<kkebreau__>We have about 13.5 thousand packages, so there's a good chance that most of what you need for a day-to-day machine is present. <azvede>boy howdy, that's a large number.... thanks! <kamil_>azvede: A beginner here. From the looks of it, writing own package modules seem no harder than composing PKBUILDs, Flatpaks etc. so if there's something you're missing, you might be able to add it yourself <mbakke>I always forget about install-file <azvede>kamil_ having never done any of that, that actually doesn't sound so horrible... and could be a nice way to contribute if I find a package that's not around <kamil_>azvede: As a person with no deeper knowledge of compiling process, I can assure you that tools, such as meson, automate the process to the degree that you can piece various bits of actual package modules/PKGBUILDs/Flatpaks together to create a package of your own. It's the shortest way possible to package software using your favourite software packaging format, without learning too much ^^ <kamil_>azvede: In case of Guix, I find its manual a very good starting point to learn a little about it from: from using the command line interface tools for trival stuff like updating your system installation to building your own packagaes. Everything you need to get started on Guix is there. If you still have questions, you can come here and ask everyone. So far I've had a hella good experience asking here for help, so should you~ <ryanprior>Has anybody looked at offering AppImage as an output format for guix pack? <kamil_>mbakke, No change on my end :( It's stuck on "Waiting for initial location to become available..." <mbakke>kamil_: did you restart the geoclue service after reconfiguring with the new redshift-wayland? <kamil_>mbakke, I did guix pull, reconfigure, and restarted a VM <mbakke>kamil_: you're not gonna like this, but I think you'll have to file a bug report :-/ <mbakke>is there a way to test whether geoclue works at all? *mbakke has no clue about geoclue ***sneek_ is now known as sneek
<kamil_>mbakke, the easiest way would to be to start over again in a new VM, install GNOME from the graphical installer, which, if I'm correct, should install and configure %desktop-services that geoclue is a part of <kamil_>Night shift functionality in GNOME uses Geoclue2 <kamil_>I'm using it myself right now ^^ <mbakke>right, would be good to check if night mode works in Guix GNOME <mbakke>kamil_: you don't need to start over though, just adding gnome-desktop-service-type and %desktop-services in the configuration should be sufficient <kamil_>mbakke, alright, I'm gonna do it <mbakke>contrary to many other distributions, removing them later won't litter your system with leftover packages :-) <kamil_>I'm taking some precautions and taking a snapshot of the VM though ***sneek_ is now known as sneek
<kamil_>mbakke, I had to delete %base-services, as there was a conflict between it and %despktop-services. It's building now, we should know the answer as to whether it works at all soon <terpri>ryanprior, i don't know, but appimage is...not good compared to systems like flatpak in my experience <terpri>much more reliance on the host machine being a traditional gnu/linux system (FHS, etc.) <terpri>of course i'm working on getting appimage programs running under guix, making appimages out of guix packages may be a lot easier <mbakke>kamil_: cool, looking forward to the result <ryanprior>terpri: I don't like appimages myself much compared to flatpaks on desktop or docker on server, but since other people do I think it would be nice to offer that as a deployment option. <ryanprior>I didn't know you were working on appimage, what does it mean to get them working under Guix? <terpri>ryanprior, i'm sporadically attempting to get unity3d (proprietary gamedev environment) running directly under guix; the gnu/linux version is distributed as an appimage <ryanprior>What does "running directly under guix" mean? Running on GuixSD? Or turning the AppImage into a Guix package? <ryanprior>Sorry if I'm being dense, that phrase just doesn't parse for me ':D <terpri>running on guixsd, yes, ideally by turning the appimage into a package <terpri>directly under guix vs. running it in a VM <pkill9>appimage is good because you don't have to install something to use it <ryanprior>What happens when you try to run an appimage on Guix System? Their reputation is that they're pretty self-contained, I take it that's not always the case? <terpri>in theory, yes, but the binaries are not really self-contained, they expect to be run on an FHS system and, i believe, explicitly don't embed common libraries like glibc <pkill9>flatpak is the worst of both world's <terpri>let me see if i can generate some examples <pkill9>you have to install flatpak to use flatpaks <ryanprior>I prefer Flatpak for desktop use cases, because I like that there's a required Flatpak runtime agent which you use to discover and install packages and updates. ***sneek_ is now known as sneek
