Race condition in test_simple_match

[jlebon]

In test_simple_match(), we have:

    journal::send(&[&filter, &msg]);
    assert!(j.match_flush().unwrap().match_add(key, value).is_ok());
    let r = j.next_record().unwrap();
    assert!(r.is_some());

But the issue is that the next record might not immediately be available. There's a delay for the daemon to process it and then it finally making it available to open cursors. Adding:

     journal::send(&[&filter, &msg]);
     assert!(j.match_flush().unwrap().match_add(key, value).is_ok());
+    thread::sleep(time::Duration::from_millis(1000));
     let r = j.next_record().unwrap();

as a test fixes it for me, though a cleaner course of action would probably be to expose the notification API and make use of it in the tests?

[lucab]

expose the notification API

Do you have a reference to which API is that?
I just double-checked in go-systemd testsuite and indeed artificial 1s pauses got inserted in all roundtrips there.

[jlebon]

Do you have a reference to which API is that?

See the functions at https://www.freedesktop.org/software/systemd/man/sd_journal_get_fd.html, though I've never made use of them myself.

[codyps]

Ya, this should use the Journal::wait() method. These tests been modified a bit since this comment: they now iterate over journal entries rather than assuming it'll get the right one immediately.