| elias_a_ | Hello! Another tired user of Lennart's cookings here. I will migrate my laptop from Debian Bookworm to Daedalus. | 00:05 |
|---|---|---|
| rwp | Welcome elias_a_ to the true path of Devuan! | 00:09 |
| elias_a_ | rwp: TY :D | 00:13 |
| Xenguy | rwp, pm-hibernate does indeed work on the X220 here : -) | 00:30 |
| rwp | Very good! And glad to hear you made it through your work day. (I am still in gdb debugging here.) | 00:30 |
| Xenguy | rwp, The X250 on the other hand just locked up ; -) | 00:32 |
| rwp | Hmm... :-( | 00:33 |
| Xenguy | restarting it now | 00:33 |
| rwp | I just verified pm-hibernate on daedalus on my x270 works okay. | 00:38 |
| ballsystemlord | Hello, On Sep 20th I sent a message to DNG ( dng@lists.dyne.org ). Then realized I was, for some reason, not subscribed anymore. I resubscribed and sent two messages on and 28th , but never got any replies. | 01:08 |
| ballsystemlord | I tried to access the archives, but I can't get to them. The page redirects back on itself. | 01:08 |
| ballsystemlord | https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng | 01:09 |
| ballsystemlord | I wanted to ask if anyone received my emails from DNG. | 01:09 |
| ballsystemlord | Just to see if things are working. | 01:09 |
| gnarface | ballsystemlord: stick around, i'm sure someone knows what's happening... | 01:10 |
| ballsystemlord | They are titled "Claws-mail stuck. Bug report to Devuan devs." and "[DNG Re (2): unresponsive machine" and "Re: [DNG Re (2): unresponsive machine" | 01:10 |
| ballsystemlord | Thanks! | 01:10 |
| ballsystemlord | I'll wait for a reply. | 01:10 |
| golinux | Hi ballsystemlord . . . try this url https://lists.dyne.org/lurker/mindex/dng@20380101.000000.00000000.en.html | 01:11 |
| golinux | Someone reported that the link on www was outdated so maybe that's the one that you tried. | 01:13 |
| ballsystemlord | My messages made it to the list! I guess they just didn't provoke a reply. I thought at least a "thank you" was in order for the last two, but it doesn't really matter. | 01:15 |
| Xenguy | golinux, The only people than can fix the 'page linking to itself' are Dyne | 01:18 |
| gnarface | Xenguy: 2-step checklist for getting hibernate working on any hardware: 1) make sure your swap partition is at least as large as the amount of physical system ram 2) regenerate your initrd.img and make sure the partition UUID it says it will attempt resume from matches the right swap partition (if not, you can set a variable in the config somewhere to override it then regenerate the initrd again) | 01:18 |
| Xenguy | golinux, Meanwhile I'm working on adding the real archives link to our web site, in lieu of help from Dyne et al. | 01:19 |
| ballsystemlord | Xenguy: thanks! | 01:20 |
| ballsystemlord | See you! | 01:20 |
| golinux | I thought you pushed that change a while ago | 01:20 |
| gnarface | i guess there's technically a 3rd step but it may only matter for very old hardware, there's a bios setting for something like "memory hole at 15MB" or 15KB or something like that, ... been so many bongs ago i don't remember if it needs to be on or off, but the setting does matter, and on such hardware it will only work in the right state | 01:20 |
| Xenguy | golinux, That was a minor gremlin update that updated a reference from 'freenode' to 'libera' | 01:21 |
| Xenguy | (we should take it to #-www I suppose) | 01:22 |
| Xenguy | gnarface, I think maybe that might be the problem, i.e. #1... | 01:24 |
| Xenguy | My RAM is ~4Gb, but so is my swap | 01:24 |
| Xenguy | Dang | 01:24 |
| Xenguy | No wait, that's incorrect | 01:24 |
| gnarface | check the size in bytes; rounding errors can screw you over on this | 01:25 |
| Xenguy | Hibernate works on this laptop, but specs are as above, hrm | 01:25 |
| Xenguy | yeah, understood | 01:25 |
| Xenguy | Is there a good way to ascertain swap file size (I tried: fdisk -l) | 01:25 |
| Xenguy | For RAM I can use 'free' | 01:26 |
| gnarface | swap space also shows in free for me | 01:26 |
| Xenguy | oh, hrm | 01:26 |
| Xenguy | Oh you're right, perfect | 01:26 |
