IRC channel logs

<ieure>I'm trying to write a home service to manage mu (the email index/search tool), and it's not working for reasons I don't understand.

<ieure>Here's the service: https://codeberg.org/ieure/atomized-guix/src/commit/35e51a45f1342fe3ef6aa19d68b6be43b6067610/atomized/home/services/email.scm

<ieure>And here's how I'm configuring it in my home config: https://paste.debian.net/1362979/

<ieure>When I reconfigure, I get: guix home: error: no target of type 'home-mu' for service 'home-email'

<ieure>Wait, wrong error.

<ieure>This: "guix home: error: service 'mu-init' provided more than once"

<ieure>Why am I getting this when I'm definitely not providing it more than once?

<ieure>There are two parts to this, one that runs `mu init', which you have to do prior to indexing any email; and a second that uses watchexec on the maildir, triggering `mu index' on changes. Both are shepherd service extensions of the same home service.

<ieure>Maybe that's wrong? Not sure. I would like to get both without configuring them independently.

<mange>I don't think you can extend the same service twice. You should just provide both shepherd-services in the one "list" returned by the function you're providing.

<mange>That is, the equivalent of (service-extension home-shepherd-service-type (lambda (config) (append (home-mu-init-shepherd-service config) (home-mu-index-shepherd-service config)))).

<ieure>Yeah.

<ieure>That does get me past that error. Okay, thank you.

<ieure>ugh

<ieure>Okay, I have a thing working, though I don't really understand why it has to be like this.

<mange>What do you mean by that?

<thema>Hi, I am trying to create an image with btrfs instead of ext4 for a server.

<thema>If I use the grub-efi-bootloader with a vfat and btrfs partition everything works fine (I am testing using qemu locally).

<thema>However, if I use grub-bootloader with only the btrfs partition, it fails with unknown filesystem. Any idea what step I could be missing?

<ieure>mange, This is where I ended up: https://paste.debian.net/1362981/ (I removed the (list ...) from `home-mu-*-shepherd-service')

<ieure>mange, 1. Why do some `...-shepherd-service' procedures return lists instead of shepherd-services? ex. https://codeberg.org/guix/guix-mirror/src/branch/master/gnu/home/services/desktop.scm#L347

<mange>Oh, I would actually write a single (define (home-mu-shepherd-services config) (list ...)) which defines both services in one go.

<ieure>mange, Much of the prior art in Guix has `foo-shepherd-service' procedures which return a list with one service in it.

<mange>Sure, but you can return more than one service.

<mange>It's just that most things only need one service.

<ieure>Right, but then like... why do so many things return 1-lists then?

<mange>Because the contract is "the procedure must return a list". If you only want one service, you have to return a list with one service in it.

<ieure>mange, Right, but why do they not have the `foo-shepherd-service' procedure return a service and compose it into a list in `foo-service-type'?

<mange>As in, use (compose list foo-shepherd-service) rather than (define (foo-shepherd-service config) (list ...))?

<ieure>Well, I don't know if `compose' is equivalent, but, sure. It feels weird for a thing calling itself `foo-shepherd-service' not to return a, you know, shepherd service.

<ieure>If it was named `foo-shepherd-services', then yes, I can see returning a list.

<mange>Okay, sure. It's immaterial in your case, though, right? Because you need two services anyway.

<ieure>Yes, it just makes it difficult to understand how the system works, and the manual on this is not very helpful

<ieure>It is also factually incorrect, it says that the guix-service-type can't be extended, when it most definitely can.

<mange>Huh. I didn't know that. That would be helpful to avoid me needing to use modify-services in my config!

<ieure>Yes, very! Strangely, you can't extend it to add channels; only substitutes, keys, and chroot directories.

<mange>Presumably a patch to allow adding channels could be merged, too. I don't use channels in my system config (what's the point?), so I don't need that. :)

<ieure>I semi-recently switched to managing channels *only* in my system config, and now think this is The Way.

<ieure>My system config needs stuff from my personal channel, which makes bootstrapping difficult if the channel isn't there; and if it's already at the system level, why duplicate into my user config?

<mange>I only use the system config for the daemon. Everything I care about is done with my user Guix (including system reconfigure).

<mange>Hence, keys and substitute servers matter, channels don't.

<ieure>Sure, that probably works fine for one machine. I have multiple, so want to share the config between them, so that has to go in a channel. And having the channel the system needs in the system config makes perfect sense to me.

<ieure>Otherwise I have to set up the system, set up my user, pull, edit the system config, reconfigure.

