IRC channel logs
2025-08-13.log
back to list of logs
<jasom>I just installed guix for the first time, and guix pull is failing with an unexpected EOF. Trying to manually clone the guix git repository over https shows a GnuTLS error when downloading the pack: (error: RPC failed; curl 56 GnuTLS recv error (-9): Error decoding the received TLS packet). Any suggestions? <f1refly>I'm trying to launch `hyprland-share-picker`, but it crashes with "Cannot mix incompatible Qt library (6.7.2) with this library (6.8.2)". I *think* I have 6.8.2 installed, does that mean that the repository version of the share picker is build against 6.7.2? <andreas-e>f1refly: It works for me on a Guix from two days ago. Maybe you need to run a "guix pull" and "guix package -u" first. <andreas-e>The error message is a bit surprising, since the binary is wrapped with a path to the correct libraries. So I hope you do not set LD_LIBRARY_PATH? This is bound to break everything everywhere. <f1refly>no, I don't set that. I'll upgrade and try again. <f1refly>while upgrading I see that guix is bringing a whole bunch of qt5 libraries into my profile, is that bad? <andreas-e>f1refly: I do not know. They should be able to live peacefully together with Qt6 libraries. <jakef>andreas-e: At the risk of going to the well one too many times, can I tempt you to look at another ITK-based package? xD <andreas-e>Ah right, I have successfully repressed this one... <jakef>hey it could be worse... it could require an older version of ITK! <andreas-e>Hehe! But it is a complex package recipe, so it requires time to look at. <jakef>oh for sure. take your time :) <andreas-e>jakef: Line 1810 and following of your patch, you should be able to use "let*", in which every next variable is computed from the previous one. <jakef>i couldn't figure out how to do it with the gexps... e..g #$(package-version itk) or (package-version itk) don't work <jakef>i'd appreciate if you could leave a comment that shows how to do it <andreas-e>Okay. I do not know, maybe I am wrong. I will give it a try later. <andreas-e>jakef: I think it might work by ungexping already the let into #$(let*. But I am not sure if this will not choke on substitute later. <f1refly>updating my profile fixed the error :) <andreas-e>jakef: It is interesting to look at the builder script. For simpleitk it currently contains "(let ((itk "/gnu/store/8nnl6jzrnp7nqwpcnk32mgfmmgj92aqx-insight-toolkit-5.4.4") (itk-version "5.4.4") (itk-version-major+minor "5.4")) (substitute* ...". So creating the derivation computes all the values. <f1refly>I have another question: I want to use greetd with its agreety greeter to launch hyprland on login which works fine. In the manual, I read that it should be able to automatically unlock my gnome-keyring on login when I add greetd to the pam entry. The fragments in my services look like this: https://paste.debian.net/hidden/e852ced9/ . Any idea what could be going wrong? I checked the pam.d/greetd in <f1refly>my systems profile and the pam_gnome/keyring.so entries indeed appear there, but there is no keyring being started... <jakef>andreas-e: so that's what we want? <andreas-e>The result is good! I was just wondering whether we could simplify our package recipe. But I cannot test yet, since I am still compiling insight-toolkit... My first idea of ungexping here and gexping there probably does not work at all. But I would like to give it a try nevertheless. <andreas-e>Does anybody else encounter an "implot problem"? <andreas-e>ice-9/eval.scm:293:34: error: implot: unbound variable <andreas-e>(The fact that it happens in electronics.scm might be random, I was testing a package by csantosb.) <andreas-e> 3 (primitive-load-path "gnu/packages/emulators" #<procedu…>) <andreas-e>./pre-inst-env guix build implot works, while interpreting the .scm files that I have not made. <csantosb>... just did a clean rebuild of all .go files, and it is gone <ekaitz>andreas-e: that kind of error oftentimes happens because of a syntax error in the guile <ekaitz>i find those a lot with googletest <andreas-e>I think it is related to dolphin-emu including the package source of implot. That is where the emulators module comes in. <andreas-e>csantosb: For me it was also gone with a make-clean go, but then it came back. <andreas-e>And the problem with "guix build implot" is with the installed/pulled guix. So there is a problem beyond my checkout. I will open an issue. <andreas-e>csantosb: From when does your guix date? The dolphin-emu commit that I find suspicious is from July 24. My pulled guix from August 10. <andreas-e>But "./pre-inst-env guix build implot" also works for me, strangely. <andreas-e>csantosb: Here we go again. I cherry-pick your commits from 1619, and "make" fails with an implot related message. Then I remove the two packages from emulators.scm, and "make" passes. Then I add the emulators packages back, and "make" recompiles the module without touching the electronics module, and I am back to working. Very strange! <ieure>andreas-e, Sounds like an import cycle. <KK>just wondered if its possible to install packages via the guix config file upon first instal <ieure>KK, It's not really "installation," but