IRC channel logs
2025-03-17.log
back to list of logs
<redacted>coyotes4ys: note that you can search through the available home services with `guix home search` <redacted>didn't notice that jA_cOp already said that <coyotes4ys>well one "warning": cannot determine provenance for current system <coyotes4ys>and "error: service 'pipewire' requires 'dbus', which is not provided by any service" <coyotes4ys>do i need use-module (gnu home services desktop)? that's what it has under location <coyotes4ys>ok can i ask some questions? i did container first, then tried reconfigure and it was really short, no downloading. <coyotes4ys>did it use the stuff downloaded during container? <coyotes4ys>redacted- do you know? just my last three lines for starters <redacted>It reuses anything that was already built for the container <coyotes4ys>but it did say warning: cannot determine provenance for current system" <coyotes4ys>i can't seem to see the search results when i click them. i am in torbrowser, are nodes blocked on issues.guix.gnu.org? <old>is it possible to remove encryption of filesystem for a configuration? <ulfvonbelow>update: seems like the farstream issue involves two different threads adding the same "rtpopuspay" filter at more or less the same time and the second one fails an assertion because it's already present <meaty>One thing I look forward to with the introduction of a forge-based workflow is the opportunity for better collaboration on a patch series, even with non-committers (via agit-flow)(?). I imagine there could even be a "help wanted" label people could put on their branches <meaty>right now I think there's no established way to post or look for unfinished work to help with <meaty>do people use it that way? It seems to be mostly people sending in finished work for review, not half-finished work they need help on <wakyct>if I have an issue with a pkg I frequently check issues.guix -- if it's unsolved I'm usually motivated to help find a solution <wakyct>I don't imagine it'd be different with a forge-based interface <wakyct>I'm not usually just looking for something to do <meaty>i was thinking a forge/agit flow-based system would make collaborating on branches easier because people's feature branches would be already part of a git tree, making it that much easier to pull, edit, and push without the added layer of mailing list protocol <PotentialUser-12>How many emails do you need manually approved before they are auto approved? <vhns>It seems passing `--dry-run` to `guix system delete-generations` does not actually yield a dry-run <ulfvonbelow>it seems that code snippets aren't being rendered in devhelp, at least for gstreamer-docs <ieure>meaty, I am excited for Codeberg. <moksh>so how to get japanese and emojis work in guix <ulfvonbelow>see "Application Setup" in the manual, specifically the "X11 Fonts" subsection <hako>and choose a font you like <moksh>hako: I install the aporetic font <moksh>but I am not seeing the font anywhere <moksh>how can I make it the default font for a user <ulfvonbelow>might have to relog to pick up a fontconfig environment variable? <moksh>so I am getting this error for every guix command <moksh>guix home: error: error parsing derivation `/gnu/store/05rjz8brap4nwq307ipz2bnb29ap1gsd-rust-dirs-next-2.0.0.tar.gz.drv': expected string `Derive([' <Rutherther>That means the derivation file is corrupted, most likely empty. To fix files like that, guix gx --verify="contents,repair" can help as long as they are substitutable. Should be the case with this one <theotherone>Hello, I am trying to set up guix home but I am having issues with reconfigure. In the configuration I am setting up zsh, dotfiles, dbus and mcron (no packages). When I want to reconfigure, it says some git hashes aren't descendants and I should use --allow-downgrades. Also I cannot do guix pull anymore with error 'symlink: File exists: current-system'. I'd appreciate any tips :) <olndrxyz>Hi, I'm trying to write a module that includes all my xorg services and config. I'm using this path for the xorg module: ~/src/guix/modules/services/xorg-custom.scm. The base os is in ~/src/guix/systems/laptop.scm. Here is what I came up with: https://paste.debian.net/hidden/8df7dd68/ (xorg-custom.scm) and https://paste.debian.net/hidden/062b6992/ (laptop.scm). The reconfiguration with guix system -L ~/src/guix/ reconfigure ~/src/guix/systems/laptop.scm <olndrxyz>fails with some cryptic errors that I'm not able to debug. Could anyone kindly tell me what's wrong in my code? <Rutherther>theotherone: where is the guix instance you are you using located? "type -p guix" <Rutherther>olndrxyz: you seem to be trying to apply(call) your variable as a function. Remove the parentheses in your system config <theotherone>Rutherther: It's /run/current-system/profile/bin/guix. When I do guix describe I am getting a different hash then the one in .guix-home/channels.scm. Also in /var/guix/profiles/per-user current-guix points to current-guix-1-link, which I assume is pretty old considering there is current-guix-134-link. <Rutherther>theotherone: