<str1ngs>daviid: I do need to test nomad with msys2 . have you tried g-golf yet? <daviid>str1ngs: no, but i hope someone will do that, I just don't have time, I can only concentrate on upstream g-golf, too many things still need to be done <str1ngs>daviid: understandable, I will look at it at some point. will let you know. <daviid>I hpe g-golf becomes an official package (it uses pacman from arch iiuc) sometime ... <daviid>i'd like guile-cv to also have an offcial package on msys2 ... <str1ngs>daviid: do you plan a alpha release soon. i was thinking g-golf is stable enough now to add to guix . WDYT? <daviid>str1ngs: not untill I solve known serious bugs - such as the one posted bu tohoyn (the list store - cell renderng example), but there are others ... <daviid>there are other know bugs, gtk-init (and all *-init segfault unless passed 0 #f ...) <str1ngs>daviid: no problem. I'll hold of on submitting my g-golf.scm for now <daviid>str1ngs: but let's move the g-golf related conversation on #guile ... so for the g-golf, guile-cv ... port on msys2 ... <str1ngs>do you mean *-init seg faults with the wrong argc length? <daviid>str1ngs: ok, please keep the name G-Golf for the guix package name, not guile-g-golf ... <str1ngs>right I was going to submit that way as per your request. <daviid>str1ngs: i don't know about 'experimental' package in guix, but if that exists, yu could certainly publish it as experimental *NieDzejkob is packaging zola even though he should probably be asleep <sirgazil>Now I have two python packages I need for Django 3 <sirgazil>I've never run guix gc since I installed the system. <sirgazil>When a package is needed both for running tests and at runtime, should it be included in native-inputs and propagated-inputs or only in the latter? <lfam>sirgazil: Usually just in 'inputs' <lfam>propagated-inputs are only necessary in special cases <lfam>native-inputs are things that are *only* used while building. If something is used while building and at run-time, it should be an input. <lfam>Propagated-inputs are more common for some languages. Please ask if you have more questions <sirgazil>lfam: Then "guix import pypi" could be contributing to my confusion, and maybe also the manual about packaging Python things. <lfam>sirgazil: Sure :) Python is one of those special languages. It needs a little help finding it's run-time dependencies so we use propagated-inputs, which means they are installed along with the package that is depending on them. <sirgazil>lfam: In that case, in the context of a python packages, when a package is required for both running tests and at runtime, should it be included in native-inputs and propagated-inputs or only in the latter? <sirgazil>Say python-abc requires python-pytz for running the test suite and also at runtime. <lfam>What "native" means here is related to cross-compiling. If you are cross-compiling, and only need some package for compiling (like make), you only need that package for the build machine's architecture, not the target machine <lfam>But whatever is an input or a propagated-input will also be available in the build machine's architecture <lfam>You would have two copies of those programs then, for the build arch and the target arch <lfam>By saying that something is a native-input you are telling Guix, "I only need this package in the build machine's architecture" <lfam>So that is why native-inputs are just build-time dependencies <lfam>The manual section "package Reference" has some more explanatoin <anadon>Anyone have an idea why I can't post to #guile ? <nckx>anadon: Is your nick registered with NickServ? Some channels +q unregistered users (we do during spam waves). <nckx>If you were able to talk in #guix a week ago, that's probably not it. <anadon>(22:39:22) Cannot send to nick/channel <anadon>Oh hold on. Password is wrong. It was working. <anadon>Got it. Looks like I need to update my email address, though. <apteryx>I have a system that fails to boot when upgrading, and wondering which of the two following errors are likely fatal? 