IRC channel logs
2025-03-26.log
back to list of logs
<RavenJoad>Heh... I avoided that filesystem thing by just not updating for 4 months... Whoops. <Tadhgmister>is there a supported way to cross compile the linux kernel on guix using --target instead of --system? <Tadhgmister>`guix build --target=arm-linux-gnueabihf -e "(@ (gnu packages linux) linux-libre)"` doesn't seem to be working and I really thought I've either done it before or saw a reference to how to do it somewhere <ieure>running `guix system image xorg-test-image.scm' (xorg-test-image.scm contains the exact contents of the paste). <ieure>It fails to build, but doesn't have any error message saying why. <ieure>The build log indicates that mcopy is failing. <ieure>Honestly had no idea what mcopy was, then was the PATH below the error... mtools! Holy crap, I haven't used that since I installed Slackware on my 386! <Tadhgmister>if a kernel module is enabled into the kernel not as a module the initRD should have access to it right? <nomike>I'm currently trying to update the package definition for PrusaSlicer. It's the first tie I'm doing something like that, so I might use a few pointers in the right direction. <Tadhgmister>I'm getting an error in guix system that it is trying to locate usb-storage.ko and it is missing because it was compiled into the kernel for arm disks and I'm trying to establish whether I should be telling it to shush or compile it as a module <ieure>Tadhgmister, If usb-storage is compiled in, you need to remove it from the initrd-modules. <Tadhgmister>nomike, if you intend to eventually push your changes upstream you may consider cloning guix and editing it there instead of a home config <nomike>It compiles and starts., the main window appears and then I get a segfault. No big deal so far, I wanted to upgrade to the latest version anyway, maybe this is fixed there. But I've ran into some issues. <nomike>Tadhgmister, ah, yes, That sounds like a good idea. <Tadhgmister>nomike, does `prusa-libbgcode` and `libigl` need to be modified as part of your change? why is copying them also necessary? <nomike>Tadhgmister, probably it's not. Especially as I now clone guix and work on it directly instead of using a home-config. libbgcode I'm trying to update though as there are also new commits. Unfortunately prusa didn´t do any tagging or releases on libbgcode, which makes packaging it tricky. <Tadhgmister>I do highly recommend setting up to develop in guix and then run the `./pre-inst-env guix refresh --update libgli` etc to auto-update the packages to latest version as the first step to updating packages <Tadhgmister>it is likely you got somewhere close to that already and it definitely won't help with the segfault but development-wise being able to do guix refresh to automatically update code is lovely <hanker>are there any plans for ejabberd shepherd service on guix? <Tadhgmister>hanker probably not but assuming I can get guix on my router I will likely setup ejabberd at some point for it <Tadhgmister>also if you are on foreign distro you may need to setup build daemon or something for tests to run properly <nomike>FAIL: tests/build-utils.scm\nFAIL: tests/containers.scm\n <nomike>Basically just a few failed test cases I guess. But I wonder why there is no log. <Tadhgmister>yeah I wouldn't worry about that, you don't have the same build daemons on a foreign distro <Tadhgmister>not sure where the log is supposed to go so not sure how to help there <trevdev>Has anybody had much luck running VMs inside of Guix? <nomike>OK. I will ignore that for now. I expected that already. I just wanted to be a nice person and report a failing test es instructed ;-). <trevdev>Tadhgmister: I'll check it out, thanks <piethesailor>I am failing to remember.. did a recent 'guix pull' news point out that external harddrives are no longer in /media/<user>/? <ulfvonbelow>if xz-mesboot is persistently timing out due to silence, should I be increasing the silent time out a whole lot or should I be decreasing it a lot and trying more? <ieure>Ah, that's me. That's the patch that allows composing xorg configs. <efraim>sneek: later tell nckx your substitute server seems to be down <viperML>that's the json that repology scrapes, I was looking into adding other channels <Kabouik>I have the following error everytime I start a shell since I installed gvfs: https://envs.sh/ppX.txt What is causing it? It's not even clear to me what Flatpak has to do with starting my shell. <Kabouik>Actually the error above is true for many other programs, including all my web browsers (I tried 4): /gnu/store/1cnaan3nhlh4j7i2pm8z4j11bwmc963z-ungoogled-chromium-112.0.5615.165-1/lib/chromium: symbol lookup error: /gnu/store/6gw41r72q8zwb0jxlplicbzd2axn3f4z-gvfs-1.56.1/lib/gvfs/libgvfscommon.so: undefined symbol: g_once_init_enter_pointer <msavoritias>has there been any bug reports about shepherd completely locking up the system lately? I get an exception with fibers and shepherd/system-log that locks up the xmpp server, and the openssh server <msavoritias>sending link because i can't copy paste from novnc and i cant ssh into the server <lechner>msavoritias / recent shepherd versions can create some similar