<bjc>indeed, running a home reconfigure with yesterday's (working) config from the store fails the same way <bjc>maybe something broke in the latest pull <alextee>hmm i still get shell command not found <zamfofex>alextee: What do you get if you run ‘guix describe’? <alextee>zamfofex, commit: a0178d34f582b50e9bdbb0403943129ae5b560ff <alextee>then guix package -u doesn't do anything <zamfofex>Isn’t that the one you had installed using your distro’s package manager? <alextee>./usr/local/bin/guix -> /var/guix/profiles/per-user/root/current-guix/bin/guix <alextee>i deleted /var/guix before running the installation so this must be the installation version <zamfofex>Ah. I think you updated your Guix, but are running the root user’s. <alextee>ice-9/boot-9.scm:1685:16: In procedure raise-exception: <alextee>error: supports-hyperlinks?: unbound variable <zamfofex>That seems related to a recent change by civodul. <alextee>how can i install an older version that works? <alextee>oh actually guix shell is doing something now <alextee>maybe it's just describe that's broken <alextee>running the bootstrap/configure commands now... <zamfofex>alextee: Ah, I forgot to instruct you to run ‘make’ after ‘./configure’, by the way. <alextee>ah thanks I was just about to run guix build <zamfofex>alextee: Don’t forget to also use ‘./pre-inst-env’, otherwise it will use your installed Guix (as opposed to the built one). <kitty2>Its been a while since I've reconfigured my home , and now when I try it keeps giving me "guix home: error: rmdir: No such file or directory" (While it cleaning up symlinks from a previous generation) ; when yall are done helping this person out anyone have any clue what I would even begin to do? lmao <zamfofex>Seems similar to the issue bjc was having. <alextee>ice-9/eval.scm:293:34: In procedure abi-check: #<record-type <origin>>: record ABI mismatch; recompilation needed <bjc>i think something broke upstream <kitty2>huh , how long have you been having the same problem? its been quite a while since ive tried guix home so just kinda curious if its been like that for a bit or extremely recently lmao <bjc>i can't figure out *what* broke. i'm sure there's logs somewhere, but ‘--verbosity=100’ didn't tell me (literally) anything new <alextee>says error: failed to load 'gnu/packages/bison.scm': <bjc>just today, but this is my first pull in a few days <civodul>alextee: the "guix describe" issue should be fixed now, my bad <alextee>civodul, no `make` was failing in guix shell on the guix cloned source code <kitty2>bjc: ah, thanks for letting me know ; I don't really have an email or anything set up at the moment to report anything but hopefully that gets worked out rather soon lmao <alextee>with error: failed to load 'gnu/packages/bison.scm': ice-9/eval.scm:293:34: In procedure abi-check: #<record-type <origin>>: record ABI mismatch; recompilation needed <kitty2>whatever it is it sounds like something really dumb to cause that type of issue with guix home <bjc>i wish i could even begin to try and figure out what's going wrong. tbh, it's more frustrating that all i get is an error and no other way to investigate, than it is that it's failing <kitty2>yeah lmao , its , quite bizarre tbh <zamfofex>alextee: I think Ludovic was just saying that the error related to the ‘supports-hyperlink?’ you were getting when running ‘guix describe’ should be fixed now. <alextee>ah i got confused because of the question mark, thought it was a question, nvm XD <alextee>yeah guix describe works now after re-pulling <alextee>`make` still fails inside `guix shell` <zamfofex>Do you have Guile installed without Guix? <alextee>are the precompiled .go files in home messing with the guix installation? <zamfofex>From the ‘guix shell’, when you type in ‘which guile’, what does it tell you? <zamfofex>Also, after using ‘--pure’, also try ‘make clean’ before running ‘make’ again. <alextee>zamfofex, I ran make clean and then make and now I get a different error <alextee>ice-9/eval.scm:293:34: error: commit: unbound variable <alextee>hint: Did you forget a `use-modules' form? <alextee>that shows right after this line: [ 11%] LOAD guix/tests.scm <alextee>zamfofex, nvm that was my code, i forgot i had some changes <alextee>i git stashed and it seems to proceed now <alextee>how can i make gtk-update-icon-cache available at build time? <lilyp>(it's in gtk:bin, but those caches are generated per profile, so adding them to the packages is not helpful) <alextee>(substitute* "meson.build" (("gtk_update_icon_cache: true") "gtk_update_icon_cache: false")))) <PaulePanter>flaminwalrus[m]: I found some of the packages in the archive. Thank you for the explanation. <alextee>hmm is there something special I have to do to be able to use the libbacktrace package as an input other than #:use-module (gnu packages debug) ? <alextee>because that throws errors when i try it <alextee>looks like some include order thing? <alextee>just using the module (gnu packages debug) in any file under gnu/packages blows up <anadon>I'm going to just have to have all my packages in a personal channel. I need those for other work at this point. <luchadoritos>Hello all! Whenever I run Guix home reconfigure I get an error, "guix home: error: rmdir: no such file or directory" right after it says "Cleaning up symlinks from previous home at /gnu/store/...-home." <zamfofex>luchadoritos: It seems other people were experiencing the same issue. They seem to think it is a bug. <luchadoritos>zamfofex: Thank you for letting me know! I appreciate it! Sorry for asking about a known bug. <zamfofex>It is fine! If no‐one asks about it, then it’s more difficult for it to be found, I think. <luchadoritos>May I ask, where should I look for bugs in the future? Is it in the Guix mailing system? I'm more used to GitHub issues. <luchadoritos>NVM I found the frontend at issues.guix and debbugs.gnu for Guix. Y'all have this figured out. <piethesailor>I have a simple question.. I just got into guix and like it so far, but I am curious. I use EXWM as my window manager. I recently installed emacs 28 through "guix install emacs-next". When I log into exwm, "M-x emacs-version" returns 27.2. Havent come across a solution yet and thought I's throw my question here? <alextee>I was trying to update the zrythm package but I'm stuck at an error related to including (gnu packages debug). so I just sent 5 patches to update other packages for now, will continue another time <MysteriousSilver>piethesailor: no idea, but maybe the emacs-exwm package creates a executable with emacs (instead of emacs-next) package <the_tubular>Uhm, I'm pretty sure I know what's happening but I don't know why <the_tubular>That wouldn't put it in it's path right MysteriousSilver ? <kitty2>after all the issues today with guix home people have been experiencing, I decided to explore around the guix issues tracker to see what there is there and <kitty2>quite interesting read of guix potluck, a thing from 2017(?) that has a lot of good ideas <luchadoritos>piethesailor: There's a default one (that I forget the location of) depending on your operating-system declaration. If you create a "~/.xession" file then xorg will use yours. <piethesailor>So should I run in the command line "guix repl" and then run that define-public sexp emacs-native-comp replaced with "emacs-next-28.0.50"? <MysteriousSilver>you can add it to an local channel or 'guix package --install-from-file=emacs-exwm.scm' ***sneek_ is now known as sneek
