IRC channel logs

2025-07-01.log

back to list of logs

<cancername_>Hi #guix, gdm fails to start my session after a reconfigure. Here's a log: https://0x0.st/8UFZ.txt . I'm not sure what to make of it.
<cancername_>It would be awesome if someone could help me figure out what's going wrong her.
<cancername_>e.
<cancername_>I did check the web, but all past issues of the sort seem unrelated.
<cancername_>okay, something else is definitely wrong. guix describe shows a guix that is definitely not the one I had prior to rebooting
<cancername_>when listing PATH, ~/.guix-home/... is present. shouldn't ~/.guix-profile be as well?
<cancername_>after setting my PATH such that guix points to ~/.config/guix/current/bin/guix, I got a "proper" guix instead of /run/current-system/profile/bin/guix. why did I have to do that? isn't guix home supposed to set PATH appropriately?
<cancername_>it's worth noting that I use fish, but even running bash -i results in a broken PATH.
<civodul>Hello Guix!
<cbaines>I'm going to turn off qa.guix.gnu.org for a bit as I still can't work out why data.qa.guix.gnu.org isn't working
<civodul>“isn’t working” in what sense?
<cbaines>civodul, I can't figure it out yet, maybe the fibers thread(s) are stuck in some kind of loop, or maybe the GC is just running over and over again
<bdju>Is there anything packaged for easily sending a magic packet to wake up a machine on the LAN? I've read about etherwake and wakeonlan but I don't see them packaged.
<bdju> https://heavydeck.net/blog/wake-on-lan-with-netcat/ I also tried this script but it complained about -b being an invalid option for nc.
<noe>bdju, try package wol
<noe>the wol package*
<bdju>Ah, nice. Thanks.
<jlicht>hey guix
<jakef>lilyp: I'm revisiting that matplotlib issue #70687. do you know if there's bug report upstream?
<peanuts>"python-matplotlib not respecting env var MPLBACKEND=TkAgg" https://issues.guix.gnu.org/70687
<wingo>guix. i upgraded recently and now my compose key is not working. gnome, wayland. what could that be, i wonder?
<wingo>it would seem that pressing the right alt, which is marked as my compose key, seems to toggle the keyboard into some other mode
<wingo>i want to compose = e -> euro symbol
<wingo>hmm, it seems it was a weird interaction between dead keys and compose
<wingo>switching to a layout without dead keys seems to have done the job
<vntsuyo>can I make a root service depend/require a home service?
<noe>vntsuyo, I don’t think so since they run in two different instances of shepherd
<vntsuyo>noe: i see, i'll find a workaround then
<civodul>cbaines: BTW did you deploy the Data Service with support for Forgejo webhooks? (could that explain the problems?)
<dsgdsg>I have a fresh install of Guix System but I can't get into Gnome from gdm. Every time it returns to the gdm login screen. Any tips?
<untrusem>dsgdsg: can you share you system config in a pastebin in bpa.st
<dsgdsg>untrusem: My config is here, with some parts elided: https://bpa.st/XW3Q
<spiderbit>So I wanted yesterday to create a modify-package to fix a bug in stumpwm: https://codeberg.org/guix/guix/issues/913 but with a bit help maybe alternatively could somebody help me to write the package new from the base and to use that because it's a window manager so I can't just install it as user and have it as option in gdm? So I need some
<spiderbit>packaging help mostly How do I add my package (transformation) to the system packages? Can somebody hold my hand a bit?
<spiderbit>So a channel is just a git folder? Is that right or do I even need git technically?
<podiki>in short define your package somewhere (can just be in the system config file) and use it there; it can be a (package... inherit)
<spiderbit>ohh so I don't even need a own "channel".
<podiki>nope, here is an example of adding a patch to the kernel to use, for instance https://github.com/podiki/dot.me/blob/master/guix/.config/guix/config.scm#L50
<podiki>(ignore nonfree please)
<podiki>you can also use the transformations machinery, for instance try something like "guix shell package <transformation> --export-manifest" and you can use that in your config too
<spiderbit>so basically I do that already: (define stumpwm-with-mylibs
<spiderbit>  (package
<spiderbit>   (inherit stumpwm)
<podiki> https://github.com/podiki/dot.me/commit/ebdd31f3fff9abf744aac7a61e3f2e6f2f1d2d9e is an example with transformation instead
<spiderbit>so I must just copy paste it as most simple the build phase as example and then modify it.
