IRC channel logs


back to list of logs

<Deltafire>is there an eta for an update from gnome 44? i think it would be part of 'core-updates'
<jmes`>Is there a way for a guix package to depend on one of a list of other packages? e.g. the same thing in Gentoo: DEPEND="|| ( app-misc/foo app-misc/bar )"
<jmes`>I'm trying to package emacs-flycheck-hledger and it requires emacs-hledger-mode or emacs-ledger-mode. Or in this case is it best to let it "fail" at run time if the user doesn't choose their own [h]ledger-mode?
<PuercoPop>jmes`: inputs are the way to mark dependencies afaiu. For example emacs-consult-eglot has emacs-eglot in the propagated-inputs
<jmes`>PuercoPop: Yeah I'm propagating emacs-flycheck as it's a definite runtime dependency of hledger-flycheck, but I'm not sure whether I should propogate the associated [h]ledger-mode(s). I guess I'll leave it up to the user for now.
<PuercoPop>jmes`: I'm not sure if there is a known pattern here, but more or less you are asking the user to make a choice here. YOu could do that by defining two packages, one for each dependency. ej. emacs-hledger-flymake-with-hledger-mode and emacs-hledger-flymake-with-ledger-mode. Or you could use expose a function that takes on of the two packages as its single argument and returns a package
<apteryx>daviid: thanks, the way their appimage is built may yield clues
<FlaminWalrus>Asked this before but at an inopportune time: =guix-home-service-type= is supposed to install the packages listed in the home configuration for the user indicated, right? It's not been doing that for me.
<FlaminWalrus>Shell file services ran just fine, but guix package -I is empty and none of the packages I requested are in the profiel.
<apteryx>daviid: I'll let you know if I can fix the one in Guix ;-)
<freakingpenguin>FlaminWalrus: Packages installed via $ guix install aren't part of your home profile if that's what you mean. $ guix package -I only shows "user-profile" packages ($HOME/.guix-profile), while $ guix home describe -I shows home profile packages ($HOME/.guix-home).
<freakingpenguin>guix-home-service-type definitely should be installing packages configured in guix home, but it has no idea about packages you installed manually.
<FlaminWalrus>freakingpenguin: aware. I have declared a system configuration in config.scm that includes a term of the somewhat new =guix-home-service-type=. This is advertised to allow you to declare and manage a =guix home= config from your system config, removing the need to run two commands. But it appears the packages in my =home-configuration= are not actually being instantiated in the home profile; this is a fresh install and I have manually
<FlaminWalrus>installed no packages.
<freakingpenguin>Mind sharing a pastebin?
<freakingpenguin>And to confirm, $ guix home describe -I shows nothing?
<FlaminWalrus>Yes on the latter; no software on the system under test obviously so I'm trying to remember how to debpaste with curl lol
<freakingpenguin>Hmm, actually guix home describe -I showing nothing might make sense since guix-home-service-type doesn't actually register new generations. Discussed here: The packages themselves are still supposed to be installed though.
<peanuts>"[PATCH] services: Add the Guix Home Service"
<freakingpenguin>FlaminWalrus: Haha, I feel ya. guix shell might help with that.
<FlaminWalrus>To further confirm, I did =guix pull= right before re-raising the issue
<peanuts>"dpaste/ZsK0M (Python)"
<freakingpenguin>Huh, something seems funky with (i'm guessing) the encoding because I get a weird error I've never seen before. error: #{\x200b;}#: unbound variable
<freakingpenguin>FlaminWalrus: I found with this stripped down config packages are installed as expected. (Had to shrink the config for the sake of my hands since copying led to the above error for some reason). guix system vm + log in as root + su -l dnw + emacs, worked fine.
<peanuts>"debian Pastezone"
<freakingpenguin>Man, I didn't have to go through all that. dpaste has a raw download option.
<freakingpenguin>FlaminWalrus: zsh prints an error while evaluating .zprofile, the "gpgconf" command is not found. This halts evaluating the rest of your zsh config meaning your path isn't updated for guix home.
<freakingpenguin>There's a few zsh warnings/errors printed so there might be some other things too.
<graywolf>Is there a way to validate a record as a whole? I see that I can use sanitizer to validate a field, however is there a way to validate record as a whole?
<fnat>I must have messed things up a bit with a recent system reconfig, as shepherd no longer wants to speak to me. E.g. 'herd status' will sit there without giving any output.
<fnat>A strace indicates there might be something wrong with my '/var/run/shepherd/socket', which I'm now investigating.
<fnat>(The file seems to be there though.)
<fnat>ACTION investigates harder. :)
<fnat>Hm... the file is there, it's a socket, it's owned by root:root, its file perms are 755.
<fnat>This would all look fine to me...
<graywolf>If I add user in an inherited operating-system, it is just ignored
<graywolf>(as in: (users (cons (user-account ...) (operating-system-users base-os)) )
<graywolf>Is that a bug or expected behavior?
<nikolar>Could it just be hanging
<nikolar>That might happen when the system clock changes
<Franciman>hi, i'm having many issues with nscd. Most of the time i find myself running "sudo herd invalidate nscd hosts", because nscd incorrectly resolves hostnames
<Franciman>afaiu it mostly happens when my internet connection is slow or brittle
<Franciman>is there a way to make nscd less stubborn?
