<florhizome[m]><nckx> "Does Guix's GNOME have a virtual..." <- Rudraveer Mandal: yes but it sucks <nckx>I tried to package Maliit today. Wasn't fun; doesn't work (yet). <nckx>Whether Guix's GNOME has a virtual, of course. <nckx>I was very surprised that Guix had none. <nckx>Or at least none that I could find. <KarlJoad>How "strong" is an inherit? I want to add an additional phase to the stumpwm+slynk package to configure the contrib-module's directory to point into the store, but I do not want to duplicate stumpwm+slynk. <jackhill>I guess there's also squeekboard and wvkbd which we don't have and <jackhill>although, I guess wvkbd is specific to wlroots? To think of it, I'm not sure squeekboard works outside of posh either <KarlJoad>Is there a way to force a package to build locally, but still use substitutes for the various inputs to the one I want to build? <KarlJoad>unmatched-paren: That was it. I just want to inspect the build directory (by also passing -K) and see what asdf-build-system/sbcl was doing. <KarlJoad>For packages, is package inheritance or building through composition of functions preferred? <KarlJoad>Ok... I will have to figure that out then... later... <nckx>KarlJoad: Inheritance is as strong as it gets: you start off with a record that is identical to the inherited one, then only the fields you specify (below it) are replaced. <nckx>The replacement is complete, though: if you write (inherit foo) (arguments '(blah)), no trace of foo's arguments will remain. <nckx>If you want to mix at the level of phases, use substitute-keyword-arguments and modify-phases on the inherited ones. <nckx>You can do this even without inheriting. <KarlJoad>nckx: Ok, cool. That was what I was curious about. StumpWM has a configure flag for the contrib-modules directory, and I am trying to figure out how to get that run so it builds. <KarlJoad>Problem is, StumpWM as it is now, build using asdf-build-system, which does not use autogen.sh, configure, make, or friends. <nckx>Ah. I'm not at coomputer, I'm afraid. <KarlJoad>No worries. I need to figure out how to run the autogen & configure scripts before ASDF does anything to build Stump, that way I can point the modules-dir to a store location. <nckx>akirakyle: Something went wrong with Shepherd. The machine probably needs to be physically rebooted. <akirakyle>nckx thx, hope it gets sorted out soon since trying to browse debbugs sucks <KarlJoad>nckx: So, I figured out that I can just change the `make-image.lisp` before the build phase and change the modules-dir. That way I do not have to mess with the build system. <nckx`>Hah. Somewhere, something was said that causes my IRC client to crash each time I try to connect to Libera. Wonder what it was. <atka>hi guix, are there guix issues again? guix pull is failing and my generations are broken somehow. went from a working system to a guix pull and reconfigure to a broken systems <atka>guix pull is failing saying dependencies can't be built <atka>fixed my generations but guix pull is failing to build guix-cli-core.drv <podiki[m]>not that I know of, pulled a few hours ago and was fine <podiki[m]>looks like I"m on the last commit I see the CI built, not sure why there's nothing newer (cd3c4ad) <nckx>Builds aren't building for the same reason issues. is down. <podiki[m]>ah (though issues. was down for longer, is it spreading??) <podiki[m]>for once I had a lucky pull then, usually I try right as things happen <nckx>I think they began failing around the same time. <podiki[m]>oh. thought I hadn't seen issues up for more than a day <nckx>I think I pushed cd3c4 late Thursday evening. <florhizome[m]>Funny cause guix has been quacking about not pulling for 9 days <gnucode>error: service-type: unbound variable <podiki[m]>ci hasn't built in a bit, issues front end down, ...not sure if that should cause problems, but that's the status at least <florhizome[m]>It's tragic this all still lives and dies with one server managed voluntarily (as i understand) <atka>if that's true, we should do something about that <dirtcastle>how to get zathura work. with mupdf extension and all. <roptat>dirtcastle, guix should have defined a variable for you if you installed zathura and its extension in the same profile (like, both with guix