IRC channel logs
2025-03-03.log
back to list of logs
<graywolf>Hi :) I am curious, is there any guaranteed order of processing for extensions of a service? fifo? lifo? Or unspecified? <sneek>Welcome back graywolf, you have 1 message! <sneek>graywolf, jlicht says: Is it okay if I drag your ngtcp2 and nghttp3 bump patches into javascript-team? They mostly impact node AFAICT <graywolf>sneek: later tell jlicht: I am using them as libraries (for building http3 enabled curl) in my channel, hence why I am updating them. If they get merged on reasonable time frame, I do not think it matters whether they go directly to master or via javascript-team. :) <coyotes4ys>hi. so my soundcard isn't in proc/asound/cards. is it the "multimedia audio controller" in lspci? <noob`>Rutherther: yes i check all of it and its fine. but it keeps not showing as booteable partition. i don`t know what else to try. <user_oreloznog>coyotes4ys: here, 'lspci' shows '00:1f.3 Audio device: Intel Corporation Sunrise Point-LP HD Audio (rev 21)' <apteryx>janneke: hello! I'd like to confirm that qemu-7.2.4 doesn't actually have a use anymore, right (it was because of "QEMU >= 8.1.0's riscv64 binfmt service is unreliable.", but we found that to not be the case anymore, IIRC) <apteryx>is there any way to prune references to output variables in inherited phase gexps? <apteryx>e.g. a variant package that has its static output removed, with the parent using #$output:static would cause problems. <vagrantc>hah. have not messed with this aarch64 guix system since 2022 ... we'll see how this goes... <vagrantc>hrm. guix pull is burning a lot of CPU with no visible activity ... maybe should try to pull up to v1.4.0 first ... <divya>efraim: Do you have an idea if this shold be packaged as a rust package or a python one? <divya>Yep, installing it through pip doesn't work. The binary isn't able to detect libgcc_s.so <vagrantc>wow ... after how many minutes, i finally got to ... Computing Guix derivation for 'aarch64-linux'... <efraim>divya: I assume as a rust package <efraim>adanska: what commit are you on? Therewas a fix pushed since the gnome-team merge <adanska>oh, i pulled like 2 hours ago, let me try again <efraim>looks like it might not have been enough, guix weather shows no substitutes <tomenzgg>What's the command to use a different version number for a package (and just hope that things haven't sufficiently changed such that it fails), again? <tomenzgg>Not sure if it's what I was thinking of but "guix build --with-commit=" works, it looks like. <tomenzgg>Alright; almost. How do I specify that I want the package to be built with the commit? Specifying the package in "--with-commit" seems to build it at the commit specified but, since it's not an input of itself, then builds the package at the version that's already in the repos. <adanska>im gonna update gpaste and see if that solves things. <drewverlee>should i run `hash guix` after ever `guix pull` if i want to run `guix upgrade`? <futurile>drewverlee: yeah, it's basically to make sure that after you do guix pull you're actually using any new 'guix' cli that was downloaded/built <mange>I've been getting a bunch of warnings saying "ambiguous package specification `glibc'", "choosing glibc@2.39 from gnu/packages/base.scm:1666:2". Are other people getting this, too? It's pretty annoying. The two definitions seem to be gnu/packages/base.scm lines 915 and 1666 (according to "guix show"). <rekado>are you installing glibc on purpose? <rekado>(I'm asking because installing it is not necessary for other packages to work.) <mange>I am not, but let me try to isolate the conditions that cause it. I feel like I see it all the time, but running "guix shell hello" didn't show the warning. <Rutherther>mange: do you maybe get it for guix install/package commands? <mange>I basically exclusively use "guix shell", "guix home reconfigure", "guix system reconfigure" and "guix deploy". I can't remember the last time I ran a "guix package" command. <mange>Ooooooooh, okay, I'm an idiot. :P It's because I have a (specification->package "glibc") in the manifest that I'm working on at the moment, which is the main "guix shell" command I've been running lately. <mange>Thanks for the help, rubber ducks! :) <jlicht>guix.gnu.org seems to have an expired TLS cert on my local machine with (AFAICT) no time settings <Rutherther>"Not After Mon, 03 Mar 2025 10:30:54 GMT" yeah, expired like half an hour ago <rekado>ACTION tries to restart nginx, but it takes a very long time... <rekado>needed to be killed manually :-/ <msavoritias>how do i specify a python version in the pyproject build system? adding python3.12 in the native-inputs did nothing and the #:python argument doesnt work <cbaines>I've run renew-certbot-certificates on bayfront, and that might have done something, it's hard to tell <cbaines>the first restart of nginx did happen, but the shepherd seems to have hung on the second attempt <cbaines>we might need to restart bayfront to get the shepherd back working again <cbaines>it still seems possible to SSH in, so I'll leave it for a