IRC Logs for #crux Saturday, 2012-06-30

tilmanjaeger: i've read "it beats uses musicbrainz" and was like "whoa whoa wait a minute" :D00:25
*** vaddi has joined #crux01:13
horrorStruckanyone having some personal notes on how to properly build a toolchain and willing to share them? tia02:27
*** mike_k has joined #crux02:27
RomsterhorrorStruck, frinnst or jaeger should be able to help you there03:14
horrorStruckin fact i'm mainly confused by the proper way to handle a glibc upgrade03:15
Romsteri'd imagine you'd need to make a basic glibc only then rebuild binutils gcc then rebuild glibc fully.03:17
Romsterbut that might be wrong for a glibc only upgrade..03:17
horrorStruckbut won't the first glibc build+install screw up everything? oh well, let's try anyway03:18
Romsteri wouldn't do it on the live system03:19
horrorStruckchroot :)03:19
Romsteri've been meaning to ask jaeger for his notes on toolchain compiling.03:26
juegood morning03:35
juehorrorStruck: the way I do a bootstrap is: glibc - binutils - gcc - binutils - glibc03:38
jueit's important to restart init after a reinstall of glibc, without that you will get a unclean unmout of '/'03:39
jueyou can do that with 'telinit U'03:40
horrorStruckjue: thanks and now prepare for the worse: what exactly do you do when you say "I do a bootstrap" ?03:42
jueto build a new toolchain03:42
horrorStruckoh ok so that's no black magic or vodoo or somehting03:43
horrorStruckyou just rebuild that on the top of your rexisting system and everything should go well?03:43
juewell, not really but replacing the core library in a runnig system is always a bit hairy03:44
horrorStruckATM, i'm building glibc 2.15 in a chroot, then i'll i'll build binutils -> gcc -> binutils -> glibc. will i be able to install those on my machine after that or should i do some more rebuilds (core?) ?03:46
jueTBH, I wouldn't use 2.1503:47
Romstermorning jue03:48
jueI'd suggest to wait for 2.16 which is announced for next week03:48
horrorStrucki'm just trying to understand how this should be done so i'm not that concerned about versions right now TBH03:48
juehi Romster03:49
Romsteri think 2.15 needs a ton of patches03:49
horrorStrucki took patches from jaeger's 2.15 + one more from LFS03:50
Romsteryeah see the patches!
Romstermight pay to check gentoos patches too03:52
horrorStruckusing +
horrorStruckbuilt fine03:52
horrorStruckinstalled fine, revdep is quiet03:53
Romsterif it runs fine with no segfaults is another story03:54
horrorStruck:( dream destroyer ^03:56
Romsteri meant it as be warry and test before you install on your system03:56
Romsteralways rebuild your shell (bash) and kernel after toolchain and revdep all of core03:57
horrorStruckkernel too? i've been building kernels on build machines with different glibc/binutils/gcc versions without any issue, do you think it really matters?04:00
Romsteronly when glibcfunctions chane add removed afaik its a hit or miss04:01
tilmanhorrorStruck: does the kernel link to glibc? :P04:02
horrorStruckdont laugh at me guys i'm learning here :P i feel like i'm the first human walking on mars04:03
tilmanwhen you have a working kernel at some point, you can replace binutils/gcc/glibc as much as you want04:03
tilmanit wont influence the working kernel ;)04:03
horrorStruckyeah that's why i was a bit surprised04:04
horrorStruckbut thanks for the explanation04:05
Romsterah i thought kernel depended on glibc too04:05
Romstermust be other way around04:05
horrorStruckwell you need kernel headers to build glibc04:06
Romsterunless you set the kernel headers to a newer version than your current kernel then it should work04:06
Romsterwatch out for live cd's i locked myself out and could not chroot on a older live cd back in as kernel on live cdw as too old for glibc headers in system04:07
horrorStruckLinux blackout 3.5.0-rc4-tip+ <-- should be recent enough :P04:11
horrorStruckah you mean the kernel on the live cd04:12
horrorStruckanyway glibc is built against 3.0, should be fine04:12
niklaswegood morning04:26
niklaswehow are you horrorStruck ?04:32
horrorStruckfine thanks, what about you?04:33
niklaswenice to hear, Im fine too, just sitting here and drink some coffee :)04:35
Romsteryeah kernel on live cd was too old for minium kernel vesion on glibc after bumping toolchain and forgetting to run lilo after rebuilding kernel04:35
niklasweand wondering if I should order 3 new disc or not..04:35
Romster don't watch if you don't want to remember spiderman this way04:36
horrorStrucki already see him like the guy from /the social network/ anyway04:37
horrorStruckniklaswe: 3 at once?04:38
niklaswehorrorStruck: yeah :) 1 ssd 60GB for system and 2 2TB for my new raid :P04:39
