IRC channel logs
2025-03-10.log
back to list of logs
<redacted>try removing the space. you can roll back your configuration changes, so don't be afraid to try things <graywolf>Am I the only one who sees `redacted ?'... <coyotes4ys>while it's going, it appears my soundcard is not libre compliant, but i need to make sure of that <coyotes4ys>no soundcard is detects in wherever it is, etc/cards or whatever <jakef>is it just me or do the guix commands show the download progress differently now? the quick ones don't show a progress bar <jakef>they end with e.g. "2.7MiB/s 00:00 | 35KiB transferred" <the_tubular>How would I go to enable the SSH EnableEscapeCommandLine option? In a guix way <wakyct>the_tubular add it to your .ssh/config managed by Guix Home? <the_tubular>Yeah, that's what I'm currently doing, was wondering if there was guile binding <meaty>what causes 'wget' to be included in the user profile by default? I want to remove it from interactive use <lfam>meaty: It's provided by %base-packages-networking <lfam>Urgh. Looks like I have a system generation composed of zero-byte files <lfam>Not trivial to switch to another system when the current generation's scripts are empty <vhns>Seems that he meant `guix build -f file.scm --no-grafts -d` <adanska>Hi Guix! I'm having trouble with gnome-46. In `gnome-settings-daemon`, the usb-protection daemon keeps crashing on startup, causing the whole session to crash (https://paste.debian.net/1362233/). I've reconfigured with the usb-protection disabled via a build option in `gnome-settings-daemon`, but since noone else is reporting this problem, i think its something wrong with my setup. <adanska>also, keybindings dont work when i reconfigure (even if its not crashing because of usb-protection), yet they work on my old gnome-44 config. <adanska>I would love some help with this, some tips to get better debug information, etc because this is really sad <adanska>its a blocker from me using Guix System further, as I can't reconfigure my system <hugohugo>Hi all, if I do a "guix pull", it gives me this error: "guix pull: error: Git error: object not found - no match for id (18e6031f7405b8b9c38255db1957e1eb064d4632)". Could one of you help me debug this? I have no clue where to start. That particular commit does exist and is from March 3 <hugohugo>I don't recall doing anything specific on March 3. Maybe I ran out of disk space, as that occassionally happens <hugohugo>I don't even know where the guix repository that "guix pull" updates is located. But maybe I can find that <hugohugo>let me try csantosb, I guess that could work because earlier I used the savannah channel. It has updated the channel and is now "indexing objects" <hugohugo>it seems that might have worked! since it got past the "Updating channel" part and went into "Authenticating channel 'guix', commits 9edb3f6 to 9866d32 (401 new commits)..." <hugohugo>through lsof I have found the git clones, they are in ~/.cache/guix/checkouts <hugohugo>and indeed something is broken, "git status" (on the savannah one) claims "error: index file .git/objects/pack/pack-de86f15d09681bdedc38dcdd4823e0a0ec1ba6f8.idx is too small". So I have removed (renamed) that checkout and now a regular "guix pull" will probably work. Thanks <guix-noob>Hi, i downloaded a Kernel from Kernel.org and tried to compile it but i am running into errors(see https://paste.debian.net/1362244/) . Is there a specific way to create the Build environment for Kernel compiles? Am i doing somethign ovious wrong? <df>so I currently use a ~/.config/guix/channels.scm to configure my channels and set up substitute servers in my guix-service config using substitute-urls (and authorized-keys) <df>I'm looking to move it all into guix-service config, but two things are puzzling me: <df>a) does channel-with-substitutes-available remove the need for a separate substitute-urls field (but presumably not an authorized-keys field)? <df>b) what is the difference in the guix field with a guix-with-channels config, and the dedicated channels field? <df>for b) the doc seems to imply that I need both but I don't know what each does <df>huh, although it looks like their is one default channel but two default substitute servers, so it seems like it wouldn't