libera/#devuan/ Wednesday, 2020-10-28

systemdlete2#join /starlinux00:23
systemdlete2sorry, got that backwards00:23
systemdlete2anyway...00:24
systemdlete2I thought I'd try out star linux just for the heck00:24
systemdlete2Installed OK, pretty much the same procedure as devuan.  But I can't get the dovecot server to work.00:25
systemdlete2It complains it does not know about "imaps" --> I added that to a file in /usr/share (why is a config normally found in /etc suddenly in /usr/share??)00:25
systemdlete2(all the rest of dovecot config is in /etc)00:26
systemdlete2star's dovecot is at 2.3.4.100:27
systemdlete2interesting.  Crowz is another distro the same 2 people built, also based on devuan, but this time, beowulf specifically00:29
systemdlete2star linux got a decent write-up on distrowatch, and it seems to be pretty good overall.00:30
systemdlete2star dovecot is the same as beowulf.   So it should work...00:37
fsmithrednormally, stuff is /usr/share is the default and any changes you make go in corresponding files in /etc00:37
systemdlete2agreed.00:37
fsmithrednot talking about dovecot, just in general00:37
systemdlete2again, agreed.00:37
systemdlete2maybe I should saunter on over to something like #dovecot and ask00:38
fsmithredyou installed the dovecot imap package?00:41
fsmithreddovecot-imapd00:41
systemdlete2how else could it be failing if I didn't have it installed?00:47
systemdlete2no worries.  I am inquring at #dovecot.  I think I see part of the problem already.00:48
systemdlete2There is no "imaps" protocol, though there are places in the config where imaps must be specified.00:48
neppuHiya, need help with OpenRC. Service still runs as root even though I did specify command_user.04:50
masonneppu: How are you quoting the command_user argument, and is it just a plain string, not a variable? I still can't find anything indicating its breaking on folks.04:57
neppu#!/sbin/openrc-run04:58
neppucommand="/usr/bin/pulseaudio"04:58
masonneppu: https://wiki.alpinelinux.org/wiki/Writing_Init_Scripts makes it look fairly straightforward. Something else I'd try is to vary the user.04:58
neppucommand_args="--start --use-pid-file=true"command_user="sample-user"04:58
neppu^ newline missing.04:59
masonAlright, so, that ought to just work. Hrm.04:59
neppuI have no idea why it doesn't work, ugh.04:59
masonAnd /sbin/openrc-run is a binary, right? So we can't just read it to see what it does.05:00
mason(I don't have an OpenRC box around to check, but my understanding is that most of its guts are binaries.)05:00
neppuIt is yeah.05:00
masonneppu: Maybe you'll see something in https://github.com/OpenRC/openrc/blob/master/src/rc/openrc-run.c I'm missing.05:03
masonOh, it's here: https://github.com/OpenRC/openrc/blob/07908be0903229a69b9e0f733ed13eeff0b55a44/sh/start-stop-daemon.sh05:07
masonneppu: I'd see if you've got a start-stop-daemon.sh kicking around, and see if you can maybe shim a "set -x" at the top and observe what it does when you start your service.05:10
masonAlternately, if it runs under their fake-bash and that doesn't work for some reason (dunno) you can probably dump env out to a tempfile and see if it's actually got your command_user, etc. The set -x will show you how it's trying to invoke start-stop-daemon.05:10
neppuI'm pretty sure it's a binary too...05:12
masonneppu: I don't think --user works like they think it does. Read the man page for start-stop-daemon. I think they really want -c.05:12
masonneppu: So, I'd look for that start-stop-daemon.sh and change this line: ${command_user+--user} $command_user \05:13
masonInstead of --user I'm pretty sure you want --chuid to implement what they say they want.05:13
neppu!05:14
neppuCHUID ACTUALLY WORKS05:14
masonneppu: Right.05:14
neppuHoly, you're a lifesaver, man.05:14
masonneppu: Next up, file a bug against openrc-run saying "I don't think --user works like you think it does."05:14
