Something changed between v3.0.0-dev14 and v3.0.0-...

# apollo-kotlin

[Doug Chappelle]

Something changed between v3.0.0-dev14 and v3.0.0-alpha01 that is now causing one of my mutations to never return. I'm looking into it more but thought I would post this in case it rings a bell with anybody involved in making those changes.

[mbonnin]

Is this on JVM or Apple ?

[Doug Chappelle]

Android, so I would assume JVM

[mbonnin]

Yup, Android is JVM

[mbonnin]

Is this using a WebSocketNetworkTransport? or regular HttpNetworkTransport?

[Doug Chappelle]

WebSocketNetworkTransport

[mbonnin]

Alright. Yep, there have been a few changes around this

[mbonnin]

Maybe https://github.com/apollographql/apollo-android/pull/3214 ?

[mbonnin]

Mmm no that mainly touches AppSync

[mbonnin]

The other candidate would be https://github.com/apollographql/apollo-android/pull/3217

[Doug Chappelle]

And the winner is... https://github.com/apollographql/apollo-android/pull/3220

[mbonnin]

Oh wow that's interesting.... I wouldn't expect this to do anything with mutations.

[mbonnin]

I'm afk right now but will dig a bit more later today

[Doug Chappelle]

np, thx

[mbonnin]

This PR is relatively small. The only thing that could maybe explain a behaviour change is the change from flatMapLatest to flatMapConcat (here).

[Doug Chappelle]

When I get a chance I'll try changing it back and see what happens... get some sleep

[mbonnin]

Thanks!

[mbonnin]

IIRC, this was made so that emissions from the ApolloCacheInterceptor would not suspend (so that we can check the cache status just after receiving the response, else receiving the response is always delayed to after potentially writing the cache)

[mbonnin]

So this changes the supension points. flatMapLatest is introducing an intermediate channel that maybe prevents a deadlock or something like this

[mbonnin]

Going to sleep too, have a good night!