IRC channel logs
2025-10-07.log
back to list of logs
<Chwoka2>ooh, that would be super-useful if so <ekaitz>wow PotentialUser-83 ended the conflict! <FuncProgLinux>builder for `/gnu/store/n7inl5dl94x4y1qaj4wzv98n3813i9zv-guix-package-cache.drv' failed to produce output path `/gnu/store/0kwx1bv90rdzk6ih457w2azlsffry6mj-guix-package-cache' oh heavens to betsy what now!? xD <FuncProgLinux>(exception unbound-variable (value #f) (value "Unbound variable: ~S") (value (xvfb-run)) (value #f)) <russell`>I'm using guix with gnome that is selected at install. When I run guix home reconfigure ~/dotfiles/config/home/home.scm my PS1 is not updating across any shells, and my flatpak repo is not sourced. However, when I log in directly to the terminal (Alt + Ctrl + F1) PS1 is correct and flatpaks are sourced. <russell`>Shells I've tried: default gnome terminal, alacritty, emacs shell. I've confirmed the updates make it into the .bashrc and .bash_profile. The .bashrc has some of the aliars in the config.scm that also work correctly <apteryx>janneke: oh, wait before resending the mail, it was in my spam for reasons <sneek>apteryx, you have 1 message! <sneek>apteryx, untrusem says: the guix gc warning appeared again when I did `guix home reconfigure` <bvk>Hi, I am trying to use GUIX on a Debian laptop. My goal is to reduce amount of package proliferation that gets accumulated overtime and keep my base Debian installatino as small as feasible and guix looked like a good fit for this purpose. However, I get confused with the GUIX_PROFILE path I must use. Some guix commands hint me to set 'GUIX_PROFILE=~/.guix-profile' and some other commands suggest to set 'GUIX_PROFIL <bvk>E=~/.config/guix/current'. What is the difference? I did guix-pull and guix-install commands as non-root user and intend to use non-root commands at all times. <nigko>bvk: in 'GUIX_PROFILE="$HOME/.guix-profile" ; \ source "$GUIX_PROFILE/etc/profile"' GUIX_PROFILE is just a temporary bash variable. Note that 'export' statement is absent. <Ironsmith>i hope that was ok, i know one should ask permission before feeding pets XD <apteryx>ACTION did a bit of cleanup on the repo's branches (mostly really old wip or merged stuff) <apteryx>sneek: later tell civodul can we get rid of most of the version-* branches, perhaps keeping only the last release (I think that's used by the website?). Rationale: it clutters the view, and we have tags for them. <kestrelwx>efraim: Seems we need to patch QtWebEngine sources after all. <kestrelwx>The thought was coming to me before bed like an angel. <allana>Hi guix! is anything written in terms of documentation on contributing via codeberg? I see that we get a form to fill out when creating issues or PRs, but are there any further guidelines that I can be following? I'm wondering about labels or tagging particular teams/users, etc. I have contributed a few patches over the years via e-mail and I've also struggled a bit with the "Changelog style". The referenced docs don't show much when <allana>compared to what's actually often present in the guix git logs. e.g. * path/to/file.scm (package) [package-property] <key>: arbitrary (?) text <lilyp>there are snippets that you can use to fill in the changelog, though only for the simple stuff <lilyp>for more complex things you might want to ask a human to review <allana>lilyp: thanks, at this exact moment I am going to try to manually do the right thing, but for future contributions can you point me in the direction of where I can find these snippets? <lilyp>we have both tempel and yasnippet <lilyp>(if you're using another snippet engine, you might want to add yours for review) <allana>I use yasnippet, so this is great <csantosb>allana: have also a look to the `./pre-inst-env etc/committer.scm` command, it is documented <allana>csantosb: Thanks! I think this is what I have been looking for. Frustratingly it seems to have crashed though, not able to generate that commit message for all of my changes across multiple scm files. <hugohugo>Thanks for the question allana and the rest for the answers. I have no clue what snippets are. Apparently they are mentioned in "The Perfect Setup" in the manual, so should read that again. Most of it went over my head the first time around. <allana>Funny thing is that I probably review the contributing documentation once every few years. I can't tell what's new or what's been forgotten. <allana>At the risk of being redundant, I've