problems but i think most have been solved <msavoritias>hmm. I am doing a full system reconfigure atm lechner to maybe fix it with an update <lechner>msavoritias / did you pull recently? <lechner>i am thinking in weeks, so the describe won't do much <msavoritias>yep. crash was reported hour and a half ago. rebooted, 10 minutes later it crashed again. which is why i am trying the pull/reconfigure. but its only 17 commits new so i dont expect it to solve much <msavoritias>everything except bootstrap is deployed in that repo <lechner>i would get rid of the timer for now <msavoritias>lechner prosody the xmpp server went offline. message i got was Error: Connection failed: unknown POSIX error: timeout <lechner>what kind of crash, please? a kernel panic? <msavoritias>lechner all i see is that crash logs are empty saying localhost -- MARK -- <msavoritias>i can see if it says something now but once that crash appears in the picture the whole system locks up. can't ssh or login at all <lechner>msavoritias / are some services simply unresponsive. can you log in on a tty? what does the log on tty12 say? <msavoritias>no i can't login from the tty either. let me check the tty12 <lechner>msavoritias / i have seen your error before. i believe it is an error in your system configuration. can you post it somewhere? <msavoritias>except bootstrap all of it is deployed currently on the server. i deploy with sudo guix system reconfigure --load-path=. system.scm <lechner>i don't know what that does. i use guix deploy <msavoritias>i have a pending pr for guix deploy. it doesnt change much what exists there, but it does make things easier yeah. but then this whole thing happened :( <lechner>the "184" error looks like a failed assertion in the shepherd to me. civodul added some checks, and I have no further problems. i would disable services and see if you get a functioning system. it helps to think in terms of binary search, i.e. enable half of those that are disabled <lechner>by disable i mean strike from your configuration and reconfigure <lechner>any that are non-essential, like mcron <lechner>what is the purpose of --load-path=. please? that does not seem like a good idea <msavoritias>i have split the system config into modules. so that it is not a huge file. so the load path loads all the modules in that specific directory. its from the guix docs <lechner>also, can you figure out where the assertion 184 vs the bounds of 20 and 193 comes from? <msavoritias>i dont know much scheme yet so just following manuals and tutorials i see :) <lechner>that assertion is from the shepherd. i have previously only seen it when i misconfigured my timers <msavoritias>lechner how do i find out where the assertion and the bounds thing comes from? <lechner>msavoritias / i think you need an expert like civodul because something is parsing the log messages from prosody, i think, and fails due to an assertion in fibers. i am not sure what fibers does, but it sure appears in a lot of error messages <lechner>it looks like a bug in (shepherd service system-log) to me <msavoritias>ahhh right, it say about prosody. hope the service is not broken now. yeah good point <lechner>can you disable prosody without losing your cell phone? <lechner>this is more about logging, i think, but disabling prosody would remove the error by brute force <msavoritias>it would but we are an xmpp server. the whole point is to have prosody there. and ejabberd is not packages as a service so prosody is the only option <msavoritias>i could add the container instead and call it a day. if it crashes again that is probably my next step <lechner>i am not sure what a container would do <msavoritias>instead of running a prosody "native" service, maybe running a container could fix it for now. until shepherd is fixed. <K0nrad> Is there any introduction or something about setting up docker containers under guix -- where GUIX gets to manage their insides? <csantosb>K0nrad: what do you mean, what the internals of `guix pack` are ? <K0nrad>I'm interested in running a different piece of software that insists on a non-guix distribtion in a container. And I'd love to at least have the setup of the container "from the outside" managed by guix. <sham1>You could use the oci-service-type or whatever it's called, and then have guix handle the container engine part <K0nrad>Yes. And as those are still all totally new concepts to me, I'm asking for a pointer to some tutorial or itroduction or something :-) <sham1>Ask if you need some clarification, ofc :) <sham1>(It's slightly annoying that this cannot be used with podman yet, but oh well) <redacted>WARNING: (guile-user): imported module (guix build utils) overrides core binding `delete' <redacted>I'm trying to figure out if that's a bug in my guix channel, but I don't get that warning when building or loading my guix channel packages with things like `guix build -L ~/guix-channel <package>` <redacted>I could add #:select to the import and fix it, but I'd like to reproduce the problem before trying that. <redacted>Correction: reconfiguring shows the warning, not pulling <msavoritias>so removing unattended upgrades and the mcron did nothing. the problem does seem