ZeroMq IRC Log

Monday May 31, 2010

[Time] Name	Message
[00:05] versificateur	hello :)
[00:06] versificateur	I have some question about compiling java bindings
[00:07] versificateur	...
[00:07] versificateur	if i could have some help i'll be happy
[01:09] versificateur	hello
[01:14] versificateur	please...
[03:37] sustrik	versificateur: hi
[04:13] CIA-17	zeromq2: 03Martin Sustrik 07master * r04fcd4d 10/ src/rep.cpp : memory leak in REP socket fixed - http://bit.ly/d6dQUE
[04:13] CIA-17	zeromq2: 03Martin Sustrik 07master * r3bb60da 10/ (7 files in 3 dirs): Merge branch 'master' of git@github.com:sustrik/zeromq2 - http://bit.ly/cgjNYb
[04:13] sustrik	zedas: done, try it now
[04:19] CIA-17	zeromq2: 03Martin Sustrik 07master * r2e9be56 10/ src/req.cpp : memory leak in REQ socket fixed - http://bit.ly/9ULDmt
[05:08] zedas	sustrik: cool i'll give it a shot
[05:08] sustrik	ok
[07:30] CIA-17	zeromq2: 03Martin Sustrik 07master * rbe6019a 10/ src/req.cpp : issue 28. - SNDMORE/ RCVMORE is dropping every other message - http://bit.ly/bQHWhC
[12:19] CIA-17	zeromq2: 03Martin Lucina 07master * r8becacf 10/ (doc/zmq_setsockopt.txt doc/zmq_getsockopt.txt):
[12:19] CIA-17	zeromq2: Documentation updates
[12:19] CIA-17	zeromq2: Add zmq_getsockopt(3), clean up zmq_setsockopt(3). - http://bit.ly/dyI3wE
[12:19] CIA-17	zeromq2: 03Martin Lucina 07master * reb9ff1e 10/ (doc/zmq_getsockopt.txt doc/zmq_recv.txt doc/zmq_send.txt):
[12:19] CIA-17	zeromq2: Documentation updates
[12:19] CIA-17	zeromq2: Multi-part messages - http://bit.ly/aVO3PG
[12:19] CIA-17	zeromq2: 03Martin Lucina 07master * r0fa73b0 10/ doc/zmq.txt :
[12:19] CIA-17	zeromq2: Documentation updates
[12:19] CIA-17	zeromq2: Add getsockopt to index - http://bit.ly/bwZY5J
[12:19] CIA-17	zeromq2: 03Martin Lucina 07master * rdfbaf4f 10/ (doc/zmq_recv.txt doc/zmq_send.txt):
[12:19] CIA-17	zeromq2: Documentation updates
[12:19] CIA-17	zeromq2: Multi-part messages - http://bit.ly/aj9jbE
[12:19] CIA-17	zeromq2: 03Martin Lucina 07master * r8a4df43 10/ doc/Makefile.am :
[12:19] CIA-17	zeromq2: Documentation updates
[12:19] CIA-17	zeromq2: Add getsockopt to Makefile.am - http://bit.ly/9i34WB
[13:42] cremes	sustrik: i'm available in irc if you want to try anything else with that zmq_poll issue
[13:54] sustrik	cremes: have you seen my comment?
[13:55] cremes	sustrik: yes; i ran the test and posted my results
[13:55] sustrik	a-ha
[13:56] sustrik	heve you sent it a message?
[13:56] sustrik	i've just posted the receving part
[14:00] cremes	no, i thought it was supposed to be self-contained
[14:00] cremes	i'll do that now
[14:00] sustrik	thx
[14:05] cremes	sustrik: fails on OSX, succeeds on linux
[14:05] sustrik	fails = reports pollout?
[14:06] cremes	the assertion fails; i'll change the program to print out revents
[14:07] sustrik	btw, zedas already reported problems when runing on netbsd... this may be the same thing as both use kevent rather than epoll
[14:08] cremes	sustrik: zmq_poll returns, but rc == 0 and revents is 0
[14:09] sustrik	that's because you've left the timeout value in place, no?
[14:09] sustrik	try -1
[14:09] cremes	i had it set for 10 seconds
[14:10] cremes	however, just saw something interesting...
[14:10] cremes	it succeeds when passed -1
[14:10] cremes	but fails with 10000000
[14:10] sustrik	see man zmq_poll(3)
[14:11] sustrik	zmq_poll() shall wait up to timeout microseconds
[14:11] sustrik	note up to part
[14:11] cremes	so it may unblock ahead of time?
[14:11] sustrik	yes, this is an optimisation
[14:12] sustrik	i can make it to unblock on time
[14:12] sustrik	but it will slow zmq_poll down
[14:12] cremes	btw, it fails in the same way on linux when given a timeout value > 0
[14:12] cremes	i set it for 20 seconds and then sent a message within 1 or 2 and it failed
[14:12] sustrik	same way = reports pollout?
[14:14] cremes	reports revents == 0
[14:14] sustrik	ah, that's the timeout thing
[14:14] cremes	on both osx and linux
[14:15] sustrik	yes, it's supposed to work that way
[14:15] cremes	but i am clearly sending it a message
[14:15] sustrik	ok. some internal details:
[14:15] sustrik	zmq_poll unblock on any internal event that affects a 0mq socket in the pollset
[14:15] sustrik	in this case what happens is
[14:16] sustrik	that "connection is established" event occurs
[14:16] sustrik	thus zmq_poll exits
[14:16] sustrik	with no messages to delivers
[14:16] sustrik	as i said this kind of behaviour can be prevented
[14:16] sustrik	but it'll slow zmq_poll down
[14:17] sustrik	back to your original problem!
[14:17] cremes	i changed the sample program to call zmq_poll a second time; it still returns revents == 0
[14:17] cremes	there should have been one in queue
[14:18] sustrik	change it to loop forever
[14:18] sustrik	afaics there are two events raised when connection is established
[14:19] sustrik	i/o thread says the app thead: "here's the pipe you can write messages to"
[14:19] sustrik	and afterwards: "here's the pipe you can read messages from"
[14:20] cremes	ok
[14:20] cremes	let me try this on osx and see if i get my original problem
[14:20] sustrik	ok
[14:27] cremes	the sample program returns the correct value; i'll try it with my test program to see if it behaves correctly now
[14:30] sustrik	ack
[14:35] cremes	it still gets a POLLOUT first; i'm going to operate under the assumption this is my bug
[14:35] cremes	i'll keep looking into it
[14:37] sustrik	i've just checked the code
[14:37] sustrik	cannot find a way how POLLOUT may have been signaled
[14:37] sustrik	unfortunately, i cannot test on OSX as I don't have it installed
[14:38] cremes	i'll keep looking from my end; perhaps i'm doing something stupid (likely) :)
[15:24] cremes	sustrik: you can close the issue; it was a bug in my bindings
[15:24] sustrik	ah, ok
[15:24] cremes	i had a logic error in one of my examples; it was doing a "ping pong" using two REQ sockets
[15:25] cremes	apparently it's possible to tie 2 of those together
[15:25] sustrik	cremes: ok, the issue is closed
[15:25] cremes	but the "receiver" REQ will get POLLOUT on the first received message and POLLIN on all subsequent ones
[15:26] cremes	sorry for wasting your time
[15:32] sustrik	no problem
[16:33] alefka	hi
[16:34] alefka	anybody?
[16:35] alefka	(