IRC channel logs
2024-09-04.log
back to list of logs
<elevenkb>I want to package xenops-mode but it depends on AUCTeX and calls (require 'tex) which ends in an error. <elevenkb>Nope I found a fix, in the definition of emacs-company-auctex. <eikcaz>Any idea why ssh might kick me out after an update (like I can connect, but the connection immediately terminates)? <eikcaz>I forgot to pass --no-bootloader, and I'm worried I might have killed my swarm of devices across the country... <mange>Does "ssh -vvv $host" give you any useful output? <mange>(I never remember how many 'v's to do.) <eikcaz>It gives a lot of output. Looking now. Thanks a bunch <mange>Another thing to try could be "ssh $host -- bash -i", which might help if the user's shell is messed up, I think? <eikcaz>^ that fails silently (connection made and terminated with no prints) <mange>Mmm. Hopefully the verbose output gives you something to go on, then! <ieure>mange, I had a weird thing like this happen, I'd pulled and started a system reconfigure inside a detached screen session. The machine started failing to accept SSH connections, and I couldn't log into the console, either. Flashed a message about "Unknown module" and dumped me out. <ieure>I rebooted and I could get back in. Didn't boot into the previous generation. Super weird. <mange>I feel like I've had something similar in the distant past, but I can't remember any of the details. I hope eikcaz has some way to reboot their remote devices. 😟 <eikcaz>well, normally I would post my logs to my website to share them, but my coincidentally, my website is down... <eikcaz>I feel like I had this happen once, and that a reboot fixed it <eikcaz>I see "debug2: chan_shutdown_write: channel 0: (i3 o1 sock -1 wfd 5 efd 6 [write])" <eikcaz>I don't have i3 installed though <mange>That won't be the window manager. <eikcaz>The torch has been lit. Someone should be rebooting one soon. <mange>Yeah, I can't see anything in the log that looks obviously wrong. Hopefully the reboot sorts it out. <eikcaz>Maybe the connection is getting garbage collected: "debug2: channel 0: garbage collecting"? Like maybe a reference count got into an inconsistent state after the live update. <eikcaz>I had accounts connected when I did the update. If the update reset the count to 0, then disconnecting from the machines would put the count at -1. Then connecting again puts the count to 0, which triggers the garbage collector to clean up any remaining processes, i.e. the one I'm using. <mange>If that were the case, I would expect trying multiple connections at the same time to have a chance of succeeding. <mange>I'm not optimistic, though. I don't have a solid reason why, but it just doesn't feel right to me. <eikcaz>it seems that they still all get closed. Maybe each time the number goes down, it reduces the number of subprocesses to match <eikcaz>Is it worth mentioning this experience in issues.guix.gnu.org? <eikcaz>I actually have two (mostly) identical machines, one still in the borked state, and one in the now rebooted state, so I could concievably debug <mange>My only concern is that I doubt it's going to be actionable without more information. If you can find a way to reliably reproduce the issue, or can find out more about what went wrong, then creating an issue would be worthwhile. <mange>This has led to me noticing, for the first time, that Guix lacks an equivalent to "nixos-rebuild boot" - that is: leave the system as-is now, but arrange to boot into a new generation on next boot. Am I missing something in the manual/--help output? <eikcaz>hmm. That's a good point. You could build with --no-bootloader, then flash the bootloader with a manual grub invocation, but I don't think there is a flag for it <eikcaz>Or, I suppose just building without reconfiguring wouldn't require --no-bootloader anyway <aswjrisp>I searched for sway in info guix and did not find results for a sway service. <eikcaz>First thing I would do is see if sway still crashes with %desktop-services. I would be surprised if sway crashed when launching from gdm. <aswjrisp>It looks like it might be an issue with MESA-LOADER being unable to mesa/lib/dri. <eikcaz>seems like its unable to find some .so files, so maybe they are hiding in %desktop-services <aswjrisp>eikcaz: Okay I will try sway with %desktop-services. <aswjrisp>I think guix probably has a way to search for the file i915_dri.so to find out what package it is in. <aswjrisp>eikcaz: guix locate tells me that file is from mesa. So I will install mesa and see if sway starts with %base-services. <aswjrisp>Installing mesa did not allow sway to start. I get the same error about i915_dri.so. <aswjrisp>I am going to try to start sway with %desktop-services. <oriansj>aswjrisp: well I know sway starts just fine if executed directly (you just need greetd and seatd setup) but you can't use certain login managers (like the guix default slim) to start wayland desktops <aswjrisp>eikcaz: Thanks for pointing out guix locate. <aswjrisp>eikcaz: Using %desktop-services was not enough to start sway. I am getting the same error messages. <meaty>people who are using the MATE desktop, how were you able to get 'preferred applications' working? for me, it crashes immediately saying i.e. 