IRC channel logs
2026-03-20.log
back to list of logs
<dajole>I'm having a silly simple Guile / Guix problem I seem to be too dull right now to figure out. I want to prefix a package: `((gnu packages dns) #:prefix dns:)` because I want the `bind` package, which name appears to collide with something else. But when I try to use the prefixed package `(list dns:bind "utils")`I get an unbound variable error. <gabber>dajole: i'm not sure i understand <gabber>the variable with that package is calles isc-bind <dajole>😅 now I'm not sure I'm following. I'm trying to install the package bind, which provides e.g. dig. But when I include it in my home config without a prefix I get an error that sounds to me like bind is being treated as a procedure. <bdunahu>Looking at the package definition, the package name is 'bind', and the symbol name is 'isc-bind', to avoid conflict with a built-in Guile procedure <dajole>Ohhh, so it's called isc-bind to avoid exactly the collision I was running into. Okay. <dajole>Thank you. How would I have been able to tell? guix show bind doesn't seem to mention thsi <bdunahu>I'm not really sure myself without looking at source <dajole>Thank you both! I'm still learning the various problem solving strategies with guix; seems like I should definitely start looking at the source more. <dajole>For looking at the source, is just `guix edit bind` the way to go? <FuncProgLinux>Hello. Does anyone know if the go-build-system has issues with network tests in packages? :( <ekaitz>FuncProgLinux: what do you mean by network tests? the tests accessing the network? <ekaitz>i think that's not possible available <FuncProgLinux>ekaitz: I see. I was testing a net/http thing I did and the network tests were failing on the 'check phase on Guix. I've did #:tests? #f and everything works. <FuncProgLinux>I just found it odd that It wouldn't resolve paths on a new http.NewServeMux() route inside guix while the pipelines and the local tests weren't suffering from that when mocking a server for tests <ekaitz>i don't know if there's any way to enable that for specific cases but the build container is isolated, so yeah... <FuncProgLinux> app_test.go:201: resolved pattern="" handler=http.HandlerFunc (the tests are "offline", no actual network access is needed, it uses mocking) <FuncProgLinux>idk why resolved pattern="" stays empty in guix. But it would be nice to have a way to "shush and build" without disabling those tests :s <ieure>FuncProgLinux, All Guix packages build in an isolated environment with no network access. <ieure>FuncProgLinux, Put the build log in a paste? I strongly suspect it's actually making network requests. <YAR_Oracool>I need some tutorials to make and maintain packages because some of the stuff I want... don't really exist on guix <ieure>YAR_Oracool, Guix is not big on "tutorials" or "guides," but you can find reference material in the "Defining Packages" section of the manual; and the packages in the Guix source are a good reference. <YAR_Oracool>Hmnmm.. .that'll be tough but I give it a shot. GUIX manual is suprisingly well made. <YAR_Oracool>I'll upload it to a repo upon success so everyone can use it. <ieure>YAR_Oracool, Common stumbling blocks: assuming package builds can download stuff (there's no networking in the build environment); assuming packages can drop configuration in /etc (or other arbitrary parts of the filesystem); assuming packages can run code on install (like Debian's hooks) -- there is no mechanism for this in Guix. <ieure>YAR_Oracool, All that said, having been both a Debian developer and a Guix committer: I think Guix's packaging mechanism is both better than Debian's, and much easier to understand and use. <FuncProgLinux>ieure: I suspect the same. I think my mocking implementation is wrong <Gooberpatrol66>i wasn't online when this was breaking, what is guix going to do about the new age verification laws? <ieure>Gooberpatrol66, We haven't discussed it at all either way, but I would be astonished if anyone so much pondered considering the remotest notion of the merest possibility of lifting the smallest finger the slightest degree towards implementing the tiniest fraction of that stupid, misguided, unworkable law. <ieure>My