IRC channel logs


back to list of logs

<mbakke>bricewge: your GPG key has expired
***catonano_ is now known as catonano
<pkill9>does anyone know a fix to this? error: ‘round’ is not a member of ‘std’
<pkill9>compiling C++
<alextee[m]>pkill9: are you compiling with a -std flag? it looks like that needs std=c++0x or above
<alextee[m]>(stackoverflow to the rescue)
<fvr>What's the simplest way to access the location of the definition for a package that guix search returns in the location: field.
<fvr>guix edit seems to be the way
<fvr>Are manifests the proper way to manage different development environments?
<raghavgururajan>Hello Guix!
<raghavgururajan>sneek, later ask g_bor[m]: You email address is your lastname at gmail dot com, correct?
<lispmacs>sneek tell sneek hello
<lispmacs>sneek tell sneek later goodbye
<sneek>Umm, I'm right here.
<lispmacs>sneek version
<sneek>Sneeky bot running on Guile version 3.0.0 using bobot++ release.2.3.1-6-a463-dirty
<lispmacs>sneek license
<lispmacs>sneek help
<lispmacs>sneek what is guix
<sneek>Someone once said guix is a functional package management tool for the GNU system
<lispmacs>sneek guix help macros
<lispmacs>sneek guile help macros
<sneek>Did not find any object named `macros'
<lispmacs>sneek guile help lambda
<sneek>No documentation found for:
<sneek>(guile): lambda
<reepca>sneek: later tell sneek: later tell sneek
<lispmacs>sneek who is sneek
<usney>Hi guys
<sneek>Its been said that sneek is a sneek sneek that sneek sneek sneek sneek sneek
<usney>guix pull is finally fast!
<usney>with my new upgrades
<usney>it seems my ram was faulty so it was slow
<usney>Or maybe the server is faster?
<usney>new ram and faster cpu made all the difference
<usney>How come the essential fonts package section has gotten smaller?
<usney>I could have sworn there was a bigger list.
<Kimapr>is there a good way to copy/move the store?
<Kimapr>i tried cp, but it fails with no space left on device, even though device is roughly same size as current one
<usney>you could use a disk usage analyzer to figure out which packages are taking up to much space and remove them for the ones you don't need and the system doesn't require
<Kimapr>no, thats not what i want. I want to move the store to another partition
<efraim>sneek: what is the cake
<efraim>sneek: botsnack?
<Kimapr>it looks like hardlinks aren't copied, so the store is copied inefficiently
<efraim>forgot to test screen unlock with enlightenment-0.24
*apteryx finally managed to boot an embedded board over NFS hosted on a Guix System
<nixfreak>I tried install icecat and guix keeps telling locales not found
<nixfreak>but when I look at the /etc/conifg.scm file I see locales set in the operation-system definition?
<nixfreak> guix pull
<nixfreak>guile: warning: failed to install locale
<nixfreak>I have en_US.utf8
<nixfreak>I install enlightment for a wm
<nixfreak> guix pull
<nixfreak>guile: warning: failed to install locale
<azvede>any recommendations on desktop env?
<nixfreak>I'm using enlightenment
<azvede>first time installing a *nix OS and have no context
<nixfreak>umm , I don't think guix is good for a new user (in my opinon)
<nixfreak>this is diffent then say ubuntu or fedora
<nixfreak>but if you want to try Gnome, or MATE is good
*efraim is popping in while at work
<azvede>I mean.... I've used prebuilt vms for a while... and an ubuntu laptop
<efraim>nixfreak: Yesterday I bumped enlightenment to 0.24 but the screen lock function doesn't work for unlocking, I'm testing a fix now
<azvede>just haven't really had to deal with different interface options... and yeah I've read it's a bit mor of an advanced distribution
<nixfreak>not even herd works though
<nixfreak> I just installed this instead of gentoo/funtoo
<nixfreak>locales need to be installed but I have the in the config file ?
<efraim>you need sudo herd to interact with the root shepherd
<fvr>nixfreak, If it's just a warning about the locales you can ignore it most of the time.
<nixfreak>oh by the way don't use e with tiling , it will crash everytime , also acpid isn't running by default
<nixfreak>I installed icecast too , and can't get that to run
<nixfreak>herd status
<nixfreak>error: connect: /run/user/1000/shepherd/socket: No such file or directory
<fvr>you can install glibc-locales if they're not available
<efraim>azvede: I haven't tried guix with gnome but I hear its working fairly well. I set my kids up with xfce though
<nixfreak>@fvr I did into that also
<nixfreak>and exported the path
<nixfreak>still getting locales errors
<fvr>nixfreak, herd should be used with sudo
<nixfreak>oh sorry
<fvr>azvede, I used gnome for a while but it crashed due to unrelated reasons. Also make sure your GPU is supported well
<efraim>we don't have an acpid service yet, I suppose you could ad-hoc one in the background with 'sudo $(guix build acpid)/sbin/acpid &'
<fvr>nixfreak, what's the locales error
<nixfreak>herd statusguix pull
<nixfreak>01:08 < nixfreak> guile: warning: failed to install locale
<fvr>I had that warning for a while, but it went away magically on its own
<nixfreak>why can't I get icecat to run
<efraim>ok, patch works, I'll push it in a minute
<nixfreak>do I need to sudo herd start acpid ?
<efraim>I meant the patch to fix screen unlocking on enlightenment
<nixfreak>but I understand that
<efraim>I'm going to try 'sudo $(guix build acpid)/sbin/acpid &' and see if that works until we have a real acpid service
<nixfreak>I just did
<nixfreak>I just did not found in the services
<reza[m]>Hi, does anyone knows how can I force guix to refrsh it's substitute list?
<azvede>is it normal to get a whole HEAP of hash mismatches when installing?
<nixfreak>I haven't see it yet
<efraim>reza[m]: you can clear the /var/guix/substitute/cache/ folder, that should get it to refresh
<cbaines>azvede, I don't think so, could you share soe of what you're seeing
<efraim>azvede: no, that shouldn't happen
<azvede>seems to just be tlsing the failed matches so I'm assuming it's not too bad
<azvede>sha hash mismatch for /gnu/store/[blah]
<azvede>then hashes, a notice that substitution failed, then attempt to dl new file
<nixfreak>I'm downloading econnman right now I don't see any mismatches
<nixfreak>maybe your time on the system?
