IRC Logs for #crux Monday, 2014-07-21

*** tilman has quit IRC00:04
*** tilman has joined #crux00:05
*** jdolan has joined #crux01:20
openfbtdpidsley, I wasn't ignoring you, I was working.01:43
*** jdolan has quit IRC02:00
*** q5sys has quit IRC02:05
*** mechaniputer has joined #crux02:09
*** jdolan has joined #crux02:18
*** SiFuh_ has quit IRC02:32
Romsterpidsley, distcc is pretty cool02:36
pidsleyRomster: i was impressed02:38
pidsleyhow do i use it in pkgmk.conf?02:39
pidsleyi think i have it working there, but i could use an expert opinion02:40
RomsterCC="distcc gcc" CXX="distcc g++"02:41
Romsterhow is your DISTCC_HOSTS= line02:42
Romsterexport MAKEFLAGS="-j$(/usr/bin/distcc -j 2> /dev/null)"02:42
jaegerI recommend using PATH=/usr/lib/distcc:$PATH instead of setting CC and CXX02:42
Romsterexport DISTCC_DIR="/var/lock/distcc"02:43
*** mavrick61 has quit IRC02:43
jaegerthat way the CC and CXX overrides don't usually cause issues02:43
Romsteri used todo that but i don't anymore jaeger02:43
Romsternot many ports ignore it now.02:43
Romsterlike 5 or so i've hit02:44
Romsterall of mine will agree to use CC and CXX02:44
*** mavrick61 has joined #crux02:44
jaegerit caused a lot of trouble during multilib dev, not as much now02:44
Romsteri use it live on multilib no issues02:44
Romsterthe only issues i ever had was if a box had a different tool chain was switched on and in the distcc pool02:45
pidsleywhat i have now is export DISTCC_HOSTS="localhost hyrax"02:45
pidsleyexport MAKEFLAGS='-j8 CC="distcc"'02:45
Romstereither do it jaegers way or set CC and CXX02:46
Romstereach var on there own line02:46
pidsleybut that doesn't work for all packages02:46
Romsterexport CC="distcc gcc"02:46
Romsterexport CXX="distcc g++"02:46
Romstermajority do.02:46
jaegereither way works most of the time but if you run into compile errors disable distcc first02:47
Romsterand the fe it don't wlel they don't take much time to compile anyways.02:47
Romsterhow many cores does localhost and h\yrax have?02:47
Romsterman cant type... *02:47
pidsleyah yes -- it was the g++ packages that didn't like the way i had it set02:48
pidsleyeach has two cores02:48
RomsterDISTCC_HOSTS="localhost/2 hyrax/2"02:48
Romsterif each have 2 cores02:48
Romster-j8 will slow you down.02:48
pidsleythank you02:48
Romsterdistcc -j would probably suggest 5 or 6 in your case.02:48
Romsterto account for tcp socket and queues02:48
Romsterdouble the make jobs is never good.02:49
Romsterlike i got 18 cores and i use -j24 but this is over 6 machines.02:49
pidsleyi do know that using it plain halved my kernel build time02:49
pidsleyi just didn't know how to set up pkgmk.conf02:50
Romsterthis is what i use02:50
Romsterexport DISTCC_HOSTS="--randomize arkanoid/4 space/4 fire2/4 chunkserver1/2 master/2 metalogger/2"02:50
Romsterexport MAKEFLAGS="-j$(/usr/bin/distcc -j 2> /dev/null)"02:50
Romsterand distcc sets the number of jobs for me.02:51
Romsteri had experimented with the number of jobs but no real gain from those that distcc suggested.02:51
Romster--randomize will scatter jobs between all hosts.02:51
Romsterelse it'll use the first item in DISTCC_HOSTS as highest priority going along as far as it'll need too.02:52
pidsleysweet. thank you very much. i was even surprised to see that adding the atom or the sempron made a difference02:52
Romsterif you don't add --randomize make sure you got the faster machines listed first.02:52
Romsteri went a step more i got ccache going though distcc.02:53
Romstermakes rebuilds faster02:53
pidsleybut adding more than the two fastest machines made very little difference02:54
Romsteryeah add --randomise and add in the number of cores on each02:54
Romsterand use what distcc -j suggests, tweak from thee if you so wish02:55
Romsterit is number of cores +1 for each box02:55
Romsterso -j6 for 2 machines of 2 cores each02:56
pidsleynone of the "pump" crap from the distcc site?02:56
Romsteri've tried it it don't do a hell of alot and it's far more problematic02:57
Romsteri was using it though02:57
cruxbot[opt.git/3.1]: alsa-lib: updated to version 1.0.2802:58
cruxbot[opt.git/3.1]: alsa-oss: updated to version 1.0.2802:58
cruxbot[opt.git/3.1]: alsa-utils: updated to version 1.0.2802:58
Romsterit required me todo this02:58
Romsterexport DISTCC_HOSTS="--randomize arkanoid/4 space,lzo,cpp/4 fire2,lzo,cpp/4 chunkserver1,lzo,cpp/2 master,lzo,cpp/2 metalogger,lzo,cpp/2"02:58
Romsterand then a wrapper for make02:58
