IRC channel logs
2025-10-06.log
back to list of logs
<dcunit3d>it's a lot more complicated to install guix after building `./configure --localstatedir`. that and trying to untangle Guix from the NixOS services.guix was tough. i'm not sure whether there's a good clean way to Guix on NixOS without too much impurity in the /nix/store & derivations <dcunit3d>the ./guix-install.sh script isn't a Makefile target. not a biggie. i eventually backed out. i just wanted to have the same layout on both systems and i kinda wanted /gnu/var/guix to be contained alongside /gnu/store ... idk why though. it's not really useful per se <dcunit3d>can the metadata from /gnu/store be regenerated in /var/guix though? <bdju>.tell kestrelwx Is it expected that using that qutebrowser workaround you can't copy things to the system clipboard? <bdju>sneek later tell kestrelwx Is it expected that using that qutebrowser workaround you can't copy things to the system clipboard? <eikcaz>I noticed we have a bunch of cloud-service-specific REST API code baked directly into guix source. Is this desireable behavior, or should it be avoided in the future? <eikcaz>(i.e. I could take the time to move such things to a seperate "guile-api-zoo" guile library repo <jlicht>Deltafire: so it's not just me /w the unclean mounts <untrusem>how to test, how long does, one's config take to eval roughly? without downloads or builds, just pure evaluation delivery failed <sneek>kestrelwx, you have 1 message! <sneek>kestrelwx, bdju says: Is it expected that using that qutebrowser workaround you can't copy things to the system clipboard? <kestrelwx>bdju: Apparently. Also, just discovered if you move around elements in the broken version they do render. <trev>sneek: tell me later (1 hour) that i'm awesome <sneek>me, trev says: later (1 hour) that i'm awesome <futurile-afk>trev: yes, mostly the 'later' command is useful for leaving a message <futurile-afk>Noisytoot: I think the reason no-one has touched it is no-one knows anything about core-boot <futurile-afk>and with a name like that it sounds serious and probably dangerous <futurile>Noisytoot: I guess you could run these patches locally and if they work for you report that they do <futurile>It's following a bunch of Guix people - if you want it to follow you tell me and I'll add you! <futurile>and if you think there's news that we should be telling people about - I'd be interested to know :-) <futurile>(is not a social media expert by any stretch) <ds7832>futurile: I'd love to boost your toot − could you point to anything showing that the account is official? <futurile>ds7832: not yet, I got the confirmation that we were official last night, but it's not working <ds7832>Ah, I've just seen the github link you shared perfect <futurile>ds7832: it said it was done, but I don't think the person doing it actually hit the publish button I guess - as I can't find Guix in the list <futurile>ds7832: I'm going to wait for the Guix site to publish the link (it's building) and then I'll reply on the ticket <ds7832>I wonder if it makes also sense to link prominently to https://guix.gnu.org/en/help/ on the Mastodon profile, in addition to the current link to https://guix.gnu.org/. For me and probably many others who ponder trying out a new distribution, one of the first things to know is how good some favorite pieces of software are supported. <futurile>I think I've fubared the site publish <sigh> <ds7832>I don't think so, and it's getting quite good traction <untrusem>futurile: yay for the official guix account :) <untrusem>you can follow my account, it has same name as here, I ramble about guix time to time, just booted the post <csantosb>futurile: Are you familiar with groups ? See @mastobikes@a.gup.pe, for example <hwpplayer1>I installed Guix inside a vm qemu/kvm but guix pull does not work <gabber>hwpplayer1: you need to configure networking <gabber>pulling on slow machines takes very long <Guest13>can someone help me set up a simple guix based Haunt blog? I got everything but the serving working with ssl on a non standard port <gabber>maybe yes, but we will need very detailed information. what are you trying to do (show us your config(s)), how is it failing (error messages), etc <Guest13>I'm new to guix, imo documentation isn't helping much <gabber>if you want help from us you have to tell us *much* more detailed what is failing and how <Guest13>as far as i know nothing is *failing*, more that i have no idea how to get ssl going with the webserver Haunt uses. <Guest13>gabber - if you're unfamiliar with Haunt, could you maybe help me with a workaround using nginx? <gabber>Guest13: usually this kind of setup is hidden behind a "reverse proxy" (usually nginx) that handles TLS/SSL for you and routes traffic between the public entry point and the "private"/hidden port (in your case 8080) <Deltafire>i don't think Haunt has an in-built ssl server <Deltafire>you can either use a proxy, or upload your site to a webhost <Guest13>how can i set up a reverse proxy