blockingFirst race condition

[RyanHurst]

Telemetry telemetry = new Telemetry();
telemetry.armed().subscribe(armed -> logger.debug("armed: " + armed));
logger.debug("blocking armed: " + telemetry.armed().blockingFirst());

so I’ve found a strange race condition. sometimes this will output:

16:52:26.985 [main] DEBUG io.dronecode_sdk.telemetry.Telemetry - Building channel to 127.0.0.1:50051
16:52:28.145 [grpc-default-executor-2] DEBUG io.dronecore.example.DroneCoreExample - armed: false
16:52:29.140 [grpc-default-executor-2] DEBUG io.dronecore.example.DroneCoreExample - armed: false
16:52:30.140 [grpc-default-executor-2] DEBUG io.dronecore.example.DroneCoreExample - armed: false
16:52:31.140 [grpc-default-executor-2] DEBUG io.dronecore.example.DroneCoreExample - armed: false

and then it will just repeat indefinitely

then sometimes this will output:

16:54:49.835 [main] DEBUG io.dronecode_sdk.telemetry.Telemetry - Building channel to 127.0.0.1:50051
16:54:51.246 [main] DEBUG io.dronecore.example.DroneCoreExample - blocking armed: false

Process finished with exit code 0

[JonasVautherin]

What happens here is that in the C++ SDK, we currently only register one callback. With your code above, it does register armed() twice, and therefore only one stays. We don't want that behavior in Java.

So the solution is, in the Java SDK, to share() the stream (Flowable) between all the callers, so you can call telemetry.armed().subscribe() as many times as you want. That's how we do in the Swift SDK already.