IRC channel logs
2024-09-06.log
back to list of logs
<podiki>anyone have an idea why tests (gst-plugins-base here) would fail with "Error creating context No GLX extension"? <podiki>this is on the mesa-updates branch, so newer mesa <aswjrisp>xelxebar: In /var/guix/profiles/ there are sysmlinks for different generations. If you know the generation you want you could do /var/guix/profiles/system-42-link/ for the 42 generation for example. <xelxebar>aswjrisp: Yeah, that's what I asked above. I guess you just joined? <xelxebar>Wondering if there's a better way to execute stuff from old generations other than grubbing around directly in /var/guix/profiles. <podiki>you can roll-back to previous generations of course <tatsumaru>hello, just running the guix qemu image in a VM and trying to test a few things, but so far can't get anything to work. installed KDE Plasma, and when I restart the VM I see Plasma loading and then a black screen, then I do ctrl + alt + f1 install a few packages like sddm and when I do sddm it just returns error about 'xauth' missing even though I also installed 'xauth'. I also installed 'lightdm' and when I type 'lightdm' in terminal it s <tatsumaru>ays 'command not found'. I am very confused. <nckx>At least that's what you'd do on a ‘real’ Guix System. I have no idea whether the VM image allows you to edit its system configuration and reconfigure. <nckx>That doesn't change the packages vs. services distinction though. <nckx><There is no ‘lightdm’ binary> not actually true, I typoed because no backspace key 😒 But the rest is true, promise. <tatsumaru>so you are not supposed to just install packages at all? <robin>tatsumaru, installing random packages is fine, but some require services to be at all useful (like x display managers) <robin>(iirc even getting a useful "startx" command is tricky under guix) <tatsumaru>robin: i see thanks., I have some reading to do, I guess it's different from other distros <robin>tatsumaru, system configuration is for sure quite different. not as much on the user level, where "guix install foo" works as you'd expect :) <rynn>nckx: Good call; I did a pull followed by a reconfigure. After the reboot, gnucash installed just fine. <podiki>gsl change on master is causing tons of rebuilds? <apteryx>When I 'guix deploy' to a remote machine for the first time, it ends with "error: unauthorized public key"; it seems this should have been handled automatically with this 2019 commit: https://issues.guix.gnu.org/36956#2 <jlicht>it seems the json feed files we use to run the CVE linter has some missing info :/ <apteryx>re key missing from ACL; could it be that the guix-daemon needs to be restarted? I think this ACL is read every time though... <apteryx>ah, there was another change recently that turned this file into a declarative one; so it's now a link to the store <apteryx>it's probably not mutable anymore, so that change no longer takes effect <dthompson>anyone else having trouble building the master branch right now? <dthompson>I'm getting failures to build the german version of the manual <dthompson>I think I did this but I will do that again in case I forgot... <apteryx>(about 'guix deploy' not authorizing coordinator anymore) <dthompson>this has happened to me in the past but never figured out why... problem just went away after awhile <apteryx>as in 'git worktree add ../other-tree' <apteryx>or 'git clean -xfdd' in your main tree if you don't mind loosing uncommitted stuff laying around <dthompson>yeah I'll start doing more drastic things... <dthompson>I thiiiink I may have fixed it by deleting the po/ directory, reverting all the git managed stuff I deleted, then re-running the build starting from ./bootstrap <dthompson>not sure why translations cause such problems but this isn't the first time <thorondir>Hi all, I'm trying to get the home-dotfiles-service to work with the 'stow layout, but feel like I'm missing something: <thorondir>if I don't mention any packages to include, it gives me the error of "guix home: error: duplicate '' entry for files/" <thorondir>and if I tell it to take everything in the dotfiles repo, via "(packages '("."))", it obviously links _everything_, including the .git folder and everything within, which is obviously not what I want. <thorondir>Exclusions happen via regex, but only on the filename, not on the dirname, so I can't exclude the whole .git folder. <thorondir>So my question is: what am I missing? The documentation doesn't help me, because it just says "you can use the stow layout", but doesn't go into details. <thorondir>For clarity's sake: I use `stow .