<ieure>I agree it probably doesn't make much sense if you don't have a setup like i do.

<ieure>Ugh, this mu service isn't going to work, you can't `mu index' when mu4e is up, because the mu server locks the index. And you can't remote control it except from the Emacs process that spawned the server.

<ieure>whelp

<ieure>Guess I have to implement that bit in elisp. The inotify interface is annoying :(

<apoorv569>anyone has any idea why this config https://0x0.st/8QHm.txt the postgres service works if I do `guix syste vm CONFIG`, but if I instead use `guix deploy` to a VM I created my self in virt-manager the postgres service doesn't work https://0x0.st/8QHa.txt

<apoorv569>not very reproducible I'd say

<ieure>apoorv569, Well, what are the permissions on the PID file it's complaining about not creating vs. the machine where it works?

<apoorv569>ieure: let me check.. using guix deploy this file doesn't exist BTW but let me check in the guix system vm..

<ieure>apoorv569, Right, yes, `guix system vm' starts from a clean slate. Deploying means there's state to deal with.

<apoorv569>in the guix system vm, `/var/lib/postgresql/data` all files in this dir are owned by `postgres:postgres` in the `vm` I created my self, all files are owned by 957:postgres

<apoorv569>I don't see any user in /etc/passwd with id 957

<apoorv569>not sure who/where that id came from

<apoorv569>in the guix system vm postgres UID/GID is 967:981 in the vm I created its 943:977

<mange>ieure: I use Emacs as a daemon and run "mu index || emacsclient -e '(mu4e-update-index)'" to do the index update.

<apoorv569>mange: you can set mu4e-update-interval along with mu4e-get-mail-command to auto update your mu index..

<mange>Yup, also a way to do it.

<mange>ieure: On the "multiple machines" bit, I use "guix deploy" which also runs as my user. When I set up a new machine I make sure I can SSH in, then I use my user Guix on a different machine to bootstrap it. Like you, I now think this is The Way. :)

<PotentialUser-69>guix install librewolf gives unknown package - normal?

<janneke>PotentialUser-69: have you run guix pull?

<janneke>librewolf should be available

<PotentialUser-69>yes i did 30 min ago - I will retry in a few minutes [but I did nor reboot]

<Hwarang99>Hello, it's the first time that I enter here, I need help with a specific problem that I have with Gnu Guix. Please explain me how this place works so that I can contribute with doubts and solutions

<PotentialUser-69>yes it is loading librewolf v136 now!

<jA_cOp>Hwarang99: welcome! It's a pretty typical IRC channel, with a mix of help questions and discussions about Guix. If you want some example conversation, you can read the public log: https://logs.guix.gnu.org/guix

<Hwarang99>Thanks for the information

<janneke>PotentialUser-69: good!

<Hwarang99>I have an intel i5-10210U cpu, currently the hardware acceleration by vaapi is not working. I have the intel-vaapi-driver package installed and “mesa”. I suspect it might be because my hardware requires the intel media driver software to work, however I have not been able to find any package with that name or similar. Please if anyone knows

<Hwarang99>about intel igpu hardware support in Gnu Guix, I would appreciate any possible solutions

<jA_cOp>Hwarang99: I only have Guix on an AMD system, but the Arch Wiki has some instructions for troubleshooting hardware decoding with mpv, maybe that could be useful: https://wiki.archlinux.org/title/Mpv#Hardware_video_acceleration

<jlicht>Hwarang99: You can search for packages across multiple channels at https://toys.whereis.social/

<Hwarang99>Wow

<Hwarang99>Interesting website

<Hwarang99>How can I add this paste.debian.net/1363011?

<Hwarang99>How are external modules added? In this case is module from nonguix of intel-media-driver

<Hwarang99>Could someone explain me why you don't add this package (the free version) in the official gnu guix repository? I don't understand it because it is an important driver to add support to any intel igpu from 2015 or newer, it is totally opensource(as long as the compile flags are chosen correctly). If anyone knows an answer, please I would appreciate

<Hwarang99>it.

<jlicht>Hwarang99: I may have been the one to package it originally, (many years and some packages ago), but ISTR that this package is not rebuilding all kernels from source

<jlicht>where kernel is not a linux kernel, but something relating to GPU stuff.

