IRC channel logs
2023-08-04.log
back to list of logs
<Guest28>Altadil: Don't have any experiences with gnome-boxes, but I use virt-manager USB passthrough several times without any issue. It is definitely possible (gnome-boxes uses AFAIK the same backend as virt-manager). <Altadil>Guest28: Thanks. I was going to use virt-manager, but for some reason it cannot find any network interface when trying to create a VM, so I got stuck there. Seems I’m just unlucky (or cursed ^^) <Guest28>Altadil: I had the same issue with virt-manager as well. Right click on that qemu:///system and click on details. Go to network and create one with the default settings. Now you should have a simple NAT configured for the VMs that it will use automatically <Altadil>Guest28: Thanks a lot ! For some reason, the guest OS installer still sees no internet connection, but I’ll investigate more <Guest28>Altadil: If your guest OS is Guix, you need to go through the installer and finish the network section, else network isn't configured on that type of ISO. If you use something else, it should work out of the box <Altadil>Guest28: it’s xubuntu 23.04, so I still have challenges ahead it seems <Altadil>Gonna give up for tonight, thanks a lot for the help ! <Guest28>yep, seems like bordeaux isn't responding <GNUtoo>Hi, I'm thinking of depending on Guix for several uses cases (for instance benchmarking) and I was wondering how to integrate guix <GNUtoo>Is it for instance possible to have a script that takes care of building and runinng Guix in standalone mode? Or if people already have Guix, would that conflict with the existing Guix installation? <GNUtoo>Yes, like compilation benchmarks <Guest28>Thought you were kinda pointing to overclocking <GNUtoo>For instance running that: guix time-machine --commit=v1.4.0 -- build --no-substitutes --no-grafts --no-offload --rounds=10 --check hello <GNUtoo>And plugging that into phoronix-test-suite <GNUtoo>Basically phoronix-test-suite tests tend to break over time but that looks easier to deal with <GNUtoo>So that works for me because I've Guix installed, so I made local tests and tested it once <GNUtoo>But then if I want to somehow contribute upstream or run my own repository, there is the question of how to integrate guix into all that <GNUtoo>Like should I just ask users to install Guix? <GNUtoo>Or can I just write a script that would run guix locally <GNUtoo>If sudo is permitted I'm pretty sure I don't need to care about compatibility, like I don't need to recompile to a different prefix and so on <GNUtoo>But I would still need to handle all the cases (guix present, guix not there, etc) <Guest28>I can't help you with that question (new to Guix). Also it is late, most people are probably away <GNUtoo>ACTION will also try publish his working test somehow <isf>hello, i know in the website say "linux based operating system" <isf>How can an operating system created since 1983 be based on a kernel released in 1991? This concept confuses me. <GNUtoo>I think it's easier to understand if you want to somehow replace Linux <GNUtoo>Even Google didn't manage to do that with Android <GNUtoo>And even some BSD kernels take code from Linux (because not all drivers are GPL) <Guest28>Why should something be created a long time ago, automatically mean it is bad? <GNUtoo>It's not necessarily bad, I only explained that it was hard to replace <Guest28>My answer was to isf. At least that is how I interpreted it <GNUtoo>And the second issue would be replacing it by what? <GNUtoo>The more you deviate the more work it becomes <GNUtoo>For instance some BSD can reuse some Linux drivers (like radeon, ath9k, etc) because they have similar designs <isf>You can put GNU the BSD kernel or Hurd himself, how could someone then call GNU a system based on a kernel that didn't exist when GNU already existed? <GNUtoo>ACTION tried to reuse a Linux driver in a context where there weren't a similar design and it was almost impossible (too much dependencies) <GNUtoo>ACTION didn't understand the question <Guest28>Ah, you mean because of the years? They worked on other stuff that an OS needs. Not the kernel itself. That is why the OS is later than the actual kernel <GNUtoo>Also note that changing kernel is not trivial <Guest28>AFAIK they had already everything for a full fledged