IRC channel logs
2024-07-16.log
back to list of logs
<Cessation>Reintalled guix on a new computer and was shocked to find that no /var/spool directory exists for my email setup to use. What do I need to get a /var/spool/ directory? <jaft`>I was going to go use ~local-host-aliases~ but see it's been marked as deprecated; what's been implemented to replace it? <bigbookofbug>how would i go about adding libgcc_s.so.1 to a container? it seems like the "gcc" package described here is no longer available, and gcc-toolchain does not contain libgcc_s <jaft`>bigbookofbug: I see libgcc_s.so.1 in one of my gcc-toolchain installs in my store (v14.1.0, at least). <bigbookofbug>hmm that's odd ... it doesnt seem to populate the emulated fhs for some reason <Cessation>ok. I just created the mbox files manually and that seems to have done the trick <Cessation>I'm always uneasy about creating files manually in guix. I'm worried that's the wrong way to do things. <weary-traveler>the guix manual mentions how in grafts the replacement package needs to have name and version that are the same length as the original package. ooc, does this get checked/verified somewhere? on a related note, where would it? <lfam>I would look at 'guix/grafts.scm' and 'guix/build/grafts.scm', possible some other files too <lfam>I haven't studied the code, so I can't say that this check is definitely invoked and its result respected. But that code does check the length of the replacement string <elpogo>efraim: Thanks for the video. Good to know that you can exclude the guix daemon and guix itself from the image. Were you able to get closer to the 480M number? <jungy>Hello. I am having issues with a GUIX system I newly set up. Cursor is missing after login, searching makes it sound like an xorg issue. Recommended I set 'SWCursor' to true, but I'm doing it wrong. I was wondering where I could see what the generated conf file looks like? <rickame>is guix's format built on s-expressions? <spenc>(which are built on top of S-expressions I guess) <spenc>(tbh dont fully understand it yet) <rickame>sorry, is guix's format built on s-expressions? <trevdev>I have my package manifest for guix home in a module. Every package is a symbol rather than a specification->package string. I want the flexibility to be able to use more explicit package references. This is going well, except that when a package isn't properly imported, or if there's some error in the syntax, I am not getting any sort of useful stacktrace. What's the best way to check a symbol is bound/defined properly in Guile so I <trevdev>can go write my own stupid stack trace? <rickame>omg yes ty spenc those are the kinda docs i was looking for <spenc>this isn't just guix's format, its all of scheme which is used in a bunch of different places, notably emacs <rickame>i'm wondering lately what's the best configuration file format for apps. even apps that aren't made in scheme/guile and don't even run on guix. i heard good things about lisp/scheme/guile, but also comparing with stuff like cue and pkl <spenc>for guix, I don't think they're going to change it anytime soon :p, but in general my opinion is that cue seems really neat, but if you have to parse the schema again inside your app (to make sure someone didn't mess with it) then it not having very many language bindings is an issue <spenc>since you'd be redoing a lot of work with the validation code <spenc>pkl seems less fancy and cool than cue but has more integrations <spenc>and both can be used as a cli if youre having trouble managing a ton of configs but i think you'd know if you had that problem <rickame>if guix was being rebuilt from the ground up you still think basing the config format on guile is best choice? <spenc>for running in the GNU system, where everything in the OS is configured with guile as one of its selling points, yes <vagrantc>rickame: given that a lot of the premise is "guile all the way down" ... i think it would be a totally different project <spenc>i think if you're trying to make an external project thats an easier to use package manager, guile definitley has a ligh learning curve that i havent crossed yet <rickame>sure but i'm saying if there was a fresh start over, is there any config lang or format that would be better? <spenc>not for this project, but if youre making an independant one <vagrantc>although i think there is a package definition format in json or something, if i recall correctly ... so stranger things could happen <spenc>yeah, if i understand correctly (probably do not) the derivations are a fixed format <rickame>like if you could magically have everything rewritten would config in any other format or lang be better than guile? <spenc>not