<dongcarl>First patch on my journey to getting RISCV64 support in Guix ready: #35888 <dongcarl>(I have the rest working locally but need to tidy up) <phi`>Hi, can anyone help me to find a package that installs "X"? For some reason it's missing preventing me from logging into a X session. <phi`>Or generally, is there some kind of a reverse-search function from binary-path to package? <phi`>str1ngs: gdm, slim, startx, or any other normal method. I just need the binary, I don't intend to start it myself. <dongcarl>vagrantc: Thanks! Planning on submitting the rest of the patches in the next week or so *vagrantc has done a tiny bit for riscv64 on debian <str1ngs>phi`: %deskop-services should install GDM. Are you using that as your base services? <dongcarl>vagrantc: hope you don’t mind if I prod you a little if I get stuck :-) <phi`>str1ngs: GDM (or currently slim) *is* installed, but when I enter username+password, the screen just blinks and restarts GDM/slim. <vagrantc>dongcarl: i've mostly been a cheerleader, not much more :) <str1ngs>phi`: what xsession are you using. aka desktop manager, or window manager <phi`>str1ngs: MATE. The funny thing is that it worked when I installed it, earlier today <str1ngs>building a vanilla kernel won't probably help you with gpu issues. they both use the same drivers. <str1ngs>phi`: it worked with GDM? but not slim? <phi`>str1ngs: it currently works with neither, but it used to work with GDM. <phi`>I suspect it's unrelated <str1ngs>if gdm or slim runs fine, then it's something related to your xsession I guess <phi`>str1ngs: I must have misexplained, I know exactly that what's missing is what's usually called "/usr/bin/X", but I just don't know what package provides it ***Glider_IRC__ is now known as Glider_IRC
<pkill9>oh hmm it isn't, maybe it got removed actually <pkill9>oh, no, the package is just "xinit" <pkill9>actually, that contains the startx script, xorg-server contains X <str1ngs>startx kinda does not work. it looks for X in the arg0 directory <phi`>pkill9: As far as I see, xinit (which I have installed) only provides startx and xinit <Marlin1113>getting this error when trying to use the ati drivers on xorg <Marlin1113>happened when i tried using alsa-service as well <krafter>How can I load kernel modules at startup? I am new to this. <Chaekyung>heh that's why I find Guix hard krafter - I am actually unsure though I am guessing it would be yet another /etc/config.scm entry :) <Chaekyung>krafter: though it's totally different from all other distros it's mostly possible to find the answer in the secret guix manual <krafter>Chaekyung: On a related note, there is a etc-service-type that can create files in the /etc/ dir for you. <krafter>Though I am asking because yesterday I tried changing that variable only to discover that my OS didn't load anymore. <krafter>Chaekyung: Have you used a custom kernel or changed modules before? <festerdam>Sorry for asking such a question but is there an IRC channel where I can request help on certain questions to set up Guix SD? <krafter>Though there is usually more activity here during european daytime I hear. <festerdam>krafter: Thanks. Thought this channel was aimed on development. Yeah, I'm in Europe. Have been up for quite some time now. <festerdam> But guix system reconfigure returns "invalid field specifier" at the line the "boot-entry" line. Should it be located within "bootloader-configuration"? <festerdam>Oops seems like first line's missing. I'm trying to add a boot entry. <festerdam>If I put outside "bootloader-configuration" but still inside "bootloader" I get the same error but at the bootloader line. <krafter>festerdam: Can you post a more complete form? <festerdam>You mean the config file itself? Is there a way to do it with irssi or with a browser without JS support? <krafter>festerdam: There is a link in topic. <festerdam>Should I send the whole file or only the bootloader part? <krafter>festerdam: You better ask someone else tomorrow, I haven't tried what you are doing. <festerdam>Yeah, I should probably get some sleep. I just got another question. Is there a way to continue the download process of a package midway if the previous download attempt was interrupted? Like wget's -c flag. My network isn't that stable and I often get early EOF's when downloading big packages. <festerdam>Ok, I was able to get rid of the error. Seems like there needs to be a "menu-entries" and "list" containing the entries. <festerdam>krafter: Thank you for your support. Going to test it now and hopefully (in the modern meaning) go sleep. ***akkad is now known as ober
