IRC channel logs
2023-09-30.log
back to list of logs
<nckx>Right, OK of course, OK, my rescue initrd now has git in it, OK, I guess that's a feature. <PotentialUser-47>The repository clones fine when using `git` from the command line, whilst this repository does not contain submodules, it is also failing for me when using a repository with submodules <PotentialUser-47>seems like those dependencies are intended to be added in `guix/git-download.scm` line 92 <PotentialUser-47>which that diff didnt change, however I may be incorrect about where they are sourced <gnucode>It would be nice to have the same bot in #hurd <Aurora_v_kosmose>Is there a supported way to add a mirror for sources like the Software Heritage is? Or does one need to modify Guix source? <AwesomeAdam54321>or add sources to the store manually like `guix download path/to/hello-1.0.tar.gz` <bumble>I'm trying to use geiser's "geiser-edit-symbol-at-point" but when I call it from emacs ... nothing happens except this message gets printed to *Messages* buffer "You can run the command ‘geiser-edit-symbol-at-point’ with ®" <bumble>does anyone know what I'm doing wrong? <bumble>if I move the cursor over, for example, home-inputrc-configuration the *Messages* buffer shows "geiser-eval--scheme-str: Wrong type argument: listp, :t" <bumble>also is there any reason to prefer either of use-modules or define-modules expressions? <lilyp>Aurora_v_kosmose: you can do (list alternative1 alternative2 …) <lilyp>note that some stuff like SWH is added implicitly even without you doing so <lilyp>PotentialUser-58: sounds like a leaf package to me, feel free to update <PotentialUser-58>Actually, what is the best way to get into Guix? Just read the entire manual? <Lumine>The best way into it is by installing Guix :) <PotentialUser-58>Well that's fore sure, but in the past I've found myself confused compared to other distros <janneke>PotentialUser-58: there are also gentle(r) introduction videos to guix if that's your thing <PotentialUser-58>That definitely sounds worth looking into; I only know the ones from System Crafters <PotentialUser-58>Those are excellent; they are the reason I even know about Guix - they are just a bit long sometimes. Thanks for the link, that looks more like what I was looking for. Anyway, it is time to install Guix ;) <Guest80>Just trying to set up tlp. I think there is a misspelling in option "disks-devices". It sould be "disk-devices". Can see the result by invoking tlp-stat -c. Hope someone can chang that. <PotentialUser-68>I then tried to reconfigure with the file from /etc without any changes. It was super slow, and failed because commit x of channel guide is not a descendant of y. <PotentialUser-37>I figured out the issue for the keyboard: the services need to be moved after bootloader and file-systems in case one wants to configure the Xorg layout according to the manual. For the other issue (with commits), I think the issue is that I forgot guix pull. <mirai>What's the equivalent of `git clone` for a CVS repo? <PotentialUser-37>Do I need to add GUIX_PROFILE to my bash_profile? The manual makes it sound like that should already be taken care off, but the variables (INFOPATH, EMACSLOADPATH, ..) aren't set. <mirai>lilyp: it doesn't look as simple as `cvs checkout URL` :( <mirai>though nvm, because I'm now pondering whether its worth spending time to write a package for this <mirai>it looks like the thing has been “abandoned” <mirai>or outdated (something something CVS being migrated to git repo) <mirai>am mailing upstream to find out <vivien>#66099 finally updates upower and eudev, for use in gnome-team! It involves guix-friendly patches that have been sent upstream, for which I am expecting reviews. In any case, this was a very complex issue (exceeding by far my usual guix level), so I’m glad it went well. <PotentialUser-45>The icecat doesn't seem to have been updated already. Is that already WIP? <AwesomeAdam54321>Does Guix fully utilise the /gnu/store/.links yet? I remember there being an issue where guix redownloads stuff that was already in the store <tex_milan>Hello Guixers! I am trying to package broot (command line file manager), and it is in our favorite language rust. I get this error: package `anstream v0.6.4` cannot be built because it requires rustc 1.70.0 or newer, while the currently active rustc version is 1.68.2. Any idea what to do? <tex_milan>Any idea when it will be merged? It is 2 month old with no comments... <vivien>My guess would be "when enough people show interest in getting rust 1.71.0" <vivien>Well, if you write a nice message to the issue, maybe that will remind other people that it exists. <tylerm>Hi all! Im currently trying to package caddy, and to do that i need to package its dependencies. One of them I need is net/netip which is in the go stdlib. Does anyone know how to get this in, as adding just go or go-stdlib to native inputs doesnt work. Thanks! <PotentialUser-22>When running make check, build-utils.scm failed. Is that to be expected? <jaeme>How does one configure gnupg on GuixSD? I'm stuck on getting a "pinentry interface." <jaeme>I have a guix home configuration for home-gpg-agent-service-type but I don't know why it's not working. <tex_milan> "pinentry-program " #~(string-append #$(file-append (S "pinentry-rofi") "/bin/pinentry-rofi") "\n") <tex_milan> "pinentry-program " #~(string-append #$(file-append (S "pinentry-rofi") "/bin/pinentry-rofi") "\n") <Altadil>PotentialUser-22: Weird, it’s working for me (just ran make check right now) <PotentialUser-22>Altadil: Yeah, it is super weird. When I tried again, only substitutes failed. When I tried once more, substitutes worked fine. <radio>Anyone here uses the modular latex setup? <radio>I've put a lot of latex packages