<privacybuilder>+1 systemcrafters youtube channel - I have used emacs for 20 years and learned a lot from him. <raghavgururajan>Anyway, after c-u --> master merge, GNOME gonna become more awesome. <KE0VVT>raghavgururajan: I miss GNOME 40. <KE0VVT>I can't figure out how to enable the Compose key in this old version of GNOME incl. with Guix. <opalvaults>singpolyma: for all the tinkering I do, I actually really dislike messing with windows managers. I like having all of the basic stuff taken care for me :P <opalvaults>it's no fun and I'd rather be coding or something instead. <raghavgururajan>I would like to do *lot* more improvements to GNOME, after c-u --> master merge. Hopefully once things settled in my life. <singpolyma>opalvaults: that's why I use a tiling WM that handles things for me instead of one where I either have to drag things to where I want them or install/configure more extensions :) <opalvaults>singpolyma: which WM do you use that does everything for you? I've only ever tried things like DWM, BSPWM, etc. *raghavgururajan is currently using StumpWM <singpolyma>opalvaults: usually DWM, I'm experimenting with others as part of my exploring, but nothing beats DWM for out of the box experience <opalvaults>raghavgururajan: that's on my radar. very cool WM. <opalvaults>singpolyma: I used to just pull LukeSmith's DWM config since it had a lot of things set up for me, but honestly Gnome has tiling and is super minimal visually these days so I like it. Material shell is great, for all its quirks, <opalvaults>the lack of gnome 40 is the only reason I haven't reinstalled this desktop with GUIX at this point. <singpolyma>The only config I ever do is some filters and one or two keys, and that's all very optional <jgart>I use sxhkd with dwm. That's all the stump I need <KE0VVT>A Guile Wayland compositor would be sweet. <jgart>I'm not a fan of exwm. I lived in it for some time. I like my 'unix as ide' hack space which emacs is also an equal part of <jgart>I like to think of emacs as an island (or shopping mall on an island) within that space that I can visit <jgart>raghavgururajan, how's st been working for you on Guix System? <jgart>raghavgururajan, that's great to hear! <rekado_>civodul: yeah, been stuck on this treadmill for much too long. <singpolyma>I ran all minimalist for years and am now pivoting to heavy bloat for everything, because I want my daughter's intro system to be as malleable/scriptable/accidentally programmable as possible <rekado_>I keep finding things that need fixing... <rekado_>and then a merge happens and there's even more... <jgart>It'd be great if there was an emacs lisp language server so I can use all the emacs features in vis without opening emacs <raghavgururajan>jgart: I did have a bug. Inside pure-env, when I do backspace, the cursor moves to the right. But you patch for terminfo fixed it. <jgart>singpolyma, I started using Big Blue Bloat. It's pretty nice! <jgart>raghavgururajan, oh that's great to hear! <jgart>I experienced that bug too! hahaha <singpolyma>jgart: why? If you want emacs, use emacs? If you want vis I totally get it but that's like the opposite choice <jgart>I want some emacs features in vis <jgart>I just want to hack it for the sake of hacking it <singpolyma>But isn't the point of using vis to not want features? <jgart>it's not practical, I know or maybe it might be in some edge cases <jgart>I think it is the opposite with vis <singpolyma>Oh, for the sake of doing it. That's always valid of course 😀 <jgart>In vis I want all the features from the outside unix as ide of which emacs is also a part of <jgart>In vis every programming language is game for me to use <raghavgururajan>civodul: Anything changed in c-u-f thats gonna cause mass rebuild? Could it use newer gobject-introspection version from c-u? <jgart>I guess the same could be said for any editor that supports piping in and out of it, etc... <jgart>singpolyma, I'm thinking of the acme philosophy here <jgart>vis has these very useful operators from acme: `<`, `>`, and `|`. And of course, '!' <jgart>you can access them from command mode <jgart>it's a joy to use them to get stuff in and out of the editor <jgart>for example I do this regularly: <raghavgururajan>There is way for software to have as many features, but not being blot. Extensiblity and Modularity. Any number of features can be added by extenstion. Any those extensions can be modded in or out. <singpolyma>I mean, vim can do that with read ! I think it is, but still nice <jgart>the acme commands above allow me to extend to any language in my environment. That's what I love about it <jgart>I like the terseness of the implementation in vis <jgart>It's cleaner and feels more first class. It was thought as to how the editor would function at its core <singpolyma>Yeah. I was about to downshift to vis when I suddenly decided to rocket to emacs instead <jgart>it's all free software so I'll consume it all <jgart>it's not free time though, I guess but it's worth the investment for a robust tool <jgart>emacs needs a unix piping library/framework to get stuff in and out of it <jgart>or a language server protocol :() <raghavgururajan>singpolyma: This was/is in the back of my head for long-time. I am thinking of doing package and service definitions for cheogram system. <singpolyma>raghavgururajan: I have a patch series for the dependencies of cheogram-sip <jgart>I'd like to get the debbugs-mode data out of emacs for example to stdout > vis <singpolyma>I need to do the dependencies of Cheogram proper soon <florhizome[m]>jgart: btw have you seen the treeedit package ? Doesn’t support scheme ofc. <singpolyma>My big series tend to not get looked at though. I think because no one has time to read so many :P <jgart>singpolyma, are you planning to write a guix system service for cheogram? <singpolyma>jgart: no. But I'm not against one being written <jgart>florhizome[m], can you link it? <singpolyma>We don't use system as of now, just foreign in Debian, so packages I need but services not yet <jgart>cool, maybe I'll try to hack one one day <jgart>singpolyma, are all the cheogram deps in guix already? <singpolyma>jgart: no. But after my series for cheogram-sip most of them would be <raghavgururajan>singpolyma: Woah, okay. Cheogram stuff are too good. I'll try to review some. <jgart>raghavgururajan, that would be awesome rg <singpolyma>I haven't tried cheogram itself yet, but all these are deps for that too <jgart>singpolyma, Would you be interested in packaging cheogram itself after the above deps are in? <jgart>I can try working on it also <jgart>I'm currently a jmp.chat subscriber <singpolyma>jgart: yes, actually I will make a package I just only submit the deps and I build the cheogram package itself in our CI, since I change it all the time etc <jgart>It was the best thing to happen to me since vis <jgart>singpolyma, What CI do you use? <florhizome[m]><tex_milan> "Looks Guix configuration is..." <- Have you tried starting x + I 3 from vt ? <florhizome[m]>I am not an expert in the x startup process but maybe something landed in your .config dir that alters the process? maybe slim and gdms defaults aren’t compatible somehow... ***lukedashjr is now known as luke-jr
<florhizome[m]>I’m getting up until the build but it then fails right after starting meson without further info. <patched[m]>Any way to access .git when the source origin method is git-fetch? Cmake (thinks it) needs it to fetch submodules. Is this easily fixable, or does the build system need to be modified? <raghavgururajan>patched[m]: Yes, there is an option for git-download to recursively fetch sub-modules/sub-projects. But it is highly avoided. Generally, we package the sub-modules/sub-projects first as dependencies. <jgart>What do people think of using repology API as a alternative backend to check for updates on packages in guix? <jgart>natewrench123456, it uses linux kernel also <patched[m]>raghavgururajan: Hmm, I think I'll have to restructure the project structure a bit then if it's to work with guix I think. Because the project is git-aware, cmake looks for .git, can't find it, and concludes it is not in git repo and exits. <jgart>would be cool if it used hyperbola's GNU <3 BSD kernel also <jgart>search through hyperbk repos <raghavgururajan>patched[m]: The recursive download is for bundling and vendoring source-code, which is very bad for security, reproducibility etc. Thats why we explicitly unbundle/unbundle stuff. This is quite normal in guix. You can look for examples in guix repo. <raghavgururajan>patched[m]: Personally, I did them in linphone.scm and telegram.scm. <jgart>In an ideal world I'd prefer to run Guix System on hyperbk kernel than on linux <florhizome[m]>Well I would need a bsd -nonguix then but I find it awesome that they work towards that. <patched[m]>raghavgururajan: It seems like a good idea. I think I'll restructure the project to allow for guix packaging. Cmake specifically complains about not finding .git (which I assume is omitted for reproducibility) <raghavgururajan>patched[m]: Sometime cmake script checks for dependencies in the system (build-environment) and if not found, tries to fetch via git. You could try packaging the dependencies first and adding them as inputs to the main package. <jgart>I think getting guile/guix to work on alternative kernels should be high priority for the future <jgart>oh, think I spotted the issue with the above syntax ***lukedashjr is now known as luke-jr
<patched[m]><raghavgururajan> "patched: Sometime cmake script..." <- It doesn't, it's a personal project :^) the cmake code is pretty hacky because the focus was on getting cmake to work. So this should be a good reason for cleaning it up <jgart>florhizome[m], I was on parabola for sometime and I used it then <jgart>These days I want a simpler browser. It's what I hope for <jgart>with easy way to get my bookmarks and history out of it <jgart>ideally there is no sqlite db for that <jgart>florhizome[m], bookmark/history UI like this one: <jgart>there's something about it I don't like. I'd prefer to go through dmenu or fzf/skim/fzy <jgart>I feel nyxt is too slow on my machine <jgart>luakit with luajit is blazing fast <jgart>but qutebrowser is a nicer UI sometimes <jgart>A well patched surf browser might be enough for most browsing for me <jgart>luakit and qutebrowser use sqlitedb <jgart>The thing I like the most about nyxt is the fuzzy tabs search <KE0VVT>pyrotek45[m]: guix install i3-gaps <KE0VVT>pyrotek45[m]: I started Sway from TTY. :P <KE0VVT>pyrotek45[m]: What's wrong with logging into the TTY and then tying "sway"? <KE0VVT>pyrotek45[m]: You could put 'sway' in your '.profile', but I don't know if your graphical terminals would try to start Sway inside Sway if you did that. <KE0VVT>pyrotek45[m]: You would probably have to add conditional statements to your '.profile', to ensure that 'sway' is only run in a TTY and not in other terminals. <KE0VVT>Personally, I like logging into a plain TTY with tmux and/or Emacs, and moving to Sway only when necessary. <KE0VVT>Log in, do stuff in tmux, detach tmux, start Sway, reattach tmux to graphical terminal. <KE0VVT>...launch web browser alongside terminal. <pyrotek45[m]>yeah but i use a lot of gui stuff , although i do enjoy working in the terminal <pyrotek45[m]>i just like having things done automatically when it can be <KE0VVT>pyrotek45[m]: I'm guessing you don't write shell scripts. <pyrotek45[m]>i also have some scripts i wrote to set up gnome with dark