IRC channel logs
2026-04-02.log
back to list of logs
<ieure>bjc, GCC stands for "Guix Compiling Compilers" <mra>do you know what version of gcc is ultimately built by the bootstrapping process? <bjc>i think it pulls you up to 14, maybe 15 now <bjc>huh. guile-git was updated to 0.11.0? but gitlab still shows 0.10 as the last release <bjc>ah, it was tagged last week <kestrelwx>Oh, well, they ended up using the installation image rather than the Nix service. <mra>watching the artemis launch while waiting for gcc to build :) <squid64>How would I go about using sddm as display manager? <squid64>If I use (services sddm-service-type) I get unbound variable error and if I add sddm to (use-service-modules) I get an error about xorg-server being provided more than once <ieure>squid64, You have to add SDDM *and* remove GDM. That's what the error is trying to tell you. <ieure>squid64, Look up `modify-services' in the manual, that's what you need to remove gdm-service-type. <squid64>The error says it's xorg-server being provided twice though or unbound variable if I don't put sddm in use-service-modules <squid64>I'll look that up though, the modify-service thing <ieure>squid64, I know. My answer is correct. <squid64>Alright. I wish the errors were more clear, because I don't think I would ever have guessed it. <ieure>It is a bad error message, yes. <ieure>The Guix display manager services are tightly coupled to Xorg, which means they provide 'xorg-server. So you cannot have two of them, because then you have two services doing that. <ieure>But it's also kind of common sense, almost nobody would want to run two graphical greeters at the same time. <squid64>I am new to guix so yeah I was just writing what made sense to me at first <squid64>Thanks ieure that worked, I can put sddm in use-service-modules now and it doesn't complain. <squid64>Which is programs not updating in the kde start menu <squid64>Anything new I install with guix install doesn't get added and anything I remove stays there as a blank icon <squid64>ok so i copy pasted the .desktop files from .guix-profile/share/applications to ~/.local/share/applications/ and they work when they're in the latter directory <PotentialUser-59>ps if someone has any idea pls reply even if i logged out cause once at home i nead the logs <bjc>PotentialUser-59: i didn't realize this was bcachefs, rather than bcache <bjc>bcachefs was removed from the kernel <bjc>do you need it for boot? <PotentialUser-59>in fact i dont think the issue is specific to bcachefs but rather loading dkms <bjc>are the modules in /run/current-system/kernel/lib/modules? <PotentialUser-59>in this case its bcachefs kut maybe just cause its the only dkms im trfna load <bjc>dkms should work as long as you don't need them in the initrd <bjc>the kernel-module-loader-service-type will use modprobe to load bcachefs <bjc>for that to work, the module has to be in a known place, which is /run/current-system/kernel/lib/modules/ <bjc>if it were me, at this point i'd start digging through code and poking at the filesystem to see if i could suss out what was happening, but that's not transferable =) <bjc>but i'm at a loss. maybe someone else can help <bjc>i used to, but not for years now, so i'm pretty rusty <bjc>you can try the guix-users mailing list <bjc>if you `guix build bcachefs` it'll output a path, which will have its build products, including its kernel module, in there <bjc>so you can just insmod it directly from there <YAR_Oracool>AH! New day new ... nothing... let's get emacso n this so I can actually use my sytem. <YAR_Oracool>but have to automate emacs doom install. Like check if this file exisets, if no, do this <YAR_Oracool>Nano is fine but it is awful to use with mr parantasys <YAR_Oracool>The first messege I saw from you was " rather make it load automatically than insmod..." <YAR_Oracool>hmmm let me see if I can find any thing with some search <YAR_Oracool>from what I see dkms is unsupported on guix but somee more experienced might know more <YAR_Oracool>I am most likly wrong. That's just what I could find <ieure>PotentialUser-59, Your configuration looks right to me. <PotentialUser-59>2026-04-01 21:03:28 localhost shepherd[1]: modprobe: FATAL: Module bcachefs not found in directory /run/booted-system/kernel/lib/modules/6.18.19-gnu <ieure>PotentialUser-59, Did you `sudo guix system reconfigure' after editing the configuration? <PotentialUser-59>yes did reconf. 