<piethesailor>so I need to make a change in ~/.config/guix/system.scm? <piethesailor>just answer that last one pls? I really appreciate the help you have provided regardless <MysteriousSilver>yes but the package defintion should be available to the package manager <vagrantc>so if something installs to libexec, but i want it to appear in PATH ... what to do? *MysteriousSilver goes away <vagrantc>can i just drop a symlink in from bin/FOO to ../libexec/FOO ? <zamfofex>Can’t you add ‘…/libexec’ to the PATH variable? <vagrantc>i mean, yeah, as a one-off, i could add that, but how would be the right way to do that in the package itself? <vagrantc>more specifically, updating libxmlb to the current version, which ships a binary in libexec that diffoscope can use... <vagrantc>i guess, diffoscope has a bunch of embedded references... <zamfofex>Yes, I think you should refer to the executable by its full path in a different package definition. <zamfofex>Otherwise you’d have to propagate the input so that it is accessible at runtime, and that’s not generally a good choice, I think. <vagrantc>the vast majority of the file formats in diffoscope don't have hard-coded values ... and there are hundreds, largely because most of the features are not strictly dependencies, but opportunistically added support <vagrantc>but so far all of the others are on PATH. <zamfofex>Is the executable in the previous version of libxmlb in ‘…/bin’? Or is it also in ‘…/libexec’? <vagrantc>don't know about the previous version, as it's been failing to build for quite some time <vagrantc>since it fixes the build failure, just pushed to master <vagrantc>last version that successfully built was mid-january <vagrantc>although it seems to also install into libexec, with a simple workaround applied (adding python to native-inputs) <vagrantc>will figure out what to do for diffoscope later <Aurora_v_kosmose>Got a package failing to build because its makefile looks for 'cc' specifically for building and doesn' <vagrantc>it might have a variable such as CC or HOSTCC or something similar that you can use to tell it to use gcc <Aurora_v_kosmose>It doesn't use a variable, I'll likely have to substitute that in during the build phases. <apteryx>rekado: uh, texlive-polyglossia has built, after some 89 packages/updates. phew! <Aurora_v_kosmose>So I have LFE building now. Given the general lack of a pre-existing codebase for other Erlang stuff, I don't think I'll be packaging rebar3. <ZhuAisi[m]>Aurora_v_kosmose: substitute 'cc' with the result of (cc-for-target), don't substitute it with 'gcc' <AIM[m]>What is equivalent of build essentials <AIM[m]>I'm compiling a emacs plugin vterm which needs them I think <ZhuAisi[m]>If you're talking about the emacs-vterm in Guix, you can use `guix shell -D emacs-vterm` to create a development environment for it <AIM[m]><Aurora_v_kosmose> "Hm, I know it includes automake,..." <- I was missing binutils as well it seems <atka>what would be a good way to hack on some system files to test things out, maybe in a repl? I want to glay around with mapped-devices.scm and get the output of a let expression <atka>guile itself plus the source, or emacs-geiser or something? <AIM[m]>I'll try out vanilla one as soon as my uni closes for summer break <AIM[m]>Till then I gotta find my way in this <Aurora_v_kosmose>Joking aside, the primary reason to avoid Doom is that it makes debugging a lot more complicated because it includes a lot of stuff & changes from default. <AIM[m]>Detecting C compiler ABI info -- failed <AIM[m]>Sys crafters has a good video tutorial on emacs from scratch <Aurora_v_kosmose>Well, I say more user-friendly with some doubt, I never actually used it myself, so I'm going off reputation and observation. <AIM[m]>I tried hello world compilation in gcc <zamfofex>If I have a program that raises SIGPIPE immediately then prints “hello”, when I execute it, it does not actuall get to the “hello”. When I use ‘guix shell’ (with or without any specified packages) then run it, it does print “hello”. Why is that? <AIM[m]>I need gcc working for some uni works, so kindly help me <zamfofex>AIM[m]: Try to uninstall it, and then ‘guix install gcc-toolchain’ <AIM[m]>Trying to compile a normal c program <AIM[m]>Like it just has printf hello world <zamfofex>(Also: For context, I’m trying to update m4, since the glibc update broke the current version. There is a single failing test, and it involves raising ‘SIGINT’ (which for some reason doesn’t end up exiting the process as expected). When I try it on the ‘/tmp’ directory left by ‘--keep-failed’, it works completely fine unless I use ‘guix shell’, in which case a different test that raises ‘SIGPIPE’ fails.) <Aurora_v_kosmose>AIM[m]: Silly question but did you re source your $GUIX_PROFILE/etc/profile ? <AIM[m]>I thought the installation did that by default? <Aurora_v_kosmose>I'm not sure. But on foreign distros if you change the state of Guix underneath the shell's nose, you need to open a new shell instance or explicitly make it aware of the changes. <Aurora_v_kosmose>You don't have much to lose, though I won't be able to explain why it works or doesn't given I don't use it the same way. <AIM[m]>I'll have to setup guix home maybe? <ZhuAisi[m]>I try `guix shell gcc`, it doesn't add binutils and glibc to develop environment now <Aurora_v_kosmose>Makefiles should actually use $(CC) which should be determined by autoconf <Aurora_v_kosmose>There's a bunch of portability reasons why. The info manual explains it. <zamfofex>sneek: later tell civodul: It seems like m4 required an update for the newest glibc version (see commit 8e99f24c0931a38880c6ee9b8287c7da80b0036b in gnulib), but now there is an strangely failing test. See <https://logs.guix.gnu.org/guix/2022-04-09.log#081848> and the few messages I sent thereafter. I’m not sure what the issue is exactly. I should take a break from this for today, though! <AIM[m]>I've tried to symlink gcc to cc in ~/.local/bin <AIM[m]>But somehow libtool can't see it <AIM[m]>It's saying cc command not found <zamfofex>Are you sure you need ‘cc’? Can’t you just ‘export CC=gcc’? <AIM[m]><zamfofex> "Are you sure you need ‘cc’? Can..." <- Thank youuuuuuu <zamfofex>You’re welcome, I’m glad I was able to help! 