mode <pyrotek45[m]>ooof trying to run sway on guix just gave me al sorts of problems <Zambyte>pyrotek45[m]: Really? Like what? I don't daily drive Guix with Sway but I play around with it on my laptop and I haven't noticed major issues <pyrotek45[m]>just ading sway to my config.scm and starting it, things just didnt work <Zambyte>I just run the command `sway` and it launches sway for me <Zambyte>That's really weird... that should never launch i3 o.o <Zambyte>Are you sure it isn't just sway using your i3 config? I think sway will look in the i3 config locations if it can't find sway config files. <podiki[m]>perhaps you'd find useful a package I have (from Arch, gotta submit to Guix) that lets you run your .xinitrc as a session <podiki[m]>so it shows up as an option in gdm for instance <podiki[m]>I use that to start xmonad, and in the past for i3, stumpwm <pyrotek45[m]>but running it with gnome stuff, i added sway to my config and revuilt it <podiki[m]>maybe GDM does, but I wasn't familiar with it (used lightdm a bit before, and before just startx) <pyrotek45[m]>i wonder if its not able to run since gnome is using it? <podiki[m]>I have that and several other random packages to submit, but have been on core-updates-frozen so was waiting for the big merge to clean them up to submit <Zambyte>pyrotek45[m]: When you run sway from the tty, do you get some output and a crash? <Zambyte>Ok. I think I had issues when I first set up my laptop becasue I didn't have elogind installed and running. Do you have that installed? <Zambyte>Note that it seems to be a part of %desktop-services, but not %base-services <zoglesby>Zambyte: I am running sway and using sddm rather than gdm for GUI login <zoglesby>oops, pyrotek45[m] I was pinging the wrong person :) <zoglesby>I *think* the issue with GDM is that it is currently running in X11 not wayland and is not configured to read the .desktop files for Wayland sessions. ***Server sets mode: +Ccntz
<podiki[m]>I haven't customized it (yet) but I see some pretty ones out there <apteryx>rekado: sorry for not seeing this earlier, no objection for me, although we've lost like 20% covereage since, it seems; reading the backlock it seems the problematic commits should have been reverted, so I'm not sure why it's still at that <apteryx>rekado: if we're going to need to go through another mass rebuild, I'd like to know so that I can slip 52269 in, and perhaps a patch to xorg. <lilyp>raghavgururajan: you now when you'll build it, there's a stupid ld error ***molybdenum.libera.chat sets mode: +o ChanServ
<raghavgururajan>So ~/.guix-extra-profiles are not mapped in /var/guix/profiles/per-user/user? <pyrotek45[m]>when i run guix to start in a tty it starts some weird "new seat0" at the gnu login screen <pyrotek45[m]>i was wondering if i could make a service that could start sway on the login <ns12>Hi, what happens to a package when its upstream disappears (i.e. the source URL gives 404)? Will the package be removed from Guix? <rekado>apteryx: a big chunk of rebuilds is due to the php upgrade. Php has since been removed from the inputs of libsoup (better now than later) and there have been changes to librsvg to avoid other problems. <rekado>so we won't be able to avoid this round of rebuilds. <rekado>at least it no longer involves the rust bootstrap. <rekado>would be nice to have a recursive "guix challenge" that takes references into account <efraim>florhizome[m]: let me know what problems you're having with enlightenment (or open a bug so it doesn't get lost!), I'm using it as my desktop, enlightenment/wayland with sddm <pyrotek45[m]>does guix have a one liner to install the package manager on another distro? <mothacehe>on some distro you should also be able to "apt install guix" <jonsger>pyrotek45[m]: there is a fedora package of guix somewhere <raghavgururajan>The `"$GUIX_EXTRA_PROFILE"/etc/profile` has lines to update only PATH? How XDG_DATA_DIRS is also gets updated? <florhizome[m]><efraim> "florhizome: let me know what..." <- So it never stopped working for you? <efraim>not any time recently, I think perhaps at some point I may have had to clear some cache <efraim>sddm session logs are in ~/.local/share/sddm` <pyrotek45[m]>the awesome de doesnt seem to work well once installed with guix <civodul>mothacehe: thanks for your feedback on the --tune patches! <mothacehe>it may bring more questions than answers but yw :) ***atw` is now known as atw
<civodul>mothacehe: right now /workers shows some idleness, even though there are lots of pending builds for c-u-f <civodul>could it be due to a bottleneck in the dependency graph? (a package every other one depends on before they can build something else) <mothacehe>it is extracted from the db-get-pending-build query <mothacehe>and it shows that no builds are ready to be submitted <mothacehe>so yes it has something to do with the dependency graph <awb99>I am trying to build a arm64 image from x86. The build takes forever with guix image --system=aarch64-linux --image-type=rock64-raw <awb99>It has ci and bordeaux as substitute servers, but still is building everything from source. <awb99>forever in my case means 24 hours and still not finished. <zimoun>c-u-f is full of red. Is Cuirass having troubles? <awb99>I tried cross compile with "--target=aarch64-linux-gnu" and it either fails because of meson-build or perl that are not cross-compile compatible. <mothacehe>on core-updates meson is now cross-compilable <awb99>what is this mysterious core-updates all about? <awb99>I have seen so many messages in the forum on it <mothacehe>that we intend to merge back in the master branch soon <zimoun>mothacehe: what is it mean ’? Canceled’? <mothacehe>it means that someone used the "Cancel pending builds" option on the