IRC channel logs
2025-06-09.log
back to list of logs
<ekaitz>lfam: i think you need to push, and then codeberg detects that the diff is similar (somehow) and then you can press the manually merged button and put the commit-id there <lfam>It does seem to be that way ekaitz <ekaitz>i wish we could do that from the commit message... <noe>Maybe it could be improved to use the ChangeIds <Noisytoot>how can I make the qtbase package not provide its own "echo" binary that shadows coreutils echo and makes scripts/makefiles that use echo behave in weird ways, other than by installing coreutils over it which I need to remember to do each time I update it? <Noisytoot>(the output of that command is /gnu/store/rr2c884lngq3pf5kabvap69sidxzh7cn-qtbase-6.7.2/bin/echo) <Noisytoot>I don't know why qtbase echo even needs to exist <kkremitzki>Sorry for the lazy/drive-by bug report, but it seems doing `guix install qtwayland` in a "Debian as foreign distro" context breaks my login <kkremitzki>*Specifically using KDE Plasma Wayland session, oops <apteryx>Noisytoot: maybe check what other systems like Debian do about that <apteryx>it could be they opted to remove it from the package output (not install it); if so we could follow suite. <mikethe1wheelnut>..hello? ..I have a question. I just installed guix. it's the only thing on my (other) computer. during the installation process, I opted to install three different desktops, because I'm not familiar with any of them. I chose awesome, i3, and the emacs option. I was given the impression that I would be provided with a choice of which <mikethe1wheelnut>desktop to launch when I booted up the computer. this does not seem to have happened. it boots into awesome. this is a dell latitude e7450 laptop. I have searched online briefly with duckduckgo, and not found anything. looking at the list of hot-keys, there doesn't seem to be anything for switching to a different desktop. ..obviously I <mikethe1wheelnut>could re-install and only choose one desktop at a time, but the installation procedure wasn't so quick that that wouldn't be irritating.. <cancername_>mikethe1wheelnut: which display manager do you use? if it's gdm, there should be a little cog icon in the bottom right, where you can choose the desktop <mikethe1wheelnut>I don't recall being presented with such an option. I'm now in the place where I can select which user to log-in as, and there is no such cog. <mikethe1wheelnut>I can try re-installing to watch carefully for a selection for display manager, but that won't be tonight. <drewverlee>Do you suggest I build my guix-system up on a VM (e.g /run/current-system/configuration.scm) then transfer that into guix-system installed on my laptop? Will I have to make modifiations to the configuration.scm? Will it be easier going that route then just going straight to the laptop? (it has nothing i care about on it). Thanks! <PotentialUser-9>Hello, i am trying to install or use the wifi in my thinkpad t490. i installed the iwlwifi-firmware in the firmware list with the full linux kernel also charge the module with modprobe but doing rfkill command or nmcli only show the bluetooth interface or ethernet and loopback respectively. I dont know what else try. Someone have an idea of what <attila_lendvai>PotentialUser-9, i don't remember having to do anything, it just worked once i added the non-free stuff using nonguix <PotentialUser-9>not even adding the iwlwifi-firmware? maybe it's not the correct one. <cow_2001>within project foo i have ./guix.scm. within project bar i want to have ./guix.scm whose package definition has as input the package definition in foo's ./guix.scm <flypaper-ultimat>either add foo to your guix load path, e.g. invoke "guix <subcommand> -L /path/to/foo" and have foo/guix.scm use (define-module (foo)), and bar/guix.scm use (define-module (bar) #:use-module (foo)). <flypaper-ultimat>the other way is to in bar/guix.scm use (define foo-pkg (load "/path/to/foo/guix.scm")) <flypaper-ultimat>cow_2001: the former is more following the guix style, the latter is easy for quick hacking, but is fragile when you move foo. <flypaper-ultimat>alternatively, turn foo your foo project into a channel, and add it to your guix channels <cow_2001>could i use it temporarily? without having to guix pull everything? <flypaper-ultimat>you could guix time-machine with a channels file, but the quickest development cycle would be to use 'guix ... -L /path/to/foo' or the '(load "foo/guix.scm")' <flypaper-ultimat>(because time-machine / pull you'd would need to regenerate a guix profile with every commit made to the foo project) <flypaper-ultimat>if you forget how to do it, you could even make the shebang of your guix.scm for bar/guix.scm #!/usr/bin/env -S guix build -L /path/to/foo -f <newline>!