IRC channel logs
2025-01-05.log
back to list of logs
<orahcio>Hello Rutherther, I could setup my autologin, and sway started by fish as default shell. My last error was to writte the sway start script on ~/.config/fish/conf.d, that folder starts scripts before the config.fish, and the (home-fish-service-type) makes a guix profile script on that file. Now I just extend this home service with a sway start script instead to copy it to fish/conf.d folder <amano>I have an idea for unlocking two LUKS2 volumes that host a RAID1 btrfs filesystem with one password during boot. <duncan>i have made that sort of crypttab setup work on guixsd by putting the key in initramfs as per documentation. <amano>Disk 1, partition 1: EFI partition on /boot. Disk 1, partition 2: 17MiB luks partition unlocked on /dev/mapper/keyfile. Disk 1, partition 3: a mirror of btrfs RAID1 mounted on /gnu/store. Disk 1, partition 4: LUKS2 volume used a mirror of btrfs RAID1 mounted on /. <duncan>unfortunately, there is a bug which means that even if the root filesystem is associated with a key, the password still has to be input in initramfs. <amano>Disk 2, partition 1: a mirror of btrfs RAID1 mounted on /gnu/store. Disk 2, partition 2: a mirror of btrfs RAID1 mounted on /. <amano>During boot, /dev/mapper/keyfile is opened with one password. /dev/mapper/keyfile is used as the keyfile for two LUKS2 volumes that hosts btrfs RAID1 mounted on /. <amano>I haven't tested this, but this should only require one password during boot. <amano>This leaves /gnu/store unencrypted, but that's better than not having encryption at all with btrfs RAID1. <amano>duncan: I don't think documentation teaches you how to unlock two LUKS2 volumes that host root btrfs RAID1 filesystem with one password during boot. <amano>After unlocking two LUKS2 root btrfs RAID1 filesystem volumes, you can unlock zfs pools without passports with keyfiles on root filesystem. <amano>duncan: How did you implement one password unlock of two LUKS2 volumes during boot? <duncan>the documentation doesn't teach you about that use case because it is generic to any encrypted volume <duncan>if you have the keyfile in a second initramfs, and then associate (one or more) luks container in your config.scm with that keyfile, the initramfs does the needful and you aren't prompted to enter the backup passphrase <amano>duncan: How exactly did you configure initramfs? <amano>I read mapped devices documentation. <duncan>I'm trying to find the documentation, I'm not sure it's there <amano>I just explained my setup in a few sentences. <amano>I know what extra-initrd does. I just don't know what your exact setup looks like. <amano>I know extra-initrd can contain a keyfile. <duncan>setup is, as per extra-initrd, put the keyfile in a cpio archive, then associate the keyfile with a mapped device. this accomplishes e.g. a second luks encrypted drive, dedicated to home directories, with only one encrypted password prompt (the root drive) <amano>That's just encrypted /home.... <amano>My setup encrypts root filesystem but still leaves /gnu/store unecrypted. <duncan>I'm suggesting that this is generic and is worth adapting to your use <fnat>Thanks homo, sorry for the late reply, just seen the screenshot from yesterday. I'll try that way. <orahcio>homo: Hello, finally my autologin runs and fish starts the sway with every guix things <amano>Is there a documentation on how to use dropbear ssh server to unlock LUKS2 volumes during boot? <amano>dckc: guix doesn't have many packages, yet. <ieure>amano, Guix has nearly 30k packages, which is in the same ballpark as most mainstream Linux distros. <ieure>Anyhow, if someone wants lazygit, they can package it and send in patches. <ieure>Not impressed with the lazygit repo starting with an ad for a terminal "with AI built in." <ieure>I'm still really struggling to get Guix System to work in an aarch64 VM. :/ <ieure>I have an image booted that's more or less working, but for some reason, it has a pretty old guix, despite the image getting built from an up-to-date machine. <ieure>Reconfigure is building Guix from source for some reason I don't understand, and also failing tests for a reason I don't understand, <ieure>I've been trying to get this VM working for like a week, I need it for the job I start on Tuesday. Going to have to use Debian instead, I guess. <amano>ieure: Somehow, I get all the packages I want from gentoo linux and nix. <amano>gentoo linux has 56k packages. nix has 128k packages. <elpogo>Do you have an amd64 machine ieure? A few months back I built an aarch64 VM image on amd64 using "--system=aarch64-linux" with the binfmt support enabled on the amd64 machine. It mostly downloaded binaries, and didn't have to compile much <elpogo>this is the command i have in my history "guix system image -s aarch64-linux --image-type=raw-with-offset ./librewolf-os.scm" . I was able to run the image on a m1 macbookair running macos <dckc>I see nix is packaged for guix... lazygit isn't a service or anything... let's try this indirect approach... <dckc>on my first attempt, I got Poetry. (error: file 'nixpkgs' was not found ...) <dckc>also: warning: '/nix/var/nix' does not exist <dckc>the nix curl-bash recipie seems to work. <dckc>I'm not sure how to get the guix