IRC channel logs
2023-03-31.log
back to list of logs
<mirai>shared-mime-info isn't one of them though you'd be missing a lot of support for the newer file formats since their mime types will remain unknown <mirai>though to upgrade shared-mime-info you will really want to integrate docbook-xml changes <mirai>since it fails to build otherwise <mirai>are the builds already going? <jpoiret>We've neen working on the branch for a month <jpoiret>We can't just keep pushing the goalpost further by adding new changes <jpoiret>All the core-updates changes since last month have been to fix stuff on the branch not add new things <mirai>is it fair for changes that do not update a package then? <mirai>as in, they fix existing ones <jpoiret>Well, that's less clear, but from what i've seen, the policy is: if it's not a regression from master, we don't care <jpoiret>But that's just while the branch is getting sorted out and then merged <mirai>if it's unsuited for the nearest core-updates window, then... what happens next <jpoiret>Hum yes, this patch series looks quite good <jpoiret>You should just keep it around until after the merge and bring it up again <mirai>sneek: later tell lfam it might perhaps require a G-Exp restyling <quidnunc>I installed guix 1.4.0 on Ubuntu Lunar via apt. When I install a package it doesn't show up in the bin directory of my profile. Does anyone know what I can do to debug? I'm at a loss. <bumble[m]>would someone tell me what I am doing wrong? I tried to make a package that "installs" a local bash script. The install process completes, but command isn't found afterward http://paste.debian.net/1275831/ <bumble[m]>"cactus file manager" requires an env var $OPENER to be defined for openeing files. My goal is to make the bash script available as an executable to be used for this purpose <mvnx>bumble[m]: I tested it out and it does install it, but it loses it's permissions. Not sure if that can be defined within an install-plan though. <bumble[m]>@Agent Kilo @mvnx thanks for helping me. I will look at #:install-plan and maybe `#:recursive? #t` for the local-file call... <bumble[m]>I don't know how to determine for myself if it was installed <bjc>‘guix build $package’ will print out the path to the package in the store <bjc>so, for instance, ‘ls -l $(guix build foo)’ will allow you to look at what was copied in <mitchell>Hi guix, what do you think about adding some code to the gnu-build-system which adds inputs inputs to native-inputs when cross compiling to help pkg-config and friends pass configuration. It is always hit or miss whether configure will succeed when cross compiling <bumble[m]>@bjc thank you guix build openersh shows package not found, but guix package --list-installed lists the package <iyzsong>michaelll: when cross-compile, build tools needed to to native-inputs, but libraries needed to be in the target arch. <iyzsong>PotentialBear-61: that depends on the build tools used (make, cmake, etc), usually via modify configure-flags or 'CFLAGS' in the package definition. <PotentialBear-61>iyzsong: so I want to apply this to packages like GAP. is there a way to add it to an existing package? <mitchell>iyzsong: I know this. What i'm refering to is when cross compiling the configure script looks for dependencies using the host tools which it fails do to the fact the cross toolchain uses different environment variables. To fix this you need to add the inputs in question to the native inputs to make the configure script happy <iyzsong>mitchell: that's sounds like the configure script should be fixed to check target dependencies instead of host ones.. <mvnx>bumble[m]: If you `find /gnu/store -type f -name sh.opener.sh` you will find it. <bumble[m]>@mvnx thank you it is appearing in the output of that command <bumble[m]>strangely there are sh.opener.sh and one of them is exectuable... the package was updated to use "#:recursive? #t" with local-file and "install-plan" as recommended by @Agent Kilo perhaps that is why <mvnx>What list of `source target [filters]` did you use for `#:install-plan`? <bjc>bumble[m]: this is just a guess, but you might want to try installing into "/bin/sh.opener.sh" <bjc>(as the final argument to the string-append, that is) <bumble[m]>though the command script is found in /gnu/store it only runs if the full path is used to call it, otherwise command not found <bjc>yeah, that sounds like it needs to be in a ‘bin’ directory so it can be added to the right place in your profile <bumble[m]>@bjc if the string-append is updated to use "/bin/sh.opener.sh" the install fails file or directory not found... I'll try updating the install-plan to use ./bin/ as the destination <bjc>ah, right, sorry. had a wee brain fart. the install plan is the correct place to edit it <mvnx>It worked in that it installed to the bin directory but still not on my path - is it on yours? <bumble[m]>@Agent Kilo @mvnx @bjc thank you :) @mvnx yes it is on my path <mvnx>From your most recent paste, all you changed was the 2nd arg (target) for install-plan to "/bin"? <mvnx>Thanks! Glad you got it solved. I was going to need to get to something like this eventually too <jpoiret>bumble[m]: as a side note if you're only adding a file like this, you don't really need a full package declaration, (local-file) is enough <bumble[m]>okay thanks I will use local-file from now probably <etienne_>Morning everyone! Can we specify a set of channels per profiles? <civodul>etienne_: hi! the set of channels can be per-user, with the ~/.config/guix/channels.scm file <civodul>or you can specify a channel file, as in "guix pull -C my-channels.scm" <etienne_>Thanks! A clarifying question: I basically want to constrain guix install to only sift through specific channels, and thought of using profiles. Is that the best way to do that? <civodul>channels specify what "guix pull" (or "guix time-machine") fetches <civodul>so you could provide ~/.config/guix/channels.scm or /etc/guix/channels.scm so that any "guix pull" gets precisely those channels <jpoiret>civodul: do you know what package provides ldd? <rekado>/gnu/store/ayc9r7162rphy4zjw8ch01pmyh214h82-glibc-2.33/bin/ldd <jpoiret>i wonder how we're going to get that zig sigsegv resolved <jpoiret>the binary seems to fault on libc's getenv, called in LLVM's init during dynamic linkage <civodul>jpoiret: sounds like an interesting puzzle :-) <civodul>jpoiret: ldd is a mere shell script that sets LD_TRACE_LOADED_OBJECTS=yep <nom[m]1>is it possible to symlink a writable config file with home-files-service-type <mirai>jpoiret: I recall you saying that you didn't like the record inherit thing (or something similar?) <jpoiret>I often just use LD_DEBUG=, it's quite useful <jpoiret>mirai: It's not ringing a bell, but I'm not saying I haven't had some hot takes <jpoiret>inherit is bad when combined with inter-field references <mirai>what alternatives do you have in mind? or what makes it “bad” <jpoiret>i don't think there are direct alternatives in Guile (except maybe GOOPS but I don't like it either :p) <jpoiret>but in general i've been happily using it, you just have to keep the caveats in mind <mirai>oh, the alternative doesn't have to exist yet, but what other abstract mechanism were you considering <jpoiret>oh, and yeah when you inherit a package you break guix refresh :) <jpoiret>but that's more Guix's fault I would say <jpoiret>records are fine, but I guess the semantics of field references inside other fields could take inheritance into account <jpoiret>that would make it more natural i guess <civodul>jpoiret: re inter-field references, there's 'this-package' <jpoiret>but really I do think they are fine as they are now, especially in a language like Guile <jpoiret>civodul: right, but that's what I meant, as long as you know the pitfalls it's fine, but it's still unclear in the beginning <jpoiret>could it be called 👀-bot? or would IRC disallow UTF-8 names? <apteryx>arg, I keep hitting: 'error: rust-cargo-lock-8: unbound variable' on staging, even after make- clean-go <apteryx>it's nowhere to be seen in the code, AFAIK <apteryx>it manifests itself when I run 'guix refresh something' <mirai>is 'spawn' unavailable within guix? <mvnx>Is it normal that user `herd status` only outputs `error: connect: /run/user/1000/shepherd/socket: No such file or directory`? <mvnx>I understand I can use `sudo herd status` to see system statuses. <mirai>mvnx: if you have no user shepherd running yes <PotentialUser-12>Hello. I can't find in the Guix documentation if it honors the "no_proxy" variable ? I have found various places where it's said that guix-daemon would honor the HTTP_PROXY and HTTPS_PROXY env variables, but nothing about NO_PROXY. Does someone know if that's the case ? <mvnx>mirai: I guess I thought it would have come with initial Guix home setup. I added it manually and then it complains because no `/run/user/1000/shepherd` which was some on-first-login error about /var/run I keep getting when I sign in - which I also thought the initial home manager config would set up. <apteryx>shouldn't it be possible to call a procedure like: #$(version-major+minor (package-version cmake)) inside a phase (which is inside a gexp) ? <apteryx>I get: error: verson-major+minor: unbound variable, although the top level module has #:use-module (guix utils) <jpoiret>mirai: I think guix uses guile 3.0.8 <jpoiret>TIL about `(pixel-scroll-precision-mode)` <jpoiret>civodul: do you still want to organize a core-updates hackathon? <jpoiret>I could try to coordinate this (with Andreas if he's available) if that's a problem. The question would be when mostly <jpoiret>advertising it in advance on guix-devel is probably essential <jpoiret>(yes I want to avoid dealing with the zig segfault issue myself, why do you ask?) <mirai>I see some gnu/packages/linux.scm comments with ";; XXX TODO: Replace with simply #$OUTPUT on core-updates." <nalaginrut>hi folks! I want to install minetest in gnu/packages/minetest.scm, not in gnu/packages/games.scm, how should I do? <jpoiret>I don't see any minetest in gnu/packages/games.scm? <nalaginrut>hmm...it exists in my machine, and there're two minetest in different scm file <nalaginrut>maybe I should reinstall guix, how can I remove the old version completely/ <jpoiret>they appear when you do `guix show`? <nalaginrut> commit: a0178d34f582b50e9bdbb0403943129ae5b560ff <jpoiret>this is v1.3, have you `guix pull`'d since? <jpoiret>it is almost two years old at this point <jpoiret>no, I just think you don't have `guix` setup correctly <nalaginrut>oops, it should be ~/.config/guix/current/bin/guix <jpoiret>you should be sourcing `~/.config/guix/current/etc/profile` inside your .profile <jpoiret>specifically, you should be doing `GUIX_PROFILE="~/.config/guix/current/"; . "$GUIX_PROFILE/etc/profile"; unset GUIX_PROFILE` <jpoiret>how did you install Guix? through debian? <nalaginrut>I didn't use guixSD, I installed guix with the binary under debian <jpoiret>yes, I don't really get why it wouldn't just work out of the box then, but alas <jpoiret>do you have a file in /etc/profile.d/ related to guix? <nalaginrut>I think it's because the path to guix, now it seems back to the correct version <jpoiret>this is getting ridiculous, how much space do i need to build my whole usual manifest? I had 33G free this morning and now i'm getting ENOSPACE <apteryx>sounds about tight for chromium? haha <apteryx>just unpacking its source consumes like 4 GiB... likewise for icecat. <mvnx>Not really a Guix-specific question but does anyone have good resources on the login process? I'm getting confused because the Arch doc for 'getty' suggests that agetty and mingetty are alternatives to each other but Guix includes both in `%base-services`. Then now is my first time trying to use a Wayland compositor and I'm reading it requires elogind instead of the base login. So I recognize a knowledge gap. <jpoiret>mvnx: agetty and mingetty don't run on the same ttys <jpoiret>basically, you could run agetty on tty1 and mingetty on tty2 <jpoiret>also, elogind is orthogonal to those: it manages who's logged in, and not necessarily through a tty. It relies on PAM to tell it who logged in and where, and then sets some env variables depending on it <mvnx>Ah that's why I didn't see it yet. I noticed no loginctl <jpoiret>you don't "need" elogind but it's useful to manage seats, permissions for different devices on your system when you're logged in, and some apps rely on it <jpoiret>for example i'm not sure dbus/polkit (maybe just polkit) would work without elogind <jpoiret>note that elogind was a part of systemd, but some people forked only that part to avoid having to rely on the whole systemd stack (and by some people I mean initially Guix and now Gentoo) <jpoiret>apteryx: looks like qt 6 is the biggest culprit <apteryx>or do you mean just the size it takes in the /gnu/store? <sneek>Welcome back gnucode, you have 1 message! <sneek>gnucode, dthompson says: wow, neat! glad haunt works on openbsd. thanks for the report! <gnucode>I wonder if guix the package manager would work on chimera linux. <gnucode>chimera uses the linux kernel with a FreeBSD userland. <mvnx>jpoiret: Thanks. But why would I want both agetty and mingetty then? <jpoiret>so idea but I guess mingetty is very minimal and doesn't support all possible configurations, so agetty is there to autostart for serial lines? <jpoiret>I'm too tired and am forgetting to type half of my sentences: "so i have no definitive answer but my idea is that ..." <gnucode>I would check my own guix install...actually I can. my servers still run guix. <gnucode>sweet qtbase is packaged at version 5.15.5. Pretty close! <apteryx>it's at 5.15.8 on staging and core-updates <mirai>do home services run only if the user is logged in? <mirai>or is it like the "persist" option with loginctl <mirai>where the user daemons will always run regardless of login <mirai>but do they launch without interaction as well? <jpoiret>oh hell no qtbase 6 just failed during tests <mvnx>What's the difference between `login-service-type` and `elogind-service-type` other than of course using elogind instead of login(1). Or maybe I'm just asking the difference between them... elogind just adds the concept of sessions/seats that some DMs need and some handy commands/policies to allow system shutdown/reboot? <mvnx>Hmm I guess that was already covered in an answer above. <mvnx>I'll just tinker and fail and see for myself - probably easiest here. Ignore thanks <gnucode>mvnx: I had a guix config that uses seatd if you are interested. <f1refly>when exploring fontawesome, I noticed that a bunch of icons of the "regular" sort aren't available in the guix fontawesome package - is there a way to see what icons are available in the guix package when browsing the website? <f1refly>oh, I think it's just the version, right? we have 4.7 in the repo and the website defaults to 6.something <mvnx>gnucode: Sure, I'll have a look. I think I need to take time to break down the full process of logging in and where all these different programs come in. Particularly greetd which claims to be a replacement for both agetty and login so I guess some cross multiple layers. I've never done a linux from scratch or anything - not even sure if it would be covered in that <mvnx>Why did you choose seatd over elogind? Just curious <gnucode>and elogind is brutally pulled out of systemd. It is hard to update. <gnucode>mvnx: please bear in mind that you cannot use network-manager with seatd <mvnx>All good to know thanks. Nice to see another more advanced config - some good stuff in here <gnucode>mvnx: A good portion of that config is borrowed. :) <mvnx>So what do you use in place of network-manager? <mvnx>And what is the underlying limitation? <mvnx>I mean what blocks it from being used? Is that something that could be fixed? <gnucode>mvnx: well I actually couldn't get the other wifi managers to work. So I just started using dhcp...and now I switched to OpenBSD... <gnucode>and I wish OpenBSD had a config.scm... <mvnx>I understand. Yeah, it's (config.scm, configuration.nix) a game changer. <jpoiret>mvnx: some people here use wpa_supplicant directly <gnucode>mvnx: You could set up a iwd service... <apteryx>it may get rid of a good class of network-related test suite failures <f1refly>but the current version in guix is 4.7, what's up with that? <ngz>f1refly: I think there is licensing issue. <f1refly>it was reverted in a later ticket i think <apteryx>interesting; 'M-x woman getaddrinfo' crashed my emacs <Michal_Atlas[m]>Good evening, how would one go about allowing derivations in a config? I'm writing a dconf service and would like to allow url-fetch or local-file, however I'm really not sure how to approach it, and reading through other guix files and the manual hasn't helped much. I'm not sure if it's supposed to magically get handled by #$ or if I directly play with monads like the home service does? <mirai>Michal_Atlas[m]: you'd be surprised to hear that there's already a dconf service <mirai>it's just: a) not documented, b) would benefit from an audit/cleanup <podiki[m]>should we be defaulting to pyproject-build-system over python-build-system? <podiki[m]>and is the error for tests failing "error: in phase 'check': uncaught exception: <podiki[m]>%exception #<&test-system-not-found> " familiar for pyproject-build-system? <podiki[m]>ah, from reading pyproject-build-system either there is a missing input to run tests