IRC channel logs
2022-11-07.log
back to list of logs
<civodul>lfam: Pjotr and Manolis are trying to get the devroom going <civodul>lfam: and the Guix Days at ICAB beforehand <civodul>i guess you can get in touch with them to show your support :-) <lfam>Ah, I didn't know if there was a plan for Guix Days at ICAB <lfam>I am going to try to be there <the_tubular>(use-package-modules linux admin bash certs compression curl emacs emacs-xyz guile-xyz guile gawk gnupg docker less file-systems man ntp ssh version-control rsync python util-linux-with-udev) <rekado>civodul: thanks for liberating goggles-bot from the confines of my screen session! <the_tubular>I don't get it ... I'm importing modules and guix is crying that they are still not found ... <podiki[m]>that is needed for the docker service, and add that to your services declaration <podiki[m]>with (service docker-service-type) and make sure your user is in the docker group (add to user-account's supplementary-groups) <kozo[m]>Hey all, does anyone know what package has libgcc_s.so.1? Thanks! <the_tubular>Both have been added, then it requires elogind I added that, then it required `(use-service-modules desktop)' <the_tubular>Yet when i run reconfigure I get : error: module (gnu services elogind) not found hint: Try adding `(use-service-modules desktop)'. <kozo[m]>nckx: cbaines Hey guys, it looks like podiki got banned from the channel for posting a github link? Are one of you perhaps an admin for this channel? <the_tubular>There has been a few bots here in the last few weeks though ... So I get it, but there needs to be more mods, even if you want to automate <lfam>Does linking to this really cause you to get banned from the channel? <lizog>what would be a good way to install guix in a non-root environment? assuming the non-root user have kernel namespace support <daviid>the_tubular: <the-bot-nick> later tell nick this and this ... <daviid>sneek: later tell the_tubular you see ... <apteryx>I've asked about why podiki[m] was erroneously banned by their bot in #liberachat <apteryx>our own litharge config may be at fault; "maybe there's a ($nick isin url) filter?" nckx? <apteryx>nckx: "/t\.me\//" *] in our litharge lspattern would be at fault <apteryx>nckx: unfortunately it seems I need more op to dig further in our litharge config or change it; I'll leave it to you <apteryx>in the meantime, I'd advise people to not post links containing their own nick :-) <sneek>Welcome back the_tubular, you have 1 message! <sneek>the_tubular, daviid says: you see ... <the_tubular>sneek: later tell podiki[m] Your sacrifice was appreciated. <robin>sneek, later tell robin to check if you understand any commands other than 'later tell' and 'botsnack'... <sneek>Sneeky bot running on Guile version 3.0.3 using bobot++ release.2.3.1-6-a463-dirty <nckx>apteryx: What's wrong with posting links containing your own nick? <nckx>sneek: later tell podiki[m]: Sorry 'bout that zap, the leading '/' got dropped (well, unescaped) in a migration to less permanent spambans. Have a botsnack to recompense. <nckx>Wow. You should be ashamed. <nckx>sneek: later tell podiki[m]: Sorry 'bout that zap, the leading '/' got dropped (well, unescaped) in a migration to less permanent spambans. I tried to give you a bot snack but sneek stole it. <xd1le>hahaha just stole it, ate it and said no :) <civodul>bad news: shepherd still seems to leak on bayfront :-/ <mothacehe>Hey civodul! Oh too bad :( Is it leaking at the same pace? <civodul>i mean, i did see a significant different with my synthetic workloads <civodul>problem is, debugging this is extremely expensive in terms of human labor <civodul>i have a screen session where i run that regularly <fiesh>the two hardest issues to debug are race conditions, resource leaks, and off by one errors... <mothacehe>it seems like a huge commitment indeed. let us know if we can help somehow :) <civodul>i'm interested in fresh ideas and reproducers at this point <civodul>for me they no longer exhibit a leak <ph03n1xaim[m]>I would be grateful if you could provide me any help. I have used emacs for long and this is a working config from another distro. <mothacehe>ok i can reconfigure berlin to compare with bayfront <mothacehe>if it can cheer you up i'm closing in on the