nix package to work. I wonder how it's expected to be used. <freakingpenguin>dckc: Are you using the nix-service-type on Guix System or via guix install? <dckc>I did `guix install nix` <dckc>nix-service-type sounds interesting... <freakingpenguin>Never touched Nix but my guess is you need the nix-service-type so the Nix build daemon exists. There's some instructions in (info "(guix) Miscellaneous Services"). <dckc>I was just going to ask where to find more details; a quick web search didn't turn up much <Guest56>Hi, after running guix system reconfigure I am unable to boot into guix. It showed some sort of error regarding kexec I think? <dckc>"runs build daemon of the Nix" <- I'll pass on that. I'm happy with single-user mode <ieure>Guest56, You can select boot into your previous system configuration, second option in the grub menu <dckc>"Here is an example showing how to use it:" is followed by fragments of scheme code. If I just type that stuff in verbatim, I doubt I would get good results... <Guest56>Unfortunately I don't have any previous configuration to boot into <ieure>I wouldn't think kexec would prevent booting. <dckc>... these docs assume I know where to put those fragments <ieure>Guest56, You just said you ran `guix system reconfigure' -- what'd you run it from? <ieure>Did you GC all your old system generations before booting the new one or something? <dckc>... ok... going up to "11.10 Services" provides the relevant context <Guest56>ieure sorry, I ran guix system reconfigure, on my previous guix installation and it did that. I have a fresh installation now and it won't boot just like after I ran guix system reconfigure on the old installation. Because this is a fresh install I don't have previous generations <ieure>Guest56, and you reinstalled because of the boot issue? <Guest56>Yeah, I thought maybe something went wrong during the initial install because this never happened before on this pc. <Guest56>On my previous install I used the iso provided by guix and it booted fine but after running guix pull and than guix system reconfigure i couldn't boot it anymore. <Guest56>This new install is using an iso with an updated guix channel. And it won't boot after install <Guest56>So I think something is wrong with the newer versions. (Sorry if I'm not making sense I'm really new to guix and on mobile lol) <dckc>reviewing my efforts here in Aug, I used to be able to get libstdc++ from gcc:lib ... <dckc>today, I get: error: package `gcc-toolchain@11.4.0' lacks output `lib' <dckc>sigh. back to: libstdc++.so.6: version `GLIBCXX_3.4.30' not found <dckc>the libc used by the FHS stuff is old :-/ <dckc>I suppose if I had reported that bug thru channels in Aug, it might have been fixed by now :) <podiki>dckc: are you on a recent commit? gcc-toolchain now includes what was gcc:lib <podiki>and the glibc version in fhs is the same as default glibc in guix <podiki>(just gcc-toolchain, no gcc-toolchain:lib) <dckc>I did a `guix pull` pretty recently... <dckc>guix (GNU Guix) ddbb59f9bc5eb6dfaf1fc24881bb62f11960d664 <podiki>dckc: ok. well anyway, as i said, gcc-toolchain now has what was in gcc:lib and glibc is same as in guix (which was updated in the past few months) <dckc>ok. I'll have to double-check what's going on. <elevenkb>Hello, I wonder why cargo-build-system is not responding to my modify-phases invocation at all. <PuercoPop>elevenkb: can you show a snippet of what you are trying to do? <PuercoPop>I'm trying to use the uuid extension with SQLITE, which is normally a dynamic library loaded at run-time. Is there a way to provide it in the system? should I try to .load it from the store after writing the package for it? <homo>sneek later tell orahcio nice, meanwhile I switched back to gnome, because it still offers more convenience <lh>is there another subcommand or a guix pull option to bump the hashes in a channel.scm file to the latest values from the git remote? coming from nix flake update, it feels a bit cumbersome <meaty>Anyone else having issues with guix reconfigure? I get this error: <meaty>guix system: error: getting attributes of path `/gnu/store/3s8nk0iqpaqck6242xr3la9clnlkmmvv-python-dbusmock-0.30.0': No such file or directory <meaty>I tried 'guix build python-dbusmock' and it just doesn't produce a store dir at all?? <meaty>like, ls `guix build python-dbusmock` says no such file, not "build failed" <lh>ci is giving me 502s again 😔 <meaty>can someone please try pulling and building python-dbusmock? <meaty>I just tried from the latest git and it does the same thing. However git blame does not show any recent changes <podiki>meaty: i'm not on the current commit but from 18463019a24a7c5acc9c2f3ddf3c0ba04a36db96 i have python-dbusmock <podiki>meaty: your error above seems like something different, maybe verify your guix store? <meaty>podiki: ah, yes. guix gc --verify fixed it. What happend? maybe ran out of space? <podiki>maybe a cosmic ray flipped some bits <podiki>maybe check your system log and hardware health to be sure (e.g. bad ram or hard drive failing, weird stuff happening can be caused by either) <meaty>don't say that, I had some wierd specific graphics bug earlier... but it was so specific i'm sure it's not my hardware <meaty>if we don't have dmesg, where do messages