IRC channel logs


back to list of logs

<graywolf>What is a difference between a guix repository and a guix package (the thing in package-management.scm which is now revision 13)? Package definitions do come from the repo, the same for for example grub configuration. What does come from the package?
<graywolf>I am looking at c65cc26baaaecdba5d64165813df3c8b5d61c3b7 and realizing I have no idea why it is required, I somehow assumed everything comes "from the 'guix channel".
<pastor>Good night. Anyone here using shepherd? I would like to know how you have it set up so you can get a service started by shepherd, such as emacs-daemon, have environment variables defined by the graphical environment such as SSH_AUTH_SOCKET started by the gnome-keyring.
<pastor>I'm starting `shepherd` in `.profile`
<pastor>Any other way of doing it?
<apteryx>pastor: I believe one way is using 'guix home' to manage user services
<apteryx>another way is hand-made simple service files and manually launching 'shepherd' in ~/.xsession or similar
<Kolev>apteryx: Guix Home is hard.
<apteryx>I'm not in a good place to tell, but I believe you
<pastor>well, now I'm starting shepherd in '.profile' the problem is that gnome won't initialize the ssh-agent until the graphical session is ready so the processes started by shepherd don't have access to that variable
<graywolf>pastor: (require 'server) (or (server-running-p) (server-start)) Is my solution to this particular problem
<pastor>apteryx: I imagine guix home won't be able to see the variable either
<pastor>graywolf: That sounds promissing. Is it delaying the start of the shepherd services or something?
<graywolf>no no that goes into init.el
<graywolf>So first time I run emacsclient it start the deamon
<graywolf>(first time == if it is not running)
<pastor>ah. You mean in emacs
<apteryx>we can't have (guix search-paths) on the build side, right?
<Kolev>RIP SSH.
<pastor>hum. I can't believe we don't have a workarround this in shepherd 🫤
<graywolf>pastor: I guess you could write a shepherd service that would scan /proc and look for a process with SSH_AUTH_SOCKET env varible. One it is found, just copy the value and lunch emacs.
<graywolf>Horrible? Yes. Working? Probably.
<graywolf>I mean, it is hard to have general solution. Env variables are in general inherited down the process tree. So it the env is not present when the shepherd starts, it is hard to get it later. Maybe you could make gnome auto-launch the shepherd instead?
<graywolf>(After the var is set)
<graywolf>That sounds like cleanest solution
<pastor>I was thinking that the login manager should have a what to control the order of processes
<graywolf>But login manager here is GDM I guess. So you could try asking in #gnome (is that a thing?).
<pastor>I'm using sddm
<pastor>maybe `xsession-command`
<graywolf>Ah, so #sddm? :) Sorry, I am unlikely to be of much help. I just log into tty and run startx :D
<dcunit3d>is git-fetch preferable to url-fetch?
<dcunit3d>and can the libre kernel run BPF programs?
<pastor>graywolf: I guess you are not using gnome
<graywolf>i3 :) I... don't like gnome.
<Kolev>I like Sway
<pastor>I don't like neither but here we are 😤
<pastor>guys. Does `xdisplay-start: Script to run before stopping xorg-server.` sound reasonable?
<graywolf>dcunit3d: Assuming you use linux-libre-with-bpf package for the kernel (it should.
<pastor>It's from the documentation of `sddm-configuration`
<graywolf>You copied description for xdisplay-stop :)
<pastor>ups. Copied the worng docstring.
<Kolev>R.n. I just want my GPG SSH auth. to work.
<pastor>haha. Yeah. I was reading it like 😵‍💫
<ajarara>hello #guix, if I want to try life without guix home I can just comment out the HOME_ENVIRONMENT blurb in .profile, yes?
<pastor>Kolev: I feel you :(
<graywolf>pastor: Quick google suggest that that is called after the X for SDDM is started, so before login dialog.
<graywolf>Not before launching the WM.
<Kolev>pastor: My Guix Home issue is on help-guix m.l.
<ajarara>yeah I gave up long ago on gpg ssh auth backed by yubikeys. Was happy to see guix supported it openssh with sk support : )
<Kolev>All my hosts use GPG auth. now. I can't go back.
<dcunit3d>thanks graywolf i'm looking to package picosnitch
<pastor>Kolev: I cannot find it
<peanuts>"Guix Home: SSH won't ask for GPG password"
<dcunit3d>Kolev: is gpg hanging for awhile when it's asking for PIN?
<Kolev>dcunit3d, SSH asks for the user password. It doesn't get handed to GPG.
