libera/#devuan/ Tuesday, 2024-10-01

fonkyhi all00:12
fonkygot one question, im using live 5.0 and removed xfce and installed lxde and while-so fdisk -l does not show usb_nand_flash 2.0 key when plugging it, while on xfce it worked, any ideas?00:13
fonkynor gparted00:13
fonkybut however it does show it is connected, port # and serial and so on00:13
fonkyoh and if it is of any concern00:14
fonkydpkg: warning: 'ldconfig' not found in PATH or not executable00:14
fonkydpkg: warning: 'start-stop-daemon' not found in PATH or not executable00:14
fonkydpkg: error: 2 expected programs not found in PATH or not executable00:14
fonkyNote: root's PATH should usually contain /usr/local/sbin, /usr/sbin and /sbin00:14
fonkythis pops out every time i boot00:14
rrqdpkg warings during boot?00:18
fonkynone00:19
fonkyand it is happening alot of times00:19
fonkyeven after entering00:19
fonkyexport PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin00:19
fonkyit works for a while then i have to repeat00:19
fonkyand sorry for exceeding lines in a certain time interval00:19
rrqare you booting live 5.0 and get those wrnings and errors?00:20
fonkyyes it is a live00:21
fonkyhappens most of the time whilst trying using dpkg -i something.deb00:21
rrqand you do that by using "su -" ?00:22
fonkyyes, with root account00:22
fonkybut not su && the command00:22
freemthe important point is the argument '-' or '-l'00:22
rrqby using "su -login" ?00:22
fonkysu and the command after00:22
freemif missing those, then you will not get root's environment, leading to wrong $PATH00:22
fonkywhen already logged as root00:23
rrqthe point is that you don;t actually log in as root00:23
rrqunless you use "su -login: or the abbreviation "su -"00:24
rrqunless you use "su -login" or the abbreviation "su -"00:24
fonkydevuan@devuan:~$ su00:25
fonkyPassword:00:25
fonkyroot@devuan:/home/devuan# modprobe usb_storage00:25
rrqthe point is that you don;t actually log in as root00:25
rrqunless you use "su -login" or the abbreviation "su -"00:25
fonkydidnt know that, thank you00:25
fonkyso if logging from account devuan 2 root using just su does not yield real root00:27
drizztyep00:27
drizztyou only substitute user and group id00:27
fonkyoh00:28
drizztbut keep (most) environnement00:28
drizztman su00:28
fonkyIC00:29
fonkyi have been doing so since well started using linux, never with - <-00:29
drizztif you want/need to become root, then login as root, which means running "su -" (short for su -login)00:29
fonkythank you00:30
rrqfonky: yes, some year(s) ago now, debian changed the upstream for "su" to one that doesn't set PATH unless you use "-"... the previouls upstream did00:31
fonkyhmm must have been inbetween 2016-201900:34
fonkythen i was not paying attention :)00:34
fonkythank you for the polite explenation00:34
freemfonky: yes, this changed relatively recently, perhaps 2 stable versions of debian ago?00:47
freemoh, was explained already, sorry00:47
fonky:)00:47
freemI got weirded by that change as well, and sometimes still forget to add the -l00:47
freemmost people probably never noticed since mostly using sudo00:48
nemodrizzt: I don't disagree with anything in your lecture, but it really wasn't relevant to my situation either...00:49
nemodrizzt: but yeah. is off-topic for here anyway00:49
fsmithredfonky, if you want to restore the old behavior, put the following into /etc/devault/su (create the file)00:49
fsmithredALWAYS_SET_PATH yes00:49
fonkyim guessing but they probably did this for security purposes?00:50
fsmithredit happened when su got moved into a different package00:50
fsmithredeither to or from util-linux. I forget which.00:50
freemI'd love to read the reasonning behind the change as well, and thanks for the hint to get back to old behavior, I might do that00:51
fsmithredsee 'man su' for more things to put in there. I have no idea what else might be useful.00:51
fonkyum any ideas however on how i could get usb flash drive seen in fdisk00:52