personal opinion is that the law is unworkable and doomed, so there's no point bothering. <ieure>But also it's also stupid and impossible and there are actually valuable things to spend time on instead. <ieure>sneek, later tell yarl I left you a review. Happy to discuss / share examples if you have questions about anything I mentioned. <nutcase>Hi. I have a USB-WiFi-Adapter "Ralink Technology, Corp. RT5370 Wireless Adapter". How do I know, whether it works on a Guix system (without additional channels)? <gabber>nutcase: i think if it's listed on h-node.org then you're fine <NaN23>Hello I'm using guix system, and since recently when I `guix pull` get the following warnings: ;;; WARNING: loading compiled file /gnu/store/qsmcq0c4xd4d36nv4gfah6203bivmdsx-guix-module-union/lib/guile/3.0/site-ccache/gnu/packages/libffcall.go failed: <NaN23>;;; In procedure load-thunk-from-memory: incompatible bytecode version (a lot of these). Also the next pull runs into an error: guix pull: error: linux-libre-support-for-Pinebook-Pro.patch: patch not found <nutcase>although it's listed, it's not fine. h-node tells that it's not compatible with GNU/Linux-libre <NaN23>I did then roll-back my system but cannot update because of this issue. Does anybody have any clue about this? <adanska>is anyone else able to see the ,build `guix repl` metacommand? it disappeared a while ago and now i could really use it... no clue where it went or if something is wrong <csantosb>Morning ! I'm trying to figure out how python-nbsphinx could broke up <futurile>csantosb: did you find any ways to get information out of the build system so you know if things are working/failing in the end? <futurile>csantosb: I take it python-tinycss2 <--> python-nbsphinx-0.8.8 are connected in some way? <csantosb>Through nbconvert, guix graph --path python-nbsphinx python-tinycss2 <csantosb>No idea what this "ERROR: nbsphinx==0.8.8 ContextualVersionConflict(tinycss2 1.5.1" means <andreas-e>Well, our nbsphinx dates from 2021... And there have been 13 releases since that time. So we cannot really complain. <andreas-e>Almost no dependencies, so a simple update of the package (and its dependencies?) could work. <csantosb>Not that easy, I'm afraid; upgrading to 0.9.8 still fails. <csantosb>I'm annoyed because building the upgrade to tinycss2 with -P1 didn't suffice to detect the breakage <futurile>andreas-e: I tried working on Ruby a bit, but the new 4.0 release have a reproducibility problem. I don't have an idea how to resolve it. I've pinged cbaines in case they have an clue from previous experience (#5166). I guess I'll work through some of the libs that gemmaro did <andreas-e>This is really strange! It looks like a side-effect of propagated dependencies; maybe nbsphinx depends directly on tinycss, but through propagation that relationship is hidden. In which case it would be an undetectable error in the nbsphinx package. <andreas-e>futurile: Nice! From time to time I pushed a few ruby updates, since they are usually clean and easy. But the recent megaupdates are just too much work for me. The ruby team is not active enough. <csantosb>andreas-e: Someone requires tinycss2 <= 1.5.0, but who ? <andreas-e>I just saw a good case for short waiting periods when removing packages: The update of python-angr exhibited 10 packages that have not been building for between 6 months and 2 years (!), and nobody has noticed it. So now we will have to wait one month for nothing. <futurile>andreas-e: do you have a script/capability to show failing builds or changes in builds on the command line? Or do you use the Web. Asking because it's slow using the Web interface ofc <andreas-e>csantosb: I refused to learn Python when the tutorial I looked at mentioned that one could create global variables anywhere in the code, in an unrelated module and out of lexical scope. So I have no idea how to check. Is it not the .toml file that writes down the version constraints? <futurile>csantosb: I think it's a constraint within the site packages for nbsphinx right? Can you patch the constraint out? <andreas-e>futurile: Here I have just used the web, the output of the cuirass bot on the PR to get the list of packages, and then the dashboard of