to be shepherd itself <msavoritias>so im going to downgrade the system to some point in january when as i remember this didn't happen and wait for shepherd to be fixed. <lechner>Rutherther / My missing service did not appear after guix pull and reboot <csantosb>Any idea if we have `getopt' binary somewhere ? How to locate it, other that installing packages locally ? <lilyp>csantosb: util-linux – currently none other than folk knowledge <lilyp>there has been a proposal to compile /bin directories into a knowledge base, so that you can search for things to install <lilyp>that would be a CI effort tho <csantosb>Thanks ! ... Yes, this or something else, would be great. <RavenJoad>Does Cuirass have a known bug right now? My cuirass-web instance fails because of '("connect" "~A" ("No such file or directory") (2))'. The cuirass service also claims an incorrect type argument in the default Guix channel entry in a specification. <RavenJoad>The only odd thing I see in the generated specifications file is that %default-channels is not expanded. Is that a potential problem? <ieure>civodul, Want to take a look at #77266 if you have a minute? <civodul>am i the only one being greeted that way, though? :-) <civodul>RavenJoad: hi! i think you just need to ‘herd restart cuirass-web’ <civodul>there’s a synchronization issue in the latest snapshot, meaning that cuirass-web can start before cuirass is ready to listen <civodul>this is fixed in ‘main’, thanks to socket activation <RavenJoad>civodul: I rebooted the machine (a VM in this case) after doing the guix deploy. <RavenJoad>Even after doing a herd stop & disable cuirass-web, and herd restart cuirass, herd enable & start cuirass-web, I get this error. This machine had Guix 5bf7d805 deployed to it today. <RavenJoad>And I pulled that Guix commit to my desktop on March 21. My desktop is the one that deployed to the builder machine. <janneke>civodul: yay! in what way(s) would this series affect guix system? <Rutherther>if service is in starting state, does it prevent shepherd from doing other stuff? I was under the impression that there is just one thread starting the services, which would mean that if a service blocks in its start function, it prevents everything else from happening. Is it like that? <Rutherther>I also have a question about the series, why is run-with-writable-store a function? <luca>`guix build -f guix-scm` <dckc>thanks! (I gather you meant guix.scm rather than guix-scm) <dckc>successfully built /gnu/store/f7pc71is15g0xrgkqmpjy13zpk0i2ar4-pumpkin-chat-0.1.0-git.drv <luca>Though to actually use it you probably want guix install or guix shell <dckc>yes, I started with `guix shell` <Tadhgmister>the guix package `linux-libre-arm-generic` doesn't include the config flags necessary to run guix, is that package intended for packaging non guix kernels for arm devices? <Tadhgmister>just trying to understand what the different kernels are for, `guix build --target=arm-linux-gnueabihf linux-libre` fails but the `-arm-generic` one succeeds, but when using `(kernel linux-libre-arm-generic)` in an OS definition `guix system build --target=arm-linux-gnueabihf ...` fails because it wants to put modules in the initrd that weren't <Kuchenzufriedenh>Are the build workers pluggable? Like can I make the workers spawn virtual machines instead of normal sandbox? <Rutherther>lechner: that is not true. I don't know how you got that interpretation from my reply. <Rutherther>you need to add (shepherd service timer) to modules of the shepherd-service <efraim>Tadhgmister: I think the linux-libre-arm-generic kernel config bit-rotted a bit <Tadhgmister>I can totally believe that, I'm just trying to figure out what should be there instead <Tadhgmister>guix specifies a list of extra config options which are applied and architecture specific config files in guix/gnu/packages/aux-files/linux-libre/ that are *not* specified for specific architecture kernels <Tadhgmister>I'm wondering if they just should be or if the `linux-libre` package should be fixed so it can successfully cross compile <Kuchenzufriedenh>ugh why does the webchat disconnect by itself. did anyone answer in the meantime? <Tadhgmister>what do you mean by getting build process to spawn VMs instead of sandboxes, like what would that even mean? <Tadhgmister>you mean like `guix copy` to copy build packages to another device? <Tadhgmister>or if the VM has guix there is a similar process to offload the build process to the host machine <Kuchenzufriedenh>I mean that guix had some template for builders, spins up a vm, builds in it, saves the artifacts outside <futurile>Kuchenzufriedenh: the builds are already built in an ephemeral builder that isn't connected to the Internet <Kuchenzufriedenh>futurile: yes, I noticed that. But putting it into VMs would make the isolation stronger. <Rutherther>Kuchenzufriedenh: guix itself doesn't really support doing that, but you can use the service virtual-build-machine-service-type if you're using guix system and offload to it <RavenJoad>What you're describing is not too far off from what a combination of the system tests do, the normal containerized build system, and hurd offloading do overall. <Kuchenzufriedenh>I