<lfam>It's weird, it works if I apply the patches "by hand" <pineapples>o/ I want to build emacs-next-pgtk from source to adjust configure-flags to my needs, but I'm unable to due to this re-appearing error during `guix build': "Server does not allow request for unadvertised object" <pineapples>Does anyone have an idea how to hot fix this package? <PotentialUser-47>Hey guix! I have some trouble with guix pull. My system lost power during a guix pull. Now I get errors all the time when I do a guix pull. "ice-9/boot-9.scm:1669:16: In procedure raise-exception: <PotentialUser-47>error: #{\x0;\x0;\x0;\x0;[...]". I already renamed ~/.cache/guix. Any ideas? <lfam>pineapples: It means the Git server won't allow you to fetch some random commit <lfam>You'll have to choose a different Git revision for the package <lfam>It's been discussed before, in the context of Guix and Savannah (it only seems to happen with Savannah) <lfam>There's a discussion somewhere on the mailing lists, maybe a bug <lfam>I don't believe it should cause anything to break, though. Git should fall-back to a full clone <lfam>So... you might share the full build log with us pineapples <pineapples>lfam: It's just stuck on that error message. That said, I'm currently compiling something else in the background so there may not be free guixbuild workers, I suppose <lfam>I think you might be mistaken <lfam>But, it's hard to help without seeing the build log <lfam>After that error message, it should fall back to cloning the entire Git repo <pineapples>Anyway, I'm recompiling the kernel, and its build process is taking up all the available CPU time, which I why I believe it was taking so long for the cloning process to begin because they could've not simply been free guixbuild workers available <KE0VVT>My Guix System is ancient. How do I update all the things? Tor is not working. <lfam>`guix pull && guix package -u . && sudo -E guix system reconfigure /etc/config.scm && reboot` <lfam>If you ever work as the root user, then do `guix pull && guix package -u .` as root as well <lfam>You can (sometimes) save some work by making sure to `guix pull` to the same commit for all users, using `guix pull --commit=...` <lfam>You can check which commit the first user built from with `guix describe` <valerii_leontiev>I've noticed that many packages in guix broken or works not perfect. Does guix community consider to add new repo for tested packages? Something like "testing" and "experimental" in Debian <KE0VVT>I'm probably just going to go back to CentOS for my home server. <lfam>valerii_leontiev: If something doesn't work, then it's a bug in the Guix packaging and should be fixed <lfam>The problem is basically one of humanpower, and creating another repo will make that problem worse <lfam>valerii_leontiev: And, it always helps to say what isn't working... <valerii_leontiev>I already said about it several times today. Even sent report to guix-bug <valerii_leontiev>I totally understand that it's free software. But model with second repo with tested packages would solve many problems <valerii_leontiev>For me personal I prefer use slightly outdated packages but with some guarantee that it works. <pkill9>there's not really enough contributers to make a testing repo <lfam>I mean, the packages are tested. They aren't tested enough in some cases. <lfam>The best way to improve this situation is to start contributing to Guix by sending patches that fix the problems <lfam>The policy of the Guix project is to not offer broken packages. If they are broken they will either be fixed or removed. <lfam>The reality is that we have a lot of people contributing new packages, and fewer people fixing bugs. It's hard to force people to do the boring work of maintenance <ngz>lfam: I think the skills involved in these two tasks are not exactly the same, tho. <vagrantc>and because of the way guix works, packages get rebuilt more often than in other distros <vagrantc>for example, i test all of the packages i update before uploading ... but if someone updates one of the inputs that the package uses, it might still build but not work ... <vagrantc>i might not test the package builds on all architectures, etc... <vagrantc>debian's workflow also hides some broken packages ... e.g. there are usually at least 300 or more packages in debian that fail to build, but built when they were uploaded ... <raghavgururajan>bdju: You are the one that had trouble loading python-based plugins in Profanity, correct? <KE0VVT>lfam: Human resources is a conundrum. I would get something like a "Red Hat Home Edition" subscription if it catered to user rights issues that are ignored by establishment. <lfam>ngz: Yeah, that's a good point. It's interesting to think about how to encourage and teach different kinds of contributions <lfam>I wonder what sort of standards we could apply to make the kind of thing that valerii_leontiev is asking about <bdju>raghavgururajan: I don't think so, but not sure. have only used profanity a little <bdju>can't remember if I tried plugins ever. mainly I use dino *raghavgururajan checks logs <lfam>KE0VVT: At least, to update the kernel. But other things may only be possible to "restart cleanly" by rebooting <lfam>You don't have to reboot, either. It depends on what you wanted to update <KE0VVT>lfam: I want to update Tor, so "torsocks wget" works. <lfam>Hopefully someone else can tell if that upgrade can take effect without rebooting <bdju>raghavgururajan: haha, thanks for confirming. it sorta rings a bell <raghavgururajan>bdju: Cool! Just wanted to let you know that it has been fixed shortly after. :-) <KE0VVT>lfam: If I reboot, I have to attach a monitor and keyboard to type in the disk passphrase. <lfam>Alright, then try reconfiguring and not rebooting <lfam>Maybe you'll need to restart some services, with `sudo herd restart ...` <KE0VVT>arescorpio: Same result with curl. <lfam>Ah, it's tricky to make python-3.8 not inherit from python-2.7. Currently, the python-3.8.2-guile-builder describes the build phases like this: <lfam>#:phases (modify-phases (modify-phases %standard-phases ...)) <lfam>So, when you remove the inheritance, the guile-builder changes, the derivation changes, and you end up having to rebuild from the bootstrap <lfam>I'd hoped it was possible to separate the packages without changing the derivation ***apteryx_ is now known as apteryx
<Rovanion>I'm trying to apply the patch mailed by Francoais here https://issues.guix.gnu.org/44178#23, pressing the download button I get a mbox-file that I think `git am` should be able to consume. But running `git am 44178-23.mbox` outputs "Patch format detection failed". `git apply 44178-23.mbox` does not work either, though that seems more expected. Any ideas? <Rovanion>Thanks, `git am --patch-format mailbox 44178-23.mbox` worked! <avalenn>Indeed mbox downloaded from debbugs contains a spurious Received line as the first line. <kondor>I'm curious, can you change UID, or GID of a file (say, in the user's directory) during the building of the `operating-system` (say using reconfigure, image, or vm-image) ? <kondor>... since I the system does not exist at the time of building <kondor>or, maybe i am mistaken. Since, more precisely, I am trying to change uid/gid from a service extending `activation-service` the system does exist and this happens at boot time, right? <leoprikler>reconfigure happens mostly inside the store, only special parts, e.g. grub, activation, etc venture outside <pkill9>no it wouldn't, i asked similar thing about multimc (minecraft launcher) <jetomit>right, thought so but wanted to check anyway :) <leoprikler>as far as minecraft alternatives go, there's always minetest ;) <dauf9i>Newer versions of rust-cargo-c require "rust-cargo" as cargo input. How can I use the "cargo" output from the rust package as cargo input for rust-cargo-c? <dauf9i>leoprikler: It works as input but not as cargo-input. <dauf9i>But it fails to compile if I set it as an input (error: no matching package named `cargo` found). <dauf9i>and I added use-module (gnu packages rust) ***DiffieHellman_ is now known as DiffieHellman
<leoprikler>dauf9i: hmm, it seems that the cargo tool is different from the cargo crate <leoprikler>you could try recursively importing that and making it work somehow, but I doubt whether that's a good idea <dauf9i>leoprikler: Yes, and the version numbers are also different (1.xx vs 0.xx). Should I create a rust-cargo package for the cargo crate? <leoprikler>I think you'll need (package-source rust) there. <leoprikler>that corresponds to nothing useful in that tree, but you may get 0.6.13 to run with it maybe <leoprikler>otoh what happens if you simply drop that line from Cargo.toml? ***rekado_ is now known as rekado