<jlicht>So as-is, that package would be free software, but since everything in it is not built from source, it's not eligible for inclusion in guix

<jlicht>Hwarang99: wrt adding extra channels: https://guix.gnu.org/manual/devel/en/html_node/Specifying-Additional-Channels.html

<Hwarang99>Thanks for the explanation, sorry for mentioning this external software in the official gnu guix chat. I appreciate your help

<Hwarang99>And another question I have, there are some software that when they are installed they need to add some groups correctly for the users to have permissions. In specific case I have installed the wireshark package, normally to be able to capture packets it is necessary that the wireshark group is in the user to have permissions to do it. I tried to

<Hwarang99>add it in the system configuration, but I get an error because it does not tell me that the group does not exist. How to solve this kind of problem in Gnu Guix? Should I create a new group (will it work)?

<Hwarang99>This is my config.scm paste.debian.net/1363013

<yelninei>janneke: tried building the hurd manifest on core-packages-team: an undefined pthead symbol error during openssl tests, some automake test failures, more packages have the same failing gnulib tests and --disable-year2038 configure flag. "which" now uses PATH_MAX. Bison is segfaulting so I couldnt get to curl, cmake and beyond

<Hwarang99>If someone knows how to manage the groups so that they do not give errors and have the correct permissions, I would appreciate it

<janneke>yelninei: oh my...

<jA_cOp>Hwarang99: there is a wireshark service for Guix Home which uses setcap instead of groups: https://toys.whereis.social/services?search=wireshark

<jA_cOp>this means all users on the system would be able to capture, though

<jlicht>Hwarang99: no biggie, it's not Voldemort or anything; it's just (by-definition) off-topic for the channel ;-)

<yelninei>The rest will need to wait until i can run a childhurd on core-packages-team

<janneke>yes

<Hwarang99>The hardware acceleration support, and that several packages are not working by default in Gnu Guix seems to me something relevant is not a minor problem. People are not born learned

<m4xxed>Hola, since I am using this IRC's logs frequenly as a source of knowledge / troubleshooting hints, I will just leave this here for anyone who (like me) recently reconfigured their system after a pull and found the new gnome-keyring not providing an ssh-agent anymore: https://paste.debian.net/1363019 . This is a home-service for gcr-ssh-agent which

<m4xxed>will hopefully be a good starting point for anyone who finds themselves in the same situation as me (being new to guix and somewhat struggling with guile scheme sometimes). This service is adapted from a snippet provided to me by lilyp (thanks again) :) have a nice day everyone

<mrvdb>Hi guix, is anyone working on making rust build on ppc64? More an more package depend on rust it seems, so things are slowly disappearing from my manifest, as they do not build (anymore).

<janneke>yelninei: i'm reverting (proposing to revert) which to 2.21 on core-packages-team

<cbaines>janneke, hey, is something up with the core-packages-team request for merging issue?

<cbaines>I'm pretty sure it was this one #74676 but the issue title now doesn't match the format

<peanuts>"[core-packages-team 5/5] DRAFT gnu: gcc: Update gcc, gcc-toolchain to 14." https://issues.guix.gnu.org/74676

<janneke>cbaines: https://issues.guix.gnu.org/75518#36

<janneke>it was decided to let gnome go first

<cbaines>ah, I see

<yelninei>janneke: great, in case this is accidental you only pushed the 2.23->2.22 downgrade

<janneke>yelninei: thanks for the headsup, i missed the 2.22 upgrade

<PotentialUser-98>hi

<kinote>Hi, I am trying to run the "git" command on a server via ssh where I have installed it using guix-home. The guix-home profile does not seem to be loaded from /etc/profile. This was supposed to be fixed in https://issues.guix.gnu.org/63009 but I do not see the fix being applied. The issue references a git commit hash but I cannot see it in the guix git repo. Am I missing something?

<janneke>kinote: it seems that something went wrong, thanks for the heads-up

<kinote>janneke: btw, i tested the patch and it also worked for me

<janneke>kinote: thanks! i've really pushed it to master this time... https://git.savannah.gnu.org/cgit/guix.git/commit/?id=2b6017dd0b031e43de7e66b3e1f3f331c0ab712c

<kinote>janneke: thanks!

<janneke>no problem, thanks for noticing and notifying!

<h4>Why no `eglgears_wayland`? The Debian hack? https://git.savannah.gnu.org/cgit/guix.git/tree/gnu/packages/gl.scm#n708

<tae>Is git.savannah.gnu down again? I'm getting 502 on guix pull.

<Guest1512>i'll give it a try. maybe it's related to my problem.

<Guest1512>i'm also getting a 502

<moksh>hello

<moksh>I just installed guix

<futurile>moksh: congratulations!