<reza[m]>@efraim thanks, is there any service needs to be restarted after deleting the cache file? since in the server part `guix-publish` service needs to be restarted
<efraim>reza[m]: no, it should just start working better after that
<azvede>nixfreak... yeah, I suppose that could be....
<reza[m]>efraim: ok thanks
<nixfreak>any other browser besides icecat I could try ?
<cbaines>nixfreak, icecat isn't working for you right? What error are you getting?
<nixfreak>nothing I just can't even find it
<nixfreak>I installed it , doesn't come up in the enlightenment menu at all
<nixfreak>I installed kitty terminal also , and I get locales error when I try to run it
<cbaines>At least with Gnome, you have to restart the window manager to pick up new apps unfortunately. Maybe the same is true for enlightenment
<cbaines>If you try to run icecat from a terminal, does it work?
<cbaines>(and is this on Guix System?)
<nixfreak>yeah I installed a fresh copy of guix distro
<nixfreak>icecat: command not found
<cbaines>Hmm, does icecat show up in guix package -I
<efraim>it could be the whole first generation needs a re-login to start working well thing
<nixfreak>guix package -I
<nixfreak>guile: warning: failed to install locale
<nixfreak>terminology 1.6.0 out /gnu/store/yhx3ybq096jjmqiyafspa1gz8rrfmf5a-terminology-1.6.0
<nixfreak>kitty 0.16.0 out /gnu/store/4ql1ijgp9yvnbpfms57cyj4nspwxrcfl-kitty-0.16.0
<nixfreak>irssi 1.2.2 out /gnu/store/b14a1mjqm29r65vihxysq5a3iwhfx3h4-irssi-1.2.2
<efraim>failed to install locale you can ignore, or install glibc-utf8-locales
<nixfreak>is this the only mirror for installs ?
<nixfreak>also is python not installed by default ?
<efraim>very little is installed by default
<efraim> is the main build farm, there's a secondary one that's underpowered at, and which has a different signing key
<nixfreak>I was just curious
<nixfreak>ok I got icecast to work now , weird I had to install twice
<nixfreak>also lock works with enlightenment
<nixfreak>say is herd like systemd ? I'm trying to enable connman
<nixfreak>Does guix not support iwlwifi drivers for intel wireless
<nixfreak>do I need to build the wifi driver for the kernel ?
<cbaines>Linux-libre, which Guix uses doesn't include the proprietary firmware for iwlwifi
<nixfreak>ok so can I look at my kernel settings using make menuconfig ?
<nixfreak>I don't see a /usr/src
<cbaines>tweaking the Linux configuration won't magically make the firmware appear
<nixfreak>I know I have to rebuilt the kernel correct
<nixfreak>cp the firmware file to /lib/firmware which doesn't exist either , how do you get to the kernel on guix
<cbaines>You specify things about the kernel through the operating-system record
<nixfreak>ok what about this ?
<cbaines>what about that?
<nixfreak>nevermind I see it only has chromium
<nixfreak>I'm just trying to get wifi on this laptop
<civodul>Hello Guix!
***apteryx_ is now known as apteryx
<efraim>civodul: did I find a typo on guix/build/utils:803, 'letter' instead of 'letters'?
<civodul>efraim: yup, that's a typo!
<civodul>you can fix it in core-updates
<civodul>that typo has been here since 2012
<civodul>which means you're the first one to read it carefully in 8 years :-)
<efraim>not the first time I end up correcting typos in non-code :)
<efraim>I'll see if it can apply to master or cause a mass rebuild
<efraim>yep that's a mass rebuild
<efraim>I'm testing a fix for building tensorflow
<nixfreak>is the /etc/config.csm the only config for operating-systems definiations ?
<nixfreak>I don't see kernel or firmware in there ?
<cbaines>nixfreak, some of the operating-system record fields have defaults
<nixfreak>hmm so would I have to add kernel and firmware sections then ?
<cbaines>nixfreak, there's some reference documentation here
<civodul>efraim: it causes a mass rebuild, hence core-updates
<civodul>because "everything" depends on this file
<cbaines>If anyone is up for changing the NGinx configuration on berlin, I have a patch...
<civodul>cbaines: yay! can you push it to maintenance.git?
<cbaines>civodul, sure, I've now pushed
*civodul reconfigures berlin
<cbaines>I think I checked the NGinx config, and those changes looked OK. I forget if the NGinx config is checked as part of the reconfigure, but it might be worth having NGinx check the config before restarting the service.
<civodul>the config is checked when starting the service IIRC
<civodul>that is, "nginx -c" (?) is invoked at some time for syntax checks
<civodul>unfortunately not at reconfigure time
<rekado_>I’ve been rebuilding ghc-pandoc from GHC up multiple times now, and I still think it should be possible for us to build a smaller variant of pandoc.
<rekado_>Fedora appears to do a semi-static link of pandoc, because when installing “pandoc” it only installs the tool, not any of the libraries it depends on.
<rekado_>but the thing is still dynamically linked to libc and what not.
<rekado_>it looks like this would require us to actually build the static libraries, though
<rekado_>Fedora just moves them to the *-devel packages
<Kimapr>is there a way to umount the store, delete it and mount another (identical) one?
<Kimapr>(on a running Guix System)
<rekado_>in my experiments I wasn’t able to remove the references in the static libraries to the rest of the package, so having a separate output with just the static libs didn’t work for me.
<rekado_>this works on Fedora because their packages install files to the *same* directory
<rekado_>Kimapr: you can just bind-mount another store over /gnu
<rekado_>you don’t need to unmount and delete
<Kimapr>then where does the old one go
<Kimapr>and would guix automatically start to use it?
***rekado_ is now known as rekado
<civodul>rekado_: sounds tricky
<rekado>Kimapr: the old one just stays in place, but is no longer accessible via /gnu
<rekado>(it remains accessible through other means)
<Kimapr>question remains: does guix daemon drop the old one from its private namespace and start to use the new one?