finalizer issue! <minima>which doesn't seem to be packaged under guix <minima>i wonder how one would go at packaging something that's part of a broader repo... <mothacehe>civodul: berlin is now reconfigured, shall we restart it to install the new shepherd? <minima>it does build (yay!) but it doesn't seem to work - i.e. emacs doesn't seem to find erc-sasl <minima>it might be because of a wrong path for erc-sasl.el <minima>it ends up being saved at /gnu/store/cn0aq0psybzs14y164vk29si7phxvnk8-emacs-erc-sasl-0.200.14/share/emacs/site-lisp/erc-sasl-0.200.14/layers/+chat/erc/local/erc-sasl/erc-sasl.el <stevenroose>Is there a way to do the powertop service like for systemd in Guix? <apteryx>civodul: looking at the leak problem from a bird's view; how that even possible in Guile? Is the GC loosing track of objects? Or is the program (shepherd) at fault, continuously creating new objects that are referenced? <apteryx>I've read your reports, it seems like it's the former (some C code managing stacks is leaking?) <civodul>apteryx: right, it's code that deals with delimited continuations in libguile that's leaking stacks <civodul>where leaking means "unduly keeping a reference to previously-allocated memory" <civodul>(it's not a leak in the C/malloc sense) <apteryx>have you made a report against Guile? it should be fixed at the core <apteryx>I don't see something related to that memory leak in Debbugs for Guile <civodul>apteryx: i had it "fixed" in Guile, but apparently it's not really fixed <apteryx>seems someone on fedora was not able to reproduce #59021, based on replies in #guile? That's interesting, if true. <apteryx>it means it could perhaps be related to the Guix build of Guile? <apteryx>I can try that today (using VMs) if that's useful. So I just let the snippet from #59021 run an check its memory growth over time? <f3n1x>hi Guixers ! ... i've just declared 'guix install ripgrep' ... package derivation is downloaded and (apparently) installedt. Then i run '$ ripgrep' ... and it says 'command not found' <f3n1x>Uh. i'm puzzled... what am i missing ? Thanks thanks thanks for any kind of help and/or direction ! <nckx>f3n1x: Is it not called ???rg???? <nckx>???guix build ripgrep??? confirms that it's called ???rg???. <f3n1x>ah.oh. yes , it is 'rg' . lol <nckx>Apparently life's too short for vowels and the majority of consonants ???? <f3n1x>on the other hand, (i3wm here) after switching to a new machine and its subseqent fresh OS install, i'm puzzled by the fact that 'emacsclient' is not launching it as expected, what could i be missing here ? <nckx>I don't know, I use it myself without issue. <f3n1x>fyi, running it on the terminal leads to ' '/home/fenix/.guix-home/profile/bin/emacsclient: file name or argument required' <f3n1x>ACTION is trying to double check wether if the daemon is running or not ! <nckx>Sure, but there's a second line ??? Try '/home/nckx/.guix-profile/bin/emacsclient --help' for more information <nckx>If you're new to emacsclient, I'd read the ???Invoking emacsclient??? part of the Emacs manual, at least. This is not Guix-specific. <nckx>If ???emacsclient -c??? still fails, the daemon isn't running. <f3n1x>'emacsclient -c' works like a charm <nckx>I've aliased it to ???emacs??? (through a one-line script, so it works when called from other tools, but it's still just an alias for ???emacsclient -c???). <nckx>But if you don't have it bound to Super+E, are you even living. <f3n1x>Let's have fun! ah... in your 'modus operandi', calling 'emacs' means that the first time (whoever does it: me or the system itself) 'emacs' is called it runs 'as emacs' and the second (and the 3rd, the 4th...etc) it is called it runs as emacsclient 'automagically' ? If that's the case... i'll be following your path. Indeed <apteryx>civodul: on my local machine it seems to stabilize after a bit; it's been at heap size has been at 87846912 for a long while in one process <apteryx>in a Debian 10 VM using guile-2.2, it increases only once at the beginning; currently stable at 8990720 <civodul>apteryx: to be clear, which one are you running? <civodul>so on Guile 3.0 from Guix, i typically see a 10x growth after between the first a second "heap-size" line <civodul>(*from Guix before commit 7138ba34fa11d3eb2d9e7e44771e698f53415dbc) <apteryx>yeah, so