IRC channel logs
2025-12-02.log
back to list of logs
<mwette>I've install binary guix via guix-install.sh. I'm trying to get the daemon to run as guix-daemon but it's not working. I don't see a User= field in /etc/systemd/system/guix-daemon.service. Shouldn't I see that? <lilyp>mwette: did you use a new guix-install.sh? If you've been using Guix for a while you might have accidentally used an old file. <mwette>pretty sure I used a new one. Do I need to "guix pull" as root first? <mwette>shasum guix-install.sh -> 7139938... <lilyp>sha1 appears to match, hmmm… <lilyp>yup, line 481 looks at ~root/.config/guix/current <lilyp>so a root pull is needed to get the correct service file <lilyp>on foreign distros, you should do root pulls anyway to upgrade the daemon <mwette>Is there a reference for how mounting of /gnu/store before/after guix-daemon is started is supposed to work? I installed with /gnu/store in /root but now want to transfer to another (ext4) filesystem. <Rutherther>mwette: the file system doesnt matter as it is mounted through bind mount <basicnpc>Has anyone used guix pkg manager for macos successfully? <csantosb>When one closes a pr, say #3655, the codeberg ui automatically closes an issue, #2189 in this case <csantosb>I wonder how this goes exactly, other than the "Fixes" word in the pr, there is no link <flurando>I set up a cuirass service to compile my own channel. But it has not succeeded in computing once, always saying fail to compute Guix channel xxx. I even renamed all my services to avoid naming conflicts with guix's. <flurando>Is there something that I missed, or does it has 0 relevation with the naming of my custom services? For example, a hard forked nix-service-type, I ended up to surfix it with my nickname, but still, the build failed. <flurando>Or maybe it actually has minimal system requirements, as my machine only has 2 vcpus and 1.7G usable RAM. <gabber>is it me or is my standard `guix shell -D guix -- make -j $(nproc)` doing more work lately? i built the tree to push some commits, changed the branch, rebased on the previously merged master (so that i am now 1 commit ahead with a simple package change) and the invocation above runs the configure script and compiles all the .go files... <flurando>gabber: No idea, personally never touched guix development <flurando>Also, services defined in personal channel does not appear visible by guix system search, why? Is there any way to make it work? Currently only packages being searchable. <gabber>flurando: cuirass jobs are a bit tricky. i had my issues when i set them up myself. maybe sharing your exact job specification here could let us help you better? <gabber>also: have you checked the logs? there's quite a bunch to read in /var/log/cuirass* <gabber>(how) can i @ a whole team in a codeberg issue discussion? <csantosb>gabber: This is usual, when you change branches and come back again, you'll need to rebuild part of the .go files again ... <gabber>csantosb: even when i effectively only changed small part of 1 file? <flurando>They are all the same content, let me open the pastebin <flurando>all 300+ error logs are all this, with different guix commit, though <gabber>flurando: your exact service specification would be good, too <gabber>flurando: so it tries, but fails evaluating. that's half the job, already (: <flurando>gabber: thanks for checking, this is my config <gabber>can't check rn, am afk for at least 1hr <flurando>Note that my machine only has 2G RAM, which is absolutely not enough if it tries to compile linux-libre or so from Guix channel. However, the strange point is the log actually does not say much, expect "report this through email" <flurando>Or maybe it is because my services definition is actually faulty, in a manner crashing official guix from been computed. Which is why locally I can guix search my packages but could not guix system search my services <csantosb>gabber: IIUC, it depends on how different your branches are; if they only depart on one file, you'll need to produce a single .go file upon branch switching <csantosb>I just checked: if changes appear in a single file, .go files are quick to build <csantosb>Do we have a lispy way to refer to #:tests? inside #:configure-flags ? The idea is to adapt the flags following the check status of the package. I'm pretty sure I've seen this somewhere. <flurando>csantosb: maybe consfigurator, the declarative configuration in Common Lisp, have similar functionality implemented. A lispy way, which might be ported to Guix / GuileScheme. <csantosb>Someone to reproduce a bug with Guix native ? See #4619. <test2020>hi, somebody can suggest useful for emacs for guix begginer packager <csantosb>test2020: try this, `./pre-inst-env guix refresh -u -s