IRC Logs for #crux-devel Wednesday, 2015-02-25

*** mechaniputer has quit IRC00:26
*** tvaalen has quit IRC01:07
*** tvaalen_ has joined #crux-devel01:08
*** nrxtx has quit IRC01:13
*** teK_ has quit IRC01:13
*** nrxtx has joined #crux-devel01:24
*** teK_ has joined #crux-devel01:24
*** jaeger has joined #crux-devel01:24
*** irclogger has joined #crux-devel01:24
*** nrxtx has quit IRC01:37
*** teK_ has quit IRC01:37
*** jaeger has quit IRC01:37
*** irclogger has quit IRC01:37
*** leo-unglaub has quit IRC01:37
*** jue has quit IRC01:37
*** frinnst has quit IRC01:37
*** pitillo has quit IRC01:37
*** tvaalen_ has quit IRC01:37
*** Romster has quit IRC01:37
*** Workster has quit IRC01:37
*** heroux has quit IRC01:37
*** novak_ has quit IRC01:37
*** Lukc` has quit IRC01:37
*** atharic has quit IRC01:37
*** irclogger_ has quit IRC01:37
*** c0x has quit IRC01:37
*** prologic has quit IRC01:37
*** erdic has quit IRC01:37
*** diverse has quit IRC01:44
*** irclogger has joined #crux-devel01:52
*** jaeger has joined #crux-devel01:52
*** teK_ has joined #crux-devel01:52
*** nrxtx has joined #crux-devel01:52
*** tvaalen_ has joined #crux-devel01:52
*** Romster has joined #crux-devel01:52
*** Workster has joined #crux-devel01:52
*** irclogger_ has joined #crux-devel01:52
*** heroux has joined #crux-devel01:52
*** Lukc` has joined #crux-devel01:52
*** novak_ has joined #crux-devel01:52
*** atharic has joined #crux-devel01:52
*** frinnst has joined #crux-devel01:52
*** pitillo has joined #crux-devel01:52
*** c0x has joined #crux-devel01:52
*** jue has joined #crux-devel01:52
*** prologic has joined #crux-devel01:52
*** erdic has joined #crux-devel01:52
*** nrxtx has quit IRC03:08
*** nrxtx has joined #crux-devel03:08
*** mavrick61 has joined #crux-devel04:11
jueprologic: still, why do we have a bigger problem? The only _real_ and current problem I see is that our rc-scripts are not very reliable and that's what ssd improves in a nice way, not more and not less07:13
jueand why do you think that ssd is doing that not very well?07:14
frinnstanybody with a lwn subscription that mind sharing the latest article regarding glibc?08:29
juefrinnst: IIRC tilman has one08:35
frinnstthanks :)09:06
Romsterglibc is what gone stale? where are the new releases09:15
Romsteronly thing i didn't like was bumping sshd rc script so soon. ssd itself i'm fine with09:16
frinnstwhat?09:16
Romsterglibc with all them CVEs09:17
Romsternot patched upstream yet?09:17
Romsteri haven't looked at there git tree.09:17
frinnstno pointrelease, no09:17
frinnstpatches yes09:17
Romsterguessing you troll https://sourceware.org/git/?p=glibc.git;a=shortlog09:22
frinnstyes09:23
Romstermaybe they think them CVEs aren't high risk to make a point release.09:23
Romsteras for crux 3.2 is there much planned?09:24
Romsternew gcc glibc possibly palm?09:24
Romsterpam*09:25
Romsteri guess also remove tcp_wrappers at that point too?09:27
jueRomster: just read a bit in the runit docs and are quite unhappy with it, besides the init stuff it's not more than a simple process supervisor and can not do most of the interesting things of ssd like gracefull shutdown, changing user, creating pids if the process doesn't it etc.10:57
Romsterjue, yeah it has basic watch dog timer that restarts a died process every 60 seconds if it's not already running. and has dependency resolution for starting services in order.11:05
Romsteras for the rest of it there is 1 2 and 3 scripts for single user multi user and shutdown.11:06
Romsterhttp://hvlinux.net/linux/ports/hvlinux/core/runit/runitscripts.tar11:09
Romsterwhat i was using.11:09
Romsteretc/runit/3 for stopping services etc11:09
Romsterand where is this graceful shutdown coming from?11:11
Romsterif services does not stop when told too. then we run SIGTERM sleep 5 and if not stopped then SIGKILL11:11
Romsterunless ssd is even smarter than that, i haven't looked. i don't care as long as it does it's job.11:12
Romstermy /concern/ was being locked out of a box miles away from it with no sh access. due to the admin forgetting to install start-stop-daemon. but you reverted the sshd rc script so that wont happen.11:13
juewell, a good example is squid, it takes quite a long time to shutdown11:15
jueand even several seconds to stop11:16
Romsterhttp://hvlinux.net/linux/ports/hvlinux/core/openssh/sshd/run and syntax for sshd11:16
jueand the script has to wait until this is done11:17
Romsterso even my old method in runit wouldn't be correct.11:17
