<morgansmith>the linter yells at me now because I want to leave the test native inputs in as comments, but that makes the parens after the comment lonely :/ <cbaines>morgansmith, I wouldn't comment out the native inputs used for the tests, just leave them uncommented <morgansmith>Nah, nckx was very mean last time I tried to put sleeper code that doesn't do anything into the codebase. I think I agree with him. <vagrantcish>f08587682a631d3fe30159af838c6766dd65205b gnu: guix: Update to commit 5918cb5. .... seems to make me unable to build guix <nckx>That's weird. It should have been reverted. <morgansmith>I did a guix pull when you guys said it was failing, and it failed for me. Then when it was fixed, I did it again and it succedded <vagrantcish>ah, must have tried pulling just before it was reverted <halfdann>Hey all. I want to try guix to package a clojure/java project. I assume a guix package must be created for every maven dependency. Then somehow the packages must be mapped to maven so projects can depend on them. But I dont see that being done anywhere... Are there examples of packaging a maven project? <cr4zyg3n3>i tried installing guix, but in the gui install it get's to the wifi and doesn't detect it and then stops what solutions are there? it is a dv5 <morgansmith>Wifi is a known issue in the installer. It was a big hurdle for me to get over and at that point I thought it was a big enough deal to launch another version. Today, even more so. Wifi works after install so if you have access to ethernet, that's the easiest solution <morgansmith>I got around the issue by doing the "manual installation for wizards" as described in the manual <cr4zyg3n3>@morgansmith than you. I don't have the wifi and when I try to continue it resets and start from the beginning. I will try pluging in the wifi to usb and see if that works. <cr4zyg3n3>if that doesn't work I will try the manual install for wizards. thanks! <morgansmith>I don't think there is a way to setup wifi using the GUI regardless of your wifi setup. I think you can setup wifi using a terminal (just follow the first bit of the manual install) and then finish by using the GUI ***qyliss is now known as qyliss|illumos
***qyliss|illumos is now known as qyliss
***catonano_ is now known as catonano
<xelxebar>Man, newsboat is like the package from hell that always updates but never has substitutes and also takes longer than the entire Cambrian period to build. <jgart[m]>I was reading this post from 2012 from edolstra about nix not being "complete until it has static typing". Would it be too much work to add static typing to guix/guile? Why would it be a bad idea? Why would it be a good idea? https://github.com/NixOS/nix/issues/14 <sneek>jgart[m], you have 1 message! <sneek>jgart[m], raghavgururajan says: : Ate yuo a vailable at yor JID? <xelxebar>How can I go about checking what jobs my build offload machines are processing? <vits-test>the dirs in /run/user/UID are created if elogind-service is enabled, but the "error: connect: /run/user/1000/shepherd/socket: No such file or directory" is due that there is no `shepherd` instance running for Ur user. <Kimapr[m]>if you run herd without sudo it will try to control your user instance. so unless you run shepherd as your user and won't to control it instead of the system-wide instance you have to run herd with sudo <helaoban>vits-test: I think I've found the problem. I use the mosh shell when connecting to remote machines. Mosh starts the connection by a initiating a short-lived ssh session to launch mosh-server on the remote machine, before switching over to UDP. When the SSH session is closed, the PAM session that was started by the SSH login is also closed, and elogind removes the user runtime directory <helaoban>everything works when I connect using regular ssh and not mosh. <helaoban>The temporary work-around for me right now is to ssh into localhost as the first thing I do after I login via mosh. <helaoban>this was particularly troublesome because elogind only unmounts /run/user/uid if all PAM sessions for a user are closed. Sometimes I have a several connections to a remote machine open in different terminal windows, sometimes running plain SSH, so I was only seeing this problem once in a while. <helaoban>I'm interested here specifically in running shepherd under my user. <Brendan[m]1>Installer just boots to a black screen with nothing on tty1. i can switch to tty2 with the documentation though. is there meant to be the ncurses installer on tty1? <helaoban>Actually I don't think this has anything to do with pam, but rather specifically elogind behavior. <helaoban>If I login to my remote guix machine using mosh with 'mosh helaoban@remotemachine' and then run 'sudo loginctl list', 'helaoaban' is not listed, only 'root'. <Brendan[m]1>i tested loading the trisquel installer and that loaded to desktop fine <leoprikler>yeah, radeon does not work nicely with guix afaik <Brendan[m]1>yeah ill get the error message soon once i can copy it to a usb <leoprikler>if you can get to tty3 you can do anything really ***nlyy is now known as nly