2026-04-01 21:03:28 localhost shepherd[1]: modprobe: FATAL: Module bcachefs not found in directory /run/booted-system/kernel/lib/modules/6.18.19-gnu <ieure>PotentialUser-59, Try adding this to the operating-system record, reconfiguring, and rebooting: (packages (cons* bcachefs-linux-module %base-packages)) <ieure>PotentialUser-59, It sounds like it's trying to load it, but it isn't ending up in the system profile; that should do that. But this is just a guess, I've never done this and am kind of shocked at how difficult it seems to be. <ieure>PotentialUser-59, Just want to say that I admire your tenacity working through this. <PotentialUser-59>2026-04-01 21:03:28 localhost shepherd[1]: Service swap-euhbeuthbeuthb started. <PotentialUser-59>2026-04-01 21:03:28 localhost shepherd[1]: '/gnu/store/uethubethubth-modprobe -- bcachefs' exited with status 1; output follows: <PotentialUser-59>2026-04-01 21:03:28 localhost shepherd[1]: modprobe: FATAL: Module bcachefs not found in directory /run/booted-system/kernel/lib/modules/6.18.19-gnu <PotentialUser-59>2026-04-01 21:03:28 localhost shepherd[1]: Service kernel-module-loader failed to start. <PotentialUser-59>2026-04-01 21:03:28 localhost shepherd[1]: Service file-system-/home started. <ieure>PotentialUser-59, The location within the operating-system record does not matter. <ieure>YAR_Oracool, What do you mean? <PotentialUser-59>2026-04-01 21:03:28 localhost shepherd[1]: Service kernel-module-loader failed to start. ieure <ieure>PotentialUser-59, Is that before or after adding the package to your system profile? <ieure>PotentialUser-59, Okay, and did adding the package work? <ieure>YAR_Oracool, In what context? ~ was originally a shell expansion, but it became very popular and other programs also understand it. But not all, and not all automatically. ex. Emacs treats it literally unless you use `expand-file-name'. <PotentialUser-59>shall i place ur line after (kernel-loadable-modules (list bcachefs-linux-module)) <ieure>PotentialUser-59, As I said, the order of the fields within the operating-system does not matter. <ieure>PotentialUser-59, If that doesn't work, you should post to the help-guix mailing list. <YAR_Oracool>I thgt I was done with these syntax issuse... I was wrong <ieure>PotentialUser-51, You need to post to the help-guix list. <mra>finally finished building gcc-4.6.4. it took ~4 hours lmao <squid64>If I want to install something, like a sddm theme for example and they come from guix packages, do I need to specify the packages in my system packages or if I just type the theme name it should download and use it automatically? <czan>Looking at the manual, sddm defaults to checking /run/current-system/profile/share/sddm/themes for themes. I expect installing a theme as a system package will put its files in that directory. You might then also need to set the "theme" option of sddm-configuration. <ieure>squid64, There don't appear to be any SDDM themes packaged. But they'd need to be in the same profile as the package they're for, and SDDM is in the system profile. <squid64>I see. I guess I have to install them as system packages in my config then and after that set the theme in the sddm config. <YAR_Oracool>Any plans to add Proton VPN GUI? or am I on my own if I want it? <mra>YAR_Oracool: the CLI seems to have already been packaged. packaging the GUI yourself shouldn't be too hard using that as a template <adanska>is there a way for a shepherd service to trigger an action in another shepherd service? or any other sort of inter-procees(service?) communication between services? <adanska>I get the impression that this isn't really the way things were meant to be done, insofar as you should do this through some sort of socket system and come up with your own ipc protocol/listening situation. <adanska>... which is probably a bad idea to begin with. <adanska>what im trying to create is a sort of 'watchdog' service that can send a signal to the service managing an unresponsive process to restart <adanska>or... there is a `perform-service-action` procedure that might hold the key... this doesnt look like something a service itself should have access to, but might be worth looking into... <Ofthemasses>Hi there, wondering if anyone has encountered permission errors regarding /tmp/.X11-unix/? <Ofthemasses>It is unfortunately causing my user to not be able to enter the wayland environment unless I am root? <Ofthemasses>Default ownership is as follows: drwxr-xr-x 2 root root 4096 Apr 2 19:54 /tmp/.X11-unix/ <adanska>civodul, i have a q for you, since you're very familiar with shepherd. For performing communication with services, i've found `perform-service-action` in the docs. The question is this: would you be able to use such a procedure in a shepherd service definition to facilitate a sort of IPC with other services? or would this somehow conflict with the fibres concurrency model? sorry if this is a big question but i think there a few better people to ask than <los7432>I'm trying to understand the differences between .guix-home and .guix-profile. I found that the /etc/profile of .guix-home is sourced in .profile but I can't seem to find where the /etc/profile of .guix-profile is being sourced. Does anyone know? <los7432>*setup-environment specifically, not .profile <nox>> I'm trying to understand the differences between .guix-home and .guix-profile. I found that the /etc/profile of .guix-home is sourced in .profile but I can't seem to find where the /etc/profile of .guix-profile is being sourced. Does anyone know? <nox>.guix-home is what you build with `guix home reconfigure`, so if you build your .bashrc or your .zshrc with guix home, it will (obviously) try to source itself. <nox>.guix-profile is what you build when you run `guix install`, which is an "imperative" way of installing packages (as opposed to guix home, where you provide an scm file), so you're expected to source it yourself- since it won't manage configuration files or insert itself into your dotfiles <los7432>yes but I am not sourcing .guix-profile explicitly but the bin is in my path, so not sure where is it being sourced, that's what I want to find out <nox>oh, I think the /etc/profile (root) tries to source all of those <los7432>ah yes it indeed is. seems obvious now that you said it :) <nox>to be fair, I had no idea it even did it before checking <los7432>yeah since that folder is in the home folder, i was just checking the profile files in home <usernew>Hello everyone I sent an email asking for help to load dkms in guix (non boot partition). Here is the link, any help is very appreciated: <usernew>civodul: I hear that you know how to deal with Shepherd ;) <kestrelwx>This has nothing to do with Shepherd. I think the kernel module is called 'bcache', so you would put that for the loader. <usernew>kestrelwx: first I wanna make sure: I'm talking about bcachefs the file system, not bcache <usernew>I wanted to point out also that I think is Shepherd-related because of the error I get (see text of my email) <usernew>kestrelwx: np, yesterday someone else said the same exact thing <usernew>kestrelwx: but if you have any ideas... Hehe. <yelninei>usernew: looking at the bcachefs-linux-module pacakge it is just empty, so it cant load a module that does not exist. MIght be a problem with the package <usernew>yelninei: oh my. Really? There were some problems with the web interface of guix packages yesterday. Did you use the non web interface to figure that much? <kestrelwx>`ls -aR $(guix build bcachefs-linux-module)` <usernew>yelninei: does it mean that your recommendation is to contact eho builds tje packages (how?) and build from source myself (how) ? <yelninei>i did guix build bcachefs-linux-module and looked at the directoy and it just contains a license and an empty lib/modules <usernew>kestrelwx: do you confirm it's empty? <usernew>yelninei: many thanks for the breakthroigh <usernew>kestrelwx: how do I go abiut reporting this and fixing this on my laptop? <usernew>I dont have , I dont want, and I can't register, a codeberg account <usernew>ieure: bjc it seems we found the root of the problem we've been discussinf about yesterday <kestrelwx>Locally you could define 'my-bcachefs-linux-module' with a fixed build and use that in the configuration. <kestrelwx>For reporting and submitting a patch I think you do need a Codeberg account. <untrusem>yeah until forgejo implements activitypub and email intergration you need an account <usernew>kestrelwx: do you have any references/recepes to guide me to do that? <usernew>untrusem: then I wont report it; but if you have an account and are keen then you can report