<dcunit3d>if it is, you can run `pgrep -fa gpg` or get the pid for scdaemon and so forth, then run `pstree -H $gpgpid`
<dcunit3d>is ssh-agent also running or just gpg-agent?
<Kolev>dcunit3d, `gpg-agent --launch` (as I did on Fedora) did not work.
<Kolev>dcunit3d, IDK how to check if they are running.
<dcunit3d>if shepherd is running alright, then you can do `herd status` or you can look at pstree if it's not too large.
<Kolev>dcunit3d, running:
<peanuts>"debian Pastezone"
<dcunit3d>have you looked at the shepherd docs? there are a few new commands like `herd detailed-status`
<Kolev>dcunit3d, I'm new to Guix System. I am coming from Fedora.
<dcunit3d>if you're trying to use gpg-agent for ssh authentication, ssh-agent can't also be running otherwise, it will attempt to intercept authentication requests
<Kolev>dcunit3d, hm. ssh-agent is running. How do I prevent this?
<dcunit3d>i'm not sure. i'm looking at your home-configuration in the mailing list. it doesn't include an `ssh-agent` service, but the output you pastebinned makes it seem like it's pulling that in as a service dependency.
<dcunit3d>i'm still not using guix home myself.
<Kolev>dcunit3d, I'm new to Guix System and Guix Home, but Guix Home seemed like the proper way to do my user config, rather than doing things manually with `guix install` and manually placing dotfiles everywhere.
<dcunit3d>if you run `guix home edit home-gpg-agent` it will pull up your $EDITOR and take you to the service definition
<dcunit3d>i think it definitely is, but i still have a lot of older scripts. i'm running on Arch & Guix SD and I'm using multiple window managers, so it's a bit hard to do all at once lol
<Kolev>dcunit3d, that brings my Emacs to a read-only file.
<Kolev> /gnu/store.
<dcunit3d>right, but that's the source for the home service. the source line that reads `(provision '(gpg-agent ssh-agent))` is what was confusing me. it's not starting ssh, it's just saying that service is already provided in case other services mention a dependency on it.
<dcunit3d>there is also `guix edit` for packages and `guix system edit` for system services.
<Kolev>dcunit3d, I'm no Lisp wizard.
<dcunit3d>and `guix graph` provides options to see the package and services dependencies for pretty much everything.
<dcunit3d>then guix graph may be what you need to use.
<dcunit3d>i didn't know scheme when i started and i'm still not that great, but the packages and services use DSL's which is fairly clear (minus some of the scheme magic)
<dcunit3d>is SSH_AUTH_SOCK set for you?
<Kolev>dcunit3d, unset.
<pastor>graywolf: I found a workaround. BTW I missed whatever you said since I lost conection had to restart to do the tests and well this is IRC 😢
<dcunit3d>that variables tell processes which SSH socket to attempt to connect to for ssh. the gpg-agent acts like the ssh-agent, so other processes still need the SSH_AUTH_SOCK variable exported/set in order to talk to the correct socket
<pastor>The secret sauce is to let GNOME start shepherd by using a .desktop service defined in '~/.config/autostart'
<dcunit3d>your home environment needs to set the variable, but you can test it by setting it in your shell. the SSH_AGENT_PID should be unset.
<Kolev>dcunit3d, should I define an env. var. in Guix Home config?:
<graywolf>ACTION likes irc but reaction emoticons for specific messages are sometimes nice...
<dcunit3d>yes, if you want processes to be able to interface with gpg-agent. (i tried using a TTY-only gpg config for awhile, but i gave up on that and just used pinentry-gtk-2)
<dcunit3d>See this guide's section on replacing agents
<peanuts>"GitHub - drduh/YubiKey-Guide: Guide to using YubiKey for GPG and SSH"
<pastor>graywolf: Yep. What would be even nicer is to be able to go away and come back later to read but, meh. Enyoing the 90s I guess
<dcunit3d>you can either set it to "run/user/$UID/gnupg/S.gpg-agent.ssh" or use "$(gpgconf ... )" to retrieve it
<graywolf>pastor: I have on my todo list to setup a bouncer but never got around to it
<dcunit3d>and your gpg-agent.conf references `pinentry` and I don't think that works in a window manager. i'm not sure though
<pastor>graywolf: Yeah. Not on my priorities neither
<Kolev>dcunit3d, I don't understand the article with regards to SSH_AUTH_SOCK.
<Kolev>export SSH_AUTH_SOCK=$(gpgconf --list-dirs agent-ssh-socket)
<pastor>Well. I have to go. That was a nice brainstorm about SSH pains. Good night 🙂
<Kolev>dcunit3d, it works after setting SSH_AUTH_SOCK! 😀
<Kolev>pastor, be well!
