<truby>Ollreee: I found I needed to have fontconfig installed from guix too, then ran fc-cache to regen the font cache and fonts from guix were picked up fine. But fc-cache from my distro doesn't pick up guix installed fonts <truby>this might be something worth adding to the application section of the guide that covers this stuff on foreign distros? <apteryx>is anyone tackling gnucash 3.5 build failures on core-updates? <janneke>sneek: later tell civodul: i would very much work to get gzip in Gash, R6 sounds like it should be doable; certainly for now as we are still using Guile <efraim>relating to the btrfs compression discussion on the mailing list, 'sudo btrfs filesystem defrag -clzo /gnu/store' fails with permissions errors. forgot about the directory being mounted read only <coldpress>how much space should one allocate to a Guix partition? <sneek>Welcome back civodul, you have 1 message. <sneek>civodul, janneke says: i would very much work to get gzip in Gash, R6 sounds like it should be doable; certainly for now as we are still using Guile <roptat>coldpress, that's really low, not even a MB :) <desmes>hey guix! If I don't want to use a display manger, I guess I just have to remove 'gdm-service-type' from '%desktop-services', should't I? What would be the preffered way to do this in guile scheme? <Gamayun>desmes: Or you could use %base-services instead. Depends if you want other desktop services started. <desmes>Gamayun: Good idea, so I should just use base-services with the other services except the display manager <desmes>Gamayun: is there a way to inspect the desktop-services so I can know what to add? I guess now I will use the guile repl? <divansan`>Hi All.Trying to install eclipse from guix. I'm not quite sure which package it is. There are 45 results. Anyone know? <roptat>unless someone managed to package eclipse <roptat>it's a huge beast, most of the results you see are actually just libraries from the eclipse project <roptat>it's somewhere in my todo list, but with a rather low priority <desmes>does anyone know how can I inspect, in the guix repl, the %desktop-services variable? <efraim>are we missing the rpcgen binary on core-updates? <efraim>i think I found it in rpcsvc-proto ***ng0_ is now known as ng0
<Gamayun>desmes: Can't remember from the repl, but the (fairly short) definition is in gnu/services/desktop.scm <janneke>hi civodul, can you give me a pointer to those gzip sources? <civodul>specifically weinholt/compression/gzip.sls <civodul>commit 5e6fc0cc48d0893a74fcfd9cfd0f2b89f14922c5 has it <civodul>that's from an old copy i have on my laptop <civodul>yeah apparently it's been removed from 'master' *janneke takes patches for Gash <civodul>there are lots of goodies in Industria <coldpress[m]>Hi, I'm in the installation and have mounted /mnt. I have a BIOS grub. Should I mount /boot ? <divansan`>any good links/articles comparing guix, nix, flatpak and snappy anyone has? <roptat>coldpress[m], on the guix system we don't need a separate /boot <roptat>but if you have one, you should mount it <roptat>and declare it in your config.scm <coldpress[m]>Is it not possible to mount a BIOS grub to a partition instead of an entire disk? <coldpress[m]>I have other boot partitions on the disk, and I don't want to overwrite them <roptat>not sure what you mean, grub should be installed on the first sectors of the disk <roptat>there's also a /boot directory, but that's for system files related to booting (usually the kernel and initramfs, plus some grub configurations) <roptat>that /boot can be on a separate partition or not, and there's no need to have the same /boot for every distro (actually it would be bad I think) <roptat>I don't think you'll be able to boot if it's not managed by guix <coldpress[m]>anw I found this command: 'parted /dev/disk set partition bios_grub on', that should be enough, right? <roptat>the kernel and initramfs are in the store, and there's no way for other distros to find them <nwww>I install GuixSD on my chromebook(using emmc) <nwww>it display sdhc_acpi.ko not found <nwww>I use the 1.0.1 installer <roptat>oh, it probably failed to detect a module properly <roptat>I think you should reboot on the installer, but don't use the graphical installer: mount your newly installed system manually, there should be a /etc/config.scm in it <nwww>I define this module in config.scm <roptat>if you used the installer you should already have it in config.scm, as "sdhc_acpi" try to set it to "sdhc-acpi" instead (a "-" instead of a "_") <roptat>I can't promise it'll work, but the modules are named inconsistently between the kernel and the filesystem <roptat>that happens if you quit the guile console after a failed boot :) <nckx>nwww: Hold the power button, pull the plug, … <nckx>Anybody on Guix System feel like testing (kernel linux-libre-5.3)? 