`multiple_inherent_impl` shouldn't trigger for `#[cfg(test)] impl`

[nickdrozd]

Summary

I have a struct with impl block, plus an additional impl block for testing:

struct X {}

impl X {}

#[cfg(test)]
impl X {}

The multiple_inherent_impl lint flags this, but it shouldn't. Combining the two impl blocks would cause a change in behavior, namely enabling the test methods outside of the test environment.

Lint Name

multiple_inherent_impl

Reproducer

No response

Version

rustc 1.81.0-nightly (6292b2af6 2024-07-02)
binary: rustc
commit-hash: 6292b2af620dbd771ebb687c3a93c69ba8f97268
commit-date: 2024-07-02
host: x86_64-unknown-linux-gnu
release: 1.81.0-nightly
LLVM version: 18.1.7

Additional Labels

No response

[Jarcho]

It shouldn't trigger if there are any attributes, not just #[cfg(test)].

[nickdrozd]

I guess in general it should trigger only for two impl blocks with identical attributes. So for example, for two #[cfg(test)] impl blocks it should trigger.

[profetia]

Looks like this already fixed in the latest nightly.

[mrchilliballs]

Looks like this doesn't trigger with #[cfg(test)] impl on Clippy 0.1.89, but it seems like it still does for other attributes, unlike what @Jarcho suggested.
https://play.rust-lang.org/?version=stable&mode=debug&edition=2024&gist=da8219de8517cd7cebf48f6ee7c5f413

[Jarcho]

My comment was that this attribute shouldn't be singled out, not about what the lint currently does.

[MikkelPaulson]

@mrchilliballs:

Looks like this doesn't trigger with #[cfg(test)] impl on Clippy 0.1.89, but it seems like it still does for other attributes, unlike what Jarcho suggested. https://play.rust-lang.org/?version=stable&mode=debug&edition=2024&gist=da8219de8517cd7cebf48f6ee7c5f413

Clippy doesn't evaluate #[cfg(test)] code at all unless run with --tests, which is not the case in the playground. Evaluating the same code using cargo clippy --tests in Rust 1.90.0 does indeed produce a warning.