IRC Logs for #crux Sunday, 2014-07-06

*** tilman has quit IRC00:04
*** SiFuh has joined #crux00:11
frinnstI dont have any firmware installed for my r8169 - other than what comes with the kernel00:39
*** jdolan has quit IRC00:56
*** jdolan has joined #crux00:56
*** jdolan has quit IRC01:01
*** dkoby has joined #crux01:02
*** jdolan has joined #crux01:15
*** jdolan has quit IRC01:19
*** hotaronohanako has quit IRC01:23
*** hotaronohanako has joined #crux01:23
*** hotaronohanako has quit IRC01:23
*** jdolan has joined #crux01:27
*** jdolan has quit IRC01:34
*** jdolan has joined #crux01:46
*** timcowchip has joined #crux02:26
*** pidsley has quit IRC02:28
timcowchip prt-get install subversion02:30
timcowchipre-installing lilo after compiling my network driver into the kernel fix my innanet02:35
timcowchipdo I need to install apache to build subversion?02:37
*** SiFuh has quit IRC02:41
*** SiFuh has joined #crux02:42
timcowchip apparently I can't install apache02:45
*** mavrick61 has quit IRC02:48
*** mavrick61 has joined #crux02:49
timcowchipI think I need to install apr02:53
*** pidsley has joined #crux03:42
*** dkoby has quit IRC03:44
*** renovaatio has joined #crux04:16
renovaatioi'm trying to setup a crux virtualbox guest on my windows box04:16
renovaatioit's tedious04:17
renovaatiodo you know if I can find a prebuilt crux guest file somewhere?04:17
jaegerNone I'm aware of. I used to keep one around but that was years ago04:18
jaegerWhich part is giving you trouble?04:18
renovaatiowell, kernel panic04:20
renovaatiokernel wont boot04:20
jaegerMost likely missing the proper disk controller support04:21
jaegerIncidentally the default config that comes with the install should support pretty much everything virtualbox does04:21
renovaatioi went a stp too far then04:22
renovaatioi must've fucked up my config04:22
renovaatioimma format and retry with the default configs04:24
renovaatiounless it's possible to revert to defaut configs?04:24
renovaatiowhere can I find the kernel sources in the install cd ?04:26
jaegerYou can copy it from the CD while booted04:26
jaegershould be in /media/crux/kernel or similar04:27
renovaatiogot it04:27
renovaatiocrux is04:27
jaegerSo you don't have to format again if you don't want to, you can boot into the CD, mount your system, copy the config, chroot in, etc.04:28
renovaatioyeah doing it right now04:28
renovaationever thought I'd find a distro that fits me that much04:29
renovaatioeverything fro crux is A+04:29
renovaatioexept the distribution of sources to be compiled instead of prebuilt packages04:29
jaegerComes with the territory (source-based distro) :)04:31
jaegerwhile you're booted into the installation media, check which storage (and network) modules are in use with "lspci -k | grep 'in use'"04:35
jaegerthen you can compare those with the kernel config to make sure you have the right things enabled04:35
renovaatiostill kernel panic04:45
renovaatioi give up04:46
jaegersorry :/04:46
renovaatioi cant belive there is just crux and gentoo who are not using systemd04:46
renovaatioand slackware04:46
renovaatioall mainstream distros are using it04:47
jaegereventually we won't have a choice, I imagine. When the eudev team gets tired of trying to avoid systemd04:47
*** Sleepy_C1der has joined #crux04:49
*** deus_ex has quit IRC04:49
*** timcowchip has quit IRC04:53
*** frinnst_ has joined #crux04:54
*** frinnst has quit IRC04:58
*** BitPuffin has quit IRC04:58
*** sh[4]rm4 has quit IRC04:59
*** sh[4]rm4 has joined #crux04:59
*** Sleepy_Coder has quit IRC05:00
renovaatiojaeger: I can't believe no one is fighting against systemd05:02
renovaatioit is clearly an attempt at taking over GNU/Linux05:02
renovaatioit's a freaking pile of shit05:03
Romsterjust go back to static nodes when that happens05:05
renovaatioRomster: i'm a noob. please explain05:20
*** darfo has quit IRC05:21
Romsterbefore udev came along we all had static nodes in /dev05:21
Romsterwith specal numbers for each device.05:22