🙂 <nwww>I need to wait for it to drain the battery to shut it down. <nckx>What horrible hardware is this. <nckx>Assuming that's true: can't you remove the battery? <nwww>chromebook power botton on keyboard <roptat>press the power button for long enough? <nwww>I need to open it to remove the battery <nwww>I need to open its case to remove the battery <nwww>I think it's a problem because the kernel module is not properly added to the initrd. <nckx>Horrible hardware: confirmed. Still, I'd expect holding the power button for 4+ seconds to power off any laptop. <nwww>How to using custom kernel to boot it? <nckx>nwww: Yes, it's a known Linux problem: they call their modules different things (_ vs. -) in different places. <roptat>nwww, just declare a custom kernel in the (kernel) field of your operating-system declaration :) <nwww>I will try to change the config.scm and redeploy it tomorrow. <g_bor>How can I output something from a shepherd service to the console? <roptat>nwww, that's what nckx meant by "Anybody on Guix System feel like testing (kernel linux-libre-5.3)?" <roptat>g_bor, no idea... if you can find its PID, probably cat some file in /proc :p <roptat>nwww, for a custom kernel, you need a custom linux package and use it in the (kernel) field like. I think there's a blog post about just that <nwww>I find a blog on guix.gnu.org to create a custom kernel <g_bor>What I am trying to do is to create a service to install guix system. I intend to do it as a gexp directly, but would like to see some output. <roptat>I'd like to talk more, but I really have to work now <g_bor>nwww: are you just simply willing to upgrade to 5.3 linux-libre? <g_bor>or you need more customization? <nckx>(I didn't mean to imply that 5.3 would fix this…? Or am I misunderstanding?) <nckx>I just need sanity testers, this is my first major kernel bump 🙂 <g_bor>nckx: what is current policy regarding kernel version? <nwww>I want to emem the emmc driver in the kernel instead of the module loading <g_bor>Do we retain the latest 5.2, define 5.3 and make it the default? <nckx>Probably, up to a month ago, ‘Mark's good judgement’. <g_bor>I see several version, and could make an update fast, but I am not clear on this. <nckx>5.2 is still its own #linux-libre release, so I kept it. <nckx>g_bor: What do you mean, make an update fast? <g_bor>what is the current version? <g_bor>did you just push an update I missed? <nckx>I committed 5.3 a day or two ago. Flipping the default variables is what I want to postpone until >1 people have tested it 🙂 <nckx>g_bor: Thank you! (x86_64, I guess? Good enough, I guess.) *g_bor building a disk-image right now <nckx>I just want to rule out any stupid mistakes on my end; next time I probably will just flip the default right away. <g_bor>nckx:yes, but the configs here are quite limited, about bare-bones, but should be enough <nckx>My test was spinning up a few minimal VMs and running uname. I really can't reconfigure any bare metal ATM. Yes, even with e-zy roll-backs. <civodul>nckx: i think Mark would keep the LTS branch(?) plus the latest version <civodul>in that spirit, i think we'd keep 5.3 and drop 5.2 <civodul>but i'm not super familiar with Linux releases i must say <nckx>civodul: *Current* linux-libre releases are: 5.3.1-gnu, 5.2.17-gnu, 4.19.75-gnu, 4.14.146-gnu, 4.9.194-gnu, and 4.4.194-gnu. <g_bor>civodul: any idea how can I output something to consol from a shepherd service start? <civodul>nckx: oh, i hadn't realized we had so many of them :-) <nckx>civodul: To be clear: ‘Our’ linux-libre, but also what #linux-libre announced. <nckx>I based the one off the other 🙂 <civodul>well that's prolly the right thing then :-) <nckx>If they drop 5.2 from that so will I. <nckx>I don't know what's ‘correct’ either, so I trust others to know that for me ;-) <nckx>R versioning: (1.10.1 >= 1.10) → true; (1.10.1 >= 1.10.0) → true; (1.10.1 >= 1.9) → false <g_bor>ok, I see it now. I can do something like the log to info from system install... <g_bor>nckx: so far it looks good to me. Boots fine, gpg, emacs, git and elogind-service works as expected. <nckx>Perfect! Thanks. I'll stop worrying and just do it. <desmes>Hey guix, does anybody else have screen tearing problems with Intel integrated cards? Usually using a compositor like compton with the 'Tearfree true' Xorg option solves this for me in other distros, but not on Guix. <nckx>desmes: I have an Ivybridge GPU (i915 driver) and videos tear. I… just didn't care enough. But I can confirm. <mbakke>desmes: try adding 'Option "AccelMethod" "sna"' to the Xorg configuration <desmes>mbakke: What the hell? That worked! I don't really understand why but ok. Thanks a lot <nckx>mbakke: Thanks too, I'll try that. <mbakke>why does linux-libre depend on 'openssl'? <coldpress>mbakke: are you hinting that