IRC channel logs
2025-03-04.log
back to list of logs
<ieure>alainalain, It's too inconsistent to say with any certainty. <alainalain>librewolf WORKS - it is showing web video on odysee (will see for yt) - THANKS <ieure>alainalain, Sure thing, glad it works for you. <alainalain>What are you using for virtual machine? gnome-boxes or virt-manager or ... [cloning dont work in gnome-boxes] <Kimapr>after updating my guix system several apps in my user profile broke until i updated them too with this error: symbol lookup error: /run/current-system/profile/lib/gio/modules/libdconfsettings.so: undefined symbol: g_once_init_enter_pointer <futurile>Kimapr: it's likely that after the update you had updated your GNOME stuff, but something was still running on the old version - logging in and out or restarting will probably fix everything <[>Does Guix contain all of the necessary packages (the free (Debian) version of the android SDK) to build android apps? <vagrantc>PotentialUser-79: might need to use agetty for serial console <apoorv569>PotentialUser-79: You can switch to using base-services instead of desktop-services <apoorv569>desktop-services is pretty heavy for what you doing.. I personally pick services I need out of base-services, don't even add that. <apoorv569>pick out of desktop-services I mean not base-services <vagrantc>wahoo! updated to a recentish guix build on aarch64 machine... <jmes>In a guix repl I can use ,build on a (program-file ...) and it gives me the store path which is my desired behaviour. But how do I go from program-file to store path in regular code? <wakyct>[: from the little I've looked into it, seems many Android packages are not in Guix. I want to explore this more though, would be nice to have a good Android dev environment in Guix <divya>It builds fine when done inside a guix shell --pure <divya>Also it's trying to use git inside the build, even though inside the guix shell it doesn't do that (it doesn't have git) <divya>Please ping me when replying. <divya>Guix has a really old version of it <apoorv569>can someone tell me why when I have openssh-service-type in my config, sudo herd status doesn't show any service for ssh and there is no ssh command available to use? <divya>mange: But that's the thing, I can't replicate this when doing `make` of the same repo inside a guix shell without git <mange>apoorv569: Obvious first question: you've reconfigured into the system, without error? The service name is "ssh-daemon", if that helps. <mange>divya: Can you paste the package definition you're working with? Are you starting with the sources that you get from "guix build -S $your-package", or something else? <apoorv569>mange: Yes, I'm testing a minmal server config in a VM. I reconfigure from a ssh session only after applying my config I cannot seem do `sudo reboot` I get error sudo: pam_open_session: Error in service module sudo: policy plugin failed session initialization Segmentation fault <apoorv569>so I hard reboot the VM then I boot in, and there is no service for ssh, sshd ssh-session nothing <apoorv569>I can share the entire config, pretty simple. <mange>Anything relevant in /var/log/messages? <divya>apoorv569: sudo herd status doesn't show ssh-daemon? <divya>Looks like you have some issue with pam, ssh-daemon needs pam <apoorv569>divya: no.. first time witnessing this.. I use same ssh config on my personal system. <apoorv569>does elogind adds pam? because I didn't had pam before.. <mange>The "segmentation fault" is concerning, for sure. <apoorv569>had to add elogind because of oci-container-service-type -> docker-service-type -> containerd-service-type <divya>Just doing openssh-service-type and includding the package and service modules should be fine. <divya>You have permit-root-login as #f? <apoorv569>also iptables -L -n -v errors out as well doesn't show my rules <apoorv569>sorry I can't copy paste the output because VM .. until I can ssh into the VM I can't copy paste <divya>Have you tried doing use-pam? #t <mange>Do things work if you build your config with "guix system vm" on your local machine? <divya>mange: Any clue with the package definition? <apoorv569>mange: haven't tried that.. can try though.. <mange>Just to isolate whether it's some state left over somewhere, or if it's actually a Guix config issue. <mange>No clue with the package definition. I started digging into the radare2 source, but it's build process is a bit convoluted and I couldn't figure it out in the time I had to focus on it, sorry. <apoorv569>divya: ‘use-pam?’ (default: ‘#t’) default is #t <apoorv569>ok its built.. same.. no ssh service in sudo herd status.. <apoorv569>my config is in 2 parts, base and per server config inherited from base.. <apoorv569>shouldn't be.. because I use similar base and per system config for personal systems as well and they work <vhns>are there broken tests pushed to master? It seems to be the case for me, can't reconfigure my system because tests in guix fail <podiki>reconfigure does not run tests on guix, or do you mean a particular package is failing tests? <apoorv569>is there anything that I'm doing that's wrong here? <meati>How can I help a cmake package to locate a dep that has no .pc or .cmake files? <mange>apoorv569: I had a quick look, and I can't see anything obviously wrong. I'll see if I can find some time to look in more detail later. <meaty>I tried searching for a solution, but all the search engines see is "cmake how to find a package" and I can't find an actual answer <meaty>The wierd thing is that I'm updating a package that seems to be able to sidestep it by setting -D$package_INCLUDE_DIRS but that doesn't work for the latest version <vhns>02:39 < podiki> reconfigure does not run tests on guix, or do you mean a particular package is failing tests? # Sorry, yes, that seems to be the case: https://0x0.st/8M9h.txt <vhns>uhhh wrong paste, one second <vhns>(I'll need more than a second as I have to reproduce the error, the logs got overwritten) <jA_cOp>Is there a good way to programmatically check if a Shepherd service is running? parsing the human-readable output of `herd status` seems hacky <vhns>jA_cOp: (shepherd) Interacting with Services <vhns>One example would be (service-running? SERVICE) <jA_cOp>vhns: thanks - I see that part of the manual now. But would I be able to get a service by name? Like if I'm just calling a Guile script from the terminal <jA_cOp>Specifically I'm using Sway with a status bar from eww, and I want to be able to start and stop certain home services from there. I assume I'd somehow have to connect to the shepherd socket? <jA_cOp>Using the `herd` CLI is fine for now, just not very machine-friendly <vhns>jA_cOp: just try it out directly through guile's repl `guile repl` `(use-modules (shepherd services))` `(service-running? foobar)` <vhns>you will need fibers installed, mind you <ieure>divya, Hard to say why exactly radare2 fails in the build jail and fails outside. Any conditionals in the directory doing the clone which would change whether it does? Is it downloading tooling with the network of the normal build env? <ieure>divya, Cool that you're working on radare2, that's one of the arcane footguns I'd like mastery over. I did use it a few years back to crack some 90s utility software for some expensive, niche test equipment. <divya>ieure: Not really, the guix build happens in /tmp anyways and the other build was inside a guix shell --pure, so no curl, no git or anything with which it can download. <divya>ieure: Yeah, I'm working on getting more RE tools into Guix. One of the reasons I was asking about Java yesterday was to get Ghidra packaged in Guix. <mange>divya: One thing I didn't think of earlier: can you try the build in a "guix shell --container"? <divya>If you do plan to hack around radare2 ieure , we might have you on board. Me and GNUtoo have been thinking of reverse engineering some of the network chip firmware blobs. <ieure>divya, Ghidra would be a big one, would love to see that. <divya>That's weird, why would it fail inside a --container? <mange>Does it fail with the same error? <mange>The same error as the build, that is. <divya>Not really, slightly different <mange>Can you be more descriptive than that? :) <vhns>04:20 < the_tubular> What's in there that requires fibers ? # Without fibers `scheme@(guile-user)> (use-modules (shepherd service)) <vhns>`While compiling expression: no code for module (fibers)' <vhns>It's also in the manual that they use fibers, I think <vhns>"Under the hood, each service record has an associated “fiber” (really: an actor) that encapsulates its state and serves user requests--a fiber is a lightweight execution thread (*note Service Internals::)." <mange>This isn't a clean build. It's not doing much building, because it's all already built I assume. This isn't really useful to debug the issue - you need to build from a clean checkout (or, even better, the result of "guix build -S $your-package", which is the source as Guix fetches it). <divya>mange: Ah, I copied th wrong thing. Let me do a clean build. Also, how can I build from the checkout of guix build? <divya>the_tubular: It sure would be, but before that we need to tackle the dragaon that is Java. <mange>"guix build -S $your-package" prints a path to the source in the store. You can copy it out of the store (and extract it, if necessary) then build it. <divya>Ah..is it advisable to go there and do a build? <mange>I think building from the sources from "guix build" is your best bet for reproducing the error when you build your package. <mange>apoorv569: I think I've found your problem. In your definition of base-server-services you close a list too early. You have (define-public base-server-services (cons* ...) %base-services). <divya>mange: Well, it says read-only file-system... <mange>You have to copy it out of the store, as I said. <mange>Everything in /gnu/store is read-only. <ieure>divya, If you put the subproject's (origin ) in your inputs, you should be able to copy it into place in the source tree, then make won't try to run the `git clone'. But it looks like you'll have to do this for several things. <divya>Good lord, who puts git inside a .mk file... <mange>It's more common than you'd think. <divya>Okay, so I should take these subprojects and package them separately and then include them as inputs to radare2? <ieure>divya, No, put origin records for them into the radare2 pacakge's inputs. <ieure>I'm not sure they can be built outside the radare2 source tree. <divya>Like a origin record _inside_ the list of inputs? okay <ieure>divya, Yeah, inputs can have either packages or origins. For origins, whatever the result of the download gets downloaded and you can futz around with its contents in your package build. <divya>I see, never really did this. Let me try for one and see if it works. <jA_cOp>vhns: thanks, I'll use herd as reference! I was just about to post that I was able to get the shepherd library working on the repl, but indeed the procedures take a struct representing the service <divya>ieure: Even after including the origin, it's still trying to fetch with git. Should I patch that .mk file and not have the git command? <ieure>divya, Did you move the provided input into the radare2 source tree where the makefile expects it? <divya>ieure: Umm, not really. I guess it expects it in the suprojects/ directory? <divya>Also how can I copy/link this directory for a guix build? <ieure>divya, Right, well, that's the important second half of my suggestion. Putting it in inputs downloads it to the store, how's the r2 weirdware Makefiles supposed to know that? <ieure>divya, You want a build step after 'unpack that copies or symlinks the store path into the r2 source tree. <divya>ieure: But how can I copy/symlink the store checkout? I mean how does one do it reproducibly? <ieure>Symlink preferred, not sure how to make one, I assume Guile has something for this. <divya>Hmm, also how do I get the patch of the store checkout? <ieure>The value in inputs the inputs alist is the store path. <ieure>Ugh, "The value in the inputs alist is the store path." <mange>Guile has a (symlink oldpath newpath) procedure that you can use to make the symlink. <divya>How can I access the patch of this particular origin record to use the in the phase <ieure>assoc-ref on the inputs, I'm not sure what the default key is in this case. I believe you can provide one to look it up as. There's some prior art in gnu/packages/ <divya>ieure: Well, I can't exactly find anything that has a separate origin inside inputs <divya>I meant that most of the inputs are direct packages, I can't find a definition that has a separate origin record as an input <mange>I think an origin used directly in an input list needs to be used with the old "named" input style, if you want to look it up. `(("capstone" ,capstone) ... ("my-origin" ,(origin ...))) <mange>Origins don't have names, so you have to use the legacy syntax to give them a name. <divya>Learning more guile-fu this morning :) <mange>This is more Guix-fu. These are the unfortunate layers of complexity that you get as a project goes on. :) Multiple subtly incompatible ways of doing the same thing. <mange>Oh, cool! I missed that, thanks for letting me know. <divya>jakef: So in that case, how can one label the origin? The discussino seems too long. <mange>You can still label things the way I said, but if you don't label things then the origin's file-name becomes its label. <divya>Let me try your way since I've already written the package definition with that, and see if the error goes away <mange>What that is depends on the origin you use, but you can also set it with (file-name "...") on the origin itself. <jakef>it means you don't have to use the deprecated style when you have origins as inputs <jakef>and you can refer to them by their auto generated name <divya>Weird, the upstream tree has subprojects directory, but not the checkout by guix <divya>Because it's funking around the files <mange>Time to try copy-recursively, from (guix build utils)? <divya>Okay I _think_ I figured out what's the issue. <jA_cOp>Anyone using Emacs rustic-mode? It seems to want the cargo clippy CLI to be installed, but this isn't packaged in Guix <yelninei>hello, after upgrading gnome the keyring no longer seems to provide the ssh agent <jA_cOp>I just tried it, and it seems not :( <Rutherther>jA_cOp: there definitely is rust-clippy in bin of rust:tools <jA_cOp>Well, `cargo clippy` reports "no such command" even after adding `rust:tools` to my home profile. I know there's a rust-clippy package, but it's #:skip-build? #t <jA_cOp>hm, I don't have that either, I must be doing something wrong... <mange>divya: That paste didn't work, so I don't know what your ld problem is. <jA_cOp>Yeah - I added "rust:tools" to the list of packages I install in guix home like (packages (map specification->package+output '("rust:tools" ... ))) but cargo-clippy is still not in $PATH. Maybe I'm missing something about how package outputs work, I haven't really looked into those <divya>I think it's an issue from radare2 <divya>Though it built fine when inside a guix shell <yelninei>jA_cOp specification->package+output is a bit weird as it returns multiple values, maybe specifications->packages would be better <Rutherther>jA_cOp: yeah, I am not sure how to add package's outputs to packages list, this what you have doesn't seem to be working right <mange>You need to compose list onto specification->package+output. I use the equivalent of (map (compose list specification->package+output) '("rust:tools" ...)) in my home config. <mange>The way to pick a specific output is to use something like (list rust "tools"), so passing the values from specification->package+output into list constructs the right thing. <yelninei>Re gnome-keyring: the ssh-agent is no longer