IRC channel logs


back to list of logs

<nckx>vagrantc: I'm going to bed, I'll read the log tomorrow. Good luck!
<vagrantc>yup, that patch seems to work
<vagrantc>probably fixes a few other variants as well
<vagrantc>well, i guess linux-libre-arm-omap2plus
<vagrantc>thought there were more that might be affected
<vagrantc>well, the version-specific variants of linux-libre-arm*
<vagrantc>not sure the arm-omap2plus needs to continue to exist...
*vagrantc just pushes
***tune is now known as bdju
*janneke -> zZzz
<reepca>odd question: how does one read the posix 2001 specification? I can find 2008 and later on the opengroup site, but nothing earlier.
<Minall>Hello guix!
<Minall>Hello guix!
***catonano_ is now known as catonano
<apteryx>question: why do we need --root in the 'boot-system' procedure (gnu build linux-boot), to specify the root mount devices, if we are already passing a list of <file-system> records as its MOUNTS argument? (which, I believe, should include the root file-system record). It seems redundant.
<apteryx>I guess it's nice that you can try booting a system from multiple devices at the Grub command line
***sneek_ is now known as sneek
<vagrantc>apteryx: my wild guess would be an ordering issue regarding what to mount first?
***freedom is now known as gnufr33d0m
<g_bor>hello guix!
<g_bor>do we have a one section per page version of the manual?
<g_bor>never mind, I've found it
***catonano_ is now known as catonano
*apteryx hacks the initrd some more
<apteryx>sneek: later tell vagrant we're already treating the file-system with mount point '/' differently, so that's not a valid reason.
<sneek>Will do.
<apteryx>Anyway, I've made the init use any root <file-system> record as a fall-back to --root
<apteryx>so eventually we can phase out using --root except where an override is needed.
<shrdlu68>Hello, I've looked at the code and it seems virtio drivers are built-in by default. However, the system can't boot after a successful `guix system init`.
<shrdlu68>It also seems I can't do `guix system init /etc/config.scm /` in a fresh CD/DVD install of GUIX. I was trying to reproduce the problem with virtio above in a QEMU VM.
<civodul>Hello Guix!
<roptat>hi guix!
<janneke>hi guix!
<g_bor>hello guix!
<g_bor>Anyone knows what timezone the blog post dates should be in?
<truby>if I build with a gcc from guix I get warnings from glibc headers, I guess they aren't being marked correctly as system headers? It causes our code to fail to build because we build with -Werror :(
<civodul>g_bor: it's unspecified, but it happens to be CEST or CET most of the time :-)
<civodul>truby: yeah, that's on core-updates, right?
<truby>it is yeah
***roptat_ is now known as roptat
<truby>civodul: so, if I go back to master I shouldn't have this problem?
<civodul>indeed, but core-updates will be merged in master Real Soon Now
<civodul>we're aware of the problem you mention with system headers, but unfortunately it's hard to fix with newer versions of GCC
<truby>is there a planned fix?
<civodul>no, not yet
<civodul>i could dig the references if you want
<truby>quite a lot of things build with -Werror in my experience so this might cause a decent amount of breakage
<truby>can't we just set `C_INCLUDE_PATH` and `CPLUS_INCLUDE_PATH` as environment variables? instead of `CPATH`?
<civodul>-Werror is often problematic independent of this issue, but yeah
<civodul>the root cause is at <>
<civodul>specifically, that's the reason why we have to use CPATH instead of C_INCLUDE_PATH & co.
<civodul>for GCC > 5
<truby>ah, ok. That's a nasty change in behaviour :)
<bricewge>Hello Guix!
<bricewge>How can I access the kernel config of my running system?
<truby>how does nix fix this? they must have had the same problem..
<bricewge>`/proc/config.gz` isn't there and `modprobe configs` fail.
<truby>it looks like it might be fixable by also setting CPLUS_INCLUDE_PATH to contain the libstdc++ header directory, before the glibc header directory? seems nasty though
<nckx>bricewge: /run/current-system/kernel/.config
<nckx>Actually, to be moar correct: s,current,booted,
<bricewge>nckx: Thank you :)
<nckx>All the goodness without having to modprobe something into kernel memory as root.