pidsleywhat does the --ramdomize do?02:58
Romstermake wrapper02:59
Romsterexport MAKEFLAGS="$MAKEFLAGS"02:59
Romsterbut like i said it's problematic02:59
Romsterit's listed in the man page02:59
Romsterit randomizes the make jobs to each host02:59
pidsleysorry -- i will read the man page02:59
Romsterthan to go from  DISTCC_HOSTS faster medium speed slower.... may not get used if nto enough build jobs or parralizing in makefile.03:00
RomsterDISTCC_HOSTS works in order of first to last.03:00
pidsleyoh this makes sense in what i saw earlier then03:01
Romster--ramdomize shares the load out equally. which makes it more important to set number of cores per a host in DISTCC_HOSTS="--ramdomize foo/2 bar/2"03:01
Romsterit might be a bit cryptic to read at first i've spent some time messing with it.03:02
Romsterwith randomise even the slower hosts get a chance to pitch in.03:02
openfbtdDistributing the load equally might not be a good thing though.03:03
pidsleythank you for sharing this -- i saw you mention it earlier, and with as many machines as i have it is silly not to try it03:03
Romsterthe only down side is if a host is doing a job that make is waiting on, it'll make the rest wait until that job is done or times out and is distributed to another host03:03
Romsterdepends on your setup of machines.03:03
Romsterjust play around for what works for you.03:03
openfbtdYeah, and with bigger builds distcc might give an atom a job that it would struggle with for hours :D03:04
Romsterthere is no right way really. but faster machines listed first always set the number of cores on each machine.03:04
openfbtdThen again, what's the reason to even include a really slow machine in the cluster03:04
Romsteropenfbtd, possibly but the timeout would then kick in and it'll goto another host or be compiled locally.03:04
Romsterafk work03:05
openfbtderdic, still losing time03:05
openfbtdRomster, *03:05
pidsleyi tried it several ways, and the best seemed to be with just the athlon x2 and core2 duo03:05
pidsleyadding the others made only a minimal differnce03:06
*** jdolan has quit IRC03:06
*** jdolan has joined #crux03:07
*** Workster has joined #crux03:14
Worksterpidsley, yeah if the other machines are way slower it'll have a negative impact should a job land on that and the rest end up waiting for that job. you'll know if oyu get heaps of not used cpus on the other machines while 1 is running 100%03:17
Worksterbeen using htop, but i should look into a distrobuted daemon to conkey03:18
pidsleyWorkster: i think i will stick to sharing just the core2 and the athlon for now -- that seems to make the biggest difference03:19
Worksteryour call but it's good to be aware of these things.03:21
pidsleythen i can share built packages betwwen the two -- correct?03:22
Worksterif they are the same -march=03:22
Worksterin okgmk.conf then yes03:22
pidsleythey are03:22
Worksteri would typically set -march for each cpu on each machine03:23
Worksteroh that reminds me do nto use -march=native with distcc03:23
Worksteralways set a arch but not native.03:23
pidsleyboth are x86-6403:24
Worksteri'd relaly be using -march=atom on the atom and core2 i think it was for the core2 machine.03:25
Worksterbut you only gain some speed ups but it wont be portable between machines.03:25
Worksteralso bear in mind if yougot a soft dependency on the compiling machine and not on the destination machine. that'll break a library link. revdep will let you know03:26
pidsleyi think for thw newer atoms they use =bothell03:27
Worksterman gcc03:27
Worksteryu'll see a bunch of options in there03:27
pidsleycould i use a shared ports dir?03:28
Worksteryou could but not alter -march fom the x86_64 though03:29
pidsleyor just install packages built on one machine on the other? better?03:30
Worksteryou'd need to use pkg-get or what was it to install built ports in crux if oyu wanted to deal with dependencies that way.03:30
Worksterpersonally i just compile on each box. for its own march03:31
Workstersince i have several different types03:31
pidsleythat's what i do now03:31
Worksterbut i could share the core2 duros between the 3 machines but a bit lazy to bother there03:31
Worksterway more effort than prt-get sysup on each box03:32
pidsleyi thought so03:32
Workstersince i have 3 differnt -march builds03:32
Worksterbut if you wanna go the extra mile03:33
pidsleyi will have to look to see if the athlon really needs a different march'03:33
