IRC channel logs
2026-05-01.log
back to list of logs
<vagrantc>first time i ever did same-day kernel updates :) <vagrantc>mwette: openscd doesn't appear to have substitutes available even without your patch, so chances are it is nothing you have done <mwette>vagrantc: thanks, odd that it got all the way to test on my desktop <mwette>Oh, but I did try --no-substitutes. <theesm>vagrantc: are the deblob scripts available yet for todays round? checked a few hours ago the last time <vagrantc>theesm: they were not ... given the severity of the security updates, i chose to push right away ... <vagrantc>theesm: although maybe i should have waited for 7.x (not present) and 6.18.x (old version not affected) <vagrantc>5.10.x is still affected, as new versions no longer build. <vagrantc>on x86_64 ... who uses that architecture anyways? <vagrantc>we have so many other kernel versions to choose from ... so anyone still using 5.10.x for some reason will hopefully be invested in fixing it :) <vagrantc>the whole reason we re-apply the linux-libre machinery against the upstream tarballs is exactly for situations like this ... so if we're going to suffer the pain of it, we may as well reap the benefits :) <lullaby>To be clear, currently there is no such file at /etc/containers/policy.json <mwette>what do people use for LC_ALL? Someone mentioned here that C.UTF-8 should be used. <vagrantc>not sure it is enabled in the build environment yet, but it ought to be :) <vagrantc>it's a reasonably agnostic country-unspecified locale (other than the obvious, uh... basically being english) <mwette>as user, as well as for guix-daemon <mwette>i ran into an error message on locale, but I'm seeing a lot of odd errors in this (openscad) issue ; thanks vagrantc <lullaby>For policy.json I ended up using a simple file that allows anything from docker.io <lullaby>{ "default": [ { "type": "reject" } ], "transports": { "docker": { "docker.io": [ { "type": "insecureAcceptAnything" } ] } } } <apteryx>the 'guix deploy' backtraces are tough <rgarcia>hi! So I added a pam-limits-service-type to the list of services on my config.scm. sudo herd status reports a pam service runing. But there's no /etc/security/limits.conf file. Is this normal? <sneek>Welcome back rgarcia, you have 1 message! <sneek>rgarcia, yarl says: guix shell -e '(use-modules (guix packages) (gnu packages glib) (gnu packages xdisorg)) (package (inherit arandr) (inputs (modify-inputs (package-inputs arandr) (replace "python-pygobject" python-pygobject-3.50))))' <nomike>I'm trying to fix an issue with the prusa-slicer package. In the prusa-slicer source they are searching for the system CA bundle. On Ubuntu this is detected as /etc/pki/tls/certs/ca-bundle.crt. Since Guix is not adhering to the FHS, this detection fails. I want to patch the path to that bundle into the prusa-slicer source code. Does anyone know how I can get that path in the package? <loquatdev>If I'm writing a service that frequently runs guix commands under a user, should I primarily step channels by using `guix pull` or run every individual command with `guix time-machine`, using gexps + the guix package? <loquatdev>I started with `guix pull` in my service because it makes things easier in some other parts of the script, but it also depends on the PATH. <GalaxyNova>Should I add the standard `guix` channel to `dependencies` in my `.guix-channel`? <loquatdev>No, you don't have to add it. I imagine you could, but big channels (nonguix, for example) don't have it in there and I'm not really sure how you'd be pulling a guix channel without, well, Guix. <loquatdev>Although, attempting that might be an interesting experiment :) <GalaxyNova>conceptually it feels wrong to have the channel depend on the user's external channels.scm file, isn't it the case that if `guix` it is not specified in channels.scm, then `(use-module (guix))` and the like will fail? <loquatdev>GalaxyNova: If I remember correctly, `guix pull` will simply refuse if no `guix` channel is specified. <loquatdev>Learning Emacs is on my TO-DO list. I'd be lying if I said I wasn't putting it off, haha <adanska>i started using doom emacs and ive never looked back. does everything i could ever want and if i need something else i can easily just write a little elisp function or a hook for some package <GalaxyNova>loquatdev: welp, there is only one way to