IRC channel logs
2023-06-06.log
back to list of logs
<bumble[m]>I tried to guix pull and guix system reconfigure today but am seeing a new error today that was not there before `error: "swaylock" invalid field specifier` would anyone here happen to have any solutions to suggest? <bumble[m]>thanks in advance for any recommendation or advice anyone may have <ChocolettePalett>I assume that specifying field names is mandatory, e.g. instead of "swaylock" you should write (name "swaylock"). <bumble[m]>when the swaylock pam setuid issues popped up a week or two ago, I tried using this newer interface but would get reconfigure errors <bumble[m]>and maybe I needed to pull and update one time before using the newer interface and updating again <bumble[m]>I tried removing this and the result was that swaylock wouldn't run and the pam setuid error would be returned instead <isf>hello, how I can run gnome-disks as super user? when I try it dont find it even when I have it installed as root too <isf>i open gnome-disks for open a hard drive automathically but just do it for super user, as I need sudo -E to run it, I want to access to the hard drive with simbolic links. Someone know how to do this better? <lilyp>Mari[m]: using udev-rules-service <Mari[m]><lilyp> "Mari: using udev-rules-service" <- how would i add the udev rules using that? <Mari[m]>there's no package for the ones i want <lilyp>you can try and see whether they are in our dolphin-emu package; if so, you can use dolphin-emu <lilyp>if not, you have to roll your own package based on copy-build-system <Mari[m]>problem is the dolphin package for guix is broken <Mari[m]>whenever i try to install it, it fails <zamfofex>Is it just me or do “bare” freedesktop.org GitLab URLs not work wit Git anymore? I get error 500 for e.g. wlroots. It seems adding a ‘.git’ suffix to the URL fixes it. <efraim>janneke: I don't know a lot about the python test suite, but I believe it runs twice, in your 177/401/4. it would be 177 passed of 401 (with the others not yet run), with 4 failed. I think it runs concurrently 1 test per core <ChocolettePalett>I assume there is a correlation between the number of cores and the number of failures <panosalevro>I get `commit ... lacks a signature` when I try `guix pull` <atuin>I am trying to debug this error when using nix-daemon service: `opening pseudoterminal master: No such device` when running `nix-channel --update`. Seems that the problem is that devpts is not mounted in the mnt ns. Who creates that ns, make-forkexec-constructor? <atuin>also restarting the daemon with herd seems to fix the issue <atuin>After restarting the daemon the devpts fs is mounted, could it be a race condition_ <atuin>after computing the deps graph seems that nix-daemon has no dependencies, similar to iptables and others, that could be the cause indeed <evilsetg[m]>Is anybody else having problems with swaylock? After the latest reconfigure I get following error: <evilsetg[m]>`[source/pam.c:17] swaylock is setuid, but was compiled with the PAM backend. Run 'chmod a-s swaylock' to fix. Aborting.` <tlecarrour>Hi Guix! Before spamming `help-guix`, I wanted to take my chances here… anyone running into `In procedure canonicalize-path: No such file or directory` when running a container built with `guix system container`? <jpoiret>evilsetg[m]: you now need to use the screen-locker-service-type with (using-setuid? #f) <jpoiret>the manual has been updated to reflect this change <jpoiret>tlecarrour: do you have a minimal working example of this? <jpoiret>if it's long you can use paste.debian.net <tlecarrour>jpoiret: actually, I had to add `(service syslog-service-type) (service nscd-service-type)` this moring. It used to work without them. <jpoiret>huh, why do i need to download linux-libre for this 🤔 <jpoiret>huh, can't reproduce myself. Do you have the full backtrace, or preferably even the full log? <jpoiret>tlecarrour: can you find out what file /gnu/store/bp029p543bvsprbnz2v0p33 is and paste it? <tlecarrour>jpoiret: no exact match, but /gnu/store/bp029p543bvsprbnz2v0p33yf1n8k8yk-system, which contains `activate boot etc kernel locale profile` <jpoiret>HiltonChain[m]: it might be because shepherd now starts services in parallel <chomwitt>if i edit my /etc/fstab to add some options that i dont have in my /etc/config.scm could or will guix on reboot overwrite fstab values with those in config.scm ? <jpoiret>civodul: huh, isn't the bug you reported this morning *really* bad? <jpoiret>chomwitt: yes, you should add the devices in /etc/config.scm <jpoiret>tlecarrour: what's your `guix describe`? <tlecarrour>jpoiret: depends on what you mean by "old" ! I installed it 4 years ago (no full reinstall since then) but updated (pull & reconfigure) it this morning! <tlecarrour>jpoiret: yes, sorry, I'm one of "those" people who updates weekly! 