<ryanprior>appimage seems pretty simple, which is nice locally, but problematic globally. <terpri>for example...with UnityHub.AppImage, step 1 is, do patchelf --set-interpreter because it's hardcoded to /lib64/ld-linux-[...] <ryanprior>terpri: afaik this channel has a blanket policy discouraging discussion of proprietary software like Unity. I'd be happy to chat about it in another venue of your choosing. :) <terpri>if you run that, you get an error: "error while loading shared libraries: libz.so.1: cannot open shared object file: No such file or directory" <terpri>it's applicable to appimages generally, i believe <kamil_>mbakke, it doesn't work on GNOME. I've enabled "Location Services" in "Privacy" in the Settings app, and also "Night Shift" in "Displays", but to no avail. <terpri>to be explicit, if you want to do gamedev on gnu/linux: use godot! <kamil_>mbakke, I know you're asleep, but I'll talk later to you when I come back. I'm disconnecting now <ryanprior>Godot is really cool, I've done a little game dev work with Godot and was impressed. <terpri>now that libz.so error is not from the appimage directly, ldd *.AppImage will tell you it's not a dynamic executable. i think it's coming from a program run inside the FUSE filesystem the appimage binary automounts <terpri>ryanprior, thanks for the link, i'll use that for testing if i end up building something generally useful <terpri>i'm only running unity to work for a client; actually i plan to port their system to godot, and am expecting them to liberate their source code if that's accomplished <terpri>no commitments yet, but they seem very interested in doing something like licensing under the GPL and selling exceptions (a business model endorsed by rms) <terpri>(the project is basically a giant library of general-purpose unity components) <terpri>actually, i'm curious if there are organizations that can provide free/cheap advice to businesses about moving to a selling-exceptions model <terpri>(i'm doing a good job of persuading the client on my own, i'm happy to report, but it's a bit outside my bailiwick) ***sneek_ is now known as sneek
<terpri>exactly the same initial error with the zulip appimage btw, it's a general appimage+guixsd issue and not specific to [censored proprietary software]'s packaging ***sneek_ is now known as sneek
<terpri>i am really optimistic about getting some useful free software released via this contract, otherwise i wouldn't be working with a proprietary software company ***sneek_ is now known as sneek
<terpri>i hope it's not *too* unethical to work on a proprietary program if that labor leads to the program being freed in the immediate future, it's a gray area for me <joshuaBPMan>terpri: You could read the book "The Cathedral and the Bizarre". It talks about how "open source" may be a good methodology for business. <joshuaBPMan>RMS disagrees with the "open source" philosophies, but those are great philosophies for convincing business people to try open source-ing software. <ryanprior>terpri: I'm not here to judge (& it sounds like you're taking a considered approach & advocating where you can, which is great!) and the rules for the Guix channel aren't intended to censor. Guix has a political mission in addition to a technical vision, and the guidelines are to help align the discussion here with the political mission of the project. <joshuaBPMan>terpri: Also, there is an RMS endorsed free software business model for games. You release the source code for the game under the GPL. You make the media assets proprietary. ***jonsger1 is now known as jonsger
<terpri>ryanprior, thank you, it helps to hear that, being in aporia as i am. i understand the political mission; i'll try to be just a little more careful discussing my vocation and avocations here <terpri>joshuaBPMan, i'll re-read that, it might give me some rhetorical ideas, thanks for reminding me it exists <terpri>i'm more of a free culture maximalist than rms, but i'll be sure to mention the proprietary-assets option to my client too, as some sub-clients might be open to that kind of thing <terpri>to quote a bit of doggerel from a brilliant lisp historian: "I will not cease from Mental Fight, / Nor shall my Sword sleep in my hand: / Till we have built the Lisp Machine, / in Linux' green & pleasant Land" <Kozo>mbakke: If by night mode you are talking about night light, it works on Guix Gnome <reepca>anyone know if there would be anything blocking packaging of terasology? <ryanprior>reepca: you might have to package Gradle first <fvr>On booting, gnome is saying "Oh no something has gone wrong." I don't see any errors in Xorg logs, there are some errors about an exception in JsApi in gdm/greeter.log. Any ideas as to how to debug this? Here's the greeter.log paste http://paste.debian.net/1148525 <fvr>I see some replies online as suggesting to rerun startx and to see what happens, from the manual I gather that (xorg-start-command) should return a startx script, how'd one execute this returned script? <terpri>fvr, could gdm be running wayland rather than xorg? (debian gdm is wayland by default these days, dunno what guix system does) <terpri>oh, i see you posted an xorg debug log, never mind :) <fvr>In all the custom.conf files I see WaylandEnable=False commented out, so I it's using Wayland <fvr>I guess it's using Wayland <terpri>fvr, maybe try slim as the login manager? (iff your goal is to just log in rather than to fix the problem) <terpri>"(.gnome-shell-real:948): Gjs-WARNING **: 08:20:41.842: JS ERROR: Error: Requiring NMA, version none: Typelib file for namespace 'NMA' (any version) not found" <terpri>those lines look suspicious. NMA is, i think, NetworkManager, is that service installed? <fvr>login is not so important as I am just doing all my work on TTY on emacs :) , I want to fix it. <fvr>How do I check it. I see the networking service up and running <terpri>fvr, if you are using %desktop-services in the config NetworkManager is enabled by default <fvr>I think it was the default ***sneek_ is now known as sneek