find out <gerogaga>Is it advisable to create Shepherd service files as part of package definitions for daemons? <adanska>gerogaga, yes, it is a welcome addition since they are not particularly useful otherwise <identity>so 5f9b8cef broke ring-racers; the package builds, but now there is no wrapper that sets RINGRACERSWADDIR, only the bare binary <identity>…so ringracers fails at runtime with «I_Error(): "bios.pk3" not found!» <rovanion1>When I reconfigure my system I'm told "warning: 'local-host-aliases' is deprecated, use 'local-host-entries' instead". When I replace local-host-aliases with local-host-entries I'm told "error: local-host-entries: unbound variable". Both seem to be defined in the same file. What gives? <identity>also, ring-racers seems to install the icon incorrectly, the desktop entry does not display the icon <Rutherther>janneke: because the 'default' versions (usually latest, but not always) are called without version - so qtbase <apteryx>is it only me, or are VMs installed with GNOME Boxes always end up broken after a while for some reasons? <apteryx>I don't think I've run a GC since I installed the broken one I'm looking at, so it's probably not about stale baked store file names <janneke>apteryx: when i still used gnome, i often had to rm -rf /var/.../cache or something <apteryx>I'll try... I'm still trying to make sense of the log file <apteryx>looks like it contains some old stuff <janneke>gnome used to use a constant error message (one of the great evils imo) "oh no! Something has gone wrong" <janneke>Rutherther: i guess that makes sense <janneke>ACTION had the impression they were using qtbase-5 at some point where they needed 5 over 4, and apparently got "stuck" on 5, so yeah <janneke>quite surprising there's no qtbase-6 now, i think linux-libre has all versions also available with their current number? <janneke>but it would "fix" packages getting stuck on an older dependency... <jlicht>hmm, could it be that the emacs-eglot package we've had in master actually depends on an as-of-yet unpackaged version of emacs-jsonrpc? <janneke>every time json is used in one sentence with a lisp, a fairy dies <jlicht>ACTION is okay with that if the fairy mostly shouts "LISTEN" <efraim>ACTION has flashbacks to Ocarina of Time <janneke>sadly, they're all beautifully curved s-exp faries (x/v/m) <rovanion1>How do I use block-facebook-hosts-service-type in combination with a normal hosts-service-type where I add my own entries? <Rutherther>rovanion1: so what's the problem you're seeing? it should work normally, you just add that facebook service along with your own extension of hosts-service-type <rovanion1>Rutherther: guix system: error: more than one target service of type 'etc-hosts' <Rutherther>rovanion1: are you adding hosts-service-type? you should be extending it, not adding it. Every operating-system that does not use the hosts-field already has this service, so you cannot add it twice <Rutherther>rovanion1: right, you are adding it. This isn't caused by the block-facebook-hosts-service-type in the first place, even if you removed that you would get the error, just extend it, ie. "(simple-service 'block-reddit hosts-service-type (list (host "127.0.0.1" "www.reddit.com")))" <Rutherther>your value going to the service is also unsupported one. It accepts list of "host" records. And those local host entries are already added by default, no need to add them <apteryx>'virsh --connect qemu:///session start ubuntu24.04-' says: error: Cannot check QEMU binary /home/maxim/.guix-profile/bin/qemu-system-x86_64: No such file or directory; normal, it's now in ~/.guix-home/profile/bin <Rutherther>apteryx: well at least it did not resolve to /gnu/store path, that would lead to even higher frequency of errors <apteryx>funny, it was baked directly in the VM definition that libvirt or gnome-boxes made; I could fix the qemu path via 'virsh --connect qemu:///session edit ubuntu24.04-' <apteryx>so now I may be able to test the installer on Ubuntu 24 again, because it looked broken according to guix deploy with a hetzner-environment-type machine <apteryx>it fails with 'chown: cannot access '/etc/guix': No such file or directory' <apteryx>hm, it drops me to an initramfs shell now. I guess that one is truly broken. <sham1>I take it that GRUB still doesn't support LUKS2 with Argon2i. Oh well, guess I'll