IRC channel logs
2026-03-19.log
back to list of logs
<ekaitz>but if it fails to build... how do you plan to have it installed? <bdunahu>can you install an older version to your profile with guix time-machine? <ieure>dajole, Either use an inferior or use guix time-machine to make a profile containing the working version of the package. <ieure>dajole, What package is broken? <dajole>The current version is working, but a new version with guix pull and then reconfigure fails. It's a package not in the main guix channel. I'll look into inferior and time-machine, thanks! <redacted>I'm working on packaging a program that has both an Emacs package and a Vim plugin. Should I package the program, Emacs package, and Vim plugin separately, even though they're from the same repo? <redacted>Emacs and vim have their own build systems, so I'm unsure whether I should be mashing them together. <identity>redacted: you can take some phases from other build systems and stuff them in the list of phases so you can build all of that in one go <redacted>Yeah, I'm kind of frankensteining them together. <redacted>Ah, looks like the vim build system just copies the files and builds helptags, but there's no helptags to build anyway. That simplifies things. <benjaminwil>ieure: just checked my battery after a guix pull and system reconfigure ~6 hours ago. battery still depleting rapidly during suspend for me. :'( <benjaminwil>too bad, but i am half looking forward to learning how to debug this in a guix context. :) <benjaminwil>identity: yes! first thing i did after system reconfigure! i have thankfully learned that lesson enough times before. lol. <ieure>benjaminwil, I don't even really know how to debug it in a normie Linux distro context. <benjaminwil>i think i have already confirmed that all of the power-related gsettings are all good and normal, so i'm hoping that maybe with list-generations i can see if there is obviously a package missing or something like that? i dunno either, though. <apteryx>hm, is it me or the new SRFI 64 implementation in Guile 3.0.11 broke our test-driver.scm; namely, --select='something' no longer works <dajole>I'm still trying to debug the issue that I am asked every time for pin entry when using my yubikey to sign git commits. I've made a little bit of progress, which is that for whatever reason guix home doesn't start gpg-agent. guix home reconfigure succeeds just fine. When something else then requests the agent it starts. I still don't have a good understanding of what exactly is going wrong. Any ideas where I could look next? <dajole>Seems like gpg never gets linked into the sheperd init...hm... <Rutherther>dajole: guix home is not responsible for starting gpg-agent. Gpg itself is responsible for that <dajole>You may very well be right, though. Honestly at this point I'm at a loss as to what's going on. I don't understand why it keeps re-asking for the pin. <dajole>In a way it works, in that it signs my git commits correctly, and when I remove it it doesn't. But it keeps asking for the pin every singly commit. <Rutherther>dajole: how does your ~/.gnupg/gpg.conf look like? <dajole>The result is the same whether using git commit in the terminal or magit in Emacs. <Rutherther>Because I think the pinentry you set up is only terminal one. But I cannot check now <dajole>I was wondering if Emacs was interfering in some way, too, but it re-asks every time in the terminal as well. <dajole>This is the yubikey's pin, I think. But in my previous distro it was somehow cached all the same. I suppose I could remove the pin from the yubikey, but that seems counter to what it's supposed to do. <dajole>All that being said, I'm definitely in the somewhat-understand-but-still-rather-muddleheadded phase of guix usage. Perhaps I'm missing something else that's obvious and assumed. <Rutherther>are you sure you did not enable forcesig on the yubikey itself? You should be able to check out in gpg --card-status <dajole>Hm... gpg --card-status says Signature PIN ....: not forced It also seems like that should have been the same behaviour in my previous distro then, as I didn't make any changes to the key. <Rutherther>Yes. Have you tried other operations such as encryption with gpg directly to see if it asks you for pin for all of them <dajole>I just tried and it asks for the pin all the same. <dajole>As in I tried something like `gpg --verbose --encrypt --armor --recipient "me@me.com" --output ./STRICT_TEST.asc guix-config/home-configuration.scm` <abbe__>with a gexp which uses '(ice-9 textual-ports)' I seem to get: In procedure dlsym: Error resolving "scm_init_custom_ports": "/gnu/store/nbn6j7qbxk2mh22bszvpy2ypgazmj12q-guile-3.0.9/lib/libguile-3.0.so.1: undefined symbol: scm_init_custom_ports" <civodul>abbe__: see “warning: importing modules … from host Guile”: what happens here is you’re trying to embed modules from the Guile you’re using into the derivation, and to build them, potentially with a different Guile version <civodul>you probably want (use-modules (ice-9 textual-ports) …) in your gexp, but definitely not in ‘with-imported-modules’ <abbe__>thanks. I needed to use source-module-closure, likely was missing another dependency :) <cbaines>I spent ages figuring out that I was getting hangs/failures from a stale ./guile, and now I'm somehow still getting workspace