make sense to have a 1:1 mapping between channels and substitute servers... <df>oh I see, I misunderstood the purpose of channel-with-substitutes-available <df>still looks like it might be quite useful but it's not what I need for now <df>ah, and I see the channels field of guix-configuration affects /etc/guix/channels.scm as used by guix pull <df>still not entirely sure what guix-with-channels is for in that case (I get that it bulds a custom guix package with the channels specified), but I'm sure it'll become clear over time <ArneBab>I just updated my system packages and then my user packages and then got hit with the error ~/.guix-profile/bin/emacs: symbol lookup error: /gnu/store/ha8xhxc71vvkp24js9xxdg6ag8wry1zg-gvfs-1.56.1/lib/gvfs/libgvfscommon.so: undefined symbol: g_once_init_enter_pointer <ArneBab>I think the update of the user packages may have been aborted because my X11 broke at some point (though I thought that the update had been finished). Is there a way to ensure that system updates do not break running logins? <m4xxed>Hey folks, just updated my guix system and I was immediately super happy to see that gnome was updated from version 44 to 46! :) however this happiness didn't last long as immediately I found that NetworkManager <m4xxed>did not work anymore: "Could not activate connection: The VPN service 'org.freedesktop.NetworkManager.openvpn'was not installed." <m4xxed>Anyone else having the same issue? :) <rekado>m4xxed: do you have a VPN connection defined? Does nmcli on the terminal list the connection? <m4xxed>Yes, all configured, worked perfectly for the last few months. This error message is from "nmtui" and "nmcli" when I try to activate the connection (which is listed there as usual) <rekado>is network-manager-openvpn installed in the same profile as network-manager? <m4xxed>I actually found that there used to be a patch file (which is still in the repository), that was removed when NetworkManager was updated to 1.50 in the commit "1f26b0eec83b5dc949900a743ed01088cb093c65" <m4xxed>Still checking but maybe this patch file was dropped while still being required to make network-manager work <m4xxed>the file is called "network-manager-plugin-path.patch" if anyone wants to take a look that has more experience with guix and git patches <rekado>I'll see if the patch still applies <rekado>I literally just added (patches (search-patches "network-manager-plugin-path.patch")) back to the origin declaration. <rekado>I don't know if it *works*, because I have no VPN connection. <rekado>better ask apteryx about the commit. Perhaps dropping the patch was not on purpose, seeing that the patch file itself is still there. <apteryx>m4xxed, rekado it was indeed my mistake; I overlooked the patch when I changed the origin from the tarball to git <m4xxed>How did you know that apteryx is the right person to ask? I cannot find that name anywhere in the commit. <apteryx>m4xxed: many years of hacking along ;-) <m4xxed>Oh, that was quick. No worries, I was very happy that I could find that issue (almost) on my own and now I learned something new :) <apteryx>.guix-authorizations can help sometimes mapping real names to nicknames <m4xxed>just out of curiosity - how will you proceed now that you saw that there is a fix needed? (I mainly ask becase I am curious how quick patches will get into the main repo and how soon I can pull them and how soon they will be available via guix pull) <apteryx>m4xxed: oh, I thought rekado had applied it already but if not I'll do so now <apteryx>m4xxed: If you'd like your real name to appear in a 'Reported-by' git trailer, you can send it to me by private IRC message <apteryx>m4xxed: alright, pushed as 7239eb86b6 after testing it still built fine <apteryx>thanks to you and rekado for the report/ping! <m4xxed>Thank you for the quick response and the fix :) <apteryx>ArneBab: re the error, I'm not sure exactly what it is caused by, but it has probably to do with a mismatched version of system glib vs profile glib. A reboot should be the cure. <apteryx>the update of glib came with the gnome-team branch that was merged recently <hugohugo>For your entertainment: it seems I really botched up my guix installation last week... Not only was my ~/.cache/guix/checkouts