okay. Something is definitely wrong with the instance you are using. Can you try just guix pull, but with guix out of the 134 folder you are talking about? <apteryx>something built for a fixed arch, such as ovmf-x86-64, can be added in native-inputs, right? <apteryx>since it's firmware for the fixed target, so doesn't depend on host, nor is it intended to run on host or target. <apteryx>I guess it's kind of N/A to be correct, it could be added to one or the other without any effect. <theotherone>Rutherther: guix pull fails with symlink: File exists: "/var/guix/profiles/per-user/alice/current-guix" <theotherone>I linked current-guix to the the 134 version with 'ln -s' <h4>When I modify a package definition it seems to want to fully compile locally the whole profile instead of said package and its dependants. It's curious, does the substitutes aswell recompile all packages in existence everytime one of them is updated? I think not looking at their job logs <olndrxyz>Rutherther: I had already tried without parenthesis and I get: "...laptop.scm: error: 'services' field must contain a list of services" <theotherone>Rutherther: guix pull doesn't complain anymore. In the output of guix pull the short commit hash of guix (fd19785) is different from the one I get when running guix describe (5058b40). Is that expected? guix home reconfigure still wants me to use --allow-downgrades... <h4>I already had every binaries in store locally, it should have only downloaded new package and their deps source code to only compile that part and re-use local derivations for all the other packages <h4>But right now it's giving me a Gentoo experience <theotherone>Rutherther: After running guix home reconfigure, guix pull stops working. <bjoli>So, for plasma I have some issues getting discover to work. Starting it shows the error "the name org.freedesktop.packagekit was not provided by any .service files". It is only for some flatpack apps, but it would be a nice thing to have... <Rutherther>olndrxyz right, your variable is a list, so now you have list inside of a list. You need to append the two lists. <Rutherther>theotherone: no, it is not expected normally. But what instance of guix are you running now? Is it the new one or still the old one from current-system? You might need to relog to get the correct path. <bjoli>Oh, no worries. I found the warehouse flatpak app. <olndrxyz>Rutherther: How do I append the two lists? Could you give me an example? <theotherone>Rutherther: hash guix doesn't return anything anymore, paths are also a bit messed up, programs aren't found. When running guix home reconfigure, it pulls a specific guix commit, the same one guix describe shows. I moved the packages into the home configuration in hopes of saving the system. Is guix home with packages meant to be used with guix package -i <package>? <Rutherther>theotherone: no, guix home command is used with guix home <h4>like it tries to download pcre-8.45.tar.bz2. But I have already 10 profiles, 2 binaries, 10 builders, 8 bz2 drv, 15 drv and 2 gz of it. It's the same code yet it retreive it for the nth time <h4>and like all the rest it wouldn't have anything to do with what I try to compile <theotherone>Can somebody explain why guix describe and guix home describe return different guix revisions? <h4>if Guix OS is an island, then Guix PM is an island into it and Guix Home yet another nested <redacted>Is there a convenient way to search the load path for a specific symbol? <redacted>For example, I don't remember which module exports the 'service-type symbol <olndrxyz>Rutherther: if I add (append xorg-custom-services) I still get the same error. Maybe I have to add that before declaring all my services? Isn't (cons* supposed to allow multiple lists? <efraim>hako: I started working on an awk script to create a Cargo.lock file from rust-crates.scm to feed to cargo-audit, it ended up being shorter than I expected <efraim>I'd like to parse the other packages too but I don't want to create a real parser for guix packages in awk <efraim>although now I think I can extend what I have and create enough of a dependency tree that we should be able to see which package has the vulnerable input <jean0t>I love guix and its purpose but software is downloading in a Kb/s rate <jean0t>even minor softwares like moc are taking several minutes <jean0t>absolutely impossible to use it as a daily driver <ieure>jean0t, Yeah, the substitute server does that, I've no idea why. Are you in North America? <divya>lilyp: Would you please finalize this series of reviewed patches? The earlier `yara` patch was for this: <divya>It's already been there for a year. <rekado>silly idea: a package with *.desktop files, all altered to run "guix shell", maybe with a little progress bar prepended, so that any application can be run from a desktop environment. <Rutherther>olndrxyz: cons* allows exactly one list as last argument, other arguments are elements. You need to merge the two lists you have, append with just one list will not change anything <moksh>but