IRC channel logs
2025-05-08.log
back to list of logs
<sturm>has anyone come across recent issues with missing Atheros firmware? My ath9k wifi has always worked out of the box, but has stopped working after a recent reboot. I thought it might be hardware, but a Debian hard drive works fine. <sturm>lspci lists the PCIe card. nmcli says "plugin missing" <PotentialUs-32>I’m glad to have found guix. I’m having an issue with installing packages though. Whenever I run guix install <package>, the installation fails and I get the error: Temporary failure in name resolution. sudo cat /etc/resolv.conf shows an active DNF configuration. Has anyone else encountered this problem? I used the installation fails script and <sturm>PotentialUs-32: Is this Guix on top of an existing OS, or Guix System as the OS? <isf>Guix on Linux sound strange <isf>becaues Linux is not a system is a kernel <sturm>Thanks PotentialUs-32 - out of my area of expertise sorry <sturm>isf: while perhaps technically correct, I'm not sure that's especially helpful <ieure>Well, it only runs on Linux and Hurd, and nearly nobody uses Hurd. <isf>Hardware run through Linux <isf>but software like GNU Guix need the GNU operating system to work <PotentialUs-32>Gotcha. Is guix running in a container? I think it can access the network on my machine for whatever reason <ieure>PotentialUs-32, Not unless you're doing something to run it in a container. <ieure>sneek, later tell andreas-e "I sent a clean v2 LibreWolf patch series." <vagrantc>well, the guix build environment is a container of sorts <ieure>Yes, but that's not relevant here. <vagrantc>hard to know without knowing more of the question :) <isf>I dont understand why there are so much people who are strongly against to speak about the operating system they are running <vagrantc>isf: i invite you to explore that in the form of genuine questions :) <sturm>could PotentialUs-32's issue related to NSS or nscd? <ieure>sturm, I was wondering the same thing. <PotentialUs-32>ld cause installed packages to have host resolution and other https errors? <ieure>PotentialUs-32, Can you resolve savannah.gnu.org, bordeaux.guix.gnu.org, or ci.guix.gnu.org? <ieure>PotentialUs-32, Many browsers are using DoH these days, so a browser being able to resolve a host doesn't say much about whether other programs on the system can. <PotentialUs-32>ieure I got disconnected. Did you reply to me? I don’t know if I receive messages sent to me when I wasn’t connected <meaty>Is there a way to use external guile packages in your guix config? <meaty>e.g. using a toml module in your config to generate toml files <meaty>I notice that the beets home service that comes with guix does almost nothing, even though the beets config is plain toml <meaty>lechner: as in, a guile module that's available as a guix package <meaty>is there an elegant way to do it? or an inelegant way <lechner>meaty / you can copy the package definition straight into your config, or "load" it from disk <lechner>sorry, you want to use a guile module in the build recipe? <meaty>I think so? I want to define configs as sexps and then use various guile modules to convert them to the correct config formats <meaty>maybe patch home-beets-service to be more useful <lechner>meaty / people often use the serialization features in (guix configuration) although I'm personally not a fan <lechner>mpd in audio.scm is a good example, i think <meaty>I theorize that it would be possible by adding guile-toml into the desired profile, reconfiguring, then adding the code that uses the module to the profile, then reconfiguring again <meaty>lechner: and yeah, ideally we wouldn't have to internalize a bunch of serialization procedures that are already written/maintained elsewhere <lechner>meaty / why can't you use guile-toml directly? <lechner>you can copy the service into your config and rename it <old>anybody else get hang after starting shepherd? I've recently pull channels and my system hang during the initialization process <meaty>I could also just vendor guile-toml or w/e into *my* system declaration... <nikolar>old: i've encoutered hangs before, when the clock gets adjuested <umanwizard>Hello, is there a way to disable tests for a particular package when running guix system reconfigure? <umanwizard>Tests in "nautilus" are hanging forever for me (aarch64) <meaty>umanwizard: see info page 9.1.2 --without-tests=$package <lechner>meaty / in order to improve the service, you can make a copy and edit it. that's generally easier than setting up your own channel or maintaining a fork of Guix <umanwizard>meaty: that is for guix build, not guix system reconfigure <umanwizard>brennan@eugene ~/code/parca-agent [env]$ sudo guix system reconfigure --without-tests=nautilus ~/config-desktop.scm <umanwizard>guix system: error: without-tests=nautilus: unrecognized option <ieure>umanwizard, Can you build