Monday November 22, 2010

[Time] NameMessage
[02:58] Steve-o that was Linux to FreeBSD though, I'm just testing Linux to Linux to see if the graph changes
[03:02] Steve-o the goal is to find out if anything special is causing Windows performance to be ass
[03:06] doug what are you measuring?
[03:07] doug and what makes you think it might be in your critical path?
[03:08] Steve-o I send a packet from one host to another, which then sends it back, I'm measuring the latency from the first second to the last receive
[03:08] Steve-o I'm trying to rule out all controllable issues that can affect the critical path
[03:10] Steve-o performance testing is always overly subject to errors in the testing method and framework
[03:11] Steve-o you always get developers complaining they cannot achieve reported performance figures, etc
[08:13] sustrik mikko: yes
[08:13] sustrik HP-UX is broken in this respect
[08:14] sustrik IIRC
[08:14] sustrik mato may know more
[09:10] demmonoid Hi. I'm trying to use Erlang 0mq plugin on Mac OS X, but the build fails. It seems that the C driver needs to be built as a bundle rather than a dynamic library.
[09:10] demmonoid The problem is that all 0mq libraries are built as dynamic libraries as well
[09:10] demmonoid Is there a way to work this out?
[09:16] sustrik demmonoid_: there are some OSX related issues in the bug tracker
[09:16] sustrik have a look there
[09:20] demmonoid sustrik, yes, I saw them but they are like 3 months old and with no response...
[09:21] demmonoid To be more exact, one of them is still with no response, just that I described here...
[09:22] sustrik well, you can try to fix the problem yourself then
[09:26] demmonoid That's what I've tried to do but it looks like 0mq doesn't work when it's libraries are just build with -bundle instead of -dynamiclib :)
[09:26] demmonoid *its libraries
[09:32] mikko sustrik: but that deletes the -g -O2 flags as well
[09:33] sustrik mikko: probably a bug
[09:34] sustrik i have no idea about how the build system works though
[09:35] sustrik demmonoid_: what exactly is the problem, does if fail? what's the error?
[09:35] mikko sustrik: i got a couple patches coming soonish
[09:36] sustrik mikko: ok
[09:36] mikko sustrik: is there a reason why i cant send pull requests in github?
[09:36] mikko would make the development flow a bit more fluent
[09:36] sustrik matter of process
[09:36] sustrik we should do it in a well defined way
[09:36] sustrik pull requests are problematic for people on exotic platforms
[09:37] mikko why is that?
[09:37] sustrik NO GIT
[09:37] mikko if they come through github
[09:37] sustrik no git
[09:37] sustrik you need git client though
[09:37] mikko you need a git client to format-patch as well
[09:37] sustrik all you need is diff
[09:38] sustrik patches don't have to be strictly formatted
[09:38] sustrik just and diff and a sign-off
[09:43] demmonoid sustrik, it gives the error "i686-apple-darwin10-g++-4.2.1: -compatibility_version only allowed with -dynamiclib" upon build
[09:46] demmonoid I removed this flag from libtool, as well as install_name which is also incompatible with -bundle
[09:46] demmonoid Will see if this works
[09:46] mikko demmonoid_: osx build for erlzmq is broken?
[09:47] demmonoid Yes, it is
[09:47] demmonoid Looks like it's broken for OS X for a long time...
[09:48] demmonoid At least for Snow Leopard
[10:02] demmonoid So, I rebuilt 0mq with -bundle, edited erlzmq's configs so it will compile with -bundle instead of -dynamiclib, but I still get the same error: "i686-apple-darwin10-g++-4.2.1: -bundle not allowed with -dynamiclib"
[10:02] demmonoid It appears that zeromq libs were built with -dynamiclib anyway
[10:03] mikko sustrik: i added --enable-debug flag to builds
[10:03] mikko sustrik: do you think something like this is useful?
[10:06] mikko as currently it's not possible to really set debug flags
[10:07] mikko -O2 will be in the build line in any case
[10:07] sustrik what I do is:
[10:07] sustrik export CXXFLAGS="-g -O0"
[10:07] sustrik ./configure --disable-shared
[10:07] mikko you need to do CPPFLAGS and CFLAGS as well, if building with pgm
[10:08] sustrik right
[10:08] mikko for normal user maybe ./configure --enable-debug might be easier
[10:08] sustrik right, you want to merge all three into s single flag
[10:08] sustrik 1. -g
[10:08] sustrik 2. -O0
[10:08] sustrik 3. --disable-shared
[10:09] mikko i will look into adding --disable-shared as well
[10:09] sustrik it helps
[10:09] mikko the problem is that i am not 100% convinced if this is really needed
[10:09] mikko as for example with sun studio the debug flag is -g0
[10:09] sustrik as if it's not set, the tests are built with a different name
[10:10] sustrik and then there's a wrapper script to launch them
[10:10] sustrik thus launching it within bebugger
[10:10] sustrik actually debugs the shell
[10:10] sustrik not the test program
[10:10] sustrik mikko: i would appreciate it
[10:10] sustrik it saves some time when debuging
[10:12] mikko sustrik: ok, will add
[15:06] CIA-20 zeromq2: 03Martin Sustrik 07maint * r83d9af9 10/ MAINTAINERS :
[15:06] CIA-20 zeromq2: Mikko Koppanen becomes a maintainer of the build system.
[15:06] CIA-20 zeromq2: Signed-off-by: Martin Sustrik <> -
[15:07] CIA-20 zeromq2: 03Martin Sustrik 07master * r83d9af9 10/ MAINTAINERS :
[15:07] CIA-20 zeromq2: Mikko Koppanen becomes a maintainer of the build system.
[15:07] CIA-20 zeromq2: Signed-off-by: Martin Sustrik <> -
[15:07] CIA-20 zeromq2: 03Martin Sustrik 07master * ra689047 10/ MAINTAINERS :
[15:07] CIA-20 zeromq2: Merge branch 'maint'
[15:07] CIA-20 zeromq2: * maint:
[15:07] CIA-20 zeromq2: Mikko Koppanen becomes a maintainer of the build system. -
[19:08] cremes anybody seeing this assertion failure against latest master?
[19:08] cremes Assertion failed: !more || pipes [current] != pipe_ (fq.cpp:61)
[19:09] cremes i'm trying to get a small reproducible case but have so far been unsuccessful
[19:09] cremes would a backtrace be useful in the absence of reproducible code?
[19:27] sustrik cremes: yes, back trace is always helpful
[19:45] cremes