enabled by default and behind a configure-flag. I changed out the gnome-keyring package in gnome-keyring-service-type and things work again <mange>divya: I don't know how to solve your ld problem. Does it build off a clean checkout in a --container? If so, are you sure you're running the same make target that the build is running? <jA_cOp>Thanks everyone! I tried divya's suggestion and it fixed it! <jA_cOp>Ah, just saw mange made the same suggestion, thanks! But this was all very good feedback :) <apoorv569>mange: Oh! so its missing %base-services then? whats in there that, it not being there could cause issue? <mange>Other way around, I think. It's missing your services, but has %base-services. I don't really understand why, but I haven't dug into it at all. <apoorv569>makes sense though.. iptables and fail2ban were also failing <mange>And why you needed to add a manual network service, instead of the dhcp-client-service-type providing networking for you. <mange>Okay, I just dug into it out of curiosity, and it's because (guix packages) shadows define-public in a way that puts its body into a (syntax-parameterize ...) form. <apoorv569>yea, I had a doubt when it asked me to add networking service.. but I though maybe dhcp-client is not a networking service.. <mange>syntax-parameterize is fancy macro stuff. It lets you pass information down the macro-expansion stack. In this case, it passes the source location of the package into the record itself, to help with error reporting and whatnot. <apoorv569>is elogin-service-type minimal enough? or is there another minimal alternative to it? <mange>That's not the problem, the problem is that it matches (define-public prototype body ...) and expands it to (syntax-parameterize (whatever) body ...). <Rutherther>elogind is not minimal for most people. But it depends on what you want to do <hanker>apoorv569, you can use seatd/greetd <hanker>but elogind fits reasonably well most use cases <mange>I think they need elogind for docker, which is a requirement for the shepherd service, or something? <mange>Oh, actually, seatd also provides a elogind shepherd service? I didn't know that! <apoorv569>I am installing a server environment, don't want any graphical stuff in there.. <divya>mange: Yeah I've asked in radare2's matrix chat. In a --container it doesn't work...because it'll need network to fetch over git. I copied the subprojects, but then also it's failing for some other reason. <apoorv569>divya: could you not use --network witthe container? <mange>The problem is getting it to work in the build environment. <divya>Guix build doesn't have that luxury <apoorv569>sorry I don't know what you are trying to do <divya>My hunch is that it's a bit of an issue in radare2 upstream <apoorv569>you want all deps to be availble in guix not rely on git pull or anything? <mange>No, it's just a package that uses the network during its build, and we're trying to cut off all the places where it does so in order to get the package building in Guix's build environment. <mange>But now it's failing for a seemingly unrelated reason at link time, with a duplicate symbol or something. <divya>apoorv569: Guix cannot do git pull while building <mange>I'm saying "we", but TBH I'm just providing commentary every few hours. divya is doing the real work. <apoorv569>oh while building.. like cmake or something trying to fetch some dep? <apoorv569>I have had a problem like that before I think.. <mange>We've sorted the network issues now, though. We're on to non-network failures, I think. <apoorv569>you have to package all the deps it tries to fetch separately and use them as input.. also need to substitute* cmake file to remove the fetch.. <divya>I think I'm close, even if I disable some configure flags, these issues go away and then there's some dynamic library issues. So yeah, probably non-Guix thing... <apoorv569>OK the config seems to work somewhat.. ssh is there now so I can at least ssh and copy/paste output <mange>Any logs in /var/log/messages? <mange>Or any clue in herd status $x for each of those services? <apoorv569>$ sudo cat /var/log/messages | grep docker just says this grep: (standard input): binary file matches <orahcio>Hello, Ihave error to start applications like nm-applet and tuba, I got this error `...libdconfsettings.so: undefined symbol: g_once_init_enter_pointer`. It is a guis issue or just some gtk configuration? <orahcio>Rutherther: I will reboot now and back again in a minute <attila_lendvai>i started to get "guix system: error: more than one target service of type 'special-files'"; is this something new? <mange>Oh, that is new! Have you included %base-services twice, somehow? Can you paste your current configuraion? <attila_lendvai>mange, my config is convoluted. let me check for %base-services... <attila_lendvai>mange, meanwhile: i do have one special-files-service-type instance in my services <mange>Oh, sorry attila_lendvai, my client coloured your name similarly to apoorv569 (who is working on their own issue) so I've mixed you up. That said, your issue might still be having two instances of %base-services or %desktop services. <orahcio>Rutherther: I am back after reboot, the error is the same <mange>special-files-service-type is added by Guix itself, so you shouldn't add