the package without tests, then reconfigure? <umanwizard>the derivation during the reconfigure would still be the one with tests, no? <meaty>umanwizard: once built, packages are kept in the store <meaty>any profile built using that package will use the already built store item <umanwizard>once built, the output of *derivations* are kept in the store <ieure>umanwizard, Yeah, I think you're right. You probably need to replace the nautilus package in your operating-system with one transformed to not run tests. <hako>meaty: if you want to use it in the builder, see with-extensions. <hako>I think there're services using it in Guix <apteryx>python-minimal is like 7 MiB larger than just python on disk. odd. <csantosb>Can someone confirm the issue with `guix shell -C python-h5py python-numpy python -- python3 -c "import h5py"` ? <futurile>csantosb: yeah blows up - numpy.dtype size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject <ruther>csantosb: you're bringing in incompatible version of numpy, either don't specify python-numpy at all, sicne it's propagated from python-h5py, or specify @1 <futurile>really wish we'd used the same package names as Debian, it would make my life a lot easier when I cross-check packages <laugh> <futurile>also Debian should sort their inputs like we do <futurile>csantosb: what's the line you end up with? <csantosb>futurile, you mean this `guix shell -C python-h5py python -- python3 -c "import h5py"` ? <sturm>anyone else had issues with their atheros wifi recently? <futurile>csantosb: yeah, and how did you specify the @1 part? <futurile>csantosb: just curious, I've never done that before <attila_lendvai_>so, i have a PR on codeberg. i have rebased and force-pushed my branch, and now the PR contains 600+1 commits. why? can i somehow avoid the foreign commits showing up as part of my PR? <ruther>attila_lendvai_: do you have a shallow clone of the repository? <attila_lendvai_>this does not happen on github when i rebase and force-push into the PR's branch <ruther>attila_lendvai_: okay, then I think the answer is as simple as this: the civodul's fork hasn't been updated for a long time <civodul>attila_lendvai_: oh sorry, i guess i didn’t notice (was away last week and catching up since) <PotentialUs-32> I’m having an issue with installing packages. Whenever I run guix install <package>, the installation fails and I get the error: Temporary failure in name resolution. sudo cat /etc/resolv.conf shows an active DNF configuration. Has anyone else encountered this problem? The error I’m getting is: https://paste.debian.net/1373609. <ruther>PotentialUs-32: are you running nscd? It shouldn't be completely necessary according to the manual, but I can't think of anything else <csantosb>Weird question concerning #78297, how is one expected to send a v2 patch ? <csantosb>With a `git send-email -1 -v2 --subject-prefix='PATCH electronics-team' --to="78297@debbugs.gnu.org"` I get #78313 <csantosb>I creates a new issue, which is not what one would expect, right ? <PotentialUs-32>ruther no I’m not running nscd. I thought that’s only needed for packages not for the install command to work <ruther>csantosb: I think it's because you have `Cc: Cayetano Santos via Guix-patches via <guix-patches@gnu.org>` in the headers <ruther>PotentialUs-32: did guix pull work initially, fine? <PotentialUs-32>I’m running guix pull now. Let’s see. I literally just installed it and tried to run guix install hello like the docs show <csantosb>ruther: I wonder why do I have this "via Guix-patches" thing to begin with <hako>SPF policy in your mail server is too strict <ruther>I mean it doesn't really matter anyway since Codeberg switch GCD has got approved <csantosb>It does, since email based workflow is still alive until 31th december <PotentialUs-32>My system complains that /home/user/guix-profile is a broken symlink to /var/guix/profiles/per-user/user/guix-profile because only guix-profile.lock exists in /var/guix/profiles/per-user/user/. What’s this about? <ruther>PotentialUs-32: what does complain exactly? Why would the system care that a symlink points to a location that doesn't exist? <PotentialUs-32>Because of that broken symlink guix pull printed the error: -bash: /home/user/.guix-profile/etc/profile: No such file or directory <PotentialUs-32>I assume .guix-profile is necessary, since guix probably needs the path it points to <ruther>no, .guix-profile is not necessary for using guix <identity>PotentialUs-32: i do not even have that in my home directory... i am pretty sure it is only used by guix package (i use guix home instead) <PotentialUs-32>What do you mean by guix home? If it’s used by guix package then would it not be important? <PotentialUs-32>By the way, what are the reasons for installing guix at root rather than in the usual places such as /usr/local or /opt or in /home/user? <identity>PotentialUs-32: guix home is the subcommand that lets you easily manage dotfiles and the packages installed for your user, a more declarative alternative to using guix package and something like GNU Stow. guix package is the subcommand with which you can install packages for your user. i do not use guix package. see (info "(guix) Declaring the Home Environment") and (info "(guix) Invoking guix package") <ruther>PotentialUs-32: what do you mean installing it to root, do you mean the default /gnu/store path? <identity>PotentialUs-32: Guix needs root priviliges for build daemon isolation. There is an ongoing effort to drop the need for root priviliges in favour of Linux namespaces, you already can switch to rootless Guix during installation/later during the configuration process <identity>if you mean the store directory being in /, then there is no particular reason i know of <PotentialUs-32>identity thanks for explaining. So what’s the difference between guix home, guix package, and guix install? Before reading your explanation, I planned to only install packages using guix install. Is this wrong? <ruther>guix install is alias to guix package -i <ruther>PotentialUs-32: installing WHAT... <ruther>the difference between guix package and guix home is that guix package is just a simple profile that you install packages to, whereas guix home's architecture is more akin to guix system, where you have services that you can declare how your home should look like in guile. <PotentialUs-32>Files owned by guix. I don’t know exactly what’s included in /gnu. Maybe binaries if I use substitutes, and also config files <identity>PotentialUs-32: basically guix home is an entirely different way of managing your user packages, but also lets you do things like starting user-level services and putting dotfiles in your home directory <ruther>PotentialUs-32: everything guix builds is in the store <ruther>PotentialUs-32: the store has a default location that most users will prefer as the substitutes are then only for that location, so it cannot be under one's home, otherwise there would be no substitutes available for most people as most people differ in usernames. <ruther>you can of course adapt it if you are fine with not using substitutes, but that means building for days to get through the bootstrap <PotentialUs-32>In other news, guix pull finally failed with the error: executing SQLite statement: unable to open database file. <ruther>seems your guix installation is broken completely <PotentialUs-32>Any idea what this error is about, and where this database file is located? <ruther>The database is at /var/guix/db/db.sqlite <ruther>you will probably need to reinstall, something went wrong with your installation <ruther>no, you shouldn't have, running guix pull as root on foreign distros makes sense for updating the guix-daemon as long as the service is made to point to the root's home, but if you want to install packages to your user's profile, you pull as your user, pulling as root is going to have no effect on your own packages <PotentialUs-32>I was thinking guix could not access the database file because it’s owner and group are root <ruther>PotentialUs-32: guix as your user doesn't access the database file, the daemon does. What user does the daemon run under in your system? <identity>PotentialUs-32: it is in the name, Gui(le Scheme), though the daemon is in C++, it is a fork of the Nix daemon and was not (yet) rewritten in Guile Scheme <clemens3>wow, i think i just failed with the installer <clemens3>or it is the worst installer i ever used <ruther>clemens3: what installer are you referring to, like the TUI installer for Guix System? <identity>clemens3: i am pretty sure you can skip to any step you already gone through by pressing some F# button like F3 (it should tell you that, does it not?) <ruther>clemens3: can't you edit the config manually at the end? you could just remove the two excessive users. If that happens again <ruther>disclaimer: I haven't used the TUI installer, ever <PotentialUs-32>ruther I’m not sure how to check that. What I do know is that it’s path is /etc/systemd/system/guix-daemon.service. So I assume root since systemd services are run as root, I think <ruther>PotentialUs-32: no, systemd runs services as root by default, and switches the user if the service has User field set <ruther>So you need to look in the service file <clemens3>its not that i added three users, i keyed in three times to add one user, in a loop <ruther>for User= line, if there is none, then, yes, it's root <ruther>PotentialUs-32: okay, so as root. Then it should have access to /var/guix/db <PotentialUs-32>So then the only option is to try reinstalling? Is that install script I linked broken? <clemens3>i am in qemu, so ctrl-alt-f2 for documentation does not work, however it is laudable to offer documentation. but why not f1 or f2 for