getting unbound variable kmonad-service-type <moksh>trying to get a service for kmonad <dariqq>was there a big gtk rebuild in the last few days? Had to recompile both webkitgtks for i686 today? <csantosb>`mumi search emacs` is not supposed to be the same ? <ieure>jean0t, See "(guix) Getting Substitutes from Other Servers" <jean0t>but I dont know where to get the links <efraim>hako: I didn't want to figure out matching guix-names to cargo-names so I dropped that, but I did get get an easy check for regular packaged crates <jean0t>using tor makes the download faster <jean0t>2 options, or the server is halving the conection to me or it is my isp <rekado>jean0t: which of the servers is slow for you? <rekado>ci.guix.gnu.org is located in Berlin/Germany, and while it has a good connection we know that from some ISPs connections are throttled. Not clear where exactly this happens, though. It appears to be out of our control. <jean0t>I used this command in the other terminal <jean0t>guix time-machine -- weather --substitute-urls="bordeaux.guix.gnu.org" <jean0t> llvm-for-mesa-18.1.8 33.9MiB 9KiB/s 03:42 ▕█ ▏ 5.9% <jean0t>some packages go 500Kb/s others go 10Kb/s <rekado>and with ci.guix.gnu.org it's not better, I presume? <jean0t>I was wishing to substitute my system to the guix system because the idea of this package manager was amazing, I even started reading the manual <jean0t>and to update the first guix pull was half a day to a simple thing <jean0t>I want to change my software and all to libre software, but I tried trisquel in a VM and same internet issues <jean0t>I thought that it was an ISP thing so I changed dns and used even tor, but same network speed <jean0t>probably focused in european and american users <jean0t>I thought it wouldnt be a problem since I installed void linux before from a server in california and was fast asf <jean0t>but Ig the gnu servers arent as good as them <wakyct>maybe Guix could fundraise to set up substitute servers in other parts of the world <decfed>how do I undo a guix home reconfigure? And I do not mean roll-back but going back to managing dotfiles manually <moksh>jean0t: I want to try that but It is not packaged for guix <moksh>and packaging a rust application is not a easy tast <jean0t>rust can dream, but C will be superior <moksh>I have yet to contribute a patch to guix <jA_cOp>decfed: whenever you run guix home reconfigure and it replaces one of your dotfiles with a symlink, the original dotfile is backed up to ~/<timestamp>-guix-home-legacy-configs-backup. You can simply copy the file from there (or your guix home repo) to overwrite the symlink <moksh>I am trying to upgrade this package for guix <jA_cOp>decfed: then of course, make sure to either 1) not run guix home reconfigure again after, or 2) remove those files from your guix home configuration so they don't get symlinked to the store again <ieure>jA_cOp, I think you have to delete the symlink, then rename the original. I've tried this before, and copying over the symlink tries to replace the target, which is in the store, which you can't write to. <jA_cOp>yeah I can imagine many tools working like that, it all depends <moksh>can anyone guide me with this patch <moksh>I have updated the commit in the uxn patch <moksh>guix package: error: cannot install non-package object: #<unspecified> <moksh>here is my package defination <ieure>moksh, What command are you running? <ieure>Well, if that's what moksh is running, try `guix build -f uxn.scm' (or whatever). That explains what's wrong; odd that installing doesn't. <untrusem>why does build show what stuff was missing and install doesn't? <ieure>I don't know! Worth filing a bug IMO. <ieure>untrusem, It's just sending an email. <untrusem>how to easily see build logs, I need to unpack the tar.gz and lookup each time <ieure>untrusem, Not sure what you mean. The failed build will print the path to the build log, it's a gzipped text file. <ieure>untrusem, Emacs will transparently decompress it if you open it. Or zcat, or zless. <untrusem>I need to unzip it everytime to a temp di <ieure>If you're using Emacs, find-file-at-point is your friend here. <ieure>untrusem, You're using modify-inputs in #:phases. You want modify-phases. <ieure>modify-inputs is, unsurprisingly, for package inputs. <untrusem>but why did it worked for the previous version <ieure>If a previous version built, it must not have had that error. I don't believe it's possible for it to build with that error. <untrusem>I should also sent a patch for this as well <untrusem>what package do you use for email in emacs <ieure>I hear good things about notmuch, but haven't tried it. I have tried Gnus and find it impenetrable. <ieure>I used to use vm for it, that was a very long time ago. <ieure>mu4e is alright. The search is awful. <untrusem>I would like to manage mail setup with guix as well <ieure>The MUA options are more or less fine, but I haven't found an IMAP sync tool that I think is actually good. <untrusem>If I use gaijm it asks for password everytime <ieure>Both offlineimap and mbsync