it yourself. If you want a specific special file you use (extra-special-file "file-name" gexp) <Rutherther>orahcio okay and did you perhaps update only system, but not home/user profile? <orahcio>Rutherther: I will try a guix pull and reconfigure the system and the home profile, maybe I did a system reconfigure yesterday and today the home profile :) <attila_lendvai>mange, that was it, thanks! (it's a surprising constraint to me, but so be it) <mange>The problem is that service extension happens by type, so if you have two of the same service then you can't extend either of them (because it's ambiguous). Lots of services might want to create "special" files, so it's not a good idea to break that. :) <apoorv569>could it be that wireguard needs network-manager-service-type instead? <mange>I run wireguard with dhcp-client-service-type with no issues, so I don't think it's that. <apoorv569>in my personal config I do use network-manager-service-type <apoorv569>mange: Did you manage to get a look? if you are busy then no issues. <mange>I've had a bit of a look, but I couldn't see anything obvious, sorry. <apoorv569>because alot of the issues seems kernel related (maybe?) <mange>I use linux-libre on my VPSes/VMs and haven't had any issues. I do things which look similar to what you have (wireguard network, reverse proxy, iptables, docker containers, and fail2ban). <mange>Okay, I've got to go. Sorry apoorv569, I hope you find solutions to your problems! <apoorv569>could you share your config? perhaps I can find something in there? <simendsjo>I can no longer start libvirtd and get the error "could not find a usable firewall backend". Anyone seen this? <apoorv569>overlay and wireguard module are indeed missing <apteryx>ieure: by the way, I can reproduce this old bug#49940 with librewolf! <apoorv569>OK I added overlay wireguard ip_tables iptable_filter module but still docker wireguard and iptables won't work.. <apoorv569>I don't have a separate machine to test this.. <apteryx>hm, bumpy upgrade. Some glib apps stopped working after upgrading my system. I updated my user profile and reboot hung (shepherd, apparently). Then I rebooted in a previous generation, and couldn't login in gnome-shell because it was too new for the previous system generation, so had to login via tty and do 'guix package --roll-back'. I'm back! <apteryx>updated my system profile and rebooted * <apteryx>lilyp: did you have that glib broken for a bit for current applications after updating your system to the new GNOME? <apteryx>librewolf wouldn't run anymore for example <apteryx>lilyp: say, until you reboot into the new GNOME system <ieure>apteryx, Oh, interesting. Seems like an upstream thing if it's happening in both downstream forks? I will see if I can repro. <apteryx>ieure: maybe! I think it used to work, a very long time ago <apteryx>I doubt I could repro that in vanilla firefox, but it'd be good to try <ieure>Yeah. I don't run Firefox on anything anymore, but easy enough to try to repro on it. <apteryx>so yeah, everywhere, at least as built with Guix libraries <rekado>I don't know details but some update on master has broken gdal. <jaadu>I have been using guix package manager on a foreign distro and had the intent on switching to guix home as well. <jaadu>There are some issues with some programs with guix on a foreign distro, but I generally feel it works well. <jaadu>Do you have all your settings/ configurations in scheme? <jaadu>It seems a bit suboptimal to me. <jaadu>If you have a .bashrc, wouldn't it make more sense to write it in bash? <Rutherther>guix home services from guix channel support only files for bashrc that will be merged and put in place (~/.bashrc), so they aren't written in scheme at all <apteryx>ieure: Ctrl+Alt+R also crashes firefox as packaged by Nixpkgs <rekado>jaadu: I have plain .bashrc, .bash_profile, etc files that get placed in $HOME by "guix home" as they are. <ieure>apteryx, Half inclined to download plain Firefox on my work machine and see if it repros there. <apteryx>that would be a good test! from what I understand, nixpgks wraps picutils/lib in LD_LIBRARY_PATH on the firefox binary itself (wrapper), so it should be globally available to its child processes <jaadu>rekado: I already have a well structured dotfile repo that creates symlinks for everything I need. I don't fully want to swap over into guix and be completely dependent on guix in the future. <apteryx>(and that doesn't help with this crash) <JodiMcJodson>would anyone know why my system is building glibc-mes? it seems like it's some sort of bootstrapping process, but it hasnt done that before (only been using guix for a few days) and I can't imagine what I could've done to trigger it <apteryx>anyone else having Guix System hang at boot? Seems like it happened after nginx started in my case (I have lots of services though). <apteryx>It's probably Shepherd-related, I assume, but haven't dug yet. <apoorv569>do I have to use `modify-services` as mentioned in the docs? <ieure>apoorv569, You don't need the guix field / guix-for channels. <Rutherther>they do different thing. Channels will put them to /etc/guix/channels.scm, guix will change the guix instance used. If you don't change guix, still default guix with only guix channel is used <jas>hi! how