evaluation probably <efraim>pingpongball: the snap daemon has a hard dependency on systemd so I don't think anyone has tried packaging it for Guix yet <mothacehe>zimoun: and yes cuirass is having a lot of issues caused by different factors (gc, publish ...), i posted something earlier today about that <zimoun>mothacehe: you mean bug#52258, right? <mothacehe>i mitigated #52258 but %51787, #50040 and #52182 are still problematic <pingpongball>i mean less bloated and have unix philosophy "do one thing and one thing well" <zimoun>mothacehe: publish (#50040) and GC related (#52182), right? Ah, so the merge is not for tomorrow. ;-) <rekado>pingpongball: no. "bloat" is not a useful concept. <rekado>pingpongball: shepherd is written in Guile so we can easily use Guix things (like containerization) through Guile modules. <rekado>is this a question or a statement? <rekado>booting takes longer than in an ideal situation because Guix System "checks out" the system from /gnu/store on each boot. <rekado>it creates user accounts anew, links up /etc, creates directories in the activation services <rekado>so if raw speed when booting is your primary concern: don't bother. <mothacehe>also shepherd doesn't start services in parallel i think <pingpongball>:( ok i'll not come to guix , but i hope this would be soon fix in future :) <rekado>what I wrote above is a feature. So it's very unlikely it will be "fixed". <rekado>FWIW typing my disk encryption passphrase is the slowest part of booting for me. <rekado>even on the old Thinkpad T60 booting is pretty fast; barely an inconvenience. ***taylan2 is now known as taylan
<zimoun>is it possible to just display the ’phase’ starts and ends? Which verbosity level? <civodul>mothacehe: re CI, thanks for the clarification! <civodul>mothacehe: the publish issues seem to be gone now? or at least i don't recall experiencing it lately <civodul>GC performance has improved but we need to keep monitoring and to get a breakdown of time spent in each phase (finding roots, traversing the db, rm -rf /gnu/store/trash, and "deleting unused links") <efraim>Would it be safe to deduplicate file system extents in the store? <jlicht>bung6[m]: 'USB/DVD ISO installer of standalone Guix System' <jlicht>bung6[m]: If you mean a bootable image that runs Guix: yes. There is no GUI to speak of, if that is what you mean <bung6[m]>jlicht: no. I want to use Guix without setting up <jlicht>bung6[m]: I don't know what you want to do, I'm sorry. Are you running into any problems trying to use the image? <rekado>bung6[m]: the bootable installer is a full Guix System. You can use it right away. <nckx>Interesting, a master-rebuilds commit that becomes c-u-rebuilds when merged into c-u. I wish I'd come up with that on purpose. <rekado>Guix System doesn't come in pre-configured flavours, so the barebones installer system is a valid live system. <nckx>bung6[m]: It's live, it just doesn't include a desktop. <rekado>there is no bootable live system with an installed desktop environment <nckx>It's a live system for installing GNU Guix. <nckx>rekado: Did reverting it fix everything? <nckx>I don't know what more to say at this point. <jlicht>bung6[m]: it is a live installer; live + installer <nckx>You can. That's what ‘live’ means. Have you booted the installer? You can switch to a text terminal with Ctrl+Alt+F3. <nckx>From there, assuming you have enough RAM to contain new packages or start the cow-store service on a mounted writable partition, you can install whatever you like. Possibly reconfigure? I haven't tried that. <nckx>That's the only live system we offer. Maybe others offer others with more packages pre-installed. <zimoun`>efraim: I used fdupes and other similar tools, to avoid «-80 fridge effect» on shared ressource. Well, I do not remember exactly but the resulat are not as good as expected. ***zimoun` is now known as zimoun
<nckx>Ctrl+Alt+F3, bung6[m], as noted above. <nckx>Then you saw the live system? <nckx>There is no ‘live mode’. The installer is a live system. <nckx>efraim: It's transparent deduplication, so Guix won't know or care, but you'll sliiightly increase fragmentation. Not generally an issue on SSDs though. <nckx>bung6[m]: If a VM's an option, you can try the qcow2 link above that efraim originally suggested. It's also live but it includes a desktop too. <nckx>bung6[m]: If you don't want to install, you can just ignore the installer. 😉 <bung6[m]>but ı can't install GNU Guix on the vm :d <bung6[m]>trsiquel is easy but GNU Guixis not easy <bung6[m]>s/trsiquel is easy but GNU Guixis not easy/trsiquel is easy but GNU Guix is not easy/ <nckx><but ı can't install GNU Guix on the vm> I think you can, but I agree, it's not easy. <nckx>But then we have an installer which can :o) <bung6[m]>How do I know if the system is compatible with hardware? <nckx>If it ran Trisquel, it should run Guix System just fine. I also recommend having a few 10s of GiB free for /gnu/store. That's not a minimum at all, but it makes working with the system whilst getting to know it much easier. <nckx>If it didn't run Trisquel or some parts need non-free firmware, it might not run Guix System well or at all. <roptat>I've been working on offlate this weekend, I hope I can make a release soon :) <roptat>the new version will be able to interact with weblate <roptat>but I still get errors sometimes when pushing changes, and I was hoping I could work on interactive conflict resolution, when updating files from weblate <roptat>but I can't find a way for qt to stop a thread and wait for user input <mothacehe>civodul: re publish issue: it is less frequent but still there <mothacehe>and i'm reading gc.cc to see how to introduce some metrics <civodul>mothacehe: nice; it already prints something at the