IRC channel logs
2024-06-06.log
back to list of logs
<anthk_>any news on the freedink package? it still doesn't build <user11037>trying to install guix and in early boot I'm getting "in procedure statfs64: not implemented" from the statfs call at gnu/build/file-systems.scm:1235 (it's trying to create a bind-mount in my initramfs) and I'm wondering if the early init guile is linked against musl? it certainly does not seem to be aware if so, see guix/build/syscalls.scm:933 <user11037>selecting the gnu libc symbol statfs64 (it's a macro in musl for statfs) <user11037>then that error is very puzzling to me, and I'm not sure how I'd begin debugging it <wigust>i think we should document that, not an obvious thing <wigust>it will not be happened in the initram though <user11037>wigust: unfortunately, it needs to happen in initramfs. I am trying to set up a tmpfs root (as implemented in bug#65335) and I have a file-system with needed-for-boot? #t in the operating-system, hence the early bind mount. I followed "3.10 Setting up a bind mount" from the GNU Guix Cookbook, and get that error in the early init Guile <Tre_brock>When I try to update the guix package manager it seems like it updates, but when I search for packages they are all outdated, from around 2022. Does anyone know how to fix this? I tried running guix pull. <wigust>Tre_brock: try `hash guix` command <wigust>user11037: isn't better to write a service which removes everything except specified files and directories? What's the reason to use tmpfs except "fast" (which is not always true) and loosing RAM? <user11037>wigust: disk wear, in fact, the GNU Guix Cookbook example I mentioned is also pertaining to disk wear. It's also going to be wiped each time I power off regardless, as RAM does. Removing all the files (except the ones to keep) takes time to write to disk, potentially, while with tmpfs it's a natural part of powering off the system <wigust>user11037: i don't think disk wear could be a reason to do something like that, except really bad situtations like sdcards on raspberry where you have a small RAM in addition <user11037>wigust: I have 96GiB of RAM, which makes me more inclined to use it as the root tmpfs, really. If I were to create a "delete everything script" then it'd need to run in the initramfs also, but after the LUKS unlock happens from (mapped-devices ...). However this does not really solve the initial problem, which is certainly a bug in Guix that I was <wigust>user11037: i wonder how iso (guix installer) works, maybe it can help you to figur out the solution? <roelj>hello guix! I'm wondering if there's a specific Python group within Guix to upgrade Python and Python packages. <jakef>by group do you mean a group of people responsible for maintaining python in guix? <roelj>Ehm, I'd like to get involved in updating packages and building/testing them. Is there a specific mailing list to do so? <mgd>Hello. I suspect there is an issue with the package Mixxx which i use. Where is the best place to report this? <wigust>mgd: Hello. Email to bug-guix@gnu.org mailing-list. <roelj>peanuts: jakef: wigust: Thanks I've run the command to figure out who to contact. <efraim>polyedre: mount, umount and fusermount are in /run/setuid-programs/fusermount, so I don't remember it ever being a problem for me to use mount <Kabouik>Has anyone ever managed to use screensharing in Guix & Wayland with web videomeeting applications, like Jitsi/Zoom and co. ? I can only share a tab of the browser running the video call, not any other application. <wigust>Kabouik: i think it could be possible to use OBS with a virtual camera and wayland capture plugin, but not all apps allows that virtual camera feature to be used <wigust>avalenn: i don't see it, probably no <Kabouik>wigust thanks. I think I had tried that too, to no avail. Will try again. <Kabouik>i have obs-wlrobs installed, which I think is the plugin you mentioned for virtual camera. <daviwil>fnat: Awesome, thanks for fixing that (assuming you are the author)! <wigust>Kabouik: obs-wlrobs is for capture wayland. Virtual camera is built-in in OBS, but if you use Guix you probably need to load a kernel driver for virtual camera, e.g. with sudo modprobe v4l2loopback <wigust>Kabouik: also maybe need to add (kernel-loadable-modules (list v4l2loopback-linux-module)) to a sytem config first <daviwil>wigust: hey! have you tried bumping OBS to 30 yet? I was trying it a couple weeks ago