little while at least <Rutherther>msavoritias: could you clarify what doesn't work with the #:python argument? <Rutherther>So you seem to be putting quoted python into it, you need to put in the actual package, not just a quoted symbol / string. But note that replacing python is not as easy, since you would need to replace also in all dependencies. <msavoritias>the problem is that the package itself requires at least python-3.11. dependencies haven't changed. <Rutherther>also it's not right to put python to this argument directly, you actually need to put in wrapped python package that will provide binary called "python". There is wrap-python3 for that <Rutherther>msavoritias: well python-next should point to a package, not to a quote, quotes are like strings. You need to unquote to point to the symbol you have in scope <Rutherther>msavoritias: I am not talking about the dependencies being updated to newer version. I am talking about them being made for python-3.12 as well. If they won't be, your package won't work, it won't be able to find them. Guix doesn't support a good way to just pick different python versions all over the place. So your best bet is waiting until whole guix updates to 3.11/3.12 unless you want to have a lot of work to do to replace python in all the... <msavoritias>do you have any examples of #:python being used anywhere? I am a scheme newbie so visuals help ^^' <msavoritias>but I am guessing that is because it needs more changes to propagate to the whole package as you said <Rutherther>that seems like the python build system module has not been imported <Rutherther>no, it should start building, it shouldn't fail like this. But the build will fail <msavoritias>Using 'setuptools.build_meta' to build wheels, auto-detected 'setuptools.build_meta', override '#f'. <msavoritias>Prepending '[]' to sys.path, auto-detected '#f', override '#f'. <msavoritias>the package in question uses setuptools and wheel as a native input that are not build for 3.12 <Rutherther>yes, this is the part about using wrapped python, even if you did that, it will not find dependencies <msavoritias>right. okay then. thank you. i just open a PR and wait for Guix <cbaines>right, I'm none the wiser as to why the shepherd is stuck on bayfront, I'm assuming it's something about the NGinx service controller as the first command that hung for me was herd start nginx <cbaines>and the last nginx related log message in /var/log/messages was "Service nginx is now stopped." <cbaines>I'll try and reboot, unfortunately reboot hangs as well... <cbaines>bayfront is back up now, unfortunately the certificate issue still seems to be there <cbaines>I think the guix.gnu.org certificate has been managed manually since moving the website to bayfront, I'll try and add system confguration for it as that seems like a good idea <cbaines>ok, guix.gnu.org is back, I'll push the config in a bit <futurile>cbaines: nice job, thanks for sorting it! <z572>civodul: now core-packages-team is the first branch again. Do we have anything else to do besides un-grafts? <z572>And I didn't come across #75658 <civodul>z572: yes, fixing that bug; unfortunately i haven’t been able to spend time on it since reporting it <civodul>for ungrafting i think i have a patch that i could push <z572>It seems that many people are unable to reproduce this problem <SpanishInspector>Hi, I'm trying to access Weblate to fix some problems in the Spanish translation of the website. However, trying to log in show 400 - Bad request error: invalid transaction id. Anyone else experiencing this? <SpanishInspector>Ah, never mind, it seems it was a VPN issue. I disabled the VPN and could log in. <divya>I just checked that Gradle hasn't yet been upstreamed into Guix, are there any plans for this? <divya>The latest relevant discussion I found was #70886 <df>should geiser be able to navigate into guile builtins? <df>I've tried adding this, but I'm not sure whether it should be necessary: <df>(add-to-list 'geiser-guile-load-path "/run/current-system/profile/share/guile/3.0") <df>oh... it seems to work for some stuff, like functions from srfi-1 <ieure>divya, Don't know of any plans for that. I don't have firsthand knowledge, but I'm pretty sure Gradle isn't packaged for Guix because it needs a ton of work to bootstrap. <df>and that doesn't seem to require an extra entry in load-path <ieure>divya, All Gradle-using projects bootstrap themselves from binaries downloaded over the internet with their gradlew script. And Gradle is self-hosting. <divya>ieure: Indeed, that's also what the issue mentioned... <divya>*sigh* I wish more engineers cared about bootstrapping ieure <ieure>divya, Oh yeah, also Kotlin. <divya>Noè isn't here, right? I might ask him if he made any progress. <aure84>Hello everyone. Does anyone know if `inputs` given as `source` with `git-fetch` method delete the `.git` directory on checkout? <aure84>My inputs are libraries that will be cloned locally (e.g., git clone file://....) by a CMake superbuild system <ieure>aure84, Don't know for sure, but I doubt it does. <aure84>It seems to be the case. only the code is available on /gnu/store/.....