<civodul>not sure, but what are you trying to achieve? :-)
<Kimapr>move the store to another partition
<Kimapr>already copied it, but want to delete old copy
<haha>I am going to use guix system container to create my buildroot development environment, is there any good way to keep the container's ssh key? (/etc/ssh/ssh_host_*)
<sturm1>I'm trying out `guix deploy`. Is the `host-key` a string with spaces between the three parts? eg. "ssh-ed25519 AAAAC3Nz..."
<Kimapr>i can boot up in another system, but htats tedious
<haha>I used guix environment before, but it doesn't look very powerful.
<sturm1>ah, figured it out, I'd missed the "ssh-" prefix
<haha>Ok, I shouldn't ask such a stupid question. It is easy to solve this problem by using the --share parameter.
<r-7>Some observations. The GUI font issue seems to have been partially fixed for me after the latest update - Emacs and Icecat no long have numbered squares in place of text in popups. And after running 'sudo fc-cache -rv', dmenu also has normal text. Some textual icons (e.g. folder symbol) are still absent from GUI popups, though, all replaced by a square icon with one corner folded in.
<mbakke>is anyone using gexps with shepherd user services?
<mbakke>i tried creating a service with a start command like (make-forkexec-constructor #~(list #$(file-append mcron "/bin/mcron")))
<mbakke>shepherd accepts it, says it starts the service, but the process is nowhere to be found
<efraim>I assume you added (guix gexps) to modules at the top of the file
<efraim>I haven't gotten gexps to work yet but I haven't really tried
<civodul>mbakke: should be #~(make-forkexec-constructor (list #$(file-append mcron "/bin/mcron")))
<civodul>or something along these lines
<nckx>Mornin' Guix.
<civodul>for sure, make-forkexec-constructor should be within the gexp, and there shouldn't be any nested gexp
<civodul>hey nckx!
<mbakke>civodul: right, then I get 'Wrong type to apply: "#<gexp (make-forkexec-constructor ...'
<civodul>i guess we'd need to see the whole code
<civodul>but otherwise you can start by pasting an existing service :-)
<mbakke>civodul: nothing fancy:
<mbakke>migrating my "monolithic" shepherd configuration to efraim's very nice blog post setup and figured I'd improve things in the process :P
<civodul>mbakke: ah, no gexps then!
<civodul>you're using the shepherd API directly in that case
<civodul>there's no Guix, no store, no gexps
<cbaines>civodul, how's reconfiguring berlin going?
<janneke>civodul: remind me, the signalfd story on the Hurd is covered?
<civodul>cbaines: i had forgotten about it :-)
<civodul>i've restarted nginx now
<civodul>lemme know!
<civodul>janneke: no, there's no signalfd there (yet), but the thing still works without it
<civodul>well, we'll double-check on the Hurd before releasing :-)
<cbaines>civodul, thanks, that looks to have worked. Here's the before and after
<janneke>i had missed the "should still work without" bit
<efraim>mbakke: yeah that's basically what I would've tried (
<civodul>cbaines: good
<efraim>doh, no icons on netsurf
<mbakke>civodul: did you also update berlin? I think that will fix
<jlicht>hey guix! Does anybody have a way to set php.ini files for the php-fpm service?
<rekado>I’m getting let’s encrypt warnings for the * certs
<cbaines>Haha! Looks like repology is actually fetching data now
<nckx>Yep, it's working. Projects using these things are going green:
<Kimapr>so is it possible to replace the local store with another one on another partition?
<civodul>mbakke: ah! i used the same channel commit as before
<civodul>should i pull?
<civodul>ah yes, for the fontconfig fix, right?
<jlicht>jonsger: quite some time ago, you mentioned doing *stuff* with php on guix. Do you still have some info on what you did?
<nixo_>Hi guix! Just saw the updates on repology. It seems like 1% of all our packages have known vulnerabilities
<nixo_>Also, only 65% of our packages is up to date (in comparison, nixpkgs-unstable has 80%)
<cbaines>At least relatively to other things tracked by Repology, 1% for "potentially vulnerable" is pretty good!
<cbaines>as for keeping Guix packages up to date, well that's an area that can always benefit from work
<nixo_>cbaines: my reference is always nixpkgs unstable, which has 0.38% with 4.1X our total packages
<haha>Hello, everyone, does anyone use buildroot on GuixSD? It seems that it conflicts with some environment variables set by Guix. Is there any better way to solve this problem?
<cbaines>nixo_, in terms of security, I'd make the claim that I percive Guix to be making a better effort in this area.
<cbaines>The "potentially vulnerable" number Repology reports is lower for Guix, 128 to 214
<cbaines>This shouldn't be a "us" vs "them" discussion though.
<PurpleSym>Is there anyone looking through CVE’s on a regular basis? CVE-2020-12762 has not been addressed as far as I see?
<nckx>jonsger: Super late, but evince seems to support PDF forms. At least it insisted on doing so when it wasn't supposed to.
<mbakke>civodul: updating would be great, yes :-)
<nckx>nixo_ 1% vs 0.38% is just noise, and repology doesn't even know about security patches or grafts. So it's a fun toy (and useful to get cheap update notification) but those numbers mean nothing at all. Even ‘up to date’ is iffy.
<mbakke>PurpleSym: thank you for the notice, can you file a bug report? we try to stay on top of CVE's, but don't have enough people working on it :-/
*nckx updates dia to ‘fix’ a CVE but Repology will never know.
<mbakke>perhaps Repology could be taught to parse the 'patches' field if we ask nicely
<PurpleSym>mbakke: Sure.
<nckx>mbakke: Absotootely. We have (or had) a CVE-foo.patch policy so…
*civodul views part of Repology as "guix {lint -c cve,refresh}"-as-a-service
<nckx>Still won't fix things like dia since it's not a patch, and R'y makes no attempt to understand non-straightforward version numbers.
<cbaines>I believe Repology is publically available free software, so it should be possible to poke around and have a look
<nckx>And that's fine but it is the distrowatch of packages.
<civodul>cbaines: but since we already have the tools, i'd rather see us improve them
<civodul>an example: how does Repology map package names to upstream/CPE names?
<civodul>that's kinda hard to guess, except perhaps if they have a mapping from source URLs
<cbaines>civodul, yeah, we should definately have some way of providing users with a view over CVEs/security issues, without them having to run the CVE lint checker on a regular basis
<civodul>that's an example where we should be able to do better, because we know that mapping
<civodul>cbaines: like "guix health"?