directed to it go? <meaty>my gpu driver crashed (i think) and it said "for details see dmesg" <divya>./pre-inst-env guix style package is saying not found inside guix checkout <amano>Does /gnu/store contain any sensitive data like password and LUKS2 keyfile? <amano>If not, I can implement btrfs RAID1 root filesystem on two LUKS2 volumes unlocked by one password during boot. If I leave /gnu/store unencrypted, I can only type one password during boot to unlock two LUKS2 volumes. <graciousgnu> 77 | #error "Please port gnulib fseterr.c to your platform! Look at the definitions of ferror and clearerr on your system, then report this to bug-gnulib." <homo>divya "guix shell -D guix --pure", "./bootstrap && ./configure && make && ./pre-inst-env guix style package" <amano>chatgpt 4o mini says /gnu/store doesn't store any sensitive information. <homo>unless you want it to store sensitive information such as your company's secrets <divya>homo: I've done that, still shows "unknown package" <amano>Does anyone know how to run dropbear ssh server in custom initrd to remotely unlock LUKS2 volume for root filesystem? <homo>divya hm, what about "guix style -f path/to/file.scm" <divya>That would style the whole module, and might change things that aren't required for my commits <homo>is that file registered in gnu/local.mk? <homo>that's still strange error <homo>it's even more mysterious error when you run it inside shell <meaty>mf why would chatgpt know about guix <meaty>guix stores its profiles, i.e. synthesized system config files, in the store dir iirc. So as long as you don't write anything sensitive in plaintext in the first place when configuring guix, you should be good <meaty>keep all your passwords, keyfiles, etc. in userspace <meaty>or don't specify them in config.scm i think <divya>homo: Indeed. For now I'll send the patches. 44 commits is a bit too much to fiddle around. <homo>divya I just tested that "guix style inferno" formats without error in file that I added as "gnu/packages/inferno.scm" <divya>Also efraim the patch series is sent, check it when you get time <homo>but the style it generates is hard to read <divya>I was doing guix style package-1.2 when it should be guix style package@1.2 <divya>So there inevitably are indenting issues. <divya>ACTION panics about restructuring the 44 patches <homo>yes, so it's best to not run "guix style" and instead manually follow style of other packages <amano>Can I execute any code in G-expression? <divya>homo: I did, but sometimes you have them nonetheless. <amano>For example, I want to generate a password prompt. <divya>There should be an Emacs command that fixes the style in marked-region <divya>Also should there be some indicator in the patch to which branch it's directed? I sent everything from the rust-team branch but just to confirm <divya>Yep it should have rust-team alongside PATCH <divya>Should I send the patches again? <amano>It seems I can execute arbitrary code and also create files and directories in g expressions. <amano>Meaning I can insert scripts as packages into initrd and execute dropbear ssh server for unlocking LUKS2 volumes in initrd. <rekado>homo: "guix style" is in fact recommended (or the use of an editor that follows the rules in .dir-locals). You just don't run it on the whole module. If there are concrete bugs in the style it produces please report them. <rekado>it is not useful to develop a culture that learns to avoid tools because they have unreported bugs. <divya>Indeed, rekado guix style is a good tool. I just became clumsy with these commits. <divya>rekado: Can I use guix style in Emacs to fix marked regions? <amano>How do I start and stop a shepherd service in initial RAM disk? <homo>rekado 1. this particular bug seems to be obvious to developers of style 2. I don't follow the culture of using tools to write code anyway <divya>homo: You don't use linters? <homo>divya I use "guix lint" only before I submit patch, not while writing code <divya>homo: Well that's the point of these tools, they don't help while writing, they just ensure whether your code is formatted as it should be <homo>and I judge programming languages by how easy it is to write them without such tools <divya>sneek: Tell efraim to let me know (in mail) if I should send the patch series again with the `rust-team` prefix included. <sneek>efraim, divya says: to let me know (in mail) if I should send the patch series again with the `rust-team` prefix included. <dckc>secret-tool ... I'm struggling to find it... <lilyp>it's "later tell" if you don't want sneek to echo immediately <Kolev>I'm having issues with LibreWolf. I can't share screen in Jitsi Meet in Sway. It doesn't remember my logins, either. <homo>divya recently the patch I sent was merged into kernel-team branch without me doing any prefixing <divya>Okay, thanks for letting me know homo. Last time I had the prefix but forgot this time. <divya>Kolev: You might have librewolf's history and cookies disabled. Check the configurations <Kolev>divya, idk i just installed it. using defaults <homo>divya if you already sent patches, I recommend to contact rust-team and ask them to merge those patches into their branch <divya>homo: Indeed, rust-team has only efraim so yeah <divya>I just sent the patch to