-git-checkout directories generated <aure84>I've tried to call a git init using a (snippet) clause, but I'm not sure how to refer to the chekcout directory. Any tips? <Rutherther>aure84: it's the current cwd. But why do you want that anyway? <ieure>aure84, Inputs are immutable, so, don't do that. <ieure>aure84, What are you actually trying to do? <aure84>I'm trying to build a software that uses CMake to pull external dependencies (superbuild). Since the build daemon does not pull external dependencies, a solution is ato add them as inputs and trick the superbuild system to do a clone from file:// instead of https:// <ieure>aure84, What do you need the .git directory for, though? <anticomputer>hrmm when zsh config is managed out of home-zsh-service-type shouldn't it be creating .zprofile to source .profile on its own accord? for some reason my ~/.zprofile remains a regular file and if I delete it and reconfigure Guix home, it stays deleted, that doesn't smell correct <aure84>you cannot git clone a directory that is not a git repository <Rutherther>anticomputer: so what are you using to manage .zprofile if not home zsh service type? <aure84>and the superbuild script trying to do git clone won't clone anything <anticomputer>Rutherther: note: I'm assuming an empty list zprofile defaults to the bits and bobs Guix Home requires by default for .zprofile, this might be wrong <ieure>aure84, I see. You need a different approach here. The Guix build-system clones the repos for you. You need to make CMake use those, not try to clone from them again. <aure84>That mean's big changes in the cmake infrastructure of that project, while passing -DLib_GIT_REPOSITORY=file://.... would do <Rutherther>anticomputer: so are you using home zsh service type now? I thought you said you were trying to manage it outside of this service now <anticomputer>Rutherther: nvm, .zprofile lives in the XDG config path because that defaults to #t, it's just an old file I'm looking at, it's working correctly out of the service type <anticomputer>Rutherther: no the opposite, I was moving from out-of-service to in-service, but it's resolved, zsh moved into XDG config path so I was looking for .zprofile in the wrong location <aure84>it is remarkable: (modules '((git))) (snippet (repository-init ".")) will fail with "./.git" directory exist. If I remove that code, I can see that the /gnu/store/....-git-checkout does not have any .git directory. I wonder what happens with git fetched sources between pulling the inputs and the building process... <civodul>roptat: hi! i had to turn off Cuirass on lieserl because bandwidth is so low that it ends up taking up “download slots” for ages, slowing down the entire process <apoorv569>I'm trying to use oci-container-service-type but getting this error when trying to build the system config via geiser repl, `1. &message: "primary group 'docker' of user 'oci-container' is undeclared"` <apoorv569>what does this mean? I need `docker-service-type` as well? <podiki>maybe? i have docker-service-type but just to set some environment, so i never tried without <futurile>the wording in the manual for that service is a bit unclear - it seem's to indicate that it will set-up the docker group as default <apoorv569>Hmm.. I see. Also there is no way to automate the creation of a docker network via the service right? a network has to be created manually before? <apoorv569>OK.. I added docker-service-type it asked me to add containerd-service-type which asked me to add dbus-root-service-type which asks me to add elogind <apoorv569>I was trying to keep this config minimal.. but this is getting bigger and bigger by sec. <podiki>yes, you need those 3 services (no config necessary though) <podiki>correct that there is no network you can create for your images, but see the patch set I linked above, with that update you can (from what i can see, i need to test myself) <podiki>i think it has a "network" field you can specify in the oci service <podiki>i believe the use case would be just like docker-compose where you put all your containers on the same network so they can talk to each other <podiki>that's what i do with my badly hacked together modification of oci service, which is why i want to try to proper looking patch set :) <podiki>will get us a nice step closer to doing what docker does but in our declarative system config <apoorv569>the oci-service-type you mentioned there.. that? so any network mentioned in the (networks (list (oci-network-conifguration will be created if doesn't exist? <podiki>sorry i can't parse your message <apoorv569>I was asking about the patch you mentioned.. oci-network-configuratin will create any mentioned network if not exist? <podiki>i haven't tried, but that's what i thought (or in oci-container-configuration) <apoorv569>interesting.. that would be really nice to have BTW. <podiki>there is documentation in that patch as well as a home service <worni>After upgrading to gnome 46 the extension app is broken. All it says is "It was not possible to