<leoprikler>Alternatively, you could try to debpaste your error <leoprikler>The installer should come with facilities to enable network connection. <Brendan[m]1>im assuming the clock being set to 2007 isn't a big deal? <leoprikler>2007 seems to be the year this mainboard was built AFAIC <Brendan[m]1>i think there is a little battery on the motherboard that keeps the clock alive isnt there? maybe it finally died <Brendan[m]1>i found this laptop on the side of the street. it looked like it was stolen by teenagers and they kept the hard drive and dvd drive <Brendan[m]1>I also get a warning in early boot guile. pthread_getattr_np or pthread_attr_getstack failed for main thread <Brendan[m]1>with nomodeset, i managed to get do the installer, yay <Brendan[m]1>Some of the languages are written in english, others are written in the native language <mroh>hmm, geiser bump broke M-x guix. <bdju>anyone here using syncthing? how should I make it run in the background? I ran syncthing-gtk which seemed to start the daemon, but after closing it, my phone doesn't see my pc anymore, so I think it didn't keep the daemon running <bdju>hm there's a setting to make it keep the daemon running on exit instead of terminating it, but I still wonder if a user service would be better <leoprikler>I'm not quite sure how the GTK side handles things, but you could try writing a shepherd service for it <bdju>it just starts the daemon if it's not already running, and then whether it kills it or not on close is configurable. so if I have a shepherd service it should work fine with the gtk thing <bdju>I've never made a shepherd service before, though <Brendan[m]1>just run the shepherd command. I added it to my .bash_profile <bdju>thank you, I'll look at that then <rekado_>jgart[m]: why would it be a good idea for Guix? <rekado_>the nix language is a language with only one purpose, and it appears to be modelled after Haskell (which suggests static types). Guile is a general purpose language. <rekado_>I for one have not missed static types in Guix, except for monadic functions. <rekado_>I don’t see what types would bring to the table that would be beneficial for Guix <Brendan[m]1>i lightly bumped my laptop and it caused the USB to cut out fail the installer after a couple hours :/ i was almost done. <leoprikler>Static type checking in Guile/Guix would be super nice. You could check at compile time, that a list of packages is indeed a list of packages, and not e.g. a list of packages with a random thunk thrown into the mix. <Brendan[m]1>ive heard it can improve error messages and editor integration. <ani>Outreachy Intern here. I installed GUIX and now I am going through contribution section from manual what else shall I do <asterope>Just noticed that stumpwm's info pages are incomplete, its 'install-manual step has a "we can get away with much less" in the comment <asterope>I'll look at the commits to see why it was done this way, but it's weird anyway <nckx>asterope: That sounds very, very much like something I'd write. It's quite possible it's no longer true. <nckx>asterope: Can you give a concrete example of the incompleteness? <nckx>Aha, the code snippets are fucked up. <asterope>It looks like it's the work of Pierre Neidhardt, along with adding the whole stumwm package <nckx>I think they just moved it to a new file. <asterope>nckx: The variable and function index are empty. Some nodes have "scrambled" function/variable names <nckx>Yep, the !!! / ### stuff. <asterope>I had to use the docs on the web, but it's not ergonomic <mfg>does someone know why flycheck-guile always complains that it cannot create a path for the auto-compiled file in procedure compile-file? Do i need some additional configuration? <numerobis>Hi #guix! Does anyone use the tailon service? With only the line (service tailon-service-type) in my config file, I obtain a "500 internal server error" on visiting http://localhost:8080 <cbaines>numerobis, I don't at least unfortunately. In terms of debugging that error, I'd suggest running tailon from the command line <cbaines>if you look at the command that's being run, then run it with sudo -u tailon ... you might get