IRC channel logs


back to list of logs

<Veera>joshuaBPMan: run "ip a"
<Veera>joshuaBPMan: see if your device has been assigned ip addresses
<Veera>In Guix how to manually stop services started by shepherd init
<NieDzejkob>`herd stop some-service`?
<Veera>dhclient may be started by shepherd init
<NieDzejkob>it's called the "networking" service
<Veera>may be if you want to run it manually stop it and then try
<Veera>joshuaBPMan: hey I am new; the proper way to do it, some Guix officials may say
<joshuaBPMan>Veera: I actually am trying to set up the manual installer. Though I could pop in a harddrive that has guix installed already.
<joshuaBPMan>"dhclient -v [interface name]" is what the manual installer says to do. I believe that is the command that networking does as well.
<joshuaBPMan>Verra: So I can boot my T400 into an installed Guix System (I installed Guix System on a Macbook, and then put that harddrive into the T400).
<joshuaBPMan>and I did try running sudo herd restart networking.
<joshuaBPMan>Still the same error.
<joshuaBPMan>ping -c 3 does shows I have no networking connection
<Veera>I could not get it. Are you running it inside a VM emulator
<Veera>If so which VM emulator are using?
<lfam>joshuaBPMan: QEMU doesn't support ping by default
<lfam>I would use `guix download` to test connectivity
<lfam>Just pass it an URL
<lfam>Sorry, I thought you were in a VM but now I'm not sure if you are
<Veera>joshuaBPMan: you can try traceroute if available
<joshuaBPMan>lfam: I'm not in a VM.
<joshuaBPMan>I turned on the T400 and booted from the guix System installation usb stick.
<Veera>then how you are running it
<Veera>directlt in mac hardware
<joshuaBPMan>nope. The T400 is running from the Guix System installation image.
<joshuaBPMan>I am at the manual install and cannot get an internet connection.
<joshuaBPMan>ifconfig shows my ethernet device. enp0s25.
<Veera>Okay I get it. First installed in Mac then put that drive in T400 system and then boot
<joshuaBPMan>dhclient -v enp0s25 gives me an error.
<joshuaBPMan>Veera: Yes.
<joshuaBPMan>Right now though, I booted from the usb stick.
<NieDzejkob>is that running as root?
<joshuaBPMan>Though I could reboot into the hard drive.
<joshuaBPMan>NieDzejkob: yes sir!
<joshuaBPMan>The error I am getting says:
<Veera>Do you have network connection to that network card/device
<joshuaBPMan>DHCDISCOVER on usb0 to port 76 interval 7
<joshuaBPMan>Unable to set up timer: argument out of range.
<Veera>Have you tested the network works in some other OS
<Veera>Else then what is your network/device card
<NieDzejkob>huh, building guix from a clean checkout of the repo is failing when I run "make":
<joshuaBPMan>Veera: The T400 is plugged into an ethernet port. I could plug in another computer into the same port (with the same ethernet cable) and make sure the port still works.
<NieDzejkob>hmm, is your date/time set right?
<Veera>Is the other system Linux. If so find out which kernel driver/module it is using
<Veera>see it is loaded in guix or load load it using modprobe
<Veera>may be that linux-libre does not supports your card.
<joshuaBPMan>NieDzejkob: interesting... "date" returns a date from this january!
<Veera>please find out
<joshuaBPMan>past january.
<TempeVolcano>Thank you brendyyn, my confusion is why you need to mount the same disk that the root of the Linux file system is on.
<joshuaBPMan>Veera: it's possible that a kernel module was not loaded properly.
<Veera>is T400 x86/x86_64 one; intel/amd
<joshuaBPMan>e1000e I think is the module name for the ethernet. Let me see if it was loaded.
<joshuaBPMan>T400 is Intel.
<NieDzejkob>I'd try setting the date to something closer to reality
<NieDzejkob>the error talks about a timer, after all
<joshuaBPMan>oh dear Lord! date actually says it is 2120! hahaha.
<Veera>NieDzejkob: i don't think it is about system rtc time
<joshuaBPMan>That could be causing some real issues!
<Veera>set it in bios
<NieDzejkob>Veera: I don't know much about DHCP internals but if there's some absolute timestamp flying over the wire then it might be trying to wait for a negative number of seconds
<NieDzejkob>it's also possible you're hitting a Y2038 bug
<joshuaBPMan>Let me change the time real quick. They probably happened because I unplugged my BIOS battery during the laptop tear down.
<joshuaBPMan>also what's CET? Central Eastern time?
<Veera>please see dhclient is reporting that fro usb0
<Veera>does your network device uses usb0
<Veera>may be usb network device is causing problem
<joshuaBPMan>I now have an internet connection! THanks guys!
<joshuaBPMan>Verra: I was using usb0 to connect to wifi. I was tethering my phone's wifi.
<joshuaBPMan>That's why I was trying to use usb0.
<joshuaBPMan>Let me reboot into Guix System (the hard drive) and update guix!
<Veera>so it got solved by disconnecting wifi
<nikita`>wrt 37207, do you really think playing catch-up with repology is a good use of time? he's been so uncooperative with some of us that at least one person started writing a replacement which does not paint pkgsrc as months outdated
<pkill9>i made a hack around the problem with pulseaudio shutting down
<pkill9>i made a script that watches to see if pulseaudio is running, then if it doesn't it starts pulseaudio and connects to my bluetooth headphones
<joshuaBPMan>Veera: so I solved it by date -set command. The date on my machine was 100 years in the future.
<joshuaBPMan>Also, weirdly enough, my T400 appears to be booting via UEFI....I'm pretty sure that the T400 is not supposed to support that.
<Veera>T400 is a recent system?
<Veera>Recent system boot via uefi; there may be an option of diabling it in bios
<Veera>but uefi supports gpt partition table
<Veera>gpt has advantages
<Veera>like big partition sizes
<joshuaBPMan>Veera: I would prefer to use uefi. I am just surprized that 2008 hardware supports it.
<Veera>then it definitely will have option to disable it
<Veera>something called compatibility mode
<Veera>in bios
<Veera>or legacy mode
<anadon>Wait, what's the deal with repology?
<atw>I'm doing some python packaging and the pypi importer seems to use propagated-inputs. Could I get a little explanation as to why? just for my curiousity :D
<brendyyn>atw: Python modules look for their dependencies through PYTHONPATH, and so they will often fail to find them if they are not propagated. I think it is usually better to put them in inputs, and use wrap-script to make them available though
<brendyyn>atw: I'm also not 100% when something should or shouldn't be propagated. I think generally you want to avoid it unless it is really required
<ngz>atw: There is an explanation in "python-build-system.scm" headers
<atw>thanks brendyyn, ngz :)
<TempeVolcano>Hey mbakke when I try to follow the first step of the "continuing installation" instructions and enter the command "herd start cow-store /mnt" I get this error "filesystem on '/mnt/tmp/guix-inst' not supported as upperdir"
<Veera>TempeVolcano: can you tell me your starting setup
<TempeVolcano>I made 1 MB partition for GRUB, then a 500 GB partition that's encrypted for the root system.
<brendyyn>Is 1MB enough for grub?
<Veera>TempeVolcano: I think no
<Veera>TempeVolcano: In my system it is using 9MB
<TempeVolcano>Uh oh will I have to start the manual install over again?
<Veera>TempeVolcano: to be safe/reduce afterwards problems make it 128MB
<Veera>TempeVolcano: see I am new as well
<Veera>TempeVolcano: usually /boot holds grub files plus linux kernel image + initrd
<Veera>TempeVolcano: But here i see it holds only grub files
<Veera>TempeVolcano: and linux images + initrd are stored in /
<Veera>TempeVolcano: in other distros linux images + initrd may hold much space
<Veera>TempeVolcano: Guix uses the store
<Veera>also see it that the filesystem does not has anything in it.
<Veera>usually mkfs cleans up while creating a fresh install
<Veera>Also make a partition for swap
<TempeVolcano>Looks like I can't expand my grub partition.
<Veera>may be a swap file may be sufficient
<TempeVolcano>I think I'll probably need to start over.
<Veera>If it the disk does not holds anything valuable start from again
<Veera>it can also be done otherwise
<Veera>what are using to make partitions
<Veera>parted / fdisk
<Veera>Do you know how to use it
<Veera>type help shows it
<TempeVolcano>I only kinda know how to use it.
<Veera>in parted you can start fresh or delete the partitions 1 2
<Veera>see just delete partition 1 will not do
<Veera>partitions are contiguous
<TempeVolcano>I'm sorry the 1MB was for a BIOS based boot partition.