get the list of installed extensions". I dont even know where to begin looking and google searches comes up empty. Any pointers as to where I might start looking at whats wrong? <futurile>worni: gnome-team only just landed their series, so it might be a new bug - I would search on issues.guix.gnu.org - if nothing shows open a new bug <lilyp>extensions did work the last time I checked them, but maybe I missed something <vagrantc>wheee. updated all the way up to v1.4.0 on my aarch64 system ... and have my profile updated to current master ... but updating the system to current master fails to build samba (much to my surprise, the usual culprit, linux-libre, had substitutes available) <simendsjo>I'm in the process of updating my system. Started with the system, which worked. Then home environment which ran out of space. Now emacs seems broken: `/home/simendsjo/.guix-home/profile/bin/emacs: symbol lookup error: /run/current-system/profile/lib/gio/modules/libdconfsettings.so: undefined symbol: g_once_init_enter_pointer` <csantosb>Hello, Guix ! I have a doubt about channels. <csantosb>I’m building package A (in channel 1, which depends on channel 2); A depends on package B (in channel 2). <csantosb>Now, when I fix commit of channel 1 to build package A (with time machine), which version of B am I using ? <futurile>simendsjo: revert back and run gc to save space? then update maybe? <lilyp>simendsjo: as a workaround you can run `guix shell emacs -- emacs` <simendsjo>futurile: I ran gc and started reconfiguring home again. But I think it's strange that it broke. It should flip the symlink first after everything has succeeded, right? So how can it be broken when it failed during update? <Rutherther>csantosb: there is no definitive answer. You can be using whichever commit of the channel 2, which in turn changes the version of B <lilyp>did you log out and back in again? <lilyp>I'm getting similar errors for a bunch of packages and haven't rebooted yet <lilyp>probably a search path (fingers crossed) <csantosb>Rutherther: but there is a huge ambiguity in here, right ? User of channel A is not even aware of the existence of channel 2. <Rutherther>no, reproducibility is not broken, using different inputs leads to different outputs, sure, that doesn't mean reproducibility is borken. Reproducibility means using same inputs leads to same outputs <lilyp>great news: emacs is working exactly as expected <futurile>why would you only fix/pin one of the channels, if you're using two channels and you want reproducibility then you'll fix the versions of both channels <lilyp>to work around a particular bug in one channel for example? <csantosb>Rutherther: I install channel 1 (which depends on channel 2, but this is transparent to me). I fix commit in channel 1, so I'm always using a given version of package A. Except that I'm randomly using package B from channel 2. <Rutherther>or no, not really, not randomly. Probably latest, that is not random <Rutherther>you're misunderstanding what reproducibility means <csantosb>Reproducibility: get same results at any given point in the future. <Rutherther>it's like expecting unpinned guix to always provide the same package. Nope, it doesn't, if you guix pull, you will get new package, hence reproducibility broken? No. Inputs changed <csantosb>But I'm pinning guix with a channels.scm file <csantosb>guix time-machine --channels=channels.scm -- package -p $GUIX_PROFILE -m manifest.scm <Rutherther>if you want same inputs, you pin all chanenls you are using, if you don't, you don't have same inputs guaranteed <csantosb>channels.scm and manifest.scm are always the same. They never change. <Rutherther>it doesn't matter if channels.scm change, channels.scm can contain unpinned channels <csantosb>I don't even know about the existence of channel 2. This is my point. <Rutherther>the same goes to manifest.scm, you could for example start depending on the current year. You don't do that if you want same inputs. Same way, you do pin ALL your channels, INCLUDING dependencies <Rutherther>You know about its existence, it is written in the output of time-machine <csantosb>When I install guix-science, I dont check which other channels guix-science depends on (if ever, this is just an example) <csantosb>If guix-science depends on guix-science-other, how can I know ? <Rutherther>either: 1. output of time-machine, 2. the channel definition file <csantosb>Ok. So better advise users of a channel to go check (and pin) this channel dependencies; same for the dependencies, etc. <podiki>channels declare their dependencies in their channel file no? <podiki>for pinning a channel, yes usually best to pin all other channels it depends on (including guix) <csantosb>I assumed until now that this is supposed to be transparent for the channel user <podiki>well if you pin you should just pin everything i would say <podiki>but a channel has no way of knowing if it will break a dependent channel, so the maintainers of a channel need to take care of that <csantosb>But having to manually check the full chain of channel dependencies is far