Workstergcc '-###' -e -v -march=native /usr/include/stdlib.h03:33
Worksternot sure if that works still but should list what the gcc compile would of picked for native cpu options03:34
pidsleyit is my main machine, and i would just use the core2 for distcc03:34
pidsleythe atom will soon be nfs server, probably running debian stable03:35
pidsleydpn't hate me03:35
Worksterwont be ably to distcc with debian and crux unless you used a chroot or moved to icecream03:36
pidsleyno, and i don't care -- the atom really doesn't add much03:37
Worksterws just an exercise eh03:37
pidsleyi really just wanted to see what timcowchip was up against03:39
pidsleyand he has a better atom than i do03:39
pidsley35 minutes to build a kernel on mine03:39
pidsleyi can only imagine what a glibc upgrade looks like03:40
pidsleytimcowchip has bigger problems than his machine though ;)03:42
*** SiFuh has joined #crux03:44
pidsleyhis atom 510 at least has two real cores03:45
pidsleymy 41o is only single core03:46
pidsleyi did put crux on a single-core sempron03:46
pidsleyi should try it on a Pentium D03:47
Worksterwell i think half of timcowchips problems are more than just a slow pc.03:54
pidsleyyou guys here are very polite03:59
pidsleyi like that04:00
pidsleysince i am a noob and no one blames me for it]04:01
*** ukamai has joined #crux04:09
*** ukamai has left #crux ()04:09
*** diverse has joined #crux04:11
*** mechaniputer has quit IRC04:18
openfbtdpidsley, basically done:
*** pidsley has quit IRC04:49
*** SiFuh_ has joined #crux05:01
*** SiFuh has quit IRC05:02
*** pidsley has joined #crux05:11
joacimpidsley: 35 minutes for a kernel doesn't sound that bad to me05:50
joacimI was ok with a C2D E6300. Took me about 15 minutes to compile a kernel on that one05:50
joacim2 hours for firefox05:51
joacimmy current i5 compiles my kernel in about 90 seconds05:52
joacimjust leaving it over night for compiling stuff. Don't think I would bother with that on a laptop tho05:54
openfbtdpidsley, so now it's up to you to create the Ember project06:50
openfbtdA CRUX spinoff :)06:50
*** mike_k has joined #crux06:57
*** timcowchip has joined #crux07:33
timcowchipWorkster: are you saying the other half of my problems are PEBKAC? :D07:36
Romsterthat or the manual your reading is wrong07:44
Romsteror maybe you tinker as much as i used to do and brake everything in the process.07:49
Romsterthen learned how to fix it07:49
timcowchipI learned to cut back on the tinkering a little07:50
timcowchipbut thats because I'm not as good at fixing my own breakage as you07:53
Romsteri only learned by lots of reading07:54
*** timcowchip has quit IRC08:14
diversetimcowchip, the only way you are going to get good at fixing stuff on your own, is by learning from what you break. Don't be afraid when something breaks. When you do break something, try to solve the problem, go through a troubleshooting process, narrow the problem down, research and apply the solution. From doing that you should be able to get the idea and solve the problem when it occurs again.08:40
diverseand if you can't find a solution, then that's when it's a good time to pester Romster or someone else08:45
diverseand if you are too afraid to break things or learn from them, Crux isn't your distro of choice08:48
Romsteri don't clam to know everything either.08:52
diverseme neither, that's what research is good for08:53
*** SiFuh_ has quit IRC09:34
*** BitPuffin has joined #crux09:36
*** sh[4]rm4 has joined #crux09:47
*** BitPuffin has quit IRC09:48
*** BitPuffin has joined #crux09:50
*** nurax has joined #crux09:50
*** hotarono` has joined #crux10:37
*** hotaronohanako has quit IRC10:37
*** SiFuh has joined #crux10:47
*** jdolan has quit IRC11:19
*** jdolan_ has joined #crux11:20
*** jdolan_ has quit IRC11:29
*** dkoby has joined #crux11:32
*** jdolan has joined #crux11:35
*** mike_k has quit IRC12:06
*** jdolan has quit IRC12:06
*** phant0mas has joined #crux12:07
*** mike_k has joined #crux12:15
*** SiFuh has quit IRC12:53
*** SiFuh has joined #crux13:00
*** BitPuffin has quit IRC13:03
*** BitPuffin has joined #crux13:06
*** Pingax has quit IRC13:18
*** SiFuh has quit IRC13:37
*** SiFuh has joined #crux13:40
*** SiFuh has quit IRC13:47
*** jdolan has joined #crux13:52
*** SiFuh has joined #crux13:56
*** pidsley has quit IRC14:08
*** SiFuh has quit IRC14:13
*** mike_k has quit IRC14:15
*** dkoby has quit IRC14:19
*** mechaniputer has joined #crux14:27
*** mechaniputer has quit IRC14:31
*** SiFuh has joined #crux14:35
*** deus_ex has quit IRC14:37
*** deus_ex has joined #crux14:38
*** mechaniputer has joined #crux14:39
*** SiFuh has quit IRC14:45
*** syncn has joined #crux14:45
*** mike_k has joined #crux14:49
nwe any idea whats wrong.. I got it after I upgrade cyrus-sasl16:17