more information on what the problem is <cbaines>unfortunately the service doesn't handle logging very well as far as I can see... <numerobis>cbaines: the cause of the error was that I was already using the port 8080, duh! Thanks! :) <cbaines>numerobis, ah, OK, glad you figured it out :) <nckx>asterope: Is it complete now. <vits-test>`nft` not working on arm64-generic kernel. Worth to open a bug? <nckx>Of course, provided you've made sure the module (if it's modular) is actually loaded & other such basic troubleshooting. <asterope>Found the commit, it was like it from the start <vits-test>nckx: yes, the CONFIG_NF_TABLES "is not set" in my /run/current-system/kernel/.config <asterope>nckx: ah, you've changed it. I'm checkit it... <nckx>vits-test: Right, the -generic one is the ‘shitty upstream defaults’ one, I always forget that. I don't think that's a/our bug to fix, sorry. NF_TABLES is set in all standard kernels: https://paste.debian.net/plain/1166603 <nckx>If you're unable to use the normal (linux-libre[-X.Y]) kernels on your hardware, I think that should be fixed instead. <nckx>Sorry for the -generic confusion. It's still not quite clear to me why that exists but I'm not an ARM expert. <nckx>It should really be called -noconfig or -linuseswhims. <vits-test>nckx: fun with defconf is that the conservative, schedutil, and ondemand cpufreq-governors all has the performance governor as a fallback. But performance the "just set all to maximum freq" is default one. <rekado_>mfg: check the permissions on ~/.cache/guile/ccache (or delete it) <nckx>vits-test: ‘performance == peg @ maximum freq’ hasn't been true on x86_64 for a long while; it just means ‘get out of the way of the CPU’. In reality it probably results in the *most* complex algorithm being used. Maybe it's still true on ARM. <nckx>What do you mean by fallback? CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE? <ani>rekado_ : I have installed guix what next shall I do? <rekado_>ani: have you set up a development environment yet? The Contributing section in the manual tells you how. <ani>rekado_ : also when I do guix install emacs it doesnt install it <mfg>rekado_: thanks a bunch :D <roptat>ani, why not? is there an error message? <roptat>or do you not have ~/.guix-profile/bin in your $PATH? ***rekado_ is now known as rekado
<clone11>Does anyone run mpd as its own user? I can't get it to talk to pulseaudio when I do <leoprikler>There is quite some setup involved if you want to do something like that. <leoprikler>I think canonical docs have mpd talking to pulse "over the network" <nckx>clone11: That's not unexpected: if pulse is running as its own user it shouldn't be able to talk to your PA daemon by default. <leoprikler>you need to configure pulseaudio-service-type to load the native-protocol-tcp module and configure mpd to talk to pulse over tcp <nckx>leoprikler: Is there a hack similar to chmod +? /run/user/$UID/pulse/... && MUH_PULSE_SOCKET=/run/user/blah mpd? <nckx>Or is that $UID hard-coded somewhere. <leoprikler>I'm not sure if there is, but this is something that should be solvable using config files alone. <leoprikler>IOW by writing some different things in /etc/config.scm <nckx>IOW the Turing-complete file that can do anything & more, but I see the point you were trying to make. However, should doesn't mean can. <leoprikler>If I had a use case for it, I would write a mini blog post on how to do it. <clone11>Sorry stepped away. But yeah I did switch it to native protocol and tcp. Here's a paste with what I put into my config.scm. http://0x0.st/iGKG.txt <leoprikler>of course it would be nicer if the script-file was purely in scheme, but that's cosmetical <clone11>When I do that mpd launches on mpd user but when I do mpc play it gives an error about not being able to connect to pulse. Let me pull up the exact error real quick <leoprikler>Your pulseaudio is still using the old config file <leoprikler>(this should no longer be a problem on your next login or reboot) <clone11>Hmm so applying the config, manually loading the tcp module, and restarting mpd, it worked, but after rebooting with the new config, I get "ERROR: Failed to open audio output" in mpc, even if I manually load the module in pacmd and restart mpd. <leoprikler>Okay, first things first, you do actually have pulseaudio running after the reboot, right? <clone11>Yeah, and applications running on my user work with it <leoprikler>`pacmd list-modules` also lists the tcp module, right? <leoprikler>(If push comes to shove, try `pulseaudio -k; pulseaudio -vvvv` then mpc play and check for errors in pulseaudios output) <clone11>I'm not sure