IRC channel logs


back to list of logs

<nckx>mbakke: Oh, super cool progress on the man front! To think of all the random things I tried...
<mbakke>nckx: I found by chance that it worked with groff in the environment, but I'll take it :P
<darkpsi>hey guys can not install cabal-install
<darkpsi>reporting missing dependences on guixsd
<catonano>darkpsi: recently there was an issue with some guix modules being miscompiled. It's fixed now so an update to your guix installation cold help
<catonano>darkpsi: I might be wrong, of course
<catonano>darkpsi: so if updating is demanding (maybe because of slow hardware) you couuld wait for some more feedback
<darkpsi>i did a guix pull and guix package update not long ago
<darkpsi>do i need to do a system reconfigure as well?
<darkpsi>just today
<darkpsi>same as before :(
<mbakke>darkpsi: cabal-install fails for me too. Can you file a bug report? Or better yet, a patch? :)
<darkpsi>new to guix and only have some haskell experiance so i will file a report tomorrow
<kkebreau>Anyone here use the Shogun library for machine learning?
<kkebreau>I'm trying to update its Guix package definition, but it tries and fails to download optional third-party libraries during the process.
***jonsger1 is now known as jonsger
<IntoxicatedHippo>Is there a way to change the order that kernel modules are loaded during boot? I'm trying to set up GPU passthrough so I need to load vfio-pci quite early
<IntoxicatedHippo>Is there an easy way to copy derivation outputs between machines?
<catonano>I have an environment with some python packages in it. I'd need several terminals working on the same environment. Is it possible ?
<catonano>maybe I need a profile ?
<janneke>catonano: that's what i do for such projects
<catonano2>janneke: thanks
<mbakke>IntoxicatedHippo: for copying closures, try `guix copy`.
<mbakke>i686 GRUB fails spectacularly on core-updates:
<mbakke>efraim_: By providing "gpgme" with a newer libgpg-error, the build succeeds with gnupg 2.2.7.
<mbakke>Not sure if that's preferable to disabling the broken test.
<mbakke>Currently building all gnupg+gpgme dependents, will push later unless anyone protests!
<davidl>I can't get gdb to work on GuixSD for some reason
<davidl>for example, I try to run: p dup2(open("/dev/pts/5", 1), 1) which errors with unknown return type. Then I tried with p (char *) dup2((char *) open ("/dev/pts/5", 1), 1) which errors with $1 = 0x1 <error: Cannot access memory at address 0x1>. Not sure what's wrong.
<efraim>mbakke: apparently i'm back to efraim :) libgpg-error has over 5000 dependants, for just a version bump on gnupg it didn't seem worth adding a separate one
<efraim>if you're interested there's also a newer gpgme
<efraim>re grub: i hate assembler errors
<mbakke>For the second time in a few days, I had to resurrect a commit by `git reflog | grep title`. I should improve my workflow :P
<pkill9>anyone know of a way to bind-mount /gnu before the guix daemon loads?
<pkill9>i'd like to have the store on a different partition
<mbakke>Oh no, cmake failed for armhf on core-updates.
<pkill9>zybell: not sure, because you specify mounts in the system configuration, so i thought it would be generated
<pkill9>yeah it gets generated
<pkill9>also it needs to be mounted before the guix daemon starts
<zybell>when dou you think / gets mounted? After daemon start? /var ?
<pkill9>basically the daemon starts before /gnu/store gets bind-mounted to /gnu/store as read-only
<pkill9>so basically i think it starts before anything in fstab gets mounted
<zybell>no,the most must be mounted,because the daemon depends on it.
<pkill9>the most?
<zybell>manually mounted backup drives for instance not.
<pkill9>what do you mean 'the most must be mounted'?
<zybell>ok, daemon needs postgres,postgres needs network,network needs /run,/dev postgres needs /var, all run from /nix(rw)
<dustyweb>anyone else in here use hexchat?