# <csantosb>Hi Guix ! Correct me if I'm wrong, but installing guix in a foreign distro still requires root access, right ? <ruther>csantosb: that's not easy to answer. With the default permission on most distros and using the official installer, yes. But there are definitely ways to install completely rootless if you wanted to do that <attila_lendvai>guile printing an _ in the backtrace instead of the very value that would help me close this debugging session is... infuriating. <Lumie>Getting an SSL error for git during giux pull on a newly downloaded VM image <Lumie>luca: hah, thanks, that was it <attila_lendvai>again, shepherd is hung after a reconfigure and not responding to herd... nothing of interest is in /var/log/messages <mikethe1wheelnut>I figured it out! The cog -is- there, but -only- once I select a user! I select the user, and it gives me the field to put my password in, and at the same time, the cog appears! (confusing, but if the system is setup so that different users have the option of different combinations of desktops, then it makes sense. I -would- add it to the <cow_2001>if i want to connect to a guix repl with geiser, how do i do that? there is guix repl --listen, but i don't think it does what i think it does <cow_2001>guile --listen does what i think it does <mawumag>hi everyone, is there any limitation on how long paths should be? I have this strange bug where the build process fails with a cryptic 127 error, but the command that fails can be run manually no problem (after sourcing the build env variables of course) <ieure>mawumag, What paths, specifically? $PATH / $PYTHONPATH / etc? Or do you mean how long the path to a file can be? <ieure>mawumag, Can you put the build log on a pastebin? <mawumag>of course hopefully it's not too long <ieure>mawumag, Depends a lot on the context. I don't think there's a fundamental limitation, but some programs could have a fixed-size buffer for them or something. <mawumag>funny thing is that the cargo build works just fine from the build directory manually :/ <ieure>mawumag, These look like your actual problems: <ieure>error: failed to run custom build command for `tikv-jemalloc-sys v0.6.0+5.3.0-1-ge13ca993e8ccb9ba9847cc330696e02839f328f7` <ieure>process didn't exit successfully: `/tmp/guix-build-python-snapatac2-2.8.0.drv-0/source/target/release/build/tikv-jemalloc-sys-6d913cb8910cb421/build-script-build` (exit status: 101) <mawumag>ieure ah yeah that's true, just did not check that directly because of all the env variables that need to be <flypaper-ultimat>cow_2001: guix repl --listen does do what you want but the syntax for the option is slightly different than the one for guile, 'guix repl listen=tcp:37146' 'M-x geiser-connect 37146' <cow_2001>flypaper-ultimat: what about a fifo file? <cow_2001>in guile's --listen i can provide a path for a fifo file to be created <flypaper-ultimat>don't blame you, its one of those things you expect the same between guile and guix repl and arent. (similarly i'd like guix repl to have an -e option to guile so that i can set an entry point for my scripts, rather than just calling it at the end of the scipt) <mawumag>ieure thank you! I now figured it out... the env variable CARGO_PKG_VERSION is not set but it is expected by that script <mawumag>I will try to figure out how to set it during the build process <PotentialUser-17>Hello. I was running down on storage on my machine and decided to install `rmlint` to try and regain some free space. <PotentialUser-17>However, this tool, which is a small and fast command-line tool pulled in so many dependencies that left me dizzy. <PotentialUser-17>I guess the documentation is to blame, but still... is this "normal and expected" for Guix? <ieure>PotentialUser-17, It looks like most of those deps are coming from the GUI for it. <mawumag>ieure I figured it out btw: it was actually some other flag that had to be set for jemalloc (and it had to be unbundled from the tikv-jemalloc-sys package). Thank you again for the help! <ieure>mawumag, Nice! Happy to help. <PotentialUser-17>I mean, it even pulled in cups, elongind, wayland, polkit, vulkan-loader and mesa! <flypaper-ultimat>PotentialUser-17: main culprit seems to be python-sphinx, which is for building the docs <flypaper-ultimat>if you do e.g. 'guix graph --path rmlint mesa' you'll see the (first found) path from 'rmlint' to 'mesa' <flypaper-ultimat>so, you could make a variant that excludes python-sphinx and skips the doc making phase <flypaper-ultimat>mutiple outputs would not solve this though, because those multiple outputs are share their inputs, you'd want a package variant (like e.g. git-minimal vs git) <ruther>flypaper-ultimat: multiple outputs do solve this, the 'shared' inputs part of multiple outputs is true only for building it. Afterwards only the actually used dependencies are gc rooted, while the rest is free for garbage collection <flypaper-ultimat>yeah, but in this case it was a native-input, so there shouldn't be any gc roots anyway <ruther>then the issue of taken up disk space is to run gc <reedm>I have seen that some services which are implemented as shepherd services allow a "shepherd-requires" field, which makes it easy to start a custom dependency. Is there a standard way to achieve this with a service that doesn't expose a "shepherd-requires" field? As a specific example, If I have a shepherd service which provides a symbol <reedm>'run-this-before-starting-database, how would I modify postgresql-service-type to make sure the services are started in the correct order? <csantosb>Just noticed I'm getting emails from codeberg.org/guix/guix not necessarily related to me; to my surprise, my