Friday February 25, 2011

[Time] NameMessage
[02:10] amacleod Are ipc:// URIs expected not to work under Windows? We are getting a "Protocol not supported" error when trying to bind to "ipc://listener".
[02:11] amacleod Ah.. there it says right in the manpage. It's currently only implemented on OSs that provide UNIX domain sockets.
[03:01] brazilnut000 hello, i was wondering if I could get some assistance trying to figure out some difficulty I'm havnig with multithreaded code...
[03:03] brazilnut000 basically, I'm spawning a thread to subscribe to incoming data and update some variables while my main thread does some other stuff (including talk on other ports)
[03:03] brazilnut000 they're both using the same context, but my thread doesn't seem to get anything from the publisher
[03:04] brazilnut000 however, if i make the same calls to create, bind, and subscribe in my main thread it works...
[03:05] brazilnut000 anyone?
[03:24] brazilnut000 hrmm...i guess not...
[03:25] Steve-o morning all, looks like o'reilly doesn't actually archive webcasts?
[03:32] amacleod brazilnut000, you are not sharing sockets between threads, are you? Just contexts?
[03:32] amacleod brazilnut000, what transport are you using, and what ZMQ pattern?
[03:39] brazilnut000 oh, sorry, stepped away.
[03:40] brazilnut000 I'm sharing contexts only, tcp, ZMQ_PUB and ZMQ_SUB
[03:44] amacleod Ok. So you shouldn't need to be messing with any addressing/routing.
[03:44] amacleod How many I/O threads? 1?
[03:44] brazilnut000 1
[03:47] amacleod Dunno. Is it possible that the subscriber is trying to connect before the publisher has had a chance to become established?
[03:48] amacleod (I'm not sure how timing issues are supposed to work with PUB/SUB.)
[03:50] brazilnut000 it's possible, though unlikely....also, as I understand it, order shouldn't matter. It's possible to lose the first 'x' messages, but I don't ever get any...
[03:56] brazilnut000 well, thanks for giving it a shot. Gonna go get some dinner...
[03:57] amacleod You're welcome. Sorry I couldn't get any further.
[07:44] pieterh sustrik: g'morning
[07:44] sustrik morning
[07:44] pieterh just curious if you found that leak yesterday
[07:45] sustrik well, i found one delayed deallocation
[07:45] sustrik but afaics it applies only to sockets with no connects/binds
[07:45] sustrik so it's probably a different issue
[07:46] pieterh a'ight
[07:46] sustrik anyway, i'm patching it now
[07:50] sustrik pieterh: the master documentation doesn't seem to be updated automatically
[07:50] sustrik should it?
[08:00] CIA-21 zeromq2: 03Martin Sustrik 07master * rd4e418f 10/ (src/reaper.cpp src/socket_base.cpp src/socket_base.hpp):
[08:00] CIA-21 zeromq2: Socket with no owner objects is deallocated immediately
[08:00] CIA-21 zeromq2: Till now the deallocation of such socket was delayed
[08:00] CIA-21 zeromq2: till zmq_term() thus creating a "leak".
[08:00] CIA-21 zeromq2: Signed-off-by: Martin Sustrik <> -
[08:00] sustrik pieterh: the patch is in master
[08:01] sustrik can you check whether the leak persists?
[08:01] sustrik just in case...
[08:18] pieterh sustrik, re
[08:19] pieterh master docs are not yet updating automatically, but I'll check with mikko, everything is ready for that
[08:19] pieterh i'll check the leak later, need to use a different box (multicore)
[08:21] sustrik ok
[08:21] pieterh though in the test case, sockets were definitely being used (recv) and closed systematically
[08:52] pieterh sustrik: good news
[08:52] pieterh as far as I can tell from initial tests, you've fixed the leak
[08:52] sustrik interesting
[08:53] pieterh I'm doing more tests now
[08:54] pieterh I retract that
[08:54] pieterh it works over inproc but not over tcp
[08:55] pieterh and I think inproc was getting weird test results if run after tcp
[08:55] pieterh cremes' original test case still leaks memory (and consumes a whopping amount of CPU)
[09:05] sustrik ok
[09:23] pieterh do we have an open issue for this?
[09:23] pieterh i've stripped down chuck's test case and would like to post that somewhere
[09:27] sustrik no idea
[09:27] sustrik there's none unless chuck opened one
[09:28] pieterh ok
[09:28] pieterh I'll ask him to do so
[09:30] pieterh hmm, back with 'too many open files'... strange
[09:31] sustrik well, the test case is creating undetermined number of sockets
[09:32] pieterh yes, it's a fault in my test case
[09:34] pieterh ok, I need to leave for a while bbl
[12:03] mikko ian did well
[12:14] cremes pieterh, sustrik: i opened a ticket on this leak a few days back:
[12:14] cremes example code is in a gist attached to the ticket
[12:14] cremes brb...