install) <roptat>you should have ZATHURA_PLUGINS_PATH <roptat>if not, try GUIX_PROFILE=HOME/.guix-profile; . "$profile/etc/profile"; unset GUIX_PROFILE <roptat>you'll need that in your .profile <roptat>sorry, typo: GUIX_PROFILE=$HOME/.guix-profile; . "$profile/etc/profile"; unset GUIX_PROFILE <roptat>note that it's different from .config/guix/current, you need both <roptat>comment je pourrais traduire "seat management" ? <roptat>pour la description du démon seatd <roptat>j'ai trouvé « gestion de séance » je trouve ça pas mal <vhallac>Hello. Is guix pull broken at the moment <vhallac>And where do I start trying to figure out what is causing the problem? <vhallac>I get the build log with the error on guix-cli, which eventually says "error: service-type: unbound variable", but I am not even sure which file causes this. <vhallac>It looks like it is in gnu/home/services/symlink-manager, but I am not sure ***Dynom_ is now known as Guest3330
<vivien>roptat, je ne connais pas bien le logiciel en question, mais le mot « session » revient souvent dans ce contexte, non ? <vhallac>yea. I am trying to build from source, but I cannot find the make target <vhallac>It tries to build everything, which is taking too long <vhallac>make-cli-go built successfully. Not sure if it should have failed <vhallac>I had a look at changes in symlink-manager.scm and services.scm, but couldn't see anything suspicious <vhallac>make install completes with /usr/local/bin/guix in DESTDIR <vhallac>make as-derivation fails. I don't know how to proceed. <vhallac>Maybe a git bisect with successive make as-derivation <vhallac>I mean I started with make-cli-go before I wrote here <vhallac>Oh well, any bug squished is good news. :) <roptat>vivien, visiblement non, mais séance n'est pas adapté non plus <vhallac>unmatched-paren: the link to paste.sr.ht has some extra chars at end. <vhallac>Oh well, if person looking at it cannot figure correct link, they can't fix the problem either. ;) <paulbutgold>Dear all, I somewhat followed the discussion about teams on guix-devel. Are teams already active? Is there some way to point out that some patches require some love? I'm not in a hustle as I'm already using those packages through my own channel but I'd like for them to be in Guix as I sent those patches around January the 4th <sneek>Welcome back paulbutgold, you have 1 message! <sneek>paulbutgold, nckx says: corefreq is another package that combines build systems (disclaimer: I wrote it). <roptat>vivien, j'imagine que siège correspond, finalement <vhallac>unmatched-paren: If I didn't mess up, ba22560627f848f40891a56355ff26b6de1380bc is the first broken commit <vhallac>I cannot see how it breaks this, though. <vhallac>Oh well, at least I have a good commit to guix pull from. :) <vhallac>It may be incorrect. Although I was able to do a "guix pull --commit=5df8f71" <vhallac>Can you try "git revert ba22560627f848f40891a56355ff26b6de1380bc --no-commit" on master? It looks like it is building <unmatched-paren>Actually, I'll do `guix pull --commit=ba22560627f848f40891a56355ff26b6de1380bc`. <vhallac>That's a good idea. I didn't try that <vhallac>brilliant. Thanks for investigating. <vhallac>I noticed that the gc.scm is only using modules from guix before the commit <vhallac>Now it uses (gnu home). That can easily intorduce a cycle <vhallac>Ye, I don't know anything about the code base, so I should not really say anything. :D <unmatched-paren>> Looks like an &assertion-failure is being obscured, so I guess the error handling in (guix build compile) is wrong <vhallac>It could be a response to the error you've seen during bisect *attila_lendvai is rather frustrated with the error handling in the guix/shepherd codebase <tewi>hi, i have a question - if many packages were installed with guix install separately (think about 70), and all the generations were to be deleted, would the installed packages be preserved? <clever>tewi: if you keep the latest/current generation, yes <clever>if the current generation is deleted, then its like you never installed anything <tewi>clever: nice, so it just collects all of the stuff into one generation i guess <PotentialUser-13>Hi, so I was tinkering with the config.scm file trying to split it into multiple files, and I have