<kirisime>You inch your cursor closer to the 'done' button and it grays out if it notices you approaching... <leoprikler>Now just clean up your inputs and you can earn yourself a copyright line. <kirisime>leoprikler: But I feel like I'll never get to mailing my packages to guix. <kirisime>I have my mail already, my work is done. <leoprikler>Sounds like one more reason to write a patch for that. <kirisime>So how I should do this is, get the guix repository, set up a basic development environment for it, add my package, test it in the pre-inst-env, then mail the diff to the right mailing list... <leoprikler>it's just slightly more difficult than maintaining a channel <pkill9>oh, they already exist on my machine <pkill9>i am running a slightly older kernel (5.0.10), maybe updating to the latest could fix the error ***butleria1 is now known as butlerian
<leoprikler>what exists on your machine? those folders or the cgroups? ***vagrantc_ is now known as vagrantc
<leoprikler>which doesn't necessarily imply that the cgroups do, though <leoprikler>perhaps updating the kernel will work, but does guix build the kernel with cgroup support? <pkill9>can't even reproduce error after restart, still doesn't run though <leoprikler>guix' kernel flags are known to be minimal, so you might have a look at that <pkill9>i can't find any flags mentioning 'cgroup'' <leoprikler>where do you find those flags in the first place? <pkill9>i was searching in the file that is opened with `guix edit linux-libre` <leoprikler>good idea, but you have to take kernel-defaults into account as well <vagrantc>it should install the .config into /gnu/store/*-linux-libre-VERSION/ <nckx>Even better: /run/{booted,current}-system/kernel/.config. <nckx>No need to waste unswappable memory on a gzipped blob with Guix! <nckx>leoprikler: 'ndeed, that was a less genius choice. <stikonas>am I right in assuming that guix system init can only preinstall system packages and not any user packages? ***catonano_ is now known as catonano
***bsima1 is now known as bsima
<efraim_>ramajd: try adding 'netdev' to the supplementary-groups <efraim_>although it seems possible that our cmst is built with the dbus group set to network. i'm going to see about updating cmst to the latest release ***efraim_ is now known as efraim
<nixo>Hello Guix! Does gimp "open location" works for you on guix? On other systems I can use it to directly fetch images from a given url <ramajd>efraim: unfortunately adding the 'netdev' to list of supplementary groups of my current user wouldn't solve that issue! do you have any suggestion about how I can perform this group change for dbus? I can test and submit a related patch for it. <efraim>i'm not sure how to change it for dbus <efraim>i do know that my user is in the netdev group, and i'm using econnman on enlightenment and it's working <jayspeer>I stumble here once again with my everoccuring problem <jayspeer>So... Once again I am struck with "failed to install locale" message. glibc-utf8-locales is installed and $GUIX_LOCPATH is set <zacts>Is anyone actively working on packaging KDE? <valignatev>jayspeer: Haha, hello :) Welcome to the club. I happened to fix this one for me thanks to folks here explaining me how exactly guix works. Do you use guix on a foreign distro, or is it Guix System? Does this error also happen under root? What's your login shell (bash, zsh, fish?) <valignatev>Btw, I managed to break my INFOPATH somehow, and now stuff like "info guix" shows "info: No menu item 'guix' in node '(dir)Top'" <ramajd>efraim: I'm trying to have a test with enlightment and econnman to see if the issue resolves for me. I also tested with gnome+enconnman and received same error <efraim>ramajd: I saw in the bug report you tried with econnman, not just with cmst <valignatev>Looks like something set explicit INFOPATH in my ~/.guix-profile/etc/profile: export INFOPATH="${GUIX_PROFILE:-/gnu/store/i5rs6cwnzxc1kjq9b4fs9avcvxnbrsfc-profile}/share/info${INFOPATH:+:}$INFOPATH". <valignatev>And since there wasn't anything in my INFOPATH before, it ended up having only this single entry <jayspeer>valignatev: It'n on foreign system, fedora more specifically <jayspeer>It happens under root, both my user and root use bash <valignatev>jayspeer: What do you have under your GUIX_LOCPATH dir? <jayspeer>GUIX_LOCPATH: /home/jayspeer/.guix-profile/lib/locale <valignatev>One of the possible reasons is that guix is linked to the different glibc than locales. One other thing you can try is to enter the "guix repl" and execute (getenv "GUIX_LOCPATH") and see if it matches your GUIX_LOCPATH <jayspeer>it's set to /home/jayspeer/.guix-profile/lib/locale:/gnu/store/n79cf8bvy3k96gjk1rf18d36w40lkwlr-glibc-utf8-locales-2.29/lib/locale <valignatev>Hm, if you just call "guix" without any parameters, do you get warning about locales? If no then it looks like only guix daemon doesn't have access to them and you may want to try to update guix and/or locales under root <jayspeer>I did get a message when running guix pull as root: substitute: /gnu/store/q19l04vd2za80mk1845pz7r8cz29qk43-bash-minimal-4.4.23/bin/bash: warning: setlocale: LC_ALL: cannot change locale (en_US.utf8) <jayspeer>I'm not really sure what fixed the issue but it seems to work now. I defined LC_ALL in /etc/locale.conf on fedora install. Now guix only complains that bash can't change locale LC_ALL. No warnings other warnings are present <valignatev>Interesting, I don't really know what that means. My /etc/locale.conf under Arch only defines LANG. I was about to ask whether your bash-minimal-4.4.23/bin/bash is linked to the correct version of glibc ¯\_(ツ)_/¯ <nixo>Hi! Does guix environment --container support a rw overlay fs over an "--expose"d folder? <jayspeer>is it possible to install specific package version in guix? f. e.: python-2.6 for some arcane reason like creating script targeted for centos6 <jayspeer>I know it's possible on gentoo and other systems <bdju>I believe it's possible, but I haven't done it myself. There's a way to install specific commits of things. <jayspeer>I know I can overwrite package manifest. But can I run, I dunno, something like this: guix package -i guile-2.0 <jayspeer>huh. Can you show me how? I haven't figured it out :( <bdju>sorry, my first response still applies. I'm pretty sure I've heard of this being a thing but I wouldn't know how myself. I'm skimming the manual but nothing is jumping out at me so far <bdju>maybe search old irc logs <valignatev>jayspeer: General syntax for this is guix package -i packagename@packagever. Like guix package -i python2@2.6, but since 2.6 is not available in guix, you won't be able to do this <valignatev>But for something that guix has it is entirely possible. Like, for example, guix package -i guile@1.8.8 <jayspeer>valignatev: Thank you! that's very helpful. I've only given python as an example, so no worries. <jayspeer>I tried it. I cannot have both guile 2.2.6-1 and guile 1.8.8 in my profile. Installing one overwrites the other. Can I have both, with different links to each one? <valignatev>Right, you can try to install to the different profile <valignatev>Also, regarding versions that are not available in guix, you can try --with-source option <jayspeer>This is helpful, thanks. However I'll be needing two different versions of a package for a stupid project. I was planning to create a profile which only contains packages that I need and that includes both of those package versions. I guess I can just copy package manifest and rename it or something like that. <jyscao>Hi guys, I submitted a patch for a new package over a week ago. So just wondering, when can I expect it to be reviewed? <nixo>jyscao: which package was it? <mehlon>does anyone run guix on their desktop/laptop? <mehlon>cause I actually installed it on my PC but I must say the default install is kind of bare bones <jayspeer>how can I import package from nix? Info page is rather lacking <mehlon>kirisime: what browser do you use? and what fonts? <mehlon>icecat doesnt render numbers for me <pkill9>here's an idea of a guix dependency hell: packages made up of package inferiors, which are used in other packagea <mehlon>the only dependency hell I have so far is that theres too many of them! <kirisime>mehlon: ungoogled-chromium, and unifont might be the only font I've installed myself from the main channel. <mehlon>oh, I usually use firefox bc of snyc <mehlon>sync*, I like to easily share bookmarks and passwords <mehlon>also bootup to desktop seems to take quite a while <kirisime>Oh yeah, something I've been wondering about to that effect. <kirisime>Populating /etc/ can take forever, what makes it happen when booting? <raingloom>hmm. gpodder ui is bugged in gnome but works in i3. insufficient wrapping? <raingloom>nix version works. let's see how they differ. ***ng0_ is now known as ng0