'glib-gio-error...org.gnome.desktop.ally.applications' is not installed <aswjrisp>oriansj: Thanks for pointing out greetd, seatd and the issue with login managers. <meaty>is this a mistake in its packaging or does that org...applications string correspond to a package I should've installed <meaty>specifically the message is 'GLib-GIO-ERROR **: 22:11:15.391: Settings schema 'org.gnome.desktop.a11y.applications' is not installed' <meaty>there are *many* /gnu/store/*-glib-schemas directories <mange>The internet tells me that that settings schema is usually in gsettings-desktop-schemas, so you could try installing that to an appropriate profile? <meaty>mange: as in my user profile, right? <mange>Probably yes? If that doesn't work then maybe try the system profile as well? <meaty>aha! thank you. What internet resource did you use? <mange>I just searched for "settings schema org gnome desktop a11y applications" and saw people mentioning that in the Ubuntu instructions. Guix had a similarly named package, so I guessed that it might work. :) <aswjrisp>oriansj: I see that both seatd and greetd are in the guix manual. I will read about them in the manual. Thanks <PuercoPop>How do I list the derivations that is backing a service? installed by guix home? I can use guix home describe --list-intalled for packages, but not services <eikcaz>PurecoPop: it might be obvious from "guix home edit <home-service-type>" <PuercoPop>eikcaz: That seems to take to the service definition itself (as in where it is defined in guix). I'm trying to find out where an executable is, as it doesn't seem to be in bin <PuercoPop>I tried finding it using find as well, find ~/.guix-profile -name '*redshift' <mange>It's common for services to reference things directly, to avoid polluting system or user profiles. If you want to run it you can use "guix shell redshift" to enter a shell with it, or you can install it yourself into a profile. <PuercoPop>I want to understand why shepherd is failing to start the service <PuercoPop>Ok, looking through the ~/.config/shepherd/init.scm I found a reference to the derivation <PuercoPop>Ok, I found the issue, it is trying to call something call geoclue, which is not installed in my system. If a [default] configuration option requires a package/service and it doesn't pull it itself is it a bug? <mange>Just to be super pedantic on terms, a "derivation" is an instruction to the daemon on how to build something (documented in the manual: "(guix) Derivations"). Build results are the result of running a derivation to produce a result (in the store), and are sometimes called "outputs". <mange>I'm only mentioning this because your use of "derivation" confused me a bit. :) <PuercoPop>mange: Yes, I meant derivation. As in there in the /gnu/store is the stuff. It helps me debug issues <mange>The derivation file ends in ".drv", and is just build instructions to the build daemon. I expect you were actually looking for a build result in /gnu/store, which is the output of the derivation. <PuercoPop>Ah ok. I see what you mean. You are right. The drv is the set of instructions for the daemon to run, the output is the content-addressed folder in the /gnu/store? <mange>Also, the documentation for the location-provider field of the home-redshift-configuration object says "[if the value is geoclue2] the Geoclue system service must be running; it will be queried for location information." Unfortunately the default is geoclue2, so the dependency is not highlighted in the configuration itself. <mange>See "(guix) Desktop Home Services" in the manual. <PuercoPop>mange: yeah, I found that and now trying to move to manual. But my question is whether it would be considered an improvement to conditionally setup the geoclue service? <mange>I think it would be an improvement, but there are two problems with that at the moment: 1. Guix doesn't support conditionally setting up other services based on configuration. It's something I want to have, but it's non-trivial given the way the service graph is constructed. 