<civodul>truby: dunno, worth a try!
<bricewge>nckx: I can't find a reference to it in the manual. It should be written somewhere to avoid tribal knowledge.
<nckx>Sure, no opinion.
<nckx>It's only for humans anyway. Proggies have no legitimate reason to parse kernel configuration.
<bricewge>I'm willing to write it it, but I'm not sure in which section of the manual to put it.
<nckx>‘Kernel configuration’, which doesn't exist, because everyone seems to agree that the current MAKE-LINUX-LIBRE is suboptimal and should be replaced with… something. Or maybe that it doesn't belong in the manual at all. It's currently in doc/guix-cookbook.texi on the wip-cookbook branch.
<nckx>The manual focuses on processes: installing Guix System and configuring it declaratively. There's no ‘tour of your System as it actually exists on disc’ section AFAICS.
<bricewge>If it's in the cookbook it's good engouth for me then.
<nckx>s/processes/initial installation and Scheme abstractions/
<nckx>bricewge: Great 🙂
<truby>civodul: hmm. Simply setting CPLUS_INCLUDE_PATH and C_INCLUDE_PATH to $CPATH and then unsetting CPATH does work for me. As in, without also fiddling with the libstdc++ directories
<truby>I can't get the bad behaviour to appear when doing that. Probably I'm not hitting the specific case somehow?
<shrdlu68>Reposting: I've looked at the code and it seems virtio drivers are built-in by default. However, the system can't boot after a successful `guix system init`.
<shrdlu68>`lspci -k` shows virtio_pci, i2c_piix4, and ata_piix.
<shrdlu68>I'm trying to install Guix System from an Ubuntu live disk, which is the only available method to me in this scenario.
<shrdlu68>Since /gnu gets filled up, I move it to the disk and remount it as `mount --rbind /mnt/gnu /gnu`. I do the same with /var/guix.
<shrdlu68>That all seems to work, and everything installs and it asks me to reboot, but it can't seem to detect the disk upon booting.
<roptat>shrdlu68, do you get to the guile console? or not even grub works?
<shrdlu68>roptat: Grub works, and then it says it can't find `/dev/sda1`, or that it's waiting for a partition with uuid <uuid>, etc. Then I'm dropped into the guile console.
<roptat>ok, so you may be missing a kernel module
<roptat>ata_piix looks related to a disk thing
<roptat>but the required module is probably built in on ubuntu, so it's going to be hard to find out
<agohoth>hello what kernel si guix 5.2 libre?
<agohoth>and does it have icewm package?
<agohoth>and chrome?
<agohoth>I am thinking of making the switch
<agohoth>on my beloved desktop
<roptat>agohoth, kernel linux-libre 5.2, according to the name ;)
<agohoth>how how about icewm?
<roptat>no icewm it seems
<roptat>there's chromium
<agohoth>how would I find out what guix has?
<agohoth>where are you searching
<agohoth>how about jwm thats almost asnice as icewm
<roptat>from the command line, "guix search icewm"
<roptat>no jwm either
<roptat>you may want to have a try at packaging them
<roptat>also, you can install guix as a separate package manager on any distro if you want to give it a try
<roptat>it's probably not too hard to package, but you'll need guix to work on that
<agohoth>what lightweight wm do they have?
<pinoaffe>agohoth: I don't know what you consider lightweight, but i3, sway, herbstluftwm and dwm are all packaged
<shrdlu68>awesome, too.
<shrdlu68>Alright, I've added `(append (list "virtio-pci" "virtio-blk") %base-initrd-modules)`, hope this works.
<roptat>agohoth, openbox?
<shrdlu68>After `guix system init`, how an I inspect the kernel's .config that's been generated?
<shrdlu68>Curses, now it can't find `/gnu/store...linux-modules/virtio-pci.ko`
<civodul>truby: we'd have to review past discussions on this topic with mbakke, who handled it
<civodul>but i'm pretty sure C_INCLUDE_PATH is no longer an option with GCC > 5
<nckx>shrdlu68: Try virtio_pci just in case.
<nckx>I think we haven't fully worked around Linux's inconsistent module naming yet.
<pinoaffe>how are programs that have plugin systems dealt with in guix? how would one package plugins?