😅 <jpoiret>i just end up spending more time using my git checkout that the guix pull'd guix so I always forget to update <chomwitt>jpoiret, i added all the devices in config.scm . but i just tweaked some minor options to a certain filesystem but i noticed on reboot that fstab is again as before. <jpoiret>you need to add these options to the config.scm in any case, fstab is generated from it and changes to it on disk will not persist <chomwitt>jpoiret, ok. thanks for the confirmation. <jpoiret>tlecarrour: let me try with a fresher guix if I can reproduce <evilsetg[m]>jpoiret: I think there is a small typo in the swaylock example for screen-locker-service-type. The example for swaylock uses the xlock executable. Here is a patch to fix that: <jpoiret>cbaines: reading your patch about the branch workflow and QA, how does one register a branch for QA? <civodul>ACTION would love to master his agenda <jpoiret>today is the swedish national holiday, so I have time :) <civodul>it's due to the recent innocent-looking change in 'modify-services' <civodul>modify-services would no longer preserve service order <jpoiret>tlecarrour: can you run`guix gc -R "$(readlink -f /gnu/store/bp029p543bvsprbnz2v0p33yf1n8k8yk-system/activate)" | grep .scm | xargs tar cvzf /tmp/archive.tar.xz` and send /tmp/archive.tar.xz somehow? <jpoiret>this basically creates an archive with all the .scm files the activate script depends on <jpoiret>i'm thinking there might be an ordering problem with the activation scripts <jpoiret>i always feel bad about using multiple reverses <jpoiret>although I think it's the same bug that civodul is fixing :) <jpoiret>evilsetg[m]: thanks, i'll look into it later <jpoiret>not yet! and it's not even his code :p <tlecarrour>too bad! … by the way, thanks for your time and help! swedish national holiday or not! 😉 <jpoiret>can you also paste the script that `guix system container ...` gives you? <jpoiret>yeah, so it's definitely the same bug: the activation-service-type's extension to boot-service-type is ordered after shepherd's, meaning that shepherd boots before the activation script runs <jpoiret>so /run/current-system is never properly symlinked, hence the error <civodul>the issue of ordering totally escaped me when i reviewed that code <civodul>jpoiret, tlecarrour: lemme know if it fixes your problems, whatever they are! :-) <janneke>"check for stale shepherd socket and remove it" <jpoiret>it's a shame we depend on the ordering of the service extensions <jpoiret>what's weird is that I couldn't reproduce with my checkout <jpoiret>evilsetg[m]: how did you send your patch? it doesn't apply, and it seems that the whitespace has been replaced by nonbreaking space <bjc>speaking of service ordering, can anyone look at #62726? it moves setuid-program activation into shepherd so it can be ordered properly. currently, things like wireshark don't work because the ordering between group and setuid activation is undefined <attila_lendvai>pulled master just now, but it fails with "In procedure struct-vtable: Wrong type argument in position 1 (expecting struct): #f", while compiling guix-system-tests.drv <evilsetg[m]>jpoiret: Hmm, I used `git send-email -1 -a --base=auto --to=guix-patches@gnu.org` and copied the output from the terminal into icedove. <tlecarrour>civodul: I pulled the latest version… and… `guix pull: error: build of `/gnu/store/lcl2pvk4rw6b3rkg76sxc8yckvm1qf1k-profile.drv' failed`! 😱 <tlecarrour>`gnu/services.scm:348:48: In procedure struct-vtable: Wrong type argument in position 1 (expecting struct): #f` <tlecarrour>somewhere in `In gnu/services.scm: 348:48 0 (_ (#<service-type mingetty 7fffe0d70680> #<procedure 7fffdc904fa0 at ice-9/eval.scm:333:13 (a)> ((filename . "gnu/tests/desktop.scm") (line . 123) (column . 