🙂 <Alex[m]1>Hi, is there a way oft getting an Intel Card to Workon the T500? <Alex[m]1>Hi, is there a way oft getting an Intel Card to Workon the T500? <bost>Hi. Does anybody know how to get the `at` command installed? `guix search at` returns nothing useful. <xelxebar>Needing ncurses with versioned symbols, but it looks like our packages doesn't configure with --with-versioned-syms=yes. <xelxebar>It's easy enough to create a wrapper package, but are versioned symbols something we should consider in the main package? ***taiju is now known as Guest5
***taiju_ is now known as taiju
***taiju is now known as Guest1857
<taiju>I have set the locale to en_US.utf8 and did not notice it, but when I set the locale to ja_JP.utf8, GDM's font cannot display Japanese. <taiju>I use GNOME as my window manager. <taiju>Any hints on how to solve this problem? <taiju>When I log in with GDM, everything is in Japanese. ***Xenguy_ is now known as Xenguy
<fiesh>ok,, so the bisection process that mbakke helped me with is finally complete. 400c9ed3d779308e56038305d40cd93acb496180 is the commit that breaks me being able to input the password for my luks encrypted home partition and instead prints "Nothing to read on input." repeatedly <fiesh>what do you know, it seems to have been fixed while I was bisecting >_> <fiesh>joke's on me then I guess, let's see if it works now <taiju>I have set the locale to en_US.utf8 and did not notice it, but when I set the locale to ja_JP.utf8, GDM's fonts cannot display Japanese characters. <taiju>I am using GNOME as my window manager. <fiesh>hmm nope, still broken, so I guess my bisecting was at least good for something <taiju>Any hints on how to solve this problem? <taiju>After logging into GDM, everything is in Japanese. <zamfofex>taiju: Do you prefer for things to be in Japanese or in English? <taiju>zamfofex: I myself prefer to use GNU/Linux in English, but many Japanese users prefer to use it in Japanese, and I would like to help them. <zamfofex>civodul: I had trouble with glibc because I needed to update m4. TL;DR: The issue is a failing test in m4 which uses signals, but those seem to be a bit wonky in the Guix build environment. (I prepared a message with sneek with a bit more detail.) Ought I to just disable the tests for now? <sneek>Welcome back civodul, you have 1 message! <sneek>civodul, zamfofex says: It seems like m4 required an update for the newest glibc version (see commit 8e99f24c0931a38880c6ee9b8287c7da80b0036b in gnulib), but now there is an strangely failing test. See <https://logs.guix.gnu.org/guix/2022-04-09.log#081848> and the few messages I sent thereafter. I’m not sure what the issue is exactly. I should take a break from this for today, though! <civodul>zamfofex: maybe we could apply that Gnulib patch if that's all it takes to have the test pass? <zamfofex>taiju: If the issue is that Japanese characters don’t show well, you might be able to resolve it by installing a font that has them. <zamfofex>civodul: I think theses were tests that didn’t exist in the previous version. <zamfofex>It is strange that the raised signals aren’t actually causing the default termination effect, though. <zamfofex>I feel like that is a bug in Guix, or maybe I’m missing something. <taiju>zamfofex: I have added Japanese fonts globally and reconfigured, but it is not working. <civodul>zamfofex: ok; perhaps it's ok to skip this specific test, though it'd be good to document the details in a comment <taiju>zamfofex: There have been cases where it works if specified at the time of installation. However, I would like to use reconfigure instead of reinstall. <bjc>taiju: what's the value of LANG in /etc/environments? <taiju>TZDIR=/gnu/store/jj5aj3zxf65b7zgismwgy2wad7kbkcqz-tzdata-2021e/share/zoneinfo <taiju>LINUX_MODULE_DIRECTORY=/run/booted-system/kernel/lib/modules <taiju>SSL_CERT_FILE=/etc/ssl/certs/ca-certificates.crt <taiju>GIT_SSL_CAINFO=/etc/ssl/certs/ca-certificates.crt <taiju>GTK_DATA_PREFIX=/run/current-system/profile <taiju>GUIX_LOCPATH=/run/current-system/locale <taiju>PULSE_CONFIG=/etc/pulse/daemon.conf <taiju>PULSE_CLIENTCONFIG=/etc/pulse/client.conf <taiju>NM_VPN_PLUGIN_DIR=/gnu/store/s4j534jy2y6y4b5xff5adgwijxcrgjdl-network-manager-vpn-plugins/lib/NetworkManager/VPN <bjc>hmm. is gdm just displaying tofu characters for the japanese code points? <zamfofex>civodul: In the build environment, raising ‘SIGPIPE’ works as expected, though, but raising ‘SIGINT’ seems to be ignored. <taiju>bjc: Yes, that's right. But it's just a GDM screen issue. <bjc>i'd guess it's using a poor font, then <bjc>i don't have gdm configured on guix to test, unfortunately, but iirc, there's also a way to modify things through the gnome-settings panels (or maybe gnome tweaks?) <bjc>either way, it may give you a hint as to how to configure gdm <unmatched-paren>as i've previously mentioned here, my laptop seems to have a bios bug that makes it occasionally reset its EFI settings. this seems to make it forget about the grub in my disk. ***taiju_ is now known as taiju
<unmatched-paren>but i can't figure out how i got to grub-install last time... where should i look? <unmatched-paren>it's not in /var/guix/profiles anywhere, but surely it must be in the store, since it runs every reconfigure? <unmatched-paren>i tried looking on ci.guix.gnu.org for the grub-efi package's full path and typing it in -- didn't work, no such file or directory <taiju>bjc: I tried specifying Noto Sans CJK JP in the GNOME Tweaks font settings, but the result did not change. Strangely enough, the lock screen can display Japanese. <unmatched-paren>if i try to do anything guixy like installing the package, the guix-daemon socket refuses the connection ***taiju_ is now known as taiju
