<mekeor[m]>gym_lover: are you an emacs user? did you install a desktop environment? <EMax`0Mancer[m]>is there an example of a really simple root-level shepherd service configuration? I just want to do the equivalent of running "sudo XXX". <mekeor[m]>e-max-0-mancer: there's a small example in the manual: (guix) Shepherd Services <mekeor[m]>"The example below defines a Shepherd service that spawns ‘syslogd’"... *mekeor[m] got hpcguix-web running locally. <mekeor[m]>next step: create a guix system image with hpcguix-web including a long list of channels <civodul>mekeor[m]: yay, glad that it works for you! <civodul>if there are lessons learned or ideas you have to make the process smoother, that'd be welcome! <mekeor[m]>how can i avoid boot-loader and file-systems in a os-declaration for a vm? <civodul>mekeor[m]: i think you can have (bootloader #f) and (file-systems '()) <mekeor[m]>having to keep holding ctrl while pressing C-v in order to scroll a document is worse than having a non-ergonomic keyboard, it seems to me <bjc>emacs will use page-up/down, too <bjc>personally, i use a mouse with a fly-wheel style scroller, so i can just set it spinning if i need to ***rekado_ is now known as rekado
<mekeor[m]>damn, i just started guix system inside a vm on guix system, whoa <EMax`0Mancer[m]><mekeor[m]> "it's chapter 10.18.4 Shepherd..." <- I saw that. Should it be fairly generalisable? <mekeor[m]>e-max-0-mancer: sure, why not? maybe you can give it a try and tell us when you run into difficulties :) <mekeor[m]>yes, i'll take notes and see where i put the knowledge <mroh>pactl is provided by pulseaudio. <f1refly>how can I set up nmcli so that it can be interacted with from a normal user account? <horizoninnovatio>Greetings, 2 question seemingly related: how do I enable ./ in terminal to run scripts & how do I get appimages to run? <meo>horizoninnovatio: what do you mean how to enable ./ <AwesomeAdam54321>horizoninnovation: ./ in the shell is used to refer to a program in the current directory. It depends on the execute permissions of the program in question as ./ doesn't change it <horizoninnovatio>I have a script "example.sh" which can be executed with sh or ./ but in guix only sh works. An install script requires ./ to work on .sh scripts. <horizoninnovatio>This is the problem I'm having, as well as appimages "no such file or directory" <AwesomeAdam54321>horizoninnovatio: You need to do chmod +x to add execute permissions to your shell script <horizoninnovatio>Example.sh will run fine with "sh example.sh" but with ./example.sh output is: /bin/bash: bad interpreter: no such file or directory. <meo>horizoninnovatio: shells are in a different location in guix <meo>try this: which bash <meo>if you can't/won't edit the script to point to the guix-specific shell path, you can create ~/bin, and symlink the shell into it, and then add ~/bin to shell PATH <rekado>horizoninnovatio: they probably expect the glibc’s loader to be at a well-known location such as /lib64/ld-linux-x86-64.so.2 <rekado>you could just arrange for it to be there <rekado>but in my experience appimages make even more assumptions than that, so in many cases you may still not be able to run them. <horizoninnovatio>If I can cure the wine stack overflow, I won't need appimage! The knock-on effect of windows apps! <jpoiret>horizoninnovatio: about the shebang, i think the portable way is to have #!/usr/bin/env bash instead <jpoiret>you could manually patch the shebang of all the scripts you want <maximed>lechner, AwesomeAdam54321: php-fpm _is_ available in Guix, try guix shell php -- php-fpm --help <tom[m]1234>Hello . I would appreciate your opinion.Which ROM is now the most libre after Replicant, more hack-proof, more protected ? Thanks. <f1refly>what can lead to ~/.guix_profile/bin not being in PATH on a guix system? <bjc>you're not sourcing the .profile at login? <f1refly>I think there was no default .profile in my ~ <bjc>guix home will put one there, i think <bjc>if you're not using ‘guix home’, you'll need to do it by hand <f1refly>okay, now only my fonts are garbled, progress has been made <bjc>you're sourcing ~/.guix-profile/etc/profile? <bjc>i don't know if that'll have any effect on fonts, but it might *f1refly sucks at computers <f1refly>I'm not sure where I had that line on my old machine, I must've buried it somewhere <f1refly>Is there documentation on setting up autofs with guix or would I just define the whole configuration in my config.scm and then add it as raw file to be put in /etc? <attila_lendvai>i'm trying to get a binary to work that requires libffpeg.so, but none of the guix ffmpeg packages have that file, only libav.so stuff. what am i missing? *civodul is in patch-review mode <apteryx>I saw your activity there, thank you! <apteryx>I've been experimenting with a script to automate the