IRC channel logs
2023-12-12.log
back to list of logs
<cow_2001>i am writing some guile-srfi-n package definitions. i see that the srfi authors had testing in mind. is there a way in the guile build system to optionally run the tests? <Kolev>cow_2001, are you a Guile programmer or something? <cow_2001>i don't see how one would tell guix in the command line to run the package's tests, but i see that the check stage is involved with it <Kolev>cow_2001, sadly, IDK anything about packaging Guile libraries. <cow_2001>it's installable, but... uh... what about tests? <Kolev>I did package one program, but due to historical mishaps, it's impossible to make the program free software. <lfam>Making a new signing key for myself. I seem to have `guix git authenticate` working, but `make authenticate` complains that the new key is missing, although it is in the local 'keyring' branch <lfam>lechner-: Hi! I saw that. I tried `GUIX_GIT_KEYRING=keyring make authenticate` but it fails similarly <lfam>I must be holding it wrong, since these two commands are the same thing under the hood <Guest40>I installed guix in a VM and wanted to add spice-vdagent to my config.scm but documentation is unclear to me <Guest40>where do i put 'spice-vdagent-service [#:spice-vdagent]' in the config? <cnx>I believe you should use (service spice-vdagent-service-type) now <PotentialUser-5>is guix pull working fine for everyone, when you have nonguix channel <Kolev>PotentialUser-5, I have non-standard channels added and can pull fine. <Guest40>cnx i edited my config.scm( https://paste.debian.net/1300842/) and used what you proposed(service spice-vdagent-service-type). system reconfiguration threw no error. Rebooted. copy-paste Fedora-host<->guix-guest works perfectly. <Guest40>cnx As i am new to guix and it is somewhat confusing a Question: can you explain the rational of yours for referring to (service spice-vdagent-service-type)? Did you check somewhere? <cnx>what error did you get, Guest40? <Guest40>cnx with what you proposed i got no errors. Everything worked and system got reconfigured. Prior to that "error: space-vdagent-service: unbound variable" and "invalid field specifiers" ...but then again...i am not sure yet when hwat to put where <Guest40>cnx i see you pointed to "manual/devel/en". I got my from "manual/en" ...is "devel" the documentation for what is refered to 1.4.0 ? <cnx>sorry, the manual/en is for the stable release <Guest40>@cnx i used as installer.iso "7s3yr6b7sf2hbii3lmjf5ch7m8p55b2q-image.iso" and not "guix-system-install-1.4.0.x86_64-linux.iso". Would my results been differetn with standard installer? <RavenJoad>I have a shepherd service which depends on another being started first (Emacs depends on dbus). I have added the symbol for dbus to the requirements list for the emacs shepherd-service. Should I use-module dbus module into Emacs? More generally, how should use-modules be used in cases like this? <mirai>RavenJoad: if you're just adding 'dbus to requirements, no <mirai>it doesn't even have to “exist” <RavenJoad>If the dbus symbol never exists in the final system (even though I know it never will), and the 'dbus symbol is in Emacs' requirement list, what hapens? <mirai>RavenJoad: the service doesn't get launched <adanska>Hi guys! whats up with ci failing evaluations? i'm looking at the build logs but honestly its all gobbeldeygook to me lol <efraim>adanska: Ludo seems to have found the cause was related to avr not being a triple, I'm looking through to fix some more of the other cases <adanska>efraim: ah, ok! thanks for letting me know :) <adanska>id love to hear what the root cause is once you find it <bdju>the amount of things I have to hold back to not upgrade gajim is increasing and I still haven't heard if it's fixed <efraim>"add avr platform" added avr with a target of "avr", compared with "x86_64-linux-gnu". There are a couple of places that parse the target based on the hyphens <adanska>efraim: ahh. how are you solving it? changing the platform name to a triple and then adding cases for the avr architecture? <efraim>for make-crust-package I'm removing the option of running the host tools from avr with a note to revisit if necessary. still looking for others <adanska>bdju: whats wrong with the upgrading process? <bdju>I saw on the mailing list a while ago that latest