on my manifest and can't even use pdflatex to build a pdf from a simple .tex <pastor>Guys. I'm playing arround fixing some issues on `emacs-guix`. I have to rewrite some function from `gnu packages` and I cannot wrap my head around this simple error. When defined on the `gnu packages` module everything works fine. This is the code: https://paste.debian.net/1293624 <pastor>This is the error: socket:6405:2: source expression failed to match any pattern in form (let* ((not-colon (char-set-complement (char-set #\:))) (environment (string-tokenize (or (getenv "GUIX_PACKAGE_PATH") "") not-colon)) (channels-scm channels-go (current-guix-package-path-entries))) (set! %load-path (append environment %load-path channels-scm)) (set! %load-compiled-path (append environment %load-compiled-path cha <pastor>nnels-go)) (make-parameter (append environment %default-package-module-path channels-scm))) <pastor>If I remove the `channels-go` directive everything works fine. But I don't get how the let binding variabloe can be a syntax error if it is used on `gnu packages`... <pastor>I'm not sure If I'm being clear. What is weird is that in `gnu packages` they seem to be using this kind of syntax: <pastor>`(let* ((var1 var2 value)) ...)` <PotentialUser-22>Is there a difference between `guix git authenticate` and `make authenticate`? <graywolf>PotentialUser-22: Former is using your current guix, the latter the one you build from the git. <graywolf>make authenticate just calls guix git authenticate, but with guix from the git repo <efraim>tex_milan: I'm taking another look at the rust update patches, IIRC 1.71 didn't compile for me (at all) so I'm seeing about starting by bumping it to 1.70 <pastor>Answering my own question seems that the problem with the let binding I was having is due to not using `(srfi srfi-71)`. Do you guys have any way to have a bit more insight on what the different `srfi`s are doing? <RavenJoad>Is there someone here who could take a quick look at Issues #64778 and #66206. Both are small patches adding a package or two. <gnucode>PotentialUser-22: a little anxious for some reason. <gnucode>but yesterday was a fun day. So that's nice. <efraim>tex_milan: I will say it won't be a quick process, I'd suggest seeing if anstream@0.6.4 builds with rust-1.68.2 or trying an older version of broot doesn't require such a new version of anstream <tex_milan>efraim: will look at those options, but anyway good there is hope :) <somenickname>How can I retrieve the latest ISO since the link of the website is still broken? <koollman>somenickname: which link on which page ? <RavenJoad>None of them work actually. I just tried them all. <somenickname>iso9660-image.x86_64-linux is that the correct job name? <RavenJoad>somenickname: Yes, that is the right one for x86. <somenickname>Isn't KDE now supported as a DE? I don't see it as an option in the installation. <RavenJoad>somenickname: plasma-desktop-service-type is probably what you are looking for. I'm not sure how well supported it is though. It historically has not. Unless something changed recently I do not know about. <somenickname>RavenJoad: Yep thanks. I just want to play around with it. <jaeme`>I'm still scratching my head over pinentry for gpg, I followed the instructions in the guix manual but gpg still tsays that it can't find pinentry when I try to import one of my keys <gnucode>hey friends! hats off to whoever made/maintains/worked on the graphical guix installer. It's working beautifully on the Dell Optiplex 7020. <gnucode>hmmm. Well maybe I spoke too soon...I just rebooted and I now have some issues. Namely I cannot boot. haha. <gnucode>but at least the graphical gui worked! that's kind of cool! <gnucode>grub loads up, but I see an option to boot elemntary OS...which is what guix system should have replaced... <gnucode>I think there is a grub command...that I can run that should let me load grub.cfg from the hard drive... <gnucode>but I encrypted the harddrive so that might be hard to do... <gnucode>well I think the simplest option would be to just re-install without encryption. Weird. <gnucode>hmm...maybe I booted guix incorrectly from the installer disc. I chose to boot via legacy boot, but I think the optiplex supports booting uefi <frog_>if you want an encrypted system, you have to do a separate unencrypted /boot partition so the grub config file can be read. you can try to manually decrypt your partition using the grub command line but it may be hard. the commands are something like this: https://paste.debian.net/plain/1293634 even if you manage to boot your system, youll need to figure out how to repartition your disk and add a separate partition for /boot. then <frog_>edit system config to use it. <nckx>ACTION does not recommend a separate /boot partition. You'll be in the minority, it's not well-supported by Guix System, & you'll be even more on your own than you feel now. <frog_>out of curiosity, how do you do an encrypted system then? <nckx>Most of my systems use the standard Guix System configuration with a single encrypted partition (so no ‘FDE’, but no /boot partition) that makes you enter your passphrase twice for double security. My laptop uses a separate /boot partition that then needs to be kept in sync ‘manually’. It's an extra potential point of failure. <frog_>in the first case, where does grub read the config file from? <nckx>The encrypted volume, hence the double passphrase entry. <nckx>Guix doesn't have support for hacks like key files in the initramfs yet. <Altadil>nckx: Do you know if it will happen eventually ? I would love not to have to enter the passphrase twice. xD <nckx>I don't think it will ‘happen’ until someone takes the lead & does the work (that's not a dig at anyone; I'm not planning on working on