it if you wish <kestrelwx>You can see the source for the original package with `guix edit bcachefs-linux-module` and then the manual sections (guix) Defining Packages and (guix) Defining Package Variants on how to make your own variant. <usernew>Fantastic, let me try and I will message again if I get stuck. Once I have defined my package (variant), I suppose it will be built somewhere in the store and I just replace the name in the system config and reconfigure... <kestrelwx>You can have it either in a channel, a module somewhere in the load path or you can have it present in place in the configuration file and then refer to it by the variable name. <ieure>usernew, Yeah, I was poking at this last night and independently discovered that the package was totally broken. <ieure>Haven't had a lot of free time lately, so I didn't get very far debugging it. Updated to the current version, same issue. <usernew>kestrelwx: would there be a section in the manual for that as well? <usernew>ieure: did you try building from source? Is the result ampty or legit? yelninei <ieure>usernew, Yes, I built from source, and no, it didn't work. <usernew>ieure: what is there to debug about this? Is it because you suspect the package sources produce the empty package,m <ieure>usernew, Yes, the package only contains a COPYING file, no kernel module. <ieure>usernew, How do you do what? <usernew>How do I achieve loading the bcachefs kernel module? Did you build from the sources that guix procided or from those that bcachefs provides? <ieure>usernew, Right now, you cannot, the Guix packaging for that is completely broken. <usernew>ieure: can I ignore guix and build from sources thay bchachefs provides? <usernew>ieure: is there anyone in this channel who is in charge of these guix things, so w3 can try and fix this? <ieure>usernew, I genuinely don't know, but this kind of approach generally -- trying to subvert the Guix system -- at best, only works temporarily, because of the way Guix manages things. <usernew>identity: maybe the source code guix provides is rubbish and I can bypass by compiling the source code that bcachefs website provides <usernew>identity: is there someone in this channel which we can escalate this matter to? <ieure>I don't really know how kernel module stuff is supposed to work in Guix, but from what I was seeing, the `make install' for the bcachefs-tools (which is the source used by bcachefs-linux-module) doesn't actually build the kernel module, it wants to install the source for the module to build with DKMS. <ieure>I don't know if that's right or not, but it doesn't feel right. And even if it's supposed to install the sources, it doesn't do that either. <usernew>ieure: hold on bcachefs tools is completely different and I think has nth to do with the dkms <ieure>So I really don't know. It needs an issue filed and for someone who knows how any of this is supposed to work to fix it. <usernew>Bcachefs tools installs and works xorrectly <usernew>I think its a completely separate thing <ieure>usernew, Re-read what I wrote. <ieure>Instead of jumping to conclusions. <ieure>The same source is used to build multiple Guix packages. <usernew>Build-farm admins or maintainers? ieure <ieure>usernew, You should file an issue on Codeberg. <yarl>Something is bothering me, let me paste. <yarl>Is it doing what it's supposed to do? <apteryx>how do I change the status verbosity to see the build output directly at the REPL when using ,build something ? <apteryx>from the shell that be like 'guix shell --verbosity=10'; I tried ,build-options '(#:verbosity 10) but this does not do what I expected (shows internal daemon stuff, like getting a lock file, etc.) <ieure>apteryx, Actually, I think it's just build-options in the REPL, that calls set-build-options under the hood. <apteryx>oh, yarl is right, it must be ,verbosity directly <apteryx>,help guix mentions; ,verbosity LEVEL - Change build verbosity to LEVEL. <apteryx>maybe it should be noisy by default when using (guix), though that's just an opinion <apteryx>kind of annoying to have to open logs to see what's going on <ieure>yarl, Your paste is definitely weird, I don't know why that procedure wouldn't be bound. I see the same behavior as you. <ieure>yarl, If I C-c C-k reconfigure.scm, the procedure becomes visible in the REPL. I wonder if this is a Guile compiler bug? <noe>is