gajim is catastrophically broken and it's something I use daily so I've held off doing a full upgrade <bdju>but gajim's dependencies are shared with a million things so skipping it during upgrades is quite messy <bdju>I had a --do-not-upgrade list of like 20+ things now <efraim>you could try it out with 'guix shell gajim -- gajim' <bdju>ah that's an interesting idea actually <bdju>does that grab the latest one even if I haven't upgraded yet? assuming I've done a guix pull <efraim>it grabs whatever you would've gotten from 'guix install gajim' <adanska>bdju: seems to be working fine for me! although im not using any plugins so take that with a grain of salt <adanska>i connected to the guix disroot XMPP channel fine <adanska>you can test it out with `guix shell gajim gajim-omemo -- gajim`, for me it detects and enables just fine <bdju>weird. okay my current one and the guix shell one are both 1.7.3. also somehow omemo seemed to work in the shell one without me adding gajim-omemo to the command. is the shell using the gajim-omemo already on my system? <bdju>hm. maybe I was worried for nothing. I don't know why it wouldn't work for that other user then <bdju>thanks adanska and efraim for the guidance with guix shell. I think I'll try upgrading now since it seemed to work okay <adanska>shell sorta just puts a profile with the packages you specify on top of your user profile unless you use --pure or --container without --link-profile <bdju>so in that command what is the part after the -- doing? I notice when adding gajim-omemo it went on the left side and the right side stayed the same for the command <bdju>is it just a name for the shell? <adanska>the `--` executes that command in the `guix shell` right after its done configuring, and then exits the shell after its done. for guix commands, like `guix shell foo -- guix system reconfigure bar.scm` you can omit the `guix` in the second part to do `guix shell foo -- system reconfigure bar.scm` <adanska>so what `guix shell gajim -- gajim` is doing is making a profile with gajim in it, then in that environment executing the `gajim` command. when you exit gajim (via C-c), the shell will also exit and return you to your base user profile without gajim present <adanska>or in your case the older gajim version <adanska>also bdju, with guix you shouldn't be too afraid to upgrade packages in that situation since you have profile rollback to help you if something goes awry <adanska>and the power of guix shell to test things out is so invaluable <bdju>oh I think I get it. neat. <bdju>I am aware of rollbacks but I wanted most of my system to be upgraded still <bdju>sadly I don't think there's any shorthand option for "skip everything you have to in order to hold these packages back" <bdju>nor a way to skip some upgrades while applying a manifest <adanska>no. in guix, packages are tracked on the one guix repo (also other channels), and profiles are tied to a specific commit from that repo (see `guix describe`). to get around that issue you can use `guix time-machine --commit=foo -- install bar` to install a specific version of something to your profile, but yeah everytime you do a `guix upgrade` it will try and bring it up to date. <adanska>the easy way to get around this would be to use something like `guix home` to manage your own packages and keep the default guix profile purely for that odd package and avoid running `guix upgrade`. this seems pretty fragile and restrictive though. <adanska>this would be a more sustainable option in the long term, and would let you continue using the imperative `guix package` commands <adanska>vaguely i remember someone discussing a better way to deal with this in the IRC... but maybe that was a dream or something :p <adanska>also please if i'm mistaken here someone let me know haha <adanska>nothing much, i've just been procrastinating for the last hour, gonna finally knuckle down and do some work haha <efraim>looks like the latest derivation got closer, now it's pointing out avr-libc is missing a static output <civodul>BTW, you can try “make cuirass-jobs” locally <efraim>certainly faster than waiting for CI <civodul>those serialized backtraces are wonderful, aren’t they? :-) <efraim>avr isn't quite to the level of the "Schrödinger's cat" release name <efraim>As i remember the story it caused