<terpri>there is also network-manager-applet, which presumably has those typelib files, but i'd be surprised if gdm didn't indirectly depend on that package <fvr>But are we sure that these errors are related directly to why it's not starting <terpri>".gnome-shell-real" crashing sounds bad though <terpri>fvr, do you have gdm debug logging enabled? (mentioned in (info "(guix) X Window")) <fvr>Yeah I think that's responsible, it's followed with `gnome-session-binary[858]: WARNING: App 'org.gnome.Shell.desktop' exited with code 1` which I think is serious <fvr>It's at the default log-level info <fvr>I'll try setting it to debug and check <terpri>"(.gnome-shell-real:948): Gjs-CRITICAL **: 08:20:41.848: Attempting to call back into JSAPI during the sweeping phase of GC." is interesting <terpri>in the sense of "how the heck could that happen", not necessarily relevant to your problem :) <fvr>My bad, I was looking at networking. No I don't have debug set to true, I'll try doing it now. Is reconfiguring and restarting the xorg-server enough after setting it or do I have to reboot the system? <terpri>fvr, i would think a restart is unnecessary, but not certain <fvr>I am confused, how do I set the gdm-configuration? I tried doing appending this, `(service gdm-service-type (gdm-configuration (debug? #t)))` but it saysmore than one target service of type `gdm` which I am thinking is from the %desktop-services <rekado>fvr: use (modify-services %desktop-services (gdm-service-type config => (gdm-configuration (inherit config) (debug? #t)))) <rekado>your code would add a new service of type gdm-service-type <fvr>Okay, so it modifies existing gdm-service-type to add debug <rekado>whereas using modify-services picks out the service of type gdm-service-type, binds its current configuration to “config” (or whatever you want to name it) <rekado>and then you give it a new configuration. <rekado>using “inherit” you reuse all of the previous fields, and override the debug? field <fvr>Yeah looking at the example in the manual I just realised that <rekado>I think it may be good to add a little example like this to the cookbook <rekado>it would be less ambitious than the other things in the cookbook, but I think it would be a valid addition as a demonstration <ryanprior>rekado: have you seen the bootstrap package I added for v language? I'm waiting for feedback on that before I continue in that vein. I'm thinking about getting a git-bisect based system doing to discover what the path is to bootstrap from the latest release to the upcoming release. <ryanprior>v development moves so fast that the master branch already doesn't build with the latest release, you need a more recent compiler to build master. X.X <rekado>I’m still neck-deep in Haskell / lightdm / documentation stuff at the moment :-/ <ryanprior>That sounds good! I hope you eventually get a chance to take a look, I'm interested in opinions from people who have more experience with building bootstrap systems. :) <fvr>terpri, the main error still is the same JS Error. There are other errors but they are all logged as Warnings, so I guess they're not that bad. <boeg>Do i need some supplementary package to ghc-pandoc to be able to use the texinfo format with pandoc? <fvr>terpri, I changed the login manager to slim and when I log into gnome still facing the same problem. But xfce seems to be working <terpri>fvr, if it's reproducible, i'd be happy to take a closer look. not a gjs expert (my experience is primarily with spidermonkey; i implemented BigInt for firefox) but i might be able to figure out where it's going awry <terpri>maybe some version mismatch between networkmanager applet and gnome? just speculating <terpri>i'd hope gnome could manage to run without networkmanager around, too <fvr>I doubt it's a version mismatch, I am thinking some config files went missing or somethign similar <fvr>I am not sure how it can be reproduced because it was working and guix was failing for some reason and I restart and the problem occurs <terpri>by version mismatch, i was thinking of a hardcoded typelib file path from a wrong version or something, so we're thinking along the same lines <terpri>isn't there a way to verify the store contents? maybe in "guix gc" or something? <terpri>i've gotten a corrupted store on a few occasions <fvr>Let me see in the manual if there is some way to check the integrity of the store <terpri>(in my case, manifested as sqlite errors, so not helpful to your problem) <terpri>fvr, guix gc --verify[=contents,repair] <terpri>not sure whether --verify and --verify=contents are different (maybe the former doesn't actually hash the files?) <terpri>unfortunate that there's no obvious way to, say, just verify everything in %desktop-services <fvr>It's taking a real long time to verify the hashes <terpri>fvr, it wouldn't shock me if it took a whole day, on an average guix system desktop install <terpri>(maybe underestimating sha256 performance, that's just my intuition) <terpri>fvr, maybe could speed it up a little by cleaning out the store: "guix gc" and then delete as many old generations as you're comfortable with <terpri>("guix gc -d 1m" to nuke generations more than a month old, for instance) <fvr>Nah, it's better I'll just leave it running in the background. If it deletes some libraries that I immediately again need for building something else, have to wait for it download again. And the download speeds aren't great, looks like they're limited by the server <terpri>my /gnu/store is around 200gb, not sure how long it should take to sha256 all of that <fvr>Mine is a new install, I think it's under 30G <terpri>hopefully hours not days to verify (again, just a guess) <fvr>It verified the contents successfully, so no luck here <sammich>Hey, when i try to do open file dialogs with flatpak on guixsd i get a long error related to: 'Can't open portal file chooser' Anyone know what i need to get this working? <marusich>sneek, later tell dftxbs3e FYI I was successful in reproducing 4 out of your 5 powerpc64-linux binaries - gcc-stripped-7.4.0-powerpc64-linux-gnu.tar.xz differs. <bdju>I've just copied the emacs init.el I used on NixOS and placed it in ~/.config/emacs/. When I launch emacs I get this error in the minibuffer, and my config is seemingly not being applied: Cannot open load file: No such file or directory, seq <bdju>I've tried to install my emacs packages all through guix. on NixOS I was getting them via home-manager, so I'm hoping I can get it to work similarly. <bdju>oh, and I'm using emacs-next <fvr>bdju, I placed my init.el just at the usual ~/.emacs.d/init.el and it's picking it up <fvr>I had to relogin after I installed emacs though, don't know why <bdju>I suppose I could try that, but I wanted to use the new config location <bdju>Same error/behavior when I copy it to the old spot also. I left it in the new spot, though. <bdju>I guess I can also try a re-login, although I have some stuff I'm in the middle of so not until later. <fvr>I think reopening the shell should have the same effect <bdju>okay, I confirmed now that I can use the xdg config dir but it's over-ridden when I also have the old one <bdju>also I now have a different error, but it's at least applying part of my config <boeg>when submitting a patch to guix-patches@gnu.org, how long before it appears on issues.guix.gnu.org? <fvr>bdju, Can you just M-x eval-expression and enter (set-default-font "Terminus-9") and check what happens <fvr>Also is your $EMACSLOADPATH pointing to the correct locations, it should point to the elisp directories in .guix-profile <bdju>ah, possibly related to your question, I do get an error when launching from a terminal <bdju>Warning: Lisp directory '/home/brad/.guix-profile/share/emacs/26.3/lisp': No such file or directory <bdju>when using eval-expression I get "Symbol's function definition is void: set-default-font" <bdju>I don't see EMACSLOADPATH in my .profile at all actually <bdju>It mentions both emacs 27 and emacs 26, but I removed the normal emacs package, so maybe I should remove the emacs 26 stuff somehow <fvr>On my system it's defined at ~/.guix-profile/etc/profile <bdju>exporting the variable with the old emacs stuff taken out removes the errors in the terminal on start, but then I still get the same thing in emacs complaining about my set-default-font <fvr>When you $EMACSLOADPATH what's the path it prints? <bdju> http://ix.io/2ncK but here I redefine it and it removes the warning from before, but doesn't fix the main problem <fvr>You shouldn't have to redefine it yourself, guix should define it appropriately <bdju>seems minor anyway, was a warning rather than an error... the font thing on the other hand is stopping my init being loaded so evil-mode and such are not activating which is very bad <fvr>Yeah I think they're related because emacs has to get those function definitions from the lisp files in there <bdju>but the thing is, I've removed the normal emacs package which is emacs 26, and I'm only using emacs-next which is emacs 27. both the value it had and the one I just set myself include the equivalent path for emacs 27 by the looks of it <bdju>the warning was specifically the emacs 26 directory <bdju>do you think it needs it? I wouldn't think so... <fvr>No I was just asking if the variable is correctly pointing to the files, because it shouldn't tell function undefined otherwise <bdju>well, you probably know what's going on better than me I guess. what should I try to do next? <travankor>does dmarc work with the guix mailing lists? <fvr>I have just started using guix but faced the same problems yesterday. I would first verify in a new shell session whether $EMACSLOADPATH is being correctly set or not and if emacs minus the init file loads correctly or not. <bdju>I did open a new shell a while ago and these results are in the new shell <bdju>it was getting less far before the new shell <fvr>So now the EMACSLOADPATH doesn't have the removed 26 version? <bdju>in the new shell but the same session, yeah, it hasn't removed the 26 version except where I set it differently, besides the warning the behavior afterwards is the same <bdju>I've just closed all my shells and opened a new one (again) and so the emacs 26 no such file warning is back, but it doesn't seem to change the font error <fvr>Is it set-frame-font instead? <bdju>set-default-font is still what it's complaining about <fvr>Yeah I see that it's been renamed to set-frame-font and the former made obsolete (but I think it should work nonetheless) <bdju>I've