corrupted, but apparently I now have many .scm files in /gnu/store that consist of only NULL bytes. That is, they seem to have the correct size, but all bytes are \x0;. Maybe my laptop battery ran out of power and didn't properly flush/sync or something like that... And the annoying thing is that these broken files seem to propagate. <hugohugo>I fixed my guix checkout, but for example the codeberg mirror guix pull (that csantosb asked me to make) resulted in /gnu/store/mk9zc25g5mxnb20pkzji2kp5bydw7wn7-guix-mirror-9866d32 which also contains all-NULL scm files. But these files are fine in the git clone, so they were copied from the store somehow. I still see ways paths to fix this though <meaty>Is nonguix a good example of a guix project operating on a forge-type system? I’m tracking the “migration to codeberg” thread and was wondering if it would be constructive to mention it <andreas-e>hugohugo: Maybe you could try "guix gc --verify=repair". <hugohugo>Thanks andreas-e. Just "guix gc --verify=repair" worked fine, but the NULL-bytes scm files are still there. I could try --verify=contents, but that will take a while it says. So maybe I'll run that overnight (because I already have a script running that checks all of /gnu/store for more NULL-files, as so far I only checked the scm files) <sneek>civodul, you have 1 message! <civodul>apteryx: could be, but IEE754 is “complicated” <civodul>would be worth discussing with the Python and maths teams though! <civodul>this reminds me that roptat once submitted patches speeding up CPython in Guix <civodul>some of them had to be dropped though, IIRC <ieure>My last `guix system reconfigure' deleted my user account, this is 100% my fault -- but if I reboot and pick the previous generation in GRUB, my account still isn't there. <jackhill>civodul: sorry to ping you directly, but do you have any thoughts on the authentication problem I ran into: #76660 ? <ieure>Is it expected that users are purely stateful and span across all system generations? <apteryx>civodul: I think roptat's patches were against 3.10, and 3.11 ships/configures by default with these speedier settings <cbaines>civodul, QA is back, although substitute availability for master is still in a bad way <cbaines>the newish problem I've picked up on yesterday is that data.guix.gnu.org has some derivations missing inputs and/or sources, which is causing trouble <civodul>but really, we have a triage problem <civodul>the underlying bug was reported 2y ago <civodul>cbaines: oh; any idea what could be causing this? <cbaines>civodul, I was looking through the data service code to see if any changes looked likely, but I couldn't see anything <cbaines>so my best guess currently is maybe it was running with a broken guile-knots for some period recently <cbaines>since fibers including the knots parallelism helpers are used heavily now when processing revisions, problems there could cause issues <cbaines>I've reverted some "speedup" changes that while they probably work, the slower approach is more likely to pick up on problems and fail, rather than just not quite working <jlicht>civodul: I sent a bulldozer-patch disabling all tests that seem to fail due to teeny tiny rounding issues for python-numba. Now the fine folks in python-team get to take a look at it ;-) <cbaines>at least this seems to be caught when the derivations are substituted on bayfront <cbaines>it's a slow and painful process to go through and correct the data in the database though :( <civodul>jlicht: excellent :-) make sure to ping’em if needed! <rekado>speaking of the python-team, I think it is recruiting. I'm not even a member but it feels rather lonely. <rekado>for a software collection as big as all the Python things the team is much too small. <apoorv569>Rutherther: Hey, I was having some issues during build of my project so had to fix that, its fixed now now I am another issue with guix build for the package I was working on. <apoorv569>after the build completes, it is supposed to run `npm ci --omit dev` but it then fails again with bad interpreter `/usr/bin/env`.. so I tried adding `--ignore-scripts` to `npm ci` command as well and rebuilding the `bcrypt` lib again after that.. but no luck its still errors out after `npm