IRC channel logs
2026-07-08.log
back to list of logs
<dcunit3d>apoorv569: make sure you update/close your issue on codeberg <noxi>hi hi, I'm trying to use an HP printer and I get a message "Printer <printer> requires the '/gnu/store/[...]-cups-2.4.14/lib/cups/filter/hpcups' program but it is not installed". obviously, cups isn't supposed to look for the filter there, but why does it do? <dcunit3d>i hadn't tried upstreaming this bc the main issue for me is that there isn't proper release tracking. <dcunit3d>noxi: on nix i have ./lib/cups/backend/rasterto{epson,hp,...} <noxi>hpcups is from hplip/hplip-minimal <noxi>what confuses me is that hplip is already by default in cups-service-type's config <dcunit3d>is it in `tree -a /gnu/store/[...]-cups-2.4.14/lib`? <noxi>> is it in `tree -a /gnu/store/[...]-cups-2.4.14/lib`? <noxi>no, it's another package <switchy>is there any configuration of the initramfs? if I wanted hibernation to/from a swapfile on an encrypted root, what would I have to do? or would it kind of "just work" becuase the luks unlock is early enough? <dcunit3d>it implies that hibernate works from an encrypted ramdisk, but idk. i never got that working (i didn't really try) <dcunit3d>so you would need the `resume` kernel argument, some elogind configuration for hibernate, potentially some power service configuration for hibernate/suspend. <dcunit3d>oof... those procfs/luks modules are for grub actually, so idk. <dcunit3d>so it would take some work & validation switchy <switchy>dcunit3d: awesome, thanks! I'll see if I can cobble something together <folaht>Where can I find a list of argument values I can give for the cargo-build-system in defining packages? <qzdl>:( when you run out of disk in /tmp/guix-inst <qzdl>time to get weird with mounts <jlicht>Recently I keep seeing "warning: channel 'foobar' is not trusted" in my time-machine invocations; any (new) manual steps I missed? <Alavi_me>I did the steps. I checked for vulns, I had 1,3,4 I think, I upgraded with `sudo --login guix pull --no-substitutes` <Alavi_me>and then restarted my whole system (guix is installed on top of arch) <Alavi_me>I run the test again, 4 is still vulnerable. What do I do? <futurile>folaht: Cargo/rust build-system is a bit different. Have a look in the manual at the Gnu build system details, and then you'll have to look at the source code (as trev said). There's also some details in the Cookbook article <futurile>Alavi_me: check where your guix daemon's systemctl is getting Guix from. I assume it will be the root one but just in case <futurile>Alavi_me: look at the service file for the guix service <futurile>it's normally /etc/systemd or soemthing like that <Alavi_me>ExecStart=/var/guix/profiles/per-user/root/current-guix/bin/guix-daemon ? <futurile>ok so you know it's the root guix-daemon <futurile>so you could check what commit that root's guix is on with guix describe <futurile>to make sure you'r guix pull as root was successful <Alavi_me>it points to commit: 9a34a395501c19b0661d1b4a3460c12b95f60f20 <futurile>Alavi_me: ok so your guix pull didn't work <Alavi_me>I'm gonna try `sudo --login guix pull --no-substitutes` again then? <Alavi_me>but it didn't show any signs of failure the first time <futurile>guess so - something didn't work if you're not on the latest commit - I would sudo up into root so you can do it all directly including doing a guix describe after the pull to check it <Alavi_me>Wierd. `sudo guix describe` shows commit: 436aaa67d19457e9a4ce0ce5d4d6fd792953eb66 which I can't find in guix commits! <ngz>Hello. I encourter a nightmarish problem. I use the unpriviledged daemon on a foreign distro, and I updated it during the time frame where it was broken (substitutes unavailable). And… I guix gc’ed before realizing it. So I cannot revert the daemon to a previous state. I tried sudo -i guix pull --no-substitutes, but after hours of compilation, it consistently fails on guile-3.0.9 tests. <Alavi_me>futurile: pulled again, `sudo guix describe` shows commit: a575cefb3ab29ce6b36eb84757b05b96027fc701 and `guix describe` still shows commit: 9a34a395501c19b0661d1b4a3460c12b95f60f20 <Alavi_me>kestrelwx: `sudo --login guix pull --no-substitutes` <ngz>Is it possible to switch back to a priviledged daemon? It could solve my problem, I guess. <futurile>Alavi_me: you're probably doing "guix describe" for your standard user; you need to do "sudo guix describe" right? <csantosb>ngz: Is reinstalling an option ? This is what I did. Way simpler. I just had to re create my guix home from the files. <ngz>csantosb: honestly, I don’t know how to reistall properly. Is it just about running the install script again? <ngz>csantosb: something along the lines of GUIX_ALLOW_OVERWRITE=1 ./guix-install.sh <alavime>futurile: umm. I'm not sure what you mean <alavime>> pulled again, `sudo guix describe` shows commit: a575cefb3ab29ce6b36eb84757b05b96027fc701 and `guix describe` still shows commit: 9a34a395501c19b0661d1b4a3460c12b95f60f20 <alavime>commit 9a34 is old and a575 doesn't exist as far as I understood! <ngz>alavime: a575cef is current HEAD <ngz>(gnu: weechat: Update to 4.9.3.) <ngz>alavime: what is the output of `which guix`? <alavime>ok so my `sudo guix describe` shows the currrent HEAD and `guix describe` is not on HEAD <ngz>alavime: again, what does return `which guix`? <alavime>ngz: /home/alireza/.config/guix/current/bin/guix <ngz>Did you `guix pull` in addition to `sudo -i guix pull`? <csantosb>To uninstall completely; if you do so, export the necessary files before, to create your home / profiles afterwards <ngz>I have a home.scm somewhere, I guess I can just guix home reconfigure $MY_HOME_FILE afterwards? <csantosb>You have maybe also a default profile, or anything else. <alavime>ngz: Should I `guix pull`? I'm confused by the vuln. will that be safe? <ngz>alavime: Of course you need to. `guix pull` updates your Guix, `sudo -i guix pull` updates system’s Guix. <gabber>we're trying to pull and/or use the time machine on a fresh installation (Ubuntu host) but we get "guix time-machine: error: clone: Permission denied" <gabber>isntallation was done through the installation script <gabber>`guix build hello` actually works <futurile>gabber: I wish I had a full "this is how to solve it", I looked into it a bit, but haven't figured out the details. I don't think the policy that we have in Guix fully solves every problem (but as I say I didn't manage to track down a solution). <futurile>gabber: if you can't figure out a specific policy you can try turning apparmor off temporarily - but it's a weird service as things stay in memory <futurile>and that was the point that I kinda gave up on it <gabber>i think one gets asked during installation to set up apparmor rules? IIUC they did not want to install these rules (?) and are now asking genAI to fix their installation (unsuccessfully) <gabber>futurile: temporarily as in: for an initial pull? or every time pull/time-machine is used? <gabber>now we run into "time-machine: error: store is read-only" (or similar) <gabber>uninstalling (from the script) does not work (Error executing command as another user: Not authorized [USER=root]), neither does installing (previous installation found) <gabber>should they just replace the machine? <gabber>or maybe just rebooting? though i don't believe in these kinds of esoterics <futurile>gabber: if they didn't install the rules it definitely _will not_ work <futurile>gabber: guix uses stuff like /tmp which the default AppArmor rules for Ubuntu will not allow <futurile>gabber: guix shell uses capabilities that users aren't allowed either <futurile>gabber: so the options are (a) turn apparmor off (bad idea) (b) have an apparmor policy for guix <futurile>gabber: the install script does (b) - that is _definitely_ required - and beyond that my impression is that even with the policy we don't have it totally right because I've seen problems for Ubuntu users <gabber>how bad is turning off apparmor temporarily? <Rutherther>gabber: if the store is mounted as read only for the guix-daemon service, try "systemctl stop guix-store.mount && systemctl restart guix-daemon" <futurile>I can't imagine why they _don't want to install the rules_ - that seems mad to me. If they don't want new rules they definitely will not want to turn apparmor off <gabber>Rutherther: "Failed to stop guix-store.mount: Unit guix-store.mount not loaeded" <gabber>futurile: i don't think there is much reason to it <gabber>this fails, too (journalctl shows "Error creating IO channel for /proc/self/mountinfo: invalid argument) <dcunit3d>gabber does the systemd unit have the ExecStartPre & Post? <gabber>this should be in /etc/systemd/system, right? <gabber>i can see gnu-store.mount there, but nothing named guix* <janneke>o how, a compilation by guile-build-system creates other (incorrect, imho) code than the guild compile