do people deal with substitutes not being available after a fresh 'guix pull'? is it a good idea to use 'guix pull --commit=xyz' with some commit that is say 10 days old? does that help? <jas>ieure: how long does it usually take for the servers to catch up? <Rutherther>jas: something like that could help, but make sure to choose a commit that actually has substitutes. You can look through ci.guix.gnu.org / qa.guix.gnu.org to see which one that is. Not every commit has substitutes built, only the ones that are pushed as heads <ieure>jas, It's too unpredictable to, uh, predict. <Rutherther>apoorv569: I don't know. Depends on what you're doing. I would never recommend doing that as it's going to significantly slow down all your guix system builds <ieure>jas, You can use `guix weather' to see if they're available. <Rutherther>yes, even pinned. Because the guix derivation has to be computed every time. But yes, it will be a significant speedup compared to unpinned <jas>is there some way to automate this? i would like to pick a commit that is max 30 days old that have as much substitutes available as possible <FrenchNewbie>i've encoutered a problem i need your help to resolve, its the forst time i encounter this on linux <ieure>jas, I don't think so, you could maybe use `guix time-machine -- guix weather' inside a shell for loop, but this will be very slow. <FrenchNewbie>i ran a guix pull and got "no space left on device", after a bit of investigation i discovered i ran out of inodes on my device <ieure>FrenchNewbie, An inode is the on-disk structure containing metadata about files on disk. <jas>ieure: okay, it is still churning on my laptop so i'll see if I'm able to understand its output when it finishes... <apoorv569>I was writing this config for server use.. which I will use `guix deploy` for. <ieure>FrenchNewbie, If you ran out of inodes, that implies that you have a very large large number of very small files. The simple approach is to figure out what those are and delete them. <Rutherther>apoorv569: if you are going to use guix deploy for the target system, it doesn't matter what guix instance it has, so I don't see a point in configuring its guix service with channels <anemofilia>jas: I think the number of substitutes isn't really relevant for you, as I imagine you don't use most of the packages available on guix. What are the packages you use that are missing substitutes? <apoorv569>you mean guix deploy uses the local system's channels? <ieure>FrenchNewbie, You should delete some old generations and gc, then. <ieure>FrenchNewbie, On most Linux filesystems, changing the number of inodes available on disk requires reformatting the device. <Rutherther>apoorv569: well it uses the guix instance you are using to call guix deploy, like every other guix command does. (except time-machine where you can switch which version to use) I don't know what you mean by 'local system's channels' <apoorv569>local system meaning the system running guix deploy.. but I get what you mean. <Rutherther>jas: the green count isn't really informative as it depends on how many changes there were since last job <Rutherther>apoorv569: I know you mean the system running guix deploy, but it doesn't make sense to talk about it like that. The guix instance being used is important, and your system probably has many of those instances, so it's arbitrary saying 'current system's guix'. My guess would be that most of the time you're using "~/.config/guix/current/bin/guix" <ieure>FrenchNewbie, Probably not, it mostly won't do anything unless you also periodically delete older generations, as long as those exist, not much in the store will get collected. I wouldn't recommend automating this. <FrenchNewbie>is 'guix system delete-generation' similar to 'guix gc --delete-generation' ? <ieure>FrenchNewbie, Yes, but system generations instead of self-generations. <emil67>Hello! Thanks for Guix. I am a rather new user. I have a problem with git when trying to clone / pull a https repo. <emil67>I get the error `fatal: remote helper 'https' aborted session`. <emil67>- I have not nss-certs in system, since it's included in %base-packages I think. <emil67>- I have git installed in home.scm <Rutherther>emil67: please don't send so many lines, the bot even thinks you're spamming and deleted some of your last messages, so I see only the first point you did <emil67>Ok sorry. I am using irc for the first time. I run `git clone https:.../repo.git` <apteryx>interesting... passing --debug to udevd causes a hang in shepherd <apteryx>the debug messages are supposed to be printed to stderr; would that be a problem for shepherd? <emil32>I have a "fatal: remote helper 'https' aborted session", when running git clone on a https repo. nss-certs are installed system-wide, the env vars point to them. Git is installed in home.scm. Can someone help please? thanks! <Rutherther>emil32: it's definitely not a cert issue. My guess would be wrong url supplied, but as you didn't send it, I can't say <FrenchNewbie>is TLS error « write_to_session_record_port » common too ? <ieure>I don't know if it's a serverside or clientside issue, but the TLS stuff Guix uses breaks often. <emil32>Rutherther thanks. I am at work, I dont want to disclose company info. But I opened the gilab