IRC channel logs

<apteryx>ieure: I struggle with mpd as a system service before. Gave up with it.

<apteryx>doesn't seem to be designed to be used that way, at least not for a system where users login and also use the audio card

<apteryx>log in*

<apteryx>a home service makes more sense, I guess.

<ieure>apteryx, I think I have it working, but it's taking nearly as long to `guix system reconfigure' with my local guix changes as it did to figure out the problem. :/

<ieure>apteryx, It depends, I have a dedicated headless system for running MPD hooked up to a USB DAC, so I want mpd to run when it boots and don't use pulse/etc.

<ieure>For running on a desktop type system, yeah, it makes more sense as a home service.

<apteryx>maybe if it supports pipewire this could make things smoother?

<apteryx>(on a desktop type system, to better share the audio hardware)

<ieure>apteryx, Yeah, for a machine you're logged into directly, have a GUI, etc., a home service makes more sense.

<ieure>System service is what I want, though. Needs to start immediately on boot.

<ieure>Okay, that didn't even work. How can I test a Guix fork? This is way harder than I thought.

<ieure>Tried `guix time-machine -C mpd-channels.scm -- shell -- sudo guix system reconfigure ...', that didn't work, I got a generation that didn't use the commit in mpd-channels.scm

<flurando>hi, how to write the match-criteria in home-openssh-service-type?

<flurando>the document is unclear, and it kept saying expect an exact integer

<FuncProgLinux>It's possible that the port section is the one failing there.

<flurando>well, I checked the src of openssh service, we need "host ..." stuff

<flurando>stange, still complains about expecting exact integer in match-criteria?

<flurando>any idea? I am suspecting there is a bug here

<flurando>for this (openssh-host (match-criteria "all")), it complains gnu/home/services/ssh.scm:201:21: In procedure match-criteria?:

<flurando>Wrong type (expecting exact integer): #f

<flurando>when the match-criteria? clearly returns #f when str is "all"!

<ieure>flurando, I can't reproduce: scheme@(gnu home services ssh)> (match-criteria? "all")

<ieure>$6 = #t

<flurando>strange

<ieure>flurando, Can you put your config in a pastebin?

<flurando>yeah, I am opening a pastebin for the home-openssh-service-type part

<flurando>wait a second

<flurando>well, there is no need anymore. It works, I just forgot to tangle the emacs block after I changed to "all" the last time. Looks like I was spamming, sorry.

<ieure>Aha! Glad you found it.

<flurando>:)

<FuncProgLinux>:/ latest mate-panel won't have some users very happy with a certain applet lol

<flurando>true in guix now, you just have to reconfigure

<Aurora_v_kosmose>Does the unprivileged daemon use secondary users for builds?

<Aurora_v_kosmose>Hm, I think it doesn't, if I get the docs right.

<lilyp>nope, it uses the unprivileged guix-daemon user

<lilyp>that's like the whole point of the unprivileged daemon: it never has any permissions it would need to drop

<Aurora_v_kosmose>I see.

<cbaines>I'm looking at migrating consensus.guix.gnu.org, packages.guix.gnu.org and 10years.guix.gnu.org over to the Hetzner web server this evening...

<Hamled>I want to do some development on the source for a package, and use tooling to help with that. Since this project uses the ninja build system I can apparently get it to generate the compilation command JSON file which my C language server tool wants to use. But I will need to have the project configured the same way as it is when built as a guix package -- is there an easy way to get an environment

<Hamled>where the configuration step of the package build process has already run (or can be run from a command)? I think I have done this before with Nix

<Rutherther>you can add a phase that will just (exit 1) and use --keep-failed

<Hamled>Oh good idea thanks

<civodul>cbaines: yay, thank you!

<cbaines>I can see traffic coming through in the logs now, I think things are working

<cbaines>I missed some configuration relating to consensus.guix.gnu.org, but I think those 3 sites are migrated over successfully, do let me know if you notice any issues

<Guest92>Why do I get a conflict about lua when trying to install mpv and celestia? I thought Guix could handle different versions of one package installed at the same time.

<ekaitz>Guest92: when they are propagated dependencies that's not possible

<ekaitz>if two packages propagate lua, which one should run when you execute `lua`?

<Rutherther>you cannot have different versions of the same package installed in one profile. You can definitely have them in the store. And you can definitely use them simulatneously, out of different profiles, ie. through guix shell

<Rutherther>ekaitz: the right one, of course :P

<ekaitz>hehe

<Rutherther>here it's kind of unfortunate how many packages mpv propagates

<ekaitz>Rutherther: can we do something about that? maybe hide them and wrap?

<Rutherther>I don't think there is much to wrap here, they are propagated because they are Required in pc file. This is important only for packages using mpv as a library

<Rutherther>it's not important for users installing mpv as an application

<ekaitz>:S

<Guest92>Why are packages propagated? Can't we install lua explicitly if we want to run it?

<identity>Guest92: but do you want to install 50 packages because they are not propagated, but required for some program to run?

<Rutherther>Guest92: there are multiple reasons, but generally because something requires them to be, be it lack of enough work on the package to do it other way round or some constraints that make it really hard to do otherwise, ie. for python libraries and their dependencies, they are just propagated and it would be very hard to do it other way. That's because of how python works - just loads everything from a single PYTHONPATH, so it's not really possible to...

<Rutherther>... mix them anyway

<Guest92>So, a package that is propagated is a package that is not installed explicitly and is present in PATH?

<Rutherther>when installing mpv that propagates, among other things, lua, you get mpv and lua installed. As in, propagated packages are installed the same way as if you did install lua as well. That's what the propagation means

<Guest92>Rutherther You mean, not in /gnu/store/...-mpv-0.40.0 ?

<Rutherther>if we are talking about guix install, then I mean that both packages are put to ~/.guix-profile that is a union of all the packages that are installed via guix install

<Guest92>So the problem is just where does /usr/bin/lua under the profile points to?

<Rutherther>the problem is that there are two different versions of lua being put to ~/.guix-profile, but only one can be put there as there cannot be two ~/.guix-profile/bin/lua and so on

<Guest92>In theory, each package could define their own filesystem overlay over the profile, so that both mpv and celestia could each see only their own exclusive lua. But I think I am entering "containers" territory here...

<Rutherther>that's completely unnecessary as most of the time the propagations can be gotten rid of in much easier ways

<Rutherther>but no one has gotten rid of them in the packages you are using

<Rutherther>...yet

<Guest92>It would probably be easier to patch the path to lua on the code of the program or do some other environment tricks.

<Rutherther>yes

<Guest92>I see. I have so much to learn yet.

<Guest92>Thank you.

<gabber>do we have something like a patch-elf phase in proper guix? i have a large-ish package at hand that creates its own shared object library files and needs some patching before it is installed

<simendsjo>Hmm.. Setting up a new system, and I get an empty module-import.drv on system reconfigure.

<simendsjo>I can do system build, system container, but not system reconfigure. The last time I got an empty file, I had to chroot and recreate my system :/

IRC channel logs

2025-11-09.log