separate out my store and my /home so I can have stronger encryption for the latter <mwette>nomike: find . -type f -exec grep pki/tls/certs/ca-bundle.crt {} \; YIELDS src/clic3r/Utils/Http.cpp <rovanion1>Is it possible to get a shell inside a failed build environment? <identity>rovanion1: see (info "(guix) Debugging Build Failures") <apteryx>OK, the image used to deploy to hetzner was not ubuntu 24, but debian 13 <apteryx>so apparently guix 1.5.0 can't be installed with guix-install.sh on debian 13 <apteryx>same error as reported earlier: chown: cannot access '/etc/guix': No such file or directory <apteryx>hm, just tried this in a debian 13 VM, and yes '' | sudo guix-install.sh completes successfully, so it appears something in the hetzner environment <apteryx>I just noticed the first trouble appears to be "Job for guix-daemon.service failed because the control process exited with error code." <yarl>There's a certificate problem for berlin.guix.gnu.org <vagrantc>everybody got their up to date kernels yet? :) <nckx>yarl: There is no certificate for that domain. What are you expecting to be there? <nckx>crt.sh is being super flakey (as always) but seems to agree that connecting to that hostname is not expected to work. <nckx>vagrantc: (Quake 3 headshot voice) Impressive. <cdegroot>(exception unbound-variable (value "module-lookup") (value "Unbound variable: ~S") (value (rust-aligned-0.4.3)) (value #f)) <cdegroot><-- is anyone getting that error trying to pull guix-science? Is there stuff guix-science depends on? <mwette>I had to shut down my (foreign) desktop because kernel update not available: canonical is currently under "sustained, cross-border" ddos attack. <cdegroot>(nm my question, I had the Guix channel pinned. That'll do it) <nckx>vagrantc: That message was… uselessly vague. I meant to praise your above & beyond packaging efforts. <Tadhgmister>I somehow messed up my harddrive partitioning and there are entries in the guix store that throw IO errors, very critically packages that are being read and copied to a new drive by a `guix system init` command. any suggestions on how to recover enough to initialize a new system? from trying to do some `guix gc` commands it seems that at least some of them are live roots of the booted system which worries me <vagrantc>nckx: heh. I assumed it was overall positive. :) <vagrantc>nckx: did send me off on wikipedia researching the precursors to quake that i actually played <ieure>Tadhgmister, Save your $HOME and system / home configs. If you don't use Guix Home, export your per-user profile to a manifest. Reinstall, recover the previous state from those. <Tadhgmister>"enough to initialize a new system" -> use an install media, I swear sometimes the most obvious solutions just fly right past me <ieure>That's cool. I like radare2, I have used it a little bit. It helped me crack some ancient abandonware niche proprietary corporate software with a hardware device serial number check. <ieure>First time you boot from the media containing the software, it writes the machine's serial number back to the disk, making it refuse to work on any other. Until I patched it out. <bdunahu>nice! iaito frontend + this decompiler makes reverse engineering much more beginner friendly for me <vagrantc>theesm: the linux-libre scripts didn't change, fwiw :) <yarl>nckx: As it is reachable, a certificate. <Rutherther>yarl: but what do you use it for? it has been superseded by ci.guix.gnu.org <yarl>Rutherther: Why is it still reachable? <Rutherther>yarl: no one has removed the dns record I presume? <yarl>Btw, it makes sense to me that as bordeaux.guix.gnu.org is legit that berlin.guix.gnu.org is too. <yarl>uh, do you get an error on qa.guix.gnu.org ? <yarl>graphQL related exception <mwette>ieure: Did you try my openscad patch? I assume if the substitute works it should be wrapped in a aarch64-specific cond form (e.g, `(cond ((target-aarch64? (%current-target-system))))'). <kestrelwx>jfred: Can we have aarch64 substitutes for Zig packages at all? I am not familiar. <jfred>(granted I'm not all that familiar with Zig in particular, but I don't think it's any different from packages in other languages in that regard) <jfred>it's the fourth line on that table/second line in tier 2 <kestrelwx>Probably need to see if the newer versions build with Zig 0.16.0, it also fails on master too seems.