<dcunit3d>your home environment config should actually default to setting that. adding the home-gpg-agent-service brings in an "environment-variables" service extension.
<dcunit3d>are the aliases like `grep --color=auto` working?
<dcunit3d>to see where the home-gpg-agent-service is declaring that the SSH_AUTH_SOCK is being set, see the line (define (home-gpg-agent-environment-variables config) ...
<dcunit3d>you may not use emacs or write scheme, but you need to at least read the source. otherwise it can be tricky to figure out what's going on. also, i occasionally need to write my own packages to get things.
<dcunit3d>Kolev: guix is worth it though. i don't know that there's a faster way to learn linux. all the logic for building packages and systems is in one place. there's basically one repository where that's all defined.
<graywolf>After using guix for a bit, there is only one thing I fear. Working with "legacy" distros at work... I do both container and ubuntu packaging at $dayjob and, well, it sucks hard.
<Kolev>I'm close to just going back to Trisquel. 🙁
<graywolf>I am so not looking forward to 11th... :/
<Kolev>graywolf, 11th?
<graywolf>HIGH severity Curl CVE will be revealed together with the fix. So we will likely be rebuilding the world at work...
<dcunit3d>yeh i wish more distros would change how they package things, but i think a lot of what defines canonical/redhat's business /is/ the packages that automate components of system configuration
<dcunit3d>what is it? i think even a "curl with no vulnerabilities" is bad, honestly
<dcunit3d>or could be.
<graywolf>curl is cool no? it is really handy tool
<graywolf>Kolev: I would recommend to just push through the pain, in my opinion it is worth it if you have the time. Albeit that might just be Stockholm syndrome talking :D
<Kolev>My home-configuration.scm won't even build anymore.
<graywolf>What error?
<Kolev>graywolf, error:
<dcunit3d>Kolev: i would just recommend reading ahead a bit (in the manual and in source) and finding people you can ask for help. the author of guix home has a youtube channel with some good videos on guile/guix
<peanuts>"debian Pastezone"
<Kolev>dcunit3d, Tropin is even harder to understand. The codebase of rde is huge.
<dcunit3d>and also, the systemcrafters channel has a good matrix server
<Kolev>dcunit3d, I thought System Crafters was on IRC.
<dcunit3d>they had a bridge to their matrix server, but i think that's offline for now
<Kolev>I use XMPP, and the bridges to Matrix suck because of how inefficient Matrix is.
<dcunit3d>do you have a bridge to IRC from XMPP?
<Kolev>dcunit3d, I have access to IRC.
<Kolev>dcunit3d, I'm talking to you right now from XMPP.
<dcunit3d>you can just use the Elements client (although tbh, i hate having to maintain 50 chat clients and esp. being on discord)
<dcunit3d>yeh, i've been meaning to get XMPP configured, since GNU has a server for members
<Kolev>I refuse to use Matrix's bloated piece of garbage.
<Kolev>Get the XMPP bridge working.
<dcunit3d>then use ement.el or something
<Kolev>Ugh, I'm getting hungry and cranky.
<peanuts>"GitHub - alphapapa/ement.el: A Matrix client for GNU Emacs"
<dcunit3d>and no one's forcing you
<dcunit3d>yeh i gtg as well
<graywolf>Time to sleep, be well everyone.
<whereiseveryone>Is there currently an easy way to answer the question: What python packages have tests enabled and passing?
<whereiseveryone>s/and/and are/
<apteryx>kindof? you can find all the packages uing the python-build-system and whose #:tests? argument is true and which have substitutes
<apteryx>it should all scriptable via Guix API, starting from fold-packages
<Kolev>:-( My Guix Home.
<ulfvonbelow>well now this is an interesting thing. I have a local commit that adds rust-asn1-derive-0.12 - that is, one singular rust package that takes up all of 22 lines - to (gnu packages crates-io), and after rebasing it git seems certain that that commit's diff should now take up 18000 lines.
<ulfvonbelow>I know diff isn't terribly smart, but I didn't think it was that, uh... unpredictable?
<ulfvonbelow>is this the result of somebody going through and alphabetizing?
<Kolev>I'm so lost.
<peanuts>"csh/dotfiles: My public configuration files. - dotfiles -"
<apteryx>hm, 'mumi current 65924' hangs
<apteryx>bad gateway 502 error
<apteryx>ah, that's being hung up or something
<apteryx>I've restarted it
<apteryx>mumi send-email is handy
<apteryx>only accepts patches as arguments though
<apteryx>unlike 'git send-email'
<apteryx>cbaines: is qa down?
