IRC channel logs
2023-02-04.log
back to list of logs
<mvnx>RE: my above message... Maybe it was just a mirror issue? I ran `guix pull` as my user, it updated from `git.savannah.gnu.org` instead of `archive.softwareheritage.org` and then I ran the same `sudo guix system reconfigure ~/.config/guix/system.scm` and it still gives the provenance warning but updates guix channel from a working mirror, I guess? <mvnx>Is the first `guix pull` not necessary? Anyway, I didn't change the configuration file, just copied to a new file and tried a `guix system reconfigure` with it and when I reboot I get an e2fsck bad magic number in superblock error - . It's tough to keep testing this because the guix system reconfigure takes like 30 minutes. Any suggestions? <rekado>“guix pull” will always fetch from git.savannah.gnu.org, because that’s where the “guix” channel itself is hosted <rekado>archive.softwareheritage.org is only ever used as a fallback when source code origins cannot be reached <rekado>“guix system reconfigure” without changes should be pretty quick. < 1 minute. <mvnx>Thanks, that's helpful. Maybe just bad timing because all my guix pulls were going to softwareheritage. So my bigger issue is the `guix system reconfigure` using the configuration file found at `guix system describe` (copied to a new file) took a long time and put the VM in an unbootable state <sammm>hi, relative guix newb here. has anyone built a guix system for an odroid xu4 sbc? i am playing around with defining images/bootloaders etc... <sammm>thx mekeor[m] , looks like the upstream u-boot doesn't have a board definition for xu4.. but there is for xu3, etc.. will play around <mekeor[m]>what would be a good approach to implement a feature for guix-search to search for a package within the whole git-revision-history of a channel? <mekeor[m]>the goal would be that you find out where to time travel in order to install the specific package with the specific version you want <mvnx>`sudo guix system reconfigure $config` reliably reaches out to the fallback archive.softwareheritage.org, and subsequently fails, on the 1.4.0 QEMU VM. Is this expected? <johnabs[m]>Hey all! Anyone here happening to use sbcl-coleslaw on guix? I've run into a funky issue that seems to stem from the read-only file system when I try to create the Clack server to preview my edits. <oriansj>the_tubular: disable the service or not even install it in the first place <oriansj>the_tubular: you need to do herd stop as root <oriansj>herd stop ${service_name} to be more precise <apteryx>rekado: you are right, it's because of fluidsynth/lash <apteryx>lash: Building the following 434 packages would ensure 832 dependent packages are rebuilt <apteryx>a bit of a stretch for master, but we can do it on staging, it seems? <apteryx>seems there exists something called ladish which is a drop in, actively maintained replacement <apteryx>I could build fluidsynth without lash fine <yewscion>Hey all, anyone know why GNU Cuirass would be reporting that its schedulers are blocked when no one else is using the system? I have an instance that has both of its schedulers blocked for 4000+ seconds at this point. <apteryx>you can report it to bug-guix with a [cuirass] prefix <attila_lendvai>this tripped me up just now: the variable 'fontconfig' holds a package named 'fontconfig-minimal' <pjals>Hey, has anyone attempted to make a NanoPi M4 image with/using Guix? I tried to make one based off RockPro64 but it didn't go so well. <pjals>I skimmed over the sources and it looks free (libre). <klm_>The font rendering / layout is a bit strange when I'm using `evince`. What could be causing that, a missing font? I did not install all of gdm, just evince (I'm using just xorg-server and i3-wm). <jonsger>klm_: could you post a screenshot somewhere? <klm_>and some of the icons on the side-menu are missing. So I'm missing a package somewhere, I'm pretty sure - but which one? <jonsger>klm_: the font rendering inside evince is some issue of itself, AFAIK. It can be much worse for some PDFs. The missing icons is some Guix stuff <jonsger>are you using Guix on a foreign distro or as Guix System? <klm_>But I'm starting x11 using `sx` which means I need to add fonts and all sorts of things to get basic functionality working. <klm_>I suspect `evince` would have worked properly if I'd had `gdm` and all its friends. <Parnikkapore_m>I'm trying to run uxnemu (a basic SDL2 app) in 'guix shell -C' for funsies; it says it couldn't find the display. What flags should I pass to get it working? <jonsger>klm_: I know it was an issue for myself, but don't ask me how I fixed it (maybe by using SDDM)... <jonsger>ACTION using `guix build --system=aarch64-linux hello` on a foreign x86_64 distro still feels like magic... <klm_>jonsger: ok, thanks anyway :-) I'm actually trying to learn how to use MuPDF instead. <pjals>oriansj: isn't that only for the rpi? <apteryx>rekado: I have a patch series removing lash; jami no longer depends on python2, sizes is reduced by about 50 MiB <apteryx>next target to remove would be gtk+@2 <dcunit3d>how do i get straight to clone repos properly if i? <dcunit3d>if I'm starting emacs from within a guix shell? <dcunit3d>i'm getting certificate verification failed CAfile: none CRLfile: none <dcunit3d>magit also fails to fetch for the same reasons <apteryx>dcunit3d: try adding nss-certs and git or another package that causes