from ideal <vagrantc>futurile: huh. i think i had a cut-and-paste problem <csantosb>podiki: there is no `guix describe --format=channels` equivalent in this situation <futurile>vagrantc: fat finger is definitely the best option for something weird in our commit chain <vagrantc>indeed... although i still am confused as that was the last successful build of samba 4.17 ... but 4.18 has been in guix since 2023 ... <podiki>i think the best solution is for users to always inspect what channel they add to they know what they are getting into (good advice anyway) and for channel maintainers to document such things <podiki>frequently get issues where someone pins one channel and it doesn't work since they didn't pin guix too. maybe that should be clearer in the manual too from a user standpoint <futurile>vagrantc: although sometimes I find the UI a bit confusing, so I don't totally trust I understand when something "actually builds" <cbaines>unfortunately the data is quite patchy, but it could mean that 4.17 didn't reach master <vagrantc>cbaines: how would i go about finding the last successful build on the data service? <vagrantc>i feel like i always understand the data service results better ... if i can find them :) <cbaines>both ci and the data service processed cf34ba66cf614358dfac925212d6a9074e318d80, but the data service doesn't see samba@4.17, at least it's not a public/visible package <vagrantc>last successful build of samba on aarch64-linux <vagrantc>cbaines: and if you have other things to poke at, by all means :) <csantosb>podiki: all depends on how deep is your tower of channel dependencies ... I need to play with pinning deps in .guix-channel; I hope this is dynamic with channel history <podiki>i guess in the end there's always the usual dependency problems in some form <vagrantc>of course, then soemthing else might fail ... but that is pretyy recent if it works :) <podiki>i know i've seen this problem before but blanking on why <podiki>trying to update godot and it doesn't find linux/errno.h <podiki>i thought it was also part of gcc toolchain and so usually in a build system <podiki>(when explicitly including linux-libre-headers) <apoorv569>is this the right way to define `post-up` command for wireguard-service-type? `(post-up '("iptables -A FORWARD -i wg0 -j ACCEPT ; iptables -t nat -A POSTROUTING -o enp1s0 -j MASQUERADE"))` <apoorv569>does the `iptables` here need to be file-like? <podiki>maybe something with CPPPATH (godot uses scons, which i know nothing about) <podiki>ah got it, we do modify the scons build file to use the environment where we set these things <podiki>ye olde substitute* not matching anything throws no warning/error :( <ieure>I bet a lot of packages would break if we changed the behavior. <podiki>yeah it is annoying and wastes time <podiki>i think we should just make that change though, at least a warning <podiki>(that somehow doesn't get lost in the build output noise...) <apoorv569>not sure what's going on but I'm trying out my server config in a VM and I have openssh-service-type but for some reason the ssh command is not present <apoorv569>I use same config for iptables wireguard and ssh on my personal system <apoorv569>sudo herd restart sshd says service sshd could not be found <apoorv569>even building a container out of the config there is `ssh` in the profile/bin of the built container path. <apoorv569>im inheriting different server configs from a base one and the openssh-service-type is in the base config <alainalain>any guix web browser that can play video on web pages - I tried most and none work - recent install <ieure>alainalain, Which did you try, specifically? What's not working? <ieure>alainalain, I use LibreWolf in Guix, it plays videos fine. Did you try that one? <alainalain>I am installing librewolf now, it is compiling...I will let you know in few minutes or hours! <ieure>alainalain, Do you have substitutes turned off? You don't normally need to compile it. <ieure>alainalain, It takes a long time to compile and needs ~20gb of RAM or it'll fail. <ieure>alainalain, If you don't have that much RAM (ex are on a <=16gb machine without swap), kill the build now and add swap to save yourself some time and compute. <ieure>alainalain, I think you're confused. You want substitutes enabled, they let you download prebuilt binaries instead of compiling them yourself. They're enabled by default. <ieure>alainalain, Are you on ARM or x86? <alainalain>I tried icecat, web pages dont show most of the times, I tried the default "web" application, no video. <ieure>alainalain, Okay, you ought to have substitutes. Maybe the gnome-team merge that happened recently forced a LW rebuild. <alainalain>So turning off substitutes should allow me to download binary librewolf <alainalain>We call these pre-built items substitutes - Ha I am confuse it does not load binary? (because it is being recompiled on the server?) <alainalain>guix weather librewolf - should I wait for the binary being available? <ieure>alainalain, Yes, I recommend you wait until `guix weather' shows substitutes are available..