IRC channel logs

2017-09-13.log

back to list of logs

<civodul>maybe!
<civodul>(don't hold your breath, i don't plan to start working on a TLS lib any time soon!)
<dustyweb>:)
<oriansj>just a stupid question but does guix have a linter for package/system definition files?
<civodul>oriansj: yes: https://www.gnu.org/software/guix/manual/html_node/Invoking-guix-lint.html
<catonano>lfam: I was refering to this :-) https://gnunet.org/bot/log/guix/2017-09-11#T1499062
<catonano>it was about posting my code on paste.lisp.org or in the body of the mail message
<oriansj>sneek: later tell civodul that the link provided earlier is about checking existing packages rather than package/system definition files
<sneek>Okay.
<oriansj>sneek: botsnack!
<sneek>:)
<pmikkelsen>hey guix, sorry if I am asking a stupid question, but is there a way to run a vm for another architecture on my amd64? I just made some changes i would like to test on arm
<efraim>I suppose it could be built with `guix system build --system=armhf-linux foo', but if you don't have a way to build armhf packages you're at the mercy of the substitute servers for some/most(?) of the packages
<marusich>Won't cross-building work?
<efraim>I don't know that all of the build-systems support that
<marusich>Anyway, pmikkelsen, assuming you already have a disk image for architecture foo, look into running qemu-system-foo
<marusich>there are a bunch of qemu-system-* commands that let you run a virtual machine of a specific architecture on your machine.
<pmikkelsen>okay thanks, I'll give it a try, otherwise I will see if i can find an arm machine. I think I have a beaglebone black and a raspberry pi somewhere, but they are different ARMs right?
<pmikkelsen>Not that i expect you to know ;)
<marusich>I don't know.
<buenouanq>official arm support is greatly desired
<buenouanq>I want nothing more than to be able to install GuixSD on the EOMA68 when it finally ships.
<pmikkelsen>ACTION Successfully build a small program consisting of a bin and a lib using meson-build-system
<jonsger>:)
<efraim>buenouanq: two things I've found so far in my working toward GuixSD on aarch64 are that our grub packages need an overhaul and (gnu services xorg) uses xf86-video-intel, which is intel only
<efraim>i'm sure there were more, but those are the two I can first think of
<buenouanq>efraim: https://lists.gnu.org/archive/html/guix-devel/2016-07/msg00455.html
<civodul>Hello Guix!
<sneek>Welcome back civodul, you have 1 message.
<sneek>civodul, oriansj says: that the link provided earlier is about checking existing packages rather than package/system definition files
<jonsger>is it normal that "guix --version" doesn't show any version number/git commit hash?
<civodul>jonsger: no, it looks like a bug
<civodul>do you have ~/.config/guix/latest ?
<jonsger>no
<efraim>I don't either on my newly rebuilt aarch64 build machine
<efraim>i've never run guix pull on that machine
<civodul>jonsger: what does "readlink -f $(type -P guix)" return?
<pmikkelsen>Since meson is going to me the "default" for many gnome packages, should it have the same extra phases as glib-or-gtk-build-system, which can be enabled with a keyword argument? or should we have yet another build system called glib-or-gtk-meson-build-system? Ideas welcome :)
<civodul>glib-or-gtk-meson-build-system-with-python-bits-and-vala-as-well
<pmikkelsen>haha ;)
<civodul>those things don't compose well
<pmikkelsen>no they don't
<civodul>we should do something about it at some point
<civodul>but for now, you're right: it may be best to add the glib-or-gtk phases to meson-build-system
<civodul>which GNOME packages use Meson currently?
<pmikkelsen>allright, I will do that
<pmikkelsen>For example i just tried to build nautilus 3.26, and it uses meson, but i am not sure it is the only option
<pmikkelsen>But i think i heard that in the future, more will switch as they think meson is faster than autotools or something
<efraim>Adding it in to be removed as necessary is easier for package creation than adding in specific phases, like adding emacs phases to some of our packages
<buenouanq>in the locale definiton, is it .UTF-8 or .utf8 or will either work?
