<the_tubular>Well my first comment with podman is that it does not create a policy file by default.
<the_tubular>Otherwise it seems to work, but I've done very little testing
<podiki[m]>question about unbundling (program has dependency source): should I replace the source files with a guix package of the source? or try to remove the source and have it build with the guix package of the (compiled) library?
<the_tubular>Umm, trying to run an image from docker.io I get : Error: writing blob: adding layer with blob "sha256:a2abf6c4d29d43a4bf9fbb769f524d0fb36a2edab49819c1bf3e76f409f953ea": ApplyLayer exit status 1 stdout: stderr: potentially insufficient UIDs or GIDs available in user namespace (requested 0:42 for /etc/gshadow): Check /etc/subuid and /etc/subgid:
<Kolev>singpolyma: Yes, but it's more overhead than just using Stow, IMHO.
<Kolev>singpolyma: Feel free to send a patch to put all the Stow files into the home config.
<arjan>managed to create a issue with the cover letter, but now my email provider refuses to deliver the rest of the patches to debbugs.gnu.org because that does not support TLS, will try again later with TLS requirement disabled
<the_tubular>I got podman as root working :) . But I think the root-less could use some work
<podiki[m]>but maybe askthe list about removing all root's guix stuff; you won't be the last to do that I'm sure
<Kolev>I have IceCat in the system decl. but not every user has IceCat. 😕️
<akirakyle>I somehow am not encountering that issue anymore after doing a guix pull as root, then doing sudo guix reconfigure, then deleting old generations, then deleting the gc root for root at /var/guix/profiles/root (and the symlink at .config/guix/current). I can now sudo guix reconfigure just fine and no longer have any guix as root! Don't know why this
<akirakyle>particular sequence of operations fixed that issue, but it did
<arjan>managed to submit all remaining patches to debbugs
***califax- is now known as califax
<akirakyle>Anyone here use sway? I'm having trouble starting sway: I get the error 00:00:00.017 [ERROR] [wlr] [libseat] [libseat/backend/logind.c:267] Could not activate session: Interactive authentication required.
<lfam>I didn't go through the CI interface, but simply did `guix build $hash-guix-$version.drv` on the appropriate system
<lfam>You can run only this set of tests with `make check TESTS=tests/guix-system.sh`
<lfam>That build you linked to is pretty old and I doubt it will actually be run at this point
<lfam>It's revision 13 of the Guix 1.3.0 package, and we are currently on revision 17
<akirakyle>Isn't it necessary to be run because of the boot-strippping process?
<akirakyle>Is there some mechanism that the CI uses to ignore outdated but pending builds?
<lfam>For your first question, that package would be necessary for users who haven't updated.
<lfam>To your second question, all builds are supposed to be carried out. The aarch64 build farm support is still nascent, so we will have to cancel old builds that users are probably not actually going to use the results of anymore
<akirakyle>Well I'm hoping to make a full switch to guix from nix for my dev environment and if my emacs habits are any indication, I have a lot more fun digging into and debugging lisp than the nix dsl!
<akirakyle>On that note, you wouldn't happen to use sway?
<akirakyle>I'm trying to be a wayland citizen, but life can be hard on the bleeding edge :)
<akirakyle>I think this update (https://issues.guix.gnu.org/51970) that happened between 1.3.0 and now between may have broken something since now sway won't start for me on guix and I get this error: [ERROR] [wlr] [libseat] [libseat/backend/logind.c:267] Could not activate session: Interactive authentication required.
<akirakyle>(among other errors but I think they're all due to this one)
<akirakyle>I really can't wait for extra aarch64 build capacity. I tried to install xdot and guix started trying to build the whole rust bootstrap chain which I imagine would take quite a long time to compile
<lfam>Is use of GDM hardcoded in the gnome-desktop-service?