or package may not have them <pharcosyle>Does anyone know how to create a release tarball of gcc? I'm experimenting with changing the default gcc to 13 but since it isn't released there's no available tarball I have to make it myself. There's some instructions online but they're all different, I want to know what the "official" make commands / flags / whatever are to go from a source <pharcosyle>checkout of gcc to the release version which the Guix bootstrap expects. <quidnunc>I'm trying to install emacs-next-pgtk. I see here that it appears to have a substitute built for the latest version. When I do guix install it says it needs to download 200MB+ of packages so I assume it is trying to build from source. Why isn't it using the substitute? <quidnunc>Or how do I know whether it is using the substitute? <pharcosyle>It looks like a substitute isn't available for emacs-next-pgtk and none is queued. I guess the build farm isn't building one, perhaps it's a transient problem but in any case you're stuck building from source <podiki[m]>lechner: I see an update to x-net (for go), is there a reason for 0.5.0 instead of the latest (0.8.0 I think)? I'm not up on go stuff at all (i have a local package that depends on it) <podiki[m]>lechner: as in I see you made the patch, thought you may know the go versioning :) <podiki[m]>lechner: ah, it was from an older series and I guess was the version from there originally. <apteryx>was --expose=/etc in 'guix shell -C' now supposed to work? <apteryx>I get: guix shell: error: mkstemp: Read-only file system <the_tubular>Is there a way to use guix to download text files ? And drop them somewhere specific on the FS ? <devmsv>Hi! how can I copy a file to etc with different owner and permissions than the defaults? I have try [..]etc-service[...](local-file "file" #:recursive? #t) with file having different owner and 0600 perm but still gets installed in the store as 444 <devmsv>I have try to use comptued-file with I can wrap my head around gexp currently <devmsv>I was thinking of using a activation-service but don't know if there is something simpler <devmsv>ryulash: thanks. I will see if I can do something similar <bumble[m]>I did not guix install bash-completion but when I press <tab> my bash utterances are completed in the shell for me. How is this possible? <jpoiret>devmsv: in the store it will always end up as 444, there's no way around this <devmsv>civodul: The first time I booted guix that error message was weird. After that first time it was fun to see it. <Rampoina>are there any dotnet applications packaged in guix? <next4th>Rampoina: none, even mono not packaged. <Rampoina>is there any reason (like licenses or so on) or just nobody willing to go through dotnet packaging hell? <civodul>devmsv: so hopefully you'll enjoy the new service <jpoiret>ngraves: there's probably a module import cycle somewher <ngraves>jpoiret: How would you fix this? Is there a resource / manual page explaining how to tackle this? <jpoiret>but as civodul said one important thing to do is to always inherit packages in the same module <jlicht>what are the hardware specs of our armhf build nodes? <cbaines>at least for the bordeaux build farm, but armhf builds take place on the aarch64 hardware <cbaines>so either a Honeycomb LX2 or an Overdrive 1000 <jlicht>is there some kind of RAM usage limitation in the configuration or setting the armhf builds run in? <cbaines>nope, other than the limitations of running 32bit stuff <uniavix>Can someone help me installing sof-firmware? <uniavix>I added the nonguix channel but then I don't know how to get it from there. <oriansj>uniavix: we don't discuss that here, you may wish to try the IRC channel dedicated to that <uniavix>oriansj: how can I install everything from source without substitutes? <uniavix>oriansj: found the #nonguix channel. Thanks! <n9>Hello, everyone. I am not sure if it's right channel to ask, but I have a problem with ssl certs for irc: (default) warning Could not verify TLS servers certificate: unable to get local issuer certificate <n9>I am having that problem only on Guix "bare bones". <devmsv>n9: do you have "nss-certs" package installed on config.scm? <devmsv>well I have taken a look at how Guix populates files like sudoers which has restricted reading permissions and I still can wrap my head around it <n9>devmsv: wait few mins please, i have emergency <rekahsoft>Hi folks! I have been using 'guix deploy' for all of my local vms for some time now (about 14 vm's total). I've noticed that frequently I have issues where the deploy fails part way through due to "[GSSH ERROR] Parent session is not connected: #<unknown channel (freed) ...>". A couple of retries will result in a successful deploy, however this is blocking me from automating it (as I would prefer if 'guix deploy' ex <rekahsoft>sting with an error would mean the deploy actually failed). Any ideas on what could be causing this lack of reliability when using 'guix deploy'? <rekahsoft>(sorry if I double posted the above (had an issue on my end and wasn't sure if it was sent or not)) <n9>devmsv: No, I don't have it on the config.scm, but I installed it using 'guix install nss-certs', it's incorrect? <devmsv>n9: yes, most applications expect certificates on /etc/ssl if you install nss-certs on your profile it is actually inside your profile folder <n9>Can you please tell me how to add it to my config or send link where I can read? <n9>Just switched to Guix today. <devmsv>here you have examples of config.scm <devmsv>no porblem, we are here to help and be helped <n9>Tried it yesterday on my laptop, installed to second computer, and now installing to my main computer! <n9>Also want to test it on my server. <n9>devmsv: Are you using guix system? <devmsv>yes. on my laptop and pending reinstall to my personal server <n9>devmsv: And how do you like it? <ekaitz>hi guix! since a recent update Dino does not render in dark mode but all the rest of the gnome apps I have do, any idea why this happens? is it a problem with dino specifically? <n9>ekaitz: Jabber client? <n9>Solution is simple, use profanity! <n9>I had so much troubles with Dino <n9>ekaitz: It worked before update? <ekaitz>n9: sure! it was working alright! <n9>Have you checked logs? <ekaitz>(dino:2944): Adwaita-WARNING **: 17:11:10.548: Using GtkSettings:gtk-application-prefer-dark-theme with libadwaita is unsupported. Please use AdwStyleManager:color-scheme instead. <n9>Well I guess it tells you what to do. Can you use that instead? <ekaitz>and also, why does evince show in dark mode too, even if it does show the same error? <n9>I think it can help, but I can be wrong. <n9>ekaitz: Oh, interesting. <n9>ekaitz: Maybe a bug of new dino? <ekaitz>n9: i don't know, maybe it's something with gtk4? <ekaitz>also there are some (minor) rendering issues under i3 <itd>ekaitz: How did you configure the dark mode? <ekaitz>in the gnome configuration screen <n9>ekaitz: did you set dark mode in Dino settings? If it's there. <n9>Just install gajim or profanity, and enjoy! All I can suggest, sorry. <itd>ekaitz: I'm (not using gnome and) exporting GTK_THEME=Adwaita:dark; that seems to work here. <ekaitz>itd: that works, but there must be a way to keep that set in gnome config <itd>ekaitz: Glad to hear it. Can't help you with gnome, sorry. (Could you please describe your rendering issues?) <ekaitz>itd: looks like what I had to do was to set .config/gtk-4.0/settings.ini and set Adwaita-dark there <ekaitz>the prefers-color-scheme doesn't work but the other key does <itd>Nice. Thanks for sharing. :) <mirai>what keywords do the lambda procedures accept for package phases <PotentialUser-76>Does anyone know a workaround for restoring the keyboard/mouse after a suspend in GNOME. <mirai>Can the service-type name appear more than once in a system declaration? <civodul>mirai: hello! the 'name' field is purely informational, it doesn't have any semantic role <mirai>civodul: What's the actual behavior of #:pid-file in make-forkexec-constructor ? <mirai>The shepherd manual says: When pid-file is true, it must be the name of a PID file associated with the process being launched; the return value is the PID once that file has been created. <mirai>so... #:pid-file "/run/daemon.pid" is bogus then? <apteryx>haha, my rust-cargo-lock-8 problem was from a stray rust-xyz.scm file not checked in <mirai>the constructors should be returning a string to a path? <mirai>that's what I'm understanding from the description <mirai>it would explain many of the strange issues with shepherd losing track of daemons <apteryx>mirai: some services can produce a pid file, which helps the service manager (shepherd here) to know when the service is ready <mirai>apteryx: yes, but the manual is saying that #:pid-file is actually boolean <mirai>and the path to the pid-file is to be returned by the constructor <mirai>none of that is what happens within the existing forking services <mirai>everyone puts the path in #:pid-file and leaves the constructor to return whatever <apteryx>no, the path to the pid file is not expected to be returned by the constructuro <apteryx>it just returns the PID value based on the docstring you copied above <mirai>rereading that segment is confusing me even more <civodul>in short, #:pid-file indicates the PID file produced by the daemon <mirai>I've no idea what that _it_ refers to (a string value is considered "truthy", yes, but that's an extremely odd way to word it) <mirai>right, so everything is working as intended then <n9>Anyone knows how can I enable "Natural Scrolling"? <n9>Ah! Done by: xinput set-prop "Wacom Intuos PT M Finger" "libinput Natural Scrolling Enabled" 1 <mirai>if a shepherd service B depends on service A, will B get terminated if service A crashes or if A gets restarted/stopped? <bjc>if it crashes or is stopped manually: yes. i'm not sure about restart, but i suspect B will terminate and not start up again after A starts <PotentialBear-86>Hi, so Im trying to get jupyterlab working. I didnt see a package for it in guix, so i installed the desktop application with flatpak. My question is then, where should I point the desktop app to find python? <PotentialBear-86>Running 'which python' gives me a wild path into the store, that I assume is going to change when I update <PotentialBear-86>Ah wait that channel is still free software. To anyone looking in the logs in the future: jupyter-lab can be found in the guix-science channel <Guest19>Out of a sudden if I install packages, the output is to large to fit in my vterm window. I use EXWM. Therefore I can't see the whole line. Some of the progress is cut. Does someone else have that issue, too? <kmarker>I am new to guix and need to add i8042.dumbkbd=1 to my grub but not sure how. I think it goes in config.scm? <RavenJoad>I am trying to build a gexp in a Geiser REPL to verify my serializer's output. I have tried a (with-store s (run-with-store s (mixed-text-file "name" #~("thing")))), but it does not build. It just yields a #<<computed-file> ...> <RavenJoad>What procedure takes a computed-file and builds it? <mirai>,build <THE COMPUTED FILE HERE> <NewUser-Basic-Qu>Hi...If i build something in Guix shell via pre-inst-env (sway/wayland/mesa) do i have to make extra steps to install in my system? <rekado>NewUser-Basic-Qu: anything you’ve built can be installed with “guix install /gnu/store/…” but for desktop things you’ll probably want to use package transformations to inject your variant into an existing package dependency graph <NewUser-Basic-Qu>rekado oki....let me quickly google some of the terms you dropped. :) ---package transformation and injections <mekeor[m]>not sure why package transformations are needed here. what "variants" are we talking about? <Guest19>I can only see my printer through the system dialog in Firefox. How can I fix that to print from the dropdown menu. Shows only "Save as PDF". <podiki[m]>slime working for anyone currently? I get Cannot open load file: No such file or directory, ../lib/macrostep <podiki[m]>can reproduce with guix shell --pure emacs emacs-slime -- emacs -Q as well <Guest19>podiki[m] Running the command works for me without any errors <podiki[m]>running the command? in emacs have to evaluate (require 'slime) and then M-x slime if that's what you did then boo for me <podiki[m]>(no error in just starting emacs, is specifically when trying to run slime) <Guest19>Ah you didn't mentioned that. Trying <podiki[m]>perhaps time for more time-machine action to see when this started <podiki[m]>ah the issue is slime needs macrostep (emacs-macrostep) <podiki[m]>patch submitted, suppose I could just push it? anyway, i'll wait for QA at least, not sure the current expectations for a quick fix like this <apteryx>ACTION will merge staging into master in a bout <apteryx>is there a way to filter for newly failed builds in cuirass? <apteryx>podiki[m]: if there are not much rebuilds and you've tested that they don't break, and that the fix works, I'd say it's OK to push <podiki[m]>I think it's just a handful of packages, should be quick enough to rebuild locally <podiki[m]>Happen to have a handy one liner to take the refresh -l output into guix build? :) <apteryx>it's not perfect, but often saves me a bunch of time <apteryx>be sure to not get tricked by inherited packagaes, which wouldn't show as impacted <podiki[m]>apteryx: how do you like to handle patches (say short series or one off) locally? do you use debbugs through emacs, git worktrees, directly from the QA branches? and then pushing (I assume cherry pick from the local branch)? <podiki[m]>(I actually wanted to get some community knowledge on a guix-devel discussion, pick up some workflow tricks and tips) <apteryx>podiki[m]: currently I apply stuff in Emacs Debbugs <apteryx>I have two git worktrees tracking master, others for staging and core-updates <podiki[m]>so you apply patches from debbugs directly to your master worktree? <podiki[m]>oh you said 2, so master master and temporary master basically? <apteryx>rigth, I often have 1 thing cooking, so I use another worktree for review <jgart[m]>Are there any plans to make pipewire the default for gnome? <rekado>jgart[m]: better ask the Guix gnome team about this. <pranavats>Hello. I'm trying to get hibernation to work on my guix system. What should be the value of the "resume" kernel parameter if my swapfile resides in a Logical Volume in an LVM on Luks setup? <pranavats>I tried setting it to the path to the logical volume as in "resume=/dev/mapper/vgroup-vname" with an appropriate resume_offset value. Hibernation did work once, but I got some file system errors on next reboot and now I get a kernel panic while booting. <pranavats>Curiously, the only files that have been overwritten with garbage values are from my emacs configuration. <jlicht>pranavats: unless you have backups, I wouldn't put a lot of trust in the state of your filesystem <jlicht>or rather, the contents of the files on said FS <pranavats>jlicht: I see. I do have older backups of the important files. <pranavats>Earlier I had set the resume device wrong. And had tried to hibernate, which resumed unsuccessfully, but I think the state was stored in the swap. When I did get the resume device (logical volume) right, the system resumed to the previously stored state, and on reboot, my file system had errors. <jlicht>pranavats: sorry to hear that. I'm no expert either, just a jaded survivor from my initial attempt to do something similar :/ <jpoiret>hibernation is hard to get right, especially in situations like this! <pranavats>I also have lvm-on-luks2 with grub2 using your patch jpoiret. Still works. <jpoiret>they still haven't got merged iirc but they should soon be <jpoiret>is your swap directly on lvm or in luks2? <pranavats>jpoiret: I have a swapfile in a logical volume. <pranavats>I do think my setup is correct now. It seems that the older misconfigurations caused the data loss. <pranavats>jpoiret: Currently when I try to boot, it gets stuck at "populating /etc from /gnu/store/<hash>-etc" <pranavats>This is after I ran fsck on the concerned file systems, and "fixed" errors. <jpoiret>well, at least it goes past the early userspace. But I guess your fs is bricked now <jpoiret>maybe you could repair the store via a chroot? <pranavats>Do you mean "guix system reconfigure" from a chroot? Or "guix system init"? <jlicht>hmm, how can I get (guix http-client)'s http-fetch to "deal with" UTF-8 data? Specifically, a web page with a 🐈 in there (emoji of nice kitty) <jlicht> I get "hello kitty ð\x9f\x90\x88\n" instead of "hello kitty 🐈\n" in my responses <n9>Hi hackers, I probably have stupid question: Anyone knows how to install growisofs? Seems like it's not in repos, and I cannot find other ways. Compile from source? <n9>I needed "dvd+rw-tools" <mirai>does it make sense to define some manifests, concatenate them and <perform ???