module:"(gnu packages emacs-xyz)"` <csantosb>Then, pick your favourite package and try to upgrade it <test2020>no, i mean emacs user. i planned to package another package <csantosb>Filter the guix log history by "emacs", and see previous contributions to have an idea <test2020>mean that i am new packager for guix, using emacs and want useful tools for help with packaging <test2020>looks similar. geiser working under tramp on remote machine? <gabber>csantosb: thanks. that's how it used to/is supposed to work. now i'm wondering why it does not do so on my end (: <gabber>flurando: i think there is a (list) missing. e.g. it should be (build '(channels (list flurando-channel))) somewhere where `flurando-channel' can be evaluated to something, like in a let form. let me show you how i did it on my end. <gabber>well wait. now that i check, my cuirass doesn't build that job either, but at least it doesn't crash.... <gabber>let me check something real quick <flurando>gabber: Thanks, I am looking at your example. <flurando>The document would need an update as my config is copied from the guix devel manual online <flurando>Have deployed the new config with (channels (list xxxxxx-channel)) instead of (channels xxxxxx-chanel), would check later <flurando>I used to think, judging from the document example, that cuirass would looke for the symbol as name of given channels? I am wondering if only add list without using let to define local variable would be fine. <flurando>Here let is actually meaningless, as all variables are only used once and build seems to accept a quoted (no interpretation) anyway. <dcunit3d>does someone have a minute to take a look at this system specification? it runs, but there are a few wierd issues. i don't use if often, but i am close to using it to manage x509 stuff with smallstep <dcunit3d>i think the specified services are way too minimal, but `herd status` never returns for the root user. when i reload gpg-agent instances that start up under the 1000 user, it leaves defunct processes that i can't clear out <dcunit3d>#72166 mentions a lack of RTC battery... which this laptop also lacks. <dcunit3d>i usually need to address that when i boot it up. it's on AC power now. i set it with `hwclock --systohc` and IIRC it holds until i pull power <gabber>sneek, later tell flurando: would wou mind opening an issue on Codeberg about that documentation bug? <dcunit3d>yeh, seems like the RTC battery was the issue <look>For running 'make pdf' do I need the whole texlive? Which package do ya'll shell for just compiling the guix manual pdf? 4.3GB seems a bit excessive... <csantosb>Let me share the recipe to build the epub <csantosb>TEXINPUTS=".:.:" MAKEINFO="makeinfo --force --enable-encoding -I ." texi2any --epub3 ~/guix/doc/guix.texi <gabber>i'd have guessed that we make good use of our atomically separated texlive packages for that <fhang>Hello all, I'm new to Guix and I'm wondering if there is an official Guix Docker image? <meaty>fhang: no official one I believe <fhang>meaty: Thanks. I guess I'll make my own Docker image using the tarball. <kestrelwx>For `ant-build-system` if my `build.xml` is inside a subdirectory, do I change directory in a phase? <csantosb>fhang, you can easily build a custom docker image using `guix pack` <fhang>csantosb: Awesome, I'll take a look. <basicnpc>Has anyone used guix pkg manager for macos successfully? <nikolar>guix only supports linux (and hurd experimentally) <ieure>basicnpc, It doesn't run on macOS and many very difficult things would have to change for that not to be the case. <ieure>basicnpc, Issues are, Guix only wants to target glibc, and there's no glibc port to Darwin; and the toolchain to compile it is bundled with Xcode, which is non-free and requires acceptance of a EULA to install. <ieure>basicnpc, Here's a thread about it, it's old, but none of the fundamental situation has changed. <ieure>basicnpc, There's a thing someone has made, "Guix Subsystem for macOS," I believe it's a VM and wrapper to run Guix inside it. <basicnpc>I wonder why anyone would like to have VM in their macbook.. <Deltafire>for docker, or windows, or many other reasons <meaty>what would be a good scheme for declaratively deploying "self contained" web apps on guix <meaty>where the "source" files, configuration, and some state are all in the same tree <ieure>meaty, You're asking for "a good scheme" here, like, a plan? Not "a good Scheme," right? <ieure>meaty, The easiest thing to do is run official Docker containers with the OCI services. <ieure>If you want to not do that, you need to start by packaging the applications. I assume there are no packages for them yet. <meaty>Yes, what I'm thinking to do is to package the "clean" tree and keep it in the store, then use a