where to find pulse or mpd logs. I didn't see anything glancing through /var/log. The only output I have is the error mpc gives me (ERROR: Failed to open audio output). I don't think I can kill pulseaudio since, I tried pulseaudio -k as my user and root but get "Failed to kill daemon: no such process" while my sound still works. <leoprikler>what is the output of `ps aux | grep pulseaudio`? <clone11>gdm 740 0.0 0.0 830360 10972 ? Sl 10:32 0:00 /gnu/store/19wcjfwdr3hzq1a4wcpld1zdrjichck0-pulseaudio-13.0/bin/pulseaudio --start --log-target=syslog <clone11>gdm 744 0.0 0.0 235532 7392 ? Sl 10:32 0:00 /gnu/store/19wcjfwdr3hzq1a4wcpld1zdrjichck0-pulseaudio-13.0/libexec/pulse/gsettings-helper <clone11>clone 3628 0.0 0.0 830404 12976 ? Sl 10:48 0:00 /gnu/store/19wcjfwdr3hzq1a4wcpld1zdrjichck0-pulseaudio-13.0/bin/pulseaudio --start --log-target=syslog <clone11>clone 3660 0.0 0.0 235228 7252 ? Sl 10:48 0:00 /gnu/store/19wcjfwdr3hzq1a4wcpld1zdrjichck0-pulseaudio-13.0/libexec/pulse/gsettings-helper <clone11>It momentarily kills it and then it comes right back up. <clone11>Isn't shepherd what starts services? <leoprikler>in the case of pulseaudio, it restarts itself, so it's a race ;) <leoprikler>if you don't like races, you can also disable autospawn through the config, but that would be another reboot <ArneBab>Does anyone else get direct segfaults with winecfg? <clone11>So after some testing I found that mpd works fine when pulse is ran with pulseaudio -vvvv but not when the system launches it. <clone11>not very helpful for getting log info <leoprikler>and if you kill that and let the system run it again? <clone11>Actually I just restarted mpd after killing pulseaudio -vvvv and it works. I just tested restarting mpd on a clean reboot and it didn't work <lafrenierejm>Does anyone have examples of how to write a recipe for a Python (PyPi) package with optional dependencies? Is there a way to do it similar to how Git is split? <leoprikler>if there's a need to split them, have one with a minimal set of inputs as python-some-package-minimal and inherit python-some-package from it <lafrenierejm>leoprikler: How should one determine whether it needs to be separated? <leoprikler>If another package needs it to build and pulling in everything would just add unnecessary bloat to the closure <leoprikler>As an example: git is used in a lot of packages and therefore keeping a git-minimal makes sense. <leoprikler>Same with emacs-minimal, as it's used for byte-compilation <lafrenierejm>This package would be included as a propogated input for anything that depends on it. <leoprikler>keeping a youtube-dl-minimal on the other hand would not make much sense. <lafrenierejm>I'll read through the package source to determine what those optional dependencies are actually used for. <leoprikler>For more context, what are you trying to package? <leoprikler>correct me if I'm wrong, but grandalf doesn't look like it has much optional stuff, does it? <leoprikler>I'd say for now go for the full package with numpy and ply. <raghavgururajan>ryanprior: Nah! I meant to say that I am running to #guix screaming guix. *raghavgururajan orders food via SkipTheDishes, under the name 'Guix User'. <mfg>is it possible that the pypi-uri builder sometimes builds the wrong uri? <mihi>has anyone been able to run "guix pull" under any Hurd system, being it the VM image, a child hurd or another Hurd distro with guix as package manager? The error I'm getting seems to be coming from libgit2, and I would think it is not caused by libgit2 incompatibility and not by anything I'm doing wrong. <mihi>Have tried several times, different things, each time starting from a fresh VM image; and the result is always "Git error: invalid version 0 on git_proxy_options" <mihi>s/it is not caused/it is caused/g <Gooberpatrol66>I made an etc-service-type that adds an /etc/nanorc file, but nano seems to ignore it. Is there a way to get nano to notice the file? <mfg>which of the Qt Quick related packages actually contain the Qt Quick Module? <mfg>i guess it's qt declarative? the synpsis says it's Quick 2, does it make a difference? <mfg>i get lots of errors that modules like QtQuick.Controls.Styles are not installed even though i think i added the correct input packages... <leoprikler>Gooberpatrol66: you probably want that nanorc in /etc/skel <leoprikler>Guix patches its software so that stuff that is read from /etc is usually read from package-prefix/etc <GNUtoo>Are there more substitute servers for ARM than the default ones (berlin.guix.gnu.org.pub ci.guix.gnu.org.pub ci.guix.info.pub) <GNUtoo>(just to do a basic image on ARM I needed to install