<nckx>kmod hacks around it, but last I knew Guix doesn't use kmod to load modules.
<shrdlu68>nckx: Ah, yeah. That's probably it. But it shouldn't have built if there's no such module.
<shrdlu68>I'm close to giving up at this point, I just don't know how to introspect the system.
<nckx>By inconsistency I didn't mean that some modules use _ and others -. Some use both depending on how you're asking. I think. Used to, anyway.
***MinceR_ is now known as MinceR
<roptat>shrdlu68, actually, guix is happy with virtio-pci, it installs the file, but at load time, it tries to load virtio-pic.ko which doesn't exist (probably, you have virtio_pci.ko installed though)
<roptat>shrdlu68, "guix system build config.scm" builds the system, but doesn't install it. it returns a store path in which you should be able to inspect the content
<roptat>I remember using the guile repl from a broken system to find and load modules
<roptat>something like ",use (gnu build linux-modules)" and then "(load-linux-module* "/path/to/module.ko")"
<nckx>Heh. Who doesn't.
<roptat>although it's terrible and I had to look at the sources to find that :/
<nckx>(mount …)'s nice that way too. Just different enough from what you'd type in a shell that I have to look at the source every once in a while to get it right. Or try random permutations.
<nckx>shrdlu68: You're not wrong, early Guix debuggability leaves much to be desired.
<nckx>Most, even.
<roptat>once you have it installed, it's almost impossible to break (it took me a disk failure to break the system :p), but when you have trouble installing it, the experience is not nice :/
<agohoth>any wm NOT tiling and NOT gnome or kde?
<roptat>openbox, lxde, xfce
<nckx>twm 🙂
<nckx>evilwm (now I'm just grepping (gnu packages wm)).
<shrdlu68>I hope stumpwm is there :)
<nckx>shrdlu68: It is.
<nckx>…in (gnu packages lisp) 😛
<shrdlu68>Nice. I'm thinking Emacs+Stumpwm+Guix will make a nice environment.
<agohoth>archlinux and freebsd are best ive tried
<agohoth>how is guix better?
<roptat>I won't say it can't break, but it's trivial to recover
<roptat>as long as you have one correct generation and you don't remove it
<roptat>it provides reproducible and bootstrappable builds
<roptat>binary transparency, an API to directly work with package definitions
<roptat>or with the system, or service definitions
<roptat>and well, os declaration is really different from what you will find anywhere else
<roptat>there's just one file you need to copy from computer to computer to reproduce the same system configuration
<agohoth>devops is a poorly executed scam
<agohoth>apps are about database programming
<agohoth>not algortihms
<apteryx>nckx: what make me not want to use the Guile repl is the lack of readline; yeah, I'm lazy.
<roptat>not sure what you mean?
<agohoth>project maangers and program managers are usless as is scrum and agile
<agohoth>and I dont even liek kbernetes
<agohoth>just have ha proxy load balancer and boxes running arhclinux
<apteryx>and sadly it can't be fixed until Guile modules can be compiled in statically (IIRC).
<agohoth>I heard guile had a web frameowrk wwith guile 2
<agohoth>glitter or something
<roptat>artanis maybe?
<apteryx>nckx: I should specify, the initrd REPL
<bricewge>Is there a way to have a specific module loaded before building a package?
<bricewge>I have a package with a test that fail if a module isn't loaded.
<roptat>you mean a kernel module? I don't think so
<nckx>apteryx: Oh, no, it's totally horrible.
<bricewge>roptat: Yes, a kernel module.
<roptat>bricewge, so no I don't think it's possible. can you skip that test?
<bricewge>Thanks roptat, I'll skip it.
<bricewge>I should have done that first and not loosing hours trying too find the source of the issue...
<roptat>but now you can add a comment saying why you need to skip the test :)
<joshuaBPMan>morning guix!
<joshuaBPMan>I'm guessing that the guile-lint package is too old to properly support newer guile versions...
<joshuaBPMan>It was designed for guile 1.6.4....
<apteryx>joshuaBPMan: yeah, I researched a bit about guile-lint once, only to find out most checks had been incorporated into Guile proper
<apteryx>although I couldn't make such check working -- there's supposed to be one telling you if you have unused top level definitions
<joshuaBPMan>apteryx: oh. That's kind of cool. So does emacs support some sort of linting on guile code?