<Gooberpatrol_66>how do I get libgcc in an FHS container?
<dcunit3d>I'm trying to package picosnitch, which uses ebpf to modify. what does it mean when a python executable (intended to be built from setuptools) references PYTHON_USER_SITE?
<dcunit3d>... is this a bad sign? Here's my package with a few relevant links/notes
<dcunit3d>it's packaged for nix by the auther, but that version includes some caveats about running it as a service. it looks like a cool utility and i would send a patch upstream, but i wouldn't bother if the quality isn't there
<apteryx>dcunit3d: the Guix packaging of the python package should provide everything it needs
<apteryx>it shouldn`t look in PYTHON_USER_SITE (~/.local/lib/python3.10/site-packages), which is populated by 'pip install something'
<apteryx>if you don't bother about the quality it may be best to send it to a channel with lower standards
<dcunit3d>k, thanks for taking a look
<dcunit3d>i'm running into environment issues with it on my system anyways. I haven't run into PYTHON_USER_SITE before. that would seem to make assumptions about how a system is configured.
<podiki>Gooberpatrol_66: provide the relevant package to the shell command. though if it is in gcc:lib currently need to add -e '(list (@@ (gnu packages commencement) gcc) "lib")' as an argument (gcc:lib is not accessible otherwise)
<vhns>Hi. Is there any way to get guix to spit more verbose info about errors it encounters? I'm currently getting a `guix system: error: failed to load '/mnt/etc/config.scm': Invalid argument" on the following config.scm . I do believe it to be written properly.
<bumble>hello guix!
<Gooberpatrol_66>thanks, that worked
<PotentialUser-85>does guix/%desktop services come with some kind of mac changer? i have worked out the reason I keep needing to rerun dhclient every few minutes is because my mac address keeps changing
<minima>hi, there's a package foo that i install from my personal channel bar; the package installs fine with "guix install foo"
<minima>i'd like to use foo in a gexp, so i was expecting to just do something like "(string-append #$(foo) "/bin/foo")"
<minima>if i do that, though, i get an error about foo being an undefined variable
<minima>i imported foo as a module at the beginning of the file though
<minima>what is it that i'm missing here?
<snape>minima, could you provide your code snippet?
<minima>hi snape, sure, a min example is that:
<minima>this works: (use-modules (guix gexp)(gnu packages bash))#~(#$bash)
<minima>while this doesn't: (use-modules (guix gexp)(foo))#~(#$foo)
<minima>but foo as a module is loaded just fine, and i can call public things that are defined in foo
<snape>ok, this is because your guile doesn't know about the module where foo is defined
<minima>hmmm interesting but not sure i'm following
<snape>yes yes plz wait ^^
<minima>it does load foo though? how is that
<minima>oh sorry :)
<snape>could you paste your foo module in minima?
<minima>snape: sure, but it's just a hello-world kind of thing
<snape>does foo refer to a package?
<minima>yes, foo is the name of the guix package as well as the guile script/executable
<minima>if that's the correct answer
<minima>in other words, i do "guix install foo" and then i can type "foo" from the CLI and i get some sort of hello world
<snape>can you paste at least a file that i can use to reproduce the issue?
<peanuts>"View paste LN7A"
<snape>minima: i can't see the package foo
<snape>in what I see there is one module called foo, that contains the main procedure, that's all
<minima>snape: sorry, let me isolate the package too
<minima>(almost done)
<minima>snape: sorry it took so long,
<peanuts>"View paste LIBA"
<minima>that's foo-package.scm, which goes together with the prev snippet (foo.scm)
<snape>but it's not loaded within foo.scm?
<snape>or you didn't update the entry-point file?
<minima>snape: the folder foo/ contains foo.scm and foo-package.scm and it's the package (plus some other file that makes this a channel)
<minima>in a separate file i try to use foo within a gexp, and that's where i get the unbound variable error
<snape>so you're trying to use foo within a gexp in the first file right?
<snape>can you paste an updated version of that file, where you call foo within a gexp, and that doesn't work?
<snape>(it's very difficult to help if I don't have an easy way to reproduce the error ^^)
<minima>snape: the gexp is a separate file, my home configuration - i'm sorry if this is difficult to reproduce and i feel i've taken a lot of your time already!