<Veera>for bios start the partion sectors from 1MB
<Veera>leave 0-1MB free
<Veera>usually 0-2048 sectors
<Veera>parted may be difficult for starters
<Veera>use fdisk it take care it
<TempeVolcano>okay will do
<mroh>NieDzejkob: try `make SUBDIRS=` as a workaround.
<Veera>what is fsdg compatible license
<brendyyn>look it up:
<Veera>is it free software debian guidelines
<Veera>oh. ok
<brendyyn>There are fine details between what GNU and Debian consider "free".
<brendyyn>It's why the GNU doesn't list Debian as a free distro
<Veera>oh remember it slightly
<Veera>yes the binary blobs
<lfam>Veera: This is the FSDG:
<Veera>lfam: i saw it. thanks
<lfam>Basically any free licenses from here:
<Blackbeard>Veera: is not just the blobs, it is also about making simple to install proprietary software and helping users to do so
<Veera>Oh. I will read it after sometime.
<Blackbeard>Let's say if Guix added instructions to replace linux-libre by the official one, even if guix didn't ship it
<Veera>Well does guix accepts proprietary software
<Blackbeard>It would be against the guidelines
<Blackbeard>Or if Guix didn't ship any proprietary software but had a repository that you can enable to install software
<Blackbeard>I think NixOS does this
<Blackbeard>They have an opt-in proprietary software repository
<Veera>non-copyleft should cover mit and bsd ?
<Blackbeard>So even if it is disabled and they ship only free software, they made it really easy to install proprietary software
<Blackbeard>Veera: I think by the GNU guidelines MIT is a group of licenses, so each one is named differently
<brendyyn>Veera: Please understand we are not hostile towards you using proprietary software on your own system. Our goal is to provide a system that, by defaul[02:25:20] <Veera> guix does not ships proprietary software but also does not stops user from that
<brendyyn>I just wanted to clarify because if you read certain garbage on the internet, many ignorant people think we are crazy and want to take away their freedom to give up their freedom or something silly like that
<Veera>anyone can install anything if they can
<Veera>others can force it on others;
<Veera>others cannot force it on others;
<brendyyn>the MIT license is called expat in our licenses module. This is because Richard stated that MIT is ambiguous due to their being many licenses coming from MIT University. So it's called MIT Expat. Likewise, there are multiple "BSD" licenses. In guix you would refer to them with bsd-2, bsd-3, bsd-4
<brendyyn>non-copyleft is for when you have some obscure permissive license that is not in the list.
<brendyyn>It's a procedure that allows you to specify a URL to the license text.
<Veera>oh I missed it in licenses fiel
<drakonis>that sounds kinky
<drakonis>freedom to be a slave
<Veera>brendyyn: ok
<Veera>can xplanet be included
<Veera>any previous discussions on it
<Veera>web search shows irc log of roptat
<Veera>just one word by ng0
<Veera>its gpl2
<brendyyn>It's in Parabola, so I assume so
<Veera>xplanet has site in sourecforge
<Veera>last release 2017
<roptat>what did I say?
<Veera>nothing ng0 said release around leap years
<Veera>sorry just the previous line perhaps not connected
<roptat>you pinged me so I'm curious :) but I don't really understand the context
<Veera>parabola is different distro based on archlinux
<Veera>so it may be packaged here
<roptat>if xplanet is in parabola, it's probably free software, so it can be packaged
<roptat>unless it's not free software, in which case you should report it to parabola
<Veera>I already checked the original site
<roptat>looks like it's gpl licensed, so it should be fine
<roptat>not sure which gpl exactly, but it's fine
<Veera>original site and pkg sources say GPL2
<brendyyn>My geiser-guile-load-path is set to my source directory but in emacs guix-edit still takes me to the immutable store, not my git repo. any idea why? I even launched emacs from a terminal with ./pre-inst-env emacs
<mroh>`C-h v guix-load-path` is nil?
<brendyyn>yes its nil. i guess i need that one too
<joshuaBPMan>Hey guix, I'm trying to guix system init, and I am getting a TON of output. Is there a simple way to try to understand all of this?
<joshuaBPMan>I think I may have found a solution
<Blackbeard>joshuaBPMan: did it work?
<joshuaBPMan>Blackbeard: not quite yet.
<joshuaBPMan>But I had the same issue a moment ago. I'm essentially using %desktop-services, but I am not using gdm
<pkill9>joshuaBPMan: it's probably outputting everything being built
<joshuaBPMan>It's not building anything.
<joshuaBPMan>At least it doesn't appear to be building anything. I fixed this a moment ago. I think it's the same issue.
<joshuaBPMan>It's outputting a backtrace. I'm currently running guix system init /mnt &> error.txt so that I can read the error message.
<roptat>you might need 2>&1 too (after &>error.txt iirc)
<roptat>so stderr is also redirected
<joshuaBPMan>I am trying to set up an encrypted / (including /boot) apparently libreboot can do this.
<joshuaBPMan>The bootloader section is weird though
***heath is now known as h
***h is now known as heath
<joshuaBPMan>(bootloader (bootloader-configuration (bootloader (bootloader (inherit grub-bootloader) (installer ~#(const #t))))))
<joshuaBPMan>I don't even know if that works.
<joshuaBPMan>so I'm getting a backtrace...
<joshuaBPMan>the last backtrace says in gnu/services.scm
<joshuaBPMan>826:27 0 #<<service> type: #<service-type screen locker ?> ?) ?)
<joshuaBPMan>hmmmm, maybe I need to remove that from desktop services
<roptat>that snippet has weird indentation...
<roptat>what did you change from it?
<roptat>it seems that the bootloader section is not at fault here, but the services have an issue
<joshuaBPMan>I think I'm just going to try to get a bare bones working for now. Well basically I modified %desktop-services... I removed gdm-service-type avahi-service-type and a few other services.
<roptat>ok, with the manual installation, I tink it's advised (at least, that's what I would do) to install a bare-bones system, guix pull and install the desktop system only after that
<roptat>at least you have a basic running system :)
<joshuaBPMan>yup. I'll give that a try. thanks roptat
<roptat>and you can roll back to it in case of troubles
<joshuaBPMan>man... I'm still getting a backtrace.
<joshuaBPMan>It looks like it has an error with service-type login
<roptat>that's weird, did you change anything?
<roptat>what is you services field now?
<Veera_>joshuaBPman: is your base vm.img
***Veera_ is now known as Veera
<joshuaBPMan>Veera_: I'm not certain that I can test this configuration on a vm right now.
<joshuaBPMan>I'm manually installing it.
<joshuaBPMan>my services field looks like
***Veera_ is now known as Veera__
<joshuaBPMan>(services (append (list (extra-special-file "/usr/bin/env" (file-append coreutils "/bin/env")) (service-dhcp-client-service type) %base-services)))
<raingloom>anyone else can't use (gnu[04:24:11] <raingloom> joshuaBPMan, i want to test some earlyoom configuration modifications
<raingloom>it's easier in a geiser repl, because i have proper autocomplete and stuff.
<joshuaBPMan>raingloom:that certainly makes sense.
<joshuaBPMan>what earlyrom stuff are you tweaking?
<raingloom>minimum RAM. 10% kills things way too early.
<joshuaBPMan>fair enough
<raingloom>huh. (load "~/Projects/Guix/guix-source/gnu/services/linux.scm") errors with No such file or directory: "/home/raingloom/~/Projects/Guix/guix-source/gnu/services/linux.scm"
<raingloom>why is ~ in the middle...
<Veera>In Add pkg patch what should be Subject line; gnu: Add this; or gnu: image-viewers: Add this
<raingloom>ah. seems to be a guile bug. guix repl (which uses guile 3) can import the module.
<joshuaBPMan>hmmm. Well libreboot doesn't seem to be able to unencrypt the guix system....
<joshuaBPMan>that's annoying.
<Blackbeard>joshuaBPMan: can you share your config.scm
<Blackbeard>And what error are you getting
<joshuaBPMan>I'm not certain how I would share it at the moment...It's on a machine that currently won't boot. I can boot into the machine via the usb stick. But I'm not certain how to copy the config on disk and share it with you via the terminal
<joshuaBPMan>Blackbeard: I'm not getting any error. Libreboot loads. I select "o" load operating system.
<joshuaBPMan>and I have a blinking cursor.
<joshuaBPMan>it is just sitting there.
<joshuaBPMan>it is not asking me for the luks passphrase
<Blackbeard>joshuaBPMan: that happens to me with libreboot sometimes, even though mymlaptopt has parábola and it is not encrypted
<joshuaBPMan>hmmmm. That's odd.
<Veera>joshuaBPMan: so it is a preinstalled one now booting from T400
<Blackbeard>joshuaBPMan: to solve it I reboot it a few times
<joshuaBPMan>Veera: It is a pre-install booting from a T400.
<joshuaBPMan>Blackbeard: I've rebooted 3+ times.
