IRC channel logs


back to list of logs

<oriansj>odd that guix uses ${XDG_CONFIG_HOME}/guix/channels.scm but uses ~/.config/guix/channels.scm in the documentation. perhaps for ease of understanding for people who don't mess with their XDG_CONFIG_HOME environment variable?
<rekado>yikes, the patch I just sent to fix “guix deploy” has unintended side effects
<rekado>the system I successfully deployed now cannot seem to be redeployed, because again the host architecture ends up somewhere in a lookup, causing the changes to be rolled back.
<apteryx>what are the guidelines of guile applications relying on guix Guile modules (library) ?
<apteryx>I hope it's not 'guix install guix'
<peanuts>"Problem with Guile Hall's `dependecies` section in `hall.scm`. (#85) ? Issues ? Alex Sassmannshausen / guile-hall ? GitLab"
<festerdam>Hi, all.
<festerdam>Has anyone tried GNU Guix on a low-end machine? How was the general user experience? I know it is possible to offload some guix tasks to another more poweful machine on the network (though I can't remember the name for that, but substitutes are okay), but I'm more interested about the experience without using that feature.
<festerdam>*(though I can't remember the name for that), but I'm more interested about the experience without that feature (substitutes are okay, though).
<apteryx>can package-name be used as a procedure?
<apteryx>e.g. in (map package-name packages)
<ulfvonbe`>(use-modules (guix packages))
<ulfvonbe`>$5 = #<procedure %package-name-procedure (s)>
<ulfvonbe`>survey says yes
<apteryx>I'm reproducing this report though:
<peanuts>"Problem with Guile Hall's `dependecies` section in `hall.scm`. (#85) ? Issues ? Alex Sassmannshausen / guile-hall ? GitLab"
<apteryx>maybe autoloading it plays a part in that?
<apteryx>well, this works:
<peanuts>"debian Pastezone"
<ulfvonbe`>is there a good way to run multiple tor instances on guix system? yt-local really wants to be able to access the control port, but I don't want it to be able to have control over anything tor-related except for what it itself uses
<below_turtles_>will 3d acceleration work on a GeForce GTX 1650 with linux libre?
<adanska>Hi Guix!
<efraim>go-github-com-miekg-dns FTBFS when build with gccgo
<adanska>is there a way to incude a package in a system config that is defined in a file? i want to make a vm that incudes a package i have defined in a file.
<efraim>if it's a file that returns a package it would be easier to copy the package into the VM, if it's in a local channel that defines packages then you can use the regular -L to add that path and then add the import in the VM
<iskyd>Hi everyone, I'm really new to guix and I'm trying to install packages I need using guix to create a reproducible build of my system. I use alacritty but using guix search I see that the latest version is pretty old. Is there a way to update this packages?
<viperML>hey, yesterday I asked about a file "environment-variables" created with the copy-build-system + local-file, why is this file created?
<viperML>or how can I change whether it is created or not?
<civodul>Hello Guix!
<adanska>hi civodul!
<adanska>efraim, okay. by copying, do you mean copying the file into the vm and building it there? or is there a way to access it from the shared store?
<adanska>if its built on my system already
<nmeum>when updating a patch in a patchset do I resend the whole patchset or just the patch that changed?
<mfg>Hey, how do i manually delete a system generation? roll-back does not work since the creation of the most recent generation failed and produced an empty file where a non-empty file is expected. I can boot olde rgenerations but reconfiguring from those results in the same error
<jpoiret>mfg: what's the exact error?
<mfg> it's on another machine so i can't copy-paste it, but it's like "guix system: error: unrecognized boot parameters at '/var/guix/profiles/system/parameters'" which makes sense, since this file is empty
<mfg>and in other generations it isn't empty, so i suppose it should not be empty but is a result of a failed reconfigure
<jpoiret>can you boot from an older generation, then manually delete the new generation in /var/guix/profiles/system?
<mfg>i booted an older generation, but what exactly do i need to delete then? just the (i guess symlink) in /var/guix/profiles/system?
<futurile>Morning Guixers
<futurile>Anyone want to kick-off their Tuesday by reviewing a teeny-tiny first Clojure library? (
<peanuts>"#69091 - [PATCH 0/1] Adding - GNU bug report logs"
<jpoiret>mfg: yes
<mfg>jpoiret: since system is itself a symlink to the newest generation i set that to the most recent working one and deleted the symlink to the faulty generation, now it reconfigured again, just need to reboot.