` to link the dotfiles to where they need to go, so I thought that should suffice. <rekado>I built the installer image from the latest commit. Even after fixing a bug (missing zstd, just reported it) there's a problem with the locales, leading to a backtrace as the installer starts. <rekado>(it complains about "tok", but I haven't investigated more) <xelxebar>Huh. guix system reconfigure keeps erroring with "error: failed to load 'config.scm': No such file or directory" <xelxebar>The file definitly exists. Same thing for /run/current-system/configuration.scm <rynn>xelxebar: The file exists, but is it in the current directory, or are you putting in the full path? <xelxebar>rynn: config.scm is in the current directory, but same issue happens even for absolute paths. <cbaines>is there a way to have a system without guix-icons? <cbaines>given it depends on both rust and go, it's pretty expensive and I don't think it's ever built on riscv64-linux <cbaines>ah, I see, it's in %base-packages-artwork <port19>amd and nvidia gpus don't work with linux-libre. Are modern Intel iGPUs supported on linux-libre (and GuixSD by extension) ? <cbaines>port19, some work well, but I don't know if all do <port19>What about the Intel UHD 7x0 (11th - 14th gen) or Intel UHD 6x0 (8th - 10th gen)? The lineup is pretty small and doesn't change very often, making this quite convenient <cbaines>I think 11th gen intel works fine, I can't speak to other generations though <xelxebar>Hrm. The ENOENT error above looks like it references the incorrect path. <xelxebar>Throwing `guix system reconfigure' at strace, something appears to be trying to read a nonsensical path <xelxebar>openat(AT_FDCWD, "my-hostname/etc/dhclient-enter-hooks", O_RDONLY) = -1 ENOENT (No such file or directory) <apteryx>port19: I was surprised to have basic video at all using an AMD iGPU a few days ago <apteryx>granted, it doesn't seem to do any opengl, so things requiring it like GDM will fail <xelxebar>I do have a `/etc/dhclient-enter-hooks' on the host (to make dhclient use resolvconf). <xelxebar>Not sure how that's directly relevant to a reconfigure, though... <podiki>yeah even a few years ago i could at least get basic video with nomodeset on linux-libre for a current amd gpu <podiki>i spent a while last night figuring out why gst-plugins-base was failing tests on mesa-updates, disabled a bunch of tests...and then saw a very fresh mesa version, which does indeed fix those issues <podiki>anyway, rebasing and updating mesa-updates with the latest libdrm, wayland-protocols, vulkan fix, and mesa; patches/discussion to be sent shortly <cricri>hi there, I've got a question about a guix qcow2 image I created for qemu. When I launch the image with qemu it seems to boot fine (you can find the output log at https://paste.rs/RWiOx). However, there is no tty I can access. Normally, I would type alt-ctrl-f[1-6] in qemu to access the ttys. In this case, nothing happens when I send those keys to qemu. What might be the problem? This is the guix configuration I built: https://paste.rs/qfnhm. <rynn>I wish I had known I was going to be going the Guix route when I built my current system; it has an AMD igpu, and as previously said I can only get basic video with nomodeset on linux-libre. <podiki>rynn: i don't know if anything from the past many years works fully without some binary firmware or similar. sadly the current state <podiki>i will say, you will have a much much easier time than if you have nvidia, if you do go down a route for full functionality (true more generally in linux, not guix specific) <rynn>Yeah, my next system is going to be Intel almost entirely because I should be able to use Linux-libre with it. <rynn>Though, I don't think I'll be building from scratch again next time; I'll likely grab something like the Purism Librem Mini instead. <rynn>That way I can be reasonably confident it's as free-software friendly as possible. <rynn>Franciman: A chuckle for sure <podiki>still depends on the chip right? i'm not sure about iGPU but i believe their discrete graphics need firmware of some sort? <Rutherther>ekaitz hello, I have another update on the arm-none-eabi toolchain. I've fixed the libstdc++ paths, and also noticed another issue with getentropy that I found a patch for. Apart from that I also discovered the cplus include path was wrong. The C should come after C++ include dir. That is because <cstdlib> uses include_next to include <stdlib.h>. This is also an issue for the toolchains that are already in Guix! I am able to compile examples from... <Rutherther>if I want to reply to an e-mail from guix-devel list, I suppose I should add the guix-devel at gnu.org to CC so it appears in the archive? if the e-mail is from Dec 2023, will it pick up properly? - How does it even know the e-mail I am replying to? <aswjrisp>I am happy that there are mailing list archive. I was trying to get x to start manually form the linux console. I started searching through the help mailing list archive and found a guide by Alex Kost with the subject [HOWTO] Start X server manually instead of using a login manager. <aswjrisp>The helpful part I was missing was the options to use to start x with xinit. An example that worked for me was in that guide. <Rutherther>that's quite an outdated solution, there is startx-command-service-type service that makes a wrapper around xinit <freakingpenguin>Are home-environments in a channel supposed to be evaluated during $ guix search? One particular function in my channel is printing when I don't expect it to. <ieure>Is there some way to inspect what guix-daemon is doing? My build box has been spinning fans all morning, but Cuirass claims there are no running builds. ps(8) shows guix-daemon is building some webkitgtk thing. Why would it be doing that, how can I figure out what triggered it? <ieure>oh... 17439 ? Z 0:00 [cmake] <defunct> <aswjrisp>Rutherther: I had tried startx-command-service-type before trying the approach in that guide. I guess I can try again to get startx-command-service-type to work. <Rutherther>cross checking arguments in startx-command-service-type against the ones in the e-mail