*** mechaniputer has quit IRC16:31
nweI have fixed it now :D17:25
frinnsthow did you fix it? (in case someone else sees the same error)17:34
nwewith a patch.17:34
nwe2 sec17:34
frinnst19:36 <nwe> with this patch
frinnsttilman: lol17:37
*** _odie has quit IRC17:50
teK__why does everything break with the new cyrus-sasl, it's been around for quite some time18:38
*** u7knv9h has joined #crux18:40
u7knv9hcan i ask for help?18:41
teK__you can try18:41
teK__don't ask to ask18:41
u7knv9hoh, well, i just installed crux on my laptop, but i'm not able to connect it to internet with an ethered connection and neither with wifi18:43
u7knv9hduring the installation i selected all the packages from core, opt and xorg18:44
teK__the next steps depend onwether you use static or dynamic (via dhcp) IP address assignment18:44
u7knv9hi use a dynamic ip address addignment18:45
teK__is your card listed WITH the kernel driver used if listed with lspci -k?18:46
teK__maybe you skipped the kernel driver during kernel configuration?18:46
jaegerdoes it show up in "ip link" or "ifconfig -a" output?18:48
u7knv9hduring kernel configuration i selected the wireless support, and yes, it shows in "ifconfig -a" output18:49
jaegerboth wired and wireless or just one?18:50
u7knv9hboth wired and wireless18:51
jaegerWhat if you run dhcpcd manually? "dhcpcd -t 10" or similar (without specifying the interface name)18:51
u7knv9hdhcpcd[31467]: sending commands to master dhcpcd process18:53
jaegermight want to kill the original process, then18:58
u7knv9hand then what should i do?18:58
jaegerthen run the dhcpcd command I listed and see if it behaves properly18:59
u7knv9hi don't know if it's behaving properly but now the output is completely different19:00
u7knv9hsomething like19:00
jaegeryou can check ifconfig or "ip addr" to see if it got an address. also you should be able to see dhcpcd messages in /var/log/messages19:01
u7knv9hi think it got an address19:02
u7knv9hbut it still not connecting to internet19:03
jaegerdid you modify /etc/rc.d/net or is it the default one?19:04 the output of ip r and ip a19:04
u7knv9hteK__ i can't do that, it's not connectin19:07
u7knv9h the /etc/rc.d/net it's the default one19:08
jaegercan you ping your gateway? does it work if you specify the ip address, netmask, and gateway manually?19:12
jaegerYou could save the output from those commands to a usb drive and paste them from another computer, also19:14
u7knv9hspecifying ip address, netmask and gateway where? in /etc/rc.d/net? anyway, no it doesn't ping my gateway19:18
jaegerWhat address did the interface get? Does it start with 169?19:19
jaegerok, then DHCP didn't work19:19
jaegerSo back to making sure the interfaces actually exist and are what you expect rather than tunneling or sit0 interfaces, etc.19:19
u7knv9hi'm pretty sure the interface exists, it has always worked with other distro (even though on debian and some distros it had a different name, on arch on others the same name)19:21
jaegercan you copy the "ip a" and "ip r" outputs to a usb drive or something?19:23
u7knv9hnow i try19:23
u7knv9hsorry, now i must go, thanks anyway, see ya later19:31
*** u7knv9h has quit IRC19:31
*** mechaniputer has joined #crux19:58
*** looop has joined #crux20:35
*** phant0mas has quit IRC20:53
*** jdolan has quit IRC20:58
*** mike_k has quit IRC21:04
*** Henesy has joined #crux21:12
*** Feigr2 has joined #crux21:14
*** Feigr has quit IRC21:15
*** nurax has quit IRC21:16
*** mechaniputer has quit IRC21:18
*** mechaniputer has joined #crux21:19
*** mechaniputer has quit IRC21:34
*** looop has quit IRC22:08
*** SiFuh has joined #crux22:08
*** thetornainbow has quit IRC22:08
*** thetornainbow has joined #crux22:15
*** hotaronohanako has joined #crux22:28
*** Henesy has quit IRC22:56
*** hotaronohanako has quit IRC23:04
*** hotaronohanako has joined #crux23:04
*** jdolan has joined #crux23:07
*** Henesy has joined #crux23:07
*** hotaronohanako has quit IRC23:35
*** hotaronohanako has joined #crux23:36
*** mechaniputer has joined #crux23:55
*** SiFuh has quit IRC23:56

Generated by 2.11.0 by Marius Gedminas - find it at!