for the GNU project, and there is a lot of value having it be a full language <vagrantc>rickame: probably need to define "better" ... better. <spenc>^^^, nix also has a full language for their config, often considered online less good than guile (though i dont know much about it, so take that internet opionion for what it is) <rickame>ya i know about nix. so what i'm wondering is if there's a: ? > guile > nix <vagrantc>guix could probably be (re)written in any language, and it would have advantages and disadvantages ... <vagrantc>rickame: you're also asking in a channel with a strong bias :) <vagrantc>i feel like i am one of the few who loves it despite guile, not because of it <spenc>rickame, if i understand correctly ( i almost certainly do not) , theres a derviation format thats relatively fixed, so anything in theory coud get it there <rickame>ya but that's an intermediate format, not what ppl use directly right? <spenc>(again, i probably dont know) that's what the guile code produces <spenc>and then the guix daemon turns that into packages <vagrantc>i think the .drv / derivations are what the daemon ingests ... the format largely inherited from nix <rickame>ya, i'm talking about what humans touch, the config files <spenc>so yes, i think you could use cue of pkl instead and (imo) it would probably be more clear for people who dont already know scheme (me) <vagrantc>(and the guix daemon as well, though there is work in progress to move to an all-guile implementation) <spenc>but that this project would and should never do that, and there are advantages to having both a ful language and one that allows a lot of interesting things with data-as-code <spenc>vagrantc, nice presentation btw <rickame>is guile typed? like can you restrict data to only a certain type or is it the full dynamic typing anything goes? <spenc>i mean, i have not sucessfully made a package definition yet, so not me <rickame>can you verify that a guile config file is correct without running it? <spenc>its not hard to run, its a short program that doesnt do much <spenc>it seems like you're fishing for people to say "yes! guile was a mistake, we should use a type safe language' which has huge rewrite-it-in-rust energy and its never going to happen with this project <rickame>i want a file format that i can validate against a schema without having to run it and see <spenc>in my opinion, writing a guix package is much more difficult than an npm one, and the type safety is not the reason why <rickame>what makes a guix package so hard to make? <spenc>type safety is nice for big projects where running the entire program exhaustively is hard / impossible <spenc>for guix all it would get you is better ide integration maybe, and also would require not using scheme which has a lot of advantages <spenc>imo: it's doing a lot more: npm knows your build system, doesn't apply patches, etc. <spenc>to me i guess really the comparison isnt npm, its a dockerfile, which wins on easyness because I just use the exact same commands ive always been using and are in the documention, and guix requires rewriting it using their libraries and functions instead <rickame>what makes the extra work of guile worth it over dockerfile easiness then? <spenc>docker images aren't reproducable in the same way, you can keep one around and keep using it but you dont keep the same chain of 'this source made this product' <OwOday>Hi, I'm trying to package a program thats currently in a git repository, it has a requirements.txt but the catch is it uses python3.8 , whats the sanest way to do this in guix? <nutcase>Hi Guix. Please, can someone send me a link to the recording of last week's Patch Review Session? <xelxebar>This should end with source pointing at target2, right? ln -s target source; ln -sf target2 source <xelxebar>The -f is effectively just making ln a no-op for me... Without -f, it correctly complains that source already exists. <xelxebar>Not sure why it's not showing up as a patch either <rhuijzer>I'm constantly getting a hash mismatch for guix core when running guix pull this morning <rhuijzer>Is there server outage again or should I file a bug report <b0ttymacb0tface>I have taken from the 64669 issue that the ddclient-service-type is broken and has been removed. <b0ttymacb0tface>(However, it is still mentioned in the manual which is lowkey confusing). <Oleand3r>Hello everyone, I installed gnome recently and text is missing from main titlebar in gnome-console. Is it the same for you? <civodul>Oleand3r: hi! you installed GNOME on Guix System, right? <civodul>i don’t use it personally but i’ve seen it work fine recently <civodul>guix refresh -l -e '(@ (gnu packages python-web) gunicorn-bootstrap)' <Oleand3r>the