submitted a few PRs between yesterday and today and would love some feedback. My general approach is to pair it with an issue in Codeberg, but I can't figure out how to label things to assign it to teams or if I should even do that at all. <hugohugo>I think we cannot assign labels and such ourselves, as non-committers. There is hako who often assigns team labels to my PRs. <hugohugo>I've learned to always do the `./pre-inst-env guix build -k -P1` thing and add that output, because it allows the reviewer to a) know that all dependents build, and b) compare those hashes with the hashes they get <allana>hugohugo: thanks, I actually did do that <allana>hugohugo: Now I have added the output of your suggestion to the PR and I'm ticking off each checkbox I can. Unfortunately I have to stop at cross builds because "build system `pyproject' does not support cross builds". Again, thanks! <hugohugo>I don't think all checkboxes are always necessary, as most PRs don't have them all. Just indicating the 'build -P' one because it seems to be useful. The lint one is also useful, also for old packages. <hugohugo>As for the log message, I usually search existing commit messages for something similar and copy that. <hugohugo>And depending on who will probably review my PR I add a space between "[arguments]" and "<#:phases>" or not. j/k, I alternate between commits :) <allana>I also wonder if I should even submit a codeberg issue at all. It might make sense to just issue a PR if I have a fix for a problem that I've encountered. <gabber>can we do shallow (--depth 1) git clones in package builds or is that even the default? <gabber>allana: if you send me a link i can label your issue/PR (: <kestrelwx>gabber: I think it tries to download only the relevant revision and otherwise does full checkout. At least from output I get when making mistakes in reivsion specification. <mroh>gabber: I think --depth 1 is default, see git-fetch() line 73 in ~/guix/build/git.scm <kestrelwx>`url-fetch` for that purpose makes sense to me. <gabber>ACTION is not surprised by whatever github is doing there <Deltafire> probably because it doesn't link the the actual file <futurile>so you actually have to figure out the 'real' place - a lot of the packages have a slightly different url to the one you see in your browser bar <csantosb>Substitute search is particularly slow for everyone or this is just me ? <csantosb>./pre-inst-env guix build ... takes ages <csantosb>I feel guilty every time I run my ci tests ... Pulling takes resources <cbaines>the nar-herder on bayfront has some memory leak thing going on <cbaines>it's not a quick leak, but over a week it'll use up most of the memory on the system <cbaines>I've got no idea what's causing this yet, the problem started a couple of months ago <csantosb>Me 😥 ... "1,046.6 MB will be downloaded ..." <cbaines>with the bayfront instance, it's the heap size that keeps growing, with the other instances, there's also an issue but it's not the heap and it's a much slower leak <kestrelwx>mroh: Oh, you're we should just probably patch `dlopen()` everywhere there. <mroh>kestrelwx: we should patch it in glibc: if(path.startsWith('/')) {fprintf(stderr, "you are evil, this will most likely not work in guix.Please fix your code, your buildsystem and everything else.");} <kestrelwx>mroh: If we had our `dlopen()` that searched inputs instead, like in `config.h`. <futurile>FuncProgLinux: we don't remove stuff very often (too slow in fact), it has to have an issue on Codeberg and wait a month <FuncProgLinux>futurile: I'm updating my system and I'm getting a channel incompatibility error. I investigated further and it turns out it's on my MATE channel lol. The mate-menu@22.04.2 is the faulty package and the thing that died was xvfb-run <FuncProgLinux>It tells me that guix repl: warning: ambiguous package specification `python2@2.7', but...that menu is Python3 only :L <ieure>FuncProgLinux, xvfb-run is probably just used for the tests, yeah? <ieure>xvfb is a headless X11 server. <FuncProgLinux>ieure: No tests upstream, it's listed as a dependency though <FuncProgLinux>Worst case scenario I'll drop the package since It's not ideal for upstreaming to guix because it's not "official" from MATE it's a third party applet <ieure>FuncProgLinux, Want to paste your build output? <ieure>FuncProgLinux, Variable is there and hasn't changed locations since 2021. <ieure>FuncProgLinux, So something's probably wrong with your