*** darfo has joined #crux05:22
Romsterudev made this dynamic and added/removed with udev rules then along came systemd shoved down everyones throats and udev was in that but could be compiled on its own. then they stopped that.05:23
openfbtdudev can still be used on its own05:24
Romstersystemd makes me mad on every level05:24
openfbtdYou can't build it separately, but you still can use it05:24
openfbtdI doubt that will continue for long though05:24
Romstereudev is a fork of that effort05:24
*** BitPuffin has joined #crux05:24
openfbtdYeah, I know, I use eudev on my arch desktops05:25
Romstercrux 3.1 has eudev05:25
*** mike_k has joined #crux05:25
Romsterbut its a relaly big fuck you to all other init systems when systemd is forced down everyones throats.05:25
Romstersystemd makes me so mad -_-05:26
openfbtdIt used to make me that mad05:27
openfbtdNow I'm just... meh05:27
openfbtdThe community will write their own udev if it comes to it05:27
openfbtdThey hate systemd too :)05:27
openfbtdSome of them use mdev to manage desktop systems05:27
Romsterjust continue on eudev05:27
Romsterbut the kbus kernel thing is another thing...05:28
Romsterhal/dbus mess05:28
openfbtdI don't understand the idea of putting dbus into the kernel05:28
openfbtdBut if they make it accessible via a virtual file system, that'd be neat05:29
Romsterthe mess i wanna do is punch the heck out of pulseaudio/systemd05:29
openfbtdDBus that is easily integrated with shell scripts is better than the current one.05:29
Romsteri bet it'll require sysemd, all apps will want systemd i'll go insane and just say fuck linux at that point.05:29
*** darfo has quit IRC05:30
Romsterand probably move to bsd or something05:31
*** BitPuffin has quit IRC05:31
openfbtdWell. Right now kdbus is a reimplementation of dbus in the kernel05:31
openfbtdWhat apps that you use _need_ dbus?05:32
openfbtdAside from, say, pulseaudio05:32
Romsterbet it forces the use of systemd unless you run another userspace daemon.05:32
openfbtdIt doesn't and I don't see how you would justify doing that05:32
Romsternotifications on desktops05:32
openfbtdIt's just dbus in the kernel05:32
Romsteryeah but userspace will be in systemd so then someone /else/ has to write that bit out of systemd.05:33
openfbtdAFAIK it's planned to be completely compatible with dbus05:34
Romsterand knowing them they'll constantly change the API to make it impossible and give up on maintaining05:34
openfbtdLinus will thow kdbus out of the kernel if they do that05:34
Romsterwish he threw systemd away05:35
Romsterwtf was he thinking05:35
Romsteris not the unix way05:35
Romsterone tool one job. not the damn kitchen sink05:35
Romsterseriously, linus the guy that wrote the kernel allows that junk in?05:35
Romsterkay has been nothing but trouble.05:36
openfbtdLinus can't do anything about a userspace program05:36
openfbtdsystemd isn't a part of the kernel05:36
openfbtdAnd kdbus is actually a good idea05:36
Romsterno but the systemd crap thrown into the kernel...05:36
Romsterkdbus should of been done ages ago05:37
openfbtdCompared to having dbus in the userspace05:37
openfbtdI personally don't like dbus _at all_05:37
Romsterhal/dbus is a mess. and the heck run systemd as the main 0 pid?05:37
openfbtdBut I'm a very strong supporter of the KISS way, and dbus is not KISS05:37
openfbtdPID 1. It's 1.05:37
Romsterthe once stable history of linux is about to become a pile of unstable crap like windows.05:37
Romsterthought crap starts at 0 in computers.05:38
Romsteranyways got to the point that i don't care too much anymore.05:39
openfbtdBut the pids start with 1.05:41
*** darfo has joined #crux05:42
Romsterjust too much shit to remember anymore. so i don't remember everything.05:45
*** timcowchip has joined #crux06:02
timcowchipconfigure: error: Package requirements (vdpau >= 0.4.1 x11-xcb xcb-dri2 >= 1.8) were not met:06:02
timcowchipcan't update mesa3d06:02
Romsterinstall libvdpau06:10
Romsterlook at prt-get deptree mesa3d see the missing packages.06:11
timcowchipok thanks romster06:12
timcowchipI see you got a tint2 port06:12
timcowchipmine's broken06:12
Romsteryeah newer that has the panel icon support.06:13