2. Guix doesn't have a geoclue-home-service-type. <PuercoPop>mange: What is the difference between a home service and a regular service? Would it be hard to use the geoclue-service-type as a reference for the geoclue-home-service? <mange>There are some tools to turn a system service into a home service (see system->home-service-type, and an example use in gnu/home/services/syncthing.scm), but I'm not aware of their limitations. I'm also not sure what geoclue needs to run (e.g. whether it run as a regular user). <wizard>is there a way to return the amount of ram a system has? <wizard>i want to make a swap file that's 80% or whatever of my system's total ram capacity <wizard>i could manually type it in but that wouldn't be super portable across computers <mange>You could read it out of /proc/meminfo? The first line looks like "MemTotal: ??? kB" with the amount of memory the system has. <wizard>my first thought was smth like that but i'm always worried there's a more idiomatic way that i'm missing :p <mange>There might well be, but I don't know it. :) <omar_b>Can some one help me figure out why this patch is not passing the QA. I tried applied the patch locally on the latest master branch and it worked seamlessly. This is my first time submitting a patch to guix so I don't know how to know what's wrong with the patch it self or with the email I sent. <cbaines>omar_b, the key bit of information from QA is: error: corrupt patch at line 44 <omar_b>cbaines: Yes I've seen that but I don't understand what it means. <cbaines>omar_b, I'm guessing you didn't send the patch with "git send-email"? <omar_b>No, I didn't I just copied the patch content to my gmail and hit send. Do I have to use git send-email? I haven't got it set up before and it was getting me errors. <cbaines>omar_b, right. git send-email takes care of sending patches correctly. GMail is wrapping the content which is corrupting the patch. <omar_b>Yes but when I look at the attached file. It doesn't look like it's breaking the lines> <cbaines>it doesn't, although I'm confused by the  (capital A with hat) characters I see <cbaines>it's also confusing to include the patch twice in the same email <omar_b>cbaines: what do you mean by "include the patch twice"? I also see the patch included once. <cbaines>omar_b, I was assuming you'd attached it as well, but it just looks like you've sent a plain text version as well as a HTML version (which is what issues.guix.gnu.org is displaying as an attachment) <omar_b>I cbaines I finally was able to use git send-email but the issue is not updated yet. <omar_b>cbaines: I did recieve an email that my patch was submitted but the issue tracker doesn't seem to reflect that <cbaines>things can take time to update, don't expect it to happen instantly <omar_b>cbaines: OK. I just wanted to make sure that it works. <jpoiret>the fwupd service in guix upstream is very basic and doesn't work <jlicht>jpoiret: yeah I figured we needed one of two workarounds as specified in 60065. Might be a false premise, perhaps there are other options too :-) <jpoiret>I think there's also a one-time init step that I haven't gotten around to implementing, just did it manually <jpoiret>disclaimer: I haven't done any updates using it unfortunately, and I haven't tested whether the fwupd-efi stuff works at all <jpoiret>but the basic interface and update checking does work <jpoiret>jlicht: notice how I just disable polkit rules checking, because the executable has no chance of knowing whether the polkit rules are enabled system-wide <jpoiret>on Guix they're not in a well-known location IIRC <jlicht>jpoiret: yeah that makes sense, I'll play around with your stuff over the weekend. Thanks! <jpoiret>there also was a nasty gpgme bug that got fixed in core-updates, where the gpgme library didn't refer to gpg by full path and so you had to have gpg available in the PATH (!!!) <jlicht>It's good it's been fixed by now then :) <h4>chromium.scm: `ungoogled-chromium`: %chromium-version <bost>`guix system reconfigure syst-config.scm` returns "guix system: error: service 'ssh-daemon' requires 'networking', which is not provided by any service", however I can't find any 'ssh-daemon' service and the syst-config.scm contains `(use-service-modules networking ssh)`. Can anybody explain me what's going on? <ds-ac>ssh-daemon is provided by SSH services (e.g. openssh). It requires a service providing networking (e.g. NetworkManager or a static networking configuration). <ds-ac>My guess is that in the "service" field of your OS record, you have an SSH service, but no networking one. <andreas-e>bost: I suppose you have openssh-service-type in your system configuration, but are lacking something providing networking, such asnetwork-manager-service-type. <ds-ac>NB: "(use-service-modules a b c)" is an abbreviation to import modules (gnu services a), (gnu services b) and (gnu services c). It is not part of the configuration, which is the record (operating-system ...) at the end of syst-config.scm <meaty>When using the full guix system, do I need to define and source $GUIX_PROFILE/etc/profile? and if so, what's the "canonical" way to do it? <jpoiret>meaty: on guix system /etc/profile does that among other things <meaty>jpoiret: yes, but what about sourcing /etc/profile itself? <meaty>should I just include the lines in `home-environment-vars-service-type`? or if it's something all users would have to do, set them in system config somehow? <futurile>can someone see if they can build python-apprise on master - I can't seem to find the right place to check if CI is building it <futurile>it was working a few days ago, but now gives a failure on the tests phase for me <fnat>futurile: Yeah, it seems to be failing here