dependent behaviour :( <cbaines>yay, I've finally got ./guile --version to be 3.0.11 everywhere <cbaines>unfortunately this seems to break the containers tests, but maybe that means the changes I was looking at don't break the tests <snamellit>I need to switch the sleep mode of a PCIe device (the disk in my lil' macbook). How can I best do that at boot time? A simple shepherd service and write the change config in the #:start instructions? <charlesroelli>Howdy all, I have a Guix VM whose "herd status" hangs indefinitely, even after reboot. I got a Shepherd REPL open and can produce the hang with ,use(shepherd logger) ,use(shepherd service) (logger-recent-messages (service-logger (lookup-service 'root))) ... so the code interacting with the root service logger is hanging. Any ideas where to go from here? <efraim>it sounds like one of your services isn't working correctly so I'd enable and disable any services that you added or changed until you can narrow it down <efraim>that is, in the os-config and then relaunch the vm with the modified config until you know which service it is <snamellit>efraim: it is `echo 0 > /sys/bus/pci/devices/0000\:01\:00.0/d3cold_allowed`. I do not think this is mapped to sysctl <efraim>otherwise I'd make a one-shot service to write 0 to /sys/bus/... <charlesroelli>efraim: thanks, I did manage to run "herd status <service>" for each service besides the root service, and that did not hang. I'm not sure if I can stop any of them, given the hang related to the root service logger, but I will check. <efraim>snamellit: `sysctl --all` will show all the options <snamellit>efraim: nope, I'll just make a quick shepherd service and be done with it. <gabber>i ./bootstrap, ./configure and make my guix sources (fresh pull) and keep getting ";;; WARNING: loading compiled file /home/gabriel/src/guix/gnu/packages/FOO.go failed:" for multiple packages. is this new or was i just really good in ignoring those? <efraim>it's new from changing to guile-3.0.11 <gabber>so i guess these are bound to stay? <identity>you might need to do a make clean (not clean-go) or something <efraim>I ended up running `git clean -dfx` instead of just `make clean-go` <gabber>-dfx does not work (no target x) <andreas-e>So you currently build guix with guile-3.0.11 in this way? <efraim>All I can say is I have it memorized as my magic incantation and it seemed to work for me <kestrelwx>Is this a problem on master? guix/ui.scm:920:18: error: go-github-com-pkg-sftp: unbound variable <efraim>no, it looks like I'm still using guile-3.0.9, from checking config.status <andreas-e>efraim: Thanks! I moved back to it when I noticed problems, but thought they may have been solved. <andreas-e>Does anybody remember which service to restart on berlin when CI times out? <andreas-e>I would like to file a removal request to test a notification option ;) <kestrelwx>Nvm, I think that's just a compilation issue. <andreas-e>civodul: Timeouts on the cuirass web server. I have restarted cuirass-web, but that was not enough. <civodul>andreas-e: that would be ‘sudo herd restart cuirass-web’ <charlesroelli>abbe: ah yes, I saw that issue too. strace showed shepherd receives the request and then nothing happens, just like what civodul mentioned. Except the epoll_wait calls in this case were with a zero timeout argument. <civodul>it’s not new, we’ve been struggling for a while <andreas-e>But I think I have the answer to my question already. <andreas-e>There was a suggestion to use the RSS feed to get notified of deprecation issues. But labels do not seem to be covered. <andreas-e>I think we might need a custom web hook that reacts to label changes. <andreas-e>I assume that when one opens an issue with a label, this is one action and not two. <PotentialUser-18>Could someone please help me understand what I did wrong and how to address it? <efraim>downloading all the gcc versions to try using different ones is going slowly <czan>Hi PotentialUser-18! It looks like python-uv-build doesn't exist in Guix? <PotentialUser-18>czan There is a package for uv. And that uv call did not came from me or the setup of the project. Hence my confusion. <czan>Where did the (native-inputs (list python-uv-build)) part come from? Because that seems to be referencing a package that doesn't exist. <czan>Generally you need to use --recursive to make sure you pull in all the dependencies. I'm trying to build a manifest where I've done that to check it works, but it might take my machine a little while. <czan>On the import: guix import pypi ass --recursive <czan>But the generated python-uv-build doesn't build, because it needs some rust setup. I'll try for a few minutes to see if I can get something working. <PotentialUser-18>czan Oh. That option seems to be undocumented. But it seems to produce something, yes. <czan>It's documented in the manual: (guix) Invoking guix import, under the pypi importer. Also in "guix import pypi --help". <PotentialUser-18>czan I thank you for your time and effort. I wonder why this more complex route is taken. I recall that `uv` was added recently to Guix and now seems to be the default for python package imports. <czan>This wasn't added from Guix, that's from the package metadata that came from pypi. Guix doesn't currently support building using uv_build in its pyproject build system, so the