<podiki>if you have your "stumpwm-with-mylibs" in your system config or somewhere it can see it (like a channel), yeah just use that in your system packages rather than plain stumpwm
<spiderbit>but I don't need to copy everything from the package then.. right I mean a have modify-inputs as example it doesn't replace the inputs just adds mine or merge them...
<spiderbit>so with arguments / phases it would be the same?
<spiderbit>Well I guess I have that: (package-inputs stumpwm)
<spiderbit>so therefor the normal things are there...
<spiderbit>is there something similar for arguments? or do I not need it like package-arguments stempwm
<spiderbit> https://codeberg.org/guix/guix/src/branch/master/gnu/packages/wm.scm#L2774
<spiderbit>most of the code in argumetns is about desktop-file do I need to copy all that...
<spiderbit>(package-arguments stumpwm) maybe I need to add that...
<podiki> https://guix.gnu.org/manual/devel/en/html_node/Build-Utilities.html see modify-phases
<podiki>or better yet, look at examples in the guix code
<podiki>git grep #$phases gnu/packages is your friend
<podiki>for modifying phases of an inherited package
<spiderbit>would this be the right phase to change / add: (add-after 'unpack
<spiderbit>does build-program even care about environment variables?
<physikerwelt>Hi, I'm new to guix and try to build a basex package. To to that I need to have openjdk-jdk and maven. The builld work when I use the normal debian versions of but I could not find a suitable combination of guix versions. (see https://github.com/physikerwelt/guix-package-basex/blob/main/README.md)
<spiderbit>so I need 2 build phases added one after unpack to capture the home path and then after to after it for some reason forgot it...
<spiderbit>I guess that is the key line:
<spiderbit>(uiop:register-image-restore-hook 'build-restore-hook/set-sbcl-home)
<spiderbit>    Call the hook functions registered to be run when restoring a dumped image...
<spiderbit>shouldn't it be backed into the image or is guix backing some sort of environment around the image
<podiki>sorry i don't know the details here, but remember the build environment is isolated; you can indeed set/unset environment variables in phases (plenty of examples of that) and yes some programs do capture this somehow in their final output, just depends
<podiki>e.g. git grep setenv gnu/packages
<spiderbit>guix system: error: failed to load '/etc/config.scm':
<spiderbit>ice-9/boot-9.scm:2254:24: #. read expansion found and read-eval? is #f.
<spiderbit>what a great non-saynig error...
<civodul>uh
<civodul>spiderbit: does the sequence #. appear in /etc/config.scm?
<spiderbit>yes
<spiderbit>as it should :D
<civodul>ah well, i guess that’s the culprit, but the error message is indeed quite obscure
<spiderbit>   (arguments
<spiderbit>    `(#:phases
<spiderbit>      #~(modify-phases %standard-phases
<civodul>(please use a pastebin like https://paste.debian.net)
<spiderbit>for 1 liners?
<civodul>i thought you were going to copy more
<civodul>there’s no #. in there :-)
<spiderbit>ohh the point...
<civodul>you can paste the entire config if that’s fine with you
<spiderbit>#-#.(hu.dwim.asdf
<spiderbit>well that is in the code...
<civodul>you found it :-)
<spiderbit>it's just copied from here: should be valid clisp: https://hub.darcs.net/hu.dwim/hu.dwim.environment/browse/bin/build-development-image.sh#213
<civodul>spiderbit: apparently that bit is being parsed as Scheme, hence the error
<spiderbit>ohh I forgot... why can't not everything be done in emacs-lisp :D
<civodul>heh
<civodul>in other news (unrelated), i posted a forgejo-runner service based on what dthompson had done before 👉 https://codeberg.org/guix/guix/pulls/956
<civodul>should be pretty fun!