ripple effects across many websites, some with the space, some with the umlat o, some with the apostrophe. <efraim>it turns out file@5.45 also fails its test suite on i586-gnu on core-updates <efraim>master is still on 5.44 I thought <efraim>I never actually bothered testing on armhf, I should see if that one also is broken <efraim>I also didn't see anything obvious from Debian, so I think its something we've done <gabber>even though the Webs claim that Flameshot supported wayland (i'm using sway) from version 12 (Guix offers 12.1) it complains: `qt.qpa.plugin: Could not find the Qt platform plugin "wayland" in ""' is there a quick+dirty hack or a long-term solution applicable? <civodul>NixOS spends 9,000 USD per month on S3 <efraim>now we're getting into "fake static output" territory <civodul>efraim: i was thinking in (gnu ci) we could have ‘packages-to-cross-build’ return the empty list for AVR <civodul>because it doesn’t make sense to cross-build Coreutils or Guix for AVR <efraim>... I certainly wouldn't be against it <civodul>let’s do that and ask Jean-Pierre whether there are relevant packages to add to the list <efraim>sounds good. I'll run make cuirass-jobs locally to make sure it works <Franciman>is there a guide to contribute to packages? I'd like to update kakoune to the latest release <Altadil>Franciman: yes, you can take a look at part 22 of the Guix manual. <mirai>check with issues.guix.gnu.org first <mirai>perhaps someone might already have sent something for kakoune <rekado>and yet someone calls it “heavily subsidized”, so I guess it could be even higher. <jonsger>ACTION dreams about what can be rented for 9.000 USD a month at Hetzner :) <gabber>i am running into a "/gnu/store/<hash>-grub-2.06/bin/grub-mkimage: error: cannot open `/gnu/store/<hash>-grub-2.06/lib/grub/i386-pc/moddep.lst': No such file or directory." when i invoke a $(guix system vm) for a definition that i could swear worked before. <gabber>the path of course doesn't exist (i try to emulate an aarch64 system) <devmsv_web>what can I do when system shepherd does not respond to commands (sudo herd status or sudo herd restart /service/) ? I can't even reboot the remote machine... <gabber>hako: it is possible when everything is set up correctly. can you confirm you defined your modules the right way? can you load the files and (use-modules) with them as well? <gabber>i'm not sure about this exact error message, but setting your own channel up is somewhat tedious (when you don't have a ton of experience with it) <devmsv_web>thanks hako that would allow me to reboot. before that is there something I can do to recover shepherd? <gabber>devmsv_web: shepherd has pid 1 - killing it means killing your machine.. if it is not responsive than there may be an issue that really needs a reboot <gabber>can i use (with-parameters ..) from within build phases in a package definition? <hako>Ahhhh, I know now, it segfaults... "Dec 12 22:07:50 localhost vmunix: [ 157.759025] guix[2149]: segfault at 7fffe0164778 ip 00007ffff7f4638c sp 00007ffffffdc6f0 error 7 in libguile-3.0.so.1.6.0[7ffff7eac000+bb000] likely on CPU 0 (core 0, socket 0)" <hako>actually invoking `guix pull` segfaults too, I was using `/run/current-system/profile/bin/guix pull` when pasting the log. <hako>Adding pcspkr and snd_pcsp to modprobe.blacklist? <hako>in kernel-arguments of operating-system definition. <Franciman>ah there is guix home's inputrc service, where i can disable the bell <hako>TIL it can be disabled this way 😄 <janneke>wait whut, a world rebuild on master? <hako>try deleting gnu/packages/bootstrap.go? :P <janneke>ACTION was afraid building a hurd image was broken (again) <janneke>and now i can continue to check that <ekaitz>hi, we are providing autoconf 2.64 but the automake we provide is not compatible with it? <gabber>i am able to ,lower and then ,build the (with-parameters ) package in question. what am i missing? <peterpolidoro>can a local guix.scm package file be used to create and modify files in the local directory? or does it only create and modify files in the store? <gabber>peterpolidoro: you can use any .scm file to do whatever you want - but $(guix build) or $(guix shell) commands tend to simply add entries in you /gnu/store <peterpolidoro>I would like to use the guix.scm package to create local