freemdoes not dmesg give you the file to feed fdisk with?00:53
fonkyas if i dont touch the livedvd and leave it as is it works00:53
fonkyyes, showing and lsusb -v shows extensivly, dmesg | tail 30 shows basically it is there00:53
fsmithredam I to understand that you change the desktop environment while running the live system, as opposed an installed system?00:54
buZzsomething like 'fdisk /dev/sdc'00:54
freemwhen I plug something, I usually run dmesg, since not running a DE nor anything to auto-mount, so that I need the info about the exact file.00:54
buZzfonky: tail -n 10 ;)00:55
buZzunless you got some funky tail00:55
fsmithredI would guess it's some policykit issue. Is lxpolkit installed? (not sure if I spelled that right)00:55
freemwhat is the error message?00:56
freemlet's start with that :)00:56
drizztfsmithred: PATH is not the only difference between su and su -, whole env is concerned : you keep old env, meaning any exported var00:56
drizztwhich is a security problem00:57
fonkyhttps://pastebin.com/JtfAMLLE00:57
fonkynono it does not show it in fdisk00:57
freemand you stay on the folder you were in when issuing the `su` command, instead of having to move back00:57
freem(which is why I might get back to old behavior, even if less secure)00:58
freemfonky: what is the error that `su -l -c 'fdisk /dev/sg1'` gives you?00:59
fonkythe fdisk line give me cannot open, no souch file or directory00:59
fonkygives*00:59
freemwith that exact command?00:59
fonkyfdisk -l /dev/sda01:00
drizztfreem: "su" is the less secure behavior01:00
drizzt"su -" is the secure one01:00
freemdmesg reports your disk is /dev/sg1, not /dev/sda01:00
freemdrizzt: I get this, but security and usuability are 2 opposite things. One needs to pick the balance that suits them, depending on the risks01:01
fonkyfdisk: cannot open /dev/sg1: Illegal seek01:01
drizztfonky: does "cat /proc/partitions" show the partitions you are looking for ?01:01
drizztor "lsblk" ?01:01
fonkypartitions sr0 and loop01:02
fonkyonly01:02
freemillegal seek is something I don't remember seeing01:02
fonkylsblk again loop and sr001:02
freemdid I misread the log?01:02
drizztfreem: agreed, but I would have chosen security as default behavior ;)01:02
drizztthen kernel does not hendle your disk01:03
freemwell, see, urxvt did that. Now, you need to validate when you want to paste several lines in a terminal. It is... very annoying. That su change is similarly annoying, for someone like me, with my uses and knowledge. I don't claim I'm smart and very knowledgeable, but I believe I know the basics enough :)01:03
freemI *do* understand I am not the average computer user though01:04
freemso I understand the choice for security by default, especially when you see so many "wget foobar | sudo bash" on the interweb :S01:05
freem(which hurts my eyes for several reasons, not just one)01:05
freemfonky: do that device work on other systems?01:05
fonkythey work if i dont ditch xfce on the livedvd01:06
freemflash storage are not exactly what I'd call sturdy01:06
fonkyon same pc same session01:06
freemoh, interesting01:06
fonkywhat i did was init 1, then init 301:06
rrqdoes init 3 re-run udev?01:07
fonkylemme check01:07
fsmithredinit 3 should be the same as 201:07
fonkysame things start yes01:07
fsmithredunless it was something other than official desktop-live01:07
fsmithredsome make 3 non-graphical multi-user01:08
fonkyi installed lightdm instead of slim and lxde instead of xfce01:08
fonkyand did a dist-upgrade before that01:08
fsmithredin a live session? How much ram you have??01:08
freemsilly question, but could something make `chmod 000 /dev/sg1` in the back? Would this give that weird lseek error?01:08
fonky24gig01:09
rrqI'd say the first option is that some needed kernel module fails to be loaded (for handling the device)01:09
fonkyi tried modprobe usb_storage01:09
fonkyreturns blank01:09
rrqdg1 is a generci scsi interface; not a block device01:09
rrqwhich *hci* modules do you have?01:09