<civodul>buenouanq: ".utf8" is preferred (it's the canonical name as mentioned in the manual)
<buenouanq>for some reason the other had slipped into my configs and I've no idea why or where it would have come from
<jonsger>civodul: /gnu/store/vir3lrwqy50pr8fkaf3m091dgbrja2n6-guix-0.13.0/bin/guix
<jherrlin>hi, just started out running GuixSD and have some questions
<jlicht>hello jherrlin
<jherrlin>how can i disable the PC speaker beep in my config.scm?
<jherrlin>is that possible?
<jlicht>jherrlin: of course :-). If I am not mistaken, this is due to some kernel module being loaded
<jlicht>jherrlin: so you would want to blacklist the pcspkr and snd_pscp modules (if I am not mistaken).
<nee``>jherrlin: I don't think there is a setting for it yet. There are a few solutions here: https://wiki.archlinux.org/index.php/PC_speaker
<jlicht>see the `kernel-arguments' in https://www.gnu.org/software/guix/manual/html_node/operating_002dsystem-Reference.html
<jlicht>jherrlin: how I used it some months ago: http://paste.lisp.org/display/355720
<civodul>"modprobe.blacklist" is documented at https://www.gnu.org/software/guix/manual/html_node/Initial-RAM-Disk.html
<jherrlin>thanks you very much! i'll go with jlicht solution
<jherrlin>will reconfigure and reboot, be right back
<jlicht>civodul: interesting. What is the practical difference with using `kernel-arguments'?
<civodul>jlicht: oh you do have to use 'kernel-arguments' like you showed; the page above just explains this particular "modprobe.blacklist" kernel argument
<civodul>so yeah, the solution that jlicht showed is the right one
<civodul>i was just trying to complement it, not to confuse people ;-)
<jherrlin>thank for the help! :)
<jlicht>jherrlin: yw
<jherrlin>i added: http://paste.lisp.org/display/355721 to my consig.scm, reconfigured, rebooted and the pc speaker is no more :)
<jherrlin>i have another problem, this is my current config:
<jherrlin> http://paste.lisp.org/display/355729
<jherrlin>i would like to get the 00-keyboard.conf config into my running X session
<jherrlin>so i can use TAB as CTRL and change keyboard layout
<civodul>dustyweb: recently i watched https://www.youtube.com/watch?v=OyfBQmvr2Hc which you recommended a while back
<civodul>so, what was i saying
<Apteryx>what happened? :)
<civodul>dunno, network connection lost
<civodul>dustyweb: i had seen the program-by-example demo at the Scheme Workshop last year
<civodul>i think we should do GuixSD config by example
<civodul>somehow :-)
<civodul>that would be really awesome
<civodul>why do i lose connectivity in the middle of such a great idea? :-)
<Apteryx>eh! I have no idea what programming-by-exampe looks like, but examples are usually very useful to validate our understanding of the docs.
<dustyweb>civodul: yes re: guix config by example!
<efraim>civodul: I saw the new evaluation had 4000+ new packages and I thought it was from my gcc changes. Then I saw it is for staging
<jonsger>ACTION having fun with go and bundler :P
<jonsger>ACTION DDoSes mirror.hydra.gnu.org. At least his terminal looks like that :P
<Apteryx>what to do with coreutils such as 'find'? Do these need to be patched in packages using it?
<Apteryx>s/coreutils/findutils
<Apteryx>but the question remains the same
<Apteryx>OK, I found an answer from Mark from 2013 that says the implicit inputs are defined in (gnu base system), as %final-inputs. Those are (tar, gzip, bzip2, xz, diffutils, patch, coreutils, sed, grep, findutils, gawk, make, bash, ld-wrapper, binutils, gcc, libc).
<bavier`>I'm surprised no one on the HN Caddy thread has brought up the issue of corresponding source w/r to the EULA on binaries
<janneke>a system reconfigure after update to master rebuilds gcc and guile...
<janneke>iwbn to first update to a fairly recent commit that does *not* require such heavy local recompiles...what am y doing wrong?