<Rutherther>anticomputer: thanks for the protonmail-bridge script, turns out I can't use protonmail bridge since I am not paid user. Interesting that I can use heisenbridge :) I actually have their password manager paid, but not the e-mail. I thought that was why heisenbridge is working for me, but seems not. Seems that anyone can use heisenbridge, not just paid users

<tae>Guest5382: Thanks, was down earlier the year too.

<moksh>but I am having a hell of time understanding stuff

<moksh>first thing I would like how to change my default shell

<moksh>for a user

<futurile>moksh: do you have much Linux experience generally? any experience with Nix previously?

<futurile>moksh: everything in your system is currently defined in a file /etc/config.scm - that's where you change things like the shell

<ieure>moksh, User shell configuration goes in your operating-system configuration. See the `shell' field part of "(guix) User Accounts."

<ieure>(in the manual)

<futurile>moksh: you can do some searches, or go through the manual: https://guix.gnu.org/manual/devel/en/

<moksh>yep I was using linux as my main os for arount 2.5 years

<moksh>I tried nix one time, didn't get into it

<moksh>have tried guix a few times

<Guest5382>tae i ran guix pull a couple of days ago and it worked fine. so it don't think it's related to my endless SWH update problem.

<moksh>used to have it on a foreing disto

<moksh>but didn't use guix at that time much either

<moksh>when I got my main refurbished thinkpad t480 some days ago

<moksh>I installed guix

<ieure>Nice, great machine.

<moksh>I am putting my stuff in a git repo

<moksh>so that I can ask for help easily

<moksh>btw one major problem I have currently is XDG_RUNTIME_DIR problem

<moksh>I have elogind as well

<moksh>lemme make a repo first

<ieure>tae, Guest5382, I also get a 502 when pulling.

<moksh> https://codeberg.org/moksh/verito

<moksh>here is the repo

<Guest5382>ieure is it possible the 502 is related to my endless "SWH vault: requested bundle cooking, waiting for completion..." when running guix system reconfigure? I don't think it is, but I might as well check.

<ieure>Guest5382, No.

<ieure>Guest5382, Do you have any packages you've created in your system config?

<ieure>Guest5382, Can you paste the complete output of `guix system reconfigure' on a paste service like https://paste.debian.net ? At least, complete up until you kill it.

<moksh>ieure: so If I installed fish with guix install fish, do i just need to add (shell (file-append fish "/bin/fish"))

<ieure>moksh, You don't need to install it, but I think that is the correct configuration.

<ieure>moksh, `guix install' installs stuff in your user profile, that's not visible to the operating-system.

<moksh>or do I need oh

<moksh>oh I see

<moksh>I will try with guix system container first

<moksh>do I need to do cow-store on it

<moksh>?

<Guest5382>ieure no extra packages added to system config after the initial install. Only includes %base-packages and 4 window managers provided by guix

<moksh>ok I edited the config.scm

<Rutherther>moksh: do you really want to make your login shell fish? it's not posix compliant, so it doesn't work with guix-home

<Guest5382>ieure i'm sure i can provide the pull full output since it contains references to a non-free repo. let me know if i should cut that part out. i think it might be related to my problem though

<Guest5382>i'm not sure**

<moksh>Rutherther: I want to use fish as a user shell as I have been using it for years

<ieure>Guest5382, It's fine.

<hako>home-fish-service-type uses fish-foreign-env

<hako>to source ~/.profile

<moksh>ok i edited the config, should I test with `sudo guix system container <file> ` first or does it work without sudo

<Rutherther>moksh: that doesn't men you need to use it as a login shell. Most terminals just respect SHELL env var. But as hako is saying that home-fish-service-type takes care of it, so it's probably fine then.

<moksh>I don't about services yet, and while configuring guix home

<moksh>I am met with an error

<Rutherther>moksh: cow store is used when your system is booted in ram so that your ram doesn't fill up, so the store gets backed on disk instead of in ram. So you use it only if that is the case. You don't need it in installed system, you typically use that only in live installation iso

<moksh>ok, thank you

<Guest5382>ieure https://paste.debian.net/1363090

<moksh>you can see my home config > https://codeberg.org/moksh/verito

<ieure>Guest5382, Huh.

<moksh>and this is the error https://0x0.st/8QKm.txt

<ieure>Guest5382, Can you also paste your operating-system config?

<moksh>even though I have elogind

<Rutherther>moksh: did you reboot after installing elogind? and since you're just a beginner, I am going to ask to make sure: after adding elogind, did you reconfigure your system?

