IRC channel logs

2024-06-01.log

back to list of logs

<mihi>New month, new logs :)
<rekado>freakingpenguin: I can't view the log above, but you'll likely have more of these lines when logging non-interactively.
<daviid>rekado: hi! i did read your email, which i deleted, but later thought that one thing you might kindly do, is training someone to handle the guile/guix logs if 'it fails', as it happens from time to time ... tx!
<daviid>*to handle/to restart the logging of guile/guix channels ...
<freakingpenguin>Hmm, looks like (guix scripts substitute) is invoked by the daemon directly. This is sure going to make testing it a pain! :)
<freakingpenguin>Trying to give it a "simple-reporter" to use when isatty? is #f.
<freakingpenguin>I think the easiest way to test this is create an os config, customize the system-wide guix, and use a system vm, although that uses "guix the package" and not "guix the source tree".
<freakingpenguin>Hmm if the daemon is invoking the substitute script directly it might not be writing to a tty normally. If that's the case then this just became a very interesting problem to solve.
<apteryx>ieure: I got it building on 8 cores... retrying with 24
<meaty>hooo real high roller over here damn
<meaty>threadripper?
<mirai>freakingpenguin: you might be interested in reading https://issues.guix.gnu.org/62056
<peanuts>"[PATCH] guix: Only issue erase-current-line sequence for ttys." https://issues.guix.gnu.org/62056
<mirai>It's been a while since I've looked into it and there was some issues pointed out in review
<freakingpenguin>mirai: Thanks, that looks relevant. Sounds like the technical challenges aren't as bad as I think.
<freakingpenguin>My strategy of making a different progress-reporter when not a tty seems like it'll pass review.
<sughosha>Hi! Is it possible to emulate a guix container as a root user?
<sughosha>one program requires to be run as root, but I want to run it in an isolated container as a root user.
<PotentialUser-81>Sorry I'm sure this is a common question but I couldn't easily find an answer. What is my best option if I want to explore using GUIX on an M2 mac?
<futurile>PotentialUser-81: there's no MacOS version of Guix. It's Linux only.
<PotentialUser-81>That makes sense. Though in lieu of another laptop I think I'll try something like https://n8henrie.com/2022/10/guix-system-guixsd-vm-on-an-m1-mac/
<PotentialUser-81>I see there are also efforts such as https://gitlab.com/cnelson31/msg
<peanuts>"Chad Nelson / msg ? GitLab" https://gitlab.com/cnelson31/msg
<futurile>PotentialUser-81: Nix runs on MacOS I think, as another option - don't know aobut M2
<futurile>oh cool blog post
<PotentialUser-81>🙏
<the_tubular>What's recommended for running OCI containers on guix? Docker ? Podman ? Docker has a pretty old version available and Podman is missing cool features due to missing systemd to manage pods
<civodul>the_tubular: hi! there’s now a Guix System service for that
<civodul>lemme see
<civodul> https://guix.gnu.org/manual/devel/en/html_node/Miscellaneous-Services.html#OCI-backed-services
<peanuts>"Miscellaneous Services (GNU Guix Reference Manual)" https://guix.gnu.org/manual/devel/en/html_node/Miscellaneous-Services.html#OCI-backed-services
<the_tubular>I fail to see which issue does that service tackle
<the_tubular>I haven't tested, but Docker's guix seems to be vulnerable to multiple CVE, just looking at the version number
<the_tubular>And yeah, podman auto update and stuff like this wouldn't work even though i use that guix service right ?
<futurile>the_tubular: the service makes it easy to run containers using a declarative approach - which is the 'special sauce' of guix/nix
<futurile>the_tubular: if you just want to run docker containers then you can use the package, as you would on another Linux distribution
<futurile>the_tubular: as a new user, don't forget you can run VM's with another linux distribution - so you could run your old linux distribution in a VM and run Docker in that while you're learning Guix
<the_tubular>Right, that might be the best way
<anthk_>when will gnome 46 hit guix?
