TCK Improvement: Throw meaningful error in white box subscriber verification when Probe.registerOnSubscribe isn't called

[jroper]

Currently, none of the the SubscriberWhiteboxVerification tests check whether WhiteboxSubscriberProbe.registerOnSubscribe is invoked before it starts trying to use the puppet supplied by that invocation. So, if it wasn't invoked, it just throws an NPE:

java.lang.NullPointerException
	at org.reactivestreams.tck.SubscriberWhiteboxVerification$1.run(SubscriberWhiteboxVerification.java:83)
	at org.reactivestreams.tck.SubscriberWhiteboxVerification.subscriberTest(SubscriberWhiteboxVerification.java:497)
	at org.reactivestreams.tck.SubscriberWhiteboxVerification.required_exerciseWhiteboxHappyPath(SubscriberWhiteboxVerification.java:80)

It would be nice if this was checked, and a meaningful error message (saying that registerOnSubscribe hasn't been invoked) was thrown.

[ktoso]

Sounds like a good improvement, are you giving it a shot or should I try to next week (won't be able this week)?

[jroper]

I'll take a look today.

[viktorklang]

@jroper If the promise hasn't been completed at the time of trying to use it, it should yield a test failure:

public T value() {
  if (isCompleted()) {
    return _value;
  } else {
    env.flop("Cannot access promise value before completion");
    return null;
  }
}

Perhaps that failure is hidden by an NPE? In which case, we might want to both flop AND throw?

[viktorklang]

@jroper are these tests running without runDefaultInit? https://github.com/reactive-streams/reactive-streams-jvm/blob/master/tck/src/main/java/org/reactivestreams/tck/SubscriberWhiteboxVerification.java#L541

[viktorklang]

@jroper I'm planning on issuing the 1.0.2-RC1 within 24h and would love to include this improvement as well. Will you have time to take a stab at it, or can provide a "failing test" someone could use to verify if it gets fixed? Thank you!

[viktorklang]

@jroper 6h to go

[akarnokd]

@viktorklang Do you want the fix for this included in RC1? If so, I can have a look and post a fix today.

[viktorklang]

@akarnokd That'd be very nice. If the fix is small then it should be quick to inorporate (in time for RC1)

[akarnokd]

The OP is not very detailed on how to trigger this failure so I have to figure that out too. Plus, "none of the the SubscriberWhiteboxVerification tests check" indicates all of them is affected and should be tested/modified.

[viktorklang]

Ideally the fix can be applied in one place, but let's see what it turns out to need.

[ktoso]

Please assign milestone @viktorklang ?

[viktorklang]

@ktoso Done!