just commented out the line and now it launches without issues! <bdju>but while I'm at it I'll try set-frame-font <bdju>no error and the font actually changed <bdju>I wonder why they couldn't just build in an error to tell me what to change it to... <fvr>That's how it is on my version <fvr>Can you check emacs NEWS for version 27 whatever you're using <bdju>uh yeah, but I've never used the news before. what should I be looking for? I just opened it <bdju>oh, I searched "font" and notably they changed to HarfBuzz for a text shaping engine <bdju>lots of font stuff actually <fvr>But I don't know if it would be properly catalogued in emacs-next <bdju>I searched set-default-font now and it's in a big list of functions removed, says they were all "obsolete" since macs 22 <bdju>but that's the only mention of it <fvr>That's when they made it obsolete <fvr>and also in the lisp/frame.el.gz where it's defined <bdju>oh, yeah, I see it there if I open it <bdju>I don't usually read this stuff. I guess maybe I should ***wxie1 is now known as wxie
<efraim>marusich: I'm pretty sure gcc is known to not be reproducable for bootstrap binaries <mbakke>the latest core-updates round changed the timestamps on compressed archives from epoch 0 to epoch 1, I guess that might explain the difference <mbakke>kamil_: thanks for testing, I guess geoclue does not work at all on Guix? :-/ <janneke>maybe %default-theme => (grub-theme) <bdju>agh my screen locking broke after a reboot <bdju>looks like my setuid thing for swaylock refers to /bin/swaylock which doesn't exist... I wonder why <bdju>oh, is this a guix change? I see if I do `which swaylock` I get `/run/setuid-programs/swaylock` <bdju>nevermind, now it outputs ~/.guix-profile/bin/swaylock <bdju>do you use something else instead? <bdju>okay nothing I'm trying is working here, this really sucks. I lock my screen 10+ times a day <bdju>it worked the way I had it on another machine for months, and on this one until recently... ugh. <bdju>I think I'm doing it somewhat right with how I had it. It creates the /run/setuid thing because of me putting /bin/swaylock in my config <bdju>but my key bind does nothing, running the command manually does nothing, no error <bdju>nevermind, doing it with swaymsg exec has no error. by itself I get "2020-05-24 07:07:24 - [source/shadow.c:19] swaylock needs to be setuid to read /etc/shado" <bdju>oops didn't copy the last character <mbakke>bdju: have you done anything with $PATH? /run/setuid-programs is supposed to be the first entry to prevent these kinds of problems <mbakke>bdju: you can also change your hotkey to refer to /run/setuid-programs/swaylock directly <bdju>I'm not sure. I haven't done too many reboots since installing. Maybe my .profile hadn't taken effect until the reboot. <mbakke>bdju: the export PATH line is what breaks things <bdju>well, I meant in that line... am I not supposed to change my path at all? I had that on my other guix system machine for ages <mbakke>bdju: actually most of those variables (C_INCLUDE_PATH, PYTHONPATH, GIT_EXEC_PATH) etc should probably not be there: they will be defined automatically when you install the relevant programs <bdju>it used to always print out these after installing new things and ask me to put them somewhere <mbakke>bdju: yes that warning is a bit misleading: logging out and in again is the correct way ... <bdju>oh dear, well now I know <bdju>I'll remove most of these and log out and back in then <mbakke>great, good luck! improvements to that warning message welcome :-) <bdju>looks like it does work now. thank you for the help mbakke <joshuaBPMan>bdju: I do not use swaylock. When I close my laptop and reopen it, sway just starts as it was. <bdju>joshuaBPMan: ah I see. I leave my laptop docked all the time like a desktop and I lock it whenever I'm leaving my chair. it also makes my monitors dark. just part of my routine. <bdju>I find putting computers to sleep to be annoying and just slow me down. plus then I can't remotely access them <bdju>I like to ssh into my computer from my phone a lot <krusnus>I'm trying to remove gdm from my system but using "(remove (lambda (service) (eq? (service-kind service) gdm-service-type)) %desktop-services)", as specified in the wiki seem to do anything. Here's a paste of how my services-setup looks in my system configuration https://paste.debian.net/1148564/ any help would be appreciated :) <cbaines>set-xorg-configuration I believe interacts with the gdm by default <cbaines>which will be adding in back in to your set of system services <krusnus>cbaines: well that sucks. Do you know of any way i could stop it from adding it back? <cbaines>well, what's your intent with using set-xorg-configuration? <cbaines>are you trying to use something else in place of the gdm? <krusnus>swaping caps and escape. No, i want to use startx to login <cbaines>I've seen people occasionally asking about using startx, but I'm not aware of how (or if) that works <krusnus>cbaines: I've had it work on a previous instance of guix but i didn't use set-xorg-configuration. <krusnus>I'm having trouble choosing which one i want the most though <bricewge>cbaines: set-xorg-configuration is only to be used with services that support it (ie. gdm, slim, ...) <cbaines>I don't follow, I thought that's what I said? <bricewge>krusnus: As for startx I didn't saw someone successfully using it, tho it would be