masonneppu: A lifesaver? I've just enabled you to use PulseAudio. That's not something to feel good about. :P05:14
neppuI mean, yeah.05:15
neppuAs I've already said, I can't completely get away from Poettering bullshit because I'm just too used to PA.05:15
masonTo each his own. I'm mostly content with ALSA here. :P05:15
masonAlright, it's past my bedtime and I'll turn into a pumpkin soon. Sorry I didn't dig into this earlier.05:16
neppuIt's fine. Night.05:16
masono/05:16
neppuUgh, why can't this thing just work?05:42
neppuDamn PA.05:42
danuananyone know how to start nfs-kernel-server without portmapper running , trying to do nfsv4 shares only08:27
clortit used to run fine without portmapper08:28
clortbefore portmapper08:28
gnarfacei'm pretty sure the only secret is to uninstall portmap08:31
gnarfaceit is included by default for convenience because a bunch of that old Sun era network stuff uses it for peer discovery08:32
gnarfaceobviously it's hugely insecure but they don't expect you to care if you're using nfs08:32
danuani get  (Starting NFS kernel daemon: nfsd  Not starting: portmapper is not running ... (warning).)  when i manualy stop portmap08:33
clortthe low-effort file-sharing for me is sshfs08:33
clortall the access control is already there, and it doesn't barf on disconnect08:34
danuangnarface you think uninstalling would change that message ?08:34
danuanclort , i have root on nfs stuff , and crossmounting inside shares , file locking  locking etc.. ,   sshfs might not cut it08:42
clortinteresting ok08:43
clortnice to have options08:44
clorton the NeXT nfs didn't lock-up my life on disconnect.  thanks linux.08:44
danuanyes , sometimes i think smbfs handles things way more gracefully then nfs, when things like reboots and errors of unable to unmout remount etc go08:44
clortprobably someway to get around that though08:45
danuangoing to try and comment out the test for rpcbind  in init.d/nfs-kernel-server08:46
danuannope  (Starting NFS kernel daemon: nfsdrpc.nfsd: writing fd to kernel failed: errno 111 (Connection refused) \ rpc.nfsd: unable to set any sockets for nfsd)08:48
clortlooks like some maintainer started assuming portmapper08:53
alv_...09:05
golinuxalv_: You're back?09:11
alv_yes i'm here !!!!09:11
danuanclort , the funny thing is , after you start nfs-kernel-server  you can kill portmap and all  rpc.stuff   , and you can mount and unmount all shares without problems untill the next reboot of the server09:23
clortsounds like you are sysadminning danuan :)09:28
clortit's been decades since i did that09:29
gnarfacedanuan: sorry, i did indeed think uninstalling it would change that message, but i didn't think just having it stopped but installed would cause any errors either.  i must be forgetting something.09:31
danuanand cat /proc/fs/nfsd/versions always reports -2 +3 +4 +4.1 +4.2  , no matter if i try to disable things in /etc/defaults/nfs-kernel-server09:34
gnarfacehmm, you might actually have to disable nfs versions by altering which modules actually load, or at least the options they load with09:47
gnarfacelast time i messed with it i had to rebuild the kernel to make it do what i wanted, but that was back in the sarge -> etch era and i also had to support OS X clients09:48
gnarfacei think i might have been trying to force it to v3 not v409:48
gnarfacenot sure09:48
gnarfacebut you should check if there's an /etc/defaults/portmap or /etc/defaults/rpc or something like that09:48
gnarfacelook for a line like disable=0 and change it to disable=109:49
gnarfacemaybe if they made a bunch of dumb script checks for the presence of portmap they also built in a override09:49
danuanneather of those files exist on my sys09:49
gnarfacehmmm09:53
gnarfacethey could have changed it but it's an odd move considering their recent behavior patterns09:53