deletion of all leaf python2 packages <civodul>it feels like emptying the ocean with a spoon :-) <apteryx>I was a bit surprised that the reader transforms sexps like (quasiquote [...] (superseded . ,python-pytest) into (quasiquote [...] (superseded unquote python-pytest)) <jpoiret>attila_lendvai: im not sure that exists <civodul>apteryx: ah yes, it's peculiar: (a . (b)) = (a b) <jpoiret>f1refly, bjc: i'd even recommend sourcing /etc/profile instead if you're on guix system <civodul>which makes sense when you think of it in terms of pairs, but the read syntax is a bit... surprising <civodul>is there still anything preventing us from reconfiguring? <apteryx>I've replied to Arun, it's waiting on me to revert a few commits on the hydra config to revert the root fs; I think it'll be able to look into it this weekend. <apteryx>(the new array is not bootable on berlin yet, for reasons to be investigated). after the revert the plan is test the config on a similar setup on node 129, which is the same hardware <civodul>well, "great", it's kind of a bummer that things didn't go according to the plan :-/ <civodul>thanks for working on this upgrade, you're a hero! <apteryx>yeah, it's been met with more resistance that I had hoped :-) <apteryx>but we've made some progress on a few things, such as allowing external sysadmins to see the live boot logs via serial (through iDRAC) <apteryx>any idea how I could synchronize the script so that it doesn't try to call committer.scm while an edited module is in the progress of being saved? <apteryx>hmm, it seems the file update is supposed to be atomic, so if that's true 'edit-expression' shouldn't need to be manually synchronized I guess <apteryx>perhaps it's the source-location-map that goes stale <peterpolidoro>I am trying to import a pypi package and it says it needs both python-pyqt5 and python-qtpy and that python-pyqt5 does not exist. Since the quix python-pyqt5 package does not seem to exist, does that mean that only python-qtpy is really needed? <jpoiret>peterpolidoro: if it's not in guix, you'll need to also import it <peterpolidoro>I am assuming that is a package that lots of people might need, do you think it is not already included in guix because of licensing issues or something? <jpoiret>maybe noone ran into that dependency <peterpolidoro>interesting, I will go ahead and add it and submit it for contribution then <peterpolidoro>the pypi importer seems to think those need to be two separate packages, but I will go try ignoring python-pyqt5 and see if it still works thanks <euandreh>WesterWest[m]: I see that your patch was applied upstream :) It sometimes takes a couple of days, but maintainers do a good job with keeping up with contributions <bjc>i'm still waiting on a one-liner that lets me use .dir-locals over tramp =( <meena>how do i get newer software from guix? do i have to compile it myself? <bjc>‘guix pull’ will get you the latest that's been packaged <meena>the packaged version of pandoc is 20.14.0.3; the current version is 2.18 <maximed>(after ‘guix pull’, don't forget to ‘guix package -u’) <bjc>if it hasn't been updated in guix, you can try your hand at doing it yourself and submitting it <jpoiret>hmmmm, haskell software... might be annoying to update <maximed>meena: To update to the upstream version, you'll have to make a git checkout, build it (./configure --localstatedir=/var && make) and run "./pre-inst-env guix refresh -u ghc-pandoc" <maximed>Though "guix refresh" unfortunately does not always work perfectly in 100% of the cases. <maximed>(after that, you can submit it to guix-patches@gnu.org) <bjc>in particular, the instructions in “Running Guix Before It Is Installed” are handy but you will probably need to prefix some commands with ‘./pre-inst-env’ in ways that aren't documented <maximed>civodul: I've made a v2 for the ‘Replace some mocking with with-http-server*, avoid hardcoding ports.’ series <maximed>(keep-lingering? has been removed, + other simplifications) <meena>i'm confused. doesn't guix have build recepies? shouldn't i just be able to bump a version somewhere ?? <maximed>meena: Yes, that's what "guix refresh -u" does. <maximed>"guix refresh -u" bumps the version field <maximed>and also changes the sha256 field appropriately. <maximed>"guix refresh -u" is just some tool for doing these things automatically. <maximed>But you can also manually modify the package definitions. <jpoiret>meena: you might have to update the dependencies though, which might be annoying <jpoiret>since haskell packages often have many deps <meena>jpoiret: i've been building software on unix since i started using it… 22 years ago <meena>in theory, i know how it works. <bjc>guix is /very/ different <jpoiret>depending on the language and its tooling, guix takes very ... unconventional paths <jpoiret>see go or rust for the more