it not just getting inlined? <ieure>noe, Don't you have to explicitly use `define-inlinable' for a procedure to get inlined? <noe>no, in Guile inlining happens all the time <ieure>Hmmm, I guess that forces it, but it may choose to inline in other cases. <bjc>the constant inlining is one of the things that makes the backtraces so difficult <ieure>The stack in yarl's paste shows channel-relations in the stacktrace, though. <bjc>i got a 404 on that paste. i assume it timed out <noe>I thought the paste shows bindings, not backtraces <ieure>It does look like some stuff was elided from the paste, maybe the .scm was loaded in between paragraphs. <noe>When they call “,re” that stands for ,reload right? <noe>if at this point the module is loaded directly without compiling then there would not be inlining <noe>in any case, I remember that you can use something like (set! channel-relations channel-relations) if you want to prevent inlining of that function <yarl>Well, I guess this kind of behavior should be documented in the manual. <yarl>(guile manual, of course) <noe>Is it not documented in @@ <noe>since it warns about declarative modules, which is about inlining <yarl>Well, there's no mention of @@ in "Declaratives Modules" <avigatori>does anyone here have guix forked on codeberg? I asked, because I am missing the keyring branch on mine. I specifically recall using the "fork everything" option, because I caused Error 500 issues on Codeberg <noe>yarl, yeah I suppose it could be made clearer <los7432>during the output of guix reconfigure, i sometimes e <los7432>see it downloads the same package same version multiple times <avigatori>ieure: thank you. I guess I have to crash codeberg once more <noe>avigatori, you can push that branch manually too <PotentialUser-18>I understand that it is a security measure, but I can't find the what, where or why. <PotentialUser-18>Oh, wait, I found the "where": there is a `.dirs-locals` I did not create. But I did create it. <ieure>PotentialUser-18, That's an Emacs question, not a Guix one. #emacs or look at "Per-Directory Local Variables" in the Emacs manual. <PotentialUser-18>I know. But perhaps this file was created by guix somehow. It seems like it is a negative there. <PotentialUser-18>Here is a Guix question: Battlezone 2100 is failing to update. I got the following log: <ieure>PotentialUser-18, What directory is the .dir-locals.el file in? <PotentialUser-18>It is in ~/.config/emacs. Looking at this file makes me believe it was created by some Doom Emacs experiment I made long ago, and since it is hidden file, I may have overlooked it. <PotentialUser-18>Since you know nothing of it, it gives me confidence that I can delete it. <ieure>PotentialUser-18, Guix doesn't randomly drop files into user $HOME. The only way that file could have gotten there is if you put it there by some means. This could involve Guix Home, but you'd have had to explicitly configured the file's contents, location, etc. <ieure>PotentialUser-18, Yes, looks like the warzone2100 (not sure why you called it "Battlezone 2100") package is broken. Please file an issue about it on Codeberg; seems like it uses some bundled libraries which are out of date and using deprecated APIs, which cause warnings, which are treated as compiler errors due to the options provided. <PotentialUser-18>So it _was_ the "treat warnings as errors" that was causing the problem! <kestrelwx>I think it's been broken for a long while, so you might be its first recent user. <PotentialUser-18>By the way, those messages about "patch-shebang: ./po/scripts/extractGuideJsonStrings.py: warning: no binary for interpreter `python3' found in $PATH" in the build log are important? <ieure>PotentialUser-18, Usually not. It looks like a script used during the dev process which isn't needed for the build. <dajole>I sometimes come across packages mentioned in the documentation that no longer exist. E.g. https://guix.gnu.org/manual/1.5.0/en/guix.html#Using-TeX-and-LaTeX mentions a comprehensive texlive package that doesn't seem to be there anymore. How do I 1) go about finding out when and why the package was removed and 2) contribute to the docs so they can get updated? <avigatori>is there a way to skip compiling the documentation when compiling guix? <vagrantc>sometimes i just wish to be able to pause updates to master so i can actually