<joshuaBPMan>It may just be an issue of how I configured it.
<Veera>so could you get console or it has not yet passed through
<Veera>mean in pre-init phase asking for LUKS for root
<joshuaBPMan>Veera: I can only get to a console, if I boot to usb
<joshuaBPMan>it does not ask for LUKS password
<Veera>you have installed to usb also
<joshuaBPMan>After I boot the T400 from the harddrive I only see a blink cursor.
<joshuaBPMan>Veera: yes. I just installed "bare-banes.scm"
<Veera>In hard disk which version of guix is there
<joshuaBPMan>It's perhaps a week old. No more than 1 month old.
<Veera>did it booted fine when in Mac
<joshuaBPMan>what I'm trying is only possible with libreboot.
<joshuaBPMan>I am encrying the whole harddrive.
<Veera>Does Grub boot screen shows up
<joshuaBPMan>Grub does show up, but libreboot includes grub as a payload.
<joshuaBPMan>I did not install grub in guix (at least I don't think).
<Veera>In Mac there is no libreboot
<Veera>I want to know how you initially set/installed the guix os
<Veera>assume T400 has the libreboot
<Veera>And with Mac you used grub
<Veera>just type 'c' in grub boot menu
<Veera>it goes to command line
<Veera>see that in new T400 system partitions of hard disk are shown
<Veera>perhaps you can chainload the grub installed in hard disk
<Veera>That blank/blink screen sometimes happens because of mismatch
<nckx>I haven't kept up with development the past 2 weeks :-/ Do you mean there's a bug?
<nckx>Yeah, I think the transition on c-u to Guile 3 is complete.
<nckx>G'night o/
<Blackbeard>nckx good night :)
<Kimapr>i want to have ssh daemon on guix system, but reconfigure keeps complaining that variable `openssh-service-type` is not bound
<Kimapr>i'm very confused
<guix-vits>Kimapr: (use-service-modules desktop ssh) ?
<Kimapr>oof. documentation says nothing about that ssh is a submodule of networking
<guix-vits>Kimapr: `cd guix-git/guix/gnu/services/; grep 'openssh.service.type' *` :)
<guix-vits>grep prints "ssh.scm:(define openssh-service-type"
<Kimapr>that doesn't count as documentation
<Kimapr>^ it's not clear from that page that `ssh` is a module
<Kimapr>the page misleads me to think that i only have to (use-modules (gnu services networking)) to use `openssh-service-type`
<guix-vits>Kimapr: try contact NieDzejkob; or: rekado_ suggests filling a bug reports in such a cases.
<guix-vits>documentation is great, but only if looking in code at the same time. At least Scheme is a lot easier to read than XYZ.
<guix-vits>Kimapr: btw, (use ... ssh) solves this?
<Kimapr>but the reconfiguration is not done so i don't know what errors there can be
<Veera>the os configuration doc says it clearly
<Veera>with some basic examples
<Veera>ready to deliver docs are not there
<guix-vits>Kimapr: :) -- recently modified the console-font-service-type, so understand you. One misplaced paren will spit a bunch of strange errors... i'd solved this temporary by defining a @my-font for `(file-append font-terminus "/share/consolefonts/ter-v24b.psf.gz")` for `(cons tty my-font)`
<Veera>We have to find our way by learning, asking and trying
<guix-vits>Kimapr: i'd just read in Manual: "password-authentication? (default: #t)" -- isn't the password-based auth insecure (#t mean "true")?
<guix-vits>maybe you'll want to set it to #f to use key-based auth (idk).
<Blackbeard>Veera: it works
<Kimapr>guix-vits: probably not, password-auth is secure enough for my use-case
<str1ngs>guix-vits: it probably defaults to #t for ease of use. but it's not as secure as public key.
***apteryx is now known as Guest39296
***apteryx_ is now known as apteryx
<brendyyn>oh wow, debian has 107 patches for calibre....
<str1ngs>almost a complete book in patches :)
*str1ngs hides
<brendyyn>but the good news is it works with python3
<brendyyn>most the patches are just applying stuff from the python3 branch, not just debian messing with it
<guix-vits>str1ngs: idk about security, just seen the folks complaining somewhere (about pass-based). thanks.
<str1ngs>guix-vits: weak passwords can be brute forced.
<nikita`>Veera: I recently updated xplanet for pkgsrc and had a chat with the maintainer. It's too bad they don't do releases anymore and just recommend pulling from svn. So we have release + relevant patches afterwards
<Veera>I thought of pulling the src tarball and packaging.
<Veera>pkgsrc is which distro openbsd/freebsd
<nikita`>I'd just use the release and some selected patches.
<nikita`>No distro
<Veera>are you also pkging for guix
<nikita`>It's a project by netbsd but right cross operating systems
<nikita`>Yes, but not very active these days
<nikita`>*runs cross OS
<Veera>so i can pull from the source plus patches from netbsd
<Veera>what about getting from debian
<Veera>they also bundle it
<Veera>they also provide source + patches
<nikita`>You can do in guiz what I did in pkgsrc, there's no need to vendor the patches
<Veera>is it readily available with pkgsrc; i mean easily downloadable using std http/ftp tools
<nikita`>I mean I'd go back to rereading what I've written. Just use the release and pick the patches as additional downloads in the package definition, iirc it's possible from the original hosted. If not, sure debian would work
<Veera>ok where is the pkgsrc hosted
<nikita`> there's a link to the CVS directory iirc. No idea if you understand completely what I wrote, but I want to get up soon
<Veera>is there a way to pull from cvs just like git-fetch
<bricewge>Veera: Looks like there is a cvs-fetch
<Veera>guix info pages does not shows one
<bricewge>“rg 'def.*cvs-fetch'”
<bricewge>It's in (guix cvs-download)
<civodul>Hello Guix!
<civodul>hola rotty! :-)
<Veera>bricewge: I can't find it in my system
<Veera>bricewge: grep of scm files does not shows it
<rotty>Veera: there is also git-cvs, but it's intended to replicate (or rather compute) full history, so it might be overkill, depending on what you want to do
<Veera>Is there a easy way to fetch from pkgsrc cvs repos and pkg them for guix
<Veera>Eh, I was just checking pkgsrc cvs repos have only commands for buidling
<Veera>in xplanet case
<Veera>am I right?
<Veera>there were some Makefile's and patches
<Veera>else the web interface is not showing all
<rekado_>hmm, we missed a warning from because the only user account we have set up did not have a preferred medium configured.
*rekado_ sets it to “IRC”
<civodul>i want to see an alert! :-)
<rekado_>it’s disappointing that it hasn’t retried sending out the alert
<rekado_>the problem first appeared on the 24th.
<guix-vits>Veera: about searching in git: `cd guix-git; for x in `find ./ -type f -name '*.scm'`;do grep -l 'SOMETHING' ${x};done`
<Veera>I used a simple version
<Veera>find -type f -name "*.scm" -exec grep -H "*something*" "{}" \;
<guix-ci>civodul: This is a test message from Zabbix
<rotty>guix-vits: why not just use "git grep"?
<guix-vits>rotty: idk how to use it; thanks.
<guix-vits>Veera: cool.
<rotty>I think this would be equivalent to your loop: `git grep SOMETHING "*.scm"`
<brendyyn>How about `rg -t lisp SOMETHING`
<brendyyn>guix package -i ripgrep
<brendyyn>grep is outdated and there are many shinier tools one can use.
<Veera>yep could have used git ones
<Veera>but learned git afterwards
<brendyyn>it adds colours too
<guix-vits>thanks folks.
<guix-vits>is `(copy-recursively` better than `(invoke "mv" ...`?
<rekado_>guix-vits: yes.
<rekado_>don’t run external programs when you can do it with just Guile.
<guix-vits>rekado_: thanks.
<guix-ci>Problem started at 09:58:35 on 2020.03.27
<blackbeard`>civodul: ping
<blackbeard`>civodul: I've been reading a bit about init systems, shepherd and systemd
<blackbeard`>as you told me, the systemd api is quite big and it seems like there are many things I need to understand and implement
<guix-ci>Problem has been resolved at 10:05:35 on 2020.03.27
<civodul>what's wrong with guix-ci?
<civodul>guix-ci had a problem
<rekado_>hmm, that’s not very helpful.
<rekado_>I also don’t see anything corresponding on monitor.*
*rekado_ just ACK’d one reported problem
<guix-vits>should i rename "mapc" to "neverball-mapc" (''map compiler''; The Debian packages it as "mapc")?
<kawerte>Hey guys! Is there a way for me to add flatpak to guix?
<brendyyn>there is flatpak 1.4.3 in guix. i have never used it though.
<blackbeard`>kawerte you can do "guix install flatpak"
<blackbeard`>but I am not sure if something else is necessary, as I've not used it either
<blackbeard`>kawerte: is there a package you need from flatpak?