<lfam>That would explain it, since gnome-desktop-service-type is used in desktop.tmpl
<akirakyle>That makes sense to me, although I'm still getting used to all this in guix so not the expert here
<apteryx>I think civodul had replaced gdm by lightdm for non-x86_64 systems; but haven't checked the details
<lfam>I think he used SDDM in commit 49599fab564f203b8e92d32e9b28c99e99849bfb
<lfam>akirakyle and I are trying to figure out #52908 / #52943, where the Guix test 'tests/guix-system.sh' fails on aarch64 because xorg-server is provided more than once in the service graph of desktop.tmpl
<lfam>This breaks the build of the guix package on that platform
<lfam>Somehow it looks like GDM is still being used in that case
<lfam>I always use `guix environment / guix shell` to make it available for this importer
<akirakyle>lfam I think I maybe figured it out "set-xorg-configuration" is used explicitly in desktop.tmpl and it includes gdm-service-type in desktop.tmpl independent of the login manager set in %desktop-services
<lfam>Should it set that option in all cases, or only when not on x86_64?
<akirakyle>I'm new to guile too, so can the #:optional support a conditional, since that might be the cleaner way of doing this so the desktop.tmpl isn't touched with a conditional
<akirakyle>Otherwise I don't see a way of avoiding a conditional on not x86_64 in the desktop.tmpl file, which is maybe not desired since it gets included in the docs
<apteryx>akirakyle: #:optional can take any valid expression, unless I'm missing something
<lfam>I think we should feel free to adjust desktop.tmpl
<lfam>If I understand correctly, this file is only used by the set of tests under discussion, so it shouldn't be too hard to find regressions
<lfam>And I think it's the right place to put such a conditional, rather than in the set-xorg-configuration procedure. Like, these kinds of workarounds for growing pains of the distro should be in things like examples, templates, and not in service definitions themselves
<akirakyle>Does it make sense though for gdm-service-type to be the default argument for set-xorg-configuration?
<akirakyle>I feel like this may be a trap someone can easily fall into themselves when trying to write something like desktop.tmpl and customizing %desktop-services
<lfam>Whether you trust a PGP signature is totally up to you; it's a social issue. All the software does is create a mechanism to prove that a signature was made by a person that had access to the private key
<DiffieHellman>Sure, but if a person uses gmail and adds it to their signature, I feel that they're incompetent.
<lfam>Of all the commits made to Guix, the top email server used by committers is gmail.com (18 committers) followed by gnu.org (12 committers) and then a few dozen personal domains, many of which use gmail
<akirakyle>lfam: unfortunately I jumped the gun a bit on blaming set-xorg-configuration, I still get the build failure when the services for desktop.tmpl is just set to %desktop-services, so there's at least more going on here
<akirakyle>I'm also hitting my debugging limit for the night, so I think I'll pick this up tomorrow. I feel like I'm understanding things better though so hopefully with some more debugging I can send it a patch that fixes this
<jpoiret>i don't think there's an easy and general way to do that, packages using the gnu-build-system could very well be using a compiler that's not gcc/clang
<mekeor[m]>jpoiret: a different topic: you have been writing about polkit, dbus and network-manager on the help-guix mailing list. your e-mails went in depth but i still don't have a clue what you are actually suggesting we can practically do about the issue that we can't get non-root users on guix system to run network-manager tools using user groups.
<jpoiret>mekeor[m]: well, the forwarded mail didn't include what you were actually trying to do, the replies only go so far
<mekeor[m]>jpoiret: i did not start the mail-thread. i didn't respond to it either. but i face the same or a similar issue: e.g. i can't connect to a wifi per `nmcli device wifi connect SSID password SECRET`. it results in `Error: insufficient priviledges`.
<mekeor[m]>jpoiret: yes. (i start xorg as a user and run xmonad as window manager.)
<jpoiret>I suggest running /libexec/polkit-gnome-authentication-agent-1 from polkit-gnome alongside it then
<jpoiret>(it shouldn't pull in anything gnome related)
<else->in the `channels.scm` there is the possibility to reference a specific commit – i would like to use a tag instead but it does not seem to be possible. any other way i can "set" a tag as version for a channel?
<mekeor[m]>jpoiret: is that included in the polkit-gnome package?
<else->could `guix channel update` be a possible remediation?
<mekeor[m]>jpoiret: it works. thank you. would you like to add this to the mentioned e-mail-thread or should i do it?