started by moving out the users block into a module. I copied config.scm to a new file base-system.scm. in the same folder I have users.scm. This are the files content: https://pastebin.com/DPrp5qZd . currently when testing them with `sudo guix system <PotentialUser-13>build guix/systems/base-system.scm` it returns the error `uix system: error: failed to load 'guix/systems/base-system.scm': <PotentialUser-13>no code for module (systemUsersModule)` not sure what Im doing wrong, any tips? <unmatched-paren>PotentialUser-13: before i answer your question: why are you using camelCase instead of kebab-case? <clever>tewi: yeah, each time you add/remove/upgrade something, it takes the current generation, mutates it, and saves the new state as the next generation <clever>tewi: rollbacks are just switching which generation is current <muradm>something broken after guix pull? <unmatched-paren>Well, I can't see any cycle in (gnu home services) or (gnu home services symlink-manager), at least <unmatched-paren>workaround is `guix pull --commit=5df8f7802e8b45855e8f84830677001dc631e4f7` <PotentialUser-13>unmatched-paren uh its what I'm used to put :P didnt realize it. Prob should get used to use that with this files <unmatched-paren>and you can merge (use-modules (systemUsersModule)) and (use-module (gnu)) into (use-modules (gnu) (systemUsersModule)) <unmatched-paren>The name of the module has to be the same as the name of the file without .scm. <PotentialUser-13>so the path in the use module (`(use-modules (gnu) (system-users-module))`) is relative to the path where I call the system build? (Im guessing it should be then `(use-modules (gnu) (guix system system-users-module))`) <PotentialUser-13>no, Im trhying to make a config for my main pc and tinkering with the configs however I can :D <PotentialUser-13>So far from what I understand with guix and guix home I should be able to have declarative configs for my pc and servers, but there isn´t as much info/tutorials/examples as I would like out there, but so far Im liking what im learning <unmatched-paren>hako: I was thinking you could set some FC_CACHE env var, but apparently such a thing doesn't exist <Maya[m]1>hi, is the function find-files from (guix build utils) recursive? <Maya[m]1>bcs if not, i think that the function patch-shebangs in %standard-phases has a bug, i have discovered that it does not patch all shebangs and the pcsc-lite package is broken <unmatched-paren>hako: aha! try adding a phase that does (setenv "XDG_CACHE_HOME" ".cache") <unmatched-paren>if that doesn't work, try adding (mkdir-p ".cache/fontconfig") after it <hako>unmatched-paren: Greeeeeeaaaaat THX! <Maya[m]1>unmatched-paren: when I looked at it it didn’t seem that way <Maya[m]1>but im not exactly great in guix methods <Maya[m]1>imo the patch shebangs should be recursive and i don’t understand why the python script from pcsc-lite doesn’t get patched <unmatched-paren>thanks for sending a patch for i3-autotiling, i wrote a package for it and put it in guixrus but never upstreamed it (and i didn't enable the tests) <Maya[m]1>unmatched-paren: no, it is a source file <Maya[m]1>i think it is somehow modified but the shebang is there <hako>FAILED test/test_shutdown_event.py::TestShutdownEvent::test_shutdown_event_reconnect <hako>FAILED test/aio/test_shutdown_event.py::TestShutdownEvent::test_shutdown_event_reconnect <hako>FAILED test/aio/test_workspace.py::TestWorkspace::test_workspace - RuntimeErr... <unmatched-paren>Maya[m]1: Could you post a link to the file in its git repo or whatever? <Maya[m]1>and it is put into /run/current-system/profile/bin/pcsc-spy <Maya[m]1>and you can find that file in src/spy/pcsc-spy <hako>ok, btw that's my site... <unmatched-paren>hako: btw (which "true") or (string-append (assoc-ref inputs "coreutils") "/bin/true") is probably faster than search-input-files <Maya[m]1>yes it already is in guix but it is broken <hako>unmatched-paren: OK, thanks : ) <Maya[m]1>unmatched-paren: i wanted to submit a patch to fix it but i didn’t know how patch-shebangs works <Maya[m]1>unmatched-paren: oh! i haben noticed the space <unmatched-paren>^[[:blank:]]*(/[[:graph:]]+)[[:blank:]]*([[:graph:]]*)(.