***rekado_ is now known as rekado
<guix-ci>Problem started at 11:09:35 on 2020.03.27
<civodul>good to know, guix-ci
<guix-ci>Problem has been resolved at 11:16:35 on 2020.03.27
<efraim>is it possible to build the .go files in the repo piecemeal? i'm playing with a machine with 1.5GB of ram
<civodul>efraim: it's possible but you have to do this manually
<civodul>but basically you could run "make" and C-c it when it becomes too much :-)
<civodul>or you could pass make a list of .go files
<blackbeard`>civodul: where should the cgroup support for shepherd be included?
<blackbeard`>I am not sure if service.scm would be the right place
<civodul>rekado: there's a "guix gc --list-dead" process running that holds the GC lock
<guix-ci>Problem started at 12:20:35 on 2020.03.27
<guix-ci>Problem has been resolved at 12:28:35 on 2020.03.27
<guixer>Hej all. I like to extend some user services for shepherd. Unfortunately I don't know how to restart shepherd for my user.. Invoking "herd load shepherd" gives me error: connect: /run/user/1000/shepherd/socket: No such file or directory
<guixer>Any ideas?
<guixer>Already killed the user process for shepherd..
***PlasmaSt` is now known as PlasmaStrike`
<Veera>Hi guix
<bricewge>guixer: Looks like shepherd isn't running, try `shepherd'.
<bricewge>And the command you tried to run is probably `herd load root service.scm'.
<rekado>civodul: yes, I started it earlier today
<rekado>(I’m collecting dead disk images)
<rekado>the zabbix alert should be better now
<rekado>(the order of arguments was wrong, so it only displayed the subject)
<guix-ci>Problem: Zabbix discoverer processes more than 75% busy Problem started at 13:31:35 on 2020.03.27
<guixer>bricewge: I figured out that I can run 'shepherd' as user.. then I can also debug bc I can see stdout of shepherd. I thought there is a smarter way to run shepherd than rebooting or typing 'shepherd &' in the shell...
<guixer>bricewge: If I run `herd load root service.scm', is that the "smart way" I was searching for?
<guix-vits> -- is this proprietary?
<guix-ci>Resolved: Zabbix discoverer processes more than 75% busy Problem has been resolved at 13:39:35 on 2020.03.27
<guix-vits>"The Octocat ball `data/ball/octocat/` uses the Octocat design with permission from Github ..."
<guix-vits>Ok, a Debian package contains ball/octocat, so...
<raghavgururajan>Hello Guix!
<bricewge>Hello raghavgururajan!
<raghavgururajan>I cloned guix git repo and did `./bootstrap` and `./configure --localstatedir=/var`. When I did `make`, I get this error.
<raghavgururajan>rm -f && /gnu/store/f2mrwn4r6zfnk3lxw782j137b97v6mk3-profile/bin/msgfmt -c --statistics --verbose -o en@boldquot.po
<raghavgururajan>en@boldquot.po:4584: format specifications in 'msgstr[0]' are not a subset of those in 'msgid_plural'
<guixer>guix-vits: I would say it is not free, since it is a registered trademark. But if you refer to a product of github on your website or in an app, it seems ok with the codex.
<guix-vits>guixer: probably, as neverball is in Debian's "main".
<bricewge>guixer: I dont know if it's "smart", it just allow one to load code into a running shepherd.
<guixer>bricewge: ok thanks.
<apteryx>civodul: bah, my team now requires we use Nothing seems to work with it in Guix :-/. Icecat freezes and the sounds cut, ungoogled-chromium doesn't seem to want to allow microphone.
<NieDzejkob>you mean It works quite ok on IceCat for me
<apteryx>NieDzejkob: yeah, this
<apteryx>oh really? With sound and all?
<brendyyn>apteryx: have you tested
<brendyyn>for me icecat gives no mic their either, but chromium does
<brendyyn>as for webcam guix seems to completely fail to recognise my webcam
<apteryx>brendyyn: yeah, it was a bit buggy as well in icecat, although workable for sound
<apteryx>but the team seems to prefer discord to it
<NieDzejkob>huh, I mostly use it for text chat and the laptop I have guix installed on only has one jack, so testing the microphone isn't exactly easy...
<apteryx>NieDzejkob: I had it working for a short time and then it becomes unresponsive and I stop hearing others
<apteryx>maybe cause of my old machine
<starcrATI[m]>Matrix is a good alternative to Discord ...
<starcrATI[m]>but maybe you could try to install the Discord desktop client via flatpak ?
<brendyyn>starcrATI[m]: does it do group voice calls and group video chat?
<NieDzejkob>Is there a way to specify that a package needs the binutils for x86 even when building on and for arm?
<civodul>apteryx: i used Jitsi with ungoogled-chromium and it works fine
<starcrATI[m]>brendyyn: Didn't try it out, I use Jitsi and BigBlueButton personally
<brendyyn>i didnt like riot it was ugly
<brendyyn>and my contacts couldnt find me, or theyd accidently find my irc account and it would bridge over
<starcrATI[m]>brendyyn: but if you mean Matriy, yes you can do group calls and video chat via Jitsi woth Matrix
<brendyyn>via jitsi??
<brendyyn>wut the heck
<brendyyn>i hate python how many hours do i have to spend just trying to make it find some stupid module
<anadon>Good morning, all.
<brendyyn>starcrATI[m]: thats cool
<starcrATI[m]>brendyyn: what module ? xD
<brendyyn>although i doubt its that reliable
<brendyyn>i wonder if perhaps i should become a javascript developer instead
<brendyyn>but i just dont get the impression ill be able to build tools how i want them
<starcrATI[m]>where did you find a python module for jitsi ?
<brendyyn>i didnt
<Veera>How to build only a specific package in image-viewers
<NieDzejkob>just do "./pre-inst-env guix build feh" or whatever
<apteryx>civodul: discordapp worked after all, on a beefier machine with more RAM (even with the old Icecat 60 there was on that machine)
<guix-ci>Problem: Zabbix discoverer processes more than 75% busy Problem started at 14:43:35 on 2020.03.27
<civodul>rekado: this message is nicer!
<mbakke>apteryx: the (format "~s" (which "sh")) line from emacs' patch-program-file-names phase fails on core-updates, likely from the Guile 3 switch
<mbakke>format: bad destination `~a' "~s"
<guix-ci>Resolved: Zabbix discoverer processes more than 75% busy Problem has been resolved at 14:50:35 on 2020.03.27
<civodul>apteryx, mbakke: it should be (format #f "~s" (which "sh"))
<guix-vits>what is "Bitstream Vera Font (c)" converted to `(license`?
<mbakke>civodul, apteryx: indeed, but the substitution expects the returned string to be surrounded by double quotes, so I'm not sure how that works on 'master'.
<mbakke>trivial fix in any case :)
<efraim>Should probably add the timezone to the message
<civodul>mbakke: the first arg to 'format' must be a port
<civodul>i think it was deprecated (thus allowed) in 2.2
<NieDzejkob>Looks like berlin time based on the timing
<civodul>and presumably finally unsupported in 3.0
<kawerte>How can I see the sturcture and possible options for settings in the config.scm file?
<mbakke>kawerte: see the operating-system reference in the manual:
<kawerte>I saw it, that is why I'm asking
<kawerte>for example I have no idea from reading the manual how to have multiple input methods
<kawerte>any help with that?
<brendyyn>where is the input method bit?
<kawerte>How should I know what #options there are and what possible things I can write for that function?
<kawerte>(if it even is a function, I am new to lisp)
<brendyyn>im interested in improving the manual in this regard, just havent got around to it
<roptat>hey, found a japanese blog post about guix: :)
<brendyyn>"Keyboard layouts are represented by records created by the keyboard-layout procedure of (gnu system keyboard). Following the X Keyboard extension (XKB), each layout has four attributes: a name (often a language code such as “fi” for Finnish or “jp” for Japanese), an optional variant name, an optional keyboard model name, and a possibly empty list of additional options. In most cases the
<brendyyn>layout name is all you care about. Here are a few example: "
<kawerte>and this doesn't say anything
<kawerte>"additional options" hahahaha
<brendyyn>see records here
<rekado>FWIW I’m using Gnome and didn’t add any input method configuration to my operating-system record. I just selected my input methods in Gnome.
<rekado>the “additional options” are the same that are supported by xkb
<rekado>(like dead keys, control moved to caps lock, etc)
<brendyyn>kawerte: As rekado says. You say you want to define multiple layouts, but actually every layout is already available regardless. this is just setting your default one
<brendyyn>I have a customised layout, but it's set in my users config, not with guix
***jonsger1 is now known as jonsger
<jeko>hello guixters !!