<lfam>janneke: Can you stop and restart with '--dry-run' and share the results?
<janneke>lfam: i can..but already at BOOTSTRAP GUILEC language/tree-il/primitives.go
<janneke>is stopping wise?
<lfam>It's up to you
<lfam>If it's only been going for a few minutes, you won't lose much time. Those bootstraps steps take a loooong time
<efraim>could it be dd4efefdb0dc3240cd394a79f648cd9b171e83c0 ?
<janneke>i cannot oversee the consequences...losing an hour vs ...?
<lfam>janneke: We will have to revert the change anyways, so even if you finish building, the result of the build may not be used
<janneke>but i do want to help others even if it costs some time... i do need a reconfigured system tonight and failed to mark my previous master
<janneke>lfam: OK, thanks
<lfam>efraim: It could be, since gcc-5 inherits from gcc-4.9. So, if the derivation of 4.9 changed, it would also change 5
<janneke>ACTION runs --dry-run
<efraim>also gcc-4.9 is used to build gcc-5 in the bootstrap
<lfam>I'm going to bisect with --dry-run now
<happy_gnu[m]>Hello \\o/
<janneke>lfam: http://paste.lisp.org/display/355764
<janneke>i already built gcc-5.4.0 and quite some other stuff
<lfam>I love git bisect
<janneke>yeah...and irc :-D
<janneke>efraim: reconfiguring on dd4efefdb0dc3240cd394a79f648cd9b171e83c0~ --dry-run gives me no package rebuilds, only .drv's
<lfam>Our Git repo seems to be intermittently inacessible :/
<happy_gnu[m]>Searx is not part of Guix
<happy_gnu[m]>I would like to package it
<happy_gnu[m]>It has python dependencies
<lfam>janneke: How long has it been since your last update?
<lfam>Approximately
<lfam>Hi happy_gnu[m]: That would be cool :) We have lots of Python things already packaged and it's usually easy to add more
<happy_gnu[m]>I've been reading Packaging Guidelines
<janneke>lfam: lrwxrwxrwx 1 root root 33 Aug 8 21:08 system -> /var/guix/profiles/system-51-link
<lfam>janneke, efraim: janneke's dry-run report looks okay to me. It plans to download most of the low-level things like coreutils etc, which means this isn't a recent rebuild-the-world
<lfam>Ah, janneke, I think you are getting the results of the recent core-updates merge, but for some reason your Guix is trying to build instead of download things
<janneke>with efraim's guess now I only rebuilded ImageMagick, at least that helps my reconfigure
<lfam>That would explain why your dry-run report wants to download coreutils 8.27, for example
<lfam>But if something goes wrong with the network, Guix mirror, etc, you'd build instead if you passed --fallback. Did you use --fallback?
<janneke>possibly I should learn to first do a reconfigure *without* a git update, and only then try upgrading git
<janneke>lfam: no, i didn't use --fallback
<happy_gnu[m]>Basically is just writing a file with dependencies and another with metadata?
<lfam>happy_gnu[m]: In most cases, yes. Sometimes you have to do a little extra but we can help you with that
<lfam>janneke: What is your `guix version`?
<happy_gnu[m]>lfam: OK I will do that today
<happy_gnu[m]>But I was thinking on using a fresh install of guix
<happy_gnu[m]>Do you think that's better?
<lfam>What do you mean?
<janneke>lfam: what do you mean? i switched to latest git just now, then switched back to dd4efefdb0dc3240cd394a79f648cd9b171e83c0~ ... but you mean something else?
<lfam>Ah, I didn't realize you were working from Git
<janneke>oh sorry, yes i do
<janneke>ACTION just "bought" a puri.sm phone, let's make sure it runs GuixSD
<jherrlin>nice!