| gnarface | oh, i always use swap partitions though... it might matter that it's a partition and not a swap file | 01:27 |
| Xenguy | So yeah on this laptop where hibernate works, swap is slightly larger than RAM; let's check the other laptop | 01:27 |
| gnarface | some time ago i heard Linux gained the ability to use a dynamically scaling file as swap space from within another regular filesystem partition, the way Windows does it, but i never tested it and i don't know what sort of implications it might have for hibernation functionality | 01:28 |
| Xenguy | gnarface, Well the ratio between RAM and swap is similar but the newer laptop has double of both, and it's the one where hibernate doesn't work | 01:30 |
| Xenguy | So, still a mystery I suppose | 01:30 |
| gnarface | off the top of my head, i don't know how to check list entry #2 without regenerating the initramfs | 01:31 |
| rwp | cat /etc/initramfs-tools/conf.d/resume | 01:31 |
| gnarface | but it's possible if the swap partition's UUID changed since install, it could be wrong | 01:31 |
| gnarface | the bios setting should be easy to check thoguh | 01:31 |
| gnarface | *though | 01:31 |
| gnarface | rwp: afaik that is only set if you manually set it though, it has a default that never shows up in the config files unless you do | 01:32 |
| gnarface | (and the default i think is just whatever swap partition it finds first when the initramfs is generated) | 01:32 |
| rwp | That's set if one is using a swap partition. Xenguy said using LVM in the normal way. So it should be set to /dev/mapper/$VG-swap with $VG-swap being whatever is the swap partition. | 01:33 |
| gnarface | hmmm... | 01:33 |
| * gnarface also isn't using LVM | 01:33 | |
| * Xenguy knows jack about initramfs ... | 01:34 | |
| rwp | "swapon -s" will list the swap device which will be /dev/dm-1 for me and ls -l /dev/mapper/ will show (for me) v4-swap -> ../dm-1 there | 01:35 |
| rwp | LVM uses UUIDs internally so using LVM LV names effectively use UUIDs too. | 01:35 |
| Xenguy | ^^ Do I need to run that as root? | 01:35 |
| * adhoc prefers to set the swap device to a floppy | 01:35 | |
| adhoc | or tape | 01:35 |
| rwp | If it gives an error as non-root then yes you will need to run it as root. | 01:36 |
| Xenguy | fair enough | 01:36 |
| adhoc | that way you know when your machine has run out of ram and you need to go kill some crap | 01:36 |
| rwp | Works for me as non-root. But I did run it here as root. | 01:36 |
| Xenguy | Doesn't work as root here | 01:37 |
| Xenguy | Let's see if I get /kicked | 01:37 |
| Xenguy | # swapon -s | 01:37 |
| Xenguy | Filename Type Size Used Priority | 01:37 |
| rwp | Reading the man page it says we are supposed to use "swapon --show" now. Who knew? | 01:37 |
| Xenguy | /dev/dm-2 partition 4075516 2101880 -2 | 01:38 |
| rwp | So dm-2 and then look at "ls -l /dev/mapper/" to see what is pointing to dm-2 | 01:39 |
| Xenguy | # swapon --show | 01:39 |
| Xenguy | NAME TYPE SIZE USED PRIO | 01:39 |
| Xenguy | /dev/dm-2 partition 3.9G 2G -2 | 01:39 |
| Xenguy | rwp, https://paste.debian.net/hidden/f1ba52da/ | 01:41 |
| rwp | The line is "devuan--vg-swap_1 -> ../dm-2" so /dev/mapper/devuan--vg-swap_1 | 01:41 |
| Xenguy | I have no clue about this stuff | 01:41 |
| adhoc | is this thing a VM you are putting swap on? | 01:42 |
| Xenguy | This is a bare metal laptop | 01:42 |
| adhoc | ok | 01:42 |
| rwp | If you run "lvs" it will list out the LVs logical volumes and you will see "swap devuan-vg" listed there. But a single - is escaped to -- through that so devuan--vg is what it becomes. | 01:42 |
| rwp | cat /etc/initramfs-tools/conf.d/resume | 01:42 |
| Xenguy | Again no clue | 01:42 |
| adhoc | why mess with LVs and VGs if you are not using extenable storage in a laptop ? | 01:43 |
| rwp | Mine says: RESUME=/dev/mapper/v4-swap | 01:43 |
| Xenguy | # cat /etc/initramfs-tools/conf.d/resume | 01:43 |
| Xenguy | RESUME=/dev/mapper/devuan--vg-swap_1 | 01:43 |
| rwp | Because I don't like the long dashed names and renamed devuan-vg into v4 on my machine. | 01:43 |
| adhoc | that is a lot of over head in block mapping | 01:43 |
| rwp | Xenguy, That's perfect! All good. | 01:43 |