rrq(that's the USB hu support)01:10
rrqhub01:10
fonkyusb_common01:10
fonkyusb core01:10
rrqand *hci ?01:10
rrquhci is USB 1, ehci is for USB 2 and xhci for USB 301:11
fonkyhttps://pastebin.com/tYn7933101:11
rrqok. looks like it should havnde the drive01:12
rrqwhat's in /sys/class/block/ ?01:12
freem<rrq> dg1 is a generci scsi interface; not a block device01:13
freemso I *did* misinterpreted the logs, my bad01:13
fonkyloop from 0->7 and sr001:13
drizztrrq: init 3 should'nt re-run udev, which should be part of rcS01:13
rrqok ta01:13
drizzt(sorry if I miss some answers, I have connections problems this evening :( )01:14
fonkyoh works01:14
fonkyrestarted eudev01:14
fonkymy thanks01:14
drizztwell, it shouldn't have stopped either01:14
drizzt(udev/eudev)01:15
freemwhy would restarting eudev (restarting: stopping and starting, to me) be necessary, though?01:15
freemit should be sitting around all the time for that exact purpose of discovering plug'n play stuff after all?01:15
drizztfonky: what kind of usb stick is it ?01:16
drizztand anyway, devfs should handle this without udev/eudev01:16
drizztwell, devtmpfs, not devfs01:17
rrqinit 1 does run killprocs01:19
fonkyhttps://www.kingston.com/en/usb-flash-drives/datatraveler-kyson-high-performance-usb-flash-drive01:20
fonkythis exact one01:20
fonky128gig01:20
fonkyoh and it is ntfs formated01:21
fonkyand that high performance is 20mb/sec on a shiny day01:21
fonkyread is fine though01:21
rrqright (my) init 1 does run killprocs.. which includes killing udev01:21
fonkyread is 127MB/sec01:24
drizzt(missed last 8 minutes ...)01:24
freemdrizzt: https://p.mort.coffee/X78 those are all you possibly missed, I suppose01:26
drizztfreem : thanks01:28
freemyw. Unstable connections are a pain, and libera does not yet have IRCv3's history, which would handle that problem.01:29
drizztI missed some parts earlier I think, I did not notice the lags01:29
freemdrizzt: ah, yes, around 1:10 a.m. I guess01:32
drizztalso between 0:33 and 0:46 am01:33
drizztand maybe other shorter periods01:33
freemmaybe this chan is logged, though?01:33
drizztas for this udev problem, I have the message about "init 1 then init3" but I missread it :(01:34
* freem fails to parse it as well01:34
freemmaybe about runlevels01:34
fsmithredChanlogs http://reisenweber.net/irclogs/libera/_devuan/01:36
drizztbut even with udev killed I still have sda and sda1 showing up when plugging in a USB stick on one of the test systems I have on my desk01:36
drizztand it should behave the same with kernel starting from 5.X at least01:36
Xenguyrawr01:36
drizzt(it's sda there because the system is running from an µSD card seen as mmcblk)01:37
freemhm... last time I read about this stuff, I understood that it is (e)udevd which is responsible for populating /dev... but I don't know much on that topic, sadly (otherwise I'd have gotten rid of udev for the discovery of the static hardware since long...)01:38
drizztwhat does "mountpoint /dev"01:38
drizztwhat does "mountpoint /dev" say ?01:39
drizztand/or "mount |grep devtmpfs"01:39
rrqif I stop udev and unload some modules (esp "uas") it fails to set up the dev node for the usb stick01:47
freembut why would changing the DE unload modules?01:48
rrqthe kernel handles "immediate" dev nodes for the hardware but not the "indirect" ones01:49
rrqmost likely fonky didn't juggle with the USB in a possibly working order01:50
freemI don't think I understand what you just said :/01:50
freembut it's ok, I know it's not my domain at all01:50
rrqI meant that, yes, maybe the USB issue wouldn;t have come up of it was inserted with the first DE and then stayed in place during the DE shift... maybe01:51
rrq(because then the running udev would have installed the required modules early enough)01:54
rrqall guesswork of course, but it now works for fonky, which is good.01:55
freemindeed01:55
fonkya quiet camper is a happy one :)01:59
fonkylog about if daemons got killed and or restarted is in syslog?02:03