<zamfofex>civodul: What do you recon? It is really strange behavior regarding signals. Could it be worth reporting? Do you know what could be the cause? I disabled that specific test with a patch for now, I’ll see how it goes. <ZhuAisi[m]>The gdm user cannot read the fontconfig file, you need to add it manually <ZhuAisi[m]>IIUC, the problem is introduced in `core-updates` <blake2b>sneek: later tell zimoun: i sent you my blog submission yesterday but it was blocked by gmail for whatever reason. I've tried to resend it from another email address, but its unclear if it reached you. lmk, thanks! <unmatched-paren>hmmm.... i'll try manually setting up the wifi, then start a daemon and open a shell with grub <unmatched-paren>does guix store a wpa_supplicant.conf file anywhere? or will i have to write one manually? <faust>i have a trouble with podman <faust>Error: OCI runtime error: cgroups in hybrid mode not supported, drop all controllers from cgroupv2 <unmatched-paren>wifi up! now i have a grub-efi shell... i'm going to attempt grub-install now :) <gnucode>I wonder how hard it would be to make a sway config for guix system using %base-services instead of %desktop-services... the goal being to have minimal dependencies. And to make updating super short (in terms of number of packages one has to build). <gnucode>blake2b: just 'cause I'm a little nosy...what was your blog submission about? <gnucode>faust does is podman? Is that a podcast catcher ? <gnucode>unmatched-paren: just of luck with installing. :) <unmatched-paren>try looking at the definition of %desktop-services, and picking out what you need <gnucode>unmatched-paren I think I have done most of that.... but I am still missing some things...probaly the abliity to launch sway as a regular user and some other stuff. <unmatched-paren>because my BIOS has memory loss and likes to forget that grub exists <unmatched-paren>gnucode: podman is something to do with containers, like docker i think <gnucode>unmatched-paren I don't nkow how many times I have had to re-install grub. Since I am not a real technical user, that usually means, just re-installing. haha. <gnucode>unmatched-paren also what kind of hard ware do you have that has BIOS issues with forgetting grub is installed? <unmatched-paren>so i have to re-disable Secure Boot, change SATA type back to AHCI (the default is RST with Optane, which only seems to work with Microsoft(R) Windows(TM)), change function keys to actually work, and reinstall grub <gnucode>geez. That's annoying. you ought to get yourself a thinkpad. That's where the action is @ ! <unmatched-paren>but grub's reinstalled now, so i guess i'll hold my breath and reboot... <Alex[m]1>Is there a way to use Guix with Linux and not with Linux-Libre? <unmatched-paren>gnucode: i think i've been spoiled a bit by the modern-ish CPU and NVMe in this one, but I've heard good things about thinkpads <unmatched-paren>you don't need to use mainline linux, you could buy a wifi dongle that is compatible <Alex[m]1>But I have an Atheros Card Flying Around. Do you know if a T500 boots with an Atheros Card? <gnucode>Alex[m]1 I would have to google it.... but I don't actually know. <unmatched-paren>there is a way, but strict GNU policy in this channel is not to talk about it (personally, I think that if someone absolutely *needs* to use nonfree firmware, then we should help them so that they can use as much free stuff as they could on their hardware) <Alex[m]1>unmatched-paren: Could you please send me PM? <unmatched-paren>The thing that *really* annoys me is not people using nonfree stuff when it's necessary for their hardware to work, it's when people install e.g. Steam and claim they 'like [open | free]-source sofware, but I'm "pragmatic" about it' <Alex[m]1>unmatched-paren: 100% agreed. I think people like Torvalds are more reasonable about things like that. Stallman has great values but he's a little extreme in some aspects. <unmatched-paren>Both of them can be a bit shouty, from what i've seen of their emails ;) <singpolyma>Honestly *using* nonfree software isn't unethical. Only producing or spreading it. And even RMS has famously said if your neighbour asks you for a copy of a nonfree program you should give it to them <unmatched-paren>I'm not entirely convinced that Torvalds cares very much at all about free software (afaik he uses Ubuntu, which actively advertises proprietary software) <gnucode>looks like my guix pull is trying to build ghc. I think that is the haskell compiler...that's odd. I wonder how I pulled in that dependancy. Also I have disabled substitutes, which may have been a mistake. <unmatched-paren>but I don't know much at all about Linus, so i have no idea if he believes in the philosophy <unmatched-paren>especially since Linux used a proprietary VCS for a while until git was developed <gnucode>unmatched-paren: I kind of wonder if guix developers would care if I signed the copyright to my code to the FSF... <singpolyma>gnucode: indeed. Substitutes are pretty important :) <unmatched-paren>pandoc is probably one of the more common programs written in haskell <singpolyma>gnucode: never assign copyright for code. Especially copyleft code. More copyright holders is better <gnucode>singpolyma I respect your opinion, but I still also really respect Bradley Cohn. from software freedome conservansy. <gnucode>It's a double edged sword... because guix doesn't have a policy of assigning copyright, we have LOTS of developers. It's easy to start helping out. <gnucode>But how are we going to discourage people making guix a proprietary thing that some company sells? <singpolyma>My discussions with bkuhn are part of what made me realise assignment is even worse than I thought <singpolyma>For example, if you assign your copyright to FSF then the GPL will never be enforced on your copyrights even if you wanted them to be and were willing to put in the work <gnucode>singpolyma really? I'm game to hear your opinion. I honestly liked when someone on this channel sent me a linux chain email. Basically suing someone was a great way to p*ss them off. And why would they help your software project after that? <gnucode>singpolyma have we had some examples of that happen in the past? <unmatched-paren>If the copyright holder goes bad (unlikely for the FSF but it's better to stick to principles 100% of the time instead of making exceptions) they can just take it proprietary whenever they like, since they hold all the copyright <singpolyma>Anyone who needs to be sued is already a parasite and not helping :) don't sue your friends obviously <unmatched-paren>examples of what I said: elasticsearch, apollo federation (those are just the ones I saw mentioned on Drew DeVault's blog) <singpolyma>They can only take future versions nonfree since they can't