Romsterbut i haven't ran runit since 2010 as you can see.11:17
jueno11:17
juethe --retry option of ssd was the most interesting feature for me11:18
jueyou can say with it "send a TERM wait 10 seconds and if the process still exist send a kill ..."11:19
Romsteri haven't looked into finit if this does wait/retry11:20
Romsteri'm not particulary keen on all the symlinks runit did have but it did run a faster boot up process and made sure services stayed running.11:20
jueI took a quick lock at finit and got the impression that it is not what we want11:21
Romsterand services scripts were very simple.11:21
Romsterit does not fork them to the background they run under the sv super vision11:21
Romsterie not daemonised11:21
Romsterso sv knows when its running or not.11:22
jueyeah, the scipts are simple but they are doing nothing11:22
jueif you have to change the user it will be as complex as our current scipts at least11:23
Romsterthey shouldn't have too the daemon should take care of the work and even do the waiting and retry.11:23
Romstersu ...11:23
juesure11:24
juebtw, Void Linux looks really interesting11:25
Romsteryes i've been poking around and it's really clean written. i have not ran it yet.11:25
Romsterdare i say it. it looks more structured than crux. at first looks11:26
jueme neither, just read a bit their docs11:26
Romsterthere tempate fles for building are a little more abstracted. but not by much11:26
Romsterlike a variable for all the configure options11:27
juefor sure it's more complex than CRUX, but obviously that's what most people like to see nowadays11:28
Romstertemplate*11:29
Romstercomplexity comes at a cost.11:29
Romsterharder to debug11:29
Romsterpretty much like how i could get a LFS example and throw it in a Pkgfile and only do minor changes like adding DESTDIR=$PKG pkgmk11:30
Romsterno other distro afaik allows this.11:31
Romstermaybe slackware comes close11:31
Romsteri think they might be doing something similar to frugalware's up2date and my version sorting11:34
Romsterplan to dig deeper into that. and there buildbot11:34
Romsteri've got semi automated docker builds working now.11:34
*** crash_ has joined #crux-devel12:02
juethat's really nice -> http://curl.haxx.se/docs/vulnerabilities.html12:55
Romstermore projects should do this13:04
frinnstwow, imagine that for glibc  for example13:41
jaegerlearned something odd yesterday: core/zip can create large files but core/unzip can't unzip them13:57
jaegermight just be a build flag missing or something but it scared me at first, thinking a big zip had been corrupted13:58
frinnsthuh14:03
frinnstit could also be that unzip is shite :-)14:03
jaegerpossibly :) p7zip had no trouble extracting the files, fortunately14:03
*** tvaalen_ has quit IRC14:19
*** tvaalen_ has joined #crux-devel14:19
juejaeger: can you please test this instead -> make -f unix/Makefile generic14:20
jaegersure14:22
jaegerjue: feels much slower but it works14:40
jaegerI think it's probably not worth changing core/unzip for just this use case14:40
jaegeras an amusing side note:14:41
jaeger19Gwoods14:41
jaeger9.5Gwoods.zip14:41
jaegerthat zip file is the "woods" dir with no compression, simply stored14:42
jaegeryay for huge filesystem block sizes14:42
frinnsthaha14:50
jaegerjue: I guess the slowness is my imagination. only 15s difference between 7z and unzip when extracting that same zip file. I can't really compare "generic" vs "linux_noasm" unzip on this file since it won't unzip correctly with the latter14:53
*** Workster has quit IRC16:17
juegeneric runs a configure scipt to determine a couple of flags e.g. -DLARGE_FILE_SUPPORT, I'd say we should use that target and don't think that there's any speed difference between linux_noasm16:18
jaegerno objection here16:31
juewith a zipped kernel tree (234M zip file) I cannot measure a noticeable difference16:35
jaegernice, good to know16:39
*** diverse has joined #crux-devel17:14
*** diverse has quit IRC17:15
*** greduan_ has quit IRC19:13
*** diverse has joined #crux-devel20:04
*** Feksclaus has joined #crux-devel20:27
prologicjue: it's fine; I think we're going with ssd21:32
*** leo-unglaub has joined #crux-devel21:35
*** Workster has joined #crux-devel22:37
*** Feksclaus has quit IRC22:43
*** crash_ has quit IRC22:51
*** leo-unglaub1 has joined #crux-devel23:01
*** leo-unglaub has quit IRC23:03
*** leo-unglaub has joined #crux-devel23:49
*** leo-unglaub1 has quit IRC23:53
*** c0x has quit IRC23:59

Generated by irclog2html.py 2.11.0 by Marius Gedminas - find it at mg.pov.lt!