<mfg>guix tries to replace the shebangs but says it can't find a python binary in path if i'm using python3 <mfg>No problem with python-2 on that end, but the configure script fails with python-2 because argparse is missing, adding python2-argparse to the input list does not change anything <mfg>the exact python error is: filename too long, import command not found. Faield to import argparse. python is too old ... <mfg>yes i know sry, i mean it doesn't work with python in the input list <mfg>because then the patch shebang phase doesn't find the python binary <mfg>warning: no binary for interpreter `python' found in $PATH, which doesn't make sense to me <mfg>python-2 is no problem as i said, but then the configure fails <str1ngs>maybe you need to add python to native-inputs? <iyzsong>mfg: you can try 'python-wrapper', which has 'python' point to python3. <mfg>maybe ... I don't really understand the difference between those two <mfg>iyzsong: python-wrapper does work, but the error remains the same <str1ngs>does ./configure --help . list an option for python at all <mfg>ah lol, it isn't a autotools project <mfg>the configure script is in python <mfg>i'm not sure, do i have to use the trivial-build-system then? <str1ngs>mfg: is the project a python py project? <roptat>mfg, you could add a phase before 'patch-shebangs <mfg>roptat: so running the configure script by hand before patch shebangs and then let the normal phases to the rest ? <roptat>like (add-before 'patch-shebangs 'patch-python-shebang (lamdba _ (substitute* "configure" (("#!/usr/bin/python") "#!/usr/bin/python3")))) or something <str1ngs>I'm just not sure if you can tell auto patch what version of python to substitute or not <roptat>then the 'patch-shebang will find python3 <mfg>roptat: when i use python-warpper it even does find the correct python, but the script itself stops and says python is too old it can't find argparse (it substitutes to python 3.7.0 which should have argparse ?) <str1ngs>mfg seems roptat understands the patching order better then idea. <roptat>oh, I didn't know about python-wrapper <str1ngs>mfg: does the finally script substitute the right python <str1ngs>seems maybe it is still using python 2 ? <mfg>how would i check this? <roptat>or is ./configure run with bash instead of python? <pkill9>it looks like it's running configure with bash instead of python <str1ngs>mfg: run build with -K then see /tmp/guix*/ <str1ngs>mfg: forget the actual directory format. should be guix-build blah blah <roptat>yes, gnu-build-system uses (apply invoke bash (string-append srcdir "/configure") flags) <str1ngs>but if it's shebanged properly that should be ok <roptat>so maybe replace the configure phase: (replace 'configure (lambda _ (invoke "python" "configure"))) or something similar <roptat>no, the shebang is useless if you call bash directly <roptat>it's only useful if you call ./configure <pkill9>so gnu-build-system calls `bash configure` <str1ngs>yes, that makes sense. it avoids hardcoded #!/bin/sh issues *mfg thanks everybody for helping me :) <mfg>how do i reuse the configure-flags keyword if i replace the configure step ? <rekado>bind it: (lambda* (#:key configure-flags #:allow-other-keys) … configure-flags …) <mfg>thank you i'm now getting further with in the build process. <mfg>the build wants to download additional assets via wget, but it says name or service unknown, how am i supposed to get those assets? <mfg>i can download them without problems from my normal shell <rekado>there is no network access in the build container. <rekado>add the assets to the native-inputs instead <mfg>hm, so i need to make packages for them in order to add them to the native-input list? they are just .tgz archives which are unpacked in the source tree. Or can i modify some phases to do that? <roptat>you can use an origin record as an input <vup>hi everyone, is there a general way to "override" a package in guix? (something similar to "packageOverrides" in nix) <roptat>vup you're looking for --with-input (from the common build options) I think <vup>more specific: i am searching for a way to add my own patches to xkeyboard-config (because i am using a custom keyboard layout), however xkeyboard-config is used at various places like keyboard-layout, as input to xorg-server and various other packages, so just using package-input-rewriting is not so straight forward <roptat>ah mh... you'll have to replace every occurence of the default package then I think <vup>so how would i go about doing that for keyboard-layout? <roptat>maybe there's somewhere in the configuration where you can change the xkeyboard-config package passed to xorg? <roptat>or you'll have to change the xorg package used in the config, so it used your xkeyboard-config instead of the default <vup>system/keyboard.scm just does <vup> #:use-module ((gnu packages xorg) <vup> #:select (xkeyboard-config console-setup)) <vup>so i don't really know how to override that <vup>(other than actually modifying gnu/packages/xorg.scm) <mfg>How do i modify multiple phases in a single modify-phases form? <cap>Greetings. I would like to ask how far the lvm support did go. I searched in the archives and issues but I wasn't able to find up-to-date informations. Thx <mfg>Is there a complete functino reference for Guile and/or Guix somewhere> <roptat>vup if you look at where xkeyboard-config is used in that file, you'll probably get a hint <mfg>i don't find an explanation for the sustitute* form ... <vup>ok it seems like what i want to do can be achieved with module-define!, how bad would it be to use that? <vup>which leads me to the question: is there a file like `init.el` that gets executed on every run? <mfg>thx kmicu let's see if i understood it corerctly ;D <nckx>vup: It's not ‘functional’ but that doesn't make it bad. In fact it's what I missed in ‘purely functional’ Nix. But adding a # <nckx>* #:package keyword for both in the service is probably the righter approach here. <ngz>I