guix on an ARM machine, and it seems to build a lot of stuff, including qemu-minimal for isntance) <GNUtoo>(It's been compiling for days on an I.MX6Q with 2G of RAM) <GNUtoo>Hopefuly It now uses the 4 cores to compile <Gooberpatrol66>leoprikler: Do you have more information about /etc/skel? I'm not really finding anything. <OriansJ>leoprikler: they were cursed the day they lost the Uli purchase to ATi <OriansJ>which is why AMD merged with ATi instead of nVidia <OriansJ>leoprikler: different sort of curse though <OriansJ>AMD always has the Via escape route thanks to the global foundries spinoff deal but nVidia got stuck in a grow or go bankrupt cycle <leoprikler>Everything is stuck in a grow or go bankrupt cycle, that's just capitalism. <OriansJ>leoprikler: no, there are businesses that reach equilibrium with the market demand and just enter a cycle of slow refinements. You see it in businesses that are more than 200 years old. <leoprikler>I can't recall a single business that has existed since the dawn of capitalism for a purpose other than pleasing its shareholders. <OriansJ>leoprikler: capitalism doesn't require shareholders <leoprikler>Well, technically you're right, but… I feel like you mean this in the opposite sense of the way that would be correct. <OriansJ>leoprikler: perhaps you are just using a corrupted definition of capitialism; could you define what you mean when you capitialism without pointing to examples? <leoprikler>capitalism: an economic system characterized by private or corporate ownership of capital goods, by investments that are determined by private decision, and by prices, production, and the distribution of goods that are determined mainly by competition in a "free market" <leoprikler>not sure what point you're trying to score here tbh <leoprikler>If you're going to "commies had markets", then yeah, so did feudal lords. <OriansJ>leoprikler: yes thus it by itself can't be used to define a distinct economic system <leoprikler>Yeah, sure, but "free market" are literally two words in this definition and I've put them under scare quotes to indicate how bullshit this part of it is. <leoprikler>where as "private ownership of the means of production" is what it's actually about and "investments determined by private decision" is literally talking about shareholders <OriansJ>leoprikler: here is my definition: "A system of resource distribution where individuals are reworded equally to the value they provide to others" ***terpri_ is now known as terpri
<OriansJ>Now mind you, no society yet actually meets that definition but I can see the natural attraction to it <leoprikler>The word you're searching for is "meritocracy", which capitalism is not. <OriansJ>leoprikler: that is because money makes for a poor tracker of merit <OriansJ>Does the man who borrows $10M and places 2 phone calls more meritious than a woman who spends 80 hours a week helping children? <apteryx>civodul: any clue as to why 'guix build guix --with-git-url=guix=file://$PWD' can't work? It seems to introduce a cycle. <apteryx>GNUtoo: if you don't specify --cores, it'll use them all. <apteryx>have you considered using the qemu-binfmt integration with the guix-daemon? It allows to emulate ARM and other architectures on a beefy machine if you have one. <GNUtoo>apteryx: how do I do that? my host distro has already qemu-binfmt on x86 <apteryx>ah, you need to be on a Guix System to have that integrated nicely in Guix with the qemu-binfmt-service-type service. <apteryx>I think Mathieu experimented with some success on foreign distros but it's a bit of manual fiddlig to expose the qemu entries in the build chroot correctly, IIRC. <apteryx>I wouldn't recommend it unless you have a lot of time on your hands <apteryx>with the guix integration, it's a matter of doing 'guix system build --system=armhf-linux your-system.scm' <apteryx>then it builds all the packages for ARM as if it was done on natively (no cross-compiling), via emulation. *GNUtoo tried with target but one build system didn't support cross compilation, so I'll probably just wait for the build on ARM to finish at this point <apteryx>OK! I hope it succeeds! Sometimes the lack of substitutes can mean the package failed to build on the CI <GNUtoo>Is there a way to only use substitutes? <apteryx>you can use 'guix weather' to get a feel of how much substitutes coverage you'd get <apteryx>usually using a day couple hours or a day old guix commit will give you more substitutes <leoprikler>OriansJ: it never was, but that also invalidates your "definition" <OriansJ>leoprikler: how? I never implied merit in the definition. <leoprikler>fair