and made some progress but still wasn't able to get it working <Kabouik>Thank you wigust, reconfiguring my system now with the kernel module you mentioned, and then I'll try with modprobe. Hope it works, and Jitsi supports virtual cameras. Hope Zoom does too, because for some reason everyone at works wants to use it over free and simpler alternatives. <fnat>daviwil: Not the author here, I was just keeping my eyes on Zheng's patch as an interested user. :) <Kabouik>So I now have the "Start virtual camera" button in OBS indeed wigust, but nothing happens when I press it, it does not even seem to be toggled on (I guess it would turn into "Stop virtucal camera" if it started anything). I tried all corresponding settings (internal, scene, source). <Kabouik>How weird, I had just rebooted and ran the modprobe command, and it would not work. Ran it again, and now it works. It's werid because I can see in my fish history that I did run it already after the reboot. Now I need to undrestand how to use it in a videocall context. <Kabouik>Hum, now even the built-in webcam refuses to work, mpv can no longer stream /dev/video0, /dev/video1 or /dev/video2. This used to work. <wigust>Kabouik: if you use webcam (/dev/video{0,1,2} in obs, then you cannot use it in mpv unless pipewire is used (which is probably not because of configuration requirement). In any case, you could use webcam in obs directly instead of mpv (webcam will be faster in obs from my expirience) <Kabouik>Got to go to lunch, I'll pick this up later, but just in a few words: I don't intend to use the webcam in OBS, it's just that when trying the Virtual Camera feature in a videoconferencing webapp, I noticed that my normal webcam no longer works in the webapp, and then tried it in mpv too, where it also no longer works. This used to work so I suspect it may be due to the new kernel module. Also, starting the virtual camera in OBS did not offer me to <Kabouik>select another source in the videoconferencing webapp, so I think something is still missing. <wigust>Kabouik: also videoconferencing webap can be the reason for taking controll on a webcam, so it should be launched latest <wigust>if you have a led on webcam it's possible to notice which app takes the control after launching it <taeaad>How do I copy my current profile into a new one? <taeaad>I have e.g. $HOME/.guix-profile, and want to create a profile at $HOME/.config/guix/profiles/new_profile. <wigust>taeaad: i think it's a bad practice, but you could symlink it <wigust>taeaad: better just create a new profile with guix built-in commands <avalenn>taeaad: guix package -p $HOME/.guix-profile --export-manifest >/tmp/manifest && guix package -p $HOME/.config/guix/profiles/new_profile -m /tmp/manifest <efraim>ERROR: ld.so: object '/gnu/store/zdkpnlgkdj4fbd85krgjw26v235r8v8h-libfaketime-0.9.10/lib/faketime/libfaketime.so.1' from LD_PRELOAD cannot be preloaded (wrong ELF class: ELFCLASS32): ignored. <Kabouik>Nope that's not the reason wigust, I launched the webapp after starting the virtual camera, and the native webcam no longer works either (it has no LED, but I can see nothing uses it), be it in mpv or in the webapp (individually, not simultaneously) <wigust>Kabouik: you could 'sudo rmmod v4l2loopback' or reboot, though i don't understand why this can happen. Also it's strange that the module didn't load for the first time without issues. Do you see it in the 'sudo lsmod'? Also, do you some errors in the dmesg during loading the module? <taeaad>I've been trying to understand "guix import pypi <x>" better, what is the intended modification to make it a standalone manifest file, e.g. python-<x>.scm? <taeaad>In this example it would only contain the prerequisites for python-<x> and install exactly and only a minimal environment for python-<x>. (If I understand correctly.) <taeaad>singpolyma: Thanks, will add it. Will trying working with "guix import pypi <x> -r | guix package --export-manifest" for now. <Kabouik>As soon as I sudo rmmod v4l2loopback, the native webcam works again wigust <Kabouik>I see v4l2loopback in `sudo lsmod` after modprobe, yes. <Kabouik>Having made rmmod then modprobe, the native webcam still works, by the way. Maybe what broke it was trying to use virtual webcam in OBS. <Kabouik>No. Turns out what breaks the native webcam is having OBS running, Virtual camera being toggled on or not. Closing OBS and the native webcam works again. Definitely something