IRC channel logs

<untrusem>btw how does one package rust apps who are not available on crates.io

<apteryx>would the `--as-needed` linker script option causes validate-runpath to report missing entries in the RUNPATH in case a library is not actually needed? c.f.: https://gitlab.gnome.org/World/planner/-/issues/274#note_2580869

<attila_lendvai>is the libxml2 issue resolved? is it safe to pull to the head?

<Rutherther>Yes libxml2 has been resolved like a week or two ago

<attila_lendvai>thank you Rutherther!

<efraim>finally opened an issue about my pine64 issue

<kestrelwx>e

<apteryx>any ideas how we could load the definitions from ld-wrapper.in in a unit test validate them?

<apteryx>maybe primitive-load?

<attila_lendvai>qtdeclarative-6.9.2-debug 867MiB... i wonder whether it can be avoided and it's a mistake on my side

<peanuts>"23.0.60; isearch string not quoted in M-s h r" https://issues.guix.gnu.org/867

<kestrelwx>Oh, the Cuirass bot is a first time contributor, that's exciting.

<untrusem>I mean until its contributes any code, it will always be the first time contributor :P

<Badeaffe42069>Hey, I try to use the autossh system service. But it starts before internet is reachable and does not (or not long enough) try to restart on failure before giving up. How can I make the service wait for connectivity?

<Badeaffe42069>fixed it, it is gatetime. but gatetime was misexplained in the commit, the manpage was clearer

<apteryx>attila_lendvai_: it's because of grafts, and a recent revert of old code that was avoid some cases like this, at the cost of having the very same libraries with multiple hashes in the store breaking glib and other things.

<apteryx>Dariqq summed it up better; "Always use all outputs s.t. there is always a single unique grafted variant of an ungrafted package."

<apteryx>Some good news, among the obvious downsides (large debug outputs being downloaded for grafts) https://codeberg.org/guix/guix/issues/347#issuecomment-5852905

<apteryx>I was trying to write a test for ld-wrapper, but it seems "load" does not make available its private definitions, e.g. `library-files-linked'

<apteryx>start of a test: https://paste.guixotic.coop/ld-wrapper.scm.html

<apteryx>I'm suspecting the peusdo-module definition in ld-wrapper

<civodul>“lugaru” :-)

<civodul>private definitions may be inlined, that’s the reason

<civodul>that said, ld-wrapper is sufficiently excercised that we can live another decade without tests :-)

<mroh>^^

<apteryx>civodul: actually, I think it's becausue it puts the definitions in the module namespace; I can access the module with (resolve-module '(guix build-support ld-wrapper)))

<civodul>ok

<apteryx>Since the module is not built, is there a way to ensure the optimization for load won't inline anything from it?

<apteryx>the reason I'd like to test it is to implement the missing edge cases that it doesn't support yet

<civodul>missing edge cases? i’m curious :-)

<civodul>i thought response files were the very last missing edge case :-)

<apteryx>I think there was a report on the debbugs about lld using some different link directives or something

<apteryx>also this kind of thing with plugins: https://gitlab.gnome.org/World/planner/-/issues/274#note_2580869; this could be avoided if the ld-wrapper script could produce an rpath for the project-local plugins

<apteryx>from: -Wl,--start-group libplanner/libplanner-1.so.1.0.0 /gnu/store/m7anhr7191smrvsysyzk3jyjpbhdfxfh-glib-2.84.3/lib/libglib-2.0.so -Wl,--export-dynamic /gnu/store/m7anhr7191smrvsysyzk3jyjpbhdfxfh-glib-2.84.3/lib/libgmodule-2.0.so -pthread /gnu/store/m7anhr7191smrvsysyzk3jyjpbhdfxfh-glib-2.84.3/lib/libgobject-2.0.so /gnu/store/szn3gpkz53w75zy1x7ihm6qin0m30699-libxml2-2.14.6/lib/libxml2.so -lm -Wl,--end-group

<apteryx>I guess if it links a relative library like: libplanner/libplanner-1.so.1.0.0, it could infer that this is relative to the project and add an -rpath directive for $ORIGIN or sometIng

<apteryx>then all these pesky loadable plugins would have the RUNPATH check verified... even though it makes little sense it practice.

<apteryx>(because they are loaded from the main application which already has all the needed symbols defined)

<apteryx>can't easily find the old ld-wrapper report, but I see some about unicode breakage, e.g. bug#65723 or bug#59031

<peanuts>"GCC/Clang toolchain package ld wrapper on utf8 paths with LANG enviroment variable not utf8" https://issues.guix.gnu.org/65723

<peanuts>"ld-wrapper replaces unicode character with question marks" https://issues.guix.gnu.org/59031

<jlicht>hey guix!

<apteryx>jlicht: o/

<apteryx>civodul: some interesting ideas for handling the plugins situation better: https://gitlab.gnome.org/World/planner/-/issues/274#note_2581256

<civodul>apteryx: i see, thanks!

<civodul>not sure we need to do something in ld-wrapper but this needs more thought

<civodul>for the many packages with plugins that we have, we ensure RUNPATH is properly set, to placate ‘validate-runpath’, even if that’s not strictly necessary

<civodul>that’s been a reasonable tradeoff in terms of maintenance burden IMO

<civodul>ACTION has to go

<civodul>ttyl!