timcowchipyou where right about using svn checkout not working06:13
timcowchipthanks for doing the port for it06:14
Romsterno worries its a nice panel06:15
timcowchipyeah, I like it06:15
timcowchipI'll take my port down when I finish configuring rc306:17
renovaatio[01:29] <Romster> the mess i wanna do is punch the heck out of pulseaudio/systemd06:17
renovaatioi'd fight Poeterring06:17
renovaatio[01:37] <Romster> the once stable history of linux is about to become a pile of unstable crap like windows.06:19
renovaatiocouldn't agree more06:19
renovaatiosuckless is surprising06:20
renovaatioimma follow them more06:20
Romsterhope some good comes out of it06:25
renovaatioi don't know06:25
renovaatioi looked into bsd, and it does not look like a good choice06:26
renovaatiowhat remains?06:26
renovaatioslackware/gentoo ?06:26
renovaatiothose are a weird bunch06:26
Romsteri dunno i might of gone to reactos but that's been alpha for what 30 years -_-06:26
Romsterplus the decoded dick in that channel, decided to kick me out of the channel, oh because i was idling too muc, and not long before that i got asked by aother guy if i wanted a reactos cloak.06:27
Romsterand then decoded was a prick in #winehq and i felt like banning him but i didn't.06:28
Romsterthat was ages ago though too06:29
Romsteridiots everywhere in every system06:30
Romstercan not really win06:30
renovaatiowhen they are tamed, things are alright06:30
renovaatiobut when their stupidity takes over, we're doomed06:31
renovaatioi'm thinking of that Poeterring guy here, or whatever the fuck his name is06:31
RomsterLennart Poettering06:32
renovaatioi'm boiling06:33
renovaatiomust go to sleep06:33
Romsterokay g'night06:33
*** timcowchip has quit IRC06:54
openfbtdSome guy has forked my watchman right in the middle of a transition and a bugfix fest07:17
openfbtdSo he has an extremely buggy fork now :D07:18
Romsteryou should tell him to rebase it07:23
openfbtdIn fact, the fork he has now just doesn't work07:23
openfbtdAt all07:23
openfbtdAlready done07:24
openfbtdSHould all watchman-related tools share a version number07:25
openfbtdOr even should they report it07:25
Romsterupto you you.07:27
Romsterthought --version usually works on all binaries/scripts07:27
openfbtdThe thing is I will then have to change the version in four different places >_>07:29
openfbtdI could just make all the scripts call watchman and make it report the version though07:29
Romsteryeah or when you package it use and then sed -e 's/@@VERSION@@/1.2.3/' foo07:30
Romsteras part of a Makefile07:30
*** tilman has joined #crux07:30
openfbtdI don't have a makefile right now >_>07:31
openfbtdIt's a bash script07:31
Romsteror make it all call a common header file or some function for version.07:31
openfbtdThen again, I really do need to.07:31
RomsterMakefiles are pretty easy to knock up.07:31
openfbtdToo lazy for now07:31
openfbtdNever dealt with em07:31
Romsterthink i got one as a example here.07:32
openfbtdI have the knowledge of the fact that makefiles are not complicated07:32
openfbtdBut still, if I want to do it right, I need a configure script also07:33
openfbtdWhich is also not a problem by itself07:33
openfbtdBut meh, not now07:33
Romsterlazy <<07:34
nrxtxnah no configure, and no autotools stuff, just plain makefile07:34
openfbtdnrxtx, the script has a confdir07:34
openfbtdEven two07:34
openfbtdI want that to be configurable with a configure script07:34
openfbtdBecause that's how nice unix tools behave07:35
Romsteri said Makefile i never mentioned autohell07:35
diverseDo one thing, but do it right07:35
nrxtxhave a look at the suckless projects they are doing it very nice without autotools stuff07:35
openfbtdI could do it like they do, hm.07:36
nrxtxjust simple configurable makefiles07:36
tilmanconsider using a too07:37
RomsterUSR ?= /usr/local07:37
Romstermake USR=/usr07:37
Romsterif USR is unset uses /usr/local07:37
Romster? on ?= means you can override the variable on command line.07:37
Romsterand use install -d $(DESTDIR)$(USR)07:38
openfbtdAhem. How do I escape the $ sign07:56