horrorStrucksounds cool, order now! :P04:41
niklaswehorrorStruck: I got 2 TB  disk at home :)04:41
niklasweso totally i will get 4 TB disk in my raid 504:41
horrorStruckyou never have enough (if like me you DL tons of HD content that you never watch)04:42
niklaswehorrorStruck: hahha yupp :p04:42
niklaswebut its good to have them ^^04:43
niklaswemaybe sometime i will look at it :)04:43
horrorStruckyes, for "later" :P04:43
Romsteri got like a total of 5tb here atm and i need more space.04:44
horrorStrucki'm close to that04:45
horrorStruckand all disks are full04:45
niklaswehorrorStruck: haha yeah :P04:46
horrorStruckoh 3.6TB to be precise04:46
niklasweurk.. I hate duty :/04:46
Romster3tb for au $179 but no 4tb disks in fav store yet04:47
horrorStruck~ same price here and they are produced few km away from my house. errf...04:52
horrorStruckso i'be built glibc2.15 -> binutils2.22 -> gcc4.6.3 -> binutils -> glibc in a chroot, all went fine, revdep is mute, i can re-renter the chroot without issue, am i safe to install those packages on my machine?04:58
horrorStruckor anything else i could check?04:58
Romsterrelaseding in April i don't see anywhere i can even buy 4tb yet.04:59
Romstermost likely i would rebuild bash for good measure and install that too.05:00
horrorStruckok thanks, i will do it05:00
Romsteri got locked out after a toolchain upgrade that bash wouldn't work for me.05:00
Romsternot sure why.05:00
Romsteri want with a external enclosure.05:02
horrorStruckinstalled, reboot, all fine so far05:09
*** vaddi has quit IRC05:36
RomsterhorrorStruck, check ports some might fail with gcc 4.6.306:18
horrorStrucki'm building glibc from 2.15 git branch now to get some few more fixes, does it mean i have to rebuild the entire toolchain again or being the same branch makes it unnecessary?06:20
horrorStruckRomster: yes i was using 4.6.3 for some time, had to collect few patches here and there06:20
horrorStruckthanks for the heads-up06:22
joe9why doesn't crux have a check function in the Pkgfile similar to this Arch file:
joe9isn't it a good idea to check/test the package?07:08
Romsteryeah but i don't know if crux wants that /feature/07:15
Romsterthere is plenty of things i'd like but then we would be closer to arch than crux too.07:16
Romsterhmm for the price of 1 4tb hdd i can get a 3tb disk for half it's cost. so i could get 6tb for the price of a 4tb disk07:22
Romsterstupid prices.07:22
Romster4tb should be 1/3 more expensive not 1/2 more.07:23
Romsterthan a 3tb07:23
Romsterno actually than 2 times as.07:23
joe9how does crux deal with the .pod files to which perl installation information is appended:07:53
joe9Appending installation info to /var/pkgmk/work/ls++-git/pkg/usr/lib/perl5/5.12/linux-thread-multi/perllocal.pod07:53
joe9i tried this approach:
joe9any thoughts, please?07:53
joe9or, is it a good idea to just rm that file?07:55
Romsterfind $PKG -name perllocal.pod -exec rm {} \;07:58
joe9ok, thanks.07:58
Romstergrep though ports if not sure.07:58
*** cruxbot has joined #crux09:00
*** spider44 has quit IRC09:36
*** spider44 has joined #crux09:36
jaegerRomster, horrorStruck, jue: I usually use the crux iso tree to build toolchains, honestly. It takes a bit longer but works very well. That way the toolchain, kernel, shell, everything gets rebuilt multiple times10:02
juejaeger: a bit longer :)10:13
tilmanhorrorStruck: i thought we were talking about upgrading the toolchain in a working system10:14
jueindeed, that was my understanding as well10:15
Romsterit is on a working system but it's safer to use a new directory or chroot before building/moving the toolchain to the running system10:16
jaegerJust passing on how I do it... that way not only is the entire package set bootstrapped (built 3 times) but it's done in chroots10:17
jaegerIf the bootstrap succeeds I've never had problems with the toolchain afterwards10:17
jaegerThe time investment is worth it, in my opinion, because I'd be testing it by recompiling a bunch of things later anyway10:17
juesure, that's the safe and preferred way. What I've outlined above are the minimal steps to get a proberly bootstrapped toolchain, not more10:22
jaegerI'm not trying to correct anyone, just outlined it because of this: 03:26 < Romster> i've been meaning to ask jaeger for his notes on toolchain compiling.10:23
Romsteri thought jaeger may have had a specific compile order from scratch.10:31
jaegerIf I do the toolchain only I do it in the same order as jue mentioned10:32
Romsterok that's good to know.10:34
Romsteri was expecting a bit more detail on it but the LFS guide is good enough.10:35