in web browser and clicked on the copy button. <ieure>emil32, Is it a repo that requires authentication to clone? <Rutherther>emil32: okay if you copied it, it should be fine. Though I would expect ssh usage for companies. But on the other hand I would expect gitlab to not give you https url if https wasn't supported protocol <ieure>emil32, That's likely related to your issue. Authenticated repos should really be cloned with SSH, not HTTPS. <emil32>Normally I always use ssh. It is a long story. I tried `guix shell -CP git`, still same issue. Funny think two days ago I did the same clone and it worked.. <orahcio>Rutherther: after a guix pull and reconfigure both system and home nm-applet and another packages finaly run, thank you <Rutherther>container shell doesn't have network, you would have to use --network, but better yet, don't use container for stuff like this at all <Rutherther>orahcio: great to hear, it sucks though, this dbus stuff can problematic if user and system get desynchronized libraries it seems <emil32>but it cannot be that I didnt install a package output like git:xxx, or that I need some other package installed like curl oor openssl? I tried installing many of them, did not solve the issue. <Rutherther>I doubt that, cert issues should look differently. Have you tried cloning something like a public repo from github? <FrenchNewbie>managed to get my reconfigure done, just had to retry multiple times as you said, i was starting to get worried i did something bad again <emil32>I think I messed something up. Because two days ago on fresh guix install I could clone that exact repo even. Thanks Rutherther and ieure, I will try troubleshooting myself. I mean I can add a temp ssh key to the repo, but I want to learn guix and why I does not work :) <wakyct>emil32, do you need to pass a username or pass to GH to clone a public repo? <emil32>I dont use GH I use plain old git <emil32>And I cannot clone a public https repo <ieure>emil32, Can you reach the site at all? `nc -vzw1 the.host 443' will tell you. Can you curl the URI you're trying to clone? <ieure>Is there a verbose / debug flag for git so you can better see what's going on? <ieure>Just some suggestions for things to look at. <Rutherther>"guix shell -CN git nss-certs curl -- git clone https://github.com/magit/magit.git" works for me fine. (reason for curl is to get the search paths for ca, it's not strictly necessary, I would have to export the CA env vars) <orahcio>Rutherther: yes, I will try to separate as maximum as possible, now `unable to connect to libvirt`I will see if I erase some configuration <emil32>ieure, yes I can (which package installs nc?), I have it in librewolf open. Rutherther for me this command does not work. maybe it is the network, you guys. I am running guix on Windows -> VirtualBox VM in NAT network <apteryx>anyone else stuck with a tiny looking emacs on a 200% scaled monitor with GNOME Wayland? <orahcio>Rutherther: I disabled libvirtd some days ago, I need just enable it :) <Rutherther>then your problem probably lays in wrong xwayland scaling <apteryx>oh, pgtk is actually the more modern gtk variant? <apteryx>any reason why this isn't enabled for our default emacs package? <ieure>Meaning Emacs doesn't talk to X11 at all, only uses GTK, and GTK handles the underlying protocol (X11 or Wayland). <emil32>github.com [140.82.121.4] 443 (https) open <emil32>(again ieure, even the default github https repos do not work) <ieure>emil32, Sure, just want to start from the bottom up to make sure you're troubleshooting at the right layer. <apteryx>Rutherther: works better now, thank you <emil32>okay, w/e i will find out eventually. I just wanted to say that i am really enjoying using guix, it is so much fun to use, thanks for making it. sorry that i came with a problem first, instead of before. good night! <meaty>Is there a way to get a look around the filesystem of the profile a package is built in? <ieure>meaty, Add a build step that fails and build it with `guix build -K', it'll keep the build dir and print its path. <lilyp>sneek: later tell apteryx I did have GIO modules related crashes that result in a one-liner immediately on program launch; if it's not that then no <Rutherther>meaty: could you be more specific? packages aren't built in profiles <meaty>so, I'm upgrading a package which, as part of its build process, tries to clone a couple repos of templates and put them adjacent to the source dir. It has functionality to avoid the clone if you already have the data (I think), but I'm trying to figure out how to put that data in the right place in the build dir (or in the build dir at all). What I've done so far is turn those template repos into locally-define packages with copy-build system <meaty>avogadro2libs specifically (after that, the actual avogadro2) <wakyct>OK I'm catching up on the thread at devel <apoorv569>I also face this same issue without deploy, meaning if I directly inside VM run guix system reconfigure same issue then as well <Rutherther>elogind is quite a heavy change to a system, so it's not so surprising and I wouldn't expect a fix soon, it's probably going to be quite hard to come up with a fix <apoorv569>how are people using docker on their headless servers then I wonder? no one ever faced