Flaky console-subscriber integration tests

[hds]

What crate(s) in this repo are involved in the problem?

console-subscriber

What is the issue?

The console-subscriber integration tests show flaky behavior locally.

The tests were originally introduced in #452.

How can the bug be reproduced?

Run all the framework tests locally. Sometimes the self_wakes test fails.

To reproduce this, I ran the tests in a loop until they failed. On my machine it typically takes under 5 minutes to produce a failure.

echo "Start!" && time while [ $? -eq 0 ]; do cargo test -p console-subscriber --test framework; done

Logs, error output, etc

No response

Versions

• main branch.
• console-subscriber v0.2.0

Possible solution

It appears that the root cause for this issue is the one described in tokio-rs/tracing#2743.

Additional context

No response

Would you like to work on fixing this bug?

yes

[Rustin170506]

Here is another one: double_sleep_wakes
https://github.com/tokio-rs/console/actions/runs/6989256116/job/19017617547?pr=494

[hds]

These two tests are so flaky that I'm going to delete them until we fix the problem. From console-subscriber/tests/wake.rs:

#[test]
fn sleep_wakes() {
    let expected_task = ExpectedTask::default()
        .match_default_name()
        .expect_wakes(1)
        .expect_self_wakes(0);

    let future = async {
        sleep(Duration::ZERO).await;
    };

    assert_task(expected_task, future);
}

#[test]
fn double_sleep_wakes() {
    let expected_task = ExpectedTask::default()
        .match_default_name()
        .expect_wakes(2)
        .expect_self_wakes(0);

    let future = async {
        sleep(Duration::ZERO).await;
        sleep(Duration::ZERO).await;
    };

    assert_task(expected_task, future);
}