<taeltydes>Is anyone aware of any documentation for declaring guix shell args declarativly? I'm aware of the oci services for declarative docker containers but I would like to use direnv with --emulate-fhs
<futurile>taeltydes: I've created bash scripts that do this, that then call guix shell, it's a trick from kaelyn. See the FHS example in this post which creates a shell and calls a manifest: https://www.futurile.net/2023/04/29/guix-shell-virtual-environments-containers/
<futurile>taeltydes: unfortunately, I don't have any example of doing this in guix home etc
<apteryx>ieure_: librewolf update pushed! thank you
<taeltydes>futurile: thanks. this should do the trick!
<apteryx>meaty: it's one of our build machines at the MDC (EPYC)
<apteryx>hm, my patch update doesn't work as planned: gnulib@2024-05-30-1.ac4b301 -> patch@2.7.6 -> cl-asdf@3.3.5 -> clisp@2.49-92 -> gnulib@2024-05-30-1.ac4b301 -> patch@2.7.6
<apteryx>ACTION tries to copy the gnulib source instead of using it as a native input
<apteryx>that worked
<jonsger>does someone know if bayfront (bordeaux server) has SSDs or HDDs or both?
<civodul>anthk_: re GNOME, it all depends on volunteer work; if you’d like to help, you can get in touch with the GNOME team here
<civodul>jonsger: it currently has HDDs, and they’re quite slow
<jonsger>civodul: so does it hurt the performance serving NARs? Compared to SSDs
<jonsger>apteryx: what is your opinion on the icedove update: https://issues.guix.gnu.org/67849 I don't have the time, motivation and skills to get the localized icedove updated to 115
<peanuts>"[PATCH] WIP: gnu: icedove: Update to 115.5.2." https://issues.guix.gnu.org/67849
<civodul>jonsger: not sure about serving nars, specifically; it does hurt performance under high load, since like guix.gnu.org, hpc.guix.info, etc. can get slower
<civodul>problem is that the machine often builds stuff, in addition to running these services
<jonsger>apteryx: my proposal would be to deprecate and replace the localized icedove with icedove-minimal@115. I'm running icedove-minimal@115 for a few weeks now with no issues...
<taeaad>Suppose I am packaging something, and it will run on two computers, on which the code will be slightly different in the package. Will it be bad practise for "guix install <x>" to mean two different things on the two computers?
<jonsger>civodul: maybe the approach we're doing at the nonguix build server could help. We are running a VM on the build server with the workers in it and dedicated resources. So the services on bare-metal (outside the VM) never run out of CPU or RAM...
<dariqq>do i need some extra settings for stumpwm to find the extra modules? I have sbcl and sbcl-stumpwm-kbd-layout in the same profile as stumpwm but it complains about missing modules
<apteryx>jonsger: I'll try looking into adapting the locals (and perhaps modernizing how it's done -- upstream compute them for us now, so we can grab them as data)
<jonsger>:)
<civodul>i’m looking at the relative apathy on guix-devel & co. and think we need signals from maintainers and old-timers to counter that
<civodul>in particular growing newer contributors into leardership positions
<Googulator>Excuse my ignorance, but is there any way to force the build of a hidden package from command line?
<Googulator>I tried guix build -e '(@ (gnu packages commencement) glibc-mesboot)', but to no avail
<Googulator>(glibc-mesboot is failing to build as part of a system image build, and I'm trying to debug that without having to wait for Guix to compute derivations for the whole system image for each attempt)
<civodul>-e is the way indeed
<civodul>you need (@@ …) in this case because ‘glibc-mesboot’ is a private variable
<civodul>(that is, two @ instead of one)
<Googulator>this is a bit of a double mystery btw, since 1. this failure came quite late in the image build, well after glibc-mesboot was supposed to be built (indeed, even Rust bootstrap was complete by this point)
<Googulator>and 2. there same step of glibc-mesboot previously succeeded within the same build session
<Googulator>it looks as if (getcwd) is returning #f for some reason
<stephen-loom>I've just installed Guix with Gnome as the desktop environment, how do I uninstall gnome-boxes and other Gnome default packages *using just config.scm*? I'm thinking something like NixOS's environment.gnome.excludePackages. If anyone needs to know my config.scm is the default after a fresh graphical install with Gnome as desktop.
<apteryx>has anyone tried using the new sysdig package?
