IRC channel logs
2026-04-30.log
back to list of logs
<ham5urg>what will happen after some DE are going wayland only but now are not? what will happen to an installation which runs X for now? <ieure>ham5urg, "An installation" does not run exclusively X or Wayland, you can pick the session type in the DM. So the only thing that will happen is the entry for the GNOME (or whatever) X11 session type will be removed. <mwette>For editing files from a package are there options other than subsitute*? (Outside of guix I usually use `ed'.) <mwette>(with hereis : ed -s file <<EndOfInput) <apteryx>ah, default values in guix-record-type* can't yet refer to another field value, right? <apteryx>maybe using a thunked field combined with this-record <apteryx>hm, it looks like the sanitizer runs after a default value is set, for define-record-type* <apteryx>looks like ssh-key for hetzner-configuration is required, unlike what is documented <JazzJackalope>Well... I've been trying out Guix in a VM, and I think I hosed the qcow2 image. I installed LXQt and now I'm in a black void with a window that doesn't work. (It didn't show anything in XFCE either.) I have no idea how to fix this, what's wrong, nor how to go into a different DE <JazzJackalope>There's probably documentation, but I have to find where to start <czan>Can you boot the VM into a previous generation? <czan>I mean a Guix generation. <czan>Whenever you run "guix system reconfigure" Guix will write a new configuration, but leave the old one there as well. When you reboot you should be prompted for a generation (which will default to the most recent, and boot after a second or two). <czan>Huh. That surprises me. Have you been using "guix system reconfigure" to change your system? <czan>I don't really understand. You used "guix install" to install LXQt/XFCE? Because that will have installed them as user packages, not as system-wide DEs. <JazzJackalope>Oh... I didn't know there was a difference because when I installed MATE in between, it started booting into it with no issues <czan>At any rate, if you can get to a shell you can use "guix package --roll-back" to undo your most recent change to your installed packages, or "guix package --switch-generation" (along with "guix package --list-generations") to switch back to any previous package generation. <JazzJackalope>In fact, LXQt prompted me to choose a WM between Marco, xfwm4, and Openbox, just like normal <JazzJackalope>Before you advised about using guix packages, I started reconfigure on an /etc file like the command prompt suggested. Is this something I should stop with Ctrl+C, or let it finish? <JazzJackalope>I was sort of expecting an editor, not packages to start downloading... <czan>The reconfigure command says "make my system match the specification in this file". If the specification in the file matches what you want your system to be, then great, you can let it finish! If not, then stop it. <JazzJackalope>Excellent, package --roll-back got me back into MATE. Thanks a lot for the help <JazzJackalope>I saw a tidy little list from --list-generations of exactly the most recent installations <czan>"guix install" and "guix remove" are really just shortcuts for "make a new generation with this package added/removed", so Guix keeps tracks of all your changes for you. :) <czan>Very helpful for when an upgrade goes wrong. Both at a user-package level and at a whole system level, being able to roll back has saved me a fair bit of time over the years. <JazzJackalope>Now I just need to learn how to properly install a DE with it... My goal is to use Guix on my daily driver to update LXQt to the most recent version <ieure>That's going to be a big change. <ieure>IDK what Noble is. The Ubuntu release from 2024? <Noisytoot>how do I get the location of a package as a string from scheme? <ieure>JazzJackalope, Gotcha, nice. Love to see the less-popular GUIs get some love. <Noisytoot>I'm trying to add the guix equivalent of "install algif_aead /bin/false" to modprobe.d, but I need to get the store path of coreutils for that <ieure>Noisytoot, (string-append coreutils "/bin/false") <ieure>Noisytoot, If this is a thing that's dropping a file on disk, you'll need to use the mixed-text-file procedure. <czan>I think you need to use some gexp magic rather than just string-append like that. Either (file-append coreutils "/bin/false") or (string-append #$coreutils "/bin/false"), right? <ieure>mixed-text-file is the best way to do this if you're creating text files on disk. <czan>If the coreutils variable is a package and not a string then I don't see how string-append can work. :) <ieure>Noisytoot, Disgust at the LLM slop all over that page. <Noisytoot>ieure: they even hallucinated RHEL 14.3 (the latest version is 10.1), but the vulnerability is real <JazzJackalope>Noisytoot: I recognized your goal just from seeing the install command <ieure>Noisytoot, I'm sure it is, I just find LLM output viscerally repulsive. <Noisytoot>although it looks like they fixed that. it said RHEL 14.3 a few hours ago <ieure>Noisytoot, Can you just blacklist the module instead of screwing around with modprobe configs? <Noisytoot>blacklisting it also requires changing the config <Noisytoot>or a kernel command-line parameter, but my goal is to not reboot <ieure>Noisytoot, The "Kernel Module Loader Service" section of the manual has almost exactly what you want. <Noisytoot>the problem is getting the path to coreutils but if blacklisting