<jayspeer>I'm trying to build new package (terraform). However I get an error: "guix package: error: cannot install non-package object: #<unspecified>" <civodul>jayspeer: you ran "guix package -f file.scm", right? <civodul>in that case, your file needs to return a package object <civodul>so if you wrote: (define-public foo ...) <civodul>you need to add "foo" on a line at the end <jayspeer>thanks, now I get some other error - so signs of progress :) <mehlon>can I use grub without it changing my efi variables after every reconfigure? <mehlon>I already have systemd-boot from nix and I'd like to keep that <dejanr>hi all, i have been using nixos for few years now, and would like to check guix, anyone know of a nice comparison between two, for example, do you have configurations, sandboxes, etc <dejanr>mehlon: thanks, how do you provision your machines, do you have github repo i could take a look? <mehlon>I currently only have a few nix configs on a secret gist <mehlon>haven't really figured out how to properly share settings <mehlon>one thing however you might find is that guix can be slow <mehlon>guix pull can really take a while, and also the default web browsers aren't always available as substitutes(?) so it can take a while to compile icecat or ungoogled-chromium <dejanr>can you sandbox projects like with nix? <stikonas>dejanr: guix builds everything in containers <mehlon>idk probably? Guix is based on the nix codebase is all I know <dejanr>but i am just not happy with nix language, even after few years of using it, i forget syntax, and there so many different ways to do the same thing <mehlon>that is definitely a disadvantage, for me I sometimes forget the semicolons <nckx>dejanr: I don't know what nix-shell looks like nowadays but we have ‘guix environment --container’, which might do the same. <mehlon>also I heard that this chat is where you request packages so umm, if someone's willing to add systemd-boot support to Guix... :) <mehlon>this is roughly similar to my experience using it so far <mjw>civodul, solved it by supporting git smart http, now you can also use: <mjw>apparently libgit2 doesn't support the dumb git http protocol. The smart protocol is a lot faster and efficient though. So I am glad I finally added it on the server. <nckx>From https://zge.us.to/guix.html: ‘My existing GnuPG scripts don’t work as expected, since it cannot find pinentry, even though it is installed’ — I wonder what they mean. Something's not being said. *vagrantc found getting pinentry to work with GnuPG on guix very challenging ... so it's unsurprising <jayspeer>so I build terraform 0.12.10 as a guix package... should I publish it? it's under Mozilla public license <mjw>It is also challenging on other distros. Especially if you want to combine it with gpg-agent. *mjw hugs his little nitrokey <mehlon>what category would I put systemd-boot in? <mehlon>in the wishlist.... system -> bootloader? <nckx>mehlon: Sure, you can create it if it doesn't exist. <mehlon>but with icecat sync not working I think I'll just put this off for now <nckx>jayspeer: Yes please? Why not? 🙂 <nckx>mehlon: I'll add it for ya… <nckx>mehlon: I'm not familiar with systemd-boot at all though. Does it require supporting systemd init at all (we currently don't)? You're immediate problem is just that ‘guix system …’ changes EFI vars, correct? <mehlon>yes, grub likes to reinstall itself every time <mehlon>systemd-boot does not require any other part of systemd <vagrantc>mjw: i seem to recall the gnuk on debian "just works" with gpg-agent without any fiddling. <mehlon>it used to be called gummiboot, a uefi-only bootloader <mjw>vagrantc, on Debian yes, but I was/am (still, sorry) running centos ***reepca` is now known as reepca
<mehlon>on nix the bootloader is only installed with nixos-rebuild --install-bootloader switch <mehlon>installed to efi vars/mbr that is <efraim>Sneek later tell jayspeer I don't know of a package definition for python 2.6 but I have one for 2.4 if you need <nckx>mehlon: Inversely, Guix has a --no-bootloader option. I just don't know if it skips too much for your use case. <nckx>mehlon: Oh, I remember it as Gummi. Didn't know it got monolith'd. <dejanr>so guix is not using systemd, what's the advantage of dmd over it? <mehlon>dejanr: shepherd is written in guile scheme so it can reuse code or sth <nckx>What mehlon said, so it's very easy to integrate with Guix System, and it works. Forcing Guix services into the systemd straightjacket would be a lot of work but not impossible (I guess). systemd makes a *lot* of assumptions about systems. <mehlon>nckx: --no-bootloader is good enough, although it's too bad it doesn't update grub either (?) <nckx>mehlon: I've added your wish as ‘Support booting from pre-existing systemd-boot, i.e. "grub-install --no-nvram".’ <nckx>I guess that's what you mean above: update GRUB, but don't install it? <mehlon>I manually added an entry in my systemd-boot to chainload GRUB so I ofc dont need grub to hijack my bootloader every time <kirisime`>nckx: There's a wish for tewi font on that list and I have a package. <nckx>kirisime`: Great! Don't hesitate to post it to guix-patches@gnu.org, even if it's not perfect. Assuming it's free and built from some kind of source. <mehlon>is there some way to check which substitutes are available? <nckx>mehlon: Look up ‘guix weather’ in the manual. <nckx>Jinkies, I knew and completely forgot about the #t hack. raghav-gururajan, thanks for writing that, & please consider adding it to the cookbook or whatever's appropriate. <janneke>jje: your inner remove has an extra (network-manager-configuration ...) argument <nckx>jje: That indentation is correct, you have written (remove (lambda …) (remove …) (name-service-switch …)), while remove takes 2 arguments. <nckx>See (guix)Using the Configuration System. <bandali>leoprikler, i tried with the updated exec-path patch you sent me, and guix gc --references for emacs-next still showed gcc :/ <bandali>/gnu/store/2plcy91lypnbbysb18ymnhaw3zwk8pg1-gcc-7.4.0-lib <bandali>ah, it's gcc lib. is that different? <str1ngs>bandali: gcc lib is an output of gcc. could that be why? <leoprikler>compare existing emacs, which also references gcc lib <mehlon>I'm installing gnunet and the 'check' phase seems to take a very long time <bandali>if that's fine, and all else is well, i'll send in the patch today :) <bandali>leoprikler, btw, how did you generate that updated patch? it wasn't a git patch was it? <leoprikler>The existing patch is not more than a commented diff either, so I thought this was slightly more appropriate. <mehlon>how well does the nix package work on guix? <raingloom>mehlon: mostly ok. i've had trouble with some graphical applications. but i just used it to launch gpodder, which is broken in guix. so. ymmv. <mehlon>since icecat is going to take ages to run I'm trying to install firefox in nix or flatpak <mehlon>although I guess I could simply be using my existing nixos.... oh well <mehlon>can I run flatpaks on guix system? <mehlon>simply running flatpak gives me a segmentation fault <sneek>Welcome back g_bor, you have 1 message. <g_bor>I have received response from the nginx devs <jonsger>hm. There seems to be no way around a glib update for gnome 3.34 <erudition>mehlon: Yup that happened to me without root <jonsger>g_bor: I did not instruct sneek to tell that to you :P <g_bor>The problem observerd on berlin is really a bug, and it was really in the commit I mentioned :) <g_bor>I got quite near to the reason also. <g_bor>They are looking for a proper fix, but there is a configure flag that bypasses this on systems without kqueue. <g_bor>As far as I know we support kqueue nowhere, is that correct? <raghav-gururajan>sneek later tell mehlon : flatpaks require FHS, which is not supported in guix yet. I have seen some conversations regarding a service for that. If you are looking for featured web browsers, you can use icecat and ungoogled-chromium. If substitutes are unavailable and you do not have high CPU power, you can wait for the build farm to finish building those substitues. <erudition>I have gotten flatpak to work, no problem. Just required to be run as superuser. <erudition>I use it for packages like Pithos, not something available in guix otherwise <erudition>so it doesn't matter how much cpu power or time I wait lol *raghav-gururajan was about to type FSH (Follicle Stimulating Hormone) instead of FHS (File Hirearchy Standard). This what happens when someone is into both biological and computer sciences. xD <erudition>so far that's the most compelling use of guix, yeah. High Performance Computing <kirisime`>And I'm here only because I didn't want a partial upgrade to break half of my system again. <optima>Hey, I want to start taking emacs, but I can't get modulea to work <optima>According to the docs modulea should just work™ <optima>But I installed telega, started emacs and it cant find the telega command <optima>The files arw there tough, what am I missing? <str1ngs>optima: you may need to require that module <optima>Should that be in my init.el? That's still more or less empty <leoprikler>Then head over to paste.debian.net and paste the contents of the help buffer. <leoprikler>so you installed both emacs and emacs-telega in your user profile? <optima>I alao installed magit and use package, but both do not seem to work <leoprikler>Are you launching Emacs from the command line or from your WM? <optima>Cli, but with both fish and bash it wont work <nckx>g_bor: \o/ Thanks so much for taking care of this from start to (hopefully soon) finish. <leoprikler>Okay, then log out and back in, and telega should work <optima>Omg as always the turning it off and on again dis the trick <leoprikler>Well, the "real" trick is sourcing ~/.guix-profile/etc/profile. <leoprikler>But there's no way of doing that