<cbaines>civodul, indeed, being able to work with a profile is good
<cbaines>and some of the provenance work done should enable tooling to look for security issues as well
<civodul>"guix health" could now map a store item to a live <package>
<civodul>regarding updates, i think we should have a bot that roughly does "guix refresh -u X" and "guix build $(guix refresh -l X)"
<cbaines>I'd really like a package update bot
<cbaines>although I've been thinking about is as something enabled by automated patch review
<cbaines>as then you could have the bot submit patches, and the testing of the patches would do the "guix build"
<cbaines>then someone can "adopt" the patch and merge
<PurpleSym>Is `guix lint` currently able to tell if we fixed a certain CVE? (i.e. CVE-*.patch exists)
<civodul>PurpleSym: yes
<civodul>cbaines: that'd be nice
<dragonker>hi, good morning. Am too far from guix servers
<dragonker>I want to set up a CDN for personal use
<dragonker>what do I need?
<cbaines>dragonker, what region of the world are you in?
<dragonker>south america <(
<dragonker>a guy in the mail lists says that he set up two cdn mirrors, when the main mirrors were on hydra
<dragonker>so I guess it's possible
<dragonker>though I don't know much at all about the topic
<cbaines>dragonker, Ok, I remember something about using rsync to setup mirrors using rsync, but I don't know the details
<cbaines>You could also setup a caching HTTP proxy, using something like NGinx
<dragonker>sounds like a cool project
<rekado>rsync daemon is not currently running, but if it were you could sync the “guix publish” cache via port 873
<cbaines>dragonker, do you know what's bad about your connection to Is it the throughput (the peak speed), or the latency?
<civodul>dragonker: see as an example of an nginx caching proxy configuration
<rekado>the reason why it’s not running is that the files created by “guix publish” have the wrong permissions.
<nckx>I don't think rsync works yet because the guix publish cache isn't {world,daemon}-readable or something. A caching nginx HTTP proxy works fine but won't do much for a single user with a single machine (unless you GC too much).
<civodul>rekado: they're 400?
*nckx sighs.
<rekado>today we will be running some more tests in the data centre to diagnose the low speed when connecting from the outside
<rekado>we will use one of the new build servers, new cables, new IP, new core switch port
<civodul>re 'guix publish', looks like we can easily fix it and chmod everything in /var/cache/guix/publish?
<dragonker>cbaines: i guess is the latency, peak speed it's totally inconsistent. Sometimes it's 100kbs, other times its 20kbs
<rekado>if none of this performs any better I’ll pass this on to the networking team who will have to get the switch vendor involved
<dragonker>In the same package install
<rekado>civodul: yes, the fix should be simple
<dragonker>thanks all
<rekado>civodul: I’m having a weird problem with “guix build -f doc/build.scm”
<rekado>it’s as if file-append no longer works
<rekado>in guix-manual-text-domain po-directory is defined as (file-append* source "po/doc"), but inside of “build” it is identical to “source”.
<rekado>I can only make it work by appending “po/doc” inside of the “build” gexp.
<lprndn>Hello guix! :)
<rekado>lprndn: hello! About lightdm: I’m currently stuck after I implemented my changes. lightdm fails to start now and I haven’t been able to debug it yet.
<lprndn>rekado: I can look at it and try to debug it if you want. ;)
<civodul>rekado: ah yes, same problem here
*civodul looks
<rekado>lprndn: I’ll publish the diff in a moment
<janneke>mothacehe: so, i think i found it
<janneke>copying the file /servers/exec from the vm-image over to the disk-image, fixes it
<civodul>rekado: it's a regression introduced in d03001a31a6d460b712825640dba11e3f1a53a14
<janneke>i suspect somehow, someone creates a translator there -- i suspect grub-install but havent' found anything yet
<lprndn>rekado: I'll look at it.
<mothacehe>janneke: Oh, nice!! I'm not familiar with translators. Searching for settrans in Grub doesn't output anything.
<janneke>mothacehe: yeah -- i'm still confused
<janneke>i just found that make-disk-image is populating the root file system in the store, is that right?
<janneke>and only later doing mkfs -t ext2 -o hurd and populating it?
<janneke>it seems like my make-hurd-device-nodes is running too early: it tries to create the servers/exec translator in the (linux) store, which won't work
<janneke>it can only run on a "-o hurd" mounted file system
<mothacehe>Ha! Yes, first I create a store directory which contains the whole root file system.
<janneke>unless, we find out how it works right now, and use the "old" inode embedding of translators -- yuck
<mothacehe>Then, I create a file system image out of it with mke2fs
<mothacehe>Mounting is not an option (requires root permissions)
<mothacehe>So we need to find another way indeed
<janneke>is there a way to move the make-device-nodes call to after creating mke2fs and mounting it?
<janneke>ah wtf -- yeah
<mothacehe>If an other way there is
<janneke>hmm, my linux patch hasn't been accepted or even commented on, upstream
<janneke>we could, conceivably, rewrite it and propose to allow gnu.* xattr attributes on linux ext too
<janneke>i wrote it very conservatively, only allowing gnu.* xattr's for Hurd type file systems
<mothacehe>I'm reading this:, to try to understand the concept.
<janneke>but i'm not sure if that's desired (it might be better, even -- dunno)
***mekeor- is now known as mekeor
<PotentialUser-38>hi, all. I'm using Guix on my old laptop (x220). where is almost impossibly compile in reasonable time. Every time that I want upgrade some package I run "guix pull && guix upgrade -n" and after I run guix upgrade "with only package with substitute" there is a way to do this with one command?
<janneke>oh, missing civodul on #hurd :-/
*janneke assumes too much
<mothacehe>janneke: could we set translators at boot time instead? In an initrd/activation script fashion?
<janneke>mothacehe: not in a script; it's the exec translator that is missing
<janneke>having no /servers/exec means we cannot run bash
<janneke>to run guile, we need /servers/socket/1 (pipes) on top of that
<janneke>my idea was to use this new xatt (setxattr/getxatt) code which is nice and clean and could work on linux (with a patch for the linux kernel) ...