enough, then what do you label the "value you provide others" <OriansJ>leoprikler: a metric the "others" have to determine for themselves <nckx>Good Randian evening, Guix. <leoprikler>I would need to get hammered, but then I'd break a 25 year long record of not drinking. <nckx>sneek: Randian botsnack that you don't deserve. <civodul>apteryx: no idea; is that a new issue? <civodul>i've used that with the 'guix-daemon' package a few days ago <OriansJ>nckx: atlas forgot to mention every attempt at a Randian utopia ended up a warlord hellhole in less than 2 years; over thrown ultimately by any organization with a functional government. <civodul>apteryx: always run "guix build -S guix --check" after "make update-guix-package" :-) <civodul>as written in update-guix-package.scm, it's designed to allow you to refer to private commits <civodul>but you'd rather be careful if that's not the intent <civodul>("guix build guix --with-git-url=guix=$PWD -n" works for me) <apteryx>so perhaps I've broken something with the above change. Will test, thanks. <apteryx>Referring to private comimts is fine, what is problematic is that it uses a potentially dirty checkout to compute its hash <OriansJ>nckx: that which is true on individual scales, does necessarily not hold true at larger scales <civodul>apteryx: yeah, one has to check anyway <apteryx>dirty checkout sounds scandalous but it's a trivial situation to fall into without really being aware, if you've ever used 'git update-index --skip-worktree' for example. <civodul>but maybe we could have the script perform that check by default <mfg>i'm trying to package cura and when i start it it crashes with a segfault - reason is: futex(0x209ed70, FUTEX_WAIT_PRIVATE, 0, NULLFatal Python error: Segmentation fault. Does this seem like a python bug or what does it mean? <apteryx>that's what dannym suggested; I suggest to go a step farther and have a pristine tree recreated from the referred commit. Using git worktree that's easy and cheap. <civodul>sounds more difficult to do (does guile-git have worktree bindings?) <civodul>but anyway, as long as the default is to check, we'll have made progress :-) <nckx>OriansJ: Sure, we agree, although I don't pretend to know what AS did or didn't mention. I've tried three times to finish it and just can't. It's. So. Badly. Wr*goes to do something fun* <OriansJ>nckx: well you know what they say; "There are two novels that can change a bookish fourteen-year old’s life: The Lord of the Rings and Atlas Shrugged. One is a childish fantasy that often engenders a lifelong obsession with its unbelievable heroes, leading to an emotionally stunted, socially crippled adulthood, unable to deal with the real world. The other, of course, involves orcs." <civodul>apteryx: oh! i don't see a patch there, maybe it's just mumi keeping it for itself? <alextee[m]>does guix collect any stats on # of installations for packages? something like debian's popcon <roptat>alright, I can encode/decode ipv4 and ipv6 addresses, I only need to understand how to make the request properly, and guile-netlink will be able to set an IP address on a link <apteryx>civodul: seems to be a MUMI issue, yes. I can see the patch I sent in M-x debbugs-gnu <apteryx>civodul: oh, it seems to work now ./pre-inst-env guix build guix --with-git-url=guix=$PWD <apteryx>not sure why it seemed to hang earlier <kmicu>jgart[m]: it’s not even difficult to add it (whatever approach we choose e.g. solutions from Typed Clojure, Typed Scheme, Typed Racket) but optional and gradual type systems have big costs and little benefits. <civodul>apteryx: probably the initial clone took too a long time, and since there's no progress report, you assumed it was stuck <str1ngs>sneek: later tell guix-vits. I'd rather keep %search-provider-format and %search-providers in (nomad web) <joshuaBPMan>OriansJ Have you read Beren and Lutherien? I read that a few days ago, and it was possbibly the most beautiful short story I have ever read. <roptat>mh... I don't get it, I'm asking for every IPv6 address I have, and I only get ::1 for lo <roptat>and I'm trying to use strace on iproute2, but it looks like we do the same thing... <roptat>ah! I actually receive an array, but I look at only the first element ^^' <clone11>Is there a way to declare user packages in an operating-system declaration? <roptat>I'm not really sure how I can find out I receive multiple messages <roptat>clone11, I don't think so, you can only specify system packages <roptat>you can use manifests to have a declarative interface for user packages <erkin>I got Guix running in a VM and I'm trying to configure it