IRC channel logs
2025-10-03.log
back to list of logs
<gorilla>can guix manage NetworkManager connections on foreign distros? <gorilla>Or is that not even a thing on Guix System. I guess it isn't guix that manages that in System? <Deltafire>doesn't seem to, NetworkManager manages itself <FuncProgLinux>would it be better if instead of submitting one MR per feature on #2997 I submitted a big patch on mate.scm? <FuncProgLinux>(At least for updates, new variables are flour from another harvest) <robin>is there existing practice for defining dependencies on a non-extensible service? (maybe just documenting it, or adding bogus extend/compose procedures?) <robin>oh, i think i can just "extend" the dependency with (const #t) <vhns>Is there a way to programatically change the default C/C++ compiler flags in Guix? I don't care if I'll have to compile most things locally <apteryx>cbaines: bordeaux is sending me source archives at < 100 kB/s; is something up with it? <apteryx>e.g. gcc-cross-aarch64-linux-gnu-14.3.0 63.4MiB 30KiB/s 09:31 ▕████▊ ▏ 26.6% <apteryx>are absolute file names for #include in C possible? I don't think so, right? <apteryx>implemented-defined, but does what you'd expect in both GCC and Clang <bdju>guix was building chromium, which is frankly unrealistic on my machine, so I just canceled it. It had been going for a few hours I think. My CPU is still very hot and things are lagging. Is there possibly a stray process I need to kill? <bdju>I see some guix offload, guix substitute, and guix daemon --build-users-group guixbuild stuff in htop when I filter to guix. Can I kill these? <bdju>Please can anyone tell me what to kill? My CPU is at 94C still but I'm not doing anything. <bdju>Okay, I killed the two results for "build" in htop and things are finally cooling down. What the heck was up with that... That seems like a nasty issue. <bdju>I keep getting stuff saying "no space left on device" but when I run df -h nothing is out of space that I see. <bdju>Ah jeez I think everything is horribly broken. <bdju>guix install: error: setting synchronous mode: unable to open database file <bdju>My root partition has 145GB free. <robin>bdju, do you maybe have a tmpfs or similar of limited size? (or hitting memory limits? not sure what behavior that has) <bdju>I'm worried I need to reboot but I'm also worried if I reboot it will make things worse. I wonder if I'm out of inodes or something. qutebrowser broken from my partial upgrade and running the --no-grafts thing isn't working because of this database file error. <bdju>I haven't done anything strange or non-default that I'm aware, the only recent event was deleting my old channels.scm file in /etc/ then doing a pull and starting to upgrade all my packages. I then came back to my PC a few hours later and saw chromium was building so I C-c'd out of the build. It left a stray build process that was burning up my CPU so I killed it. Now here I am. <robin>(it looks like the default configuration doesn't use tmpfs for much) <bdju>I am not out of RAM either. <bdju>Something went awry with the attempted/aborted upgrade I just did I think. I know it's transactional package management but things didn't end cleanly so maybe some crud was left somewhere. <robin>"unable to open database file" sounds potentially bad; there's a comment suggesting it's an sqlite error <bdju>Well, the upgrade didn't finish. Would I need to rollback? <robin>(exactly once, using guix since 2018, i managed to crash the system in a way that corrupted /var/guix/db/db.sqlite. in that case it involved some reference counts being 0 and could be fixed with some careful sql) <untrusem>check if its the correct, I have borked enough time to know it by heart, but just double check <bdju>It's taking a long time. Is that normal? <cbaines>apteryx, regarding download speeds from bordeaux, I'm unsure, but feel free to create a guix/maintenance issue for it <cbaines>aside from bayfront just being slow, I'm not sure what could be wrong <apteryx>cbaines: I think it recovered. Will give it some time. <apteryx>but yeah, if it keeps coming back I'll report an issue. <adanska>apteryx, thanks for fixing the `perl-xml-libxslt` issue, i appreciate it :) <sneek>Welcome back civodul, you have 1 message! <civodul>csantosb: hey! yes, that’s a good idea <civodul>would you like to make a pull request against guix/artwork? <Ironsmith>hey guix! i know this isn't a guix issue but i was wondering