Tests, ignore with reason

[blonk]

Is it possible to "report back" to the test framework that a test has been ignored for a reason?

We have a client library that has a few tests that require a (test) server. If there's no server available, I want a discreet ignored (server is not running) message rather than an error dump.

I saw some brainstorming somewhere a while back that had some kind of predicate function used to determine if a test should run. IIRC it looked something like this:

#[test]
#[skip(predicate = some_function)]
fn dostuff() {
}

This looks super sweet -- but did it ever go beyond brainstorming?

[erelde]

The ignore attribute takes a reason argument

https://doc.rust-lang.org/reference/attributes/testing.html#the-ignore-attribute

That would allow you to run the test when you know the service is running. But I guess what you want is a dynamic check.

[blonk]

Yeah, the dynamic part is important, unfortunately.

We tried simply failing the test when no (test) server is available, but it causes too much noise in the output.

[jofas]

tracing logs aren't captured by libtest per default. So a quick and dirty workaround (that you could abstract into a self-made #[skip] attribute macro) would be to emit a warning when the test is skipped:

#[test]
fn test_requiring_server() {
    tracing_subscriber::fmt()
        .event_format(tracing_subscriber::fmt::format().without_time())
        .with_target(false)
        .init();

    #[tracing::instrument]
    fn test_requiring_server() {
        tracing::warn!("skipped function because predicate was true");
    }

    fn predicate() -> bool {
        true
    }

    if predicate() {
        test_requiring_server();
        return;
    }

    // here the test code
    assert!(true);
}

Output:

 WARN test_requiring_server: skipped function because predicate was true
test test_requiring_server ... ok

[blonk]

That's a pretty neat trick!