"field is never read" warning despite `Debug` being used explicitly in the code

[fjarri]

use core::fmt;

#[derive(Debug)]
pub struct Error(String);

impl fmt::Display for Error {
    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> Result<(), fmt::Error> {
        // warning
        write!(f, "{:?}", self)

        // no warning
        // write!(f, "{}", self.0)
    }
}

pub fn api_call(x: usize) -> Result<usize, Error> {
    if x == 0 {
        Err(Error("Error".into()))
    }
    else {
        Ok(x)
    }
}

Let's start from the case where impl Display is commented out. clippy gives a dead code warning about the unused String field in Error (see the discussion in #88900).

The suggested workaround is an explicit Display impl. But if I add that, using the Debug internally to generate the displayed string, the warning is still there:

warning: field `0` is never read
 --> src/lib.rs:4:18
  |
4 | pub struct Error(String);
  |            ----- ^^^^^^
  |            |
  |            field in this struct
  |
  = note: `Error` has a derived impl for the trait `Debug`, but this is intentionally ignored during dead code analysis
  = note: `#[warn(dead_code)]` on by default
help: consider changing the field to be of unit type to suppress this warning while preserving the field numbering, or remove the field
  |
4 | pub struct Error(());
  |                  ~~

In order to get rid of the warning, I must explicitly use the .0 field in the Display impl. While it is alright for this simple case, for error types which are enums with a lot of fields, this becomes very tiresome. Is it the intended behavior?

Meta

rustc 1.77.1 (7cf61eb 2024-03-27)
binary: rustc
commit-hash: 7cf61eb
commit-date: 2024-03-27
host: aarch64-apple-darwin
release: 1.77.1
LLVM version: 17.0.6
(seems to have been introduced in 1.77.0)

[jieyouxu]

Looks like duplicate of #123068

[fjarri]

Yes, sorry, didn't find it when searching. Commented there.

[jieyouxu]

Closing this as a duplicate of #123068.