wrong on OBS side, especially as even when Virtual camera is on, it is still not available in the videoconferencing webapp anyway (after restarting it). <bost`>Hi. `sudo echo "foo"` running in a guix shell returns 'sudo: /gnu/store/<checksum>-profile/bin/sudo must be owned by uid 0 and have the setuid bit set'. I dunno what do I need to do about it. <Kabouik>I'm sorry this may sound like an issue independent from Guix, but it might be Guix related too since there doesn't seem to be such issues in other distros. <efraim>in Guix System sudo is /run/setuid-programs/sudo <efraim>since we don't store executables as setuid in the store <bost`>efraim: ??? Fo I need to start the shell with `guix shell ... --share=/run/setuid-programs ...`? <efraim>I guess you could use --expose=/run/setuid-programs <singpolyma>Are you using --pure I guess? Otherwise I'd expect it to work... <efraim>I would start by just not adding sudo into the list of packages <bost`>singpolyma: efraim I'm trying to hack on the guix-daemon, so I need to run `sudo -E ./pre-inst-env guix-daemon --build-users-group=guixbuild` <bost`>efraim: yeah. It doesn't pull it in. But I need it and I pull it in. Either way having the --expose=/run/setuid-programs doesn't help. (The full command is `guix shell --container --network --pure --nesting --expose=/run/setuid-programs --development guix sudo`) <efraim>do you not have /run/setuid-programs/sudo? <singpolyma>Definitely don't include sudo in the packages for your shell <bost`>`/run/setuid-programs/sudo echo "foo"` returns 'sudo: /run/setuid-programs/sudo must be owned by uid 0 and have the setuid bit set' <bost`>singpolyma: efraim OK I excluded sudo from the shell. <bost`>guix shell: error: '--nesting' cannot be used without '--container' <bost`>HA! this works `guix shell --network --pure --expose=/run/setuid-programs --development guix -- /run/setuid-programs/sudo echo "foo"` <singpolyma>Probably don't need --network without --container but it won't hurt anything <bost`>singpolyma: yeah sure the --network is superfluous. thx. <bost`>efraim: singpolyma thank you both. <cbaines>does guix build icecat-l10n work for anyone? <cbaines>for me, fetching substitute information seems to hang (or at least take far too long) <lechner->Hi, I'm a Rust noobie. How do I deal with this error on 'cargo build', please? /lcl/lechner/pam/libpam-guile-rust/git/target/debug/build/clang-sys-600b28517f365aa7/build-script-build: /home/lechner/.guix-profile/lib/libc.so.6: version `GLIBC_2.34' not found (required by /gnu/store/6ncav55lbk5kqvwwflrzcr41hp5jbq0c-gcc-11.3.0-lib/lib/gcc/x86_64-unknown-linux-gnu/11.3.0/../../../libgcc_s.so.1) <wigust>Kabouik: do you have sources which use webcam in obs? also, you could temporary reset obs with kill it and moving configuration with mv ~/.config/obs ~/.config/obs.1 and start from fresh <cancername>Hi! I'm trying to use my mediatek mt7612u wireless adapter on Guix, it worked well on Debian. the `mt76' module is modprobed. in dmesg, I see `new SuperSpeed USB device number 2 using xhci_hcd'. the device does not show up in iw. I would greatly appreciate some advice for getting it up and running. <cancername>after connecting the adapter to a different computer, it looks like it's actually using the mt76x2u driver, which doesn't seem to be in linux-libre :/ looks like I'll have to use mainline <jonsger>cbaines: "guix build icecat-l10n" works for me, it just fetched the substitutes from bordeaux <Kabouik>wigust you're right, I had a webcam source in OBS. It was disabled, but apparently that was enough to "reserve" it for OBS and break it for other applications. Thanks a lot. I haven't managed to use the virtual camera yet though, it still doesn't seem to provide any new source in the video conferencing applications I tried. <Chubbymomo>Hello again! I can report back that after the patch regarding oci-container-service-type was merged, it fixed my issue! <Chubbymomo>With that being said, I now run into another issue, breaking the oci-container-service-type declarations into other files because my config.scm will turn into spaghetti if I put everything in there... <Chubbymomo>I managed to get the module to load in the use-modules declaration after adding a directory to the Guix package path, however I'm for some reason I cannot reference the defined "pihole-service" as it's an "unbound variable" <ieure>No idea. Man Guile's errors are just the absolute worst. <ieure>Me: please can I know what file and line the error was on? <Chubbymomo>I feel like an idiot, which I might be, when trying to debug Guile. <cbaines>Chubbymomo, sometimes Guix can get in the way of Guile errors, so I'd try evaluating your configuration with Guile <Chubbymomo>Okay, how would you go about that, I just tried the naive approach of running "guile config.scm" and it couldn't find the gnu module, so that can't be correct. <dariqq>Chubbymomo: if i evaluate you pihole-service.scm i get In procedure raise-exception: <dariqq>error: service: unbound variable <cbaines>Chubbymomo, maybe try guix shell guile guix -- guile config.scm <cbaines>you might also be able to do guix repl config.scm <Chubbymomo>'=D dariqq was on the right path... After adding the use-module for (gnu services), it works. <Chubbymomo>Thank you dariqq cbaines and ieure! Would any of you have any tips regarding two things: <Chubbymomo>2) Best approach to keep custom modules that are frequently changed (I'm using the environmental variable approach right now, seems kind of weird). <dariqq>i've been using the -L option in guix to add directories to the loadpath as needed <old>is there a trivial way that I can take my shepherd-service in home and put it in system services? <taeltydes>what I really want is just to boot directly into a tty. <taeltydes> as an aside, would it be worth adding a guix-help or guix-issues channel or something? Feel a bit cheeky asking for advice here but not sure where else to ask. <ieure>taeltydes, There's a help-guix mailing list. <taeltydes>ieure: thanks. I'll check that out. I've switched to mingetty instead but it's still booting with gdm. strange. <ieure>taeltydes, You probably need to delete the gdm-service-type from %base-services in your operating-system. <wakyct>hi folks, has anyone had problems with bluez.obex and found a fix? I can't send files from host to device because "name org.bluez.obex was not provided by any .service files", though I have a foggy memory of this working before <wakyct>I haven't changed my BT setup since I added it to config <wakyct>it's weird because looking at info and plugins via Blueman seems to show that obex is there <wakyct>a few years ago someone on IRC said "I added bluez to my profile and the org.bluez.obex.service became known." not sure if that's different than adding bluez via the system config? <Kolev>I dread running `guix pull` because it takes forever. <cancername>taeltydes: do you happen to be using `desktop-services'? if you are, you may have a better time using just `base-services' <cancername>Kolev: Same! I suppose the solution is to make a cron job for it :P <Kolev>cancername, that could work. <taeltydes>cancername: this might be an idea. I tried that earlier though & find I had no networking, though I guess it might be easier to add what I need instead of removing what I don't <cancername>taeltydes: there are several options. I've been happy with `dhcp-client-service-type', but you may want to choose `network-manager-service-type' or something else <cancername>speaking of networking, I've asked this before, but I've been wanting to run a DNS->DoH proxy locally, and have been using dnss for that, but it doesn't seem to be available in Guix. what's the recommended Guix package for this? <cancername>I just ran `echo mem >/sys/power/state' and all hell broke loose, lots of null derefs and "pcieport ... multiple correctable error" errors, even tearing the screen <wigust>Kabouik: if it works on a some site for testing webcam like https://webcamtests.com/check (make sure to select video input on the website, otherwise your real webcam could be selected) then everything is configured correctly and your app just does not support it <fnat>Anyone using ZNC (the IRC bouncer) under Guix? There doesn't seem to be a ZNC service or am I missing something? <fnat>There's a ZNC home service but I was expecting something I can run on a server, so I'm slightly puzzled as I'd be expecting that to be a system and not a home service. <cbaines>I use it, but I just ssh in to the server and run znc, then it forks and runs in the background <wigust>probably can be rewritten as a system service and converted to a home service like other home services, so we could have both <fnat>cbaines wigust: Ok, great, thanks, I'll look into it then. <Kolev>Does the current version of yt-dlp work? <blewn0se>kolev if you mean from the guix repo, I don't know. I am