| Xenguy | Meaning what? | 01:43 |
| Xenguy | Oh wait, crap, I'm giving you the info from the laptop that works, hah | 01:44 |
| Xenguy | Gah | 01:44 |
| rwp | adhoc, The strategy is to use exactly one LVM PV physical volume encrypted with LUKS and that way there is one passphrase query at boot time. Then that results in the ability to have everything encrypted. The LVs then can have root and swap and both are encrypted with one passphrase. | 01:44 |
| Xenguy | Okay, taking a break for a bite to eat, and then will try that sequence again with the laptop that *doesn't* work | 01:45 |
| rwp | Xenguy, You now have the working A case. Look at the non-working B case. The A-B comparison will hopefully show something not configured to work on the B case. | 01:45 |
| adhoc | the whole poiont of VGs and LVs is that you *can* extend storage combined across remote SCSI LUNs on large arrays or fibre channel | 01:45 |
| rwp | That may have been an original point of LVM but we have abused it into the current strategy for fully encrypted laptops. | 01:46 |
| adhoc | rwp: is this enterprise requirement ? | 01:46 |
| Xenguy | rwp, Okay, soon enough then, but time for some chow | 01:46 |
| rwp | adhoc, It's everyones' requirement. | 01:46 |
| adhoc | i hope you keep backups. | 01:47 |
| rwp | Everyone should also keep backups too. | 01:47 |
| rwp | When installing the debian-installer dialogs basically ask you questions and depending upon the answers if you answer that you want encryption then this is what you get and if you answer that you don't want encryption then you don't get encryption. The choice is your choice when you install. | 01:48 |
| rwp | Everyone should have backups regardless though. | 01:49 |
| jjakob | how would you approach building preinstalled images of devuan for raspberry pi 3's? years ago I used https://github.com/TheSin-/rpi-img-builder for raspbian (debian) and made a custom config to mount root read-only with an overlayfs | 08:33 |
| jjakob | I guess I can take that and see if I can make it work with devuan | 08:33 |
| jjakob | maybe I can do a one-off install and manually configure the overlayfs but it's nice to have an automated build | 08:34 |
| gnarface | jjakob: someone has the build scripts we use, or you can just use one of our own already-built rpi3 images: arm-files.devuan.org | 08:35 |
| gnarface | i think fsmithred has the build scripts somewhere... | 08:35 |
| jjakob | I looked at the arm build scripts on devuan's gitea | 08:35 |
| gnarface | maybe someone in #devuan-arm knows | 08:35 |
| gnarface | there's buildscripts on git.devuan.org too, but i think the one they're actually using for the arm images may be listed elsewhere | 08:36 |
| jjakob | https://gitea.devuan.org/devuan-sdk/arm-sdk | 08:36 |
| gnarface | i've seen the link but i don't remember it, but i'm pretty sure the channel logs will include fsmithred posting the link | 08:37 |
| jjakob | it's not immediately obvious how to customize the build though, like installing more packages, editing/installing config files into the target, changing cmdline and so on | 08:38 |
| gnarface | yea, i'm sure he'd know that too | 08:39 |
| jjakob | grepped through channel logs and only saw http://arm-files.devuan.org/ and refracta-installer-base being mentioned, and #devuan-arm | 08:45 |
| gnarface | hmm, maybe i'm thinking of the x86 or refracta build scripts or something | 08:46 |
| gnarface | well, slow channels. be patient. | 08:47 |
| jjakob | pbuilder, sbuild, cowbuilder, #arm-img-builder | 08:51 |
| jjakob | seems like those are just package building, not whole bootable os, other than the last one | 08:52 |
| gnarface | well, there's not much to it aside from getting a working kernel and running debootstrap, especially for the rpi images | 08:53 |
| jjakob | https://arm-files.devuan.org/README.txt | 08:53 |
| gnarface | it wouldn't be so difficult for you to make your own image from scratch if it came to that | 08:53 |
| jjakob | > For Raspberry Pi Beowulf 3.1.0 images, the build system that generated the images is at https://github.com/pyavitz/rpi-img-builder, configured with Raspberry Pi Foundation non-mainline kernel built from source on RPi3 arm64 running Beowulf 3.1.0. | 08:53 |