<mdevos>Could someone review <https://issues.guix.gnu.org/46216>, in particular the last patch in the series? It removes hardcoding of ports in tests/publish.scm (instead asking the OS for any port), and I would like to do something similar for tests/challenge.scm. <mdevos>(tests/challenge.log hangs in my modified tree) <benoitj>been running guix package manager for some time, and made the switch to GuixSD yesterday. been really good so far. not much issues and could reload working config when it would not boot <benoitj>my only problem so far is when I try to mount my luks encrypted /home on another device than my encrypted / <benoitj>I tried adding the (device-mapper, and the file-system as it should be, but cant make it work, it refuses to mount / and boot <benoitj>was looking for /etc/crypttab, but obviously, now I understand it does not exists. any hints on how to troubleshoot that problem? <leoprikler>Pardon me if I misunderstand something, but to summarize: On machine A everything works, but trying to mount the same partition on another machine B does not? <benoitj>ok, let me clarify: I installed guixSD on uuid matching /dev/sdb2, a luks encrypted device. I previously had another linux on this system and have the content of my /home in /dev/sda1, the other drive. today my config.scm only contains the setup for /, I can just cryptsetup luksOpen manually and mount /home. now I want to add it to my config.scm so that I dont have to do this. I added a (mapped-device (source (uuid <benoitj>"my sda 1 uuid")) (target "home") (type luks-device-mapping)). that seems to work and I dont have any warnings. but it complains on my file-systems /home telling me /dev/mapper/home does not exists. if I reboot, it fails and I end up in init prompt <leoprikler>ahh, because you're missing the other half of that <leoprikler>you need (file-system (device (file-system-label "home")) (mount-point "/home") ...) as well <benoitj>no matter what, I dont get prompted for the luks password <leoprikler>just to be sure could you paste your config.scm – at least the relevant sections? <benoitj>sure, trying something first. it's complaining about /run/cryptsetup <KE0VVT>I need to set up a torified SFTP server. <benoitj>telling me /run/cryptsetup is unusable, not a dir or missing. failed to aquire lock on /dev/sda1, and then /dev/sda1 is not a valid luks device. I will need to fix that /run/cryptsetup <benoitj>that /run/cryptsetup is always present when boot loading tries to decrypt it. it almost look like it tries to use /run/cryptsetup from the init drive, before / is mounted. <leoprikler>benoitj: perhaps try ordering them the other way round. Guix is sensitive w.r.t ordering sometimes <benoitj>leoprikler, ok, will try that. I'm also going to try similar setup on a vm to see the resulting config.scm <KE0VVT_>Guix is not a normal GNU/Linux system. I have to relearn everything. I can't be bothered with it right now. Going back to CentOS. <pretzel>(I'd like to patch one of the packages listed in my config.scm. If possible, the patch should be part of config.scm as well.) ***ece7 is now known as ece
<leoprikler>For simple patches, you can inherit the original recipe, add a phase after unpack and substitute* your way to profit. <leoprikler>if you really need a diff however, you better store that separately at a known location and refer to it via local-file <mdevos>(but if you really want to include it in config.scm, you can use plain-file) ***ece4 is now known as ece
<cybersyn>has anyone tried guix on a Beaglebone AI? i'm curious how it compares to the Beaglebone Black <mihi>cbaines, you wrote on the mailing list (i found it in web archives) that your Hurd building lock up sometimes? May I ask if they use a swap file/partition? Reason is that I have seen Hurd VMs lock up when they don't have any swap, and even adding a 50 MB swap file made these issues go away. <mihi>(as there is no mkswap in childhurds, you'd have to use "swapon -n" so it accepts files without signature) <cbaines>right, they are running without any swap, so adding some sounds like a great suggestion <cbaines>mihi, do you happen to have an easy way to do that? <mihi_>cbaines, I generally ssh into the VM and do "dd if=/dev/zero of=/swapfile bs=1M count=...; swapon -n /swapfile". Will not survive reboots, though. <cbaines>I've got all the configuration happening through the service, so ideally I'd be able to set out a swap file/partition that way <cbaines>I'm not quite sure how to have the image contain a partition though... <mihi_>so we have a chicken/egg problem. I'm unable to try the proper way as I never got any guix hurd reconfigured without substitutes, and without substitutes I cannot tell you how to persist a swap file..., so you cannot build substitutes :) <cbaines>mihi_, haha, substitutes are coming :) <cbaines>I think I see how the service goes about creating the image as well... ***mihi_ is now known as mihi
<carrothopper>was the bug ever fixed in the guixsd installer that would cause the luks full disk encryption to fail if the drive was over 1TB? <jackhill>Which of (sha256 …) or (hash (content-hash …)) is preffered for new packages? <lfam>carrothopper: Do you have a reference for that bug? Like, a bug number or previous discussion or something?