revoke the license on already released code <unmatched-paren>hm, i've been chatting all this time and haven't rebooted yet :) i'll do that now <unmatched-paren>singpolyma: yeah, which is how amazon apparently forked elasticsearch <singpolyma>But they can definitely be like the FSF and sit on a big pile of copyright and not defend or enforce with them <gnucode>does the FSF hold the copyrights to elastic search ? <singpolyma>gnucode: no. But to many other commonly-violated things <unmatched-paren>it's a shame that the FSF took so many wrong turns, because the overall mission of GNU is really admirable... <singpolyma>We can still work on the mission. Concentrating in one org isn't healthy even if they were doing well <unmatched-paren>anyway, i'm going to quit and rejoin from my laptop, since this is irssi running in termux <unmatched-paren>singpolyma: many people seem to have become adverse to the free software philosophy precisely because of the FSF's wrong turns; they go 'oh no, free software purists' <Kalq[m]>hmm, after a recent guix pull, I'm getting the following error reconfigured guix home: <Kalq[m]>guix home: error: rmdir: No such file or directory <Kalq[m]>there's been no change to my configuration <singpolyma>unmatched-paren: yes, that is a problem for sure. people are also confused as to what the philosophy is even about because FSF and GNU and GPL all get conflated <gnucode>I'm heading out to work. Talk to you all later. <unmatched-paren>singpolyma: didn't GnuTLS pull out of GNU precisely because of these reasons? <flaminwalrus[m]>One of the really annoying "wrong turns" in my opinion is the GCC IRs (GIMPLE &co). There was an explicit design decision to make them as not-modular as possible, by separating necessary components between abstraction levels in an unnatural way, precisely because they wanted to make sure corporate tools couldn't use GCC as part of a compiler toolchain. This makes projects like Emacs' native-comp much harder than they would be if done for <flaminwalrus[m]>LLVM IR, which is sanely designed. I do get that the FSF is in a unique place because they were the _only_ ones doing free and open operating systems at the time, but the compromises to usability made in that name are quite frustrating. <flaminwalrus[m]>unmatched-paren: I'll see if I can find a source. I like clang and LLVM languages a lot more for that reason <singpolyma>Technical solution to legal problem, what could go wrong? <flaminwalrus[m]>If a corporate tool takes GCC IR input for e.g. extra compiler passes, it wouldn't violate the GPL, but would encourage users to adopt nonfree tools <unmatched-paren>But then LLVM came along and made that all pointless anyway, since all the corporate compilers now use it -.o.- <singpolyma>And worse because gcc is rapidly losing relevance even where it used to dominate as a result <singpolyma>But at least we forced AAPL to spend a lot of money in the meantime, which is good <singpolyma>unmatched-paren: the stock symbol for apple computers <singpolyma>Well, but that's the while point of copyleft. To force them to waste their money rebuilding things <fiesh>I don't think gcc is "rapidly losing relevance"... it's made tremendous improvements over the last years, and arguably is still the best C and C++ compiler, both in terms of compile time and of performance of executable, even though the margin with clang's gotten small. plus libstdc++ hasn't really gotten any competition from libc++ either, it's still a good deal superior <fiesh>I'm actually somewhat surprised this is the case, I guess it's a relatively small die-hard team vs. tons of people in a much more cumbersome review process? <bjc>Kalq[m]: yes, the rmdir issue has been going on for at least a day. it's been reported to the mailing list, but there hasn't been a response yet (that i've seen, anyway) <fiesh>although of course gcc is mainly backend by redhat, which belongs to IBM ;) <fiesh>if gcc had been a more open platform, it may very well be though that llvm wouldn't have gained traction and the companies would have stuck with gcc. I don't think that the gpl is really the issue there to be honest <unmatched-paren>yeah, more GNU projects should follow guix's lead of accepting contributions readily :) <Kalq[m]>ah thanks bjc! Nice to know it's not some obscure issue just with my setup <KE0VVT>Question: Can `guix-install.sh` be run non-interactively? Is there an `-y` option? <Ox151>hello, can anyone help with appending entries to the /etc/hosts file? I see there is host-name and hosts-file in operating-system, but i am unsure how to use them to append to the current file creating the localhost entires. <KE0VVT>jgart[m]: I'm trying to write an Ansible playbook to install Guix. <Alex[m]1>Hi, Is there a way oft getting a fingerprint Sensor (T500) to work with the T500? <jgart[m]>hey guixers, how can I append a manifest to my current profile? <jgart[m]>when I try, it ends up switching out my current profile with the manifest <jgart[m]>I want to just append to the current one <KE0VVT>I can't yet do everything on Guix System, so I'm using Ansible in place of a system declaration, and using Guix as my package manager and home manager. <jgart[m]>are you using this on Guix System or another distro? <KE0VVT>jpoiret: I'm using Guix on RHEL 8.5. <jgart[m]>I wish matrix had a way to forward a message in the Guix matrix channel to the Guix irc channel in a nice way <singpolyma>That doesn't sound like a nice way :P if you send a message in a pastebin link, I probably won't even read it <jgart[m]>I'd like to bridge it over and "pipe it" though a pastebin first <singpolyma>Oh, so you mean turn just the code block into a pastebin link. Maybe <jgart[m]>not sure how that looks on the other side <KE0VVT>I dream of `#guix:guix.gnu.org`. <unmatched-paren>jgart[m]: i once had to join a matrix channel because the project had no irc channel, and i frankly hated it <singpolyma>Is that an alias, jgart[m] ? Having trouble joining at the moment <unmatched-paren>maybe that's just because i couldn't be bothered to use an actual client, so i used element on the web... <unmatched-paren>i'm probably not being fair, but it just looked like IRC with more features, only one of which was useful <jgart[m]>I'm afraid to change my element client. That's what a new encryption protocol that you don't understand yet can do to you <jgart[m]>once that is working: goodbye element and gajim <unmatched-paren>oh, i remember now: i did try to set up a client, but i couldn't get weechat-matrix working <atka>unmatched-paren: no