<jeko>preparing my new package patch, what is scheduled Software Heritage archival ? from guix lint ??
<civodul>jeko: nothing to worry about!
<jeko>cool :p
<mbakke>jeko: guix integrates with the Software Heritage to ensure package sources are available even after the upstream is long gone :-)
<jeko>ok so I let it that way
<rekado>on /var/log has reached 83G. I think we can remove some of these log files…
<civodul>do we have a job for /var/log/guix/drvs/?
<raghavgururajan>I cloned guix git repo and did `./bootstrap` and `./configure --localstatedir=/var`. When I did `make`, I get this error.
<civodul>with 3.0.2 we should be able to get rid of guile-2.0 for grafts
*civodul runs torture tests
<raghavgururajan> en@boldquot.po:4584: format specifications in 'msgstr[0]' are not a subset of those in 'msgid_plural'
<raghavgururajan>What should I do?
<NieDzejkob>oh, I've hit that yesterday
<NieDzejkob>[01:59:21] <mroh> NieDzejkob: try `make SUBDIRS=` as a workaround.
<raghavgururajan>NieDzejkob Me?
<raghavgururajan>Thanks! It's working
<jeko>Patch sent. Hope I didn't miss anything !
<roptat>jeko, well done :)
<Gooberpatrol66>is there a way to regenerate the program icons in GNOME so I can launch a program right after installing it?
<jeko>roptat: thank you ! but I should wait for the validation before yaying hahaha
<guix-ci>Problem: Zabbix discoverer processes more than 75% busy Problem started at 15:55:35 on 2020.03.27
<joshuaBPMan>Hey guix! Last night I installed guix onto my T400. with an encrypted / partition (/dev/sda2 btrfs) and unencrypted /boot (/dev/sda1 ext4).
<joshuaBPMan>I do not seem to be able to boot from it via libreboot, which has a grub payload...maybe i need to mark /dev/sda2 as the boot partition...I do not believe that I did that.
<sirgazil>raghav-gururajan: regarding the "make" error: That's a known issue:
<joshuaBPMan>and I suppose I see the error in my config now. my mapped-devices has a different target than what my filesystem does.
<NieDzejkob>guix-ci seems to be messaging here about problems with a 10 minute delay, but even if the problem got resolved within the 10 minutes, it still complains here. What gives?
<guix-ci>Resolved: Zabbix discoverer processes more than 75% busy Problem has been resolved at 16:01:35 on 2020.03.27
<apteryx>mbakke: eh, good catch
<joshuaBPMan>Also, how can I run guix system reconfigure from the usb stick such that the command reconfigures the hard drive?
<joshuaBPMan>Do I need to run herd start cow-store /mnt?
<joshuaBPMan>I think I do need to run herd start cow-store. At least the manual seems to suggest that.
<apteryx>has anyone managed to get Icecat to show right-click context menus? Such as should be possible when right-clicking on issues from that page:
<roptat>do you mean the generic "open in new tab, ..." or something specific to that website?
<roptat>I get the former
<apteryx>roptat: something specific (context sensitive) to the web app (website)
<roptat>I get lots of Content Security Policy: The page’s settings blocked the loading of a resource at inline (“script-src”). for that page
<apteryx>like; actions for an issue (set its status such as 'close', 'open', etc.). I've seen this work on other collegues' browsers, so I'm curious as to why I don't get this in my Icecat :-)
<apteryx>roptat: yeah, I've whitelisted/unblocked the page (Redmine is free software)
<sirgazil>apteryx: I see a context menu specific to the website in IceCat.
<sirgazil>But only if I click in the row of an issue, not in the issue link
<roptat>oh yeah, after allowing scripts and xhr, I see the specific context menu too
<roptat>right-clicking on the links show the default menu though
<apteryx>weird. Can't get it to work on my Icecat, even after whitelisting the domain in LibreJS.
<apteryx>roptat: how do we 'allow xhr' ?
<joshuaBPMan>bummer. It doesn't really seem that I'm able to do that.
<roptat>ah sorry, it's because I use umatrix
<roptat>it should be allowed by default
<apteryx>I'll try updating my Icecat, I'm on an older 68.4.1.
<apteryx>hm. Same thing.
<apteryx>ah, it worked after starting in safe mode (all add-ons disable).
<apteryx>Now to figure out which one it was :-)
<asdKovind>Hi, I am running `guix install next` but getting error message `guix install: error: next: unknown package`
<rekado>asdKovind: have you run “guix pull” recently? Not all versions of Guix have a package called “next”.
<rekado>asdKovind: you can see what packages are available with “guix search”
<asdKovind>I searched I am getting sbcl-next
<asdKovind>But it is not working
<rekado>what do you mean by “it is not working”?
<asdKovind>may its the package issue, I will ask on there channel
<apteryx>roptat, sirgazil: the behavior was caused by the 'Searxes' Third-party Request Blocker', even it didn't show anything being blocked for that particular page. Strange.
<apteryx>ah, but I found a way to configure it: on the right-click with TPRB enabled, you have some TPRB entry, where you can whitelist the action. After doing so it works :-)
<apteryx>thanks for the testing
<guix-vits>i'm just switched to copy-build-system... i'm happy.
<civodul>"Guix as an apt replacement"
<civodul>(no pun)
<jonsger>long article and 40 comments. Not bad :)
<civodul>they write that is unusable due to the lack of a search box
<roptat>most of these comments are about the author suggesting using "wget | sudo bash" though
<janneke>if only guix had a search box, it might be usable
<janneke>i would even consider trying guix, then
<roptat>it wouldn't harm to implement that search box though
<janneke>yes, and what about this curl stuff; surely it should be" curl ... install.scm | sudo guile"
<civodul>we could take it from hpcguix-wegb
<jonsger>^ +1
<civodul>BTW, if you can comment roptat, you could point them to
<civodul>the article is interesting, it even has a fair "What about Nix?" section, which is uncommon
<janneke>and talks about bootstrappable!
<civodul>it's a salient feature
<civodul>many people realize that it really matters
<civodul>and that's good!
<guix-ci>Problem: Zabbix discoverer processes more than 75% busy Problem started at 17:05:35 on 2020.03.27
<roptat>mh... isn't this 75% due to package building? or is there something else?
<guix-ci>Resolved: Zabbix discoverer processes more than 75% busy Problem has been resolved at 17:13:35 on 2020.03.27
<sirgazil>I'm getting "Error downloading release information through the GitHub API." when I try to lint a package I'm defining...
<sirgazil>"This may be fixed by using an access token and setting the environment variable GUIX_GITHUB_TOKEN"
<NieDzejkob>sirgazil: you must've been linting a lot of packages
<joshuaBPMan>geez, I cannot figure out how to install guix with an encrypted root partition
<sirgazil>NieDzejkob: About five, since yesterday.
<sirgazil>Python packages I need.
<sirgazil>I wonder how long you have to wait to avoid that error...
<apteryx>civodul: I tried jitsi with ungoogled-chromium, but I won't allow my microphone. Did you need to do something special for this to work?
<raghav-gururajan>sirgazil I see.
<NieDzejkob>sirgazil: hmm, that might be an actual problem with the GitHub API, then
<NieDzejkob>are you sure you aren't re-linting the entire package repository by accident?
<watman>Is there any actual way the see all the available options for the system configuration (config.scm file)
<sirgazil>apteryx: requires you to use Chrome specifically. Maybe it only works [well] in Google Chrome.
<bandali>works for me on firefox derivs
<apteryx>bandali: it does, but firefox does not implement RTX which can lead to breaking audio
<bandali>but they do seem to recommend chrom{e,ium}-based browsers sadly
<bandali>apteryx, oh yeah? interesting
<apteryx>apparently, if your connection is saturated
<NieDzejkob>watman: see info "(guix)operating-system Reference"
<apteryx>(in a multi-users conference setting, say)
<bandali>i might have experienced something like that; not a deal breaker so far thankfully
<mbakke>apteryx: there was a bug with older versions of ungoogled-chromium that prevented websites from accessing your microphone
<anadon>Is the `nginx-upload-module` just missing from guix?
<apteryx>mbakke: just visiting triggers the error for me
<mbakke>apteryx: to fix it, either create a new browser profile, or remove the "audio_capture_enabled: false" line from ~/.config/chromium/Default/Preferences
<apteryx>it says: "Failed to get access to local media. Error name was AbortError. Continuing without sending a stream."
<apteryx>mbakke: will try
<apteryx>mbakke: should be good if I delete ~/.config/chromium ?
<apteryx>I don't have anything important there
<mbakke>apteryx: that's should work yes
<civodul>apteryx: re Jitsi, maybe check what pavucontrol says
<civodul>and also the "Settings" menu of the Jitsi interface
<apteryx>mbakke: works perferctly now :-) thank you!
