Jobs and coroutines are typically expected to comp...

# coroutines

[Uli Bubenheimer]

Jobs and coroutines are typically expected to complete quickly after being cancelled. Is there a built-in way to track that they do, essentially a kind of runtime assertion? I believe this would help me find leaks.

I can do this on an individual Job/coroutine basis by adding some tracking code at the Job creation site, but I am looking for an easy way to do it at a broader level, like the runtime or Dispatcher.

[Zach Klippenstein (he/him) [MOD]]

cancelAndJoin?

[Uli Bubenheimer]

Umm, no

[Uli Bubenheimer]

Here's a complete example. job1 is cancelled by its parent, but never completes, because it does not suspend. I can detect this situation with something like the second launch. However, I have to use a check like this individually in every possible place where this could happen to detect it. I am looking for a hook to check for this situation automatically. I imagine I could likely do it with a custom Dispatcher, but that's not really feasible. Is there any built-in debug-like thingy that could detect this, or a Dispatcher or runtime hook, or maybe someone created a custom solution?

coroutineScope {
    val job1 = launch(<http://Dispatchers.IO|Dispatchers.IO>) {
        // bad job: runs for a long time, and fails to suspend
        while (true) {}
    }

    launch {
        // check that job1 completes soon after parent cancellation
        try {
            awaitCancellation()
        } finally {
            GlobalScope.launch {
                delay(5.seconds)

                if (!job1.isCompleted) println("Not completing")
            }
        }
    }

    // Wait a bit, then cancel our parent scope
    delay(5.seconds)

    this.cancel()
}

[Uli Bubenheimer]

cancelAndJoin() does not help, because in general there are many ways that job1 could end up getting cancelled.

[Uli Bubenheimer]

I want to detect that job1 is leaking, like a LeakCanary for coroutines