| gnarface | ...but i'm certain someone around here will have actually good answers once they wake up | 08:53 |
| gnarface | yea, so beowulf was a long time ago, and that information could be out of date | 08:54 |
| gnarface | we're on daedalus (5) now | 08:54 |
| gnarface | heh | 08:55 |
| gnarface | yeti in #devuan-arm just now suggested also that you could simply use the debian one and then transgrade to devuan | 08:55 |
| jjakob | yeah I guess it wouldn't be that hard to adapt any builder for debian to devuan | 08:57 |
| jjakob | anyway I went to #devuan-arm for further discussion | 08:58 |
| xrogaan | Something weird is going on with apt. It tells me updates are available for libre office, from the backport channel. Thing is, I didn't install the backport version of libreoffice | 10:39 |
| xrogaan | libreoffice/stable-backports 4:24.8.2-1~bpo12+1 amd64 [upgradable from: 4:7.4.7-1+deb12u5] | 10:40 |
| xrogaan | not only that, but the versionning seems off | 10:41 |
| xrogaan | I don't understand why backports priority became the same as stable for this packet | 10:43 |
| gnarface | i just typically keep backports commented out in the sources.list unless i'm planning on using it | 11:00 |
| xrogaan | I'm using some backports, not all | 11:04 |
| xrogaan | It's just forcing me to install libreoffice from backports and git from backports staging. | 11:04 |
| xrogaan | they're both from the stable channel, so I don't get why it happens. | 11:05 |
| xrogaan | I mean, the version I have installed isn't a backport. So the priority should take precedence over the version. | 11:05 |
| gnarface | hmm, not sure, maybe you've installed the backports version of something it depends on? | 11:05 |
| gnarface | only other guess i have (besides packaging error obviously) is maybe it's a "recommend" somehow? | 11:06 |
| gnarface | do you have recommends disabled? | 11:06 |
| xrogaan | do you know how to get the value of a config variable from apt? | 11:09 |
| xrogaan | yeah, it's not set on true | 11:11 |
| gnarface | i forget how to get the defaults | 11:12 |
| gnarface | you can override them in /etc/apt/preferences.d/ though | 11:12 |
| gnarface | i keep being told that the defaults should prevent this, but sometimes they don't so that's why i keep it commented out | 11:14 |
| xrogaan | oh, maybe the package was removed from stable? | 11:22 |
| xrogaan | holding the package does nothing | 11:33 |
| xrogaan | It's just weird. | 11:33 |
| xrogaan | So it's the repository that determines if packages are automatically installed (NotAutomatic: yes) and if updated are automatic also (ButAutomaticUpgrades: yes). So there has to be something that requires the backport version, but here is no other package installed or updated than libreoffice. | 11:46 |
| xrogaan | (the suite of libreoffice, should I say) | 11:46 |
| gnarface | are you sure about that? | 11:54 |
| gnarface | dpkg -l |grep bpo | 11:54 |
| gnarface | i've seen nvidia drivers pull extra stuff when installing from backports, it stands to reason libreoffice could do it too | 11:55 |
| gnarface | it'd be really weird for them to remove a package from stable | 11:56 |
| xrogaan | gnarface: I have bpo stuff installed. libreoffice isn't one of those. LO is kind of self contained too, without broad dependencies. | 12:03 |
| xrogaan | a depends on libreoffice only show it being suggested by spelling packages | 12:03 |
| xrogaan | ah, might be ure | 12:12 |
| xrogaan | okay, went nuclear and purged libreoffice from my system | 12:16 |
| xrogaan | and reinstalling install from stable, not backports ¯\_(ツ)_/¯ | 12:18 |
| kairu | Hi, I am also investigating the problem with the libreoffice backports. | 13:37 |
| kairu | Maybe it's because version 4:7.4.7-1+deb12u5 was installed but is no longer available. | 13:38 |
| kairu | It was installed by daedalus-security main. Somehow this repo seems to be empty. | 13:38 |
| xrogaan | oh? | 14:21 |
| xrogaan | right | 14:21 |
| xrogaan | Who's in charge of the repo, onefang? | 14:22 |