<mbakke>apteryx: cool, sorry for the breakage!
<apteryx>no issues
<apteryx>thanks for making chromium available!
<alextee[m]>does it make sense that package URLs include the version?
<alextee[m]>they're not very future-proof if you want to link the package URL somewhere, unless it auto-redirects to the newest version
<alextee[m]> for example
<alextee[m]>what happens when it changes to 1.4? 404?
<civodul>alextee[m]: i agree, the version shouldn't be in the URL
<civodul>although there are packages with several versions
<civodul>maybe that's the reason
<alextee[m]>well, they could be listed as "Available versions: ..." or something on the page i guess
<alextee[m]>or "Other versions: ..." and have the latest one shown in the title as like now
<guix-ci>Problem: Zabbix discoverer processes more than 75% busy Problem started at 18:17:35 on 2020.03.27
*jonsger enjoys now the heroic work of rekado on R :)
<guix-ci>Resolved: Zabbix discoverer processes more than 75% busy Problem has been resolved at 18:24:35 on 2020.03.27
<civodul>alextee[m]: yeah, that would work
<civodul>would you like to give it a try? :-)
<civodul>it's a good excercise for a short hacking session
<alextee[m]>civodul: sure, good chance to practice my guile skills!
*alextee[m] clones the repo
<bricewge>anadon: Yes, it's not yet packaged.
<rekado>don’t know what Zabbix discoverer processes are, but … should we allow them to be busier on
<civodul>alextee[m]: heheh :-)
<civodul>let us know if you need guidance!
<civodul>rekado: difficult to answer that question :-)
<civodul>it's good to hear from guix-ci once in a while, but we don't quite understand each other
<civodul>speaking of bots...
<civodul>cbaines: there's guile-mastodon
<civodul>would it make sense for the Data Service to send messages?
<civodul>like when there's a push with new packages, etc.
<cbaines>I would like that, and it's something I've been thinking about :)
<bavier`>that would be fun
<cbaines>Just need to get around to writing that publish/subscribe system that the Guix Data Service can send events to...
<rekado>cbaines: use redis with guile-redis?
<cbaines>that's definately easier technically than what I have in mind
<cbaines>I've been thinking of doing a websub like thing, the advantage of that being it's a bit more federated. I think just having a service that sends out HTTP requests for the events would make it easier to allow anyone to just write software and subscribe.
<civodul>sounds like a plan
<jsoo>is there a configure flag for emacs to avoid needing texinfo?
<mbakke>jsoo: sometimes you can pass "TEXINFO=true" to trick the build system into calling the 'true' binary instead, no idea whether it will work for emacs though
<civodul>Guile 3.0.2 is upon us!
<jsoo>hmm, nice! i'll give it a go. i'm trying to make emacs-next-minimal
<mbakke>civodul: yaay :)
<mbakke>civodul: tangentially, do you think it's safe to use libgc 8 for Guile 2.2 and 2.0, or should they stick with libgc@7?
<raghavgururajan>Hello Guix!
<jsoo>o/ hi raghavgururajan
<raghavgururajan>I get "ld: undefined reference to `CU_trace_handler'". What package provides this? I am unable to find the correct package.
<jsoo>man i really like the general programming style in guix. i tend to really like types but guile really feels right most of the time
<civodul>mbakke: it should be safe, but worth checking
<Blackbeard>Good morning guix :)
<alextee[m]>no code for module (syntax-highlight)
<alextee[m]>i installed guile3.0-syntax-highlight and guile-syntax-highlight. is that not it?
<roptat>are you missing the guile environment variables?
<roptat>ah, tor doesn't work over IPv6 only :/
<alextee[m]>roptat: im not sure, what would those be?
<roptat>I have them set to ~/.config/guix/current/share/guile/site/3.0 and ~/.config/guix/current/lib/guile/3.0/site-ccache respectively
<alextee[m]>they're in the environment: /run/current-system/profile/lib/guile/2.2/site-ccache and /run/current-system/profile/share/guile/site/2.2
<roptat>did you install guile-syntax-highlight globally? otherwise, you might need guile itself so guix knows you want the variables to be set to your personal profile too
<alextee[m]>oh didn't know you were supposed to set them manually for your user
<roptat>no, guix takes care of that, but you need to install whatever package declares them (in general, the one that uses them, here guile itself) in the same profile
<alextee[m]>roptat: what would globally mean? sudo -E guix install .. ?
<roptat>I mean, in the packages field of your operating-system declaration :)
<roptat>sudo guix install would only install packages for root, not very useful
<alextee[m]>oh no i just installed everything in my user's profile, including guile
<Blackbeard>bavier`: hey! Did you give up with BOINC?
<roptat>Blackbeard, I tried to run boinc yesterday, but it couldn't access the internet for some reason, with no log whatsoever
<roptat>alextee[m], you might want to source ~/.guix-profile/etc/profile then
<roptat>it should contain definitions for guile and other stuff
<roptat>also I think you need the same version of guile as the version used by haunt
<alextee[m]>roptat: hmm weird, i thought i was already sourcing that. sourcing it now worked, thanks! will add it in my zshrc
<Blackbeard>roptat: yes, me neither :/
<guix-ci>Problem: Zabbix discoverer processes more than 75% busy Problem started at 19:28:35 on 2020.03.27
<guix-ci>Resolved: Zabbix discoverer processes more than 75% busy Problem has been resolved at 19:35:36 on 2020.03.27
<bavier`>Blackbeard: gave up for now, yes
<cbaines>Haha, guix-ci does not like to stick around!
<bavier`>I was thinking of installing it to a fedora vm, but my virt-manager is refusing to start right now (think I need a reboot)
<cbaines>bavier`, virt-manager broke for me recently as well
<cbaines>the package seems to break quite often
<cbaines>I just started it from a previous generation, so maybe check if you have a previous generation of your profile you can start it from
<cbaines>On a somewhat related note, you can ask the Guix Data Service for the "output" history for a package, like virt-manager:
<rekado>guix-ci doesn’t stick around because my silly little script just makes it pop in, drop a message, and return.
<bavier`>ha, yup, older version starts. yay time-travel :)
<cbaines>good good :)
<cbaines>If the data in the Guix Data Service is correct, there have been 14 different outputs for virt-manager this month
<rekado>one could implement it as a service that runs a bot loop and listens on a socket for new messages.
<guix-vits> -- LC_ALL=zh_CN.utf8 neverball not working. Can someone help?
<cbaines>I'm interested to have a look at Zabbix, I need SSH access to berlin right?
<rekado>cbaines: you need SSH access only to generate your client certificate.
<rekado>guix-vits: do you get an error?
<guix-vits>rekado: i see "squares" instead of chars.
<NieDzejkob>do you have that font installed in your profile?
<NieDzejkob>maybe it needs to get picked up by the fontconfig cache...
<NieDzejkob>does it work in other applications?
<mroh>maybe its only me, but applying grafts feels faster now! (2b6fe60599 packages: Use Guile 3.0 for grafts)
<guix-vits>NieDzejkob: i see Chinese chars in Web-browsers
<NieDzejkob>does the issue occur if the font isn't added as a dependency?
<guix-vits>i added this font later, when seen the issue
<guix-vits>the game, actually, have some fonts boundled
<NieDzejkob>are they ttf?
<Blackbeard>rekado: I made one like that with 8sync o couple years ago :)
<guix-vits>NieDzejkob: yes
<guix-vits>i'll try outside of `guix environment` now.
<NieDzejkob>guix-vits: the game seems to be bundling its own chinese font, see data/lang/zh_CN.txt and data/ttf/wqy-microhei.ttc
<NieDzejkob>is anything printed to the console when the game tries to display chinese characters?
<guix-vits>no (i not yet looked in zh_CN.txt; thanks)
<cbaines>daniela18, hi o/
<guix-vits>actually microhei is in repos, maybe it'll fly?
<alextee[m]>is adding dependencies to the guix website code acceptable?
<alextee[m]>thinking to use this to compare versions
<lfam>How can I take advantage of the recent GnuPG pinentry change? I am using GnuPG and pinentry-gtk2 from Guix, but even updating to include that change, I still need to manually set pinentry_program variable in gpg-agent.conf
<joshuaBPMan>lfam: what is the advantage of the recent GnuPG change?
<joshuaBPMan>What is the change?
<lfam>It's commit c7af9d0b5ebaa1fdb08ff5d8a56004998bcd8103
<lfam>The idea was to make our GnuPG find our pinentry packages automatically
<vagrantc>that sounds nice :)
<lfam>But it doesn't work for me :( I restarted all of the GnuPG daemons
<lfam>It looks like it's based on having the correct value for 'gnupg_homedir' in the GnuPG code
<lfam>Right, it works if you export GNUPGHOME
<lfam>Can't GnuPG just read $HOME?
