Clarification on multiple subscribers needed ..

[RuedigerMoeller]

I implemented multiple subscribers using a "wait-for-slowest-subscriber" policy. This means if of 2 subscribers only one sends "credits" via request(N), no message will be sent. So credits = Min(credits of subscribers)

this makes

required_mustRequestFromUpstreamForElementsThatHaveBeenRequestedLongAgo()
and required_spec104_mustCallOnErrorOnAllItsSubscribersIfItEncountersANonRecoverableError() fail.

But "wait-for-slowest-subscriber" (with optional timout/forced cancel policy) actually does not violate the spec, isn't it ?

[ktoso]

Hi @RuedigerMoeller, thanks a lot for catching this - you're absolutely correct.
These tests assume too much about the Publisher - namely that it just signals the elements once demand comes in. I've skimmed the impls now and it seems that amending them won't be too hard - I'll send in a PR soon to fix this.

[RuedigerMoeller]

Thanks for the quick response :). Kudos for the test suite, it discovered an awful lot of issues in my initial implementation ..

[ktoso]

Addressed one part of the issue in this PR: #287
Will try to make time and figure out required_mustRequestFromUpstreamForElementsThatHaveBeenRequestedLongAgo soon.

Thanks again for reporting! Such reports are super useful in making the tests generally useful for all future implementers :-)

[RuedigerMoeller]

Hi, afair required_mustRequestFromUpstreamForElementsThatHaveBeenRequestedLongAgo had the same problem in the test prologue.
Will unignore the testcase and retry when out of office ;)

[viktorklang]

@ktoso Was this fixed and merged?

[viktorklang]

@ktoso Sorry, I see the PR…

[ktoso]

Yeap, it's that one which I need to get back to: #287

[viktorklang]

@RuedigerMoeller Please see #414. Will be released in 1.0.2 which is due before end-of-year!