service to copy it over to the desired spot in a web server and configure it. but this feels naive <ieure>I agree. It looks like there are build systems and dependencies involved; phpBB has a node package.json file. So copying the source tree into the store probably won't work <ieure>Also you shouldn't make a copy at all, you should point the web server at the stuff in the store. <meaty>does guix have a way to "capture" these? Can a store tree symlink to /var/...? <meaty>i.e. capture attempted changes to a tree and put them somewhere else <ieure>It would be an exceptionally poor idea to do that. <ieure>I assume these have to have some way to configure where that stuff goes, your service will need to twiddle those knobs to point it to somewhere in /var/run. <ieure>meaty, I assume there's already a solution for these webapps, because I assume they have to run in containers; and the problem needs to be solved in that case, too. <meaty>Oh, containers are immutable? I'll look in that direction then <ieure>meaty, Containers aren't really immutable, but they don't persist any changes written to the FS, and it's very poor practice to do so. <basicnpc>Deltafire: Could you elaborate more? What do I get to put a guix VM on my macbook!? <basicnpc>I'm quite noob, and am genuinely interested in that. <getstate>Is there a way to use an inferior in a channel? I'm trying to use it to grab rust-1.90 for a package, but using it in cargobuildsystem causes the channel to fail to build <getstate>I'm guessing this is not possible? It tries to get the package definition during the channel build eagerly <ekaitz>getstate: isn't that what time-machine does? or I didn't understand what you mean? <Rutherther>...are you trying to put rust from inferior to #:rust argument of cargo-build-system? I am afraid it will not accept an inferior package <getstate>Rutherther: that's exactly what I'm doing inside my channel. <Rutherther>it has nothing to do with your channel. If it is unsupported, it's unsupported in general <getstate>It's trying to eval it, I get an exception with "store-connection-error" during the channel build phase <getstate>how would one normally do this? Copy/paste and vendor the package? <Rutherther>have you considered not making it a package but rather a procedure that would output a package? That way the inferior shouldn't be evaluated right away, I think, but I am not sure now. But definitely first try without any channel builds, because as I am saying, I expect cargo build system is not going to be accepting inferior-package <getstate>No, but that sounds like a good idea! Is there an example you know of? <Rutherther>getstate: no, but it's just wrapping the thing you have in a procedure and returning the package as a result. Make sure to also put the inferior for channels and so on in the procedure <dckc>I'm struggling to push to codeberg. anybody else see related symptoms? <fhang>dckc: It seems Codeberg is down. The site is currently saying "Service unavailable". <dckc>thanks for confirming that it's not just me <gabber>how can i set a LD_LIBRARY_PATH when installing a package? i have a python package that needs to know where libngspice is <Rutherther>if it's packaged through guix, then it's wrong. It should've been patched. If not, still, I would recommend to refrain from setting LD_LIBRARY_PATH globally as that WILL almost definitely lead to issues in the future for you as you will be getting wrong library versions in Guix programs you start. So just setting it manually when starting the scripts/app / wrapping what you are starting with LD_LIBRARY_PATH. <gabber>Rutherther: they are both packaged in Guix (this is what i'm doing right now) <gabber>so you say i should look into the sources and patch the occurrence of libngspice.so with its full path? <gabber>but if it's not more than the usual (substitute* ...) i'm fine <Rutherther>right I haven't realized I copied from a private branch <Rutherther>just substitute that fix/mbr-hybrid part with master <gabber>i don't think the line numbers match still <gabber>nice. this seems to work, except <gabber>running the package's own `pyspice-post-installation --check-install` manually in a shell with only that package works just fine, but doing the same in a separate build phase after 'install fails. any ideas on why? <ieure>gabber, What's the error when it fails? <gabber>it doesn't find its own module(s)/package <gabber>"importlib.metadata.PackageNotFoundError: No package metadata was found for PySpice" <ieure>Sounds like PYTHONPATH is messed up in the build env. <gabber>yeah, codeberg seems to be having issues atm <gabber>not even their status page is available <Rutherther>the resulting package will not be in pythonpath, you would have to add it manually