<jherrlin>would be awesome if the Librem 5 project get the foundings
<jherrlin>i have a question regarding time in GuixSD, i have =(timezone "Europe/Stockholm")= and =(services %desktop-services)= in my config.scm
<janneke>jherrlin: never got my ubutu edge :-( but this project's scope is much bigger
<jherrlin>as i understands it, %desktop-services contains the ntpd application
<lfam>janneke: Upgrading my long list of installed packages from the latest Git revision, I don't see Guile or GCC (or similar low-level things) being rebuilt
<jherrlin>and it runs on my system, if i run =herd status ntpd=
<jherrlin>but my time is not correct
<janneke>lfam...so possibly i hit network troubles?
<jherrlin>janneke , okey never heard of that project
<jherrlin>was it in the the ubuntu phones collection?
<lfam>janneke: Yes, or whichever mirror you queried returned 404 for some reason. The main Guix mirror will return 404 if it doesn't have the substitute cached, and then it will request it from the upstream server. This 404 is cached for a few minutes. You could try again and see if it starts downloading instead of building
<janneke>jherrlin: yes, some similarities with librem 5: no android but gnu/linux, add monitor and have your pc
<lfam>Also, https://hydra.gnu.org can't handle many users. That's why we cache it via the mirror
<janneke>lfam: thanks...
<efraim>`sudo ntpd -g -q'
<janneke>I'm using (only) mirror.guixsd.org atm
<efraim>there's also maruOS for AOSP + debian
<janneke>usually that's fine
<lfam>jherrlin: What efraim said. You can set the '-g' permanently as documented: https://www.gnu.org/software/guix/manual/html_node/Networking-Services.html#Networking-Services
<happy_gnu[m]>So if I have the lucky of macking package work what do I do next?
<lfam>jherrlin: It's the 'allow-large-adjustment?' option
<efraim>-q is 'once and quit', -g is 'oh god my time is really off! fix it now!'
<lfam>happy_gnu[m]: You will need to set up a Guix development environment, which is documented in the manual under Contributing: https://www.gnu.org/software/guix/manual/html_node/Contributing.html#Contributing
<jherrlin>efraim, it complains about a config file */etc/ntp.conf*
<efraim>boo
<jherrlin>does it needs some ntp servers?
<lfam>jherrlin: Do `ps aux | grep ntp` to learn the name of the configuration file
<lfam>Then, pass that file to the command
<efraim>hmm, our ntp package doesn't have an $out/etc directory
<jherrlin>lfam 'allow-large-adjustments?' is not set. I dont know how to set it, i think the =%desktop-services= contains all of the ntp stuff
<lfam>jherrlin: `ntpd -q -q -c /path/to/config-file
<lfam>`
<lfam>Sorry, `ntpd -q -g -c /path/to/file`
<jherrlin>can i do this from my config.scm?
<lfam>Do you know if this computer has a "real-time clock" with a battery, so it can retain the approximate time while powered off? If not, you'll want to set allow-large-adjustments? so that you don't have to manually adjust the clock every time you boot
<lfam>Yes
<jherrlin>the computer have a battery, i am 95% sure on that. (it's called a cmos battery and the cmos stuff that holds the time?)
<lfam>Does anyone have a snippet that modifies the ntp-service within %desktop-services they can share?
<lfam>Okay, then you shouldn't need to keep 'allow-large-adjustments?' all the time
<jherrlin>if someone have it would be awesome!
<efraim>didn't lfam push a change for the guix ntp pool?
<efraim>also, maybe we should patch openntpd to use those same guix ntp servers
<lfam>Yes, all the various services that use the NTP protocol need to use those NTP pool names
<lfam>I don't see a service for openntpd, so we don't need to do anything for it
<efraim>there's no service, but there is an etc/ntpd.conf in %out
<lfam>The pool name is about device vendors and operating systems playing nice with the pool. By using a specific DNS name, the pool can effectively handle misbehaving clients.