<dthompson>civodul: yesssss!
<civodul>dthompson: and thanks for doing the heavy lifting!
<spiderbit>so now I need to start a clisp interpreter in the building thing...
<civodul>dthompson: i added a bunch of options that look commonly useful and changed ‘register’ from 4 to 2 arguments
<spiderbit>build-program requires a list of Common Lisp expressions to be passed as the #:entry-program argument.
<spiderbit>so it does accept clisp code...
<dthompson>civodul: looks good! I had only provided an interface to the options that I needed to tweak
<spiderbit>the question is what the entry-program is supposed to do...
<attila_lendvai>sneek, later tell spiderbit you shouldn't need/use any of the hu.dwim.* stuff when porting over the logic to the guix builder
<sneek>Okay.
<spiderbit>what emacs setup do you use normal guile mode... because auto identation is in my view horrific each time it uses like 20 chars long for each level further to the right.
<sneek>spiderbit, you have 1 message!
<sneek>spiderbit, attila_lendvai says: you shouldn't need/use any of the hu.dwim.* stuff when porting over the logic to the guix builder
<attila_lendvai>spiderbit, i've removed the backwards compatibility from that code, but apparently i can't push to hub.darcs.net for some reason (that site is pretty much in a zombie state). https://dwim.hu/opt/darcs/hu.dwim.environment/bin/build-development-image.sh
<spiderbit>ok
<spiderbit>I just fight to get even unmodified a version to compile with given arguments...
<spiderbit>ok at last something compiled ;D
<spiderbit>ok adding even a sbcl home before "build-program" brakes the building
<spiderbit>          (setf (getenv "SBCL_HOME") ("/tmp"))
<spiderbit>so it compiles without a path set but doesn't if I set it to /tmp
<attila_lendvai>spiderbit, try using uiop:getenv instead of getenv
<spiderbit>does scheme know uiop too?
<attila_lendvai>spiderbit, that must run in the CL runtime
<spiderbit>and how can I do that? over #:entry-program?
<spiderbit>couldn't I hardcode it into the entry program somehow?
<attila_lendvai>spiderbit, also note that the SBCL_HOME env var is pretty much an obsolete stuff. the "modern" accessor is to use (sb-int:sbcl-homedir-pathname), and to (setf sb-sys::*sbcl-homedir-pathname* sbcl-home)
<spiderbit>that helps me still not much... I don't know where and how I can execute clisp in the scheme config
<spiderbit>I guess I could set it in the entry program I just need to catch the right path than somethere else to make it work right...
<attila_lendvai>spiderbit, you probably can add something along these lines to the entry-program: (setf sb-sys::*sbcl-homedir-pathname* #$(system* "sbcl-used-for-building" "--eval" "(sb-int:sbcl-homedir-pathname)"))
<attila_lendvai>spiderbit, make that (princ (sb-int:sbcl-homedir-pathname))
<spiderbit>but I never set the "sbcl used for building" so... this is way to completaced for me either code it for me or I need to do it step by step in the simplest way.
<attila_lendvai>spiderbit, just get hold of the sbcl that is in the package inputs
<attila_lendvai>i.e. something like (search-input-file inputs "/bin/sbcl")
<spiderbit>I rather have now a partial solution that works than the perfect solution that does not compile and should theoretically be perfect...
<attila_lendvai>ACTION takes a look
<spiderbit>               #:entry-program '((setf sb-sys::*sbcl-homedir-pathname* "/tmp")(stumpwm:stumpwm) 0))))
<spiderbit>ok that compiles at least...
<spiderbit>stumpish "eval *sbcl-homedir-pathname*"
<spiderbit>The variable *SBCL-HOMEDIR-PATHNAME* is unbound.Backtrace
<spiderbit>but that does not fix anything :D
<spiderbit>maybe because I did
<attila_lendvai>spiderbit, you can use this to read that variable: (sb-int:sbcl-homedir-pathname)
<spiderbit>restard-hard and and that is not enough
<attila_lendvai>or use full package prefix for internal symbols
<spiderbit>attila_lendvai I can't set currently anything if it's not the fault of restart-hard
<attila_lendvai>spiderbit, not sure i follow, but eval *sbcl-homedir-pathname* should be eval sb-sys::*sbcl-homedir-pathname*
<spiderbit>do I have to log out to see it.