have a question about wireshark (packet sniffer) package. Currently, it can almost only be used as root. There is an option to use dumpcap so that any user in the "wireshark" group can capture packets. However, when activating the option, the build reports the error "chgrp: invalid group: ‘wireshark’". It in understandable, but how could we work around this in Guix? <mfg>I would add that group in my systems config /etc/config.scm and add the necessary users to it, i guess <ngz>But the problem is at the build level. <ngz>The "wireshark" group cannot be activated <mfg>afaik groups can only be added via config.scm, so if this group didn't exist before and the build tries to create it it will fail <ngz>IIUC, the package output will then depend on the groups available on the system at build time. <ngz>If you build the package in a system where there is no "wireshark" group, the permissions will not be identical. <mfg>ngz: makes sense. don't know how to solve that then <nckx>vup: Yas, if that service is doing what it looks like it's doing. Most other services provide such an option (‘#:opensmtpd my-custom-opensmtpd-package’, for example). *nckx doesn't have the Guix source handy rn. <vup>yeah, i understood that part, but about which `two` packages is both referring? only xkeyboard-config would be relevant, right? <nckx>vup: Why not console-setup? <ngz>mfg: Is there some standard, fixed, always-existing, user group that we could use instead of "wireshark"? <vup>nckx: ah, because console-setup is only used for ckbcomp (which is used converting the xkb keyboard description to a console keymap) and i assumed not many people would need to replace / override that <vup>but yeah, wouldn't hurt to add that too <mfg>hm i think the users group does always exist, but i'm myself very new to guix so it might be better if someone who knows more about this answers :D <nckx>vup: Ah, possible, does sound like a niche thing to replace. You're in a better position to judge. ☺ <ngz>OK. It might imply securities issues, tho. Otherwise, any use can sniff packets entering or exiting the machine. It might not be a good thing. <vup>nckx: right, well i agree the #:package solution feel's righter, maybe i'll try to add that once i get a bit more a accustomed guix ;) <mfg>ngz: ah yes that's right. <ItsMarlin>I gotta get my pc to use the ati drivers, get linux vanilla if that doesn't work and package the digimend drivers for guix <mfg>in which guile module is the string-match function? <mfg>Thanks. btw what does ice-9 stand for? <mfg>i know that srfi is scheme request for implementation <nckx>mfg: (The name itself is just a Vonnegut reference, doesn't ’stand for’ anything.) I also wonder whether, 9 years later, it's any closer to disappearing… <festerdam>I've a question. In "menu-entry", should the location given to "linux" be the same as the location given to "linux" in a grub.cfg or is the path relative to something? <nckx>festerdam (yum): It's an absolute path. <nckx>And a GRUB path at that: you can use things like ‘(hdX,y)’, and if your kernel isn't on the same partition as Guix's store, probably must. <nckx>So in that sense it's relative, but not really ;-) <juh>Is there already a feature request that the guix installer checks for the wifi card before formatting the disk? That would save a lot of work. <festerdam>Oh that may explain the error I'm getting. I did (linux "boot/vmlinuz-3.13.0-141-lowlatency) and GRUB would if tried to boot declare it invalid and tell me I needed to load the kernel first. Should get into grub syntax to get it working... <ItsMarlin2>hey, can i use emac's org mode to edit the config.scm file? <nckx>juh: Change the ordering of the steps? <nckx>ItsMarlin2 (I count 5 Marlins so far ;-): Sure, as long at the result is a valid Scheme file. <juh>I wouldn't have erased my old OS, if I knew that guix won't support my wifi. <nckx>juh: Sounds like a reasonable suggestion. *nckx can't read that in not-Trump's voice :-/ <nckx>juh: Could you send a mail to the bug tracker? That's better suited for such things than IRC. <nckx>I haven't used the installer successfully, but I'm a bit surprised at the order myself, in this day & age. One could want to search for something on the Web while partitioning, for example. <nckx>ItsMarlin: I don't know, I presumed you were familiar with Org-mode 😛 <nckx>The same way people use it for their .emacs, I guess. <ItsMarlin>i use it for emacs, and the init.el file is what runs the .org file <ItsMarlin>but how would i make it so the .org file is ran for the .scm? <juh>I guess there is no installation image with non-free wifi drivers? <nckx>ItsMarlin: ☝. I hope I wasn't too enthousiastic in my response. It's definitely possible (your system configuration file is just a programme, after all) but might not be as easy as I thought. I only use Org for Scheme within emacs. <nckx>juh: No, and there never will be ;-) <ItsMarlin>oh, nckx, i have found a bug when trying to install some packages with guix <ItsMarlin>i'll send you a dpaste link when i get to my pc <juh>nckx: This is sad as it limits possible users. <ngz>ItsMarlin: You could do it by tangling the Org file into a Scheme file, I guess. This is the same as storing Emacs config in an Org file. <nckx>ItsMarlin: You can send it straight