<minima>my home configuration makes use of g-exps in a few places
<snape>but you see, this is an important information ^^
<minima>sure, absolutely
<snape>i'm not confortable with home configuration but i  imagine the environment in which you call your code there is not the same as the one in the first file you pasted
<minima>hm i reconfigure my home in a shell where foo is available as a command
<snape>h ok
<minima>but yeah i have a vague intuition of what you say, i guess that's the reason
<snape>but foo may be available as a command because it's installed
<minima>yes, it is installed, true
<snape>and in the same time Guile would not know about where foo is
<snape>the definition i mean
<minima>but my home config file contains a (use-modules (foo))
<snape>guile knows about things in GUILE_LOAD_PATH
<snape>from there it finds things you call with (use-module whatever)
<snape>but the (foo) module i see is empty
<minima>foo exports main
<minima>maybe there's something else that i should define as public
<snape>does it contain (use-module (foo-package)) ?
<minima>ah, no, it doesn't!
<minima>let me try with that
<vhns>Has there been a workaround to not being able to use GuixOS with LUKS2 for the root filesystem?
<minima>snape: i think this is great progress! still getting an error but that might be unrelated
<minima>snape: thanks so much, i'll let you know how it goes
<snape>minima: nice!  sure let me know
<snape>you see foo is part of the foo-package module
<snape>but it's (counter intuitively) not part of the foo module
<snape>in your example
<snape>maybe next time send a minimal home configuration with which one can reproduce the issue, because it would allow to see things like that quiclky
<minima>snape: you're right, sorry - it was more painful (for you, not for me) than it should have been
<minima>this now works the same as for bash: (use-modules (foo-package))#~(#$foo)
<snape>not at all, i havent been playing with guile for a few years
<minima>i was missing the '-package' bit!
<snape>it's good to learn again
<minima>oh cool, have you been on guix in the meanwhile already or have you joined the guix world only recently?
<snape>i was active in 2017-2020, then made a small break
<snape>now i have some time
<minima>that's cool
<minima>thanks for all the help today
<snape>you're welcome
<PotentialUser-85>does the live usb us gnu/hurd kernel?
<Altadil>PotentialUser-85: No, it uses Linux-libre.
<PotentialUser-85>Altadil thanks, i think I'm finally getting on top of this wifi issue, seems like it's a kernel issue and that's why it works on the liveusb
<PotentialUser-52>Hello GUIX, any idea where to add the path to dmenu to have my guix binaries available?
<PotentialUser-85>how can i see logs for hurd?
<snape>PotentialUser-52: i guess it depends on how you start dmenu
<snape>probably ~/.profile?
<PotentialUser-52>snape: dmenu is available in i3 after I installed suckless-tools
<PotentialUser-52>snape: guix runs on top of Debian
<snape>did you try adding your PATH in .profile?
<snape>and restarting the session maybe
<PotentialUser-52>source $GUIX_PROFILE/etc/profile
<PotentialUser-52>This is the end of my .profile
<PotentialUser-52>And I even restarted Debian
<PotentialUser-52>The binaries are available in the terminal, but not in dmenu
<vhns>PotentialUser-52: IIRC PATH is hardcoded in dmenu
<vhns>PotentialUser-52: Not exactly hard coded, but I assume "/bin/sh echo $PATH" wouldn't return what you expect it to
<peanuts>"dmenu_path - dmenu - dynamic menu"
<vhns>Could you test so?
<vhns>s/echo $PATH/printf $PATH \\n/
<PotentialUser-52>Why in the world the dmenu path is hard coded? Will look for alternatives. Someone suggested to set the PATH in /etc/profile. But this stinks for user paths
<vhns>you could just modify the file we linked you up above
<vhns>not sure if you missed it
<vhns>so I'll send it again
<vhns>PotentialUser-52: IIRC PATH is hardcoded in dmenu
<vhns>PotentialUser-52: Not exactly hard coded, but I assume "/bin/sh echo $PATH" wouldn't return what you expect it to
<peanuts>"dmenu_path - dmenu - dynamic menu"
<vhns>One way out of it would be to change that or just edit the cache manually
<vhns>I'm sure you'd also be able to edit guix as to apply a patch to it during compile time or to just apply a change to that file
<vhns>tho I'm not smart enough for that
<PotentialUser-52>vhns: I can manually add entries to .config/dmenu_run and then it appears in the search bar but will not be executed because its still not in the PATH
<PotentialUser-52>dmenu comes from debian. I might try the guix version, but I am unsure if it plays well with the rest of the desktop
<PotentialUser-52>Its a debian inside VMWare Fusion running on Macbook M1.