nice to have it <bricewge>cbaines: Yes definitly. Sorry, I used the wrong username :s <hulten>Hello! I am confused about mounting USB flash drives' filesystems on Guix System; confusion started after I decided that NTFS would be the best filesystem (for reliability and cross-platform). <bricewge>krusnus: Opening an issue about it would a good start <hulten>I like to mount (or automount) as a normal user. Why am I not allowed to use 'mount' for this even as it is setuid? <hulten>And then there is 'fusermount' that has neither a man nor info page. It complains about «old style mounting not supported». <hulten>Other tools that seem interesting are 'ntfs-3g' and 'udisksctl'. <hulten>The former is great as it can mount NTFS writable. <hulten>The latter is great as it can do it as a normal user, but only ro. <bricewge>hulten: The manual for fusermount is in the fuse package, install it <hulten>I'll do that. Why is it not provided by 'fusermount'? <krusnus>bricewge: my problem isn't getting startx working though. I've succesfully done that before on guix. What's sad is that set-xorg-configuration doesnt work if i set want to set up startx and disable the login manager ***dingenskirchen1 is now known as dingenskirchen
<cbaines>Maybe there could be a startx-service-type that somehow makes that work... <hulten>bricewge: I see 'fusermount' is not a normal package, because it resides as an executable in /run/setuid-programs/. <hulten>So, it is part of the 'core system' (in so far that exists in Guix)? It is at least not in /gnu/store/. <bricewge>Yes we need a startx-service-type or a procedure taking a xorg-configuration that reutrn a properly wrapped startx <bricewge>hulten: /run/setuid-programs/ is populated by taking non setuid binaries from packages and putting them there as setuid <bricewge>hulten: In your case the fusermount binary comes from the 'fuse' package, which also contains the manual <rekado>bah, looks like GHC really does need to have a reference on GCC, because it tries to use its record of the build-time GCC when packages run “cabal configure”. <rekado>annoying that it won’t take GCC from the environment <rekado>please show us your configuration file <bricewge>krusnus: I played a bit with ntfs-3g, and it seems that I get a read-only filesystem when using 'mount' instead of 'mount.ntfs-3g' <rekado>you have a list of packages where something else is expected <bricewge>hulten: udiskctl is probably using the former <cbaines>rndd, it looks like your issue relates to %rndd-packages <cbaines>you've got a list of packages, but you've including %base-packages at the end, but also within the list as well <rekado>you’ve got two %base-packages there <rekado>because what you have is a nested list <rekado>you want a plain list of packages <rekado>I wonder if we can catch these kind of user errors and print something nice <cbaines>it would be really nice to go round an try to do more input validation <cbaines>well, nice in terms of the user impact, not so nice in terms of code complexity <rndd>sometimes hard to catch it <bricewge>hulten: I managed to get a NTFS partition to be {,auto}mounted by installing ntfs-3g in the system profile instead of just in an environment or a user profile <bricewge>Using 'mount' or 'udisks' via 'udiskie' get me a RW filesystem <guixy>Some packages containing game data are not defined publicly <guixy>This makes it difficult to debug. <pkill9>guixy: this is deliberate, because they're only used for those packages <pkill9>guixy: you can access them with (@@ (gnu packages module) variable) <guixy>I noticed there's a function called hidden-package in guix/packages.scm <pkill9>guixy: and you can build them on the command line with `guix build -e (begin (@@ (gnu packages module) variable))` <guixy>It says it defines a package hidden from the interface. <pkill9>e.g. guix build -e '(begin (@@ (gnu packages games) xonotic-data))' <pkill9>oh yea, i'm used to always using that <guixy>Why aren't those hidden packages defined with (hidden-package)? <pkill9>probably because no other packages outsid ethat module references those packages <pkill9>hidden-package prevent shte package being accessed as a 'specification' <pkill9>e.g. the internal search sort of system <guixy>So if packages are defined with hidden-package, they won't be found at all? <pkill9>but still exports the package as a variable for other packages to use, i think <pkill9>they won't be found with commands like `guix package -A <package>` and `guix build <package` <pkill9>or the guile function like specification->package <pkill9>but if you write a file that imports the module that the package is in, it will be available as a variable <guixy>Ok. I think I understand. hidden-package will export the variable, but hide the package. That way other packages can refer to it. Correct? <guixy>Provided it's defined publicly or exported <guixy>Ok. When I'm debugging a hidden game engine, I will use guix build -e (@@ (gnu packages module) variable) <guixy>I thought guile 3 stopped supporting @@ <samplet>Hello all! Is anyone else having trouble with GDM? I pulled recently and now it won’t start. <samplet>X tells me that “Switching VT failed”. <samplet>I checked the logs and I see that someone brought it up and worked around it by using SLiM.... <civodul>could it be stateful bits in /var/lib/gdm or similar? <samplet>civodul: Good guess. I’ll try