ci` command.. <apoorv569>I feel like this is way to hacky.. and probably going to bite me later someday. <Rutherther>apoorv569: could you send the full log so I see the command and the file? <apoorv569>files in node_modules/.bin are symlinks BTW that's why I am subtuting the file it points to <apoorv569>if we can set `/usr/bin/env` during build it will solve the problem.. we won't need subtitutes and multiple npm install rebuild etc.. <fommh>guix environment: error: build of `/gnu/store/jqymw455xxzl7vgpaw28k2y5rrmnps1l-gnutls-3.6.15.drv' failed <ArneBab>sneek: later tell apteryx: thank you for the info (re gnome branch) <apoorv569>csantosb: The only other place I saw guix instructions was tauri v1 <apoorv569>v2 is out but they don't have instructions, probably because of outdated packages in guix <csantosb>apoorv569: when I contribute a package, I try to update upstream instructions on installing it with guix <apoorv569>I do add guix instructions for my personal project though.. manifest and whatnot. <apoorv569>I am writing a service definition for intalling my project on guix right now actually. <csantosb>Packaging is great, making people aware of its availability with Guix takes a couple of minutes ... and maintainers are most usually grateful with packagers for distributing their work. <lfam>On my Guix System installation, the scripts that compose the active system generation are all zero-byte files. I'm able to boot into other generations via the GRUB menu, but I can't actually switch generations using `guix system switch-generation` because the code that is supposed to run is simply missing <lfam>For example: activating system... --> guix system: error: error parsing derivation `/gnu/store/10bkzzqykymz4kxx0ilicjd5kn4x52v8-module-import-compiled.drv': expected string `Derive([' <lfam>Any ideas for how I can manually change the active generation, at least enough so that I can switch to another one properly <[>guix pull: error: Git error: unexpected http status code: 502 <[>Is there any way to have guix pull ignore that and pull only from my channels? <[>Although savannah is no longer 502'ing <lfam>[: You could specify a commit of the Guix channel in your channels.scm <lfam>I'm not sure if it would actually prevent Guix from trying to fetch itself, but conceptually it should <dariqq>lfam: did you try guix gc --verify ? <lfam>Good question dariqq. I haven't tried it. I'm pretty sure it will tell me that something is broken but I'd like to get onto a working generation as my first move <df>is there a build-system for packages that use plain make/make install (without autotools)? <lfam>We just remove phases from the gnu-build-system for that, df <lfam>Specifically the configure phase <lfam>dariqq: Looks like maybe it was actually root's Guix that was corrupted. When I used a different Guix I was able to switch systems <lfam>Like `/var/guix/profiles/per-user/root/current-guix-25-link/bin/guix system switch-generation 90` <hugohugo>Oh I should have thought of that too lfam . I had several scm files that had only NULL bytes in them (so they were not 'empty', but still useless). I rebooted into an earlier system generation, and earlier generation of `guix pull` and `guix package`. Then I could `guix gc --delete` all affected files from /gnu/store, and now I'm upgrading everything again. (I hope the broken files don't somehow return.) <Rutherther>apoorv569: you can't really do that, there is a way to add stuff to the build container, but it is not configurable per package, only per guix-daemon instance. So your package would start depending on a certain way the daemon runs <lfam>Huh, I hope it's not affecting too many people! <hugohugo>probably a different issue though (not sure what caused mine, my guess is the battery went out, or I force shut down the machine myself) <hugohugo>In a way I am quite happy actually. For some reason I botched an upgrade, and it was possible to recover from that with only a little help <lfam>Yeah, it could be worse. Still, I'm not used to this with Guix! <lfam>Currently I'm running `guix gc --verify=contents,repair`, and it is finding lots of problems <lfam>I guess at the end I'll do some `guix gc --referrers` to figure out if any live generations are corrupt <hugohugo>I