<PotentialUser-52>GUIX does not run native on M1 yet. But I am fine with that
<vhns>I don't ever expect it to run on the M1, I'm pretty sure the currently working Linux kernel for that has binary, non-free blobs all over the place
<snape>PotentialUser-52, dunno if this will help, but you can access dmenu env with "cat /proc/<pid>/environ | tr '\0' '\n'"
<snape>maybe this will help you understanding some stuff
<snape>like if there is a PATH env
<jlicht>rekado: good call on using esbuild for minification! It makes a lot of sense
<wonko-th1-sane>ahaha xkbcomp just printed a variation on a classic theme, an error: "Error: success in unknown" … made me smile and think of you fellows
<nmeum>could someone explain to me why the QA status for is unknown? am I doing something wrong? what does unknown even mean in this context?
<peanuts>"[PATCH] syscalls: Add support for musl libc"
<vivien>My issues start to look good :) #66099, #66339 and #65879
<peanuts>"[PATCH gnome-team 0/3] Update upower"
<peanuts>"Gnome-team dbus socket in /var/run/dbus, not /run/dbus"
<peanuts>"[gnome-team 0/1] Update gnome-control-center"
<lilyp>vivien: oh, nice
<lilyp>so it's 66339 v6[.5] + 66099 v10 + 65879 v3 or do you prefer another order?
<vivien>66339 and 66099 commute, and in my local branch, 65879 is applied last
<vivien>If you can rename the loop and restyle for 66339, that will save some internet bandwidth, at least a quarter of a polar bear worth of CO2.
<apteryx>lilyp: shell completions get installed to git-minimal; I think that's reasonable to do (it's lightweight, and make the package much more useful to git-minimal command line users)
<vivien>(all of them should commute though, but I didn’t try all orders)
<lilyp>it makes sense to apply series as series tyvm :)
<lilyp>will restyle, but first I've got to waste some polar bears on c-u
<lilyp>apteryx: fine, then it should go to c-u anyway
<vivien>The polar bears wish we still had core-updates-frozen
<apteryx>lilyp: had you looked at 3/3 ?
<lilyp>of which patch?
<apteryx>that's the one fixing 'git submodule' in a pure environment
<peanuts>"git searches coreutils and util-linux commands in PATH"
<apteryx>if the 3 look good to you, I'll merge to core-updates
<lilyp>I'd search coreutils and sed by command names, but otherwise LGTM
<lilyp>there's probably also a better way of writing PATH-variable-definition, but if it works it ain't stupid
<apteryx>yeah it works; the idea to *not* patch the exact commands in the code is lower the maintenance in the future
<apteryx>you can't just substitute something like 'basename' safely
<apteryx>so you need to start substituting things like '$(basename', and that's fragile and the day the code changes it's silently broken
<apteryx>so augmenting PATH should be more robust
<apteryx>poll: how many of you use a 'commit-msg' hook of your own in .git/hook/commit-msg ?
<lilyp>I understand that, but what I mean is to use (search-input-files) with known binaries ("/bin/sed" and "/bin/yes" spring to mind)
<apteryx>you mean instead of using evaluate-search-paths?
<lilyp>inside evaluate-search-paths with a dirname
<apteryx>that is handy to format the: export PATH="/gnu/store/x0h85ra58syy0zs6lgxiy23j1q4ypn9m-coreutils-minimal-9.1/bin:/gnu/store/fyy3wkjkix16sb1ginqw2kbji74cwl2b-sed-4.8/bin${PATH:+:}$PATH" snippet automatically
<lilyp>so that one could swap out coreutils-minimal for a regular coreutils for example
<lilyp>(or gash-utils)
<apteryx>I think that's a limitation of input rewriting that's handled by bug#65665
<peanuts>"package-mapping with #:deep? #t doesn't get all the implicit inputs"
<apteryx>it now rewrites gexp inputs found in arguments as well
<apteryx>which I suggested should be merged to core-updates as well (waiting on the core team to say something about it)
<lilyp>is it? I don't think it'd easily let you swap labels, or does it?
<lilyp>ahh, my bad, it'd be agnostic of the label
<lilyp>instead replacing the package itself IIUC
<vhns>I'm very sorry for asking this again, but it seems that there's no easy way to get a LUKS2+LVM root on GuixOS that doesn't involve exposing /gnu under a LUKS1/unencrypted partition or some major undertake in writing stuff for guix, right?
<vhns>(or getting grub to learn luks2)
<stikonas>vhns: grub knows some luks2, only argon2 is not supported
<stikonas>though I don't know why upstream maintainers don't finish grub2 support... argon2 support patches were published a few years ago but are still not merged
<mirai>stikonas: were they reviewed?