this issue? <apoorv569>in any case.. is there a workaround until a fix arrives? <Rutherther>apoorv569: because most people either don't use elogind, or they use elogind from the start, so there is no time they would switch from no elogind to elogind <apoorv569>I am not modifying my system.. I am testing in a VM a new config I wrote. or are you saying elogind didn't had this issue before until recently? <Rutherther>apoorv569: yes there is a workaround, just once copy the system definition and channels definition to the server, time-machine and system reconfigure instead of deploy. After you've configured elogind it will be fine <Rutherther>apoorv569: I don't know what you mean by that, it being a vm or not doesn't matter, what matters is the guix deploy usage <Rutherther>and that you didn't have elogind and now want to go to a system that has elogind <Rutherther>orahcio: did you send the correct issue? that issues is about initial elogind reconfigure with deploy, so I don't see how it would have anything to do with libvirtd <apoorv569>Rutherther: I already mentioned I tried without guix deploy by running guix system reconfigure directly in VM same issue <apoorv569>I didn't say VM matters here.. you said most people use elogind from start and all.. so I was clarifying that I am not using elogind suddenly on my existing system or something and I am installing it on a new config which happens to be running in a VM. <Rutherther>apoorv569: I thought you were having exact same problem as in the issue that explicitly mentions they could reconfigure (and I also remember since the person was here asking about it) <Rutherther>apoorv569: what I meant is that you didn't guix system init with elogind, but you are trying to reconfigure from no-elogind to elogind <Rutherther>most people init their system to elogind, that's why they can't have issues like this <apoorv569>Rutherther: Yes that problem plus "<apoorv569> I also face this same issue without deploy, meaning if I directly inside VM run guix system reconfigure same issue then as well" <Rutherther>are you logged in as root or as non-root using sudo? <Rutherther>I would try logging out of all accounts except root and reconfigure with root <Rutherther>orahcio: re "how to confirm if I have the same problem?", so you're asking how to confirm you have the same issue as was reported in the issue you sent? <orahcio>Rutherther: yes, I want to confirm if I have the same issue, I have the problem, libvirt stopped after my last system reconfigure <anemofilia>wakyct: Did you manage to package the rust program? <orahcio>Rutherther: I could to confirm the issue here, I could see the log mensage on tty12 when I try to start libvirtd <orahcio>Hi simendsjo, I will try your solution, thanks <orahcio>nice simendsjo, now libvirtd is running again, thank you again <simendsjo>orahcio: Good to know the solution worked for you too. I have no idea exactly why though, or how the libvirtd service should be changed to avoid this. Hopefully someone will address the issue. <chio-chico-banea>2. the ./guix-install.sh script reports an unexpected syntax error at && on line 621 <chio-chico-banea>3. ~root/.config/guix/current/lib/systemd/system/guix-gc.timer doesn't exist, no matter how many times you replicate previous steps <apoorv569>OK, I give up.. no idea why iptables, wireguard and docker won't work in the VM. tried adding missing modules no luck. maybe its a vm issue, maybe config not sure. <Irvise>Arg, it seems someone pinged me a while ago but my backlog does not reach that far... So I got the notification but I cant see the message... <Irvise>Whomever it was, feel free to ping me again or PM me, sorry for that :/ <mange>apoorv569: I found the problem with your server definition. Your base-server-packages variable doesn't include %base-packages, so the system is missing things that it expects to have available (like modprobe). If I add that in, I can instantiate the machine and everything starts except iptables (you can't use "--dport 80,443", it needs to be two separate rules). <civodul>i’m getting “arm-trusted-firmware-rk3328: unbound variable” <civodul>whiel attempting to build firmware.scm <vagrantc>so, before i pushed that, i got it ... and then i re-ran "./bootstrap && ./configure && make -j5 check-channel-news" from a clean checkout and it worked. <vagrantc>i squinted very hard to try and figure out why else that might have occurred ... but feel free to poke me if it persists? <civodul>vagrantc: i found a problem in make-u-boot-rockchip-package: it refers to arm-trusted-firmware-rk* from the top-level <civodul>so this introduces the dreaded circular dependency <civodul>i’m checking but it seems that moving that reference within the ‘inputs’ field solves it <civodul>(because ‘inputs’ field as thunked, or delayed) <civodul>these issues can be hard to spot :-/ <vagrantc>help appreciated ... finally got around to reviewing and pushing some old commits ... and things seemed to check out fine running on my rockpro64-rk3399 (which was over 3 years out of date!!) <civodul>doing some more tests and will push the fixlet soon if everything goes well <vagrantc>the other aarch64 machine i had been using ... had a bad ram module ... which kind of killed me working on this stuff since december