to bug-guix@ if you prefer. <nckx>I'm leaving in a few minutes and won't be back for hours. <ngz>juh: Guix is free software, you can create and distribute an installation image with non-free wifi drivers. Some external channels already allow that. However, Guix, as a project, doesn't encourage this. <ItsMarlin>making an install image that works with any hardware wouldn't be too hard with the config.scm <ItsMarlin>i still have to figure out if i'll need the firmware blobs for my gpu <mfg>i'm currently in an ad-hoc environemnt with my favourite compiler and build tools, is there a way (other than starting emacs in that environment) to let emacs know about it so that flycheck and such find the necessary binaries? <mfg>How do i use a local channel? i need some packages being available via guix environment and this doesn't have a -f option ... <notnotdan>there is also the -L option for your local package definitions <mfg>Ah yes that is even better notnotdan <mfg>is there a way to name environments? that the GUIX_ENVIRONEMNT var gets set to that? <janas>Is there a guide or example somewhere of modifying existing packages in your config.scm? <janas>I'm just trying to change the default vim to add a few extra features, like clipboard support <Marlin1113>when i try to use (modules '(%default-xorg-modules xf86-video-ati))(drivers '("ati")) <tune>how do I check if a program needs an environment variable? <tune>when I have not recently installed it to get the helpful hint <pkill9>or look at the package definition which specifies them <nckx>tune: guix package --search-paths. *nckx missed pkill9's answer. <tune>hm okay so I just see one thing for minetest <tune>I had asked about a minetest problem the other day and then forgot to keep looking into it, so I just searched logs to recall what someone said <tune>12:36 < samplet> tune: Do you have to set an environment variable so that “minitest” can find SSL certificates? Many programs need “SSL_CERT_DIR” and “SSL_CERT_FILE” set. <tune>not sure why he said minitest, I checked and I definitely said minetest... anyway I'm not sure if I'm checking the right thing <tune>looks like he just left considering the "Adios" <Marlin1113>the xorg-configuration seems to be giving me errors <tune>I don't think it's recommended but it would probably work <Marlin1113>hnode says my gpu will work on linux-libre if i use the ATI drivers instead of amdgpu <Marlin1113>but it's a bummer there are no instructions whatsoever <Marlin1113>i don't wanna have to use the firmware, i want muh freedom :c <tune>ATI drivers... somehow I doubt that. could be awkward wording <tune>iirc there's "amdgpu" and "amdgpu-pro" which works on top of the former and adds non-free bits <tune>but amdgpu should be free as far as I know <Marlin1113>amdgpu is free, but still requires non-free firmware blobs it seems <tune>yeah because ati doesn't exist anymore... so I'm not seeing how it would support a fairly recent GPU. I don't know that much about this stuff, though <tune>guix often makes me wanna cry too. I've managed to learn almost nothing in my 1.5 years using it... but I still enjoy it <tune>your use-modules section looks odd to me <mfg>Err, in which ackage is the llvm linker ? <mfg>or is binutils required in guix? <Marlin1113>i'm liking guix, the issue is more the binary blob stuff <tune>yeah, I find that if you need to do anything that someone didn't already figure out for you, guix gets complicated fast <mfg>Marlin1113: Yeah, i have a quadro inside my notebook but can't use it with things like opencl and such ... that's a bit sad, but tbh i don't care because linux-libre and guix is fucking awesome :D <tune>Marlin1113: your error paste doesn't seem to have much info, maybe the build log would say more <Marlin1113>mfg: yeah, i need proprietary drivers for opencl too, but that i don't need completely. But having a rx580 that can't even do the 3d accel would be throwing the money i spent on the trash can heh <tune>it says in your error paste the path to the build log <tune>> View build log at '/var/log/guix/drvs/cm/7j07f5yg74nz33lfjin0s2pm0kd2sa-etc.drv.bz2'. <tune>You can run bzcat on it to view it without extracted it <mbakke>Hmm, `guix pull -p /tmp/foo` wrongly updates my ~/.config/guix/current in a special environment ... maybe because HOME is missing? <mfg>Marlin1113: agreed!. I haven't searched for it, but are there any unofficial channel which try to solve exactly that and just package drivers and maybe the non-free linux kernel? <reg[m]>Default installation with i3 does not include: a terminal, i3status, dmenu. Is that intended? <mfg>which are also maintained (relatively) <mfg>reg[m]: i think the example in the docs states them explicitly, so you might need that too <Marlin1113>i did see some unnoficial channels, but i wanna try all i can before resorting to that <mfg>i'm currently in an environment and executed a program which says it couldn't load a shared library, looking into the profile folder under /gnu/store the library is linked. what else could influence this ? <mbakke>mfg: Due to a longstanding bug in Linux-Libre, it's not possible to use a custom (free or not) firmware without also changing to the vanilla Linux kernel. <Marlin1113>i also wanna stay on libre because i don't like how the vanilla is going <mfg>funny thing is clang links against the lib without problems so is able to find it but the resulting binar doesn't ... <mfg>Which package contains ldd? isn't it in binutils ?! ***Server sets mode: +cnt
