Hi Martin, That sound reasonable, but I don't think it is the case here. I bumped up the sleeps, and add printlns and flushes. What I am seeing (with the latency test) is that the error happens before local_lat even gets to the sleeping part.
Again, the fishy thing is GC. As soon as GC happens on the local_lat side, boom! Stu > Hi Stuart, > >> If I take the Java bindings, and run the latency or throughput test, >> things work fine. But if I jack the iterations up by an order of >> magnitude, I get the following error: >> >> Exception in thread "main" java.lang.Exception: 156384765 - 0x9523dfd >> - Context was terminated at local_lat.main(local_lat.java:45) >> >> at org.zeromq.ZMQ$Socket.recv(Native Method) at >> remote_lat.main(remote_lat.java:49) >> >> This error can also be reproduced by increasing the message size >> (instead of the iteration count). >> >> If I turn on verbose:gc, the error seems to happen immediately after >> the first GC, which would suggest a bug in the native code cleaning >> up a resource. >> >> I am running 2.0.7 plus the latest jzmq, on Mac OS X Snow Leopard. >> >> Any suggestions? > > That's because remote_thr sends the messages and waits for 10 secs > before exiting the process. If there's too much messages to send they > may not make it in 10 secs. Try increasing the sleep interval in remote_thr. > > Martin > _______________________________________________ > zeromq-dev mailing list > [email protected] > http://lists.zeromq.org/mailman/listinfo/zeromq-dev _______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
