`missing_docs` triggers for integration tests

[tv42]

Summary

If you warn on missing_docs, every integration test causes a warning from clippy --tests.

I think integration tests should be exempt from missing_docs.

My editor (https://zed.dev) (and maybe rust-analyzer by default?) shows clippy --tests messages too (and not just clippy), so having any integration tests in the tree makes missing_docs practically unusable.

Lint Name

missing_docs

Reproducer

Using the fn add example from cargo init --lib

Add to Cargo.toml

[lints.rust]
missing_docs = "warn"

add new file tests/two_plus_two.rs

use rust_clippy_missing_docs_for_tests::add;

// literally just a copy-paste from the in-library test

#[test]
fn it_works() {
    let result = add(2, 2);
    assert_eq!(result, 4);
}

And now clippy --tests complains:

$ cargo clippy --tests
warning: missing documentation for the crate
 --> tests/two_plus_two.rs:1:1
  |
1 | / use rust_clippy_missing_docs_for_tests::add;
... |
8 | |     assert_eq!(result, 4);
9 | | }
  | |_^
  |
  = note: requested on the command line with `-W missing-docs`

warning: `rust-clippy-missing-docs-for-tests` (test "two_plus_two") generated 1 warning
    Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.01s

Version

rustc 1.85.0 (4d91de4e4 2025-02-17)
binary: rustc
commit-hash: 4d91de4e48198da2e33413efdcd9cd2cc0c46688
commit-date: 2025-02-17
host: x86_64-unknown-linux-gnu
release: 1.85.0
LLVM version: 19.1.7

Additional Labels

No response

[lapla-cogito]

Note that missing_docs is a lint in rustc, not in clippy.

[tv42]

Similar: rust-lang/rust-clippy#11024

[tv42]

Note that missing_docs is a lint in rustc, not in clippy.

Yes, true. Should I move this issue elsewhere? I'm not very clear on the whole clippy-vs-rustc thing, apart from the work started in rustc and then got expanded as clippy.

[lolbinarycat]

Workaround is to enable it in the root module of your crate, instead of in Cargo.toml.

[ehuss]

This looks like it changed recently in #130025. cc @Urgau

I think not honoring missing_docs on tests was intentional, as missing docs on test code is usually not something that you really want.

[ehuss]

Though there was a related issue at #24584, where the behavior before 1.82 was a little awkward. If you are running cargo check in your editor with --profile=test (so that you get diagnostics for unit tests), the missing_docs lint would not work at all, which is not really what I wanted.

There seems to be some conflicting desires on how that behaves.

[Urgau]

The change in #130025 was a bug-fix to make #[expect(missing_docs)] work when --test is passed. That bug-fix was necessary, and in general we should not silence lints based on secondary compiler flags.

The change in behavior was in the "Compatibility Notes" of 1.83.0:

The allow-by-default missing_docs lint used to disable itself when invoked through rustc --test/cargo test, resulting in #[expect(missing_docs)] emitting false positives due to the expectation being wrongly unfulfilled. This behavior #130025, which allows #[expect(missing_docs)] to be fulfilled in all scenarios, but will also report new missing_docs diagnostics for publicly reachable #[cfg(test)] items, integration test crate-level documentation, and publicly reachable items in integration tests.

I don't think there is anything (on rustc's side) to do here, the behavior is expected and IMO wanted, lints should be predictable and consistent.

I disagree that it makes "missing_docs practically unusable", IMO putting description of what tests are supposed to do is IMO a good thing.

In any case, one case always put #![allow(missing_docs)] in integration tests.

[lolbinarycat]

I don't think there is anything (on rustc's side) to do here, the behavior is expected and IMO wanted, lints should be predictable and consistent.

I didn't realize rustc doesn't know if its running an integration test or unit test. It's easy to forget how relatively simple rustc's interface is when compared to cargo.