IRC channel logs
2025-07-29.log
back to list of logs
<podiki>anyone ever use modprobed to customize the kernel modules in guix? <nomike>luca, I figured out once and for all that: When I start an application in gnome (e.g. by Alt-F2 and typing in the command name) the process will be invoked by "gnome-shell". And "gnome-shell" does read ~/.zshenv. <nomike>So I can source the guix config in that file. The problem is: When I do this, my mouse cursor get's broken. It's no longer a an arrow but a white square. <nomike>So, I can source ~/.confiug/guix/current/etc/profile in ~/.zshenv, but if I source ~/.guix-home/profile/etc/profile in there, I'm getting that weird issue with the mouse pointer. <kkremitzki>This is strange--after doing more testing on what would be needed to backport the fix for the recent CVEs--the provided script reports 'closed' for 1.4 through 1.0, and in particular, after bisecting it seems an Apr 5 2023 commit is the first one it reports 'open' for, even though the commit doesn't look problematic at all to my untrained eye <kkremitzki>(checksum: bbc4a8d238ee2f73ae3e16065fe4be4fd0e639aa) <kkremitzki>I suspect I'm doing something wrong but basically that previous commit reports 'open' and the one right after it on `git log` from Apr 4 2023 reports 'closed' (6311493d7a6271bfbc51f4693857f9a12fe9965d) <apteryx>lilyp: I'm sorry I just pushed another world rebuilding change on gnome-team ^^', but it'll help debug things (meson-build-system using '--verbose') <apteryx>otherwise we don't see the compiler command line <apteryx>was trying to debug directfb, which I've superseded to directfb2; and it has some issues with validate-runpath I'd like understanding <jfred>Hm, after a recent `guix pull` on my (aarch64) laptop `glib-networking` is failing to build with some test failures, and it's a dependency of gtk =/ <jfred>that's on guix b585cc6b699006a03f7b3125484d8ad66810fddd <untrusem> when I try to build it i get this error <untrusem>guix system: error: failed to load 'systems/verito.scm': <untrusem>/home/untrusem/test/verito-2.0/systems/verito.scm:26:26: Wrong type to apply: #<syntax-transformer bootloader> <untrusem>I didn't change anything in the bootloader <untrusem>ahh should have pasted line by line 😅 <lilyp>apteryx: it would be nice if we could coordinate world rebuilds better <lilyp>also, I think we should order non-workaround patches according to their position in the rebuild chain, so that we don't get stuck rebuilding the same package over and over at different points in the commit history <lilyp>(this will become important for later rebasings) <apteryx>does it matter? Guix cares not about git commit ID <apteryx>perhaps the workarounds should appear at the bottom of the rebase pile? <apteryx>then at least it's easy to find them <apteryx>the only commit we're carrying on gnome-team at the moment which shouldn't be merged is the rust bootstrap shortcut <apteryx>is rust-team has their own, more complicated version of it <apteryx>(which accounts for the fact that on some systems the new shorter bootstrap doesn't work yet) <lilyp>Guix does not care about git commit ID, but it absolutely does matter when you're building things commit-by-commit for whichever reason <lilyp>workarounds at the bottom are fine I guess – what I do want to avoid are meaningless rebuilds on my already weak machine <lilyp>(don't forget that we got to build clang before rust for example) <lilyp>issues with validate-runpath are typically a result of missing linker flags – probably -Wl,rpath=some/lib/subdir is needed <ArneBab>(this is the one build failure that’s critical to my setup so it prevents me from updating) <ColdSideOfPillow>I noticed that there's an emacs package called `emacs-nov-el`, which a name unlike other packages <ColdSideOfPillow>The only other package to end in `-el` is `emacs-prodigy-el`, which is deprecated in favor of `emacs-prodigy` <ColdSideOfPillow>I just wanted to consult the people here before putting out a change that may be rejected. <human_equivalent>Hello. I've spent a few days setting up dovecot and opensmtpd and stuff, but I want opensmtpd to also check the database for users - not a file on the filesystem. This requires opensmtpd-extras, which doesn't build for me. Can somebody check if ~guix build opensmtpd-extras~ works for them? Any tips on how to troubleshoot it? Can't find anything about it in the email lists I've checked. <Deltafire>human_equivalent: doesn't build for me either, configure: error: *** compiler cannot create working executables, check config.log *** <Deltafire>the config.log was no help, just "configure: exit 1" <human_equivalent>I don't want to learn how to troubleshoot this right now, I just want to fix my mail thing, haha. <Deltafire>"conftest.c:46:1: note: include '<stdlib.h>' or provide a declaration of 'exit'" <Deltafire>i think that's the issue, exit() is not defined <Deltafire>added the missing include and it seems to build okay <ArneBab>human_equivalent, Deltafire: there are currently a number of build failures. At least two connected to gcc-14 changes. <Deltafire>i guess it could be patched to build, but the archive message on its upstream repo makes me wonder if it should actually be removed <apteryx>wasn't there an issue open for the wrap-program taking basename out of argv[0] ? <human_equivalent>If it's unmaintained I think it should probably be removed.. for my part I'll probably just set postgresql to export to files on changes to the tables, I think it can do that. <human_equivalent>I'll also add checking out opensmtpd/table-postgres and see if I can package that some time in the future, but that's probably ways off <Rutherther>also note that the easiest fix is just to add gcc-11 to native-inputs <jakef>hi Guix, in a package definition, is it possible to use a phase from a different build system? i'd like to copy the sanity-check phase from pyproject build system and run it on the "python" output of a package <identity>jakef: you can just grab the function from (guix build pyproject-build-system) or wherever it is <jakef>identity: sure, but not copying code is nicer <ArneBab>If you’re going to remove all unmaintained libraries, I expect that plenty of programs will fall apart. The typical number of maintainers of FLOSS is less than one. Because code that works usually doesn’t need much maintenance. <ArneBab>(at least once it’s finished and does exactly what it’s intended to do) <ArneBab>Sovereign Tech Fund isn’t yet in position to provide funding to enough people to adopt all those old packages. <ArneBab>(usually distributions are the de-facto maintainers of such packages, so as long as there are up-to-date patches in debian or Gentoo the packages aren’t actually unmaintained) <ArneBab>jakef: there are 15 packages with a python output. Could a central method for that be useful? <jakef>ArneBab: yeah maybe! it's generally not ideal having a python output though, because of how python packages often have propagated-inputs, and you don't want to propagate them to non-python outputs <nikolar>ArneBab: the difference is that it's usually local pacehs <mrvdb>hi guix. my guix pull fails, apparently because of a missing valgrind for diffutils /check/ phase (see: https://dpaste.org/xvDTa ) I have valgrind installed, what am I missing? (guix runs on top of arch power kernel on ppc64le machine). <efraim>diffutils needs valgrind on ppc64le? <yelninei>mrvdb: The test is being skipped because valgrind is not found, but this should not cause the failure. Could you grep the log for ^FAIL <mrvdb>yelminet: FAIL test-float-h (exit status: 134) is probably the culprit then? <efraim>I know test-float-h fails, then there's more too after that one <mrvdb>efraim: XFAIL large-subopt (exit status: 1) you mean? <efraim>not sure if there's more after that one <human_equivalent>ArneBab: yeah you're probably right, I don't want to pretend that I have much of a say in what get's removed or not - maybe in a few years if I get more involved with guix :) <human_equivalent>I'm having some issues with rspamd too - I've added pkg-config and gcc-14 to native-inputs and that makes it go further but it doesn't go all the way. ld returns exit status 1 and I get stuff like undefined reference to `std::condition_variable::wait(std::unique_lock<std::mutex>&)@GLIBCXX_3.4.30' <gabber>how can i hint meson-build-system towards a specific input? i am trying to fix #1678 by updating gtk+ input for gimp but now (after updating to 3.24.45) the gimp build fails to find the dependency <gabber>> Run-time dependency gtk+-3.0 found: NO (tried pkgconfig) <sham1>Oh lord, Rust