<attila_lendvai>spiderbit, if you want to work on this then you should find a way to start your gui with a custom stumpwm package
<attila_lendvai>or maybe find a way to start a second gui session... but that's beyond my knowledge
<spiderbit>stumpish "eval sb-sys::*sbcl-homedir-pathname*"
<spiderbit>NIL
<attila_lendvai>spiderbit, either way, once setting "/tmp" works, then try this: (setf sb-sys::*sbcl-homedir-pathname* (uiop:parse-unix-namestring #$(system* (search-input-file inputs "/bin/sbcl") "--eval" "(princ (sb-int:sbcl-homedir-pathname))")))
<spiderbit>I love to kill my browser / emacs session and everything just to restart X to test this... great...
<attila_lendvai>spiderbit, and if "/tmp" breaks things, then try (uiop:parse-unix-namestring "/tmp")
<attila_lendvai>spiderbit, actually, you should straight out go for (uiop:parse-unix-namestring "/tmp"). that value should be a pathname, and #P"/tmp" will not pass through guile...
<spiderbit>ok I try it
<spiderbit>well I then have to restart my session because it did not set the variable still...
<spiderbit>ok that did the trick: ~$ stumpish "eval sb-sys::*sbcl-homedir-pathname*"
<spiderbit>#P"/tmp"
<spiderbit>now I try the princ search-input-file stuff... attila_lendvai
<spiderbit>does it keep the create-desktop-file stuff even if I not add it because of the (inherit stumpwm)
<attila_lendvai>spiderbit, it should, i guess
<spiderbit>haha 2 step forward 2 back
<spiderbit>1st no removing those build phase things... means they are out..
<spiderbit>so no desktop file had to revert
<spiderbit>but also the line does not work
<spiderbit>black@satan ~$ stumpish "eval sb-sys::*sbcl-homedir-pathname*"
<spiderbit>NIL
<spiderbit>and yes I roll-back but I first compiled the line with all the build things and the new line and then I removed them in another build so this is with the new line
<attila_lendvai>spiderbit, you can play with this in a guix repl: (system* "sbcl" "--eval" "(princ (sb-int:sbcl-homedir-pathname))")
<attila_lendvai>for me it prints: /gnu/store/iw0f8ldxbq2w1l1yyxvw2l3sizc4149z-sbcl-2.5.2/bin/../lib/sbcl/
<spiderbit>for me something similar
<attila_lendvai>spiderbit, you should invoke sbcl with --no-userinit and probably also --no-sysinit
<spiderbit>but that does not work in the config
<attila_lendvai>and --noinform, too
<spiderbit>A solution to get rid of the other add-after blocks would be cool... but whatever...
<spiderbit>also is there no sane setup for scheme in emacs that the identation is not this horrific... do I need this geiser thing?
<spiderbit>geiser-mode
<attila_lendvai>spiderbit, i use geiser and the indentation is usually not as horrible as you describe
<spiderbit>how do you quit guix repl
<ieure>,q
<spiderbit>(exit) (quit) ...
<ieure>Or C-d
<spiderbit>C-d also refuses here don't know why
<attila_lendvai>spiderbit, the system* sbcl interferes with that. use C-z and then kill %1
<ieure>spiderbit, Geiser connects your scheme-mode Emacs buffers to a Guile REPL, and I don't believe it changes indent behavior at all.
<ieure>spiderbit, What indent issues are you seeing? I've found that stuff to be fine; my only gripe is that M-q to fill a comment doesn't work at all.
<spiderbit>,q does also not work...
<spiderbit>killing the terminal will work...
<ieure>spiderbit, Paste what you're seeing? All of those should work.
<spiderbit>I closed the terminal
<spiderbit>ohh I am done
<spiderbit>just posted my whole system config when I did not want to ...
<spiderbit>but got it deleted I tihnk is nothing bad in it but didn't look through it...