<dustyweb>it's suddenly been crashing a lot, I wonder if that's due to the recent update
<jahb>tip: to use native duply with guix duplicity package on foreign distro: set PYTHON="" in /etc/duply/<profile-name>/conf
<civodul>dustyweb: were you able to get a core dump?
<dustyweb>civodul: nope, it's just hanging
<dustyweb>until I kill it
<civodul>perhaps if you can attach gdb to it we can try and see if the Internet knows about it
<dustyweb>civodul: embarassing admission: I've never used gdb!
<civodul>that's fine you know!
<civodul>just get the pid of the faulty hexchat process
<civodul>then start gdb and type "attach THAT-PID", and the "bt"
<az`>hello dear guixers
<az`>I have trouble with installing inside VirtualBox
<civodul>hello az`! what troubles?
<az`>grub: embedding is not possible
<az`>guix system: error: failed to install bootloader
<civodul>sounds like you're trying to use a partition instead of a disk as the GRUB target
<az`>no : (target "/dev/sda")
<civodul>then it should be fine
<civodul>which config.scm did you base yours on?
<az`>I step through installation docs
<civodul>are you sure the file you passed to 'guix system init' has (target "/dev/sda") and not, say, (target "/dev/sda1")?
<civodul>could it be that VirtualBox maps /dev/sda to a partition instead of a disk?
<civodul>also using the qcow2 image at might be more convenient
<az`>there I trying to make luks-encrypted image
<roptat>I recently created a virtualbox VM running guixsd
<roptat>I based my configuration on bare-bones
<roptat>no problem
<pkill9>is there any interest in packaging firejail?
<vagrantc>having trouble with guix pull on armhf-linux (running on debian with a binary install of guix 0.14.0) ... eventual goal is to install guixsd ... but guix pull segfaults
<az`>roptat: you use separate boot partition in ext2, or only one ext4 root partition?
<roptat>only one partition
<az`>roptat: me too...
<roptat>but I didn't try to encrypt anything
<civodul>vagrantc: could you get a backtrace?
<vagrantc>civodul: sure
<vagrantc>civodul: the short of it: # drEagle <>
<vagrantc># Rick Thomas <>
<vagrantc>wrong paste buffer ... gah.
<vagrantc># drEagle <>
<vagrantc># Rick Thomas <>
<vagrantc>apparently it's just segfaulting, no traceback: compiling... 20.1% of 407 filesbuilder for `/gnu/store/r8b08w1i024zyhqw1w2799h4sj10jm8s-guix-packages.drv' f
<vagrantc>ailed due to signal 11 (Segmentation fault)
<vagrantc>though on previous iterations, there were some tracebacks... let's see what the latest pull finishes out to
<civodul>oh so it's a segfault in the build process
<civodul>i'm debugging a bug in this area, it could be related
<vagrantc>ok, good to know
<vagrantc>i'm trying to get guixsd running on the novena... quad-core with 3.8GB of usable ram... probably beats running on a beagleboneblack or something :)
<vagrantc>civodul: is there a way to run "guix pull" with more verbosity and logging?
<vagrantc>ACTION notes --verbosity
<civodul>"guix pull --verbose" maybe
<civodul>--verbosity is usually unhelpful, you don't want that one :-)
<mbakke>az`: That error message typically means GRUB does not have enough space to embed itself.
<mbakke>If it's a GPT disk, try adding a tiny (~1MiB) partition and set the "bios_grub" flag.
<grafoo>hi! is there a way to symlink /var/guix and /gnu?
<grafoo>setting NIX_IGNORE_SYMLINK_STORE=1 does not seem to work.
<pkill9>grafoo: you need to do a bind-mount if you want to be able to use build substitutes
<pkill9>on a foreign operating system, just put in an fstab entry
<pkill9>so far I don't know how to bind-mount the gnu store on GuixSD because it needs to be done before the guix daemon starts
<grafoo>pkill9: great idea. thx!