*)$ after eating the #! <unmatched-paren>I wonder if it's because python isn't given as an input to pcsc-lite. <unmatched-paren>patch-shebang: ./src/spy/pcsc-spy: warning: no binary for interpreter `python3' found in $PATH <Maya[m]1>unmatched-paren: thanks for the help! <3 <Maya[m]1>if i made a change in local git clone of guix, i ran bootstrap and config, how can I test-install my changed package? If I run ./pre-inst-env guix install <package> it doesn’t reflect the change in the repository <sneek>nckx, raghavgururajan says: Test Message <nckx>raghavgururajan: I found no evidence of thought anywhere on that Web site, so it would not be fair for me to have any. <unmatched-paren>It tries to argue that lack of sandboxing on Linux makes it insecure because malicious applications blah blah *nckx read the Linux page now. <unmatched-paren>But the security model is not to stop malicious applications doing bad things, it's to prevent the user from installing them in the first place! <unmatched-paren>For one, most applications under linux are free-source so they're unlikely to be malicious <nckx>No, you need to pay Apple to do that. <nckx>You didn't read hard enough. Read harder. <Maya[m]1>unmatched-paren: did i do something wrong with pre-inst-env? <teythoon>% guix shell --pure gcc -- /bin/sh -c "echo 'int main() {}'|gcc -x c -" <teythoon>gcc: fatal error: cannot execute 'as': execvp: No such file or directory <teythoon>i guess that distinction is very helpful for all the people who just want to run the preprocessor <nashdidan[m]>Hi guys. New to Guix. Any idea why "git pull" fails with this right now for me? 10x! <nashdidan[m]>builder for `/gnu/store/wz251qlhkh17zd1zlf5xi0qfi69iyp04-guix-cli-core.drv' failed with exit code 1 <nashdidan[m]>build of /gnu/store/wz251qlhkh17zd1zlf5xi0qfi69iyp04-guix-cli-core.drv failed <unmatched-paren>it's also very helpful for all the applications that reference libgcc :) <teythoon>but those don't use gcc-the-binary, so would they not be better off depending on libgcc instead? <PotentialUser-13>Hi again! when trying to create a module in which I declare the system users its giving me <PotentialUser-13>hint: Did you forget `(use-modules (gnu packages package-management))'?` <nckx>teythoon: gcc-toolchain provides more than just gcc (including binutils, as you learnt), all patched to actually work together to produce working binaries. <sneek>raghavgururajan, you have 1 message! <sneek>raghavgururajan, raghavgururajan says: Test <nckx>PotentialUser-13: The issue is (add-to-load-path (guix(systems(system-users-module)))) — which incidentally happens before any module imports, but the real question is why you're calling a procedure named ‘guix’ without knowing what it is. <raghavgururajan>nckx, unmatched-paren: Btw, I was wondering if the claims or so-called claims extend to today? and whether to icecat and ungoogled-chromium as well. <nckx>sneek: botsnack for special bots <unmatched-paren>raghavgururajan: I think there may be some truth in the idea that firefox isn't as private as many people think it is. <nckx>Which is a prejudice wrapped in a vague assertion. <nckx>(Not saying you're wrong, it's probably true, but wow has it absolutely nothing to do with ‘browser security’.) <unmatched-paren>and of course there's the modified variants like icecat and librewolf <raghavgururajan>My line of thinking was like if firefox is indeed less secure than chromium, then how come it missed by tor-project's review and their choice of firefox over chromium. Their whole area of research and study is security and privacy right. <nckx>Use a pastebin (such as the one in the topic) for multi-line (>2 or 3 lines) pastes, PotentialUser-13. <unmatched-paren>and anyway, even if it did accept a module spec it would be `(guix systems)` <nckx>PotentialUser-13: That is not how Guile modules are name (it would be (guix systems system-users-module) if so), but yes, this needs to be a stringy file name anyway. <raghavgururajan>Google tool telemetry of chromium further to "Don't do anything without my knowledge". <PotentialUser-13>that did it, the moment I changed it to a string it worked, thanks! (add-to-load-path "guix/systems/") <teythoon>nckx, unmatched-paren: my point was, that is rather unexpected and not very user friendly <nckx>teythoon: I know, I was just toying with the fact that