Closed kxfang closed 5 years ago
The code indicates that to be impossible, no?
Yeah, I spent like an hour trying to reproduce with different tests (we work together with @kxfang), it seems impossible to me too 🤷♀️
Are we sure the problem is here and not RxRelay? It's in the original trace.
On Thu, Feb 14, 2019 at 2:14 PM Artem Zinnatullin :slowpoke: < notifications@github.com> wrote:
Yeah, I spent like an hour trying to reproduce with different tests (we work together with @kxfang https://github.com/kxfang), it seems impossible to me too 🤷♀️
— You are receiving this because you commented.
Reply to this email directly, view it on GitHub https://github.com/JakeWharton/RxReplayingShare/issues/30#issuecomment-463754609, or mute the thread https://github.com/notifications/unsubscribe-auth/AAEEEcMkEhZFt1vxckCLj8VcRYFfGJMoks5vNbWsgaJpZM4a8Gty .
@artem-zinnatullin can correct me if I'm wrong, but we've had RxRelay in the code base for years and haven't had any similar issues. It's only when we started introducing ReplayingShare recently that we started seeing this crash affecting thousands of users. Maybe it's possible that the combination of the two is somehow causing issues?
Concurrent emission from upstream seems to be the only way to get into this state and replayinShare
is prob just a part of the chain here
It's however a bit weird that all stacktraces we have have replayingShare
in them, but that might just indicate a error pattern on our side
We stopped seeing these crashes after https://github.com/JakeWharton/RxReplayingShare/issues/31
Originally reported here: https://github.com/ReactiveX/RxJava/issues/6224
There's a possible protocol violation in ReplayingShare that calls
onNext
on the downstreamswitchMap
beforeonSubscribe
is called.