CI itself to check whether the packages already fail on master and if yes since when. <andreas-e>My experience with CI is that it is close to binary. Either it works, and then it is sufficiently fast; or it almost consistently times out. <csantosb>Greeping for tinycss2 in nbsphinx only returns uv.lock <csantosb>Which is ignore when not using uv, I guess <csantosb>andreas-e: I'm neither a big fun of Python; but the rest of the world is, for unknown reasons (I'm trying to package openfasoc) <futurile>csantosb: did you check whether python-team branch is at the same package levels. They have been doing a lot of updates. <futurile>csantosb: it seems like the underlying problem is the nb-xxx which is going to be one of those big jobs where you have to update a whole group of things <csantosb>Yes, this was my last resort, the up to date python branch <kestrelwx>adanska: I think you have to do (use-modules (guix)) for ,build to be available. <adanska>omg yes that worked!! thats going to help me so much, i appreciate it :) <attila_lendvai>csantosb, hi, have you seen my question about llama-server yesterday? <attila_lendvai>csantosb, i'm new to this, so i'm not sure whether this should work or not. or maybe that thing should be used some other way than browser to localhost:8080, i don't know. <csantosb>attila_lendvai: Still trying to install llama-cpp ... downloading 1GB qtdeclarative <csantosb>Wondering why qtbase (a ui framework) requires cups (a printing system) <csantosb>`guix size llama-cpp` still returns ~7GB <attila_lendvai>csantosb, what do you mean not available? it's in master right? if your guix has pulled it, then it should have reduced the closure, no? <csantosb>It takes time to build in the servers, like a couple of days or so ... <csantosb>Doing a `/gnu/store/7g6jqn2vlikcfjr7cfc9fvmvhmf0mp6q-llama-cpp-0.0.0-b8411/bin/llama-server -hf ggml-org/gemma-3-1b-it-GGUF` now ... <csantosb>And a beautiful white canvas in the browser <attila_lendvai>csantosb, AFAIU it should load a chatgpt like interface, but i get an empty page and the 404 in the console <attila_lendvai>maybe some js lib is not mapped properly to the web entry points? maybe due to the special path magic of guix? but this all is just speculation... <csantosb>Oh, I see ... srv log_server_r: done request: HEAD /cors-proxy 127.0.0.1 404 <csantosb>This look correct "lddtree /gnu/store/7g6jqn2vlikcfjr7cfc9fvmvhmf0mp6q-llama-cpp-0.0.0-b8411/bin/.llama-server-real " <csantosb>Let me try to add python-gguf back to its inputs <attila_lendvai>csantosb, damn, we managed to snapshot on a commit that is broken in upstream? that's unluckly <csantosb>Let's see; I'm building tag "b8445", could you try from your side ? <attila_lendvai>csantosb, i'm busy with other stuff right now, but i started the builds in the background <csantosb>Same for me; I'll push an update later on <bjc>does anyone know if we have examples of things being built with qmake? <bjc>ah, thanks. don't know why i didn't try that <andreas-e>csantosb: Thanks for the denemo update! Very welcome. <civodul>csantosb: heh, i must say i’m happy this one is finally solved :-) <bjc>having spent the last year or so doing nix stuff, it's been a real joy to use scheme again <ieure>Scheme is my least-favorite Lisp, but the Nix stuff I've seen looks atrocious. <NegateThis>I started my journey into Lisps with Clojure, so I'm still getting used to the amount of parentheses in Scheme. But I'll take Scheme over Nix any day of the week <NegateThis>My hope is that one day there will be a Clojure equivalent to Guix <noe>or you could implement clojure in guile <ieure>That's probably challenging, as a lot of the Guix stuff leverages Guile's C compatibility. I think there's a native-compiled version of Clojure these days, maybe that could help. <ieure>But really, my dream Lisp steals all the good stuff from both Clojure and CL. <ieure>Immutable by default, destructuring, maps from Clojure; CLOS, native code, conditions, loop, format from CL. <ieure>I also do like how Clojure handles dynamic variables, and its package/module system is IMO better than CL's. <ieure>At least easier to use without needing to care about systems. <FuncProgLinux>For me I would like something