package definition. <ieure>FuncProgLinux, As for the warning, there are probably >1 package which satisfy "python@2", are you using a manifest or something? <FuncProgLinux>ieure: Yes but that error happened on a guix pull && guix upgrade <FuncProgLinux>the manifest is indeed on the channel but no python2 dependencies on packages or the manifest itself <wlgreet-issue>hi all, sorry for my username, used it for asking about a seatd-greetd-wlgreet isse with sway but this is not about this. Host system: Fedora42. Guix running in a VM via virt-manager. I can not get sound to work in the VM. I choose ICH9 for sound under libvirt(virt-manager wrapper). Wondering if maybe others hit similar issues. I tried the <wlgreet-issue>Guix-VM which comes with a XFCE-Desktop. works there but there is no home service defined there. goal is to use pipewire with a home service. Anyone hit similar issues? <Rutherther>most programs are compiled with pulseaudio support... so why wouldn't sound work without a home service? <wlgreet-issue>Rutherther good Question. I don't know. Just wanted to ask if someone stumbled over similar issues running guix via a libvirt-wrapper(virt-manager). <Rutherther>I really don't understand what the issue is... having sound is an issue? <wlgreet-issue>Yes...i'm on my work-pc. i don't use the host-system for private stuff so i thought i setup a guest-vm running guix. <Guest24>Is there any webpage centralizing some more guix examples? For instance, I am setting GUIX_PROFILE and sourcing the profile within it in my .bashrc, but it really doesn't seem to work reliably. I'll login and type in git, no git available, so of course, go and source the file making them available. This whole process is clunky and I am <Guest24>wondering how other people do it. I will be really surprised if people's .bashrc has something like ```   GUIX_PROFILE="/home/smoky/.guix-profile" <Guest24>``` and it actually works. Do people use a daemon or something like that to update the $PATH? <csantosb>Guest24: See "7.1 Invoking ‘guix shell’" <csantosb>"... make sure that environment variables are set or modified in ‘~/.bash_profile’ and _not_ in ‘~/.bashrc’ ..." <Rutherther>Guest24: .bashrc is not supposed to update environment variables. So most people won't have that in their .bashrc. But yes, they will have it in their .profile, it's how you use the profiles in guix. Don't forget to unset the GUIX_PROFILE at the end, though. <Guest24>Ok, I have moved it to my .profile. Maybe sourcing it in .bashrc instead of .profile was the source of my headaches. I'll look at this futurile.net, thank you! <trev>Guest24: on foreign distro? pretty much exactly what you wrote <trev>except `GUIX_PROFILE="$HOME/.config/guix/current"; source "$GUIX_PROFILE/etc/profile" <Guest24>It seems that guix package -m like to install thing in $HOME/.guix-profile, while guix install puts them in $HOME/.config/guix/current. <trev>but that is in my shell profile. in the rc you need to add a different snippet <Rutherther>Guest24: no, both guix package and guix install install programs to ~/.guix-profile. guix install is a wrapper around guix package <Rutherther>Guest24: only guix pull puts guix to ~/.config/guix/current. And only guix should be there, nothing else is installed there. <FuncProgLinux>(exception unbound-variable (value #f) (value "Unbound variable: ~S") (value (mbedtls-apache)) (value #f)) another error by Python2 <trev>the install.sh script usually adds that stuff for you if you're using bash...i had to do it all manually cause i use zsh <Rutherther>trev: that doesn't make sense, guix shell already sets environment properly, you shouldn't source its etc/profile <FuncProgLinux>would things break bad if I move to a local checkout and just delete all python2 packages? <trev>Rutherther: hm...could be wrong. i thought i ripped it from the install.sh modifications <ieure>FuncProgLinux, What is your hypothesis of the problem here? <ieure>FuncProgLinux, I find that taking action without having a strong understanding of the problem rarely results in the results I desire. <FuncProgLinux>ieure: I'm pretty sure it's a python2 thing now. I've dropped the package from my channel and re-pulled but now the package failing lives in the games category and it's from guix itself <ieure>FuncProgLinux, What specifically about python2? <ieure>FuncProgLinux, It looks like mbedtls-apache has been removed. <Guest24>trev: what is the snippet you