fonkyim just happy because i do all the stuff in linux and use usb constantly02:05
fonkywell except for cubase and reason02:06
drizztback ...02:06
fonkywb02:06
freemthat dark elf is stubborn I see, keeps going back after the kicks :p02:07
drizztleft after my 01:39 message according to chanlogs02:07
fonkyit was really an itch because i looked with a browser, at least first x pages google serves02:07
fonkyand it didnt yield any results02:07
freemI don't remember if udev logs in syslogd by default or if they have their own logger. I have it's output captured by svlogd, but my setup is not standard02:08
fonkybtw what happened with /var/log/messages02:09
drizztyou apparently missed my last messages :02:09
freemI have no specific option in my udev/run file though, so I suspect it will just pick what it prefers, syslogd if one is running seems likely02:09
drizztdevtmpfs is a kernel subsystem which maintains (populates) a dynamic /dev02:09
drizztudev/eudev is usefull for the "practical" links like those found under /dev/disk/*02:09
freemreally? Only those? If I do not have udev running on my startup, kernel will stay in the VESA graphical mode, though02:10
gnarfacefonky: removed from the stock upstream /etc/rsyslog.conf, but you can put it back in yourself02:10
drizztbut it's now too late for me to try to understand this any further, and if it's now running fine, that nice :)02:11
drizztfreem: udev/eudev is not useless, but it does not handle basic stuff like /dev/sd** anymore if you have devtmpfs mounted on /dev02:12
drizztfreem: also, you may not have seen my message about "-confirm-paste" ?02:13
freemthat is good to know! i'll tinker with that tomorrow, that will be one less fragile componenent on my system: % upl /etc/sv/udev/run -k02:13
freemhttps://p.mort.coffee/ZNJ02:13
drizzt(for the solution to the anoying urxvt paste message)02:13
freemthis script is *really* ugly...02:13
freemoh, right, I missed it, thanks02:14
freemlost patience the other day, picked the easy solution: downgrade.02:14
freemI guess it's a command-line parameter to pass?02:14
drizztfreem : nope02:15
drizztfreem : for this paste problem with urxvt, add this to your .Xdefaults : "URxvt.perl-ext-common: -confirm-paste"02:15
freemthanks02:15
drizztif you already have ""URxvt.perl-ext-common: something", then make it "URxvt.perl-ext-common: something,-confirm-paste"02:15
drizzt(the messages)02:15
freemthanks, I'll upgrade it anew and see if that works when I'll boot, after a good sleep :)02:16
drizztI also had an example for what "unstable" does not mean : apache2 in ceres is Version: 2.4.62-102:17
freemI should clean my $HOME and dotfiles someday thuogh, it's a giant mess of past tinkerings with various random stuff02:17
rrqdrizzt: afaict nowadays module loading is outsourced to udev; the kernel might know which module(s) are needed, but it doesn't load, handling some drives require a small stack of modules. Those get loaded by udev.02:17
drizztwhich is current ... stable apache2 version from apache (https://httpd.apache.org/)02:18
drizztrrq: yep02:18
drizztbut it was seen before init 1, so I would have thought that modules would have stayed02:19
drizztI do not know if init 1 unloads any module02:19
drizzt(I use very few modules on my systems, I usually build a kernel which matches my hardware and needs, and no more)02:20
rrqit wouldn't... so I think the case came up when the USB was added after the DE shift (killing udev)02:20
drizztyep, in this case the modules wouldn't have been loaded before, and would be missing02:21
freemcompiling a kernel matching the actual hardware indeed seems the best to do, but I remember being completely lost in the kernel's configure interface, ~10 years ago. It's probably even worst those days02:22
drizztwhich is coherent with the syslog messages missing the sda: sda1 parts02:22
drizztfreem: that's part of my job, so it helps me stay tunned with the interface, but it's not something averyone can do02:23
drizzt*everyone02:23