os, but no kernel. So they started working on their own and later found the linux kernel <GNUtoo>BSD have different libc for instance, and glibc doesn't have official support for other kernels than Linux and hurd <Guest28>That is why the project is older than the actual kernel itself, if that answers your question <GNUtoo>And even with HURD + glibc I guess that in Guix not everything works with HURD <nckx>GCC wasn't even released until 1987, so GNU is obviously a hoax. <GNUtoo>Ah and there was no distribution like Guix either <GNUtoo>The issue is that there are circular dependencies in all that so it takes time to fill the gaps <GNUtoo>You need a kernel to run the compiler and a compiler to build the kernel <isf>it would even be more correct to say that GNU is an emacs-based system. <isf>since emacs is older than GNU <GNUtoo>People doing that didn't even know if it was going to work out, but they wanted to try anyway <nckx>Emacs and Linux are two different operating systems, although they are often used together. <isf>Linux is not a operating system <isf>is just a kernel, a program with drivers and firmware today sadly with nonfree firmware inside <GNUtoo>ACTION has a joke on that. People told me I should try the Linux operating system, so I tried and my computer said "kernel panic, cannot find rootfs", so I abandonned and used GNU/Linux instead <GNUtoo>+ Android is also very different from GNU/Linux and they both use mostly the same kernel <nckx>I wish GNU had device drivers. I think it would help it compete with Linux. <GNUtoo>I think that HURD can use some Linux drivers but probably very very few <GNUtoo>The big issue when doing that is to have similar frameworks <GNUtoo>For instance if you import a GPU driver, that depends on many other frameworks <GNUtoo>like the backlight driver class for instance <gnucode>The Hurd is working on reusing netBSD's device drivers currently. They can boot a HDD (maybe an SSD even) via the NetBSD drivers. <nckx>GNUtoo: or just say f— it and rump someone else's kernel. <GNUtoo>ah ok, that makes sense since they can run in userspace <GNUtoo>Linux also somehow works in userspace but I'm unsure if it's possible to plug it to real hardware easily <nckx>ACTION -> zzz. Please troll politely, friends. <GNUtoo>It can be used to test the USB stack for instance <iyzsong>%build-inputs is not available in qt-build-system's configure-flags, but available in cmake-build-system, interesting.. 🤔 <RavenJoad>podiki: I managed to get a local system VM running using the qcow2 image method in (guix) Invoking guix system, under the image action. <RavenJoad>system vm mounts something from /gnu/store directly as / in the spawned VM, which breaks some assumption Cuirass has. <cnx>how do i add python module for test phase of a package not built be python-build-system? <nmeum>I just installed Guix 1.4.0 on a foreign distribution. I want Guix to use the substitutes servers and I believe I authorized them correctly. However, if I run `guix shell --dry-run cowsay` for example it seems that all dependencies would be build from source https://tpaste.us/LRdp why is that? can I somehow "force" Guix to use substitutes? <iyzsong>does `journalctl -u guix-daemon` have something interesting? <nmeum>I am using OpenRC, but there is nothing interesting in the stdout/stderr of guix-daemon <nmeum>also tried running `guix shell` with verbose and debug output but it doesn's seem to output any information as to why substitutes are not used <iyzsong>what's the arguments for guix-daemon? does it use --no-substitutes or --substitutes-urls with a wrong value? <nmeum>according to ps(1) it is running as follows: /usr/bin/guix-daemon --build-users-group=guixbuild --discover=no <iyzsong>um, do you have nix and some NIX_* envs? <nmeum>oh! is that known to cause issues? I previously had nix installed but I believe I removed all of it and I don't see any nix-related environment variables in the environment of my non-root user <iyzsong>i don't know, sound should not be a problem.. <nmeum>maybe substitutes are just available for the versions it tries to fetch? <nmeum>so for example in the output I referenced above it seems to download coreutils-8.32 but the most recent available version is coreutils-9.1 <nmeum>can I check that somehow? my system is x86_64-linux btw <iyzsong>maybe try 'rm -rf ~/.cache/guix/substitute/', and 'guix pull' <nmeum>oh! but guix pull will itself build all kinds of stuff from source and does not use substitutes for some reason <nmeum>seems to me that it is building gcc-mesboot (and probably the other stuff too) from sourc <nmeum>is .cache/guix/substitutes supposed to look like this: https://tpaste.us/d18j ? specifically the `(value #f)` looks suspicous. can I somehow rebuild the substitutes cache? <nmeum>i think i figured it out: i am missing guile-zstd which is listed as an optional dependency in the guix manual but seems to be required for substitutes to work ':D <lissobone>Howdy. I have set up an apache HTTP web server with the document root somewhere in my user home (I don't know if I am supposed to do so, so correct me if anything). Howerver, apache doesn't seem to be able to access the document root, even in spite of all the permission bits being set to readable+executable. <lissobone>It presents the viewer with a 403 forbidden page. <lissobone>I might be an ignoramus and just not know something very simple, but nevertheless important. I beg thee, enlighten me. <iyzsong>lissobone: i think the apache httpd run as a specified user, which do not have access to user home. <attila_lendvai_>lissobone, test the permissions using `sudo -u [apache-user] bash`. the parent dirs also require +x. <lissobone>Where is it declared in shepherd service configuration files? <iyzsong>lissobone: %httpd-accounts (httpd:httpd) in gnu/services/web.scm <lissobone>Thanx bruv (now I know that the accounts are specified like this). <lissobone>Oh, the default for document root is /srv/http! <lissobone>Welp, there's also a httpd-config-file-user field. <lissobone>I better set it to my user, since it's in my home directory. <iyzsong>that will work too, but be aware of security impact <lissobone>I don't really care since I want to share a meme. <iyzsong>if you make it public, anyone can access all your users files under document-root... <lissobone>All user files under document root? Relax, it's in /home/lissobone/www/memes/ <lissobone>No one will EVER know how to access /home/lissobone from within there! <lissobone>Also, in gratitude, I will share my meme with you, folks. <lissobone>Beware, it's HTTP and hosted on my computer. <iyzsong>haven't played minetest yet, can't get it lol <lissobone>cuz there's a radioactive material in nodecore (a game for minetest) called lux and it does bad things to u (like any ionizing radiation lol) <iyzsong>if apache httpd have some serious bugs, in worst case, running it as a normal user will allow anyone to remote run any commands as that user.. <lissobone>And? If I could, I would have removed all passwords and permissions from my computer. <ulfvonbelow>anyone know a good way to run X11 programs in a container with hardware acceleration without giving access to the entire display? <nmeum>apparently my "guix uses no substitutes"-issues did not fix itself after installing guile-zstd :( <nmeum>now kind of poking around in guix/substitutes.scm and guix/narinfo.scm to figure out what's going on but haven't had much success so far <nmeum>let me know if someone has an idea what I might be doing wrong or any pointers regarding code to look at <nmeum>for cowsay specfically, it seems that a substitute is available for zqjfh5r8zmagcl097szfg13f7pykc76g-cowsay-3.7.0 but not 52riipmf2ha1p9qz5nijmdy99dvll8aj-cowsay-3.7.0 (which is what my guix is requesting). how can I figure out the difference between the two? and determine why my guix is requesting the latter? <jx0>morning guix! Hope everyone's friday is going to plan... <zmhanham[m]>hey all! was curious if there was a way to identify unused imports in guile scripts? Don't rly want to delete one at a time and see if it still works and also too lazy to scan through the imported scripts to check usages. ( I really wish guile had namespaced imports... ) <unwox>zmhanham[m]: you can specify #:prefix for imports <unwox>or use #:select to only import needed symbols <zmhanham[m]>ah select sounds like just what I need. Will just delete one at a time in my script and use select moving forward. <unwox>but i'm curious if there is a tool for detecting unused imports too <zmhanham[m]>been awhile since I looked at my guix stuff so i forget what