<apteryx>joshuaBPMan: If I recall, the checks would occur at compilation time. I don't know of an Emacs package readily making use of such checks to lint Guile sources.
<joshuaBPMan>apteryx: gotcha.
<joshuaBPMan>I remember once configuring emacs to linting some php files. It was super awesome! It would make suggestions to make split up large functions into smaller ones, etc.
<joshuaBPMan>I wish I could do the same for guile
<apteryx>joshuaBPMan: yes, I'm used to this in Python too; it's quite useful.
<apteryx>although with Guile you have the REPL, so you can often just try out stuff and figure out if it works or not
<joshuaBPMan>also, does does emacs suppert an edbug for guile code? I don't actually see how to step through guile code...
<joshuaBPMan>apteryx: true
<apteryx>(there's a REPL with Python, but it's not as neatly integrated compared to Emacs+Geiser+Guile
<joshuaBPMan>agohoth: that link isn't working
<apteryx>civodul: still reworking my btrfs submodule patches, following your comments
<apteryx>joshuaBPMan: perhaps try 'guild lint your-guile-scheme-source-file.scm'
<joshuaBPMan>apteryx: thanks.
<apteryx>it's quite primitive for now. You can read the header of the (scripts lint) source to learn about what it does.
<joshuaBPMan>I'll give that a try.
<joshuaBPMan>hmmm, that's weird. guix lint file.scm just gave me a backtrace on a program of guile that compiles and runs normally.
<apteryx>joshuaBPMan: sorry if I mislead you, it's *guild* to lint random program; 'guix lint' is to lint a specific guix package, such as in 'guix lint emacs'
<apteryx>'guix lint' is really about linting guix packages, not Guile source code.
<quiliro>Minall: Saluton!
<quiliro>Saluton Giksujo!
<joshuaBPMan>apteryx: I actually ran guild lint web.scm. That's where I'm getting a backtrace...
<apteryx>joshuaBPMan: on my system it outputs this:
<apteryx>it seems to be confused about Guix records.
<joshuaBPMan>apteryx: yup. I get the same output.
<roptat>saluton quiliro
<joshuaBPMan>I have created my own file web.scm...It just a scheme file that runs a website.
<joshuaBPMan>locally running website. Not a site that you could point your browser to.
<joshuaBPMan>I suppose that would be an error. I guess I could file a bug.
<apteryx>I see! Yes, erroring out on a valid files doesn't sound right.
<quiliro>roptat: Mi volas konstrui la servojn de Freedombox-o en Gikjo. Kiel mi faros tion?
<roptat>vi devas krei pakaĵojn
<roptat>poste, vi kreos servicojn
<quiliro>roptat: dankon al vi!
<Minall`>Hello guix!
<Minall`>After doing a guix pull and tried to do a reconfigure, I had to shutdown my machine, my question is... Do I have to make another guix pull after reconfigure?
<quiliro>No you don't unless you want to upgrade the guix package (and the package list) fer the current user.
<apteryx>guix system reconfigure is always used locally, right? I mean, what does 'guix deploy' does (I haven't kept up to date with it, sorry) -- send the config.scm to the remote than reconfigure it there?
<apteryx>what would be your optnion on probbing a local file system during 'guix system reconfigure' (on the host side). Any problems in sight?
<apteryx>the issue is that Btrfs subvolume IDs must be translated to Btrfs sobvolume names so that GRUB can prepend the necessary submodules names on the paths of its kernel and initrd lines.
<apteryx>*to the paths
<civodul>apteryx: translation must happen on the deployment node i guess
<civodul>can it be done at activation time or as part of the 'file-system' service?
<Minall`>Hello guix!
<quiliro>Minall: Saluton
<Minall`>Hello guix!, any other information about the ssh-daemon not starting? the output that I have with 'herd status ssh-daemon' is:
<Minall`>ssh-daemon state:
<Minall`>It is stopped:.
<Minall`>It's activated.
<Minall`>Gets (ssh-daemon).
<Minall`>Requires (syslogd loopback).
<Minall`>Has a conflict with ().