freemI guess02:23
freembut I also think I simply tried to configure things too much back then, for a 1st time :D02:24
freemthose years I no longer have much patience for this kind of tinkering, thuogh02:24
drizzt(not about knowledge, which anyone could get, but rather about time to get the knowledges + do the task)02:24
freemdon't worry, I got that02:25
freemI lived from c++ coding for more than few years, I know I would be able to setup that, if I had the motivation.02:26
drizztwhich is exactly why we have linux distributions, and why we do not like package maintainers to change the defaults02:26
freemnot only package managers, upstream devs as well02:26
drizztyep02:26
drizztone of the core reason behind devuan :)02:27
freemwell, I have the advantage compared to many that 1) I have a minimalistic system and 2) I can git revert and understand what I do when it's about C or C++02:27
freemthat's likely offtopic though02:27
drizztI have /etc under git, pretty usefull to check what got changed by an update02:28
drizztthough it requires to spend some time after each upgrade02:30
freemwell, most of the PID1/PID2 stuff I have, I wrote myself in any case. It won't change without me noticing it. Then, only few stuff remains, and dpkg will tell me about those I care about, since I modified them, so it asks what I want to do02:30
freemrunit *is* great for writting the service stuff yourself, it's almost only one-liners, lovely and trivial02:30
drizztthe problems do not come from these, but from where the default changed ...02:31
freemtrue02:31
freemI shuold probably version them as well, indeed. So far, I simply barely had any bad surprise... except for su and urxvt, I don't remember any, actually.02:32
drizztsysv init is not that bad either for all the base startup02:32
freemI only have the daemons handled by runsvdir, not all the one-shot stuff02:32
freemI never got to the point of migrating those. I should, though, I'd master my system more then.02:33
drizztI have my own script which does all of rcS for my embedded systems02:33
drizztas I always know what it has to do, and it never changes, it is way quicker than the original "parallel" scripts02:34
freemI don't use runit for speed, seriously. I use it because it's simple, for a start, and 2nd, because you don't want to travel 200km each time you need to restart, say, ssh, pppd, or other important daemon because the system would not start it anew02:35
drizztbut as I said, I have to get some sleep !02:35
freemwhich is something people thanked me to bring in my last IT job.02:35
freemI need to do that as well02:35
freemgn02:35
drizztbye :)02:36
onefangdrizzt: Are you using etckeeper to keep /etc under git?  It can do everything automatically, by hooking into the apt system.  So no need to "spend some time after each upgrade ".04:37
drizztdid not know about etckeeper, gonna give it a try !11:23
Xenguydrizzt, Please report back on etckeeper14:43
AlverstoneHi. What's my slap-slap-and-ready strategy to install Devuan without making installation media? You don't distribute a rootfs, but maybe there's a way with debootstrap?21:28
fsmithredyes, you can use debootstrap21:30
fsmithredwhat are you running now?21:30
AlverstoneBut apparently it wouldn't make sense if it installed systemd, and IIRC debootstrap's notion of basic system implies that21:30
AlverstoneI'm on Debian 12 now21:30
AlverstoneI don't want to migrate, I want a separate fresh install to see how it goes21:31
fsmithredmaybe you can install devuan's debootstrap21:31
drizztAlverstone: nope, devuan's debootstrap won't install systemd21:32
Alverstonewhat exactly do you mean by "debuan's debootstrap"? Can I just shove Devuan's mirrors at Debian's debootstrap, or that's not how magic works?21:33
drizztdo you have /usr/share/debootstrap/scripts/daedalus ?21:33
drizztor /usr/share/debootstrap/scripts/excalibur or /usr/share/debootstrap/scripts/ceres ?21:33
Alverstoneyep, got only ceres21:34