<civodul>hey hey janneke! :-)
<janneke>i am not sure if the trick to create /servers/socket/1 could be pathed into a C startup script -- and even less sure if we'd want to do that
<civodul>janneke: previously, /servers/socket/1 would be set up by or whatever it's called
<civodul>that solved the bootstrapping issue, although we had to have Bash before Guile
<janneke>yeah, i wrote a patch to remove that and start straight into guile
<civodul>yes, but it doesn't work unless we have xattr like you write?
<janneke>well, yes -- it's a tad more complicated
*janneke wrote in #hurd...
<civodul>so maybe that's a future milestone?
*civodul just joined
<janneke>civodul: maybe...however we just found a very similar but more serious bootstrap problem: the /servers/exec translator
<civodul>there's no such problem on master, right?
<janneke>right; "someone" creates /servers/exec on master (grub-install??)
<civodul>not GRUB, GRUB isn't that smart ;-)
<civodul>i forgot, but it's either or startup.c
<janneke>i couldn't find it there...but mkfs, on linux maybe?
<civodul>from the root /hurd/ext2fs process
<civodul>so it happens at boot time
<civodul>is it true?
<civodul>ah yes, diskfs_start_bootstrap
<janneke>hmm...that makes sense...but something's weird
<janneke>we need a canary somewhere
<janneke>it only does that under a certain condition?
<mothacehe>janneke: seems that Yocto is patching mke2fs regarding xattr support
<mothacehe>I'm trying to understand what is does
<civodul>mothacehe: i think that's for an old version of e2fsprogs
<civodul>current version of mke2fs(8) says xattrs are copied by default when using -d
<mothacehe>civodul: oh ok, sorry!
<mothacehe>trying to catchup here!
<civodul>heh, np :-)
***gf is now known as gf_
<janneke>so, any idea civodul how diskfs_start_bootstrap works, i.e., why copying /servers/exec from the qemu-built vm-image to the new disk-image "fixes" booting into ?
*janneke tries to read boot-start.c
<civodul>janneke: do you know when it stopped working?
<civodul>or maybe, what does "guix system vm-image" do differently compared to the ad-hoc tool in (gnu system hurd)?
<civodul>i don't remember us doing anything special for /servers/exec
<janneke>civodul: eh,... stopped working?
<janneke>guix system vm-image finally works, on wip-hurd-vm \o/
<janneke>having solved all cross build issues, and all
<civodul>so, how does that compare to the image from (gnu system hurd)?
<civodul>normally, exec is the 2nd multiboot module in the GRUB menu entry
<janneke>...but that point now somehow co-incides with guix system disk-image /almost/ working for the first time
<civodul>it's started by the root fs server
<civodul>which makes it an active translator on /servers/exec automatically
<civodul>i think that's how it's supposed to work
<janneke>afaik, both images are pretty identical
<janneke>(gnu system hurd) and system vm-image
<civodul>in hurd-directives (on master), we create /servers/exec
<civodul>because the dir_lookup call for _SERVERS_EXEC in boot-start.c assumes it already exists, i think
<janneke>yes, i do something similar in either make-hurd-device-nodes (for disk-image) or using hurd-directives in system/vm.scm
<janneke>that "just" works
<janneke>the weird thing is that this new "guix system disk-image" does not work, until i cp /servers/exec from our qemu vm-image over to the disk-image
<janneke>note that disk image already has a /servers/exec ... so i was suspecting attributes/inode tranlators
<janneke>i bisected the working and failing images until reaching this file ...
<civodul>i think /servers/exec just needs to exist as a regular file on the root file system
<civodul>so if it's in hurd-directives there, it should be fine?
<cbaines>anecdotally, I'm seeing faster download speeds from, lots of 2Mib/s and 3MiB/s
<cbaines>I think the guix publish changes have really helped
<bdju>has there been work on converting bitmaps to opentype or whatever so they can continue to work?
<bdju> nix is making good progress on their end, not sure if someone could re-use the work for guix packages
<mbakke>bdju: no one is actively working on it, but contributions in this area would be appreciated
<bdju>I'm sure they would be appreciated... I don't feel able to contribute in most ways, though. I certainly have the time, just haven't had luck with this sorta stuff and so I'm not confident I can figure it out.
<janneke>civodul: that's almost right, possibly maybe -- but it's tricksy
*janneke goes to write a mail
<mbakke>civodul: brice is missing from build-aux/git-authenticate.scm, should he be added or are we phasing out that script?
*mbakke did not pay enough attention to the authentication changes
<mbakke>bdju: FWIW I'm confident you can figure it out :-)
<cbaines>Seems like is unreachable
<lprndn>bdju: Just in case mbakke's faith in you is not effective, a bug report would already be nice. ;) Even better if you can include font packages that need a fix.
<civodul>rekado: is the networ test on-going?
<civodul>mbakke: brice is in .guix-authorizations
<civodul>you folks should pay attention to those changes or i'll put it a bunch of backdoors muahaha
<mbakke>civodul: lol :-) I had to add him temporarily in in build-aux/git-authenticate.scm in order to pass 'make authenticate' on the staging branch yesterday night
<mbakke>hmm, maybe I should have used './pre-inst-env make authenticate'?
<mbakke>or upgrade my installed guix
<civodul>no no, just "make authenticate"
<civodul>we should never add anyone directly to build-aux/git-authenticate.scm now
<civodul>if "make authenticate" failed, maybe check whether bricewge is in .guix-authorizations on that branch
<civodul>i'd like to know if it's a bug
<civodul>if there's an unauthorized commit on that branch, we'll have to rewrite it
<nckx>rekado: I hope someone at the MDC is working after 5…? No route to berlin.
<mbakke>civodul: the key in .guix-authorizations looks correct and is the same as I added in build-aux/git-authenticate
<civodul>not reachable over Tor either, it's definitely off-line
<civodul>mbakke: how can i reproduce the issue?
<mbakke>I suppose the commit that failed was pushed before .guix-authenticate was merged? hmm
<rekado>I’m chatting with someone in the data centre…
<mbakke>civodul: try to check out staging, purge your authorized keys (where?) and run 'make authenticate'
<bricewge>mbakke: I pushed my refreshed gpg key yesterday to Savannah
<bricewge>mbakke: Is it about stagging-next?
