It seems as if this is not always reproducible. I think what is happening is the "Hello" message gets sent to the server in some cases, then the server never sends a response (it was killed), and the client is left looking for a response.
Other times, the server is killed before the client sends the "Hello" message, so when the server comes back it sends this message and then works properly again. This is a valid error condition right? -- Wolf On Sun, 2010-08-22 at 13:17 -0400, Wolfgang Richter wrote: > I was wondering if this was expected behavior. > > Working with the REQ-REP Hello World example in C with 0MQ release > 2.0.7(beta), I have noticed that zmq_recv in my client will block > forever if I kill the server prematurely. > > Even if I bring the server back up. > > I was expecting that once the server came back, the client would > continue, or somehow signal an error condition (or an error condition on > the server). > > I have noticed that if I use ZMQ_NOBLOCK on my zmq_recv and loop > whenever I get errors, when the server comes back up things continue > making progress as expected. > _______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