deleting it. <bricewge>I broke gnome-shell 2 days ago with the libnma patch series <bricewge>e24d311f9c4787333b136b8bad504ae74f679259 fix gnome-shell <samplet>bricewge: Thanks for letting me know! I’ll try that. <samplet>(Deleting “/var/lib/gdm” did not help, BTW.) <bricewge>samplet: It's expected, if you look in the logs there should be a mention about libnma missing <samplet>bricewge: Thanks! Pulling and reconfiguring did the trick. <bricewge>samplet: Good to ear, sorry for the inconvenience. <hulten>bricewge: How did you mount the NTFS filesystem? <hulten>I have it now in my system profile ($ which ntfs-3g <bricewge>Then I tried udiskie to automount it and it worked too <hulten>Something went wrong in hexchat when pasting more than one line. <hulten> /run/current-system/profile/bin/ntfs-3g <hulten>Right, starting with a '/' doesn't work, of course :-) <hulten>In any case, $ ntfs-3g /dev/sdb1 /mnt <hulten>Unprivileged user can not mount NTFS block devices using the external FUSE library. <bricewge>hulten: 'udisksctl mount -b /dev/sdX1' worked as a user <hulten>Yes! Thanks! But why ntfs-3g need to be installed as a system package? <hulten>Neither ntfs-3g nor udisksctl are setuid-root. <bricewge>Try 'sudo LIBMOUNT_DEBUG=all mount /dev/sdX1 /mnt' <bricewge>By default 'mount' don't know how to mount a NTFS file-system in RW mode only RO, so it tries to look if a 'mount.ntfs' binary is present on the system to be able to mount in RW <bricewge>'ntfs-3g' provide /run/current-system/profile/sbin/mount.ntfs <hulten> /run/current-system/profile/sbin/mount.ntfs ... found <hulten>So, this means that when setuid-root is done, it gets a 'full root environment', i.e. similar to 'su -' instead of just 'su', right? <hulten>(Assuming this hypothesis is right) I did not expect this; rather that setuid would imply a very minimal action of elevating the permissions. <bricewge>setuid is just a flag on file that means that when executed it will be done as the owner of the file <bricewge>So when you execute /run/setuid-programs/mount, it'll be done as root instead of your user. <hulten>In what or whose PATH does it search for mount.ntfs? <hulten>'mount' will try to fork a process 'mount.ntfs'. <hulten>I understand that a PATH is set when the user runs something for hirs shell, <hulten>but I actually don't know what happens when one process forks another "on its own". <bricewge>I don't think it use a environment variable, as it is provided by the user you can't trust it otherwise she could run any arbitrary binary as root <guixy>I changed freedink-engine to a public package to make debugging easier. <guixy>When I call `guix build --source freedink-engine`, extract and cd to it, then `guix environment --pure freedink-engine` `./configure` and `make` it is successful. <guixy>But when I call `guix build freedink-engine` it fails. <guixy>guix environment --pure gives an environment almost identical to what the builder has, correct? <lfam>guixy: It's close but not close enough to say it's "almost identical" <lfam>It doesn't limit the programs in the environment from going outside of it <lfam>All that it does is change environment variables <lfam>If you want to try to reproduce the build environment, you should try `guix environment --container` <lfam>Again, it's not quite the same, but it's closer than --pure <lfam>Ultimately, you can't reproduce the build environment exactly, at least due to idempotency <guixy>Is there a way to get a list of commands guix calls? I want to see what's different about the build phase. <boeg>I have installed gcc-toolchain - shouldn't I have a cc executable available? <boeg>cbaines: i just thought there would be a symlink called cc pointing to gcc <cbaines>I've heard that Guix just provides the things upstream does, which is gcc in the case of gcc <apteryx>I suspect our nfs server (in Guix System) may not work with U-boot. Is someone using this combo successfully, e.g., expose a rootfs via the Guix System NFS server and attempt to nfsboot an embedded board with it? <bricewge>apteryx: IDK but I'll start looking into it in the coming days for GSOC. <rekado>turns out GHC keeps references to GCC and binutils because it wants to make sure GCC is available before compilation of Haskell code. <rekado>so we can’t just remove the references without also patching in a new discovery mechanism <ryanprior>civodul makes the point that it's not hard to alias cc=gcc in your own environment, if you want that. <ryanprior>My angle is that, coming from an environment where I'm used to having cc available, that's an option I'd like to be able to specify in my manifests. Whether it's appropriate to create a bunch of packages which just have a symlink is dubious, so I'm interested in hearing superior solutions (that preserve reproducibility and don't involve everybody having to manually manage their own aliases) <Kozo>nckx: Thanks. Pidgin is not producing an error anymore =] *mbakke finally migrates off OfflineIMAP, feels good <apteryx>bricewge: nice! I'll keep an eye on your work :-) <rekado>I tried mbsync but I’m worried about losing emails <mbakke>rekado: I found that I don't really need two-way synchronization and went with "mpop" <rekado>I have this comment in my init.el: “Rename files when moving. This is NEEDED when using mbsync or else there is a problem of duplicate UIDs!” <mbakke>I was able te reuse all maildirs <ryanprior>I've been using mbsync with protonmail bridge & and it fails to download emails. 