1. udevd conflicting device node link to /dev/dm-1 will not be created 2. CP437: Invalid argument <apteryx>After these messages, the file-systems fail to mount and chaos ensues. <Kimapr>Download and deploy the latest version of Guix. <Kimapr> -s, --system=SYSTEM attempt to build for SYSTEM--e.g., "i686-linux" <Kimapr>this is what i wanted a few days ago <apteryx>with Guix, your wishes come true (TM) <Kimapr>i'm doing system update and user profile update simultaniously <anadon>Guix does a lot of neat, not obvious at first things. I'm caught almost daily by this. <nckx>apteryx: Sheesh. Both are possibly fatal. My guess is the CP one. Try disabling fsck (check #f) for your file systems (especially FAT ones; is this an UEFI machine? maybe even set needed-for-boot #f for your ESP then). <apteryx>nckx: UEFI yes. This UEFI thing keeps biting me somewhere every now and then, it seems ;-) <nckx>Did you mess with your initrd-modules at all? <apteryx>nckx: will try this first thing tomorrow, sounds easy! <apteryx>nope. I thought I had messed my btrfs subvolumes patches, but they are doing well on another (non-uefi) system. <nckx>apteryx: That's just a first random debugging step, I don't know if it will fix anything. <apteryx>I suspect the UEFI fat partition is dirty, there was an electric failure very recently. <nckx>It ‘feels’ FAT-related is all. <nckx>The feeling intensifies. <nckx>apteryx: *In addition* to the above try adding (mount? #f) to the /boot/efi file-sytem entry as well, reconfigure, and reboot. <nckx>apteryx: You can also give fsck'ing the ESP from a rescue system a go if that's an option. <PlasmaStrike`>is there any video url links for guix days that happened after fosdom 2020? <raghavgururajan>nckx: Oh shoot, sorry, I slept. I just meant prefer xmpp as it is extensible in design. <guix-vits>raghavgururajan: nckx fall asleep near an hour ago. ***sneek_ is now known as sneek
<brendyyn>if i have a let that sets foo in one build phase, how can i make that foo available in another phase after it? <guix-vits>brendyyn: i'm not a developer. if you `(define-public x (let ...` and then `(define foo x` and then use @x? i'm wonder... <brendyyn>i know i could use setenv to pass it outside via the environment, but seems kinda wrong <brendyyn>im also confused why package-with-python2 is failing to rename this package from python- to python2- <guix-vits>brendyyn: but you want to pass some local value to another procedure, anyway? <guix-vits>so `(define-public` (which will not be an SHELL env) may work (hope it'll not broke another package, though). <brendyyn>maybe define would work, ive just never seen it in a phase before <brendyyn>i see. i think package-with-python2 only works if its python-build-system <guix-vits>(define-public python2-pyte (package-with-python2 python-pyte)) <brendyyn>yeah i know i cant use that since its gnu-build system, so ill use inherit <brendyyn>cannot link `/gnu/store/.links/1adkpr52z6bz53r0sri8nfkbkddf15zsfwhvfhmz4gk8rbz96mhb' to `/gnu/store/bg22kysvl3f337l7wzx5kc38niia466f-GConf-3.2.6.tar.xz.drv': No space left on device <brendyyn>/dev/sdb1 61054976 9906310 51148666 17% / <guix-vits>about python2-pyte: maybe the definition should be removed from terminals.scm? <guix-vits>returns only ./NEWS and ./gnu/packages/terminals.scm <guix-vits>only if there python2-pytest and pyte was on one line, though. <PurpleSym>My main issue is that I don’t understand why guix complaints in the first place. Any hints? <civodul>PurpleSym: the problem is that you indirectly end up with two different python-nbformat packages in your profile <civodul>they are pulled via propagated inputs <PurpleSym>Hey Ludo :) Ok, but I don’t have a second python-nbformat defined anywhere. <civodul>the message shows the chain of propagated inputs that brings it in <civodul>perhaps the two different variants are because not all the packages come from the same Guix revision? <civodul>you can address that by first running "guix upgrade" <civodul>or (better) by using "guix package --manifest" <PurpleSym>I’ll try `guix upgrade`, because I’m experiencing the same issue with a slightly different chain when using manifests. <civodul>if you have the problem with a manifest, then it means there's a "real" problem <civodul>that is, if you look at the dependency graph, you should see the duplicate node <Blackbeard>I tried to install Boinc before going to bed but boinc-server is not compiling <Blackbeard>It is 3 am so i'll check in the morning again. But if anyone can confirm please let me know <guix-vit`>Blackbeard: civodul here, maybe you can post the error... <guix-vit`>PurpleSym: maybe you'll install those in different profiles, or use `guix environment` (in the mean time)? <PurpleSym>guix-vit`: Strangely `guix environment -m manifest.scm` works. Maybe it does not check for collisions? <civodul>PurpleSym: right, so if you run "guix graph python-jupyterlab", you should see two different python-mistune packages <civodul>and yes, "guix environment" doesn't check for collisions <civodul>(and you could also use '--allow-collisions' here, but that's cheating :-)) <guix-vit`>civodul: aren't this can broke the generation? <brendyyn>i hate deciding which module a package should go in. <PurpleSym>civodul: Ok, it’s tricky to look at the actual image, but if I’m reading the .dot file correctly it looks like the first mistune is pulled in by python-nbconvert and the second by python-m2r? <civodul>BTW you can use xdot to navigate the graph <PurpleSym>Thanks. What do you mean by “because not all the packages come from the same Guix revision”? The packages come from two different guix channels (my own and guix proper). Is that a problem? <civodul>what i meant was if you did: "guix install python-nbformat; guix pull; guix install python-jupyterlab" <civodul>problems you don't have if you use --manifest <civodul>but here you're using --manifest, so that means there are really two different python-mistune packages in the graph <civodul>just to be sure: does using --no-grafts make any difference? <PurpleSym>Ok, I understand. No difference with --no-grafts, still two colliding entries for python-mistune. <civodul>PurpleSym: do you have code somewhere that inherits from python-mistune and creates a variant or something like that? <civodul>or a package-mapping call or something? <PurpleSym>I’m not touching python-mistune at all. jupyter-zpid is created using package-input-rewriting. <civodul>i realize you have a pending patch for jupyterlab, BTW :-) <civodul>it's a bummer that there's all this JS code <PurpleSym>Yeah, probably not going to make it into guix proper because of that… <civodul>but probably we can include everything up to jupyterlab (excluded, for now) <civodul>the package-input-rewriting thing is likely the culprit <civodul>it probably introduces a python-mistune variant that you don't care about but that's there anyway <PurpleSym>Hm, you mean the #f case in package-input-rewriting? <PurpleSym>(Changing it to (#f p) does not make a difference.) <sturm1>Hi folks, had anyone else had issues recently with programs like Evolution and Dino not being able to login? I'm seeing DBUS errors like "blah blah was not provided by any .service files" <civodul>sturm1: that's with GNOME on Guix System? *civodul hasn't seen it but does not use GNOME <brendyyn>do we have python package with the 'gi' module? <sturm1>civodul: that's right, Gnome on Guix System <sturm1>civodul: yes has worked fine previously - perhaps I should try rolling back some versions to see if I can pin it down. I didn't notice when it actually started hapenning <Formbi>I think there's something wrong with the libgccjit package <brendyyn>im wondering with python how do i decide if an input should go in propagated inputs or not <brendyyn>A requires B requires C. putting C in B's propagated inputs works, but putting C in both B and C's inputs also works i think. which is better? <civodul>sturm1: could you send the details to bug-guix? like: what applications fails with what error message, the output of "guix system describe" for both the "broken" generation and the one that worked <guix-vit`>brendyyn: i'm not a developer, but probably "putting C in B", because DRY. *raghavgururajan is primary nick *raghav-gururajan is secondary nick ***guix-vit` is now known as guix-vits