painful examples <maximed>jpoiret: FWIW, rust should become less painful with the WIP antioxidant-build-system <jpoiret>for haskell it should just be a matter of updating all dependencies to the needed versions, but that might take a bit of time <meena>i can see why no one would wanna bother… <jpoiret>yes, i've been reading your emails with anticipation :) <civodul>maximed: noted, thanks for the heads-up! <bjc>meena: guix offers significant advantages for some people over traditional unix methodolgy, but it's not for everyone <jpoiret>i'm just worried it might become an additional maintenance burden when upstream rust tooling changes significantly <jpoiret>meena: the thing is, when building haskell packages manually, you'd also need the newer versions of the dependencies, its just that haskell has its own tooling for that that we can't use on guix unfortunately <peterpolidoro>It seems when I only include python-qtpy it cannot find PyQt5 <jpoiret>apparently, pyqt is just a small wrapper for pyqt5,6 and friends <WesterWest[m]>euandreh: it was! I am really happy that I am now a contributor of guix :D <apteryx>oh, another bug with package-definition-location; try it on python2-pytest-warnings. <apteryx>it gives the line next (under) the define-public ... binding <guixsd-n00b>Hi! Is there any way to introduce symbols definitions from a context into a gexp? I know there is 'with-imported-modules, but I was hoping that there was an alternative for simple things, like using a variable; maybe something like '(with-context-symbos symbol-list gexp-body ...) <bjc>^- i would be interested in an answer to this, as well <civodul>apteryx: (package-definition-location python2-pytest-warnings) says line 2403 for me, which looks correct <apteryx>sneek: later tell civodul weird, here also now <apteryx>perhaps because I'm editing the sources at run time and the information is not refreshed <maximed>antioxidant-build-system update: rust-pkcs5 and all its deps build <apteryx>congrats, i can't wait to see rust packages have normal Guix inputs <apteryx>it will even help the bootstrapping effort for other platforms <maximed>apteryx: Because lower build times, or some other reason? <apteryx>because cargo was needing proc crates (thus needed to be dynamically linked), but with your work this may not be the case anymore <apteryx>for cargo I think, not for rust itself (I hope) <maximed>They are still dynamically libraries. Curiously proc-macro crates only build when built as shared libraries, whereas the inverse hold for other crates. <apteryx>so with your work we may be able to build a statically linked rust more easily <apteryx>yes, proc-macro crates rely on dynamic linking as an implementation detail <maximed>antioxidant-build-system update: rust-pkcs1 and all its deps build ***chexum_ is now known as chexum
<peterpolidoro>how might you show all packages that have the same name in multiple channels? <peterpolidoro>I first put packages into my own custom channel and then submit them to be contributed to guix proper <peterpolidoro>it would be nice to periodically run a check to see if that package has been included in both guix proper and my custom channel so I can then delete it out of my custom channel <alextee>actually, that libadwaita patch needs another update to 1.2.0 <alextee>but I can update it in another patch set <maximed>peterpolidoro: Submit the package definition to guix-patches@gnu.org, even if only as a [WIP PATCH]. When it is applied to guix proper, the issue will be closed by whoever applied it (unless they forgot to do that). You will be notified by e-mail (by ‘Debbugs’) that the issue has been closed. <maximed>(basically, notification instead of polling) <maximed>(Though not exactly what you asked for) <peterpolidoro>yes that is how I am doing it now. that works great until I get behind, then I have to scroll through emails while pruning my custom channel packages and it gets a little tedious <peterpolidoro>I was kind of hoping I could just run a command and see that three packages have duplicates then remove those from my custom channel <maximed>peterpolidoro: Does your e-mail client has a search and folder feature? Maybe you <maximed>(nevermind, didn't mean to send that) <maximed>AFAIK no such command currently exists. <peterpolidoro>I read my email in Emacs so maybe I can write some helper code within Emacs <unmatched-paren>does anyone know if mercury is bootstrapped? i can't tell from the package definition, but it doesn't seem like it <reza[m]>does someone know how I can open files in gnome with my already running emacs instance? <reza[m]>I start the emacs server in my config, but I am not able to tell gnome to start emacsclient when opening a file with gnome... <reza[m]>probably I need to create a desktop file for emacsclient, but I have no idea how to do that with