IRC channel logs
2025-05-29.log
back to list of logs
<vagrantc>takes maybe 40 minutes to actually build the kernel... <civodul>noe: that deletes the “review” (message) though, not the assignment <civodul>we certainly don’t want to delete all the messages <noe>There is a delete button on review assignments in the UI <civodul>then you want DELETE … /requested_reviewers i guess <noe>Well, you want it because I don’t have the permissions for that :P <noe>But yeah looks right <civodul>noe: question is, on which PRs would we do that? <civodul>also, if you come up with a code snippet for that, i can give it a spin <civodul>(right now rebooting berlin which seems to have hit the shepherd bug that bayfront hit earlier) <noe>Any that was in the range of the CODEOWNERS file, so about all? <noe>Yeah I’ll add it to the teams file and send you the patch <noe>Its public holiday today after all :) <civodul>this will probably be tomorrow though; if i’m unavailable, feel free to ask another “owner” <meaty>Is there a resource I can refer and cite to determine whether or not a non-text file is "kosher" to be redistributed? e.g. CBOR files are human-editable and understandable with foss tools, but they're also "binary" files <meaty>I think it'd be okay if the interpreting software it's meant to be used with is also available? maybe? <noe>Nice, can you remove the review requests? <noe>and… codeberg is down <civodul>… maybe there’s a hotfix being deployed? <civodul>and no, fj.el doesn’t show reviewer assignments <noe>oof, I guess we will see tomorrow <mra>yep, it seems to be back <mra>hmm, are issues being migrated from debbugs? I can't say I've kept on top of the migration news <ieure>mra, There's no automated process to move bugs/patches. <mra>alright! i have a long train ride tomorrow, so i'll move over the stuff for mfs :) <noe>civodul: we’re cooked. The api shows no requested reviewers for these issues, and you need a name to delete them. But the name was deleted. <getstate>Hi, I can't reconfigure my system anymore, and guix doesn't find the configured channels (don't know when this broke). <noe>getstate: what’s the error? <getstate>The channels show up in the current manifest, in `pull list-generations`, but not `guix describe` <getstate>it's `no code for module (nongnu packages linux)` <noe>So, guix describe says you’re on guix without the nonguix channel, right? <civodul>noe: perhaps you could summarize the situation in the issue and ask if they have an ETA for the fix? <civodul>ACTION sorry to have caused this mess <noe>getstate: try guix pull --switch-generation to one that shows the right channels <noe>civodul: I will, also a post to guix-devel would be a good idea IMO <getstate>noe: switched back to an earlier one, same thing. Although the current generation does show all the channels. <noe>Are you using sudo for reconfigure but not for describe or the other way around? <getstate>Noe: Yes, the sudo is because I'm trying to use `guix system`. <noe>But what does sudo guix describe say <getstate>same thing, just the guix channel alone. <noe>Try sudo -E, this should keep your environment variables and use the user’s guix <Tadhgmister>I feel like you need to do a `guix pull`, just having the channels setup doesn't help if you have never pulled from them <civodul>‘sudo guix’ also uses the user’s guix <getstate>I have pulled, it was working until a while ago. <civodul>‘guix describe’ and ‘sudo guix describe’ should print the same thing <noe>civodul: fix is “Tomorrow, no promises” <Tadhgmister>then what do you mean by the nonguix channel shows up in the manifest? <getstate>it shows up in ~/.config/guix/current/manifest <Tadhgmister>in the manifest it contains a pointer to the data in gnu/store right? might be worth trying `-L /gnu/store/....-nonguix/share/guile/site/3.0/` to explicitly give the path to the modules and see if that works <Tadhgmister>not at all a permenant fix but I know I've gotten very used to doing these kinds of one-off fixes since I started putting my channels in my home-config <hanker>can I change the exact timing of the `file-database-update` and `package-database-update` timers? <civodul>Guest70: it’s a fallback in case it’s down <getstate>Tadhgmister: It worked, but it did not fix the issue <apteryx>yay! /gnu/store/10janvrpvhfarr4mk44xgcsf7d6sjp0m-git-repo-go-0.9.0 built successfully, with full integration test suite <apteryx>bayfront appears to be struggling to serve the website <apteryx>Is Ubuntu the only github hosted runner options for GNU/Linux? <apteryx>ieure: hi! good news, I have access to berlin now and the certs are still working (thanks andreas for having comfirmed it on their end) <podiki>apteryx: where can i find how to generate/add a certificate for cuirass? last i looked i didn't see it in the docs <podiki>i take it back, the directions are there but in the cuirass package output i don't see an "etc" directory or those scripts <podiki>just in the source, maybe something to be included or clarified in manual <apteryx>podiki: hey! I should document this in the infra-handbook.org in the guix-maintenance repo, but one has to run 'sudo -E /etc/ssl-ca/new-client-cert.scm name' on berlin to generate these. <apteryx>do you want one? I can send you one by encrypted email. <meaty>is there a way to set environment variables or start processes with a manifest file? <apteryx>manifest.scm