in the GNOME Shell <leoprikler>(Anyone interested in learning JS to write an extension for that?) <optima>Yeah I'm just on gnome because there's so much new stuff on guix and I don't want to get overwhelmed ^^" <str1ngs>I not confident this is exclusive to gnome. this issues probably with most DE's <leoprikler>Perhaps there is one minimal WM that lets you mess around with its internal variables. <str1ngs>I have a potential fix for this. but it's non standard approach and use LD_PRELOAD <str1ngs>it's theatrical at this point but potentially we could override getenv such that it returns the updated variable without having to restart the parent process. example bash or DE <str1ngs>but it's kinda counter to UNIX. it's just something have been musing on. <g_bor>I recently got an interesting assignment. <leoprikler>That sounds like the most evil thing I've ever read on this channel. <g_bor>I should create a fully automated guix system installer. I hope that someone on the installer writer side has some insight on how to achieve that. <nckx>str1ngs: That's so… triggering in so many ways and interesting at the same time. (And hey, Guix and really GNU itself aren't wed to UNIX anyway.) <g_bor>I could do a sell script and execute it as a service like we do it for the marionettes, but it would seem to be a better option to reuse something from the installer. <nckx>g_bor: How automated is it? <g_bor>Well, what I hope to do is to incorporate simple service running a script that is premade on the first boot, to configure the system to a state where I can guix deploy to it. <g_bor>one thing is that it include partitioning the local storage. <g_bor>Something similar to a pre-seeded hands-off installer for debian. <nckx>Damn, never used Guix deploy. <str1ngs>leoprikler: almost as evil as having to restart bash or DE because you did guix install :) <nckx>It's trivial to tweak install.scm to start SSH at boot, and to include authorized keys, so isn't all that's missing a (very site-specific) one-shot partitioning service? <leoprikler>str1ngs: My personal suggestion would be to try using setenv from gjs <g_bor>And even more, the partitioning schemes provided by the install would suffice. So I hope that I can reuse them. <nckx>Who'd have thought that Unix would be too functional and we'd need a way to set! environment variables. <g_bor>In the first trial setup I would only need a simple single partiton for all on the local hd. <nckx>g_bor: Ah, now I get it, I didn't understand why you wanted to keep the installer at all. <str1ngs>leoprikler: that's a good idea for gnome. but doe's not resolve other DE or processes. either way it's good to consider away to make this work better. <g_bor>so my first thought was that the backend part of the installer might have a nice api to that for me :) <str1ngs>right and you could potentially use dbus to ask gnome shell to set the environments <str1ngs>probably gnome shell is already a service. so that might not be hard to do <str1ngs>service in the context of a gnome application service <str1ngs>leoprikler: atleast with this approach people might only need to restart emacs and bash <leoprikler>It doesn't seem as if /org/gnome/Shell is particularly useful <str1ngs>leoprikler: maybe not on it's own. but you could add service methods potentially <str1ngs>in the case of unique applications new local instances emit signals on the main instance. so it's just a matter of probably creating a custom signal. though I don't know how you would do this with GSJ. I'm use to C myself <leoprikler>Okay, so the good thing is, we can get GNOME Shell to execute arbitrary GJS over D-Bus <kirisime`>Speaking of GNOME, what's the state of guile-gnome? <str1ngs>leoprikler: well when you word it that way. I don't know if it's a good thing :P <g_bor>ok, so the partitioning interfaces are on the intermediate levels... the newt frontend does a lot of things. <leoprikler>so we need to get a series of (format #f "imports.gi.GLib.setenv(\"~a\", \"~a\", true)" var val) <leoprikler>kirisime`: it is pretty outdated and limited to Gtk-2 among other things <leoprikler>which we can probably do easily in any language except JS <leoprikler>So we can finalize this to a three-step process: <leoprikler>1st: launch a clean shell and source /etc/profile <str1ngs>you could sub shell varibles, then format them as json <leoprikler>2nd: Synthesize D-Bus queries to org.gnome.Shell.Eval <leoprikler>(Obvious step 1.5 extract the variables from the shell) <kirisime`>leoprikler: If you do this whenever I install software, will the software appear in the launcher without re-login? <leoprikler>You don't need to relog for that even right