latter does not occur only with gnome. <civodul>Oleand3r: weird; can you reproduce it in ‘guix system vm’? if not, that would indicate a stateful issue <civodul>as for the shutdown issue, that doesn’t ring a bell <cbaines>civodul, QA prioritises branches, so it's been focusing on go-team, kde-team and core-updates recently and hasn't had time to process many patches <civodul>re core-updates, i guess the main issue at this point is i686, right? <cbaines>yeah, nss blocking things on i686 and armhf, I've been busy lately, but I don't think using datefudge was successful, and I believe the nss tests segfault with libfaketime <cbaines>I'll hopefully be able to take another look at core-updates later this week <civodul>hopefully i can take a look at the branch as well <RavenJoad>Has anyone tried to run Guix as a VM in XCP-ng with a UEFI boot setup? I am having an issue where once GRUB and co. load the initrd and the initrd is almost done, the virtual screen freezes. <civodul>RavenJoad: hi! does the VM have enough RAM to load the initrd etc.? <RavenJoad>sneek: later tell civodul The VM should have enough RAM, yes. I gave the machine 8GiB. The weird part is, I cloned the UEFI VM to test with BIOS too, and the BIOS one works fine. All I did was clone the UEFI VM, change the "Boot firmware" to BIOS, and reinstall. <efraim>elpogo: Unfortunately in the intervening time the image size grew to over 2GB, so I need to work on it again. Excluding guix and the daemon itself was an AHA moment for sure <jackhill>hi, I have an avahi-daemon service that fails to start. How can I troubleshoot? <zamfofex>Hmm, I’ve been reading backlog, and I found a (somewhat recent) conversation about “language package managers” (I assume things like Cargo and npm). I want to say: I don’t think package managers being language‐specific is very good, but I feel like communities (of programming languages) want to have a way to manage their programs without having to wait some downstream to package them. <zamfofex>The problem is that there is no unfied “source of packages” that people can simply publish their programs to, and people want their programs to be readily usable by others. (I don’t know how to solve this problem satisfyingly, it’s a bit social as well as technical, I think.) <ieure>Sure. npm or whatever wants to work on whatever platforms their users use, there is no unified system across Linux/Windows/macOS, so they make a thing that solves the problem for their language. Then everyone else also does that and Life is Pain. <ieure>Zero percent chance Microsoft or Apple will get their shit together on this, so that's just going to be... how it is... forever <ieure>Or until people stop using the platforms holding everyone back. Which also seems like it has a zero percent chance of happening. <dlowe>please contribute more guix import patches :) <zamfofex>Well, my point is that I don’t know what would be a solution to the issue I brought up. Even within “just Linux”, there are various package managers across distros. But people want to create a program (be it a library or a runnable program) and just publish it themselves for others to use, without having to wait for package managers to pick it up. <craigbro>guix, and nix, are not package managers, anf they are not in any way compliant with linux package standards, and never will be <ieure>Of course Guix is a package manager. <craigbro>and beyond just laying out files on a system, they are also build systems themselves <ieure>It's even in the manual: "Guix, as a package manager..." <craigbro>correct, I should have said, `not just...` <ente>I've had my kernel hang on guix when I tried to connect to wifi - something about hwrng missing entropy <ente>I've never seen that on any other linux. Is anyone else aware of this? <ente>like, I would click on the nm-applet (NetworkManager tray icon) and the whole system would more or less freeze <ieure>zamfofex, I don't think there is a solution, is the thing. People building language have to support platforms without a package management facility, so they will continue inventing their own. <craigbro>zamfofex: langauge dep managers are also dealing with the runtime composition of a system during development and build time, preparing deployment artifacts etc... in short, a much larger set of use cases and design goals than any package manager across OSs will ever provide <ente>and a much more specific use case at that <ente>dealing with how that language in question loads extensions etc <ente>fwiw FreeBSD does a really good job at integrating perl (even when installed through CPAN instead of ports) into its package infrastructure <ente>you can install packages via CPAN and