with guix? <gabber>Guest13: this should be described in the manual <gabber>have a look at "nginx-location-configuration" <gabber>you'll have to set up an nginx-service-type. `proxy_pass' is where you want to configure something like "localhost:8080" <ente`>build of /gnu/store/0r4jxl77z3frcz7s5zi793h98sscwygh-libspatialite-5.1.0.drv failed <gabber>ente`: that's written in the output of the build command invocation <ente`>didn't see the bold white text under the bold red text, followed by more normal white text <ente`>seems the issue got fixed 15h ago <ente`>do I have to do "guix pull" both as root and as user or can I use the same repository clone for both? <gabber>ente`: you usually don't have to `guix pull` as root <gabber>but it somewhat depends on what you are trying to do (and how you are doing it) <ente`>I thought for "guix system reconfigure" I need to do "guix pull" first so it reconfigures with the up-to-date packages <gabber>if you do a normal `sudo guix system reconfigure` it uses the profile of your (unprivileged) user <ente`>how do I delete the extra clone? <gabber>i think they usually live in ~/.cache/guix/checkouts, so in your case probably: /root/.cache/guix/checkouts <ente`>ok I guess for my next trick I should start packaging my window manager <futurile>ente`: check that another channel hasn't already done it <untrusem>yep its down, we should make a cli program of it <untrusem>I also found https://searchix.ovh , a guix equivalents would be cool, I talked with author, they might even do it themselves because of how the things are going in nix land, :p <futurile>followed you back on the Guix account btw <futurile>csantosb: no not familiar with groups, I do post and follow #guix on Mastodon, but definitely not an expert <csantosb>Channels avoid the need to manually boost individual posts, maybe worth considering <csantosb>Don't forget @guix@lemmy.ml and @guix@infosec.pub <wingo>hey question. is anyone able to get -flto working with clang-toolchain on guix? i am seeing weird errors like "failed to load library libLLVMgold.so" coming from the linker <wingo>ACTION finally making some things compile with clang in addition to gcc <ente`>untrusem: how are things going in nix land? <ente`>(I have nixos on another machine. Do I have to be concerned?) <untrusem>naah, the steering and moderation committee drama, it doesn't affect the end user (for now) <tusharhero>Hello, I am trying to work on the emacs-guix package, and I have it installed in my user's profile via "guix install emacs-guix", how can I test my changes to emacs-guix? <tusharhero>the question of how to load it into emacs for testing still remains though. <gabber>tusharhero: why does `emacs -q -L . -l file.el` not work? <tusharhero>gabber: It should work but I am not sure what to load. <tusharhero>I would load the autoload file but I can't seem to find it. <gabber>for the path in the store you can invoke `guix build emacs-guix` <tusharhero>but that wouldn't be the emacs-guix I am working on right? <gabber>i think (but am not sure) that the packages installed with `guix install` get to reside in ~/.guix-profile/ <gabber>tusharhero: it would be the same thing that you installed with the command above <allana>Hi guix! What is the preferred way to submit patches now? I see a lot of activity on codeberg <untrusem>> Hello, I am trying to work on the emacs-guix package, and I have it installed in my user's profile via "guix install emacs-guix", how can I test my changes to emacs-guix? <untrusem>make a emacs-guix package with your changes; and use "guix shell -L . emacs-guix emacs` to test it <tusharhero>can I maybe just my local repository as the source code directly? <allana>gabber: thanks! I seem to have missed this. What about older patches on debbugs? If they've been unanswered for a long time, is it better to issue a PR from codeberg? <gabber>allana: IIRC debbugs will be deprecated by the end of the year <Guest13>gabber - where can i find more information about how to set up the nginx service as a proxy? For the life of me i can't find anything helpful <Guest13>btw am i right to think that my best bet to host a nginx service as a container on a non guix system, system? <Guest13>where can i find more information about how to set up the nginx service as a proxy? For the life of me i can't find anything helpful <Guest13>btw am i right to think that my best bet to host a nginx service as a container on a non guix system, system? <Rutherther>Guest13: why would that be your best bet? I mean ... maybe if you're unexperienced with Guix System, but at that point why would you even use it at all if you don't want to learn it <ieure>Guest13, I don't understand the question at all. <Guest13>Rutherther - my vps provider doesn't support installing Guix system. And yes i am very inexperienced, learning is harder without thorough documentation <Guest13>ieure - im currently running debian with guix installed and im not sure how to run the nginx service. Having watched systemcrafter's (minimal