libera/#devuan/ Tuesday, 2020-08-11

malade_mentalHi, I just installed a minimal netinstall of devuan 3.0 (on a virtual machine), but when I log in with root, I don't have the /sbin, /bin in my PATH variable. Is this normal?00:53
gnarfacemalade_mental: yea that's a new change from upstream.  it's stupid but you can just put it back.01:09
malade_mentalmwarf01:13
malade_mentalin /etc/default/bash ?01:14
malade_mentalthe problem is that by default i don't have "rc-service" etc...01:14
gnarfacei think /etc/login.defs and /etc/profile01:17
gnarfacemalade_mental: ^01:17
malade_mentalty gnarface !01:21
malade_mentalis it considered as a bug?01:22
gnarfacemalade_mental: not sure, but no i don't think so, unfortunately01:24
rwpI assume that is part of the movement to have a combined /usr/bin & /bin and /usr/sbin & /sbin?02:26
fsmithredno, it's a product of su being moved from the shadow package to util-linux02:26
rwpOh, su...  So it's only when you su to root?  (I am looking to see where this PATH problem is happening...)02:27
fsmithredTo revert to the old behavior, put this in /etc/default/su02:27
fsmithredALWAYS_SET_PATH yes02:27
rwpI haven't seen the problem myself it.  Looking to see if I can recreate it.02:27
fsmithreddefault setup in beowulf is this: if you use su, you don't get the sbins in your path. If you use 'su -' you do get root's path.02:28
rwpHow about in ceres?  That's where I am looking at the moment.  I'll switch to looking at beowulf.02:28
fsmithredI assume ceres is the same as beowulf.02:29
fsmithred99% sure it is02:29
rwpIf this gets tedious just ignore me...  I look in /etc/profile and PATH there includes PATH="/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin"02:30
gnarfaceif you upgraded, you probably kept your old version, but people with new installs are getting it without /sbin i think02:30
rwpI do not have /etc/default/su on the system.  util-linux is installed version 2.33.1-0.1+devuan1~beowulf202:30
gnarfacemalade_mental: i forgot to mention the default behavior of su changed too, see this conversation here^02:31
rwpRight.  It's an upgrade.  However I probably somewhat unusually upgrade with -o DPkg::Options::=--force-confnew -o DPkg::Options::=--force-confmiss02:31
rwpThat should replace all of my previously existing conffiles with the new version of the conffile from the package.02:32
rwpI also look for obsolete conffiles and remove them too.  So I won't have anything left behind from the previous that doesn't exist in the new package.02:32
rwpHopefully.02:32
gnarfacemaybe i'm wrong and that's not in beowulf yet02:32
fsmithredyou would have to create /etc/default/su02:32
rwpFYI on the obsolete conffiles: dpkg-query -W -f='${Package} ${Conffiles}\n' | less02:33
rwpI do not have a /etc/default/su file.  Not anywhere that I look on the various systems.02:35
fsmithredright. You would have to create it.02:36
rwpI do usually have my own custom /root/profile file installed overwriting the system one.02:36
fsmithredthen you're all set02:37
rwpSo if PATH was inappropriately set in /root/profile then I would have automatically fixed it with my overwritten copy.02:37
rwpRight. But I tend to be local admin support. So when odd and unusual questions arise in IRC it is good if I understand them because I can guarentee you that if it is a possible problem in the wild that I will eventually get asked to help someone IRL about it.02:38
fsmithredwhat annoys me about the new setup is that I usually want to stay in the current directory when I su, but that no longer gives me root's path.02:38
fsmithred'su -' gives the right path but changes directory to /root02:39
fsmithredlol02:39
adhoc'su -' is the right thing to be doing?02:40
rwpI prefer to change directory and reset HOME. Because otherwise many utils chown stuff in ~rwp to be owned by root causing me problems later.02:40
adhoci very rarely use just 'su'02:40
fsmithredadhoc, I think there's some room for opinion02:40
rwpadhoc, "Right thing" is a loaded question. Since it all depends.  And there is more than one way to do things.02:40
adhocjust su mess up permissions on local files in the users directory02:40
rwpBut "su -" and "sudo -i" are both good ways.02:40
adhocsudo is rarely "good", IMHO.02:41
adhochowever, grok "sudo -i"02:41
adhoci tend to use "sudo su -" from habbit02:41
rwpReally?  I use 'su' *a lot* but I have no grudges against 'sudo' and also use it a lot.02:41
rwpMe too.02:41
adhocI only use sudo on systems that are outside of my admin sphere where accounts are centrally managedy02:42
adhocsome linux distros enforce sudo, which I feel is less than ideal.02:42
rwpI don't know of any distros that enforce using sudo but I do know some corporate environments that do so as a local admin policy.02:43
adhocanyway, I gave up "su" as it would change the "user:group" of dot files in the home directory of the user. makes like hard later02:43
rwpFor say Ubuntu and Mint and those one can always set a password for 'root' and then use su if one wants to.02:43