remove them with "pkg delete" <ente>I've never seen that on anything else <ieure>ente, At least one Guix machine will freeze for a moment when I enable/disable WiFi, but it doesn't hang the system, just pauses for a moment during the connect/disconnect process. I don't love it, but it's a minor thing I haven't looked into. <zamfofex>ieure: Yeah, that’s what I mean. It seems what people already do is the best that can be achieved to solve the problem. Even if it has issues, it seems that it’s “good enough” for what people want. But note that it’s not only about “systems without package managers”, it’s also about “being agnostic to system‐specific package managers”. <zamfofex>E.g. Even if someone targets Linux specifically, when they write a Rust/JS/etc program, they don’t want to have to submit and wait for the maintainers of *every* package management system to accept their changes. So e.g. they just upload it to npm, and everyone can just use it immediately. <ieure>ente, Sounds like your system freezes hard and you have to power cycle to revive it? <ente>ieure: maybe it would have stopped hanging if I had had more patience. I waited a couple of minutes and then hard-reset it <ente>I was able to vt-switch and move the mouse still, just not click on anything and ctrl+alt+del had no effect <zamfofex>ente: What happens if you try to connect with nmtui from a tty? <ente>in that situation it would hang <ieure>ente, Gotcha, that's not the behavior I see. I forget if the box I'm thinking of has AX200 or 9370, but it's blobby Intel WiFi <ieure>I put an athk into the X60 Libreboot build I did. <ieure>Loved the formfactor, didn't love the slow performance and heat. <ente>I haven't really noticed any problems with it <ente>the intel card that was in there before was slower <ente>dunno about heat - the whole machine heats up <ieure>Yeah. The Core Duo in the X60 is the source of most of the heat, but the ath9k also got pretty warm, and it's right under the wrist rest on that machine. <ieure>I don't know how ath9k compares performance-wise to Intel, but it did get pretty warm. <ente>I guess I'll measure that <OwOday>anyone know why I get "&test-system-not-found" when I'm not trying to use a test system <ente>also the intel card was old. this is a 2012 era laptop (hp elitebook 8470p) and I think the old wifi card was one that only supported 802.11b/g <ente>I wanted 802.11n so I bought a new card <ente>at least iirc that was the reason <OwOday>wait it doesn't cache anything from previous attempts right <OwOday>ok, adding python-pytest and a test_null.py that asserts true got me to the next phase, I guess maybe pytest is required? or something? <ieure>OwOday, Builds are isolated, previous builds shouldn't have any impact on later ones. <OwOday>now I'm failing sanity check, which tracks <ieure>OwOday, It seems like the package has no tests? If that's the case, set #:tests? #f in your argument list. <ieure>Not really a Guile thing, it's a Guix packaging thing. <OwOday>paste.debian.net/1323471/ this is my failed sanity check, going to set tests to what you said now <spenc>zamfofex, yea exactly, as someone new to linux / system package managment this confused me a lot <evilsetg>Is there a difference between a guix repl and a guile repl beyond the %load-path? I am trying to run some guix code in a guile repl and it fails but works in a guix repl. I did set the load path for the normal guile repl to my current guix checkout and profile. <spenc>@ having upstream be just the code and the distro packagintg it eventually <ieure>OwOday, Not sure what's up with that, seems like a package dep missing from inputs? Looks like upstream doesn't actually have proper Python packaging at all, just a weirdware install shell script. <OwOday>Im trying to make a pyproject.toml for it <OwOday>it doesn't help this is my intro to guix <spenc>i think thats a lot of the appeal of flatpacks, you get the .exe style experience of, 'heres a file, run it if you want' which is quite different from the package manager 'see if its bundled, else write code yourself' <spenc>i was trying to get python2 the other day and they removed it in my distro cuz its old, so actually getting it became a huge hassle with the reccomendaton being to compile from source <ieure>A lot of the Flatpak / Snap / whatever stuff seems intent on replicating the "App Store" experience. I do not like the "App Store" experience, though. <spenc>yeah app store is kinda annoying, and having the source code and all the features of guix is so lovely <spenc>but there really is something to, as someone building a thing, being able to package it up yourself and be confident that someone else will actually be able to run it <OwOday>I