now, M-F2 r should suffice <str1ngs>leoprikler: alternative you could use something like systemd's set-environment but using a shepherd user service instead. then you can use guile directly and probably avoid a sub shll <str1ngs>I kinda do this now, where shepherd launches my WM <leoprikler>In my own setup, shepherd is launched by GNOME, not the other way round. <str1ngs>I use xfce4-session to start my user shepherd. same as you. I guess there would be more flexibility if the sessions where launched by shepherd <str1ngs>I need to think more on this, I think it has potential though; <leoprikler>Writing this as a herd service (or action) sounds nice, though. <str1ngs>maybe we could start with say a shepherd X session. then users can use ~/config/shepherd/init.scm to start there X services <leoprikler>Btw. how would you avoid the shell? (use-modules (guix ...))? <str1ngs>maybe that would include a gnome session etc <str1ngs>I guess you would avoid the sub shell by using the same API as what generates ~/.guix-profile/etc/profile <str1ngs>sorry not familiar with that side of guix yet. <leoprikler>well, the thing is, that you don't have only ~/.guix-profile <leoprikler> /etc/profile sources the system profile, ~/.guix-profile and ~/.config/guix/current <str1ngs>understandable, maybe the sub shell can not be avoided <leoprikler>also, with multiple profiles people edit their ~/.profile to source those as well <str1ngs>though the shepherd X session. is still useful on it's own I think <leoprikler>I should probably suggest a change to that that requires less editing <leoprikler>sure, because that would be WM-agnostic (probably) <leoprikler>meaning it would work not only in gnome, but also XFCE, KDE, ratpoison, etc. <str1ngs>right, and now users could have declarative user X11 services <str1ngs>leoprikler: also maybe guile-gi or g-golf could use GIO to do RPC over debus. then your are writing only scheme. <str1ngs>I guess some other RPC could be used. <str1ngs>from my understanding no. but it has the potential in the form of guile-gi and maybe g-golf <leoprikler>sure, but that sounds like hitting a nail with a tool shed <leoprikler>also, both are still experimental to varying extent, so rolling them out large-scale in guix is probably not the wisest thing to do <str1ngs>I agree, it was just food for thought <vagrantc>janneke: fwiw, have mainline u-boot + 2 patches working on the pinebook pro, though i haven't successfully booted a mainline kernel, which would require some additional patches to support <janneke>vagrantc: but those patches are probably readily available, right? *pkill9 wants anbox to work <vagrantc>janneke: pretty normal for supporting a new board <janneke>vagrantc: i can imagine, that's all pretty new to me <bandali>thanks again valignatev and leoprikler <civodul>i noticed that some commits of yours are signed by 7E8FAED0094478EF72E64D16D889B0F018C5493C but i don't have that key (it's not the one on Savannah either) <nckx>civodul: I'm not sure. It's just the RSA signing sub-key for F5BC 5534 C36F 0087 B39D 36EF 1C9D C4FE B9DB 7C4B. I've been using it for at least a year and probably more. https://paste.debian.net/1121591/ <nckx>Let's see what's on Savannah (but if it's not expired it obviously post-dates that key)… <nckx>civodul: I'm no GPG guru but <nckx>*Very* happy to know that you're actively monitoring these though 🙂 <civodul>nckx: i'm monitoring but apparently i'm overlooking something :-) <valignatev>bandali: Thanks for submitting the patch! Have you checked EMACSLOADPATH that it generates though? <nckx>civodul: What does ‘gpg --list-keys me@saibot.rg’ tell you? <nckx>Shart, I didn't want to post my mail here. Oh well. <valignatev>I'm afraid that after installing as it is, it might miss built in emacs /share/version/lisp <civodul>i thought i had imported it in a separated keyring, but had forgotten --no-default-keyring <bandali>valignatev, cheers! hmm, i did not check <bandali>i have not kept up to do date with the recent changes to emacs concerning EMACSLOADPATH <valignatev>After installing, it exports EMACSLOADPATH in the .guix-profile/etc/profile <valignatev>It's not recent changes, it's just the way version is set in the package definition :) <bandali>ah yeah, i vaguely remember you mentioning something about that variable and the package version <bandali>does it set EMACSLOADPATH to a "wrong" value? <valignatev>.guix-profile/share/emacs/27.0.50/lisp should be on this path, but it isn't because our version is generated with commit string, but it's hardcoded as "27.0.50" when it mkdirs the directory. It doesn't set the wrong value. I guess it just silently ignores the entry if it doesn't exist <valignatev>Does emacs-next works well for you after sourcing /etc/profile that exports EMACSLOADPATH? <bandali>i'll try to do that if i find some free time later tonight <bandali>otherwise, i'd really appreciate if you or leo could look into this please and help get the patch past the finish line ? :p <bandali>civodul, would you please restart issues.guix? <nckx>…and maybe document how this is done if you know how? 