IRC channel logs
2025-01-06.log
back to list of logs
<ArneBab>Is there a service or such with which I can donate some CPU hours every night to build packages at pseudo-random to check whether the build works here? <ieure>ArneBab, Unfortunately not that I'm aware of. I've been thinking that having some kind of collaborative DHT for substitutes would be awesome. <ieure>But, not sure how to implement something like that without it being a gigantic security risk. <ArneBab>we could have trusted builders, but that’s actually only the second step <ArneBab>we could even have multiple builders and only accept a substitute from them if several trusted builders get the same hash for the substitute. <ArneBab>the first step would be to have a collaborative CI by building packages on many different computers and alerting someone if a build fails. Before I try to update my system and get build failures. <meaty>When should we be putting a copyright entry into a Guix file? <podiki>meaty: when you make changes; some people say if it is just a line or two (or a version and hash change) that you shouldn't, but i say it is up to the author <podiki>and of course how meaningful such small changes are from a legal standpoint will vary in different places, so who knows what 'counts' as a change <Guest94>Does anyone know why when I try to run guix home reconfigure I get the error No such file or directory: /run/user/1000/shepherd? <Guest94>It works fine when I run guix home container, but I get that error when actually trying to reconfigure <podiki>does that file exist? is shepherd running? (what's "herd status" say?) <Guest94>No I get the same error when doing herd status and no the file doesn't exist <Guest94>I just did a new install of Guix system and this is the first time I'm trying to configure Guix home <podiki>maybe because shepherd isn't running, special case for very first reconfigure? not sure maybe you can just try running shepherd manually and then reconfigure? <podiki>(would be good to report as bug) <Guest94>I get the same error running shepherd manually. I'm not sure if it has to do with it being my first reconfigure. I ran into this issue the other day on a guix system I've had for a while after I wiped my home config back to the default. <the_tubular> Using podman, I can start a webserver, but only the host is able to reach it <podiki>Guest94: does /run/user/1000 exist? might be an environment thing as that is XDG_RUNTIME_DIR <Guest94>Yep that's just what I was looking into. I haven't set up a window system yet. When I log in I get a warning that XDG_RUNTIME_DIR doesn't exist and the on-first-login script for guix home isn't being ran. /run/user/1000 does not exist when I first log in <podiki>that is probably some basic shell setup? not sure, but you'll want to deal with that first I think <podiki>gotta run, but good luck, there are experts around usually (not so much this hour, daytime europe more) <Guest94>podiki Thanks for the help. I think I can set it up manually or use something like elogind <Guest71>Now that zig 0.13 has been added is there any examples on how to deal with zigs package manager <rekado>Kolev: I replied to you earlier: you need portals. <jlicht>is there a way to force `guix build XYZ` to also build propagated-inputs of XYZ? <Rutherther>jlicht: propagated inputs are required during build, so they have to be built <jlicht>Rutherther: yeah that's what I would think; yet I'm downloading all texlive packages during 'guix install texlive-scheme-full', even though 'guix build texlive-scheme-full' is already in my store <jlicht>but thanks for the sanity check; there must be something else going on. <Rutherther>jlicht: are you sure it is the same texlive scheme all derivation as you have in store? <jlicht>Starting to doubt that or my sanity by now, but I'll double check <Rutherther>Also another thing is that if texlive scheme full doesnt reference them, but are still propagated, the paths may not stay active and gc will collect them <jlicht>ah; so could the inverse be the case? <jlicht>I'm getting a substitute for texlive-scheme-full, which probably does not contain references to its propagated inputs. (so guix build texlive-scheme-full => fast!) <jlicht>yet when I actuallly try to install the thing, the propagated-inputs are now actually required as well on my local system. <Rutherther>jlicht: no, if it references them can be checked only at the end of the build <jlicht>Rutherther: I know, but `guix size texlive-scheme-full` lists it as being an empty package, with no references. Subsequently building this with substitutes enabled, it makes sense that simply ends up being an empty directory in the store <jlicht>yet when I actually want to install it, the propagated-inputs will now obviously also be downloaden. What you're saying seems to be that were I to build texlive-scheme-full without substitutes enabled, I would (obviously) have to have the propagated-inputs available locally before the build can even start <Rutherther>Could you clarify what you mean by guix build being fast then? Do you mean it succeeds without building maybe? <Rutherther>Or also substitutes... Yeah, I agree, that could be the case then. Those inputs are required only if you are