<Rutherther>(and by installing elogind in this case I mean adding the elogind service type in your services / using %desktop-services that contains it)

<moksh>I didn't add elogind by hand

<moksh>neither did I add the elogind to services

<moksh>when I run `sudo herd status`, it should elogid there

<moksh>and XDG_RUNTIME_DIR is /run/user/1000

<moksh>I don't remember installing elogind manually either

<Rutherther>moksh: it is in %desktop-services that you have in your config

<moksh>It was configured by the graphical installer

<Rutherther>moksh: so what is your XDG_RUNTIME_DIR set to?

<Rutherther>oh sorry you already sent that

<Rutherther>moksh: and when you check now, the directory still doesn't exist?

<moksh>it is there

<moksh>when I echo $XDG_RUNTIME_DIR it shows that dir as well

<Rutherther>moksh: okay and does the error happen on every login you make or was it just one time?

<moksh>lemme just retry

<Guest5382>ieure https://paste.debian.net/1363092

<Guest5382>very little modification from the result of the basic graphical install

<moksh>I just did `guix home container <File>` and the error is still there

<moksh>but the changes are applied in the container

<moksh>but when I do `guix home reconfigure <file>` changes are there but packages listed in the file do get installed in home but I don't get the bash prompt

<Rutherther>moksh: oh... in container. Then it's kind of expected. You need to ensure that directory is coming from somewhere in container. You don't have elogind in container.

<Rutherther>moksh: but as long as you don't need shepherd services in the container, you don't actually need this directory

<ieure>Guest5382, Yeah, humm. I don't know. It's failing to download something, and falling back to SWH, which is also failing.

<Rutherther>moksh: re bash prompt - did you open a new bash?

<moksh>Rutherther: yes

<moksh>but the prompt is not applied

<moksh>packages are there but not prompt

<Rutherther>moksh: oh I only now see you put it to environment-variables... PS1 shouldn't really be an env variable, it should be set from bashrc as regular bash variable. But it doesn't matter. The only thing is, for env vars to get applied you need to get the login shell applied, so one way is to relogin.

<moksh>h

<moksh>oh

<Guest5382>ieure i guess i will wait until i can run another guix pull to see if that fixes the issue? the main reason i'm doing system reconfigure is because my most recent guix pull/upgrade broke logging into awesomewm, and i thought doing this may fix it.

<ieure>Guest5382, Yeah, I'd try waiting. You can also roll back to the last working generation.

<janneke>alas some life here, i was beginning to wonder if pi-day was a bank holiday

<janneke>maybe i should say american-pi-day, as its only them who write 3/14

<oriansj>is there any reason why fbterm isn't packaged yet? https://salsa.debian.org/debian/fbterm

<ieure>oriansj, Because nobody's done it.

<janneke>sneek: non-american pi-day is july 22th, i.e., 22/7

<sneek>Got it.

<janneke>sneek: botsnack

<sneek>:)

<snihil>janneke: ha, never realised that before

<janneke>ieure: or everyone who did it didn't care/dare to share it?

<ieure>Sure, maybe.

<moksh>Rutherther: so I logged back in

<moksh>no prompt is not there

<Rutherther>sneek: later tell moksh: I think that is because it is being overriden by bashrc. 1. make sure to set it in bashrc, not in bash_profile (that's where this environment-variables goes) as bashrc is loaded after bash_profile, 2. make sure to disable "guix-defaults?" in home bash configuration as that is what is giving you different prompt.

<untrusem>I am here

<moksh>I have a bouncer setup via xmpp

<moksh>any examples config for beginner to learn from

<moksh>also is there a blog to setup flatpak in guix

<Rutherther>moksh: no need, just install flatpak (guix install or put it to home packages, up to you) and use it with --user flag.

<moksh>Rutherther: thanks I removed the prompt successfully

<moksh>Rutherther: I installed flatpak

<moksh>and installed gajim but can run it with either flags

<moksh>what I did

<moksh>installed flatpak with home

<moksh>added flathub ref as --user

<jA_cOp>only additional thing I did with Flatpak was add $HOME/.local/share/flatpak/exports/share to $XDG_DATA_DIRS, but that's only so app launchers will find the desktop files

<moksh>ahh I see

<moksh>yes it is installed now :)

<erik``>When declaring the source in a package definition, is it possible to specify the subdirectory to use? For example, I want to package the vosk-api go bindings, and this package is found under "go" in the vosk-api repo.

<Rutherther>erik``: no, not really, because you instead make a phase that will change dir to the subdirectory inside the build process

<ieure>That.

