#[test_log::test(test_case::test_case)] doesn't work with multiple cases

[tamird]

Spin-off from #30.

It is not possible to write #[test_case(...)] tests with multiple cases. For example

#[test_log::test(test_case::test_case(-2, -4; "one"))]
#[test_log::test(test_case::test_case(-4, -2; "two"))]
fn it_works(x: i8, y: i8) {
  // ...
}

does not work.

I think the approach of wrapping other test macros such as tokio::test and test_case::test_case in the
test_log::test macro ends up being inflexible for such cases and it's probably better to abandon it in favor of
parsing other attributes on the test function and inserting #[test] only if nothing else has already done it. This
would mean that ordering of attributes would start to matter but this is already the case when test_case::test_case is
composed with tokio::test; test_case has to come first or else compilation fails with "functions used as tests can not have any arguments".

[d-e-s-o]

it's probably better to abandon it in favor of parsing other attributes on the test function and inserting #[test] only if nothing else has already done it

I am not sure I completely understand the proposal. How would we be able to detect that fact? Aren't both tokio::test and test_case::test_case just arbitrary strings? Could you sketch the proposal out some more, please?

[tamird]

I'm afraid I waited too long to file this proposal and have forgotten some of the details. That said it's empirically true that test_case plays well with tokio::test; we should be able to use the same technique they do.

[jtojnar]

Alternately, maybe it would be possible to offer a separate attribute for those cases when passing it as an attribute is not sufficient.

For example tracing_test uses a separate attribute. Of course, user will then need to be careful to order the attributes correctly.