get an update in ... have kernel-updates essentially ready to push, but by the time i rebase and run the appropriate checks someone has already pushed <vagrantc>at least this is only 5 commits to sign, rather than the usual 7 or so <avigatori>finally! After 5 hours I managed to create a package variant ;_; \o/ <Rutherther>vagrantc: are you entering your pin for each commit or why does the number of commits matter? <Rutherther>vagrantc: also why does it take you so long to iterate? If you just rebase the files to compile should be only the ones touched by your branch, so it should be fairly quick unless you're on a branch that touches a lot of files <vagrantc>Rutherther: yes, pin is entered for every commit <Rutherther>I see you wouldn't like to be rebasing a team branch xD <vagrantc>Rutherther: i usualyl try to at least run guix pull <vagrantc>ugh. forgot to add my signed-off by ... oh well, minor problems <vagrantc>not sure if caching the pinn would be possible with the security token i use ... but i also am a bit dubious of caching the pin ... i like to know what i am signing <avigatori>just my luck, when I try to test it it's time to rebuild linux :< <vagrantc>at least for x86_64-linux substitutes should be available <vagrantc>and bordeaux usually picks up aarch64-linux after a bit ... <vagrantc>can't recall the last time i saw an aarch64-linux linux-libre substitute from CI <vagrantc>it's in theory there ... but rarely finishes <avigatori>ah, I forgot, maybe I ran into that IPv4 bug again <civodul>vagrantc: pulls.ci will always run ‘guix pull’ for you and complain when something breaks :-) <ieure>vagrantc, My YubiKey PIN gets cached. I forget if this is a token thing or a gpg-agent thing. Might be a token thing you can change. Unplugging the token wipes the cache. <vagrantc>civodul: that's the on that comments on pull requests? <vagrantc>if so, i think that also only does x86_64-linux? <civodul>i mentioned it cause you said you’d always run ‘guix pull’ by yourself <vagrantc>i mean, it is useful, although the comments often seem to com at pretty random times and it is not clear which commits it is commenting on <vagrantc>usually what i does is do a guix pull to a profile, reconfigure the system with all the various kernel generations, reboot into each kernel ... and then rebase over and over again, running guix git authenticate, check-chennel-news, etc. to check if the derivations changed ... and then ideally run guix pull to make sure something else didn't go sideways <vagrantc>admittedly ... i just pushed the last one without running guix pull locally ... as i need to get something else done today :) <vagrantc>but pretty sure updating electrum wouldn't have affected anything i was doing :) <vagrantc>i'm sure others are less ... obsessively fastidious about this than i am ... but whatever. :) <csantosb>civodul: The comment in `guile-3.0-latest` is still relevant ? <RavenJoad>My university is moving to a PKI-based login for eduroam (WiFi). Just to get things working in a hacky way, I need to put my university's root certificate in the system store (/etc/ssl). That's RO on Guix system right now. Is there another spot I should put it instead? <mra>third day of my bootstrap, things are coming along! just finished building gcc-14! <mra>what an adventure this has been lol <ekaitz>mra: in riscv just building mes takes 3days :) <mra>ekaitz: lol, probably faster to bootstrap a cross-compiler on x86 and then use that to bootstrap risc-v <ekaitz>surely but that's not the point of the bootstrapping <ekaitz>cross compiling is not always obvious <mra>i wonder who actually developed the bootstrap for risc-v. seems like a nightmare to test <ekaitz>mra: want to guess why I know why it takes 3 days just to build mes? <ekaitz>it's not totally ready, but I worked on it for 2 years <mra>i assumed it was just risc-v chips being slow <ekaitz>but so yeah, for your interest: it was mostly stikonas from live-bootstrap and me <ekaitz>but of course many more people took part and made hugely important contributions <stikonas>it's not too bad CPU there in terms of speed <stikonas>and mes is really heavy on memory bandwidth <ekaitz>oh yeah, and the cache is not huge either <mra>oh, interesting! i have to admit i don't really understand all that well how mes actually works. the bootstrap seeds seem kind of like magic to me