unused_parens generates invalid code

[benStre]

Summary

Clippy transforms structs in combination with modular_bitfield and derive(Debug) macros into invalid rust sytnax.
See minimal reproducible example: https://github.com/unyt-org/clippy-bug-example

Reproducer

Given the following struct definition:

#[bitfield]
#[derive(Debug)]
pub struct Flags {
    val: u8,
}

cargo clippy --fix generates the following code:

#[bitfield]
#[derive(Debug)]
pub struct Flags {
    al: u,
}

Version

rustc 1.90.0-nightly (3014e79f9 2025-07-15)
binary: rustc
commit-hash: 3014e79f9c8d5510ea7b3a3b70d171d0948b1e96
commit-date: 2025-07-15
host: aarch64-apple-darwin
release: 1.90.0-nightly
LLVM version: 20.1.8

Additional Labels

No response

[blyxyas]

@rustbot transfer rust

This also happens cargo check, and it doesn't come from a Clippy lint. So I moved it to Rust

[blyxyas]

bisected: #142237 is the origin for this regression
cc @benschulz @fee1-dead

[benschulz]

From what I can tell, the expanded bitfield macro contains types with unnecessary parentheses. Because the field's span is used, we end up here.

AFAICT, there are two possible ways forward.

1. Declare this an issue with the bitfield macro. It emits types with unused parentheses without #[allow]ing those.
2. Revert to a stringly check for parentheses before emitting the lint.

I don't know what the policy is on this, so I'll let someone else decide.

[blyxyas]

From the Clippy side we usually check if the given span comes from an external expansion /proc macro and just skip it if that's the case. (Because more times than not, either the user cannot change the original macro, or the issue in question doesn't affect macro-generated code.

In this case, being that unused_parens is a strictly stylistic, I'd say that taking this route is appropiate.