<guix-ci>Problem: Zabbix discoverer processes more than 75% busy Problem started at 20:40:35 on 2020.03.27
<bavier`>Blackbeard: well, bit of a copout, but I installed boinc on a fedora vm and have it running tasks for rosetta@home
<Blackbeard>bavier`: is it exactly the same?
<Blackbeard>bavier`: I wonder what's happening
<Blackbeard>Did you check the certificates that you told me about?
<bavier`>fedora has a slightly older version packaged (7.16.1)
<bavier`>idk if that'd make a difference
<bavier`>my debug-fu fell short in debugging the network issue in guix's package
<guix-ci>Resolved: Zabbix discoverer processes more than 75% busy Problem has been resolved at 20:47:35 on 2020.03.27
<sirgazil>alextee[m]: regarding adding dependencies to build the website, I don't see why not.
<sirgazil>alextee[m]: What are you working on, by the way :)
<joshuaBPMan>may I ask, is there anyone working on updating sway?
<vagrantc>ooooh, core-updates looks to have some very nice features coming!
<vagrantc>can it cross-compile hurd systems on x86_64 ?
<jonsger>rekado: what happend with I try to use R on Guix System and have problems (certs not found)
<mbakke>vagrantc: the focus has been on cross-compiling for ARM, but it should be possible to cross-compile for Hurd as well
<jonsger>I don't understand R. Does it need a special SSL_* variable
<mbakke>joshuaBPMan: not to my knowledge, I'm sure many people would appreciate it though! ;-)
<vagrantc>really cool to hear the cross-compiled system support is working ... that was one of the biggest annoyances with arm systems and guix ... the initial install took *forver*.
<mbakke>vagrantc: right, it should be possible to cross-compile the initial system now and convert it to a native system after flashing
<alextee[m]>sirgazil: trying to make the URL of packages that are at their latest verison not have the "-{version}" suffix, so they are future-proof
<mbakke>raghavgururajan: your linphone patch seems to be missing something important
<alextee[m]>and add a "other versions: " to packages that have multiple versions
<mbakke>1 file changed, 0 insertions(+), 0 deletions(-)
<mbakke>raghavgururajan: also I suppose we could use telephony.scm for linphone?
<alextee[m]>sirgazil: im currently checking with semver, and when the versions are not semantic, i do a plain string>? to check which is later
<guix-vits>NieDzejkob: update, if i choose language in GUI, things works with Chinese fonts... yeah.
<aidenholmes>i was told to bring up this idea here
<aidenholmes>guix is currently slower than nix in terms of download bandwidth speed, this is because, iirc, you only have a few donated servers
<aidenholmes>guix could be faster by using p2p filesharing
<mbakke>hmm, apparently libgc needs libatomic-ops when cross-compiling, wonder why that is
<civodul>aidenholmes: yes, this has been on our minds for some time!
<joshuaBPMan>aidenholmes: there is a wip branch for the universal filestystem
<str1ngs>i'm optimistic for ipfs support :)
<str1ngs>gnunet would be cool too.
<aidenholmes>i used to be huge fan of ipfs but in practice it's much slower than gnunet
<aidenholmes>or simply, torrent
<str1ngs>I've not had a problem with ipfs
<aidenholmes>if a p2p option is added, i hope it's optional because i wouldn't be able to use it at school
<aidenholmes>oh? maybe it' improved since i last used it
<str1ngs>in theory gnunet could be used but it would be way more work then ipfs.
<civodul>we also had a GSoC on GNUnet integration back in 2016 IIRC
<str1ngs>I think the way ipfs does content addressing translates well to guix
<str1ngs>I wrote an offline ipfs hasher at one point. could be useful guix eventually
<guix-vits>NieDzejkob: probably it's not an UTF?
<rekado>jonsger: nobody read the documentation and the difference between the deployment at the MDC and vanilla Guix had almost completely disappeared, so I replaced it with
<lfam>aidenholmes: It may be that our servers are slower than Nix's, but the big slowdown for Guix is that things are downloaded sequentially, and setting up and tearing down the download session takes a long time
<jonsger>rekado: ah okay. I hoped there would be some doc about how to use R on Guix
<lfam>aidenholmes: Some of us abuse the server (our abuse our private mirrors) by setting --max-jobs=N and letting downloads occur concurrently
<lfam>This is a nasty kludge but it works
<rekado> should be able to handle quite a few simultaneous connections.
<lfam>I feel like
<lfam>I feel like there was some preliminary work on pipelining these requests but I'm not sure what the status is
<rekado>(I think “guix publish” might be the limiting factor here)
<lfam>Perhaps, but it's very slow for me just setting up connections to my private nginx mirror, which doesn't use guix publish
<joshuaBPMan>lfam: how are is it to set up a private mirror? I've got a 1GB RAM server with 20 GB harddrive....
<rekado>aidenholmes: we have 30+ servers that build packages. There’s one head node with a dual 10G connection.
<rekado>jonsger: I there something particular you are looking for?
<lfam>joshuaBPMan: I adapt our upstream mirror code:
<lfam>joshuaBPMan: Specifically mirror.conf and mirror-locations.conf
<rekado>jonsger: using R with Guix is pretty straight forward: “guix install r-minimal r-…”
<aidenholmes>oh, could the slow downloads be on my end then , rekado
<aidenholmes>i doubt it because apt downloads faster
<joshuaBPMan>lfam: thanks. I sort of forgot that we are still using hydra.
<lfam>We aren't using the hydra software.
<lfam>We still use the name as a catch-all for our build farm stuff
<rekado>aidenholmes: I do think there are performance problems with the software. I just don’t think it’s a problem with our hardware.
<joshuaBPMan>lfam: oh. I thought we were calling it curraiss (I probably spelled that incorrectly).
<aidenholmes>ah, well there are other reasons p2p would be useful as well
<lfam>We are calling the software Cuirass
<lfam>Don't worry about what that directory is called. It's not importnat
<lfam>It's just a name in a dark corner of our support code
<jonsger>rekado: I do but I get errors when accessing https sites: I have no clue why. I have set all those SSL_* env vars we have
<aidenholmes>imo it just makes sense that package managers, which deal with a lot of publicly available and widely downloaded files, should use p2p file sharing
<lfam>Well it's tricky because most clients don't have access to much upstream bandwidth
<guix-ci>Problem: Zabbix discoverer processes more than 75% busy Problem started at 21:51:35 on 2020.03.27
<jonsger>I guess the download speed could be a problem of peering or so (between the relevant ISPs)
<joshuaBPMan>Also, I just read on this channel that gnunet may be faster than IPFS. I thought IPFS was I wrong?
<civodul> saw a release this month, the GNU package manager :-)
<aidenholmes>the last time i tested it gnunet was much faster
<lfam>I think the low-hanging fruit is the Guix download code. It could be made a lot more efficient
<lfam>That and using a CDN
<lfam>There is a discussion in this thread:
<lfam>It spans multiple months so you'll have to dig to find later messages
<lfam>Here is a message with some examples:
<jsoo>Is there any news on the content addressed protocol for substitutes? That really caught my eye
<lfam>Plus the previous discussion:
<bdju>in my experience ipfs is extremely slow, but I don't know the speed of gnunet at all
<rekado>jonsger: I don’t know how r-xml2 connects to the web. I know that I had to patch at least r-curl (or was it r-rcurl?) to respect the SSL_* vars. It’s possible this needs to be done for r-xml2 as well.
<rekado>using a CDN (again) would be good. It’s just pretty expensive.
<guix-ci>Resolved: Zabbix discoverer processes more than 75% busy Problem has been resolved at 21:57:35 on 2020.03.27
<rekado>then again it’s not like we’re lacking money
<lfam>Yes, the CDN was very expensive. I think the best course of action is to improve the Guix download code to pipeline requests. And also to profile it to see if anything is slower than necessary
<lfam>We have money but do we have a way to replenish that money when it runs out?
<civodul>it would be good :-)
<civodul>it's actually useful
<lfam>I feel like the handshake foundation was sort of a special situation
<lfam>We are still not pipelining nar requests, right?
<rekado>(personally, I’d prefer to buy build machines for ARM + POWER for the money)
<civodul>also, there's slowness on the nginx side of things on the first hit
<civodul>i investigated a while back but didn't understand
<jonsger>I could provide 0.5T fast storage for a distributed CDN/rsync mirror what ever...
<rekado>jonsger: xml2/R/xml_parse.R says that read_html accepts URLs “that will be converted into connections either using `base::url` or, if installed, `curl::curl`”.
<lfam>How much bandwidth could you provide jonsger?
