fix: append #[::core::prelude::v1::test] only if it does not exist

[kezhuw]

This way if preceding test macros add #[::core::prelude::v1::test] by appending, then we can avoid duplicated test runs.

See also frondeus/test-case#101, frondeus/test-case#143

Closes #35.

This pr depends on frondeus/test-case#143 and tokio-rs/tokio#6497.

[d-e-s-o]

Thanks for the proposal! I am fine with the approach. Left a few nitpicks.

This pr depends on frondeus/test-case#143 and tokio-rs/tokio#6497.

Can we please wait until these are merged? I'd rather not include any temporary patches.

[kezhuw]

@d-e-s-o Thank you for reviewing. I am good with the merge timing.

I have pushed with a fixup commit to solve your concerns. Also, I checked that there is no duplicated test runs for new test cases. cargo test with_existing runs 2 tests, cargo test with_append runs 7 tests including ones generated by test_case.

[d-e-s-o]

It seems as if the tokio change has made it in and test-case can't get its act together. @kezhuw I think we use test-case only for exercising attribute parsing with special-sauce syntax. If you know of a different crate supporting similar non-standard syntax, then we can just switch existing tests away from test-case.

[kezhuw]

I think we use test-case only for exercising attribute parsing with special-sauce syntax.

Do you mean tests like this ?

#[test_log::test(test_case::test_case(-2, -4; "my test name"))]
fn with_inner_test_attribute_and_test_args_and_name(x: i8, y: i8) {
  assert_eq!(x, -2);
  assert_eq!(y, -4);
}

This pr does not depend on this. I can drop some tests so that we don't need to patch test-case crate.

[kezhuw]

I can drop some tests so that we don't need to patch test-case crate.

Those tests should pass once frondeus/test-case#143 published. Or I could #[ignore] and doc them.

[d-e-s-o]

Right, I understand it's just a development thing and all that, but my point is that we don't want to loose the coverage nor do we want to depend on some unpublished snapshot (or add ignore broken tests, for that matter).

So the easiest way out that I can think of would be to just use some other proc macro that supports more or less arbitrary syntax in test-case's stead. That would then unblock this PR. That it would (presumably) break test-case interaction is something I can live with, given that at this point that crate seems to be abandoned and we cannot freeze the entire eco system because of that.

[kezhuw]

we don't want to loose the coverage

For test coverage, the following two methods are identical in test-log side except that it will fail without patch to test-case. Besides, tokio::test also accept attributes.

#[tokio::test]
#[test_log::test]
async fn with_append_test_attribute_and_async() {
  assert_eq!(async { 42 }.await, 42)
}

#[test_case::test_case(-2, -4)]
#[test_case::test_case(-2, -5)]
#[test_log::test]
fn with_append_test_attribute_and_test_args(x: i8, _y: i8) {
  assert_eq!(x, -2);
}

So the easiest way out that I can think of would be to just use some other proc macro that supports more or less arbitrary syntax in test-case's stead.

Let me figure out.