IRC channel logs


back to list of logs

<g_bor>you can try make the generic pot writeable, and see if the problem persists...
<ngz>Indeed. It doesn't seem generated.
<g_bor>I see that a file handler is opened for writing here, but can't see the value of fpath...
<ngz>Oddly, building it from the repository doesn't generate this error.
<ngz>It seems specific to the package definition.
<g_bor>This is not so surprising, the tarballs often have more restrictive permissions set.
<g_bor>You can find some examples in our packages already.
<ngz>But I use git-fetch
<ngz>So there is no tarball.
<g_bor>and you can build it from the git-checkout created in the store, or from a clone?
<ngz>from a clone
<ngz>But the commit is the same.
<ngz>Could it be related to CMAKE_BUILD_TYPE?
<ngz>I tried Release in the clone, but I think Guix default is RelWithDebInfo
<rekado>ngz: you need to make the checkout writable
<ngz>The whole repo has a mode of 644
<ngz>But it is not sufficient because someone else (guix-builder) tries to write files, not me. Am I right?
<ngz>That would explain why I can build the software without problem, but cannot from Guix.
<g_bor>rekado: do you have an example in mind where we do this already?
<ngz>I found something like this in coq.scm
<ngz>OK. It worked. Thank you. I lost my hair on this one.
<g_bor>ngz: yw! Nice that it works now :)
<g_bor>Good night, I have to go now.
<ngz>Good night g_bor[m]!
<dongcarl>Going to show off Guix to colleagues tomorrow, anything in particular that you guys think I should touch on?
<wednesday>guix system init is giving me the same error as this guy now, guixsd just dosnt want me to install it
<wednesday>OK my issue in that case was guix system init didnt like the fact that I already had /mnt/var/guix and/or /mnt/gnu, makes it die with that error, someone should report that bug ;)
<wednesday>after a little test the problem is not exactly what I said above, guix system init just crashes if /mnt/var/guix is a btrfs subvolume it seems, it was fine with /mnt/gnu being one
<Copenhagen_Bram>my ~/.guix-profile/etc/ssl/certs/ directory is empty, how do I fix that?
<Copenhagen_Bram>oh, installing nss-certs seems to have fixed it, although I thought I had installed nss-certs before
<g_bor>hello guix!
<g_bor>I just had a look at #23118.
<g_bor>I don't see these duplicates in the environment variables any more.
<g_bor>Anyone still experiencing the issue?
<efraim>I seem to have a working perl6 and build-system, should I package some more libraries for more testing or just submit it as-is for now
<roptat>hi guix!
<Richard[m]>hi, is the gtk+-2.24.32 build failing for anyone else or just me?
<jlicht>hey guix!
<wednesday>Does anyone here have experience of having /gnu and /var/guix on btrfs subvolumes? having /var/guix on one causes guix system init to crash, but getting past that error, when I boot I get grub complaining about missing /store/stuff-linux-libre/bzImage
<roptat>wednesday, I don't think it's supported yet :/
<roptat>the kernel is in /gnu/store, so on the btrfs partition. Does grub support that?
<wednesday>roptat: well I used to have separate boot subvolumes on gentoo I believe(some years ago), don't entirely remember, but I think in some mailing lists I think I saw someone with a separate /gnu subvolume
<roptat>mh... you could "insmod btrfs"
<roptat>not sure how to do that with guix though
<wednesday>also grub says /store when it gets the error instead of /gnu/store, is that typical?
<roptat>yes, that's because you have a separate partition for /gnu, so there is no gnu directory on that partition
<roptat>the first directory is store directly
<wednesday>ah makes sense, but in the same menu entry both --system and --load refer to /gnu/store, but initrd and linux reffer to /store, I assume that's how it's meant to be
<wednesday>even if its not changing the entry doesn't fix anything so I may just have to abandon the /gnu subvolume
<roptat>but even so, I don't think guix can start the system properly
<roptat>I don't know
<wednesday>i'll try keeping my /var/guix on a subvolume and put /gnu on root and see if that'll do anything
<nckx>Subvolumes aren't partitions. If /gnu/store is on a subvolume, Guix putting /store in grub.cfg is wrong and buggy.