it's mostly the same, so I would expect the service to work <fnat>aswjrisp: I have a simple '(service startx-command-service-type)' plus a '.xinitrc' file - this works well for me, I don't think I'm relying on anything else other than that. <aswjrisp>fnat: Are you using %base-services or %desktop-services or your own set of explicit services? <aswjrisp>Rutherther: Okay I will try agian to get startx-command-service-type to work for me. Maybe I added some packages or services to my config and it will work now. <aswjrisp>With the startx-command-service-type in my list of services and a .xinitrc that I know works if I call xinit manually, when I run startx I get this error output https://termbin.com/gakm <aswjrisp>Rutherther: `which startx` outputs /run/current-system/profile/bin/startx. That symbolic link points to /gun/store/longhash-xinit-1.4.2/bin/startx Did you want the long hash? <Rutherther>no this suffices. That means you are not actually using the command coming out of the service. That is, I suppose, because you put xinit to your packages list, and its files won over the files from the service. Easiest will be to just remove xinit from the list of packages <Rutherther>to verify you are using the package, the package name in tha path should be"xorg-wrapper", not "xinit" <aswjrisp>longhash is 9174sqv0xxyg2a51py6107hn7mmbmf7f <aswjrisp>Rutherther: fnat: thank for your help startx is now working. <aswjrisp>Rutherther: All I needed to do was remove xinit from my config as you suggested. <ekaitz>Rutherther: so! i was right! if it was unable to find the libc properly those functions were not found by the linker! <ekaitz>Rutherther: do you have a link for me? <ohyllad>i still don't understand how service extensions work it seems. i want to add a file in /etc, but using etc-service-type directly results in a collision, how are you supposed to extend the etc service from a os definition really? <ohyllad>the documentation speaks of "passing" it a list, but does not mention how or where <Rutherther>you make your own service-type that will extend the service - when extending a service you say which service to extend, and give a function that takes value/configuration of your new service, and returns back something that the other service should add. So in this case you make it a function that will return a list with file names and contents <Rutherther>You can use "simple-service" that will make the "service-type" for you in case you just want to make a simple extension of one service <ohyllad>ok that is starting to make sense, i feel like my mental model for a lot of stuff in guix is ever so slightly off <ekaitz>Rutherther: could you please link me to the arm toolchain? <cow_2001>can't get one of my packages to compile ~_~ <ekaitz>cow_2001: it's probably your syntax somewhere <ekaitz>or it was expecting a value and you have an empty list or something like that <ekaitz>take a look to the way you made your phases and gexps <rekado>you can generate the derivation and look at the builder <aswjrisp>I do not have notify-send on my system. When I do `guix locate notify-send` it tells me it can not find it. How else should I search for the package that provides that program? <aswjrisp>guix locate: error: file 'notify-send' not found in database '/var/cache/guix/locate/db.sqlite' <ekaitz>i don't know any other way to locate things but the program is packaged in `libnotify` <aswjrisp>ekaitz: thanks for telling me what package it is in. I was having trouble finding it with guix search as well. I guess you just need to know what package for this program. <ekaitz>yeah this is something we should improve <ekaitz>it has bother me a lot in the past, too <Rutherther>cow_2001: I was able to build the package you sent with guix from like two weeks ago, but I cannot with latest. Maybe a regression after the core-updates <cow_2001>looked into the guile-build-system and turns out the phase i add is added after install phase, which is deleted in the %standard-phases of guile which itself depends on gnu-build-system that has the install phase <Rutherther>oh right the ordering of the phases when using modify-phases was changed somehow by core-updates. I don't know the details though <futurile>wait what, the standard phases ordering changed? <ieure>I saw some recent patches around this. <futurile>bloody hell - that sounds absolutely mad <ieure>It was a core-updates thing. <ekaitz>Rutherther: arm-none-eabi-gcc: error: unrecognized command-line option '-rdynamic' ?? <ieure>Would be nice if changes like this were communicated better, though I understand there was so much stuff in core-updates when it merged that this may have been difficult to do. <Rutherther>ekaitz: what's the issue? Dynamic linking is not supported for the arm-none-eabi toolchain <ekaitz>Rutherther: how did you make pico sdk not to use it? <futurile>ieure: that corrects it, happen to know when/how - most importantly why - it was changed? <Rutherther>ekaits: I don't know pico-sdk, I just used the examples, I didn't configure anything myself. Just did cmake and make as you are suggesting <futurile>I have an entire blog post I'll have to redo then I guess <Rutherther>ekaitz: in pico-examples root I did "mkdir build && cd build && cmake .. && cd blink && make -j8", inside of a shell "guix shell arm-none-eabi-toolchain make python gcc-toolchain" <ekaitz>oh i'm missing