far i can't say I reproduce; 2 guix systems with vastly different x86_64 CPU (core 2 duo and ryzen); both stabilize at different points <apteryx>have you tried that script on berlin? <civodul>guile-2.2 from Debian is very different from what we're targeting though <civodul>i haven't tried that script on berlin, but i did try lots of other things <civodul>and we can't deny that shepherd leaks on berlin and bayfront :-) <apteryx>I'm just wondering if we've got the correct problem cornered <civodul>yeah these things are always hard in part because it's not deterministic <civodul>and while you can demonstrate that there is a leak, you cannot prove lack thereof <apteryx>it's been stable on my desktop for more than 15 minutes <f3n1x>/me tries to mimic nckw way of invoking emacs... <f3n1x>ACTION tries to mimic nckw way of invoking emacs... <civodul>apteryx: with Guile from Guix, right? which store item? <f3n1x>nckx ... i would love to implement it, too. how do you do the Super+E you mentioned ? Thanks, thanks, thanks ! <apteryx>yes for the test on my system; /gnu/store/1jgcbdzx2ss6xv59w55g3kr3x4935dfb-guile-3.0.8/bin/guile <f3n1x>ACTION ah... nckx , forget my question... may i just try a keybinding/shortcut in the i3wm config file ! <apteryx>(both which seems stable in the heap size test) <civodul>if there's no significant peak early on, it's probably not going to happen later <apteryx>it's very strange that it's not deterministic <apteryx>shouldn't the heap size numbers be the same on each run? <apteryx>I wonder where the non-determinism comes from <civodul>it's expected that growth is logarithmic i guess, in that the allocation rate remains the same whereas each heap expansion is bigger <apteryx>civodul: it did grow, but it stabilizes <apteryx>at least for a time window of about 30 minutes <civodul>yes, but that's still a problem i think <apteryx>civodul: I can leave it to run for hours, but I suspect it won't grow anymore <apteryx>if that's the case, it's not really a problem? at least it's not the exact cause of the leak happening on berlin <minima>anyone can advise on this error: (file-missing "Opening directory" "No such file or directory" "/gnu/store/w3da091h97syqdia9kb9gz8gkb7vyfjh-emacs-...") - which is raised during the make-autoloads phase <civodul>apteryx: i don't believe in that theory but i think you're right: we must keep an open mind and look for other leads <Guest23>I've tried booting up guix VM under qemu and I'm getting confused. I've tried to upgrade to system (guix pull, guix upgrade), but guile --version shows 3.0.5; guix install guile tells me that I've installed 3.0.8, however guile --version still shows 3.0.5. Then I've tried guix shell guile -- guile --version and that does give me 3.0.8. <Guest23>I feel like I'm missing something fundamental. How do I actually upgrade my (VM) system? <Guest23>Never mind, hash -r. Sorry for noise. <minima>this one only contains the relevant extension, as opposed to the spacemacs one - so packaging it should be easier <minima>what should be used as version in those cases where tags are not available? just the commit hash? <nckx>minima: See corkscrew for an example of what to use in such cases. <minima>excellent example, thanks, may i simply use "0" here? (git-version "0" "0" commit) <nckx>Do let-bind ???revision???, though. Mainly because it's convention, so people bumping the package both expect it and will be less likely to miss it if it's right under ???commit???. <nckx>(No hard technical reason to, I know.) <nckx>Better luck with this repo! <apteryx>nckx: hi! I'm just checking that podiki[m] should be able to join again, right? <apteryx>I'm not exactly knowlegeable with the litharge bot <nckx>Didn't you (or someone I'm confusing you with) unban them? <nckx>litharge is pretty baroque, yeh. But (unlike ChanServ) it directly manipulates the main banlist (/mode #guix +b), so that should always reflect the truth. <tricon>"baroque" - great word and great music. <nckx>???ignoring ChanServ, but litharge doesn't use it. <nckx>I'm doing a very poor job at explaining it but the upshot is ???yes, they should be able???. <apteryx>I thought litharge had its own banlist that I didn't know how to manipulate <tricon>I was just thinking