<nckx>So no, that's not normal.
<roptat>ah, ok
<nckx>I can't afford to break a machine right now, but I'm pretty sure GRUB still sees /gnu/store. In which case having a symlink /store → /gnu/store would allow you to have /gnu on a subvolume, but it's ugly as hell. Also just an assumption that nothing else is broken too :-)
<wednesday>putting /gnu and /var/guix in the root subvolume doesn't seem to fix anything, get same error about needing to load the kernel first, would I need to put my store on an ext partition or something nckx?
<kmicu>wednesday: there is an issue about that on bugtracker. Currently Guix generates wrong paths for btrfs subvolumes.
<nckx>Stilll guessing: store-mount-point in bootloader/grub.scm is probably "/gnu" when in the case of btrfs subvols it should be "/". Probably an easy fix™.
<nckx>kmicu: Oh cool. Link handy by any chance?
<kmicu>Let me check but grepping btrfs in the ml archive will be faster xD
<kmicu>cbaines: ^^
*nckx wishes mu4e were fast enough that never deleting mail were actually an option.
<kmicu>(Cuz cbaines even had some patches for subvolumes handling)
<wednesday>i've just found this, which might actually help
<wednesday>i'll have take a look
<kmicu>(Generally you can inspect what guix generated for your config and you will see that in case of subvolumes it generates garbage paths.)
<jonsger>Sleep_Walker: :)
<Sleep_Walker>jonsger: \o/
<jonsger>don't know why it went through this time, maybe because it was submitted by a bot :P
<wednesday> thats it on the bug tracker
<wednesday>kmicu: I can get partway through booting with /gnu on my root subvolume, by adding the subvolume name to the start of the paths in the grub menu options. I also have to add insmod all_video to get graphics, but I end up with a kernel panic it seems
<nckx>The (wrong) path is probably also used in early user space. I wonder how. Is it… trying to mount subvolumes?
<kmicu>wednesday: iirc there are more places to fix than only grub config. For example mounting logic is also incorrect. I recommend installing stock config and then play in a VM to fix btrfs subvolumes otherwise installing Guix System is very frustrating process with a long feedback loop after each adjustment.
<kmicu>(Don’t ask me how I know that xD)
***atw` is now known as atw
<nckx>wednesday: Have you tried my stab-in-the-dark symlink kluge?
<wednesday>nckx: well right now i'm not using the store subvolume, I started trying with just a root subvolume, the only other subvolume it should be mounting after boot is /home
<nckx>wednesday: Mounting? So you'r /home subvolume isn't actually at /home?
<nckx>*your, wtf.
<wednesday>it should be getting mounted at /home
<nckx>Any reason you're doing it that way and not just making /home itself a subvolume?
<nckx>Because that works.
<nckx>(The other might, just never tried it.)
<wednesday>the way i've always done things is, a btrfs partition, with root subvolmes, home subvolume, and whatever else I want, that get mounted at boot
<wednesday>when I've done it with other systems it lets me easilly swap out parts of the fs, like the whole root, or share things like home across other os's installed
<nckx>So something like ‘mount /@my-home-subvol /home’?
<wednesday>nah it's like mount -o subvol=home /dev/sda3 /home
<nckx>Yeah, that's probably beyond Guix's limited knowledge of the world for now :-(
<wednesday>well in file-system it has mount options, and i've stuck the subvol= stuff in there which afaik should work
<kmicu>What is in config.scm is not important. It can look good but Guix code generates invalid configuration for subvolumes.
<wednesday>kmicu: so it has trouble mounting subvolumes after boot?
<nckx>In any case, it shouldn't panic on /home.
