IRC channel logs
2024-05-02.log
back to list of logs
<coyote>ran out of space in the installer in the middle of a `guix pull' oh no <coyote>don't know what to do now, can you even "expand" an overlayfs?? <wdkrnls>I have noticed that the CRAN importer has stopped working for me :( <wdkrnls>I tried importing a random python package from pypi and that worked, so it seems CRAN specific. <wdkrnls>matrix(c(1, 0.5, 0.5, 1), nrow = 2, ncol = 2) <Tadhgmister>hello, I am trying to use `guix system image` on an image that specifies an arm platform and it fails to cross compile glibc-mesboot and I was wondering if someone could help me ensure I'm doing the right thing <Tadhgmister>specifically `guix build --target=arm-linux-gnueabihf -e "(@@ (gnu packages commencement) glibc-mesboot)"` fails with the same error message and I'm wondering if it should be cross compiling that package or whether it is an issue with that package <adanska>btw, going offline so sneek me if anyone has any ideas :) <jpoiret>sneek, later tell adanska: It's in (gnu tests install), you can use `make check-system TESTS="..."` as detailed in "(guix) Running the Test Suite" <zamfofex>Hey Guix! Just here to say thank you. Hope you’re all doing well! <civodul>hey zamfofex, hope you’re doing well too! <kimjetwav>Question: If I wanted to update emacs-next to use a specific commit, would it be enough to use `--with-commit'? I know emacs does some funky stuff in terms of already having the base package be a transformation of the minimal package. <zamfofex>ciThank you! Sometimes I wish I could be more helpful. But contributing (to any project) always feels so daunting to me. So I’m glad to be able to show appreciation, at least. <zamfofex>civodul: Sorry, my tab-completion didn’t work. 😅 <zamfofex>kimjetwav: I’d try it, at least. If you use it on the Emacs package itself, it will be applied *on top* of such “funky stuff” instead of on the base package. <civodul>zamfofex: i can sympathize wrt. the energy it takes to contribute! <civodul>kimjetwav: i also think it should work, but it will discard patches that are applied to emacs-next, so expect some functionality loss <civodul>or use a series of ‘--with-patch’ options on top of it… <kimjetwav>The reason I ask partially is that I've already been trying it and it doesn't seem to do what I want. <kimjetwav>I'm just kind of phrasing my question as "is this expected behaviour?" <kimjetwav>I could of sworn I had this doing what I wanted to in the past, but it's been a while since I've messed with it. <civodul>kimjetwav: i don’t know the specifics about emacs-next, but what the fact that ‘--with-commit’ discards patches that emacs-next uses could explain that it “doesn’t do what you want” <KE0VVT>I forget how nice manifests are. <picnoir>Morning! Roses are red, issues.guix.gnu.org is broken :) <kimjetwav>I've also been trying it with a variant in my load path, for the record. I really can't figure out what I'm doing differently than I used to. <kimjetwav>I'm sure it's something really small that I've messed up. <civodul>picnoir: oh, how come? i’ve restarted it and it’s happy, but i suspect it won’t last <civodul>instead of waiting for Big Corp to fix its problems on its zillion-server network, random folks type random commands on the one server of the project <futurile>kimjetwav: maybe try a manifest that inherits the current emacs-next and defines a specific commit? that way it won't throw away the patches that the base package uses <kimjetwav>The worst thing is it's not like it's erroring out or anything, it's telling me "nothing to do" no matter what commit I suggest it use. It accepts my currently installed eamcs-next as the one. <jpoiret>nat-418: I know that, but I don't really get why it would be an input to other packages <civodul>picnoir: for this particular issue, we need a less random person i’m afraid <kimjetwav>jpoiret: is it input or is it like a super transitive input? Lots of things might depend on something that depends on `patch' or something. <picnoir>civodul: too bad. Thanks for restarting debbugs :) <kimjetwav>Guix reduced bootstrap seed binaries soon to be reduced further by the introduction of guile-ed. <kimjetwav>Now you too can boot to the shepherd repl and get it to issue a string of question marks as you try to edit code before loading. <civodul>cbaines: moin! qa-frontpage chokes on invalid JSON, but i’m not sure where that comes from <kimjetwav>KE0VVT: I'm joking, jpoiret asked about guix packages depending on ed, and I was referring to like gash and gash-utils, which are guile implementations of a subset of bash and awk etc., for use in bootstrapping primitive unix utilities with only guile. <kimjetwav>KE0VVT: Specifically guile-ed in the joke is meant to be a guile implementation of ed. <aiena>does guix use GNU Hurd or the linux kernel <kimjetwav>aiena: Linux-libre by default but Hurd is worked on, and there are systems and services in place for using Hurd underneath Linux, currently. <KE0VVT>Is there a guide to using Gash to write scripts in Guile? <kimjetwav>KE0VVT: You can just guix install gash and gash-utils, you can run it as though it's a normal shell, but you'll find it's missing many features of bash. They exist principally for early-early bootstrapping where most of the OS doesn't exist and you need a shell-like entity to run build-scripts, and an awk-like language to process text files. <jpoiret>cbaines: what was your method to remove the duplicate commits on c-u? <jpoiret>i agree with you that it's probably better to start from a clean slate <jpoiret>i've got 1 week and a half of holidays <KE0VVT>kimjetwav: Basically, I'm trying to stop writing Bash scripts and start writing Guile scripts, but I've found it a little awkward. I find myself using string-append a lot. <kimjetwav>KE0VVT: What you want isn't gash per se, gash is just a subset of bash implemented in guile. Inside gash you don't have access to normal lispy-schemey syntax as that would conflict with shell syntax. <civodul>rekado: i think we can’t add gfortran-12 etc. due to ABI incompatibilities <civodul>there’s a comment above ‘gfortran-toolchain’ <kimjetwav>KE0VVT: Gash is literally just a bash-esque shell running on top of guile, so that when guix needs to rebuild the world, it can leave bash out of the initial seed state until later. <kimjetwav>KE0VVT: If you wanna get into writing guile vs shell, just like, give yourself some simple functionality you want to implement, and script it in guile. Maybe write a little command that would in some way benefit from free recursion, or whatever. <kimjetwav>Yeah that certainly does look like it's trying to be bash lol. <jpoiret>hmm, looking at my reflog, core-updates' been broken for a very long time <kimjetwav>Also, kinda tangential to guix, but has anyone worked with an implementation of srfi-166? <kimjetwav>I found myself about to suggest to KE0VVT using `format' for complex string outputs, but hesitated. Guile's implementation is fairly old and also not necessarily portable to other schemes. <futurile>jpoiret: that would be amazing if you could figure it out - I really struggled when I explored it last week <jpoiret>I'm going to write a small script to filter commits based on whether a commit with the same name was merged into master, that should narrow it down a bit already <futurile>jpoiret: the only thing I started doing was comparing to merges on master - oh you're there already heh <civodul>kimjetwav: i’d love to have that too! the pretty-printer in (guix read-print) could benefit from it, or at least from this approach <civodul>jpoiret, futurile: also, what about starting the rewrite work right before the first set of “duplicate” commits cbaines identified? <kimjetwav>civodul: I got reading about it a while back after comparing Guile and SBCL's implementations of format, and thinking "hmm I wonder what the current best recommendation for generalised string generation in Scheme is?" <kimjetwav>"Turns out, it's monads" -- some haskeller, probably <jpoiret>civodul: okay, I figured out where the big chunk of duplicated commits is, I think it'll be doable to clean up core updates in a couple of days <jpoiret>still, ~750 commits to rebase will probably hurt a bit, esp. since some of them need to be cleaned up because other updates on master have rendered them a bit outdated eg. disabled tests. <civodul>jpoiret: so you’re planning to start from the first chunk of duplicate commits, right? <jpoiret>i think there's only one big chunk of duplicated commits <jpoiret>between "72d13578ca * gnu: glibc: Add $TZDIR native search path." and "725b2528a65edf7425762b99330ab1c9bd33bc86 gnu: psmisc: Update to 23.6." <civodul>futurile: what are your thoughts? this is kinda putting on hold efforts to fix the packages you mentioned in your last update? <jpoiret>ah, I didn't see the request for merging <civodul>well, not really putting on hold, because fixes made now will still be relevant on the “cleaned up” branch <jpoiret>i'm a bit surprised though, because I could build up to gnome 1 month ago <janneke>is there a handy way to say something like `guix weather system.scm' <janneke>ACTION esp would like to know beforehand if substitutes for libre-linux are available <jakef>i was thinking about that too, but for guix home <apteryx>jpoiret: last I tried I got stuck on librsvg segfaulting during its test suite <apteryx>I suppose perhaps the merges of master into cu, which included rust upgrades perhaps played a role in it, but I'm not sure really <futurile>jpoiret, civodul: personally I would rather fix the commits so it's clean - it's only a couple of weeks - and then we don't have to live