<noe>stephen-loom you should be able to change it in gnome-desktop-configuration
<dariqq>stephen-loom: Could you upgrade first because the installer is quite old and gnome-configuration has changed since? (guix pull and then sudo guix system reconfigure /path/to/config.scm)
<Googulator>Is there some way to force a bootstrap package like glibc-mesboot to force an already built tar/xz for extraction, instead of bootar? (On the system I'm working on, bootar is absurdly slow.)
<Googulator>*to use an already built tar/xz
<stephen-loom>noe: this looks promising, I'm a bit embarrased my efforts with the search engine didn't lead me to this before. Is there some good example online of how this works? I've just found some examples of defining a custom gnome package inheriting from gnome with packages removed in some way, is this what you mean?
<stephen-loom>dariqq: Good idea, I tried running this earlier but quit out because the command was taking forever to build python-jupytext and then python-scipy. I'll report back when this is finished.
<noe>stephen-loom to find it, I searched for gnome in the manual. Search engines are pretty bad with guix information unfortunately. I'm not that good at scheme, if you use %desktop-services I think you want to use modify-services to modify the gnome service
<dariqq>stephen-loom: you could use something like this https://paste.debian.net/1318822/. It is a bit ugly but i think it should work
<peanuts>"debian Pastezone" https://paste.debian.net/1318822
<Googulator>alright, it wasn't getcwd that was #f, but something else - but as soon as I added debug (display) statements - the issue disappeared!
<Guest44>Does GNU Guix system support bcachefs?  Want to try it out and wonder if it works out of the box
<stephen-loom>dariqq: back at my pc and the reconfigure has completed and after a reboot I no longer have gnome-boxes! I still have other things like gnome-music and the default gnome web browser I'd like to take care of now though.
<dariqq>stephen-loom: https://paste.debian.net/1318822/ and replace gnome-calcular and gnome-weather with epiphany and gnome-music (and potentially add more alist-delete statements) . Depending on how many things youd want to remove it might be easier to instead specify packages directly instead of removing from the default
<peanuts>"debian Pastezone" https://paste.debian.net/1318822
<dariqq>not really pretty i know
<stephen-loom>dariqq: my apologies I must have missed your previous message, I'll get on with this and report back
<stephen-loom>it can be prettied up later ofc
<stephen-loom>dariqq: amazing stuff, that worked and now I am without gnome-music, thanks. I did get 20 or so warnings that 'gnome' is deprecated, I wonder if anyone on here has an idea about this?
<stephen-loom>ignore the above about the deprecation warnings, that was some old (inactive) code I had littering up the config from a previous attempt
<ieure>apteryx, Thank you, appreciate the review and discussion.
<gnucode>a hurd developer is live coding right now if ya'll want to join
<gnucode> https://jitsi.member.fsf.org/Hurd-code-jam
<peanuts>"Jitsi Meet" https://jitsi.member.fsf.org/Hurd-code-jam
<gnucode>he is currently trying to port the Hurd to ARM. And making progress.
<Googulator>Probably my weirdest ever commit to any Git repository so far... https://github.com/Googulator/guix/commit/19c6f111f6c11e5f62922139321d55fd7b60f93f
<peanuts>"WTF? Fix glibc-mesboot build failure... by introducing debug prints? ? Googulator/guix@19c6f11 ? GitHub" https://github.com/Googulator/guix/commit/19c6f111f6c11e5f62922139321d55fd7b60f93f
<tacocatgirl>is anyone else having issues with the latest latexindent package?
<tacocatgirl>I get an error that says it cant locate YAML/Tiny.pm
<tacocatgirl>ok, so it works when I run the script using perl on my host distro
<ieure>Everyone loves a heisenbug.
<tacocatgirl>so I guess the issue is that perl-yaml-tiny and perl-file-homedir are not being passed as inputs to the latexindent package. can anyone else confirm?
<freakingpenguin>Bah, so my naive hope #71229 would be easy is wrong. The substitute script isn't passed the invoker's stderr fd directly. Instead the daemon creates its own, non-tty fd pair and we just assume it's always a tty.