a moment ago of how I wish to learn Latin. <abrenon>how can a call to `guix search` trigger the building of a package ? <nckx>litharge has rules, and takes action on those rules, by manipulating the standard IRC channel modes. It does not do ???if see X, kick them and set +q???. At least I haven't configured it to and won't. <nckx>* +b, but whatever, same point. <nckx>That might have been them. The problem is, that's a unique ID all right, but there's no way to map it back to a nick[m] AFAIK. <nckx>So unless you set the ban or saw it happen, good luck with that. <apteryx>nckx: I thought litharge had its own list because after they were kicked yesterday they didn't appear on the akick list, so should have been able to rejoin, right? <apteryx>doesn't strike me as something normal; which package gets built? <nckx>apteryx: The akick list is another thing entirely. <apteryx>one day I'll know a thing or two about IRC <nckx>I should update that little guide I sent y'all to include litharge, but IDK, it might just sow confusion. <nckx>Anyway, that (synthetic) IPv6 above was indeed podiki, so they should be able to rejoin. I've pinged them in the ironic channel, since we share no other. <nckx>kozo[m]: You can request the channel ACL with ???/msg ChanServ flags #guix??? to see who's an op here. <tricon>nckx: i hope you know that i see the Elmo Fire GIF when you do that. <nckx>I'm guessing you're not talking about St. Elmo, so a-ducking we shall go. <nckx>Yes, that is me adminning IRC. <abrenon>I was trying to look for a package, loading a local package source <podiki[m]>I'm back! oddly had posted the same/similar links before and iddn't have a problem <sneek>Welcome back podiki[m], you have 2 messages! <sneek>podiki[m], the_tubular says: Your sacrifice was appreciated. <sneek>podiki[m], nckx says: Sorry 'bout that zap, the leading '/' got dropped (well, unescaped) in a migration to less permanent spambans. I tried to give you a bot snack but sneek stole it. <abrenon>nckx: I don't know what this is about but good luck ! <podiki[m]>the_tubular: haha thanks, hope you got it figured out! <nckx>abrenon: Thanks! Had I been unlucky, I wouldn't be talking to you now. <apteryx>nckx: ah! it was the literal '.me' in the URI? <abrenon>(previous topic continued) I had this ABI mismatch error regarding python-setuptools <podiki[m]>I did see that sneek was naughty, stealing snacks when I was down <abrenon>so I thought I'd just delete the ~/.cache portion related to this local source <nckx>It should have been /t.me/ , and was, but see how that got mis-ex-imported, being a /regex/ ? <podiki[m]>the_tubular: the config that got me is mine, and have used docker; sometimes needed to tell shepherd to launch dockerd but I believ that bug was fixed <abrenon>I remembered reading something about python build systems in the recent news, so I guix pull --news to read about it again <abrenon>I identified some potentially problematic module in the local source and bypassed it before running the search <the_tubular>I'm still not there yet, I haven't guix system reconfigure this server in a while and having trouble with the library, it complains that it needs X library that is already imported <abrenon>now I get this weird build before my search ^^ <podiki[m]>the_tubular: I'll be away for a couple of hours, but happy to look at your config and messages then <apteryx>nckx: so what is the intent of the corrected "/\/t\.me\//" regexp? ban any text containing a tab followied by .me/? <podiki[m]>ACTION away for now though, back later guix friends <nckx>apteryx: It's been replaced with a more robust regex now. t.me is a Russian(?) site popular with crypto scams. All bots in the past few months have used it, AFAICT, which is extremely convenient and will probably end soon. <nckx>It's called Telegram. Some mobile thing. <nckx>I think it's like the dodgy version of Discord or whatever, which is really saying something. <unmatched-paren>nckx: originally developed in russia, yup: "Telegram was founded by the brothers Nikolai and Pavel Durov. Previously, the pair founded the Russian social network VK." <nckx>Thanks. I don't know why I added that. It's not relevant, of course; just an artefact of me thinking via TLDs. <nckx>Seems popular in the ???Android ROM??? space to. So we're safe for now, and we can remove the regexp if it ever becomes a problem. <apteryx>civodul: the 59021 repro script reported heap size is still stable on berlin after about 2 hours <civodul>apteryx: ok, thanks for testing! we may have another thing going on <civodul>now i'm back to testing in a berlin.scm VM <danisanti>unmatched-paren: hello! I found your match.... it's name is ')'! <nckx>danisanti: To webkitgtk-next, you mean? <nckx>Hmm, I can't find that package. <nckx>It currently uses webkitgtk (2.36.7), which is why I asked about -next (2.38.0). Indeed, simply updating webkitgtk is not an option. <danisanti>this unupdated package causes nyxt to not work on some websites. Another person that compiles nyxt source said that in his/her pc the problem does not occur <nckx>I read ???WebKitGTK dependency??? as swapping, not upgrading. <nckx>danisanti: Do you happen to know which version they used? <matched-paren>danisanti: perhaps we should change nyxt to use webkitgtk-next? <danisanti>nckx: ok. please clarify me why is updating not an option. <nckx>danisanti: Well, it is, we'd just have to make it specific to nyxt, or it would rebuild too much. <nckx>danisanti: Why was webkitgtk-next not an option? <unmatched-paren>danisanti: try this: guix shell --with-input=webkitgtk=webkitgtk nyxt <danisanti>it is an option. I just needed to understand <nckx>building /gnu/store/dhqv5a771fzrgggr0i4fmdjl11dbr8gm-webkitgtk-2.38.0.drv... <nckx>(I dunno, Rust or whatever.) <nckx>I'm building it on berlin. <nckx>I got distracted by the client CA still not working, anyway. <tricon>kudos to all the devs and supporters that have worked on guix-emacs and its companions. works quite well. <nckx>There has to be a way to fix it without breaking existing certs, which are still valid??? <Gooberpatrol66>i've been experiencing a pretty brutal bug where i get asked to enter a password every time i change the brightness <Gooberpatrol66>if i do it from the panel it doesnt like the password popup and i can't click on anything and i have to go to a tty and kill the power manager to get control of my desktop back <tricon>Gooberpatrol66: what is getting executed when you change the brightness? <tricon>or to rephrase: what are you changing the brightness with? <Gooberpatrol66>tricon: the function keys (or the panel if i forget it freezes my desktop) <munen>Hi everyone. I'm trying to find out how to configure the location of the store. The reason is that I want to use `guix shell` in a CI setting and don't want to hammer the Guix servers. To cache the store, I need to move it into the user directory, unfortunately. <tricon>Gooberpatrol66: probably a groups issue, where it's asking for elevation because you're not in a given group. what groups are you in? <tricon>has this worked correctly before? <Gooberpatrol66>i think it might be happening if the system generation and the user profile are using different versions of guix <tricon>Gooberpatrol66: that is a good insight. <munen>unmatched-paren: Thanks for the hint with regards to `guix-service-type`, anyway. It's new to me anyhow and good docs to check out(; <cbaines>the store needs to be accessible at /gnu/store (unless you want to build everything with a different store path) <cbaines>you might be able to take some other directory and mount it at /gnu/store if you just want the files to be put somewhere else <munen>cbaines: I want to build everything with a different store path. The use case is to use `guix shell` on a CI server. <tricon>cbaines: i was wondering the same. <munen>I've got the `guix.scm` file, altogether with inferior configuration, ready. It works. But I don't want CI to go to the Guix servers anytime I'm starting a CI job. It just doesn't feel very nice to create that much trafic. <unmatched-paren>munen: it won't go to the guix servers if you've already downloaded the file before <munen>unmatched-paren: Not on a normal machine. On a CI server, which is stateless, it will. Unless I cache the state (the store and other dependencies) explicitly. <cbaines>munen, using something different from /gnu/store will be a very effective way of not fetching substitutes, since you won't get any, and will have to build everything from source <munen>cbaines: Doesn't matter. It's always the same stuff. I don't care if CI has to build the packages once. <cbaines>as I say, I'd consider trying to have the files somewhere convinient, then mount to /gnu/store <munen>cbaines: That was what I was considering as a fallback, too. But since the docs mention that `/gnu/store` is the default path, I was strongly hoping that there's a configuration flag for it(; <Gooberpatrol66> /etc/polkit-1/actions/org.xfce.power.policy refers to a different xfpm-power-backlight-helper than the one in my profile <cbaines>munen, I cannot stress enough how much this isn't something you "configure" when using Guix, but rather some core aspect of how packages/the guix-daemon/various tools work <cbaines>another approach is to run your own instance of guix publish somewhere, and then query that first and fallback to the default substitute servers <Gooberpatrol66>it's way too fragile. it's too difficult to keep the system and user profile on the same guix version <nckx>I didn't follow, but it is. <munen>nckx: Are you saying the path to the store is configurable? ???? <munen>If you know how to configure it, I would be grateful for a hint(; <nckx>Yes, if you're OK with building every single thing from source for the entire life of the system. <nckx>You'll need to set --localstatedir=/var as well, as that is not the default (which is not very sane, but has its historical and other reasons). <munen>With that I can happily use `guix shell` on CI (without hammering the Guix servers on every CI run^^). <nckx>munen: Everything cbaines said above applies, and you've explicitly disclaimed any interest in warranty claims ???? That said, if you find bugs, they should be fixed or the option removed. <tricon>munen: would be curious to hear of your experience as you go through with this. <nckx>munen: I'm not entirely clear on what you're doing, and one can simply disable substitution, but you clearly have a plan (what with this ???caching??? in ~). Good luck. <cbaines>I'd also be interested in what you're planning to use as the store path? <cbaines>I do wonder if there are bugs lurking around that depend on what characters you use, and how long the store path is <cbaines>also, given how many times /gnu/store shows up if you search gnu/packages/*.scm, I do wonder if that's something we're doing a poor job of keeping out of package definitions <cbaines>I was just recently looking at a bug in r-minimal that was related to the store path being included in the derivation <munen>I'll happily expound my reasoning(; I want to use `guix shell` on a self-hosted Gitlab CI instance. Gitlab CI is stateless, hence it's notorious to 'download the internet' on every run. Doesn't matter if it's OS packages being installed ad-hoc or dependencies of your actual project. There's a solution, though: To explicitly cache certain folders. However, Gitlab CI only supports caching of folders within the users home directory. <nckx>munen kindly offered to put on this guinea pig costume and go poke that sleeping dog, it would be rude of us not to accept their sacrifice and learn from it. <nckx>Even if what we learn is ???yeah never ever do that again???. <nckx>Godspeed, my fuzzy friend. <nckx>munen: I didn't think this through (at all), but is there no way you could symlink/bind-mount/??? your way out of that predicament to get the local maximum of both worlds? <nckx>Well I did think it through but I know 0 about GitLab. <nckx>So it was a very brief affair. <munen>nckx: Using a bind mount was my first idea, too^^ I just wondered if there was a more Guix native way since the docs mention that `/gnu/store` is a default. And any default can be configured, right? ???? <nckx>Technically, yes. I don't think it's ???more Guixy??? to do so though. It's??? orthogonal. <munen>nckx: Thank you for your kind support ???? <lechner>Hi, why would /run/user not be available, please? I am getting a variety of errors when logging in, including warning: XDG_RUNTIME_DIR doesn't exists, on-first-login script https://paste.debian.net/1259885/ <lechner>Hi, I'm on a different client. Can anyone see this? Also, did I post something five minutes ago? <lechner>podiki[m]: that's okay. i wasn't trying to be pushy, but a misconfiguration knocked out my weechat in