<wednesday>because editing the menuentry before I boot can get me to a point where I kernel panic, but I can't see enough of the message to gleam much from it
<kmicu>I was inspecting that logic line by line. Today I only remember that it generates wrong paths (for GRUB and fstab) and mapped-devices doesn’t understand subvolume dependencies.
<wednesday>Just before the kernel panic I do see the words btrfs pop up for a second
<wednesday>I'll try see if it's done something stupid in my fstab
<kmicu>After that I installed regular btrfs cuz editing files in grub shell or guile repl is PITA 😺
<wednesday>kmicu: is fstab meant to be in a store? I found it in one and it's empty ha
<kmicu>wednesday: there is a file for mounting (a guile one iirc) you can find it by generating new config. I don’t remember its name.
<kmicu>I basically inspected any new file created by guix system init /confi /mnt.
<wednesday>your root mountpoint is meant to be in the fstab right? I found my fstab and I see home in there but no root
<kmicu>On my Guix System laptop in fstab I only see some systemd/cgroups things and /gnu/store entry.
<nckx>sneek: later tell wednesday: ‘your root mountpoint is meant to be in the fstab right?’ N.
<nckx>sneek_: later tell wednesday: ‘your root mountpoint is meant to be in the fstab right?’ No.
<nckx>Evil shapeshiftin' robot.
<nckx>Evil shapeshiftin' robot ignoring me.
<kmicu>wednesday: in Guix land expect Guile/Shephard stuff.
<nckx>wednesday: Oh. You were gone (or so my client said) so I left a message.
<wednesday>yea last thing i saw is what i said about is root meant to be in the fstab ha
<kmicu>The best way to figure out what really happens is to diff filesystem after config changes or to read Guix source code. (Both are painful on live half-booted OS).
<nckx>Anyway. No need for / in fstab. That's for ‘mount -a’, and guess what Guix doesn't use.
<wednesday>kmicu: well right now I got no system to boot into ha, so it's either try fix this or give up and use ext for now, which will be a pain for me later on
<kmicu>wednesday: regular btrfs also works. But I understand. Only want to point out that fixing the issue is not trivial and will take days when performed on half-booted OS. So you can have a less frustrating path with installing a working config, fixing subvolumes on it (in a VM) and then reinstalling the system. (Personally, I’m doing exactly that cuz I also have a NixOS box with subvolumes setup similar to yours and wanted to recreate it on Guix System.).
<kmicu>In other words: I was in the same place, I gave up after 2 days xP
<jonsger> this is maybe an interesting repo, as Efraim uses btrfs
<wednesday>so your saying just have it with btrfs without the subvolumes? i'll probably end up doing that, but that makes me sad inside, maybe one day I can fix it ha
<wednesday>yea he seems to be doing what kmicu suggested, doesn't look like he is using subvolumes
<kmicu>wednesday: With a delcarative configuration we can reinstall OS in 5 minutes. That’s my plan - fix subvolumes and then reinstall OS in my desired file system layout.
<efraim>I didn't bother with subvolumes, I just wanted the compression and checksumming
<wednesday>it's been about a year since i've been on linux, been on bsds for a while, but back when I was on gentoo for several years I had a setup like this, so I thought it shouldnt be too hard to redo heh
<nckx>I can confirm that Guix System on btrfs without (early-mounted) subvolumes is a breeze :-)
<nckx>I've never had a compressed and heavily used btrfs last more than a year before panicking, but that's not Guix's doing.
*jonsger is happy that Guix doesn't use btrfs as default file system :P
<nckx>The checksums help me sleep slightly better than their history keeps me awake. It's a trade-off ;-)
<wednesday>I remember having one issue with btrfs some years ago but btrfs restore and such saved the dat
<wednesday>well I removed all the subvolumes and now I got a bootable system
<nckx> /znc rehash
<nckx>Oops. Sorry.
<nckx>wednesday: Huzzah! It's something™!
<wednesday>now I need to figure out the other issue I was trying to solve yesterday, my usb 3 not working with immou on in the bios, and usb 2 not working with it off
<kmicu>wednesday: it’s not too difficult just core devs don’t use subvolumes so it’s up to us to contribute that.