that? <PotentialUs-32>I wonder if these errors should have been caught during testing. <ruther>clemens3: it can work, just switch to the qemu terminal (in menu or ctrl-alt-2), and `sendkey ctrl-alt-f2` <clemens3>then "graphical install using a terminal" well, that is a contradiction in itself <PotentialUs-32>You know, testing that guix pull and guix install are successful <clemens3>i offer myself for testing, if this is not a good moment <ruther>PotentialUs-32: the script should be fine. But no one can guarantee it's going to work in all distros/environments <clemens3>i am talking the os not the package manager <clemens3>ok, i think i know what some logical problem is... <clemens3>on every other distro i used so far, in the dialog it says ok back and above you set the stuff <clemens3>and enter means go into the dialog that is mentioned above <clemens3>so in the guis installer, enter can mean exit or abort instead of continuing, depending what tab did before <clemens3>now its installing stuff, but the step with the partitioning for me was quite random behavior <futurile>clemens3: <enter> means either activate the <component> or <press the button>, it's the same thing. I guess you're not that familiar with TUI installers - most distro's use full graphical live-cd ones now. <futurile>clemens3: you can <tab> around the components on that screen <clemens3>i just used the tui installers of all devuan, netbsd, openbsd, freebsd in short order <futurile>clemens3: just to re-express what ruther was telling you - at the end of the install it gives you a file - this is what the installer actually consumes - the graphical installer is really asking you some choices it puts into that file - so you can edit everything at the end if you want <clemens3>i will redo and report again where i think it goes astray <clemens3>some dialog with partitioning at the bottom it just says exit <futurile>clemens3: fair enough, I wasn't throwing shade - I was just explaining <clemens3>you are supposed to hit enter to enter the selected thing above i guess <clemens3>no problem, i also just give free feedback <futurile>clemens3: exactly, maybe it should say <Next> <clemens3>in the other installers you select something, then you progress with a next or back button <futurile>clemens3: so are you on some massive try-every-distro project? <clemens3>here enter seems to be more like entering different rooms/dialogs <futurile>yeah it's - enter-dialog / activate-dialog <clemens3>regarding the hopping, i do my own distro and I look at different distros to learn different things or get documentation ideas <clemens3>guis is interesting because of bootstrapping java and other strange languages <futurile>ah makes sense. Basically, everything is "declarative", so most users will use the installer once to define their ideal system, and then can install N machines using variations of that config. <nomike>I'm working on some packages for guix, each one in their own branch of the guix repo. I jave submitted some of them already, but it takes time for them to be approved and merged. <nomike>In order to be able to use them in the meantime, I have just created my first local channel. I'm still struggling with it a bit, learning how to do that, working on technical challenges, figuring out how I want to structure it, etc., so I'm doing a lot of updates and thus need to run `guix pull` quite often. <nomike>But `guix pull` allways pulls the `guix` channel and my local channel, ad this usually takes ~20 minutes. Is there a way to `guix pull` just from my channel and leave the `guix` channel untouched? <sham1>You could pin the guix channel to a commit, and that should probably help <ieure>nomike, You can pass -L/path/to/your/channel to most Guix commands to run your code before you've committed and pushed it. <ieure>nomike, That's much, much faster than committing, pulling, reconfiguring etc. <identity>nomike: i personally just put my custom packages in a module file that i `use-modules` in my home config <nomike>sham1, I want to avoid that, because this will lead to me forgetting to remove that pinning and then running out of sync. <nomike>ieure, yes, that's an option, but it quite verbose having to type this all the time <nomike>identity, as I ultimately want to use my packages in my home-cofnfig, that's probably my best option. <nomike>Thanks a lot peeps for your feedback. <sham1>In my personal config I have a specific `modules`-directory where I put all my actual modules, and then I use direnv to set ${GUILE_LOAD_PATH} to prepend that while in my config <ieure>nomike, I suppose? Sure is less of a drag than `guix pull'ing all the time. And you can't make a shell alias to make `guix pull' faster. <sham1>direnv means that I don't need to do -L "${PWD}/modules", at the expense of having to be in the proper directory to use the configs <nomike>What I was initially googling