can be any arbitrary code, but it's not always evaluated because of caching, so you can't use it for side effects, at least reliably. <podiki>apteryx: thanks, i do have one for berlin already! was looking for another channel. the scripts are in the source, but not in the cuirass package output <icy-thought>Question, do you find yourself building a lot of packages locally while using guix or not? Because I am missing some packages and it's making me a bit hesitant regarding the switch. Laptop is in a terrible state and building locally is not an option unless I want to sacrifice some programs <apteryx>I'm amazed how inefficient it must be for Github to run their CI runners. Basically nothing gets cached, unlike if I was to use Docker or something similar. <apteryx>running the same workfile 100 times results in 100 times downloading the guix binary/installing it <podiki>sounds like they could use some reproducibility focus ;-) <apteryx>maybe the caching is hard to observe; I had a new run where installing guix took just 23 s, so I'm assuming some steps were cached, but the output doesn't seem to suggest that <meaty>any rust-heads in the mood to help me figure out how to regenerate the binary files in hypher (https://github.com/typst/hypher) ? I don't know rust very well and I don't want to bother them by filing an issue if it's something obvious <meaty>there's a file called "./tests/generate.rs" which I thought would do the trick, but removing tries/* and doing 'cargo test' seems to reveal that it's actually to generate code from the binary files <efraim>meaty: it looks like the cargo release tarball removes the pattern directory and the binary files in tries/ get generated during the test phase. certainly not an easy one to fix <meaty>efraim: so running 'cargo test' from a git checkout would generate them? I have tried running it from a cargo release and it complains about missing tries/*.bin files... or perhaps it is that generate.rs does in fact generate the tries, but as a test it does not run until the main lib is built, which requires the tries to be there...? <meaty>what a strange design choice <efraim>I'm not sure running the tests before the build would be enough or if we'd need to copy the test script to build.rs and adjust it but it's certainly not a great situation <meaty>efraim: do you think there is any convincing them <meaty>if they get "miscompiled", then can't there be a more rigorous way to compile them? idk, i don't want to be arrogant seeing as idk rust too well <efraim>I think they'd be open to walking us through manually rebuilding the files but unlikely they'd switch back to building them automatically <meaty>so, file an issue asking for documentation on how to do it? <efraim>yeah, that's probably the best option. And then we can figure out what to do after that <mra>filed a pull request to resolve 41602! really enjoying the migration thus far <mra>forgot to shallow clone and was promptly reminded that the train's wifi is not that good lol <ruther>Guest70: there isn't really a concept of package maintainers. There are only teams that can have a certain file under them. But this file is not under any team. <yelninei>gcc-final for x86_64-gnu failed again after 4.5h with qemu just becoming unresponsive. Trying again with a bit more disk space <ruther>Guest70: so the most you could do is write an e-mail to the author to find that out, or make an issue on codeberg and maybe someone will know <ruther>Guest70: but imo it is qiute likely it is not supported either because dependencies for that are not packaged, or that the author didn't really care abou the use case to package it with that in support <yelninei>i feel like my guix machines are slowly turning into hurd vm launchers <Guest70>Is there anything I need to do before creating a merge request to the guix repo? <ruther>Guest70: apart from the usual of having an account on Codeberg? No, anyone can submit PRs <old>I wonder why Guix is depending on pkg-config. Can't it parse itself any *.pc files? <flypaper-ultimat>old: because its configure.ac uses PKG_CHECK_MODULES to check for guile ? <old>that would make sens <untrusem>so I have added `emacs-next` and `emacs-magit` in my home config, but the packages don't get loaded in emacs, when I run `guix-emacs-find-autoloads` it shows `guix-emacs-find-autoloads: Opening directory: No such file or directory, /usr/share/emacs/site-lisp` but I Install `emacs-next` and `emacs-magit` in default non-home profile, its gets loaded even though `guix-emacs-find-autoloads` shows the same thing <flypaper-ultimat>untrusem: 1. how do you run `guix-emacs-find-autoloads'? 