you chose to be snarky instead. <teythoon>but i have to be honest, i have been using linux for 24 years now, i have used my fair share of distributions <nckx>I was just typing my confusion re: hiding, then unhiding, apparently, gcc on a yearly basis. <teythoon>but everything is somewhere between hard to frustrating <teythoon>guixsd installation, configuration, every day usage, debugging problems <lilyp>Let's start by being modern and calling it Guix System :) <unmatched-paren>but i'm not sure about configuration/everyday usage. What do you take issue with there? <teythoon>i have to hope that the documentation covers my needs exactly, otherwise i'm lost <nckx>Should have kept the GuixSD name, it obviously has preternatural mind-burrowing powers that will outlast us all. <teythoon>tweaks to the configuration, udev rules, stuff like that <unmatched-paren>teythoon: Yes, I think lack of documentation is the leading cause of confusion <teythoon>it breaks every web site on the planet in the name of my freedom <teythoon>and helpfully includes a calculator for us postal rates <teythoon>sure, i'm using ungoogled0chromium and firefox now <lilyp>LibreJS has a literal "disable everything for the current page" button <teythoon>every one of these things has a fix, or a good reason to be that way <teythoon>but the overall user experience is terrible <teythoon>saying ah, teh package you want is gcc-toolchain totally misses the point <lilyp>you're missing the point – gcc without binutils is the greatest nothing burger imaginable <lilyp>If we hide it, folks complain that gcc does nothing <unmatched-paren>and make `guix (shell|install|...) gcc` a special case that warns you :) <lilyp>if we don't then folks complain that they have to type "-toolchain" as though that was so hard <unmatched-paren>Or maybe we can add some wrapper: `(hidden-package-with-redirect "PACKAGE-NAME" ...)` ***deba123[m] is now known as debm[m]
<lilyp>Our documentation for `guix shell' literally mentions gcc-toolchain <lilyp>Also @section The GCC toolchain <teythoon>if i apt install gcc, i have a working compiler <teythoon>surely that holds for fedora and opensuse, and all the other distributions <teythoon>i understand that guix is not like other distributions <teythoon>that doesn't mean it has to be unintuitive and difficult just to set it apart <teythoon>you don't have to listen to user feedback <teythoon>in fact, i worked for 2.5 years for a project that never listens to user feedback, yet is very successful <teythoon>everyone i speak to agrees that it has a terrible ux tho, and everyone is looking forward to replace it <unmatched-paren>Seems like the philosophy is "if X doesn't need Y to build successfully, X should not depend on Y" <clever>teythoon: i dont know guix as well, but the same problem exists within nix <clever>and the basic problem, is how does #include <foo.h> find foo.h? <clever>there is no /usr/include to search within <clever>nixpkgs/stdenv uses a special env var, and a shell script wrapper around gcc, to inject -I args into it, for each dependency you defined in the package <lilyp>yeah, lots of projects sadly assume that /usr/include is a magic place that holds everything <clever>with the extra -I flags, just doing #include <foo.h> works <clever>then packages install to $prefix/include/package-123 to avoid conflicting with other versions of itself, exactly what nix/guix prevents <clever>and they wind up making nix/guix harder to use with that package :P ***debm[m] is now known as deba123[m]
<Maya[m]1>unmatched-paren: ok i have submitted the patch! <Maya[m]1>how can I create a new pam rule? of ‘auth sufficient <name>.so’? <lilyp>specifically for the module, use (file-append package "/path/to/sofile") <Maya[m]1>lilyp: i want a service more than a package pam entry <lilyp>is it not a pam module (which is part of a package)? <lilyp>(on a higher level, the service you're looking for is probably pam-service) <Maya[m]1>lilyp: i want to add pam entry for fprind as it does not currently work for authentication <lilyp>okay, and your definition so far is...? <lilyp>hmm, it uses dbus and polkit <lilyp>especially the latter should be enough for granting permissions, no? <lilyp>anyway, assuming it does not, you'd add a (service-extension