IRC channel logs

2025-02-23.log

back to list of logs

<divya>We really need better debugging information in Guile. It's a pain to figure out where things are going wrong.
<aure84>Hello. I'm using Fluxbox and Thunar (w/thunar-volman), but I can't manage my usb drives. Anyone has experience with similar setups?
<divya>aure84: What's the issue happening with Thunar?
<aure84>I can't see the usb volumes in the left panel
<aure84>they should appear there when the usb is plugged in
<aure84>divya: I'm trying my setup with a qemu vm. I'm forwarding usb through spice and can verify that lsusb in the vm sees the usb device
<divya>Weird. I use thunar with volman and it works
<aure84>do you use xfce?
<divya>No, I use a window manager.
<aure84>any dbus gvfs special setup?
<aure84>do you automount usb drives?
<divya>aure84: Not really, no gvfs shenanigans. You have volume management enabled in Thunar, right?
<divya>And yes, I have a script that automounts usb and other disks when plugged in
<aure84>volume management should be enabled by default, right?
<aure84>have any snippet for the automount part of your setup?
<divya>Wasn't in my case. Check Advanced tab in Preferences
<aure84>do you also get the message that gvfs is not found?
<aure84>on the preferences tab?
<divya>Nope, because I have gvfs installed. You need that
<divya>It's the virtual file system needed
<divya>And here's the bash script I use: https://paste.debian.net/1354211
<aure84>yes, I have enabled the volume management in thunar
<divya>And installed gvfs?
<aure84>yes I have gvfs. Do you have udisk or gvfs-service-type?
<aure84>how about groups? do you remember to have added any special group to your user?
<divya>Wait let me check
<divya>aure84: Nope, no gvfs service
<divya>Nope. No special groups either.
<aure84>in fact, if I don't add (service gvfs-service-type) lsblk does not even list the device in the vm
<divya>aure84: Could be because of the vn
<divya>s/vn/vm
<RavenJoad>Hey Guix! I am packaging something with an OSI- & FSF-approved and GPL-compatible license, but Guix does not have a license entry for it. It is The Universal Permissive License (UPL), Version 1.0 used by the Souffle language's compiler. How should I proceed?
<ieure>RavenJoad, Include a patch adding it, send your package as a patch series.
<RavenJoad>ieure: That's what I figured. I just wanted to confirm that the UPL is actually something compatible with Guix, since I am _very much_ not a lawyer.
<ieure>RavenJoad, Me either, but I assume the FSF does, and if they say it's okay, it is.
<RavenJoad>The GNU site lists it and says its compatible (linking to the FSF site too). GNU says it prefers Apache, but this isn't mine. https://www.gnu.org/licenses/license-list#UPL
<cobra>just downloaded the latest i686-linux image off ci.guix.gnu.org and shepherd fails saying `Unbound variable: vileged`
<cobra>and then kernel panic
<apteryx>fun, avatars on mumi (<https://issues.guix.gnu.org>)
<amano> https://packages.guix.gnu.org has been broken for a while.
<apteryx>hm... in import-html-release, when updating qemu, the directory argument ends up being "/", because in its default computed value, (package-upstream-name package) apparently returns "" instead of "qemu"
<apteryx>I wonder what this could be caused by.
<apteryx>ah nevermind, it gets passed explicitly
<meaty>could anyone help me out with packaging https://github.com/uben0/tree-sitter-typst ? I made a naive attempt using tree-sitter-grammar, but I immediately came up against a wall 'cos npm can't find the right tree-sitter version? See http://paste.debian.net/1354250 for the build log
<rekado>meaty: https://codeberg.org/guix-science/guix-science/src/branch/master/guix-science/packages/typst.scm
<rekado>oh, sorry, does not contain the tree-sitter thing
<meaty>yah
<meaty>sorry I'm not sure how to proceed, tree sitter and node is all foreign to me
<meaty>but I would like to have tree-sitter-typist working to use the only emacs mode for typist
<meaty>which is a tree sitter mode
<meaty>There is no prebuilt binary for it either
<futurile>hmm is bordeaux offline? I'm getting bad gateway to https://guix.bordeaux.inria.fr/jobset/nonguix
<csantosb>Seems so
<nckx>Just to note that's not ‘our bordeaux’. The machine we call {bordeaux,bayfront}.guix.gnu.org looks all right from here.
<nckx>(The jobset name's a bit of a hint, but still.)