actually building it <jlicht>Rutherther: It downloads a substitute with size 0.0 MiB, which goes quite fast even on my slow connection ;-) <Rutherther>Then to substitute propagated inputs as well, I am not aware of a simple cli option. Dependency manifest could be used (what guix shell -D gives you, you can make the same thing and run build on it). But it woud download native inputs as well <jlicht>Rutherther: I'll probably just build the actual profile (e.g. guix build --manifest=profile.scm --profile=/tmp/my-actual-building-profile) <arni>hello! i wanted to add stumpWM to the list of DE/WMs i can choose on login on Guix System, but even if i downloaded stump i can only choose between XFCE and EXWM, which i've both installed on the graphical installer. how can i add stump to the DEs one can choose? i'm guessing it might have to do with manifest files, and the DEs i installed in the graphical login do show up. <arni>(i'm very new to this distro! so perhaps im missing something obvious) <arni>guix install stumpwm, normally <Rutherther>arni: you cannot install wms like that if you want for them to appear in your login manager. You have to put them to system profile <arni>oops, thank you for the correction! i'll look into system profiles then <Rutherther>arni: system profile is built from system services and packages list in your operating-system record <arni>gotcha :). i added ("stumpwm") to the list sexp inside of (packages ( append ...)), is this the correct way to add packages? <Rutherther>arni: yes, that is a way to add a package to your system profile <arni>nice, thank you! i'll reconfigure the system and see if this works then <arni>it worked! thank you again :) <paul_j> Good morning Guix! I am having some fun with fonts, and suspect a gap in my knowledge when it comes to Guix (as compared with other distributions I am more familiar with). If I install a font with "guix shell font-vazir", I would expect to see the font when running fc-list, but that doesn't appear to be the case. However, other fonts installed on my system through my configs are there. I recognise I could be doing something wrong, and <paul_j> the presence of the other fonts I have configured is incidental as they could have been pulled in by other packages. Have I missed something? I will keep exploring, but post this question in case I have missed something basic! Thanks in advance for your comments. I should add - this is GuixSD, not guix running on a foreign distro. <blu132>Hello, I just started trying to do anything with guix. The system I am experimenting on has an APU and I wish to blacklist the amdgpu and radeon kernel modules. I am trying to create a menu-entry in config.scm, in the bootloader area. I am stuck on setting (linux ...) to a path to my bzImage. The Bootloader-Configuration section on custom menu entries says I should do (file-append linux-libre <blu132>"/bzImage"). But the result of that is "/etc/config.scm:80:42: error: linux-libre: unbound variable". How do I get the path to my bzImage? <Rutherther>paul_j: try clearing your fc-cache. Also, what is your XDG_DATA_DIRS env var set to? <Rutherther>blu132: that means the variable called linux-libre is not in scope. So you haven't imported the module that uses it <blu132>Rutherther: Is there a module list where I can see what might fit my needs? I assume that there is a module that defines linux-libre. <Rutherther>blu132: not really. You can use guix search for packages, though it's not 1:1 mapped to symbols. Other than that you can just grep guix's source for define symbol <paul_j>Rutherther: XDG_DATA_DIRS="/home/paul/.guix-home/profile/share:/run/current-system/profile/share:/home/paul/.guix-profile/share:/run/current-system/profile/share" <paul_j>I'll try clearing the cache now... <blu132>Rutherther: Ok thanks for the pointer, I will have a look around <blu132>Before I dig too deeply: Is there a general way to get the filepath of a file in /gnu/store if I know file name and package name? <neox>blu132, yep: guix locate <blu132>... that I can put into config.scm as an expression in a place where a filepath is expected <blu132>neox: thanks, I am looking at it anyway <Rutherther>blu132: there isn't really a general way, depends on where you want it, usually either gexps with either ungexping the package directly and appending string, or file-append. Or if in a package phases, then search-input-file iirc <blu132>Rutherther: Thanks, I'll keep looking around <jonsger1>has someone an idea why KVM is not working for me on Guix System? My user is in the kvm, libvirt, netdev and wheel group. The `kvm` kernel module is loaded. Yet `qemu-system-x86_86 -enable` gives <jonsger1>Could not access KVM kernel module: No such file or directory <jonsger1>qemu-system-x86_64: failed to initialize kvm: No such file or directory <Rutherther>jonsger: did you try manually loading the module that provides kvm? do you have virtualization enabled on your system? <jonsger>Rutherther: I did a `sudo modprobe kvm` => kvm 1269760 0 <jonsger>what do you mean with "virtualization enabled on your system"? I have the `qemu-binfmt-service-type` in my config.scm <Rutherther>jonsger: virtualization can be disabled via bios. qemu binfmt doesn't do virtualization, it