automatically imported package fails to build. I'm trying to see if I can get it building with setuptools instead. <PotentialUser-18>czan Yes, I see the option in the manual. Those options seem to be importer-specific, and the importers are not documented in the man or --help output. Still learning things. <czan>I don't know what you mean by "the importers are not documented". The manual lists them in "(guix) Invoking guix import" (https://guix.gnu.org/manual/devel/en/html_node/Invoking-guix-import.html), and "guix import --help" lists the importers, with --help working for each of them to give importer-specific help. If there's some way we can improve things, I'm certainly open to helping to make that happen. <civodul>PotentialUser-18: the man page is mostly empty actually; documentation is available via the “info guix” command or online <PotentialUser-18>czan Beautiful! I see that you added an "argument" to override the "build-backend". I wish to be able to do that one day. Thank you very much for your help. <czan>Yeah, I found that by searching the Guix repo for uv-build and seeing that python-snakesay did that with a comment about not having python-uv-build yet. Someone else solved the problem, I just copied their work. :) <PotentialUser-18>czan I am still used to use `man` and not `info` . I understand that in Guix the `info` source is the preferred option, but I keep forgetting the many keys it uses. I need a crib sheet. <PotentialUser-18>The `--help` output lists the importers, which is good if you know what you are looking for (I do not know half of those sources), but the 'ARGS' induced me to think of "name of packages" but not of "further options". <PotentialUser-18>The Guix manual explains everything well. It does demand a lot from the reader. Often times it has what I need, I just don't know what I need or where to look for it. :-) <PotentialUser-18>czan Still a solution out of my reach at the moment. I hope to learn a lot from you capable people. <PotentialUser-18>Ah, I finally understood the source of some of my confusion: the repository I saw on Github used setup tools (is has not been updated for many years). The one tracked by Pypi is updated and uses uv. <kratacoa>PotentialUser-18: `info --vi-keys` helps :) <PotentialUser-18>kratacoa Oh, I know about the Vi keys. ;-) (But I'm learning Emacs, so they are a bit forgotten.)Minor things like you search with C-s (like in Emacs), but you continue searching with '{', unlike Emacs and Vi. <PotentialUser-18>You have previous and next for top sections (perhaps also a next entering subsections -- at least there is one for "follow first link"), and previous and next for history navigation, like in web browsers. <PotentialUser-18>Info want to be both a book and a website. It makes sens, I understand it, but the interface is not intuitive (and I don't mean that it needs a GUI). <PotentialUser-18>I still need to investigate better the Emacs interface for info. When I tried it I was in a bus and Emacs was beeping like crazy and I decided to stop and return to the `info` command! <allana>I am bit baffled. As of today this works: "guix build -m manifest.scm --system=aarch64-linux", but this does not "guix pack -f docker -m manifest.scm --system=aarch64-linux". When using guix pack it seems to think the target is amd64. I wasn't facing this problem until this week. <allana>To provide some context I am running on an amd64 system with the qemu-binfmt service, and I think things were working normally ealier in the week. I will probably come up with a reproducible example for an issue, just sharing here first. <jcabieces>Is it worth opening an issue of am I doing something wrong ? <redacted>jcabieces, does that command work without creating a container with guix shell? <redacted>Looks like it's trying to access /tmp and can't. <jcabieces>redacted csantosb ok, let me try without -CPW (I just followed bluntly the documentation Building From Git on this) <muaddibb>Greetings! How do I remove a home config completely? My plan is to delete `~/.guix-home' log off then in again. It's a simple configuration, just some packages. <gabber>csantosb: is BUILD_NAJA_PYTHON a configure flag? how do i set it to "on"? <kestrelwx>Should I hope that pull from `file://` will finish? <kestrelwx>I guess I could just do my stuff from the checkout in the first place. <gabber>muaddibb: i guess you could just move it? and if all goes well remove it? <muaddibb>gabber: I don't know, it's a symlink to a store object, Guix probably detects it <gabber>if you just move .guix-home and fail to log in you'll probably see an error message <snamellit>muaddibb: you have to remove the stuff activating the home profile in .profile too to avoid errors. <gabber>just make sure there's another (root?) login that still works <muaddibb>Ok. Thanks snamellit and gabber! I'll do this later and report my results. I was just wondering if someone went through the same thing. <snamellit>muaddibb: I went several times back and forth circling the drain which is the perfect GUIX config. <me`>How is sway-configuration->file supposed to be used with greetd-wlgreet-sway-session? Nothing I include seems to have any effect, even I am able to sucessfuly run a reconfigure. For example, providing a list of packages does not install them, unlike when using sway-configuration with