3))) #f ())` <chomwitt>In config.scm trying to provide options to an ext4 fs mount (options "rw,relatime,nofail 0 2") the filesystem wont get mounted and i see in dmesg [ 41.928441] ext4: Unknown parameter 'relatime' <chomwitt>i see that message after many attempts of shepherd shepherd[1]: Service file-system-/home/chomwitt/MyCalibreLibrary-SSD failed to start. <chomwitt>strangely if i dont supply any options then the 'defaults' will be added to my fstad and the filesystem will be mounted ok with rw,relatime <chomwitt>but wanting my system to not hang during booting if my external storage is not connected i made a couple of tests with 'nofail' but with no luck. <jpoiret>evilsetg[m]: icedove messes up with formatting <jpoiret>you should let git send-email send the mail itself <rekado>make as-derivation fails with the above error <jpoiret>tlecarrour, attila_lendvai, rekado: i'm on it <rekado>it’s the builder for guix-system-tests <jpoiret>just need to test the easy fix I have <jpoiret>make as-derivation is pretty long though <jpoiret>bjc: haven't taken time to reply to it, but here's one blocker for it: nothing depends on the shepherd service <jpoiret>so you can't guarantee that eg. login is available setuid <jpoiret>ah, maybe login uses PAM, but the point stands <jpoiret>we'd have to make a careful review of what services depend on setuid programs being available <bjc>jpoiret: yeah, the lack of dependents occurred to me, too, but i didn't see anything in ‘/run/setuid-programs’ that wasn't meant for interactive use. doesn't mean i'm not missing something obviously <jpoiret>even if they're for interactive use, we have to make sure they're available before the user is able to interact with the system <bjc>that's relatively easy, right, by making user-homes dependent? <jpoiret>i'm not sure that's the right synchronization point <jpoiret>also, you might have looked at your own setuid-programs directory, but not all of the ones provided by Guix, right? <jpoiret>i see at least some dbus stuff on mine, so I would expect the dbus service to be impacted <jpoiret>eg. people with autologin or with complicated DMs like GDM would have problems I think <jpoiret>rekado, tlecarrour, attila_lendvai: just pushed a fix <jpoiret>if only we had static typing to prevent easy mistakes like this <bjc>jpoiret: yeah, i didn't do a thorough audit of all setuid-programs. it didn't even occur to me, tbh <bjc>the “right thing” would be to have each service that needs it be dependent on it, rather than a single synchronization point <jpoiret>yeah. Still don't know about ensuring they're available for interactive use though <bjc>it doesn't look like there's too many, less than 12. i'll see about modifying them <jpoiret>maybe adding it to each DM/login process is possible, but it's quite annoying and prone to forgetting <bjc>well, one of them is ‘pam-root-service-type’, so that should cover the greeter bases <jpoiret>greeters are not the only dependents of pam though <bjc>i don't think i'm understanding you. if pam requires setuid-activation, then anything that requires pam would, as well <bjc>i'm not suggesting only changing pam, just mentioning that adjusting pam would allow greeters to work in a better defined fashion <jpoiret>pam doesn't require setuid-activation, it's orthogonal afaik. What I mentioned before with login was maybe imprecise <bjc>pam does require it, though, for ‘sbin/unix_chkpwd’ <jpoiret>ah, right. Still, I have the feeling that using only this dependency to ensure setuid-programs are setup for user interaction will be flaky <evilsetg[m]>jpoiret: I resent the patch again via git send-email. Is it possible to apply it now? <jpoiret>evilsetg[m]: It does! Although, note that you shouldn't modify anything above the -- line, as that will be part of the commit message (except for the automatically inserted From:) <bjc>jpoiret: the main problem is the dependency graph is under-specified. i'm not sure how to fix that for everything in one patch. we *can* fix it for (hopefully) everything that has a shepherd service (such as pam), which covers a lot of uses. the rest will have to be done as they're discovered, i think <bjc>but having setuid activation itself be underspecified is also a problem, as evidenced with wireshark <jpoiret>in general you should follow the ChangeLog format to write commit messages, also make sure to capitalize