[12:25] sustrik mikko: you've been on the coference?
[12:26] sustrik there's a lot of tweets about the lecture
[12:28] sustrik hi ian
[12:28] sustrik how did it go?
[12:30] mikko sustrik: here at the moment
[12:30] sustrik no haste
[12:30] ianbarber hey! pretty good i think
[12:31] sustrik people seem to be happy according to twitter
[12:31] sustrik any feedback, questions?
[12:32] mikko
[12:32] mikko there is some feedback already
[12:34] sustrik nice :)
[12:36] ianbarber good comments, had some questions about persistance, telehash, whether it could run on an arduino, some zero vs activemq stuff, and a bit on whether you can do pgm on ec2
[12:38] sustrik good question
[12:39] sustrik especially the last one :)
[12:39] sustrik i have to admit i have no idea
[12:41] ianbarber me either :) the next talk was about ec2 though, so that's handy
[15:56] mikko pieterh: here?
[15:56] pieterh hi mikko, yes
[15:56] mikko should i run apiall or apione in the docs builds?
[15:57] pieterh apiall, that's the main script
[15:58] pieterh sorry...
[15:58] mikko hmm
[15:58] mikko ZMQ_DIR seems to have changed as well
[15:58] pieterh I've made apiall do the whole job
[15:58] pieterh there is no ZMQ_DIR any more
[15:59] pieterh apione is to update a single version from some repository
[15:59] pieterh does this make sense?
[16:00] mikko yes, ill check the code to see what exactly is happening
[16:01] mikko so it expects to find ../../zeromq2 ?
[16:01] pieterh yes
[16:01] pieterh and ../../zeromq2-1
[16:01] mikko and ../../zeromq2-1
[16:01] mikko ok
[16:01] pieterh and eventually ../../zeromq2-0 when we make that
[16:01] pieterh etc.
[16:02] pieterh does that match how you use the repos on the build system?
[16:02] mikko no, this far i've used branches
[16:02] mikko i think i need to update all maint builds at some point
[16:02] pieterh right, but the relative locations of ztools and zeromq2
[16:03] mikko ah, thats different as well
[16:03] mikko but i can make it to use the new structure
[16:03] pieterh let me know if you need any changes to the apisite stuff, happy to do whatever's needed
[16:03] mikko i wonder if git submodules would make sense here
[16:04] mikko they are a bit like svn externals
[16:05] pieterh oh, that gets really complex real fast, mikko...
[16:05] ianbarber i've heard they are often a bit of a pain
[16:05] mikko i meant using submodules in this way:
[16:06] mikko ztools/apisite/source/master -> zeromq/zeromq2/master
[16:06] mikko when you checkout ztools you can recursively take submodules if you want
[16:06] mikko and if not then you can put the code into right place
[16:07] mikko not sure how jenkins handles that
[16:07] mikko because currently jenkins is configured to use zeromq/zeromq2 as source repo for docs build
[16:07] mikko and it polls that repo for changes and builds if changes are detected
[16:08] mikko i'll look into this as soon i can get onto a bit faster internet
[16:08] mikko conference wireless is a bit painful
[16:14] pieterh hmm
[16:18] pieterh ok, I've made a PHP version of the Guide,
[16:23] sustrik pieterh: some of the examples are in python and c++
[16:25] sustrik and c
[16:30] pieterh only the very first ones, deliberately
[16:30] pieterh and some fragments here and there which are never translated
[16:31] pieterh this is still experimental, I'm sure we can improve it, but I don't want to have to modify the text itself to make a language-specific version
[16:31] pieterh my idea is that any language which hits 100% gets its own version
[16:32] sustrik ok
[16:59] lt_schmidt_jr :Steve-o, there was nothing on it about ZMQ
[21:57] DOBRO hi
[21:58] apexi200sx hi
[21:59] sustrik hi
[22:07] apexi200sx anyone got any use cases for zeromq
[22:08] cordor is there precompiled version somewhere?
[22:11] sustrik condor: there are snapshots
[22:12] sustrik apexi200sx: there were few use cases mentioned on the mailing list
[22:13] cordor zeromq-2.0-alpha1.exe?
[22:14] siggimoo apexi200sx: we're using ZeroMQ for network monitoring. Individual processes monitor something and push their readings to a central analyzer, which in turn publishes its findings to zero-or-more subscribers, one of which is a pager/alarm system.
[22:15] sustrik condor:
[22:19] cordor thanks
[22:33] erickt Hello #zeromq. Could 2.1.1rc1 be tagged in git? It would be handy for those of us tracking git to know what is considered stable-ish
[22:35] erickt oh, look at that, I guess 2.1 has been forked into it's own repository