> to use them in the packages field in home-environment ? <unmatched-paren>so it would make sense to define a list of packages and then lower those to a monifest <unmatched-paren>oh, never mind, it looks like you can retrieve the MANIFEST-ENTRY-ITEM <unmatched-paren>(map manifest-entry-item (manifest-entries MANIFEST)) might work, mirai <mirai>I'm getting a list of #<package …> records <mirai>so I presume it should work? <civodul>PotentialUser-47: hi! Guix itself is committed to user freedom and does not provide non-free software <n9>PotentialUser-47: It doesn't works with anything non-free. <civodul>other projects out there do provide non-free packages though, in the form of channels <n9>I would suggest to avoid hardware that requires non-free software. <mirai>guix doesn't like service canonical name changes upon guix reconfigure <n9>$ sudo herd restart myself <n9>PotentialUser-47: Technically there are, but no. <mirai>so… a physical restart is required? Sounds suboptimal that if shepherd encounters some problem in the shutdown sequence actually results in shepherd refusing to attempt it again or “voiding” the reboot/shutdown commands <n9>We will *not* help you with non-free software. <n9>I am not angry, just saying. <n9>I've never seen/used Nix. <PotentialUser-47>It's guix if the configuration language sucked and the documentation sucked <n9>I didn't liked Guix documentation, so I've written down things I need, in way I will understand. <PotentialUser-47>Which Is unfortunate because i think if nix had these 2 problema corrected <n9>Just don't use and calm down <mirai>system reconfigure and reboot <civodul>right, but the canonical name change? <civodul>IOW, what was the difference before and after reconfigure? <mirai>I think it's the NetworkManager change? <mirai>let me see if I can pull the commits <civodul>there's a message about NetworkManager <mirai>can I get the guix commit associated with a system generation ? <civodul>i did reconfigure since the NetworkManager change, but i've been using Shepherd "master" :-P <mirai>from: f01b5299db6031174f05124b843c936388cd872a --> to: 380faf265b0c3b231ab8b69597d161be5e704e18 <civodul>so the system that changed canonical names is NetworkManager, right? <civodul>if that's the source of the problem, i'm surprised nobody complained before <civodul>it's been a couple of weeks since that change, no? <mirai>well, I'm only complaining now, 31d since last boot 😂 <mirai>it's the only thing I can think of that could be the cause of it <civodul>could you send the relevant /var/log/messages excerpt (showing the reconfigure phase followed by the reboot attempt), along with the commits, to bug-guix? <mirai>well, that assert message seems to be where things fall apart? <mirai>Assertion (eq? (canonical-name new) (canonical-name old)) failed. <civodul>which doesn't mean it wouldn't crash ;-) <mirai>it never expects that a service might gain a canonical name at some point in the future <civodul>i wonder where the wrong-type-arg/for-each thing comes from (Tor?) <mirai>the plague of the sporadic connection losses just disconnected my SSH session <mirai>oh ok cool, I managed to log into a tty <mirai>civodul: what other info do you want me to pull from this machine before I powercycle it <mirai>occasionally I get “NIC Link is Down” messages in dmesg <mirai>I... don't know, it's very erratic <mirai>one machine loses less often <mirai>but I can hold a ssh connection to non-guix machines for much longer <civodul>well, i do experience it with berlin.guix.gnu.org for instance, but that's not because of Guix, rather due to some tricky networking issues on the way <mirai>civodul: speaking of that, that reminds me of an issue that might solve the problem with the absurd nginx log sizes <mirai>that occur with *.guix.gnu.org <mirai>it should reduce the multi-TB log files into something more manageable <civodul>we'll also need log rotation for nginx <mirai>log-rotation is a bit subtle <mirai>it will require a 'reopen action to shepherd <mirai>'reopen action for nginx I mean <mirai>but the ownership of the logs is weird <mirai>the first non-rotated logs is root:root <spiderbit>Hi, I try to get stumpwm in guix ready to import normal cl libraries :D <spiderbit>basically defined my own stumpwm version with <spiderbit>i this inputs (modify inputs) only for build inputs <spiderbit>oder does it somehow puts this package inside the other package as output or so <spiderbit>and if that is the wrong way to be able to load stuff from inside stumpwm is there another way <spiderbit>here is some example to add stuff to stumpwm, but I can't really understand that fully and or adapt it to my problem <PotentialUser-76>Any idea if there is an option to play youtube on GUIX? I am using GNOME-web browser. <spiderbit>it should probably work but just to give a easy fast solution <Aurora_v_kosmose>I'm thinking of a view of services similar to how systemd or Erlang supervision trees work. <RavenJoad>Does it make sense for the serialize-configuration procedure to also accept a delimiter, so it behaves like interpose? Or should each field's serialization function provide the delimiter? <mirai>RavenJoad: there's a thread in guix-devel <mirai>methinks it should be a transducer <RavenJoad>Ok. At least I am not alone in that thought process. <RavenJoad>mirai: I have been getting those on several different Emacs instances for a while. I think Emacs' warning checker got stricter in a recent release. <spiderbit>I wonder if this information is still correct <spiderbit>I tried that and guix system reconfigure does not refuse this <spiderbit>but stumpwm doesn't allow me to require ttf-fonts this way <spiderbit>also there exist a sbcl-stumpwm-ttf-fonts packages <RavenJoad>Is there a way to perform a map over/into gexps? I have gexp<string>, where the strings do not have newlines. I want to map (and maybe flatten) by adding a \n to the end of each gexp's string, forming a new gexp. <spiderbit>Well I restate my comment about the cookbook entry <spiderbit>it's partially false/outdated, that is a fact there is no sbcl-ttf-fonts package anymore so it must be wrong <mirai>has git undergone some update recently? <mirai>gitk launches but git gui doesn't <spiderbit>I just thought I figured my problem partially out <spiderbit>sbcl-stumpwm-ttf-fonts asks for sbcl-clx-truetype and sbcl-zpb-ttf <spiderbit>when I add them in stumpwm config file with (stumpwm:add-to-load-path "/run/current-system/profile/share/common-lisp/sbcl/zpb-ttf") <spiderbit>one by one which is stupid but the errors go away :D <spiderbit>but it doesn't install in this sbcl parent folder <spiderbit>ok that seem to work to require str into stumpwm <spiderbit>but why can't guix not setup this load paths automatically <Guest701> does anyone have a config, where firefox is installed with? <Guest701>I looked at some configs, but i could not find, how to install firefox <ngz>Guest701: Guix provides Icecat, which is morally equivalent to Firefox. <rmnull>Q2. some of my PATH come from user installed packages. e.g: ruby -e 'puts Gem.user_dir' gives the gem directory and then that is appended to the PATH. What's the guix way of doing this ? <RavenJoad>rmnull: To link a configuration file into place that does not currently have a guix-home service-type defined, you use a simple-service. (simple-service 'your-service-name home-xdg-configuration-files-service-type ...). See (guix) Essential Home Services in the manual. <mirai>someone™ should adapt mpd-service-type into a home service <mirai>it probably only needs the accountservice extension stripped <RavenJoad>rmnull: The PATH thing is slightly more complicated. But same concept, look at that section in the manual. It has a good example, but in summary (simple-service 'service-name home-environment-variables-service-type <alist of env-vars>). I recommend if you toy with PATH "manually", you use "<addition>${PATH:+:}$PATH" syntax. <rmnull>RavenJoad, thank you for the suggestion, I've looked into the manual it was kinda confusing I'll go through it again after going through the source whereiseveryone provided. <whereiseveryone>rmnull My config linked about shows what RavenJoad is talking about in a real config that I'm currently using <rmnull>whereiseveryone, thank you for the source, It looks something I can copy and build upon. I don't use rde, the other example looks just like what I want, thanks!! <rmnull>I'll get back here if I get stuck. thanks <whereiseveryone>I'd avoid rde for now until you feel comfortable with the core home services <RavenJoad>I found the examples in the manual more useful than the dialogue that goes along with it for guix-home. <whereiseveryone>And you might not even need rde depending on what you're trying to do <whereiseveryone>See the guix channel search engine results I linked for an example of rde features <msavoritias>so the difference between home service and this is that rde is for applications and home is for home environments? <whereiseveryone>feature-emacs-evil, feature-emacs-denote, feature-emacs-org-roam, etc... <msavoritias>the reason i am asking is because i was looking into learning home <whereiseveryone>For example, (feature-emacs-evil), will enable an opinionated config for emacs-evil stuff in your emacs <whereiseveryone>(feature-emacs-evil) is just a function call and there are function keywords that you can modify to change the feature's (that is, function's) functionality <PotentialUser-76>Does anyone know why the download button in gmail does not auto download in GNOME-GUIX <PotentialUser-76>i am guessing the backend javascript is missing but not quite sure if this is the right answer. As there may be an opensource library that is missing? <Guest701>ngz I triied icecat and run into many problems (youtube not working, dropbox not logging in). I just need a solution, that works <Guest701>whereiseveryone can you share your config please? <whereiseveryone>There might be random useful stuff in the other branches also that are not gnome. That is, b, sway, and root <Guest701>whereiseveryone where do you define your screen resultion? <Guest701>whereiseveryone I want to change my resolution in i3, but i just think i have to use xrandr with addmode and try it <Guest701>whereiseveryone will check it first myself, thank you so much <Guest701>whereiseveryone sounds good for now, i am just building my config in qemu to switch to guix hopefully soon. <projectmoon`>hi, i am experimenting with guix as a build tool and trying to get it to install something from a local file i have lying on my filesystem. but it seems guix runs as another user, and thus cannot see the .tar.gz. what is the proper way to handle this? <projectmoon`>using guile-hall, which i cannot really understand the guix.scm files it generates... can't i just use guix itself to build the entire guile package? <lucas_>hi, is there a way, typically based on logs, to have a look at how many users in my local installation are using guix? (I could not find any logs for what the daemon receives, but maybe I did not look at the right place) <spiderbit>RavenJoad any idea why I have to setup all paths of all dependencies in sbcl / stumpwm to load any module? like ttf-fonts as example not like in the example from the cookbook where I have to set none like here: <projectmoon`>whereiseveryone: i think that's exactly what i am trying to do. i have a scheme project, and i want to build and 'install' it locally using guix. <projectmoon`>not sure if that's what i'm SUPPOSED to do, but that's what i'm trying to do <spiderbit>to make that work I have to add many paths add-to-load-path statements like that: <spiderbit>(stumpwm:add-to-load-path "/run/current-system/profile/share/common-lisp/sbcl/zpb-ttf") <spiderbit>which would not be that bad because at least it stays in the same path <spiderbit>(stumpwm:add-to-load-path "/gnu/store/sixajfiirrw79zijnlc2n75mvqfa4wg6-sbcl-cl-ppcre-unicode-2.1.1/share/common-lisp/sbcl/cl-ppcre-unicode/") <projectmoon`>whereiseveryone: well that definitely gets it to run. but now it cannot see ./configure, because i assume guix is running as a separate user that doesn't have access to my source directory <projectmoon`>whereiseveryone: i am not sure if what i'm trying to do is the actual best way to go about iterative scheme development. but dunno how else to test the packaging locally o_O <RavenJoad>spiderbit: If you are manually adding to the stumpwm load-path, that makes me think StumpWM does not have those packages available to it during compilation (if you are compiling it)? <projectmoon`>i dunno. preferably a command i can enter that pops out an installable thing <whereiseveryone>the guix developer workflow experience is WIP and people are dogfooding it in different ways <projectmoon`>it is not entirely clear to me what the workflow with hall is supposed to be <spiderbit>raven, well I made another package but it did not help <RavenJoad>spiderbit: I use the stumpwm-with-slynk package, which adds slynk to the StumpWM you run. You might want to look at that. <projectmoon`>doesn't have to be hall. just needs to get me from scheme code to installable thing <msavoritias>if you just start with scheme I think hall is overkill right now <projectmoon`>whereiseveryone: yes i saw it mentioned in the documentation <projectmoon`>currently pkg has just a few guile dependencies, all of which are available inguix <spiderbit>I am just a noob in scheme / common-lisp and guix, I am just good in emacs-lisp :D author of a few packages <whereiseveryone>guile-build-system just compiles the guile file to a guile object (.go) and copies the sources to output <whereiseveryone>You can write guile files and then use guile-build-system in your project repo <spiderbit>or do I have to logout is "restard-hard" not enough <RavenJoad>I am not deeply familiar with the asdf-build-system/sbcl, but to make StumpWM see slynk stuff, the stumpwm-with-slynk package needs to change some program dependency things. I don't know if that is required here. <whereiseveryone>See those two packages and the video from today if you'd like to have a similar experience to quicklisp without needed to restart a guix shell <spiderbit>I stumbled over that, but looks a bit a overkill especially because the packages are all in guix <whereiseveryone>in quicklisp you can load new dependencies in a repl without restarting the repl <RavenJoad>spiderbit: It seems like the StumpWM lisp image that is being built does not get the store-path of the run-time dependencies embedded into the image, which is why you have to manually add them yourself right now. You might need to explore how other CL apps get load-paths embedded. <spiderbit>whereiseveryone I copied 1-2 times some quicklisp line <spiderbit>but looks a bit ugly when it's not distribution packages :D <spiderbit>maybe it doesn't do it because it also don't makes a link towards: /run/current-system/profile/share/common-lisp/sbcl/ <RavenJoad>If you are building a package, it should not rely on /run/current-system (unless there is good reason to). You should expand the path into the store. It helps prevent issues when reconfiguring. <spiderbit>but the package itself should make such link <spiderbit>RavenJoad the video buffers horribly like 20 sek loading then 2 sec playback or so :D and I can't lower resolution <RavenJoad>Whether or not the package does that may require some massaging of inputs. The stumpwm-with-slynk package needs to set dependencies and dependency-prefixes manually, so perhaps investigate stuff like that? <RavenJoad>As for the video thing, I have no idea what that is about. I do not know what you are doing there. <projectmoon`>whereiseveryone: do you know if there's a way to get emacs to auto-complete guix's guile functions? like is there a guix source pkg i can add to guix.scm to get auto-comlpetion with company-mode? <msavoritias>I was looking in the docs but it doesnt mention anything except paredit and geiser <RavenJoad>projectmoon`: I just need Emacs and a running Geiser instance, albeit the Geiser instance needs the Guix Guile API to be in the load-path though. <projectmoon`>i have geise running inside emacs, but maybe i am missing the load path <RavenJoad>projectmoon`: That is more-or-less what I use. I have not done a ton of configuration to get Guix auto-completions working, beyond loading the modules in an attached Geiser instance. <projectmoon`>ok. now i just need to see if the guix ebuild for gentoo installed the source files somewhere <projectmoon`>i changed geiser-load-path to point at a git checkout of guix, but no dice <RavenJoad>To be fair, my auto-complete is not great either. I do not have a good Guile setup yet. Most of the time, I still need the manual open for searching. <RavenJoad>In Geiser, if I need something I use either (use-modules ...) to bring things into scope or ,m to switch to that module. <rmnull>what's the recommended way to inject user-bin directory in guix home? I have collection of bash scripts that all reside in a local ~/bin folder I want it to be part of guix home shell invocation. what service would i need to use without painfully mentioning a file one by one <RavenJoad>If you want script-name.sh to be available as a "regular program" on the command-line, make sure the ~/bin directory is at the front of the PATH environment variable. Just another simple-service to prepend to PATH. <rmnull>RavenJoad, so I just copy the whole bin directory as it is and then prepend "${HOME}/bin" to PATH? <rmnull>jgart[m], gexps look cool, I'll take a look at them in future, currently I'm getting the existing setup to build. Without having to write additional code :sweat smile: <RavenJoad>In the long-run, rewriting them to be Guile scripts might be the most Guix-y thing to do. (That's what I am doing). But in the short-run, yes. Just copy ~/bin and prepend to PATH with a simple-service that extends home-environment-variables-service-type with ("PATH" . "$HOME/bin${PATH:+:}$PATH"). <rmnull>jgart[m], haha, you