<roptat>I'm trying to test something in a VM with guix system vm, but the command is stuck at [ 1.728111] IPv6: ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready
<civodul>roptat: yeah see
<zeronineseven>Hi! I'm trying to install guix on archlinux but it's failing due to error during "tests/"
<zeronineseven>Here is test-suite.log content:
<zeronineseven>Could somebody please explain to a poor guix newbie what's going wrong? ;)
<zeronineseven>I assume that guix test suite is trying to mount something without root permission. But why is it in the testsuite? Isn't this operation supposed to always fail?
<davidl>zeronineseven: I am not sure why. Maybe these INSTALL-notes can give you a clue:
<davidl>more like I have no idea why, really.
<catonano>I installed python in a profile and it is installed as "python3" so some scripts fail, because they use "python"
<vagrantc>civodul: guix pull --verbose didn't add anything more to the output ... still getting a segfault
<davidl>catonano: if you install python@2.7 as a user, you still get it as just "python".
<vagrantc>civodul: if it's at all helpful, here's the output:
<davidl>catonano: it should perhaps be like that since python3 isn't backwards compatible
<mbakke>catonano: You can use 'python-wrapper' if you want the "python" executable to be Python3.
<mbakke>zeronineseven: Can you try to run ` sudo sysctl kernel.unprivileged_userns_clone=1` and run the test again?
<mbakke>If that works, we should probably skip the test if user namespaces are disabled.
<catonano>mbakke: davidl thanks
<jahb>az`: I recall having the same problem and it was because I didn't choose to create a "GPT partition" in cfdisk
<zeronineseven>mbakke: thx for response! unprivileged_userns_clone was set to 1 when this error occurred. Actually there are unprivilleged lxc containers running on this very host now
<zeronineseven>davidl: thx! will have a look =)
<mbakke>zeronineseven: Oh, interesting. Since you're running Arch, I guess you have kernel 4.16?
<zeronineseven>mbakke: yep
<zeronineseven>Meanwhile I've tried to build guix inside arch lxc. Now more tests have failed:
<pkill9>i think the cause of the large number of 'updating list of substitutes' messages is when the terminal window is smaller than the message, or something
<pkill9>actually no it isn't, nevermind
<civodul>zeronineseven: could you use blocks Tor users
<zeronineseven>civodul: test-suite.log from arch host: ; test-suite.log from arch in lxc:
<civodul>zeronineseven: thanks! looking at the 1st one, it looks harmless
<civodul>we'd need to see the details of the kernel and config on that host to find out what's happening
<zeronineseven>civodul: my current kernel config:
<zeronineseven> doesn't support large files =/
<civodul>zeronineseven: sorry i won't investigate further right now, but perhaps you could also check with the maintainer(s) of this package in Arch?
<civodul>this test has to do with "user namespaces" and bind-mounts of files (not directories)
<civodul>if that helps...
<zeronineseven>civodul: ok, thanks!
***dijong_ is now known as dijong
<rekado_>ACTION fixed the circuit, gets back to working on guix
<civodul>rekado_: heh, congrats! :-)
<grafoo>has anyone built emacs with xwidget support?
<grafoo>installing the webkitgtk package does not seem to cut the deal.
<kvetcher>Where can I find a full guide to config.scm options?
<kvetcher>Do I just include packages from the repos in it and enable services as described here?
<lfam>kvetcher: That part of the manual is main guide to writing a config.scm
<lfam>There are also some examples:
<mbakke>Hmm both mesa and wayland provide libwayland-egl.
<kvetcher>lfam: Thanks, I'll take a look at the examples.
<lfam>kvetcher: Please feel free to ask more questions as necessary. Either here or on <>
<mbakke>Wowza. The man-db fix apparently does not work outside `guix environment`. Oops.
<axd-v>Hey, so I'm trying to install guixsd onto my system and I want to set up encryption. The installation manual doesn't really go into details on how to set it up. I guess it assumes that the user will know already. Well I have done it in the past and there is always arch wiki, but I just want to hear specific advice for people that have already set it up on guixsd.