jaegerIt's been some time since I did it that way, though, I like the thoroughness of the bootstrap method10:35
jaegerwell, there's not much detail to give, honestly... the iso.git Makefile does all the heavy lifting10:35
Romsterunderstood and is more reliable.10:35
jaegerunrelated to the toolchain stuff, I see OCZ SSDs mentioned above, avoid OCZ like the plague10:36
jaegerget something from crucial, corsair, samsung, intel, or kingston instead10:36
Romsterhmm bad quality?10:36
jaegerthey have a track record not only for very bad quality but for lying to customers and horrible RMA service10:37
jaegerTo back that up with some (anecdotal) personal experience, the first two SSDs I bought were OCZ. One of them got RMA'd twice and the other is now dying slowly but I can't bring myself to work with OCZ's RMA support again to get it replaced10:38
jaegerNow I have an intel and 2 samsung SSDs at home and another intel one at work and I love those10:39
Romsteroh man thanks for the top, i'm not buying any yet but i am looking around.10:39
jaegerI realize in AU it's tougher to get things shipped but I think it's worth avoid OCZ if you can10:39
Romsterjsut costs a bit for freight but there is others besides OCZ here.10:40
Romsteri think my ram is OCZ branded...10:41
jaegerIt's a shame that amazon and newegg won't ship to AU10:41
Romsterbut i've not had any issues at all with that ram.10:41
Romsteroh you kidding me...10:41
jaegerOCZ used to make decent RAM, just not SSDs10:41
Romsteri was looking at that SAS card ^10:41
jaegerPerhaps I'm wrong but that's what I've been told10:41
Romsterhave been for a while now and when i got some spare money i was going to order it. as i can't find it any cheaper anywhere else.10:42
Romsterguess i'll find out when i do try to later on.10:42
jaegerI like LSI's hardware, got quite a lot of it10:42
jaegerraid controller at home and lots of SAS HBAs at work10:43
Romsterif i can't i can do a deal with a trusted usa friend to post it to me.10:43
Romsterworks with recent kernels10:43
Romsterunlikesome other 4 ports sata pci card i bought a few years ago sitting here collecting dust...10:43
Romsterbecause it said it's linux compatible but only with there binary blob i think it was an adapotec or something brand10:44
Romsterhad a driver in the kernel that was reversed engineered but it was buggy as hell10:44
jaegerI haven't used that particular one but the ones I have at work are quite well supported... 1068e or something10:45
Romsterfound it on there.10:46
*** nthwyatt has quit IRC10:46
jaegerlooks like that has a sas1068e controller, should work great10:48
jaegeractually those are two different cards, sorry, I looked at the wrong one10:49
*** nthwyatt has joined #crux10:49
jaegeranyway, the mpt2sas driver works well10:49
Romsteryeah that's what i was hoping on all the reviews i've read.10:50
joe9do you set this flag: # PKGMK_COMPRESSION_MODE="gz"10:58
joe9to .xz?10:58
cruxbot[contrib.git/2.7]: powertop: also depends on libnl10:59
Romsteri odn't but you can if you want smaller built packages11:03
Romsterthat takes a little more time than gz to compress.11:04
joe9Romster: ok, thanks.11:04
joe9Romster: would you mind pasting your pkgmk.conf?11:05
joe9and prt-get.conf11:05
Romsteris that wise it's not exactly clean :D11:05
joe9that's ok. I just want to see how you do stuff.11:05
joe9romster, that is cleaner than mine.11:08
joe9btw, what is multilib? you have it before the x86_64?11:09
Romsteri'm using jaeger's multilib iso11:11
joe9Romster: is that better than x86_64?11:11
Romsterruns ports on top of x86_6411:11
joe9ok, thanks.11:11
Romsterbetter as in it also supports 32bit as well as 64bit11:11
joe9good, thanks.11:11
Romstermostly for wine as that needs 32bits for 32bit PE11:12
Romsterbut some emulators also use 32bits too.11:12
Romsterand some binary only ports11:12
joe9i do not use wine. should I still bother with multilib?11:13
joe9or, any emulater that I know of.11:13
Romsternot unless you want some emulator or binary only program that's 32bit only to work11:13
joe9ok, thanks. no multilib for me then.11:14
Romsteronly some emulators need 32bit not all. those that do are generally got -32 on the end of the name=11:14
joe9i like how you have nice in the pkgmk command.11:15
Romsterit's not for everyone.11:15
Romsteryeah i do that on all my boxens :D11:15
juewrt xz vs gz compression, 'littel more time' is a bit understated. It's much slower for big packages, see
Romsterok alot more time on large ports.11:17
jueRomster: no offence, I just remembered that I did some tests :)11:18
joe9Romster: why are you not using pretty-make?11:18