packages pull so many deps <identity>generally rust packages do not pull in many deps, it is the deps that pull in so many deps… <Deltafire>rust/cargo isn't alone, i looked at packaging a nodejs application is that's a similar nightmare <ieure>NPM is patient zero for the dependency explosion. <identity>program A pulls in library B, which pulls in library C, and all of a sudden you run out of keys on your keyboard for metavariables <sham1>I'm just trying to package a singular rust program, but it feels like the ecosystem is a tarpit where one struggles getting out of <luca>have you seen the new rust packaging thingamajig? <sham1>Can I use that already or is that just in some branch like rust-team <luca>I think it's still in rust-team. Coming soon :tm: <Rutherther>It is still not merged. You can still use it though <sham1>Well sure, I can use it through various means. Welp, time to get myself a new channel file so I can experiment with this stuff on a guix shell <sham1>Hm. I'm slightly confused now, but I'll investigate <sham1>And as one might expect, that worked out spectacularly. Certainly beats out writing out (package) forms by hand for each dependency <efraim>oh wait, do we still need a utf8proc without julia? <efraim>not yet, only on rust-team so far <Kabouik>Trying to use tree-sitter in Emacs, I installed `emacs-tree-sitter` and `emacs-tree-sitter-langs` and when I (require 'tree-sitter), I get a prompt to run cargo to install something. I'm guessing this is not the Guix way. Am I missing a package? <ieure>Kabouik, Possible you have a non-Guix tree-sitter package kicking around somewhere? <Kabouik>They might be installed from `guix-emacs` channel instead of `guix`, maybe the definitions on the Guix channel are different. I'll try requesting the versions on the Guix channel specifically. <Kabouik>(Well no, there are no versions of these two packages on the Guix channel) <identity>Kabouik: what is ‘emacs-tree-sitter’? iirc, it is no longer needed since emacs got built-in treesit <identity>just installing ‘emacs’ and ‘tree-sitter-LANGUAGE’ for every LANGUAGE you use should be enough <identity>and the major modes if they do not come with emacs, obviously <BakedKraken>Hey, I want to have a script run after guix system reconfigure. What is the best way to do that? This is crude hack territory, I don't want it to be elegant. <BakedKraken>I do not want to have all the bells and whistles needed for a repo. I just want to adhoc fix something after bootloader deploy so I can actually work on it and make it a package. <Rutherther>Reconfigure is for running scripts, this is no crude territory, you just make a service and extend the activation service type. This will be ran on system activation - reconfigure, boot, switch generation etc. <Rutherther>Buf if you want to modify the bootloader you should rather modify the bootloader install script <Rutherther>Calling the old one and then modifying what you need to modify <BakedKraken>I want to detach the active bootloader from the gnu store. After every group deploy, just make a copy of the active linux and initramfs and grub config things into the ESP instead of linking to the (encrypted) gnu store <BakedKraken>> Reconfigure is for running scripts, this is no crude territory, you just make a service and extend the activation service type. This will be ran on system activation - reconfigure, boot, switch generation etc. -- This is an absolutely opaque thing to me <untrusem>does anybody uses niri window manager in guix? <untrusem>if yes, is there some snippets i can look at to actually make it work <BakedKraken>Rutherther: So you essentially already did what I want to do and have it build into your config? <Rutherther>@irc_libera_bakedkraken:ditigal.xyz: yes, I did exactly what you want some time ago <lynnn>it feels like guix home is putting the files into a dotfiles environment and then just not symlinking it to my user. <lynnn>yup just added "qutebrowser" to my home configure, reconfigured, and it does not show up in my bin at all. in the past this would work. i recently recovered my system, any tips on how to figure out why my home broke? <ieure>What do you mean, "putting the files into a dotfiles environment?" <lynnn>sorry that was word salad. it is putting it into a GUIX environment. <lynnn>i