mean its fine mostly but then it got all tangled up in build scripts and package managers <OwOday>yeah and if you break your system version, surprise apt had a macro it was depending on python for <OwOday>and I had the same problem with nix <spenc>ppl hate on npm but that thing actually works, hence why every project gets to have 10,000 dependencies. python's requirements.txt is so much lamer <ieure>OwOday, Python's package management stuff is *so* bad these days. Really hilariously awful. <spenc>yeah for sure, that was getting me too, one thing needed 2 and one 3 in the same build <ieure>I say this as someone who's put *a lot* of Python in production. I like the language. But holy crap the packaging story for it is terrible these days. <OwOday>its getting to the point its so bad I think making a new also bad alternative will be a valid idea in developers heads for a decade <spenc>the language itself i still find lovely <OwOday>yeah, I use jupyter even for quick dirty things and higher abstraction problems <OwOday>but it needs to stay the fuck away from systems <ieure>Still prefer a Lisp, but Python is about as good as non-Lisp gets. I like the progressive typing it's got these days. <spenc>thats kinda the dream of guix right, *everything* gets its own venv equivalent <OwOday>I dont even think docker can get this repos python3.8 setup right <spenc>i need to really learn a lisp soon <spenc>did the christine webber tutorial on guile and it was very cool <ieure>Scheme is pretty different, I have 15+ years of Emacs Lisp, 10+ years of Clojure, ~3 of Common Lisp. <spenc>gotta say: as a beginner there really is something to being able to look at a hunk of python code and at least be able to say 'this is an array, at least i know that' , looking at scheme i feel like until i know what every single function keyword does I know nothing <ieure>Mostly I find Guile to be kind of a crappy implementation of Scheme. <spenc>also, this is a little emabarrasing <spenc>but im trying to join the xwindows IRC on OFTC <spenc>wow irc is convoluted tho, compared to like slack <ieure>Yes, Slack is like a child's crayon drawing of IRC. <ieure>I do not think much of Slack. <spenc>i get that being an open protocol is very cool, enough to forgive everything else <spenc>and not being electron app is nice <ieure>I honestly like almost nothing about Slack. Just a miserable system that keeps getting worse. <ieure>Ability to directly upload images is about the only thing I like. Protocol is ludicrous, official client is horrible and continually getting worse, and their TOS forbids anyone from making a client that's, you know, good. <spenc>yeah the TOS of those things is horrendous <spenc>id like to root for matrix but its still pretty far behind on usability <ieure>I run a Matrix homeserver. It also sucks, but it's at least a more open flavor of sucking. <ieure>Most of the stuff I dislike about Matrix is that they want to be "Slack, but Open Source." And since I do not think much of Slack, I also don't think much of the Slack-like features which they keep adding. <ieure>I was using emacs-slack for a good while, it's also awful, but preferable to the official client for most things. It might be fixed again? It's a neverending cat and mouse game of Slack breaking shit for no good reason and a bunch of people trading patches in the comments of the emacs-slack repo, because it's almost entirely unmaintained. <spenc>yeah i hate that about closed protocols like that, no fun at all <spenc>and discord freaks me out, because theres no way it should be free. something sinister is happening there <ieure>All comes down to control. I'm of the belief that I know best how I want the software I use to work. Proprietary software is made by people who believe that I shouldn't worry my pretty little head about any of that, they know best. <OwOday>lol discord is serving way too many people <spenc>it really does get the new user UX so much better, no fussing with NickServ before youre allowed to join, search the full logs so you can jump into the convo right away, etc.. <ieure>Assume anything you write in Discord or Slack is fed directly into their LLMs. <ieure>spenc, Yeah, this part of IRC is not great. It used to be 1000x easier than closed stuff, but then spammers happen, the auth features to cope with it have been shoehorned into a thing that was never designed for it and all is not right. <spenc>yeah and they host the images theselves, i dont get how they survive <spenc>ieure, yeah :c spam prevention is one of those things big companies really do better, as someone who much prefers open platforms its sad <OwOday>idk mastodon communities moderate pretty well, about as well or better