ldconfig is not libre? <mbakke>coldpress: not at all (it comes from the GNU C library after all); just that Guix does not need it <coldpress>yeah i gathered, installed 'gcc-toochain' and now have 'ld' <nckx>coldpress: ldconfig has no use on Guix System. <coldpress>well, Haskell's 'stack' wants to use 'ldconfig' <nckx>coldpress: It should be patched to invoke ‘true’ instead. <nckx>So ‘stack’ administers your system libraries as root to install some Haskell packages? Big hm. <nckx>Or does it parse the output of ‘ldconfig -p’? (Other, different hm.) <nckx>coldpress: You can try creating a simple executable file containing ‘true’ and putting it somewhere in (root's?) $PATH. If stack is well-written, it will find it. <civodul>then i guess roptat and others can provide guidance once you got started! <PotentialUser-92>Is there a way to get rid of locale warnings? I've set up needded environment variables and installed glibc-locales, but still get those <bavier>katco: I think additional outputs would make sense <desmes>Has anyone here managed to use X without a display manager (just using 'startx')? <katco>bavier: thx for the opinion. the current package decends into the `cpp` sub-project specifically, so it is going to be interesting to figure out how to write this. off the top of your head, can you think of any examples to follow? <bavier>katco: in practice, if the sub-projects are all rather independent, and don't need shared build state, then creating separate packages may also be easiest, since you might be able to keep different sets of inputs. <katco>bavier: i think that is an accurate characterization. there is some interdependance of _outputs_, but not shared build state. <roptat>It's easier with separate packages, see maven for instance <bavier>katco: for a monolithic package, you could redefine the build phase so that it loops over the set of subdirs and calls the original build function, while maybe also redefining the installation prefix appropriately for each. <katco>from a user's perspective, i like the idea of all of this output being grouped under `apache-arrow`, but from a build/logical perspective, i think it might make sense to have separate packages. i'm really on the fence. <mbakke>I suppose you could do both, by creating separate derivations for each sub-project, and then combine them into one package :-) <coldpress>we need a symlink from xorg-server/bin/X to xinit/bin/X, but the filesystem is read-only <desmes>coldpress: But if that's all that's lacking, it should be easy to work around it, shouldn't it? <sneek>Welcome back vagrantc, you have 1 message. <sneek>vagrantc, nckx says: Yes. Not just for ARM. 5.3 is untested on bare metal. <vagrantc>nckx: not untested anymore ... i've tested it on both x86_64 and armhf :) <desmes>mbakke: Thanks for the resource, gonna check it out <coldpress>so what's up with the default guix login display? <desmes`>How come that in this distro you can't open another tty without the tty1 Xorg crashing? <desmes>I keep crashing the tty1's Xorg when opening a new tty :/ <coldpress>anw i managed to log in from the gdm screen, but not from xinit <coldpress>well, given that the graphical login is still there, I'd say we haven't removed all the desktop-services <quiliro>What happens is not that gdm crashes. It just moves to tty8 and opens another login in tty7. <sneek>sirmacik1, you have 1 message. <sneek>sirmacik1, g_bor[m] says: I've checked my config on the sway machine. The only thing I have in the config.scm is elogind-service and %base-services that relates to sway. Sway is installed in my user profile, and I start it from the command line. <sirmacik1>sneek tell later g_bor thanks! will you be able to show me your config.scm anyway? <sneek>later, sirmacik1 says: g_bor thanks! will you be able to show me your config.scm anyway? <coldpress>and how is one supposed to add *.pc files to PKG_CONFIG_PATH, if all the *.pc files are in separate directories? do we set the PKG_CONFIG_LIBDIR? <sirmacik1>sneek later ask g_bor thanks! will you be able to show me your config.scm anyway? <coldpress>i'm currently specifying all the packages i need in PKG_CONFIG_PATH, which doesn't seem like the correct solution <pkill9>PotentialUser-47: you could symlink ~/.config/guix to the other user <pkill9>e.g. symlink /root/.config/guix to /home/<your-user>/.config/guix <pkill9>then root would use your user's guix <truby>out of curiosity, how often do people usually run guix pull, and how long do you expect it to take? It feels quite slow at least on AArch64 <pkill9>truby: i probably run it about once a month, if substitutes are available then it takes about 3 minutes <PotentialUser-47>pkill9: is there a man page for how to do that particular symlink? sorry if this sounds very newbie <coldpress>anyone else's icecat only