| rrq | libreoffice=4:7.4.7-1+deb12u5 is in daedalus-proposed-updates | 14:39 |
| xrogaan | I had u5 installed, but got u4 on reinstall. | 15:07 |
| xrogaan | deb12u5 should be from security | 15:07 |
| xrogaan | https://tracker.debian.org/news/1566837/accepted-libreoffice-4747-1deb12u5-source-into-stable-security/ | 15:08 |
| Alverstone | abuse me | 21:18 |
| Alverstone | I got a working devuan chroot via apt-rdepends, ar/tar and Debian's buster repos | 21:19 |
| Alverstone | don't know why I bothered with buster | 21:19 |
| Alverstone | probably though to use DPKG_ROOT at first, but it didn't work out | 21:19 |
| Alverstone | So I ended up using ar/tar | 21:19 |
| Alverstone | It's the ugliest thing I've ever done to a system | 21:19 |
| Alverstone | why I'm smiling is beyond me | 21:20 |
| Alverstone | It's also so minimal that I'm afraid I'm gonna spend a few days making it functional | 21:21 |
| Alverstone | But here are serious questions | 21:21 |
| Alverstone | Is testing reliable on Devuan? | 21:21 |
| Alverstone | How quickly are security fixes merged? | 21:22 |
| fsmithred | no longer than a week, I think | 21:22 |
| fsmithred | that's the usual time for things to move down from ceres/sid | 21:22 |
| fsmithred | or maybe it's 5 days | 21:22 |
| fsmithred | sometimes it's faster if it's important | 21:23 |
| Alverstone | Is it OK for downstream to have security fixes delayed? I though maybe there's some coordination | 21:23 |
| Alverstone | Also what about testing? | 21:24 |
| fsmithred | and I want to add that I don't understand half of what you said above. | 21:24 |
| Alverstone | They say it doesn't break often on Debian, so I suppose I can safely use it on Devuan as well? | 21:24 |
| Alverstone | > I don't understand half of what you said above | 21:25 |
| Alverstone | You don't need to | 21:25 |
| fsmithred | for a chroot tarball, I would just do a debootstrap and then tar it | 21:25 |
| fsmithred | yeah, I know | 21:25 |
| Alverstone | I can explain though | 21:25 |
| fsmithred | 99% of the packages devuan provides come straight from debian and are unchanged | 21:25 |
| Alverstone | apt-rdepends coreutils binutils libc-bin $whatever | awk '!/^ .*/' | xargs -I, apt download , | 21:26 |
| fsmithred | so testing is pretty much the same as testing | 21:26 |
| Alverstone | and then ar/tar unpack into new root | 21:26 |
| Alverstone | ugly isn't it | 21:26 |
| fsmithred | I could see doing something like that just to know if it works | 21:26 |
| Alverstone | Yeah it works, at least half of it | 21:27 |
| Alverstone | I just got an error saying unknown user root:root | 21:27 |
| Alverstone | I guess that's what you get for being dumb) | 21:27 |
| fsmithred | probably something was missing | 21:28 |
| fsmithred | wild guess | 21:28 |
| Alverstone | I didn't want to install Devuan's debootstrap into my Debian that badly. Reason isn't applicable | 21:28 |
| fsmithred | I've thought about providing a root tarball for people to start with. | 21:28 |
| fsmithred | but I don't suppose many would go that way | 21:29 |
| fsmithred | about testing: until freeze, there are many updates | 21:29 |
| fsmithred | so you can do a bunch every day or do hundreds every week or month | 21:30 |
| Alverstone | Yes I can | 21:31 |
| Alverstone | But do I have to? | 21:31 |
| fsmithred | I have a print server here that's running excalibur and it works | 21:31 |
| fsmithred | no, if you have a working system, you don't have to keep doing the upgrades | 21:31 |
| fsmithred | but if you want security fixes fast, you'll probably want to upgrade every day | 21:31 |
| Alverstone | You can do something like Void Linux does, providing a very small rootfs. It will take about 40-50mb of space, so can't be an issue for your servers | 21:32 |
| Alverstone | Install apt/dpkg there and the most basic things, like hostname etc etc | 21:32 |
| fsmithred | or do a mixed excalibur/ceres and pin ceres to lower priority. Then if something gets fixed that you need right away, you can install that package from ceres and get it a few days earlier. | 21:32 |
| Alverstone | who's testing? excalibur? | 21:33 |
| fsmithred | yes | 21:33 |