<peanuts>"30.0.50; [PATCH] rename comp-run.el and comp-cstr.el private functions" https://issues.guix.gnu.org/71229
<freakingpenguin>oops, #71299
<peanuts>"Better handle "updating substitutes" messages in CI pipelines and dumb terminals" https://issues.guix.gnu.org/71299
<ieure>freakingpenguin, This link in the issue doesn't work for me: https://builds.sr.ht/query/log/1238029/update-readme/log
<ieure>I get a JSON payload complaining about not having a bearer token, so I guess you have to have some kind of OAuth thing for it to work.
<freakingpenguin>ieure: Ah, guess I should've tested the link in private browsing mode.
<freakingpenguin>Interesting that srht lets you view the build job publicly but not the raw log.
<ieure>I tried Sourcehut and honestly fail to understand its appeal.
<ieure>(You do not need to tell me why it appeals to you.)
<freakingpenguin>Don't worry, I only evangelize about emacs and guix
<freakingpenguin>Much to the dismay of my friends
<ieure>I don't bother, all my tastes in computing are so niche that any time someone *does* ask me, I usually tell them that what I'm doing almost definitely isn't going to be their jam.
<freakingpenguin>They do the same about their niches so in my case turnabout is fair play :)
<ieure>Yeah, I'm uh... I like to tell people I've crawled up my own ass and found that I like it here. My daily driver runs EXWM and has a split ortholinear keyboard with blank keycaps.
<ieure>I was downloading something for my dad, he said "it'll be on the desktop." "I don't have a desktop," I told him, "it's a failed metaphor."
<ieure>I once offered my elder child $5 if they could figure out how to open a web browser on my setup. They didn't collect.
<freakingpenguin>Ooh, split ergo keyboards are fun. I love my 36 key one.
<ieure>Ergodox over here, ever since the early days when you had to buy a $350 box of components you soldered yourself.
<ieure>Felt like kind of a dick for spending so much on a heap of parts at first, but it's been night and day for managing RSI.
<ieure>I no longer feel like a dick about it.
<guix-noob-2k>Hi, I've got a quick question with a probably super simple answer: I'm trying to test out guix system + guile, but I'm not able to load a module installed with "guix install" in the guile REPL. Is there some other setup needed for that to work.
<guix-noob-2k>The package in question is "guile-chickadee", but running examples only result in "no code for module".
<ngz>guix-noob-2k: Do you have guile and guile-chickadee in the same profile?
<guix-noob-2k>ngz: very good question. running "guix package --list-profiles" I've got two, so maybe I've got them in different. Is there a way to check what binaries come from what profile?
<Isaz>You could try 'guix shell guile chickadee' and see if that solves it
<ngz>guix-noob-2k: You can check what packages are installed in a given profile with `guix package -p /dir/of/your/profile -I'
<Isaz>guix shell guile guile-chickadee
<guix-noob-2k>Isaz: that works it seems, now I'm getting better errors.
<guix-noob-2k>ngz: none of those list guile as installed, so I guess that's an issue
<ngz>Indeed.
<ngz>guix-noob-2k: Installing guile in your profile populates GUILE_LOAD_PATH, which is used to collect the locations of the other Guile libraries you may have in your profiles.
<guix-noob-2k>Very good. Thanks! I'll have to go through the docs more and understand profiles and how they work :)
<civodul>Googulator: re glibc-mesboot, what build issues were you experiencing?
<civodul>(perhaps you reported it already?)
<fnat>I've got this machine where the 'herd' command has become "unresponsive" after a stuck 'guix deploy ...'.
<fnat>Any 'herd ...' command will simply stay there without producing any output. I can no longer SSH into the machine, but the machine is local anyway.
<fnat>I guess I can try with strace.
<fnat>Oh, ok, yeah, strace says something about '/var/run/shepherd/socket'.
<fnat>I think I might have seen something similar before. (Although not sure if I remember what I did to fix it.)
<trev>can someone take a look at this patch? https://lists.gnu.org/archive/html/guix-patches/2024-05/msg01728.html
<peanuts>"[bug#71228] [PATCH] gnu: guile-ncurses unicode fix" https://lists.gnu.org/archive/html/guix-patches/2024-05/msg01728.html
<fnat>'/var/run/shepherd/socket' is regularly there, hm.