*kmicu loves subvolumes cuz they are super handy during debugging or backups so rest assured that use-case will be supported in Guix System.
<jonsger>nckx: what is the reason to hardcode the name in the source URI?
<bavier>every time I 'guix pull' I have to rebuild libreoffice, and it doesn't honor --cores :(
*bavier wonders why the --with-parallelism flag is no longer effective ':/
<kmicu>bavier: if you could paste libreoffice’s hash we could check whether its substitute is available on the build farm. Maybe it’s not ready or some local override demands different hash.
<bavier>kmicu: good point, 'guix weather --coverage' didn't list libreoffice as missing. I don't have any local changes
<bavier>"building /gnu/store/q93l80yr25g6m9rk1phll4pvxhk2s9vh-libreoffice-"
***ng0_ is now known as ng0
<kmicu>bavier: there is no such substitute on the farm.
<bavier>I'll let this build go through and then try to see why --with-parallelism is having no effect
<bavier>ha, oops
<kmicu>So I assume weather gives wrong status.
*bavier forgot to pass --cores...
<kmicu>Maybe recent update broke libreoffice and weather doesn’t handle broken builds correctly. 🤷
<bavier>needs some investigation, yeah
<nckx>jonsger: Using NAME there is just parametrisation for the sake of it, it serves no useful purpose (and I was as guilty of it in the past as anyone :-). I remove it when I find it. IIRC there was also some consensus on the ML but my duckduckfoo sucks.
<jonsger>oke, I understand. Though I wouldn't mass change it...
<nckx><kmicu> it’s not too difficult just core devs don’t use subvolumes → exactly.
<nckx>jonsger: I ‘fix’ it when I come across it, hence the pairing with updates (and if they're on their own it just means the update broke). I agree that we shouldn't mass-sed anything.
<kmicu>Is there a hydra interface for Cuirass doesn’t have searching. 🤷
<kmicu>bavier: yeah, it’s broken
<kmicu>(Though it has different hash so dunno, maybe you could have better luck.)
<bavier>kmicu: it's on the 'install' phase for me right now
<g_bor>hello guix!
<g_bor>hello guix!
<g_bor>I've noticed that a commit caused a lot of rebuilds on berlin. What do we know about it?
<g_bor>I believe it was the sbcl update, guix refresh says it's within the threshold.
<g_bor>(149 dependent packages)
<bob20190227131>join #freenode
<g_bor>hello guix!
<nckx>g_bor: How many rebuilds?
<g_bor>I have a lagging connection now.
<g_bor>nckx: about 1000 derivation.
<g_bor>Is that ok?
<efraim>Sbcl is ok
<g_bor>ok, fine.
*vagrantc has found next browser to be rather unstable
<vagrantc>gets locked up and unresponsive to keyboard ... anyone else using it? with/without problems?
<jackhill>vagrantc: yes, for me, it does that as soon as I try to navigate off the homepage.
<vagrantc>when it works, i'm liking it a lot ... but...
<jackhill>I haven't had the time to do any bug hunting though. And there is still thy missing references bug (so next-webkit-gtk got collected the last time I gc'd)
<jackhill>vagrantc: interesting. I wonder what is different about my setup that is causing it to behave worse for me. Maybe that's a clue.
<jackhill>As far as I know, I'm running a fairly vanilla Guix System with GNOME desktop.
<vagrantc>i'm running in i3, but nothing too novel otherwise
<vagrantc>the worst is it's unpredictible
***renich_ is now known as renich
<bob20190227131>How safe is it to use guix?
<bob20190227131>Pretty much thinking of using it because Guile and Lisp library management is a mess.
<jackhill>bob20190227131: Hi! What do you mean by safe?
<bob20190227131>jackhill: Have there so far been instances of malicious definitions or inserts like have happened with say... npm.