joe9i see that you have it commented out.11:18
joe9just wanted to check if it is really "pretty"11:18
Romsteri did hack pkgmk to use pigz that so sped up gzip on a quadcore cpu11:19
Romsterit's pretty but slows the compile down a little.11:20
Romsterbut that hack was back when i was using hvl11:20
Romsteri'm all stock atm11:21
Romsteri've done many hacks in the past but lately not much at all.11:23
Romsterlots of effort for small gains.11:23
Romsterbe nice if libarchive could do stuff in parallel11:24
joe9romster, why not add this? export LDFLAGS="-Wl,-O1 -Wl,--hash-style=gnu -Wl,--sort-common "11:24
joe9to your pkgmk.conf?11:24
Romsteri could just haven't bothered and i did have a patch to binutils to enable that by default.11:25
Romsteras not all ports pass LDFLAGS11:26
Romsteri should add that patch as a feature request for crux 2.8 but crux's policy is not to patch unless it's broken.11:27
Romsteri don't bother so much with -fomgoptimise11:28
juejoe9: an official maintainer should more than less use a default system11:31
juefor me such "optimizations" are a no-go11:32
Romstermost of them break stuff.11:32
Romsteri know first hand i've tried them.11:32
Romsterwait for it...11:41
RomsterEarlier this year, official timekeepers from across the world discussed  whether to eliminate the practice of adding leap seconds. They decided  they needed more time to think about the issue and will next debate the  issue in 2015.11:41
horrorStruck2 hours later... thanks all for your advices on toolchain building. to my surprise, the whole process has been very smooth.11:41
cruxbot[opt.git/2.7]: smartmontools: update to 5.4312:03
*** tilman has quit IRC12:10
*** tilman has joined #crux12:12
*** ChanServ sets mode: +o tilman12:12
*** vaddi has joined #crux12:36
joe9do you have a separate partition for var/ccache or var/cache like folders?13:05
joe9or, in general, put the whole /var folder in a different partition?13:05
joe9i notice that these folders are the main reason I run out of space on the disks. I can set the limits with ccache.13:06
joe9but, just curious if any of you actively manage that folder with a partition, etc.13:06
mike_kjoe: not really. /var's size depends on the installed/actively used daemons. I tend to have FHS-friendly paths in custom ports like /srv/db/postgresql for non-trivial amount of data13:19
mike_kI' go for 500MB-2GB var (depending on amount of logs to keep), plus separate /var/lib/someapp if it takes too much space13:22
mike_kdata fragmentation might be an issue for some apps or access patterns too13:24
joe9mike_k: ok, thanks.13:24
tilmanfor a 'desktop' system, i like having 500 mb minimum for /var13:31
tilmansome distros require quite some space there (hello debian)13:31
krueI usually have a partition for programs and a partition for data.13:31
krueThe program partition is usually the root.13:31
krueThe data paritition is mounted somewhere like /mnt/data.13:32
krueBoth /home and /var are symlinks into /mnt/data.13:32
krueThis scheme separates the physical partition files are stored on and the pathname used to access them.13:32
*** jue has quit IRC14:22
*** mike_k has quit IRC16:10
*** vaddi has quit IRC16:11
*** joacim has quit IRC16:35
*** joe9 has quit IRC16:48
*** nogagplz_ has quit IRC17:47
*** nogagplz_ has joined #crux17:48
*** nogagplz_ has quit IRC17:53
*** joe9 has joined #crux18:30
joe9i am getting these errors on a new machine that I installed with crux:
joe9/bin/sh: warning: setlocale: LC_ALL: cannot change locale (en_US.UTF-8)18:32
joe9in my shell profile file:
joe9i see some places where this is defined in /etc/rc.conf.18:33
joe9is that so?18:33
joe9is there anything I can do to generate these locales?18:33
joe9locale -a
joe9 is this relevant to me?18:36
joe9got it: localedef -i en_US -f UTF-8 en_US.utf818:38
joe9is what I needed18:38
joe9ruby build, this warning
joe9jaeger: thanks, I missed it. sorry about that.19:44
jaegerIt happens, just figured I'd point it out so you know it's there for future reference, if you reinstall, etc.19:46
joe9ok, thanks.19:46
*** joacim has joined #crux19:48
joe9does gcc-fortran even build on x86_64?
joe9beginning of the pkgbuild log: :
horrorStruck <--- more experiments today :P21:33
joe9horrorStruck: when will it come to crux? is it a replacement to glibc?21:42
horrorStruckwhen i have no idea but i guess it will be the version devs will pick up for 2.8. it's not exactly a replacement, it's a new release21:45
*** mavrick61 has quit IRC21:45
*** mavrick61 has joined #crux21:46
*** nogagplz_ has joined #crux23:38
*** nogagplz has quit IRC23:39
*** SiFuh_ has quit IRC23:46

Generated by 2.11.0 by Marius Gedminas - find it at!