IRC channel logs
2025-06-27.log
back to list of logs
<ieure>Is there a Guix Home equivalent of `guix package -I'? <ieure>I would like to see the specific versions of packages installed in different home generations. <ieure>Okay, I see, `guix home describe --list-installed' <meaty>what's a good program to fill in PDF forms? <hlp>hello, I am trying to build a Guix package for a finalized SRFI that doesn't have an existing definition. I cannot get the package to compile with guild due to my lack of familiarity with how the SRFI code is structured. Are there any write-ups on how this works? <bavier>meaty: Gnome's Evince usually fills in PDF forms well for me. <dtx>when using "guix shell" when should I use -f vs -m ? They seem like they accomplish the same thing? <ieure>dtx, Similar, but different. `guix shell -m' lets you reproduce some other environment, ex. one created with `guix package --export-manifest'. `guix shell -f' lets you create a shell with arbitrary packages, including ones defined in the provided file. <ieure>Unless you're doing arbitrary-package stuff, probably easier to write a manifest and use that. <dtx>is there a way to define environment variables inside of the guix shell from manifest.scm? I would like to append to PATH but not wipe out what's there, and I want to call commands in the shell directly not just open up a subshell in it. <dtx>in a container shell I mean <ieure>dtx, I'm not aware of any facility which lets you define environment variables from either a manifest or a file of packages. You can use --preserve / --expose to set environment variables from the parent shell in the container, but you can't append to the $PATH the shell creates. <ieure>dtx, This sort of thing is generally unnecessary, since `guix shell' is designed to run guix-packaged software, and any packages available in the subshell get put on $PATH automatically. <reepca>two things I have learned today: 1. in 2025, glibc still uses socketcall on x86 unless explicitly configured to assume a kernel newer than 4.3 2. strace will lie about this and pretend it's using whatever the multiplexed system call is instead <reepca>looks like glibc by default doesn't assume any kernel features newer than 3.2.0 <reepca>I had assumed that it would be designed to first try the dedicated system call, and then if that returns ENOSYS, fall back to socketcall <untrusem>dtx: I also had to use php for a project, but it seems no one does php in guix, resources are very scarce on php <janneke>ACTION works to trying to resurrect cross build of guix [to the hurd] <janneke>adding a guix-for-build to the guix package... <Noisytoot>guix pull: error: aborting update of channel 'guix' to commit bc3ce40ee72d10e180afd32786d6b5fc7ad02dfd, which is not a descendant of ad8cb7af8fc572bb3d11cd0344bd2bed3b9d653f <Noisytoot>hint: This could indicate that the channel has been tampered with and is trying to force a roll-back, preventing you from getting the latest updates. If you think this is not the case, explicitly allow <mange>There was a commit accidentally pushed to Savannah after the Codeberg migration. I think you can either allow downgrades, or you can "guix pull --rollback" then "guix pull" again. <janneke>yay, by guix-for-build guix-1.4.0-39.837ddb0 finally built...now seeing if the hurd cross build fix actually works... <janneke>hmm; build: Do not generate 'CODEOWNERS' when cross-compiling. <janneke>that's also a "solution" of course... <janneke>oh welll, why not; probably better than my configure/make/guix guix-for-build work <janneke>it would be nice if they would align things on irc... :( <civodul>oh you were also working on fixing this? <z572>janneke: I think we should communicate under codeberg's issues, such as setting "Assign to me". Since no one was going to fix this problem, I went to fix it <janneke>civodul: yeah, someone advised me to reconfigure asap...but i my childhurd's don't build yet <janneke>z572: ah, there already was a bug opened for this...my bad then <janneke>ACTION hasn't really gotten into the codeberg thingy <janneke>ACTION agrees that aligning on irc can only be a courtesy and hardly be mandadet...but it used to work pretty beautifully before, most of the time... <z572>janneke: I think it's mainly because our time zones are different. It's 4 p.m. here now. :) <janneke>z572: yeah, that probably also makes alignment more interesting :) <janneke>i've just pushed a guix package update so that it may actually be effectuated for a system reconfigure <z572>i think a idea is add a page on guix.gnu.org about current gcd status? <futurile>z572: yeah, Noe has worked on something, I think they're close to being able to do that <civodul>we could have it at consensus.guix.gnu.org <z572>I'm very glad to know this <z572>maybe add a WIP pr to #guix/artwork? It has always been very difficult for me to search for content in irc <z572>daemon: conditionally disable seccomp filter on socketcall systems <noe>We can transfer it to Guix <noe>civodul, if we set the gcd frontend on Guix infrastructure how do I update the site? Would I need to ask every time for someone to push the changes? <futurile>noe: you'd be added to the infrastructure repo (or website one) so you could push updates <yelninei>janneke: thanks for the libgit2 revert and the guix update. Also cross compiling guix should now be a lot faster and it is not taking 30mins to translate the manuals <untrusem>How do we folks use guix shell? I mainly use it to open program like librewolf, keepassxc etc. <untrusem>What would be the benefit of them opening in guix shell rather than using them in a home manager? <futurile>untrusem: if you're concerned about security then guix shell isn't really designed for that. It might provide a bit of protection since you can limit what apps can see. But, it's not a security tool like some of the others out there <futurile>guess it depends on how secure you want to be <untrusem>One advantage I see of that is that their updates won't hinder reconfiguring update when we have no substitue <ieure>untrusem, I use guix shell for developing guix itself, for testing packages I write, for running software I use infrequently enough that I don't want to install it. <ieure>untrusem, I wouldn't use it for programs I use every day, since it makes the startup slower and much more variable. <untrusem>i see, So it's not isolating like flat back <futurile>untrusem: correct - I think it would be fair to say that Flatpak was designed with isolation in mind. I assume someone in RedHat / SUSE has looked at it reasonably deeply. <futurile>untrusem: whereas guix shell was/is really designed for running an application easily - technically it's not a security tool <untrusem>Can we have a security related program based on geeks? <futurile>untrusem: honestly, I think people go off the deep end iwth regards to security, so I'm personally comfortable using guix shell and I actually don't limit my browser <untrusem>futurile: I see, the reason being I am very fascinated by qubes <futurile>untrusem: we could if someone was willing to do the work - they're both using the same underlying Linux tech (namespaces etc) <untrusem>so it was either qubes or guix for my thinkpad t480 and I chose guix :) <futurile>untrusem: yup, qubes is a nice idea - I read their web pages, I just haven't spent any time on them myself. <futurile>untrusem: I loosely looked into integrating bubblewrap or one of the security systems into Guix and decided I wouldn't be in it for the long-haul <futurile>untrusem: funnily enough it wasn't something that could be done on a Friday afternoon over a cup of tea and a biscuit heh heh heh <ieure>Qubes is a cool idea that really doesn't fit how I like to use the computer at all. <futurile>do you run any of your apps jailed in some way ieure? <untrusem>ieure, qubes is very cool thought, but would need a hefty machine <untrusem>it doesn't very much with permacomputing.net <futurile>yeah landrun looked nice from an hour of browsing <identity>running everything in a vm sounds like a bad idea unless you have a lot of ram and do not know what to do with it <ieure>Well, I assume they're using built-in namespaces to run programs directly, which is about as lightweight as you can get for this sort of thing. <ieure>It seems like a fine tradeoff for those who want the strong isolation it offers. <reepca>having "a lot of ram and not knowing what to do with it" actually prompted a rather funny thought experiment <ieure>Just a taste thing, but I prefer the big blob of junk where all the bits can talk to each other. <reepca>which is cheaper on resources: running modern software bare metal or running software from 2005 in a vm? <janneke>yelninei: yw, hope it helps. trying to build both guix's at the moment as it seems the build farm hasn't picked/caught up yet? <identity>ieure: Qubes runs on Xen, which is essentially a bare-metal VM manager; you can run whatever OS under that, Debian, Fedora, windows, whatever <identity>(including multiple of them under any combination at the same time, if that was not clear) <ruther>wondering what a good way to not crash services when compositor exits... currently it kills my session. Should I not use exec and let bash be the managing process, or are there better ways with elogind? Seems that simple enable-linger doesn't allow persisting shepherd, shepherd is still killed <futurile>viaken: yes, I think firejail is still around. When I searched firejail, bubblewrap and landrun were all mentioned - it was difficult to tell which one(s) were best or most integratable <jakef>civodul: i nominated you to review a PR updating insight-toolkit because image-processing.scm doesn't have a team, and i saw that you and Ricardo are the ones who've contributed to ITK. not meaning to come across as impatient; there's no urgency whatsoever <janneke>hmm, libcap doesn't [cross-]build any more for the hurd <janneke>...or is it somehow a new dependency <reepca>looks like it's a dependency of slirp4netns, but slirp4netns should only be included in inputs when (target-linux?) is true <g_bor>I am trying to set up a build from git on my side, and I have two test failures: <g_bor>guix pack with appimage+localstatedir with this error: Failed to run /tmp/guix-build-check-appimage-with-localstatedir.drv-0/appimage_extracted_3489228c7b247db83317da0bda48c8a0/AppRun: No such file or directory <g_bor>location: /home/gabriel/guix/tests/derivations.scm:99 <g_bor>+ (build-derivations %store (list drv)))) <g_bor>also from this second one, how could I get more info about what went wrong? <ieure>g_bor, You need to use a pastebin service, you'll get kicked if you spam the channel. <janneke>ACTION cannot find another drv dependency that wants the slirp4netns derivation <g_bor>do you have a recommended pastebin? <ngz>Hello. From the Codeberg web UI, how do you comment about a particular commit it a PR page? <yelninei>janneke: I have seen something similar with guile-avahi which should also not be in the manifest for a hurd target <g_bor>I hope I managed to catch all relevant info <g_bor>ngz: do you have link to a PR with a comment? <janneke>reepca: found it...i'm including in my chilhurd-devel-os all guix depedencies...to do guix development; something like <janneke>(fold alist-delete (package-direct-inputs guix) '("graphviz" "guile-avahi" "imagemagick" "po4a"")) <z572``>maybe `guix graph` should support --target. <apoorv569>Is it possible to create a LXC image using `guix system image`? I tried building tarball image and I was able to load it on Proxmox, but it fails to boot from the image. <apoorv569>something filesytem related? or init system? <g_bor>ngz: on a regular comment on the discussion flow I use 'quote reply' from the ... menu on the comment, but I don't see one with a review comment <reepca>janneke: or better yet, wrap (package-direct-inputs guix) with (parameterize ((%current-target-system "i586-pc-gnu")) ...) <g_bor>ngz: this looks to have only one commit, let me try something <ngz>Yes, but the comment points to a precise part of that commit <g_bor>It was merged last week to forgejo, I think it would take some time to deploy this on codeberg, but I might be missing something <g_bor>ngz: I might be misunderstanding what you would like to achieve, would you like to know how to create these comments that point to a single line in the change? <g_bor>so, on the ui where you are on the diff view there is a column where the + and - is displayed for the added and the remove lines. This on hover changes to a blue +, and if you click on it then the comment dialog pops up <g_bor>once you have the comment you have to options to go on, either you can start a review, which delays the comment submission until you do a finish review, alternatively you can add it as a single comment <ngz>g_bor: I’m looking into it. Thanks. <ngz>This does help. Thank you! <oriansj>guix perform-download silently stalls forever behind certain types of proxies and doesn't have a timeout <oriansj>the stall seems to be most common with mirror: packages <Deltafire>pdfarranger can no longer save pdf files.. anyone else noticed this? <g_bor>Hello, I am looking into a warning coming from our build. I have a package with a custom configure, and I get the warning that enable-fast-install is unrecognized. The package builds all well, so this is aesthetic only. We add this flag in gnu-build-system unconditionally, wdyt about having a way to disable this? <ieure>g_bor, Seems straightforward, put up a PR? <g_bor>configure: WARNING: unrecognized options: --enable-fast-install <ieure>g_bor, Can you modify #:configure-flags though? <ieure>Or does it get appended to that? <g_bor>ieure: sounds good, I will create a PR. Actually in the build system we build up flags, where we add this, and then `configure-flags` is appended to the end <ieure>g_bor, Yeah, was just looking at the code. <ieure>g_bor, You can also replace the whole 'configure phase. <g_bor>ieure: yes, for this package it is super simple, why I think this is something we might consider is that there might be way more complicated configures, where it would be nice to be able to keep it. Most probably the best way would be to make this smart. The comment says it is there when we use libtool. It should be relatively straightforward to <g_bor>put a condition on that, which would result in it being usually the correct setting. If it happens to be bad nevertheless we can replace the phase in those specific cases. <ieure>g_bor, I think you should open a PR! <g_bor>ieure: I assume against core-packages-team branch, this should trigger almost a world rebuild I think, is this correct? <ieure>If configure-flags changes, I think that will cause world rebuilds, but if you change how that value is computed, I think it only changes for packages that trigger the behavior. I could be wrong!