it). <gnucode>frog_: and nckx I got it to work now. I installed guix via legacy boot the first time. The second time I tried I installed guix via grub-efi. <gnucode>guix made me enter my password twice to properly boot. <vivien>Close to that idea, I would like to have a way to reboot the system without having to enter the passphrase. <nckx>gnucode: Yep, get used to it. <Altadil>nckx: makes sense. I wish there was some kind of bounty program for Guix, because that’s the kind of thing I would definitely help fund. But I doubt there are enough people interested in the first place. :( <frog_>i was under the impression that grub needs a readable config file with commands like cryptomount in order to know to mount encrypted partitions. based on what you said, i guess grub is smart enough to decide on its own to do that when there's no config found otherwise. huh. <vivien>lilyp, for #66099, I’m using the v3.2.12 tag as a source, but one of the patches that get applied overwrites the version in configure.ac to 3.2.14. If I declare eudev to have version 3.2.12, then it is misleading, because its configure.ac reads 3.2.14. I have never met this problem before. <vivien>So I naively declared the version to be 3.2.14.beta, because it is not released yet, and still have 3.2.14 in configure.ac <lilyp>You are free to modify the patches you submit. Choose one versioning scheme and commit to it (heh) <lilyp>Do comment on the fact that you altered the patch if you do so, however <nckx>frog_: That configuration file is embedded. <nckx>(Standard GRUB feature, not a Guix thing.) <frog_>nice, didnt even think about that possibility. still recovering from arch linux brain rot from years ago of only installing grub once. <mirai>Is it possible to have two X servers running for two different users? <PotentialUser-53>Hello, everyone! I wanted to install Guix but ran into a problem. With Russian ip I am not allowed to download packages, no access to official repositories. Can you tell me who to contact to get the blocking lifted? I think it's not fair to my fellow citizens, because everyone, regardless of nationality, deserves freedom <mirai>PotentialUser-53: rekado may know something about this <jaeme`>PotentialUser-53: Does that mean you can't connect to GNU savannah nor the substitute servers? <efraim>The institute hosting the berlin substitute server is currently blocking Russian IPs but the bordeaux substitute server should still be accessable <vivien>There are a few services in guix that expect a raw configuration file. This makes it unextensible, because how do you merge raw files into one? A popular approach is to include configuration files by their file name. For instance 10-initial-config.cfg comes before 20-optional-stuff.cfg. What would you say if, instead of a plain configuration file, the configuration for these services were to be an alist from priority values to files? That way <vivien>the services could become extensible. <mirai>vivien: through clever scheme code? <mirai>you can certainly merge lines or files into a bigger file <vivien>mirai, on the guix side, it would be sufficient to merge the main alist configuration and all alist extensions, sort by priority value, and them concatenate them. <mirai>regarding `service-extensions', that's a bit trickier <PotentialUser-53>The bordeaux servers don't work for me on Fedora 38, but they do work on a Gnu Guix virtualization, it's all strange <mirai>vivien: you can tinker with how service extensions are “merged” <vivien>Another idea would be to still have named configuration bits, but the extensions would be forms such as (add-after '10-initial-config '20-optional-stuff (plain-file "optional-stuff" "the actual configuration...")) <vivien>This is how phases work in guix build systems <vivien>The problem arises when no extension provides '10-initial-config of course <mirai>IMO I think the services need to be rethought a bit (for instance, how to express _instances_ of a service) <mirai>just because nginx-service exists and can be extended doesn't mean that only 1 running nginx is enough <mirai>and as you're saying, capturing the notion of priority for service extensions <vivien>The problem I see with nginx is you can only extend the server blocks, not the upstream blocks <mirai>I used nginx as an example, the idea is to have the systemd foo@instance-1 foo@instance-2 analogues <vivien>Maybe this is more of a problem for syncthing <mirai>I think there's room for improvement or redesign here, especially if we're going ahead with using our own init here (shepherd) <mirai>how shepherd meshes with guix and its way of configuration <vivien>My dream for an extension mechanism is if services could be extended with pure scheme functions. You would then extend the nginx service with a function that take a partial nginx-configuration and return a more complete nginx-configuration <vivien>I have never had to run multiple instances of the same service to be fair <mirai>looks like it could turn into something unrecognizable quickly <mirai>if each service is allowed to muck up the configuration from another, the order of the transformations or who changed what <vivien>Correct, but some transformations commute, such as adding a server block in a nginx configuration <mirai>yes, those kinds of things are alright (important note: the nginx directives have ordering) <mirai>now that I think about this, can't you already do something similar? <vivien>I agree that it would be a mess if an nginx extension were designed to remove a server block. Then the order of evaluation would matter dramatically. <mirai>you extend the service with a lambda + the data <vivien>I don’t think any services can be extended with a lambda <mirai>then you edit how the service “merges” the