize`, found `Option<_>`
   |       `None` is interpreted as a unit variant, not a new binding
   |
  --> $SRC_DIR/core/src/option.rs:LL:COL
   |
   = note: unit variant defined here
   |
   = note: expected type `isize`
              found enum `Option<_>`
help: introduce a new binding instead
   |
LL -   let None: isize = 42;
LL +   let other_none: isize = 42;
   |

and

error[E0308]: mismatched types
  --> $DIR/issue-33504.rs:7:13
   |
LL | struct Test;
   | ----------- unit struct defined here
...
LL |         let Test = 1;
   |             ^^^^   - this expression has type `{integer}`
   |             |
   |             expected integer, found `Test`
   |             `Test` is interpreted as a unit struct, not a new binding
   |
help: introduce a new binding instead
   |
LL -         let Test = 1;
LL +         let other_test = 1;
   |

Which I feel is the exact case we want to address, right? The only other thing I'd consider is making it so that in the cases where we're suggesting changing the case, we don't include the other_, but that's a very minor thing.

[nnethercote]

This one feels marginal, and likewise the other ones in this test. I can imagine a user thinking this would match against b"012". But it's sufficiently broken thinking that it's hard to say for sure, so I can live with it.

[estebank]

I'm currently looking at a completely different case, and maybe the suggestion I'm adding might also be relevant here?

error: constant of non-structural type `S` in a pattern
  --> $DIR/match_ice.rs:11:9
   |
LL | struct S;
   | -------- `S` must be annotated with `#[derive(PartialEq)]` to be usable in patterns
...
LL |     const C: &S = &S;
   |     ----------- constant defined here
LL |     match C {
LL |         C => {}
   |         ^ constant of non-structural type
   |
   = note: see https://doc.rust-lang.org/stable/std/marker/trait.StructuralPartialEq.html for details
help: if `S` manually implemented `PartialEq`, you could check for equality
   |
LL -         C => {}
LL +         binding if binding == C => {}
   |

Note that the binding suggestion is pre-existing, it's just being made more obvious now.

[nnethercote]

@bors r+ rollup

[rust-bors]

📌 Commit 408066f has been approved by nnethercote

It is now in the queue for this repository.

[rust-bors]

⌛ Testing commit 408066f with merge 501c798...

Workflow: https://github.com/rust-lang/rust/actions/runs/23190523244

[JonathanBrouwer]

Stuck
@bors yield

[rust-bors]

Auto build was cancelled. Cancelled workflows:

• https://github.com/rust-lang/rust/actions/runs/23190523244

The next pull request likely to be tested is #154001.