shared? is this what you have in your .bashrc? I don't think that gets put in today's install script <FuncProgLinux>ieure: How do you find when that happened? I always try to search the package with guix edit first but I still get lost when tracking down commits <trev>Guest24: i will defer to Rutherther cause mine may be incorrect (but somehow works on foreign distro for me) <Guest24>It does add something different to the end of the .bashrc, but it is a hardcore soup of characters. <ieure>FuncProgLinux, So, yeah, this is why I'm asking what your theory is. Because "depends on deprecated package which was removed" is not a python2 thing. <FuncProgLinux>ieure: I based my theory on the first output lines of the log. I get the repeated package guix repl: warning: ambiguous package specification `python2@2.7' which I thought was caused by my mate-menu package using xvfb-run <FuncProgLinux>I dropped it because it doesn't run anyways and it requires a lot of post-install patching for hardcoded /usr paths <ieure>FuncProgLinux, That is a thing, but those are warnings, not errors. I think the only way you'll get that is if something is using package specifications, which is something a package definition shouldn't be doing. <ieure>Which is why I asked if you were using a manifest. <FuncProgLinux>ieure: oh, no. The channel has one at root-level repository but no packages of mine use specifications <ieure>FuncProgLinux, What's in the repo root manifest? <FuncProgLinux>that channel is just used to get MATE updates and test them on my machine before upstreaming them <Guest24>Rutherther: Thanks for linking me to this, this .sh file is in the location you specified /etc/profile.d/zzz-guix.sh, but things installed in my guix profile are not available, for example commands like direnv, which, or git. If I am understanding correctly, because I have this file, I should not even need to modify my .bashrc, .profile, or <Guest24>anything like that in order to have access to this programs and libraries installed through guix? I did specifically say "yes" to allowing guix to modify scripts. <Guest24>I am on a fresh Debian install so this should be unchanged <ieure>FuncProgLinux, Just curious, why do you need the development manifest for Git? <ieure>FuncProgLinux, FYI, `guix shell -m manifest.scm' doesn't give me any warnings about Python. <ieure>FuncProgLinux, I'm on Guix commit c024cbffa3d97d620c4932f90ab1eebcb4dc5fbe. <luca>does guix offer any prebuilt chromium-related browsers? <FuncProgLinux>ieure: Cause I'm dumb and didn't realize the first argument was for the package you wanted the development deps installed for. I just realized that like two or three days ago when someone asked a shell question 🙃 <Guest24>I am wondering if it is something with the default debian shell configuration. I'll try to force sourcing this zzz-guix.sh file, I wonder if debian doesn't source files in /etc/profile.d by default. <ieure>Guest24, I think they have to have a numeric prefix, not alnum? <FuncProgLinux>I generated it with guix shell -D git guile emacs .... --export-manifest > manifest.scm but the "-D" wasn't needed <Rutherther>Guest24: Hm. If you run "run-parts --list --regex '^[a-zA-Z0-9_][a-zA-Z0-9._-]*\.sh$' /etc/profile.d", do you get zzz-guix.sh listed there? If so, I think there are two possibilities: 1. Your bash doesn't automatically source /etc/profile and you might have to source it yourself. This depends on how bash was compiled. Not sure what Debian ships. 2. You are overriding PATH and such in your ~/.profile / ~/.bash_profile somehow <kestrelwx>luca: Ungoogled Chromium and QtWebEngine bunch. <FuncProgLinux>ieure: I think I found the error. It's on a third party channel, I've removed all channels but the "guix" one and the pull doesn't show errors anymore <Rutherther>wlgreet-issue: I have ran guix system as guest system multiple times, yes. I didn't have problems with libvirt that I would remember <Guest24>Rutherther: zzz-guix.sh appears there, so I think there is some weirdness with the Debian/Ubuntu shell, because this is a brand new fresh debian with only guix installed. I will try forcing this to be sourced. <Rutherther>Guest24: and you have bash as your login shell, is that correct? <Guest24>I believe so, how would I verify this? <FuncProgLinux>oh wait it is pulling the third party after deleting it...what? <Rutherther>Guest24: look into /etc/passwd for your username <Guest24>Rutherther: Yes, bash is the