<futurile>ah OK - I just assumed they were resolving to the same hosts
<nckx>Yah. s/but still/but still understandable/ is what I meant. It's confusing.
<kimapr>i was trying to troubleshoot an nginx problem (it no longer binding on a specified port but not erroring either) and now even shepherd seems to be broken...
<kimapr>any herd command simply freezes with no response
<simendsjo>Kimapr: I had a similar issue which was fixed in Shepherd 1.0.2
<Kimapr>simendsjo: i see. it seems i should update then as i am still on 1.0.0 apparently
<apteryx>ACTION updates qemu to 9.1.3
<elevenkb>How do you change your username on guix?
<futurile>elevenkb: on guix system you can define your user account in the system config file
<elevenkb>Sure, but then you probably have to at least rename + chown your old home folder right?
<elevenkb>futurile: and reset the password.
<elevenkb>but that's it methinks?
<futurile>elevenkb: most people just reset their passwd manually (e.g. passwd blah) because that's kept between reboots and avoids having the password in the gnu/store. You can also do things like add the user to primary and supplementary groups (I do this in my VM's so that I have the same uid on the vm and on my main system)
<futurile>elevenkb: most systems (e.g. Ubuntu ) default to creating the first user as uid 1000 - Guix does the same by default. So if you just want a different 'username' but the same uid that should work
<futurile>otherwise there's a (uid xxxx) you can do
<elevenkb>alright, reconfiguring rn.
<elevenkb>I had to chown -R my home folder since it was created using my new username as the owner.
<elevenkb>It took 3 minutes or so, so I'm happy.
<meaty>Any good way to sort package definitions in a file?
<meaty>Do I have to bother inserting all my imported packages in sorted positions in crates-io? Is there an automatic way to do it
<futurile>meaty: there's no automatic way to do it, and yes you need to do it. It's annoying!
<meaty>lawd
<futurile>meaty: yeah, also because they're split over multiple files, sometimes there's one hidden somewhere else that we already have
<futurile>meaty: so make sure you do a quick search - I install ripgrep - and then do "rg public name-of-lib" - to make sure it's not there already
<futurile>it might warn you - if you do a build - but I'm not sure
<meaty>word, rg helps a lot--I was using './pre-inst-env guix search' before and it was so slow
<meaty>futurile: do you know, is crates-io a "dump" file like python-xyz or is it supposed to be just for "I/O"
<futurile>meaty: AFAIK it's a place for "misc doesn't fit anywhere else". Efraim split it into the web, networking etc.
<futurile>meaty: I'm not sure how you choose which location it should go to - other than using judgement
<futurile>meaty: if in doubt I'd put it into crates-io.scm and Efraim will move it if he thinks it should be somewhere else I'm sure
<civodul>ACTION formally submitted GCD 002! 🎉 https://issues.guix.gnu.org/76503
<futurile>whoop whoop
<futurile>Can I actually vote for it if I'm not in a team? (or is my vote worth less if I'm not in a team?)
<civodul>futurile: there’s deliberation, not voting :-) and it’s only open to team members
<civodul>but you can surely join one! ;-)
<civodul>also note that we’re in the discussion phase for 30 to 60 days
<civodul>deliberation comes after that
<Rutherther>civodul: how is the deliberation closed in case not everyone agrees with one of the option?s
<civodul>Rutherther: check out https://git.savannah.gnu.org/cgit/guix/guix-consensus-documents.git/tree/001-gcd-process.md
<civodul>essentially we have to come to a consensus
<civodul>so it’s not so much about agreeing/disagreeing than it is about building a solution that addresses all concerns
<df>should I expect emacs-lsp's magical server-install mechanism to work on guile (in this case, for erlang - it claims to be able to install whatsapp's elp, but not the more familiar (to me) erlang-ls)
<df>or am I likely to need the server packaged for guix?
<df>it seems to get as far as downloading *something* to to .emacs.d/.cache/lsp/erlang/elp.tar.gz, but it's not a tarball
<df>in fact it's just a 'Not Found' messsage, so maybe this isn't a guix problem at all...
<redacted>Anybody already got a Synapse homeserver running on their Guix system?
<redacted>Wanted to check if the work had already been done before I make a custom service
<df>actually, looks like emacs-lsp is installed via MELPA, are things likely to run more smoothly if I install it via guix?
<lechner>Hi, what is an ungrafting run, prior to a release, please?
<potatoespotatoes>is anyone here using zfs-on-guix? is there a simple guix-ISO floating around that is zfs-ready?