beginning of each phase <roptat>civodul, ah I forgot about it. I think I fixed the issue already, so I'll make a release tonight <civodul>i think reversed IPv6 addresses was the main issue <civodul>but i wonder if vivien and others had other problems <civodul>mothacehe: i think Cuirass commit 6dcf2f65cea920b9b1c265de3e2b0abe0048a08e is not quite correct because 'catch' is too high <civodul>it should instead wrap individual 'register-gc-root' calls <vivien>I didn’t have any other issue, and having static networking would help me a lot :) <awb99>pine64-plus-installation-os .. will this installer work for the "old" pinephone? There is the pinephone-pro also, perhaps the plus means pro ? <mothacehe>civodul: i'm not sure to follow you. The derivation-path->output-paths can possibly raise an exception if drv doesn't exist. So the exception need to be ignored at least at register-gc-roots procedure level, right? <samplet>civodul: The PoG reports are back up. Thanks for the heads up! <civodul>it's surprising that the miss rate is higher for tar.xz than for tar.gz <samplet>A lot of the missing XZ tarballs are from older commits. Starting with commits from last year, the hit rate is more like 90%. <samplet>I wonder how much of that is older versions of linux-libre. <csantosb>I get this running: guix time-machine -C channels.scm -- shell -m manifest.scm <samplet>We churn through a lot of versions of that package, and if SWH didn’t crawl it, it would represent quite a lot of missing .tar.xz sources. <roptat>samplet, what do you use to generate the graphs? <csantosb>But not this: guix time-machine -C channels.scm -- pack -m manifest.scm <roptat>how do you create the relative graphs? <csantosb>Seems that python-pytorch and tensorflow cannot be installed together. <samplet>roptat: Scheme -> gnuplot -> graph – most of the heavy lifting is done in Guile. <roptat>so it's not gnuplot that takes care of that, you just generate data that are between 0 and 1, cool :) <rekado>csantosb: what's the issue with pytorch and tensorflow? Some conflict? <csantosb>rekado: yeah, a conflict with python-protobuf. <csantosb>So 'guix install python-pytorch tensorflow' issues the same error message as 'guix pack -m manifest' <csantosb>But 'guix shell -m manifest' does it ... <zimoun>samplet: some linux-libre related things had been recently added to sources.json but I do not think nixguix loader ingests it. IIRC, nixguix loader filert out based on extensions. <samplet>We also changed to running the linux-libre deblob scripts locally, too. <samplet>For PoG, that means that we are checking the upstream linux sources instead of the linux-libre sources. In a forthcoming report, I have data on the preservation of the deblob scripts (and other plain text sources). <attila_lendvai>in a gexp i can write #$(file-append some-package "/bin/foo"), but how can i refer to a non-default output of some-package? <attila_lendvai>i can use (string-append #$geth:clef "/bin/clef"), but that happens at runtime. i'd prefer the former. <civodul>attila_lendvai: currently file-append doesn't let you choose a specific output <civodul>roptat: we need Translation of Guix reports! :-) <attila_lendvai>civodul, thanks! i'll settle with the latter. but then... does it make sense to package several binaries into different outputs? even though these binaries are rather different... (geth -- ethereum node, clef -- a key managemend daemon, evm -- the ethereum vm, etc). but still, does it make sense? <civodul>mothacehe: it could be one of OUTPUTS that doesn't exist, in (for-each ... (derivation->output-paths drv)) <notmaximed>maybe file-append can be extended with an optional #:output "clef" argument? <civodul>attila_lendvai: dunno, it depends on the package i guess <attila_lendvai>is there a general principle when deciding about multiple outputs? maybe this with geth: a user who wants clef in his bin/ may be unhappy to get 5 other binaries? <civodul>attila_lendvai: in general you'd decide based on size <notmaximed>The typical reason for multiple outputs is to reduce disk usage. E.g., no need for 50M of html documentation if the user usually doesn't read it <attila_lendvai>these binaries are 20-40 MB's, so i guess i'll keep them in different outputs <civodul>attila_lendvai: are they statically linked? <attila_lendvai>civodul, yep, golang static binaries (although they link to libc.so, libpthread.so, etc) but all user code is static <winning-luser>`guix-daemon --build-users-group=guixbuild' as root and we're good to go. but i dont get why <roptat>civodul, you mean guix domains, or translation of packaged software? <notmaximed>Does anyone know if the standard go supports dynamic linking? (Searching for 'go dynamic linking' doesn't turn up anything useful) <winning-luser>i should just get off fedora and install guix fully already. sigh <notmaximed>winning-luser: Failed to locate executable /var/guix/profiles/per-user/root/current-guix/bin/guix-daemon: Permission denied <notmaximed>I think, but I'm not sure, that guix-daemon is supposed to run as root. Does the SystemD service run guix as root? <winning-luser>that service is ran as root from my understanding of systemd <winning-luser>and the first thing i did was check the permissions, which seems fine: lrwxrwxrwx. 