ZeroMq IRC Log

Monday December 13, 2010

[Time] Name	Message
[12:09] mikko	sustrik: hi
[12:13] sustrik	hi
[12:16] mikko	again, thinking out loud here: extending uris to allow shortcut for setting socket options
[12:16] mikko	tcp://127.0.0.1:5555?linger=1000&identity=mytestsocket
[12:17] mikko	although this could be done in the user application as well
[12:17] sustrik	can be useful but should be implemented as a layer on top of 0mq
[12:17] sustrik	exactly
[12:17] sustrik	the problem is that this kind of thing means wandering far away from POSIX
[12:17] ianbarber	maybe a standard function in zfl for creating sockets from a url like that?
[12:18] sustrik	for example
[12:18] sustrik	it's useful, so it's worth of implementing it
[12:24] sustrik	mikko: btw, i've added your idea about swap->device to 3.0 roadmap
[12:24] sustrik	http://www.zeromq.org/docs:3-0
[12:25] mikko	sustrik: cool
[12:26] mikko	i was thinking about the control channel
[12:26] mikko	it's not necessary to do it over a remote management
[12:26] mikko	as if there was a device handle (void *) like with sockets then people could build their own if that is a requirement
[12:26] mikko	lunch ->
[12:29] sustrik	i don't follow
[12:29] sustrik	people can build their own device even now, no?
[12:30] mrm2m	Hey ho!
[12:31] ianbarber	sustrik: i think the idea was to allow control flow outside the device, so the device runs in another thread, and the control is done through the parent process.
[12:32] sustrik	hi
[12:32] sustrik	ianbarber: sure
[12:32] sustrik	what i meant was that you can do such thing even today
[12:32] mrm2m	Is there a one directional socket type? I've got some data I'd like to send to a server, but I'm not interested in any message if the data was received correctly or even if it was received.
[12:33] ianbarber	push
[12:33] sustrik	mrm2m: PUB/SUB presumably
[12:33] sustrik	it's completely uni-directional
[12:33] ianbarber	sustrik: yeah, you can, but only if you create the device yourself - though I guess you could wrap a device available from elsewhere with one your own
[12:34] mrm2m	Ah - I thought that was for one PUB sending to several SUBs.
[12:34] sustrik	yes
[12:34] sustrik	and what do you need instead?
[12:34] mrm2m	I need to send from several PUBs to one SUB.
[12:34] sustrik	that would work as well
[12:34] sustrik	ianbarber: ah, the plan is to move devices out of 0mq core anyway
[12:35] sustrik	and make them separate projects
[12:35] ianbarber	ah, ok. that makes sense
[12:35] sustrik	so you could have different types of devices
[12:35] sustrik	the one with control socket propsed above could be one available device
[12:35] sustrik	etc.
[12:35] mrm2m	sustrik: Ok. Then I'll have a look on SUB/PUB.
[13:19] CIA-20	zeromq2: 03Dhammika Pathirana 07master * r22b2b9a 10/ (src/tcp_listener.cpp src/tcp_listener.hpp):
[13:19] CIA-20	zeromq2: fix overwriting errno on bind failure
[13:19] CIA-20	zeromq2: Signed-off-by: Dhammika Pathirana <dhammika@gmail.com> - http://bit.ly/eB0FKT
[14:17] ianbarber	yay, that's all the zguide examples translated into php :) committed the version one I needed mikko's patch for
[14:26] mikko	sustrik: pipe.cpp:56
[14:26] mikko	i reckon that should exit the device rather than assert
[14:26] sustrik	let me see
[14:27] sustrik	i have a comment on that line :(
[14:27] sustrik	can you paste the line here?
[14:27] mikko	errno_assert (false);
[14:27] mikko	after rc = zmq_poll (&items [0], 2, -1);
[14:27] mikko	i got a signal handler in my process that catches SIGINT
[14:28] mikko	it seems to however cause zmq_poll to return
[14:28] mikko	and causes following assertion:
[14:28] mikko	^CInterrupted system call
[14:28] mikko	false (device.cpp:56)
[14:28] mikko	Aborted
[14:28] mikko	if i don't catch sigint then the program exits immediately
[14:29] sustrik	device or pipe.coo?
[14:29] sustrik	cpp
[14:29] mikko	sorry, device.cpp
[14:29] sustrik	ah, have it
[14:29] sustrik	yes, it works that way
[14:30] sustrik	what should be presumably done
[14:30] sustrik	is to return EINTR from the zmq_device call
[14:31] mikko	yeah, would allow clean termination of parent program
[14:31] mikko	as running device in a thread and then doing pthread_cancel doesn't seem particulary clean either
[14:32] mikko	im surprised that it doesn't assert on pthread_cancel
[14:34] sustrik	i assume that just kills the thread at the spot
[14:34] sustrik	so it has no chance to assert
[14:37] mikko	looking at the code, i think it shouldn't assert at all
[14:38] mikko	as the caller can handle device exiting
[14:41] CIA-20	zeromq2: 03Dhammika Pathirana 07master * rf749f2d 10/ (src/socket_base.cpp src/socket_base.hpp):
[14:41] CIA-20	zeromq2: add basic uri validations
[14:41] CIA-20	zeromq2: Signed-off-by: Dhammika Pathirana <dhammika@gmail.com> - http://bit.ly/hptlRa
[14:41] sustrik	mikko: good point
[14:43] mikko	you could even send the original errno back
[14:46] sustrik	yes
[14:59] mikko	sustrik: https://gist.github.com/cc8ca1efb295254bc634
[15:03] sustrik	mikko: looks good
[15:03] sustrik	can you send it as patch to the ml?