add me in it as well <Kolev>I can't find a browser that will use Jitsi Meet screen sharing. <dckc>where does python get its TLS cert root stuff? I'm getting CERTIFICATE_VERIFY_FAILED from diigo.com <divya>Kolev: I've used librewolf and ungoogled-chromium with JITSI <Kolev>divya, but no screen sharing <divya>Kolev: It's a very much librewolf setting thing. I've encountered it previously as well. <homo>efraim I noticed there is libpijul package, is there any blocker for actual pijul program? <efraim>I just packaged pijul on Friday pushed it last night <efraim>I was really hoping they'd have a stable versioned release, but they've been at 1.0.0-beta9 for almost a year now (for pijul itself), so I suppose that's close enough <dckc>hmm... my `guix shell p1 p2 p3` command is saying: building CA certificate bundle... <dckc>blech: secret-tool: Failed to execute child process ?dbus-launch? (No such file or directory) <dckc>adding dbus made that go away, but now: <dckc>secret-tool: The name org.freedesktop.secrets was not provided by any .service files <dckc>ACTION gives up on --pure for now <amano>Can I run arbitrary commands in raw-initrd's pre-mount? <amano>pre-mount runs a G expression. <amano>I want to run things like dropbear ssh server in initial RAM disk before devices are mapped. <efraim>divya: why package bottom from the git repo instead of from the tarball on crates.io? <divya>efraim: Rutherther recommended, it's due to the absence of some test files. But alas as it turned out, it only fixed some but not others. <efraim>ok. I'll try it both ways and see how it works out <efraim>I'll also see about installing the completions :) <efraim>there's a couple of clap- crates, so that normally means completions, and a clap-mangen, so we can generate the manpage(s) too <Rutherther>ggVGc: I don't know why there aren't prebuilt packages for arm none eabi, probably just no one added any yet. There is for example qmk built for avr architecture, so I suppose arm could be added there as well. I personally use qmk, but build it only with guix deps, not using guix <Rutherther>divya: I recommended that only insofar as the test files are missing, if they aren't or are not used because you skip those tests, I don't see a point in getting it from git. <sneek>cbaines, you have 1 message! <hubslade> I am trying to run a guix VM on Ubuntu aarch64 but it fails at grub-mkimage with the arguments: ("-O" "i386-pc" "-o" "core.img" "-p" "(hd0,msdos1)/boot/grub" "biosdisk" "part_msdos" "fat" "ext2") since there is only /gnu/store/8sq0gmxs18w682m0snf8xww2wdc8nww8-grub-2.06/lib/grub/arm64-efi available on aarch64. <efraim>is that with 'grub' or with 'grub-efi' as the bootloader? <efraim>for aarch64 IIRC you have to use grub-efi-bootloader or grub-efi-removable-bootloader <hubslade>I am not sure if there should be a match case %current-system in grub.scm when calling grub-mkimage <efraim>there should definately be something, but right now it hardcodes i386-pc <hubslade>i know, i am looking at the src right now building a system where i hard coded arm-efi into it <efraim>perhaps not (%current-system), but we should probably pass a system parameter somewhere since there's no ambiguity about which hardware it's being run for <hubslade>but it always takes forever on this hardware <efraim>for my pinebookpro I'm using grub-efi-removable-bootloader <hubslade>is it aarch64 as well? Can you run a VM atm? <efraim>pinebookpro is aarch64. Currently the aarch64 kernel config doesn't support using qemu-binfmt and I haven't tried running any VMs on the pinebook pro or of aarch64 machines <hubslade>I just want to run guix on an aarch64 VPS and have to somehow install from Ubuntu <hubslade>Is there a commit thats known to work on aarch64 or does anyone have some tips for me, maybe? <efraim>I would go with 'guix system reconfigure' to switch the machine while it's running from Ubuntu to Guix, although you'll have to run it more than once as youll need to move /etc out of the way at some point <hubslade>Does guix system reconfigure work on a foreign distro? <janneke>hubslade: what would you expect it to do? <hubslade>I would expect it to say that its not on a guixSD and that there is no current-system <hubslade>But apparently I was wrong, It only tells me that it cant determine provenance <x8dcc>Hello, I am not sure if I understand something related to 'guix home'. Every time some of the packages I installed through my home config (and not through 'guix install') get updated, I should 'guix pull' and then 'guix home reconfigure' again? As far as I know, it doesn't let me upgrade packages normally with 'guix upgrade ...'