SSL_CERT_FILE to be set <apteryx>are you working from a --pure guix shell? otherwise it should pick SSL_CERT_FILE from your environment and juts work, if your environment already does <dcunit3d>yeh, i'm launching from within an xsession on arch <dcunit3d>it seems to work when i back out to a vty <dcunit3d>no, it's just `guix shell -m manifest.scm --check` <dcunit3d>one sec, i didn't copy the manifest over <apteryx>OK, but this being launched from a foreign distro's xsession it probably lacks a pre-configured SSL_CERT_DIR/SSL_CERT_FILE <apteryx>so add git or curl or whatever has a search path for it to your profile <dcunit3d>git is in the profile. i think once i run `guix pull` for the profile, it will work <dcunit3d>i'm trying to mix straight and guix dependencies, following the guix-home branch on github.com/daviwil/dotfiles <mirai>is there a way to troubleshoot why this file-system herd service fails <mirai>is it possible for /etc and / to reside in different partitions? <lfam>Any plans for dinner in Brussels tonight? <sneek>Welcome back lfam, you have 1 message! <sneek>lfam, nckx says: Using the patented debugging technique of ‘hopelessly dig around for relevant logs; give up and herd restart the world’, evaluations have been resumed. <mirai>it's not your usual setup but I'm suspecting the file-system-service runs before account-service-type and this doesn't work if both the account and file system are created as service-extensions <mirai>the uid=...,gid=.... options will make the mount fail? <mirai>what provides lua-http ain guix? <Obikawa>Hi. Is it possible to use git-fetch to use a >local< GIT repository? I have been blindly trying to install a patched version of emacs-org-fc without any further success after hitting a "fatal: '/home/marek/src/org-fc' does not appear to be a git repository". <Obikawa>Thank you. I had the feeling I was on a wrong track. I can work with this. <mirai>alternatively, you can spin a git-daemon-service-type instance <Obikawa>mirai, will I be able to call localhost in the git-fetch + git-daemon-service-type solution? <mirai>as long the host can be resolved, it can point to anything <Obikawa>mirai - I am a programming noob, so I would rather ask first before diving in. <tedium>is there a way to configure guix system to attempt to decrypt every hard drive with a given password? I don't want to put in the same password 4 times for my root drive and 3 RAID drives. <a12l>I've a question about rust development on Guix, but I suspect that the people that knows how to do that isn't active on IRC when I'm. Where is the best place to send such questions? The general mailing support list? <a12l>Also can't find any relevant information when I search on Google, so I suspect that there isn't a big guix-rust ecosystem out there. :( <mvnx>Fresh 1.4.0 QEMU VM. I try `guix system image --image-type=qcow2 $(find /gnu/store/ -type f -name vm-image.tmpl | head -n 1)` it consistently tries to pull from `archive.softwareheritage.org` and then fails with a giant stacktrace. Any way to force it from `git.savannah.gnu.org` instead? <unmatched-paren>is there anyone around who could review dissecting guix part 2 v3? :) <mirai>I'd be interesting if guix could get some sort of opt-in telemetry <mirai>data points of immediate interest would be: services and packages used (naturally, it should only report the ones distributed with guix) <mirai>This could be used as some sort of early warning for potential bitrot in services/packages <tasankovasara>btw. is there a repository of packages (compiled or not) that I could rsync to have a local mirror (like with Arch and Arch ARM presently)? <unmatched-paren>tasankovasara: guix doesn't really have a repository of package sources <unmatched-paren>every package is built from source (or downloaded from the CI, which compiles them automatically) <unmatched-paren>using the package definitions in the Guix source code, which is locally downloaded every time you do ``guix pull' <unmatched-paren>i'm not sure whether you'd be able to do something like that for the compiled packages on the CI, but even if there was i don't really see the point <unmatched-paren>because they'll become outdated really fast and anyway it's probably less work to just compile things when you're offline <unmatched-paren>because you need to be able to download the tarball/git repo from the original source... <unmatched-paren>cloning every single ``source'' object into your store is, needless to say, A Bad Idea :) <tasankovasara>unmatched-paren: I'm kind of prepping here for the big Internet down event, as you figured out :D <tasankovasara>but guix is really sexy, might be tempted to switch even if there's no easy way to have a selection of software locally available and updated <unmatched-paren>tasankovasara: i'm afraid there's nothing much you could do other than maybe set up your own Cuirass (the CI infrastructure we use) <unmatched-paren>tasankovasara: well, if you do have an unused server lying around, then you can just set up Cuirass on it (disclaimer: i have no idea how to do this) and make it build only the master branch <unmatched-paren>and then you can set that server as one of your Guix's substitute sources <unmatched-paren>not entirely sure how you'd get your Guix to treat it as a "local" thing, not just as any other substitute server accessed through the internet :( <unmatched-paren>i *think* there