<rekado>jonsger: so perhaps you could try installing r-curl and see if that fixes it for you
<jonsger>lfam: don't know if there is a limitaion
<lfam>We used *a lot* of bandwidth in the months we paid for a global CDN
<lfam>I feel like storage is not the limitation for us
<lfam>Here is a report on the CDN usage:
<Rovanion>Is it possible to find out which package provides some binary?
<lfam>Okay, it's not like "a lot" but it's significant and usage will only grow
<jonsger>rekado: a simple r-curl example works. adding library(curl) to rvest doesn't help
<lfam>Rovanion: There's not a "Guix way" yet
<jonsger>lfam: hetzner usually garantees at least 20T free each month
<Rovanion>Does anyone happen to know which package provides alsamixer then?
<lfam>Is that a CDN jonsger? Or a single point of presence?
<Rovanion>lfam: Is there another way?
<rekado>Rovanion: alsa-utils
<rekado>rekado: botsnack
<Rovanion>rekado: Thank you!
<lfam>Rovanion: Google, `guix search ...`, even `echo /gnu/store/*alsamixer*` can help
<lfam>There has been some discussion of adding this feature
<jonsger>lfam: single point of presence. a hosting provider
<Rovanion>Tried Google and guix search but they failed me :/
<lfam>jonsger: It might not be an improvement over our current peering
<lfam>The CDN was good because it helped people in all continents and different internets
<rekado>(and for China the way forward might be the rsync daemon)
<jonsger>lfam: yes. I know. But I have usually 1.5MiB/s from, could become faster...
<lfam>Can you say approximately where you are jonsger?
<rekado>jonsger: we also see much lower download speeds than theoretically possible within the same network
<jonsger>lfam: Nuremberg or Stuttgart
<rekado>so we think it’s a problem with our nginx configuration or guix publish
<lfam>I think the kind of slowness we have discussed previously was kilobytes per second. At least in the USA, 1.5 MiBs is not terrible
<rekado>still, it’s a fraction of what it could be, given the hardware.
<jonsger>lfam: it's not terrible but faster would be nicer :)
<lfam>It's a reason I use my own mirror :)
<lfam>Only do the slow request once
<cbaines>One thing I'd like to prototype is a more static site approach to substitutes
<cbaines>So rather than using guix publish, generate the nars and narinfo files, and then copy them somewhere to be served
<cbaines>Serving straight from the storage should be pretty quick
<jsoo>cbaines: is that what nix does?
<rekado>cbaines: I’d like that. This will be needed by the rsync mirrors anyway.
<jonsger>cbaines: so then they could served by every webserver
<rekado>we’re exposing the guix publish cache via rsync daemon.
<jsoo>would that fit with the content-addressable substitute protocol?
<lfam>The Guix download code would still download things one at a time, and it takes forever to set up and tear down those connections
<lfam>IMO that's the place to focus efforts
<jsoo>do you mean the one at a time problem or the setup/teardown time lfam?
<lfam>The one at a time thing
<lfam>If you only do it once it doesn't matter very much that the setup is slow
<jsoo>gotcha. can we do some asynchronous io stuff for that?
<cbaines>I remember roptat posting something about parallelising fixed output derivations I think
<roptat>yes, there's a patch that changes the daemon
<roptat>it's not really polished though
<civodul>this is to overlap downloads with builds
<civodul>what lfam suggests would also be nice but requires more important changes in the daemon
<roptat>the idea was to count builds and downloads separately, so you could have both in parallel
<cbaines>right, OK
<lfam>The whole discuss I linked to previously is important, but this exchange is key:
<pkill9>that would be good
<lfam>"but if the current code
<lfam>> waits until a NAR has finished downloading before asking for the next
<lfam>> one, that's an issue that could be fixed by use of HTTP pipelining,
<lfam>> without multiplying the memory usage."
<lfam>"There’s HTTP pipelining for narinfos but not for nars. Worse,
<lfam>before fetching a nar, we do a GET /nix-cache-info, and in fact we spawn
<lfam>a new ‘guix substitute’ process for each download (for “historical
<jsoo>what are the reasons, do you know?
<civodul>we no longer do GET /nix-cache-info
<lfam>It's terribly inefficient and means that downloading 1 megabyte of files can take minutes
<jsoo>ok that's nice at least
<lfam>That's good!
<lfam>I think it's late in the application cycle but could this be a good area for one of our paid internship programs?
<rekado>jonsger: the problem might be with r-httr
<jonsger>rekado: should I open a bug?
<rekado>jonsger: yes, please!
<rekado>I was just about to suggest that but couldn’t type fast enough with the baby on my arm
<jonsger>guix search kindergarten gives results :P
<joshuaBPMan>Hey guix! I'm using the graphical install on a T400. There seem to be some changes fromthe last time I used it. It looks pretty nice. Keep up the good work!
<civodul>lfam: for Outreachy we'd have to have a rather clear idea of what we want to do
<lfam>Right, we'd need something concrete and composable
<civodul>what are the main milestones towards our goal
<lfam>Maybe for next time
<civodul>yes, it sounds like it could work
<alextee[m]>anyone know any package early in the alphabet with multiple versions?
<joshuaBPMan>alextee[m]: may I ask why you ask?
<alextee[m]>joshuaBPMan: trying to make the package with the latest version not have a -{version} in its URL
<NieDzejkob>why does "early in the alphabet" matter, though?
<alextee[m]>NieDzejkob: because the development environment only uses the first 100 or so packages
<rekado>jonsger: I can reproduce the problem with r-curl alone. It should respect CURL_CA_BUNDLE, but it doesn’t seem to work.
<NieDzejkob>jonsger: sounds like something that I've fixed on core-updates
<jonsger>NieDzejkob: a76a343 or?
<NieDzejkob>jonsger: yup, here's a variant with a graft:
<NieDzejkob>you can apply it on master and it'll only need to build curl
<jonsger>meeh. I'll wait until core-updates gets merged into master. it's not urgend
<rekado>this shouldn’t be necessary for r-curl
<rekado>we patch it to accept CURL_CA_BUNDLE (which it normally only accepts on Windows)
<NieDzejkob>huh, don't mind me then
<guix-ci>Problem: Zabbix discoverer processes more than 75% busy Problem started at 23:01:35 on 2020.03.27
<alextee[m]>i did it!!
<rekado>jonsger: this seems to be a change in r-curl. CURLOPT_CAINFO is now only set when the curl backend is openssl. It’s gnutls in Guix.
<rekado>it’s also behind an #ifdef _WIN32 now
<guix-ci>Resolved: Zabbix discoverer processes more than 75% busy Problem has been resolved at 23:09:35 on 2020.03.27
*alextee[m] uploaded a video: Screencast from 27-03-20 22:19:37.webm (970KB) < >
<alextee[m]>^ is this ok?
<lfam>So what does this guix-ci message mean?
<lfam>guix-ci: Help
<jonsger>lfam: I guess that the ci's CPU's are running :P
<Blackbeard>alextee: I can't see the video
<rekado>jonsger: got a fix
<rekado>lfam: this is triggered by, which runs a Zabbix instance monitoring and connected build nodes.
<alextee[m]>Blackbeard ( weird, it works in riot on another machine
<alextee[m]>maybe download it first and play through mpv
<rekado>lfam: any event produced by Zabbix is reported via a dumb IRC script that logs on to freenode as “guix-ci” and prints the event message.
<rekado>I don’t know what “Zabbix discoverer processes” are, but I guess this might be server auto-discovery, which we don’t need.
<rekado>I’ll see if I can disable this (after fixing r-curl)
<Blackbeard>alextee: I am using riot on android, let my try from my pc
<alextee[m]>where should i send patches for guix-artwork? same as for packages
<alextee[m]>rekado: thanks
<Veera>Hi guix
<Blackbeard>Veera: hi
<Veera>Blackbeard: Hi
*rekado disabled Zabbix discovery for the local network around
<lfam>Thanks rekado
<jonsger>rekado: isn't a proper .patch saver then substitute* for r-curl. So that it does not silently break?
<Blackbeard>Veera: :)
<rekado>jonsger: maybe. I still think that substitute* should throw an error when a clause doesn’t match.
<rekado>silent breakage should be avoided
<rekado>in this particular case the problem was not that the patch wasn’t applied; it was that the patch became ineffective.
<rekado>(due to other changes in the code)
<Blackbeard>alextee: it works from desktop :)
<lfam>I agree rekado. If it never matches it should fail
<rekado>changing this would probably break a bunch of packages; we should make sure that is equipped to show us those new failures before we merge this.
<rekado>(this is a world-rebuilding change)
<lfam>I like the new UI message "dependencies changed". It should reduce confusion. But it also prints when only the package definition changed — no changes in the package dependency graph. Maybe we can amend the message somehow...
<lfam>Maybe just "dependencies or package definition changed"
<mbakke>rekado: I think NieDzejkob was working on something along those lines