configs <vivien>I’m talking about the configuration value being a lambda <mirai>store the lambdas somewhere, perhaps a field and invoke the lambdas before the serialization <vivien>You would have to rewrite the service type though. <mirai>you could have a service whose record looks something like (define-record-type … (mutations (list (lambda (f) …)))) <mirai>with mutations invoked when opportune <mekeor>ACTION runs X as user on guix system <singpolyma>vivien: discussion of nonfree software is off topic ;) <mekeor>ACTION ... using this command: xinit -- ~/.guix-profile/bin/Xorg :0 vt1 -keeptty -configdir ~/.guix-profile/share/X11/xorg.conf.d -modulepath ~/.guix-profile/lib/xorg/modules -logfile ~/some/path/x.log <mirai>mekeor: I'm looking to have one of the Xs connected to a physical monitor for user foo and user bar running a VNC/X server to be used remotely <vivien>I mean, I thought some parts of X were published (but I don’t know if the license is free) <mekeor>mirai: idk maybe you can tweak with :0 vt1 -keeptty in particular <crutches-adieu>"Network Services Aren't Free or Nonfree; They Raise Other Issues <mirai>mekeor: is this launched automatically by some service or are you using a shell alias here? <mekeor>mirai: yes, i have X as shell alias to that command :D <vivien>It’s amazing that guix build -c 8 chokes on webkit, because it compiles 8 huge C++ sources at once, but guix build -m 8 does not, because most other packages are reasonable. <mirai>you need to download more ram <mekeor>ACTION , if ever, uses -c 1 -M 1, lol <efraim>julia packages eat ram, when I build them on my machine they each try to use all 24 cores and about 15GB of ram <user363627>What's the approach/mechanism to run some guix shell commands via the repl? <daemonspudguy>I've been wanting to try GuixSD again. Last time I tried to install it, it hung at the bootscreen. <lilyp>user363627: ,use (guix scripts whatever) (guix-whatever "string" "arguments" "as" "if" "you" "were" "doing" "command" "line" "stuff") <lilyp>add a newline before (guix-whatever …) <gnucode>mam guix makes it so easy to install your system the way that you like it. <gnucode>dthompson: if you are who I think you are...then thanks for your last few emails. <dthompson>might be a different person unless we're talking private emails. I haven't written to mailing lists much lately. <jaeme`>How does one set environment variables at build time for a package? <jaeme`>rust-wgpu requires an environment variable WGPU_BACKEND to have a value. <apteryx>(setenv "WGPU_BACKEND" "some-value") in a build phase <apteryx>do we have a 'hid_listen' tool in guix? <jaeme`>Whelp it turns out to compile swww, I may need a newer version of the Rust compiler <jaeme`>it seems that guix only has rust-1.68 while latest upstream stable is rust-1.72 <jaeme`>I might just go to an earlier version of swww in that case <jaeme`>Or just package rust-1.72 myself in my channel <jaeme`>I can't believe my first foray into guix is contributing to rustacean idiosyncrasies smh. <efraim>I have rust at 1.70 on the rust-team branch, having trouble getting 1.71 to build <efraim>I might try again after seeing about adding more outputs to the rust package <abhiseck>I have full disk encryption. After doing a system reconfigure, when I reboot, grub asks for password (as usual) but when I give the password, it says no cryptodisk with given UUID found. <abhiseck>how to recover from this? I am currently in a Trisquel live environment <rua-k>Hi :) can anyone help me with a basic setup of sway/guixos working on an aero laptop (nvidia/intel graphics...)? <anthk_>on guix I wish I could understand packaging. I used guix import egg to import some chicken scheme srfi (203 and 216, SICP related) <anthk_>and then I added the pkg definitions with guix edit and placing some defines on top of each section, but I had no luck <janneke>anthk_: you had no luck, what does that mean? <xelxebar>Man, I can't type gui anymore without typing guix and then deleting the x. Darn you guys!! <janneke>xelxebar: you let your typing skills deteriorate and then look for someone else to blame? /S <lilyp>To be fair, you don't need a GUI if you have Guix :) <xelxebar>Speaking of no guix and high utility, any of you willing to share your childhurd config? <xelxebar>Kind of wish the childhurd store was shared with the guest. <xelxebar>Using a hurd-vm service means that reconfigure potentially nukes your image, right? <xelxebar>Will take janneke's sage wisdom by whining, complaining, and blaming others :P <wingo>hello guix :) question, is there an easy way to get an environment that includes full debug info on one specific library? <wingo>i want bdw-gc with debug symbols and ideally source. <janneke>xelxebar: yes, so anything (of value) i create in a childhurd, i `guix copy' back to the host <xelxebar>janneke: How much space do you give your instance and do you run without -snapshot? <janneke>xelxebar: 25GiB (i'm using 2GiB of swap -- which may or may not work) and i usually bake an image that i start manually, without --snapshot, to set it up <janneke>hopefully, that will soon no longer be necessary <PotentialUser-22>Are there any packages that need packaging that are suited for a beginner? <xelxebar>janneke: Interesting. What are you banking in at the start that can't be part of the operating-system def? <janneke>xelxebar: can't be is possibly a bit strong (and a good question!); .bashrc aliases, clone + compilation of guix repo + store population for `guix shell', /swapfile, ... <xelxebar>civodul: Woah. What a friggin' cool patch! Seems like a huge QOL improvement among other groundwork that I don't fully grok yet. <xelxebar>janneke: Ah. I see. Sounds like you're actually using the hurd. Naively, I'd think some of that