2. when you run in a shell (that has your home config active) whats the value of EMACSLOADPATH ? <flypaper-ultimat>3. could you share your home-config / a minimal bit of it ? then i could help test out. <untrusem>its my home's profile/share/emacs/site-lisp and same path for my normal profile <untrusem>ohh yes, gimme a min, I will push to repo <flypaper-ultimat>a good way to debug would also be doing =guix home container YOURHOMECONFIG=, then you could check if something from outside of your config (env vars, files) is effecting it <flypaper-ultimat>ah, btw since recently, guix home can take an operating system configuration as well, and it would lookup the home configuration associated with your user, so if you like to have things in one file, see (info "(guix) Guix Services") <untrusem>installing magit via guix also solved a major problem I was facing <sham1>guix home being able to take an operating-system and just fenagling the guix-home-service-type out of it seems like a very neat capability. Would make configuring systems a breeze, since things could be self-contained <untrusem>what approach do you recommeng to install rust in guix, through rustup package or installing rust and cargo separately, of course all of there will be installed via guix <yelninei>my childhurd running on core-packages-team completed the build with the extra space in 3.5h, the one on master has troubles with random hangs. Next trying to build the derivation from master on the core-packges-team machine <janneke>yelninei: while it's great to have childhurd to be fully functional on master, isn't spending [most] effort on core-packges-team more effective? <janneke>ACTION still hopes we can "merge" core-packages-team fairly soon <yelninei>janneke: yes but the effort just amounts to pressing enter and running it in the background so its not that big of deal (apart from "fixing" the host-hurd64? hack) <yelninei>janneke: I encountered a broken {i586,x86_64}-guix-gnu-g++ in gcc-boot0 several times, it would just be indentified as 'data' and then unable to compile libstdc++. Have you seen this before? <ghodawalaaman>I want to install guix on OptiPlex 3050 but don't know if the libre kernel will support it or not <ghodawalaaman>I tried it but it just hung up on "setting ethernet/internet" <luca>using ethernet with a cable? That usually works when wifi doesn't in my limited experience <ghodawalaaman>Windows is installed on that machine currently but if I can figure out how to connect it to the internet I will be switching to guix <ghodawalaaman>luca: yeah same here, but for some reasons ethernet isn't working in OptiPlex 3050 <ghodawalaaman>I can live without sound card and wifi card but internet is essential <yelninei>hmm, sadly i had to delete the evidence to continue. It was usually fine after retrying but a bit concerning that I encountered it multiple times now <apteryx>I think I managed to create my first forgejo PR with the git-repo-go tool <csantosb>apteryx: great ! what is the advantage with respect to a 'git push ...' ? <jakef>should we resubmit PRs that were affected by the teams label issue, or are they going to be recovered? <noe>jakef: The patch was merged in forgejo, so I think it will be working soon <noe>e.g. the PR will be recovered <yelninei>janneke: I have hello on core-packages-team without any additional changes \o/ <nico_>On Guix System, how can I create icons for installed flatpak applications? <ruther>nico_: you don't 'create' icons, you use existing icons. You just need to tell your app launcher to use them by adding the flatpak share directory to XDG_DATA_DIRS... <guix-newbie>i'm trying to update neovim to 0.11.1 and am encountering issues finding a luajit module. The configure step is looking for luajit-v2.1-0.538a821/share/lua/5.2/jit/bcsave.lua but the file is at luajit-v2.1-0.538a821/share/luajit-2.1/jit/bcsave.lua <luca>Hi, when running `guix shell swaynotificationcenter -- swaync` I get the error `Settings schema 'org.erikreider.swaync' is not installed`. Anyone got any tips as to how I could get around this? <ruther>luca: the package is packaged wrong, it doesn't have wrapped binary. You need to set GSETTINGS_SCHEMA_DIR to its schema dir - under share/glib-2.0/schemas... <luca>Hmmm ok. Should both swaync and swaync-client be wrapped? <podiki>i use swaync without issue, though it is not through a service; no GSETTINGS_ set, though maybe it gets it instead from an XDG_ env? <luca>My understanding is that if you install it (guix home or guix install) then it does work. But if you `guix shell` or run the file directly it won't <luca>And I'm now going on a rampage of shepherd-ing all the things <yelninei>maybe the glib-or-gtk? option of meson-build-system is enough? <ruther>podiki: yes, XDG_DATA_DIRS and default paths are going to be used. So naturally when you install it it will work. But still it is packaged wrongly, it should work without installing... <ruther>this is what the glib-or-gtk-wrap phase is for <podiki>presumably then it will work also with e.g. glib in the profile? <podiki>(something to set XDG_ in the profile) <ruther>podiki: but the package itself should work outside of profiel <podiki>i think that will be true of nearly anything that uses XDG_ env <ruther>it is not, because most packages like this use the glib-or-gtk build system <ruther>where it gets wrapped automatically <podiki>ah, and this one doesn't despite needing gtk stuff <podiki>so yes, adding in that phase would seem to be a good idea <ruther>but it does use meson which starts with glib or gtk, so I am not sure why it's not wrapped in the first place <podiki>i'm trying to remember, is this one of those phases that looks for certain files to decide if wrapping maybe? <yelninei>ruther: if #:glib-or-gtk? is #f (the default) the wrap phases are removed in (guix build-system meson) <ruther>yelninei: oh... interesting, I didn't know mechanism like this - deleting phases based on argument - was a thing <podiki>are we able to force push branches on codeberg? or is that disabled? <ruther>podiki: someone already mentioned here they could force push <podiki>oh nice; though hopefully disabled on master branch? <ruther>that I saw thanks to ieure, who wasn't able to force push a fix to master <ruther>and Ludo mentioned in their e-mail that the protection had to be disabled for a force push <podiki>probably just "owners" group can do that or change that setting <podiki>i feel like we never ended up doing anything with the change-ids unless that is used by QA? probably not needed with a forge now <luca>At least it works locally for me /shrug <podiki>indeed, if that works, that's all there is to it :) <podiki>i don't have a codeberg workflow set up just yet, but i'll leave a quick comment <luca>What's a changelog in this context? 