with confusion in future <apteryx>our librsvg is really old; I was trying to update it, but rust is a lot of churn. <jpoiret>futurile: I agree, but also it will be easy to move new commits to the cleaned-up branch. The hard part are old commits <jpoiret>so you shouldn't worry too much about waiting for that to happen, it can be done in parallel i think <futurile>jpoiret: OK - I'll keep testing core-updates - I think kaelyn sounded like they had fixed one of the issues we found <futurile>jpoiret: and then any 'new' fixes can be put onto the new branch when you're ready <blum>Hey everyone! I wrote a little chat program for fun and i'd like to run it over the network. I prefer securely isolating this in case it has an RCE. I'm already running this on a cloud vm, but i don't wanna allocate a full vm just for this. I'd prefer something lightweight and low ceremony. What would you guys recommend :D <futurile>blum: on a guix system or something else? Guix has system containers though I've never used them <pldmk>Question, has there been any influx of people due to the strife in the Nix community? <blum>i'll look into guix containers though <futurile>pldmk: don't think so - depends on your definition of influx - there's a couple of people on Mastodon that I've seen asking questions - but not that many I would say <futurile>pldmk: I follow #guix as a tag - seen a few things <pldmk>I'm not so tuned into all of the drama. All I saw was that apparently users/maintainers are leaving the project which begs the question "well where do they go". Guix, in my mind, would be only other similar alternative. <futurile>pldmk: I haven't followed the details of the issue - but a small number of people who are considering leaving mention #guix is what I've seen. I don't think these sorts of big kuffufles drive take-up in other projects that much - normally the community find a equilibrium <picnoir>I guess it's a better option for single-process containers <picnoir>It seems to be docker-based through, so your containers will likely be run as a privileged user. It'd be nice to have a bubblewrap backend for shepherd :) Could be a fun project if you have time/energy for it. <freakingpenguin>It'd be really neat if oci-container-service had support for different backends. I'm sure that's a ton of work though. <jpoiret>futurile, blum: the nix foundation has actually taken action and they've imo done enough to restore trust in the project. it's going to take a bit but I don't think there's going to be any exodus out of there <futurile>jpoiret: that's good - community is hard! <rekado>I also think that the response was good; this is a reminder that we need to continue our work towards better community processes. <jakef>wouldn't they sooner just fork nix than move to guix anyway? <blum>i might have copy/pasted a message by accident cus i didn't mean to post about the nix-thing :^) <jpoiret>blum: ah no my bad, I read the wrong username :) <jpoiret>woke up early and haven't had my tea yet sooo <futurile>oh that reminds me I need afternoon coffee <freakingpenguin>Morning coffee, second breakfast coffee, brunch coffee, lunch coffee, tea time coffee, dinner coffee, and bedtime coffee. <nat-418>I am working on a Nix meetup for June in Nice. Anyone from Guix want to come? <Tadhgmister>hello! I'm trying to cross compile `guix system image` to an arm platform by specifying the platform in the image declaration but it is failing with the same message that `guix build --target=arm-linux-gnueabihf -e "(@@ (gnu packages commencement) glibc-mesboot)"` throws, am I doing something wrong or can you just not cross compile glibc? <Tadhgmister>*can you not cross compile glibc or should a system image not require that to be compiled <futurile>freakingpenguin: come to the patch review session tomorrow which is online - it's lots of fun ;-) <civodul>nat-418: June is too busy for me but it’d be nice to see someone join y’all! <civodul>(also, Nice is the furthest city in France in terms of train travel duration :-)) <coyote>seems i've messed up and typo-ed the keyfile on my mapped-devices configuration <coyote>i would've expected it would ask me for the passphrase but it just panics :s <coyote>at least simply removing the extra initrd from the command line invocation lets me boot <coyote>hm, it tells me /keyfile.cpio not found, i wonder if this is because i have root in the "@root" subvolume <coyote>i wonder if doing `/@root/keyfile.cpio` would work <bienjensu>is this to workaround having to enter pass twice? <futurile>freakingpenguin: whoop - we can fake some sort of meeting if anyone turns up and sees you - I'll wear a tie or something! <twn>has anyone here used the service discovery feature of publish? trying to pinpoint what i'm doing wrong but not having much luck <zilti>twn: Checked all the firewalls that