could be setup with guix home? Swapfile... is also something I'd like to autogen. Will give this some thought. <PotentialUser-22>While trying to update the calibre package, I get that uchardet was not found. My guess is that I have to add the uchardet to one of the inputs. But how do I decide which? I am aware of the differences between them, but not sure how to go about choosing them. <janneke>so, guix system reconfigure on the hurd creates and activates a new system profile, installs a new boot.cfg but then fails with <janneke> /gnu/store/b0ani8jjgp21qkgr514880081hizyap5-grub-minimal-2.06/sbin/grub-install: error: cannot find a GRUB drive for part:1:device:hd0. Check your device.map. <janneke> (bootloader (bootloader-configuration <janneke> (bootloader grub-minimal-bootloader) <janneke>what is a device map, where does it live, where could the Hd0 come from? <efraim>what is the command you're using to boot the hurd VM? what are you feeding it as a harddrive? <janneke>... --hda y80kpqvb2ifrkl0x0i02ig450bp27cr7-disk-image <efraim>I remember hda being the first IDE drive <efraim>does the hurd have a /dev that you can look for /dev/wd0 or /dev/hda or /dev/hd0 to see if they exist? <janneke>root@guixydevel ~/src/guix/master [env]# mount <janneke>typed:part:1:device:hd0 on / type ext2fs (rw,relatime) <janneke>and quite possibly i booted without "noide", so config.scm _should_ say hd0 even <efraim>is something assuming grub1? I'm trying to remember the last time I had to refer to devices and partitions <janneke>i'll be generating hdX device nodes first <efraim>I think it was when I was dealing with the longsoon machine and its firmware had me calling binaries hd(0, 1)/boot.cfg or the like <janneke>and after that possibly booting using "noide" and try it with /dev/wd0 <janneke>it's just so weird that config.scm and the error don't match? <efraim>I'd consider checking with grub or hurd or even bsd people <efraim>it is weird, but I assume that something in the hurd is translating from /dev/wd0 to a device that it wants, or thinks it should have <efraim>can I deprecate a package using a specific output from another package? <civodul>efraim: you mean deprecating X in favor of Y:whatever? <efraim>I want to deprecate rust-analyzer for rust:tools <civodul>i guess you could deprecate for “rust”, as a very rough approximation <civodul>and maybe add something about ‘rust-analyzer’ in the description of rust <efraim>I guess I could make a 'fake meta-package' using union-build of the rust:tools output and provide that <efraim>I'll have to think about that one, the current one links to rust-src and I don't have a way to test the one from rust-1.70 <mirai>PotentialUser-22: looks like GCC changed something <mirai>have you checked with upstream? It looks like something that should be fixed from their side <mirai>or perhaps has already been fixed but you'll need to cherrypick <PotentialUser-22>Gentoo is stuck on the same version as well (apparently because of qt-6; but someone got it to compile) <radio>I'm trying to clean up some warnings that appear to me during the boot process, but couldn't find any useful information about how to adress them <radio>Some open issues relate that GC Warning, but I didn't find any answer in how to solve it <mirai>PotentialUser-22: IMO the message is not as literal as it appears to be <radio>But seems guix's eudev doesn't even provide a file in hwdb that contains the word `zoom' <mirai>I get that upstream might be unwilling to expend effort troubleshooting toolchain issues from the users side (due to varying levels of expertise) <mirai>but the messages you got in the build log are things that upstream has to address at some point <mirai>unless they're going to freeze their GCC & openssl forever <mirai>my suggestion is to build from their latest git and see if it happens <mirai>if it doesn't then you'll have to find out what commit(s) to backport <mirai>or simply use that commit instead of a fixed release (with a comment in the package definition explaining why a release wasn't used) <mirai>if the issue still occurs you should give a heads-up to upstream so they can get things up to speed <mfg>Does someone know what the current state of the modular texlive is regarding the ls-R database? I believe it should have been fixed such that building with modular texlive is not as slow anymore, but it is still slow for me. <mfg>So am i using it wrong somehow? <civodul>vote for me and i’ll make sure not a single worker remains unemployed <janneke>yeah right, but will they be paid a living wage? <janneke>and what about pension, i didn't hear great things about that lately <civodul>what’s better in life than a sense of purpose? <civodul>building software for the benefit of society as a whole <janneke>having a purpose in life in nice, once you have eaten <civodul>(in all seriousness, i have a fix for the bug above that i can’t wait to have deployed) <mirai>building software is nice and all but is it _free_ software? <civodul>ACTION seizes power, leaves workers idle <efraim>who elected berlin, to speak on behalf of the works? <janneke>PotentialUser-22: for a well-behaved package it is next to trivial <janneke>with an upstream that sucks or is otherwise clueless, well... <efraim>sometimes its a case of It Works On My Machine™ <janneke>just share the actual problem that you're encountering <janneke>PotentialUser-22: yeah, chances are bigger you get some useful help ;) <janneke>ah you were struggling with calibre, sorry :) <janneke>does nix have the latest version packaged? <janneke>sometimes it can help to cheat and look at nix or even debian <PotentialUser-22>I just saw the Gentoo folks being stuck at 5.44 as well and thought "guess that's that" <user11>Hi, I'm on Devuan (sysvinit) and