😅 <podiki>just describes what the commit changes, it is a particular format <podiki>if you search the commits for glib-or-gtk you find some similar ones, easy enough here <luca>Ah so just in the body of the commit message <podiki>and if you use emacs (and i think vim), there are helper snippets for simple things like package updates <podiki>hmm i wonder if the mirroring to savannah will handle the force push to a branch (so far hasn't updated) <luca>Ok, I can dig this whole codeberg thingymajig <podiki>i'm not there yet, gotta figure out a new workflow but at least i can comment on issues on the web :) will need to try out the emacs interface next <podiki>hmm maybe i should delete branch on savannah manually <luca>Well I'm no maintainer so I don't _do_ anything. But from a pure from-patch-to-push it's much faster than email. Really impressive <podiki>oh wait, no one has rights to savannah anymore i htink <podiki>glad to hear it! that's the idea to get more drive by and quick contributions <ruther>podiki: definitely not no one, but it's just a few people, yeah <podiki>right, i mean us regular folks :) <luca>Originally I wanted to contribute more to guix (and learn email workflow), but I put it off for so long that guix switched away from email :P <podiki>sneek later ask civodul: does the codeberg->savannah mirroring handle force pushes to branches? namely for mesa-updates which i did on codeberg (updates and rebase) <podiki>haha that reminds me when i put off stuff for so long they get done/no longer relevant...often works out nicely <luca>I wanted to learn guix so I installed it on an old laptop, then that laptop became a gaming server by accident and I couldn't touch it for a few months. Really slowed down my learning experience :P <guix-newbie>I managed my first significant pull request: updating neovim to 0.11.1 <ieure>guix-newbie, It gets a little easier every time! <podiki>quick notes: will need to be separate commits for each change, and utf8proc could be updated (it has ~330 dependents but i think that's okay directly to master branch) <podiki>(after checking no issues from that update for dependents) <scut>guix-newbie: got my first PR the other day too :) <noe>Did you start contributing because of the move to Codeberg? <guix-newbie>noe: No, I would have use the mail-based system too. I moved from Nix to Guix because nixpkgs is on github. <guix-newbie>I'm a fan of flakes though and have many repos using them, so I'm still using nix, but Guix System works pretty well. Now at least I can help to report and fix. <podiki>their loss is our gain apparently <madage>let's see how great is their chunk of users that have an issue with that <guix-newbie>The main thing I like about flakes is the ability to easily lock to a specific version of nixpkgs in a file in a project's repo. <guix-newbie>so i hope to figure out how to easily do that with scm files <madage>I'm not familiar with flakes, but you can certainly pin package versions on guix <identity>guix-newbie: (info "(guix) Invoking guix time-machine") <meaty>The upstream repo is on 0.13.12, crate is on 0.12.19-rc4... and that is the only version <meaty>I want to package it but there's no point if the crates version is an rc and not even up-to-date with typst <scut>guix-newbie: you do not contribute to nixpkgs because there are on github ? I don't think it's a new thing though :) <meaty>I could package it from git but maybe there's some rust reason why I shouldn't do that? Maybe it uses depencency versions that themselves aren't in crates yet? or did the dev just forget to push a crates release? <scut>(because I do so I only know PR-based workflow :) ) <meaty>on second thought, why is a "release candidate" released, anyways... <identity>meaty: probably they just did not put it on crates.io, it should be fine, and if not, well... <guix-newbie>scut: yes, they've been on there for years, even before Microsoft bought it. But I think it's not good to force devs to use a Microsoft platform to contribute to FOSS <scut>I agree but all the software I use are on Github too :( <guix-newbie>identity: i'm reading the time-machine page. I think it looks good <scut>Hope more people will move to Codeberg <scut>all the software I package for work* <scut>Also, got started with nix, so... <guix-newbie>identity: if a channels.scm has two channels A and B that both package foo, how can the right one be chosen? <identity>guix-newbie: for package variables it is obvious, the package that corresponds to the variable 'foo' is chosen; for package specifications, no idea, but at the very least the one with the highest version should be chosen from packages with the same name