you are correct, the only feature matrix brings to the table is e2e imo <jgart[m]>everytime I start an e2e room with someone new with information that I care about I invite them to a new unencrypted room <unmatched-paren>e2e is a good feature, but when loads of extra bloat is included with it... i dunno... <jgart[m]>the issue is that most matrix clients don't support the e2e well enough <jgart[m]>if you're not using an element client that is packaged by upstream <singpolyma>I'm no matrix fan, but it's federated so not really comparable to a silo like irc <akonai>is there even a point for using e2e in public chatrooms <jgart[m]>so don't use element from your distro's package manager <atka>I'm just saying matrix doesn't offer really anything in forms of improvement and its quite annoying when a project has an official irc and someone makes a splinter group on irc and discord etc <singpolyma>jgart[m]: is their anything unusual about the matrix room you can think of? xmpp:#guix#matrix.org@aria-net.org?join just staring at me <jgart[m]>I've had issues with all element client's not packaged by element upstream <singpolyma>bjc: IRC is not federated at all. It's centralized <unmatched-paren>also it seems like you have to use their javascript-infested web client to get an account in the first place <unmatched-paren>singpolyma: IRC has closed federation, which i agree is not the same at all <bjc>you can connect your own irc server to a network, assuming the network allows it <jgart[m]>I'm only talking about the desktop version of element <unmatched-paren>libera.chat is a federated network consisting of e.g. lead.libera.chat, palladium.libera.chat, ... <singpolyma>bjc: no network is going to allow a random IRC server in though. I can't just set up and go like with a federated system <jgart[m]>unless making it difficult for everyone is a goal for that public chatroom <singpolyma>unmatched-paren: that's not federation that's just internal fault tolerance <bjc>the irc protocol allows random servers to connect. the irc servers may not <KE0VVT>jgart[m]: I tried Fractal, the GNOME app for Matrix, but it froze up and would not connect sometimes. <bjc>i have worked on xmpp servers at corporations that also didn't allow federation, but xmpp *is* federated <jgart[m]>I think parabola started doing exactly that in their matrix room <unmatched-paren>what does federation actually get you with chat rooms though? the only advantage i can think of is sending private messages between servers, <jgart[m]>or maybe one of the contributing reasons <singpolyma>bjc: the difference is that it's not even safe for an IRC networks to allow s2s with random servers. So zero networks allow it <bjc>federation allows you to run your own, or choose to support one you like for whatever reason. it's not just about privacy, but about features and choice <bjc>i'm old enough to remember a time when a thousand irc networks bloomed, and people did, in fact, run their own servers to connect to others without pre-clearing it <singpolyma>unmatched-paren: there are alternate models for chatrooms and forums that work just as well as federation. But if you already have a federated system (for 1:1 real-time for example) then using it for the chatrooms too is easier than making them a separate identity protocol IMO <bjc>was it a good idea? no, not really, but it was a heady time when anything was possible <akonai>the nice thing about federated IM protocols is that you don't get the vendor lock in problem that happens with every single proprietary IM app <bjc>that's why proprietary im apps don't allow federation, even if they are a thin gauze over xmpp under the hood <unmatched-paren>bjc: didn't most of the proprietary apps abandon XMPP at some point? <singpolyma>akonai: yup. I join everything from my one client and identity <bjc>no, they embraced and extended it, while turning off federation because "spam" <singpolyma>unmatched-paren: no. Many of them still use it internally <singpolyma>Most of them never had federation on to begin with <bjc>if email became popular in 2000 we'd all have a half dozen email applications <bjc>i vaguely remember other places (yahoo?) but google was the only one that mattered <singpolyma>Yahoo had a deal with AIM for corporate to corporate federation <bjc>there were gateways being run at aol and msn, i think, too? this was a long time ago <bjc>i hate that reddit is where i have to go for anything similar <akonai>bjc: with gmail having the majority of mail addresses while filtering most smaller mailservers we're halfway there <singpolyma>I worked on a project for that for awhile. But time <unmatched-paren>another example of the 'decentralized systems accidentally become centralized through EEE' phenomenon: git and github (and recently gitlab) <bjc>if i'm going to have to use a single web forum for usenet-ish things i'm just going to use the popular one <singpolyma>I just find the community better on lobsters. And the source is free of course <Aurora_v_kosmose>I'd also like the option to use my own client and not be forced into webapps. <bjc>the webapps are all terrible, too <singpolyma>Aurora_v_kosmose: there are definitely Reddit clients <singpolyma>I found out the other day that slashdot is somehow still going <unmatched-paren>ok, apparently gtalk still uses XMPP, but google effectively deprecated it in favour of hangouts... <singpolyma>unmatched-paren: the bigger xmpp users these days is WhatsApp <unmatched-paren>oh, "The Google Talk App for Android and the Google Chat tool in Gmail were discontinued on June 26, 2017 and no longer functioned." <- wikipedia only mentions this at the end... <bjc>xmpp uses signal e2e crypto these days, too <unmatched-paren>singpolyma: "WhatsApp uses a customized version of the open standard Extensible Messaging and Presence Protocol (XMPP)" <- you're correct :) <bjc>i'm honestly miffed that matrix didn't use xmpp <Aurora_v_kosmose>Did someone just decide to ask rebar3 because I mentioned it the other day? <bjc>is matrix vc funded? <bjc>great. can't wait for that poison pill to take effect <unmatched-paren>no, it's NOT an XMPP clone, because it uses YAML instead!!! Communication streams are human readable!! *Aurora_v_kosmose looks at encrypted stream <bjc>excuse me, but anyone who doesn't like xml in the context of xmpp probably hasn't been working with xmpp too long <bjc>i don't normally like xml, but it works well for xmpp <singpolyma>unmatched-paren: there's a joke XEP for JSON streams. YMPP should probably map via JSON-LD thi <Aurora_v_kosmose>And