am considering how much work it would be to port Qubes, as in the whole stack into one of the declarative OS' <Kuchenzufriedenh>Does Guix have features for deploying into static A/B images with boot integrity, android style? <Rutherther>Kuchenzufriedenh: could you clarify what 'deploying' into an image means? like building an image or something else? <Kuchenzufriedenh>Like something that generates an UKI, a basic rootfs, signs the parts with a self generated secureboot CA, + an efi application that will be put into ESP removable path, that provisions the keys into a firmware in setup mode <Tadhgmister>guix image definitely does most of that, not sure about certificates <Rutherther>currently guix channel doesn't support uki, doesn't support signing for securebot <Kuchenzufriedenh>this is going to be the easiest to get together from parts. from an immutable architecture perspective, uefi secureboot is hella complicated and the android bootloader concept would be way better architecture <Rutherther>I think there is a patch series for secureboot signing <Rutherther>lechner: that is not true. I don't know how you got that interpretation from my reply. <Rutherther>you need to add (shepherd service timer) to modules of the shepherd-service <RavenJoad>civodul: I learned from the Xen people that QEMU & KVM together can emulate Xen hosts. I would use this to give the xen-guest-utilities/-agent-service-types a system test. But the make-marionette function does not use KVM. Obviously KVM is provided by the kernel, so the xen tests can only run on a host kernel with KVM support. <RavenJoad>This is still a huge WIP, but how should I go about handling this case? <lechner>Rutherther / it wasn't from your reply. i thought you said it here. in any event, why do my timers work all work if that declaration is missing? <lechner>Hi, does 'herd status' use SSH even on localhost? I see some blocking behavior of it and SSH at the same time <Rutherther>lechner: as I stated in the e-mail, it seems to me that the already running shepherd's modules are somehow altered already as opposed to the ones used when it boots <lechner>Rutherther / i'm not sure. i have been using most of the timers ever since they came out <lechner>i just wish 'herd status' would never block, ever <Rutherther>Kuchenzufriedenh: could you clarify what 'deploying' into an image means? like building an image or something else? <civodul>RavenJoad: ‘make-marionette’ merely spawns QEMU (which uses KVM) <RavenJoad>I thought QEMU needed a flag like "-enable-kvm" or "-kvm" to have qemu-system actually use KVM. <lechner>Rutherther / okay, now i got it. thanks for the hint to import the module, and sorry i was so stubborn <RavenJoad>Realistically any build machine Guix uses will have KVM enabled (it is not much of an ask for CPUs to have virtualization extensions today). But I'd like this to be as portable as possible (however much of a bad idea that is). <civodul>RavenJoad: tests typically do (make-marionette #$vm), where #$vm is a script similar to what you get with ‘guix system vm’, and that includes ‘-enable-kvm’ <RavenJoad>I was looking at my make-marionette in (gnu build marionette) and kvm only shows up as a comment with a URL. So it is the generated script surrounding the marionnete that does -enable-kvm? <RavenJoad>Because I need to add to add (make-marionette (list #$vm "--accel" "xen-stuff")) to make the emulated CPU present itself as a Xen vCPU. <Rutherther>lechner: no worries. So now it works for you on boot? <untrusem>is there recording for guix social session by fishinthecalculator on 26? futurile? <civodul>RavenJoad: yes, it’s the generated script that has ‘-enable-kvm’; what you write above should work <meati>Hello guix, I am stumped on an issue where I've defined an in-tree definition for a Qt live-development tool and running the program with `guix shell -f quickshell.scm -- quickshell` runs it perfectly, but `guix install quickshell -L . && quickshell` causes the program to fail, reporting that QtQml.WorkerScript (part of qtdeclarative iiac) is not installed <meati>However qtdeclarative is named as a propagated input in the package def <meati>I've also tried installing qtdeclarative explicitly alongside it but no dice <RavenJoad>civodul: Are we assured that KVM will always exist on the targets we wish to run these tests on? Really I want to have a sort of feature flag to disable these tests on hosts which are incapable of emulating their vCPU IDs. <meati>Does Qt use some extra software to find its modules? Do (non-development) shells add some extra software that lets it work only then? <meati>Could it be that guix shell is clobbering some search path for Qt, and with it un-clobbered (when installed instead) it doesn't work? How would I check such a thing? <RavenJoad>Obviously I will have mess with the exact flags to use, since QEMU & KVM have a billion flags for Xen emulation between them, but I just don't want to introduce spurious erros for no reason. <RavenJoad>meati: Easiest thing I can think of off the top of my head to check environment variables is to just go into the shell and then do an "export" to look at all environment variables defined by the