documentation files based on skeleton documentation packages stored remotely on a channel. <peterpolidoro>I know regular scheme can manipulate local files, do I just mix those scheme commands in with guix code in the local package? <peterpolidoro>the local package would have data specific to that project, it would combine that local data with the skeleton package and then generate documentation for that project in the local repository <gabber>i wouldn't recommend to just mix and match, but you can - guix is (just) GNU Guile with some modules loaded ;) <anthk_>I guess Ac97 support, VESA and that's it? <peterpolidoro>I guess I could copy files out of the store after they are generated <gabber>peterpolidoro: this sounds like a reasonably easy path to go. use copy-build-system (or whatever you need) and then copy the relevant files after the process. <nikolar>how do i get a location of package's files using guix command <ieure>nikolar, `guix build PACKAGE-NAME' will print its store path. <zamfofex>anthk_: Maybe ask in #hurd! 😄 Last I saw, they were planning to revamp the boot system. <PotentialUser-43>hello people, I am a nixos user, I just installed guix on a virtual machine to give it a try, and I have a few questions, i tried the `guix import` command and it's very very cool, you can import rust crates from crates.io with `guix import crate <crate_name>`, but can I import a crate that is *not* on crates.io? If it's on github for example <dissoc>what is the best way to start docker container on boot? <gabber>dissoc: have a look at docker-service-type <apteryx>hm, why does attempting to build hello on master attempts to build the world <apteryx>maybe because I had hacked things to use the latest Guile, which would change derivations? <apteryx>like so: lrwxrwxrwx 1 maxim users 32 8 sept. 23:40 guile -> /home/maxim/src/guile/meta/guile <civodul>apteryx: try “rm gnu/packages/bootstrap.go && make” <civodul>a record ABI change that the machinery does not detect <apteryx>it'd be nice if we have a 'guix build --explain' or something that traces the changes that triggers the builds <apteryx>civodul: that resolved my problem, thank you! <cmiller>I ran "git pull", "guix shell -CWP (from root dir)" and make, but it errors out because of contributing.de.texi has references to nonexistant nodes. <lechner->ekaitz / hi, what issues are you seeing, please? <janneke>okay, apart from quite some rebuilds, hurd-vm still builds on master \o/ <piethesailor>Hello all! Anyone try to install guix on a newer System76 machine? <rekado>sphinx still has three failing tests on python-team. This package has a big impact. IIRC this broke when I pushed the pygments upgrade to the branch. <rekado>does anyone plan to upgrade sphinx on that branch? <ekaitz>oh someone answered but I didn't have that automake <ekaitz>do you guys have automake 1.9.6? <ekaitz>janneke: so then we have an automake that doesn't work well with the autoconf we have <janneke>yes, we have (probably more than) one version of autoconf that doesn't work with the only automake <ekaitz>and automake says it needs 2.69 or newer <ekaitz>3 versions that don't work together <janneke>that doesn't make your statemen true <janneke>the default autoconf and automake work together <ekaitz>yeah yeah, I didn't explain myself well <janneke>apparently, no-one really needed an older autoconf _and_ automake <ekaitz>aren't they normally used together? <janneke>probably, in many (or even most) cases, they are <janneke>lilypond only uses autoconf, for example <ekaitz>i'll spend a really good time tomorrow packaging this all... <reedm>Is there a guixy way to set up pipewire? (i.e. some sort of home service). Or, do I just add the package to my home config, and make sure "dbus-run-session pipewire" gets added to the right init script? <Franciman>reedm: you also have to make sure pulseaudio doesn't get started before pipewire, iiuc <Franciman>btw there is a pending patch for adding a pipewire-service in guix home <reedm>Franciman: thanks! I'll keep an eye out for that. If it's going to be added pretty soon, I'll just wait <podiki>rekado: regarding sphinx...is that one where newer versions need maybe npm for javascript in themes? i was looking at something sphinx related some time ago and hit some needed basic themes which wanted node for javascript parts :(