but I can't for the life of me figure out where the `operating-system' value is kept in. <rekado>erkin: what do you mean ‘where the value is kept in’? <erkin>In which file is the default kept so that I can edit it? <rekado>did you build it yourself or download a pre-built VM image from the website? <erkin>It's the prebuilt QCOW2 image. <rekado>the configuration file is an input, so if it is contained in the output at all it will only have been copied there <erkin>I want to change the keyboard layout, disable X.org and enable SSH. <erkin>Where does it look for it at startup? <rekado>I don’t know if the VMs come with a service that stores the configuration file somewhere in the file system <rekado>it’s only an *input* at build time <erkin>But how does it know it should start Xfce then? D: <rekado>once you’ve built a system using that input it’s no longer needed <roptat>erkin, maybe /run/current-system/configuration.scm <erkin>Oh goodness it's going to be really hard to write a config file from scratch in an unfamiliar layout. <rekado>roptat: oh, that’s new-ish… right? <roptat>yeah, not sure if it was present in 1.1.0 <roptat>do you have a /run/current/system/provenance file? <rekado>I remember that we had a discussion on the list ~ 1 year ago about whether to include the config file via a service <roptat>otherwise the file that generated your VM is also present in the guix repository <erkin>boot etc initrd kernel locale parameters profile <roptat>that should be explained in the manual, let me find it <erkin>/gnu/store/*-guix-*/share/guile/site/2.2/gnu/services/configuration.scm <roptat>I don't think that's what you're looking for, it sounds like a shepherd configuration file maybe <erkin>It still baffles me that it knows which services to start without a config file. <roptat>that's because the config file is used by guix to generate the actual system config <roptat>but at runtime, it doesn't need it <roptat>it's like your system is a package: once it's built, it doesn't need the recipe anymore, it just runs :) <erkin>If I leave a field undefined in `operating-system', does it leave the previous value unaltered or does it overwrite it with blank/default? <roptat>like a package: if you change the recipe, it doesn't care what it was before <roptat>so to predict what your system will look like, you only need to look at the current configuration file, not all of the history <erkin>I was hoping I could do something like `guix configure -e '(keyboard-layout (keyboard-layout "en" "dvorak"))'' <roptat>it's really not meant to override things at runtime, even "guix install" and such are more of a hack on top of declarative package management <erkin>Maybe I'm approaching things the wrong way. <roptat>it's a very different paradigm from other distros/package managers <erkin>I've always been dissatisfied with the Unix way. <roptat>so the main difference is that it's declarative and functional. That means, your system is generated from a single configuration, and the generation takes only that configuration into account <roptat>no changes to the system that depends on its history like other systems do <erkin>Okay so, let's see if I get it right: In order to obtain the ideal setup, I need to build the VM image from a configuration file tailored to my interests. <roptat>but you don't need to build the VM completely, you can reconfigure <roptat>and guix will keep intermediate products in cache too, so it's not like rebuilding everything everytime <roptat>I think I misread a manual page, the returned size seems to be the size of the first element, not the size of the array :/ <roptat>"These calls return the number of bytes received, or -1 if an error occurred" now I'm confused... <erkin>Wow, there's vi but no emacs or mg. >:P <erkin>In order to hack on Guix itself, is the recommended workflow like this: Have a regular stable Guix binary install + Git checkout, hack and bootstrap + guix environment to test <roptat>have a regular guix installed, git checkout, build it, hack, and use pre-inst-env to test it <erkin>I think I'll skip the VM and go with the foreign method then. <roptat>haha! I got it! it was my decoding going to far in the answer <roptat>interpreting other messages as additional attributes <erkin>Is there a known solution to the `setlocale' error? <erkin>Oh, I had to set $GUIX_LOCPATH in ~/.profile <erkin>I think the reason it didn't work last time is that I used sudo to run Guix and the envvar was pointing at the local directory even though it was installed as root. <roptat>it works! I have the list of IPv6 addresses on my system, associated with the interface number!