openfbtd\$HOME becomes \OME lol07:57
Romstererr \\07:58
openfbtd-e "s%@CFGDIR_USER@%\\OME/.watchman%"07:59
openfbtd$$ works07:59
Romsteroh yeah forgot that...08:00
Romstermake is a little different.08:00
openfbtdSo, the first stage works08:01
openfbtdNow to think how to structure the thing08:01
cruxbot[core.git/3.0]: lzo: update to 2.08 (security related on 32-bit systems)08:22
*** SiFuh has quit IRC08:23
*** dkoby has joined #crux08:34
openfbtdmakefile with a config.mk08:49
openfbtdThis was easier than I thought08:49
openfbtd suggestions?08:55
tilmanye, call it Makefile09:10
openfbtdThat's the branch09:11
openfbtdAnd it doesn't exist anymore09:11
openfbtdIt's been merged into master09:11
openfbtdThe makefile is called Makefile09:11
tilmaninstall -m755 tools/* $(DESTDIR)/$(USR)/bin/09:21
tilmanimo it's preferable to define a list of stuff to copy/use/we instead of using globs09:21
openfbtdIn this case the glob is predictable09:22
tilmanno it's not09:22
openfbtdBut I do not disagree that's not optimal09:22
Romstertold ya openfbtd09:23
openfbtdtilman, pushed09:24
openfbtdinstall -m755 tools/watchman-{respawn,service,runas} $(DESTDIR)/$(USR)/bin/09:24
openfbtdAlso this will make writing a crux poert for watchman easier :309:25
openfbtdI'm still too lazy to set up CRUX 3.1 in a VM and fo that myself09:25
Romsterthe best way is a variable of programs and do install -m 0755 -t $(DESTDIR)/$(USR)/bin/ tools/$(TOOLS)09:25
*** SiFuh has joined #crux09:26
openfbtd-t is handy09:26
RomsterTOOLS="respawn service runas"09:26
Romsterbut that wont quite work the best way is a Makefile in each subdirectory wth a install: target09:26
Romsterand the main Makefile SUBDIRS="tools" and in install: call the $SUBDIRS to install.09:27
openfbtdI kinda like having one makefile09:28
Romsterbeen a while since i've done this tilman can help you there.09:28
Romsterthen it'll be just TOOLS="tools/respawn tools/service tools/runas"09:28
Romsterinstall -m 0755 -t $(DESTDIR)/$(USR)/bin/ $(TOOLS)09:28
tilmanwhile kill -0 "$job_id" 2>/dev/null; do sleep 1; done09:29
Romsteris the proper way but you can cheat with globs * but globs can bite you when you least expect it.09:29
tilmandoes that need a timeout?09:29
openfbtdtilman, yes, otherwise it runs kill too quickly and is a noticeable cpu hog09:30
Romsterthat could go into a endless loop?09:30
openfbtdRomster, in a race condition, possibly09:30
*** BitPuffin has joined #crux09:30
openfbtdIf the pid dies and immediately is taken by something else09:31
tilmani was thinking if there's a situation in which the loop wouldn't terminate09:31
openfbtdThen respawn will still think the service is running09:31
openfbtdBut pidfiles are a hack anyway09:31
Romsterpgrep for the process pid(s) kill those if $PID is not empty?09:31
openfbtdkill -0 is much better than pgrep09:31
Romsterah k09:32
openfbtdStarting from the fact that kill is a builtin09:32
openfbtdNo I meant that as in “Let's start from this point”09:33
openfbtdAnd end there because builtins are nice and you should use them when possible :D09:33
*** hotaronohanako has joined #crux09:35
*** BitPuffin has quit IRC09:36
*** Pingax has quit IRC09:36
openfbtd added a big boys section on manual installation09:54
openfbtdSo that a noob could use that to fuck up their system09:54
openfbtdNot really though, it won't overwrite anything09:54
openfbtdbtw this makes watchman easier to install on FreeBSD, mm!09:55
openfbtdThat's a plus09:55
openfbtdwatchman by default uses /run/watchman for pidfiles and such10:03
openfbtdAnd freebsd doesn't have /run10:03
openfbtdThat's literally the only incompatible thing10:03
openfbtdAnd now it's gone10:03
Romstereasy enought o use /var/run/10:03
Romsterenough to*10:04
openfbtdIt is a link to /run most of the time, yeah10:04
openfbtdBut I prefer to pint to /run specifically10:05
openfbtd/var could be not present on boot, for example10:05
openfbtd/run too, but less likely10:05
nrxtxif you are going after latest systemd changes /etc will be dumped too and generated by systemmd on demand after booting :D10:08
openfbtdThe reason watchman exists is because fuck systemd :D10:08