[15:20] sustrik	mikko: checking your swapdir patch
[15:20] sustrik	what's the point of having 3 test functions?
[15:20] sustrik	+ static bool test (const std::string& directory_);
[15:20] sustrik	+
[15:20] sustrik	+ static bool test (const char* directory_);
[15:20] sustrik	+
[15:20] sustrik	+ static bool test ();
[15:29] mikko	sustrik: overloaded for different types i guess
[15:29] sustrik	i mean, do you use all 3 of them?
[15:29] mikko	sustrik: yes
[15:29] mikko	what i think about now
[15:29] mikko	does it make sense to amend the swap now
[15:29] mikko	if it goes to heavy refactoring soon
[15:29] sustrik	well, depends on you
[15:30] sustrik	if you need it asap, i'll apply it
[15:30] mikko	i don't need asap
[15:31] mikko	i can work around it and document for php extension
[15:31] mikko	at the moment doing const char *tmp = getenv("TMPDIR"); if (!tmp) { tmp = "/tmp"; } chdir(tmp);
[15:31] sustrik	ok then
[15:32] sustrik	maybe just write a comment about it on the ml
[15:32] sustrik	so that people know what happened to the patch
[15:34] mikko	i'll write comment and send device patch today
[15:34] sustrik	ok
[17:37] mikko	sustrik: call to device is not consistent between C and C++
[17:37] mikko	seems to be void in C++
[17:37] mikko	and int in C
[17:37] mikko	return type that is
[17:39] mikko	error_t is int?
[17:42] mikko	no, wait. got it
[18:15] myraft	complete noob - quick question - where do I start? I am not sure if I even installed it correctly. Is there a zmq on linux that should be in the path?
[18:17] myraft	Nevermind - found some Java examples on Github - will scan and ask the questions again.
[18:17] myraft	Thanks though.
[18:18] cremes	myraft: make sure you run ldconfig so that the zmq libs get added to your linux library paths
[18:19] drbobbeaty	myraft: it's probably best to read The Guide (it's that important): http://zguide.zeromq.org/chapter:all
[18:19] drbobbeaty	It contains a lot of good information as well as a lot of examples in all kinds of languages.
[18:21] davetoo	with topic exchange-like behavior :)
[18:23] cremes	davetoo: that's described in the guide too ;)
[18:27] myraft	thanks folks - just found the guide as well
[19:13] mikko	sustrik: https://gist.github.com/848e949008e67ca542f7
[19:14] sustrik	hm, that may be the problem that dhammika fixed
[19:14] sustrik	i'll apply his patch tomorrow
[19:15] mikko	cool
[19:15] mikko	ianbarber stumbled on this with the php extension
[19:15] sustrik	it's a heisenbug, right?
[19:15] mikko	nope, can reproduce reliably
[19:15] sustrik	ah, then it may be something different
[19:16] mikko	if you take the .c file attached to the paste
[19:16] mikko	compile and and
[19:16] mikko	compile and run
[19:16] sustrik	right
[19:16] mikko	it should segfault / show memory errors
[20:32] mikko	i to the b
[20:33] ianbarber	m to the k
[20:40] myraft	when trying to run java program I get "UnsatisfiedLinkError: no jzmq in java.library.path" - I have done ldconfig -p \|grep zmq and see they so are a loaded. I have tried recompiling the code. Any ideas ?
[20:51] mikko	myraft: it seems to be complaining about jzmq
[20:51] mikko	rather than zmq itself
[20:52] myraft	ok
[20:52] myraft	but not sure how to proceed - followed the instructions -
[20:53] myraft	nathanmarz - if I do ldconfig -p\|grep for zmq I see libzmq.so along with ligjzmq.so -
[20:53] myraft	I meant to direct that at mikko -
[20:54] mikko	myraft: is the directory where libjzmq.so is in java.library.path ?
[20:55] myraft	mikko : libjzmq.so is in /usr/local/lib - "sheepishly" - not sure if it is java.library.path, how do I set it ?
[20:56] mikko	myraft: i've no idea tbh, i don't use java myself
[20:56] mikko	myraft: what version of zeromq are you using?
[20:58] myraft	mikko: 2.0.10
[20:58] mikko	myraft: can you test with current github trunk?
[20:58] mikko	just in case
[20:58] mikko	git clone https://github.com/zeromq/zeromq2.git
[20:58] mikko	cd zeromq2 && ./autogen.sh && ./configure && make install
[21:00] myraft	mikko: will do and postback - thanks
[21:04] myraft	mikko - did that and see libzmq.so.1 created in /usr/local/bin - running the Java program still gives the same error.
[21:04] myraft	I did go thru installation instructions - and was confused about setting java.library.path - I did CLASSPATH instead in the .bashrc
[21:08] myraft	I wonder if I need to restart the system - since there is libconf daemon - and that will not see the changes.
[21:08] myraft	Anyways - thanks for the attempt - let me know if you have any more ideas.
[21:14] myraft	I will leaving the workstation for a little bit and will be online later.
[21:19] mikko	i dont think that makes a different
[21:19] mikko	difference*
[21:19] mikko	i can give it a go later
[21:35] mikko	sustrik: still there?