too. <cbaines>would be pretty simple to add build statuses there though <ieure>Hmm, upgraded another box last night, and ran into the same failure mode I was describing yesterday -- when I try to log in, the screen flashes "Module unknown" or something similar, and logs me back out immediately. <ieure>But I did the guix system reconfigure interactively, so I found that my local KDE session also completely broke -- couldn't launch any software, couldn't log out, couldn't reboot. <ieure>Had to use the power button to get it to reboot. Worked fine after. Something with the upgrade process is very busted, though. <meaty>this is prolly in the manual but is there a quick way to search the installed/stored packages and get the install prefix (i.e. store location) of one? <andreas-e>Can you be more precise? If you know the package name and are interested in the one corresponding to your current Guix, then "guix build PACKAGE-NAME" prints the directory name. <meaty>actually, that's exactly what I need, thank you <andreas-e>(which works only for packages with only one output) <Rutherther>meaty: keep in mind the path can disappear in the future, when gc is ran, unless it has a proper gc root <meaty>and that's ok, I usually do this in an interactive shell <ieure>I generally prefer $() for shell substitution these days. <meaty>I was doing `find /gnu/store/ -maxdepth 1 -name 'PACKAGE'` before <Rutherther>Either adding it to a profile that you keep or stuff like build and shell accept -r. I am just mentioning this to note you should not store the path anywhere unless with these options <meaty>nikolar: actually this is my first time hearing of `` and it seems more convenient <nikolar>also posix defines just $() if i am not mistaken <Rutherther>Ive met some people on nix channels that did this and got surprised when the paths disappeared, so just wanted to quickly mention it. But this can also be important with some build caches like what cmake does, when you use it in a shell <zge>Does anyone here use Guix as an operating system together with Flatpak (for specific GNOME software)? <cbaines>Flatpak is more about proprietary software, if it's being used to distribute free software then there's a significant overlap with Guix <zge>I know, but there are gnome tools I use that are only distributed via flatpak, and aren't part of the guix repo yet (and I am not well verged with guix enough to take it on myself to add them) <zge>the matrix client, fractal, is one example I couldn't find for guix <wizard>sometimes i use flatpak when i want software that's not in the guix repos and would take a significant effort to package <meaty>what does "cannot determine provenance for current system" mean and how can I fix it? <ieure>zge, element-desktop is in nonguix; I believe it's free, so I'm not sure why it's in there. <zge>wizard: so you are saying it works without any greater issues? <wizard>zge: yeah with a lil bit of config it works just fine <ieure>zge, I looked into packaging FluffyChat, but it depends on a non-free framework & build tool. <zge>wizard: what kind of config? <zge>ieure: non-free build tool? <wizard>stuff like sourcing the environment variables properly and whatnot <ieure>zge, Flutter, the framework comes with build tooling. <ieure>Maybe this is free, or has become so since I looked? <ieure>I couldn't find anything but non-free binaries last I checked. <podiki>element-desktop is an electron app, falls under the "impossible to bootstrap/package from source fully" javascript <podiki>not a question of free, but of how we (guix) package things <cbaines>although it is a question of "free", as feasibility matters and is a freedom issue <cbaines>"free software" only distributed as a flatpak which you can't modify is less "free" than equivalent software which you can modify easily <zge>a general question: when downloading a stable guix release, is the idea not to do guix pull? <Rutherther>You shouldnt really use releases, at least for now, since even the newest is very outdated <zge>Rutherther: debian-stable outdated, or seriously outdated? <podiki>cbaines: yes, of course. though here i think it is more an issue of bootstrapping/source dependency as the ecosystem doesn't lend to our packaging (agree though that it does in some sense impinge on freedom, but also "build from source" is apparently very different in js-land) <podiki>zge: Guix is rolling, consider "releases" just a more tested installer <Rutherther>The latest release is like 1.5 years and doesnt get even security updates. As far as I know debian gets security updates on stable releases. So it's worse. <zge>podiki: ack, so the expectation is to guix pull, even if I install a release iso <podiki>we should call it something else as it is a bit confusing <podiki>it is a snapshot for installing that is never touched after it is tagged <Guest66>hello all i'm a little desperate, i am locked out of my main computer because my guix reconfigure won't work <Guest66>i posted to the mailing list, it's