it works too I don't need that <ieure>Noisytoot, Like I said, mixed-text-file. <ieure>Noisytoot, (mixed-text-file "modprobe-conf-example" "install algif_aead" coreutils "/bin/false\n") <ieure>,build (mixed-text-file "modprobe-conf-example" "install algif_aead" coreutils "/bin/false\n") <ieure>Returns "/gnu/store/5wxh2yba8fasfms6skn4ycsn5d1md1ah-modprobe-conf-example", which contains: install algif_aead/gnu/store/2vzq53b4gmsr4w4dyddjbk5lwwh16im3-coreutils-9.1/bin/false <ieure>That's precisely what you need, yes? <ieure>Missed a space there, but, you get the idea. <Noisytoot>modprobe -vn algif_aead still finds it though... <Noisytoot>it seems to be using /run/booted-system instead of /run/current-system I think <ieure>Noisytoot, `guix system switch-generation'. I am not sure if this will actually work. <czan>How are you loading that modprobe config? Even looking at that copy.fail site, it suggests running rmmod as well as adding something to modprobe.conf. <ieure>When you reconfigure, it creates a new generation, but does not switch to it. <czan>I wouldn't expect the modprobe config to be read if it's changed after boot. <ieure>czan, I would, doesn't modprobe(8) read its configs when you run it? <czan>Ah, it looks like you're right, according to the man page. <Noisytoot>it seems it doesn't actually check any of those paths <Noisytoot>newfstatat(AT_FDCWD, "/gnu/store/4cl2mzyp2ry715vb4y56gh56pi77h77f-kmod-29/etc/modprobe.d", 0x7fff1b63c840, 0) = -1 ENOENT (No such file or directory) newfstatat(AT_FDCWD, "/run/modprobe.d", 0x7fff1b63c840, 0) = -1 ENOENT (No such file or directory) newfstatat(AT_FDCWD, "/usr/local/lib/modprobe.d", 0x7fff1b63c840, 0) = -1 ENOENT (No such file or directory) newfstatat(AT_FDCWD, "/lib/modprobe.d", 0x7fff1b63c840, 0) = -1 ENOENT (No such file or <Noisytoot>I will just manually create /usr/local/lib/modprobe.d and put it in there for now <apteryx>does 'make check TESTS=tests/machine/hetzner.scm' work for someone? <apteryx>ah, because I had GUIX_HETZNER_API_TOKEN set; it's the integration tests that have a problem apparently (skipped otherwise) <apteryx>how do sanitizers and default values interact for define-record-type*, if there's any interaction? <apteryx>reading the code, they seem independent; i.e. you can't expect the default value to be sanitized <efraim>I believe we have a larger than comfortable amount of unvalidated fields and generated configs for services <Kabouik>Woops, sorry for the spam, I mistyped. <noe>Kabouik, yeah it worked on my machine <Kabouik>Thanks noe. I'm upgrading right now but maybe we don't have the fix yet? <efraim>noe: which kernel version are you running? wilko and vagrant updated the kernels yesterday <efraim>it's possible that took care of it <efraim>*in the batch from yesterday in Guix <noe>6.12.80 from Apr 27, so it makes sense that it wouldn’t have the patch <Kabouik>I am on 6.18.19 and didn't upgrade since April 13. Is there anything that can cause you to be on 6.12.80 still? <nnnn20430>i just tried the deobfuscated version on my guix system and also my archlinux system <noe>Kabouik, i use the LTS variant from the channel that shall not be named <nnnn20430>i used the non lts kernel from the channel that shall not be named <civodul>apparently 6.18.22, which we’re currently shipping, is safe <civodul>(the blog post doesn’t even give that info) <civodul>all the other versions we ship are probably affected <kestrelwx>Wonder if that vulnerability is good enough for Android devices. <theesm>civodul: iirc the vuln hasn't been fixed upstream in all kernel versions only 7.x (which we don't ship yet) and two of the 6.x kernels? (but maybe my info is outdated on this, will check) <theesm>oh, there's been a big round of updates for all currently supported kernel versions today, will create a PR as soon as the linux-libre deblob scripts become available for those <dariqq>has someone tried packaging gnustep? <dariqq>Also this is not the first time I have been seriously disappointed by nixos recipes that are just completely broken <cdegroot>nixpkgs is not a single project, more like an ecosystem, so variable quality is to be expected. <n|Phreak>Can you upgrade just the kernel and not the userland ? <civodul>theesm: the oss-sec post i mentioned above suggests 6.8.22 and 6.9.12 include the fix <dariqq>cdegroot: One of the main selling points is the huge amount of available packages, but running a gnustep hello world immediately complains about not finding tzdata. <loquatdev>Hello, everyone. What's the proper way to run a script and/or several commands with `guix time-machine`? Is it `guix time-machine -- shell -- /path/to/script`? <loquatdev>I'm trying to do this because I will be running several Guix commands and moving some things around based on their output. <jab>also, does anyone know if the guixotic developers are on vacation ? They are being a little slow responding to my emails. <marmar22>hello, seems like on my recent installation of guix, some applications are not able to be accessed from the application launcher, although it can be launched by cli. icecat is in the application launcher, but calibre, libreoffice and other kde applications aren't. config is <dariqq>marmar22: how did you install these and what is the value of XDG_DATA_DIRS <dariqq>do you see libreoffice things in /run/current-system/profile/share/applications/ ? <loquatdev>I'm writing a service that runs some commands under a custom user. I'm trying to