<mfg>and ldd is unnecessary as readelf -d provides the same information <mfg>Marlin1113: is the config you pasted the one you use ? <mfg>why is there the drivers "amdgpu" form? <Marlin1113>but the error is happening on the add modules part <Marlin1113>i'm trying to set my gpu to use the ati drivers as the hnode page indicates <Marlin1113>it says there is a symlink somewhere that shouldn't be there <mfg>ah yes i've seen that paste. i'll be back in 30 mins or so maybe i'm creative and get an idea :D <Marlin1113>so it either is really stretched, or it has two gigantic black bars on the screen <mouldysammich>hey, i installed sway on guixsd, how can i select to use it in gdm? there isnt any options <Marlin1113>it looks for filews on /run/current-system/profile/share/xsessions too <mouldysammich>Hmm, i see, i think gdm isnt picking up .../wayland-sessions as the gnome-wayland option also does not appear to show up in gdm <mfg>Marlin1113: did you get any further? <kmicu>mbakke: please could you link to that bug? (I thought that stripping modlues from ability to load firmware is a design decision.) <ItsMarlin>i think i'll try reinstalling with my current config <mfg>do you have an idea how i could investigate the strange linking issues i have ? <mbakke>Specifically this quote: "Indeed, I became aware that some users have got the idea that blocking the loading of blobs is a feature. It's not; it's just a bug that's quite difficult to fix. The decision on whether or not to use a piece of software, be it Free or not, should belong to the users, and it's not our intent to make that difficult." <mfg>Ahhh, what is the LD_LIBRARY_PATH variable in guix used for? How are libraries added to it? is it possible/necessary to include in a package definition that it should add it's libs to that var? <ItsMarlin>weird bug: sometimes the user and root's passwords reset after a guix system reconfigure <mfg>i read that sometime before <mfg>the only fix is hardcoding the password in plaintext <mfg>which no one wants :D <mfg>and the linking thing is resolved if i manually add the lib to LD_LIBRARY_PATH ... so how does i.e. gdk-pixbuf+svg do this? <mfg>the package definition seems not to contain anything special in that regard <mouldysammich>while trying to get gnome/wayland up and running im replacing gdm with sddm, however https://paste.debian.net/1083070/ this causes problems in that its very difficult to log in the screen flickers etc due to gdm still running, any idea whats wrong with the service list? <janas>mouldysammich: Are you trying to remove gdm from the services list? <janas>mouldysammich: That's strange, your code looks pretty similar to the code that I used to get rid of gdm <mouldysammich>janas: hum, should i be doing anything special in the reconfigure? im just doing guix system reconfigure conf.scm ***apteryx_ is now known as apteryx
<apteryx>rekado: thanks for the information about timestamps <janas>mouldysammich: Yes, but for gdm I think you also need to restart your machine <janas>mouldysammich: You're sure that gdm is running, right? <mouldysammich>yeah the issue arrises when i restart, sddm flickers for a while and gdm shows up and lets me log in properly <janas>mouldysammich: alternatively you could also use a (write ...) statement to debug and make sure that gdm is being properly removed from the list <janas>or a (print ) or something similar <mouldysammich>janas: it seems to get removed, but it still starts on startup for some reason :( <janas>mouldysammich: That's really strange... <mouldysammich>im gonna try just manually specify the list of services and see what it does; not including gdm <str1ngs>mouldysammich: to use sddm I used %base-services and manually added other required services <str1ngs>I'm not sure if this is the right way, but it works <mouldysammich>strings: thats the way i currently have it now, did you have to change where it checks for .desktop files? <str1ngs>nope, it lists all my xessions. I use xfce if it matters <str1ngs>instead of modify-services you can just use %base-services in place. <mouldysammich>ah, im a fool, and forgot the (service gnome-desktop-service-type), thanks for the help :) <str1ngs>I think you can get away with xfce-desktop-services-type. which desktop manager do you plan to use? <mouldysammich>i was gonna use gnome or sway if get around to it so its all working as plannned now <str1ngs>ah yes, if you are using gnome. gnome-desktop-service-type makes more sense. though I think there is some overlap <str1ngs>maybe I'll switch to sway sometime. I'm use xfce for session management, and i3 for my window manager <nckx>str1ngs: What does ’session management’ mean (I don't know nothin' about XFCE except that I thunar sometimes). <mouldysammich>strings: sway is very cool if you do get around to it, i was using up until today and im setting up guix for the first time so i need to reconfigure some things <nckx>ItsMarlin: Yo. Mayhaps. Is ‘backtrace’ what's in ’bzcat /var/log/guix/drvs/cm/7j07f5yg74nz33lfjin0s2pm0kd2sa-etc.drv.bz2’? <str1ngs>nckx: it handles