everything belonged to <zmhanham[m]>would imagine it would be not too difficult to implement unless there're some caveats i wouldn't know about but you could do like so: for each import build a set of it's exported symbols and then scan the symbols in the mains script. If no symbols found then it's unused <unwox>static analyzer for guile in general would be great <apteryx>is someone using keepassxc with a yubikey? <nckx>bug-guix at gnu dot org. <nckx>Did it used to be accessible? <Altadil>oh, so the same as for a bug in the software, ok ! Yes, it used to, but I do not know since when it doesn’t anymore. <nckx>Be prepared for ‘datacentre firewall blocks Tor, CANTFIX’ but we'll see what we can do. <nckx>Yep. All bugs are good bugs. <nckx>It's probably the firewall. CI is also inaccessible. <Guest28>I use Guix with my own channel. It is a local directory. Now, how would be the workflow if I want to reinstall the system, since home manages my packages and is setting ab the channel for guix. Would that mean that I first install the system via my system config and after that boot in the new machine, manually add my local channel and then <Guest28>run guix home reconfigure home.scm? I wonder if I could run it while still being on the live cd so that I only need to run system init [...] and guix home reconifugre [..] and could boot directly in the new machine with everything setup. <Guest28>Basically 2 questions, can I use guix home init or something in the live cd or is it required to boot in the new system and how I could fix the problem that i use guix home to add the channel it requires but at the same time that channel is required to install additional packages <nckx>Guest28: That sounds like a great feature request. If it is possible already, it is not documented AFAICT. <nckx>As for the channel: any reason you can't edit /etc/guix/channels.scm on the live system, pull, and continue happily onward? <nckx>I guess that, would I ever switch to Guix Home, I'd still not use it to manage my channels for exactly this reason. <Guest28>No, I can do it but thought maybe it is possible to kinda fix that bootstrap problem <Guest28>Feels just kinda wrong to do it that way <nckx>Probably little consolation, but doesn't feel wrong to me. <nckx>And as for the bootstrap problem, I was serious about the feature request. You could send it to guix-devel at gnu dot org if you like. <Guest28>I wonder, would it be hard to fix this? Basically if channel is updated, run some function to pull the new repository and after that go in with the reconfiguration <nckx>With ‘great feature request’ I meant some way to point ‘guix home’ to an inactive system, just like ‘guix system init’. <Guest28>Do you may now when guix-hardware is checked? I wrote an email about some cloud ARM server I would like to donate, since if not usable, I would delete that instance. <nckx>I'm less sure about the ‘guix system init based on channels stored a Guix Home configuration’. <nckx>Guest28: Nope, I don't even know who's behind that alias. Which is mildly embarrassing. <nckx>apteryx, efraim: Do you? <nckx>I never bothered getting fencepost access. They wanted to video chat with me… <Guest28>For that feature request, do I need to add some special stuff to the subject? Like [feature] guix home init? (don't actually now a good subject name to be honest) <nckx>Where are you sending it? If guix-devel, no. If bug-guix (because you mentioned ‘ticket’ above’), yes, but I'd prefer to keep new feature requests and bugs separate. <Guest28>I am going to send it to guix-devel as you said <nckx>Cool. No special conventions, write from your heart. <Altadil>nckx: yep, I just saw the email. I had completely missed that, thanks a lot ! <podiki>we haven't switched to linux-libre-6.4 as the default, does anyone know why? (if there is a reason) <unmatched-paren>podiki: my laptop's graphics don't work on 6.4, though i have to use the nonfree kernel <podiki>unmatched-paren: thanks. a known regression on 6.4 vs 6.3? <unmatched-paren>because ``guix shell -Df guix.scm'' does, but the former prints: ``no packages specified, creating an empty environment'' <podiki>yes, at least when I use that with the scm file ending in a package variable <unmatched-paren>guix shell -D -- meson compile -C out -> no packages specified again, even