<mixnmatch>Hello, I'm trying to package a kernel module for a usb sound driver (definition here: https://paste.debian.net/1401820/ ). It builds and installs, but the device does not seem to work when plugged in nor is the module listed in /proc/modules. Am I missing a step?

<Rutherther>mixnmatch: so how did you install it?

<mixnmatch>guix package -f Software/snd-usb-zoom.scm

<Rutherther>that installs to ~/.guix-profile, there is no way kernel would be loading modules out of there

<Rutherther>for kernel modules, use the linux-builder-service-type / more friendly linux-loadable-module-service-type

<Rutherther>inside the services of your operating system definition

<mixnmatch>Very helpful, thank you

<PotentialUser-24>Good day. I am trying to change gdm to sddm. But when I am trying to reconfigure I get error xorg unbound variable.

<Rutherther>PotentialUser-24: So how does your config look like?

<dlmayhem>&quit

<PotentialUser-24>How to convinient to share it? Because may pe it is connected I updated to kde 6.4 now I can not move windows at all they are stucked. May be kde ran in wayland, but gdm now does not allow to choose kde xorg or wayland option.

<Rutherther>I don't know, maybe something like 0x0.st that works out of the shell without any windows

<PotentialUser-41>Ok, so I loaded more earlier configuration. About Installing sddm here is my blocks of code that I changed:

<PotentialUser-41>(use-service-modules desktop sddm networking xorg)

<PotentialUser-41> (services (append

<PotentialUser-41> (list (service plasma-desktop-service-type)

<PotentialUser-41> (service sddm-service-type

<PotentialUser-41> (sddm-configuration

<PotentialUser-41> (xorg-configuration (xorg configuration

<identity>do not paste multiple lines in the chat, use a pastebin

<PotentialUser-41>Sorry, will do.

<PotentialUser-41>Ok so here it is https://pastebin.com/gQ4AeqnE

<Rutherther>you have a typo, instead of "xorg-configuration" you have "xorg configuration"

<Rutherther>and the same with keyboard layout

<PotentialUser-41>Now I have: error: 'services' field must contain a list of services

<Rutherther>that is because you have put modify-services inside of the "(list ...)", but it shouldn't be in it...

<Rutherther>modify-services returns a list, so now you have a list of services inside of a list

<PotentialUser-41>So where should I put it? Inside append or inside services?

<PotentialUser-41>I found searching for an answer and they use (cons * (service... What is the difference?

<ieure>PotentialUser-41, append appends two lists; cons* prepends elements to another list. (append (list 'a 'b) (list 'c 'd)) is equivalent to (cons* 'a 'b (list 'c 'd))

<ieure>It's a personal preference. I like the latter, for brevity.

<ieure>PotentialUser-41, Least error-prone is to do: (services (modify-services (append (list (service ...)) %desktop-services)))

<ieure>PotentialUser-41, But you could also do (services (append (list (service...)) (modify-services %desktop-services (delete gdm-servie-type))))

<PotentialUser-41>Thank you for the answers looks like config.scm is working now.

<Rutherther>PotentialUser-41: services is a field, a field accepts a single value, so _inside services_ cannot work, ie. (services (list ...) (modify-services ...)) would lead to a syntax violation

<PotentialUser-23>How to fix this error.

<PotentialUser-23>I tried lots of things, guix shell in containers, removing all gcc, and using clang, etc...

<PotentialUser-23>and nonethelss still the same issue compiliing C++ file

<PotentialUser-23>that size_t is not defined...

<PotentialUser-23>"In file included from /gnu/store/jb4szkjkmlqdc92nnhxvm9ypq6hvk9vw-gcc-14.3.0/include/c++/memory:80:

<PotentialUser-23>In file included from /gnu/store/jb4szkjkmlqdc92nnhxvm9ypq6hvk9vw-gcc-14.3.0/include/c++/bits/shared_ptr.h:52:

<PotentialUser-23>In file included from /gnu/store/jb4szkjkmlqdc92nnhxvm9ypq6hvk9vw-gcc-14.3.0/include/c++/iosfwd:42:

<PotentialUser-23>got no clue how it does to find gcc, when my manifest doens't specifiy any gcc ...

<PotentialUser-23>#!/usr/bin/env bash

<PotentialUser-23>set -ex

<PotentialUser-23>exec guix shell --verbosity=3 --container --network --preserve='^DISPLAY$' \

<PotentialUser-23>--preserve='^XAUTHORITY$' --expose=$XAUTHORITY --preserve='^DBUS_' --expose=/var/run/dbus \

<PotentialUser-23>--expose=/dev/dri --share=/dev/snd/seq --share=/dev/shm --expose=/sys/class/input \

<PotentialUser-23>--expose=/sys/devices --expose=/sys/dev --expose=/sys/bus/pci --expose=/run/user/$UID \

<mra>PotentialUser-23: use a pastebin, please

<mra>,paste

<mra>oops, wrong channel for that command, sorry

IRC channel logs

2025-10-20.log