<Franciman>ah by incorrectly resolve i mean that it says that an hostname resolves to
<Franciman>or other times it hangs
<tex_milan>Hello, anyone using virt-manager and getting 'NoneType' object has no attribute 'conn'? Gentoo says to install x11-themes/adwaita-icon-theme but I am running Guix SD and wayland. Any idea how to make it work?
<tex_milan>Nevermind, installed the theme and it works.
<cnx>Hi, I want to update Mesa to 24.1 (for the new Intel Xe driver
<cnx>how do I test it on my system?
<cnx>use pre-inst-env guix system reconfigure?
<freakingpenguin>graywolf: I've not encountered that behavior with users not being added.
<graywolf>Based on my debugging so far (it is not fun) it seems to do with the fact that I reify essential-services before inheriting from the operating system.
<graywolf>Ok, worked around. I think.
<graywolf>However this it *not* an intuitive thing to have to do:
<peanuts>"debian Pastezone"
<freakingpenguin>What's the motivation behind capturing os-essential-services vs. inheriting just os-user-services?
<graywolf>The essential services contain the account-service-type, which is responsible for creating the user accounts. Usually it is calculated via default value in (thunked) default field in operating-system. However if you modify the essential-services in base-os, then (operating-system (inherit base-os) (users (cons* (user-account ...) (operating-system-users base-os))) does not do anything.
<graywolf>Because the account-service-type was already calculated from the base-os, and the (users) field change in new os is ignored
<graywolf>Well that are 2 hours I am not getting back
<freakingpenguin>I follow, just was curious as to why do it like that?
<graywolf>I mean, if there is less retarded way, I am all ears :)
<graywolf>I will probably ping guix-devel for help here, the current work-around, well, works, but is horrible
<freakingpenguin>If the base os /needs/ to modify essential-services manually I'm not sure of a better way. If it doesn't then you should be able to add your user account + user-services and essential services would be instantiated with the new user.
<graywolf>It sadly needs to, the system is LXC-based VPS so I have to take out firmware-service-type and (service-kind %linux-bare-metal-service)
<freakingpenguin>Ah, I see what you mean. Basically what's going on in container-essential-services
<graywolf>Yeah, looks like it :/
<graywolf>Thanks for the sanity check, off to write the email...
<freakingpenguin>Best of luck. Sounds like an untrodden path.
<graywolf>I seem to often hit those. Not sure why.
<dmirth>Good morning. I'm trying to wrap my brain around the guix-home-service-type and I was wondering if anyone knew of anyone who had posted their file where they had successfully implemented that service?
<peanuts>"~freakingpenguin/rsent (master): rsent/machines/lan/pathfinder.scm - sourcehut git"
<dmirth>Thank you very much, freakingpenguin!
<dmirth>freakingpenguin: Just to be clear, instead of doing `guix system reconfigure pathfinder.scm` you do a `guix home reconfigure pathfinder.scm`, correct?
<freakingpenguin>dmirth: Nope, I do system reconfigure. guix-home-service-type basically registers a script that runs on startup and activates the given home environment.
<freakingpenguin>So since it's a system script and part of an operating-system record, you need to do system reconfigure.
<dmirth>So would that be run with sudo? (From the documentation, it looks as if it matters)
<freakingpenguin>As in do you need to run system reconfigure with sudo? Yes.
<dmirth>Yea, I guess thats right.
<dmirth>Thank you very much!
<dmirth>freakingpenguin: Do you mind if I look at the rest of your config files on
<freakingpenguin>Haha, I wouldn't post it publicly if I minded!
<Guest44>halt and reboot are now in my .guix-home directory.  It seems that shepherd is in my home profile but I don't know why.  It was /run/current-system/profile/sbin/halt.  Does someone may know why?
<Guest44>I don't have a problem with it, I just want to know why it changed because I can't resolve why
<freakingpenguin>Guest44: Guix Home uses a user level shepherd to run services when you log in (run not as root $ herd status). I imagine that's where it's coming from.
<Guest44>Yeah maybe they changed something.
<freakingpenguin>If you didn't have any home shepherd services before and now you do, that'd be my guess.
<ngz>sneek: later tell cbaines I noticed that bordeaux has started building two "tex-team" packages with a priority of `#f', which means they are stuck and take up a slot for nothing. There is one in "hatysa" and one in "milano-guix-1". I don't know if that's a benign artefact or a bug.
<sneek>cbaines, you have 1 message!
<sneek>cbaines, ngz says: I noticed that bordeaux has started building two "tex-team" packages with a priority of `#f', which means they are stuck and take up a slot for nothing. There is one in "hatysa" and one in "milano-guix-1". I don't know if that's a benign artefact or a bug.
<cbaines>The priority of #f is probably an interface bug
<cbaines>The tex-team builds started as QA was trying to build the top two branches, but I've changed that now so it's just building the first branch
<cbaines>The logic for when to build additional branches needs to be smarter
<ngz>IMO, I think it would be nice to have least spare a tiny priority to other branches. It helps finding bugs early.
<ngz>cbaines: But my initial report was more: "should these two aborted builds still appear on".
<peanuts>"Activity Build farm"
<graywolf>Does the fact that essential-services and services are thunked mean that they are re-calculated on every access? So if my (services ...) is somewhat compute-expensive, reconfigure/deploy could get quite slow?
<tex_milan>Hello, did anyone get virtiofs working with virt-manager?
<freakingpenguin>I have once again underestimated how long qemu riscv builds can take when substitutes don't exist
<freakingpenguin>My poor server has been maxed for 24 hours