Question: Observable stuck in case StackoverflowException thrown inside Task {}, is it expected?

[GrigorievNick]

import monix.eval.Task
import monix.execution.Scheduler.Implicits.global
import monix.reactive.Consumer
import monix.reactive.Observable
import org.scalatest.FunSuite

import scala.concurrent.duration._

class MonixExceptionTest extends FunSuite {

  test("fatal") {
    Observable
      .fromIterable(0 to 10)
      .mapParallelUnordered(4)(i =>
        Task {
          try throw new StackOverflowError(s"$i")
          catch {
//            case x: Throwable => throw new RuntimeException(x)
            case x: Throwable => throw x
          }
          i
        }
      )
      .consumeWith(Consumer.foreach(println))
      .runSyncUnsafe()
  }

}

this code will be stuck forever.
But if we wrap Stackoverflow with RuntimeException, Observable will fail.
Do this expected behave?

[ghik]

Yes, this is because StackOverflowError is considered fatal and unrecoverable.

Traditionally, all Errors are considered fatal (as opposed to Exceptions). Scala has a slightly refined definition of what is fatal (see scala.util.control.NonFatal extractor) but StackOverflowError falls into both definitions so Monix will not try to catch and recover from it.

[GrigorievNick]

OK, but still I expect to fail observable, rather then just silently stuck.
Or I miss understand something?

[ghik]

For an Observable to fail on StackOverflowError, it would have to catch it at some point. But fatal errors should never be caught. They should be allowed to crash the process.

Although I remember that at some point NonFatal did not recognize StackOverflowError as fatal so it may be somewhere in a gray area.

[GrigorievNick]

I am not sure that problem on Observable side, look like Task is never complete with fail or success - no matter.

[ghik]

Yes, but Task doesn't catch it for exactly the same reason.

[GrigorievNick]

OK, so what is way to deal with this issue: improve error handling on Monix side, or always wrapp code inside Task with try-catch, or add timeout to every Task?

[GrigorievNick]

Do you happen to have any news on this?
Code must throw an exception and not be silently stuck even without logs.