shell. <RavenJoad>Also, double-check that you have propagated Qt inputs to your package in the relevant environment variables (which may need to be done with wrap-program). <ryan77627>Hey all, I'm trying to define a package that has code I would like to build in a subdirectory of a git repository (specifically the C library for BLAKE3). Anyone have any references off hand on how to transform the package source properly to build within the "c" subdirectory specifically, since the Rust stuff is in the base repo? <meati>RavenJoad: Okay, it seems that the variables QMAKEPATH and QML_IMPORT_PATH are defined in the shell, but not in the post-install environment (I don't have to restart or re-log in to build the new environment, right?). Why would they be defined only in the shell and how can I get them set in the post-install environment? <ieure>ryan77627, Just making sure I understand, BLAKE3 is one project which has both Rust and C components, and you only want to build the C bits? <meati>RavenJoad: also, what does it mean to "propagate" an input "to" an environment variable? <RavenJoad>meati: In that case, I believe you will want to end up wrapping the main program(s) and libraries in a post-build phase that set the necessary environment variables to make that program run. <ryan77627>ieure: Yeah, at least I didn't see anything already made when checking the guix sources. I need the C bindings as a dependency for another package I am making. I am attempting to update nix since I noticed its a few years out of date now. <RavenJoad>In this case (wrap-program <built-prog> '(("ENV_VAR" ...))) See the documentation for the exact format. <RavenJoad>ryan77627: Just add a chdir phase after the unpack phase? <ieure>ryan77627, You want a build step after 'unpack which runs chdir. <ieure>I was thinking that there might be an argument you could give to gnu-build-system that does this, but I'm not seeing it. <ieure>Def. just chdir'ing into the right place will work fine. <ryan77627>Cool, will give that a shot. Didn't know if there was a specific thing I needed to do in terms of the source itself. Will report back, thanks <ieure>ryan77627, Nope. You'll over-download vs. what you actually end up needing, but nothing you can do about that. <RavenJoad>If you build the rust side of things too, you wouldn't technically over-download, but that seems to be a future issue. <ryan77627>True. Other reason I asked was to see if there was a convention I wasn't privy to since I plan to upstream this if I get it working. <ryan77627>RavenJoad: That's the funny thing, the rust version is already defined <ieure>ryan77627, Don't think so, but, get it working best you can and let that stuff shake out when someone looks at the patch. <lfam>ACTION reboots into linux-libre 6.14 with the rootless guix-daemon <ieure>ryan77627, You can find prior art for this in the xdelta package. <ryan77627>RavenJoad: Well, technically I am making a separate package at the moment. Maybe when I get it building I'll rework the package to follow the rust package's source to ensure the share the same files when downloaded. <ryan77627>i do not know if that actually matters or if guix detects the same source is being downloaded <ieure>ryan77627, The source is reused between any package definitions that use it. <RavenJoad>It's not terribly important because sources can be GC'd fairly freely. Don't worry about it. <lfam>ryan77627: If there is a file in /gnu/store that has the same hash, it will not be downloaded again <janneke>ACTION goes to rebase core-packages-team once more <ieure>janneke, Is there a good way to get eyes on a patch for the core-packages-team branch? Tag it team-core-packages maybe? <ieure>I sent #76798 a few weeks back and have another that I need to send in the next couple days. <janneke>ieure: pinging (like this) is always a good idea <ieure>janneke, It can be hard to tell who to ping, since IRC nicks aren't always obviously derivable from etc/teams.scm entries. <janneke>i've been intending to be very conservative with package updates to core-packages-team in order to first get a working gcc-14-based branch out <janneke>others have been starting to add all kinds of package updates, that imho has been starting to complicate matters <ieure>Maybe QA is working good enough now that I can push to topic branches. <janneke>hmm, it doesn't seem to be really down, just flaky/slow <lfam>janneke: You could tuink about rejecting changes to non-core packages. We don't want to recreate core-updates now that we have given ourselves permission to do teams branches <janneke>lfam: yeah, it's kinda difficult to communicate to *everyone* who [wants to] work[s] on it <lfam>There are plenty of teams that still need to be created :) <lfam>I guess it's a communication problem <janneke>and i don't want to be non-appreciative of nice work that people have done <janneke>and also, the branch lived a lot longer than i had hoped <lfam>janneke: Why do you think the branch has grown so old? <ieure>I wonder why xorg-configuration is embedded in a bunch of display-manager services, instead of having an xorg-server-service-type that takes xorg-configuration which the display-manager services extend. 