fsmithredwe fork debootstrap. Here's the one for current stable: https://pkgmaster.devuan.org/devuan/pool/main/d/debootstrap/debootstrap-udeb_1.0.134devuan2_all.udeb21:34
fsmithredoops21:34
fsmithredthat's a udeb21:34
Alverstone>udeb21:34
Alverstonewhat's this21:35
fsmithredhttps://pkgmaster.devuan.org/devuan/pool/main/d/debootstrap/debootstrap_1.0.134devuan2_all.deb21:35
fsmithredudebs are for the installer. You want the .deb21:35
drizztcheck wether your ceres file mentions systemd21:35
MarcFPhi21:35
fsmithredif only ceres, then make a symlink to daedalus21:35
drizztif it does, uninstall debootstrap, download the one pointed by fsmithred, and dpkg -i /debootstrap_1.0.134devuan2_all.deb21:36
drizzt(without the "/")21:36
Alverstonedrizzt, https://paste.debian.net/plain/133101421:36
Alverstonefsmithred,21:37
fsmithredit should replace the debian version. We fix the versions so they supercede the debian package.21:37
drizztAlverstone: it mentions debian keyring, remove debootstrap package21:38
drizztthen get the one from the link pasted by fsmithred21:39
drizzt(wget)21:39
drizztand dpkg -i debootstrap_1.0.134devuan2_all.deb (as root of course)21:39
drizztAlverstone: but from what I remember, current "base" package list has a dependency problem as it does not resolve alternatives21:41
drizztI had to modify the script for my latest runs to get opensysusers instead of systemd or systemd-standalone-sysusers or systemd-sysusers21:41
drizzt(cron-daemon-common dependency)21:42
fsmithredwas that ceres/excalibur?21:43
fsmithredthere's a cron-daemon-common problem there, but not in daedalus21:43
drizztyep, ceres21:44
fsmithredexcluding cron-daemon-common and logrotate helps21:44
drizztit was during my tests with arm64 segfault in ? (maybe libc ?)21:44
fsmithredand using mmdebstrap instead avoids the problem21:45
drizzt(which i did not finish yet :(21:45
drizztfrom what I can recall excluding cron-daemon-common was not possible (did not work)21:47
drizztbut I do not manage to get my hands on the notes I took back then :(21:48
drizztfound !21:49
drizztdebootstrap --foreign --arch=arm64  --include=binutils,vim,bsdutils,openssh-client,openssh-server,locales,file,net-tools,bzip2,devuan-keyring,diffutils,findutils,iptables,isc-dhcp-server,bridge-utils,bsdextrautils,bwm-ng,eudev,fdisk,file,git,hostapd,iw,make,net-tools,nfs-common,patch,screen,python3,usb-modeswitch,wireless-tools,wpasupplicant21:49
drizzt--exclude=ed,nano,tasksel,tasksel-data,vim-tiny,libsystemd0,systemd-standalone-sysusers --verbose daedalus bstrap.d/ http://fr.deb.devuan.org/merged21:49
drizztit was daedalus21:49
drizztwell these were the include/excludes I needed, adapt to your needs21:50
drizztand do not use " --foreign --arch=arm64" if you're running it on host with same architecture21:50
fsmithredI could be remembering wrong, but I think you can't exclude libsystemd0 in daedalus. You have to replace it afterward with libelogind0.21:54
drizztI think I managed it by modifying the file in /usr/share/debootstrap/scripts/22:02
drizztlook for work_out_debs () function22:07
drizztand in this function : # Fixup dependencies: debootstrap doesn't resolve alternatives, so we need manual fixes.22:08
drizztyou can put your modifications there22:11
drizzt(well, that's not a file we're supposed to modify ... but ...22:11
jonadabSo keep your modifications as a .diff22:14
freemgood old cp.old, ofc, it's doable22:17
freemit's also painful, and actually easier to just manage the fork yourself in many cases22:18
freemthis, because you don't have to handle a community22:19
freemI mean, a skilled person's interests are hard to keep in sync with a community's interest22:20
freemsorry, should be offtopic22:20
freemI apologize22:20
Alverstonethank you22:30
AlverstoneI don't feel good though, so I'll figure it out sometime later22:31
Alverstonehave a nice night ;)22:31
fsmithredthere's also dpkg-divert for alternate files so they don't get clobbered on upgrade22:34

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