the same issue of "the group guixbuild" does not exist <Guest66>the problem persists across a re-install <ieure>Guest66, This is Guix System? Not Guix on a foreign distro? <rynn>Which mailing list did you send it to? I haven't seen any new posts in a bit, though the mailing lists can be a bit slow. <ieure>Guest66, What installer are you using? <Guest66>the official installer from the website <Guest66>it seems like the error happens after the switch, and each subsequent guix command just errors out the same way <Guest55>webchat is flakey, new guest nick same issue <ieure>Guest55, What happens if you reboot? <ieure>I ran into some weird failures after `guix system reconfigure', but the system worked after rebooting. <Guest55>i can reboot to the previous configuration <Guest55>but same error if i try to update again <Guest55>using system build works just fine, it breaks as it tries to switch the generation <Guest55>re-creating the group, along with its users make the process repeatable, so it seems like the group and users somehow end up getting deleted? <Guest55>i'm using pure guix without any extra channels <omar_b>Can some one explain to me what does it mean that the QA status is "unknown" ? <cbaines>omar_b, you can get more information if you click through to QA <cbaines>in this case the data service is currently working on processing the revision, so there should be some more information in a bit <omar_b>cbaines, I did but I don't understand what's missing? <cbaines>the information is missing, so QA can't tell you a status yet, it's unknown <ieure>omar_b, I thik I've seen a non "unknown" QA status, like, twice. <cbaines>also, as that page says, substitute availability is low at the moment so builds for patches aren't currently happening <cbaines>March is quite a while ago, I'd send it again (to that same issue) as that'll at least bump it in people's inboxes and get QA to hopefully take a look <ieure>cbaines, Rebase the patch and send a v2? <vagrantc>it is hard to know weather to just wait and see if QA picks it up ... or if needing to actually do something. <cbaines>vagrantc, yeah, I think it should be possible to improve the "Issue not found" case, at least to check the issue status and give a status dependent message <PuercoPop>omar_b: If you are using gmail, it might be better to send patches as attachments intead of inline. To generate the file(s) to attach you can use git format-patch. ej. if you want to generate the patch for the top-most commit, `git format-patch -1`. That is what I've been doing for the patches I've sent so far. <cbaines>vagrantc, for that issue, at least QA knows about it. It's just quite a long way down in the queue to be processed. <omar_b>PuercoPop, My first submission was done manually using pasting the patch in gmail in the browser, but then I sent the patch again using git send-email to make sure it's formatted properly. <cbaines>I've been putting in more time and effort to try and optimise the QA data service, and I think it has helped, but I've been focusing on trying to make best use of the limited resources. It probably could use a more powerful machine to go faster. <vagrantc>yeah, seems like there is just a lot to build! when it works it works great! <cbaines>if each revision takes even an hour or two to process, and there have been ~10 new revisions on master just today, there's just a lot of stuff to process before even getting to patches <cbaines>and QA has only managed to process just 4 of those master revisions <vagrantc>when submitting patches i use git format-patch --base=master which then encodes which commit the patch is built against ... but sometimes wonder if it wouldn't make sense to at least just try to apply the patch on current master rahter than tying it down to a specific old version <vagrantc>not sure how QA approaches that quandry ... <ieure>git format-patch is very irksome, I don't understand why it can't determine the base commit. I have never been able to make this work, and always just use `-5` (or whatever) to pull the most recent commits. <ieure>It supposedly can autodetect the base commit, but I have never once gotten that to actually work. <cbaines>vagrantc, QA doesn't use the --base thing, it always applies against the latest master (at the time it applies the patch) <cbaines>then it fails and tells be it needs --base <ieure>Feel like I'm dragging myself up a mountain of broken glass by my lips using the email patch flow. <vagrantc>it is useful to know what revision it was generated against, if, say, it does not apply to master ... makes it easier to rebase correctly <nckx>ACTION uses format.useAutoBase = whenAble <nckx>I've never typed --base. <eikcaz`>Just set up build offloads from my 2.8G RAM aarch64 device to beefy dekstops. The question is: how did I survive without this? <eikcaz`>(Thanks Guixers for making such an amazing set of tools!) <ekaitz>Rutherther: you need to add libstdc++ to CROSS_* <ekaitz>yeah the cross compilers in guix search for libs