nrxtxwhat is watchman?10:08
openfbtdThe thing I was writing a makefile for for the past hour :)910:09
Romsterseriously they can't get rid of /etc was that a joke10:12
openfbtdsystemd is slowly eating everything into its own config files10:12
openfbtdLike you can already completely ditch /etc/fstab if you use systemd10:13
Romsterso now it's its own registry like gconf and dconf in gnome...10:13
openfbtdIt's not as convenient, but you still can do it10:13
Romsternow i have seen everything -_-10:13
Romsterbbs food10:14
nrxtxRomster: they want to create stateless systems, meaning different etc for different use cases just google for the announcment :D10:16
openfbtdSo they want to intagrate puppet/chef/etcd into systemd basically10:19
nrxtxyou have to ask this way: is there something they don't want to implement?10:20
tilmanprinter daemon10:21
tilmanthen again... now that cups is owned by apple maybe somebody should think about an alternative! ;)10:21
nrxtxdon't ever think about telling them :D10:22
openfbtdWhy did I not know about the ctraltdel utility10:24
*** BitPuffin has joined #crux10:32
*** Pingax has joined #crux10:35
*** BitPuffin has quit IRC10:36
Romsterwhat apple owns cups?11:11
openfbtdApple owned CUPS for a long time now11:11
openfbtdIn March 2002, Apple Inc. adopted CUPS as the printing system for Mac OS X 10.2.[4] In February 2007, Apple Inc. hired chief developer Michael Sweet and purchased the CUPS source code11:12
Romsterbut it's still available for linux right11:12
Romsterlike we re using it now11:12
Romsterso whats the big deal?11:12
Romsterit's like saying apache owns um apache :D11:12
*** hotarono` has joined #crux11:27
*** hotaronohanako has quit IRC11:29
*** BitPuffin has joined #crux11:33
*** hotarono` has quit IRC11:35
*** hotaronohanako has joined #crux11:35
*** BitPuffin has quit IRC11:38
*** Tezu has joined #crux11:43
*** Tezu has quit IRC11:44
*** doomicide has joined #crux11:44
*** heroux has quit IRC12:38
*** BitPuffin has joined #crux13:01
*** heroux has joined #crux13:22
cruxbot[xorg.git/3.0]: xorg-libpixman: updated to 0.32.613:54
*** hotaronohanako has quit IRC13:54
*** hotaronohanako has joined #crux13:54
openfbtdYay! It works on FreeBSD14:01
openfbtdYou need gmake to build it though14:01
*** jdolan has quit IRC14:06
*** jdolan has joined #crux14:26
*** diverse has quit IRC14:41
*** pidsley has quit IRC14:48
*** diverse has joined #crux14:48
*** jdolan has quit IRC14:48
*** dkoby has quit IRC14:55
*** pidsley has joined #crux14:58
*** SiFuh has quit IRC15:10
*** diverse has quit IRC15:10
*** diverse has joined #crux15:10
*** jdolan has joined #crux15:11
*** jdolan has quit IRC15:23
nweopenfbtd: are you using watchman on crux?15:27
*** jdolan_ has joined #crux15:27
openfbtdI have a WM that has an old version of watchman manually installed15:28
openfbtdWith CRUX 3.015:28
openfbtdI'm being lazy about writing at least an initial port of it for CRUX15:28
nweis it good?15:29
*** SiFuh has joined #crux15:30
openfbtdMy opinion would be highly subjective15:30
openfbtdI'm the author.15:30
openfbtdBut I personally like it15:30
nweoh :)15:31
*** jdolan_ has quit IRC15:31
openfbtdnwe, basically it's what I always wanted from a service manager. Simple, easy to use, easy to write scripts for15:38
openfbtdYou can look at the scripts here:
openfbtdWell. It's not as simple as just plain old standalone initscripts15:39
openfbtdBut those are a hell to mantain15:39
cruxbot[opt.git/3.0]: nvidia: updated to version 331.8915:44
cruxbot[compat-32.git/3.0]: nvidia-32: updated to version 331.8915:46
Romsteropenfbtd, can it be made non bash so it can work on say dash?15:46
*** jdolan has joined #crux15:47
Romsteri do like the directory approach where a Pkgfile can drop a service file in place.15:47
openfbtdRomster, you could for ik and strip all bashisms15:52
openfbtdI can't be bothered15:52
openfbtdfork it and*15:53
openfbtdOne of the reasons I use bash is the ammount of useful builtins it has15:53
openfbtdWith calls to external binaries all the time watchman would become slower and less portable.15:54