<jpoiret>it reconfigured? great, that'll do it then
<jpoiret>I was just a bit worried about deleting the symlink to the current generation but I guess what matters is the target of /run/current-system/, not /var/guix/profiles/system/current
<mfg>tho rebooting is problematic itself on that machine. It seems shepherd does not actually shut down the system but runs into a deadlock. civodul said there might be some features to debug this but they are not released yet (?).
<jpoiret>mfg: is it a laptop?
<mfg>No, it's a AM5 desktop system :D
<jpoiret>I have the same problem with mt7921e, I just rmmod it before shutting down
<jpoiret>but yeah might be the same root cause
<mfg>lol, it might be, since i'm also using that driver :D
<jpoiret>ah! have a try then :)
<mfg>amazing, will do
<mfg>jpoiret: Thanks for the tip, rmmod'ing the mt7921e driver actually does make this machine reboot :D weird, but at least i now do not have to power cycle on every reboot!
<jpoiret>mfg: that driver creates a kernel thread that shepherd misidentifies on shutdown, and it just waits for it to exit
<mfg>jpoiret: thx for the explanation :)
<janneke>jpoiret: is your glibc-2.39 work available somewhere? i'm trying to hunt down a 'guix copy'/offload problem with a core-updates hurd-vm
<jpoiret>not yet, I have some trouble with python and libxcrypt
<rekado>issue #69092 contains a workaround to the “guix deploy” troubles..
<peanuts>"[PATCH] machine/ssh: Invert WITH-ROLL-BACK and MBEGIN."
<rekado>it’s odd that PARAMETERIZE apparently doesn’t stay effective over the whole expression.
<dariqq>is it possible to fetch channels over ssh? found some blogs saying that it is but i am getting an error for starting the ssh session (Unable to exchange encryption keys). Do i need some extra setup?
<dariqq>There is also this open issue which looks like a similiar problem
<peanuts>"Can't clone a git repo over anonymous SSH"
<jpoiret>janneke: i'll try pushing it somewhere soon, I think I've got the Python thing down
<janneke>jpoiret: great!
<apteryx>Hello Guix!
<rekado>are you able to build the “hurd” package on the core-updates branch?
<janneke>ACTION really only tried cross-builds
<futurile>Q: does anyone have permissions on to create pages, or give me permissions to do so?
<futurile>Q: I want to create a page to co-ordinate the patch review sessions we talked about at Guix Days
<andreas-e>futurile: I have just created a page:
<peanuts>"Group:Guix/BugSquashing - LibrePlanet"
<andreas-e>But I am not sure whether I did it correctly; it does not show up under "Subpages" of
<peanuts>"Group:Guix - LibrePlanet"
<andreas-e>And I suppose that one needs an account for editing it.
<shayla>Hello everyone!
<futurile>andreas-e: ah that works - can you rename it to 'PatchReviewSessions2024' please?
<futurile>andreas-e: I have an account so I think I'm allowed to add content - just not create new pages - it's oddly locked down even with an account
<shayla>Quick question, here I see that the latest version of alacritty is 0.12.3 But when I do guix search alacritty I see as latest version the 0.9.0
<peanuts>"alacritty 0.12.3 ? Packages ? GNU Guix"
<futurile>shayla: did you just install Guix, have you done a 'guix pull' recently?
<shayla>futurile yes I did a guix pull today
<ebrasca>Helle, how to stop the system from spamming the TTYs with ACPI Error mesages?
<ebrasca>I am on GuixSD
<futurile>shayla: and guix search alacritty doesn't show 0.12.3?
<jpoiret>ebrasca: what's `command -v guix` for you?
<ebrasca>jpoiret: /home/ebrasca/.config/guix/current/bin/guix
<shayla>futurile no I just see 0.9.0
<jpoiret>abrasca: ah my bad that was meant for shayla
<jpoiret>ebrasca *
<jpoiret>shayla: what's `command -v guix`?
<andreas-e>futurile: I have "moved" the page with a menu item, which in fact only creates a redirection. I do not see how to remove pages. Maybe one needs more magic power to do so.
<ebrasca>How to stop the GuixSD system from reporting errors to the TTYs? (Like ACPI Error ...)