<jackhill>bob20190227131: I don't know, but a difference between Guix (and other GNU/Linux distributions like Parabola and Trisquel) and self-upload repositories like npm, rubygems, pypi is that changes to Guix go through a review process.
<nckx>bob20190227131: Not that we know of. And people do pay attention to commit logs, although we don't have a formal process for reviewing commits direct to master (as opposed to those posted to the ML first).
<bob20190227131>I see.
***ChanServ sets mode: +o nckx
***ChanServ sets mode: -o nckx
***ChanServ changes topic to 'GNU Guix | | videos: | bugs and patches: | paste: | Guix in high-performance computing: | This channel is logged:'
*nckx removed ‘FOSDEM is coming!’. ^
***vagrantc_ is now known as vagrantc
<ebrasca>Hi my icecat display rectancles instead of text in my guixsd.
<apteryx>ebrasca: I had that issue before. It may have to do with glibc-locales. Try synchronizing your system and profile (guix system reconfigure your_config.scm and guix package -u .)
<ebrasca>Do I need to add "." after guix package -u ?
<rekado>ebrasca: “-u” takes an optional argument, a regular expression. You only need the argument when “-u” is followed by other options.
<ebrasca>I have done it and icecat is broken
<ebrasca>Why if I repeat guix system reconfigure your_config.scm it downgrade my system?
<ebrasca>Why you can't make it easy to update like in other systems (gentoo , debian , arch , parabola ...) ?
<rekado>The state that is instantiated depends on the version of Guix you use.
<rekado>if you use an old version of Guix you will get a system that corresponds to the instantiation using that old Guix.
<ebrasca>Why it can say someting like "Noting to update"
<ebrasca>It don't make any sence
<nckx>ebrasca: Easy.
<rekado>I think you are misunderstanding.
<rekado>“reconfigure” does not deal in terms of updates.
<rekado>it instantiates a system configuration.
<ebrasca>How to update?
<rekado>and then switches the current system to that instantiated configuration.
*rekado has to go
<ebrasca>How to update my system and how to add new packages to my config.scm file?
<ng0>is this on GuixSD or Guix on another system?
<nckx>ebrasca: Updating is as easy as: guix pull && sudo -E guix system reconfigure
<ng0>new packages are added to the packages section, or you install them to the user profile. I only ever added window managers and things like that to the system packages
<nckx>This is exactly like in the other distributions you mentioned (apt-get update && apt something).
<apteryx>ebrasca: adding new packages is done per user with 'guix package --install'
<ebrasca>If I like to install my system in other PC with all packages
<ng0>for example in netbsd one way I can update I do a cvs update of the sources, run, do the manual upgrade steps after the build,.. this is "reconfigure + guix pull".. and the user package updates via pkgsrc are "guix pull (as user) and guix package -u"
<ebrasca>How to make it if I don't have updated my config.scm
<apteryx>ebrasca: then define a manifest of your packages, then issue 'guix package -m your_manifest.scm' on those other systems.
<apteryx>the Guix info manual covers manifests, I believe.
<ebrasca>nckx: With your steps it tries to downgrade my sbcl...
<nckx>ebrasca: I doubt that. From/to which versions?
<ebrasca>nckx: 1.4.16 to 1.4.13 ...
<nckx>Which commits do ‘guix describe’ and ‘sudo -E guix describe’ display?
<ebrasca>ebrasca@Ebrasa-PC:~% guix describe guix describe: error: failed to determine origin
<ebrasca>ebrasca@Ebrasa-PC:~% sudo -E guix describe Password: guix describe: error: failed to determine origin
<ebrasca>nckx: ^
<nckx>Cool. And yet ‘guix pull’ runs without error?
<nckx>What does ‘which guix’ say?
<ng0>what about guix --version ?
<ebrasca>guix (GNU Guix) 0.16.0-10.2637cfd
<nckx>ebrasca: Do you have /home/nckx/.config/guix/current/bin/guix?
<ng0>hm, that's short
<ng0>or is this the new format now?
