Forwarding input arguments of #[tokio::test] annotated tests

[rubdos]

Test fixture libraries such as rstest tend to use a proc macro to inject fixtures in tests, based on function parameters of the test function. Example:

#[fixture] fn fix -> u32 {2}
#[rstest] fn foo(fix: u32) { assert_eq!(2, fix) }

Currently, there is an ongoing discussion in rstest on how to best integrate with test-suites other than the default Rust #[test], in particular those that use async, such as #[tokio::test].

In order to transparently integrate and make the #[rstest] and #[tokio::test] attributes "commutative", the function arguments of a test should simply be forwarded by the tokio::test proc macro. Currently, adding function arguments to a test triggers a custom error. As far as I know, Rust itself would generate an error like that, even without that part of code.

This brings me to my main question: dropping that check would allow tighter cooperation with rstest, and I believe with other, future fixture frameworks too. Would Tokio accept a patch, dropping that check, like the one actix-net just accepted? As far as I can tell, everything else is already in place.

[hawkw]

Would Tokio accept a patch, dropping that check, like the one actix-net just accepted? As far as I can tell, everything else is already in place.

I think that would be fine. You're right that if a #[tokio::test]-annotated function is used without #[rstest] or another attribute that provides arguments, the #[test] attribute will generate an error anyway, so Tokio doesn't need to have its own error. I'd get input from other Tokio maintainers as well, but this seems reasonable to me.

[rubdos]

As discussed in the issue at rstest, this is not required for rstest to work at all. It could, however, be useful for different fixture frameworks, and it makes sense IMO to default to the default Rust error either way, instead of making Tokio's own error.