`#[repr]` on zero-variant enums leaves the reader with questions

[Veykril]

Code

#[repr(isize)]
enum Uninhabited {}

Current output

error[E0084]: unsupported representation for zero-variant enum
 --> src/lib.rs:1:1
  |
1 | #[repr(isize)]
  | ^^^^^^^^^^^^^^
2 | enum Uninhabited {}
  | ---------------- zero-variant enum

Desired output

error[E0084]: zero-variant enums do not support `#[repr]`
 --> src/lib.rs:1:1
  |
1 | #[repr(isize)]
  | ^^^^^^^^^^^^^^
2 | enum Uninhabited {}
  | ---------------- zero-variant enum

Rationale and extra context

The current output leaves the user questioning whether there may be some valid repr while not being told which ones are valid options. The diagnostic should probably be more clear about the fact that the attribute is generally not supported on zero-variant enums

Other cases

Rust Version

rustc 1.86.0 (05f9846f8 2025-03-31)
binary: rustc
commit-hash: 05f9846f893b09a1be1fc8560e33fc3c815cfecb
commit-date: 2025-03-31
host: x86_64-pc-windows-msvc
release: 1.86.0
LLVM version: 19.1.7

Anything else?

No response

[hanna-kruppe]

I’m not sure it’s true (and will remain true) that absolutely no #[repr(…)] attribute can be applied to empty enums. For example, I would have expected #[repr(align(N))] to be valid since it has no relation to enum discriminants (the rationale for E0084). Apparently it’s rejected right now, but I don’t think there’s any reason to prohibit it, it probably just falls out of an overly broad check somewhere in rustc.

[fmease]

Lol, even #[repr(Rust)] gets rejected on empty enums, that's a bug.

[SciMind2460]

I’m not sure it’s true (and will remain true) that absolutely no #[repr(…)] attribute can be applied to empty enums. For example, I would have expected #[repr(align(N))] to be valid since it has no relation to enum discriminants (the rationale for E0084). Apparently it’s rejected right now, but I don’t think there’s any reason to prohibit it, it probably just falls out of an overly broad check somewhere
in rustc.

Perhaps" #[repr(isize)] is not supported for zero-variant enums" is better?

[Veykril]

I'd be fine with putting a currently into my proposed message as well.

zero-variant enums currently do not support #[repr]

or, given fmease's message

zero-variant enums currently do not support representations other than #[repr(Rust)].

My main point here is that the current message does not inform the user about what the valid options are (which currently is none)