IRC channel logs
2026-08-17.log
back to list of logs
<apteryx>hm, why is #:cmake marked as private in build-system/qt.scm? bug? <bdunahu>I think including the `cpp.scm` module to `profiling.scm` is creating a cycle leading to weird behavior with missing symbols for me...: profiling.scm > cpp.scm > mpi.scm > profiling.scm. how do I avoid that? <niqc>Guys, yesterday i ask about run lsp and others on guix, has been recomended that run inside a fhs container, but it is very slow, has other way to make it? The lsp make my emacs slow on typing. <ieure>niqc, What LSP? It would only make sense to run in a FHS container if the LSP isn't packaged for Guix. <ieure>I have used LSPs under Guix, they have packages, I included them in my Guix Home profile and it was painless. <ieure>Performance definitely suffers if the LSP itself sucks. My current job is writing Go (it sucks lol), the Go LSP had its performance absolutely tank around the Go 1.24 / 1.25 release. Typing isn't slow, but saving a file takes like 2-3 seconds now. I hate Go. <perryprog>But it has generics now what more could you ask for /s <ieure>Absolute clownshoes language. <niqc>I need the nightly version of rust analyzer and others, i'm using uv too, and nice to have typos-lsp, i'm using eglot. <perryprog>I was installing the gh cli (written in go) on gentoo yesterday and it took like. Ten+ minutes to install. Meanwhile emacs will build in less than a minute. The binary is 40 megs. <perryprog>niqc, huh, rust analyzer and eglot should definitely be good performance wise. <perryprog>(I don't have that much experience with rust analyzer but I used it for a small thing a week or two ago and it felt fine) <niqc>i try too with python using zuban lsp, but the same result, i'm runing it with guix shell inside emacs <ieure>niqc, Use package transformations to get the latest versions? <niqc>and others, this package is only a example. <ieure>Hmmm. I wrote a service that extends activation-service-type to put some stuff in /boot which is needed at boot time (this is for a raspi, so it's a config.txt and... some other files). I added the service to a system config and used `guix system image' to make an image. But when I look at it, none of the stuff I expect is in /boot. <ieure>I think what's happening is that the generation is activated when it's booted, and that would populate the stuff I expect. But it needs to be present for the system to boot in the first place, so, that's not going to owrk. <ieure>Is there a reasonable way to get those files in place when I make the image? <fishflake>anyone getting downgrades with the most recent commit? <fishflake>i just pulled like 2 seconds ago and got guix system: error: aborting reconfiguration because commit e343ff040092cd3428f7d35423add6fdb2939f53 of channel 'guix' is not a descendant of 869fe0bf833d62ea0cc788bd8599fcf75cb03d0e <ieure>fishflake, I cannot reproduce. Codeberg says both commits are in the history of master. <fishflake>yeah this is a fresh install i just updated yesterday :/ <fishflake>do i need to intervene by cloning the new commit manually <fishflake>luckily the first thing i added to my config.scm was git because i knew this would happen lol <ieure>fishflake, You're getting the error on `sudo guix system reconfigure'? <ieure>fishflake, And e343ff040092cd3428f7d35423add6fdb2939f53 is the commit of guix shown in `guix system describe'? <fishflake> commit: 869fe0bf833d62ea0cc788bd8599fcf75cb03d0e <ieure>fishflake, Okay, and `guix describe' reports commit e343ff040092cd3428f7d35423add6fdb2939f53? <ieure>fishflake, The problem is that your guix binary is built from an older commit than the one used to build your current system generation. Something is wrong with your Guix profile. Did you `guix pull' as root or something? <ieure>fishflake, You were in here recently fixing a problem where you had no user account after the install? <ieure>fishflake, Gotcha. You need to `guix pull' as your user, not as root. <ieure>fishflake, If you ran `sudo guix pull', you may need to fix permissions in your user's $HOME/.cache/guix <fishflake>i ran it as root user instead of sudo would that do it <fishflake>for some reason sudo doesnt work on my system not really sure why <ieure>fishflake, The problem is the same either way, just if you used `sudo guix pull' you may also need to fix permissions in your user's $HOME for `guix pull' to work. <perryprog>Is there a reason there isn't a warning or similar when running guix pull as root? <fishflake>maybe u can do that in some cases but not the one i was doing <ieure>perryprog, Because nobody has added one. It's not a bad idea. <perryprog>Figured; wasn't sure if it was a Chesterton's fence <ieure>fishflake, Guix is not like Debian/RedHat/etc. There is not a system-wide list of available packages which root needs to update. The recipes to build packages are built-in to the `guix' binary, and each user has their own, because they may have different channel configurations with