<Minall`>It will be restarted.
<Minall`>But it isn't restarted, one has to start it manually
<civodul>Minall`: it's investigated at
<civodul>maybe you can help gather debugging info?
<Minall`>Thank you civodul, on it
<Minall`>Using gnome one can't see in the details section guix
<Minall`>In the 'SO' it shoul be Guix System or such
<Minall`>In the problem with ssh I see that Leo claims that the directory '/var/run/sshd' is not 'existent'
<Minall`>But I do have that folder
<Minall`>Without anything, of coures
<agohoth>damn im banned from postgresql
<agohoth>damn that toad
<Minall`>agohoth: what do you mean by gnome lol? jja
<roptat>Minall`, what is the details section?
<Minall`>In the details section it is the gnome version and all
<Minall`>The memory
<Minall`>The processor, graphics, and the name of the OS
<Minall`>And other things as the bits and the memory
<Minall`>In 'name of the OS' there's nothing
<agohoth>I enjoy iceWM to me gnome is an aircraft carrier
<roptat>I don't use gnome, so I don't really get what it is... could it be taking its info from /etc/os-release maybe?
<Minall`>agohoth: i didn't knew about icewm, I'll check it, it looks cool
<Minall`>roptat: If you want I can ask on #gnome
<agohoth>This whole job thing is outa framworks poluting linux and bsd cd agiel scrum barf
<Minall`>I don't have a '/etc/host-release'
<agohoth>is there no one with taste anymore?
<Minall`>Taste about desktop environments or what?
<roptat>so what if you add one in your config? or you could directly create the file for a quick test (although not in guix style :))
<Minall`>You mean create the file and add Guix to it?
<Minall`>And then restart gnome?... But what template should the os-release be like?
<roptat>yes, just create the file and restart gnome
<roptat>you can also add it to your config with an etc-service-type I think
<Minall`>Thanks, I'll try it
<Minall`>If we found the way we can add it so it does it automatically when installing gnome
<roptat>that's my reasoning, yes :)
<Minall`>Bad channel lol
<Minall`>I'm creating the file right now
<Minall`>Does guix has version names or is just 1.0.1?
<Minall`>What output can I put on 'ANSI_COLOR'?
<Minall`>I can't find to make orange color
<agohoth>did u do the gui install?
<agohoth>Im wondering if audio is setup by default
<Minall`>And on CPE_NAME?
<Minall`>Audio setup is by default agohoth , it uses pulseaudio
<Minall`>You can tweak it to use alsa if you want
<Minall`>One can add more details but, the only thing showed is 'Pretty-name'
<agohoth>like if I play youtube will I hear anything
<Minall`>Now it says 'Guix System'
<Minall`>agohoth: Yes, by default you will
<Minall`>roptat: How can I add it to the gnome default?
<Minall`>When installing it already creates the file
<Minall`>Just as a test of course, it would be better to use the other options, such as bug report, in order to make gnome more adapted to guix
<Minall`>roptat: I'm watching etc-service-type, thanks
***kristianpaul is now known as krispaul
***krispaul is now known as kristianpaul
<Minall>Hello guix!
<Minall>What is the native host for gnome on guix?
<Minall>Hello guix!
<Minall>How can I install gnome native host connector?
<Minall>The package is chrome-gnome-shel
<Minall>But I don't see any package like that on guix
<bavier>Minall: guix probably doesn't have it packaged
<bavier>Minall: do you have an upstream url you can share?
<Minall> git clone
<Minall>One can install it with cmake
<Minall>But one should make a package definition on guix right? since guix is different
<bavier>a local install is worth a shot
<Minall>What to you mean a local install bavier ?
<Minall>One can install things in guix without a package definition?
<Minall>That could be awesome in order to test it, and then make the package definition
<bavier>Minall: do the git clone, 'guix environemtn --ad-hoc python python-pygobject', maybe 'pip install requests', then run cmake
<Minall>THanks bavier, I'lk try it
<Minall>I'm trying to create an etc file, is this not the correct syntax?
