IRC channel logs
2025-10-26.log
back to list of logs
<libvirt-question>Hi Folks, I wanted to use libvirt which was relatively straight forward. Creating VMs and all of that worked in Bios mode. Once i tried to set up a UEFI VM the issue got strange. virt-manager complained that no firmware could be found. I inspected the xml and checked the firmware path defined there but the file( ovmf_code_x64.bin)which where <libvirt-question>referenced to be in store did not exist in this path. Where does libvirt get the path from. Is it possible to rebuild only one package of the system configuration? <libvirt-question>like .... installing ovmf package then afterwards rebuild only libvirt or virt-manager in hope the correct path gets pocked up? <redacted>package builds can't run a server on localhost and connect to it during tests, right? <redacted>I'm packaging something that wants to connect to a running Redis server during testing. <tomenzgg>Thanks, ArneBab; unfortunately, I think it's a little too low-level for my goals as it seems to be specifying particular generations of profiles, as well, rather than just the profile itself. <vhns>Is there some way to test whether a channel's files/packages are defined correctly without pushing it to the remote, pulling and waiting for the derivation to build <untrusem>vhns: you could ask hako to enable substitutes for your channel <vhns>I meant more in the sense of "How can I know I haven't broken the channel before pushing it upstream?" <vhns>or to the remote, rather <vhns>I know that locally "guix build -L . foobar.scm" works <vhns>Is there a simple way to include an entire directory into /etc? <decfed_>vhns: I did that once using the etc-files-service-type or such <vhns>I wanted to include all files from my system's repository into the install media such that I can boot it an instantiate a new install more easily <vhns>so I have a clone of my systems' configurations under /etc/ in the bootable usb <vhns>I initially thought of doing so by just having a channel with the multiple system configurations defined such that I could do something like 'guix system install -e '(@ (foo bar) bar)' /mnt' but I could never really get it working <futurile>vhns: so I guess you had a package for each host you were trying to configure then, so you installed a package that had all of /etc in it? <alanz>I am confused. On both my debian and guix machines, if I do `guix gc` and then `guix home reconfigure` it has to download a whole lot of stuff. But `guix gc --list-roots` shows the home profiles as included. Is there some config I am missing? <Rutherther>no, not at all. There are two types of paths that aren't gc rooted, but still necessary for reconfigure - 1. the build tools that build the profiles, services etc., 2. non-grafted paths. So yes, when you gc, you then need to download those things <voroskoi>I try to package tinymist based on rust-team branch. Is that normal that guix tries to rebuild half of the world with bash, gcc-4.3.4 and such? Seems a bit overkill <Rutherther>it's normal on team branches that have recently got updates somewhere close to bootstrap <voroskoi>happened something in that branch and I should just wait for ci? <voroskoi>can I check somewhere how long should I wait? <Rutherther>why are you packaging it on rust-team in the first place? does it require rust newer than 1.85? <voroskoi>and usually nobody merges my rust PRs against master, so I have thought thats the way of doing it <Rutherther>as long as it doesn't depend on something new on a team branch, it should go to master <voroskoi>what should I look at on that linked page exactly? The failded build number? That is quite high all the time <voroskoi>hm, I can try an older version against master <srbaker>Heya folks. I'm curious about deployment of clusters and server management in Guix. Specifically, I'mlooking to move a bunch of services I have in a small kubernetes cluster to be on Guix instead. <srbaker>I'm hoping someone has some articles or docs on how they do server deployment and management with guix. <Rutherther>voroskoi: if you look in built packages for rust-team, rust 1.90 is already built, so all bootstrap had to be built as well <thanosapollo>should the emacs package include a patch to fix #'tramp-get-ls-command for guix? It seems tramp has some issues finding 'ls in guix. <thanosapollo>will send an email to emacs-devel, gonna check the git history to figure out when the breaking change occured, with previous versions of tramp from 2018 it seems to be working fine tbh <voroskoi>Rutherther: I am sorry, I do not fully understand guix. Does it matter when I ran guix pull for the last time or it should just work? <Rutherther>voroskoi: if you are using pre-inst-env, what's in the checkout matters, guix pull doesn't matter at all <home-service-que>Hi all...playing with home services. Observed tat from yesterday onwards8not sure exactly when) the home-services configurationdoes not survive a reboot. Here is whats happening: <home-service-que>- i reconfigure home after adapting a config file which goes int "~/.config" <home-service-que>- i check the adapted config file and it does not have the expected content but