might be something you can do with the Avahi LAN daemon, which Guix supports in some way? <unmatched-paren>disclaimer: i may be talking complete nonsense. it's possible none of this will work <tasankovasara>I haven't got a Guix system to play around with yet, but I'll be looking into hosting a package source and such <vagrantc>tasankovasara: there is *some* sort of mirroring possible, i know it was/is done with at least one mirror in china <vagrantc>another option would be to run "guix build --source" on all the stuff you are interested in while you're still online ... then you would at least be able to build any of those things <vagrantc>while offline ... although there are probably some surprises along the way <gabber>i'm trying to bump sway but it refuses to acknowledge my update of wayland to 1.21.0 and keeps saying it only finds 1.20.0 <gabber>*wayland-server that is. is that not part of the package `wayland`? <mirai>can someone test if `guix build lua-cqueues` succeeds? <mirai>so this is to be reported as a non-reproducible bug? <mirai>ah, my bad, I meant to say broken package <gabber>unmatched-paren: is that supposed to be a binary? i cannot find it within `guix build wayland` <gabber>mirai: :) so i guess yes, reporting seems appropriate (if it hasn't been reported already) <gabber>wow yummy, that's hella helpful /s <unmatched-paren>pretty sure you also need to update libdrm... lots of fun all round :) <gabber>the funny thing is: i think i checked all sway's dependencies for anything matching version 1.20.0 (the one it complains about no being 1.21.0) -- only package wayland matched. so i updated that and am able to build it but sway just... ignores that fact? <unmatched-paren>what is the output of ``cat $(guix build your wayland here)/lib/pkgconfig/wayland-server.pc? <gabber>./pre-inst-env guix build wayland <gabber>the really weird thing is, though, that i *only* updated the version field in the wayland definition. <gabber>it is built successfully, though <neshamon[m]>Does anyone else have issues adding sddm to their config? <gabber>and i honestly don't know what to change it to if guix doesn't tell me the expected hash value <neshamon[m]>I keep getting an error saying xorg-server is provided more than once when I add it with set-xorg-configuration <gabber>weird.. i did as you said, unmatched-paren but the hash wasn't correct. but altering the hash made guix re-download the source, check the hash and tell me that the one i entered was wrong <gabber>shouldn't guix try to download new sources when i change the version field (which changes the download url)? <gabber>sway seems to build fine now, so thanks to all'yall <gabber>neshamon[m]: did you add more than one xorg-service? would you mind sharing your config (or at least the (services ..) s-exp)? <gabber>maybe lxqt and sddm both provide an x-server? i guess providing the sddm server ought to be enough and just adding lxqt to the system wide packages might make it available within sddm <gabber>the new sway needs wlroots version >=0.16.0 but only 0.14.1 has been released? <gabber>ACTION just realized they moved *all their repos* to gitlab <vagrantc>gabber: if a matching hash is already in the store, guix will in some cases just use that regardless of the version field <vagrantc>i've definitely been tripped up on that beffore <vagrantc>the version numbers in guix are more for the user interface ... when it comes down to it, guix really just relies on hashes <gabber>it's still weird. the other packages i'm updating rn trigger a source re-download... is it possible there's a bug because the new source (with the version-string in the url) cannot be found? <vagrantc>maybe you don't have the exact hash of the other sources in your store already? <gabber>would you mind trying just setting the version string of the package wayland to "1.21.0" and see whether guix asks for a better hash? <unmatched-paren>here "inputs" means ``inputs'' plus ``native-inputs'' plus ``propagated-inputs'' plus the source code plus the builder script <gabber>thanks! haven't read that one, but i'm looking forward to! but: changing part of the recipe of the build usually also triggers a rebuild (bc the hash of the package itself changes), right? maybe i found a bug -- would you mind checking for me what happens if you alter your wayland package version string to "1.21.0"? <nckx>gabber: If the file exists upstream, guix will download it, using the 'basename' of the URL, which won't exist yet; then it will check the hash and fail (suggesting the right one). If the file does NOT exist upstream, guix will also use the upstream URL's basename BUT download from a content-addressed mirror--it will create a new file with the old contents. <nckx>No real way around that with only a CA mirror AFAIK. <unmatched-paren>gabber: the version string is **not** part of what goes into the derivation <nckx>Why would it not be part of the derivation? <nckx>In wayland's case I mean. <nckx>It's not a git repo with forgotten file-name. <gabber>unmatched-paren: but the source of the package does, right? and when the version's a part of that (which it often is) <nckx>If it's part of the basename (part after last '/') yes. <nckx>It's true that the 'version' as a discrete value does not get hashed. The file-name does, and since it includes the version, the effect is the same. <nckx>Anyway TL;DR you did not find a bug.