1 root root <zimoun>samplet: I do not think the deblob scripts are preserved, yet; at least no ingested yet via sources.json and not reachable IIUC. Some discussions were floating around to move to git-fetch; which would easy both ingestion and lookup, IMHO. <notmaximed>winning-lusers: looking at the 'l', it appears to be a symbolic link (which is fine), so you might need to check the permissions of the file it points at <notmaximed>(also, maybe check the permissions of the parent directories?) <jpoiret>winning-luser: can you do a paste of `systemctl cat guix-daemon.service` as root? <mothacehe>mbakke: thanks for the django/ganeti fixes :) the ganeti system tests are repaired <mothacehe>patchwork is still broken though, any taker? <winning-luser>notmaximed: the real file's permissions: -r-xr-xr-x. 1 root root <jpoiret>any eta on c-u-f, are there any remaining blockers? <jpoiret>i'd like to submit some xdg-desktop-portal patches, but i'd need the pipewire bump on c-u-f. <jpoiret>are you running with SELinux perchance? <winning-luser>i think fedora might have it enabled by default but im unsure. ive never interacted with it directly myself <zimoun>samplet: yes but who knows how they are saved. :-) <jpoiret>`getenforce` should tell you if that is the case <jpoiret>you could disable it if you understand what that entails <winning-luser>well, switching to root and manually running the daemon works for now <jpoiret>maybe that's just systemd doing its own things <winning-luser>yeah im running guix pull fine now from my regular user account <winning-luser>i hardly understand systemd so i dont know how to troubleshoot this myself sadly <jpoiret>maybe journalctl could tell you more about that? <jpoiret>iiuc, there might not be a proper SELinux policy for systemd to execute binaries outside of /bin and /usr/bin <jpoiret>it should work properly with SELinux disabled I think <jpoiret>the systemd unit file is pretty simple <winning-luser>yeah, ill try doing that and report back. cant right now as it'd require a reboot iirc <mbakke>winning-luser: did you install the SELinux policy and relabel your filesystem as outlined in the manual? <rekado>our SELinux policy likely won't work <rekado>it hasn't been kept up-to-date with Fedora releases; since the base policy in Fedora has changed it's unlikely that our policy will work without modifications. <rekado>we'll need some Fedora users to install it, set SELinux to permissive, and then check the audit log as they run Guix commands. <mbakke>I'm testing it only on RHEL8, but don't think Fedora should be much different. <mbakke>there are still packages that fail to build with the current policy (last updated in May), but day-to-day operations with substitutes should be OK. <res0Nanz>Hi! How do I specify the minimal version for a dependency input? in a package definition. <notmaximed>res0Nanz: guix always uses the version specified in the inputs <res0Nanz>for example (native-inputs `(("cmake" ,cmake))) gives me cmake-3.16.5 <notmaximed>So if you use `(("coreutils" ,coreutils)) in the inputs, it uses the version of coreutils specified in the variable ‘coreutils’ <notmaximed>To use a different version, do something like cmake-3.16.6 (except that is not defined) <notmaximed>I don't know if other versions of cmake are currently packaged <notmaximed>Due to technical reasons (world rebuilds), the ‘standard’ cmake package cannot be updated directly, but you could define a cmake-next variable with version 3.16.6 of cmake, and use that in your package definition. <notmaximed>res0Nanz: is your guix up-to-date? I have version 3.21.1. <res0Nanz>Yes `guix shell cmake` load cmake 3.21.1 <notmaximed>Are you using cmake-build-system? If so, you can choose which cmake to use with #:cmake <notmaximed>‘It automatically adds the ‘cmake’ package to the set of inputs. <notmaximed> Which package is used can be specified with the ‘#:cmake’ <acrow>Sadly -- I am confused about this oughta be basic home configuration detail. <tex_milan1>Hello, I have bluetooth service enabled in system config, autostarting it, but when I run Bluetooth Manager as a user or root it complains about org.blueman.Mechanism not provided by any .service. Bluetooth works. Any ideas? <acrow>tex_milan1: Yep. I see that too with xfce but it appears to be extraneous because it works. <tex_milan1>@acrow Yeah, but really very annoying to see it every boot... Also what bothers me is it is not (manager) started automatically. NetworkManager applet is started automatically for example. <tex_milan1>Also see that a2dp-sink profile connect fail due to protocol not available. Do you see it too? <tex_milan1>Dec 6 20:39:36 localhost bluetoothd[837]: src/main.c:parse_controller_config() Key file does not have group “Controller” <tex_milan1>Perhaps guix bluetooth service generates invalid config? What is "Key file"? <acrow>Hmmm... I agree and have struggled with it but not cracked it. Among the complications that I've encountered are that I run linux-libre on a machine with integrated bluetooth that is not supported. So, I use a ZEXMTE USB bluetooth adapter instead. I'm not certain that the error is not due to kernel mischief surrounding an incompletely supported bluetooth adapter. Could you have a similar complication? <acrow>Is your bluetooth adapter supported by the kernel? <acrow>My native bluetooth is incorporated on an unsupported: 03:00.0 Network controller: Intel Corporation Centrino Advanced-N 6235 (rev 24) <acrow>So, I have resigned to the inevitability of this error message bc the first bluetooth adapter encountered is necessarily broken. <raghavgururajan>tex_milan1: Blueman works despite that error "org.blueman.Mechanism". <tex_milan1>I now that blueman works. But the error is here every time I boot and this is something I cannot live with :) <acrow>NixOS doesn't run a linux-libre kernel, does it? *raghavgururajan is using CSR8510 <tex_milan1>yes, BT devices works as good as they can with pulseaudio (swith to pipewire is my next thing to do) <lfam>I do not believe that NixOS uses linux-libre by default, although I do think it is available <Myrcy>that was the situation I last heard of with NixOS as well <raghavgururajan>tex_milan1: We cannot provide support for stuff