. <jakef>x8dcc: guix upgrade is an alias for guix package -u, which upgrades packages in the default user profile, not the home profile <x8dcc>tschugg: Yes, I know it's enough, I was wondering if it was "too much" :) <x8dcc>jakef: I see, thank you. I am still not completely familiar with the "profile" concept <yarl>This is not possible to declare a home configuration for an user inside a system configuration, right? Could be nice for 'system image' or 'deploy'. <df>it was originally complaining that my explicitly-installed version of ffmpeg was the latest one whereas mpv depends on 6.1.1 <df>so I thought explictly installing ffmpeh 6.1.1 would help... <df>but somehow I seem to have ended up with different versions of ffmpeg 6.1.1 <hubslade>--allow-collisions Allow colliding packages in the new profile. Use at your own risk! By default, guix package reports as an error collisions in the profile. Collisions happen when two or more different versions or variants of a given package end up in the profile. <yarl>df: What were you trying to do? <df>it's not really a problem, I just want ffmpeg and mpv both in the profile, I don't really mind what versions <yarl>df: Then just install mpv <df>yarl: install both ffmpeg and mpv (which depends on the former) <yarl>ffmpeg is a propagated-input of mpv so ffmpeg will end up into your profile if you install mpv <df>yarl: ah ok, that didn't occur to me, thanks <df>I thought maybe the propagated version of ffmpeg would only be visible to mpv <yarl>df: That's the meaning of 'propagated', see "(guix)package Reference" <yarl>And that's why you get a collision. <df>finding it a bit of an uphill learning guix, but it's fun, reminds me of the first time I installed redhat (with no prior unix knowledge) <yarl>df: Well, this particular mpv/ffmpeg might be consider a bug, kind of, if you take a look to the package definition <df>because it doesn't actually specify an ffmpeg version? <yarl>df: Because ffmpeg among many other packages are in the propagated inputs <df>oh I see (I think), just saw the comment <yarl>df: If I understand it correctly, it's just a trick to please the build, but this has the consequence of putting many undemanded packages in your profile. <df>ok, cheers for the help, will have to read up on propagated-inputs <yarl>So if you try to install one of these, either your are very lucky and it's exactly the same, or there is a collision. <yarl>df: And, just to close my answer, because it's guix (see "(guix)Managing Software the Guix Way"), 'the same' doesn't mean the same version only. <yarl>I have to leave, see you. <hubslade>I have another aarch64 question. Trying to run an aarch64 VM on x86. <hubslade>I get an internal compiler Segmentation fault trying to build gcc-11.4.0 <rekado>Kolev: screen sharing under sway is a little tricky <rekado>and in my system config I globally install these: xdg-desktop-portal, xdg-desktop-portal-gtk, xdg-desktop-portal-gnome, and xdg-desktop-portal-wlr <graywolf>I noticed that there is core-packages-team branch. If there is anyone from the core-packages-team, would you please consider 73660 for inclusion? The patch was sent 6 Oct 2024 and there is no feedback yet, but I still think it would be useful. Due to the potentional for causing lot of rebuilds (I am unsure how to tell how many) I am hesitant to just apply it to my guix tree... <janneke>graywolf: as far as i'm concerned, that would be a (good) candidate for core-packages-team but i don't want to trigger a world rebuild just now <janneke>also, i'd like for civodul to give an ack on it <graywolf>"i'd like for civodul to..." that makes two of us ^_^ <meaty>I have noob questions about packaging rust programs <janneke>graywolf: what are your thoughts about the commit message? <meaty>if a build fails in the middle of compiling a bunch of dependencies, how do I know which package failed? <janneke>ACTION applied it to core-packages-team with a "avoid world rebuild revert" back-to-back but i noticed the message is a bit bulky <graywolf>I always felt like commit message should explain the whats and whys, so while I admit it is fairly long, I still think it is a good idea <graywolf>However I will not get offended if you decide to truncate it somehow <graywolf>I admit it is (I think?) longed than the diff itself <janneke>graywolf: if it's intentional, then I'm fine with it <janneke>meaty: there should be a "guix build: error: build of `/gnu/store/...package-x.y.z.drv' failed" <janneke>View build log at '/var/log/guix/drvs/xx/yy-package.drv.gz' <meaty>janneke: so if the log is named after the "trunk" package, then the compilation error is either in the package itself, or compiling something vendored? <janneke>meaty: yeah, so "wallust" failed and you can check the log why <janneke>ACTION goes for a fourth iteration of icedtea-7 <meaty>I tried to fix it by adding rust-libgit2-sys-0.17 to the package inputs, but that didn't do anything <meaty>janneke: yeah, the log is just the same thing <meaty>wait, maybe i should try putting it in #:cargo-inputs? <efraim>meaty: try adding pkg-config and libgit2-1.8 to your package <meaty>efraim: that fixed it, thank you! <homo>running "guix shell hedgewars" downloads a lot of "ghc-*" packages, even though substitute for hedgewars is available and so I didn't build it <homo>and output of ldd shows that those are not dynamic libraries, hedgewars links with them statically <User-42>i'm trying to define a simple home-service to run a daemon, however i cannot find any examples online or in the docs on how to actually do this. When trying to copy what exists in (gnu home services sound) for example, i'm left with this definition: https://pastebin.com/1cUDbPDi - however, when i try to include this in a home profile with (service <User-42>home-ydotoold-service-type) i get the error "Wrong type to apply: (#<<shepherd-service> documentation [...]". Other services