gcc-toolchain probably <ekaitz>i was using the wrong compiler for some things <ekaitz>that's why it was failing to find those libc thingies <ekaitz>because I was building some tool using the cross compiler instead of the native one <ekaitz>Rutherther: it is building now, god bless you <Rutherther>great! I will try to make it into a patch soon, also with some bugfixes to the existing toolchains <ekaitz>this has been broken for long time <ekaitz>now i can make some new mice, trackballs and keyboards :) <Rutherther>cow_2001: install phase was not deleted recently. That was a long time ago. Maybe even when guile-build-system came about. What happened is that modify-inputs was changed. Before the thing was just added at the end when the phase was not matched, but now it errors <Rutherther>"(modify-phases %standard-phases (add-after 'non-existent-phase 'something (lambda (x) #f)))" this adds 'something as last phase before core-updates. But after core-updates it throws the error you've shown <Rutherther>so as to where to add it, maybe you can just append it <ekaitz>Rutherther: built, you made my day! <lynn_sh>hello, im defining a package and using some patches using search-patches. i have a patches folder in my 'packages' folder, with the patches. when running guix build on this particular package it says it cannot find the first patch in the list. is there any documentation on how to structure patches? <ieure>lynn_sh, I've never gotten the patch facility to work outside the Guix repo. It hardcodes the path it searches in as `gnu/patches` -- I don't have that in my channel, because it's a separate module. <ekaitz>lynn_sh: guix searches them in the main folder if you have a channel <aswjrisp>ekaitz: When you build your keyboards do you use qmk on guix? <Rutherther>it adds %load-path to the %search-path, and channels have only root folder in there <ekaitz>except some from a package i didn't do myself, that handles them by hand <ieure>It feels weird to commit patches to git to me anyway, I just use (substitute* ). <ekaitz>aswjrisp: i have done it before with qmk, yes <ekaitz>aswjrisp: i have nothing else than guix, yes <ekaitz>aswjrisp: do you have any question about it or...? <lynn_sh>ieure i get your point about substitute. the reason i am doing it this way is because the patches are easily accessible as git patches and i am lazy. :) <meaty>which package contains the 'gsettings' command? <jpoiret>meaty, Rutherther: glib's bin output more specifically, so "glib:out" on the command line <meaty>jpoiret: yeah I found that out lol :P <meaty>two things on my guix wish list: the abillity to search build servers for packages providing a binary (if not arbitraty file) and the ability to view an info/man/etc. database from the 'perspective' of a certain package/profile <Rutherther>what do you men with that man/info/etc.? like to tell man/info which package/profile to use specifically instead of having all used at once? <jpoiret>meaty: `guix shell a b c man-db -- man -k foo` <jpoiret>same for info but replace man-db with an info viewer <Noisytoot>Plasma seems to be broken on Guix with X11 (it just hangs on the loading screen with a mouse cursor). <galois`>What is my best option for monogame and c# on guix? <galois`>I wouldn't know how to package monogame as that is something I havn't done. <Noisytoot>Mono used to be packaged in Guix but was removed due to requiring unbootstrappable binaries <galois`>So I should use a VM for it? What would you recommend? <galois`>Emulation on my laptop is pretty slow though. <galois`>Wait this allows me to run commands specific to a version that are not packaged anymore? <Rutherther>it allows you to go to arbitrary commit of guix and get the packages from that time, so also supports this, yes <galois`>Ok, I will try and go that route. Thanks for the help! <Rutherther>also see the manual for Inferiors in case you wanted to get just one or a few earlier packages and keep the rest from newer Guix commit <podiki>if cuirass build fails with "Throw to key "lzlib-error' with args "(lz-decompress-read 5)'." does that mean it maybe got a corrupt download or something like that? <podiki>(i would just restart it but need a renewed cert) <Noisytoot>Why does plasma-desktop-service-type not work with X11? Can anyone else reproduce this? <ieure>Noisytoot, Probably because it's hard times for us X11 fans. <Noisytoot>and how can I make plasma reload .desktop files so it sees new ones when I install new packages? weirdly, even logging out and logging back in does not do this, but plasmashell --replace does (but that breaks the keybinding to open the application launcher) <galois`>I know this is a noob question. But how can I find out which last guix version had support for monogame? <Rutherther>hmm, I am afraid there aren't really Guix tools that would do this. So probably by grepping over git history. I am also not sure if there was a "monogame" support, probably just a mono package, and from there mono should take care of it <Rutherther>the file was gnu/packages/mono.scm, so "git log --full-history -- gnu/packages/mono.scm" gives the commits that edited it along with the latest that deleted it, in Sep 2022 <Rutherther>did you update it? it still has add-after 'install <cow_2001>turns out they didn't remove the install-documentation phase, so i just threw my script wrapping phase after that