the cloud <unmatched-paren>lechner: i believe you need either elogind or seatd+greetd to get that fixed <podiki[m]>no worries! having been on the other side of wonky connections I get it <efraim>are you using %desktop-services? that included elogind <lechner>efraim: no, i'm not. it's a cloud instance <efraim>I believe unmatched-paren is right, you'll need to add elogind to your os config <lechner>efraim: thanks! was this a recent change? <efraim>don't think so, but something might've changed when seatd+greetd services got merged <lechner>and should elogind not be part of %base-services? <unmatched-paren>lechner: probably not, as it's usually used with a desktop, and a console session mostly has no need for it <lechner>unmatched-paren: i don't know about that. i get login errors every time, plus some software like weechat plainly refuses to run <dthompson>anyone here happen to run guix system on a recent gen thinkpad x1? I'm curious about any problematic hardware due to proprietary drivers. I know I can expect the intel wireless chip to be an issue. <sneek>Welcome back dthompson, you have 1 message! <sneek>dthompson, old says: that the concept in Catbird look nice and that I will try to hack with it <dthompson>sneek I really wish you would remember which channel that message was from because it was from #guile! <unmatched-paren>dthompson: unless it's intel xe integrated graphics, you can expect an intel laptop to work well except for the wifi <dthompson>unmatched-paren: ah I was looking at 12th gen <dthompson>thanks, though. I thought there might be something sneaky like that going on. <dthompson>I haven't looked at new laptops in ages so I am definitely behind the times. <efraim>with the pandemic I went with a desktop and a pinebook pro <dthompson>used to be that an intel cpu with integrated graphics required no blobs <lechner>Hi, I would like to file a bug to move elogind from %desktop-services to %base-services. Does anyone here disagree, please? <lechner>or should i really use something as complex as %desktop-services in a cloud instance? <lechner>ACTION hears echoes of the cries when Debian mandated systemd <nckx>That's not the alternative. The alternative is adding only elogind when you need it. Or fixing guix home not to mandate it in the first place. <tricon>it's a good question; but perhaps it's best to let those that need elogind with %base-services add it manually. else, upgrading Guix and reconfiguring could cause an unwanted surprise on established systems. <lechner>nckx: when is elogind not needed, please? <nckx>I don't know how to answer that question without being tautological. <lilyp>imho guix home relying on elogind sounds like a bug <lilyp>then again, guix home itself is basically an open beta <lechner>lilyp: but weechat is not a graphical program either. i get Error: cannot create directory "/run/user/1000/weechat" <nckx>lechner: Why the ???graphical??? distinction? It seems to come out of nowhere or I missed a context. <lilyp>because we're pitting %desktop-services vs. %base-services <lilyp>note that you don't need weechat on a server either <nckx>The only reason I use weechat at all is because I run it on servers. But that's neither here nor there; as long as the ???graphical??? criterion makes sense to you it need not make sense to me. <lechner>nckx: well, graphical environments tend to pull in a lot prerequisites, such as gdm and pulseaudio, don't they? <nckx>I don't understand why that's relevant. Anyway, weechat does not require /run/user at all. <nckx>You're running it on that busted ???guix home??? server that set XDG_whatever in the first place. <nckx>This seals the ???guix home bug??? diagnosis for me :) <nckx>ACTION disappointed there's no seal emoji. <tricon>we need a neural network that can learn to describe GIFs in plain text for IRC. <jab>dthompson: MNT reform should be selling laptops that have 8GB of RAM soonish...their latest news articles talked about that. <lilyp>actually, seal is part of 13.0 <dthompson>jab: 8GB is not nearly enough for what I do. <lechner>Hi, what's an easy wasy to reformat a Scheme one-liner (from the Home service) so that it becomes readable, please? <lilyp>for anything more fancy you need to actually think <jab>dthompson: what stuff are you doing? 