if someone could help point me to the right direction. i have 2 video cards: GTX 570 and GTX 1060 so using nouveau for both. the older one has no issues, however the newer card can't run any wayland compository, i tried hyprland, sway, and river. is this a nouveau issue? a wlroots issue? <Ironsmith>btw both these cards have the same features in the nouveau compatibility matrix (except for some small differences like for laptops) <Ironsmith>also, i'm able to display the gnome login screen just fine with both cards, it's only when the wayland compositor kicks in that it loops me back to the login screen <Ironsmith>(for the latter/newer card, not the older one) <futurile>hello Ironsmith - interesting question about your video cards, sorry I have no clue - my workstation is so vanilla you wouldn't believe it heh heh <futurile>maybe ArchLinux wiki is a place to look, they often have good info <apteryx>was a bit puzzled as to why only the native (not cross) builds have the with-build-variables wrapper. <janneke>wget ... => HTTPS support not compiled in. <kestrelwx>Ironsmith: I'm not sure if yours is new enough, but newer series cards need firmware. <bdju>Is it normal for `sudo guix gc --verify="repair,contents"` to take 4+ hours or should I give up on it? <bdju>If I stop it, do a gc to delete old stuff, then start it again, will it be faster since there'll be less in the store? <bdju>Currently my main browser is not usable, and everything is a bit laggy while this runs so it's not a fun state to be stuck in. <untrusem>yep, it depends on how big your /gnu/store in bdju <kestrelwx>'qutebrowser -s qt.force_software_rendering chromium' <untrusem>bdju: `guix gc --verify=repair, contents` also delete stuff iirc <untrusem>but I am not 100% sure, not at my machine currently <bdju>kestrelwx: Starting it that way does not work, unfortunately. I still get the same popup. <bdju>I know the breakage is from grafts and I had undone the grafts the other day to work around it, but doing a guix pull today redid them and now I can't undo them due to the DB error since I had to cancel a build and my system started acting up. <bdju>Whatever I reported on the issue tracker the other day, not a new problem or anything, I just haven't been able to update anything in over a week. <bdju>Maybe it's fixed in master now, I just cannot upgrade at the moment. Everything is horribly broken. <bdju>Alright, I'm gonna cancel the verify thing and try something else. <untrusem>I mean, guix verify would have reported if something was corrupted in your store <bdju>It had been running over 4 hours taxing my CPU a lot with no clear signs it was really doing anything or when it would end... <bdju>I'm gonna clear out old stuff from the store with gc and then possibly run that again. <untrusem>bdju: can you link to your codeberg issue <bdju>I don't have a working browser at the moment. The issue was closed I think anyway. <untrusem>so your main problem is qutebrowser not working? <bdju>No, that's mostly unrelated. <bdju>I was simply trying to upgrade my packages but after several hours away from my PC I came back to see chromium was building and everything was slow, so I canceled the build, with the plan to skip the package and start the upgrades again. <bdju>However when I C-c'd out of the build it didn't stop all the processed, I had to kill -9 a build process to free up my CPU. <bdju>After that, a bunch of stuff says my disk is out of space, or the db is messed up. <bdju>So because of the broken state I'm in after attempting upgrades, qutebrowser is grafted and broken again, and I can't use the same workaround as before because I get the db error I pasted earlier. <bdju>Yeah, guix gc seems to be working. <bdju>Since the build didn't abort cleanly I suspect something didn't get cleaned up properly, but I don't know what to do about that. <bdju>None of my disks show as actually full. <untrusem>guix gc verify will verifx and repair the corrupted drv in gnu/store <bdju>How long is it supposed to take? A whole day? <untrusem>or if it can't repair something, it will throw an error <bdju>Like I said, I let it run for 4 hours and it seemed to not be getting anywhere. <Deltafire>i had a db error a couple of weeks ago and had to take some steps to repair it (i think it was basically exporting and re-importing the sqlite database) <Deltafire>not sure if guix can rebuild the database if it's deleted <Deltafire>i'm still getting