IRC channel logs
2026-04-10.log
back to list of logs
<ieure>usernew, That's how Unix works, thoguh. <ieure>How it's worked for decades. <usernew>But my question was does emacs ever take for granted that some env variables are set and instead they wouldnt be if emacs is my login shell. ieure really? Other examples of this behaviour? <ieure>usernew, Literally everything! Every program you run after login can make use of the state set up by bash. That's what it's for. <bjc>it's not really true. graphical logins throw a wrench in the works <bjc>that's why systemd-environment exists <bjc>you have to jump through hoops for things like sway to work <bjc>this emacs thing is kind of nuts, but it's not that different from what happens when you use sddm to log in to wayland <Rutherther>it is, because sddm uses the login shell to start the wayland session, to obtain the environment variables <bjc>gdm, or greetd, or whatever, then <bjc>it's a standing problem that isn't universally solved <usernew>By the way I know that not having env variables set wont break emacs <postroutine>gabber, I finally found an old USB drive I can boot on with the APU2 and it work. The problem come from the SD card, after all. I use this USB drive to write my image on an m-SATA disk. Now, the PC Engines APU2 work on Guix System. But the CPU is very slow for a `guix pull` command. <postroutine>While updating the guix channel, in the indexing of objects, I got a lot of errors like this one: `[ 791.695734] perf: interrupt took too long (2529 > 2500), lowering kernel.perf_event_max_sample_rate to 79000`. <postroutine>At each message, the last number decrease. It's now at "39750" <gabber>why does invoking `rsync` raise "bash: line 1: rsync: command not found"? invoking the same command from within a `guix shell rsync` does not change a thing <gabber>postroutine: awesome! with "updating the guix channel" you mean a `guix pull` call? yes, these are quite work heavy—i tend to not do them on my weaker machines <gabber>we can update them remote through deploy commands, though <gabber>or just do it regularly (the first one may take ages) <dajole>Huh, I was just about to start digging into figuring out how to not have to enter the disk encryption password twice, but I got preempted by upstream :D <eikcaz>Is there a preferred way to include one repo as a subdirectory of another during a build? <eikcaz>Well, I suppose preferred would be to make it a git submodule upstream, but assuming I can't touch upstream <czan>In the past I've added the other repo as an origin to inputs, then copied it into the right place in a build phase. <eikcaz>Hmm. when I try to prepend an origin object to my inputs with modify-inputs, I get "In procedure string-index: Wrong type argument in position 1 (expecting string): #<procedure version ()>" <czan>Can you show your code? Are you doing something like (string-index version #\-) somewhere? (I can see that pattern a few places in Guix.) <czan>Ah, the issue is the (commit version) in your origin. There isn't a bound version name, so it's using a top-level procedure called version and failing. <eikcaz>oh lol, that's what I get for copying code so hap-hazardly. Lets see where that gets me... <czan>Also, since you're using gexps, instead of the assoc-ref and inputs you can use #$(this-package-input "foo") (you can set file-name for you origin to match the string), or even just #$origin to use the origin directly in the gexp and sidestep the input bit. <eikcaz>I tried that at first hoping gexp magic would save me, but I think I was stopped by that (commit version) issue you pointed out. Lets try good-ol (copy-recursively #$my-origin "sub/dir") <trev>we need a yasnippet for changelog commits for a package ([inputs], [phases], etc) <apteryx>trev: there's some automation of that by means of etc/committer.scm, but it only covers very basic situations. <trev>no way that i can grok that right now 😫 <apteryx>that's running 'make check' locally in a 'guix shell --pure -D nar-herder' environment <apteryx>my guess is that this has to do with cc1e9052785de0a08b05f85a7c7c400d95e339e7 and that trunk is tested against a newer guile-knots than I have available <apteryx>hm, but guile-knots in guix appears to be newer than that in guix-dev.scm of nar-herder, so it should work (at least the nar-herder guix package is using it and its test suite passes) <jlicht>many thanks to whoever fixed the type-your-password-twice grub/FDE dance <trev>it sounds too good to be true. is it one of the fixes where i may or may not blow away my partitions? <trev>woah i found it. that's huge...this probably stopped like 60% of people from using guix :D <jlicht>trev: I was so confused, walking up to my still-booting machine but finding it at the login screen. Started doubting my memories before reading up on the recent changes :) <postroutine>gabber, yes, with the `guix pull` command. I run it in a screen session while I was sleeping. <postroutine>> we can update them remote through deploy commands, though <postroutine>I have read the `guix deploy` documentation. But it need to set a user account with the right to run sudo with no password. And I'm not a fan of it. <nox>what's the reason behind the (opinionated) UI choice of making the verbose flag `-v3` insteead of `-vvv`? <trev>off the cuff guess: how getopt-long works <cbaines>apteryx, I think locally I'd removed that pinned guile-knots from guix-dev.scm, but just not pushed that change, so yeah, testing with the newer guile-knots from Guix should hopefully work <futurile>parra: nice. Someone else is working on CI images these days as well. <nox>parra, interesting. perfect would be Codeberg. <parra>nox: does it have GitHub actions for free and as powerful as GitHub? <parra>futurile: people always thinks my project is new or something but it's the one that most time have been developed and one of the first that appeared and it's being used in production <parra>you can see all releases here: <futurile>parra: yeah, I did find it when I first started using Guix, at one point I was going to do it containerized. I didn't quite get around to it heh :-) <futurile>parra: I only mentioned it because there was a discussion on guix-devel and maybe there was a possibility of people collaborating so we could do a "Guix"-blessed release. <futurile>parra: guix-devel goes past so fast I don't always read it - so I don't know if you were in that discussion - apologies if I'm repeating something you