<ebrasca>nckx: I don't have nckx
<ng0>apparently it is
<ng0>didn't notice, sorry
<nckx>ebrasca: Hm. I promise that updating Guix is usually much easier than this ;-) What does ‘guix pull’ say when you run it? I don't understand.
<ebrasca>nckx: "lsa .config/guix" -> ". .."
<ebrasca>ebrasca@Ebrasa-PC:~% guix pull Migrating profile generations to '/var/guix/profiles/per-user/ebrasca'... guix pull: error: symlink: El fichero ya existe: "/var/guix/profiles/per-user/ebrasca/current-guix"
<ebrasca>nckx: ^
<ebrasca>nckx: I can't see that caracter
<ebrasca>directory .config/guix is empty
<nckx>Oh, right, you had Unicode troubles. It's just a thumbs-up. To be fair, you could have mentioned a fatal error when complaining earlier.
<ebrasca>Can I have spanish keyboard and english text?
<nckx>I've seen that problem before but am not finding anything on the Web. Sorry.
<ebrasca>My browser show everiting as rectangles ...
<ebrasca>They have some hexadecimal numbers
<ng0>wasn't solved by installing some fonts and fc-cache -vfi?
<ng0>i mean on other systems you can combine this.. i have C locale and german keyboard layout
<nckx>ebrasca: Sure. You'd set (locale "en_US.utf8") in your system configuration. I guess you have it set to Spanish now? It has no effect on the keyboard. But let's not try that before you can reconfigure without downgrading...
*nckx has (locale "en_GB.utf8") and (console-keymap-service "dvorak-programmer"), for example. Same for X.
<ng0>what are the permissions and ownerships on your .guix folders?
<ebrasca>nckx: What is dvorak?
<ebrasca>nckx: Is it like qwerty?
<ebrasca>ng0: drwxr-xr-x 2 ebrasca users 4,0K feb 27 23:19 guix
<ebrasca>nckx: I can't read internet links with my icecat ...
<ebrasca>nckx: sory png
<ng0>seems normal permissions
<nckx>ebrasca: …you can't view PNGs? Oh.
<ebrasca>I can
<ebrasca>I have think it is text link
<ebrasca>not png link
<apteryx>nckx: console-keymap-service! this must be a new thing :-) I'll have to try it out.
<nckx>ebrasca: Try ‘/var/guix/profiles/per-user/ebrasca/current-guix/bin/guix system reconfigure …’.
<nckx>apteryx: New in Guix 0.11.0 😛
<jackhill>Hi, I've created this package in my channel: but when I build it, substitute* calls mkstemp which fails with
<jackhill>In procedure mkstemp!: Permission denied
<ebrasca>guix system: error: symlink: Permiso denegado: "/var/guix/profiles/"
<jackhill>I assume that this is some interaction between substitute* and trivial-build-system
<ebrasca>nckx: ^
*nckx never uses the console but it's nice not to have to hunt-and-type QWERTY when my system breaks.
<nckx>ebrasca: Sorry, implied ‘sudo -E …’ was missing.
<ebrasca>nckx: What now?
<nckx>Did you try ‘sudo -E /var/guix/profiles/per-user/ebrasca/current-guix/bin/guix system reconfigure …’? What happens then?
<apteryx>nckx: eh, I missed it then :-)
<ebrasca>nckx: It give some text
<apteryx>ebrasca: "give some text" is not a useful information for us when we're trying to help you :-)
<apteryx>could you post said text to
<nckx>ebrasca: I'd ask to put it on if it's more than a line or two, but since you don't have a working browser...
<ebrasca>Give me some time
<ebrasca>My comand to upload to some pastebin don't work.
<nckx>ebrasca: Sure.
<kmicu>ebrasca: apteryx FIY in my case installing fonts eg. font-sil-charis solved missing glyphs (rectangles instead of fonts) in IceCat issue.
<apteryx>kmicu: interesting! If it ever comes back I'll remember it! Thanks for sharing.