related to vanilla kernel, for ethical reasons. If you are facing something regarding bluetooth-stack, blueman and linux-libre, feel free to ask us. <tex_milan1>i understand. and this has nothing to do with kernel. it is isssue in bluetooth-stack <tex_milan1>oot@t14s /home/milan# cat /etc/bluetooth/main.conf <tex_milan1>No wonder it complains it cannot find [Controller] section there... <raghavgururajan>bluetoothd's configured in guixy-way via bluetooth-service definition. <raghavgururajan>It should be under network services and/or desktop services section in the manual. <lilyp>apropos bluetooth, does the applet finally work in GNOME 40 or do we still have to debug it? <raghavgururajan>tex_milan1: I am afraid you'll have look at the source code of bluetooth-service-type for options. <tex_milan1>yeah, going to hack guix. that is what bring me here, I was using NixOs, great and stable system, perfect desktop, but I wasn't able to learn their DSL... Guille is extremely simple comparing it to it. <acrow>Any thoughts on getting the home-mcron-service-type configuration working? I imagine this has to do with properly embedding the gexp. <acrow>Cake, in comparison to the BT issue. <acrow>I've been fiddling different combinations for quite a while now and am frustrated. It's probably super simple. <lilyp>Probably is, but we're short on Gurus <lilyp>You might need to stick around for a while before someone with knowledge sees this. <podiki[m]>have you looked for any other guix home configurations someone has? there are some out there... <acrow>It is odd because it reconfigures without error on a foreign distro. This problem seems to only appear on GuixSD. I think maybe shepherd is implicated. <apteryx>configure flags as gexp: can these be further refined via substitute-keyword-arguments? <apteryx>I'm getting: /gnu/store/2fn2csfqrdy63clsmfx2wl078ywm27lc-llvm-next-12.0.1-builder:1:3989: Unknown # object: "#<" <acrow>Hmmm... I haven't seen that one. <podiki[m]>I ran into the same thing the other night (trying to modify from an (inherit llvm-11)) but didn't figure it out <apteryx>I think we can extract the raw sexp from a gexp, but that seems... an unsatisfactory solution <lilyp>acrow: you did already try the commented MWE, did you not? <lilyp>(if you don't have mbsync, you can omit the first job) <civodul>apteryx: yes you can use substitute-keyword-arguments regardless of whether the original flags are a gexp or not <podiki[m]>would also like to hear how the mcron-service config goes, still want to move at least some of my config to guix home <sneek>Welcome back civodul, you have 2 messages! <sneek>civodul, RhodiumToad says: re. (aif), it's worse than that, it also doesn't work at all if used inside another macro, e.g. (define-syntax-rule (foo bar) (aif bar (display it) (display "no"))) can fail with 'it being unbound <sneek>civodul, RhodiumToad says: you're correct that a syntax parameter is the correct solution of course <acrow>I've tried to emulate 10.9.2's example, using "guix gc -d 1m" lazy way. <civodul>apteryx: see for example petsc-openmpi ↑ <lilyp>you don't want to start with guix gc <acrow>Well, as I described I had already seen successful reconfiguration on a foreign distro -- putting the same in GuixSD seems to be the problem. Looking over the module requirements again. <apteryx>civodul: ah, so ungexp then proceed as normal (in an enclosing gexp). Thank you! <civodul>apteryx: yes; the only pitfall is if you mistakenly insert a gexp in an sexp, in which case you get the error you showed <acrow>Of course, maybe I need to provide the whole thing. Maybe the excerpting is problematic, but I suspected this would be quickly diagnosed as a misquoting problem... <podiki[m]>civodul: is that the same basic idea if I wanted to inherit llvm and change a configure flag? (one of the regular strings, not the #$ parts) <lilyp>pro tip: always read the generated files :P <podiki[m]>I still haven't spent the time neede for gexps <acrow>Ok, I think I've found my error. <acrow>Lemme confirm by getting back to minimal and then I'll share. <tex_milan1>How to test my changes done to service on GuixSD? sudo ./pre-inst-env guix system reconfigure /etc/config.scm doesn't work... <acrow>Problem was incomplete module listing. Without (guix) and (gnu) things fail and you don't get the thoughtful easy to read suggestion that you've left out a module. <lilyp>acrow: you probably don't want to require (guix) and (gnu) *inside* your mcron code, though <lilyp>tex_milan1: friendly reminder that you can install Guix on media other than SD cards <tex_milan1>eh, maybe I said it wrong. I have Guix distribution on my laptop. <tex_milan1>I followed the manual, I have checkout of guix, pre-inst-env can make packages, but how to apply it to the system reconfigure? <attila_lendvai>tex_milan1, try without pre-inst-env. or do you have change to the service's *code*, or only its config? <tex_milan1>change to the service's code. in gnu/service/desktop.scm... <acrow>The ix.io posting, I now can see, has several issues. If anyone wants to see my full, working home-configurations.scm, it would only create marginally greater embarrasement. Let me know. <attila_lendvai>tex_milan1, if you changed code in the guix repo, then you need to commit it, set up channels that it points to your local repo checkout (as opposed to the official repo), then guix pull it, and only then guix system reconfigure <attila_lendvai>tex_milan1, you'll need to add your pgp key to the keyring branch, add a channel-intro commit, and whatnot. it's not trivial, but once it's set up, it works. alternatively you can use qemu to test your changes, which might actually be simpler. <lilyp>tex_milan1: sorry, but attila_lendvai is exaggerating here <attila_lendvai>tex_milan1, the general idea is: $(./pre-inst-env guix system --no-graphic vm foo.scm) -m 2048 here foo.scm evaluates to a marionette-operating-system instance <lilyp>you can absolutely do guix system from pre-inst-env, but I'd be careful with reconfigure <lilyp>hence their suggestions to use channels and vm <lfam>I use `./pre-inst-env guix system reconfigure ...