work fine in the same place, what am i missing? <graywolf>User-42: home-ydotoold-shepherd-services should be a procedure returning the list. <graywolf>It takes one argument, the configuration value. In your case you can just ignore it. <User-42>urg, i knew there was probably something trivial i'd missed. <meaty>janneke: I see it, what about it? <graywolf>TIL anyone can restart the failed build on ci.guix. <graywolf>janneke: I saw no error in log, have no idea why it was marked as failed <meaty>if creating a package via import leads to a bunch of packages upgraded/added, how should we break up the patches, if at all? <meaty>Also is there a way to at least partially automate the creation of standard commit messages <efraim>also if you're working on rust packaging you should checkout the rust-team branch <graywolf>meaty: There are some helpers for tempel and yas in etc/snippets. I sent a patch adding also snippets in the built-in `skeleton' mechanism, but abbandoned and closed the bug due to no interest. <graywolf>For the rest of the commit message, magit-commit-add-log helps as well. <meaty>Is there a way to combine importing packages and refreshing them? when I recursively import I end up with some duplicate packages with slight upgrades, and I have to manually correct them <hatim>I'm in the early stage of getting my Emacs config working. I have Guix Home install Emacs, and am using the RDE emacs home service. Now I'm feeling it has added friction for quick development and testing out my Emacs config. I'm relatively new to Emacs/Guix/GNU world. I feel I'm missing something. How do people balance strictness and flexibiliy with their custom Emacs configs in a Guix system. <graywolf>I personally just do not use emacs home service. I have init.el in my git, and install it using home-files-service-type. That allows me to use emacs --init-directory my/git/files/emacs for quick iteration when needed (but usually I just C-M-x in the init.el). <graywolf>Probably not "the guix way", but for something of emacs's complexity, it seems to work the best (for me). <graywolf>The init.el itself takes care of starting the emacs daemon <simendsjo>Is there a service type to extend PATH? I want it extended through a service in my home configuration to avoid having the knowledge in my bashrc. <sneek>orahcio, you have 1 message! <sneek>orahcio, homo says: nice, meanwhile I switched back to gnome, because it still offers more convenience <dckc>so I hacked-and-slashed my way to getting my iPhoto/datasette web service working in guix. Now I'd like to properly engineer it as a shepherd service. <dckc>anybody got an example of a python web services running as shepherd service to recommend? <dckc>bonus points if the service is isolated: it gets some RAM, some CPU, a specific part of the disk, and shepherd hands it a network port to use <dckc>read-only disk might be good enough <dckc>well... logs... let's see... best if shepherd handles those too. <dckc>"Since syncthing already has a -logfile flag and it has built-in log rotation that sounds better than using shepherd's #:log-file option." odd. <jason>I'm trying to `guix pack -S /bin=bin -RR guile-next artanis guile-srfi-197 ...` just a handful of guile modules but `bin/guile` does not already have the extra modules on it's load path. Is that expected to be working or should I be putting the modules on the guile load path myself? <david95>Hello everyone. Please tell me what this error means? <graywolf>Would you happen to have a pastebin service that does not require javascript? <graywolf>Interesing, guix copy *first* copies the 5 GB only to later tell me I need to authorize the public key. <david95>On the contrary, I left it, because there are even more trackers and other garbage on this site <homo>paste.debian.net - no javascript required, compatible with netsurf <homo>I noticed a bug - if both qt5 and qt6 are installed as dependencies of different apps, then qtwayland can be installed for only one of them, otherwise guix throws an error about qtwayland 5 and qtwayland 6 being in same profile <efraim>I'm not sure if I've submitted it upstream yet :/ <orahcio>Hi homo, gnome is very convenient, maybe I come back to Plasma again (I was on KDE side before the WM), but now the sway is being very nice for me <homo>orahcio kde 4 was my favourite de, but now it's completely abandoned :( <homo>after kde 5 it's pure dissapointment with kde <homo>efraim thanks, currently it's not upstream according to "guix search" (and grepping *.scm files) <homo>ideally qt5 apps would just use qt5compat from qt6, but I guess it's not as compatible as sdl12-compat <orahcio>homo: in a oposite way, I changed the side between gnome 2 to gnome 3, and now just I left Plasma to go to Sway <homo>orahcio except gnome 2 was forked into mate, on top of that gnome has gnome classic that imitates gnome 2 experience, but nothing like that happenned to kde 4 <homo>it's a puzzle why projects that make such major breakages don't change their names to avoid confusion <podiki>i just saw the "in-vicinity" function in some guix code, but can't find a definition of that in guile docs at all (other than it used in an example for something else) <orahcio>homo: I think KDE changed a lot but the philosophy about