Command completion race condition

[d4rken]

Command completion race condition

[d4rken]

From Observable.zip(...)

The operator subscribes to its sources in order they are specified and completes eagerly if one of the sources is shorter than the rest while disposing the other sources. Therefore, it is possible those other sources will never be able to run to completion (and thus not calling doOnComplete()). This can also happen if the sources are exactly the same length; if source A completes and B has been consumed and is about to complete, the operator detects A won't be sending further values and it will dispose B immediately.

I think zip disposes one of the Harvesters prematurely before onErrorReturnItem can get us completely out of the OnErrorNotImplementedException situation.

[d4rken]

In hindsight I think this change wasn't actually the solution.

io.reactivex.exceptions.OnErrorNotImplementedException: Upstream completed prematurely.
	at io.reactivex.e.b.a$af.accept(Functions.java:1704)
	at io.reactivex.e.h.k.onError(LambdaSubscriber.java:79)
	at io.reactivex.i.b.onError(PublishProcessor.java:2325) <----------
	at eu.darken.rxshell.cmd.Harvester$BaseSub.onError(Harvester.java:96)

This wasn't a bug in the library but in a supplied Processor that didn't have an error listener.

No harm done though, the current code is still better than previously.