through the build system <dcunit3d>it probably does, but i needed to power on/off after i added that <janneke>hopefully/possibly some optimizations, let's see what they are and how to influence them... <Rutherther>gabber: if the gnu-store.mount cannot unmount, that's an issue, because then guix-daemon will see the store as read-only on startup. So presumably reboot <Rutherther>gabber: or before reboot, is there /gnu/store bind mount in the output of "mount" <gabber>Rutherther: there's an entry for /gnu/store but i'm not sure it's a bind mount <gabber>the line reads: /dev/sda4 on /gnu/store type ext4 (ro, relatime,errors=remount-ro,stripe=16) [root] <dcunit3d>if you do `mount | grep gnu`, then the /gnu/store mount will have "ro" set as an attribute. i think <dcunit3d>also, for me `systemctl cat gnu-store.mount` shows "Options=bind,ro" <dcunit3d>and `systemctl status gnu-store.mount` would clarify whether it's systemd that's mounted it <gabber>hrmmm.. rebooting is not an option atm <gabber>dcunit3d: i can confirm the setup as you describe it <dcunit3d>so do `systemctl status gnu-store.mount`and `systemctl status guix-daemon.service`indicate that they're both active? <Rutherther>gabber: then remount the store as rw without unmounting it, and restart the guix-daemon service <dcunit3d>i think that not having the apparmor rules installed will make things difficult <futurile>it won't work without either the rules or turning apparmor off. And I have a nasty feeling that it's hard to do either without rebooting <futurile>or at least my notes are full of confusion that I can't get the rules out of memory, when I'm trying new rules out heh <gabber>we'll get that sorted out somehow <folaht>"Deprecation warning: Parameters #:cargo-inputs and #:cargo-development-inputs, previously used for specifying source inputs, are deprecated. Their support will be removed after Dec. 31, 2026." <- What has this been replaced with? <Alavi_me>Are there any good, auto-updated mirrors of guix outside of codeberg? There is one Millak/guix in github but it hasn't been updated in a week. <Alavi_me>futurile: It doesn't list any git mirrors <Alavi_me>codeberg seems to be blocked in my country <kestrelwx>podiki: So Wayland has switched to mdBook, which is a Rust project, to generate their HTML docs. Should we not build these then? `guix size wayland` is 180 MB right now, and I'm pretty sure nobody has ever seen the HTML docs from our package. I am fine with packaging mdBook, though. <kestrelwx>I guess on a typical desktop you would end up with Rust either way, so the closure increase is not particularly real. <noxi>can someone using %desktop-services (and who doesn't mind logging out of their session) tell me if 'loginctl terminate-session self', 'loginctl kill-session self' or 'loginctl kill-user self' works for them, at all? <noxi>any clean loginctl loggingout command <noxi>kestrelwx: what'd they do? <kestrelwx>`sudo herd status elogind` doesn't have any logs, but perhaps it's not meant to. <noxi>what de/services are you using? <kestrelwx>Just %desktop-services and I'm on a Wayland session without a greeter. <kestrelwx>Maybe not having `gdm-service-type` makes a difference. <futurile>Alavi_me: ah sorry, don't know. Trop.in used to run one, but I don't know if it's still active <futurile>Alavi_me: and there's Hako's guix.moe - but I can't figure out if they have a git mirror <ngz>csantosb: It seems I somehow recovered my Guix installation, thanks. However, there are cheesy signals such as install-license-file phase failing when trying to reconfigure home… <yelninei>I started rewriting the C parts of (guix build syscalls) in C but I am completely lost on how to integrate it with (guix build self) <yelninei>sham1: It relies on a lot of the configure checks for headers to do conditional compilation and (guix self) does not use the configure script <yelninei>but I think it is better than hand translating libc structs, constants and syscall numbers <sham1>It certainly sounds like the more sane approach <Alavi_me>shouldn't installing openjdk@25.02 provide java and javac ? <sham1>You need to use the jdk output to get javac. With the default out output you only get the JRE, and thus only java and the like <Alavi_me>I'm new to guix I don't quite get what "the jdk output" means sorry :)) <sham1>Ah, that's fine. Basically, try openjdk@25.02:jdk <sham1>Yeah, so the thing after the colon is basically the