Sinks.One<Void>.tryEmitValue no longer accepts null in 3.4.25/3.5.0

[programmatix]

Calling tryEmitValue(null) on a Sinks.One<Void> worked without issue in 3.4.24, but in 3.5.0, we can see a NPE:

java.lang.NullPointerException: e
               at java.base/java.util.Objects.requireNonNull(Objects.java:233)
               at reactor.util.concurrent.SpscArrayQueue.offer(SpscArrayQueue.java:51)
               at reactor.core.publisher.FluxPublishOn$PublishOnSubscriber.onNext(FluxPublishOn.java:230)
               at reactor.core.publisher.SerializedSubscriber.onNext(SerializedSubscriber.java:99)
               at reactor.core.publisher.FluxTimeout$TimeoutMainSubscriber.onNext(FluxTimeout.java:180)
               at reactor.core.publisher.FluxFlatMap$FlatMapMain.tryEmit(FluxFlatMap.java:543)
               at reactor.core.publisher.FluxFlatMap$FlatMapInner.onNext(FluxFlatMap.java:984)
               at reactor.core.publisher.Operators$MonoInnerProducerBase.complete(Operators.java:2744)
               at reactor.core.publisher.SinkOneMulticast.tryEmitValue(SinkOneMulticast.java:63)
               at reactor.core.publisher.SinkOneSerialized.tryEmitValue(SinkOneSerialized.java:38)
               at com.couchbase.client.core.transaction.util.ReactiveWaitGroup.lambda$done$1(ReactiveWaitGroup.java:99)

The Javadocs for tryEmitValue still indicate that null is a legitimate value:

Params:
value – the value to emit and complete with, or null to only trigger an onComplete

EmitResult tryEmitValue(@Nullable T value);

The important change may have been in SinkOneMulticast.tryEmitValue. in 3.4.24 this has special handling for null:

		if (value == null) {
			for (Inner<O> as : array) {
				as.complete();
			}
		}

while in 3.5.0 it doesn't, it just passes the null through:

		for (Inner<O> as : prevSubscribers) {
			as.complete(value);
		}

I can't find any mention of this change in the release notes, and it's a breaking one for us. Can you please confirm if this is intentional or a bug, e.g. do I need to patch our code?

[mikereiche]

This was working in 3.5.0-RC1 but is broken in 3.5.0. The change is 849be9a207958b111ee7a06f4b21596b0122972f.
Previously, if value was null, as.complete() was called with no argument.

[simonbasle]

this is actually a regression introduced in the 3.4.x branch (even though 3.4.25 was exceptionally released after 3.5.0).
re-targeting this for 3.4.26, which will be forward-merged to 3.5.1 as well

[simonbasle]

Regression introduced in #3260

[simonbasle]

simplest workaround:

T value;
EmitResult er = (value == null) ? sink.tryEmitEmpty() : sink.tryEmitValue(value);

[dnault]

Thank you for jumping on this so quickly, @simonbasle