Thursday November 25, 2010

[Time] NameMessage
[00:06] quentusrex Is there a particular socket type that would be best for Request-reply type messages, but from one server to each of the many clients?
[00:12] lestrrat is there a working code example for ZMQ_FD ?
[00:40] mgc hi all. apologies for a drive by question: load testing the latest from zeromq2/master I am hitting "Assertion failed: inpipes [current_in].active (xrep.cpp:229)". Any suggestions?
[00:40] mgc looks like the pipe suddenly goes away while getting a multi part message?
[07:47] sustrik lestrrat: look at implementation of zmq_poll in src/zmq.cpp
[07:49] sustrik mgc: looks like an bug. can you report it and provide a simple test program to reproduce it?
[07:49] sustrik quentusrex: hm, doesn't make much sense to me
[07:51] guido_g sustrik: nice support for the ømq :)
[07:51] guido_g ok, it's from 2007, but we can ignore that
[07:53] sustrik yep, i've seen that presentation before
[07:53] sustrik or maybe a very similar one
[07:54] guido_g ah ok
[08:04] sustrik guido_g: if you believe it's interesting for general audience, just send a link to the mailing list
[08:56] quentusrex sustrik, basically I have a single server that will accept event notifications from many clients,
[08:57] quentusrex and the server needs to respond to the client to let them know if the work was accepted.
[08:58] quentusrex I don't see anything that mentions how to determine from which client connection each message came from.
[08:59] sustrik how is that different from standard REQ/REP?
[08:59] quentusrex In my case the delay is ~30 seconds until we know if the results are accepted,
[09:00] sustrik if you need something like, delayed replying, use XREQ/XREP
[09:00] quentusrex so I was looking in the documentation and examples for something that would allow me to send a response back to the client from which the message came from.
[09:01] quentusrex I'll dig into the XREP more.
[09:01] sustrik yes, that'll help you
[09:01] sustrik check the guide for examples
[09:03] quentusrex aah, that is exactly what I was looking for.
[09:04] quentusrex thanks sustrik
[09:04] sustrik np
[10:34] CIA-20 jzmq: 03Gonzalo Diethelm 07master * ra97e15a 10/ (src/Socket.cpp src/org/zeromq/ (log message trimmed)
[10:34] CIA-20 jzmq: Changes to better handle poller timeouts and the C glue code.
[10:34] CIA-20 jzmq: The flags parameter to send / receive were assigned to a long
[10:34] CIA-20 jzmq: variable, which was received in the C glue code as a jlong,
[10:34] CIA-20 jzmq: which was finally cast to an int when being passed into
[10:34] CIA-20 jzmq: zmq_send() and zmq_recv(). I changed them into int / jint and
[10:34] CIA-20 jzmq: got rid of the cast.
[13:06] CIA-20 zeromq2: 03Sebastian Otaegui 07master * r92618fd 10/ builds/redhat/zeromq.spec :
[13:06] CIA-20 zeromq2: Small fix for the rhel6 spec
[13:06] CIA-20 zeromq2: Signed-off-by: Sebastian Otaegui <> -
[13:06] CIA-20 zeromq2: 03Martin Sustrik 07master * r734624b 10/ src/i_engine.hpp :
[13:06] CIA-20 zeromq2: Typo fixed in a comment.
[13:06] CIA-20 zeromq2: Signed-off-by: Martin Sustrik <> -
[14:22] mato mikko: hi, are you around?
[16:13] mato sustrik: could you add the individual test executables in tests/ to .gitignore at the top level please?
[16:13] mato sustrik: (this is not worth making a patch to the ML...)
[16:13] CIA-20 zeromq2: 03Martin Lucina 07master * rc958409 10/ (5 files in 3 dirs):
[16:13] CIA-20 zeromq2: Control symbol exports using -fvisibility
[16:13] CIA-20 zeromq2: On systems using GCC 4.0 or newer which support symbol visibility in shared
[16:13] CIA-20 zeromq2: libraries, use -fvisibility=hidden and only export explict API functions
[16:13] CIA-20 zeromq2: defined in zmq.cpp. We do not enable -fvisibility on MinGW since this uses a
[16:13] CIA-20 zeromq2: separate mechanism (__declspec).
[16:13] CIA-20 zeromq2: Signed-off-by: Martin Lucina <> -
[16:36] mato sustrik: ping?
