IRC channel logs

2024-06-02.log

back to list of logs

<fnat>Ok, system restarted and now it's ok. I think I'd messed things up a bit with my previous 'guix deploy'.
<freakingpenguin>Does it mean anything if an unclosed patch disappears from QA? #70542 was there before and now it's not.
<peanuts>"[PATCH 0/4] Improve Shepherd service support for networked file systems" https://issues.guix.gnu.org/70542
<mirai>freakingpenguin: no and it shouldn't matter much since this is not a package related change
<mirai>btw, when you propose calling that field 'shepherd-requirement', wdyt on pluralising it instead? i.e. shepherd-requirements
<freakingpenguin>mirai: I agree, plural is preferred. I resubmitted a v3 that uses the pluralization just a few minutes ago, I think me saying the singular back thenwas a typo.
<mirai>I'm aware that there are some services using one or the other (which is partly my fault 🙂) and have been thinking on rectifying that
<freakingpenguin>Either a typo or it was singular because shepherd services use the singular even though it's a list.
<mirai>freakingpenguin: does the 2/3 patch handle link-local addresses?
<freakingpenguin>169.254 addresses? I don't see why not. host-to-ip seems to be a fancy noop if "host" is just another ip.
<mirai>I'm asking this because the address I use looks like this
<mirai>(define nas-address "[fe80::265e:beff:fe7d:6064%enp5s0]")
<freakingpenguin>I haven't tested it with ipv6, but getaddrinfo in (info "(guile) Network Databases") seems to say it can handle them.
<freakingpenguin>It might not like the [] or %enp5s0
<mirai>hmmm… the fstab or nfs manual mentioned that the address has to be written like this iirc
<mirai>my NFS mount looks like this: https://paste.debian.net/1318854/
<peanuts>"debian Pastezone" https://paste.debian.net/1318854
<freakingpenguin>mirai: Part of the challenge may come from the fact that mount.nfs may do its own preprocessing of mount options and on the contents of fstab. However, (gnu build file-systems) invokes the mount syscall directly.
<mirai>iirc a fstab entry is generated from this file-system record
<freakingpenguin>That's why mount-cifs captures the guest option and converts it to user=,guest=
<freakingpenguin>I'm guessing here since I haven't looked into the nfs code
<freakingpenguin>Personally I'd prefer if we use the userspace tooling instead of rolling our own. The only reason I can think of for not doing that is mounting file-systems in the initrd, but how can we do that for nfs or cifs if those are networked.
<RoundDuckKira> https://i.imgur.com/xVKkSFI.jpeg
<RoundDuckKira>when I use guix pull and then system reconfigure on root, it complains that guix is newer than the needed version of guix
<RoundDuckKira>this is after following instructions after fresh install to update system
<RoundDuckKira>is it safe to do "--allow-downgrades"?
<RoundDuckKira>now it looks like it's basically reinstalling the system
<RoundDuckKira> https://i.imgur.com/uKG4bVx.jpeg
<RoundDuckKira>I mean this looks familiar to me
<RoundDuckKira>why is guix downloaded an rc2 version of 1.4.0 oof
<RoundDuckKira>also is there a right way to do encrypted boot install on libreboot
<RoundDuckKira>well let me say it better, is there a way to configure libreboot to recognize guix and properly boot into that from grub instead of seabios or waitinng a century for grub to detect guix every boot
<civodul>ACTION is annoyed by the conflict of interest with rde
<Lumine>Tell us more, I'm curious.
<jakef>i second that
<thanosapollo>+1
<civodul>heh :-)
<civodul>changes made to Guix Home services just to avoid incompatibilities with rde
<civodul> https://issues.guix.gnu.org/71111
<civodul>it feels like a one-way street
<peanuts>"[PATCH 0/1] services: home: Use pairs instead of lists." https://issues.guix.gnu.org/71111
<civodul>ACTION reconfigures berlin
<civodul>uh, knot fails to build
<efraim>--enable-xdp=no?
<Franciman>hi, is anybody using zathura on guix?
<Franciman>in 0.5.4 there is a bug that makes the autocompleter show everything but pdf files in the autocomplete menu :P
<Franciman>is anybody getting bitten by this bug?
<Franciman>anybody else, since i am :P
<oriansj>Franciman: sounds like a bug with the zathura autocomplete script; look in ~/.guix-profile/etc/bash_completion.d/ for the script
<oriansj>as that can be used to confirm the bug and report it properly
<Franciman>oriansj: uhm no, there was, wait. It should be this problem: https://github.com/pwmt/zathura/issues/439
<peanuts>"The autocomplete for `open` excludes nothing but pdf's ? Issue #439 ? pwmt/zathura ? GitHub" https://github.com/pwmt/zathura/issues/439
<Franciman> https://github.com/pwmt/zathura/commit/d74091a7452846112c0bb11e1412cd2af48ee7cd it was fixed here
<peanuts>"The function file_valid_extension should return True if plugin is dif? ? pwmt/zathura@d74091a ? GitHub" https://github.com/pwmt/zathura/commit/d74091a7452846112c0bb11e1412cd2af48ee7cd
<civodul>efraim: i thought about --enable-xdp=no, but since it was added on purpose, i thought i’d rather bisect to find when it broke…
<civodul>according to https://ci.guix.gnu.org/build/4774395/details it’s probably recent breakage
<peanuts>"Build 4774395" https://ci.guix.gnu.org/build/4774395/details
<efraim>the asahi drivers for aarch64 on core-updates pull in libclc and with it a newer llvm and so on ...
<civodul>bisecting with ‘guix time-machine’ is not… fast
<efraim>I think that's one where I might actually use it on berlin since there's so many machines to offload to
<civodul>not sure it would help since the derivations for Guix itself are mostly built sequentially
<civodul>(they depend on one another)
<Franciman>what is the correct way to refer to Guix system?
<Franciman>GNU Guix System is ok?