just checked, my ~/.guix-home is a different canonical link than my guix home describe is. <lynnn>im not sure how that affects the dotfile service, since that seems to create a separate canonical link than the home one (looking at some files that still exist) <ieure>Does the profile under "canonical file name" have qutebrowser? <ieure>Are the permissions on $HOME/.guix-home correct? Maybe your UID changed when you recovered the system and it can't change the link and has bad error handling around that? <ieure>I think symlinks always have 777 permissions, but do have user/group ownership. <lynnn>if i LS inside of ~/.guix-home it is all owned by root <ieure>That's normal, but not what I was asking. <lynnn>i just id -u'd myself im still 1000 <lynnn>i ran a mimimal config that just tried to symlink and it just rebuilds font cache and that's it. <lynnn>i wish i knew how guix home worked more, it feels like it is completely disconnected now. <lynnn>not sure how else to fix this than reinstalling at the moment. <ieure>lynnn, I'd rm / move .guix-home away, `guix home reconfigure' and see if the link is correct now. <ieure>Possible Stow is overwriting ~/.guix-home? <lynnn>i removed ~/.guix-home, ran my guix home reconfigure, there is now no ~/.guix-home <ieure>Make a symlink for it that points to the canonical location from `guix home describe'? <simendsjo>lynn: can you paste the full output from reconfigure? It's also possible to increase the verbosity, but try without first. <lynnn>oh this time it is giving output ill paste it <simendsjo>What's the exit code? `echo $?` after running the command. Should have more output, but maybe it crashed? <simendsjo>Looks like 130 is SIGINT, maybe you aborted it by mistake? Try running with verbose output to see if it logs more. <lynnn>the doc says it requires a verbosity level but doesn't give examples of what levels exist <ieure>Add as many nines as you like. Just to be sure. You never know. <ieure>Not a Guix thing, I do that with any software that has a verbosity option. <simendsjo>Try adding the verbosity before the reconfigure. The documentation says `guix home [OPTION ...] ACTION [ARG ...] [FILE]`, so maybe the argument after file is ignored. <simendsjo>You can try running under strace and see what it does before it shuts down. <lynnn>i can run the strace but i am not technically adept enough to read it. <lynnn>home worked flawlessly when i had a fresh install, is there no way to just force it to treat it like a new install? i would like to work using my laptop tomorrow without having to reinstall guix.. <noe>maybe delete the symlink and some environment variables <noe>(might break everything idk) <ekaitz>oh lynn i saw your toot, did you break guix? :) <simendsjo>lynnn: Try pasting the entire strace even though it's large. <lynnn>rebooted and now the system has nothing, so i can't easily paste anything <lynnn>guix pull does not properly update the guix describe, so i'm missing my channels which provide my software. <lynnn>i did that and it worked but then never updated. that also doesn't fix the issue of config files <lynnn>ok well you tell my system that :) i have no channels now despite my channels.scm being in config <namewillforget>Hey everyone! I tried to get bluetooth running. So I added a declaration `(service bluetooth-service-type)` to by system config and installed `bluez`. Now `herd status` shows that the service is disabled and stopped. If I try to start it with herd, it immediately stops again and `herd status bluetooth` shows five messages `- 2025-07-29 21:20:52 Process exited with code 1.`. Is this a bug or should I have configured something I w <lynnn>i guess i will reinstall and lose things. laptop is not usable and no one seems to know how to fix it <identity>lynnn: have you tried deleting the symlink and let guix home recreate it? <lynnn>i just don't have that anymore, + not having a proper guix describe <simendsjo>lynnn: What about `guix home import` to create a new home configuration? <lynnn>i can't use my guix home at the moment because it requires these channels i can not add <sham1>Do you not even have time-machine? <lynnn>you'll have to give me more information on that <lynnn>i used the grub menu to recover after a bad system config and everything has slowly broken