Frédéric Déléchamp

It would be useful to see such a C code. Meanwhile, I'll PR a test demonstrating the loss of subscriptions when far exceeding HWM, together with a fix made by...

I would vote for libzmq, as Trevor stated this is the reference implementation.

Hi, NPE at this precise location would mean that a mailbox has been destroyed. I would suspect either: - socket has been closed - IO thread or reaper has been...

Hi, did you have a look on the test PubSubHwmTest ? It seems that PUB sockets seems to be not lossy, with option ``` ZMQ.setSocketOption(pub, ZMQ.ZMQ_XPUB_NODROP, true); ``` when sending...

You did not use Socket.setXpubNoDrop, in order to disable lossy behaviour. As a general rule, errno should be checked only in case of error. That's the current behaviour that could...

It's difficult to reproduce on my machine, I need to reduce the sleep between calls drastically (10ms). I tried to change the transport to inproc:// and I still observe the...

I still see something here: ``` private static final long THRESHOLD = TimeUnit.MICROSECONDS.toNanos(500); private static final long MAX = 10_000_000; @Test public void testSporadicLatency() { zmq.Ctx ctx = zmq.ZMQ.createContext(); zmq.SocketBase...

I made the upgrade to 11, and used ` -XX:+UnlockExperimentalVMOptions -XX:+UseEpsilonGC` options. No latency observed on both tests.

That's interesting. As far as I remember, once a socket goes through the configuration path (bind, connect,...) and enter communication path, the Signaler is mostly used for commands related to...

Hi, do you have an example of code demonstrating the problem?