shows white numbers on a white background? <truby>I guess there's fewer substitutes for AArch64 then? it takes me roughly an hour or a bit more on the aarch64 server I use <coldpress>also, is there a way to change localtime, other than by reconfiguration? <pkill9>PotentialUser-47: you just need to move /root/.config/guix to /root/.config/guix.bak, then run `ln -s /home/<your-user>/.config/guix /root/.config/guix`, then root will use your user's Guix. You'll want to avoid running `guix pull` as root though, as it will create the symlink "current" in that directory as root and your user won't be able to change it <PotentialUser-47>pkill9: is there anything I need to do to revert running guix pull as root? <pkill9>oh, if your user was able to run `guix pull` then it probably doesn't matter <pkill9>maybe just check that your user owns the symlink `~/.config/guix/current` <pkill9>also possibly the symlinks in /var/guix/profiles/per-user/<your-user> <pkill9>PotentialUser-47: i just run `ls -l` and it tells you which user and group owns the file <PotentialUser-31>does anyone know of any blog entry that explains how to keep guix up to date? <PotentialUser-31>I'm having trouble making analogies between pacman and apt package managers with guix <truby>PotentialUser-31: my understanding is that (broadly) guix pull is anlagous to apt-get update and guix package -u is analagous to apt-get upgrade. But, they are individual to each user so updating your users stuff doesn't affect other users on the same system <truby>PotentialUser-31: EDITOR=zile guix edit <truby>in your bashrc you can export EDITOR=zile, or you can just put EDITOR=zile in front of guix edit each time you run it. Probably you want to export it though :) <pkill9>PotentialUser-31: in my experience no it doesn't break if you don't update it for a while <PotentialUser-31>truby: Hey thank you! that was simple. I should of thought of that as I've done that before <pkill9>PotentialUser-31: when you say break, do you mean that it is unable to update? Or that it is unable to upgrade packages? <PotentialUser-31>I am thinking like what happens when you don't upgrade arch linux for two months <truby>if I'm not wrong, the same breakage won't happen because each package gets installed into a new profile before that gets switched to. So, you can't get the problem you get on Arch where a package getting unpacked breaks the currently running pacman binary <truby>but, I am by no means a guix expert so I could have understood that completely wrong :) <PotentialUser-31>I make a tweak to my command line file manager in the source code to change the leader key to semicolon. What would be the ideal way to use guix to package this modification? <kmicu>Hi PotentialUser-31, pacman/apt are like svn, guix/nix are like git. git pull cannot break your repo and in the same way guix pull cannot break what already works. <nckx>vagrantc: Awesome! I'm about to push it as default. <nckx>kmicu: wait svn breaks repos what <rekado_>nckx, civodul, the DNS names should be tracked. There should be no need for an SSH tunnel. Is this not working? <nckx>rekado_: That's what I thought. Thanks for clarifying. The problem yesterday was IPv6 being blocked by my ISP, though, so it was as simple as removing AAAA records. <civodul>rekado_: apparently not, per "ssh -vv -4 -p 5552 hydra@dmitri.tobias.gr" <nckx>civodul: Oh, both boxes are idle again? :-/ <mbakke>civodul: I've tested the boot entropy patch, and while the patch seemingly does the right thing, boot still hangs! <civodul>mbakke: are you sure the hang relates to entropy at all? <civodul>i wonder why sshd needs entropy as soon as it starts <nckx>I wonder why it dies and doesn't just wait. <nckx>Or is that a different bug? <mbakke>nckx: on 'core-updates' it waits....forever. <mbakke>(or until you mash the keyboard in frustration) <nckx>mbakke: OK, that's happened to me once (on c-u), but then afterwards it died, I guess I was just ‘lucky’. <civodul>mbakke: and does pressing keys :-) actually help? <mbakke>civodul: pressing keys, moving the mouse, ping flooding...all allows the system to continue booting. <nckx>‘broken user space that was asking for secure random numbers when it really didn't need to’ — guess I should've just read your link 🙂 <nckx>Does this mean that only people using ext4 were guaranteed boot entropy? <mbakke>booted instantly with the kernel patch (+ civoduls patch) *nckx .oO The work-around in that patch sounds a bit like… haveged. <civodul>mbakke: you'd have to try several times to be sure though, right? <mbakke>it stalls for about 0.5 second where it used to hang <mbakke>I've hit the hanging boot every time today until now :-) *mbakke prepares reply + patch <civodul>mbakke: glad you were able to go this far! <amz3>hello, did anyone consider packaging kde falkon ?