NPE during NettyClientTransport.start() if keepAlive is enabled

[bobwenx]

What version of gRPC are you using?

GRPC 1.1.2

What JVM are you using (java -version)?

1.8.0_112

What did you do?

Just simply call io.grpc.netty.NettyChannelBuilder#enableKeepAlive(boolean) during channel build. and the grpc always throw NPE exception.

java.lang.NullPointerException: null
	at io.grpc.netty.NettyClientTransport.start(NettyClientTransport.java:169)
	at io.grpc.internal.ForwardingConnectionClientTransport.start(ForwardingConnectionClientTransport.java:44)
	at io.grpc.internal.TransportSet.startNewTransport(TransportSet.java:233)
	at io.grpc.internal.TransportSet.obtainActiveTransport(TransportSet.java:203)
	at io.grpc.internal.ManagedChannelImpl$3.getTransport(ManagedChannelImpl.java:739)
	at io.grpc.internal.ManagedChannelImpl$3.getTransport(ManagedChannelImpl.java:677)
	at io.grpc.PickFirstBalancerFactory$PickFirstBalancer$1.get(PickFirstBalancerFactory.java:129)
	at io.grpc.internal.DelayedClientTransport$2.run(DelayedClientTransport.java:271)
	at java.util.concurrent.ForkJoinTask$RunnableExecuteAction.exec(ForkJoinTask.java:1402)
	at java.util.concurrent.ForkJoinTask.doExec(ForkJoinTask.java:289)
	at java.util.concurrent.ForkJoinPool$WorkQueue.runTask(ForkJoinPool.java:1056)
	at java.util.concurrent.ForkJoinPool.runWorker(ForkJoinPool.java:1692)
	at java.util.concurrent.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:157)

and inspect the code(io.grpc.netty.NettyClientTransport#start):

 @SuppressWarnings("unchecked")
  @Override
  public Runnable start(Listener transportListener) {
    lifecycleManager = new ClientTransportLifecycleManager(
        Preconditions.checkNotNull(transportListener, "listener"));

    if (enableKeepAlive) {
      keepAliveManager = new KeepAliveManager(this, channel.eventLoop(), keepAliveDelayNanos,
          keepAliveTimeoutNanos);
    }
    ....
}

i believe the field channel is never initialized right now, so NPE is throw by field call channel.eventLoop().

[ejona86]

Yeah... that looks broken. channel is initialized in the same method on line 197. @lukaszx0, what to make a fix, and figure out why a test didn't catch this?

[lukaszx0]

@ejona86 yes, I was planning to look into that as soon as I'll find some free time. Hopefully this week.

[bobwenx]

@ejona86 When is this fix be released? KeepAlive support is a important feature for us(and that's why we upgrade to 1.1.x), Hope this fix release quickly.