<lfam>I don't think we should bake GuixSD stuff into the packages. Once there is a service we should do it there. Also, the number of people using openntpd on GuixSD without a service is probably too small to affect the pool
<lfam>Like, I might use openntpd from Guix on my Debian systems. They shouldn't use the GuixSD pool
<lfam>The pool "vendor zone" guidelines say that individual users can use whatever NTP URLs they want. It's only people shipping running NTP clients that are supposed to use the zones
<efraim>using those guidelines, we don't ship an ntp client (service) using openntpd
<lfam>jherrlin: It will be something like this snippet (untested!): http://paste.lisp.org/+7MIE
<lfam>But, replace %base-services with %desktop-services, and also replace anything else that is wrong ;)
<efraim>i almost have gnucobol packaged, just for fun
<jherrlin>awesome, thanks lfam! :)
<efraim>thats what happens when they send an email to the info-gnu mailing list
<lfam>jherrlin: There is a section in the manual on modify-services. It could be helpful
<lfam>efraim: info-gnu ML -> guix importer
<efraim>yep
<efraim>thats also how we got motti
<jherrlin>thanks, i am slowly working my way through the docs
<efraim>except for libffi, I get personal emails for those
<efraim>s/libffi/libffcall
<lfam> http://paste.lisp.org/+7MIF
<lfam>hash mismatch of the libgit2 source code
<lfam>For an older version than what is currently packaged
<lfam>Could those github snapshots be generated nondeterministically?
<lfam>That would be really disappointing
<efraim>I found a typo in guix/packages.scm, it wanted to rebuild gcc
<civodul>lfam: weird, were you able to diff the two tarballs?
<lfam>civodul: I lost the old tarball to gc (that's how I noticed the issue)
<lfam>I just sent a question to github support, asking if these archives are expected to be generated deterministically
<civodul>lfam: did you try the /file URL at mirror.hydra.gnu.org?
<civodul>it may have it
<civodul>lfam: it works: https://mirror.hydra.gnu.org/file/libgit2-0.25.1.tar.gz/sha256/1cdwcw38frc1wf28x5ppddazv9hywc718j92f3xa3ybzzycyds3s
<lfam>Great, I'm using diffoscope
<civodul>drum rolls...
<lfam>Diffoscope is slow. Your arms might get tired
<civodul>:-)
<civodul>i often use "diff -r" in fact
<civodul>that avoids comparing mtimes, for instance
<bavier`>I've often thought it'd be nice for diffoscope to give a way to disable certain checks
<civodul>yeah, same here
<lfam> https://groups.google.com/forum/#!topic/linux.debian.bugs.dist/AiFlsKCM0Kg
<bavier`>lfam: :)
<lfam>Also https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=814057
<lfam>So... I guess it's my fault that diffoscope doesn't have this feature
<lfam>I never opened the new bug report with the test case he asked for
<bavier`>my last interaction with diffoscope must have been a while back then, if version 77 stopped comparing ctimes and mtimes. :/
<buenouanq>so, https://wiki.archlinux.org/index.php/SLiM#Change_Keyboard_Layout is how I've been changing the layout of the login manager - Is this the best way?
<buenouanq>I put that in a file that's called by my os config.
<lfam>Good question buenouanq, but I don't know the answer. It's something we should document, however, since being able to use the login manager is pretty important
<lfam>Ideally that screen would have a little "national flag" icon that you could choose from
<buenouanq>well, what I do certainly works
<buenouanq>I'm just wondering if there's a more guixy way to go about it.
<buenouanq>(guixie?)
<lfam>I like 'guixotic' ;)
<buenouanq>guixonian
<lfam>Heh
<civodul>which reminds me i haven't yet tried GDM after wingo's work
<lfam>Good news civodul. I made the Julia build tools do their thing without the .git repos
<lfam>They complain a bit, but I think it's okay ;)
<civodul>sounds good!
<civodul>we can tolerate them complaining a bit
<lfam>Hopefully it doesn't become a problem later
<buenouanq> https://rectilinear.xyz/p/175a8fe83e here's how I do it if you want to save this to help others or whatever
<buenouanq>do I have to include --no-bootloader every system reconfigure?
<buenouanq>or is that only a system init thing?
<marusich>The manual makes it sound like it's used for both the "init" and "reconfigure" sub-commands. See: (guix) Invoking guix system
<marusich>Probably, you need to specify it or not specify it every time, according to your needs.
***lx0 is now known as lxo