since. <sham1>Well, if you can use `guix time-machine`, then you can specify channels with the `-C` <sham1>Well that might be a bit more involved than just getting your channels unscrewed <lynnn>channels unscrewed is a new problem though after guix pulling to see if i was just out of date <lynnn>i wish i heard about guix home import before being told to read strace logs <lynnn>i'm backing up what i can that matters <dariqq>lynn: For your guix channels is your guix pointing to ~/.config/guix/current/bin or are you using the /run/current-system/profile one. I lost my guix once due to disk corruption <lynnn>i am very evangelical about guix but just having everything fall apart with no one knowing why really makes me hesitate to reinstall <lynnn>would love to know what is happening but unfortunately i need my system to work <simendsjo>You're just having bad luck with the people online I think. There are some real experts here, but you got to reach them at the right time. I usually send a mail to the mailing list, but I understand it's difficult to wait without a working system. <lynnn>i am just lucky i have a smartphone so i can connect to irc at the moment <simendsjo>.. again, strace will show you exactly what channels.scm it tries to load (and fails to find) and might give some clue on what to source or similar. <lynnn>and how do i find that in the mess of strace <simendsjo>strace have an -o to print to a file, and you can paste that file online. <dariqq>i guess that explains why your channels are missing. If you source the .config/guix/current/etc/profile (and maybe `hash guix`) you should get your regular guix back. Why this is not happening automatocally is something to investigate further <sham1>I think I have a slightly easier issue to ponder now. For whatever reason the `man` command on my system seems to be just ignoring my environment's $MANPATH when doing `man -k` which affects things such as the Emacs man command <ekaitz>I'm again with the guix locale issue, why is this again here... aaaaaah <sham1>Like, I have sway installed on my home profile, and ~/.guix-home/profile/share/man has the appropriate man-page files in there, and the path is in $MANPATH, yet `man -k sway` just gives me the "nothing appropriate" message. Like if I do `man sway`, it works just fine, but I'd like for the apropos to also work <simendsjo>sham1: Are you in a custom environment? You have to include mandb for it to be built. <sham1>Surely `man` should just follow the $MANPATH envvar despite everything else <lard> https://github.com/rndusr/stig/ Any chance someone wants to package this (or tell me how to get it running locally)? Looks like we have most of the dependencies packages already, and it's python. <ekaitz>lard: import it with the python importer then <lynnn>annnnd it immediately broke again after trying to do guix home import (the profile linking) <lard>I will try that, thanks. <simendsjo>sham1: but maybe `man foo` will look up it directly, while `man -k foo` will search a created database? So you need to build the database (using mandb)? I'm guessing a bit based on earlier problems. <simendsjo>sham1: Try running mandb to see if it fixes the issue. <namewillforget>Also, could someone take a quick look at my bluetooth issue and in particular tell me whether I should mail it to the bugs mailing list? <sham1>I ran mandb, and it did nothing. It's just complaining that MANPATH is set and so it's ignoring the config file in the store, and then 0 man subdirectories contained newer manual pages, 0 man pages added, "0 stray cats added", 0 old database entries removed <sham1>I'll try to strace mandb and see what it does <sham1>…It's looking at /usr/man, /usr/share/man, /usr/local/man etc. What even <ekaitz>sham1: that's probably hardcoded in `man` <sham1>Well, mandb does also seem to read the store config file unconditionally <sham1>I got a bit closer by running `mandb -u`, but it went into `cannot create catdir ${HOME}/.guix-home/profile/share/man/cat1` and such. So now it's at least looking under my MANPATH. Progress! <simendsjo>sham1: reconfiguring system and home should create the man database as far as I know, so it's strange that it haven't. I usually get into this problem when running `guix shell foo bar` without adding `man-db` too. <sham1>Do I perchance