adhocdefault ubuntu does not set a functionally available root password and you must sudo02:43
adhocso when you try to support those platforms for other users, life is fun. in the dwarf fortress sense02:44
rwpAgreed on the changing of ~/ dot files.  That is mostly due to $HOME not changing.  aptitude is an example of a program that writes stuff in $HOME as root. But many others do too.02:44
adhocvi02:44
adhocvim02:44
rwpBut remember if you have sudo you can always "sudo passwd root" and set a password for root.  Then there is a password for root moving forward.02:45
adhocanyhow02:45
adhocsure I can, but other folks new to linux, not so much.02:45
rwpLearning curve for new users is always going to be a problem.02:45
rwpIt is the same as the tyranny of the default.  Whatever is the default then the newbie will never change from it.  So the default becomes the majority.02:46
adhocI've been using linux for a while, but my students who have hours, I try to avoid problems where they mess things up too much02:46
adhocmost my students run winodws, 10% mac, 1% linux.02:47
rwpUnfortunately that is about typical demographics these days.02:47
adhocWe provide them with VMs to do their technical tasks.If I can avoid it, I don't use Ubuntu.02:47
rwpmalade_mental, In your minimal install is /bin a symlink to /usr/bin?  If so then that would explain why you have /usr/bin in PATH but not /bin due to that consolidation.02:53
rwpWhich is a separate issue from 'su' not changing PATH for root to include /usr/sbin (and /sbin) by default anymore.02:53
adhocrwp: is there a move to combine all the things into one "bin" path?03:16
adhoc(or am I reading too much into between the lines)03:16
golinuxadhoc: FYI What's new in this release03:56
golinuxChanges in su03:56
golinux - The behavior of su has changed. Use 'su -' to get root's path or use03:56
golinux   the full path to commands if you use only 'su'.03:56
golinux03:56
golinux - There are several ways to get the old behavior. The easiest is to03:56
golinux   edit /etc/default/su to add the line:03:56
golinux    ALWAYS_SET_PATH yes03:56
golinuxMuch more in the Release Notes http://files.devuan.org/devuan_beowulf/Release_notes.txt03:57
rwpadhoc, https://wiki.debian.org/UsrMerge03:58
meep_____https://oc.cil.li/topic/2226-flatdb-an-sql-implementation/04:42
meep_____Wrong chat sorry04:42
tuxd3vmeep_____, does you like to mine?05:16
meep_____Yes05:16
tuxd3vwhat do you mine usually?05:17
meep_____Ncp+05:42
malade_mentalrwp: gnarface: many information to read after a night of sleep :]09:36
malade_mentalI think I have to debootstrap more packages so the installation is complete09:37
malade_mentaland I'll retry this after having a correct devuan on my physical machine (cause everything i tests was on VM)09:37
openbsdtai123why TED editor is not in the repositories yet? This is X11 though :(12:10
malade_mentalok guys, this just is REALLY sucky.12:17
malade_mentalI netinstall with the last iso, install just OpenSSH server and base system, and choose openrc as init system12:32
malade_mentalthe /bin and /sbin are not symlinks but I have them in the path12:33
malade_mentalif I login into root directly from login only, if I login from my normal user, I don't have the "sbin" directories, and if I "su", I still don't have them.12:35
fsmithredmalade_mental, see the release notes for changes in su and other goodies.12:38
fsmithredyou can use 'su -' or 'su --login' to get root's path.12:39
fsmithredor set it to revert to the old behavior12:39
rrqmalade_mental: you don't get /bin in $PATH ?12:42
openbsdtai123"su -" is the debian increased security. It is crucial for lot of users. Prefer sudo - which allow to increase security instead of using "su" and root account.12:42
rrqmalade_mental: is it that you don't get /bin in $PATH ?12:44
fsmithredthat would be very weird12:44
fsmithrednot sbin makes sense12:44
openbsdtai123fsmithred: during instlaltion it should be sysvinit or openrc, then another dialog: "debian linux" or "unix philosophy".12:46
malade_mentalrrq: sorry I did not saw i have it12:46
malade_mentalyeah only "sbins" are missing12:46
openbsdtai123sbins are necessary to have.12:46
malade_mentalso  there is no usrmerge by default? that's a pretty good point!12:46
fsmithrednot in devuan12:46
malade_mentalsbins is not in the PATH12:46
fsmithredwe fixed usr merge so it's optional12:46
malade_mentalfsmithred: phew =] sorry I was alarmed too fast12:46
malade_mentalhave to it12:47
fsmithredREAD THE RELEASE NOTES12:47
openbsdtai123export can add them.12:47
fsmithredor read 'man su'12:47
openbsdtai123Dont read, ask us... we arethere.12:47
fsmithredyor export12:47
fsmithredor12:47
fsmithredI get tired of typing the solution over and over again when it's right there in the release notes12:47
openbsdtai123"set" will tell you what is the PATH=12:49
openbsdtai123export PATH... can remodify it according to your own needs.12:49
openbsdtai123fsmithred: this is why there is kde and ubuntu.12:49
malade_mentalI have to eat, thanks for your answers ;)12:50
openbsdtai123anytime12:51