guix <bjc>the emacs package should have installed an ‘emacsclient.desktop’ file <podiki[m]>yea, probably a .desktop file (you can put that somewhere in home, I forget what places are sourced by default) to run emacsclient <bjc>you may need to be sourcing your profile's init script during login to see it <reza[m]>bjc: but how do I select it as my default application in gnome <bjc>sorry, i misunderstood. i'm not sure how you do that in gnome, sorry <podiki[m]>sorry cant help either, but is a more general gnome question I think of default applications (somewhere in settings?) <bjc>sometimes you can right click on a file in the file browser (thunar?) to do it <reza[m]>but there is no option to select emacsclient... <bjc>in the application launcher, do you see “Emacs” and “Emacs (Client)”? <bjc>how'd you install it? with ‘guix package’? from the system config? home config? <bjc>you should have a file ‘~/.guix-home/profile/share/applications/emacsclient.desktop’ then <bjc>you should also have a ‘~/.profile’ which pulls in your home configuration's profile via a source/. construct <bjc>ok, that's your issue, then. you should have both <bjc>you can, but i'm trying to figure out why it's missing <bjc>that directory is handled by guix, and since *i* have the client desktop entry, i'm wondering why you don't <bjc>i'm using lillyp's patch for emacs 28.1, though, so maybe that's the difference <reza[m]>there is no emacsclient package which I'm missing? <bjc>so it turns out that emacs is still 27.2 in mainline guix, and the desktop entry for emacsclient is new with 28.1 <reza[m]>do you know when this will be merged? <bjc>i am not sure why emacs is still 27.2, maybe dependency hell? but you can use the ‘emacs-next’ package to get 28 <bjc>no idea, sorry. i honestly thought it would have been done by now. normally guix is pretty fast for these kinds of things <bjc>you can create your own desktop entry to do what you want (it's what i used to do pre-28), and just stick it in ~/.local/share/applications/emacsclient.desktop <bjc>that directory isn't managed by guix, so it's safe for you to use <lilyp>bjc: if you use my patch, make sure that you don't mix environments willy-nilly <lilyp>e.g. have you emacs-28 available when starting emacsclient <bjc>i don't. i've only got the one emacs install, and managed by home <lilyp>in that case ./pre-inst-env guix home --container should give you a pretty good approximation (though without graphics unless you're willing to go through pain) <lilyp>at the very least, it should make it so that inside the container everything is 28 <lilyp>adding your own desktop entries outside of guix is obviously a bad idea if you don't know very exactly what you're doing (even experts usually don't) <bjc>it'd be nice if the guix container commands were smart enough to preserve your display/sound settings and bind-mount the required devices <bjc>that's probably asking too much, but i like what it looks like to user <lilyp>I made the suggestion --graphics --sound during guix days already, someone just has to implement it :P <bjc>graphics isn't too hard: it's just x and wayland. i have scripts that do that with nspawn containers already, so i think i could manage that if i get the time <bjc>i dunno about sound though. there's a bunch of backends for that <lilyp>even if you assume sound == pulseaudio, it's nontrivial <bjc>i've done it for pulseaudio (and by extension pipewire, given the compatibility package), but it's definitely more of a pain than display ***benjamin- is now known as benjaminwil
<bjc>lilyp: it's a work-of-art, unfortunately, but as i recall i had to swap the pulse daemon to running single-instance as root, then allow any local connections (ie, over the unix socket), and finally bind-mount the /run/pulse/native socket over <bjc>part of the issue is that at least some of that work was so i could get pulse audio also working with qemu's sound emulation, and i can't remember which was for what <lilyp>yeah, I think we need better solutions than running pulse as root :) <bjc>i agree. the current situation is very frustrating, and it seems like all the powers-that-be have agreed there's no good reason to run your audio server as root <bjc>pipewire is no better. it makes me wonder if i should try just going alsa or something <lilyp>are you saying that POLA is bad? <bjc>when everything is astonishing, nothing is <jab>bjc you could try to run the openBSD sound server. <jab>I ran openbsd for a week or so. It had some quirky behavior but it mostly worked. <bjc>how's compatibility? my uses for sound are basically so my windows vm can use it, so i can listen to music with mpd, and firefox <bjc>i can't switch to a bsd. i need my vms too much, and bhyve doesn't cut it right now for me <jab>bjc: My main issue was that it was fairly quiet. I had to use headphones