content before adaption <home-service-que>is there a clear chain i an follow to try to debug the befroe-home-service-reconfiguration and after-home-service-config-reconfiguration ? <home-service-que>- before home reconfigure /gnu/store/h69pjyi4gnz7apqj3nwlb7jv7lc0gla3-sway <home-service-que>- after logout/relogin still /gnu/store/hkfrnn5yxb2nbsc83i96kijcigjhpcjj-sway <home-service-que>- after reboot /gnu/store/h69pjyi4gnz7apqj3nwlb7jv7lc0gla3-sway again... <csantosb>Great to have deprecation messages with make 🙏 <trev>anyone ever see an issue after doing a guix import on a rust crate and it missing crate dependencies that are not in a lockfile but perhaps a dependency of a dependency? <home-service-que>My Problem with the home servie is gone since i did a system reconfigure. Hmmm....can this happen due to a "guix gc"? <rrobin>trev: only when the rust pkg is a workspace <rrobin>i.e. if the dependency is a crate part of the workspace manifest in the repo you are importing <rrobin>if that is the issue #:cargo-install-paths would help there <home-service-que>seems like i always have to do a system reconfigure to make it permanent...is that ok? ... i thought home reconfigure would take care about everything in my home folder. It does so if i carry out the command but not persistent? <rrobin>how do people usually test new services they add to their channels locally? i'm trying setup a container/vm with guix system to test it, but am failing at it. The service is in the os config, but absent from herd in the running system. Feels like I'm failing to setup properly (but guix describe has the right channels/commits) <redacted>rrobin: Can you share your container/vm config? <rrobin>the host-name is juste due to rootless containers (does not seem to matter) <redacted>And is the radicle service the one not showing up in herd? <rrobin>not in herd and not in logs either <rrobin>ah wait i see something in log from the vm <redacted>Nothing jumps out at me but I'm getting on a call in 5. I can take another look in an hour or so. <rrobin>no worries - now at least i have an exception in the logs - just strange that herd does not show anything and no logs in the vm <redacted>you're doing `sudo herd` in the vm, right? <rrobin>yeah the surprise here was mostly no output in herd and the missing logs in the vm <rrobin>herd start radicle still says 'service radicle could not be found' <rrobin>but it did try to start it at startup - apparently failed on some unbound variable <rrobin>ah probably some gexp bits in make-forkexec-constructor call <Rutherther>yes, if you have unbound variable, the service cannot be recognized by shepherd at all <rrobin>makes sense, the lack of logs/errors is what tripped me off here (felt like the service was not there at all) <Rutherther>just say you wanted to reboot all the bots in this chat :) <rrobin>well, i hope your password is not 'ooops' <home-service-que>nope..it is not...still trying to get behind the "huix home reconfigure" not being persistent. Am i halucinating? isn't a home generation once created and set as current not supposed to be reboot persistent? <rrobin>it is supposed to persist ... you happen to have the guix home service enabled too or something <Rutherther>it depends on your setup. If you use system guix-home service type, then home is activated on boot from system generation and you need to rebuild your system to switch to new home generation. If you just use guix home through guix home reconfigure, then the only activation is done when you run the command <rrobin>ah my issue is improving, service now shows up in herd - it was indeed a gexp in the service definition that was wrong <ngz>Hello. Do you know where (if) the pyside6-uic binary is provided? It seems to be required to update plover package. <sneek>Welcome back ngz, you have 1 message! <ngz>apteryx: I fixed those a while ago ^^ <home-service-que>hi...i get "guix system: error: service 'file-system-/sys/fs/cgroup' provided more than once" ... how can i see which are the ones providing it? <Rutherther>not an easy tool for that, you could evaluate them in a repl. But this is probably docker and rootless-podman <ngz>Odd. Building Calibre fails on CI but succeeds on my machine. <andreas-e>ngz: I think calibre has some non-determinism. <ngz>andreas-e: At least, it seems so in its tests. <andreas-e>Failed today, succeeded two days ago; similarly for the previous versions. <ngz>OK, I’m going to disable the faulty test. <ngz>andreas-e: BTW, I looked at Openboard, and all seems fine. <andreas-e>voronoi: To understand gexps, I would recommend to read about quote, quasiquote and unquote in scheme. And then look up what the corresponding gexp syntax is. (I do not find the choice of symbols intuitive and always have to look up which is which.) <andreas-e>In short: quote means that something is to be evaluated during the package build; quasiquote means the same thing, except that inside a quasiquote one can then unquote and inject a value from the package context. <vhns>09:41 < futurile> vhns: so I guess you had a package for each host you were trying to configure then, so you installed a package that had all of /etc in it? # Pretty much, yes