run `guix pull` as that user but `/var/guix/profiles/per-user/git` (the user is called git) is apparently not owned by the `git` user. Any pointers on how to change this? <dariqq>what to with tests that expect a user with name root and that dont like that the home directorey is just / ? <ieure>dariqq, Ideally, contribute a fix upstream. <dariqq>ieure: I just skipped them with a patch for now. <dariqq>gnustep is in general a bit weird as it wants to have the same prefix for all its packages so I am already at patch 5 to make it sane :( <marmar22>dariqq: yeah libreoffice, calibre is in /run/current-system/profile/share/applications/ <marmar22>dariqq: oh yeah it works now! I assume this would take a while to be fixed upstream? <dariqq>marmar22: in the issue there is a link to the kde bug, so atleast they are aware of it <loquatdev>Does creating a user with `system? #t` cause the /var/guix/profiles/per-user/(user) folder to be owned by root? <ieure>loquatdev, It should not, but I'm also not sure the profile is created at the time the user is created. <loquatdev>I'm writing a service that uses various `guix` commands under a custom user, but the per-user folder is owned by root for some reason. I can't pull or use time-machine as this user because of this. <ieure>loquatdev, Possible you're running those commands as root? <ieure>And not your service's user? <loquatdev>ieure: I'm logged in as the user via `sudo su - user`. There's nothing in the folder at all, and no Guix commands have been run as that user before, to my knowledge. <loquatdev>Perhaps it was created by accident somewhere. Maybe it's worth a shot to delete the folder and reconfigure. <ieure>Seems a reasonable place to start. <ieure>You can also make a tar.gz system image using your system config, and inspect that to see what you get. <ieure>...pretty sure you can make a tarball system image. <loquatdev>I did a little digging, and it looks like the folder's permissions were messed up. It showed root but upon inspection it's got nonsensical group and user ids. I used the same user name for a service I was testing a while ago. I must've never cleaned it up. <loquatdev>I jumped to conclusions a little to quickly... Guix is not the culprit, as usual :) <loquatdev>On another note, is there some variable I can set to disable the progress bars displayed by `guix pull`? I'm displaying the output somewhere else and I get multiple lines of the progressbar symbols. <ieure>loquatdev, Hmm, would expect `TERM=dumb guix pull' to do this, but it doesn't. <mwette>janneke: nyacc-3 does not need bytestructures; can you add `(propagated-inputs (list))' to mes.scm? <janneke>thanks for the headsup!, that's for nyacc >= 3, right? <mwette>on a aarch64 laptop; it works on my aarch64 desktop (both foreign); I ran into issue building openscad so I created a package file in my own channel and called it openscad-local. It seems to be having issues getting git submodules. <mwette>I'm using the openscad package def from engineering.scm but adding a sub* to comment out a test that is failing. <ieure>mwette, Does the test only fail on aarch64? I pulled earlier today, and `guix build openscad' works fine on my x86-64 hardware. <mwette>ieure: thanks for testing; I don't have an x86_64 machine to test on. The errors are in export-svg. And, IIRC, the failures were different on the two aarch64 computers (desktop Ampere Altra, laptop Snapdragon). <ieure>mwette, Well, let's see, I had pulled out this Snapdragon ThinkPad earlier this week, sounds like that's a close match to your setup. <mwette>OK. linux-on-snapdragon is a bit bleeding edge right now. <ieure>I know :( This week is the first time audio or sleep has worked on this machine. <mwette>you're ahead of me and my asus zenbook A14: no suspend, no audio, kbd backlight is random <mwette>but I have sway, guix, emacs going :), laptop is under 1 kg <ieure>That's nice. I have a T14s, some guy bought it with minimum storage and a cheaper third-party SSD, but they couldn't figure out how to reinstall Windows and gave up. So I got it for like 60% off. <ieure>Yeah, though it's been collecting dust for a year while I wait for the software to catch up. <ieure>Still the fastest hardware I own. Faster than the MacBook work makes everyone use. <ieure>mwette, I can reproduce your aarch64 build failure, same deal, it fails on the tests. <ieure>Wouldn't think your package tweak would change how submodules are handled. <ieure>mwette, I'll try your patch in a bit here. <mwette>I don't know. The odd think is that it worked (and runs) on the desktop. <vagrantc>mwette: does it build using ... --without-tests ? <mwette>Not right now. `guix pull' on both will get them synced, right? Do I need to upgrade also? <mwette>vagrantc: I didn't know that option. I'll try. <ieure>mwette, It can happen, but is not guaranteed. Ideally, find the one with the older commit and `guix pull --commit=...' with the ID of whichever is newer. <ieure>mwette, But my guess is that your desktop is on an older commit, and laptop is newer, and something in between broke openscad. <ieure>If it was the other way around, I wouldn't have been able to reproduce the issue. <mwette>Note the patch is needed for the desktop. <trev>a wonderful non-building php has appeared <trev>i hate that i even have php on my system somehow <ieure>My ideal version of PHP also does not build at all. <trev>how do i figure out what is pulling in php? <trev>nevermind...gc --referrers