gnarface(usually they just amputate stuff like this because "nobody under 18 is using it anymore")09:54
rrqdoesn't "man nfsmount" tell how to do it?10:01
danuanrrq , problems is trying to disable portmap and still have nfs-kernel-server start10:05
rrqwell "nfs" includes 2 processes and they use portmap to learn how to talk to each other10:05
rrqyou'll have to fix their ports in their configurations10:06
rrqrpc.nfsd and rpc.mountd10:06
rrqman ...10:07
rrqI believe it's possible, but I believe many things10:07
danuanevebut i am able to kill portmap and all rpc services after nfs-kernel-starts   and things run just fine , and i just assumed based on information online that nfsv4 does not need it10:09
rrqyes, once nfsd has larnt mountd's ports it keeps using them10:10
rrqlearnt10:10
rrqso "just" tell them to always use those ports through their configurations10:12
dynebotHi all !!!!10:13
alv_hummm....10:14
dynebot<alv> ....10:17
alv_ok that's better10:18
alv_:D10:18
danuanrrq  which configuration?  there is only that i found /etc/default/nfs-kernel-server and it seems  options supplied on RPCMOUNTDOPTS=  line for setting --port and  --no-nfs-version  but they had no effect10:24
rrqwell, "man rpc.mountd" seems to suggest that you'd declare its port in /etc/services ... it was a while since I did this10:29
rrqmaybe enough to add eg "--port=2050" for RPCMOUNTDOPTS in /etc/default/nfs-kernel-server10:38
danuanagain trying to run just rpc.nfsd for nfsv4 without anything else which i thought was possible  , and starting rpc.nfsd with -p or --port=2050  gives same results10:39
danuanrpc.nfsd: writing fd to kernel failed: errno 111 (Connection refused) \ rpc.nfsd: unable to set any sockets for nfsd10:40
rrqmmm maybe needs entries "rpc.mountd 250/tcp" and "../udp" in /etc/services as well10:42
rrq205010:42
danuanbut no other rpc services again , rpc.mountd is not needed eather for v4 as i understand10:43
rrqmmm too long ago :)10:44
danuani can live with this , but would be nice to have minimal working system with least open port or services running , like having smbd without nmbd10:47
gnarfaceit's not just a permissions thing is it?10:50
gnarfacemaybe the issue is nfs-kernel-server10:51
gnarfacei seem to remember not needing it, too10:51
gnarfacebut i was just doing static mounts, maybe you are doing something different10:52
danuanseems like portmap opens some socket that nfs-kernel-server grabs  , and after it starts portmap can be killed10:52
gnarfacejust out of curiosity, did you check whether it behaves differently when portmap isn't installed?10:54
gnarfacenot just stopped, but not installed?10:54
gnarfaceif it was an actual requirement i would expect it to not let you uninstall it without also uninstalling nfs-kernel-server10:55
gnarfaceif there's some hidden sneaky check that's the easiest way to verify10:55
gnarfaceit shouldn't be a problem to just reinstall it afterwards if you don't have any luck10:55
danuanok , just got it started  with rpc.nfsd -N 3 , but now clients complain about having to use -o nolock10:57
gnarfacehah10:57
gnarfacewell that may be expected behavior10:58
danuanso /etc/default/nfs-kernel-server option were not being honored at all for some reason10:58
gnarfacedoes seem like a bug10:58
dynebot<alv> ...11:02
alv_...11:03
jibbilyhas anyone else had an issue with Sublime Text going back to unregistered after every reboot?12:29
gnarfacejibbily: is that something from the devuan repos?12:32
gnarfacejibbily: i dunno anything about Sublime Text but fyi 3rd party software is notorious for having wrong paths or environment variables12:34
jibbilyno, but I just found some IRC logs that suggests it's to do with the machine-id being changed12:34
gnarfaceah, well if it's that here's the deal12:34
gnarfacemachine-id in concept is massively flawed and insecure12:34