<rekado>nckx: it’s back now
<rekado>that was due to the *addition* of a cable
<rekado>nothing was unplugged
<mbakke>bricewge: no, the current staging branch, not related to the expired key (I think)
<mothacehe>janneke: you may want to have a look to: :)
<rekado>hmm, looks like something at the core switch isn’t properly labelled.
<rekado>this will be fun…
<bricewge>It's the kmod commit
<civodul>the power button?
<mbakke>right, commit 8229ce3116c1f522c7157ab2dcd50dc2d765686a was signed before .guix-authorizations was updated with the signing key
<janneke>mothacehe: not so fast!!! ;-) /S
<civodul>mbakke: so we need to remove that commit
<civodul>well, we can discuss
<civodul>but, that's how it's gonna be
<mothacehe>janneke: hehe, I'm too glad to have a first user for system cross-compilation :)
<janneke>mothacehe: i fixed the hurd issue puzzle, trying to compose an email
<civodul>wait, it's from Apr 4?
<mothacehe>janneke: !!!
<mothacehe>janneke: can't wait to read it!
<civodul>mbakke: when was it committed? is it feasible to remove it?
<mbakke>civodul: it was pushed may 10th, but before .guix-authorizations had been merged to that branch
<nckx>rekado: Extra cables & (a few days ago IIRC) duplicate acks? Hrm.
<mbakke>I guess the easiest way is to delete staging and cherry-pick all the commits to a new staging branch
<rekado>nckx: it also looks like the redundant uplink isn’t actually redundant
<bricewge>mbakke: I signed that commit may 10th, while I was given access to master the 4th
<rekado>trying to connect an arbitrary node to the core switch and see if it’ll be fast
<civodul>mbakke: create staging2, cherry-pick d074f73aacc5a39aed0202d6e45721f53f34a8c0, rebase staging on top, rename staging2 to staging
<civodul>even better would be to cherry-pick d074f73aacc5a39aed0202d6e45721f53f34a8c0 right before the kmod commit
<civodul>there have been master -> staging merges, right?
<mbakke>I suppose I'll finally get to try 'git rebase --rebase-merges' :-P
<mbakke>civodul: yes :/
<mbakke>nckx: want a merge challenge?
<civodul>yeah, we don't want to lose those merges
<nckx>‘Guix-devel post from requires approval’
<nckx>I dunno, sounds fishee.
<mbakke>I already approved it
<nckx>Flexing your new addy, hm? Now I want one too.
<nckx>mbakke: I haven't been paying attention to your thread so I'm going to say no.
<nckx>It's a trap.
<bricewge>mbakke: staging-next has the same issue
<nckx>(But if I can seriously be of assistance do ask.)
<janneke>mothacehe, civodul: we have a working disk-image for the hurd! (see email)
<civodul>janneke: woohoo! \o/
<mbakke>janneke: wooow, congrats!! :D
<mothacehe>janneke: congrats!!!
<nckx>janneke: ♥
<mothacehe>spamming 'g' in Gnus
<civodul>well done, it's a huge step forward
<mbakke>spamming = in notmuch :P
<civodul>mbakke: for 'staging', how d'ya want to proceed?
<civodul>we should probably send a heads-up to guix-devel
<janneke>mothacehe: would you like to help splitting/fixing/properly integrating my "WIP hurd-directives" patch?
<mbakke>civodul: I'll read up on REBASING MERGES in git-rebase(1) and do some tests, and write a heads-up
<civodul>mbakke: awesome
<civodul>"apologies for the inconvenience"!
<mbakke>civodul: I think I'll insert d074f73aacc5a39aed0202d6e45721f53f34a8c0 before the kmod commit if that is practical/possible
<mbakke>indeed :)
<mothacehe>janneke: sure, I will. I can start by taking some of my commits to master, then maybe add mke2fs "-o" support.
<civodul>mbakke: on the flip side, it's a good reminder to all of us
<janneke>mothacehe: that'd be grand!
<jonsger>nckx: oke, next time I'll try evince. But to my knowledge KDE Okular has to most advanced from support. A good hack is with inkscape :P
<mbakke>civodul: also a good test that the mechanics are working properly :-)
<janneke>mothacehe: note that we just decided to postpone (the already reverted):
<janneke>cd57c138bf system: hurd: Remove bash from boot process.
<janneke>and you definitely want to have a look at
<janneke>43df8616f6 bootloader: Add `<hurd-menu-entry>'.
<mbakke>jonsger: I have Okular in a branch for a similar purpose, but could not get it to work and ended up painting the PDF with xournal instead :P
<janneke>we may want to implement that slightly differently, dunno
<civodul>mbakke: yes, and the pre-push hook will help avoid that, but server-side checks would be even better
<bricewge>civodul, mbakke : Sorry for this mess... I should have run 'make authenticate' and getting that it was missing from the branch
<mbakke>bricewge: no worries, it was a lesson for all of us :-)
<bricewge>Hopefully the new pre-push hook will avoid such mistakes in the future
<mothacehe>janneke: ok, noted!
<mbakke>it will, that's actually how I found it
<bricewge>mbakke: Can I help in any way?
<mbakke>bricewge: I'll need a coffee, but that's probably impractical unless you happen to be in the neighbourhood! :P
<jonsger>jlicht: at the beginning of using nextcloud I had a custom php package, but somehow that isn't required anymore. In general I think it would make sense to move some modules out of the php package, e.g. the database connectors as one usually needs only one or maybe two of them...
<nckx>jonsger: TBH I don't know what PDF forms are supposed to *do*. I remembered your question when I opened a completed .gov doc in Evince and noticed I could edit some fields.
<civodul>rekado: berlin is back, thanks!
<jonsger>nckx: yeah that is what it's supposed to do. And if you can save or print the filled out PDF everything is fine :9
<bricewge>mbakke: Hum, yes it's gonna be difficult
<nckx>jonsger: Yep, tested some basic editing & that works.
<jlicht>jonsger: thanks! I simply had to compile with --enable-soap, and I decided to add a php.ini escape hatch to the php-fpm service for my other needs!
<nckx>(Both saving & printing.)