like Guile (kwargs) combined with things from Janet like the distributable binaries and the package system of Go because where's the fun without typosquatting in package registries xD <NegateThis>ieure: There's Carp. Idk if it's actually a Clojure, but it has the synax. Then there's jank, which is Clojure hosted on the LLVM with C++ interop. That could work to make a Clojure Guix <avigatori>what does `use-service-modules` actually do? <noe>avigatori, it transforms into (use-modules (gnu services arg1) (gnu services arg2) etc…) <ieure>NegateThis, Yeah. But I don't want all of Clojure, just the good bits from it and CL. <yelninei>efraim: Adding x86_64-gnu support to gcc-11 would rebuild everything as also a variant of 'pre-x86-configure is needed to use /lib instead of /lib64 which gcc-12,13,14 inherit. I guess i could move that into the patch as well <dajole>When I run Anki in flatpak I get `[2:35:0320/103821.565805:ERROR:bus.cc(408)] Failed to connect to the bus: Failed to connect to socket /run/dbus/system_bus_socket: No such file or directory`. It looks like in Guix the socket is in `/var/run/dbus/system_bus_socket`. Is the solution here just a simple symlink or is there a better way to address this? <ieure>dajole, I have `/run/dbus/system_bus_socket' on my Guix System. <ieure>dajole, /var/run/dbus is a symlink to /run/dbus on my Guix System machine also. <ieure>dajole, Does `sudo herd status dbus-system' show a system dbus-daemon running? <dajole>Huh, sorry, you are right. That's what it is on my system, too. <dajole>Hm...I wonder why flatpak can't see it then <ieure>I think flatpaks run in a containerized environment? <ieure>Yeah. Probably something with that. Hate flatpaks. <dajole>Yeah, I'd like to avoid them, too. The problem is I had a newer Anki version on my previous system than what is available in regular guix and I was hoping to mitigate potential issues with transferring my cards to an older version <ieure>dajole, What's the package name in Guix? I don't see one named anki. <NegateThis>You may have to add system bus access to the Flatpak permissions for Anki <NegateThis>KDE has a built-in GUI for doing this, or you could use the Flatseal Flatpak <dajole>Oof, I feel silly now. I can't seem to find anki either, I could have sworn I saw it yesterday. Time for more coffee! Yeah, I'll look into permissioning. Thank you both. <NegateThis>Looking at their manifest, they don't have `--socket=system-bus` set <NegateThis>Toys says anki 2.1.16 from 2019 has a Guix package, but I don't see it in `gnu/packages/education.scm` <cdegroot>is there a list of drivers/devices supported by the libre kernel? I want to swap my laptop's Wi-Fi card but want to make sure first that that would, indeed, allow me to boot linux-libre into a network (I'm eyeing the Intel AX210) <ieure>cdegroot, ath9k is the most recent blobless non-dongle WiFi, but it's expensive and IMO not very good. All the wireless hardware Intel's shipped for the last 15+ years needs closed blobs. <cdegroot>Hmm... here I thought that my server (which has an AX200) booted into networking witht the libre kernel; I must have been mistaken. <ieure>Definitely not, it needs blobs. <cdegroot>I guess I had the wrong USB stick then :). One day, I'll learn to properly label my USB sticks, promise. <ieure>Oh, I don't know if USB dongles need firmware. I have one Netgear one that doesn't. Never used an Intel USB WiFi dongle before. <ieure>But I also don't like dongles. <cdegroot>I mean USB boot stick. I recased my server and messed things up and had to boot from USB with the whole thing being far away from an Ethernet port, so booted into Wifi and was surprised it worked. I _thought_ I had the Guix 1.5 USB stick booted, must've been one with a non-libre kernel then :). The card I'd like to replace is mini-PCI of course <andreas-e>My Atheros dongle does not work on 5.4GHz networks. It is really annoying that there does not seem to be modern wlan cards without non-free firmware. <cdegroot>Oh well, "I give up", then. Indeed, really annoying. <sneek>yarl, ieure says: I left you a review. Happy to discuss / share examples if you have questions about anything I mentioned. <yarl>sneek: later tell ieure well, thanks!