IRC channel logs
2026-11-10.log
back to list of logs
<nomike>I'm trying to install Guix System on a ThinkPad using the latest nightly installer ISO and I have a question about the locale: <nomike>At first, I'm asked which language I want to use, where I select "English" because that's my preferred one. But I'm located in Austria and on the next page there is only a narrow selection available where Austria is not one of them. Any suggestions? <nomike>On Ubuntu for example I'm able to chose English and Austria at the same time, why wouldn't I? <lechner>nomike: you can change the locale later, although I'm not sure how to do it. It's probably part of the configuration file <gnufairy>Well, yeah go with it, then in the final step there is an option to edit the file. Do that. Near the top, there will be a section for locale, timezone and keyboard-layout. Edit that.... <nomike>I guess I will file a bug report for this on Codeberg. For now I will chose something. <gnufairy>You can just edit the config file before installing.... <nomike>Yes, but IMHO this should be more user-friendly. <gnufairy>Also, there's this annoying bug on the "latest" ISOs where the install does not work, you must complete the final install step manually in tty3. With herd start cow-store /mnt && guix system init /mnt/etc/config.scm /mnt --fallback <gnufairy>Well, nomike, I guess you will have to edit your system config.scm at somepoint... <aeillol>I don't want to ask to ask, but is it okay if I ask for support on packaging here? I have a font I want to contribute upstream to guix, and I don't know who to give credit to at the start of the file. <czan>For sure you can ask for packaging help. I'm not sure what you mean by credit at the start of the file, though - do you mean the copyright lines in the Guix source? <untrusem>aeillol: if its changer longer than 15 lines you can assign your copyright to the file <aeillol>Yes, I do mean the copyright lines at the start of the guix source (this is a package that has not been added to the repo yet, and looking at other examples I did notice an email being added) also the commit is longer than 15 lines long. <untrusem>answering your question you give credit to the person who is contributing the package defination, in this case, *you* <untrusem>cnx: haha was gonna share this guix-devel <cnx>thanks, and it'd be cool if we get our build farm's energy consumption published <cnx>s/published/monitored/ <Deltafire>i was wondering, if a dependency of a package changes, then the dependant is rebuilt - even though it may not result in any changes to the dependant binary itself <Deltafire>so a 3 hour librewolf build might be triggered for some minor change to one of its dependencies <untrusem>wait what, is it your local modifications because librewolf package have substitutes <untrusem>or something changed with the inclusion of mesa-updates branch? <Deltafire>i mean in general, was just using librewolf as an example of a package that is time consuming to rebuild <Deltafire>i know there is that hack which for now i've completely forgotten the name of, but usually is the last step guix does during an update <Deltafire>for some reason i can't find it in the manual either :( <Deltafire>hidden under "19 Security Updates" in the manual <pinoaffe>i think the npm importer is overdue for an overhaul, but I started small :) <untrusem>you can ping jlicht (but he will be busy with release) or apteryx <jlicht>pinoaffe: had a quick peek, and decided to leave one of those annoying "that's a great idea, but have you considered XYZ?" reviews <jlicht>pinoaffe: in general, I think it would be nice to take inspiration from the way rust/cargo stuff is imported in guix for a future refinement of the npm importer <pinoaffe>jlicht: good to know, so far I'd mostly looked at the pypi importer for inspiration, but I'll look through the cargo code too <jlicht>oh don't let me hold you back; any improvement over the status quo that doesn't bloat the code by thousands CLOC is great <vntsuyo>has anybody fully ran Guix System on a server/VPS before? <cnx>what sorta reference do you need? <cnx>ACTION was asking if it's regarding pricing or config <vntsuyo>yes it's me lol, i forgot my original nickname's password <cnx>i have a desktop that is port-forwarded but i guess it doesn't count <cnx>i don't think it should be any different from guix system on pc tho <cnx>no, i just raw dog shepherd <cbaines>vntsuyo, there's lots of servers out there running Guix System, both physical ones and VPSs <Icy-Thought>Is there a better approach to rebuilding some rust stuff instead