customization is the same. Like Mate, KDE 3 is now Trinity (TDE) <podiki>and for whatever reason srfi-59 is not in the guile docs; i never understand how that thing is organized <homo>orahcio except I complained about death of kde 4, not kde 3, I have no interest in trinity <homo>also it's very hard to make both kde 5 and kde 6 look like kde 4 <homo>the end result is broken appearance, already tried that <homo>but thanks to gnome I don't miss kde 4 anymore, I'm content with what I have <homo>except having 3D widgets for everything, and both haiku and oxygen icons are incomplete for today xdg standard, as there are some missing icons in apps <orahcio>homo: I think there are just a couple of themes on Plasma that make different apearence withou broken something <ieure>Okay, what all do I need to do to build aarch64-linux binaries on amd64 hardware? Manual "(guix) Additional Build Options" says: Similarly, when transparent emulation with QEMU and ‘binfmt_misc’ is enabled (*note ‘qemu-binfmt-service-type’: Virtualization Services.), you can build for any system for which a QEMU ‘binfmt_misc’ handler is installed. <ieure>I added this to my operating-system services: (service qemu-binfmt-service-type (qemu-binfmt-configuration (platforms (lookup-qemu-platforms "arm" "aarch64")))) <ieure>The service is running: Status of qemu-binfmt: It is running since Mon Dec 30 20:47:07 2024 (6 days ago). <ieure>But, if I run `guix build --system=aarch64-linux guix', I get; while setting up the build environment: a `aarch64-linux' is required to build `/gnu/store/6wwbzsry0c3sm5s0aifgj4sfzahmkk30-guix-1.4.0-31.121e96d.drv', but I am a `x86_64-linux' <dthompson>why node 22 and not 23? last I checked 23.4.0 was the latest <jlicht>dthompson: running non-lts versions of node is something I wouldn't wish on my worst enemy ;-) <homo>orahcio broken appearance is the result of constantly evolving standards, avoiding broken appearance requires participating in race to create new icons, new widgets and even migration to new API, which means adwaita and breeze will always have advantage over everything else, oxygen is the best theme created by kde, but kde decided to abandon its further development <dthompson>jlicht: in any case, the latest 22 should be fine. we found a wasm bug in some 22 version but I think it's fixed in the 22 series <jlicht>with less sass: we could have a node and node-lts package; I would just have node-build-system depend on node-lts by default though <jlicht>mumi is still doing its thing, so might be some minutes before the patches come through <dthompson>nice to have the latest for having a relatively fresh v8 <homo>orahcio worst case there is something hardcoded that forces adwaita and breeze look, like you've seen in screenshots I sent earlier where I installed haiku-gtk-theme, but some apps still stubbornly use adwaita theme <orahcio>homo: yes, I liked oxygen very much, I like random wallpapers and the breeze icons color changing with the wallpaper is awesome, maybe I miss just that setup of the Plasma <homo>sounds like those are translucent or symbolic icons <jlicht>dthompson: I'll most likely trick myself into working on more node up(/side?)grades, but probably only after guix days <dthompson>jlicht: hehe sounds good. you've already done the lord's work here by upgrading to the latest lts. I will try it out tomorrow, probably. <homo>orahcio I'd say creating fsdg-compliant distro of haiku is the easiest and most interesting way to go, as they don't plan abandoning 3D theme and they create new 3D icons for apps that they port <orahcio>homo: the breeze theme has folder, folder-yellow, folder-magenta, etc. for each kind of folder icons, and the Plasma changes to the color closest to wallpaper <homo>as unfortunately the rest of the world is flat <homo>of all flat icons that I've seen, paper is the best, even though it's not 3D-looking, it looks like someone was cutting colored paper and combining it into shapes and that looks very cool <yarl>ieure: Well, this works for me, maybe restart... <hatim>where do poeple maintain additionall profiles, under ~/.local, ~/.cache etc? <yarl>hatim: take a look at "(guix-cookbook)Advanced package management" <orahcio>homo: Gartoon is the best flat icon theme :) <yarl>About offloads, "(guix)Daemon Offload Setup" says "a single machine can have multiple system types [...] via emulation. Is it possible via cross-compilation? <h4>How to `sudo guix system undelete-generation 4 && sudo guix system switch-generation 4`? <homo>orahcio wow, thank's for recommendation, it's funny there are people who say that kde 4 looked very cartoonish/childish <ieure>h4, There is no undelete-generation, once they're deleted, they're deleted. <orahcio>I used a long time ago that theme, maybe I change again on my Sway, I like to use pcmanfm to manage files, and a good icon theme is very nice, the yellow folder is the best <homo>well, haiku-icon-theme comes with yellow folders <homo>what that theme forgot to bring are haiku cursors <orahcio>I don't unsderstand the blue preference, but the brown of Ubuntu or green of Mint or Manjaro are awful <h4>ieure: Even before garbage collection? <homo>orahcio we really need fsdg-compliant distro of haiku, as that system targets desktop users