<acrow>Has anyone else seen their syncthing cluster fail on the upgrade to v1.29.2? Guix allowed me to roll-back to v1.28.1 and things are now ok, but syncthing upstream seems unaware of the problem.
<acrow>Neither do I see any issues here... did I miss something?
<lechner>Also, I think the SSH "Error in push function" is back
<meaty>If I'm adding a new file with a bunch of package defs, should I make one patch for the file and contents or break it up somehow?
<futurile>meaty: one change per patch
<futurile>meaty: so for each package definition you're updating/adding to crates-io, you should have one git commit - and one patch when you do the export
<futurile>meaty: I did a video that _may_ help: https://youtu.be/PSn1_NZzQ7o?si=rN0pZiGdVBYp22Dn
<meaty>futurile: very cool, do you have the slides online somewhere?
<meaty>nvm, found your blog
<ngz>Hello. Using mumi, how can I apply a patch set without the first one, or, more generally, starting at patch N?
<janneke>headsup: i've squashed core-packages-team, merged my glibc-2.40 patches with zheng's 2.14 patches, and rebased
<janneke>about to forcee push
<ngz>It seems I need to cherry-pick them with mumi am @message-number-in-thread
<ngz>Going to be fun
<futurile>meaty: should be accessible here: https://slides.com/futurile/guix-package-updates-and-contribution
<futurile>meaty: I meant to also do a blog post, but haven't actually done one that's "in-sync" with that video
<csantosb>ngz: for single patches, you'll need to stick to @id, but ... for patch series ? 🤔
<ngz>I’m currently applying a 22-patch series calling mumi am @n one by one. My git fu is weak.
<kreijstal>does anyone know how what wrap executable does? starting phase `wrap-executable' error: in phase 'wrap-executable': uncaught exception: system-error "link" "~A" ("No such file or directory") (2) phase `wrap-executable' failed after 0.0 seconds. I am trying to make my own package but it doesn't has this phase, why does it trigger?
<ieure>kreijstal, It's defined by the build-system your package is using.
<csantosb>ngz: according to the manual `mumi am vX` should do it
<ieure>kreijstal, If you need help debugging, this part of the manual may help: https://guix.gnu.org/manual/devel/en/html_node/Debugging-Build-Failures.html
<ieure>kreijstal, Also putting your package definition and output of building it on a pastebin, then dropping the links in here.
<kreijstal>gnu-build-system, alright I'll put my package definition...
<kreijstal> https://paste.debian.net/1354341/
<ieure>kreijstal, The wine packages you're inheriting from have 'wrap-executable phases.
<ieure>kreijstal, When you're updating packages already in Guix, it's much easier to clone the Guix repo and hack on that, following this part of the manual: https://guix.gnu.org/manual/devel/en/html_node/Running-Guix-Before-It-Is-Installed.html
<ieure>kreijstal, You can make a cheap clone of the checkout you already have, with `git clone ~/.cache/guix/checkouts/pjmkglp4t7znuugeurpurzikxq3tnlaywmisyr27shj7apsnalwq/ guix-work'
<kreijstal>so I just have to constantly rebase it?
<kreijstal>pjm... is the lastest head?
<futurile>ngz: you might be interested in trying out `b4` at some point - it can download the 'latest' versions of a patch series, and you can also have it apply just some in a series etc - https://libreplanet.org/wiki/Group:Guix/GuixSocial#6._Find_the_patch_to_apply
<meaty>How can I manually send a patch series to a certain branch?
<ieure>kreijstal, It's a fixed identifier for the Guix repo, I'm not sure how it's calculated.
<kreijstal>how do I know which checkout I have?
<meaty>I make a feature branch off of rust-team, but when I go to send it in it's missing the 'rust-team' tag and the team doesn't get put in CC header
<ieure>kreijstal, I don't understand your question.
<kreijstal>the thing that makes me not want to do that is that I don't want to rebuild the universe... won't it be forced to rebuild all packages?
<kreijstal>if I have a guix fork?
<meaty>kreijstal: No, not unless you've actually changed the packages
<ieure>kreijstal, We're talking at cross purposes here. What I'm suggesting is that you make a clone of the repo to create patches updating wine, which you can then contribute.
<ieure>kreijstal, If you want to run newer Wine while waiting on the patches, you can copy-paste the updated ones from gnu/packages/wine.scm somewhere you can use them however you like.
<kreijstal>ieure, what I meant is how to get the pjmkglp.... string? so that I can clone my current checkout how can I know this? How did you know it.