openbsdtai123bon appetit12:51
openbsdtai123(bio naturellement ;)12:52
gordonDrogonHm. I see Beowulf gives us 's' in ls output now. Just read the history of that. Interesting and unexpected.15:57
gordonDrogon(as in quotes round 'tricky' filenames)15:58
fsmithredaka tricky quotes around filenames.16:21
nemogordonDrogon: the surprising part for me is that it happens even with "/bin/ls"16:30
nemogordonDrogon: I'm used to having predictable output by doing an explicit path to avoid aliases16:30
gordonDrogonyea, it's an opt-out now. Add -N16:30
nemoit's kind of irritating. since now I get double quoting16:30
nemobut yeah, opt out :/16:30
gordonDrogonlike colours - I never asked nor wanted them, but more shenanigans  needed there.16:31
* gordonDrogon engages grumpy old man mode ...16:31
gordonDrogonhttps://superuser.com/questions/1376351/why-does-ls-wrap-some-filenames-in-single-quotes16:32
nemogordonDrogon: I guess piping is another opt out16:32
nemohm. that's odd16:32
gordonDrogonwell. once upon a time... ls | sort | pr -616:32
nemo/bin/ls | less seems to not have quotes16:32
nemoyet that superuser says piping is one of the reasons they did it16:33
gordonDrogonit's for interactive ls only.16:33
nemo"ls wraps single quotes around filenames with spaces in them, for the purpose of allowing the filenames to be safely copied, used in a script or piped through another command."16:33
nemosure not the implication of that lead sentence16:33
nemobut at least they aren't breaking existing pipes16:33
* gordonDrogon goes back to the ol ddays of BCPL ...16:34
openbsdtai123how to get a directory with ftp?   get -r directory?16:38
gordonDrogonmget16:39
gordonDrogonalthough it's been years since I did that sort of thing. any reason to not use scp or rsync ?16:40
openbsdtai123on the linux machine, only the default machine is installed distant. not even wget... :( so man, ftp has this feature?16:41
openbsdtai123it seems that ftp has no recursive feature.16:44
gordonDrogonno idea. I don't even have ftp installed these days.16:44
openbsdtai123overusing the shinning graphical desktops?16:45
gordonDrogonno - mostly of the stuff I do is command-line, but I moved to scp and rsync some time ago.16:45
openbsdtai123ftp is mostly a dead protocoll. I still use it since it works well with my android and on all machines. Even a Macintosch ;)16:46
gordonDrogonalthough I did ressurect rsh/rcp in the early Raspberry Pi days as encryption was very slow for scp.16:46
gordonDrogonah, you've remonded me, I did have ftp on my old PC - just for my android phone. I've not installed it on the new PC yet. andftp I think I use, so yea, GUI from the phone.16:48
openbsdtai123I just compiled ncftp but I guess this issue with get -R -f dir, "Could not traverse directory: could not parse extended file or directory..."16:48
openbsdtai123gordonDrogon: AndFTP is magic on the phone. Wish I had a N900, this would solve many file exchnage issues and more: Unix.16:49
gordonDrogonI have an n900, but it's broken now.16:50
openbsdtai123wow16:50
openbsdtai123I would be glad to have one, even broken, to fix. Those machines are very rare.16:51
openbsdtai123I have a zaurus 3100 with debian on it.16:51
gordonDrogonit's no use now - not in the UK anyway  - it was only 3G from what I recall.16:55
gordonDrogonI moved from that to android a long while back, but yes, it was fun to ssh into it..16:55
openbsdtai123really? 3G is pretty good. In UK, shall you have only 4G or 5G? lucky cancer with the coming 6G, we will all finish in hostpital with all those magnetic fields.16:57
gordonDrogonwe have 2/3/4g here. but I guess I'm used to 4g all the time now.16:57
gordonDrogonand hurrah, it looks like Debuan Beowulf can read the files off my phone using shotwell and mpt, so I can remove andftp from it and no need to let google steal all myphotos now.16:59
openbsdtai123I try to reduce the usage of phone, bluetooth,... for medias. 5G is for bad software, when we can use better software. FTP or SSH rather than a web or google, ms for streaming... and all modern crap.16:59
gordonDrogon*Dvuan, of-course..16:59
openbsdtai123gordonDrogon: I wish people would think like you and not use google.16:59
gordonDrogongoogl has it's uses - you just need to know how to use it.17:01
gordonDrogonand not rely on it.17:01
gordonDrogonhowever - really - I'm sure stuff like that is best discussed elsewhere.17:01
gordonDrogonHere I'm liking Beowulf more and more...17:02
openbsdtai123I cannot make this completely off Unix of debian and modern devuan. I wont move to beowulf just to loose the pricing of loosing unix. Even less, ... is not there. Once we will have just no "su", "cp", .... no more "cat" ...17:03
nemosay, in debian, is there a way to cleanup the cache after a release upgrade?17:04
nemohm. say, something like "delete any archived version that is not currently installed"17:04
gordonDrogonapt clean17:04
openbsdtai123apt-get clean17:04
nemothanks17:04
nemoum wait17:04
openbsdtai123su - ; apt-get clean   or  sudo apt-get clean for newbies17:04