need to explicitly include man-db as a package in my Guix home profile? Because that seems a bit silly <sham1>Oh, it did it. That's just… wow <dariqq>lynnn: The .guix-home symlink is created by the profile activation script in HOME_ENVIRONMENT/activate. If that file is corrupted then it would explain why nothing is getting updated <lynnn>dariqq HOME_ENVIRONMENT isn't set for me <dariqq>yeah i mean .guix-home/activate or the /var/guix/profiles/per-user/lynnn/guix-home-NNN-link/activate <lynnn> /gnu/store/icfql1wd9ml62x489l8cshziz21slaar-activate: empty <simendsjo>sham1: Hmm, why is that necessary? I haven't used this. But I see %default-profile-hooks includes manual-database/optional which only runs when man-db is included. I just cannot see that I have man-db, but it might be a dependency of some other package I'm using. Even the comment says ";; Building the man database (for "man -k") is expensive and rarely used. Build it only if the profile also cont <lynnn>i need to sleep and then i will unfortunately fresh install tomorrow <sham1>simendsjo: I would think that it would be done by default for guix home at least <sham1>Like, if I was running a foreign distro instead, I would still want the man-pages for my Guix Home packages even if I didn't include man-db from Guix, since that would seem redundant. And even though I'm on Guix System, it just seems strange. <sham1>It syncs info-pages just fine without me listing info explicitly <simendsjo>sham1: I cannot see it as part of the base packages. But as the comment says, this is by design and you are an odd user ;) <dariqq>lynnn: I think it is fixable depending on what and how many files got damaged. But depending on your situation reinstalling might be easier <sham1>*shrug* that's probably true <sham1>So, for future reference. If someone asks about apropos not working, this might be why <simendsjo>dariqq: lynnn: So I didn't loose any files other than old generations <dariqq>simendsjo: It depends. lynnn has empty files in their store that should not be empty. If think this is the shepherd bug not unmounting / cleanly (cant find the bug id right now) but it could also be a bad drive or whatever <namewillforget>Oh, my bluetooth issue resolved itself by rebooting. It's running now. Let's hope it stays that way. <dariqq>It is #77086 which is still open, so I think it is still not fixed? <ieure>dariqq, I reported that, it is fixed. <ieure>The title is not actually what was happening, though the cause was proximate. <simendsjo>namewillforget: Did you just switch after reconfigure without a reboot? The service uses a dbus-root-service-type which uses an activation-service-type. So maybe the root dbus service had to be restarted for it to work? I always reboot after reconfiguring the system to avoid such issues. <dariqq>then there is #76959 with empty scheme files which points to the other one <dariqq>there is also #73249 that says it is only happening after a system reconfigure don t know if this is fixed? <ieure>dariqq, As mentioned, I reported the original issue, and it is fixed. <ieure>dariqq, The issue I was having was that e2fsck was buggy and not reliably clearing the dirty filesystem bit, which would cause it to fsck on reboot. <namewillforget>simendsjo: Yeah, I tried to use it without rebooting. I guess that's the reason? I can't really say but it sounds like it makes sense. I'm still in the initial configuration phase of my Guix usage so I have to reconfigure often, but yeah, afterwards it's probably a good idea to do a reboot after a reconfiguration. <simendsjo>namewillforget: I see. I very rarely touch my system configuration these days, mostly just my home configuration. And many changes are safe, but it's easy to get intro troubles like that on system reconfigure. <dariqq>ieure: does an fsck cause data loss/empty files if / gets unmounted cleanly? <ieure>dariqq, No. I never experienced data corruption. <dariqq>mhh system reconfigure seems to get also blamed in this thread. When I lost some data a couple of months age I noticed that when I disabled the upgrade-shepherd-services part of "system recpnfigure" the "recovering journal" after a reconfigure did not happen but i never looked further because it only happened once <dariqq>anyway I got to go now, but I'll look at this again