of alway recompiling when rebuilding a package (trying to package a pkg) <Icy-Thought>I guess I could do what we do in nix, unwrapped -> wrap or something <attila_lendvai>libvirt: the vm definition xml's contain full path to stuff like the uefi bios. how is it handled in the guix context? i don't want to insert ever-changing store paths into the config... <ampinga>Hi everyone, I want to share something and ask a question. Earlier, I asked here about a way to check the download size of a package during installation before actually installing it. I think I've found a solution for the user profile and for Guix Home: using the --dry-run option, for example guix install --dry-run emacs. It shows the download size. <ampinga>I looked for this in the manual but couldn't find it under the package management commands. So my question is: am I wrong to think I've found a solution, or does the manual lack this information? <cbaines>ampinga, --dry-run is present in the manual, but the description is quite short <cbaines>I'm not sure it's meant for providing download information, that information is provided anyway and --dry-run just cuts out the rest <cbaines>I'm also unsure how it interacts with other options, e.g. I don't see the download figure if I add --no-grafts along with --dry-run for some reason <ampinga>cbaines: You are right! I tried with --no-grafts along with --dry-run. <Icy-Thought>Question to rust packagers, how would you go on about wrapping the output package with glib or gtk ? Because I feel like I am doing it wrong by writing the manual approach <bjc>attila_lendvai: you can use the /run/current-system path for libvirt files <bjc>i can't remember what it does by default, but that has worked for me <attila_lendvai>bjc, thanks! i just removed the qemu binary, it finds it. and the firmware could also be simplified to <os firmware='efi'> <bjc>if it does use the store path by default and you have to edit the xml, can you submit a bug report? that should be an easy fix, i think <bjc>i didn't know there was a shorthand for not using explicit efi paths <aeillol>does anyone else have issues building any nerd fonts from the gnu guix repo? <jlicht>vntsuyo: I run a guix deploy'd system on my hetzner VPS <jlicht>vntsuyo: ^, where config systems hetzner has my operating system definition called "hetzner-system" <untrusem>haha, you will feel real pain installing guix on a vps where there is no recovery, no vnc, no custom iso, only ssh. <jlicht>untrusem: I have backups and can redeploy older versions on new machines. It's all just one DNS change away, so I'm not worried. I still have more uptime than Github :P <untrusem>I wasn't able to install guix on a shitty vps provider due to these restrictions <attila_lendvai>untrusem, i couldn't even install it because there's a web vritaul console, so no alt+f2, and there's no DHCP... which is a deadend for the guix installer. <aeillol>(oh sorry that was meant to be a command) <cnx>i think i'll do another round of world-rebuild branch (after 1.6) <cnx>ACTION only wants groff 1.25 but other orphaned updates have also been piling up <cnx>that ship does not show sign of sailing unfortunately <cnx>hmmm it's "only" 252 commits ahead of master <attila_lendvai>why does libvirt-service-type not require networking? libvirt-configuration's default requirement field does not include networking. should i send a patch or am i missing something here? <attila_lendvai>hrm, i guess it's possible that someone is using vm's without networking support... <nutcase>Hi. For an Emacs upstream bug report, I need Emacs built "with -g3 and without -s". How do i achieve that the Guix way (on a guix system)? <sneek>nutcase, you have 1 message! <sneek>nutcase, sham1 says: that you should check what the /var/guix/profiles/per-user/flake directory contains, and from there we could have an idea what's going on here <nutcase>sham1: Thank you. I removed the file "guix-home-322-link.lock" and that fixed it. <yelninei>--with-configure-flag=emacs=whatever_you_need <yelninei>and --with-debug-info to not strip the debug info <nutcase>yelninei: where to apply this? It's not a guix build parameter. Are you talking about modifying the package definition in a local guix checkout? <bavier>command-line flag to use with `guix build emacs` <sham1>That command line flag should work with guix build <sham1>Yeah, as per (info "(guix) Package Transformation Options") <sham1>nutcase: although you might also want to make your emacs customisation into its own scheme package definition which inherits from the base emacs, so you wouldn't need to specify