` often <lilyp>my personal suggestion would be to edit the code and then do a simple `guix system build` <lfam>It works as expected. The only thing is you lose provenance information <tex_milan1>@ifam do you have only guix channel or also the other? <lilyp>it won't work if you require third party channels <lfam>I do use GUIX_PACKAGE_PATH instead of channels <lfam>It's faster and there's no trouble <tex_milan1>oh, third party channels I need otherwise I get a brick. <lfam>But you lose provenance etc <lfam>I do use a custom kernel via GUIX_PACKAGE_PATH <attila_lendvai>yeah, i think i have this assumption because i have that other channel, too. does it also work when i have a export GUIX_PACKAGE_PATH=/some/other/repo ? <roptat>at the very least you can always use -L /path/to/other/repo <apteryx>tex_milan1: what about your hardware makes it a brick with linux-libre? <lilyp>probably "AMD works with free software" <awb99>Is it possible to add a directory to an image tht gets build with guix system image? <apteryx>tex_milan1: I see. Make sure to send a thanks letter to AMD for making your life more miserable. <tex_milan1>but they sold it without windows!!! :D and a top of that with 20% discount. and at the time of cpu shortage :D <lfam>Do we want to try to prevent crawler bots from downloading substitutes? <lfam>They aren't indexing a web page, but rather downloading software <lfam>It's a waste of resources <lfam>For example, the Bing bot is downloading substitutes <apteryx>tex_milan1: amdgpu is 125k lines of code in the kernel that is virtually useless without a binary blob ^^ <tex_milan1>well, yeah. and they are Chinese company :) what can I do, but hey, I can play CSGO well :) <tex_milan1>i would think twice before buying it, but because of problems with TSC caused by its BIOS. Because I want to play games... <roptat>tex_milan1, you don't need to justify yourself ;) <attila_lendvai>make-forkexec-constructor creates the log file as root? i.e. not as the provided #:user... is that a bug? or am i missing something? <tex_milan1>no problem. i would actually prefer something more open-hardware. but requirements are requirements :) <attila_lendvai>so, in shepherd, in exec-command, the log file creation is prior to the setuid call. is that a bug? <roptat>I think syslog also creates files as root, so I'd say it's expected <roptat>and probably, this is also the case of systemd's journal <civodul>lfam: how do these bots find links to substitutes though? <lfam>I don't know civodul, I was only looking at substitute requests <lfam>Maybe by trying to crawl the ci.guix.gnu.org web interface? <civodul>not sure it contains the actual nar links, does it? <lfam>In any case, it won't hurt to create robots.txt, right? If somebody is choosing to use these user-agents, their requests will still succeed <lfam>I see requests such as this: <civodul>yes, we should definitely add robots.txt <lfam>And the IP address that's used is registered to the semrush organization <lfam>I agree it's strange that they can find URLs <podiki[m]>do we ever put anything in /gnu besides store? <florhizome[m]>when the python importer needs like 5 minutes to produce an output, and you want to ask if it’s working for others rn and the it posts just as you want to send –.– <acrow>It's exactly then that you realize that patience is a virtue! <acrow>Especially the patience of others. <apteryx>civodul: is it possible to use package-with-relocatable-glibc from outside of (gnu packages make-bootstrap) ? <apteryx>I tried exporting it but make fails with 'error: coreutils: unbound variable' <civodul>apteryx: it wasn't meant to be exported <civodul>and perhaps there are module cycles you're seeing here? <apteryx>I'm trying to use it from (gnu packages llvm) <attila_lendvai>is there a way for a shepherd service to communicate that it depends on certain packages? <civodul>apteryx: at first sight i'd say "don't do that" :-), because package-with-relocatable-glibc is really special-purpose <civodul>but we can always discuss the use case <jeko>Is there a version of guile-gnutls compatible with guile 3 or only guile2 at the moment ? <civodul>jeko: it's compatible with Guile 2.x and 3.0 <mbakke>ooh, when did weather learn to estimate size, nice :-) <jeko>civodul: I can see the guile3.0-gnutls package is deprecated. so the only option is to build it manually right now ? <mbakke>...although 32 Exabyte seems a little off <civodul>mbakke: yes, that's because of the bug with flightgear and repeat-masker <civodul>jeko: actually "guile3.0-gnutls" doesn't exist anymore AFAICS, but it used to be a deprecated alias for "gnutls" <jeko>civodul: ok thank you for the clarification! So if guile -c '(use-modules (gnutls))' say no code for module (gnutls) it means I have to alter the guile load path by myself ? <jeko>I'm in a environment generated with guix shell gnutls <mbakke>jeko: try adding guile to the shell <mbakke>which sets up GUILE_LOAD_PATH accordingly <jeko>mbakke: bingo ! Thank you. <attila_lendvai>when i do `guix system --no-graphic vm foo.scm` then the terminal doesn't seem to have a right margin; i.e. long lines are cut off at the edge of my xterm. any ideas how to fix that? <attila_lendvai>less also doesn't work properly. i guess it's due to linux sending the output through an stty. <jeko>See you Guixters ! Thank you for your help *jeko is going to sleep zzZ