[16:55] mikko mato: yes]
[16:55] mikko am now
[16:58] mato mikko: I wanted some advice on how to check for -fvisibility support, but have that solved now
[16:58] mato mikko: (see my patch on the ml. sustrik has applied it
[16:59] mato mikko: how often do the autobuilders wake up?
[16:59] mikko mato: 5 am and 5 pm GMT
[16:59] mikko in a minute is the next build
[17:00] mato oh, good, i can see hopefully that i've not broken anything
[17:00] mikko i got a couple of patches waiting as well
[17:00] mikko need to have some time to send them
[17:00] mikko --enable-debug switch and a couple others
[17:01] mikko i had a question for you btw
[17:01] mikko for hp ux in CPPFLAGS="-D_POSIX_C_SOURCE=200112L"
[17:01] mikko is there a reason why $CPPFLAGS is not appended to that?
[17:01] mato mikko: send them soonish if you want them in the 2.1.0 release, I think we'll get that out Monday
[17:02] mato mikko: no reason, it's a typo
[17:02] mikko ok
[17:02] mikko my patch fixes that
[17:02] mikko after these patches user can override all build flags
[17:02] mato good
[17:03] mikko daily build running now
[17:15] mikko sent
[17:15] mikko bbl ->
[18:17] mikko mato: currently i get this from autoreconf
[18:17] mikko warning: AC_INIT: not a literal: m4_esyscmd(./ | tr -d '\n')
[18:17] mikko i assume this is not an issue?
[18:17] mato mikko: dunno, I've not seen it
[18:17] mato mikko: I got that incantation off the autoconf mailing list so it should be ok...
[18:18] sustrik mato: hi
[18:18] sustrik ok
[18:21] mato sustrik: hi, um, i was asking if you could update .gitignore to include tests/(the executables)
[18:21] mato sustrik: so that "git status" is clean after a build
[18:22] mato sustrik: didn't seem worth mailing a patch for administrivia
[18:22] sustrik mato: yes
[18:22] sustrik i'll do that
[18:22] mato thx
[18:22] sustrik the annoying part it has to be done on file-by-file basis
[18:22] mikko i'll amend the patches
[18:22] sustrik as executables have no common extension
[18:23] mikko so apart from DISABLE_SHARED they are OK?
[18:23] mato sustrik: yeah, well, when you add a test, update .gitignore, no problem...
[18:23] mato mikko: yeah, i'm happy, thanks
[18:26] mato mikko: one thing with --enable-debug...
[18:26] mato mikko: not sure if it's useful, but.
[18:26] mato mikko: at the moment default behaviour is to build with -O2 -g (say for GCC)
[18:27] mato mikko: --enable-debug changes that to -O0 -g
[18:27] mato mikko: I'm wondering whether or not there should also be a --disable-debug/--enable-release or something that disables -g entirely and/or in future adds more agressive optimizations
[18:27] mato mikko: but that's just a thought, no need to deal with that now
[18:30] mikko yes, i think that would make sense
[18:30] mikko do we not want -g on releases?
[18:30] mikko makes backtraces quite useless
[18:31] mikko i see your point
[18:31] mikko use can then build with --enable-debug and provide a backtrace if it comes to that
[18:32] mato well, i don't think the default should be w/o -g
[18:32] mikko i managed to amend the patch without much hacking
[18:32] mikko i think my git foo is getting stronger
[18:32] mato since that will precisely cause newbiews to report problems with no backtrace at all
[18:32] mikko i did git reset HEAD~4, git am for the two first patches, git apply for the third, edited the lines and committed
[18:32] mikko is that the best way to amend a patch?
[18:32] mato all i'm saying is it might be useful in future to provide an "optimized, no debug" build
[18:33] mikko yes, i think we can easily provide that
[18:33] mikko and it does make sense. if for example someone figures out a compiler flag for sun studio that speeds up things a lot we can easily share that to everyone
[18:34] mato yeah, for example stuff like -xtarget=native on SPARC
[18:34] mato and/or -fast
[18:35] mikko hmm
[18:35] mikko i might've messed my local build
[18:35] mikko ./configure: line 22683: syntax error near unexpected token `if'
[18:35] mikko ./configure: line 22683: `_LT_COMPILER_OPTION(if $compiler supports -fvisibility=hidden,'
[18:37] mato ok, i have to go
[18:37] mato bbl maybe
[21:29] mikko mato: did you come back?