<Franciman>or is Guix System enough?
<civodul>Franciman: “Guix System” is fine; “GNU Guix System” works too
<Franciman>ok thanks. We used it for our tests in semantic web module extractors!
<civodul>oh, sounds fun!
<thanosapollo>is "The GNU System" fine as well? I've been getting that message on my tty login prompt & been referring to guix this way
<Guest48>I have a quick question. I've got something installed that has an incorrect wm class set in its desktop file, and it's not playing well with gnome (two separate icons while running...) The .desktop files in the store are read-only and I'd normally just edit those. Is there a simple fix I'm not aware of?
<Guest48>Or, I guess, what is the guix-y way of handling it, if there is one?
<Guest48>Well, I'm a dope and forgot I could just copy the .desktop file to ~/.local/share/applications and modify it there, which has successfully fixed the duplicate icons. Brain fart!
<bjc>if upstream refuses to fix it, yuo can submit a patch to guix with the fix and why it's necessary. that'd be the best option (other than upstream fixing it)
<bjc>otherwise, you can patch your local guix w/ the fix, make a channel for yourself with the fix, and, probably the easiest: just put your own .desktop file in ~/.local/share/applications
<bjc>and i see you've come to that conclusion in the time it took me to type that =)
<Guest48>I appreciate your response anyway! I took the easy way out, but a patch might be worth looking into as well.
<amirny2205>my emacs config https://toot.io/@amirny2205/112480119143331823
<peanuts>"Mirniy Sasha: "my .emacs \w EXWM
<PuercoPop>how do I access name of a service? Should I access the slot directly? I'm trying to return a list of service names in %base-services. My first thought was to check if service-name was the name of the accessor, ie. (service-name (car %base-services))
<freakingpenguin>PuercoPop: I believe the accessor is service-kind
<freakingpenguin>Just know that returns a service-type record, not a symbol IIRC.
<ieure>Anyone have a link to an EXWM configuration with working simulation-keys? I copied all my EXWM stuff from my Debian box into ~/.exwm, including (setq exwm-input-simulation-keys ...), but none of them work and it's a real drag.
<PuercoPop>freakingpenguin: that returns the type slot, not the same. ej. (service-kind (car %base-services)) => $12 = #<service-type login 7fb2e902c540>.
<PuercoPop>But the service name is motd #<<service> type: #<service-type login 7fb2e902c540> value: #<<login-configuration> motd: #<<plain-file> name: "motd" content: "This is the GNU operating system, welcome!\n\n" references: ()> allow-empty-passwords?: #t>>
<freakingpenguin>PuercoPop: The service record doesn't have a name, just a type and a value. The type does have a name, so you'll have to first get the type, then get the name
<freakingpenguin>(service-type-name (service-kind foo))
<freakingpenguin>That's mostly meant for debugging, what are you trying to do?
<PuercoPop>freakingpenguin: half learning how to work with guile (I mostly have Common Lisp experience) from emacs, half exploring the default system config
<freakingpenguin>PuercoPop: Gotcha. I found this post helpful for understanding records fwiw: https://lepiller.eu/en/a-deep-dive-into-guix-records.html
<peanuts>"A deep dive into Guix records ??? Julien Lepiller" https://lepiller.eu/en/a-deep-dive-into-guix-records.html
<taeltydes>I don't suppose anyone has a collection of manifests for development purposes? I found something like this a while back but can't find the clone.
<makx>mhm, when I try downloading the latest x86_64 iso image I get the error
<makx>error "Could not find the requested build product."
<freakingpenguin>makx: I reported that here https://issues.guix.gnu.org/71312. It looks like CI does build it but you can't download it from the webui juding by how the "Build outputs" link behaves on this page. https://ci.guix.gnu.org/build/4746037/details
<peanuts>"Dead download links on guix.gnu.org/en/download/latest" https://issues.guix.gnu.org/71312
<peanuts>"Build 4746037" https://ci.guix.gnu.org/build/4746037/details
<cancername>It seems like i have to "guix pull" for each user individually. Can I somehow "guix pull" for all users?
<taeaad>What does it mean when something is version 0.0.0, like llama-cpp?
<cancername>🤨 it doesn't seem like they follow semver: https://github.com/ggerganov/llama.cpp/releases/tag/b3067
<peanuts>"Release b3067 ? ggerganov/llama.cpp ? GitHub" https://github.com/ggerganov/llama.cpp/releases/tag/b3067
<civodul>i’m seeing “error: %default-extra-linux-options: unbound variable”
<civodul>but it’s not actually an error
<civodul>oh, this is from hydra/*.scm in maintenance.git
<civodul>apparently those relied on this private variable, which no longer exists
<civodul>ACTION reconfigures berlin (2nd attempt)
<freakingpenguin>I'll cross my fingers for you
<makx>ACTION is making a next attempt at moving to guix from nixos; i like guix much better, not the least because, man scheme is so much nicer than the nix config language
<makx>(and I speak scheme much better than nix)
<jfred>Sigh... working on packaging asterisk, and of course its `make install` target wants to go fetch stuff from the internet. Of *course* XD
<freakingpenguin>I can see why a pre-build target might try and fetch stuff from the internet, but who thought it was a good idea to require an internet connection to install the package haha
<PuercoPop>freakingpenguin: thanks for the link. I was wrongly assuming the services where CLOS-like as in goops instead of records which are more similar to structs
<civodul>berlin reconfigured \o/
<civodul>it’s busy: https://ci.guix.gnu.org/workers
<peanuts>"Workers status" https://ci.guix.gnu.org/workers