got guix installed. Seems to work now, but for some reason the daemon doesn't run correctly and entering guix install into the console results in guix install: error: failed to connect to `/var/guix/daemon-socket/socket': Connection refused <user11>Funnily enough if I *then* start the service to start the daemon it works correctly <RavenJoad>PotentialUser-22: Some software actually depends on mutable state and is underdocumented (looking at you apcupsd) to make everything work. <superkamiguru>somenickname: I imagine if it's relating to the rpi template included in the guix package then that's fine. Not sure if discussing obtaining/adding in the nonfree firmware would be on topic though. <superkamiguru>So I have a bootloader/filesystem declaration that I am using to generate a vm image https://paste.debian.net/1293725/ (efi-raw converted to qcow2), but for some reason no matter how I define the efi partition, it doesn't get mounted in the resulting image. The vm will boot, but only the rootfs will be mounted to vda2. <somenickname>superkamiguru: I don't know it. But I know that guix system image does overwrite file-system definitions. Maybe the vm command does the same for boot? <superkamiguru>I am using guix system image to produce the vm, so that might be why. This process/config file used to produce a vm with the /boot/efi directory mounted on boot, but after a few updates it no longer works. <superkamiguru>I am mainly trying to produce a vm image with a config file, and then have the vm be able to reconfigure itself based on that same config <superkamiguru>This definition is close, but since vda1 doesn't get mounted to /boot/efi, when you try to reconfigure it throws an error that /boot/efi is not a partition. Manually mounting vda1 to /boot/efi fixes this and on reconfigure mounts correctly, but then the image is large and not practical to distribute. <pastor>Good evening. This morning I made some (janky) changes on `emacs-guix` to display also packages from user channels. Anyone would like to have this changes upstreamed? <RavenJoad>superkamiguru: The only thing I see different between my desktop's efi config and yours is the trailing "/" on the file-system mount-point. Otherwise, I do not know either. <lechner>Hi, is HEAD in a good state, please? <mirai>lechner: by HEAD I'm assuming HEAD of master. Why wouldn't it be? <lechner>it's works less than half the time for me <mirai>with a small enough sample and misfortune perhaps :-) <mirai>serious tho, when or what issues are you getting <lechner>well, reconfiguring six systems and at least one home environment with a custom eudev usually means there are compilation errors. often they are spurious but take ten or more attempts <mirai>perhaps that custom eudev is building nondeterministically? <lechner>the compilation errors are in consuming packages (of which there are many) and occur with the standard Guix version as well <mirai>hmm… do you have logs for these? <lechner>also, sometimes packages are not available due to build failures, which prevents me from reconfiguring my systems or home setups after pulling guix. that means i am unable to add those packages or even use guix shell <mirai>I've experienced spurious build failures before when I play around with the “core” ones though they seemed to be related to available memory or how many builds in parallel were happening <mirai>I recall experiencing some corrupted store items before that would seemingly pass verification <mirai>and required doing a manual guix gc -D … to clear them out <lechner>i recently corresponded with cbaines either here or on the mailing list about one package but cannot remember which one <lechner>either way, for me that means i have to downgrade guix again, unless there is another command. due to my custom guix I am not sure I can use time-machine <superkamiguru>RavenJoad: Did you generate your desktop with "guix system image" or was it setup using the installer? <RavenJoad>superkamiguru: I used the installer over a year ago, and have since refactored everything in the scheme files. I typically use grub-bootloader & BIOS for VMs. <superkamiguru>RavenJoad: ah ok. I would use grub-bootloader but I am trying to build an aarch64 compatible image and I can only get grub-efi-bootloader to work there. <RavenJoad>Have you looked at how Guix's installer defines grub-efi-bootloader for its install images? <superkamiguru>Yeah there is a lightweight-desktop template that uses grub-efi, but it defines two randomly named file-systems that get generated as vda1 and vda2 (so reconfigure with the same config doesn't work) <superkamiguru>The workaround I found was to just define the file-systems as vda1 and vda2, but the resulting image doesn't mount the efi partition on boot (so reconfigurations fail) <RavenJoad>Oh. I see what you want to do. Generate an installer with a config, embed the config in the installer for use on booted ISO, then reconfigure from inside the VM with embedded config. <superkamiguru>So I am actually not trying to generate an installer. I am just trying to generate a complete system image similar to how the lightweight-desktop.tmpl can be used to generate a working image. I would like the config used to generate the image, to also be included in the image for reconfiguration <superkamiguru>I am outputting the image first to efi-raw, and then convert it to qcow2 for booting with qemu <RavenJoad>Have you tried the qcow2 image output already? <superkamiguru>Yeah, sadly the qcow2 image type overwrites the grub-efi-bootloader declaration and replaces it with grub-bootloader <RavenJoad>When you build the vm image, does the resulting /etc/fstab make sense? <RavenJoad>Agreed. Strange that only the esp-partition is not mounting... <superkamiguru>What's