🙂 <valignatev>I decided that hardcoding the version of emacs-next is ok because python-next does the same thing :) <nckx>valignatev: Will this fix itself once emacs-next is just emacs? <nckx>If so, I think it's fine. <bandali>valignatev, i see. would you mind replying to my patch with that insight? if other agree, i'll send an updated patch; or whoever is merging it could perhaps apply it themselves <valignatev>nckx: It should be ok as long as the version is the same as the name of the dir that emacs creates in /share/emacs <anadon>Hey, back for a bit an in a rush. 1) Where do I look/what is supported for python versions 2) how can I put this into a singularity and a stand-alone container? <leoprikler>1) You mean which python versions Guix currently has? `guix show python` 2) You probably want some variant of `guix environment`, perhaps --container <anadon>I'm trying to rush together tooling on CentOS but have to keep in mind explicit Windows support. Has to be done and implemented by 10AM <anadon>Whelp, I'm going to commit a sin. A Singularity container from a docker image to run guix to package a universal standalone package. <leoprikler>Docker alone can not run on Windows, so that's what the Singularity container will achieve, right? <anadon>Just...I'm surrounded by madness. This is what I have to do to ostensibly support those users. <leoprikler>Oh, I know what you're talking about. I've had Windows as a development platform and I don't which of my predecessors I would have to thank for being allowed to use Emacs there. <valignatev>I'm trying to use emacs-guix, and when I do `C-c .u` (guix-devel-use-module), I get guix-geiser-eval: Error in evaluating guile expression: ERROR: In procedure scm-error: no code for (gnu packages some-package) <valignatev>my custom emacs-next that only differs in that it doesn't have exec-path patch <civodul>janneke: in commit 4e23d65056fff3953404693a83f090cb54e02600 and the previous one, you used a different signing key than the one on Savannah; could you email the list signed with the the key on Savannah, giving the fingerprint of that 2nd key, and confirming it's genuine? *civodul is in a signature-checking mood :-) <valignatev>my guess is that guix doesn't get into geiser search path <leoprikler>Is gnu packages some-package inside your guile load path? <valignatev>There's also `guix-switch-to-repl` which looks useless because it also wants geiser repl <leoprikler>(display load-path) is the closest approximation, I'd think <janneke>civodul: looking into this, that key has been my default since 2018-04 *janneke wonders what's going on <leoprikler>my emacs 26 version seems to work for files that are already part of current guix <valignatev>So I opened a file within the cloned guix repo and try to use that as a module <valignatev>I guess this doesn't work. Maybe I should enter the environment first <valignatev>Oh yeah, %load-path confirms that it doesn't have anything about guix <leoprikler>yeah, you have to prepend the path to your checkout to load-path <valignatev>leoprikler: Ok, thanks, I'll go RTFM on geiser a bit more then :) <nixo__>booked my flight for the guix days + fosdem! See you there, good night! :) <leoprikler>This appears to be a vanilla Guile issue though. <leoprikler>Geiser just sends the ",use some-module" command to Guile, which does not know where your file is. <janneke>civodul: on savannah there is an armour that i do not recognise; possibly my lost key? but i am pretty sure i uploaded the new armour back then. i have updated savannah, not sure what to do. i could send a new mail pretty much like the previous one? <valignatev>Ok, geiser manual sucks, it was definitely written by a poet. Too many text, not to many examples :( <valignatev>lmao, I complained and immediately found an answer: (geiser-add-to-load-path) <anadon>The instructions for adding the openPGP keys don't work on CentOS 7. <valignatev>NOW I'm a real programmer, go to definition works, yay1 <dejanr>hey all, what laptops are you using with guix? <dejanr>i was looking for some smaller form factor, and good linux support <civodul>janneke: oh you already sent a message about the key change before? <civodul>well, going to bed now but we can see that later