ZeroMq IRC Log

Friday December 10, 2010

[Time] Name	Message
[05:55] samBiotic	could someone please help me get my ruby bindings working on osx.
[05:56] andrewvc	samBiotic
[05:56] andrewvc	what's going on on OSX?
[05:56] samBiotic	zeromq has been installed with homebrew
[05:56] samBiotic	brew install zeromq
[05:56] andrewvc	also, which lib, ffi-rzmq or rbzmq
[05:57] andrewvc	hmm, but which ruby binding
[05:57] andrewvc	have you installed either of the zeromq gems yet?
[05:57] samBiotic	yes
[05:57] andrewvc	which one?
[05:58] samBiotic	gem install zmq -- --with-zmq-dir=/usr/local
[05:58] andrewvc	hmmm, I installed zmq for the tarball, and gem install zmq worked fine
[05:58] andrewvc	*from
[05:58] andrewvc	no idea about homebrew though
[05:58] samBiotic	checking for zmq_init() in -lzmq... no
[05:58] andrewvc	you could just try using the zmq tarball
[05:59] samBiotic	hmm yes i could
[05:59] andrewvc	I've never installed it through brew myself
[05:59] samBiotic	ok will try now
[05:59] andrewvc	cool, I know more about the search paths for ffi-rzmq
[05:59] andrewvc	you might want to consider using ffi-rzmq btw
[05:59] samBiotic	ok
[06:00] andrewvc	unless you have a specific need for the slight speed boost of rbzmq (just my $0.02)
[06:00] samBiotic	my aim is to build an open sound control multicaster eventually
[06:00] andrewvc	cool project
[06:01] samBiotic	i have forward scheduling of packets working but time to live reductions would be preferable
[06:02] Steve-o	samBiotic: is that like the BBCs project?
[06:02] samBiotic	what would that be?
[06:03] andrewvc	samBiotic, with that kind of time constraint, wouldn't ruby's GC cause issues?
[06:03] Steve-o	they have a music collaboration tool
[06:03] andrewvc	at least MRI
[06:04] samBiotic	andrewvc: possibly, using this http://opensoundcontrol.org/implementation/ruby-osc-implementation
[06:04] Steve-o	built upon Kamaelia http://www.bbc.co.uk/opensource/projects/kamaelia/
[06:04] samBiotic	Steve-o: looks interesting
[06:05] andrewvc	if you have GC issues btw, jRuby might be the way to go.
[06:05] andrewvc	plus, you'll get real multithreading if you need it
[06:06] andrewvc	the Kamaelia site seems abandoned
[06:07] Steve-o	try here, http://www.kamaelia.org/Home.html
[06:07] Steve-o	broken default page
[06:07] andrewvc	oh, there we go
[06:09] Steve-o	kamaelia jam from a gsoc 2008 project, http://www.bbc.co.uk/blogs/bbcinternet/2008/05/google_summer_of_code.html
[06:09] samBiotic	basically, i aim to eliminate network jitter at a slight expense of increased but static delay by forward syncronising timetagged packets
[06:10] Steve-o	more like pulseaudio then?
[06:11] samBiotic	control data only
[06:15] Steve-o	ok, similar to the pro light control system, what ever that is called
[06:15] Steve-o	DMX?
[06:18] samBiotic	yeah kind of, it's just a network procotol developed by MIT. http://opensoundcontrol.org/introduction-osc
[06:18] samBiotic	sorry CNMAT
[06:19] andrewvc	samBiotic what's zeromq bring to the table that the OSC protocol doesn't?
[06:20] andrewvc	I'm not familiar with OSC at all
[06:20] Steve-o	I'm guesing OSC over 0MQ
[06:20] samBiotic	thats what i was going to test
[06:20] Steve-o	whilst DMX is commercially licensed
[06:26] andrewvc	btw, did you ever get it up and running on OSX steve?