<mekeor[m]>jpoiret: also, i don't know what the libexec-directory is for. but i think it'd make sense for this binary to be callable from the shell. so i think it needs to and should be put into the bin-directory if possible?
<jpoiret>well, libexec is kind of a weird convention, I agree, but that's how upstream is packaged. I don't think the whole GNOME stack would behave well if we rename it
<mekeor[m]>else-: a remediation for solving your problem, i.e. for using tags?
<jpoiret>well, you can always add that to the ML, but the thing is, we don't know what they're trying to achieve. I personally thought they were working on a script that uses nmtui somehow, but it might just be them trying to use the program as is.
<dcunit3d>tissevert: you may want to put some of the language-specific stuff in a profile/environment. it depends. after my /etc/profile or gnome-shell have loaded, i don't actually have a python in my PATH
<gnoo>excalamus: love that you have a links page :)
<podiki[m]>I think more tutorials and examples in the official cookbook would be great
<excalamus>last year I was seeing if it's possible to make a step by step guide that uses guix-time machine
<dcunit3d>tissevert: in the docs, the way GUIX_PROFILE is set before sourcing the $GUIX_PROFILE/etc/profile script doesn't export the variable. that might be related to what you're seeing (it tripped me up quite a bit at first)
<dcunit3d>hmmmm, i can't remember what sources that. it gets loaded very early on. it might be the earliest shell that the system loads. that info is in the bash man pages where it talks about the order of files loading.
<tissevert>so it doesn't depend on my .bash_profile ? that'd be contrary to bash's behaviour
<tissevert>"from /etc/profile and ~/.profile, in that order"
<tissevert>I'm just mistaken about bash's specifications
<dcunit3d>you can test how guix system is modifying those variables at login with the VTTY's. login to one, insert some echo's, switch to another VTTY and login
<tissevert>it works because that's the way it's supposed to work
<tissevert>my profile shouldn't source the system's, the systems sources mine
<tex_milan>Hi people. Anyone using cups to print? How did you solve the "unable to locate printer" problem? Cups can find the printer when adding it into configuration, but then when I try to print it tells me this error. Reminds me the same problem I had on NixOS. Some timeout issues... Might be the same problem here?
<rekado_>tex_milan: I’m using Guix System with CUPS and a Brother laser printer
<mekeor[m]>i'm using guix system with cups and a kyocera printer. in my system-declaration i enabled cups' web-interface and added my user to the "lp" and "lpadmin" groups. i then used localhost:631 to add the printer. it works like a charm
<lfam>jacereda: It's likely being held for moderation if it's your first message
<lfam>It was considered to be a necessary violation of the functional model in order to help people use shell scripts
***phf-1[afk] is now known as phf-1
<karrq>hmm can anybody help me? I'm trying to use `guix refresh` to update some dependencies from upstream but when I run I get "mkstemp: Permission denied" and `strace` reveals that the denied permission is returned when trying to open the package definition file in the store with RDRW... I'm in foreign distro as user
<awb99_>my init.el files are not being executed by emacs
<phf-1>lfam: I've just rebooted and vlc works now !
***awb99__ is now known as awb99_
***iyzsong- is now known as iyzsong
***sneek_ is now known as sneek
<attila_lendvai>i think using (git-checkout (url (dirname (current-filename)))) as the source of a package used to work prior to the c-u-f merge, but it now fails with: "guix build: error: Git failure while fetching /home/alendvai/workspace/idris/idris2: reference 'refs/remotes/origin/main' not found" -- it shouldn't have any expectations wrt origin and main (the dir holds the checkout of a custom upstream, and a custom branch)
***ns127 is now known as ns12
***marwan8 is now known as marwan
***darosior1 is now known as darosior
***brettgilio3 is now known as brettgilio
<attila_lendvai>and the alternative i tried, git-reference to a local path, also doesn't work
<attila_lendvai>err, no, the error is: "fatal: '/home/alendvai/workspace/idris/idris2/' does not appear to be a git repository"
***Inline is now known as Guest1020
***Guest1020 is now known as Inline
***Noisytoot_ is now known as Noisytoot
<jacereda>I've added the following service to fix brightness on my macbook pro: `(simple-service 'fix-brightness boot-service-type #~(execl #$(file-append pciutils "/sbin/setpci") "-v" "-H1" "-s" "00:01.00" "BRIDGE_CONTROL=0"))`. Guix won't boot now, it complains IIRC with a "setpci not tainted" message... Any idea what that means?