nemothat deleted everything doh17:04
nemooh well17:04
nemomoar bandwidth burned if I need to reinstall a package ☺17:05
openbsdtai123all is into var cache apt... directory.17:05
nemoopenbsdtai123: yeah, and I've rm'd it before. I was just trying for something more nuanced17:05
nemoonly deleting stuff from old versions17:05
openbsdtai123Maybe there is some way someone...17:08
openbsdtai123someohow.17:08
gordonDrogonwell that's the first time I've ever been able to successfully use linux to suck stuff off my phone. it's just annoying that google use linux but don't actually support it on the desktop...17:10
openbsdtai123gordonDrogon: google support all nasty things for the users... they corrupt linux. they do it well ;) (+) people use it freely, without obliging them.17:11
openbsdtai123what about raspbian to phone?17:12
gordonDrogonsame issue as any linux to phone - you'll not get the drivers for the modem/dsp/gps/sensors/etc.17:12
gordonDrogonif you want a linux phone there is: https://www.pine64.org/pinephone/17:15
openbsdtai123Back to telephony with giant rpi phone : https://www.youtube.com/watch?v=R-2-wPX3G5M17:15
gordonDrogonI'd get one, but there are a few apps I need to run on a real android phone (businessy based, banking, etc.)17:16
openbsdtai123Get a Pandora with Linux, you can do anything with that.17:16
gordonDrogonwell sure - it's possible to bolt modems, audio, etc. to a Pi (or other ARM device), but you don't get the small scale or longer battery life of something designed for it.17:17
openbsdtai123the video is very cool. bit big cell phone ;917:21
gordonDrogonvery big.17:25
openbsdtai123For sure, no way to miss it.17:26
gordonDrogonalthough arguably that can be done with an arduino which would help battery but you might not get such a pretty UI17:26
openbsdtai123the samsung cells can power th e rpi zero17:26
gnarfacehehe19:04
gnarfacenews on slashdot to day about the planet Ceres19:05
gnarfacei wonder if we'll get any spillover traffic19:05
golinuxlink?19:05
gnarfacehttps://science.slashdot.org/story/20/08/11/001234/planet-ceres-is-an-ocean-world-with-sea-water-beneath-surface-mission-finds19:06
gnarface"The dwarf planet Ceres -- long believed to be a barren space rock -- is an ocean world with reservoirs of sea water beneath its surface, the results of a major exploration mission showed on Monday."19:07
golinuxWhoa . . . maybe to -offtopic?19:07
gnarfacewell you asked for the link19:07
gnarfacethat's all i'll paste19:07
golinuxThanks for that.19:07
gnarfacei was just speculating people might start googling ceres and end up interested in devuan19:08
golinuxI know.  Interesting.  Something new for humans to plunder and destroy . .19:09
meep_____I doubt it19:18
meep_____Google & Co only link crappy bloated sites trying to sell you something19:18
meep_____Not homepages and noncommercial sites19:19
meep_____You'd have to use yacy19:22
* fsmithred calls for a beach day on ceres. (Bring a mask!)19:26
furrywolfgiven the lack of atmosphere, I suspect masks, of the spacesuit helmet variety, would be fairly important.  :)19:27
ranixhowdy all, I hate systemd and am considering using devuan for my mdraid fileserver. Anything I should be aware of before just jumping in19:41
ranixdebian user since '9819:42
fsmithredranix, you will probably feel like you came home after being away for five years.19:44
ranixexcellent19:44
fsmithredif you're going to migrate your system from debian to devuan, please read the appropriate migration guide.19:44
gnarfaceranix: it works like debian wheezy, more or less19:44
fsmithredThere are always some tricky parts.19:44
ranixthis is a fresh install19:44
fsmithredcool19:44
ranixI'll get on it as soon as memtest86 gives a clean bill of health19:46
fsmithredhave you used buster?19:46
tuxd3vranix, that could take a while :)19:46
tuxd3vranix, I believe I have a error also, related with bank 0, unfortunatly :(19:47
fsmithredbiab19:48
tuxd3vdoes any one knows a solution to get MCE error exceptions on the OS19:51
tuxd3v?19:51
tuxd3vI know that there are a mcelog program19:51
tuxd3vbut I can't find it in beowulf19:52
systemdletednsmasq dhcp -- I am having a hard time making it work.  dnsmasq dns works fine.  I see from lsof that dnsmasq is listening on port 53, and I've configured dnsmasq.conf to listen for dhcp by enabling the range variable.  Still, client cannot get a lease.19:52
tuxd3vafter some time, my machine freezes on youtube, for example19:52
systemdleteI even set up wireshark to listen for dhcp traffic, and I see the requests from the client19:53
systemdletethey are on the same subnet:  192.168.56.0/2419:53
tuxd3von reboot I get a MCE error related with bank 0, but its very brief I can't read it all, unfortunatly19:53
systemdleteit seems like dnsmasq is not really paying attention to dhcp requests19:53
systemdletenext step will be to strace dnsmasq and see if it is getting hits19:54
tuxd3vsystemdlete, I believe dhcp requests need to be inserted in the dns server..19:54
systemdlete"inserted?"19:54
tuxd3vyep19:54