are whereiseveryone. I've been copying your config, it's been helpful, thanks! In your config you seem to be mentioning each script individually, that was kinda painful, so I wanted to know if there was a better way to do this, that's all <apteryx>efraim: can rust be built for powerpc64el nowadays? <apteryx>I see the latest python-cryptography requires rust... <apteryx>I'm updating the package to 40.0.1 on staging, borrowing a few of your rust changes in the process (will appear on master) <jgart[m]>rmnull I copied unwox, unmatched-paren, and various others <unwox>irfus: hi! do you mind if i publish your home-pipewire service work on my guix channel? you posted a pastebin earlier here that i copied for myself. i'll add a copyright notice with you of course <Guest7024>I added in my /home/dave/.config/guix/channels.scm the nonguix channel, but after guix pull nothing happens. Can anyone help me? I just want to install firefox, so i can upload something to dropbox. <zamfofex>Guest7024: I don’t think we’re meant to be allowed to talk about nonguix here! Though I will say: I never had any issues with websites running poorly on IceCat. What issues do you run into when tryint to use DropBox? <zamfofex>Guest7024: Conversely, you could also try using ‘ungoogled-chromium’. <Guest7024>zamfofex so youtube is workin for you? And you can login into dropbox with icecat? <zamfofex>Yes! (At least for YouTube, but I don’t use Dropbox.) Usually I disable all of the plugins that come preinstalled in IceCat, though. <zamfofex>From the “hamburger menu” in IceCat, you can choose “Add‐ons and themes”, then disable the extensions. <Guest7024>zamfofex nice, now everything works, i did not know all this plugins were blocking these sites <Guest7024>zamfofex i added to ~/.config/guix/channels.scm the nonguix <Guest7024>zamfofex but, when i do guix describe, i just see gnu <Guest7024>cool, now it work, thanks for everything zamfofex ! <zamfofex>I’m glad you managed to get everything working! <lucas_>hi, is there a way, typically based on logs, to have a look at how many users in my local installation are using guix? (I could not find any logs for what the daemon receives, but maybe I did not look at the right place) <lucas_>I see that the daemon outputs something for each connection it receives, but only with the source IP <pranavats>jpoiret: About your advice regarding chroot and repairing the store with guix gc --verify=repair,contents. Every time I try that, the same errors show up, which means probably nothing is being fixed. Some store objects are corrupted and missing. I am able to boot into older kernel configurations though. Would you recommend doing a "guix system reconfigure"? Could that update/replace those corrupt store obj <pranavats>For context, my file systems got corrupted while trying to get hibernation/resume to work. Some files got overwritten with garbage values. And I'm trying to fix the store items that got corrupted. <jpoiret>guix system reconfigure will not repair the store objects <jpoiret>guix thinks they are there and they are fine so it will not inspect them unless instructed to <TristanCottam[m]>Hi guys, can anyone help me understand why software which is meant to be run as a service sometimes only has a Guix package, but no service? <evilsetg[m]>My first idea is that nobody wrote a service for it yet. Do you have an example? <TristanCottam[m]>For instance, Synapse (Matrix.org's reference server) has a Guix package, but no service. Why would one bother creating only a package for this? <pranavats>jpoiret: would "guix system init" help? Or do I just roll-back to some previous generation, collect all the garbage and then do a guix pull and reconfigure? <evilsetg[m]>Packaging it is the first step. Without the package a service would not make sense. It would be nice though to have a service for synapse. I don't think it's on purpose there is none. <TristanCottam[m]>ardon: That's the impression I had, but what's the point of the `synapse` package by itself? <evilsetg[m]>You could probably set up synapse manually after installing the package. <ardon>TristanCottam[m]: Basically what evilsetg[m] said, without a package there's no service in the first place. <TristanCottam[m]>That's fair enough, are there any other resources than the reference manual to learn about creating Guix services? I'd love to be able to contribute <evilsetg[m]>A hands on tutorial for creating a service would be good for the cookbook. <uniavix>I could translate it to Portuguese of Portugal. <mirai>TristanCottam[m]: look at existing service examples? (preferably those that have undergone changes recently) <TristanCottam[m]>mirai: You mean plain existing services? Or are there actual existing examples? <mirai>the existing ones under gnu/services <mirai>what service do you have in mind? <TristanCottam[m]>In this case, Synapse, but ardon linked to a recent patch to RDE which would implement this <mekeor[m]>ACTION got a build error for "guix install emacs-next-tree-sitter --with-branch=emacs-next-tree-sitter=master" and wonders if it's an upstream issue <apteryx>why does cargo error out with "error: no matching package found" even for a dependency marked as optional? <unmatched-paren>apteryx: cargo requires that all packages are present, even if they are optional >:D <apteryx>so I need to patch it out from Cargo.toml <apteryx>also... the error message is not very helpful toward that goal :-) <apteryx>does getaddrinfo always require going through the libc resolver service? <apteryx>that doesn't appear to work in the guix container, even for loopback domains such as "localhost" (defined in /etc/hosts, even in guix containers) <msavoritias>I installed nss-certs just in case but it didnt fix the problem <msavoritias>I am trying to run lutris on flatpak and I am getting <msavoritias>Download failed: Could not find a suitable TLS CA certificate bundle, invalid path: /run/current-system/profile/etc/ssl/certs/ca-certificates.crt <msavoritias>is there some specific variable that needs to be added in the shell <remyd1>apteryx, may your issue be related to network namespaces ? <apteryx>it is, I'm asking in the context of enabling that scenario in the Guix containerized build environment <apteryx>(with is setup with clone and the CLONE_NEWNET option (so the process runs in its own private network namespace) <mirai>apteryx: within a container, /etc/hosts doesn't exist I think <mirai>as a result, it doesn't have the typical localhost entry <mirai>oh ok, does it have the appropriate entries? <apteryx>it has at least localhost: 127.0.0.1 localhost <apteryx>so I'd expect 'getaddrinfo' on localhost to work, although I haven't RTFM fully yet. <apteryx>perhaps it indiscriminately goes through the DNS resolver of libc, ignoring /etc/hosts <remyd1>is /etc/resolv.conf a real file ? <apteryx>real file on my system (it's generated by resolvconf, per its header comment) <henzo_midala>Hi, Today in my country the daylight saving time is not applied anymore but Guix was not taking this in count and shifted the hour. I tried ntp without success. What's suggested here? <remyd1>apteryx, can you try to override this file with (for instance) the content "nameserver 9.9.9.9" ? <mirai>henzo_midala: this is related to the TZ files <mirai>wonder what core-updates thinks on it <mirai>apteryx: has staging been merged into master yet? <mirai>or is it still feasible to include one more patch into it <cow_2001>today i had my first guix install failure. it was guix install gnash :( <cow_2001>without that we've lost much of the 00s web art <mirai>I remember there being an archive of flash games <cow_2001>and rustlang is curl | sudo sh based installation, which gives me the hibbie jibbies <unmatched-paren>i guess that boost::allocator_allocate(...) function's signature was changed since gnash was last updated... <mirai>henzo_midala: I recommend you open a issue in bug-guix <mirai>you'll want to mention that the tzdata package needs to be updated <unmatched-paren>that's why there are two tzdata packages; tzdata and tzdata-for-tests <unmatched-paren>someone would need to update tzdata making sure tzdata-for-tests doesn't change <cow_2001>unmatched-paren: should i post a bug report or do you have it under control? <mirai>unmatched-paren: so what's yielding 8138 dependent packages? <cow_2001>unmatched-paren: thank you! but……… how much of the logs should i post? <cow_2001>the last parts of the log file or the whole file? <unmatched-paren>mirai: found it! tzdata-for-tests is a native-input of python, i think that might be the problem <unmatched-paren>mirai: anyway, there's a comment above TZDATA-FOR-TESTS that explains what it's for <apteryx>mirai: no, I'm ironing out some failures still in the rust department <mirai>apteryx: could you add #62624 when you get the chance to? <apteryx>remyd1: should I use 'nameserver 127.0.0.1' instead? <mirai>it fixes some rather frightening bugs <remyd1>apteryx, I do not think so, but it might be worth testing it also <apteryx>remyd1: I tried both, I still get the error resolve: Host not found (authoritative) <apteryx>after a fresh guix-install.sh invocation, are both bordeaux and berlin substitutes supposed to be active? <apteryx>ACTION thinks yes, but a user gets the openssl expired certs time bomb problem, which suggests they aren't, or that the substitutes no longer exist for that derivation <apteryx>are the substitute serers used even when the daemon runs without --substitute-urls ? <apteryx>right; is it possible to have some limited DNS resolver working in that container? One which would know at least that 'localhost' -> 127.0.0.1 <apteryx>that should help with some test suites <Parnikkapore_m>How do you delete a GC root once it has been manually registered (e.g. via guix build)? <apteryx>guix build doesn't register guix roots <apteryx>Parnikkapore_m: of if you want to delete a single store item, 'guix gc -D /gnu/store/...