though it works explicitly specifying the file <apteryx>viaken: anything special to get the key detected? <apteryx>I have a yubikey and keepassxc says there's no key detected, although I'm using it for 2FA in browsers <viaken>apteryx: I'm having to use a VM with USB passthru because I couldn't get it working in Guix. <viaken>All the relevant /dev nodes have rw, but apparently that's not enough. <viaken>oh, and pcscd's logs don't show anything suspicous <podiki>unmatched-paren: hmm not sure; usually I would have just recette at the end (having defined it earlier) <nckx>Anybody happen to have a heisenbridge package at home? I need it for $reasons. <ulfvonbelow>is there a good way to change the console keymap back to its original setting after temporarily changing it to qwerty? <ulfvonbelow>it seems that the console keymap is buried inside the initrd, the grub keymap is in a grub format, and the X keymap is in a different format still. <nckx>ulfvonbelow: You might want to look at dumpkeys(1). <nckx>ulfvonbelow: This is as close as you'll get to ‘whatever you have now, this thing, I want that’. <nckx>fries: Yes, you're still connected. Welcome! <ulfvonbelow>I think I probably rather want to "have looked at" dumpkeys, from the sound of it <fries>looks like the matrix bridge is kinda broken right now <fries>anyways, i'll repost what i posted on the bridge <fries>whats the best way to package go apps for guix <fries>i have 2 apps that i think would be nice inclusions <fries>looking at /gnu/packages, i see 2 modules that might be of interest to me. disk.scm and code.scm. i see cloc has been shoved into there but ncdu, an existing disk usage analyzer packaged into guix has its own module. i wonder, should that be shoved in disk.scm? <nckx><matrix bridge is kinda broken right now> You have no idea. <nckx>The bridge is going away. We'll have to spin up our own, or live without Matrix users. <unmatched-paren>fries: Go apps, thankfully, are pretty straightforward; just put all the go-foo-bar-stupid-long-name dependencies into INPUTS, set the (arguments (list #:import-path "...")) to the Go import path, and set BUILD-SYSTEM to GO-BUILD-SYSTEM. <nckx>I'm about to start packaging heisenbridge. (Remember how I always said I wouldn't run a bridge? Well, I was full of bluff and bluster and lies. Apparently.) <unmatched-paren>fries: oh, also, you'll want to add #:install-source? #f to the ARGUMENTS, since installing source is a thing only libraries need to do <nckx>So, really, if anyone is sitting on a half-finished heisenbridge package or service, now ould be the time. <unmatched-paren>fries: other than that, it's pretty much the same as your average C package <nckx>fries: As for the file name bikeshed: don't take existing categorisations too seriously if another makes sense. You don't want to look at admin.scm or linux.scm if you like taxonomy. <nckx>(code and disk look good here.) <nckx>unmatched-paren: Why does this sound like a trick question… <podiki>nckx: I got as far as doing a guix import pypi heisenbridge -r and seeing it wasn't too crazy <nckx>unmatched-paren: It really should work, but I can't back that up with science. <nckx>I've used it before and am certain it's documented, so the code is wrong if it doesn't. <nckx>Oh wait. ‘guix.scm’ isn't something weird that doesn't evaluate to a package, right? <podiki>nckx: I didn't try, I was happy enough that it imported all packages without a cycle and then remembered I run matrix on my arch system anyway... <nckx>Could we technically use your server over the week-end? <nckx>ACTION bridges all of #guix* through podiki's pitato. <podiki>i'd have to see how that would work exactly, but sure <podiki>(so far so good on heisenbridge, except when i stopped/started synapse for my nightly backup and forgot to do the same with heisen, but recovered nicely on starting the service again) <unmatched-paren>"our community infrastructure runs entirely off of this one twelve-year-old raspberry pi 2 I found lying around" <nckx>‘Uptime has markably improved.’ <nckx>It works for a simple ‘(@ (gnu packages foo) foo)’ but not for your entirely reasonable monstrosity. <nckx>ACTION was responding to the ‘nope’.) <nckx>A file with that expression. <nckx>I used (@ (gnu packages ncdu) ncdu) since it was in my brain already. 