<lilyp>add quiet to the kernel command line?
<lilyp>either per kernel-arguments (operating-system record), or by manually editing the command in grub as usual
<ebrasca>lilyp: I will try!
<ebrasca>Adding quiet did nothing to fix my problem!
<apteryx>what's the state of the art of using guix as a library?
<apteryx>I equalled having the guix command available with having the guix modules available, but that's not the case
<apteryx>maybe if guix is available, I should rerun the program within 'guix shell guix -- ' ... ? That feels very hacky.
<janneke>apteryx: makes me think (again) of
<peanuts>"[PATCH] Make Guix modules available too."
<apteryx>sounds reasonable to me. should we have something similar for Guix System?
<apteryx>currently we get the modules as a side effect of having the guix-daemon service running, I think
<apteryx>ah, but problematically, the guix transitive dependencies won't be available by just seting GUILE_LOAD_PATH, so it won't be much usable without users also installing guile-git and friends
<apteryx>that's why I was considering re-execing 'hall' within 'guix shell guix' if the guix command is available
<janneke>well, note that civodul thought it was a bad idea, and suggested (inputs (list ... guix) ;)
<apteryx>can I run anything within 'guix repl' ? it failed to exec the hall script when I try
<apteryx>ACTION retries
<apteryx>I too thought having guix as an input was not desirable, mostly because of the damage 'guix install guix' does
<apteryx>but maybe it's fine since it's not propagated (if wrapped)
<apteryx>in hall though it uses guix to specify dependencies, so that could be surprising if the guix input is not as recent as the guix used by the user (a package that exists for the ambiant guix command may not exist for the older guix input)
<apteryx>maybe not a big deal
<futurile>andreas-e: cool - starting to add some content
<janneke>apteryx: yeah...
<ebrasca>I think I fixed my problem, thanks for the help%
<janneke>ebrasca: "<ebrasca> Adding quiet did nothing to fix my problem!" -- so, how did you fix your guix system?
<dthompson>ACTION tried to update node...
<dthompson>what a nightmare
<ebrasca>janneke: I am not really sure.
<ebrasca>I changed the grub file manually and that entry stop working even after removing the thing I added to it. After that I used one entry from "GNU system, old configurations..." and it worked.
<lilyp>ebrasca: Is that old configuration by chance one that has quiet?
<ebrasca>All of them have quiet
<ebrasca>I did not know it was already set before
<janneke>ACTION checks, and quiet is added by default by %default-kernel-arguments
<andreas-e>In my OS definition, can I somehow specify that a service should not be started automatically? I would like it to be there, but to start it by "herd start ..." by hand.
<ebrasca>andreas-e: You can remove it from services, no idea how will you start it later by hand.
<andreas-e>No, I want it to be there, together with all its packages, and also together with its configuration.
<andreas-e>Visible by shepherd, just not started.
<civodul>andreas-e: you can’t choose by default, but some services have an ‘auto-start?’ field for that
<andreas-e>civodul: Thanks! You mean I add a "line" (auto-start? #f) to blabla-configuration inside blabla-service-type? Unfortunately this does not work for the service I am interested in; but I can make it fail and autostop by not providing it with a crucial file it tries to open...
<civodul>yeah, few services have that field
<lilyp>andreas-e: Before resorting to hacks like that, do submit a patch that adds auto-start? as a config field
<andreas-e>lilyp: Not sure if this is worth it... I am just using it for debugging right now. And if it is of general interest, how about adding this field to everything?
<lilyp>You're now assuming it's of general interest for every service, which I can neither confirm nor deny, but I'd hazard a guess that some services are meant to be auto-started.
<rekado>systemd generally separates “enabling” a service from “starting” it.
<lilyp>But an enabled service is started on boot or whatever other trigger it has.
<futurile>Q: there's an image floating around which is Guix for 'packages, environments, containers, systems' - does anyone know where the original is? (
<ieure>I also recall seeing this in a slide deck for a talk, but the name escapes me.
<futurile>ieure: ah cool - thanks
<futurile>ieure: yeah - same, saw it in a slide deck somewhere
<ieure>futurile, FYI, I used TinEye to find the source. It's very effective for stuff like this.