the patch title. Other than than, seems good, i will apply it :) <jpoiret>bjc: yes, that's the main problem. I'm not against the idea of the patch at all, but I'm trying to anticipate the problems that it could cause <jpoiret>we're working backwards by trying to untangle the dependency graph unfortunately <jpoiret>i didn't spot it either when I read your code :) <jpoiret>trying to figure out why b4 mangled unmatched-paren's From: line <William[m]>how would you set an env var with guix shell? I tried a (setenv "HELLO" "world") after set-paths phase, but that didn't work. I'd like to run "guix shell" and be able to echo $HELLO in a shell. could someone point me in the right direction? <bjc>jpoiret: yeah, it's tricky. but thanks for the feedback; i can at least go through the existing setuid extensions and fix what i find there <cbaines>jpoiret, regarding your earlier question about qa.guix.gnu.org processing branches, it looks at guix-patches issues <jpoiret>right, but how are these branches picked? are they all of the branches? <cbaines>currently it just builds the branch corresponding to the oldest issue <cbaines>but it could do anything really, I set out in the patch I sent how to use the issues to specify the ordering in which the branches should be merged <cbaines>data.qa.guix.gnu.org processes all the branches, so the information is all available <tlecarrour>jpoiret civodul your patches solved my problem! 👍 Thaaaaanks! <dthompson>anyone have experience using chicken scheme with guix? <dthompson>I'm not familiar with it and I'm trying to build someone's project but chicken-install is trying to install stuff to /gnu/store and failing (of course) <dthompson>chicken-install is how you install stuff with their package manager. maybe that's just plain broken on guix. <dthompson>I guess probably I would need to package all the relevant libraries. <iska>dthompson: can you point chicken-install to a user folder? <iska>I know quicklisp breaks with FFI packages, so anything that uses an FFI has to be packaged and visible <apteryx>initial guix pull on raspberry pi 4: 31.5 minutes <jpoiret>ham5urg: most of the setup takes place outside the container. Currently this isn't possible in a container.scm <ham5urg>jpoiret, such feature would enhance Guix to be used instead of Nix and sometimes even replace Docker. <apteryx>hm, opencv doesn't build on aarch64-linux <lambdanil>hi, I'm currently experiencing a bit of an issue with dbus <lambdanil>for some reason my flatpaks don't connect with xdg-portals properly <lambdanil>if I start a flatpak and then run xdg-desktop-portal -r manually it fixes itself <jpoiret>lambdanil: does that fix it for good? <lambdanil>also possibly related is that virt-manager fails to connect to libvirtd completely <lambdanil>internal error: Unable to get system bus connection: Could not connect: No such file or directory <lambdanil>I think there's just something weird with dbus or something? <lambdanil>sway, I'm invoking it with the dbus-run-session or whatever it's called <jpoiret>do you have the proper XDG_CURRENT_DESKTOP variable set to sway? <lilyp>does this happen once per session or once per reconfigure? <jpoiret>and do you run dbus-update-activation-environment afterwards? <lambdanil>I have a dbus-update-activation-environment in my sway config <lambdanil>jpoiret: it does look like XDG_CURRENT_DESKTOP isn't set right <jpoiret>you need to update the XDG_CURRENT_DESKTOP variable for the desktop portal to pick up when launched automatically through dbus, otherwise it won't find the proper backend <lambdanil>because I tried fixing it yesterday and did that but the other way around <jpoiret>yes, run `dbus-update-activation-environment XDG_CURRENT_DESKTOP` after setting the env variable should be enough <lambdanil>thank you, I'll dc but I'll be back in a bit once I test it <jpoiret>i have `exec "dbus-update-activation-environment WAYLAND_DISPLAY XDG_CURRENT_DESKTOP XDG_DESKTOP_PORTAL_DIR"` in my sway config <jpoiret>WAYLAND_DISPLAY might be needed for screen capture <lambdanil>okay thank you, I'll try it out with these variables <elevenkb>Basically the geiser-guile repl doesn't work at all for me to evaluate packages. <elevenkb>Using guix repl directly does work though. <elevenkb>Nvm. Pls ignore.... something has changed. <lambdanil>jpoiret: so I added