systemdletewhat does that mean?19:54
tuxd3vat least with bind919:55
systemdletewhat is the meaning of the word "inserted" here?19:55
systemdleteI'm not following that, sorry.19:55
tuxd3vinserted means inserted, by the dhcp server into the dns server19:56
systemdleteoh.  I see now.19:56
systemdletewell, dnsmasq is both a dns AND a dhcp server19:56
systemdleteso there is no additional work there19:56
systemdlete(but I see your point now)19:56
tuxd3vOr you can run a cronjob, into the localhost that updates the server with the machine record..19:56
tuxd3vusually its a good practice to update from localhost a entry in the dns server19:57
systemdleteI'm not using bind.19:58
systemdletejust dnsmasq19:58
meep_____» <ranix> howdy all, I hate systemd and am considering using devuan for my mdraid fileserver. Anything I should be aware of before just jumping in19:58
meep_____try using ZFS instead of dmraid19:58
tuxd3vI don't know about dnsmasq, sorry, maybe been it the 2 things .. maybe it does it automatically19:58
nemowhy does /usr/lib/libreoffice have 260MiB of files according to du -hs after apt remove libreoffice ?19:58
systemdleteit does do it automatically19:58
systemdletenemo:  apt purge?19:58
systemdletemaybe apt autoremove also, idk.  I'm not a devuan expert19:59
meep_____» [10:52:39] <tuxd3v> does any one knows a solution to get MCE error exceptions on the OS20:01
meep_____are MCE's hardware problems?20:01
meep_____Machine-check-error right?20:01
meep_____» [10:54:18] <tuxd3v> on reboot I get a MCE error related with bank 0, but its very brief I can't read it all, unfortunatly20:01
meep_____sounds like a problem with a ram stick?20:01
tuxd3vmeep_____, indeed it sounds20:02
meep_____If you can't get the full error try reseating bank020:02
systemdleteblow some canned air also maybe?20:02
meep_____Clean the contacts with isoprope. The faster these circuits get the more sensitive they get20:03
meep_____Modern ram sticks running with 4GHz+ clocks20:03
tuxd3vmeep_____, indeed you gave me a Idea, from cleaning it ;)20:03
tuxd3vfrom -> for20:04
meep_____Speaking of, are there any supercomputing arches designed with only ~30MHz interconnects?20:04
meep_____Not tiny home micros20:04
meep_____But something massively parallel20:04
meep_____I've seen arches based on 4 m68k cpus but that wasn't scalable beyond 4 cpus20:05
meep_____Luna88k20:05
meep_____From OMRON20:05
meep_____I ask because if the interconnects are sub 30MHz that's simple enough to be able to manufacture boards and components at home with blank PCBs, etching solvent, and a soldering gun20:06
tuxd3vmeep_____, my problem is worst than that, since they are ddr2 800Mhz... a very rare beast nowadays, for sticks bigger than 1GB..20:06
meep_____Oh lucky you with 800MHz ddr2. My older laptop is 666MHz ddr220:07
tuxd3vbut you have tons of ddr2 at 666Mhz( were the most produced ), ddr2 800Mhz is a very rare beast20:07
tuxd3vthe prices are very high, for this memories..20:08
tuxd3vand my Desktop Xeon, likes the 800Mhz version a lot :)20:09
nemosystemdlete: I did autoremove20:11
nemosystemdlete: lemme try purge20:11
nemosystemdlete: hm.  maybe something else depends on all those libs. checking20:12
nemosystemdlete: oh. weird. libreoffice-core is still installed20:13
systemdletedid you install that one separately?  maybe apt thinks you meant that to stay20:14
nemosystemdlete: naw. this was just a basic devuan guided install. all templated20:14
nemosystemdlete: I'm just tidying it up a little for a newly duped qemu image with some stuff I don't need20:14
systemdleteok, just checking.  I've run into a bit like that here and there20:14
nemoafter removing core /usr/lib/libreoffice still has 77MiB of stuff. heh20:15
systemdletemaybe config files?20:15
systemdletetry finding out who owns those files, if anyone20:15
nemorunning autoremove now20:15
nemoprobably just need cleanup again20:15
nemoheh. nope20:15
nemoalthough autoremove had this fun line20:15
nemo"/usr/lib/libreoffice not empty so not removed"20:16
tuxd3vnemo... if you have a file open, and you remove it, the space is not freed from the Filesystem until you reboot20:16
nemo76MiB of packages20:16
nemotuxd3v: whaaaaat20:16
nemotuxd3v: no!20:16
tuxd3vnemo, yup20:16
nemotuxd3v: I mean. the space is still there. and you have a file handle20:16
systemdletefind /usr/lib/libreoffice20:16
nemobut it won't sho up in du -hs20:16
nemo*show up20:16
nemowhich is what I'm running20:16
tuxd3vhas the filesystem always cound the open file descriptors to that file, and if it is bigger than zero, space is not released20:17
systemdletewhat's still there actually?20:17
nemotuxd3v: sure. I'm just saying. it won't show up in du -hs once you type rm20:17
nemoeven if I have it open20:17
nemoand that's trivially testable20:17
tuxd3vnemo, yes it will :)20:17
tuxd3vyou need to truncate those files or... reboot20:18
tuxd3vthe space of a file is released from the fylesystem once its nr of file descriptior open reaches zero20:18
nemo$ du -hs20:18
nemo6.4G.20:18
nemo$ rm foo20:19
nemo$ du -hs20:19