could be in the way, including on your router? I haven't used it, but for other zeroconf stuff, that was usually the issue for me <twn>only thing in between would be a security group and i'm allowing all protocols and all ports within that sg. so i'm fairly sure it's not an ACL <efraim>it looks like -machine virt isn't a great option for aarch64, then it can't use kvm <bigbookofbug>im working on creating a package module for nerd fonts currently and once it's complete i'd like to submit it to the package repo, but i'm having some trouble understanding the font naming guidelines <bigbookofbug>i get that the module would be defined as "font-nerd-fonts", but for the individual fonts i'm unsure if i'm following the correct name conventions <bigbookofbug>for example, jetbrains mono is defined as "font-jetbrainsmono-nerd-font" -- would this be correct? my first package so i'm still learning the ropes :) <futurile>bigbookofbug: reading the rules in the manual - what you're proposing makes sense <bigbookofbug>additionally, is the "fonts" part of nerd-fonts redundant ? it conforms to the name of the repository itself, but i'm wondering if "font-nerd" makes more sense <bigbookofbug>ok, awesome ! i had read the manual but couldn't connect the dots in my head for some reason <futurile>bigbookofbug: a committer will have a look before it goes in - if they think there's a need to change it they'll say so in the review - wouldn't get too blocked up over it :-) <bigbookofbug>oooooh ok that's good to know :) i was stressing over getting the conventions right, but if that's the case i'll just start diving in <futurile>getting the package in as best a shape as possible - solving those things so it's easy for the reviewer will be more important - they can always give you input on the name - or you can ask about it when you submit the patch if you're unsure <bigbookofbug>hoping to complete it this weekend - then comes the fun part of actually submitting it ! <bigbookofbug>i think its in good shape. its just somewhat of a repetitive process fetching the hashes <bigbookofbug>id rather package each font individually though, as the repo size is massive and i feel like most users will only want a handful of fonts rather than the whole repository <futurile>might have to be a nonguix thing - though it's confusing as they say that the licenses they found are all fine <futurile>I guess they meant they'd found <some> of <total> and while they were all fine, the <total effort> was going to be too much <lechner>Hi, I hardly use the "reified" strings to refer to packages, but I also do not assign new "names" when I create custom versions. Now I'm trying to pull in a specific version into Geiser via geiser-guile-binary. Is there a way to check which reified versions are available? <bigbookofbug>freakingpenguin: thanks for this! i hadn't seen this in my research so this is a huge help <bigbookofbug>may use it in a custom channel for the time being in this case, and once i'm a bit more familiar with navigating licensing i'll start wading through that <bigbookofbug>from what i'm understanding, it seems like pulling the licensing for the individual fonts, and the glyphs will be the bulk of the work then <dlowe>I'm trying to use fusermount in fuse but the binary isn't setuid and it's failing <dlowe>is this a packaging problem? <dlowe>oh, maybe I need to add it to setuid-programs <dlowe>no, the problem is that I have it installed both by the system and by the user and the (non-setuid) user version is taking precedence <dlowe>hm. I can't add libfuse.so.2 without adding the fusermount binary <freakingpenguin>Hypothetical question. My understanding is reconfigure is atomic so you never wind up with a broken system if e.g. power goes out. Does that also apply to bootloader upgrades somehow? <dlowe>theoretically possible with uefi <civodul>dlowe: what do you mean by “add libfuse.so.2”? <dlowe>I can't guix package -i fuse@2 so that libfuse.so.2 is in my profile without it _also_ installing fusermount in my profile. I figured out that I could set my PATH to give /run/setuid-programs priority <dlowe>it's not elegant, though *shrug* <dlowe>probably the nicer option is to have a lib output <civodul>dlowe: ah yes; i thought /run/setuid-programs came first in PATH by default, but that’s not the case, hmm <rekado>I find the cgit-configuration intimidating and there is no cookbook entry. Does anyone here have a simple configuration? <rekado>I'm looking to replace gitile, which I can't seem to operate without restarting it daily. <cow_2001>i have my own channel. how do i guix build --check all of those packages in the command line / guile script? <freakingpenguin>cow_2001: Take a look at (gnu ci)[all-packages]. That might help you get started on a Guile script. <cow_2001>oh boy. now i'm writing actual guix scripts