Friday July 1, 2011

[Time] NameMessage
[05:43] sustrik ssi: !
[06:22] the_hulk hi, can i build shared library for iphone
[06:26] the_hulk hi
[06:26] the_hulk can i build shared lib for ios?
[06:32] sustrik i think so
[06:32] sustrik check the mailing list archives
[06:32] sustrik if you find no help there, ask on the mailing list
[06:47] the_hulk not there in mailing list
[06:48] sustrik ask then
[06:48] sustrik i dimly recall there are people who've build it on ios sybscribed to the mailing list
[06:59] the_hulk i got this list
[06:59] the_hulk thread*
[07:00] the_hulk which leads me here
[09:21] the_hulk sustrik, compliled successfully!! will test now
[09:21] sustrik :)
[13:41] the_hulk hi
[13:41] the_hulk i have build an zmq for ios
[13:41] the_hulk it gets built and i get .a file
[13:42] the_hulk but when i include in xcode i get link error
[13:42] the_hulk also while building zmq i get messages like
[13:43] the_hulk '/Developer/Platforms/iPhoneOS.platform/Developer/usr/bin/ranlib: file: .libs/libzmq.a(libzmq_la-devpoll.o) has no symbols'
[13:43] the_hulk any idea
[13:52] sustrik no idea, try asking on the maling list
[14:38] ptrb I'm using the C++ wrapper, and deleting the zmq::context_t as a method of signaling to threads waiting in blocked recv() calls to abort. However, deleting the context_t appears to be blocking in zmq_term (actually mailbox_t::recv). Should I be doing something else instead?
[14:39] dev_sa exit
[14:46] ptrb exit(3) is one option, yes...
[14:46] ptrb a bit less elegant than I'd hoped.
[15:05] ptrb oh, closing the sockets properly helps.
[15:20] the_hulk tested on ios, at least unit test succeeds..
[15:39] ssi sustrik: I hope your "!" was a "cool!" instead of a "you're a moron!" :D
[15:40] sustrik cool
[15:40] sustrik :)
[15:40] ssi hehe good
[15:40] sustrik that's something nobody has attempted so far
[15:41] ssi it's going pretty well
[15:41] sustrik do you have any idea how to accomplish that
[15:41] sustrik ?
[15:42] ssi well, I have the framework for a provider library in place.
[15:42] sustrik i mean, JMS assumes there's a central broker
[15:42] ssi yeah, well here's my thought
[15:42] ssi either a) you run a lightweight broker which is just a zmq device
[15:42] ssi or b) you pass an embedded broker flag to one end
[15:42] ssi and that provider lib runs a little embedded queueing system and binds to the url you provided as brokerurl
[15:43] ssi then the consumers will connect to what they assume is a broker, but it's just the producer
[15:43] ssi that gives me point to multipoint brokerless, or multipoint to multipoint with a dedicated broker
[15:43] ssi it's going to do S&F only, because that's all we do here
[15:43] ssi it'll be a very stripped down JMS implementation that just handles our needs
[15:44] ssi we've had no end of problems with available JMS platforms
[15:44] ssi which is how I stumbled across zeromq in the first place :)
[15:45] ssi I'm slogging through all the boilerplate JMS stuff right now, factories to create all the different types of messages, but hopefully this afternoon I'll dig into the actual transport
[15:45] sustrik is it going to be open source?
[15:46] ssi I don't know if I'll be able to do that or not
[15:46] ssi if I can, I would like to
[15:46] ssi but my company doesn't make it easy
[15:46] sustrik that would be nice
[15:46] sustrik an gateway to enterprise :)
[15:46] ssi also, I don't intend it to be a full-featured jms implementation, so I don't know how that'll go over in the wild
[15:46] ssi jms in full is COMPLEX
[15:47] ssi and that's what I'm trying to avoid!
[15:47] sustrik the nice thing about open sourcing it would be that others can add features you don't care about
[15:47] ssi sure
[15:47] ssi we have a message pipelining platform here that's homegrown over the last 10 years
[15:47] ssi and I wrote version 3 recently which is ground-up, using zmq as internal transport
[15:47] ssi and it intrinsically supports zmq in place of jms for all its edge transport as well
[15:48] ssi but it won't be ready for widespread deployment in the org for awhile
[15:48] ssi meanwhile, we have a couple of properties who are mandating that we move off of activemq by september because it's been very unstable
[15:48] ssi so as we're evaluating jms providers, I'm gonna write this one and throw it in the mix
[15:49] ssi cross your fingers for me... in a year zeromq could be running :)
[15:50] sustrik wow!
[15:50] sustrik fingers crossed :)
[15:50] ssi anyway, lunch :D
[15:51] sustrik cyl
[17:39] ssi k I think I'm up against the point where I need to start writing transport :D
[21:34] pasz can a spontaneous zmq_send to a reply be made following a zmq_poll?
[21:35] pasz I have code that polls for new messages, however, sometimes the text reply doesnt get sent
[22:58] ssi I'm not sure how to avoid using sockets in multiple threads with what I'm trying to do
[22:58] ssi hrm maybe I am