nemo19M.20:19
nemothis is with foo open in vim20:19
nemothe space is on disc20:19
nemoI can write to it20:19
nemobut it's no longer visible in the filesystem20:19
tuxd3vtill that moment, the space it ocupied and its true you can't see the files, but the filesystem has the space ocupied :)20:19
nemoit's just available through vim's open file handle20:19
nemotuxd3v: right... that was my point above about du -hs etc20:19
nemotuxd3v: otherwise what you're saying is true, but also totally irrelevant to what's going on here ☺20:19
tuxd3vbut 'du' shows the space ocupied20:20
nemo...20:20
nemoclearly not20:20
tuxd3vho yes20:20
nemojust tested. after rm, the file vanished from du20:20
nemobecause it's running standard file system operations20:20
nemotuxd3v: I think you are confusing df and du20:20
tuxd3vdoing a ls -l for example the files are not there, they don't show up20:20
nemoyes. ls and du work same way. df does not20:21
tuxd3vone simple trick..20:21
nemothus my simple test above that I pasted with du -hs (6.4 gigs) rm foo, du -hs, 19 megs20:21
tuxd3vif you do a du on a folder, and it shows its not empty20:21
nemoall while I had vim -b foo20:21
tuxd3vand you list the content of that folder and is nothing there...20:21
tuxd3vyou have unproperly removed files20:21
tuxd3vyou need to reboot, or simply truncate them..20:22
nemook. so again20:22
nemowhat does that have to do with du -hs of /usr/lib/libreoffice still showing 76MiB of files. which are real files, and unsurprisingly show up for find etc20:22
nemotuxd3v: what you're saying is true, but also irrelevant to this cleanup I'm doing ☺20:22
nemoah. uno-libs3 package20:23
* nemo removes that one too20:23
nemofinally. with that removed, no more /usr/lib/libreoffice20:24
systemdleteyou could have used find's output to ask apt who owned them... but it sounds like you have it solved now.20:28
nemosystemdlete: yeah, although I looked up how to do it on debian and said dpkg -S20:29
nemosystemdlete: or apt-file which I didn't have installed20:29
systemdletewell.  there you go!20:31
systemdletetil20:31
meep_____How can I configure Beowulf as to not spin up the disks when resuming from suspend power state?20:31
ranixtuxd3v: MCEs should log in /var/log/syslog20:31
ranixthere's mcelog(8)20:32
meep_____Whenever my computer resumes it spins up the CD-ROM tray20:34
meep_____It wastes a lot of power doing this and it's annoying having the computer vibrate when it opens20:34
gnarfacemeep_____: the cdrom hardware might do that automatically when it powers up20:35
gnarfacemeep_____: see if it's in /etc/fstab though20:35
meep_____It's in fstab, but with user,noauto20:35
gnarfacejust comment it out20:36
gnarfacesee if that helps20:36
gnarfacelike i said though i think that's something the physical drive's internal firmware is doing20:37
tuxd3v<ranix> there's mcelog(8), yeah but I don't know the package that contains it :(20:37
ranixapt-cache search mcelog20:38
ranixcollectd-core?20:38
ranixthat was from a debian system though20:38
tuxd3vhttps://paste.debian.net/1160060/20:50
tuxd3vI believe this has to do with the Ram bank 0, :(20:50
gnarfaceseems like it's time to memetest it20:51
gnarfacememtest*20:51
gnarfaceif it comes up with no errors, enable the optional bit-fade test20:51
tuxd3vgnarface, yeah, I am afraid of it since it takes a lot of time to end :/20:52
ranixlol memetest20:52
gnarfacetuxd3v: it will probably take a good day or two, yes20:52
tuxd3vgnarface, the bit-fade is something available from the menu?20:53
tuxd3vI mean easy to discover :)20:53
gnarfacetuxd3v: yea there should be an optional tests menu somewhere in there with 2 or 3 tests that aren't in the default set20:53
gnarfacetuxd3v: (because they take MUCH longer)20:53
gnarfacetuxd3v: (but they're useful for tracking some of the harder to diagnose memory faults)20:53
tuxd3vgnarface, yeah I am afraid I will need to be without this pc for some time( doing memtest ) :/20:54
gnarfaceyea there's no way around it20:54
gnarfacewell, assuming the fast tests didn't show any faults, that is20:54
gnarfaceif you haven't tried though, it might be worth it to just test each bank individually with the fast test20:55
gnarfacefast tests20:55
gnarface*20:55
tuxd3vgnarface, ranix thanks, I will be forced to do a memtest(no other option..)20:56
gnarfacetuxd3v: well, probably not, but this type of error could also be caused by temporary hardware failure due to overheating.  if it only happens under load, make sure you at least checked all your case fans first20:58
gnarfacetuxd3v: (though, that is the exact type of temporary failure that can cause a very similar looking permanent failure)20:58
tuxd3vgnarface, thanks for that option, indeed it can be..21:08
tuxd3vBut I am a bit sceptical, because I compile several kernels, in this machine, and it heats up the 4 cpus..21:09
gnarfacewell a actual physical fault in the memory i would expect to also be excited by heat or a kernel build21:10
gnarfacebecause i would expect a kernel build to fill all the ram21:11