file system not cleanly unmounted warnings when rebooting :( <untrusem>bdju: I don't know, depends on your system, if it is not throwing anything, it has not found any corrupted drv <ente`>I am trying to use system reconfigure with more or less the /etc/config.scm generated by the installer <ente`>it tells me /etc/config.scm:13:1: error: use-service-modules: unbound variable <ente`>but this use-service-models stuff was already generated by the installer? <ente`>(use-modules (gnu packages shells)) <ente`>(use-service-modules cups desktop networking ssh xorg) <kestrelwx>Yes, add `(gnu)` after `(gnu packages shells)` <ente`>I'm new to guile, what's the meaning of this module specification? I assumed I'm loading (gnu) as part of the other one <futurile>it's because `use-service-models` is in gnu.scm which itself collects together a bunch of stuff from that hierarchy <kestrelwx>No, you are loading `gnu/packages/shells.scm` only. <kestrelwx>Each entry specifies a module, and then Guile looks for the module at the matching path. <kestrelwx>`(gnu)` also exports `use-package-modules`, so you can use `(use-package-modules shells ...)` as a nicety. <apteryx>ente`: looks like your running the last released daemon, which is many years old now. there were fixes to make the tls connection more robust (retry on errors), so you should try updating it. <ente`>I'm also having trouble with the "guix pull" connection to savannah.gnu.org dying <ente`>can I already pull from codeberg? <ente`>glad to hear that's a bug that has been fixed :) <ente`>I installed guix about a year ago and then left the installation lying around until I just dug it out again, figured I want to give it a try again <ente`>didn't get familiar with nix (the language) and thought why not guix since I know a bit of scheme (and am not a fan of NIH languages) <kestrelwx>Oh, I was downloading the whole world apparently. 2k lines of derivations, 6k lines of 'looking for substitutes'. <nutcase>I'm trying to build a rust application (with close to zero knowledge about rust). I have a guix shell gcc-toolchain -D rust. When running cargo, I get " error occurred in cc-rs: failed to find tool "cc": No such file or directory (os error 2)". What should I add to my shell? <nutcase>guix shell rust doesn't give me cargo <characteristic>nutcase: --development rust is for developing *rust the toolchain*, not programms written in rust <nutcase>yeah, I thought, rust needs cargo to build itself ;) <nutcase>with guix shell rust:cargo, I get an error, which I don't get with -D rust "error: could not execute process "rustc -vV" (never executed)" <nutcase>characteristic: that's it (although -D rust also worked), thanks! <kestrelwx>You get the rust toolchain needed to build current rust, so you could've gotten an old rust. <nutcase>I'd love to use espanso and try to 1st make it run and 2nd try to package it. <nutcase>(if there was another package that does the same (like yasnippets everywhere), that would be sufficient as well <nutcase>characteristic: sure, sorry to be that business customer like. "does not work" as in "building the software does not work, since a dependency (smithay-client-toolkit) of the software does not compile" <characteristic>chances are you are going to have more luck using the crate importer <nutcase>finally this works: "guix shell gcc-toolchain wl-clipboard dbus openssl wxwidgets libxkbcommon rust rust:cargo pkg-config" <kestrelwx>See if you can package it with `guix import` that'd be useful. <nutcase>However, before putting too much effort into this. I am aware of the issue with that tool (it needs to read your keyboard). How do I set "sudo setcap "cap_dac_override+p" $(which espanso)" <nutcase>Do I have to add espanso to my setuid programs in the end or is there another way? <nutcase>kestrelwx: it does not seem to be available as crate, yet <futurile>that would have been a problem in the previous system, I wonder if the new origin crates thing will work for it <nutcase>futurile: how do I use it (aka where is the doc)? <nutcase>characteristic: guix import crate espanso did not work (as in "guix import: error: failed to download meta-data for package 'espanso'") <nutcase>@futurile is guix import "the new origin crates thing" you were referring to? <dthompson>futurile: is there a preferred donation method? in other words, who charges you the smallest fee? <nutcase>characteristic: futurile just for my understanding: Even if espanso itself is not available at crates.io (and cannot be guix imported), I have to define packages for all its dependencies (maybe a lot) and this can be done by `guix import crate <dependency>', right? <characteristic>nutcase: you should be able to point the importer at the lockfile and it will do its thing <nutcase>that worked. Then I have to explicitly add the generated crate packages via (crate-inputs) to the inputs of my package definition. Is there also way to do automate that (aka read the lockfile and manipulate the package definition)? <futurile>dthompson: for you (dollars) Stripe is the best option. <futurile>dthompson: I couldn't figure out how to let people do any number they want for a recurring donation (you can for the one-off), but you can change the quantity at each level <futurile>dthompson: I am by no means a Stripe expert so if you know of better ways very open to suggestions <futurile>nutcase: there's also a blog post about the new crate system ... <janneke>ah, wget's "HTTPS support not compiled in" is a funny bug; try: <janneke>wget https:/guix.gnu.org #notice the single / <dthompson>I've never used stripe so I know nothing about optimal setup :) <futurile>I think it's all good for a basic set-up. Fingers crossed it works for everyone and we get a good response! <jlicht>futurile: gz on the fundraising efforts! <simendsjo>I notice avahi-daemon (5353) and ntpd (123) listens on wildcard addresses by default. Is this necessary? Is it a risk? I don't want to expose more than necessary. <futurile>janneke, jlicht thank-you - feel free to spread it around on all your favourite social media location <apteryx>I don't get how I can insert unicode characters in inkscape <apteryx>I'm trying to get inkscape to draw U+1F91D in a 'guix shell inkscape font-google-noto-emoji fontconfig' environment <janneke>...but, you can also open the .svg in emacs and just change the text there :) <janneke>ACTION has been fighting with inkscape guis and unicode too, lately :) <apteryx>it's a font that I want to transform to SVG <apteryx>for the commercial support page, trying another icon as suggested by robin <apteryx>hm, I could input "1f91d" with C-u, and even saw the icon appear next to the unicode text input box, but then it doesn't draw in the main view <janneke>ACTION found that many unicode characters were missing in a boldface font that do work in normal face; but i guess you're looking at quite another problem <apteryx>I've got 'Noto Color Emoji' chosen as the text font. Odd. <janneke>ACTION found that icecat shows characters in [inkscape's] svg that inkscape doesn't show, but hasn't found a way to convert that to png with the characters in tact (imagemagick's convert didn't pick them up either) <janneke>hmm, come to think of it, scaling it waay up and taking a screenshot could work :-/ <ente`>is it normal that "guix pull" takes forever? <tesseract>git pull always downloads and builds something and takes time. is it normal? <ente`>Rutherther: how do I change the url to codeberg? <ente`>I have to get going at some point and I don't want to have to restart it later on <janneke>Issue created Jan 23, 2019 by Rik Faith <janneke>however, possibly we could blame "the browsers"/firefox for not contributing back <kestrelwx>apteryx: I don't see emojis without Noto Fonts either. <apteryx>could be that the regular color variant had only a variable font, while that monochrome one had static fonts as well <futurile>civodul: thanks, glad to launch it! Fingers crossed <Deltafire>"Admission as a member will be confirmed following a vote by the Solidary Administration Council" - how does that work? <Deltafire>bit weird, i'd expect a confirmation email from Strip at least <futurile>Deltafire: to join you have to be confirmed by the SAC which is a board that makes decisions on behalf of the members - it's designed to prevent some bad people suddenly joining and Association and changing the intent of the group. See the regulations here: https://foundation.guix.info/statutes/statuts-201602-en.pdf (this is the English translation) <futurile>Deltafire: imagine $evil-corporation had 100 people join, that would immediately outvote the 40 current members, so it's a pretty mad idea, but a useful protection. <Deltafire>is there normally a payment confirmation email? I'm beginning to wonder whether i typo'd my email address <futurile>Deltafire: I have to go click it manually, I'm doing it