different packages. <fishflake>do you know whatll happen to gnome when they reach full systemd dependence <fishflake>and gdm, i think were rolling a few versions behind current gnome right <ieure>fishflake, Guix has been adding support for the systemd stuff they need to depend on, I assume this will continue. Or not. <fishflake>ya i know someone from artix got current gnome working on whatever init he was using, and chimera also has it working <ieure>Yeah. Fairly safe to assume this will remain the case, unless/until GNOME's systemd integration becomes too tight to replicate in some other init system. <fishflake>on gentoo i had issues related to alt inits on nvidia for some reason <ieure>Well, nVidia is just The Worst. <fishflake>someone actually donated me that card to watch me suffer so i get your point <ieure>After years of trying to get the GPU in this desktop rig I otherwise like, I just pulled the damn thing and am using the iGPU. <ieure>Which, like, works. Which is an improvement. <fishflake>ill say tho if ur willing to just use a normal distro turing+ isnt that bad these days <fishflake>a few things i cant do especially on wayland but otherwise its pretty solid as a daily driver <fishflake>amd equivalents are way cheaper and probably better though <psycotica0>ieure: I'm only kinda paying attention, but on an Ubuntu system I sometimes run `sudo guix pull` to update my build daemon and stuff, whereas `guix pull` updates my user's information. Is that different on guix system? Is `sudo guix pull` just always broken there? Or is there still some equivalent case where it's important? Or, perhaps more importantly for me, am *I* doing it wrong, and I shouldn't be run <psycotica0>ning `sudo guix pull` even on a hosted guix install? <ieure>My daily driver laptop with Radeon has been trouble-free, but it's just whatever's integrated. <ieure>psycotica0, Foreign is a different case, you do want to pull as root there for the reasons you state. But this is almost always a mistake on Guix System. <fishflake>ya this thinkpad just got 8th gen intel graphics works fine even though guix said it probably wouldnt work <ieure>Yeah, I have an 8th gen Intel ThinkPad also, no problems there. Especially since I backported the tearfree patch for the modesetting driver. <psycotica0>Gotcha. And is that because you should actually be doing `sudo guix system configure`? Does that do the pull on the inside? What's the equivalent of updating the build daemon on guix system? <ieure>That's in the Guix Xorg package now, so if you use Xorg with modesetting, you can get tearfree graphics instead of beating your head against the wall of a half-dozen buggy unmaintained circular forks of broken X compositors. <fishflake>substitution of /gnu/store/n4xwlnb807vqvarwnc7cg7sxnxllj3zl-linux-libre-6.18.44-guix.tar.xz failed <vagrantc>looks like something triggered a world rebuild ... openssh update, maybe? <fishflake>guix system: error: corrupt input while restoring archive from #<closed: file 7fe4132e40e0> <ieure>psycotica0, `sudo guix system reconfigure' is how you update everything on Guix System, including the daemon. That does not `guix pull', you need to do that yourself *if* you're updating versions. You'd choose not to pull if you were, say, changing configuration without wanting to update at the same time. <fishflake>oh do i just wait for it to get resolved or something <fishflake>surprised were still on 6.18 kernel 7.0 is out right <vagrantc>7.2 was released today, 7.1 was in guix for some months now <vagrantc>fishflake: are you using linux-libre-lts or something? <fishflake>tbh i just installed this yesterday im running whatever guix told me to run <fishflake>but when it asked for a kernel i chose libre <fishflake>hurd 64 would have been cool but then i couldnt have a display server <vagrantc>guessing the ruby or openssh updates ... <vagrantc>or gcrypt ... or openssl ... so many possible world-rebuild triggers <fishflake>is there a way to change my profile to pull new kernel tho <vagrantc>fishflake: check if there is a (kernel ...) defined in your system configuration <vagrantc>if you want to avoid the world-rebuild, i'd recommend pulling to an older commit.... 21c3d67 seems to be the last evaluation before ... <ieure>fishflake, Agree with vagrantc here. Find what kernel is in your configuration, it one of `linux-libre', `linux-libre-lts' etc, then `guix show linux-libre' (or whatever) will tell you what version is configured. IF you've reconfigured your system and not rebooted, you'll still be running whatever you booted into. <fishflake>i dont mind rebuilding world if i need to i just dont know why it pulled such an old kernel <fishflake>but no i dont see a kernel defined in my config.scm <vagrantc>ah, the crypto-team branch merge was probably what triggered all this <vagrantc>curiously, 6.18.44 is the most current 6.18.x/linux-libre-lts ... so you're not on a super-old guix commit <vagrantc>default should be linux-libre, which *should* pull in 7.1.x <fishflake>ya i thought i just pulled like now so i should be on current <fishflake>i dont see lts defined in my config.scm either though not sure where thats defined or why <vagrantc>nckx: you still using bcachefs ? bcachefs-linux-module hasn't been building for a while ... probably in one of the major kernel updates or something <vagrantc>it was a really good talk, in my opinion ... simultaneously an introduction to the topic, switching part-way through into a deep dive <fishflake>im guessing i need to switch release channels yeah? <vagrantc>fishflake: there's no release channels ... just guix master and unofficial channels <fishflake>oh ok, why is master pulling such an old kernel <ieure>fishflake, It isn't necessarily, multiple kernel versions are supported, your config may be referencing an older version, or you may have not rebooted after your system reconfigure, so are still running the kernel from the generation you booted into. <fishflake>well i rebooted and i pulled and reconfigured which failed on 6.18 <vagrantc>or some other tool may be requiring the older headers <fishflake> linux-libre-6.18.44-guix.tar.xz 147.9MiB 676KiB/s 00:01 ▕ <fishflake>my only packages currently in my config are mpv git and fastfetch <ieure>fishflake, Those should not be in your system configuration at all. <fishflake>why not dont i define system packages in config.scm <vagrantc>in general, the recommendation is to include as few packages as possible in the system, and use user profiles user guix home to define the packages each user has available <ieure>fishflake, Those are not system package, they're user packages. <fishflake>ohhh is there a separate place to have user packages <vagrantc>but ... this is free software, you can do as you like too :) <fishflake>i saw you can use guix package for that but i dont think thats persistent or adds them to path <ieure>fishflake, Every user has their own value of $PATH. <ieure>And it is set for you, on Guix System. <vagrantc>the advantage being your system configuration has a smaller set of dependencies ... keeping the user and system configuration more distinct <fishflake>ok so what should i do instead to add user packages <vagrantc>fishflake: unfortunately, you've installed guix on a bad weather day ... there is a lot of compiling to do... <fishflake>its fine i can wait, this isnt my daily driver yet <ieure>fishflake, If you want to use imperative-style package management, `guix package -I mpv git fastfetch'. If you want to use declarative style, create a manifest, or use Guix Home. <fishflake>yeah whatever i shoud use just to get the packages installed tbh, thats all i want lol <fishflake>on gentoo i learned how to like organize my use flags and stuff but idk how to do it on guix yet <ieure>`guix package -I' is the simplest. <vagrantc>or the modern equaivalent, guix install PACKAGE <vagrantc>shorter to type with less odd keystrokes ... <vagrantc>guix install|remove|upgrade ... for user profiles <ieure>fishflake, It will create a new generation of your per-user profile; the profile is kept in the store and symlinked from your ~/.guix-profile. The profile itself is a sparse filesytem (ex. it will have bin/ share/ lib/ etc directories) full of symlinks back to the store items for the packages within the profile. <fishflake>but yeah bad weather days are what im here for i like being part of something :) <ieure>fishflake, Manipulating $PATH (and similar variables) points programs to the store locations, allowing you to run the programs in it. <fishflake>ok so if i guix install git i can run git from terminal right <vagrantc>you might need to reset your path or log back out and log in again the first time you install something into your profile <fishflake>yeah now that i removed those packages ive got new kerne <vagrantc>maybe you were not using the guix you pulled ... that's also a path issue, potentially on first install <fishflake>id type an emoji to convey this feeling but idk how to type emojis on gnome <vagrantc>(although 6.18.44 is quite recent, actually) <fishflake>yeah its like the kernel my risc v sbc runs lol <vagrantc>6.18.44 was just released august 9th, so it's the most current 6.18.x <vagrantc>guix provides all current kernel versions (although i will start removing the really old branches soon) <pinoaffe>i'm getting "undefined variable: gexp" even though I'm importing (guix gexp) into said module - anyone know what might be going on? <ieure>pinoaffe, You'd have to paste whatever code is causing it. <cbaines>You're mixing quasiquoting and gexp's incorrectly <pinoaffe>aaaaah, I'd completely overlooked the quasiquote! <cbaines>maybe unquoting the gexp bit, e.g. ,#~ would work, but generally if gexp's are used in the arguments, (list is used rather than `( <rrobin>does anyone have a working config for pppd? Mine keeps trying to read the pap-secrets file from the store (the pkg does have a file there). I can't seem to override this at runtime.