[06:26] andrewvc	errr Sam
[06:27] samBiotic	still at it
[06:27] andrewvc	I should mention, if you use zerocopy the ffi binding can beat out the speed of the C binding. Not sure if you need to actually open the msgs as they go through
[06:28] samBiotic	trying gem install ffi-rzmq now
[06:28] andrewvc	btw, some versions of the ffi gem don't play well with ffi-rzmq
[06:29] samBiotic	seems to have work with my zeromq homebrew install
[06:29] andrewvc	nice
[06:29] andrewvc	I think the most recent one is fine, but if you have weird hanging issues related to threads, try building it from the ffi github repo
[06:29] samBiotic	worked*
[06:30] andrewvc	cool
[07:00] samBiotic	andrewvc and Steve-o thanks for the help
[07:00] samBiotic	bye for now
[14:25] matias	hi all :-)
[14:39] Steve-o	mikko: I've backported many fixes to PGM 5.0 but ANSI strict is a bit too much work, therefore 5.1 must be used for Cygwin/Solaris x6/ANSI strict/OSX
[14:40] mikko	Steve-o: ok
[14:40] mikko	maybe we should remove the strict flags for solaris for now
[14:40] mikko	until we can use 5.1 in zeromq
[14:41] Steve-o	I need to run through all the tests then I can ship a 5.0.93 and the first 5.1 release
[14:43] Steve-o	Sun ONE is odd as ever, it wants __asm__ only in C++ ANSI strict
[14:44] Steve-o	every other combination is fine with __asm
[15:29] mikko	sustrik_: how does HWM interact with ZMQ_QUEUE?
[15:32] sustrik	mikko: you have to set it by hand
[15:32] sustrik	i.e.
[15:33] sustrik	create socket 1, sreate socket 2
[15:33] sustrik	set hwm on 1. set hwm on 2
[15:33] sustrik	zmq_device (ZMQ_QUEUEU, s1, s2);
[15:35] mikko	ok, so in order to get messages swapped out when there is no consumer on the back socket i would just set on s2 ?
[15:42] mikko	Assertion failed: rc == 0 (pipe.cpp:187)
[15:43] mikko	swap init
[15:43] mikko	makes sense now
[15:43] mikko	so permission issues on swap file causes an assertion
[15:43] mikko	it looks like
[15:44] mikko	maybe hte swap location should be a sockopt
[15:44] mikko	as for example this daemon chdirs to / and drops privileges
[15:44] mikko	so it makes sense that it cant write
[15:44] sustrik	mikko: yes
[15:45] sustrik	it's a problem
[15:47] sustrik	zmq should not fail
[15:47] sustrik	however, i am not sure what the solution should be
[15:47] mikko	in this case it would be easy to return error and set errno to eperm
[15:48] sustrik	it happens on setsockopt?
[15:48] mikko	it happens later
[15:48] sustrik	i would say it should happen during setsockopt
[15:49] sustrik	the async error later on is just confusing
[15:49] mikko	thats impossible
[15:49] sustrik	why so?
[15:49] mikko	you create sequential files
[15:49] mikko	someone might have one file present with same name which is not writable
[15:49] mikko	maybe the closes you can get is:
[15:49] mikko	a) give a sockopt for setting swap dir
[15:50] mikko	b) check that swap dir is writable
[15:50] mikko	c) leave the rest to user
[15:50] sustrik	what i mean was to crate the file would be created directly when setsockopt(SWAP) is called
[15:50] sustrik	rather than later on
[15:50] sustrik	thus the credentials would be checked immediately
[15:50] mikko	yes, that would work as well
[15:51] sustrik	that should be rathereasy
[15:51] mikko	and allow setting swap file name by sockopt
[15:51] sustrik	yes, that's an option
[15:51] mikko	if not set default to current
[15:52] mikko	a lot of daemons won't have write permissions to $PWD
[15:52] sustrik	right
[15:52] sustrik	can you fill an issue in the bugtracker so that we don't forget about it?
[15:54] mikko	sustrik_: even better, i can create an issue and try to create patch tonight
[15:57] mikko	issue 140
[15:58] sustrik	:)