docker) video i thought that using a container would be my best bet. <ieure>Guest13, Are you trying to run Guix System in a container, or use Guix to run a container of something? <gabber>Guest13: i think you want to install Guix System somewhere. if all else fails, maybe consider switching your VPS provider? <Rutherther>Guest13: you can't run guix system nginx service on debian, that's a service for guix system <Guest13>ieure - using guix with the command `$(guix system image --network [FILE}.scm)` <Guest13>could be that the command i tried in a vm is a little different, but i did get a simple nginx service running there <ieure>Guest13, That doesn't really answer my question or make sense -- that'll evaluate the output of your `guix system image' command in the shell, which, since it doesn't print shell commands, won't work. <Guest13>gabber - i tried the linode setup on my provider and failed. I'll try to get a custom iso running tomorrow <Rutherther>ieure: it does print the path to the shell script that you start with evaluation in the shell <ieure>Rutherther, Does it really? What weird behavior for that command. <Rutherther>your approach for running nginx service seems fine <Guest13>is there a better way to run the service without guix system? or is this more or less *it* <PotentialUser-63>FuncProgLinux I would say yes. The issues I had with Inkscape are gone now. <Guest13>ok, i got something compiling now... <PotentialUser-63>`Error in configuration at CPU_SCALING_GOVERNOR_ON_AC="schedutil": governor not available.`. <ieure>PotentialUser-63, In what context did you get that? tlp just talks to the kernel PM stuff, which may or may not include schedutil. Being supported by tlp != supported by your hardware, nor the kernel it's running. <PotentialUser-63>ieure My surprise is that I did not see this message earlier. I think I have had this configuraton for 2 months, perhaps? <lee-thomp>what's the current state of using qmk in guix targetting an ARM board? I've seen some posts on the mailing list from last year showing issues getting the cross toolchain working, has this been solved since? <ieure>PotentialUser-63, PotentialUser-63 Maybe the kernel configuration or tlp default changed. <henrix>Hi! Quick question: I want to install guix on my debian system. can I use have /gnu and /var/guix as links to some other partition/disk? <henrix>the reason is, of course, disk space constraints <futurile>henrix: you can mount /gnu to some other location, it's a bit fiddly as I think the installer insists on creating them - anyway I have done it <futurile>henrix: I think if they're already there it thinks you've already run the installer and says 'Uh oh stopping..." <PotentialUser-63>ieure I think I did a `guix system reconfigure` but did not restart the system. <PotentialUser-63>Oh, well... One more thing we will never know. Thank you for taking an interest. <henrix>futurile: ah, ok, so I may need to hack the script. no prob I guess. awesome, thanks! <futurile>henrix: try just creating/mounting a different /gnu and see if it blows up or not - it looks like I rsyncd stuff across after running the first run <futurile>as usual my lab notes are kinda useful, and kinda a hodge-podge of things I then can't quite make sense of later <henrix>futurile: looks like there's GUIX_ALLOW_OVERWRITE to allow me to create those 2 dirs and simply proceed with those instead of aborting <henrix>the big question is: can those 2 dirs be symlinks? I guess I'll find that out <PotentialUser-38>I don't see an `suedit` on guix system. Why? Is there a good alternative? <ieure>PotentialUser-38, Because it's not declarative. Guix is big on declarative configuration. <ieure>Okay, maybe I don't know what suedit is, lol <ieure>PotentialUser-38, I see, that's not the same thing at all. <PotentialUser-38>I think it copies the file, you edit it on your userspace, then it uses its privileges to overwrite the target file. <ieure>PotentialUser-38, You asked about "suedit," but linked the man page for "sudoedit". I see bin/sudoedit in the outputs of sudo. <ieure>*the outputs of the sudo package <futurile>QQ: does anyone know how to test adding a news entry. I'm struggling to find a way to have it show up <ieure>futurile, Commit it, then pull from your local repo. <Chwoka3>trying to switch guix's main channel to the codeberg in my channels.scm so i don't have to keep entering --url. where can i find channel introduction and openpgp, or do i not need it for the main repo? <ieure>Chwoka, `guix describe -f channels > channels.scm' then edit to change the url and commit. <Rutherther>Chwoka3: the introduction is the same as for the channel you currently have in channels.scm <PotentialUser-38>ieure You are totally right. Strangely it was the third hit on my web search for "suedit". My bad. I'm sorry. <ieure>PotentialUser-38, Is sudoedit the thing you want? <PotentialUser-38>ieure No, it was suedit. But I guess it is not on guix because you don't package nvim. <kestrelwx>I wonder last time I used