<spiderbit> https://paste.debian.net/1383895
<spiderbit>ieure well in the file it looks compacter or it changes the identation... from wm.scm as example
<spiderbit> https://codeberg.org/guix/guix/src/branch/master/gnu/packages/wm.scm#L2774
<spiderbit>that looks nice in a browser but if I copy that and Press tab...
<spiderbit>it reformats it drastically and makes much bigger tab spaces
<spiderbit>Well I try to reboot one last time then I have to cook...
<spiderbit>or better said logout/login not reboot
<spiderbit>now it does not start had to do another roll-back...
<spiderbit>now I can't even revert to the working /tmp stuff because I had to restart X and emacs... nice
<spiderbit>I also can't go back in the chat here... well this does not work... and I am so done... if I have to I need to write the path in a file output it to a file and check first the file if that produces the correct string befor I machine gun restart my X server over and over again.
<spiderbit>I can't look up the current config somehow so of the active system
<reepca>not particularly related, but a reminder that "clisp" is a common lisp implementation whose name is based on "Lisp implemented in C". If you mean "common lisp" the language, it might be clearer to use CL.
<ieure>spiderbit, Based on what you're saying, your Emacs config is causing that. Check the indent-tabs-mode variable.
<spiderbit>ieure that should be a function?
<spiderbit>ahh ok both
<spiderbit>it's value is t and I don't modify it in my config
<spiderbit>I have no problem that it can do that... I just think there are sane and less sane levels of indentation.
<attila_lendvai>spiderbit, there are logs at https://logs.guix.gnu.org/guix/
<attila_lendvai>spiderbit, search around the web, it shouldn't be too complicated to start a second X session to debug this...
<spiderbit>(uiop:parse-unix-namestring "/tmp") that was my last working stand... I am done maybe later but not for now at least I am able to set something, now I have to find out how to get the right information to use...
<Samantha252F>💓Hi babe,I'm a single girl
<Samantha252F>If you guys want to have fun,
<Deltafire>not sure if computer is slower or guix pull takes longer these days
<lfam>If you run Guix System on armhf (32-bit ARM) or i686 (32-bit Intel-compatible), please begin working on linux-libre kernel configurations for those platforms: https://lists.gnu.org/archive/html/guix-devel/2025-07/msg00005.html
<lfam>And if you use Guix System on aarch64 or riscv, you are also invited to help out, but I do plan to keep making configs for those platforms in case pitches in
<lfam>In case nobody pitches in
<wingo>good evening!
<ieure>Hello.
<wingo>i finally did "guix home"
<wingo>and so... now magit on a remote guix-home-using system doesn't work: can't find git
<wingo>why would that be?
<wingo>when i ssh over there, there is git
<ieure>wingo, Could be a couple things, probably a good idea to enable tramp debugging to see what's going on.
<ieure>wingo, But either the remote host's dotfiles are horked and not setting up $PATH correctly for remote shells, or magit is trying to use the local path to git (ex. somewhere in your local store) and that doesn't exist on the remote.
<luca>just to double check, have you restarted and/or logged out and back in?
<wingo>yep, i restarted
<wingo>re: dotfiles, surely that is the purview of guix home ?
<wingo>i just did a "guix home import ..." and didn't change anything in the profile-related blocks that it made
<wingo>(and a guix home reconfigure, obvs)
<wingo>usually magit just runs "git", so somehow when ssh'ing, the guix profile doesn't get loaded, or something
<wingo>the weird thing is that tramp works.
<ngz>Hello. How do you apply a PR from the command line or, better, from Emacs ? I found the followin git alias "!f() { for id in $(git cherry HEAD pullreqs/$1 | cut -c2-); do git cherry-pick --signoff $id; done; }; f", but it doesn’t work (unknown commit pullreqs/NNN).
<noe>ngz, on codeberg click on View command line instructions on the bottom of the discussion
<ngz>noe: sorry I meant with the agit worflow.
<noe>should be the same thing
<ngz>Ah. And do you know why the git alias pasted above doesn’t work?
<noe>No, I have no idea how it works