you can definitely have packages available on the system by putting them in the packages field of your operating-system. This is not really recommended, though. <KK>why is it not recommended? also how can i do it? ive tried doing some methods but they put up errors <andreas-e>KK: I do it for packages that I want every user on the system to have, like an editor. <ieure>KK, As I said, put the package in the packages field of your operating-system. <ieure>KK, It's not recommended because Guix is intended to manage packages per-user, so users can choose whatever mix of things they want. So having them in the system / available for all users undercuts that. <andreas-e>ieure: Yes, I thought of import cycles as well. But did not spot any. implot comes from the toolkits module, which imports almost nothing. At least none of the modules that use implot. <KK>how do i add the packages part? also there will only be one user from the system <ieure>KK, See "operating-system reference" in the Guix manual. <andreas-e>KK: So it depends on who the users are on your system. For me it is "root" and "andreas-e", and both need an editor, for instance. <ieure>KK, It is still not recommended for single-user systems. The only way you can upgrade packages is with `guix system reconfigure' if you put them in the system. <KK>upgrading wont be an issue, its moreso a single install from the config since I'll be removing the networking modules alongside <KK>I will look into it, Thanks guys! <jasom>Rutherther: thanks for asking; things are working this morning; I first tried https://git.savannah.gnu.org/git/guix.git (which is what `guix pull` defaulted to), then noticing the topic in here tried the codeberg repo. Turns out codeberg had a 20 minute outage last night right when I tried that. Things are working fine this morning. <jasom>Reading the guix pull docs, it lists where the channels come from with item #4 being "the built-in default channels specified in the %default-channels variable." how does one determine the effective value of e.g. %default-channels? <Rutherther>jasom: you can use "guix repl", specifically ",use (guix channels)" and then ",p %default-channels" <jasom>Hmm, guix repl is giving me an undefined symbol when trying to run guile; I'll wait for the system reconfigure to finish and try again... <csantosb>andreas-e: Here it comes again ... ice-9/eval.scm:293:34: error: implot: unbound variable <andreas-e>csantosb: I have it every time I touch eletronics.scm, so it is not surprising you should see it! Then I open emulators.scm in my editor, delete 341 lines from line 424 one, save the file; "make"; undo the changes in the editor, save and "make" again. <andreas-e>A bit lengthy... But it has the merit of working. <andreas-e>I do not see the connection between the files. <Rutherther>jasom: yeah, that's strange, it definitely shouldn't happen... I assume LD_LIBRARY_PATH is empty, right? <jasom>Rutherther: I did't check, but it's a nearly fresh install. After the system reconfigure finished, the problem went away. <ngz>Hello. I’m currently trying to create a PR with Agit workflow for two patches sitting in a dedicated branch named "giac" in my local tree. I tried `git pr-push giac -o "giac-update"' with pr-push script being in my .git/config, but pre-receive hook declined that. What could I be missing? <ngz>For the record, the pr-push alias is the following: pr-push = "!f() { branch=$1; shift; git push origin HEAD:refs/for/$branch/$(git branch --show-current) \"$@\"; }; f" <Rutherther>ngz: there is no giac branch, so I don't see how could refs/for/giac work? <Rutherther>there is no giac branch - as in the codeberg.org/guix/guix repo I mean <ngz>I also have the 2 commits on top of master, if needed. But I thought it was possible to push a branch and create the PR. Hmmm. Odd. Anyway, could you refresh my mind about the process? <Rutherther>are you a committer? You could need to first create the branch giac and then make PRs to it, you cannot make PRs to non-existent branches, agit workflow or not. Why do you want to create PR against giac in the first place, why not master? <Rutherther>how would forges know what commit to base the branch from in the first place? They cannot just create branches themselves <ngz>I don’t absolutely want to create a PR against giac. It was just more practical for me to stuff the commits there. But as I wrote, I also have them on the local "master" branch, lagging by a few days behind "origin/master". <Rutherther>I don't know how it is revelant on what branch you have them locally <Rutherther>if you don't want to create PR against giac, why are you putting giac as first argument of pr-push, going into refs/for/giac - saying you want to create PR against giac <ngz>So, you’re saying that git pr-push master -o "whathever" will work. But then, how do I specify it should be commit from local "giac" branch? By checking it out first? <Rutherther>as you can see in your alias, you're pushing HEAD, so yes, it needs to be checked out <ngz>Ah! I created PR 1982. The title is silly, but at least, it contains both commits. Thanks. <ngz>(I should have used -o "better title" I guess) <ngz>err -o title="better title"