openfbtdI don't see any problems with treating bash as a dependency to watchman.15:54
*** doomicide has quit IRC15:54
openfbtdYep, dash doesn't even have kill as a builtin :315:56
openfbtdAnd those handy [[]]15:56
diverseso the question is, how minimal should one go?15:59
openfbtdAs minimal as you want to while being comfortable with it?16:01
openfbtdOr as complex16:01
openfbtdIt's your system16:01
*** doomicide has joined #crux16:03
openfbtd.w 1816:05
diverseRomster: either dash with fewer features or bash with more features. :P16:05
Romsterdash is considerably faster than bash, and closer without as many bashisms.16:08
openfbtdIt's more like either bash and me being comfortable with the code or OH GOD NOTHING I'M USED TO WORKS16:09
Romsterwould prefer not shell at all tbh for init.16:15
openfbtdIt's not init16:16
openfbtdIt's a service manager16:16
openfbtdI use for my init16:16
openfbtdAnd watchman is not a daemon16:17
Romsterso you separated the two parts16:17
openfbtdThey were never one16:17
openfbtdI had a bash init16:17
Romsterso it's not a watchdog timer16:17
openfbtdBut it was not a part of watchman16:17
Romsterwhy dothat when sinit can do services16:17
openfbtdsinit can't do services. It forks exactly one thing.16:18
openfbtdAnd that thing is /bin/rc for me16:18
openfbtdWhich then sets up the system and starts the services using watchman16:18
openfbtdAnd watchman _has_ a watchdog16:18
Romsterinstead of the crux rc.d/16:18
openfbtdYou can choose not to use it16:18
Romsterso you basically redid rc.d with sinit and watchman16:19
*** deus_ex has joined #crux16:19
Romsterfair enough. still better than systemd way16:19
openfbtdI redid rc.d with rc and watchman. init is irrelevant to the setup16:19
openfbtdYou can use any init with this16:20
openfbtdEven systemd if you're that weird16:20
Romsterewww just eww16:20
openfbtdI would ask why16:20
openfbtdBut hell, go ahead16:20
openfbtdOne of my friends uses watchman _with_ systemd16:20
Romsteroh right your still using svinit16:20
Romsteri thought sinit replaced sysvinit16:21
openfbtdsinit is not sysvinit at all.16:21
openfbtdsinit does replace your PID 116:21
Romsterand sinit does /sbin/init16:21
openfbtdsinit does /bin/rc16:21
openfbtdWhich is, in my case, a bash script :)16:21
openfbtdBut it too just starts everything and dies16:22
Romsterwhich then starts watchman16:22
openfbtdWhich then uses watchman to start services, yes16:22
Romsterso uh parallel services starting? dependency resolution with services?16:22
openfbtdwatchman can do rudimentary dependencies16:23
Romsterto start in correct order16:23
openfbtdi.e. “this service depends on these, let me start them first”16:23
openfbtdAnd my rc can start services in parallel16:23
Romstermaybe even delay start stuff like crond while giving the login prompt already16:24
openfbtdExactly like the old arch rc used to work16:24
openfbtdBut with dependencies and an optional service watchdog16:24
Romsterthough then startx you'd have to make sure dbus was running by that point at least.16:24
Romsteryou basic did a basic openrc16:25
openfbtdNot even that basic16:25
Romsternow make it work with lvm2 cryptosetup and mdadm and your cooking16:25
openfbtdThat's just a matter of setting up the proper scripts.16:26
openfbtdIn proper order and/or with proper deps16:26
openfbtdJust so you understand:16:27
openfbtd1) init: sinit16:27
openfbtd2) rc:
openfbtd3) service manager:
openfbtdThree different projects16:28
openfbtdNo relation whatsoever.16:28
openfbtdExcept for the fact that I wrote the last two16:28
Romsteryou just have to watch for race conditions16:32
Romsterand test it when a FS is dirty and fsck is forced.16:32
openfbtdOnly if I watch pids from pidfiles, yes16:32
openfbtdAnd I do >_>16:32
Romstercan't get to pids in /proc?16:32
openfbtdI can, but I need to save the service or watchdog pids between watchman runs16:33
openfbtdWhen watchman spawns, it knows nothing about no one16:33
openfbtdIt needs to know what to look for16:33
openfbtdIt's not a daemon that can just sit there and know of the children it spawned because they its children.16:34