login, logout, services <nckx>I'm afraid it's truncated beyond all usefulness… I'm not sure this is ATI-driver related (actually, I doubt it). <nckx>ItsMarlin: What's ‘echo /gnu/store/z9ix8k4bxarman7sd4bwl0wmcwzny5y2*’? <nckx>str1ngs: Services? Interesting. *nckx just uses .xsesh & shep. <str1ngs>nckx: as in desktop services aka .desktop <nckx>I… have no idea what those are. <nckx>Yes, welcome under my rock, have some coffee. <kmicu>Thank you. Coffee would be nice. <nckx>str1ngs: I have a ‘polkit-service-type’ in my system.scm but it's under the ‘newfangled fd.o stuff I don't understand’ heading. <ItsMarlin>issue seems to be a symlink that shouldn't be there <nckx>ItsMarlin: Yes. The name of that truncated store file might give a hint as to which. <nckx>This is just a hack around the fact that I don't know how to set $COLUMNS in the build environment or otherwise disable truncation. <nckx>ItsMarlin: I'd just brute-forcedly delete anything like that under /etc, reconfigure, and reconfigure again. If the error reoccurs I guess we have a real problem. <ItsMarlin>just did that... the group guixbuild does not exist <nckx>…you didn't delete /etc, right… <nckx>I meant ‘delete anything like asound.conf under /etc’. <nckx>& I guess I've learnt never to use phrases like ‘everything, except’ or ‘anything like x’ in a support situation. <nckx>ItsMarlin: Try running the commands to create the build group & users in the manual. And don't close any open terminals. <nckx>It's up to you. But this isn't Windows, dammit. <mfg>ItsMarlin: i feel you, on new and to me unknown distros i also tend to just stat from the beginning instead of searching for immediate solutions :D <ItsMarlin>removing /etc was too much for me to try to fix heheh <nckx>The good news is that Guix (re)generates most of /etc; the less good news is that is does so from a file in /etc. <str1ngs>hmm I just keep config.scm in ~/config/ <mfg>i now have a local copy of the guix repo. how do ia make that into a local channel? so that i can test packages i added to it? <nckx>str1ngs: Sure, & I don't call it config.scm, but most people will follow the choices in the manual. <nckx>str1ngs: But good idea IMO. It's user data. <str1ngs>mfg: you can use file:///home/$USER/src/guix.git <nckx>And should be backed up as such. <str1ngs>nckx: mine is named by machine, and I have a common.scm I use for common configuration <dongcarl>mfg: I'd say, add the package, then `./pre-inst-env guix build <package>` <mfg>dongcarl: that's what i did <ItsMarlin>that's why i didn't freak out for deleting /etc <mfg>but a normal guix environment <package> says package is unknown <nckx>str1ngs: I used to do that, then switched to querying the hostname directly in the configuration; still not sure what I like best. <mfg>so i thought i should add a channel thus my normal guix knows it <dongcarl>`./pre-inst-env guix environment <package>`? <mfg>or is there a another better way? <ItsMarlin>i saw some dude package the blobs to a .scm file. tempting, but i won't sin! <mfg>dongcarl: this means i always have to be in that dir or write a long path :D <str1ngs>nckx: I have specific configuration per host I guess. <dongcarl>mfg: Oh you want it in your normal guix... <dongcarl>mfg: Yeah use what str1ngs said, `guix pull` on `file:///path/to/repo` <nckx>str1ngs: Mine had so much in common that it was cleaner to (when <host> …), but it's not… what it should be, yet. <mfg>dongcarl: yes exactly :) <str1ngs>nckx: I just keep common things in common.scm as variables and load that. then each host has a host.scm file for host specific <nckx>One possible gotcha for using file:// in channels.scm is: Guix will pull the last commit, ignoring any uncommitted changes you might've made. <nckx>So commit early & often ☺ <str1ngs>what nckx , which kinda defeats file:// in someways. I guess it saves pull from the interwebs atleast <mfg>hmmmm yes that's reasonable :D <nckx>I guess some find it reasonable & str1ngs does not 😃 <mfg>how do i decide in which file i should add a new package? i don't see a consistent system :D <str1ngs>it's not unreasonable. just might as well push to a remote if you are commiting :P <str1ngs>I guess you can always squash and push later this way. <str1ngs>mfg: guix search <package> will tell you what module you need to include <nckx>str1ngs: My secrets channel would disagree, and I like working off-line, but it's probably less of a philosophical decision than a side-effect of ‘channels are git repos and we pull and describe them as such’. There's no code to deal with uncommitted changes because it's not something a ‘real’ channel would serve. <nckx>mfg: To add: just stare at all the modules, and choose what's most appropriate, in the very unlikely situation that nothing is: create a new file. <str1ngs>nckx: I just ended up using a remote. since more then one machine uses the channel <nckx>kmicu: Yess! If previous Guile versions are any indication it'll take another year at least before Guix uses it, but awesome nonetheless. <str1ngs>also is there an easy way to flatten a profile. so that the current profile becomes Generation becomes 1? <nckx>No more calculating how many times I can ‘guix pull’ on one battery charge (about 12). <str1ngs>right now I roll back to 0 then delete all generations then