<erik``>Mm, but that doesn't work nicely with the go unpack and build phase. The source is unpacked in a specific structure and build assumes that you are in the root directory when it is evaluated. What I do now is to add a phase that moves the subdir one level higher and removes the original repo, but seems a bit hackish.

<hako>Same thing should be possible to do in a origin snippet

<erik``>Tried that as well, but found no nice way to replace all files in the repo with the files in one of its subdirs.

<hako>delete all files except that dir, move files within the dir up, and delete the dir?

<moksh>how do I setup keyring in guix

<erik``>Conceptually simple, yes, but in practice relatively cludgy vs. "move subdir one level up, delete old dir, rename subdir" in guile.

<moksh>and enable dark mode, I am using xfce

<hako>(copy-recursively "dir/." ".")

<erik``>The cludgy part is "remove all but 'go'"

<Rutherther>erik``: so I suppose that /go$ is not in the import path of that package, correct?

<erik``>It is, but if I specify that to #:import-path, it creates an extra level (.../go/vosk-api/go)

<hako>(for-each delete-file-recursively (remove (cut member <> '("." ".." "go")) (scandir ".")))

<Rutherther>erik``: then the go build system already got you covered. Just set unpack-path without that last go

<Rutherther>erik``: so "#:import-path ".../go" #:unpack-path "...""

<erik``>Aaah, that is just what I need. Thank you both for the help!

<Rutherther>erik``: and see this for next time: https://xyproblem.info/

<untrusem> https://codeberg.org/moksh/verito/ > I added some packages

<erik``>I know XY, but it is often difficult to know where in the chain to pick X...

<untrusem>How would I go on about this now?

<untrusem>like how to build up on this

<untrusem>For now I am adding a todo

<untrusem>what did you folks did when you were a noob in guix

<untrusem>?

<janneke>wtf, someone removed guile-emacs that i resurrected in december

<janneke>one of the most exiting projects around...

<janneke>"why wasn't i consulted" (TM)

<janneke>i know it's silly, but you could check who worked on this and cc them no?

<janneke>ACTION has to go afk for a bit

<Rutherther>janneke: seems like a mistake to me, the commit message states it's because it is unmaintained and the website hosting it no longer active, both are false as far as I can tell

<Rutherther>codeberg.org still works, and guile-emacs.org as well (but that is not the hosting website), though they seem to have expired certificate

<janneke>istm that some people just want to remove guile-emacs, this isn't the first time

<Rutherther>janneke: btw since you are saying you worked on it, why wasn't the repo changed to guile-emacs/guile-emacs where the development seems to take place now?

<old>someone know in which package is `javac' available?

<ieure>old, The jdk output of openjdk.

<old>ahh thanks

<old>what's in the other output?

<ieure>The JRE. There's also a doc output with documentation.

<second_timer>is Git Savannah supposed to be down? im getting a error 502 on a git pull

<ieure>second_timer, Supposed to be? No. Is it? Yes.

<janneke>Rutherther: yes, of course

<janneke>i'll look into it after dinner or else tomorrow

<second_timer>leure: i should postpone my guix baptism them

<second_timer>ieure

<janneke>dude, and guile-emacs "just builds", yay

<janneke>well, i'm gonna look into updating it later

<janneke>i guess i shuld add a home-page

<janneke>or something

<Rutherther>janneke: has the home page changed?

<luca>Hi, anyone got a link to a guix mirror? Main repo seems to be down at the moment

<vagrantc>luca: https://codeberg.org/guix/guix-mirror.git

<luca>Thanks!

<vagrantc>interestingly enough, i was just able to push to git.savannah.gnu.org but also have trouble pulling

<vagrantc>just pushed an example bare-bones configuration for a MNT Reform system :)

<luca>mutant ninja turtles?

<vagrantc>hah

<vagrantc>never made that connection before. :)

<lfam>Also find it reliable to push but not to pull

<vagrantc> https://mntre.com/reform.html

<luca>Oh I think I saw that before. Someone in #voidlinux also wanted to contribute some packages to make void run on it

<vagrantc>i've had it for some years, and then got the upgrade module ... and then it still mostly sat unused ... until i realized it could actually do guix on a native arm64 system and thus signifanct chunks of my past couple days

<vagrantc>it knocks the socks off of all the other arm64 systems i have tried with guix

<vagrantc>of course now i run into https://issues.guix.gnu.org/48172

<luca>is it libre? Can it run vanilla guix? (in theory at least)

<vagrantc>e.g. need the /boot partition on SD, but i want guix on NVMe

<vagrantc>anyone have any decent workarounds for a separate /boot partition?