openfbtdIt dies after executing whatever was asked of it16:34
openfbtdthe are16:34
openfbtdI am bad at typing. they are its chldren*16:35
openfbtdI'm looking at cgroups, but not yet.16:35
openfbtdAnd there are not a lot of scenarios where a service dies and its pid is IMMEDIATELY taken16:36
openfbtdThere could be that a service died quietly, a new process took its pid and is now sitting on it. So you try to restart the service and end up killing the new process instead16:37
openfbtdBut the probability of that is not very high.16:37
Romsterhrmm still exists though16:38
openfbtdEspecially if you spawn most of your services under the provided watchdog16:38
openfbtdWhich lowers that possibility to nearly zero16:38
Romsterand slight chance even if it's 0.01% is still a chance.16:38
openfbtdYeah, but the only solution is to always be up and knowing your children16:39
openfbtdAnd I don't want that16:39
openfbtdAnd the default crux scripts are certainly not addressing that particular issue at all16:40
openfbtdEven runit the Internet seems to drool over lately is vulnerable to that16:40
openfbtdNot every service you can fork() and watch16:41
openfbtdYeah, sure, it talks to its instancec via a fifo, but you can still lose a watchdog. The fifo can be deleted. The watchdog can die itself16:41
openfbtdI don't understand why people think that runit is the magic pill16:42
openfbtdIf anything, systemd is16:44
openfbtdBut it comes with so much negative stuff...16:45
Romsterbinary log files and everything else16:45
*** pidsley has quit IRC16:46
openfbtdIn any case, _every_ service in my systems is running under watchman-respawn. Which is a very simple script and has very little chances of falling on its head for some reason16:48
openfbtd(which is why UNIX way exists — the smaller the pieces, the smaller the attack surface)16:48
openfbtdAnd those run continiously. So the possibility of something taking over one of their pids is very small.16:49
*** pidsley has joined #crux16:49
*** doomicide has quit IRC17:09
*** pidsley has quit IRC17:12
Romsteryeah unlike systemd bug, wait for it an entire machine rooted in the matter of seconds a massive botnet of linux boxes.17:13
Romsterall thanks to systemd17:13
Romsterno more diversity everyone on systemd17:13
*** jdolan has quit IRC17:13
*** hotaronohanako has quit IRC17:15
*** hotaronohanako has joined #crux17:15
*** hotaronohanako has quit IRC17:20
*** hotaronohanako has joined #crux17:20
*** hotarono` has joined #crux17:23
*** hotarono` has quit IRC17:23
*** hotaronohanako has quit IRC17:25
*** pidsley has joined #crux17:26
*** pidsley has quit IRC17:29
*** hotaronohanako has joined #crux17:33
*** jdolan has joined #crux17:40
*** jdolan has quit IRC17:50
*** jdolan has joined #crux18:00
*** sh4rm4 has joined #crux18:14
*** sh4rm4 has quit IRC18:16
*** sh4rm4 has joined #crux18:17
*** sh4rm4 has quit IRC18:19
*** sh4rm4 has joined #crux18:22
*** sh4rm4 has quit IRC18:24
*** pidsley has joined #crux18:27
*** sh4rm4 has joined #crux18:27
*** pidsley has quit IRC18:29
openfbtdok systemd now has a DHCP server18:30
openfbtdThis is getting silly18:30
*** pidsley has joined #crux18:34
*** timcowchip has joined #crux18:39
timcowchipundefined reference to `yylex' trying to install tilda18:41
openfbtdI wonder when will they implement an sshd18:42
openfbtdI don't have any doubt they will18:42
openfbtdI wonder when.18:42
*** Pingax has quit IRC18:43
*** Pingax has joined #crux18:52
*** Pingax has quit IRC18:52
*** Pingax has joined #crux18:52
*** Pingax has quit IRC19:06
TLHa similar tool19:11
openfbtdI know other tools exist, yeah19:17
openfbtdAnd I knew about that one19:17
openfbtdBut depending on ubase or any other respawner in a simple bash project...19:17
openfbtdI don't think so.19:17
*** timcowchip has quit IRC19:18
openfbtdEspecially taking into account the fact that this is really easy to implement19:18
openfbtdon ubase or any other external projecy*19:19
openfbtdTLH, but you know what.19:19
openfbtdYou can use any respawner with watchman19:20