it every time <nutcase>ok, found it, I think: "guix build emacs-pgtk --with-debug-info=emacs-pgtk --with-configure-flag=emacs-pgkt=-s --with-configure-flag=emacs-pgtk=-g3" (looks good?) <nutcase>"guix build emacs-pgtk --with-debug-info=emacs-pgtk --with-configure-flag=emacs-pgtk=-s --with-configure-flag=emacs-pgtk=-g3" <sham1>-g3 sounds like a CFLAG to me <nutcase>yelninei: tbh, idk (to be honest, I don't know) <bavier>might need something like `--with-configure-flag=emacs-pgtk=CFLAGS+=-g3` <bjc>isn't the ‘-s’ cflag for outputting the assembly code? <nutcase>I was told by Emacs maintainer "Could you build Emacs with -g3 and without -s, and then run it under GDB and produce the full backtrace from the crash?" <nutcase>"guix build emacs-pgtk --with-debug-info=emacs-pgtk --with-configure-flag=CFLAGS+=-s --with-configure-flag=CFLAGS+=-g3" this starts to build (at least) <bavier>probably need to leave out the "+", I can't always remember which build systems support that. <ieure>They want it *without* -s though, yeah? <nutcase>ieure: yes, is it "CFLAGS-=-s then"? <sham1>Oh neat, caddy is finally a part of Guix. I can finally have a decent reverse proxy <dlowe>I just have a private channel for the stuff guix doesn't have <sham1>I also do that, but I do like things being upstreamed <dlowe>yeah but most of the stuff I do that, I don't think people would be interested in <dlowe>Not sure, it's pretty bespoke for my specific setup. <bavier>nutcase: I think guix doesn't typically link with `-s`, but including `--with-debug-info=emacs-pgtk` might be needed <nutcase>bavier: I have a stack trace with more symbols in it, now. I ask the Emacs maintainer, whether this is more useful. <acid-bong>evening. i'm with a bit tangential problem today: i have a manifest.scm and an .envrc with `use guix` in a directory, and plain `guix shell` works, while using Direnv doesn't. what could be the difference? <untrusem>acid-bong: haven't you enabled the dir with direnv? <acid-bong>i did, but guix says "warning: no packages specified; creating an empty environment" <acid-bong>welp, explicit `-m manifest.scm` does help, but it's redundant, and iwant to just use `use guix` in .envrc <jlicht>I'm working on some wasm crosscomp support for guix, would it be yay or nay for getting the wasm sandboxes in librewolf/icecat enabled? <jlicht>from what I can tell, there is some processing overhead, but it theoretically is more secure, in an hand-wavey way <acid-bong>untrusem: i get it now: that direnv function executes a command (`-- direnv dump`), which in turn makes `guix shell` ignore manifest if it's not specified with -m <jlicht>oh wow, reka has been packaged \o/ <untrusem>I have built it 70 times at this point, /j <gnufairy>What's a solution to slow substitute mirrors? <jlicht>I know it can be ISP related; my home-connection was very unreliable some months ago and I had to use my tethered phone for a while. Fixed now though <jlicht>as in, home connection to the substitute servers specifically <ieure>gnufairy, Your trace doesn't clearly indicate to me what the problem is, but I'd lean towards unprivileged daemon as more likely. <sham1>Yeah, the trace looks like a permission issue <gnufairy>Thanks ieure, sorry I could not give a more detailed log this has been a repeating issue with an unprivileged daemon. I'll try again with privileged? #t <gnufairy>sham1, what sort of faster mirror? Are there faster substitute mirrors for the US? I'm just using the default ci, and bordeaux. <gnufairy>jlicht, the issue is not the internet speed, I have tested it against faster internets and I get the same issue. <yarl>Where is guix? :p I mean, the guix package is not in %base-packages. what part of the code puts guix into /run/current-system/profile/bin/. What's the interaction with the <operating-system> record? <gnufairy>You can view it any time with guix edit guix <yarl>gnufairy: What I don't undersand is how it end up in the operating system. <yarl>I'm looking at what part import package-management's guix and put it into the operating system that will end up in a image. <gnufairy>I am not actually sure where it is imported in Guix, Guix's system decleration does have implied fields it may be in there.... I think the guix commands to create an image work fine without importing it. Actually while Guix may not exist explicitly, guix-service-type containing the daemon does exist in %base-services or %desktop-services. It may be pulled in as a dependency....\ <yarl>But, _something_ constructs the profile and guix ends up there. <yarl>(Sorry if I'm unclear on what I am looking for)