anyway, wireshark has already shown how you do human-unfriendly->human-friendly protocol display. <bjc>i endorse this idea :))))))) <Aurora_v_kosmose>Streaming parsers are painful, but most protocols these days are message-oriented anyway. <bjc>i'm on the fence about xmpp's xml stream method. having that perma-open often seems like more trouble than it's worth to me <singpolyma>KE0VVT: I was responding to a joke from higher up, and yes <KE0VVT>Why did Matrix become a thing when XMPP and Jingle worked perfectly fine in 2009? <bjc>yeah, but you could have done that in other ways <unmatched-paren>KE0VVT: i made a joke about YMPP, which singpolyma replied to, but Aurora_v_kosmose demolished it by noting that basically any modern messaging protocol has to be encrypted for it to gain acceptance :P <bjc>and it breaks down with start-tls already <KE0VVT>singpolyma's service partially brings those days back to life. <singpolyma>I'm thankful for them at least, since they invented jingle for me to sell to people <Aurora_v_kosmose>Matrix was inventented because XML parsers and XMPP servers were too performant, we needed cruddier, slower and more computationally-intensive implementations. <bjc>xml is flat out better for messaging than yaml. it's crazy to me that they went that route <bjc>yaml (theoretically) is a super-set of json <jgart[m]>How can I run Guix System on core-updates only? Like, I don't want to know about master <akonai>they should have just used s-expressions <bjc>and json has the same weaknesses as yaml compared to xml <Aurora_v_kosmose>Basically every improvement to JSON has just been backporting things that already existed in XML for decades. <bjc>i would be ok with s-expressions over xml <singpolyma>bjc: to be fair, while it's not better for sure, JSON-LD exists and it basically equivalent to XML. Just less good tooling <singpolyma>pushcx: yes, I'm aware. Code is still free as in freedom. BSD is free :) <Aurora_v_kosmose>I fully agree that JSON is more pleasant to write than XML for humans... but SXML already existed back then. <jgart[m]>I can't find the will to package the last two inputs <KE0VVT>singpolyma: I did not know XML was more efficient. The foolish Libertarian spartans of cat-v hate XML and love JSON. <singpolyma>KE0VVT: really comparing XML and JSON is silly, they are totally different use cases <singpolyma>XML is a RDF serialization. JSON is a syntax for strings, numbers, arrays, and maps <KE0VVT>cat-v is also a bunch of Unix worshippers. <KE0VVT>unmatched-paren: Plan 9 is the messiah of Unix-ism. :P <KE0VVT>singpolyma: Guix is a conspiracy to turn Unix into a Lisp machine. <Alex[m]1>Hi, Is there a way oft getting a fingerprint Sensor (T500) to work with the T500? <KE0VVT>Alex[m]1: Perhaps you could post to the mailing list to get more exposure. <bjc>what does a lisp machine look like in an era of virtual memory isolation? <KE0VVT>A Guile Wayland compositor would be dope. *singpolyma runs screaming from Wayland <akonai>unless you just ffi'd all the work to wlroots you would need to put in a lot of effort <singpolyma>Last thing I need is another reinvention of stuff :P <KE0VVT>singpolyma: Matrix compositors are so elegant. TTY2$ i3 # That's it. <Aurora_v_kosmose>bjc: Do you really need virtual memory isolation in Lisp Machines? They seem to be reference/capability compatible nearly by default. <KE0VVT>singpolyma: No `startx`. Just type the name of the compositor in a TTY and it Just Works. <singpolyma>Well, I don't normally have to deal with that. But also if I type startx it just works. I still just type one word (and usually zero words) <bjc>Aurora_v_kosmose: they do? i don't know much about them, to be fair, but capability support is relatively recent (compared to lisp machines, anyway), and my understanding of them was that they had unmitigated access to the whole ram <Aurora_v_kosmose>bjc: Well, in the case of the actual Lisp Machines yes, they did have ambient authority. But the general model of high level languages & security by language lends itself very well to a machine with no low-level memory addressing, only references. <KE0VVT>singpolyma: Also, I've been on Wayland for years at this point. When I went back to i3, the infamous screen tearing I thought wasn't a big deal before became very apparent. <singpolyma>KE0VVT: I'm still only 30% sure what "screen tearing" is <KE0VVT>unmatched-paren: I wrote `i3` just for simplicity. <Aurora_v_kosmose>singpolyma: Basically mismatched frame parts & lines showing up on screen. <KE0VVT>singpolyma: Go to your web browser and scroll quickly up and down a webpage. <bjc>Aurora_v_kosmose: the issue is, even with references, you need some way to direct a selected number through a capabilities layer. using caps for literally all references seems like it would be pretty slow <unmatched-paren>X also has specialized video drivers for everything, i think, whereas Wayland just uses KMS <KE0VVT>singpolyma: It's a very minor cosmetic issue, but it's glaring once you're not used to it being there. <akonai>on intel GPUs tearfree fixes tearing <KE0VVT>Aurora_v_kosmose: And it's blatantly obvious after years of Wayland use. <Aurora_v_kosmose>Consider my example, now imagine if you move your cursor 180 degrees around and it does that the whole way. <singpolyma>If Wayland works out I'll probably be forced on it eventually. But no reason for me to bother until then <KE0VVT>singpolyma: Wayland work just fine, unless you're dumb and use a proprietary software-dependent GPU. It's also, like I said, more elegant. <Aurora_v_kosmose>bjc: Performance and capabilities is a difficult thing and a large part of the effort behind seL4 & its ilk. <bjc>yeah, and those are neat systems, but they still use virtual memory <akonai>I dislike that in wayland, anything that interacts with other windows has to be built into the compositor. Seems like a weird decision to me <singpolyma>My mom used Wayland by accident for awhile (purism laptop) but I had to move her to X to fix screen sharing so she could do her job <bjc>i'll grant that virtual memory isn't necessary in future where all references are inherently capability grants, but i'm not sure when that future is <singpolyma>unmatched-paren: moving her to X also fixed that and she otherwise couldn't tell :) and then easier for me to admin in the future <unmatched-paren>"PipeWire is a project that aims to greatly improve handling of audio *and video* under Linux." <bjc>someone should port guix to l4 <KE0VVT>singpolyma: Is Mom using Pop!