<Minall>I'm getting: more than one 'etc' service
<Minall> (service etc-service-type
<Minall> (list `("os-release" ,(plain-file "os-release"
<Minall> "PRETTY_NAME=Guix-System"))))
<Minall>Actually, maybe I have to add this configuration along with gnome,
<Minall>I'll try ir
<grumbel>Anybody here using Guix on multiple monitors? I am currently trying to switch from the Ubuntu Xfce to the Guix Xfce, but when I maximize a window it gets stretched over all the monitors, instead of the current one
<grumbel>Feels like there is a libXrandr dependency not getting compiled in at some point
<bavier>grumbel: I use multiple monitors with guix's ratpoison, and things work fine. idk about xfce.
<Minall>bavier: Can you help me with my reconfigure problem, I'm trying to create an etc file
<agohoth>any light weight but non tiling madness wm?
<bavier>agohoth: I like evilvm
<bavier>agohoth: our xfce4wm package seems to have a reference to libxrandr
<bavier>sorry, for grumbel ^
<bavier>but maybe there's some configuration that doesn't get set properly by default
*bavier afk
<grumbel>problem might be in gtk+, as I am getting no match on "objdump -p /gnu/store/jr6kyk1qqhyb9wxxy7g0fp3k4yyi3rgn-gtk+-3.24.10/lib/ | grep -i xrandr", but trying same on the ubuntu libs gives a libXrandr reference
<anon987321>hi guix
<grumbel>Is there a 'apt-get build-dep ...' analogue in guix (i.e. fetch build dependency)?
<reepca>grumbel: see 'guix environment'.
<grumbel>How would one rebuild only a single package? Doing "guix build gtk+ --no-substitutes" seems to rebuild all the dependencies
<bavier>grumbel: do a 'guix environment gtk+' to fetch substitutes for gtk+'s dependencies, then do your 'guix build gtk+ --no-substitutes'
<bavier>though, the difference between a gtk+ built locally, and one gotten as a substitute will probably be minimal
<bavier>if at all
<grumbel>No luck, even after a 'guix environment gtk+', build still builds everything
<bavier>grumbel: what are you trying to do?
<grumbel>bavier: Trying to view the gtk+ build log and than maybe patching and rebuilding gtk+
<reepca>huh, yeah, that's weird. --no-substitutes seems to either cause repeat builds or changes the order of builds
<bavier>grumbel: 'guix build' has a '--log-file' option, which is maybe what you're looking for
<bavier>it has the unfortunate tendency of pointing to a log with "grafting ..." messages. In that case you probably want to include '--no-substitutes'
<bavier>but with that, my 'guix build --no-substitutes --log-file gtk+' seems to insist on building openssl. :/
<reepca>for example: 'guix build --dry-run --no-substitutes --check git' says The following derivation would be built: /gnu/store/na46y0k24d3pf50fvs70p9fmnzixvyxh-git-2.23.0.drv, but running it without --dry-run, it immediately starts building openssl
<bavier>oh, order of options seems to make a difference: "guix build --log-file --no-substitutes gtk+'
<mbakke>bavier: to get the actual build log, you need --no-grafts too
<mbakke>also, that would prevent pulling in openssl etc
<bavier>mbakke: ah, right :) to many things going at once
<grumbel>That however still leaves me with "guix build: error: no build log for '/gnu/store/f35d2zi13hvkwzxkmbb552nzll0pvz51-gtk+-3.24.10'", since I haven't build gtk+ on my machine I guess
<bavier>grumbel: it'll point you to our build farm's logs: "guix build --no-grafts --log-file gtk+" prints for me
<grumbel>ok, that works, had a --no-substitutes still in the command line
<bavier>so, our gtk+@2 has xrandr support, but apparantly not gtk+@3
<mbakke>woah, evaluations of 'master' have been broken since de3896f:
<mbakke>bavier: huh
<bavier>grumbel: do you want to check whether adding a libxrandr input to the gtk+ package fixes your multi-monitor setup with xfce?
<grumbel>bavier: yes, that's the goal
<bavier>grumbel: great, I've confirmed that the configure script detects enables support with the new input, but can't test xfce at the moment.
<bavier>and both the and libraries now reference
<grumbel>bavier: multi-monitor support is fixed by the libXrandr change
<bavier>grumbel: cool.
<bavier>grumbel: could you send a patch to
<grumbel>will do