already know <parra>what is Guix blessed release? <nox>approved by the Guix project <futurile>parra: trying to have containers as an approved "download" in the Guix project as nox said <parra>but respect to my project? or other? <futurile>parra: the discussion on guix-devel was looking for people to build a "team" and work together - as I understood it <parra>they could just use my repo xD <parra>how can I send an email to that mailing list? <ieure>parra, You just email it. guix-devel@gnu.org <parra>I have to use the same title of the list and that's all? <ieure>parra, Do you want to reply to an existing thread? <parra>or would you publish a new topic? <ieure>parra, Using a related subject won't thread with the other messages on that topic, because it won't have an In-Reply-To header. You could find the message ID and set that. <parra>I don't understand what you mean, I've never used mailing lists xD <parra>in any case I don't know how to contact guix devs, if they are interested they could fork my work <parra>maybe they don't need to reimplement everything from scratch again <parra>which has happened multiple times absurdily <avigatori>Recently I've experienced a failure mode and I am not yet sure how I trigger it: I symlink my channels to $HOME/.config/guix/channels.scm. I let guix home manage the symlink. However, sometimes I seem to manage to delete the symlink, after which I can't update my system or home configuration until I figure out again that I am missing a few channels. <avigatori>The strange thing, to me, is that I don't recall deleting anything in my home profile yet somehow keep losing the symlink from time to time <ieure>avigatori, Where does the symlink point? <avigatori>oh, I thought I had sent an intermediate message. I guess I am more confused than I thought ^^ <avigatori>You made me look again. So I saw that the symlink was still the one I manually created and not one pointing at the store or $HOME/.guix-home/... So I reran `guix home reconfigure` and saw that it was telling me that it couldn't find the directory where I told it the channel.scm file was. I had tried to use a relative path. I replaced the path with an absolute path and now the symlink was created correctly <ieure>avigatori, Gotcha. Why not use home-channels-service-type, though? <ieure>avigatori, Well... now you do! I personally quit specifying channels in my user configuration, I put it all in the system configuration. Feels weird to me to have the operating-system have the substitute configuration for channels only defined in my user. So I put everything in the operating-system now. <ieure>*everything channel and substitute related <avigatori>yeah that makes sense too. I think you also have moved your channel authorizations into it's own module IIRC. That was also something I wanted to steal <ieure>Yeah, I stick all the keys into their own module. <n|Phreak>ha ha , I just ran guix system from msys2 on windows using remote-viewer using spice <n|Phreak>why ... cause I was looking into a way to use guix containers for testing stuff , but I don't think it matters how much ram I throw at qemu its just not going to perform how it want it to. <n|Phreak>Maybe there is a better way to install qemu on window instead of using msys2 ? <redacted>I'm trying to add flatpaks to my global PATH. It looks like I should do this with home-environment-variables-service-type, but I'm unsure how I should extend the original alist of env vars. <redacted>Ah, maybe I need to consult service extensions <n|Phreak>I just add flatpak to my packages then download whatever package and create an alias in .bashrc <n|Phreak>for instance alias seafile='flatpak run com.seafile.Client&' <redacted>There's workarounds like that, but I'd like to know how to do this correctly. It looks like home-environment-variables-service-type does work, but I get warnings about duplicate definitions. <redacted>Well, I think the offending values were variables Guix sets correctly anyway, so it's working now. <avigatori>tl;dr (kde) discover can't install flatpaks without the user manually installing the flatpak package to their profile. <avigatori>The flatpak dependency is marked "optional" in the discover package source, however, I am not sure what else one would use discover for. It seem pretty useless without this option since one can't controll guix with discover <ham5urg>I tried a iso creation but it failed with grub unable to compile due to the testing phase "SKIP tpm2_key_protector_test". Can I switch off all tests for all packages? <avigatori>I guess I am looking for the opionion if the behaviour is intentionally this way (and why) <ieure>ham5urg, You *can* switch off tests, but you really do not want to. <ieure>ham5urg, Running tests or not is part of the package definition, changing that changes the derivation, upshot here is that you'd have to compile everything locally. <ieure>ham5urg, Want to put config.scm and channels.scm in a pastebin to see if someone can reproduce your error? What architecture is this? <ham5urg>ieure, it's amd64. Just a second, I gonna paste it now <ieure>ham5urg, Attempting to reproduce. <ham5urg>ieure, now it went through. Idk why. <ieure>ham5urg, The error you got is something you'd see if you were building locally. Maybe a substitute became available in between your runs. <ieure>ham5urg, What was building that produced the error? <ieure>ham5urg, Things are still running for me, it seems like it will complete successfully. <ieure>ham5urg, Everything built for me. How is this "for an Intel Core x7 CPU?" It looks like the standard nonguix install OS. <ham5urg>ieure, at the moment I can't boot with the official iso. This is my try to get an Intel X7 CPU to boot. <ham5urg>yes, the x7 went through with the new image <ieure>ham5urg, What have you done that would make it work on that CPU, though? <ieure>ham5urg, Okay. Well, I would not assume that will work better. <ham5urg>ieure, at least I could finish the graphical installer menu, but now it hangs <dlmsr>Hi everyone! Is anyone successfully using the redis-service-type? <PotentialUser-13>Have there been any statement made on age verification in Guix? I wasn't able to find anything on this issue <ieure>PotentialUser-13, There has not been any statement. I believe we are simply ignoring it and continuing to do what we do. <ieure>Ignoring it until it goes away. <ham5urg>I bet it will get embedded into the firmware of any device at some point in the future