<kmicu>‘Instalación terminada. No se notificó ningún error.’ so it lookes good.
<ebrasca>kmicu: Yea I have my icecat with text XD
<nckx>ebrasca: OK, so *NOW* updating Guix is as easy as guix pull && guix … :-P
<bavier>jackhill: you're running substitute* on a file in the store, which is a no-no :)
*nckx hopes.
<nckx>Or ‘sudo -E guix system …’ when doing rooty stuff.
<bavier>jackhill: you should copy the source script into the build directory, substitute* it, then install
<bavier>jackhill: also, doing a single substitute* would save a read-write cycle
<ebrasca>nckx: quix pull don't work
<ebrasca>nckx: Same error
<nckx>ebrasca: It probably will if you use the full path (but no sudo!) again.
<kmicu>ebrasca: if you want to update system to the latest version you run ‘guix pull’ to pull latest changes and then ‘sudo -E guix system reconfigure /path/to/your/config.scm’. To install a font you use ‘guix package -i font-sil-charis’ and restart IceCat.
<kmicu>ebrasca: Could you show what ‘guix pull’ gives you?
<nckx>kmicu: The issue is that ‘guix pull’ doesn't work.
<jackhill>bavier: thanks! I'll give that a go
<nckx>kmicu: Scroll up :-)
<ebrasca>Here output of "guix pull" with pull path :
<nckx>ebrasca: Then, if that works, I'd just reconfigure again and reboot and hope the environmental mess sorts itself out.
<kmicu>ebrasca: thank you.
<jackhill>bavier: should I use copy-file to copy the source script?
<bavier>jackhill: yes, that should be fine
<nckx>That auto-migration always seemed a bit too complicated for its own good. Does anyone know what it's trying to copy? I'd be tempted to just remove (rename) the old source.
<kmicu>ebrasca: what about ‘guix pull’ only? Gives the same text?
<nckx>kmicu: Yah.
<kmicu>Better to check whether guix is in PATH.
<kmicu>(Cuz there’s an absolute path to guix in the snippet.)
<nckx>kmicu: guix isn't in the path, that's part of the problem.
*nckx greps the sources.
<nckx>ebrasca: And? Any luck?
<kmicu>On stock Guix System 0.16 install?
<ebrasca>guix pull don't work
<ebrasca>sudo -E guix pull don't work
<nckx>ebrasca: Try ‘mv ~/.config/guix ~/.config/guix.old’ and ‘/…/bin/guix pull’ again.
<nckx>ebrasca: Asido: you shouldn't ‘sudo -E guix pull’.
<jackhill>I tried adding copy-file but I've messed it up somewhere. Now string-append complains about being called with the wrong time. I can't spot what's wrong:
<nckx>ebrasca: It might work but it's not necessary.
<ebrasca>nckx: mv guix to guix-old and sudo -E guix pull don't work
<kmicu>Maybe root owns user’s files now and that’s the problem.
<nckx>ebrasca: Drop the sudo -E and use the full path (that's the /…/bin/ I meant above).
<nckx>kmicu: …yeah. I eeked when I read that.
<ebrasca>nckx: I have reboot I don't have saved full path
<nckx>ebrasca: /var/guix/profiles/per-user/ebrasca/current-guix/bin/guix pull :-)
<nckx>Yay scrollback.
<kmicu>Literal commands are most helpful in this case.
<ebrasca>nckx: Same error
<kmicu>ebrasca: could you show ‘ls -al ~/.config/’
<kmicu>Also ‘ls -al /var/guix/profiles/per-user/’ would be useful.
<nckx>ebrasca: Please ‘mv /home/ebrasca/.config/guix /home/ebrasca/.config/guix.old’.
<ebrasca>nckx: I have deleted guix and guix.old after not working
*nckx goes to stare at the codez again.
<ebrasca>Maybe I need to reinstall GuixSD againg?
<kmicu>Maybe we can ‘rm /var/guix/profiles/per-user/ebrasca/current-guix’, Guix should recreate links anyway during guix pull/reconfigure.