account is watching this repo (I'd never do so) ... anyone knows what is going on ? <ieure>csantosb, Are you in CODEOWNERS? <ieure>PRs get assigned to teams, then the team members get notified to review them. <csantosb>I'm in emacs and electronics teams, which is fine; I'm just getting emails unrelated to these teams ... and watching the whole repo <hako>the repo is added to the team, so it's watched automatically. <csantosb>But team members don't want an email at every single thing going on in the repo, I guess <hako>unwatch the repo or change email settings. <csantosb>Just wanted to warn, as I guess this is not the intended behavior. <cancername_>Is a shepherd one-shot system service provision fulfilled before it exits? <calleha01>Hello! I have just installed Guix for the first time on my thinkpad l390 yoga. Perhaps I made a mistake by opting not to install network-manager or connman during the installation procedure. Anyway, since this computer depends on non-free firmware for wifi I used USB tethering during the installation, which worked fine. However, after booting into the system, I no longer am able to utilize USB tethering. How do I proceed? Do I have to <ruther>calleha01: so I suppose you don't have any dhcp client installed? Like dhcpcd. (you already said you don't have network manager) If not, you will have to set the ip manually for now, or as you are saying, basically reinstall/reconfigure from a live iso with a dhcp client <calleha01>I have the default dhclient from the installer which I have attempted to use, but to no success. I can see the ethernet device but as soon as I try to enable it it won't connect. Does dhclient support USB tethering? If not, I suppose I have to set the ip manually. Will look into how to do that. Thank you! <ruther>calleha01: dhclient doesn't care if something is 'tethering' or not <ruther>calleha01: it is a regular ethernet from the standpoint of dhcp client <calleha01>yeah, that's what's confusing me. I am used to arch linux and I have run into wifi problems on fresh installations before, and in those cases USB tethering was pretty much plug and play so I don't see why it shouldn't work on guix <ruther>calleha01: so what you you mean that you 'enable' it and that it won't connect? Did you start the dhcp client and you didn't get ip assigned? <calleha01>starting the dhcp client gives me "No DHCPOFFERS received." <calleha01>it runs DHCPDISCOVER 6 times before timing out <calleha01>wait nvm it worked now for some reason. I guess because I went on irc to ask for help it decided to fix itself <luca>Hi, does anyone know where the chdir function comes from/how I can import it? And maybe how I could find out myself next time <ieure>luca, Grep for where it's defined in your Guix checkout. It's in ~/.cache/guix/checkouts <luca>Sure, but chdir is a very very used function, so it's hard to figure out <ieure>luca, grep for define.*chdir <ruther>chdir is guile function, not guix function... <luca>oh that makes sense. Thanks! <ruther>and I don't think you can import it, it should be available <luca>Yep, it is. I just misunderstood the error message. Some other variable was missing <civodul>luca: if you’re lucky enough to find it a snippet in the online manual, you can click on it to view its documentation, whether in Guile or in Guix <civodul>not the most convenient way, but hey, can be helpful :-) <luca>So pretty much everything in guile is "available" and doesn't need to be imported? <ruther>luca: I wouldn't say that, there are the srfi modules, ice-9 match etc. <ruther>but I think chdir could be said to be in some sort of standard library and those might not have to be imported. But the terminology might be wrong here <podiki>the berlin issues are complicating looking at possible breakages for mesa-updates....but i'll likely merge that later today (substitute coverage nearly same as master based on QA) <vagrantc>hrm. i seem to be getting lots of codeberg notifications on all sorts of different parts of guix that i am not part of a team for and in no way really hold any expertise <PotentialUser-74>I have a script I would like to have put into /usr/, is there a system service that would let me do this? <PotentialUser-74>It doesn't have to be /usr/, I just believe that it makes most sense for me to put this script there. I also need to be able to create files in /etc/ <ieure>PotentialUser-74, You need to create a package for your script which puts it in /bin. For configuration that goes in /etc, you'll need a system service that extends activation-service-type in order to populate the files. <ieure>PotentialUser-74, If you can share what your goal is with this, we might be able to give you suggestions for better ways to accomplish them. <jmes>Hello, is there a guix-y way to add routing tables and routing rules? <ieure>PotentialUser-74, The WSL2 stuff expects /etc/wsl-distribution.conf inside the image? <ieure>PotentialUser-74, Okay, you either need to do what I said, or patch the `guix system image' stuff that makes the image to populate it. <ieure>The latter feels to me like a better way to do this... but I don't really know anything about WSL2. <jmes>Oops there is an nftables-service. I actually knew that and forgot.