gnarfacei guess unless you have like, a really large amount of it21:11
tuxd3vyeah , I also think of that option, the kernel takes some because I am compiling with -jnr_cores in parallel21:12
tuxd3vmaybe the kernel is a more randomized thing, and the codecs for video, are something less random, and more frequently hit certain areas..maybe, don't know..21:14
meep_____Maybe your cpu is going something sneaky like SMT21:15
meep_____Presenting you fake cores21:15
gnarfacei guess i don't know.  i get these type of errors from the cpu cache though sometimes when the machine is overheating: https://paste.debian.net/1160062/21:15
gnarface(usually this is a sign i need to clean the dust out of my fans)21:15
meep_____gnarface: do you have ECC memory?21:16
gnarfaceno21:16
gnarfaceer21:16
gnarfaceactually the cpu cache may be ecc21:16
gnarfacebut the system ram is not21:16
ranixinstalling the base system21:21
ranixtuxd3v: if you still have problems memtest86+ doesn't expose try SAT21:22
tuxd3vgnarface, yeah the caches have at least capabilities to detect parity errors21:22
ranixhttps://github.com/stressapptest/stressapptest21:22
tuxd3vI think wen I shutdown the system, I will cleanup the bank0 ram contacts, then experiment a bit with it, if it fails( which I believe is will), then no other option left :/ I will go to memtest fast-first, then using the gnarface proposed option to go deep inspecting..21:25
tuxd3vBut I don't rule out the heating problem... because I am running a Xeon CPU, with a normal coller :/ yeah,( at the time I haven't found a exact match)21:26
gnarfacetuxd3v: if you're not very unlucky, then any permanent damage will be isolated to just one ram bank and you can simply remove that one and run on half ram temporarily21:26
gnarfacetuxd3v: i've never had contact corrosion issues with ram banks but it's been a problem for the first 3 nintendos so i guess it's a theoretical problem21:26
tuxd3vgnarface, thanks a lot , its yet a deepr way to locate the problem for sure :)21:27
gnarfacetheoretical possibility* i mean21:27
malade_mentalopenbsdtai123: (bio du jardin!)21:35
meep_____How much dedotated wam do I need for serva?21:42
meep_____gnarface: MCEs could also just mean your CPU or chipset is worn out21:42
meep_____A decade of being plugged directly into the wall without any line conditioner21:43
meep_____My workstation from 2006 kept popping up those errors with more frequency in it's last two years of service21:44
gnarfaceMetaYan: yea, a distinct possibility, but these ones do seem to coincide with high heat21:44
gnarfacemeep_____: ^this was for you21:44
openbsdtai123malade_mental: du jardin. Ca pousse en ce moment? pas d'eau21:44
gnarfaceMetaYan: (ignore me, sorry, tab-completion error)21:44
meep_____Until one day the text on the screen was just replaced with random gibberish, sometimes weird distortions on analog VGA-out21:45
meep_____Even when using known-good ram and reseating the cpu21:45
gnarfacemeep_____: that's usually a symptom of *video* ram failure21:45
gnarfacemeep_____: (i see this happen almost like clockwork to unmodified nvidia laptops)21:45
meep_____And then 2 days later the system would just never complete a post21:45
* tuxd3v thinks that his CPU is properly working, and crossing fingers..because he loves his Xeon for kernel compilations..21:45
gnarfacetuxd3v: your error specifically referenced bank 0 though, which i assume refers to motherboard ram bank 021:46
meep_____However I really can't complain. Over a decade of service to a single set of parts, not upgrades except for more ram and a second hard drive and good cleanings over the years21:46
meep_____Oh, and added USB3 pcie expansion card21:47
meep_____gnarface: what was for me?21:47
gnarfacemeep_____: "yea, a distinct possibility, but these ones do seem to coincide with high heat"21:48
meep_____Oh i'm not the one with MCEs21:48
meep_____I'm just helping21:48
tuxd3vgnarface, yeah, I also thinks that, also I found a kernel bug, with the kernel being tainted when he tried to access a page...its for sure a memory region in bank 0 Ram sticks21:48
gnarfacetuxd3v: i think that there is a kernel feature you can use to pass dead system ram blocks to the memory mapper at boot time through the kernel command-line21:49
gnarfacetuxd3v: if you have dead blocks of ram but it's not many and it's not the video ram, you can probably retain use of the rest of it21:50
gnarfacetuxd3v: i don't know exactly how to compile that list though, i don't know of a ram equivalent of "badblocks" unless memtest86+ can be run from userspace21:50
tuxd3vgnarface, I believe that option exists yes, I will try to find it, if memtest detects the region, and now that you spoke about video memory, indeed the system reserves some megabytes for the graphics, maybe, maybe its that region, because when I experience the problem is when on youtube, its ramdon in time , some time after 1 hour sometime after 10 minutes21:53
tuxd3vwhat should be the package for memtest{ memtest86 | memtest86+ }?21:55
tuxd3vI believe memtest86+ :)21:55