😥 <mbakke>I also activated a gnu.org address that I'll be migrating the Guix lists to <efraim>I use mbsync. Spent about a week downloading but not syncing deletes before I felt satisfied i had it right <mbakke>ryanprior: first, become a maintainer of a GNU package :P <boeg>its like in the environment of the build of the package with the build file, cc not existing is not a problem, where as if i do it manually, it is a problem <cbaines>boeg, maybe CC=gcc is being passed to make, or something like that <boeg>cbaines: yes, maybe. Another theory i have is that I am installing the common lisp libraries via quicklisp, where as that build script installs them via guix package -i, so maybe they have patches that make them work properly on guix when they are installed like that <apteryx>bricewge: I think I just need to expose a flag for listening to UDP, as by default rpc.nfsd only listens to TCP <apteryx>and I see UDP requests from U-Boot's nfs command <wklew>Hi Guix, is there a graceful way to reconfigure a tty-only system into a desktop system and vice versa? analogous to switching between run-levels or systemd targets <lfam>wklew: I would use a TTY besides the default one <lfam>You can switch with CTRL+ALT+Fn <wklew>lfam: but to reconfigure a system with no desktop services into one which has desktop services? <wklew>I ran into issues when elogimd did not restart <bricewge>apteryx: Thanks! That's a strange quirck <lfam>In general you'll need to reboot after reconfiguring <wklew>I would be unable to issue any commands, and attempting to log in would result in an error. I was forced to do hard reset and that screwed me up pretty badly <alarat>Is there a nice way to run a system config container (ie. guix system container some-config.scm) as a service? <lfam>wklew: I recommend trying from another TTY, as root, and rebooting immediately after <wklew>lfam: I did that as well, all TTYs give same errors <lfam>What do you mean by "unable to issue any commands"? Specifically, what happened?> <wklew>I would try to reproduce it but last time it left me with a broken grub <wklew>So, I reconfigure from %base-services to %desktop-services, lots of output on my tty about services being started and failing, including elogind, and then issuing any commands gives me errors about /proc or something, can't quite remember <wklew>is there a way to defer service changes from a reconfigure to next boot or something? <lfam>That's how it used to work! <lfam>But now there is some effort to handle it "live" which obviously doesn't work so well <lfam>I have to go but somebody else should be able to help. Otherwise please make a bug report at <bug-guix@gnu.org> <lfam>Maybe somebody knows exactly what to do but otherwise it will be good to know what you see when trying to run commands <wklew>thank you lfam, I'll look through logs to see if I can find some records <lfam>Usually one can do some /proc trickery <apteryx>at least for the U-Boot 'nfs' command to load something into memory from NFS <apteryx>I'll see shortly if the kernel is now able to boot from NFS as well <bricewge>apteryx: Nice! Which board are you using? <wklew>hmmm... do I need to specify %base-services in addition to %desktop-services? <bricewge>wklew: No %desktop-services already contains it <wklew>bricewge: ok, still can't figure out why switching from one to the other leaves me in a broken state <apteryx>bricewge: the board is a Techlologic System TS-7970 <apteryx>note that U-Boot from master uses NFS version 3. The one on my board from 2015 only supports version 2. <bricewge>I guess all those settings are really board specific, was it provided by the vendor? <efraim>enlightenment is still complaining about no acpid service, I guess I should start writing one after the 0.24.0 update <mbakke>efraim: yay! I've been meaning to do that forever so that hypervisors can shut down and reboot Guix VMs... <apteryx>bricewge: Except being outdated, I don't think there's anything too specific here. I'd think the 'nfs' command of U-Boot would require UDP even when using a current U-Boot. <mbakke>speaking of hypervisors, would anyone be interested in a native Guix VPS service? <efraim>I've seen people express an interest before <vagrantc>hrm. building guix with 32 cores only gets me about 6-7 used cores <sneek>vagrantc, you have 1 message! <sneek>vagrantc, bdju says: I did get Sway working. I still have to get a gtk theme working and a few other things now, though. <vagrantc>hrm. building guix 1.1.0 on debian with guile-3.0 gets a lot more test failures ... i guess i'll need to go with guile 2.2 instead <vagrantc>i thought 1.1.0 was already switched over to guile 3? <mbakke>vagrantc: what kind of failures? <mbakke>I haven't seen any tests behave differently on 2.2 vs 3.0 <vagrantc>might be unrelated to the guile version then ... but so far most of the test suite has failed with my last attempt <vagrantc>actually had it down to a small number of failures last time i tried... *vagrantc tries again with guile 2.2 ... <mbakke>vagrantc: do you get any hints from the test logs? sounds like something is very wrong somewhere <vagrantc>hrm. 2.2 fails all over the place too ... hrm. <vagrantc>my last attempt was documented in #40650 ... but the toolchain has evolved a bit since then ***jonsger1 is now known as jonsger