`ambiguous_wide_pointer_comparisons` false positive for `Ord` and `PartialOrd`

[tyilo]

Code

use std::cmp::Ordering;

fn ptr_cmp<T: ?Sized>(p1: *const T, p2: *const T) -> Ordering {
    p1.cmp(&p2)
}

fn less_than<T: ?Sized>(p1: *const T, p2: *const T) -> bool {
    p1 < p2
}

fn main() {
    let s = "abc";

    assert_eq!(ptr_cmp(s, s), Ordering::Equal);
    assert_eq!(ptr_cmp(&s[..1], s), Ordering::Less);

    assert_eq!(less_than(s, s), false);
    assert_eq!(less_than(&s[..1], s), true);
}

Current output

warning: ambiguous wide pointer comparison, the comparison includes metadata which may not be expected
 --> src/main.rs:4:5
  |
4 |     p1.cmp(&p2)
  |     ^^^^^^^^^^^
  |
  = note: `#[warn(ambiguous_wide_pointer_comparisons)]` on by default
help: use `std::ptr::addr_eq` or untyped pointers to only compare their addresses
  |
4 |     p1.cast::<()>().cmp(&p2.cast::<()>())
  |       +++++++++++++        +++++++++++++

warning: ambiguous wide pointer comparison, the comparison includes metadata which may not be expected
 --> src/main.rs:8:5
  |
8 |     p1 < p2
  |     ^^^^^^^
  |
help: use `std::ptr::addr_eq` or untyped pointers to only compare their addresses
  |
8 |     p1.cast::<()>() < p2.cast::<()>()
  |       +++++++++++++     +++++++++++++

warning: `wide-ptr-cmp` (bin "wide-ptr-cmp") generated 2 warnings
    Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.01s

Desired output

No output

Rationale and extra context

The warning references std::ptr::addr_eq, however that can't be used for ordering two pointers.

Either std::ptr::cmp should be added to std (and core) which can be referenced instead or the lint should simply not be triggered by <pointer>::cmp and <pointer>::partial_cmp.

Other cases

Rust Version

$ rustc --version --verbose
rustc 1.87.0 (17067e9ac 2025-05-09)
binary: rustc
commit-hash: 17067e9ac6d7ecb70e50f92c1944e545188d2359
commit-date: 2025-05-09
host: x86_64-unknown-linux-gnu
release: 1.87.0
LLVM version: 20.1.1

Anything else?

No response

[Urgau]

The warning states "use std::ptr::addr_eq or untyped pointers" and then proceeds to explicitly suggest the "untyped pointers" approach.

warning: ambiguous wide pointer comparison, the comparison includes metadata which may not be expected
 --> src/main.rs:4:5
  |
4 |     p1.cmp(&p2)
  |     ^^^^^^^^^^^
  |
  = note: `#[warn(ambiguous_wide_pointer_comparisons)]` on by default
help: use `std::ptr::addr_eq` or untyped pointers to only compare their addresses
  |
4 |     p1.cast::<()>().cmp(&p2.cast::<()>())
  |       +++++++++++++        +++++++++++++

We should probably not mention std::ptr::addr_eq for cmp/partial_cmp cases, but the lint is still valid either way.

[tyilo]

I actually need the cmp behavior that includes the metadata of wide pointers.

[Urgau]

Then you can just allow the lint, since it's expected for you, put #[allow(ambiguous_wide_pointer_comparisons)] on your comparisons.

If we had some version of std::ptr::addr_eq for cmp/partial_cmp it would also strip the metadata. For equality we have std::ptr::eq, which also compares the metadata, but we don't have anything similar for cmp.

[Urgau]

Either std::ptr::cmp should be added to std (and core) which can be referenced instead or the lint should simply not be triggered by <pointer>::cmp and <pointer>::partial_cmp.

I can see a rational for that. Allowing the lint is also an option.

I think it would be quite unfortunate to reduce the scope of the lint simply because we don't have a "nice" std function to avoid it. Comparing pointers with metadata is also never what you want.

[tyilo]

Comparing pointers with metadata is also never what you want.

I need to use it for this: facet-rs/facet#675

[Urgau]

What if the suggested output for the cmp/partial_cmp cases was? Would you still think it's a "false positive"?

help: use untyped pointers to only compare their addresses
  |
4 |     p1.cast::<()>().cmp(&p2.cast::<()>());
  |       +++++++++++++        +++++++++++++
help: or allow the lint to compare metadata and addresses
  |
4 -     p1.cmp(&p2)
4 +     #[allow(ambiguous_wide_pointer_comparisons)] p1.cmp(&p2)
  |

[tyilo]

That would be nice 👍