than reddit <ieure>Mastodon is a company now, too. Mastodon gGmbH. <OwOday>eh, activity pub is what I really mean <OwOday>half the servers I know are all akko or misskey or w/e anyway <OwOday>I've never seen any positive sentiment towards gagron either lol <dlowe>irccloud is about as easy to get into as discord and has some nice upgrades. It just costs a subscription. <spenc>is it everyone that pays, or just the server owner? <dlowe>everyone - it's a persistent front-end to irc. <ieure>I selfhost The Lounge, it's kind of similar. Worse, but free. <dlowe>so there's a mobile client with a battery life-friendly protocol that connects to their bouncer that connects to whereever <spenc>i messed around with rocketchat in sandstorm for a bit <ieure>I use ERC to get on IRC, been using it for uh... 20 years. <spenc>thats the problem with chat apps tho, gotta get other ppl to use them too :p <OwOday>yeah I just have a qube with thelounge start on launch <dlowe>these days I use irken, which I wrote myself as a challenge to keep the base app below 1024 sloc <dlowe>the plugins add a bunch of stuff though, so technically it's not that small <dlowe>huh, it's got nice buzzwords ;0 <dlowe>sorry, I was being earnest. It looks nice. <dlowe>looking forward to genode guix <spenc>they actually have their own homebuilt package manager, inspired by guix/nix <spenc>where its at right now, there's barely the concept of 'files', since everything is a an isolated component that talks to everything else client/server style <spenc>so it uses a store like guix, but standardizes to just one build system since genode cant run anything not manually ported anyways, so configuration is just a single list of what other packages it needs <spenc>oh wow - doing the nickserv captcha rn, these things have gotten funky <spenc>asking 'which images best match the mood' <ieure>I have never had to solve a nickserv captcha, that's bizarre. <dlowe>it's all doomed. We'll be relying on a web of trust system for humanity before long <spenc>im shocked this IRC is as nice as it is, I guess it isnt worth the effort to spam? since it doesnt give linkbacks like web indexed places have <spenc>pretty much only hang out in invite-only places with people i know IRL these days <freakingpenguin>Barring the occasional bad actor I only notice spam in discord-bridged channels. It's quite nice here in my experience. <spenc>yeah discord spam used to be pretty annoying, mstly from hacked accounts <spenc>makes me kinda bitter abt humanity, like, the odds of getting something out of posting the same shady link in every single channel gotta be so low <spenc>and you just made 5 minutes of work for every mod in every channel for every account <spenc>gotta be years of combined life lost to that sorta stuff <ieure>spenc, Widespread stuff like that gets handled by the IRCops, not channel ops. <spenc>when i was part of my uni's boardgame club chat that sorta thing used to happen every few weeks, it's been better about that recently <spenc>maybe just less people getting hacked idk <dlowe>hierarchical tiers of moderation seems like the way to go for now <dlowe>irc operates this way, and so does activitypub <spenc>kinda insterested how bluesky's going to do it <ieure>Infrared Cop sounds like a sci-fi b movie from 1986. <spenc>but with some elbow grease it can still be a scifi b movie from 2025! <dlowe>My take is that bluesky is going to outsource moderation to unpaid volunteers while still keeping control of the content, ala reddit <spenc>they did make a big deal about how your data is more portable, even than activitypub <spenc>but its a big company that runs the largest / kinda only pod and agregator, so i could totally see a bait n switch <spenc>it is kinda wild that c++ code doest list their depencneces in any way, seems like there'd be some standard format <spenc>maybe it could be guix manifests <spenc>trying to build https://gitlab.freedesktop.org/xorg/xserver.git rn, and it's constantly going back and forth getting a little farther and then saying it's missing something & then I have to look up whatever debian renamed it to, or it's there but too old a version, etcetc <spenc>oh thats a really good idea actually, <spenc>was trying to build someone's fork of it, gut guix shell would work perectly <ieure>spenc, C++ dates from the 1980s, the modern notion of "dependency management" didn't exist. <ieure>autotools is more or less how you handle C/C++ dependencies. <spenc>omg --with-source is so cool <podiki>hrm, anyone ever see this error when guix tries to download something (guix download or in guix build) "Bad non-negative-integer header component: 31536000" <podiki>while same file link is fine via browser or wget