for was something like `guix pull --channel=my-channel`, but yeah, I've got some better ideas now thanks to you peeps. <futurile>nomike: out of curiousity you said at the start that you did your packages with each one "in their own branch of the guix repo" what was the idea with that? <clemens3>with the installer, "Partitioning method", "Guided - using the entire disk" is red selected, when you tab, then Exit is selected and the previous selection is blue selected. So one suggestion is to add the OK button left to exit... this is on some other dialogs <clemens3>exit means exit the installer, ok, my mistake, but why would I want this here, and not some "back" button either <clemens3>and the exit is surrounded, so does this not look like already selected <futurile>clemens3: what distro are you working on btw? Can I play with it ;-) <lechner>Hi, how did the old ext2fsck cause the widespread reboot problems, please? <futurile>lechner: I don't think there's an actual confirmed answer on the email list <clemens3>futurile: thanks. got guix running and screen setup (in qemu) <futurile>clemens3: nice - if you haven't seen them the systemcrafter youtube videos are nice for exploring the more unique aspects <shcv>hello; is it just me or is the savannah git server down? And is it possible to add a mirror to the channels configuration? <ieure>shcv, Not just you, it's been very unreliable for weeks. <sham1>Hum, some emacs packages seem to be failing tests when compiling with emacs-next. Irritating <csantosb>Around 10 of them, yes, we have the list <csantosb>They fail to build, I mean; refer to emacs-team branch <lilyp>do we have statistics for emacs-next? I thought we only had them for the emacs-team packages themselves <lilyp>(which mainly uses emacs-minimal for what I know) <Guest27>sorry if this is getting repeated ad infinitum, but is savannah reachable? timing out to git fetch/guix pull <identity>there probably should be a news entry about this... <identity>but how will people get the news with savannah being down? <ruther>news are about what's new in guix content, not about the hosting unresponsiveness, so I don't understand why it would go to news <dcunit3d>Guest27: setting your channels to pull from codeberg is an option <dcunit3d>if you're pulling substitutes from the main channel <identity>ruther: i guess that makes sense, but i think it still should be communicated, one way or another <dcunit3d>the image is nonfree, but if there are any obvious pointers with the free stack <identity>dcunit3d: TIL about guix processes. if you are building a big image, it might take a while. <dcunit3d>Guest27: also, if you specify subs from {bordeaux,ci}.guix.gnu.org (via --substitute-urls ...) that can help sometimes <ruther>civodul: linux-container: Lock mounts by default has broken both guix system container and guix home container, with guix home: error: unshare: 268566528: Invalid argument. I will try to investigate (and send e-mail either with patch or at least the error), just wanted to give a heads up <dcunit3d>i've been building similar isos for a minute <dcunit3d>but i can't recall whether it took this long. one cpu is definitely spinning hot though. (and it's hopping around from green core to slightly magenta cores) <Guest27>to optimize disk usage of the store, I've started 'locking' guix channel updates such that most profiles can share commits of their channels -- this and regular GC has helped, but I've noticed that even when the store is fully hydrated, the guix derivation during pull is never cached <ruther>Guest27: yeah, it isn't, it's always calculated from locally cached checkout <Guest27>ah interesting, do you know the limiting factor here? my naive impression is that if all inputs are equal, we can memoize the output (for subsequent self-builds) <untrusem>I can't package this app as its no for beginners <untrusem>even nix folks are in a process to package this <luca>If it's a C# app it's not realistic that it will be proper packaged on guix any time soon <untrusem>yeah that why I am running it in a guix shell <luca>Where did you get a grayjay.app appimage from? <untrusem>its not packaged as appimage but rather a zip <untrusem>when I unpacked it in a container and tried to run it, it gave the linker error as expected to I used a fhs container with proper packages to run it <luca>What's the full command you ran to get in the container? <luca>idk I can't figure out how to run it on my machine. But if I were to guess I'd try passing $XDG_RUNTIME_DIR as well <untrusem>lol my whole does not have sound I think <kkremitzki>In the logs on 2025-05-07, "isf" mentioned they tried to install Guix using full disk encryption but it didn't seem to work/doesn't show anything... looks like they're gone now but I have seen the same behavior... basically the system is waiting for a password but failing to display a prompt