<jpoiret>jacereda: if i'm not mistaken, you'd rather have to add a service that extends boot-service-type, not a new boot-service-type
<dcunit3d>why doesn't guix system ship with GCC? if I add gcc-toolchain, should I do so within a profile that is loaded only when i need GCC?
<jpoiret>yes. Many other *nix distributions don't ship with gcc by default as well
<jpoiret>you can also use `guix shell gcc-toolchain`
<jpoiret>to launch a shell with a profile containing gcc-toolchain loaded
<dcunit3d>cool, would you recommend isolating that into its own profile?
<jacereda>jpoiret: oh, so I'm overriding the boot-service-type... so I should use service-type instead of simple-service-type, right?
<excalamus>when doing a system reconfigure, I've been seeing a warning about "List elements of the field 'swap-devices' should now use the <swap-space> record". When I look at the reference that the warning gives, it only seems to explain what a swap space is. I already know that. What I don't know is how to update my config to the new syntax.
<excalamus>what I have is (swap-devices (list (uuid "29be12c9-6439-4987-9e46-02ac200761f2")))
<jacereda>excalamus: I think it's (swap-devices (list (swap-space (target (uuid ...
<Kolev>vldn[m]: error in initialization thread: Success!
<jacereda>jpoiret: this is the first thing it executes: `(begin (execl "/gnu/store/1nwlhs58rr7pxpbyi8h9apgb3r0q4c5a-pciutils-3.7.0/sbin/setpci" "-v" "-H1" "-s" "00:01.00" "BRIDGE_CONTROL=0") (begin (use-modules (guix build utils))...`, which looks right to me, but maybe it's too early in the boot sequence?
<jpoiret>boot is executed after the system has pivoted its root filesystem, so things in the store should be available without any issue
<Kolev>Perhaps I should break down and use `code` for the Matrix users.
<jonsger>is there a particular reason why Guix doesn't use something like e.g. 2015-2020 instead of 2015, 2016... in the copyright headers?
<singpolyma>jonsger: what if you need to skip a year where you don't want it to be copyrighted? :troll:
<excalamus>okay, I think I see how I can make sense of the warning hint. I think the thing tripping me up was that I didn't equate records with objects nor understand that they were basically structs.
***robin_ is now known as robin
***phf-1 is now known as phf-1[afk]
<attila_lendvai>argh! the nix sandbox doesn't have /usr/bin/env... and so did i break the idris build on nix. luckily, there's a CI.
<attila_lendvai>actually, it also breaks on guix. there doesn't seem to be a /usr/bin/env in the builder's sandbox
<jacereda`>In case anyone wants to execute some `setpci` at boot, `(simple-service 'fix-brightness activation-service-type #~(execl #$(file-append pciutils "/sbin/setpci"...` seems to work, I don't know why it failed with `boot-service-type` :(
<robin>attila_lendvai, the build process rewrites shebang lines but it may not be early enough in the build process if e.g. building with autotools
<attila_lendvai>robin, the output of idris (a compiler) is a wrapper shell script, and the tests are run using a binary produced by idris (called runtests), and this is that fails.
<robin>attila_lendvai, i use a phase, (add-before 'bootstrap 'patch-bootstrap-shebangs (lambda _ (patch-shebang FILE) ... #t), there's probably a better way
<attila_lendvai>robin, i'm afraid i'll need to do something like that between build and check. thanks!
<robin>oic, you'd want to generate a wrapper before the test phase then, i think
<jacereda`>Anyone using isync+notmuch? Do you use a mcron service for that?
<ngz->Here is a part of the Guix manual: "each extension designates [...] and a procedure that, given the parameters of the service, returns a list of objects to extend the service of that type." I cannot make sense out of it. Isn't there a verb missing somewhere?