the dbus thing in my sway config, WAYLAND_DISPLAY, XDG_CURRENT_DESKTOP and XDG_DESKTOP_PORTAL_DIR are now all correctly set but unfortunately nothing changed <jpoiret>is it set in the activation environment? <jpoiret>you can check by finding the running xdg-desktop-portal process and cat'ing /proc/<PID>/env <lambdanil>well now my portals aren't starting at all on a fresh login for some reason <jpoiret>yes, that's what they're supposed to do <civodul>sneek: later tell ngz how far are we from merging 'tex-team'? <lambdanil>well, I saved the environ from the autostarted one and the manually started one, I'll try to compare them to dinf a difference <cbaines>civodul, I see some log entries regarding ludo on hamal, do you get an authentication error? <cbaines>if so, try again, I think there was a wrapping issue with the SSH key <jpoiret>civodul: right! i'm stacking a couple of patches so that I don't push too many times <lambdanil>I don't know what's wrong but I think it goes deeper than the portal issue since virt-manager fails to connect to libvirt <lambdanil>despite the socker existing and having all the correct permissions and groups <jpoiret>you can also monitor dbus by using `dbus-monitor` (and possibly `--system`) and see if there's anything there that's going wrong <civodul>cbaines: i don't see the "Grace period" messages in /var/log/messages*, so maybe it's too late <civodul>anyway it's running shepherd 0.10.0, so you might want to upgrade <davidl_>how do you put a g-expression inside a build argument, or more specifically to a configure-flag string? <apteryx>#:configure-flags #~(list --prefix=#$output) <apteryx>err, #:configure-flags #~(list "--prefix" #$output) <bdju>The pcsxr package fails to build. I'll send in a report with the build log in a bit... <elevenkb>Hello there is there any way to invoke guix build --no-offload from the repl? <apteryx>you may be able to do so by setting the build options before invoking ,build <apteryx>see the 'set-build-options' procedure from (guix store) <elevenkb>apteryx: thanks for letting me know about this. <elevenkb>I wonder: how come it isn't set up so that if you can't contact an offload server then you build locally? <elevenkb>Or maybe the idea is that you add localhost as an offload server in that case? <apteryx>elevenkb: I think the original need was 'I never want this machine to build anything' <apteryx>at least 3 people would be interested in that feature; you, nckx and myself :-) (configurable local-fallback capability) <jackhill>I presume that any arbitrary logic can be implemented in machines.scm. Perhaps we need a module of commonly used helpers. <jlicht>hmm, (modify-services %base-services (delete mingetty-service-type)) seems to still contain mingetty-service-type services on current master. Not quite sure what changed recently <davidl_>apteryx: Thanks, I managed to make it work! <apteryx>are we already building our packages with GCC 11? <apteryx>I thought we were still with GCC 10.3.0 <f1refly>the package `kvantum` needs `wayland` in its inputs so it can run natively with the wayland plugin <f1refly>when setting QT_QPA_PLATFORM=wayland the application crashes when wayland is missing from its inputs <mirai>f1refly: can you send a patch or open a issue? <f1refly>I'm still unsure how to test patches before submitting them, would it be possible to get help regarding that tomorrow so I can create the patch? <f1refly>There are more qt packages that need the exact same fix that I'd like to submit patches for :) <cel7t>For testing patches, try using ./pre-inst-env <mirai>Unrelated to above, is it acceptable to send a patch-series that depends on other patch-series? Or should I group it up into a monolithic patch-series <mirai>I've got what roughly can be classified as 3 patch-series (that depend serially): define-configuration serializer-kwargs feature < (gnu services configuration generic-ini) module < NetworkManager refactoring/new fields <f1refly>I tried that but I had issued where it would use the regular repository instead of the local checkout. I will try fiddling with it again tomorrow I guess. <iska>I want to delete my home generations but keep everything needed to recreate them (to save space), which files do I need to grab from /var/guix/{...} <iska>I know normal user profiles just need manifest files