openfbtdJust do service_command='respawn'; service_args='command'19:20
openfbtdAnd I do have ubase installed19:21
TLHopenfbtd: yeah initially I had a respawn script as well19:22
TLHbecause I use respawn from ubase in morpheus (to restart getty) I thought I'd just put in ubase rather than carry it along in my fs package19:23
openfbtdWell, that makes sense19:23
TLHthe fifo option is to trigger it externally (think of monitoring sshd via nc or similar and doing a dummy echo)19:24
openfbtdYEah, I get it19:24
openfbtdrunit's supervise does the same19:25
TLHa nice19:25
openfbtdTLH, respawn from ubase doesn't clean up19:26
TLHhow u mean?19:26
openfbtdIf it's killed with TERM, it just dies19:26
openfbtdDoesn't kill the service it's supervising19:26
openfbtdThat's a different way to deal with it, but I don't like it19:27
openfbtdI like the fact that my watchdog kills the server if it's kileld19:27
openfbtdIf you want it not to do that, you can -9 it19:27
TLHnot sure what the proper thing to do here is but I can imagine people killing respawn to avoid respawning without necessarily disrupting the watched service.19:27
openfbtdBut ideally if I say STOP, I want you to stop. Everything.19:28
openfbtdAnd if I say DIE, you just die19:28
TLHyeah it probably makes sense19:28
openfbtdI'm not saying it's the only proper way19:28
openfbtdIt's just that I want it to behave like that19:29
TLHno I see your reasoning, you can always send SIGKILL if you want to achieve what I mentioned earlier19:29
TLHon the other hand19:29
openfbtdIt makes stopping services behind a supervisor really transparent in the context of watchman's older architecture.19:29
TLHfor processes that are reparented there's nothing you can do (well there is but let's not go there right now)19:29
TLHso no killing the child for SIGTERM is consistent with the boundary cases19:30
openfbtdI could go full systemd and adopt cgroups19:30
openfbtdIf it's reparented, I treat that as a fringe case19:30
openfbtdAnd just deal with it manually19:30
openfbtdNever encountered that though19:30
TLHthere's also prctl(2) but I don't want to go there.19:30
openfbtdIf a service doesn't immediately fork into a daemon and escape, it reliably stays there19:31
openfbtdAnd most services have an option not to fork19:31
openfbtdAnd if it does, there's almost always the option to write a pidfile and track it that way19:31
openfbtdAnd if there's no such option, YOUR SERVICE SUCKS AND YOU SHOULD BE SHOT19:32
openfbtdI'm looking at you, cjdns.19:32
openfbtdmr. “I quietly fixed the no-daemonize option without telling anyone after saying I will remove the feature completely”19:32
TLHtha ad.19:35
TLHthat's bad*19:35
openfbtdAnd it still doesn't work properly19:36
openfbtdIf you SIGTERM the master process, it just dies19:36
openfbtdDoesn't kill its children19:36
openfbtdI SAID STOP DAMMIT19:36
openfbtdIn the end I have this:
openfbtdIt finishes off the angel19:37
openfbtdThat kills the whole thing19:37
TLHright I see19:53
openfbtdThat's why I love the idea of initscripts still being scripts19:55
openfbtdBecause then you can be real flexible with em19:55
openfbtdEven if the basic ones are just declarative configs19:55
openfbtdEspecially if that19:55
*** sh[4]rm4 has joined #crux20:10
*** sh4rm4 has quit IRC20:11
*** mike_k has quit IRC20:15
*** hotaronohanako has quit IRC20:24
*** timcowchip has joined #crux20:48
*** frinnst has joined #crux20:54
*** jdolan has quit IRC21:02
*** jdolan has joined #crux21:15
*** timcowchip has quit IRC21:18
*** pidsley has quit IRC21:59
*** pidsley has joined #crux22:13
*** timcowchip has joined #crux22:41
timcowchipdoes someone have a link to the 3.1 ports?22:41
timcowchipthe future ain't what it used to be22:48
openfbtdda future iz nao!22:49
*** SiFuh has quit IRC22:52
teK__depends on the repo..22:59
timcowchipthanks tek23:00
*** timcowchip has quit IRC23:01
teK__off to bed^Wrepeat course work23:03
q5sysjaeger: Mind a PM?  (regarding a matter I believe Frinnst has already spoken with you about)23:44

Generated by 2.11.0 by Marius Gedminas - find it at!