home-sway-service-type. <jcabieces>redacted csantosb doesn't change anything with just guix shell. Trying ith --share=/tmp <redacted>if it's the same error, I don't think --share=/tmp will work <futurile>jcabieces: are you using Guix hosted on another distribution, say for example Ubuntu? <futurile>jcabieces: Apparmor is the one I run into with Ubuntu; and/or /tmp not being allowed to do something <futurile>jcabieces: there's a newer policy for AppArmor that was done for the 1.5.0 release <futurile>jcabieces: if it's something odd about tmps permissions you can try to do --share=/tmp=/home/whatever/tmp <jcabieces>futurile strange things is that I don't have right to "ls /tmp/guix-build-qgis-3.44.8.drv-0.pivot" although I'm the owner of the file and the permission are 755 ... <jcabieces>futurile I'm gonna try with your proposal and dig in this apparmor thing <futurile>jcabieces: there were some issues on Codeberg recently from one of the committers who's on Debian <jcabieces>futurile ok, I'll take a look, thanks for the hint <bdunahu>hey futurile, did you see my response to you a couple weeks ago about a reproducibility issue you found: https://codeberg.org/guix/guix/pulls/4542 and do you think a patch is worth applying for the affected rust crates (glib-macros<v0.21.4)? <futurile>bdunahu: oh sorry, I didn't, looking ... <futurile>bdunahu: so is it fixed in later versions right? But, I guess we're carrying this for a while <ente>I installed a package and I want to list the files it contains <ente>(this is impossible to search online) <ieure>ente, find $(guix build the-package-name) <ieure>untrusem, I saw, planning to PR tonight. <untrusem>i was occupied with my talks for minidebianconf, I can do it tomorrom <tomenzgg>With Cuirass, is there any means for forcing a build of a jobset? I set Cuirass up with my channel but the first attempt failed because I hadn't included the base Guix channel. I've since amended (and the Cuirass UI reflects it) but, even after a single push to my channel and 2 hours later, no additional build has been attempted since. <tomenzgg>I can't seem to find any button in the UI for it and both Restart All Builds and Retry the Evaluation both seem to do nothing (with no reaction whatsoever in the Cuirass logs, that I can tell). <ieure>yarl, No. The home team is responsible for the `guix home' code, not for the services used by that code. <ieure>I will try to review it tonight, but no guarantees. <attila_lendvai_>csantosb, if i start llama-server it returns an empty page and "srv log_server_r: done request: HEAD /cors-proxy 127.0.0.1 404" in the console. is that expected? or should that work? <attila_lendvai_>csantosb, i tried to add "-DLLAMA_BUILD_SERVER=ON", but that didn't help <redacted>me`: looks like sway-configuration->file just generates the config file, it doesn't create a profile with packages in it. <redacted>The home service uses sway-configuration->file to generate the config file *and additionally* creates a user profile with the packages you specify. <redacted>So basically, your config is correct. But that procedure is expected to give you *just* the config file, even though it accepts the full sway-configuration record. <me`>redacted: yeah, looking closer and comparing 2 sway-configs from /gnu/store, one includes the default gestures while the other omits them entirely like I configured. It's just that sway-configuration->file doesn't generate a .config/sway/config like the home service sway-config does, so it didn't seem like anything was happening. From my limited usage with the home service sway-configuration, it seems like it has additional default <me`>behavior, so it's probably better for me to just keep it simpl follow the wlgreet example of including the config with local-file. <Guest8>hi is there away to add display functions to the repl printer, for example can i define a display function for a srfi-9 record type that prints only part of the record in repl when a variable of that type is entered? <redacted>me`: .config/sway/config is probably a symlink to the config created in the store by the sway home service <me`>redacted: yeah, it is a symlink to the store sway-config. I thought it might have been generated by wlgreet, but sway-greetd as a sway-configuration record in guile looks to be what wlgreet generates. Seems like you're not supposed to use both services at the same time. <me`>redacted: I mean, it seems lik you're not supposed to use sway-configuration->file with wlgreet and the home service sway configuration at the same time. <redacted>Isn't wlgreet for using a Wayland compositor as your greeter before you login? <redacted>And the home service is for *after* you're logged in as a user? <redacted>I'd expect them to be completely separate. <me`>redacted: I just misunderstood how the wlgreet service was used to start sway and thought specifying the config in it was part of doing so regardless of approach. <SquircleSpace>Hey folks! I am trying to write a script that would really benefit from importing (guix profiles) and (guix channels), and I'd like this script to be run by a shepherd service. I'm having issues making that work well since those modules depend on (guix config) and transitively depend on a significant portion of Guix. Is there a good pattern for writing scheme scripts that import guix?