<vagrantc>luca: there is a bit of boot firmware binary blob for training the DDR (on this model) ... so not perfect, but pretty good.

<luca>dance dance revolution?

<vagrantc>luca: wip-linux-libre-arm64-mnt-reform branch has a working kernel with patches and an example configuration, assuming you bring your own boot firmware

<luca>Awesome!

<vagrantc>even *most* of the boot firmware is free

<vagrantc>but obviously one bad apple poisons the lot

<lfam>Is there *any* computer that where the DDR training is free software?

<vagrantc>did not need it for the pinebook-pro-rk3399 ... maybe it was burned in rom

<lfam>Yeah, almost certainly it's in there

<eikcaz>Hi Guix!

<sneek>eikcaz, you have 1 message!

<sneek>eikcaz, lfam says: I was just editing the text you gave for a Guix news entry. Here's what I have: <https://paste.debian.net/1362776/>. Then I saw your message in the logs about simplifying it. Tell me what you think about my proposed text (I'm unsure about the paragraph beginning "Note that..." because it doesn't seem to resolve)

<lfam>Hi eikcaz

<vagrantc>lfam: i think some specific u-boot systems have free implementations. e.g. for pinebook-pro-rk3399 it was actually needed at first, but a good enough free implementation was upstreamed

<lfam>Oh that's amazing eikcaz

<lfam>I mean, that's amazing vagrantc :)

<lfam>It's amazing all around

<vagrantc>a lot of amazing :)

<eikcaz>patch #76962 enabled me to build my entire home environment on my librem5. I think it's pretty simple/obvious. Would someone mind merging it?

<peanuts>"[PATCH] gnu: python-anyio: Disable keyboardinterrupt for aarch64-linux native builds." https://issues.guix.gnu.org/76962

<eikcaz>Hi lfam

<vagrantc>so, if someone could help me extract the kernel patches for the mnt/reform from a git repository instead of shipping in gnu/packages/patches/ I think I would have something ready for guix master

<vagrantc>or nearly so, anyways...

<vagrantc>it is nearly 1.2MB of patches ... so ... :)

<vagrantc>appears to be getting smaller over time

<lfam>eikcaz: Let me know what you think about the news entry I proposed at <https://paste.debian.net/1362776/>

<dariqq>vagrantc: afaik patches can be any file-like object. So if you define the git repository as an origin object you can then pass something like (file-append origin "/path/to/file.patch") as a patch

<eikcaz>lfam: first thing I notice: generally, upgrading will automatically remove ~/.config/syncthing/config.xml because it is just another symlink managed by guix. It only won't work if ~/.config/syncthing/config.xml became a file for one of a few but very rare situations

<lfam>Okay

<vagrantc>dariqq: i understand that in theory ... but i am struggling with the practice :)

<vagrantc>also need help figuring out how to add a couple phases ... with most packages i know how to do it, but with the linux-libre packages it is not as obvious

<eikcaz>Thats why I added the "2> /dev/null" originally to mean "it's safe to do this, and if config.xml doesn't exist anyway, that's also fine"

<lfam>I see eikcaz

<eikcaz>Otherwise, LGTM. Ship it.

<lfam>That makes sense eikcaz

<lfam>Alright!

<dariqq>vagrantc: For phases look at substitute-keyword-arguments

<lfam>Any way to make `git log --show-signatures` not complain about expired signing keys?

<lfam>Or rather, '--show-signature'

<lfam>eikcaz: Pushed!

<eikcaz>Great!

<lfam>Thanks for your work on this!

<eikcaz>No problem. I'm happy to contribute, but honestly the amount I will use this service is already enough motivate the effort.

<lfam>Making things we want to use is a great motivator

<eikcaz>I already mentioned #76962, but maybe I should also note that it (1) is only a single line addition and (2) unblocks building gtk for aarch64-linux, unlocking a lot of packages for building/deployment there

<peanuts>"[PATCH] gnu: python-anyio: Disable keyboardinterrupt for aarch64-linux native builds." https://issues.guix.gnu.org/76962

<eikcaz>lfam: Thanks for your help getting it through! This was actually my first contribution to a community developed free software project, so I really appriciate the time you spent helping move the patch forward.

<lfam>eikcaz: I'm not a strong enough Schemer to understand fully, but I looked at the definition of target-aarch64?, and it seems to already do something about current-system: https://codeberg.org/guix/guix-mirror/src/branch/master/guix/utils.scm#L802

<lfam>Can anyone help us with that and #76962 ?

