<euouae>I don't know why you specifically needed mrustc to be honest
<bdju>does dolphin-emu still fail to build? I guess I'll find out shortly
<ryan77627>nckx: I thought I lost your response bc I was moving to a different location and ZNC doesn't save mentions, but I found another instance of weechat I left open, woohoo! I do think that I'm gonna have to just bite the bullet and set up my own cuirass to do this. Really the only reason it's needed is bc I'm trying to set up a demo of Guix to some members of a LUG I'm in and I want to make the
<ryan77627>to that note, i wonder if I can just make guix publish actually advertise the exact derivation I would be building and just have guix use that. Or, before the presentation, I could just pull a sneaky and run guix pull beforehand so I hit the ground running
<luke-jr>is there a recent commit that can build all the way through? :\ or is this the norm and my issues more likely due to non-functional chroot?
<efraim>I have a use case for trying to use glibc-hwcaps but I'm haven't figured out how to get the binary to pickup the other libraries. Has anyone else experimented with them yet?
<efraim>I figured it out. The short version is that first you build the optimized libraries with the LIBDIR and RPATH (in a separate derivation) pointing to the glibc-hwcaps directory, then you build the actual package, then you copy the optimized libraries into the output of the actual package, keeping the glibc-hwcaps paths
<bost>Hi geeks, `guix pull` reports "GC Warning: Repeated allocation of very large block (appr. size 69632) May lead to memory leak and poor performance". What does it mean? What should I do about it?
<janneke>bost: thanks for reaching out, always good to get feedback
<janneke>and i hope these memory troubles can be tackled some time soon
<cbaines>zimoun, processing has started for job 49353, but it hasn't finished yet
<cbaines>if you refresh the log in a bit, hopefully it will have got a bit further
<zimoun>cbaines: all the warning or error messages are not related to #66030 which about Julia stuff.
<zimoun>I pick one: guix-data-service: computing derivation for nfs-server system test (on aarch64-linux). And it does not make sense to run system test for aarch64-linux when it is about Julia.
<zimoun>Or another example: guix-data-service: error computing derivation for system test hpcguix-web (powerpc-linux):
<zimoun>I read: “@ build-started /gnu/store/9f68l6wnh7gmpmmx7jlv38h7z3yqv31n-guile-bootstrap-2.0.drv - mips64el-linux“ And while mips64el-linux is interesting to test, the process seems to run a lot more than needed.
<cbaines>the data service isn't running any system tests, it's just computing the derivations
<cbaines>unfortunately due to graft issues, it can end up building some things though
<zimoun>and the effective build has not yet started, right? It takes several hours just for computing all the derivations, right?
<cbaines>it usually takes several hours at least for the data service to process each revision, yes
<cbaines>and part of that is computing derivations, although I don't know what proportion
<lars__>Is there a list somewhere for what hardware is recommended for guix with linux-libre? I'm thinking mostly laptops, but desktop computers with "gaming" GPUs also. And are there any embedded/ARM platforms that have good support under Guix/Linux-libre?
<lars__>I'm guessing there isn't a lot of GPUs that are really supported
<zimoun>cbaines: thanks for explaining. All is clear for me now. :-)
<lars__>I would like to switch to linux-libre, but seems none of my current computers are well supported
<efraim>If you just need something to display the desktop (which is where I find myself these days) I'm currently using a GeForce GT 730
<lars__>efraim: Is it fully supported, i.e can I run native resolution and watch youtube?
<lars__>I do have a GT 1030, but didn't test it on guix yet. Do you think it is usable with only free software?
<efraim>it works for me™, and it worked well enough that I bought another one to replace mine, which is starting to fail
<efraim>forthe GT 1030, I can give you a solid "maybe"
<efraim>I'm checking wikipedia, but h-node would probably also know
<efraim>As of 9 July 2016, Red Hat employee Ben Skeggs committed a patch which adds support for the Pascal-based GP104 chip found on GeForce GTX 1070 and GeForce GTX 1080-branded graphics cards to the Linux kernel. https://en.wikipedia.org/wiki/Nouveau_(software)#History
<cdo256>On a new install, I forgot to start the cow-store before running guix system init, causing it to fill up the RAM disk. Even after starting the cow-store, I can't run `guix gc` to fix my mistake because 'database or disk is full'. Anyone know a quick fix for this situation?
<cdo256>Ah screw it, I'll restart and go for a second approach
<makads>nckx: Because my country has a network blockade, I have to use something like VPN to proxy traffic when I want to use IRC chat rooms, which is very inconvenient，And my native language is not English, so I have to rely on translators to chat
<euouae>Hello, I'd like to read the documentation of `use-package-modules`. I thought it'd be a symbol from the gnu module, imported with `(use-modules (gnu))` but `,describe use-package-modules` gives me "unknown file:2:10: source expression failed to match any pattern in form use-package-modules"
<minima>i'm testing a home configuration record: `(serialize-configuration test-configuration test-configuration-fields)' this produces a gexp, i guess? how do i go from there to inspecting the final configuration file in the store?