<jonsger>jlicht: and for nextcloud I had to configure the php-fpm thing: I'm not sure if this is a bug
***stefanc_diff_ is now known as stefanc_diff
<PotentialUser-38>hi, all. I'm using Guix on my old laptop (x220). where is almost impossibly compile in reasonable time. Every time that I want upgrade some package I run "guix pull && guix upgrade -n" and after I run guix upgrade "with only package with substitute" there is a way to do this with one command? (resend)
<jlicht>jonsger: heh, nginx all the things! ;-)
<NieDzejkob>wdym "with one command"?
<bdju>I used guix on an x220 for several months, some stuff like rust, webkitgtk, and cataclysm-dda could take hours to build
<bdju>but aside from some big ones like that it wasn't too bad
<nixfreak>Ok , enlightenment desktop lock is still broken lol, I couldn't even find a way to unlock in another tty
<jlicht>bdju: I have a similar situation with my T400, but I 'solve' it by pinning my big packages to a version of guix that has these substitutes available in my manifest
<bdju>ah, neat approach
<jlicht>buyer beware though for stuff regarding security patches. Going around with an unpatched icecat for long seems like a good way to have a Very Bad Time :-)
<jlicht>but gimp, libreoffice and wine are some of the things that are in my `define %big-packages' list
<rekado>jlicht: do you use the cuirass API to check if substitutes are available or do you do this manually?
<jlicht>I had a gum-and-strings script to check this quite a while back, but nowadays I check this manually with `guix weather -m <a manifest with only big packages>.scm'
<jonsger>I have this problem as well, but I wonder why we don't do a different approach. Build the packages before merge/push them to master.
<jlicht>I'd rather turn the problem around: have a "mostly-built" branch that lags behind master, which is only fast-forwarded and synced up to master once certain % of substitutes is available. How one applies security patches in a timely fashion is an exercise left to the dear reader :P
<PotentialUser-38>jlicht jonsger:my opinion a good compromise is to add a option at guix upgrade --only-susbtitute #covodul
<jonsger>jlicht: but building before pushing is like traditional distros does it
<jonsger>and it works for them
<NieDzejkob>at least right now, some developers have push access, but not access to
<rekado>jonsger: is just another Guix user, though
<rekado>it also just polls the git repo and builds stuff from there
<rekado>so we would still want to push, but it wouldn’t have to be the same branch
<rekado>ideally we wouldn’t even have to make the distinction between core-updates, staging, and master
<rekado>the CI could figure this out
<rekado>and then dispatch to the three channels
<rekado>but do we really want a bot to have push access to our repository?
<rekado>or should it be another repository/
<rekado>what about signed commits?
<rekado>the bot can’t re-sign them
<jlicht>rekado: whether a commit is signed shouldn't matter for a substitute or build-farm, right?
<PotentialUser-38>jonsger jlicht covodul rekado : In my opinion a good compromise is to add a option at guix upgrade --only-susbti:my opinion a good compromise is to add a option at "guix upgrade --only-susbtitute" (resend)
<PotentialUser-38>arg ^ civodul
<rekado>PotentialUser-38: the implementation of that isn’t straight forward and needs discussion
<nixfreak>are texlive-txfonts the only fonts package ?
<rekado>nixfreak: no
<rekado>txfonts is the only package providing the TeX Live package “txfonts”
<civodul>the reason --only-susbtitutes was rejected before is for security reasons: one can easily end up never getting security updates because of that
<nixfreak>hmm when I search fonts I only get that or if I search ttf I only get sbcl
<civodul>still, we need a solution
<jlicht>how do the nix folks deal with this?
<rekado>nixfreak: what is your search query? And what is the tool you use to search?
<rekado>they build more
<drakonis>jlicht: we fetch
<rekado>but I have to say that we should have enough servers to build things in time
<rekado>but often enough I still see idle servers
<drakonis>texlive packages are acquired off texlive instead of being hosted on the cache
<nixfreak>using guix search fonts or guix search ttf
<civodul>jlicht: i think they have "stable channels" nowadays
<civodul>but i'm not sure about the security story for those
<civodul>especially in the absence of grafts
<PotentialUser-38>civodul : If this is the only reason. Could be a solution --only-susbtitutes except for %security-package
<civodul>how do you define what goes in %security-package?
<civodul>it's never simple, it depends on what you do with the package
<rekado>afaik nix doesn’t actually build texlive from source
<civodul>oh i guess i misunderstood the question
<jlicht>^ I was talking about what you mean, civodul ;-)
<jonsger>rekado: I know that cuirass isn't built for that use case. But I think we have already some bits with cbaines stuff
<jlicht>sorry rekado, I could have prefixed that it's after 7 here, so no voluntary texlive from my side :D
<PotentialUser-38>civodul : every lib that serve some cryptographic function or related on communication and all the package that is related
<jlicht>PotentialUser-38: ...and every internet facing software! And every containerisation software! And the kernel! And everything with CVE in the commit message! etc etc etc
<cbaines>jonsger, I still want to try and get some automation behind building things related to patches. Maybe that could provide a way to provide substitutes at the point when the patches are merged in to master.
<jonsger>cbaines: nice. That's what I actually thinking/dreaming about
<drakonis>rekado: it most certainly does build from source
<nixfreak>what other font packages are their ? nerd fonts ?
<cbaines>nixfreak, have you tried running guix search ^font | less
<cbaines>there's also which shows some fonts
<PotentialUser-38>jlicht this is not a world for 2GB RAM computers
<drakonis>for some specific things that is
<jonsger>nixfreak: or guix package -A font
<cbaines>I sent a long message to guix-devel about my thoughts on assisted patch review:
<cbaines>It's slowly progressing, but it's only one of the areas I'm working on at the moment
<nixfreak>ok thanks
<drakonis>PotentialUser-38: well, arm devices still have that much ram
<mbakke>the RAM requirements should hopefully be significantly lowered with the upcoming Guile 3.0.3
*mbakke is back after a HDMI cable emergency + surgery
<nixfreak>can I install libc to execute elf files ?
<lfam>You rebuilt an HDMI cable mbakke? That's impressive! HDMI is really pushing the limits
<mbakke>nixfreak: you mean run pre-compiled binaries?