also weird is that I can manually mount vda1 to /boot/efi, reconfigure with the same config, and then everything will mount correctly moving forward <RavenJoad>I'm curious if the efi-raw -> qcow2 step is messing with something. <superkamiguru>The only downside is doing a reconfigure of the vm expands the base size from like 600MB to 2+GB which isn't optimal for distributing the vm <superkamiguru>That's what I was thinking, but it also used to work fine a few months ago. Not sure what changed <superkamiguru>So I am setting up an application that will pull/run a premade guix qcow2 image and run it as a local guix environment. Similar to how podman is configured to run on non linux platforms. <superkamiguru>So for this purpose I would need to have a prebuilt vm that can be pulled relatively easy. <RavenJoad>Have you tried booting the VM directly from the efi-raw, to see if the conversion is the problem? (If qemu even allows booting from an efi-raw.) <superkamiguru>I didn't even consider that qemu supported raw images, but it looks like it does/might <superkamiguru>Thanks for all the responses by the way. I have been banging my head against a wall with this for a few months now <the_tubular>Is there a resource I could read that explains what every expression starting with a # does ? <RavenJoad>the_tubular: # is a special thing for Guile's (Scheme's?) reader. It allows you to have things bound to symbols, like ' expanding to quote. <RavenJoad>#~, #$, #+, #$@ are all g-expression (gexp) functions. See (guix) G-Expressions for an explanation. The Guix website blog has posts about it too. <nckx>#: are (used as) keyword arguments, basically named arguments to procedures so you don't have to rely on/remember a fixed argument number/ordering. That's what package arguments really are. <nckx>#\a is the literal single character ‘a’, as distinct from a string, such as "a". <nckx>Also, good evening Guix. <nckhexen>In nckx news, a script has deleted all my /dev and /etc. I wonder how long I can keep using this machine before I'm stuck. <nckhexen>(Why is this on-topic? Because the script started chomping through /gnu, which saved the rest of my file system from doom. Thank you, Guix.) <nckhexen>An ominous beginning to the month of spook. <superkamiguru>RavenJoad: So I booted the raw image with qemu and it still doesn't mount the efi partition <nckhexen>Having to chmod 666 /dev/null is just too on the nose. <janneke>nckhexen: any possibility for extundelete? <janneke>ACTION never made good use of that... <nckhexen>I don't use ext, so I don't think so. A (proper) versioning file system sounds cool right about now, but I should probably have mentioned that I have full backups and am not crying. <nckhexen>If anyone is sitting on a privilege escalation to root, however, now would be the time to share. <janneke>ACTION stopped with any and all feeble attempts at backups since guix <nckhexen>In that situation I think I'd go for init=/var/guix/gcroots/… since it typeable & guessable. <Noisytoot>nckhexen: Mount the filesystem from another computer? <janneke>ACTION still believes that store prefixes really should have been named /gnu/store/bash-5.x.y-HASH <nckhexen>Last time I tried to tab-complete the (then-Nix!) store in GRUB it didn't end well. Which is to say, it didn't end at all. <nckhexen>Noisytoot: I appreciate the efforts to help, but I'm not really stuck or worried. I've got a little rescue initrd with ssh + rsync for just such situations. <janneke>yeah, you need to know a bit of the hash prefix <the_tubular> Thanks for your help, I'll have a lot to read (and learn)! <superkamiguru>RavenJoad: I am starting to think that what you said about "guix system image" overriding parts of the file-system declaration is probably the case <janneke>probably code that needs to run during a build phase <RavenJoad>superkamiguru: Ok. I have not delved into the code. That bit about image overriding parts of file-system wasn't me; it was somenickname. <RavenJoad>It might be worth it to toy with the system in a REPL to see what the system->image procedure modifies. <superkamiguru>Yeah since using --image-type=qcow2 overrides the bootloader declaration, and using efi-raw is overriding parts of the file-system declaration, I imagine I would have to write something to generate a working/reconfigurable aarch64 image. <superkamiguru>Wonder what changed over the last few months, since this resulting image used to mount the declared file-systems fine. <vagrantc>lechner-: i think there are some parts of it that are single-threaded ... i have definitely seen it max out a single core on a slow multi-core system <vagrantc>leaving the other cores almost entirely idle ... <vagrantc>but it also builds various derivations and those can each be built independently <graywolf>Hi! Would anyone have an example how go use guix-extension? I don't have a good grasp on the service compositions yet :/ <graywolf>Hm, do I understand it correctly that the guix-extension is mainly useful if I am writing a service that needs to modify guix-service-type? And if I want to manipulate it from operating-system, modify-services is the way to go? <RavenJoad>If you want to modify your guix daemon through an operating-system record, (modify-services %base-services (guix-service-type config => (guix-configuration ...))) is your way to go. <RavenJoad>I believe guix-extension are the only fields that are available to be modified through an extension of guix-service-type by another service-type. Though I am probably wrong about that. <ulfvonbelow>is there a "proper" way to use (guix build download) in a program-file? it seems to want guile-json as an extension, and we don't have a package variable named "guile-json", only ones with version suffixes, and