now ... <futurile>you should have it now, tell me if you don't <futurile>great - thanks for supporting Deltafire :-) <untrusem>btw I think it would be nice to have some kind of forum or discussion area like github, where you could get packaging help and other thing by tagging those specific teams like @gnome, @emacs, and it won't be ephermal like irc <Guest42>Hello, I installed GNU Guix yesterday in my VM with the latest image of the website. Now I want to run guix pull but fails with operation timed out. Therefore I tried again with the codeberg URL but it timed out again. Here the full error: guix pull: error: Git error: failed to connect to git.guix.gnu.org: Operation timed out and guix pull: error: <Guest42>Git error: failed to connect to codeberg.org: Operation timed out <Guest42>I can ping domains so internet works. Maybe git problem or something with ssl? <characteristic>Guest42: you can ping domains, but can you clone the repository? <ieure>Guest42, Seems to work fine for me. <janneke>ieure: Guest42 may (still) be pulling from savannah, instead of codeberg? <Guest42>I have another VM with the 1.4.0 installer images and Guix pull works fine <Guest42>characteristig: you mean I shall git clone manually if that even works? <ieure>janneke, They specifically said they were using git.guix.gnu.org <characteristic>Guest42: yes, try doing a clone, even if a shallow one, manually and see if that works <ieure>untrusem, Builds run in an isolated environment without networking, builds cannot download random stuff from the internet. <ieure>untrusem, It looks like it's trying to clone dependencies it needs, you need to provide packages for those instead. <ieure>untrusem, You may also need to patch the package source to not `git clone' / point to your package / whatever. <untrusem>also when I clone their main repo, there is no meson build weird <untrusem>these are one of those times, when having a github account would help as I can't find any other communication source. <ieure>untrusem, Yeah, no idea. There's some Nix packaging, maybe that's similar enough to guide you. <untrusem>it just makes an python venv and install from there lol <janneke>fwiw, /me just guix pulled in real 1m48.294s <Guest42>characteristic I did manually run git clone codeberg --depth=1 and it works <RavenJoad>If I have a couple of issues open on issues.guix.gnu.org, should I move them to Codeberg? These issues haven't been looked at for some time. <Rutherther>RavenJoad: I wouldn't say you should do it, but imo you definitely can, but please don't forget to close them + leave a message that it has been moved so that it's clear. There is better chance of them being looked at on codeberg <RavenJoad>That's what I was thinking too. The issues I have open have been open since (roughly) April. So I want some more eyes on them. <RavenJoad>Could someone give me a pointer on running Guix's whole-system tests? The ones in gnu/tests/*.scm specifically. I am adding some stuff there and need to run them. <RavenJoad>make check TESTS='gnu/tests/virtualization.scm' does not seem to actually run the tests. <RavenJoad>Nevermind. I should have just kept reading the manual. make check-system TESTS='virtualization' doesn't run anything, but TESTS='basic' does. At least I have something to start with now. <tesseract>guys, icecat got update today. it is the same version number. and it told me something like "the software or its dependencies gets update" or something. does that mean, if let's say ffmpeg is a dependency of icecat, i can't see and know it was the one being updated? <ieure>tesseract, I'd recommend reading the Nix paper. <ieure>tesseract, Nix is the system Guix is based on. <ieure>Nix : NixOS :: Guix : Guix System <tesseract>well, i won't read it probably. i am just using guix for icecat <ieure>tesseract, Both Nix and Guix rebuild packages when any of their inputs change. I don't know if there's a good way to see what specifically triggered it if you're pulling/reconfiguring. <ieure>tesseract, Why are you using Guix just to get IceCat? <ieure>tesseract, Set what to download binary versions? <kestrelwx>You don't really need to pull & upgrade very often, since IceCat is on a stable release schedule. <ieure>tesseract, Yes, it's fine, you just phrased things oddly. <Rutherther>tesseract: you definitely can check what was updated, but unfortunately I don't know of a tool that would summarize it. You would basically want to