install my manifest. <str1ngs>I'd rather just be able to rename the current generation <nckx>str1ngs: Not through Guix. Guix looks at the name of the current generation file name and just blindly adds 1. You could try renaming the profile (and adjusting any relevant symlinks!), entirely at your own risk. <nckx>Whether that's worth it depends on your goal. <dongcarl>Hey all... I got riscv64 cross-toolchain working locally, and was able to successfully build `./pre-inst-env guix build --target=riscv64-linux-gnu hello` <dongcarl>However, I couldn't build `coreutils` because apparently riscv64 doesn't have std iostream? <str1ngs>nckx: I just like to clean out my generations now and again. for now I just do guix package -S0 then -d then reinstall my profile. seems to do the trick <nckx>str1ngs: (The journey begins in build-and-use-profile in (guix scripts package), if you'd like to investigate other solutions. I'm too distracted yellin' at the Yankees.) <nckx>str1ngs: If it's just for ‘superficial’ reasons, yeah, that's probably fine. <b0f0>Aynone here using posteo as email provider, I would like to know which provider works best with Gnu Guix ? <nckx>b0f0: I've certainly seen @posteo.something on the ML, not sure if it's considered polite to share who. <nckx>b0f0: That said I don't think GNU Guix should be a hurdle in your choice of mail providers as long as they use standard protocols. <nckx>There's one that does; I forget what it's called. <nckx>No, it was one of the newer (to me), hipper, ‘privacy’-marketed ones. I really forget the name. <mfg>But a local channel with the guix repo completely shadows the default chanel? even if i consed it in channels.scm, doesn't it? <nckx>b0f0: It doesn't speak SMTP nor IMAP, you need a daemon for that, and I don't think it's part o' Guix yet. <nckx>mfg: De facto, yes, because your channel provides every single variable that upstream Guix does. <b0f0>nckx: it not part of Guix yet. I am sure, but do you know some who made the Proton Bridge for IMAP work on Guix ? <nckx>b0f0: No, I don't know of anybody currently doing so, sorry. <mfg>nckx: hm channels definitively need a priority field :D <dongcarl>that's the open source protonmail bridge <nckx>mfg: It'll be per-channel, so it won't solve your problem. It would have to be per-variable *and* not a mess, not trivial to do. <nckx>Yes, ‘someone’ should ;-) *nckx really AFK promise. <ItsMarlin>well nckx, i'm getting the error on a clean install lol <krafter>What do I need to install to have notifications` <str1ngs>krafter: which desktop manager are you using? <krafter>I use i3 as a wm and add various things from xfce when I need them. <krafter>On arch which I used before there was a command called notify-send that could send notifications. <str1ngs>I just add it to my system packages. assuming you are using GuixSD. but you might be able to just add it to your profile <str1ngs>noftify-send needs a notification daemon running to work <str1ngs>with just i3 not sure. I use xfce with i3 as my window manager <str1ngs>maybe add a exec /path/to/xfce4-notifyd in i3 config. dunno <Rnytom[m]>krafter: Are you looking for a notification daemon ? <Rnytom[m]>have a look at dunst, I also use it with i3. It's simple and fits my needs <krafter>str1ngs: I tried using xfce with a i3 as my wm, but each time I started a session I had a desktop window that I didn't know how to get rid of. <Rnytom[m]>If you choose dunst and want i3 to start it you could use 'exec --no-startup-id dunst' in you i3 config <krafter>Rnytom[m]: Yeah, I used to use dunst in the past. <krafter>I tried installing libxfce4ui libxfce4util and libnotify. Apperently it all works as expected now. <krafter>But it makes me curious. I would think that the libraries would be a dependency on xfce4, but it seems they are not. <str1ngs>krafter: under applications->Settings->Session and Startup . disable Desktop in either autostart or session. if it's in session make sure to save the new session. <str1ngs>I forget which tab it's other. think it's in session <quiliro>str1ngs: yes, i know xfce is a DE. What I am as aking is if you are talking about that one. <str1ngs>quiliro: oh I misunderstood. yes in xfce <krafter>How do kernel modules work in linux? I made an attempt a few days ago. It ended up with me needing to reinstall the OS. <nckx>krafter: Er. There's a journey from A to B that needs some explanation in between… <krafter>I tried adding a few thing to initrd-modules. But when I booted it never got past "cannot find module". <nckx>ItsMarlin: Which legacy file? <nckx>krafter: I use it to load a tonne of modules ("dm-crypt" "xts" "serpent-avx-x86_64" "ahci" …, 15 others), it definitely works. Sounds like you were trying to load modules that didn't exist. The error handling can certainly be improved but the message is probably accurate. <nckx>krafter: You should also be able to roll back unless you added them right out of the (installer) gate, a reinstallation shouldn't be needed. <krafter>nckx: How could I rollback if I can't get to a shell. I still have the installation medium. <nckx>(That all sounds trivially obvious, I know, but it should just work if modprobe $foo did.) <nckx>str1ngs: That didn't used to work for me on my previous distro, had to use ‘keystatus --xxx’, but I hope