<peanuts>"[PATCH] gnu: python-anyio: Disable keyboardinterrupt for aarch64-linux native builds." https://issues.guix.gnu.org/76962

<eikcaz>lfam. I had assumed as much, yet the keyboardinturrupt test was still present and failing on my local build (on x86 with --system). Adding that line seemed to fix it, though now I see my gtk was substituted, which undermines my claim about gtk. Let me do one test with latest guix to see if this is still failing

<lfam>Where would we put a "note-taking" app?

<eikcaz>text based or drawing based?

<lfam>text based, eikcaz

<simendsjo>I spent many hours packaging the DisplayLink driver, just to find out evdi was enough :/ Just a little heads up for others who contemplate on using Guix in a corporate environment.

<eikcaz>lfam: which notetaking app is it? I see there is precedent for making a new scm like rednotebook.scm, though I assume it is best if it can fit into an existing package file

<lfam>It's this thing eikcaz: https://github.com/nuttyartist/notes

<eikcaz>how about gnu/packages/qt.scm?

<eikcaz>or maybe qt-xyz.scm?

<lfam>I feel like that's more for the packages that make up Qt, rather than programs that use Qt. I think we could create a 'notes' module. Another Guix channel has a 'productivity' module

<Rutherther>simendsjo: I would expect corporates to upgrade hw regularly and newer hw doesn't need display link at all as DisplayPort over USB-C became a thing. It sucks that you have to use this outdated technology

<simendsjo>Rutherther: We're moving to a new office in 2029, and my guess is not much upgrades will be done in the mean time. All time and money is spent on the new building and moving. But at least things is working right now, so I'm happy :)

<simendsjo>Rutherther: But it's quite depressing seeing how everybody has *a lot* of problem connecting the monitors, especially the mac folks. It takes a long time to detect the screen, and it disconnects (crashes...?) every few minutes. No idea if the evdi kernel module is more stable, but I sure hope so.

<luca>hi, quick question about the codeberg migration. Is the idea to fully stop accepting patches by email, and only accept codeberg PRs? Or would a mix of both be also happening?

<Rutherther>simendsjo: hope you will get it working eventually

<ieure>luca, This is covered in the proposal.

<luca>ah, found it https://lists.gnu.org/archive/html/guix-devel/2025-01/msg00218.html

<luca>So drop mailing lists completely, not just patches. Thanks

<Rutherther>simendsjo: hope you will get it working eventually

<lfam>luca: Where does it say that mailing lists will be dropped completely?

<luca>that's what I got out of "Issue Tracker Migration Path"

<lfam>luca: I think that guix-devel, help-guix, and other private mailing Guix-related mailing lists would continue

<edumerco>Dear GUIX enthusiasts, I'm a new guix user with some years of debian experience. Before migrating completely to guix I installed it over my old debian 10 with great success. Only a small issue remains: after installing the GUIX python, the system still uses the debian installed one instead of the guix installed package.

<edumerco>I searched a lot but found nothing and surely I'm missing the adequate keywords. Does anyone knows any reference that can point me to an explanation and/or solution? Thanks a lot in advance for your kind help. :)

<luca>hmmm ok. Still don't fully get if patches would be accepted by email. But looks like ideally codeberg would be used to validate PRs build/lint

<lilyp>maybe you want python3 rather than python?

<lilyp>add "--version" to see which one's being invoked

<lilyp>(or use good old which)

<edumerco>Ah, probably is python3... "which python" says it is in /usr/bin/python (the debian one). Version says "2.7.6". But "/gnu/store/sk0wqmsnj2brbdz26295hara147zim69-python-3.10.7/bin/python3 --version" says "3.10.7".

<edumerco>Do I try "guix install python3" then? Before, I did "guix install python" (without the 3).

<pkill9>maybe python@3

<lilyp>no, python is the right package, but the binary is called python3

<lilyp>if you want it to be called python, use python-wrapper

<edumerco>Ideally, it should be called whenever I simply call "python"... ;)

<edumerco>Ah, got it, thanks. :)

<lilyp>python-wrapper is your friend then

<edumerco>"python3 --version" gives "3.10.7". So it is there, but "python" calls the old one and "python3" the new one.

<edumerco>Thanks a lot lilyp, you solved it. :)

<edumerco>I will look into python-wrapper.

<edumerco>It is done, and it works. Goodbye everyone, an have a great weekend with your loved ones. :)

<lfam>luca: The goal is to eventually stop accepting patches by email

<edumerco>If anyone needs some interaction design help, please write me at eduardo@mercovich.net so I can give back to the guix community. Bye...

IRC channel logs

2025-03-14.log