*guix-vits only now realized.. *guix-vits just forget to quit before reconnect <sturm1>civodul: it looks like I've removed my last good generation - would you mind doing `find /gnu/store -name org.gtk.GLib.PACRunner.service` and see if you get any results for me? I don't <apteryx>nckx: I forgot to mention that the weird thing is that if I choose a generation before the last problematic one, it boots. I'll compare the dmesg. <sturm1>thanks guix-vits, do you have it in your store at all? <sturm1>neither do I really - just trying to figure out if the problem is that it used to exist and now it doesn't, or that it's a new requirement that's not present ***sturm1 is now known as sturm11
<guix-vits>sturm11: i'm also have no Evince nor Gnome installed, though. <sturm11>guix-vits: that makes sense then, thanks <brendyyn>wrapping programs is voodoo magic and i dont like it <guix-vits>brendyyn: please, explain: do you mean that the package A will have a fake dependency X + ";; needed by B", and A's true dependency B will also have X as dependency (in case someone need just B)? <brendyyn>no module nameed piper.. oh dear... the program im running *is* piper. <guix-vits>brendyyn: you'd explained you opinion about "A needs B; B needs X" (X isn't fake dependency for A). <brendyyn>now im building a python program but it requires meson, so it doesnt wrap pythonpath at all <guix-vits>maybe python will look for modules in current dir. <zogzogzogog>Hello, I'd liketo know if someone is using guix/guixsd on an armhf (arm32) SBC, I'm trying on a H3/2GB RAM, and its really slow (guix pull taking ages, f.e.) <rekado_>zogzogzogog: 2GB RAM might not be enough memory for “guix pull” with the current version of Guile <rekado_>(recent changes to the compiler have improved it a lot but also made it require quite a bit more memory) <rekado_>If you can I would suggest offloading all builds to another machine (that could emulate armhf). <zogzogzogog>I'm seeing a guile process taking between 100% and 400% of the 4 cores, but its virt/rss is only at 735/564 MB... <brendyyn>rekado_: do you think that situation will improve? it seems a little embarrasing that we can't run in 2G of ram. That's a hell of a lot. <rekado_>“we can’t run” = it is some derivations that are expensive <rekado_>we primarily depend on improvements to the Guile compiler; Guix itself doesn’t do anything terribly wrong in how it’s using Guile. <guix-vits>zogzogzogog: don't forget to run `guix weather` after `guix pull` and before `guix package -u` to see if the substitutes (binary downloads) are ready for the apps of your choice. <zogzogzogog>Here, it's a guix pull before the guix system init to install the os, I guess I'll have to cross-install ***calher is now known as KE0VVT
<civodul>rekado_: which changes increased memory requirements? <civodul>(i agree that the situation is embarrassing) <zogzogzogog>guix-vits: it is currently an armbian that I use to try guix on this SBC... <rekado_>civodul: I just mean changes since 2.0 made the compiler hungrier. <civodul>ah ok, i thought you were talking about something more recent <civodul>3.0.1 should be a bit better, but still not as good as 2.0 in that regard <anadon>Should `sh` be a meta-package to install some posix compliant shell program? I'm like to not make my projects be bash-implementation specific. <jakobrs>When I'm setting things up, is ~/.config/guix/current supposed to point to <jakobrs>/var/guix/profiles/per-user/root/current-guix or /var/guix/profiles/per-user/$USER/current-guix? <anadon>guix-vits: Just that in a manifest, I would like to specify a shell by standard rather than implementation. <joshuaBPMan>hey guix, I'm trying to respond to a guix bug report that I made a long time ago. I can't find it in my inbox. What is the secret to how I can respond to it? Essentially to say "please close this bug report". <anadon>guix-vits: That was it. I did not know that was the thing, thanks. <guix-vits>anadon: it's light (just there other tools also in that package)? <guix-vits>i'd used `guix search 'shell' 'POSIX' |less` <anadon>Odd, the `sh` command is missing, suggesting the symlink or executable are not included in the package. <guix-vits>anadon: maybe they're packed up in one binary? <apteryx>brendyyn: there's a dbus-root-service-type though <brendyyn>i want to modify the dbus-service somehow <apteryx>you can probably use the usual trick, i.e.: (modify-services %desktop-services (dbus-root-service-type config => ... <anadon>guix-vits: If that were true, then there would be a symlink. <brendyyn>why is it called dbus-root-service-type? <bricewge>brendyyn: Because it's the system bus (ie. run by root) in oposition to the session bus. <guix-vits>anadon: i'm sorry anadon, i'm `environment`, cd $PATH, cd .. -- not found by eye; but guix show `coreutils-minimal` promises shell. <anadon>`coreutils` doesn't have `sh` either <anadon>nckx: Should coreutils* have `sh`? <nckx>coreutils doesn't have a shell. That's not a bug. <nckx>Install a shell like bash, zsh, … <nckx>(Guix's default sh is bash.) <anadon>Can I specify `sh` as a standard for a manifest rather than a specific implementation? <nckx>sneek: Later tell brendynn: use the old-style (services (dbus-service) …) notation. <guix-vits>anadon: if not, "BusyBox provides a fairly complete POSIX environment for any small or embedded system.", "It is also extremely modular so you can easily include or exclude commands (or features) at compile time." <nckx>anadon: …no? A manifest is a list of packages. Guix wouldn't know what to install. <nckx>Yeah, busybox has several nice tiny shs. <nckx>anadon: I'm missing context. What are you trying to do? <anadon>OK. I just like to stick to specifying standards rather than implementations. <guix-vits>nckx: o/ ; `guix show coreutils-minimal` -- aren't it promises shell? <anadon>nckx: I'm starting a new coding project and making the `manifest.scm`. <nckx>guix-vits: coreutils are a well-known collection of command-line tools like cat, cp, ls, rm, … that are vital to using the shell but does itself not contain a shell. <nckx>I can see how our description could be confusing though. <nckx>‘These provide the basic file, shell and text manipulation functions of the GNU system.’ <nckx>Means it provides ‘shell functions’ but that's quite obtuse. <anadon>That can be interpreted either way, with the default being the use of a MLA and, omitting the oxford comma which says that the shell is an included part. <jakobrs>Does guix have something like `nix run`/`nix-shell -p`? *nckx inserts Oxford commas into Guix whenever given half the chance. This just needs to be rewritten though. <nckx>Is this the upstream description? <nckx>It is. I prefer Wikipedia's, but we tend to sync with the GNU description for GNU tools. <nckx>jakobrs: Could you reword that without using nix examples? <jakobrs>nckx: Starts bash with a specified package available, without installing it into your profile <nckx>guix-vits: Since this genuinely mislead you & you're not just nitpicking, I'd say it's a valid bug if you want to report it. <nckx>jakobrs: guix environment --ad-hoc <packages> <guix-vits>nckx: maybe anadon, as perse know about the Oxford comma (as patch to description is better)? <guix-vits>i'm just have no idea how to improve it aside of removing "shell". <nckx>Relying on something as pedantic as the Oxford comma (and again, I'm all for pedantry and Oxford commas!) isn't really acceptable. <guix-vits>i'll start then (will look up a few places for descriptions: Debian, Arch, Wikipedia as you'd told). <nckx>guix-vits: Thanks. I really need to get back to work. <anadon>nckx: "These provide the basic file, shell manipulation, and text manipulation functions of the GNU system" <NieDzejkob>file manipulation, shell manipulation, and text manipulation? <nckx>The meaning is ‘file functions, shell functions, and text manipulation functions’. <nckx>I guess they mean stuff like [ and echo. <nckx>Which most shells shadow anyway but whatevs. <NieDzejkob>can we just, like, throw in a parenthesis for grouping? <guix-vits>NieDzejkob, anadon: (?) "The basic file, shell and text manipulation utilities of the GNU operating system" -- Arch <anadon>It still doesn't parse. the nested grouping with the word shell kills readability. <guix-vits>"..., sh-utils (shell not included), and ..." <bricewge>I'm doing some bug triage since it's the theme of the week. <bricewge>Should I ask before closing or merging bugs? How does the authorization regarding control@debbugs.gnu.org works? <nckx>Logged back in because this bugged me so much: it's actually ‘file