can't cope with multi-machine setups, they make a bunch of duplicate emails. Annoying. <untrusem>so If I have pipewire service enabled in my home, does it removes and replaces the original one? <ieure>untrusem, Adding a pipewire service won't remove a pulseaudio service. <Rutherther>you basically cannot get rid of pulseaudio. It's hardcoded in the clients, when no pulse socket is found (which can be emulated by pipewire), they will start pulse <Rutherther>you would need to change the pulse audio library itself <Rutherther>(or if you just want to prevent the auto start you can also configure it with a config) <jA_cOp>the home-pipewire-service-type will configure pipewire to run the pulse-compatible server by default, so unless you have a good reason to do it differently, you can just use the pipewire service without the pulseaudio service <Rutherther>jA_cOp: imo configuring autospawn no is good reason for pulseaudio service along with pipewire <ieure>untrusem, Also delighted to see librewolf in your home config (I contributed/maintain it). <jA_cOp>Rutherther: socket-activated, then? It's probably possible with pipewire-pulse too, even if the Guix service doesn't support it <ieure>untrusem, For sure, it's my daily driver browser, so, a blocker to daily driving Guix if it's not available. <Rutherther>jA_cOp: sorry I don't know if I understan what you mean. What I meant is that the pulseaudio service puts autospawn no to its config for me, which prevents from pulseaudio spawning new clients in case pipewire is restarting for example or just a race condition <untrusem>ublock origin + Redirector are essenstial to my browsing <untrusem>I am just stealing tidbits of config from you people's config <jA_cOp>Rutherther: ah I see, so the pulseaudio-service-type is just used for the client config? <Rutherther>I guess the env var PULSE_CLIENTCONFIG is important as well <ieure>Anyone else noticed that `sudo reboot' doesn't cleanly unmount the root FS anymore? <ieure>Every time I reboot, I get a "recovering journal" fsck. <ieure>Been happening for a couple-few weeks. <untrusem>ohh then I would have to type three password <h4>When building profile I get: %exception #<&invoke-error program: "make" arguments: ("-j" "12") exit-status: 2 term-signal: #f stop-signal: #f> <Rutherther>h4: that doesn't seem like an error for building a profile, seems like an error for a package using make <h4>Rutherther: Yeah while building it a package was building itself with make and poof <Rutherther>h4: I don't really understand what you're trying to say. Anyway can you send the log? <h4>Rutherther: qtbase-6.7.2 <h4>Something ultra-long to make and ultra-prone to fail, yay <Rutherther>h4: qtbase 6.7.2 substitutes for me in latest guix, so it builds fine <untrusem>I am getting invalid field specifier error <Rutherther>untrusem: for what field are you getting this error? <Rutherther>untrusem: if services field, then you will need to send more from your config, because it's impossible to say from what you sent <necto>Hey `guix home` users! I'm wondering if any of you solved the following issue: Emacs often reaches out to modify its own config: most prominent example for me is the .dir-locals.el file in the Guix repository: when I open the repository it proposes to apply the settings and tries to save my decision. On another occasion, I want to ignore some <necto>warning from a package and remember that decision. Finally there is the famous `customize`, which I admittedly don't use. <necto>So now enters `guix home`: it shovels my Emacs config into an immutable /gnu/store directory entering into a direct conflict with Emacs. <necto>As a result, I have some unsaved config files in my session which I get prompted to save before I exit. <necto>Did anyone of you also experience this? <untrusem>I don't manage my emacs config with guix home yet <Rutherther>untrusem: the issue is that you have two things in your services field <Rutherther>untrusem: services should be one list afterall. Check the parentheses <Rutherther>h4: please be more specific - what package did you change and what compilations did you not expect? <Rutherther>and what you mean by 'build profile', because you don't seem to mean actually build a profile package that just joins multiple packages, according to your previous messages <h4>Rutherther: As said in the URL I did not expected full profile compilation <h4>Rutherther: I changed Mesa <h4>Rutherther: I mean, well, the thing you do to get new packages <Rutherther>h4: I really don't understand what you mean by full profile compilation. Profile is compiled every time a package changes <h4>Many lexic in such distribution concept that I don't get <Rutherther>profile is just collection of packages so it needs to get rebuilt each time any package in it changes <h4>Rutherther: I just mean that where I expected all package to be got from substituted, it retreived none, ordering a full local compilation, about the exact lexic, I don't have it <h4>All the time I did it it just got everything from substitutes, I expected here