does emulation <lechner>Hi, is 1.0 the standard Shepherd in master now? <lilyp>shepherd should be on 1.0, yes <dariqq>is there a general strategy when a (C) library depends on python packages? Do I really need to wrap everything that uses the library? <Rutherther>dariqq: yeah, no other way, that's how python is, you have to tell it where to search for deps via env var <dariqq>Rutherther: yes but ideally I would only need to this in the library and not in other packages that use the library <lilyp>at runtime? yeah, you have to wrap it – for compilation? native-inputs it is <Rutherther>dariqq: it's processes (so applications) that have env vars, not libraries, so yeah... and if you are asking about a mechanism in guix that would wrap the apps automatically, I don't think there is something like that generally, only for python build system itself with python deps. The most that could be done is propagating the python packages, but then the apps would work only inside of a profile, which is not very good <amano>How many package maintainers does guix have? <amano>I mean active package maintainers. <lilyp>depending on what your library does, you should be able to patch it so as to insert the necessary paths on startup <Rutherther>amano: re your recent conversation here. You might be aware, but just to make sure you really are: if you are going to use extra initrd for key to unlock your disk, you are putting the key to the store, so if you also have the store unencrypted, you have key to the disk stored unencrypted at that point <lechner>lilyp / thanks! now I can get rid of my custom channel setup etc <amano>Rutherther: I was looking for a way to remotely unlock LUKS2 volume through beardrop ssh server started in raw-initrd. Manual process management without shepherd is a bit messy. <amano>I could also use an IP KVM like JetKVM. <amano>Gentoo linux has genkernel which handles remote LUKS2 unlock with beardrop ssh server in initrd. <amano>To keep things simple, I should just use JetKVM. <dariqq>lilyp: Yeah thats probably best, will see if I can find a way. It feels a bit ugly that package A would need to know about/ keep track of some dependencies required by library B (and a maintenance nightmare if these change) <homo>(delete 'patch-source-shebangs) doesn't work, I still have broken scripts because of it... <lechner>lilyp / i have been running the Shepherd development branch for nearly a year and can get rid of it now <lilyp>iirc you're not part of the emacs team *yet*, but do you want mail that's headed for the emacs team? :) <david95>Hello everyone. Please tell me what needs to be done so that IceCat does not destroy the test, I do not see the qr code, there is an empty square in it and the text that I see looks thin, etc. What settings need to be changed? <ieure>lilyp, Ah, I was skimming the earlier version of this patch yesterday. <juli>david95: could you provide more context? what text? what QR code? what website? <juli>it's also likely the upstream project may have better support than us downstream packagers :) <david95>Hi. Last time no one helped me install IceCat, I did it myself and needed to make adjustments not in .bash_profile but in .bashrc. Now there is a problem with the text, I can only show it otherwise it is impossible to understand and describe, the text is somehow corrupted or something.. What file sharing service can I use here to show it? <david95>thanks but this is a more complicated problem than it may seem, of course I disabled all add-ons and updated the font... <david95>you see there is no qr code and the text at the top is corrupted <ieure>david95, It's not "corrupted." <ieure>It appears not to be antialiased. <ieure>How long is normal for QA to take before picking up a topic branch? I wanted it to build librewolf substitutes before merging the latest to master, but it's been three days and it still hasn't built the commit to master that my branch is based on. <ieure>I don't have a good sense of whether this is normal-slow or something-wrong-slow. <david95>What do I need to do to correct the text or sweeten it, etc.? <podiki>ieure: there was an email a week or 2 ago about upgrades to QA but it was down for some time so quite behind if I remember <Kolev>So LibreWolf folks recommend using Ungoogled Chromium but even that doesn't allow screen sharing in Sway. <fsmunoz>Greetings! Starting in Guix, apologies if this is documented but I couldn't find it: what's the approach to persisting kernel module options? I'm thinking that kernel-arguments in the operating-system definition would work, but perhaps there's something "better"? <lilyp>apart from kernel-arguments you could build a custom kernel, but… what do you mean with "persisting options" anyway? <fsmunoz>lilyp: doing something that persists "echo 1 > /sys/module/..." , since this doesn't persist across reboots, and some options are required to be present during boot. No need to build a custom kernel, I think I'll follow the kernel-arguments route first then. Thanks! <lilyp>if kernel arguments don't work for whatever reason, you could also use activation-service-type to run basically anything <old>anybody attempted to package incus? <cbaines>ieure, QA is still not really working and hasn't