| fsmithred | =trixie | 21:33 |
| fsmithred | I do think I have some problems in my excalibur install. If I try to upgrade with aptitude, it pukes. But apt upgrade works fine. | 21:34 |
| Alverstone | happens on Debian? | 21:35 |
| fsmithred | I haven't tried to figure out what the problem is. | 21:35 |
| Alverstone | I never used Debian testing tbh | 21:35 |
| freem | this apt-rdepends thing is an interesting, thx. I'll toy with it to replace the debootstrap of my scripts, just because why not. | 21:35 |
| Alverstone | Do you guys stick to openrc? I've had a brief experience with runit, so I thought about using it, but if openrc is more convenient I'll try to figure it out instead | 21:36 |
| freem | I never used openrc. I switched myself to runit when I learned about it, which is a bit before devuan was born | 21:37 |
| freem | I love runit, most of my files are just including a 35 long file for basic helpers in a 2-lines long "script", and adding a log/run symlink | 21:38 |
| freem | the only limitation it have is when you need a daemon to be launched and *ready* before another one, but that is not exactly a common need, in practice | 21:39 |
| freem | and you handle this anyway, by implementing the check script in a correct way for that required daemon, after all. If said daemon provides a reliable tool to hint that it is, indeed, ready, that is. But if not, no can do anyway. | 21:41 |
| Alverstone | I remember somebody suggesting doing the dumb fail-until-succeed | 21:42 |
| freem | I never managed a machine hosting a standard website with it, though. By standard website, I mean a CGI-able httpd, some SQL db, and some various related tools. | 21:42 |
| freem | yes, that is the traditional way in daemontools/runit | 21:42 |
| freem | well, there are no other way, actually | 21:42 |
| freem | but you do not have to actually launch the daemon, if your run script uses the "sv check $foobar" command, instead of stupidly being just "exec barfoo" | 21:43 |
| freem | anything better than that will require communication between the daemons and the supervisor. We probably all know about the last systemd fail on that point, right? | 21:44 |
| freem | well, systemd-related, the patches in question were from distros which added unneeded bloat | 21:44 |
| Alverstone | ahem testing doesn't have security and update branches right? | 21:45 |
| freem | but either you periodically checks that the target is up, or you communicate with it's responsible to know when it is. There are no other ways | 21:45 |
| * freem will stop here before being told to go offtopic though | 21:45 | |
| fsmithred | sysvinit is the default in devuan. The installer gives you a choice of that or openrc or runit. | 21:46 |
| fsmithred | openrc and runit will use sysvinit scripts | 21:46 |
| fsmithred | runit will use run scripts if they exist | 21:46 |
| freem | this ^ as a replacement for the rc.d ones, that is | 21:46 |
| fsmithred | I don't know what you have to do with openrc to replace the sysvinit scripts | 21:47 |
| fsmithred | yes | 21:47 |
| freem | and you need the folder to match the name of the init.d/ one, which can be surprising when it comes to, say, ssh (you need to name ssh, not sshd) | 21:47 |
| fsmithred | and there are a few already configured in runit-services package | 21:47 |
| Alverstone | So what benefits does sysvinit have compared to runit | 21:48 |
| freem | none. | 21:48 |
| fsmithred | it's more mature by decades | 21:48 |
| freem | well, it's historically used everywhere | 21:48 |
| freem | but I don't see any techincal one... oh, actual runlevel support, perhaps? | 21:48 |
| fsmithred | there aren't run scripts for all services yet | 21:48 |
| freem | runit-init specifically does not | 21:48 |
| fsmithred | yeah, runlevels. I miss that in runit. | 21:49 |
| freem | what is the use case? | 21:49 |
| fsmithred | init 1 instead of reboot | 21:49 |
| freem | so that init just restarts at it's entry point? | 21:49 |
| fsmithred | yeah, restarts all the services | 21:49 |
| freem | and goes through all the init sequence, except for kernel loading? | 21:49 |
| freem | I see | 21:50 |