_OS? <bjc>unmatched-paren: i think that got nixxed <bjc>too much impedance mismatch <Aurora_v_kosmose>The difficulty with Guix on seL4 is that Guix kind of assumes a more monolithic model, doesn'it? <bjc>there's a bunch of efforts to move unices to capabilities models, but it's a big ask. it's not a great fit <bjc>yeah, but it seems like its more likely to happen than creating a new os that starts with capabilities built in <KE0VVT>I'm glad I could get my uncle on GNU+Linux without him saying "I WANT WINDOWS!" I have no clue how to manage Windows at this point, and wouldn't be able to help him with it. <Aurora_v_kosmose>bjc: I mean, Genode & Qubes both do it a bit differently but they manage to enforce isolation while remaining compatible with the vast majority of unices. <Aurora_v_kosmose>Of course they do that by effectively facilitating a hypervisor-controller model. <KE0VVT>Text is easier to read under Sway, compared to i3. I don't know why. <bjc>Aurora_v_kosmose: i vaguely remember that about qubes, but i haven't looked into genode at all <KE0VVT>unmatched-paren: You should put ) on highlight. :P <bjc>tbh, qubes is *too* much isolation for me <unmatched-paren>i think it seems like pointless friction to people who haven't developed stuff on either, since it looks like the differences are generally not visible (except for tearing) <Aurora_v_kosmose>Because people don't really write GUI programs for baremetal or make them into unikernels. <Aurora_v_kosmose>So you run whole Linux instances for nothing more than a single program. <bjc>also, i'm kinda still sweet on the ideas in the hurd, and qubes puts even more distance between me and them <Aurora_v_kosmose>In theory that's fixable even for Qubes, just needs more unix-compatiblity layers to be popular. <bjc>yeah, that's a pretty big problem. it's hard to use an operating system on your desktop without a web browser <KE0VVT>unmatched-paren: I'm just getting started with GNOME Builder. Seems to be agnostic of X and Wayland. <bjc>at least i can run firefox in a container or from flatpak <Aurora_v_kosmose>Ocaml is the only language I know that has an actively maintained unikernel library. <Aurora_v_kosmose>Xorg takes a third of the memory on basically half of my qubes. Even ones that aren't displaying anything. <bjc>wayland would have the same problem, i'd think <unmatched-paren>Cage is a 'kiosk' for Wayland, it can only run one full-screen application <Aurora_v_kosmose>In an ideal world, one could take a native library implementation of qubes API daemons, a unikernel lib, and a unikernel GUI lib & just wrap any program with that. <Aurora_v_kosmose>Your dev machine for producing such images could depend on Guix. I'd imagine it'd be feasible with the same mechanism as the one used to build Docker images. <philofosser>Hello, all. Does anyone know where guix's guile packages reside in a guix system? I've been poking around so I can read what all is in %desktop-services, %base-services, etc. My current goal is to remove stuff I frankly don't want clobbering up my potato like GDM. <philofosser>If anyone can tip me off, that would be much appreciated. <atka>philofosser: if you look at the source they are in guix/gnu/services, you will have a lot of .scm files in there but you can peak at what say base.scm exports and use only those services which you want <atka>for instance I'm not using the whole desktop.scm but just using elogind from it <atka>as desktop.scm exports elogind-service-type <unmatched-paren>philofosser: there's also the manual, which tells you which services are in each set <philofosser>Odd. I was flipping through the manual and was having a devil of a time finding that information. Is this manual different than the one that comes packaged in info? <philofosser>Either way, thank you both atka and unmatched-paren. <atka>philofosser: I found the manual hard to parse for that info, the source was more readable imo <unmatched-paren>the manual is the info one... look at System Configuration / Services / Desktop Services philofosser ***slep is now known as cel_b
<zamfofex>civodul: All the packages you asked me are now done building under the new glibc and m4! (With the addition on GNU Hello for good measure.) Is there more testing I should do, or is that sufficient to submit a patch? <civodul>zamfofex: yay, that looks like a good start! <civodul>so yes, i think you can submit a patch <civodul>eventually, we'll have to check things such as non-x86_64 arches and cross-compilation to GNU/Hurd <civodul>but that can come after, with help from ci.guix <Aurora_v_kosmose>In theoretical terms there's no benefit, but in terms of crystalizing vaporware into software, there is. <bjc>honestly, that sounds very academic. i don't know of any language capable of implementing a device driver that doesn't allow some kind of memory unsafety inherently <bjc>it's not impossible for it to exist, but i don't know of any <bjc>on top of that, it doesn't allow for something like the hurd, where a user can replace things at whim using whatever system they feel like <Aurora_v_kosmose>The components of the software system with access to the raw hardware could be mediated solely by referecences to a privileged API. <Aurora_v_kosmose>Thus any program without said references or capabilities would be unable to actually act on the hardware itself. <bjc>yes, but which language are you implementing in that would allow for that and also allows for decent device driver development <bjc>like, writing a 10gb or faster ethernet port <Aurora_v_kosmose>Some variant of Common Lisp could work. Such componenets may need to suspend/avoid GC and use exclusively static allocation and explicit typing for everything. <bjc>actually, i'm arguing at cross purposes anyway. it's not like a hurd-ish system could do any better, since that automatically has more overhead <Aurora_v_kosmose>You could also have that module enable something like managed virtualization relying on hardware support. <bjc>and ultimately i'm more interested in that than ridiculously fast ethernet, in terms of home usage <vagrantc>ever since the upgrade to the newer shepard, whenever i reconfigure a system the ssh connection dies <bjc>iommu offers solutions, it's true <bjc>i need to look at how l4 et al does things, but i've been wondering if you could do message passing between processes just by allocating a shared mmu page as a channel <Aurora_v_kosmose>For the non-supporting hardware... you basically have the unassisted QEMU situation, but safer. <bjc>that'd be pretty low overhead, speed wise, at the cost of potentially a lot of memory 