<mbakke>lfam: it ultimately failed :-/
<lfam>I'm not surprised. HDMI barely works in the normal case
<mbakke>nixfreak: you can patch the binary with 'patchelf' to fix the loader and RUNPATH
<mbakke>nixfreak: alternatively you can use 'extra-special-file-service-type' to make a symlink from glibc to /lib64/
<nixfreak>also do you usually have to reboot in order for the firmware packages to work e.g. iwlwifi
<nixfreak>very cool thanks
<lfam>I wonder if there's another channel where you can take questions about nonfree software?
<mbakke>huh, I did a test interactive rebase with --rebase-merges, and while git processed 2200 commits, the end result was "up to date with 'origin/staging'"
<mbakke>not sure if I should be impressed
<mbakke>now to make some changes in that rebase
<nixfreak>I get it "its taboo" but its either install the nonfree wireless drivers or have to desolder my wireless chip and install a new one
<bdju>is it really soldered on? wow
<bdju>most I've seen were mPCIe or M.2
<lfam>I understand, nixfreak, but it's a policy of the project to not use our communication channels to recommend nonfree software
<rekado>drakonis: AFAICS they are merely unpacked, not built. There is no generic way to build them yet does not contain any special treatment for any of the packages.
<lfam>nixfreak: In general, I recommend rebooting after changing firmware things
<rekado>take amsfonts as an example. The package in nixpkgs does not appear to attempt to build the fonts from the metafont sources.
<rekado>same with the hyphenation pattern packages, which have ruby source code.
<mbakke>Rebasing (80/1922)
<mbakke>rebasing merges is fun :P
<mbakke>never really know what to expect from the end result
<mbakke>oof: Could not apply 87a40d7203... master # Merge branch 'master' into core-updates
<mbakke>I think I need a second branch in order to insert a cherry-pick into the right place
<mbakke>oh ofc, .guix-authorizations has not yet been created at 8229ce3116c1f522c7157ab2dcd50dc2d765686a
<mbakke>so I need to move that commit to after the first merge
<mbakke>it's not enough to simply cherry-pick d074f73aacc5a39aed0202d6e45721f53f34a8c0
<jlicht>mbakke: what lovecraftian undertaking are you a part of now?
<mbakke>jlicht: we did an unintended test of the new commit verification infrastructure :P
<mbakke>long story short: it works great, but we need to ensure that .guix-authorizations have been merged to any branches that new committters/keys will be pushing to
<lprndn>rekado: I won't have time to explore more your lightdm service today. But I just found out that it's was probably a PATH issue. Adding back the profile service + environment variables of the shepherd service makes it work. I think it just doesn't find the greeter executable ;)
<rekado>lprndn: oh, thanks for testing
<rekado>I’ll see if we can set PATH for the lightdm shepherd service without resorting to using the profile service.
<nixfreak>has anyone got connman to start ?
<efraim>nixfreak: you've added it to your system config, right?
<nixfreak>Ok , thats what I was going to ask if it automatically gets added to the config.csm or you have to manually put it in there
<nixfreak>I'm still learning and reading
<efraim>you have to add the service for it
<nixfreak>ahh ok
<mbakke>okay, first rebase test looks good
<mbakke>"Your branch and 'origin/staging' have diverged, and have 63 and 63 different commits each, respectively"
<mbakke>uh, 'git log' looks nothing like on staging though, fun
<mbakke>oh, git has cherry-picked all staging commits since the rebased merge, makes sense
<civodul>mbakke: woow, already!
<mbakke>git merges are second nature to me nowadays :P
<atw>"git has cherry-picked all staging commits since the rebased merge, makes sense" hmm yes I know some of those words
<kamil_>Oh noo.. Guix doesn't package xdg-desktop-portal services :(
<lispmacs[work]>I'm trying to understand how to use the feature of multiple profiles for code development purposes
<lispmacs[work]>say, I create a secondary profile
<lispmacs[work]>can I run that secondary profile in some kind of environment, without having to make it my current profile?
<mbakke>oh my, git rerere had forgotten an important merge resolution
<wdkrnls>I don't suppose there is any way to nudge guile to re-read the channels specification after writting the file in /etc/config.scm?
<mbakke>currently 'staging2' will rebuild 'python' 😅
<wdkrnls>I'm trying to make my Guix configuration self-contained within 1 file: /etc/config.scm.
<wdkrnls>So, I thought maybe I could just set my channels specification by pretty printing the guile code specifying my channels to that file as part of /etc/config.scm.
<wdkrnls>here "that file" means /etc/guix/channels.scm.
<nckx>wdkrnls: Did you run ‘guix pull’?
<wdkrnls>no: I was inside the graphical installer where I wrote out all my code in zile.
<nckx>There's nothing to nudge, you either pull or you don't.
<wdkrnls>Okay, so basically it's impossible to fully specify your configuration until after you specify your configuration less than fully.
<NieDzejkob>wdkrnls: instead of one file, I'd recommend one directory maintained under git
<wdkrnls>I guess, I just have to figure out how to do that without the installer :)
<nckx>It's a clever trick but far outside of the workflow on rails that the graphical installer is meant to guide you through.
<wdkrnls>could I write some guile code to do guix pull inside of my config.scm?
<NieDzejkob>You will have a bad time trying to configure vim or emacs out of /etc/config.scm anyway
<nckx>That said, you can always switch to a VT to do the sneakies.
<wdkrnls>Hmm... I was having a hard time getting VT's to appear. They weren't showing up with Alt-Fn3 etc...
<nckx>Erm, ‘write some guile code to do guix pull inside of my config.scm’ really triggers all the ‘pause, step back, what are you trying to do here’ alarm bells 😃
<nckx>Try C-M-F3 or Super. One of those should work.
<nckx>(Not a Guix thing, I believe, but a side effect of using kmscon.)
<wdkrnls>oh, good. That C- prefix seems to do the trick.
<civodul>i read on the swh-devel list that Software Heritage starting ingesting the Nix and/or Guix sources
<civodul>which sounds pretty sweet!
<user_oreloznog_>Yeah :-)
<civodul>so hopefully "guix lint -c archival" will complain less frequently
<civodul>source archive ✓ bootstrapping ✓
<civodul>cbaines: i think you can push
<cbaines>civodul, cool, will do :)
<janneke>todo: sunbathing :-)
<raingloom>do we not have any data files for tesseract? ,:|
<raingloom>nckx: thanks.
***sneek_ is now known as sneek