check the differences between "guix graph --type=references /var/guix/profiles/per-user/guix-profile-X-link" and guix-profile-Y-link. There could be a tool making this into a nice summary. <ieure>tesseract, Your distro doesn't provide packages? <tesseract>there is also librewolf but i trust icecat more <ieure>tesseract, There are Librewolf packages for Debian. They're provided by LW, they're not part of Debian. Never had issues, though. <ieure>tesseract, In any case, choosing to use Guix substitutes is orthogonal to inputs causing rebuilds. <ieure>The changed input will cause a rebuild, the rebuild will create a substitute you can download. <tesseract>by the way, icecat gui is more responsive than debian's original firefox somehow! <encryptedwhispe->Hi! I have hopefully a simple question: If I am trying to debug issues with my manifest.scm, and I have a package that uses a git source, is there a quick way to ignore/disregard the package cache/store entries so it can be built fresh? Can --check be used for this? <encryptedwhispe->related: if I use --check does that refresh the store when it fails and says the check failed? Or does it just keep the previous store entry? <ieure>encryptedwhispe-, Unless your package is intentionally not reproduceable, deleting stuff from the Guix cache won't do anything except waste compute to recreate the same thing you just deleted. <encryptedwhispe->The thing is I have working changes that I would like to build off. Should I modify the package so it uses a local directory, instead? <ieure>encryptedwhispe-, If the package is broken, you'd typically update the commit and SHA, or the package definition; either will cause a rebuild. <encryptedwhispe->I see. so I should commit the changes I make, and use the new hash. That makes sense <ieure>Anything using a Git source has a commit ref and SHA, unless you change the package definition, it'll either rebuild the exact same thing, or, if you're using a non-commit ID ref and force-push changes there, it'll fail to build because the SHA of the old code doesn't match the new. <Rutherther>guix refresh is for updating versions of packages and the hash of the source, yes <kestrelwx>I've been running `guix download` myself the whole time like it's not quarter century past '00 <ieure>kestrelwx, Not all packages work with `guix refresh'. But for the ones that do, it's great. <encryptedwhispe->Another question: If I have a guile package which relies on guile modules, and I'm using `guix pack` to make a docker image, do I need to include all the module packages in the manifest, or should the dependencies inside the package def make guile available? Using gnu build system. <Rutherther>encryptedwhispe-: if you have a guile application/script, it should typically be wrapped to have all the dependencies. In case it's a library, it should propagate the dependencies. In both cases, you don't put the dependencies to manifests <ieure>Agree, you generally don't have to care about dependencies with Guix. <encryptedwhispe->Currently the app has two executables (in the scripts folder) that import guile modules and do some operations <encryptedwhispe->If I don't include the dependencies in the manifest they say "no code for module xxx" <encryptedwhispe->They are defined as .in files with #!@GUILE@ --no-auto-compile at the top <Rutherther>there are plenty of examples of wrapping with guile load path in the guile-xyz in guix repo <nutcase>(cargo-inputs-from-lockfile) exists in the source, but I can't find any usage, yet. Does anyone know, if it is ready to use? <Rutherther>nutcase: should be ready as far as I know. There are no uses, because it doesn't make sense for guix. It's mostly for obtaining development dependencies for rust projects that you're working on and using a manifest.scm / guix.scm <trev>nutcase: every time i see your name i get a good chuckle or grin <nutcase>Rutherther: after reading the blog post again, I see, what I missed during my first read. <nutcase>trev: I'm happy to create some joy :-) <Guest24>anyone have a Forgejo runner? herd register forgejo-runner url token this doesn't work for me. After enter nothing happens even after minutes <Guest24>are Guix shell containers LXC containers or something own? <civodul>it uses the “namespace” functionality of the Linux kernel, which is what LXC, Docker, etc. etc. all build upon <ieure>Deltafire, Guessing, but it likely means "porting to Spritely Goblins"