sudo to edit something. <PotentialUser-38>ieure I recall suedit being used on old Debian versions, so I am even more confused than ever. <ieure>PotentialUser-38, I'm not a neo/vi/m guy, but "nvim" is neovim, yes? Guix packages neovim. <PotentialUser-38>ieure I have seen so many vi forks that I am getting confused. Then I guess it is just a matter of writing a package for it. <ieure>PotentialUser-38, Guix is designed to be used by non-root users. Pretty much the only thing you use sudo for is `sudo system reconfigure' and `sudo system delete-generations'. <ieure>If you run `halt' like that, yeah. But you can power off from GDM, and there's probably some d-bus thing. <PotentialUser-38>Ah, d-bus again. Yeah, that reminds me: I have to see that `blight` thing too. <dariqq>PotentialUser-38: there is a package for opendoas but currently no service to configure it. But i guess it would just need to write the config file and make doas suid <Rutherther>if you're using elogind which most desktop users do as it's part of %desktop-services, you can just do "loginctl poweroff" <PotentialUser-38>I saw that package but I have not tried it yet. So... you are saying it is useless as it is? =L <gabber>i can't sign commits through ssh when i use my terminal emulator (kitty) but it works fine through the standard serial console (the one i switch to via C-M-F2). does anybody have any ideas why that is? it used to work but now it doesn't (without a change in config, just updates) <gabber>well, i do in the TERM=linux standard console, not in the kitty window <PotentialUser-38>Rutherther Oh, that is new to me. Let me write that down. Otherwise you would all get a confirmation that it worked in the form of a rude "PotentialUser-38 has left" '=D <ieure>gabber, Linux console != serial terminal. What do you have in your dotfiles to set up GPG agent? <ieure>I have a VT220 right here, now that's a serial terminal. <PotentialUser-38>I know little about it, but could gabber's problem be an issue with ssh-agent that is not passing some information through the ssh connection? <gabber>ieure: and yes, it's not a serial console. but it is a .. console? <ieure>gabber, Yes, the Linux console. <gabber>PotentialUser-38: that's what i thought, but it works from the same machine when not using kitty. <gabber>just tested with `st`. it works there, too <ieure>gabber, When you SSH, the agent connection gets forwarded back to your local machine. This requires configuration on both sides of the connection. <gabber>ah, we're on an ancient(?) version of kitty.. v0.21.2 vs 0.43.1 <dariqq>PotentialUser-38: Yes, if you want to use it youd need to set it up yourself (write the services for the doas.conf and setuid doas and maybe more). I would not recommend that if you are not familiar with how system configuration in guix works. <PotentialUser-38>gabber If it is an ssh connection issue, have you tried to ssh to localhost and try it again? <gabber>it is not. it works from remote's linux console and remote's st terminal to the host. it just doesn't work with kitty <gabber>ieure, PotentialUser-38: thanks for the input! <ieure>gabber, As for different behavior in different terminals: likely because shell startup files are incredibly wack and nearly impossible to actually use right, so the stuff you need is only happening in a login shell session, which kitty doesn't stat. <PotentialUser-38>gabber I am surprised that the spartan st works fine, but fancy and featureful kitty fails. <ieure>PotentialUser-38, Reject modernity, embrace tradition. <PotentialUser-38>ieure Good thinking. Is there a way gabber could store the environment from st and load it into kitty to see if it solves the problem? <ieure>PotentialUser-38, Easier to capture and diff. <ieure>`set > console.env' and `set > kitty.env', then `diff -u {console,kitty}.env' <trev>anyone know how i could persist a file in %gdm-file-system (tmpfs) across reboots? I want to store monitors.xml in ~gdm/.config/ <gabber>ieure: that's what i thought. but a diff between `env` in the linux console and kitty only showed the obvious differences: GPG_TTY, TERM and SSH_TTY <gabber>i guess some problem within kitty <PotentialUser-38>trev copy it to disk before shutting down? Not storing that directory on a volatile filesystem? Using some bind mount on the directory? Using an overlay filesystem (there is a Fuse one that is simple)? <futurile>ieure: figured it out, to show news you have to do two pulls to a tmp profile - so pull to something earlier, then to the new commit - that way if you provide --news it knows to show news since the 'old pull' <trev>PotentialUser-38: yes looking for a way with my system config and extending the right service <PotentialUser-38>trev Please excuse my ignorance, but why do you store the file on an tmpfs? <trev>PotentialUser-38: %gdm-file-system is a tmpfs that gets mounting by the gdm service (not my choice). all i want to do is stick a config file on it so it persists across reboots. there is a comment in the code on why