pam-service-type fprintd-pam-service) <lilyp>and then implement (fprintd-pam-service config) <Maya[m]1>lilyp: i think I maybe have a little confused how the service type and service works <Maya[m]1>the service type is where i put the service-extensions yes? and the service function is one where I can through records generate the pam file? <lilyp>you generate a matching configuration for the service type <lilyp>in case of pam-service-type that'd be a list of pam-entry s <ardumont> /set irc.server.libera.autojoin "#swh-devel,#swh-team,#swh-sysadm,#swh-offtopic,#guix" <antipode>Is there any policy on backporting fixes? <antipode>More concretely, if I send a patch adding a patch (accepted upstream, but no new release yet) fixing some issues, would such a thing be accepted if it isn't a complicated patch? <mbakke>antipode: such patches are pretty common, especially for build/test/security fixes <antipode>It fixes a little thing that makes tests in the software I'm writing occasionally fail. <lilyp>If it causes a bunch of rebuilds, you'll have to graft, otherwise just go through the regular review. <Guest63>looping over %base-packages allow to do an operation per package, how do i access the package's attributes ? say its name <Guest63>it would be very helpful to know how to do some basic object inspection, when calling reconfigure to drop in a breakpoint and dump current variables, is there any such thing built in ? this is more guile related <roptat>actually, package-<name-of-a-field> to access the field in question <KarlJoad>Does guix-build have a flag for keeping the build directory, even if the build succeeds? <antipode>KarlJoad: it doesn't, only for in case it fails. <KarlJoad>antipode: I remembered I asked this before. You can add the --check flag too, and it seems to keep the build directory. <antipode>KarlJoad: it doesn't, at least when doing "guix build hello --check". <antipode>Maybe if there's a reproducibility issue (unverified) as that's a failure, but otherwise it doesn't seem to. <KarlJoad>I'm doing `guix build --check -K -f <path/to/custom/stumpwm.scm>`, and it works out. <KarlJoad>I guess it is a good thing stumpwm is not reproducible then... <PotentialUser-13>where should I add the unattended-upgrade service? when creating the %base-service list it returns `invalid field specifier` I tried adding `(gnu services base)` to the use-modules, but still not working this is my current config: https://pastebin.com/JAysQvFT <PotentialUser-13>Also, why is the openssh service added to the %desktop-services? why isn't it in %base-services? <KarlJoad>PotentialUser-13: I don't think it is in %base-services because it is not needed to make a system boot-able. <KarlJoad>PotentialUser-13: See `(guix) Base Services` for %base-services. <KarlJoad>and `(guix) Desktop Services` for %desktop-services. <pashencija[m]>cbaines: It was fixed with 0f646fc79a06c9857f5f28d5a1eebd09f2d17310 5 days ago <antipode>sneek: later tell cbains: I've written a few comments on the fibers streaming patch. <antipode>The patch I mentioned is now sent as 56728 <andrzejku>I am trying to find if I installed app and I want to run this app as service <andrzejku>unmatched-paren: this service actually has systemd rule <KarlJoad>Is anyone here particularly familiar with the asdf-build-system? <KarlJoad>andrzejku: You just need to be careful with Shepherd. It does not understand daemons that immediately fork off, which is why my emacs server uses `--fg-daemon`. <andrzejku>KarlJoad: but generally idea is that in /etc/config.scm I keep operation systema and base services <andrzejku>and in guix home I have what may change more often <KarlJoad>The way you define the service does not change if you have the service in a home environment or an os-one. I have user-specific things in my home environment. But your general idea is right. <teythoon>I have used debtakeover to replace guix sd with debian bookworm <teythoon>there was a bit of a funny moment when debtakeover decided to remove all symlinks pointing to directories from /var, killing all guix profiles <singpolyma>teythoon: obligatory "there is no such thing a guix sd maybe you meant guix system" <teythoon>also, please do go on to teach people that they want gcc-toolchain instead of gcc