<rekado>futurile: it’s from one of my talks
<rekado>here’s a version from a more recent talk:
<rekado>it’s the 31-guix*.svg files
<peanuts>"2023-dfn\talks - guix/maintenance.git - Articles, talks, and documents for Guix maintainers"
<jackhill>hey risc-v folks: what do you think of this system:
<peanuts>"Milk-V Pioneer"
<futurile>thanks rekado and ieure
<reedm>Can someone point me in the right direction for learning about how to add a custom subcommand to guix? For instance, I know that the "gwl" package adds a new command (extension?) "guix workflow". I'm happy to just look at the source for gwl, but is there a more minimal example or a place in the manual about guix extensions?
<jpoiret>reedm: simon touches on it in the second part of
<jpoiret>i don't really know how it works for a channel though
<reedm>Thanks, I'll be sure to watch that!
<lilyp>For a channel, you need to define a (guix extensions your-command) module.
<lilyp>See <> for inspiration
<peanuts>"\extensions\guix - gwl.git - Guix Workflow Language"
<reedm>I'm trying to reproduce the simple example in the video, but I'm getting an error:
<peanuts>"debian Pastezone"
<mehrad>Hello everyone, I have two computers both with Guix installed on foreign distro (Manjaro), and both are updated via. `sudo -i guix pull` when I run `guix shell r` on them, they behave differently. One manages to run R normally, and the other one throws this error:
<mehrad>/gnu/store/mad9hccdj4xajpw2kkf96kc7vidh3bdn-r-minimal-4.3.2/lib/R/bin/exec/R: symbol lookup error: /usr/local/lib/ undefined symbol: UP
<mehrad>I wonder how I can investigate this situation
<rekado>mehrad: R from Guix should never load libraries from /usr/local
<rekado>the problem here is that R from Guix loads your system’s
<rekado>since these are incompatible you get a weird error.
<mehrad>I even tried `guix shell --pure r` and it shows the same dual behaviour
<rekado>if you only want to avoid this problem without investigating more you can use “guix shell -CN”
<jpoiret>dlopen must be eradicated
<mehrad>Thanks. putting in container worked, although I'm puzzled about why -N
<rekado>mehrad: N is for network.
<rekado>you might not need it
<rekado>but people often get confused when they are suddenly cut off from the internet.
<mehrad>yes, that far I figured from `guix shell --help` but I was wondering if it has any effect on the situation at hand
<rekado>with -C you prevent R from loading stuff from /usr/local because that location doesn’t exist from the perspective of the process.
<mehrad>I tried without -N and it also worked. hence my question to know if it is a good practice or something :)
<rekado>the -N has no influence on this behavior.
<rekado>if you want to figure out why R loads /usr/local stuff in the first place, it may be a good idea to check if ~/R exists
<rekado>R loads stuff from there by default, and there is a chance that you’ve got stuff there that’s linked with libraries from locations like /usr/local/lib.
<mehrad> yes, the `~/R` exists and it is a folder (the conventional place to have the libloc for R
<mehrad>Yeah, that makes sense, thanks for the explanation. Although i have the ~/R on both computers. I perhaps have to diff these two folders and see
<rekado>generally you won’t be able to use the stuff from there with R from Guix — unless it started out empty and you used only Guix tools to create its contents
<rekado>(e.g. by running “install.packages” inside a “guix shell -CN gcc-toolchain r-minimal …” environment.)
<weary-traveler>lilyp: was the recent patch series on emacs aot compilation intended to be v9 (it was sent as v7)?
<mehrad>I'm also facing some issues with rstudio-server from guix-science channel. It tells me that "guix shell: error: rstudio-server: unknown package" where as I checked the exact commit in guix-science and it actually exists. I have put the channel.scm, manifest.scm, and the guix time-machine command I run here:
<mehrad>Any insights are appreciated
<peanuts>"Disroot Bin - Encrypted pastebin by PrivateBin"
<jpoiret>mehrad: what's `command -v guix` for you?
<jpoiret>does `guix describe` mention guix-science?
<mehrad>The guix-science was specified in the channels.scm. I'm using time-machine and so the guix describe is not describing that
<mehrad>although I used guix describe to get the latest commit hashes in the channel
<lilyp>weary-traveler: thanks for catching this
<jpoiret>ah, my bad, didn't see you were using time-machine like this. Does `guix time-machine --channels="channels.scm" -- describe` mention the channels you're expecting?