<euouae>Sure thing nckx. Is (use-package-modules certs) using the file certs.scm under guix/gnu/packages?
<minima>i seem to be able to produce a derivation from the gexp
<rekado>apteryx: I can relate. I’m switching between a bad laptop keyboard, a Kinesis Advantage2 in the office, and an atreus with a compact layout.
<zamfofex>I’m continuing to work on this: https://lists.gnu.org/archive/html/guix-devel/2023-09/msg00458.html How do people usually deal with changes that affect many packages? (E.g. build system changes.) I guess maybe a sensible approach would be to just let it run on a server, but the unfortunate thing is that I’m never quite sure about whether I’ll want to make more changes. It makes it awkward to prototype ideas and try things out.
<lispmacs[work]>hi, I have been running gnome/gdm, and recently I added lxqt service. But I don't see it showing up as an option on the gdm screen (I don't see any options). I see an lxqt.desktop file in /run/current-system/profile/share/xsessions so I'm not sure what I'm missing
<euouae>in po/doc/guix-manual.es.po there's an msgid that corresponds to `... initial RAM disk, and boot loaders ...` but in doc/guix.texi it's been modified to `... initial RAM disk, and a couple of system ...`
<euouae>Doesn't that mean the translation won't work because the sentences are out of sync?
<somenickname>euouae: Don't know if on Guix I can just edit that file. Also because GDM in this loop state I can't do anything on the computer (changing TTY and so on will always force me back to the GDM VT7 display)
<rekado>graywolf: does this work for you? guix shell -C man-db pulseaudio -- man pulseaudio
<graywolf>rekado: I see, the mandb needs to be in the shell profile itself. Will remember that, thank you.
<gnucode>hello from the hurd on real hardware people! also what's the status of running guix system on power9?
<zamfofex>gnucode: Are you on Guix Hurd or Debian Hurd? 😮
<gnucode>mirai: not your question, but I use debbugs-gnu-my-open-bugs a lot. It shows me my open bugs that I submitted and forgot about. :)
<minima>what's the best format for a commit message, say (purely speculatively) when changing /gnu/system/images/hurd.scm? `system: hurd: Done this done that.'?
<bjc>mirai: i don't know about simple, but you can use the emacs flow with patch testing if you just pretend it says "guix" instead of "emacs". personally, i just save the attachments and apply them by hand, since i can never remember the keys to do it the "right" way
<minima>is there any best practice on the `system: hurd:' bit?
<yziquel>is there any way to get left right up and down keys working in the guile REPL ? at one point in time, ocaml had ledit to make the REPL more user friendly. any similar thing for guile ? google doesn't provide a good answer.
<janneke>yziquel: what nckx says...but i guess that most just rely on emacs for their line editing support
<yziquel>janneke i like the convenience of working over ssh, raw.
<nckx>rlwrap is handy but gets confused by ‘guix repl’.
<nckx>In the sense that it will create its own .guix_history file rather than using Guile's. Might be an issue for you, might not.
<yziquel>nckx ok. I'll keep guile-readline in my todo notes
<nckx>It's an extremely minor issue, some won't even notice, it would just bug *me*.
<janneke>yziquel: sure, whatever you like :) emacs works great over ssh too
<lechner>yziquel / just install guile-readline and put (use-modules (ice-9 readline)) (activate-readline) into ~/.guile
<bumble>would there be any benefit to trying to run k3s on guix? or does guix provide things that make kube un-necessary?
<bdju>regarding dolphin-emu failing to build, sounds like there was an upstream issue with our current version of 5.0-13178 that was fixed in 5.0-13669 so a version bump may be all that's needed
<gabber`>nckx: in regard to building an interactive map of the Guix source tree of sorts ... i figure it wouldn't be *that* hard to do if we could reuse for example ETAGS.. unfortunately i have not the slightest clue of TAGS internals
<catball>no worries @nckx! thanks for the help you both. removed glibc-locales and rebooting it for good measure
<nckx>Oh oh oh, what this could be is a version mismatch. Make sure you've run both ‘guix system reconfigure’ and ‘guix upgrade’. Glibc's locale data is not stable across versions, so a tmux linked against a more recent glibc won't be able to read older (system) locales.