output. That's used to separate the package artifacts out so you can only install a subset you need. In this case IIRC openjdk has out, doc, and jdk <Alavi_me>So I install both openjdk@25 and openjdk@25:jdk ? or does installing some output other than "out" also installs "out"? <ieure>folaht, It's in the source repo, no you don't have to create it. <simendsjo>I see node was upgraded from 22 to 24. Looks like a massive amount of work! Good job to everyone involved! <ieure>folaht, Are you packaging for some channel other than guix? <ieure>The cookbook instructions are for adding packages to the main guix channel. <folaht>Ahhh... well I thought that guix might not be interested in these packages, so I'm adding them to my own channel first. <folaht>It's my first package build. I don't want to saddle guix with a beginner setup. <ieure>folaht, You probably need to make a packages/rust-crates.scm with a (define-module) and nothing else in it. <folaht>Okay, then I'll manually create a rust-crates.scm file and see what happens. <spiderbit>I have a machine (laptop) that just don't want to connect to the internet, I have gnome installed and networkmanager, but not only did wlan connect and internal hosts are mapped but for the live of it it can't resolv external host names... it can ping the ips of them... so ok anoying enough I connected lan still the same, I tried both in wlan and <spiderbit>lan to set manually the dns server but it just wont connect... <spiderbit>I try to switch to connman but without any internet I can't reconfigure. <spiderbit>is Networkmanager completely bricked in guix? <spiderbit>I guess I could boot a older system / generation <spiderbit>yes that fixed it, so something is totally wrong with network manager in the current version. <spiderbit>I booted last system configuration there all is fine <spiderbit>the only one I can think off but should be fixed is that I had used gnome-team channel <spiderbit>but removing that and "--allow-downgrade" should remove that all or not? <identity>yes, but can you ping an IP address from the generation that has broken internet? <ieure>spiderbit, There was a problem recently with DNS stuff, look at /etc/resolv.conf -- probably it has no nameserver line. If that's the case, rm the file, then disconnect/reconnect from your WiFi station. <spiderbit>I see it updated the name server I set under gnome now <ieure>spiderbit, Again, look at it first. If it has a nameserver linek you have some other problem. But if it's missing, a Guix bug is your problem and removing then disconnect/reconnect will fix it. <ieure>Okay, and is your problem resolved? <identity>now, restart the NetworkManager service, and DNS should be up <spiderbit>I just updated system so I have to reboot report back and boot back in the defect version after that <spiderbit>hmm connman with wlan isn't that easy aparently... <spiderbit>yes that fixed it somehow even I did not manually change (delete) the resolv file <spiderbit>booting one time the non defect version fixed it <PotentialUser-67>Hello guys, why can't I install sddm-service-type? It says unbound variable <ieure>PotentialUser-67, You haven't used the module which defines it. This is also not anything to do with "installation." <ieure>PotentialUser-67, (use-modules (gnu services sddm)) <ieure>PotentialUser-67, Sorry, I don't understand what you're saying. <ieure>PotentialUser-67, You do not install service types. <ieure>PotentialUser-67, The error you're getting means that there is no symbol named `sddm-service-type' defined in the namespace of your system configuration. You solve that by using the module which defines it, which I gave you the code to do. I'm not sure why you're arguing semantics. <ieure>PotentialUser-67, I usually search the source code. If you're using emacs-guix, you can use `M-x guix RET s r sddm RET' to search for any service with "sddm" in its name, and get a window with info about the service and its location. <ieure>Not sure if there's a way to do that from the CLI, seems like there should be. <ieure>There's a copy in ~/.cache/guix/checkouts/7x37p6bub2newlthjx5kk2mco2aq44vxbqp5gwa3sifqxzfb3aqq <podiki>sneek: later tell kestrelwx re: wayland and rust: i would check if wayland is available or not on e.g. i686; if this brings rust to archs it can't work then could make separate package for wayland docs (or wayland with docs) <dcunit3d>hey good job folaht. i'll take a look at it later