tuxd3vI will run the fast version now :)21:56
gnarfacetuxd3v: yea i'm not sure.  i think "86+" is the newer one21:56
tuxd3vgnarface, thanks21:56
tuxd3vI hope it should take some hours at least to complete, I only have 4 GB( 2x2GB ), but still it will have to iterate for each adress to test :)21:57
nemo32 gigs here ☺21:58
gnarfacetuxd3v: probably in the neighborhood of 5-10 hours for the base default test sets, maybe 1-2 days for the extended tests though21:58
gnarfacewoops, i was too slow21:58
gnarfacei remember it would finish in a few hours for the first 8 tests, then there were 2 extended tests that took 3x as long as the entire rest of the set together21:59
gnarfacebut those were the tests that sometimes showed the more rare errors21:59
gnarfaceit was more useful back in the early Athlon days when ram tolerances could be as problematic as hardware failure22:00
meep_____» [12:50:31] <gnarface> tuxd3v: i think that there is a kernel feature you can use to pass dead system ram blocks to the memory mapper at boot time through the kernel command-line22:07
meep_____»22:07
meep_____not really. There's an option to disable all ram past a location22:07
meep_____But disable specific bits22:07
gnarfaceoh, i could have sworn there was a way to pass a bad block list, but the problem was there's not much room in the max byte length of the kernel command-line22:08
gnarfacei just remember looking into it to try to get a nvidia card with partially burned-out ram limping along again for a few more years and found out it was only for system ram22:11
meep_____No22:57
meep_____Not badblocks for kernel memory22:57
meep_____Badblack is for storage22:57
meep_____You can only disable all memory past a specified address. Which won't help you if your bad ram appears in low memory22:58
ranixit's possible to make a badblocks map for memory22:58
meep_____Or spread out all over22:58
meep_____ranix: howso then?22:58
meep_____Also, is that new?22:59
ranixit's not yet done afaik22:59
ranixwell not that new but the project I'm aware of was terminated22:59
ranixit was working but unreleased22:59
meep_____I guess all the people smart enough to program the kernel can afford to buy working ram sticks if theirs goes bad23:00
ranixa certain company that will remain unnamed thought it was more expensive to reboot all their machines to apply a kernel patch than it was to replace bad ram23:00
meep_____ranix: I don't think that's a bad decision to make23:01
ranixI think it was a bad decision to make but it wasn't necessarily completely retarded23:01
meep_____Firstly running high-uptime servers without ECC is... Well I just think that everybody should be running ECC and it's pretty stupid of the hardware industry to make such a simple and cheap reliability factor so much needlessly more expensive23:02
ranixecc ram still fails23:02
ranixrow and column and chip failures23:02
ranixthose dimms don't really need to be replaced if you blacklist the affected memory regions in a kernel badblocks page23:02
meep_____And secondly one some blocks in your ram starts going bad, more blocks usually start going bad exponentially23:02
ranixnot really23:03
ranixit's pretty common to have a row failure that doesn't affect the rest of the chip23:03
ranixor a chip failure that's isolated to one chip on the dimm23:03
ranixit's not fatal if it's not the one dedicated to checksums23:03
meep_____It's possible yes, but usually the chips on a stick are part of the same batch23:03
meep_____Meaning they are going to fail at similar times23:04
ranixthat's not what I believe observed23:04
ranixI*23:04
ranixbut I don't have proof anymore23:04
ranixI did at one time23:04
meep_____Maybe you had a different experience23:04
meep_____I only buy memmory23:04
meep_____I only buy memory from Samsung and Corsair23:05
meep_____Because those are the only vendors I trust to not skimp out of DRAM manufacture process23:05
ranixidk if my dimms were even manufactured outside the company23:05
ranixwe might have just sourced components23:05
meep_____I used to buy ram from all over23:05
meep_____I've got an old ddr2 stick from a company that had a part in the manufacture of commodore products23:06
ranixcool23:06
ranixI still have a commodore23:06
ranixand an atari st23:06
meep_____Ha23:09
meep_____Memory used to come from all over23:10
meep_____I've got here a ram stick made by ProMOS technologies in Malaysia23:10
meep_____And DDR1 ramstick here made by AMD with a big MADE IN USA on the back23:10
meep_____Didn't know AMD used to make DRAM chips23:11
meep_____"Rev0"23:11
meep_____(C) 199323:11
meep_____http://datasheet.octopart.com/AM29F016B-150EC-AMD-datasheet-11784513.pdf23:13
meep_____The thing still works23:13
meep_____Actually wait23:14
meep_____I don't think this is ram23:14
meep_____It's just on a board designed to fit into a ram socket23:15
ranixamd did make ssds for awhile23:16
meep_____Well this looks like an SSD from 1993 designed to reuse DRAM connector sockets23:16

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