<bdju>nckx: dolphin-emu releases are weird, their last tagged thing is many years old. anyway newer is better if it actually works, but when I tried the master commit it didn't work. wasn't sure what else to try exactly besides the commit that supposedly fixed the issue our current version was having
<bdju>and if you didn't mind doing another version bump thing, sadly for another weird-release case, I've been manually installing a newer pfetch for months to get support for the PF_COLOR=0 env var so I can have it output in monochrome. sadly that guy is also allergic to cutting releases even once a year. we currently seem to package the commit where the guix logo was added
<bdju>I think I just run the master commit for that (from 2021...) and it seems okay
<nckx>OK, no, I'm not going to update Dolphin-emu this much, no. -_- I'll try your older commit.
<bdju>ah I didn't realize how old the commit I found was
<bdju>yeah would probably be good to get it working again short-term and someone feeling ambitious could sort out a newer version later
<nckx>For context, your commit would be revision 13669, where Guix is at 13178, and upstream is currently at > 19000.
<nckx>So this should be at loss less painful, thanks :)
<hrn>nkkx: I still get the same error message after 'guix pull'. Also if I try to build guix without the --pure option to guix shell I instead get "configure: error: 'guild' binary not found; please check your Guile installation." even though the build binary is in the path. I am very confused now.
<nckx>Did you make clean & configure <OPTIONS> again?
<nckx>(Most clients support tab-completing nicknames, by the way.)
<nckx>Actually, since you were working from a clean checkout, just clone a new, fresh copy and bootstrap.
<GNUtoo>hi, Is there some good practice for deploying Guix VM in libvirt qemu when the host doesn't have guix installed?
<GNUtoo>My situation is that I've some access to libvirt, and Guix on my laptop to produce an image (with guix system init)
<GNUtoo>And also a shell on the host but without root
<GNUtoo>Do people in that situation simply build the image on their laptop and copy that and import it to libvirt somehow? Or do people build and run guix without root? Or do they use VMs with the guix daemon inside to do the first installation?
<hrn>nckx: After starting from a new clean clone I can build again! Thanks for your help (I need to be building in a --pure shell, but that's ok I guess)
<minima>does `configuration->documentation' work for home services?
<nckx>euouae: There's ‘guix shell -p’, but yeah, profiles aren't really ‘switched’ and I don't think we should over-abstract them (guix shell's abstraction level is IMO perfect, but so is sourcing them directly, maybe using env -i).
<nckx>‘gnucode’: The problem with zero-friction Guix hosting is that there's little to blog about beyond ‘yo check this out’. It would be little more than an ad, but maybe that's OK? In a way, we rewarded Linode for not better supporting Guix.
<nckx>Blog aside, having a list of providers that ‘support Guix’, like Let's Encrypt has, might be useful. We'd just need to define a consistent level of ‘support’. Oh, and someone needs to write it, but that's the easy part, right…?
<euouae>Should the manual prefer guix describe -f channels or guix package --export-channels?
<nckx>guix package --export-channels should really only print channels that were used to *create* the profile, i.e., it describes said profile, while ‘guix describe’ describes the channel that constitute that ‘guix’ command.
<nckx>E.g., in my case, ‘guix describe’ lists 4 channels because that's how many go into my guix, but ‘guix package’ lists only 2.
<euouae>so export-channel is a subset of describe?
<euouae>You'd get it if you used `guix package --export-manifest`
<nckx>That doesn't mention ‘guix describe’ at all here.
<euouae>Hm. not sure. I'll just move on from this, it's not important
<euouae>ACTION should think more about the difference between --export-channels and -f channels though
<nckx>Maybe that's because my profile is a bit more, er, heterogenous than yours? Maybe if your entire profile was generated from a single commit (e.g., with a manifest) it would print something else than mine does. I dunno.
<euouae>The (define* (export-manifest line in guix/guix/scripts/package.scm contains a format port (G_ "... with the comment I'm mentioning
<euouae>I don't see a reason why it wouldn't be printed.
<nckx>Here's my output, note the lack of overlap in any commit, and the fact that you can't slap a single pretty commit on my user profile, so Guix awkwardly lists them in a comment.
<euouae>Here's two more small questions: 1) `man guix archive` says "... manual page for guix build". Should that be archive instead of build? 2) info page of guix says "This document describes GNU Guix version 0.0-git ...". Why does it say 0.0-git?
<euouae>I've noticed this version thing elsewhere as well
<nckx>Ah, I assumed that export-manifest would print that comment, in my example, I guess I was wrong.
<nckx>Yes, I *think* guix package --e-m might be more applicable here, although you'll always lose the info of exactly which package came from which commit.
<jab>civodul and nckx, I will email the guy from guix-hosting.com and let him know that you are open to a blog post/ something on the official guix site.
<nckx>lechner: I did write a linux-libre-bcachefs package that runs the deblob scripts on that, which seems to be the most elegant approach here. That was pre-hiatus, so I'll have to dig it up from I don't even know where, and I keep thinking it's not worth it because it'll be merged real soon now™.