tried using `guix gc --referrers` today for the first time, but I couldn't figure out how to use it... `guix gc --delete` said `guix gc: error: cannot delete path `/gnu/store/b32zyn2mh7wgx7b5lwzc4na6w05v3szv-guix-9866d32' since it is still alive`, but `guix gc --referrers /gnu/store/b32zyn2mh7wgx7b5lwzc4na6w05v3szv-guix-9866d32` did not give any output. Should it? <lfam>You might need to look at `guix gc --references` hugohugo <lfam>But, your store item could be alive by virtue of being a current profile <lfam>Rather than part of an inactive profile. Does that make any sense? <hugohugo>It probably was still active somehow, since I've managed to garbage collect it after I made sure I switched all current profiles to older ones. I think I tried --references too, not sure though <rafael1>Hello. I'm trying to use `guix home` to manage my dotfiles and was wondering why the `guix home container` doe snot have the `--preserve` flag like `guix shell`. Any workaround to use `guix home container` within my foreign distro to test dotfiles with guix packages? <lfam>hugohugo: In the future, you could also try `guix gc --list-live | grep $filename` <Rutherther>rafael1: as a workaround just set the env var yourself in the container <rafael1>Rutherther: thanks, I'll try that :) <apoorv569>Rutherther: So what do you recommend? what is the way I can add stuff to build container? <rafael1>Rutherter: any differences to do this in Xorg or Wayland? <Rutherther>rafael1: see what are your current WAYLAND_DISPLAY and DISPLAY and set them in the container <rafael1>Rutherther: still struggling. I have exported as in `export` DISPLAY WAYLAND_DISPLAY and XAUTHORITY with he values at the host os environment. I get: couldn't open display:0 <Rutherther>yes. Or how would you do it with guix shell --container exactly? You need the socket to be able to communicate with x <Rutherther>possibly just exposing the xuathority could work. Really, it's same as if you did guix shell which I assumed you were comfortable with as you asked first about that, but what you would use --preserve flag for, you just set yourself <tschilp>wow, be it me having borked something earlier, or some magic <tschilp> happened between 1f26b0eec83b5d and 84e7feab83af2faf -- Gnome 46 <tschilp> screencasting on wayland is on for me now :). Will stay with this <tschilp> now! For some reason I cannot see gnome-maps, will try to add these <tschilp>Ha, they're actually here and work, as are gnome-online-accounts -- I just can't see them from the applications when pressing 'Meta'. <tschilp>I found the 'culprit' regarding gnome maps (but don't really know what to do about it) -- ~/run/current-system/profile/share/applications/org.gnome.Maps.desktop~ has ~Exec=gapplication launch org.gnome.Maps %U~. This command I can't run from the shell. For ~virt-manager~ -- an application I can see Exec points at ~/gnu/store/f6n0im0lmg04a086ypw3g03jz5675ir8-virt-manager-5.0.0/bin/virt-manager~, which is in line with what ~ls -lha <tschilp>$(which virt-manager)~ actually returns! <tschilp>~Exec=gapplication...~ is also used by ~/run/current-system/profile/share/applications/org.gnome.Shell.PortalHelper.desktop~, but I do not know what this is. <df>so I was messing around with guix pack and unpacked the result... and now I have a dir tree I can't delete <df>all the docs say that by default rm should operate on a symlink by default, not its target <df>yet I get permission denied errors trying to delete these symlinks that point into /gnu/store <df>hmm... maybe it's not a symlink thing actually <df>rm -f gnu/store/87z5k84hxbqs87plgwsl2v6a4j7m3k7h-bash-static-5.1.16/bin/bash <df>rm: cannot remove 'gnu/store/87z5k84hxbqs87plgwsl2v6a4j7m3k7h-bash-static-5.1.16/bin/bash': Permission denied <df>that is a regular file owned by my ordinary user <jA_cOp>to remove a file I think you need write permission on the parent directory - maybe the `bin` directory is read-only, even for the owner? <df>the directory is not +w, but I thought rm -f would overrride that <df>yeah I'll try changing that <df>yeah that sorted it, thanks jA_cOp <jA_cOp>I guess files in the store are read-only to prevent users from accidentally removing them manually <jA_cOp>So, I wrote a program and I want to