<kreijstal>or how can I find this string by my own?
<ieure>kreijstal, As I said, it's a fixed identifier. It's the same for every Guix install, as far as I know.
<kreijstal>oh
<kreijstal>it doesn't change when I do guix pull?
<ieure>kreijstal, I opened ~/.cache/guix/checkouts in a file manager and poked around.
<ieure>kreijstal, It doesn't.
<kreijstal>ahh... weird
<ieure>It's an identifier for the repo, not for the *state* of the repo.
<ieure>The HEAD inside that clone changes when you `guix pull', the path to the repo doesn't. Just like any other Git repo.
<kreijstal>alright and after doing a cheap clone of this modified guix I would just do the "-L ." trick, correct?
<kreijstal>to build my custom package
<ieure>kreijstal, No, as I said, follow this part of the manual: https://guix.gnu.org/manual/devel/en/html_node/Running-Guix-Before-It-Is-Installed.html
<futurile>meaty: are you asking how you cc the rust-team when you send your patch series to debbugs?
<meaty>futurile: I think so? I want the subject lines to start with [PATCH rust-team XX/XX] instead of just [PATCH XX/XX], and to have the rust-team cc'd
<meaty>for some reason it isn't working automagically
<ieure>meaty, `git format-patch --subject-prefix="PATCH rust-team"'
<ieure>Or slum it, `git format-patch -o .', then edit the patches and `git send-email *.patch'
<futurile>Do the format-patch export like ieure says. You want to create a cover letter (--cover-letter) first, and then send that to guix-patches@guix.gnu.org. That will cause debbugs to send you a bug number.
<janneke>sadly, the "evaluation failed", whatever that might mean https://ci.guix.gnu.org/eval/2043375
<futurile>Look up the bug number on issues.gnu.guix.org - and find the message id - you're going to do a new export - using --message-id=<message-id-you-got-from-issues>
<meaty>futurile: Will that get to the rust-team correctly?
<futurile>no - one more step
<futurile>to get the rust team do: ./pre-inst-env etc/teams.scm cc rust
<kreijstal>I have to compile guix to build 1 custom package if I edit the guix tree or am I doing it wrong
<kreijstal>cuz I am doing ./configure && make on guix tree
<ieure>kreijstal, Yes, you compile Guix once.
<kreijstal>seems it is a lot of effort for one custom package but alright
<futurile>and then when you send your patches you do: git send-email --to=NNNN@debugs.gnu.org --cc="person@rust-team" --message-id=<message-id-from-issues>
<futurile>and if anyone thinks this is straightforward they are absolutely mad - but hopefully we'll have pull requests before we all die of email exhaustion
<ieure>kreijstal, It's not really a custom package, it's a newer version of something already in Guix. If you're packaging stuff for the very first time and don't intend to contribute back, your initial workflow is fine.
<ieure>kreijstal, Also, once you have that Guix compiled, you should try `guix refresh -u wine' and see if that's all you need.
<ieure>Did something change with issues.guix.gnu.org recently? The open/closed status is no longer indicated on anything: https://issues.guix.gnu.org/
<kreijstal>I can defintely contribute it, but I doubt it will be merged, I thought having custom .scm files would make it more accessible rather than less. anyway I'm excited to see if I can reproduce bug on full guix system
<ieure>kreijstal, Why do you think it wouldn't be merged?
<kreijstal>first 10.2 is dev not stable version, also I want to add custom patches to run cygwin programs on top of it.
<ieure>kreijstal, Ah, I see. If the patches are Free, those might get accepted. Guix has plenty of snapshot software in it, but in the case where there's also a stable release, will generally keep the snapshot in a -next package.
<ieure>Worth a shot, IMO. But also hacking directly on the packages (or adding new wine-next packages in the same file) will make it easier to see things like the 'wrap-executable phase you were confused about.
<abyss>hello
<futurile>hello abyss
<ieure>kreijstal, You also have the option to contribute wine-next patches to update it to 12.0, then keep a private package which inherits from that and adds your cygwin patches.
<superkamiguru>Had a quick question, does anyone have experience building an aarch64 guix image using grub-bootloader as opposed to grub-efi-bootloader (BIOS boot vs UEFI if I am not mistaken)
<kreijstal>ieure: hmm managed to get this working on a guix clone... it reproduces the exact same error :) just letting you know
<kreijstal>the crazy thing is that I can't see the definition of wrap-executable `git grep "wrap-executable" |grep -v add-after`