<kmicu>If guix pull complains about it ‘guix pull: error: symlink: El fichero ya existe: "/var/guix/profiles/per-user/ebrasca/current-guix"’ then why not 😺
<nckx>ebrasca: OK, let's try this. ‘rm -r ~/.config/guix && mkdir ~/.config/guix && ln -s /var/guix/profiles/per-user/ebrasca/current-guix ~/.config/guix/current’
<kmicu>(And after that better not run commands as root willy-nilly :)
<kmicu>(That can destroy any distro from Arch to Ubuntu.)
<ebrasca>kmicu: I have gentoo for years and 0 problems...
<nckx>ebrasca: You've been asked multiple times to stop running guix pull under sudo.
<ebrasca>nckx: I have first try with no sudo ...
<nckx>Now you'll have to run the first rm in my list above with sudo, too.
<nckx>ebrasca: The proper reaction to a command not working is never ’run it as root instead’ :-s
<nckx>Not by default, at least.
<ebrasca>Why 2 people are giving me 2 diferent directions?
<nckx>I don't know.
<ebrasca>nckx: I can run "rm -r ~/.config/guix" with no sudo ...
<kmicu>ebrasca: I think there is miscomunication here. We only neet sudo during system reconfiguration (that needs to change serious things like grub entries) and nothing more.
<ebrasca>kmicu: Do you say some random user can install packages?
<nckx>Also, kmicu only said ‘Maybe we can…’.
<ebrasca>kmicu: Is it save?
<kmicu>That’s the whole point. Yes.
<nckx>ebrasca: Anyone can install packages to their own profile.
<nckx>That is indeed the point (one of them) of Guix.
<ebrasca>Yea but it run in the CPU
<ebrasca>What about keylogers?
<nckx>Most things do.
<nckx>ebrasca: If you install a keylogger, it can log your keys.
<nckx>And yours alone.
<nckx>I'd advise against it.
<kmicu>Users have no access to each other profiles. They cannot mess up w/o root rights.
<nckx>ebrasca: Have you tried ‘rm -r ~/.config/guix && mkdir ~/.config/guix && ln -s /var/guix/profiles/per-user/ebrasca/current-guix ~/.config/guix/current && /var/guix/profiles/per-user/ebrasca/current-guix/bin/guix pull’ yet?
<ebrasca>nckx: You have add 1 more comand to what I have run and full path for some parts...
<ebrasca>nckx: ok guix pull for now is working
***Server sets mode: +cnt
<kmicu>ebrasca: do you have many packages installed that you want such thing?
<ebrasca>kmicu: I like to use install system and have all my packages...
*nckx overreacted. Please bear in mind we're just volunteers, and having to repeat ourselves multiple times can suck the joy out of helping. Sorry.
<ebrasca>nckx: I am stupid , I only know programing in cl ...
<ebrasca>nckx: I am lazy to lean guile.
<nckx>ebrasca: If you have a working system configuration file you're almost there. You can add a ‘packages’ field to your operating-system declaration and those packages will be available system-wide.
<kmicu>ebrasca: do you have /etc/config.scm file?
<ebrasca>kmicu: Yes
<nckx>ebrasca: You can't be stupid if you use Guix.
<nckx>This is both a conclusion and an imperative.
<kmicu>There should be a line there like ‘(packages (cons screen %base-packages))’
<kmicu>You can add there packages you want like e.g. ‘(packages (cons screen icecat vi font-sil-charis %base-packages))’ and then ‘sude -E guix system reconfigure /etc/config.scm’
<ebrasca>kmicu: Here config.scm
<kmicu>So you already added zsh, icecat. Keep adding more.
<kmicu>Is this what you want or did you ask about something different?
<ebrasca>kmicu: I have install some packages with guix package -i
<ebrasca>kmicu: Now I don't remember what ther are.
<ebrasca>kmicu: It is painfull to need to remember all packages.