in CROSS_LIBRARY_PATH and stuff like that <ekaitz>if libstdc++ is not listed there it doesn't find it <ekaitz>now that it finds the libstdc++ (adding it manually) i find this error: undefined reference to `_exit' <ekaitz>hm! yes, but it's not in there in my case <ekaitz>or at least it doesn't find it by default <Rutherther>okay, looking into it more you are probably right. But it's not that the path wouldn't be in CROSS_LIBRARY_PATH, it is that the output was to /lib instead of arm-none-eabi/lib <Rutherther>getting undefined reference to "_exit" is expected, since it's a syscall, so you would have to provide it yourself for embedded targets <ekaitz>/gnu/store/pakv32qsbfi9hdb5l3rxqfnpqpm52yzg-arm-none-eabi-nano-toolchain-12.3.0/arm-none-eabi/lib/libc.a vs /gnu/store/pakv32qsbfi9hdb5l3rxqfnpqpm52yzg-arm-none-eabi-nano-toolchain-12.3.0/lib/libstdc++.a <ekaitz>about the _exit thingie... the toolchain should provide it but I don't know why it's missing <ekaitz>when we have this fixed i'll try that further <Rutherther>looking into the gcc 7 toolchain it has libstdc++ also under lib as opposed to arm-none-eabi/lib, but I believe it's a mistake. This stuff has to be distinguished by the architecture. Otherwise it's impossible to then have libstdc++ for the regular gcc in the same profile, which should be possible <ekaitz>hmm Rutherther according to the rp2040 docs this should just work <ekaitz>and this undefined symbol problem i don't know where is coming from <PuercoPop>cbaines: would it make sense to use guix containers to have separate work queues for building master vs building patches? <Rutherther>"having a arm-none-eabi-gcc and -g++" is very inexact, since having only those, without the newlib and possibly libstdc++ cannot compile much. They just don't list much info since they already expect a fully functional toolchain <PuercoPop>And have having curiass be the way to land changes in the master branch being considered? Ideally one would run QA before integrating the change <ekaitz>Rutherther: like the toolchain coming from the debian repos, for example <ekaitz>sudo apt install cmake gcc-arm-none-eabi libnewlib-arm-none-eabi build-essential libstdc++-arm-none-eabi-newlib # this should do <ekaitz>i'll try later with a debian machine <ekaitz>when I try it properly i'll ping you back <aswjrisp>I am trying to start sway manually from a tty. With the command `sway`. When I run sway I get this output "XDG_RUNTIME_DIR is not set in the environment. Aborting". <aswjrisp>I am using %base-services. I have added seatd-service-type, polkit-service-type and dbus-root-service-type to %base-services. <aswjrisp>I have configured seatd like the example in the guix manual. <ieure>aswjrisp, When you say "from a tty," you mean, from the Linux console? <aswjrisp>seatd-service-type and elogind-service-type appear to conflic with each other. <ieure>aswjrisp, What starts the Wayland server? <ieure>I don't know much about Wayland, but I'd assume you need some kind of thing to start Wayland before clients can connect. <aswjrisp>ieure: I was assuming that sway would take care of that. Or that was my impression based on previous error message I had seen when trying to start sway. <aswjrisp>I would like to start sway without %desktop-services. <aswjrisp>I also would like to start sway manually without a display manager. So I do not think that I need greetd which was mentioned in the guix manual when I searched for sway. <aswjrisp>ieure: Looking at previous sway errors I saw wlr in the error output. I think that wlr is refering to wlroots which provides wayland compositor features. <aswjrisp>I get the same "XDG_RUNTIME_DIR is not set in the environment. Aborting" error regardless of the user I run sway as. <ieure>I'm not sure what usually sets that, the DM? <ieure>PAM as part of the DM login? <aswjrisp>I have seen examples of people starting sway just by typing sway at the linux console without using a display manager. So maybe I have to set that environment variable mannually. I am not sure what it needs to be set to though. <ieure>I think it's usually /var/run/your-user-id or something very similar to that. <ieure>It will probably complain about other XDG_ variables if you set it. <aswjrisp>When I search for XDG_RUNTIME_DIR in the guix manual I get information on greetd. <aswjrisp>So I will try setting up greetd and then sway may work. <aswjrisp>In the Base Services section of the manual. <aswjrisp>It uses %base-services and does not use seatd or greetd. <aswjrisp>So I am going to switch back to elogind-service-type instead of seatd. <oriansj>(including vm steps if you want to do dry runs prior to messing with your physical machine) <aswjrisp>oriansj: thanks for sharing those links I am looking at them I see you used seatd and greatd. <aswjrisp>oriansj: I think I will try copying your seatd and greetd parts of the config and hopefully then sway will start. <oriansj>basically my configuration is the guix default minus everything one could reasonably remove.