lurking here to learn more about guix. Though, I installed yt-dl this morning from git and it is working on my debian system <Kolev>blewn0se: I'm using Guix version. <ieure>Kolev, Whether it works depends a lot on the site, yt-dlp supports hundreds. I fairly regularly use the version in Guix with YouTube and it has been working for me lately. <ieure>But the version in Guix is very out of date and breakage is inevitable because Google is godawful. <Kolev>ieure: I was trying to DL a Facebook vid. <blewn0se>Tools like this are adapting all the time, and its hard to expect someone to keep up with it on a repo. Suggest just use the pip version <ieure>blewn0se, I mean, Guix is a rolling distro which is supposed to have frequent updates so you don't have to do that. <ieure>Kolev, No idea whatsoever. Facebook gatekeeps everything so aggressively that I'm surprised anything other than their official stuff works, ever. They're relegated to the Not Worth Even Clicking bucket for me. <ngz>IIRC, the difficulty with yt-dl update is that this requires bumping Python libraries with a lot of dependents, so those are currently in "python-team" branch. <Kolev>ieure, Mom sends Facebook videos to the family XMPP chat and I usu. download and reup the vid to the chat. <blewn0se>ieure understood. I just mean it's not easy to have every single tool updated all the time. unless its automated of course, which I guess it could be <ieure>blewn0se, Guix has some tooling to ease updates in the common case, so that is often pretty easy. But then they sit in patch review for who knows how long. <ieure>There's at least one patch to update yt-dlp submitted already, but it's sitting in patch review limbo. <blewn0se>One of my concerns with making the jump to guix, is I suspect some drivers and firmware for my laptop are going to be difficult to deal with <ieure>blewn0se, You are correct. Difficulty varies depending on what hardware you're talking about. <blewn0se>Yes. I suspect so. Although that does not deter me from following along, and installing some day soon when I have some time to play around. <nutcase>Hello guix. I want to use sftp via emacs/tramp. I get the following message "tramp-error: Package "tramp-gvfs' not supported". I read that I need gvfsd-sftp. I have "$(guix build gvfs)/libexec/gvfsd-sftp` on my system, but not on $PATH. What can I do to get SFTP working from within emacs/tramp? <wakyct>do you need to add that to emacs load path some way? <nutcase>I thought the load-path is for lisp only, isn't it? <wakyct>you're right, I'm thinking of some other search path, maybe just $path I guess <Chubbymomo>How do you guys debug shepherd services? I know there are logs in /var/logs, but I'm not finding anything that's particularly helpful. I'm trying to utilize the oci-container-service-type (Service: http://paste.debian.net/1319439/) to run SearXNG, which requires some extra arguments not natively in the service-type, so I use the aptly named <Chubbymomo>"extra-arguments" to do some extra stuff, the only thing that I'm seeing inside of the shepherd logs is exit 125, which means that command didn't execute correctly. I've done a rough translation of the service into a docker run command and it launches perfectly. <Chubbymomo>Sorry for the bombardment of needing help, thankful if any is provided. <nutcase>wakyct: ok, even with prior "export PATH="$(guix build gvfs)/libexec:$PATH"" I get the error, although vterm inside emacs finds gvfsd-sftp now <taeaad>I think my pipe was wrong, guix import pypi <x> | guix package --export-manifest, this just ignores the pipe and exports current profile's manifest. How do I package the output from guix import? <ieure>taeaad, I'm not sure you can have a manifest list packages which aren't in a channel Guix knows about, but... not sure. It's an unusual usecase. You could *maybe* take the output of `guix import' and put it in a (list ) form in a manifest; but I honestly don't know if that works. <taeaad>How should I use guix import if not like this? <taeaad>Atm it seems like I need to add the modules I need at the top manually, and then add the function definition below that. <ieure>taeaad, Typical usage of guix import is to paste the output into a .scm file in a channel, or specify `-i path/to/package/file.scm' to have it do that for you. <taeaad>ieure: The second part secify `-i file.scm', as argument to what? <ieure>taeaad, To `guix import', the thing we were discussing.