specifically, there are cool features you won't get in any xdg-compliant desktop, be it x11 or wayland <yarl>h4: I don't know if /gnu/store/*-system stays. <orahcio>Nice homo, with Guix I could to know the hackneyed-x11-cursors, now I could to config that cursor to every applications, qutebrowser was the last to insist to use the gdm default cursor <homo>it doesn't make sense for qutebrowser to communicate with gdm <homo>just like it doesn't make for claws-mail to depend on networkmanager <yarl>I guess you can try to reconfigure to the old configuration.scm you can find. using the channels.scm too. I guess if you did not gc, it should work. <yarl>But how can you tell it's the 4. <orahcio>homo: I don't know why, but I just deleted gdm and installed qtwayland, I think I need to remove qtwayland to see if the black cursor comes back again on the qutebrowser <homo>orahcio if you have both qt5 and qt6 installed, setting them up to use theme you prefer is pain, as they read same environment variables, you can't rely on qt5ct and qt6ct to setup themes <homo>orahcio that problem comes from the fact that on sway you need to have QT_QPA_PLATFORMTHEME variable, and your only available options are: windows, fusion, qt5ct, qt6ct, kvantum, lxqt <homo>so, when both qt5 and qt6 are installed, the only compatible options are windows and fusion <homo>why so? because every theme in qt is a .so (shared) library <homo>also I recommend doing "guix shell gnome-tweaks -- gnome-tweaks" to setup theme for gtk2, gtk3 and gtk4 <meaty>why do rust packages specifically get installed with source dirs buy default? is that not a huge waste of disk space <h4>yarl: I was on 2, I deleted everything despite wanting to only keep 4 <h4>yarl: I still have `$ file /gnu/store/*-system | wc -l\n59` despite having deleted all generation but one, so it's not impossible to retreive it, isn't it? <ieure>yarl, Rebooted, I get the same behavior. <ieure>Oh, wait... wrong machine lol <meaty>How can I tell guix to keep a successful build tree? <meaty>I'm wondering whether some files aren't being built or aren't being installed <ieure>meaty, You can't and I don't think you need to, the output will be in the store path printed when you build. <cluelessguixer>Apparently I can't make snapshots in virtual manager because I'm missing vfio migration? Is it as simple as just adding vfio_migration in here? (initrd-modules (cons* "vfio_pci" "vfio" "vfio_iommu_type1" %base-initrd-modules)) <ieure>meaty, You can add a build phase that fails and keep the result if you need to poke around the source tree. (/ 0 1) is an easy way to cause a failure. <homo>efraim there is a puzzle in that qtwayland-helper package, as only qtwayland@6.7.2 gets installed, but not qtwayland@5.15.15 <orahcio>thanks homo, now I could to manage qt and gtk theme and just qutebrowser was the difficult application, pcmanfm also was a little difficult to set another theme instead of adwaita, but without gdm I need just setup the gtk settings.ini file <homo>orahcio ideally qtstyleplugins would be packaged for both qt5 and qt6 and you would select "gtk2" or "gtk3" as QT_QPA_PLATFORMTHEME, that would give qt apps same look as gtk apps <homo>qt can handle gtk themes, but gtk cannot handle qt themes <homo>again, because qt themes are .so libraries <homo>I wouldn't even have qt if I wouldn't install hedgewars and jami <orahcio>Good information, I will see qstyleplugins. I like qutebrowser, also I use pass to manage passwords, and the qtpass is great <homo>orahcio well, you'll need to submit a package, good luck with that, I am still too scared to attempt compiling anything related to qt <orahcio>homo: thanks, I will need of luck, a lot of it <meaty>what is the correct path to install bash completions in on guix?/do I need to to anything special to test them in 'guix shell' with a package <rekado>looks like the least-authority-wrapper prevents the bitlbee-service-type from talking to the local mDNS server, so the bonjour protocol doesn't work. <dthompson>wellp, time to rewrite least-authority-wrapper using capabilities ;) <meaty>does anyone know how I can test that my package is installing its bash completions properly? <whereiseveryone>does gnu shepherd have any dag limitations with scheduling services? <lh>does anyone know what software logs.guix.gnu.org uses <lh>the queries seem to limit results so I would like to limit by date ranges, but can’t seem to figure out the syntax or even if it’s supported <Kolev>How do I share screen in LibreWolf? <ieure>Kolev, Not sure what you're asking. <Kolev>ieure, screen sharing in Jitsi Meet does not work in LibreWolf by default... <ieure>Kolev, Been a while since I've done this, but I think you get prompted to allow it when you start it in whatever meet thing. Tried to see if it works for me, but jitsi requires an account to do anything now. ugh <Kolev>ieure, my browser asks for permission, and I grant it, but it does not show my screen or their screen. <ieure>Kolev, Do you have canvas access enabled? <Kolev>ieure, I'm using LibreWolf defaults. <ieure>Kolev, Canvas access is disabled by default. Click in the lock and/or shield left of the location bar and disable RFP / enable canvas. <ieure>Kolev, Beyond that, probably need to ask LW upstream for help.