it's a tmpfs - something about preventing caching problems <PotentialUser-38>trev I think the simplest solution is copying the file over after the moment the filesystem is mounted and before the file becomes useful. But that is nothing new to you. <FuncProgLinux>I think I have to update asap, I still need to submit two MATE updates and one of them rebuilds almost all of the desktop components, nothing major but yeah <tesseract>what other mainstream software that are exclusive to guix like icecat and icedove? <gabber>tesseract: no biggie, just the init system ;) <luca>(icecat and icedove are packaged for some other fsf approved distros, like parabola) <tesseract>luca: i actually wanted to say exclusive to free distros <gabber>after reconfiguration my pinebook pro doesn't boot into the new system. i got the serial cable so now i have some insight: it doesn't find the dtb file. i get thrown into the ,bournish shell. is it normal that in the initrd there are only 6 paths in /gnu/store? <FuncProgLinux>I have yet to experience the blessing of 100% libre hardware <ieure>futurile, Yikes, frustrating. <tesseract>FuncProgLinux: i think wifi wasn't a libre hardware <futurile>ieure: it takes a while to test, particularly if you slightly mess up the formatting and have to do it 2x 🤦 <FuncProgLinux>btw, any haunt users, do you use syntax highlighting on your blogs? :s <futurile>FuncProgLinux: it looks like Spritely Institute have highlighting <FuncProgLinux>futurile: Oh I saw that, but they use the guile-syntax-highlighting library <futurile>ah right, I'm not much a Haunt person, just been looking at their stuff as they had a "target" in the header of their site for their donations drive <FuncProgLinux>I was wondering if someone was using something else. Like Prism or highlight.js but I think I can work with that :) <FuncProgLinux>futurile: I just tried it yesterday, i was about to try "lume" but I realized it requires deno and by the time I get a package definition for that I might as well submit a patch lol <ieure>FuncProgLinux, I started hacking on a deno package, but it doesn't build :/ <futurile>heh heh heh - yeah you'll be getting back to your blog in about 3 months <ieure>The new Rust stuff should help with that. <ieure>But I don't really know how to work it. <FuncProgLinux>Same, I use it because my boss told me to. It's nicer than node though I have only tested it in Guix using podman containers and TRAMP-ing into them <ieure>Node never made a lick of sense to me. <FuncProgLinux>And even if you compile it from source I think you may lose the "compile" feature :( <FuncProgLinux>And that feature is a heaven sent when you need to use tools that are installed via "npm -g install" because CLI's and whatnot are now a single executable file <bavier>otoh, having a deno package would be useful for a future yt-dlp release, right? <bavier>they've been parsing JS with regexs for a long time, and lately decided they need a proper JS engine to handle youtube API challenges. <ieure>That's why I was looking into packaging it. <kestrelwx>Well it makes sense on a network only application. <FuncProgLinux>deno does make a node_modules directory but it's all symlinks to ~/.cache/deno <FuncProgLinux>it is expected to be userland only. AFAIK there's no "system deno" support :s <Guest95>Guix says I should source my profile's etc profile. Should I add this to my .bashrc, .profile, /etc/profile, PATH, or what? The way I use Guix is as a package manager on top of debian. Everything goes in manifests that I install with guix package -m, no guix profiles so everything is in the default profile <futurile>Guest95: it walks you through some steps and suggests half way down that you add it to your shell start-up file <luca>it may be easier to drop deno and make node a dependency and/or patch yt-dlp to use node by default <Guest13>so i tried a few things and im quite sure nginx is running and should be proxying Haunt's serving <Guest13>but the ports set up in my nginx config are not available <Guest13>a portscan shows them as non-existent (including 8000 and 9090) <Guest13>I've followed the manual on container networking but that changes nothing. `ip a` does show that veth is online <Guest13>what can i do to open the ports 80, 443 so that nginx can use them? <ieure>Guest13, What do you mean, "are not available?" <ieure>Is something listening on those ports already? <ieure>Also, not sure why you're port scanning localhost? `sudo ss -lntp' will show you anything listening on a TCP port, and what the process PID/name is. <Guest13>8080 is curently used by Haunt's http serve using the option `--host=any` and works without nginx <Guest13>im scanning from a different machine <umanwizard>Hi, anyone know if it's possible to have the operating-system definition (that I pass to guix system reconfigure) refer to things in other files? E.g. if I have `config.scm` and `foo.scm`, and `config.scm` loads stuff from `foo.scm` <umanwizard>Is the only way to do this to put `foo.scm` into a channel?