deploy it on my system, so I wrote a Guix package for it, and a Guix service deploying it as a root Shepherd service as part of my system configuration. But in the package `source` I'm just using `local-file`. Will Guix know to rebuild my package if I update the sources? How should I best handle this situation? <jA_cOp>Would be interesting to read about how Guix caches stuff in general, if there's any info out there <eikcaz>In a nutshell, #75959 was pushed recently, but had some issues. #76379 fixes those issues (and others) but is backwards incompatible, so it would be good to push the more complete version before more people adopt soon-to-be-obsolete bindings <jA_cOp>eikcaz: thanks, that looks interesting! <ngz>qa.guix.gnu.org seems down. <kimapr>i'm starting to think my one line patch might have been lost in the fog <vagrantc>kimapr: something is amiss with the patch? non-UTF-8 encoding or something? <vagrantc>hrm. it is unicode ... but very long lines <vagrantc>anyways, a one-liner is easy enough to recreate <kimapr>I sent it from Thunderbird maybe that's a issue <vagrantc>gave me the excuse to try the mumi commandline today, which was on my TODO list :) <kimapr>actually, it could be my email provider <kimapr>i'm looking at View Source in the outgoing email in thunderbird and i can't see anything you're describing <vagrantc>well, both "git am" and "patch -p1 < patch" fail to see it as valid input <kimapr>Could it be because of the missing "From <gibberish> <date>" line? <kimapr>IDK how to include that in thunderbird <kimapr>i guess next time i'd have to learn another email client <jacobl>hello, am new to guix. I have been using guix home and have written a small home configuration. I want to push this to my git remote, so I need to set up an ssh key. Normally to do this I use `ssh-keygen`, but that program is not available, even with openssh in my list of packages for my home configuration. How can I get my hands on `ssh-keygen`? <vagrantc>Kimapr_: heh. your URL patch seems to be the same as my manually fixing git am <wakyct>jacobl strange you don't have ssh-keygen, it's on my system with probably similar config. <eikcaz>jacobl: ssh-keygen is included as part of openssh. See if specifying the whole path `~/.guix-home/profile/bin/ssh-keygen` works <wakyct>do you have the ssh service-type? <eikcaz>If that does work, it probably means you did not set up an appropriate shell service to source ~/.guix-home/setup-environment <wakyct>or I guess openssh-service-type would be the service in the system config <tschilp>I think ~home-openssh-service-type~ is mainly used for generating ~~/.ssh/config~ and ~/.ssh/known_hosts~ <wakyct>yeah, I realized I've always installed with those service types but I guess you don't need them for ssh and keygen necessarily? <tschilp>Yes, if I understand correctly, you should not need it for the ssh-client programs. <vagrantc>Kimapr_: at first glance, it is a tiny number of dependents, but each of the lua versions inherrits from it ... so it gets into a large-ish number of dependents <tschilp>the home-service is for file-generation and the system-service would be for running your openssh-server to my understanding <tschilp>so the openssh package should be enough without any service-types <vagrantc>Kimapr_: might make sense to exclude the older lua versions from the change <ieure>Is it possible to get a font in a `guix shell'? If I run `guix shell font-terminus gnome-font-viewer -- gnome-font-viewer', Terminus isn't in the font viewer. <lechner>ieure / did you try fc-cache -f -v ? <ieure>lechner, I didn't initially, but just did, and Terminus still isn't in font viewer. <ieure>Hmm, font-terminus only has PCFs, that's probably my issue. <vagrantc>i forget if 604 dependent packages sufficiently falls under "safe for master" <ieure>vagrantc, I think the number is 300. <ieure>"Changes which affect more than 300 dependent packages (*note Invoking guix refresh::) should first be pushed to a topic branch other than ‘master’;" <ieure>From 22.11.2 Managing Patches and Branches <ieure>lechner, Okay, no, still can't get it to work. `guix shell -font-orbitron:ttf gnome-font-viewer' and the orbitron font isn't visible in gnome-font-viewer, regardless of fc-cache running or not.