you're right. <str1ngs>ignore the pc speaker as it screams at you :P <nckx>The PC speaker shall not be ignored. Be sure to add it to your initrd-modules for early beep support. <nckx>Pretty sure my first PC called that ‘multi-media’. <str1ngs>this is preboot. it may scream if you spam keys <str1ngs>also try with shift down while booting <nckx>krafter: It's just a menu. If you get a prompt, it's because you mashed Esc. Try mashing downarrow first. If that doesn't work, try hitting Esc *once*, right before GRUB starts. <str1ngs>odds are you will have better luck holding down shift <nckx>str1ngs: That'll only work when ‘keystatus --shift’ is used; we don't. <str1ngs>ah, then maybe my method won't work at all <nckx>(If it works on other distributions it's because they add that explicitly. Patch patch hint hint.) <str1ngs>I need to get my current patches merge first! lol <nckx>krafter: It should be obvious if you get the pretty menu (‘GNU system, old versions…’ or something). Do you get a menu? <krafter>I have previously but there was always only a single alternative. <nckx>str1ngs: The only way to test for Esc in grub is to ‘sleep 1’ and test the exit status, ‘keystatus [--ctrl] [--shift] [--alt]’ returns immediately (hence suitable for timeout==0) but can only detect those 3 keys, not Esc. <nckx>krafter: Did you try that? Did it show the same error? Then you are out of luck. I absolutely agree that this is tetanus-level rough around the edges. Sorry. Work to be done… <str1ngs>maybe it's possible I used then when I had a short timeout. and not plausible with timeout 0? <nckx>‘[BUG] Do not crash the initrd on every error.’ <nckx>str1ngs: Well, you were talking about shift, which is possible. Just Esc that needs at least a second delay. <ItsMarlin>how do i use it on guix? you mentioned that's how you use yours <nckx>I wanted to point out the difference because it is seemingly arbitrary and stupid. BIOS limitations, I was told. <str1ngs>nckx: yes I think mabye I used this when I had a short delay then. and not timeout 0 <str1ngs>and I don't think I'd use timeout 0 to begin with now that I think about it <nckx>(So to be clear, the GRUB author's aren't stupid; PC firmware is.) <krafter>But can I do anything with GRUB? I thought all it did was just load the operating system, and maybe provide its own rudimentary shell to, load the operating system. <str1ngs>my grammar is so bad I've now tainted nckx's :P <nckx>ItsMarlin: Something like (unstested, because I'm still using SLiM which is Bad and Old and I really should Switch): <str1ngs>you can do all kinda of things with grub. though generally it's not needed. but for guix timeout 0 makes it hard to rollback <nckx>(set-xorg-configuration (xorg-configuration (extra-config (list "this is your Xorg.conf as a string" "here's more stringy goodness" "blah")))) <nckx>str1ngs: Mixing languages in different channels does not go me well off. <krafter>str1ngs: But I still had the installation medium, so I could still access all files on the disk. <nckx>ItsMarlin: Take that with a grain of read-the-manual, because this set-xorg-configuration is new to me too. I think I missed its addition. <krafter>I tried using set-xorg-configuration. It did not work for me. <nckx>Ah. Then I'm as inexperienced as all of you. <nckx>Never change a 3-year-old snippet of code that works.™ <str1ngs>krafter: think you can edit /boot/grub/grub.cfg and fix your timeout <str1ngs>krafter: then you don't need to reconfigure or anything <krafter>str1ngs: Ah, but the question then would be what I would do once I have access to GRUB. <str1ngs>krafter: then you can arrow down. and select another profile <str1ngs>it's the second option in the grub menu krafter . forget the actual name <nckx>krafter: Load the newest generation that works (there will be dates). If you really only have 2, then there's not much you can do. <str1ngs>krafter: it will list profiles by date. <krafter>Ok, and each time I use guix system reconfigure I get a new generation? <str1ngs>I think if someting has changes only. <nckx>ItsMarlin: I think I'll try migrating my setup to set-xorg-configuration as well (but I won't be able to reboot this evening). <str1ngs>though I could be wrong on that. but yes basically reconfigure creates a new system profile generation <nckx>str1ngs is utterly correct. <str1ngs>and generations only go away if you delete the profile and gc. so in theory you should have your working profile there. <krafter>Which I assume it did when I reinstalled... <nckx>krafter: …ran guix gc -d? :-/ <nckx>With the ‘faulty’ initrd module list? <nckx>If your first generation was unbootable there's nothing to roll back to. <nckx>Or not boot because there's a module missing. 😛 <krafter>Odd thing is, I can add it without issue with modprobe. *nckx spins up a VM with the stock Guix kernel. <krafter>*I can add the named modules with modprobe without issue. <nckx>krafter: Interesting. For example, I know that the ‘_’/‘-’ interchangeability in module names is actually hacked together with silly string in kmod(?) or elsewhere, and Guix tries to emulate its rules, but maybe the emulation is incomplete. *nckx still waiting for that VM. <krafter>nckx: I am speculating that it has to with me adding the new variables to the front of the list.