-item' <Parnikkapore_m>so once you have done that (or registered a new root via other means, e.g. guix shell -r), how do you remove that root? <TristanCottam[m]>unmatched-paren: Are you the author of the first 2 "Dissecting Guix" articles? <TristanCottam[m]>I'm going through those two first, then let you know about my feedback on the third one <henzo_midala>JFTR, to avoid daylight saving times shift regardless of political decisions we can set /etc/timezone to /Etc/GMTXX where XX can be -12,-11,...+11,+12. Of course, this needs to be set using config.scm. <gbot>welcome user_oreloznog_, pins for you: <html> <TristanCottam[m]>unmatched-paren: Can I DM you about my feedback to avoid polluting this channel? <unmatched-paren>TristanCottam[m]: sure, though i personally wouldn't call that pollution <TristanCottam[m]>Can you send me a message? I'm having problems creating the conversation <NewUser-Basic-Qu>Hi all. I finally build my package successfully in guix shell and see the correct version installed there. How to install it in my "real" system. <unmatched-paren>NewUser-Basic-Qu: are you doing something like ``guix shell -L .'' or ``guix shell -f FILE''? <unmatched-paren>if the former, replace ``shell'' with ``install'', otherwise replace it with ``package'' :) <NewUser-Basic-Qu>did a "guix shell -D guix --pure" on a git checkout where i adapted some files. <NewUser-Basic-Qu>i have as prompt "bash-5.1#" and it seems no "guix" available in this environment <NewUser-Basic-Qu>i have the possibility to put a "./pre-inst-env" before it to get guix but that installs not in my system. I would like to use the packages in my system <pusherofbrooms>Did you modify a package in place in the cloned guix repository? <pusherofbrooms>I'm two weeks into guix, so my advice may be bad, but in that case I think you could setup your local guix checkout as a channel. <pusherofbrooms>You might consider pulling out the expression for your package (the "(define public PACKAGENAME" part) and all of the use expressions at the top into its own package file separate from the guix checkout. You could then use "guix build -f FILENAME" to build it. <pusherofbrooms>If you break the package out into its own file, you might need to paste "PACKAGENAME" at the bottom of your new scm file. <NewUser-Basic-Qu>I would like to have it running in system. like when issuing "system reconfigure" <pusherofbrooms>ah, that sounds like you may need to make a new channel with this package (I've never done that). <Fare>Is there a Guix-friendly or Guix-based way to do deploy many hosts with Guix? Something like NixOps or DivNix, but for Guix? <ngz>Fare: Have you tried "guix deploy" command? (I never used it myself) <Fare>oh, I thought guix deploy was only for the current host. <Fare>thanks, I'll look into it <zyd>Guix warns/recommends: [1680558524.946]: [ WARN ] We recommend installing and/or starting your distribution 'nscd' service <zyd>I'm not familiar with either nscd or systemd-resolved and sssd, but I have two questions: 1. Does this break Guix packages on Fedora? 2. Is there a good way to test that name lookups are working as Guix expects/needs so I can determine whether or not I need to look more into this problem? <apteryx>arg. python-cryptography is a royal pain in the butt <jgart[m]>What should be the Guixer's approach to developing with Rust? <jgart[m]>I was speaking with John Wiegley over at #nixos about developing with Rust and Nix (John uses Nix to hack on Rust) <jgart[m]>They mostly use Nix to install rust tooling like the rust compiler, language server, linter, etc. <jgart[m]>And then they defer to cargo for the crates. <jgart[m]>I would love to atleast be able to do that with Guix <jgart[m]>But our rust and rust-analyzer is too out of date. <RavenJoad>What is the package/service that places /etc/ssl/certs/ca-certificates.crt in place? I know nss-certs gets all of the certs, but the output of guix build nss-certs lacks the ca-certificates.crt file. <apteryx>jgart[m]: you can already do that with Guix <apteryx>perhaps you could try your hand at updating it? rust packaging is typically more boring than difficult, in my experience <RavenJoad>apteryx: For a guix home config. I am working on my own services to build config files for me. I need the certs for emails. <Fare>doesn't the latest linux kernel include modules in rust? Does that mean rust is now necessary to compile it, and Guix must have a working rust toolchain? Or is the problem with library packaging, not the base toolchain? <jgart[m]>apteryx: I would like a better importer before I do that rust work <jgart[m]>the base toolchain is also not the latest iirc Fare <Fare>jgart: thanks a lot for linking me to the pinephone pro guix configuration, BTW! <jgart[m]>Fare: charje implemented a way to get the quicklisp experience but only with Guix and ASDF <jgart[m]>Fare: If you're able to try it, I'd be curious what you think given you're history and expertise with ASDF and Common Lisp. <Fare>jgart: these days I've mostly moved to Gerbil Scheme. <jgart[m]>Fare: Ya, I saw. Why Gerbil and not Chez or Chicken? <TristanCottam[m]>Hi everyone, I'm trying to create a Guix package for minetest-server, by creating a variant of the minetest package. How can I append additional #:configure-flags to the inherited ones? <Fare>Cool that you update CL servers in-place. Do you erase data by unbinding or uninterning symbols, or do you use update-instance-for-redefined-class ? <cel7t>Tristan Cottam: if I recall correctly, minetest did not support configure-flags <cel7t>Can you go to the online Guix source and check guix/build-system/minetest.scm? <cel7t>If it supports configure-flags you'll see a #:configure-flags <TristanCottam[m]>I'll try it out, thanks! What's the difference between inherit and package/inherit? <cel7t>I am not entirely sure, but I think for your use case (inherit <package-name>) would make more sense <attila_lendvai>i get a "In procedure symlink: File Exists" very early in the boot on a VPS that i `guix deploy`. does it ring a bell for anyone? <attila_lendvai>is there a way to have command history in a guile repl that comes up when there's an error in init 1? my ultimate goal is to boot a previous config, and for that to change the grub timeout value in /boot/grub/grub.cfg, because the VPS console is too slow to catch the grub menu. <davidl_>Hi, I have a displaylink device that should be able to work with udl and udlfb kernel drivers (this also shows up in dmesg), but Im getting an error from udev: <davidl_>udevd[206]: specified group 'plugdev' unknown <davidl_>I understand that Guix doesn't use the plugdev group as in Ubuntu, so what is the appropriate workaround here? <davidl_>for context, Im trying to use an USB-VGA-165 displaylink adapter to connect a monitor via usb 2.0. <attila_lendvai>can i change the default generation that will be booted from the early guile repl? <gabber>attila_lendvai: you mean other than roll-back / switch-generation? <attila_lendvai>gabber, anything that i can do from the guile repl i get very early in a failed boot <attila_lendvai>what would help me is to initiate a soft reboot from guile, so that the browser based console display doesn't re-initialize (and miss the grub menu) <attila_lendvai>if i ,q the guile debugger, then i get a kernel panic screen, and i cannot send a CTRL+ALT+DEL to the VPS <attila_lendvai>can't use the glibc reboot function, because it's not defined at that point, as per (dynamic-func "reboot" (dynamic-link))