it made me think that this was probably a seldom-trod path <mirai>ulfvonbelow: I'm cheating with (with-extensions (list guile-json-4 guile-gnutls) … <mirai>idk, never thought much about the json-4 <mirai>I've assumed its like the situation with libusb or libxml <mirai>you might need to do (with-modules … or something something source module closure to get (guix build download) <mirai>try grepping the guix source <mirai>I think there's an instance where both with-modules and with-extensions is used <veriphile>podiki: did I already ask you if you packaged exiftool <podiki>veriphile: you did not and I didn't, though I do use it. something wrong with it? <RavenJoad>Why are all the entries in my laptop's elogind/logind.conf commented out? Can I enable them? <RavenJoad>Ah. The one in /run/current-system/profile/etc/elogind/logind.conf is the one with all the commented out values. <lechner->not sure which one is being used. on my system, ps aux | grep logind.conf comes up empty <RavenJoad>I mean, when I close the lid, it seems to suspend. Now I need to figure out how to extend it to also lock all my sessions. <lilyp>RavenJoad: we should have an elogin-service-type that you can configure <RavenJoad>lilyp: Yep. The elogind-service-type already has the necessary bits for suspending when the lid closes (it has them by default). But my session is not locked. I am using StumpWM normally and the cookbook recommends xss-lock. <veriphile>podiki: ah, i didn't see it packaged. Do your tools use it, or do you just use the script on the side from a different dir? <libri>installed 'GNU Guix' in 'Gnome-boxes'/Debian testing , runs fine, miss 'update-manager' or 'Gnome-software' ... will try 'apt install guix' in Deb. <civodul>congrats to everyone who managed the upgrade, BTW! <isaneran>was very happy when emacs 29 finally landed <isaneran>my gratitude to the people who packaged it <isaneran>I didn't notice it at first so I got a little bit confused why my org agenda line looked different out of nowhere <civodul>yeah, actually i was expecting breakage in Emacs packages here and there, but so far i haven’t had to change a single line of my config <civodul>the Emacs people do an awesome job at maintaining compatibility <janneke>emacs29 is still on the legacy native elisp engine, right? :-P <isaneran>i though the native compilation was new in emacs 28 <civodul>in other news, i’m offloading a gcc build to my childhurd <civodul>works fine so far, except that i had to restart the childhurd once because the SSH connection or the QEMU bridge went off initially <isaneran>ACTION still has to give the childhurd thing a try <janneke>in other news, i've pushed a guix package update that allowed me to remove another couple commits from the hurd-team branch! <wingo>there should be a tool that takes a git checkout and produces a guix hash and a git commit <janneke>hurd-team now only holds commits that are mainly useful for native (real iron) installation <wingo>i know you can clone and guix hash -rx etc etc but it is time-consuming ;) <civodul>yeah, something like ‘guix download’ but for Git would be nice <isaneran>just to download the source code you mean` <civodul>and print the commit and content hash <janneke>and non-interactive guix edit, something vaguely like `guix edit .git/modules/foo.scm --package=foo --field=version=xy (ehh?) --origin-hash=xxx <civodul>janneke: isn’t it like ‘guix refersh’? <efraim>janneke: don't some of the tools accept json? <janneke>civodul: ah, would that work from something unreleased, a local git or tarball too? <janneke>ACTION likes to check their guix package builds before uploading a tarball release <civodul>janneke: no, but for that maybe you can use --with-source or --with-git-url? <efraim>In Search Of: guix extension that accepts some typos (refersh/refresh) and Does What I Mean™ <janneke>civodul: ooh, that would be very nice <efraim>guix build foo --with-source=file://///////////that/tarball/there.tar.gz <janneke>ah /me misunderstands; yes for testing that would work <isaneran>or is there already a commant to fetch the source code for a package? <isaneran>kind of like the apt-get source command or whatever it is called <janneke>isaneran: guix build --source <package> <isaneran>does that give me a git repo (if it is a git repo) or just the checked out source code? <isaneran>hmm no it gives me the checked out code at that point in time <isaneran>It would be nice to have a command that helps you with contributing to an existing project based on the package name <isaneran>and puts it in a subdirectory to cwd or something you tell it to <Guest78>the ungexp typo is not there in the code i am testing <Guest78>Guest78: it is being run in a stage modification gexp within a package definition <civodul>efraim: so you cleaned /var/guix/temproots on the build machines, right? <civodul>thumbs up for finding out about these! <efraim>civodul: I did, let me check them again quickly. I've been using parallel to do some of the checks <civodul>i’m planning to upgrade Cuirass on the build nodes + head node today <efraim>I had been clearing space and then saw it was filled immediately so I basically went directory by directory looking for anything that looked out of place <efraim>civodul: ok, they should be good <allana>Hi guix, after a bit of searching online and giving up, I'm curious to know if there is a way to build a package using musl-libc instead of libc? Any examples out there? <civodul>allana: hi! i don’t think there’s something that works out of the box, but it should be possible by defining a GCC+Musl toolchain (say) and then using ‘--with-c-toolchain’ <allana>civodul: Thanks, sounds like there is some work ahead of me :-) <adamnr>IRC log for guix an hurd appears to have stopped 