Friday October 7, 2011

[Time] NameMessage
[00:05] bar4ka i modified the sink class:
[00:06] bar4ka so it prints when the thread start
[00:07] bar4ka i got no prints, just the 'zmq::error_t' what(): Socket operation on non-socket
[00:33] minrk1 something weird is definitely going on
[00:33] minrk1 If I remove the close/delete, the ~Sink desctructor is getting called 8 times per constructor
[00:38] bar4ka i modified the Sink:
[00:38] bar4ka i could see the copy ctor got called 4 times
[00:38] minrk1 So you need to make sure that close is called only once
[00:39] minrk1 It would help if, instead of passing the socket to Sink, you passed the context and addr, then creating the socket in the thread
[00:40] bar4ka i was trying that but socket_t class ctor is private
[00:41] bar4ka this is the output when i run:
[00:42] minrk1 yes, because you are trying to close the same socket several times
[00:42] minrk1 (I count 8)
[01:00] bar4ka it really would be better creating the socket in the thread, the ctors and dtor is driving me crazy
[01:08] bar4ka i got free of the socket error
[01:08] bar4ka and thread finally started
[01:09] bar4ka now i got another zmq exception
[01:11] bar4ka terminate called after throwing an instance of 'zmq::error_t' what(): Operation not supported
[01:12] bar4ka new Sink:
[01:15] minrk1 you have sink/source backwards
[01:15] minrk1 sink==output, source==input
[01:16] minrk1 so you are trying to recv on a PUSH socket, which is not supported
[01:17] minrk1 What zmq are you using, btw?
[01:21] bar4ka 2.1.7
[01:27] minrk1 did swapping sink/source get you any further?
[01:28] bar4ka yes
[01:28] bar4ka the inproc is working fine
[01:29] bar4ka the tcp server did not receive, but it is a good progress
[01:30] bar4ka maybe it is something with the rebuild method
[01:31] minrk1 Are you sure you want to call rebuild?
[01:31] minrk1 rebuild closes and reinitializes the message
[01:31] minrk1 i.e. it discards any data that was there before
[01:32] bar4ka that is the problem
[01:32] minrk1 and gives you a clean new message to work with
[01:32] bar4ka i will comment that part
[01:33] bar4ka now its working
[01:34] bar4ka i just need to address memory so it does not leak
[01:35] bar4ka i will try to pass the context plus addr
[01:37] bar4ka minrk1: thank you for all the help
[01:37] minrk1 no problem
[02:18] bar4ka minrk1: do you know how can i use copy ctor with zmq::socket_t ?
[02:19] bar4ka i created another class:
[02:20] bar4ka but cant compile cause of the copy ctor:
[02:29] bar4ka looks like zmq::socket_t::socket_t(const zmq::socket_t&) is private
[02:30] bar4ka andrewhavck: hi
[03:49] abraxas good morning
[03:50] abraxas Does anyone know if it's possible to have identity-like behavior, but not for the entire connection. Instead, I'd really love to use it on a per-subscription basis.
[03:50] abraxas The subscription itself can move around my data center.
[03:50] abraxas But if publishers would keep a backlog, until the subscription is back (regardless of the machine that subscription comes from), that would solve my problem.
[06:34] aalok Greetings!
[09:55] lupine_85 so, I'm still reading through , but I was wondering if there's a common pattern for individual nodes to advertise presence to their upstream
[09:56] lupine_85 regular pings are an obvious one, I guess
[09:57] lupine_85 (say I have 20 client nodes connected to one server, and the server is making decisions about who to send messages to based on somthing other than simple round-robin - it needs to know which nodes are connected "now", surely?)
[10:13] lupine_85 hmm, ROUTER<->REP
[10:31] mikko howdy
[11:57] jd10 has now a JZMQ like API
[13:51] mile any idea how to make a listener listen to two ports with Cowboy?
[13:55] mile sorry, wrong channel
[17:21] mikko sustrik: there will be a few mails about failed builds as i am rebuilding the machines
[18:18] adymitruk socket per request?
[18:18] adymitruk too heavy?
[18:18] adymitruk for a web app..
[18:30] cremes adymitruk: sockets are pretty light; one per request is probably fine
[18:30] cremes just make sure you have enough file descriptors
[18:43] przemek Hello everyone, I'm trying to use 0MQ on Windows Server using Java
[18:44] przemek I'm following the Windows instructions here:
[18:44] przemek I've opened the project in VS2008 as instructed but I'm unsure what the following statement means
[18:44] przemek ote that ØMQ and JDK header file has to be on "include path" (Tools|Options|Projects and Solutions|VC++ Directories|Include files) and ØMQ libraries have to be on "library path" (Tools|Options|Projects and Solutions|VC++ Directories|Library files):
[18:45] przemek I'm assuming that if I Build this correctly I should get a jar file that I can then use in my Java project?
[19:44] przemek hello
[20:01] nag hi przemek_
[20:03] nag bye przemek_
[21:31] mikko evening
[21:46] michelp yo
[22:48] mattbillenstein hi all
[22:49] mattbillenstein building an rpc client where i have multiple backend rpc servers I could connect to — I'm trying to figure out how to manage failure conditions — like if one of the servers is down
[22:50] mattbillenstein using XREQ/XREP, I can always connect if a given server isn't alive, but sending a message will just hang infinitely
[22:51] mattbillenstein or rather, the receive is hanging
[22:58] mattbillenstein is there a way to tell if I can even connect to a particular server?
[23:50] espeed In the taskevent model (ventilator/worker/sink), what's the right way for a *worker* to add additional tasks?
[23:53] espeed I've tried having the worker connect to the PUSH socket the ventilator is bound to, but that doesn't seem to work.
[23:59] mikko hi
[23:59] espeed hi