gnarfacethere should be a way to disable the randomization12:35
gnarfacebut it randomizes by default as a security patch12:35
gnarfaceand you're strongly advised to neither do this nor use software that relies on it12:35
fsmithrededit /etc/default/dbus to get a static machine-id12:36
jibbilyis it a syscall or file? it'd be good if I can just give sublime a fixed one, and keep it random for everything else12:36
gnarface(and not just because this is insecure enough on its own to be used by bad actors, but because anyone who would do this to you is probably about to do something worse next)12:36
fsmithredthe actual file is /var/lib/dbus/machine-id12:37
jibbilyis there a way to get sublime to see /var/lib/dbus/machine-id as a different file?12:39
gnarfacei have no idea12:39
fsmithredme neither, but that's a good idea12:39
jibbilyI know someone who'll know, will update you12:39
gnarfaceif it's not dumb it should have a way to give a different path12:39
gnarfacethe machine-id file is a plain text file that is easy to fabricate12:40
fsmithredthere's also /etc/machine-id in systemd12:41
fsmithredmake sure it's not looking for that12:41
jibbilyso I just hex edited /var/lib/dbus/machine-id to a different path and I think it's staying registered now12:59
jibbily(in the sublime binary)12:59
fsmithredcool12:59
jibbilymy colleague suggests using nsenter + mount13:04
WafficusHi there, how do I change the timezone within Devuan? I forget. I think its like systemctl (something) but wanted to confirm15:16
kizanoWafficus, dpkg-reconfigure tzdata15:17
kizanoeven on systemd host15:17
kizanobut devuan is all about not having systemd...15:18
r3bootno, it's about init freedom, as long as it's not systemd15:19
kizano:D15:20
Wafficusthanks kizano15:21
Wafficusthat worked15:22
Wafficushave a good one for now, thanks a ton15:22
dynebot<alv> :D15:56
nemor3boot: heh.  I guess one could be some kind of ironic hipster and add back in systemd components to devuan16:11
nemor3boot: still init freedom16:11
r3bootnemo: I guess :) Imho it's a contradiction in devuan (as mentioned on their frontpage), but I admit thats wordplay16:14
GyrosGeiersystemd is a service manager, not an init system16:19
GyrosGeierthat you can abuse it as an init system is coincidence16:19
GyrosGeierthe design goal is different, the use cases are different16:20
fsmithreddevuan with systemd is available here: https://cdimage.debian.org/debian-cd/current/16:21
GyrosGeierthe boundary between init system and service manager is pretty much between the special unit types for early boot, like file systems and so on, and "service" units that assume that the rest of the system is set uo16:22
GyrosGeier-o+p16:22
n4dirxfce4; settings; session and startup i remove from "application autostart": cadence, geoclue Demo agent, PulseAudio, Screensaver and folders update.16:26
n4dirsounds reasonable or should i better add one of that again?16:26
fsmithredyeah, that sounds ok16:34
n4dirgood, thanks.16:34
fsmithredI'm not sure what folders update does16:35
fsmithredgotta go. bbl.16:35
n4diras i don't use a filemanager, i guess i won't run in it16:35
n4dirseems to create that xdg-user-dirs stuff. Lots of new folders in my home. bye bye16:35
r3bootGyrosGeier: potato potato ;) But, offtopic here!16:40
masonGyrosGeier: systemd isn't a service manager. It's a bootloader. Wait, it's a logging system. Wait, it's a resolver. Wait, it's a container management system.18:05
n4diri mean: if i want emacs i install emacs.18:06
nemoash systemd durbatulûk, ash systemd gimbatul, ash systemd thrakatulûk, agh burzum-ishi krimpatul18:06
kizanoLoL mason18:18
kizano#scopeCreep18:18
DHEthere's a well known animated gif that's like 5 years old now that sums it up18:28
observer_test18:31
clortanimated pngs are where it's at18:41

Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!