to listen to it. I believe that the OpenBSD sound server is already ported to linux. <jab>firefox can use it, and mpv can use it too. <bjc>if qemu can use it i might give it a try, then <jab>I do not know how hard it would be to package for guix system, but it might be a decent alternative. <bjc>i'm using manjaro on the desktop until i can get zfs working properly with guix <bjc>this is one of those situations where i wish i could be running guix, though. the prospect of changing a sound server is terrifying, but guix would let me roll back to pulse if it all breaks <jpoiret>bjc: tbh the issue is that there wouldn't really be a uniform interface to just have sound working™ <bjc>you're right. we need another standard! <jpoiret>the xdg-desktop-portal and pipewire shenanigans aim to support that use case, at least for Flatpaks iiuc, but it's still one more thing to wrap over with guix <bjc>sound's always been a mess on unix-alikes. i'm honestly amazed it works as well as it does <bjc>if i didn't have special needs, i would never have noticed how broken it is ;) <lechner>Hi, do I still need LDAP for user accounts once all my equipment runs Guix System? <apteryx>rekado: etc/committer.scm is pretty cool. I've teached it how to commit package removal, and combined with another script it's been happily committing Python 2 packages removal <bjc>lechner: i'm not sure what you're asking here. it would depend on why you need ldap now as to whether or not you need it in the future <maximed>antioxidant-build-system update: Renaming crates is supported (required by rust-rsa and dependents) <maximed>... though now rust-hashbrown fails to build <lechner>bjc: it seems that user accounts can also be managed directly in Guix <bjc>yes, they can. but they just use /etc/passwd et al <bjc>i'm asking because, at least on the unix side of the pond, if you're using ldap for user management you're doing that on purpose for a reason <unmatched-paren>maximed: is antioxidant-build-system a working name, or will it be renamed to rust-build-system or something once it's ready? asking because although i like it, but it seems a bit out of place among the other build systems that are just `(language-name|build-system-name)-build-system` <bjc>so you wouldn't just "get rid of it", since it's only there because someone put it there because they needed it for something <bjc>i like antioxidant; it's charming <maximed>unmatched-paren: I would like to, at some point, split-off antioxidant-build-system into a somewhat separate Guile project that can be used by other distros as well. <maximed>E.g.: "antioxidant configure --prefix=/usr --various-options && antioxidant make && antioxidant test && antioxidant install" <maximed>Found the hashbrown issue. Apparently "rust-rustc-std-workspace-..." crates are not supposed to be actually passed to rustc <maximed>That said, I supposed it could be named "rust-build-system". <maximed>But the build system is written in Guile, and the standard Rust build system is Cargo, not antioxidant ... <bjc>or all the other build systems could get fun names <unmatched-paren>bjc: hmm... I know _exactly_ what i'd rename gnu-build-system in that case :P <bjc>i suggest we cut to the chase and just name all the build systems “bikeshed” <maximed>bjc: wildebeest-build-system, honesty-build-system, parenthese-build-system, ...? <maximed>antioxidant-build-system update: rust-rsa and deps build <civodul>apteryx: just noticed "[PATCH 000/232]", uh :-) <sneek>Welcome back civodul, you have 1 message! <sneek>civodul, apteryx says: weird, here also now <civodul>perhaps worth pushing on a branch and have it built on ci.guix? <civodul>or otherwise push on staging if you're confident? <rekado>apteryx: we now have 100TB on the SAN and another 10TB for SSD-backed storage; these disks currently don’t show up on the OS yet (or at least not correctly), but we could play with those as well. <rekado>I mean: in case booting from the local SSD array with btrfs just won’t work we could put ext4 on the SAN and set up the new system there. <rekado>then move the SSDs to node 129 and reboot as often as you want :) <apteryx>civodul: it's already been built on the ci; see wip-ipython-polyglossia :-) <apteryx>rekado: OK! It'll be fun to benchmark <apteryx>the python2 great purge script is committing 10 removals per minute <bjc>i am extremely jealous of you right now <bjc>where's our federated live streaming platform so we can all share in the catharsis? <bjc>does peertube do live streaming? <apteryx>currently streaming on the rdv-jami-guix jami rendezvous point ^^ <apteryx>hm, doesn't seem to like streaming much. the script has now crashed on 'Deleting package "ptpython2"...' <bjc>i've never used jami before. i guess now's a good a time as any~ <singpolyma>bjc: yes, I've used peertube mostly for live streaming