<stikonas>mirai: I think so
<peanuts>"Re: [PATCH v2 6/6] luks2: Support key derival via Argon2"
<mirai>stikonas: you can try asking in that thread what's the status or what else is left to do
<superkamiguru>I am writing a config file for use with “guix deploy” and am targeting my sbc cluster for deployment. The last issue I am running into is how to best go about defining the file system declarations that would apply to all 4 machines
<superkamiguru>I removed the original labels from the config, and tried using the device name, but realized on two of the machines the device names differ
<lilyp>superkamiguru: why not use labels? alternatively, write the UUIDs to a file and read that
<superkamiguru>I was trying to find a way to define the file systems in a way that would be one definition for all 4 devices. Are you suggesting that I just create a list of the labels for the 4 devices and try to loop through assigning them in the deployment config?
<lilyp>that'd be one way; otherwise, ensure consistent labels
<superkamiguru>Also sorry I removed the original UUIDs from the base config not labels.
<superkamiguru>Ok. I will look more into setting/using labels for this. Thank you!
<apteryx>does 'mumi' support adding extra CCs ?
<apteryx>perhaps I could teach mumi to look that info from the commit message, compatible with 'patman' git trailers
<apteryx>e.g. Series-to: or Series-cc:
<civodul>i’ve just had to go through an unscheduled maintenance of overdrive1
<civodul>got the dreaded “Synchronous Exception” boot failure
<civodul>and following roptat’s guide on guix-sysadmin a few months ago, was able to reflash the firmware and get it up and running in 30mn
<civodul>quite unexpected i must say
<civodul>but here it is, building things again
<attila_lendvai>anyone from the python team? what is the current "pythony" way to start the unit tests? messing with is deprecated now, and i have a package (python-daemon) that fails some of its tests when it's started with `python3 -m setup test` (what guix does)
<attila_lendvai>civodul, wow, that sounds adventurous! i'm glad it worked out that quick!
<lilyp>attila_lendvai: there is no pythony way, you're supposed to ask your build system that
<lilyp>(e.g. tox, etc.)
<civodul>attila_lendvai: yes, that’s a happy surprise! it wasn’t my first time, which helps a lot
<civodul>i was afraid i’d spend the evening on it
<attila_lendvai>lilyp, too bad. upstream says i should invoke `make test`, which means each python package will have its own check phase...?
<lilyp>not really, we have infrastructure that covers most
<lilyp>"make test" is rather benign, but look at what make test does and then do that
<apteryx>attila_lendvai: use the pyproject-build-system; it already infers how to run test suite for many projects
<apteryx>e.g. run 'pytest'
<apteryx>by the way, adding support for Change-Id is very easy (see #66027); attila_lendvai, do you have an idea of how difficult it'd be to parse a git trailer from commit messages posted to mumi?
<peanuts>"Add a feature to automatically close specified bug reports upon committing."
<attila_lendvai>apteryx, i didn't know there are two python build systems. unfortunately, it's also "Using", even though the makefile mentions twine in the `make test` target.
<attila_lendvai>apteryx, re Change-Id, i think you're mixing me up with someone else (or my memory is even worse than i'm considering it to be... :) what is the context for this?
<snape>Hi, where can I see the builds triggered by sending emails to guix-patches ?
<snape>ok i found it
<snape>it says "Unable to apply patches" but git rebase worked
<aldum>>i didn't know there are two python build systems
<aldum>I think it's more than two
<apteryx>attila_lendvai: ah indeed, sorry about the noise :-)
<lilyp>aldum: for guix purposes, it's one and a half, though
<aldum>ah, ok
<minima>hi, i put this together with the idea of saving the output of a command to a file - but the resulting file is empty :)
<peanuts>"View paste AR6Q"
<minima>anyone knows what i might be doing wrong?
<boo_>Hi! I've been using autofs to mount my nfs share on my server before I switched it over to Guix, but I have no idea how to actually use autofs since I can't find any shepherd definition for it. Do I specify it in `file-systems` in my config? If so, how? Thanks!
<nckhexen>minima: It *does* work, and you'll feel a little silly (in a good way, hopefully) when you realise what's happening. :-)
<minima>ahah nckhexen - let me see...
<minima>(uneasy feeling already ahah)
<nckhexen>(What are you actually trying to list?)
<minima>nckhexen it's the output of a script, not really "ls"
<minima>but the "ls" example might be misleading?
<nckhexen>(1) Add "--version" to ls to reveal the answer (2) is you script producing stdout (as opposed to, say, stderr)?