been for some time <cbaines>I made some changes today to try and focus it on the branches in queue to be merged next <cbaines>which should at least help it build those <cbaines>the bordeaux build farm is still many weeks behind though, especially with the latest merge of mesa-updates <yarl>fsmunoz: see "(guix)Linux Services" : Kernel Module Loader Service. <wunky>Kolev: does screen sharing in sway work to begin with? i don't use wayland at the moment but i remember that being a pain point <orahcio>Hello, What the syntax to make list of vpn-plugins? The manual has no example on network-manager-service-type, I want to put the network-manager-openvpn package on that list, I try to make this https://paste.debian.net/1343032 <homo>orahcio in (inherit config) it looks like unbound variable error <homo>where "config" variable is defined? <orahcio>hi homo, that paste is inside of (modify-services %desktop-services <homo>orahcio I mean is "config" mentioned anywhere else in your file? where does it come from? <orahcio>Nice homo, it was a classical syntax error, I forgot to define the config to modify, thank you <homo>power of functional programming :) <ieure>How are y'all managing GPG with Guix? I set mine up using Guix Home, but it has the rather unpleasant property of being read-only, which makes key management difficult. <lilyp>I discovered you can do multiple GPG contexts by using GNUPGHOME – so if you don't want to manage your keys through Guix Home, just move the GPG stuff out to projects <nmeum>what's the best way to use ./pre-inst-env with an /etc/config.scm that requires custom channels? <nmeum>i.e. I want to run ./pre-inst-env guix system reconfigure /etc/config.scm but I can't because it doesn't resolve my custom channels <Guest37>Good day I am trying to figure out a strange? (I'm a newb bug) within guix, I am trying to build wine-10 within guix https://paste.debian.net/1343051/ win64 works perfectly, however when trying wine-staging.. the sources go to a 9.0 version of wine <Rutherther>nmeum: run guix time-machine -- system reconfigure or just pull with the channels specification. For guix channel use local folder as url <Guest37>It seems I generate a `/gnu/store/bzyakyn94l7aki2fbczn57xlkxnj41f2-wine-staging-10.0-rc4.tar.xz` which contains this "old?" source, but when looking at /gnu/store/fdp0yvc9yfl6a8m9djq06jzn7hgcdicy-wine-staging-10.0-rc4.tar.xz.drv the url derivation points to https://dl.winehq.org/wine/source/10.0/wine-10.0-rc4.tar.xz which .. doesn't contain a <Guest37>wine-9.0 directory, so I'm simply puzzled. <Rutherther>nmeum: alternatively if you had all the channels locally you should be able to use pre-inst-env with bunch of -L parameters going to channel folders. But it's probably easier to just use time-machine/pull with -C <lilyp>Guest37: when updating wine, you should also update wine-staging <orahcio>homo: The vpn plugin runs now, I need just privilages to add a vpn conection. Is there a way to do it as a simple user? <Guest37>lilyp: Oh I've done that, changed the version. the sha256.. everything. even hardcoded the version on some places to make sure. <homo>orahcio no idea, isn't it enough to be in "netdev" group? <Guest37>Please check the .scm file I've provided. It's the same as the native guix one, except I use 10.0-rc4 on all packages. <lilyp>you'd be wiser to hack on a local guix checkout <orahcio>homo: my user already is in netdev, but I get privileges error <orahcio>I get error if I try to delete some wifi connection <lilyp>alternatively, grab vkd3d from vulkan.scm and inline the definition <Guest37>lilyp: and constantly rebase? I am planning to add custom patches that are not upstream tho, so that's why it's decoupled... <homo>orahcio by quickly searching for documentation, I find that you might also need to be in "network" group <homo>don't ask me the difference between "network" and "netdev", I have no idea <fsmunoz>yarl: ty, reading on "-- Variable: kernel-module-loader-service-type" which seems to be what I want; I missed it because it describes a situation with no autoload but there's no reason I can't load load them explicitly. <orahcio>homo: nice, I will try it, I have no idea about that difference too <lilyp>as for wine-staging, did you really double-check your hashes? also the one for wine-staging-patchset-data? <yarl>fsmunoz: I don't know at all, I just mentioned it because I read it while searching for something else in the docs. <orahcio>It was because I want to stay on Sway, in Plasma or Gnome, everything is about just one click of distance <homo>orahcio also you need to relogin/reboot after you add/remove your user to/from groups <fsmunoz>I'll try it and report bak, thanks yarl. <Guest37>can we use guix download with git url? <orahcio>homo: Do you know if Guix has a limit of generations according the size of disk? For example, the /gnu/store can fill just 60% of the disk size <Guest37>thank you so much lmao I've been struggling for hours. <Guest37>Yeah, It was a dumb hash check.... regretful <homo>orahcio no idea, I'm as new to guix as you are <orahcio>Ok, I had a error with space left on device with more than 50G of free space, and I could to reconfigure just after delete-generations to make more space on my disk