<weary-traveler>lilyp: np; thanks for your work on this
<mehrad>jpoiret: you are absolutely right, it does not show me the guix-science despite the fact that it is in the channels.scm
<mehrad>I don't see anything wrong with the channels.scm and Emacs shows all parenthesis are matched
<jpoiret>quite weird
<trnry>So I recently moved from a profile over to guix home, and I'm a bit bummed out that now when I do a reconfigure it doesn't show me all the updates that I'm getting like it did with `guix package -u`
<jpoiret>I don't see anything immediately wrong here
<trnry>Is there a way to see what versions have changed?
<reedm>can someone explain what's happening at 7:26 in this video? In particular, I don't know what that "G_" thing is doing, and I'm having trouble searching the guix source for the definition of it
<guido_>Hi everyone, I'm Guido who was at the guix-days in BXL the other week.
<mehrad>jpoiret: I have tried the same channels and manifest on both machines and they both are behaving the same. it is perhaps too much to ask, but can you also give the files a try and see if you get the same error?
<jpoiret>G_ is for i18n in Guix, it's defined in `guix/i18n.scm` to be an alias to Guile's gettext
<jpoiret>it's described in the guile manual
<jpoiret>mehrad: I'm having a look, protip in the meantime, you can use %default-guix-channel instead of copy-pasting the whole guix channel definition
<reedm>jpoiret: Thanks!
<mehrad>jpoiret: True, I kinda have forgot about that. I wanted to be very explicit. My idea is to use time machine and freeze the time for my project so that it can be always reproduced. I wonder if that variable would always remain the same or at some point in future versions of Guix it might change.
<old>reedm: G_ is for translating string I think
<jpoiret>mehrad: it does work for me, describe does mention those channels
<jpoiret>tbh i'd first try to `guix pull` and use the guix pull'd guix
<mehrad>I did guix pull about 12 hours ago. I give it a go again and see. Thanks for confirming that the scheme files are fine and should in principal work. Cheers
<mehrad>jpoiret: Thanks for all the info, time, and help :)
<jpoiret>mehrad: hm, you mentioned that your `command -v guix` points to /usr/bin/guix, right?
<jpoiret>that means you're not using the guix pull'd guix
<jpoiret>it should point to ~/.config/guix/current/bin/guix
<jpoiret>have you set up your .profile file to load the guix profile at ~/.config/guix/current?
<mehrad>The `echo $GUIX_PROFILE` returns `/home/mehrad/.config/guix/current` and I though it is correctly set
<jpoiret>have you logged out and back in since guix pull'ing?
<mehrad>jpoiret: I did `source "$GUIX_PROFILE/etc/profile"` which made the `which guix` to return `/home/mehrad/.config/guix/current/bin/guix` (the same as command -v guix). Now running `guix pull` . When this is finished, I will try to run the time-machine again and if it didn't work, I will start a new shell session and try the time-machine
<jpoiret>mehrad: good. Usually people configure their .profile to do this on login, so that the `guix` they use is this one by default
<mehrad>jpoiret: So is it always prefered on foreign distro to have this guix binary rater than /usr/bin...?
<jpoiret>yes, the /usr/bin one is the one that is never updated
<jpoiret>even of guix system, people use the guix pull'd one almost always
<jpoiret>(although I personally use my local check-out 95% of the time)
<mehrad>after guix pull the describe still showed only one channel. I closed the shell session and opened a new terminal and still the same issue
<mehrad>jpoiret: Thanks for the help, I will spend some time on it tomorrow and see what I can figure out. Thanks for the help and the pro tip
<jpoiret>quite weird
<rekado>have you run “/home/mehrad/.config/guix/current/bin/guix time-machine -C channels.scm -- describe”? Just to be absolutely sure you’re using the expected Guix?
<ieure>Is it safe to run `guix system init' on an already-installed GuixSD? I assume that if I give it the same system config as the install uses, it is, but wanted to check before doing something I'll regret.
<mange>Why do you want to run init again, rather than reconfigure?
<weary-traveler>lilyp: regd. emacs aot patch series: there's been some improvement. uniquify-item-p is correctly noted as being natively compiled. however, backquote-process still shows as being byte-compiled despite the existence of backquote.eln
<weary-traveler>i'll respond to the issue with related details later, but figured you may be interested in knowing earlier if possible