<nckhexen>minima: Yes, because the build directory is empty ;-)
<nckhexen>No need for unease, it's an easy oversight to oversee.
<nckhexen>"-a" would've been a better answer, sorri.
<minima>nckhexen: hm (1) "--version" doesn't seem to make any diff; (2) yeah the script writes to stdout
<nckhexen>boo_: I don't see any mention of autofs in Guix System, at all.
<minima>"-a", let me try with that
<nckhexen>minima: Uh, are you sure you're testing this in isolation?
<nckhexen>Your code is definitely correct apart from the unfortunate choice of ls.
<boo_>nckhexen: Can I use some other kind of automount for nfs? Does shepherd support this use case?
<boo_>Basically what I need is for this to be mounted after NetworkManager has started, but before another service
<nckhexen>I don't think so, but I'm not a Shepherd expert.
<nckhexen>You could write your own service that uses the autofs package in Guix, to be clear, but I don't think there's integration, through (file-systems ...) or elsewhere.
<minima>ahah re ls, that was a bit disingenuous - there must be something else that i'm missing - "ls" "-a" also produces an empty file - maybe i'm not testing this properly
<nckhexen>I think so.
<boo_>Right now I have the filesystem (just plain nfs) in my `file-systems` definition, but that either renders the system unbootable, because it tries to mount before NetworkManager is started, or I allow errors and then have to restart the nfs mount after boot
<nckhexen>I'm on phone in car, so give me a sec, I'll try.
<nckhexen>ACTION not driving
<nckhexen>^ minima
<minima>i'm testing it as part of my home configuration script, maybe that's an important detail? many thanks for now, i'll brb
<minima>nckhexen: oh wow!! kudos for being so helpful even considering the context!
<vivien>You read IRC on your phone?
<Kolev>vivien, I read IRC on mobile too.
<nckhexen>Hell, I staff IRC on my phone (but not ofnet)
<nckhexen>Er, often.
<nckhexen>Punny typo.
<vivien>I don’t understand "I staff IRC on my phone"
<vivien>Do you mean, you employ people to respond to you?
<nckhexen>Uh, no, also what, but let's have lunch to discuss that proposal.
<nckhexen>I just meant network support stuff.
<nckhexen>Lots of special characters & arcane commands with no room for error, it's fun.
<vivien>ACTION is secretly chatgpt
<nckhexen>ACTION panics.
<vivien>I’m sorry, I’m just a bot, I cannot fix the issue.
<nckhexen>As a large language model, you are the bane of humanity.
<vivien>What’s certain is that bots are better at impersonating people than people at impersonating bots
<nckhexen>So... we should ask bots to impersonate humans impersonating bots and we'll know they're bots because they're good at it? That'll work.
<nckhexen>ACTION has banned a few GPTs from chat already, and it's only 2023 😕
<vivien>A few years ago we had ads for the latest cryptocurrency scam
<nckhexen>ACTION flashbacks.
<nckhexen>I think the latest cryptocurrency scam is still the previous cryptocurrency scam, namely cryptocurrency.
<attila_lendvai>nckhexen, is there a bigger scam than unbacked fiat money?
<attila_lendvai>not that i encourage a deeper divergence into offtopic land, but it's damaging to leave such statements unchallenged...
<nckhexen>It is undeed off-topic, I'm sorry.
<vivien>Is fiat money related to the auto industry?
<vivien>ACTION is going to be rich selling new conspiracy theories
<vivien>Now I realize that’s off-topic
<pkill9>hello,anyone around?
<attila_lendvai>pkill9, you just took the attention of hundreds of people in vain. if you have an actual question, then you can just start with that.
<pkill9>oh, i meant to post that in another channel lol
<minima>nckhexen: thanks for the snippet! it works! i had a parenthesis in the wrong place: (string-append #$coreutils "/bin/ls" "-al")
<minima>so that's a good starting point
<nckhexen>Good luck!
<lucypoo>just musing, a flatpak home service would be nice
<old>anybody know a self-hosting package for collaborative spreadsheets
<old>something like google spreasheets
<old>(not necessary packaged for Guix yet)
<Altadil>old: I think Nextcloud has a spreadsheet module, but I’ve never tried it.
<old>Altadil: but that's proprietary?
<Altadil>old: Nextcloud ? I’m pretty sure it’s free
<old>Altadil: Okay. I must have confuse it with another *cloud then
<lucypoo>yep nextcloud is pretty much foss self hosted services comparable to google~S
<old>I will check that thanks!
<troyfigiel>That Nextcloud module is based on Collabora Online, you can have a look there