8GB is pretty much the minimum for me. Thought since I use sway, I could probably get away with 4GB or 2GB. <jab>dthompson: actually I check the article just now, and they may be supportin 16GB soonish...where soonish is defined as I have no idea when. <dthompson>jab: 16GB is my minimum. haven't used a laptop with less than that in many years. <tricon>i'm on 16 GiB myself. presently idling at 5.24 GiB, mostly due to 2022 web browsers and sites (O_o) <nckx>lechner: So it's not setting XDG_RUNTIME_DIR. Interesting. I wonder where weechat's getting it from. <nikolar>i am on a 8gb laptop with 6.5 actually usable due to igpu <nikolar>and firefox works just fine with a bunch of tabs open :) <nckx>unmatched-paren: No mention of that in the spec, or in the code, nor does simply running weechat exhibit that behaviour. <unmatched-paren>nckx: Hm. We use (or (getenv "XDG_RUNTIME_DIR") (format #f "/run/user/~a" (getuid))) quite frequently in (home|system) services <nckx>lechner: Can you explain what that link is for? <nckx>It just getenvs XDG_RUNTIME_DIR. We knew that. <nckx>What I'm saying, but it's hard to prove a negative, is that there's no ???default value??? applications should fall back to if it's unset, as it is by default on Guix System without extra desktoppy services enabled. <nckx>Should be easy to reproduce with XDG_RUNTIME_DIR= weechat. <nckx>lechner: So is XDG_RUNTIME_DIR set on your system? <nckx>The one without elogind. <lechner>nckx: Yes, it is set to /run/user/1000 <nckx>Right. So that's my question: who's setting that, clearly just assuming elogind will always be running. That's a bug. <nckx>Somebody's telling programs ???don't use your defaults (which would work), use this (which won't), trust me???. <the_tubular>I'm having trouble doing something that should be simple... Adding docker to my guix(os) server <the_tubular>I'm getting errors relating that library that should be imported <nckx>If that was for me, I am really not the right person to ask. I don't use Docker. <nckx>By all means, paste away. <the_tubular>Here's the error right now :?? error: module (gnu services elogind) not found <nckx>There is no (gnu services elogind) module. <nckx>elogind is in (gnu services desktop). <nckx>You can use ???guix system search???, I believe. <the_tubular>Can you paste how that should be imported from gnu services desktop ? <nckx>I really don't know your context so I can't say, but (use-modules ??? (gnu services elogind) ???) or (use-service-modules ??? desktop ???) <the_tubular>This is what I have at the moment : (use-package-modules desktop ...) <nckx>Right, then just add (use-service-modules desktop) under that. <nckx>It would really help if you just shared your entire file. <nckx>My turn to be AFK a while, but there are many competent people here, and the incompetent ones are very nice. <the_tubular>Also, nckx, and probably everyone else, I'm looking for documentation on how to get better with guix, what should I read ? <the_tubular>I check out a bit "the little schemer" and the cookbook <nckx>the_tubular: OK, so. (use-service-module desktop ???) was correct, but you still need to remove the bogus ???elogind??? from that ????????? or you'll keep hitting the same error. The error mentions (gnu services elogind) because use-service-modules is just syntactic sugar for (use-modules (gnu services A) (gnu services B) ???). There is no (gnu packages desktop) module. Use (targets (list "/boot/efi")) ??? what you have will work, but is deprecated. %desktop-ser <nckx>(Answered publicly with permission.) <nckx>the_tubular: I can only answer for myself: read relevant Guix code. Documentation is great (and some Guix/Guile documentation is, too), but reading code helped me at least as much. YMMV. <nckx>the_tubular: There might be more errors after that, I didn't continue. Time for bed. <the_tubular>Thanks that seems to have worked, finally hitting another error Yay (?) <the_tubular>guix system: error: more than one target service of type 'udev' <nckx>[Mintily fresh:] Did you read my entire messages? <nckx>If you do, I guarantee your system will at least build without errors. <the_tubular>Yes, I read the full message, but you went too fast, my guess is that it isn't related to the "target' field right ? 