to get everything from it except what it hadn't and related dependencies <Rutherther>h4: in functional package managers, when dependency changes, the derivation changes => rebuild. Most stuff depends on mesa <Rutherther>if you're just making a simple change maybe grafts would be better for you <h4>Rutherther: Here literally everything was recompiled locally, even what has literally nothing to do <Rutherther>h4: send an example of what you didn't expect to recompile <h4>I mean that if that change was pushed on substitutes I'm pretty sure substitute won't recompile every existing package it hosts. Only the changed one and it's dependants, but it wasn't the case for me <Rutherther>yes, exactly, only dependants on what you changed. But since you haven't shown what change you did exactly and what was rebuilt that you think is not a dependant, I cannot tell you further what the issue was <h4>Rutherther: How can I get the logs of what was compiled? So I can search what was not supposed precisely <Rutherther>h4: guix doesn't really keep that information afterwards, it's just in stderr of the guix build/other command process <podiki>a change to mesa affects nearly 6k packages <podiki>and speaking of which, will start on a mesa-updates branch, i think it will need vulkan updates too (and will see what else needs bumps, libdrm, wayland-protocols are usual) <podiki>though with freedesktop doing a git migration and many branches waiting on the guix side, will be a while <Rutherther>there is for sure new version of wayland protocols <futurile>h4: I think what's going on is that you're thinking "the new mesa package has been compiled so why does everything have to change", but the way Guix works (because it's functional), if an "input" changes (mesa) then that package that uses mesa will also be changed (rebuilt) - that's how functional package managers work. <untrusem>futurile, is your guix packaging series still relevant? <futurile>untrusem: yes, AFAIK it's all still correct <futurile>untrusem: I did a video on my Youtube about packaging reasonably recently - didn't write an equivalent post to cover that yet <futurile>untrusem: tell me if you find mistakes :-) <janneke>it seems my (encrypted) root filesystem isn't unmounted cleanly upon reboot "lately" did anyone observe something like that? <Rutherther>janneke: ieure said the same thing few hours ago <janneke>ACTION usually never reboots but has been "playing" with building kernels since yesterday <vagrantc>i have been noticing unclean unmounts even with an entirely unencrypted rootfs <vagrantc>i've tried with halt, reboot and ctrl-alt-delete and they all seem to not quite finish ... <vagrantc>although *sometimes* it unmounts correctly ... or at least, there are no unclean fsck messages at startup <jonsger>janneke: yes, have seen this on my PC as well. Filesystem is ext4 <janneke>jonsger, ieure, vagrantc: and i guess all with shepherd-1.0.2? <janneke>it mentions that it only happens after guix system reconfigure for them...but i also haven't rebooted unless i did a reconfigure... <jonsger>janneke: nope, I'm still on shepherd-1.0.1 <vagrantc>with (services ... (elogind-service)) i get a complaint that it is deprecated and to switch to elogind-service-type ... but (services ... (elogind-service-type)) wants a configuration or something ... but not being an elogind expert, what is a sane default configuration? <vagrantc>it sure has a lot of configuration possibilities <jonsger>janneke: I rebooted now, but this time there wasn't any cleanup file systeme messages... <Rutherther>vagrantc: with service types, you need to create the service - (sevice elogind-service-type). The default config is fine <vagrantc>i have always known this was possible, but someone recently pointed out while i wait for my shiny kernel to possibly get merged, i can define the kernel package in my system configuration ... and then just use guix pull off of master ... keeping up to date without waiting for stuff to get merged. <vagrantc>Rutherther: oh wow. that was so simple. thanks! <vagrantc>it even results in the exact same system :) <janneke>jonsger: without system reconfigure? <janneke>ACTION had three clean reboots, it may wel only happen after a system reconfigure <ieure>janneke, Happens on every reboot for me, both after `guix system reconfigure' and not. <rekado>I've also seen apparently unclean shutdowns on a recently updated system. There does not seem to be a good reason for it. Possibly related to reconfigures, but I'm not sure. <quezt>how may one replace ntpd with chrony? <simendsjo>Looks like libvirt is broken (again). Probably the recent upgrade to 11.0.0. <jA_cOp>I'm interested in running chrony too, just not a priority for me right now, I'll get around to it eventually :P <ieure>I want to evaluate a (computed-file) call, but am having problems with it. The manual says I can use the ,run-in-store meta-command, but this doesn't seem to work. <ieure>But if I `guix repl' and then `,run-in-store (package->derivation hello)', I get "Unknown meta command: run-in-store"