Inconsistent warning for macros that contain unexpected_cfgs

Problem

Hi, I recently encountered an inconsistent behavior of cargo check when I have a macro which expands to something with unexpected_cfgs.

In short, I have a macro definition that expand to something with #[cfg(feature = "my_feature")]. When my_feature is not declared in Cargo.toml of my crate, I expected unexpected_cfgs to be reported whenever I use this macro.
However, whenever I use this macro in a downstream project (which also doesn't declare my_feature), no warning reported.

I originally thought this was a rust-analyzer problem, but after the discussion in rust-lang/rust-analyzer#18461, it seems this is more relevant to cargo.

I would like to understand if this is the expected behavior, because it feels like a bug to me.

Steps

Say I have a library mylib, and mylib/lib.rs has

pub fn my_lib_func() {
    println!("my nice little library")
}

#[macro_export]
macro_rules! my_lib_macro {
    () => {
        #[cfg(feature = "my_feature")]
        $crate::my_lib_func()
    };
}

#[test]
fn my_unit_test() {
    // rust-analyzer warning here
    my_lib_macro!();
}

Now running cargo check --tests, I got the expected warning:

  --> mylib/src/lib.rs:8:15
   |
8  |         #[cfg(feature = "my_feature")]
   |               ^^^^^^^^^^^^^^^^^^^^^^ help: remove the condition
...
15 |     my_lib_macro!();
   |     --------------- in this macro invocation
   |
   = note: no expected values for `feature`
   = help: consider adding `my_feature` as a feature in `Cargo.toml`
   = note: see <https://doc.rust-lang.org/nightly/rustc/check-cfg/cargo-specifics.html> for more information about checking conditional configuration
   = note: `#[warn(unexpected_cfgs)]` on by default
   = note: this warning originates in the macro `my_lib_macro` (in Nightly builds, run with -Z macro-backtrace for more info)

However, if I use mylib in other packages, say I have another crate, myapp, that depends on mylib.
In myapp/main.rs, I have:

use mylib::my_lib_macro;

fn main() {
    my_lib_macro!();
}

Running cargo check on myapp reports nothing.

Possible Solution(s)

No response

Notes

No response

Version

cargo 1.82.0 (8f40fc59f 2024-08-21)
release: 1.82.0
commit-hash: 8f40fc59fb0c8df91c97405785197f3c630304ea
commit-date: 2024-08-21
host: x86_64-apple-darwin
libgit2: 1.8.1 (sys:0.19.0 vendored)
libcurl: 8.6.0 (sys:0.4.74+curl-8.9.0 system ssl:(SecureTransport) LibreSSL/3.3.6)
ssl: OpenSSL 1.1.1w 11 Sep 2023
os: Mac OS 14.5.0 [64-bit]

@Urgau

It doesn't need to have multiple Cargo packages. Simply a src/main.rs and a src/lib.rs in a single package can reproduce it.

This also applies to some simpler lints like unused_variables. It won't report the lint diagnostic from the caller crate. Change the macro to let x = 1 and you'll see.

However, if you change the macro content to 1_i32 << 32; you'll get a lint error of when running cargo c --bins

warning: unused bitwise operation that must be used
 --> src/main.rs:3:5
  |
3 |     foo::my_macro!();
  |     ^^^^^^^^^^^^^^^^ the bitwise operation produces a value
  |
  = note: `#[warn(unused_must_use)]` on by default
  = note: this warning originates in the macro `foo::my_macro` (in Nightly builds, run with -Z macro-backtrace for more info)

Despite that, you won't get the arithmetic_overflow lint error which you could get from running cargo t --lib --no-run.

error: this arithmetic operation will overflow
  --> src/lib.rs:4:9
   |
4  |         1_i32 << 32;
   |         ^^^^^^^^^^^ attempt to shift left by `32_i32`, which would overflow
...
11 |     my_macro!();
   |     ----------- in this macro invocation
   |
   = note: `#[deny(arithmetic_overflow)]` on by default
   = note: this error originates in the macro `my_macro` (in Nightly builds, run with -Z macro-backtrace for more info)

I feel like this behavior pretty much depends on how the lint is implemented, though not a lint wizard so may need experts to answer this.

cc @Urgau do you happen to know how unexpected_cfgs is expected to work for macros from external crates?

I'm pretty sure this is because we (the compiler) don't (in general) show warnings coming from external macro, and the unexpected_cfgs is not marked as being allowed to report on code generated from external macros.

This is done to avoid awkward warnings where the users can't do anything about the warnings.

In the case of the unexpected_cfgs lint I could see it being useful to marked as reportable in external macros.

@rustbot

I'm transferring this issue to main Rust repo, as it's an issue with compiler.

~~@rustbot transfer rust~~ EDIT: nevermind I can't, @weihanglo would you mind doing it for me

Navigation Menu

Repository navigation

Inconsistent warning for macros that contain `unexpected_cfgs` #132572

Problem

Steps

Possible Solution(s)

Notes

Version

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Inconsistent warning for macros that contain unexpected_cfgs #132572

Description

Problem

Steps

Possible Solution(s)

Notes

Version

Activity

weihanglo commented on Nov 3, 2024

Urgau commented on Nov 3, 2024

Urgau commented on Nov 3, 2024

rustbot commented on Nov 3, 2024

rustbot commented on Nov 3, 2024

weihanglo commented on Nov 3, 2024

rustbot commented on Nov 3, 2024

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Inconsistent warning for macros that contain `unexpected_cfgs` #132572