login shell <FuncProgLinux>Are the channels from "guix pull" saved somewhere if you don't use a local checkout? <Rutherther>Guest24: okay. So if you put the guix profile sourcing to your ~/.profile and login again your usual way, do you get guix profile programs? <FuncProgLinux>I'd rather nuke it and work-around things than wait for updates 😅 <ieure>FuncProgLinux, Question doesn't make sense. Pull pulls from the configured channels, it doesn't configure them for you. <FuncProgLinux>ieure: it was a games channel which was failing, merge request is 6 days old, but deleting it from my channels file only delays the error, it still tries to pull it <FuncProgLinux>I thought it was pulling it under /tmp or something I could sudo rm -rf and call it a day :( <ieure>FuncProgLinux, Where is your channels file? <FuncProgLinux>ieure: I run guix pull -C <path-to-my-config-repo>/channels.scm to update <ieure>FuncProgLinux, and that channels.scm doesn't have the third-party games package, but you still get the warning coming from it? <Rutherther>FuncProgLinux: please send the channels.scm file <mra>hey, i think i'm being a bit dumb here. trying to come up with a minimal example of a derivation which takes another derivation as input. if i (define foo (derivation %store "foo" bash '())) and then (define bar (derivation %store "bar" bash '() #:inputs `(,foo))), i assumed that bar would have foo as an input, but running (derivation-inputs bar) <mra>writing a test for some new derivations stuff <Rutherther>and if you pull with this, without the channel you do not want to share, do you still get the error? <Rutherther>also the .guix-channel file in the local channel you have, are you sure it doesn't list the broken channel as a dependency or any channel that would in turn have the broken channel as dependency? <FuncProgLinux>Yes, two channel entries were deleted, the one that's not guix and the third party games one. Even after deleting the games one which is the one causing the issues, it's still getting pulled from according to the output. The error itself just takes longer to show up <FuncProgLinux>Rutherther: My local channel? It only depends on guix itself. It's meant to be only a gnu/packages/mate.scm "extension" for early updates or testing before upstreaming those packages here <trev>FuncProgLinux: do you still have it in ~/.config/guix/channels.scm or you don't use that at all? <Rutherther>it doesn't matter what is in ~/.config/guix/channels.scm if you use -C <FuncProgLinux>trev: I don't use that. I don't remember where but browsing codeberg I found someone's configuration that used a channels file inside the repository and pulled using guix pull -C and I liked that <trev>i do that but just symlink it :\ <trev>something isn't adding up though <FuncProgLinux>I'm doing experiments and I think I'm onto something, I have to confirm it <FuncProgLinux>what if you run guix pull -C <path_to_channels.scm> but the file path is wrong? <trev>it says no file or directory <FuncProgLinux>I moved the modified channels file to another location, ran the pull again and now it 100% worked. So this must be a mess up on my own configuration not on guix <FuncProgLinux>btw I know I'm not in a position to ask for any more 😅 since I probably already drained time/energy from people who stepped in to help me. But, if someone has spare time today, can you take a look at #3199 please? :) <FuncProgLinux>oh and btw, thanks ieure and Rutherther for taking time to help me :) <gabber>do subprocesses during a package build lack the ability to create directories? i get: "go: go: could not create module cache: mkdir /homeless-shelter: permission denied" <Rutherther>gabber: no, they definitely do not lack the ability to create directories in the directories they have write permissions in. They do not have write permission to root. <ieure>gabber, $HOME is set to /homeless-shelter during the build. <ieure>gabber, So that error means the build is trying to write to files in $HOME. <jlicht>apteryx: I think you ran into the same issue as I did wrt the `unshare' core files; they were dumped somewhere in the namespaced mount ether, instead of our actual disks :-) <ieure>jgart, Seems fine to me, though I have seen some intermittent 5xx errors today. <umanwizard>Anyone else notice emacs-next takes absolute ages to compile these days? <bdju>sneek later tell kestrelwx I do not have a decent machine. I think the build would take over 24 hours.