| freem | any other use case for runlevels? I'm curious. And *not* saying this is not a valid one, I can think of diskless systems for example | 21:50 |
| freem | or live CD or anything which have slow storage read, I guess | 21:51 |
| fsmithred | well, debian/devuan makes 2-5 the same, but redhat, suse and I'm not sure about slackware, use 2,3 for multi-user non-graphical login, 4,5 for graphical | 21:51 |
| freem | that stuff is easy to do with runit though | 21:52 |
| fsmithred | good, I'll ask you for help some time. | 21:52 |
| freem | IIRC it's even documented somewhere... lemme search | 21:52 |
| fsmithred | I can arrange graphical vs. non-graphical logins, but I can't change from one to the other easily | 21:53 |
| fsmithred | without rebooting | 21:53 |
| freem | fsmithred: I think this: https://smarden.org/runit/runlevels handles the use case of switching between single/multi/graphical users | 21:53 |
| fsmithred | thanks | 21:53 |
| freem | I do not know, though, I never got the usecase, so never tried. Please tell me when you will have toyed with it, I'm curious | 21:53 |
| freem | (and incidentally am planning to rewrite runit in modern C, and perhaps in C++ actually, just for my fun and maybe to put that djb code into a nice system C++ lib which does not throw nor abort() nor exit() nor allocates memory randomly...) | 21:54 |
| freem | aka: something very different from the STL | 21:55 |
| freem | so I suppose while I'll be at it, I could keep in mind some use cases which are not handled yet. Will probably never be done, though, don't hope too much :) | 21:56 |
| freem | oh, this makes me think... fsmithred: do you know if devuan's prepared runit scripts get some handling for cgroups/namespaces/[selinux|apparmor]? I know it is technically possible, but myself never got the motivation to go that far | 21:57 |
| freem | (technically just need to call the cgroups tooling and toy with the kernel's filesystem API...) | 21:58 |
| freem | (just...) | 21:58 |
| fsmithred | I don't know. If you want to see the source code, it's at salsa.debian.org | 22:00 |
| freem | oh, I thought you used them | 22:01 |
| freem | since you knew about some details of the runit integration that I would not expect someone who is not writting their own runit scripts themselves to know | 22:02 |
| fsmithred | I've made some live-isos with runit and I've written a few run scripts | 22:03 |
| freem | this explains that | 22:04 |
| fsmithred | I have only a vague notion of cgroups | 22:04 |
| freem | I know the general idea of cgroups, namespaces and selinux/apparmor, because at some point I wanted to become admin for real, but my hacker mind is unable to accept using high level tools like docker I don't understand how they work internally | 22:05 |
| fsmithred | I know nothing. I just kick it around until it does what I want. | 22:05 |
| freem | I don't even know how to kick it around :D I never implemented it myself, and by default it's out of the way | 22:05 |
| freem | being out of the way by default is not a wisdom lost by everyone after all | 22:06 |
| fsmithred | oh, I was referring to linux in general. | 22:06 |
| freem | ah lol | 22:06 |
| freem | I don't even dare touching the kernel except by poking random /proc or /sys files | 22:06 |
| freem | one of my many lacks. | 22:07 |
| fsmithred | we should be in OT for this | 22:08 |
| freem | yes | 22:08 |
| freem | fsmithred: when you will be back, I think I should mention about runit's "runlevel" situation that it does not touch the init at all (I was unsure): it only changes the directory targeted by the service-handled (runsvdir) which will thus start or end daemons depending on the difference. The runit-init scripts `1`, `2` and `3` are not executed, which means the one-shot services will not re-run. I do not know if this is how sysvinit+rc.d runlevels work | 22:26 |
| freem | as well, but I kind of doubt it? I guess one could migrate those to proper runit scripts, but I never went to do that myself as it might be more complicated for a quite small benefit (for me) | 22:26 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!