manipulation functions, shell functions, and text manipulation functions’ which makes NO sense. I stand with anadon. Maybe now I can let this go. *nckx whispers ‘there is no authorisation’ <brendyyn>how can i make the guix-edit command in emacs go to my git repo instead of the read only store? <sneek>Welcome back brendyyn, you have 1 message! <sneek>brendyyn, nckx says: use the old-style (services (dbus-service) …) notation. <nckx>bricewge: If you're sure you're correct, you don't need to ask for permission. Otherwise post ‘is it OK to close this?’ etc. to the bug number. To merge, just send ‘merge nnnn mmmm’ to the above address and debbugs will comply. <nckx>anadon: Hah, not sleep, work. I wish 🙂 <jakobrs>I thought I should point out that the version of Nix in the Guix package repository is vulnerable to CVE-2019-17365. <brendyyn>i cant really understand the nix code, but it seems to enable it as a dbus service. what should i do for guix? <brendyyn>do we having something like if a package is added to the system profile, it will automatically modify the dbus service appropriately? <rekado_>I’m giving up on upgrading tensorflow. Their CMake stuff is so *very* broken that I can’t get it to build even with extensive patching. <rekado_>I’ll try to get bazel built from source instead. <civodul>let's hope Bazel development freezes in the meantime <janneke>maybe do some social engineering to hook a bazel dev onto guix *brendyyn cat rekado_ | guix offload Hawaii <civodul>we did a bit of that, but apparently that was badly engineered <anadon>rekado_: We remember you fondly, and keep your name in contributions in your absence. <guix-vits>*excluding, of course; send it or you decided to do it best, anadon? <rekado_>argh! “Build Bazel from scratch […] Download bazel-<version>-dist.zip […] The distribution archive contains generated source files” <rekado_>that’s not what “from scratch” means! <anadon>rekado_: We already confirmed that the Bazel folks are so far into 'practical' that they're impractical. *janneke feels for rekado_ <rekado_>perhaps this could be used to break the cycle <bricewge>brendyyn: (simple-service 'ratbagd dbus-root-service-type (list libratbag)) <janneke>rekado_: would it be feasible to use this "from scratch" method to produce a shell script? <bricewge>I tried with dunst and it added org.knopwob.dunst.service to /etc/dbus-1/system-services *rekado_ removes Bazel from bootstrappable.org <brendyyn>bricewge: thanks. im kinda mystified as to how that works though *kmicu remembers that with some tweaks we could use Nix under ~512MiB of RAM so that more‑than‑2GB‑for‑guix‑pull requirement when Guix has order of magnitude less packages is not very future proof 🤔 <brendyyn>anadon: software freedom implies that something is not just open source, but you can actually get the source without going through great pains to get it and setup an environment to build the software <bricewge>brendyyn: I don't know either. We should probably look into the definition of simple-service to get an understanding. <anadon>brendyyn: But the reproducible builds project doesn't specify software freedom. It specified reproducibility. <brendyyn>bricewge: im having a look. im just wondering how the (list ...) bit gets put in to the (services...) part of the dbus service <nckx>anadon: I'm not involved with the R-B folks but I think ‘from source’ is implied. <nckx>‘This here quinary poops out perfect copies of itself’ reproducibility is not. <rekado_>if you have a binary and you copy it to other places: that’s trivial reproducibility ***drakonis1 is now known as drakonis
<svinoba>Hi, is anybody using guix standalone in another distro and able to start X? I didn't face any problem with GuixSD so far but am frustrated not being able to start x on Debian. <civodul>rekado_: how 'bout implementing Bazel on top of Guix? <civodul>it's the kind of idea that roptat would love i'm sure :-) <mbakke>svinoba: what happens when you try to start X? <svinoba>I am not able to install lightdm. The build fails. GDM installed OK but unable to use it. systemctl says no gdm.service. If I try to launch with xinit it fails to start and comes back to console with some error code. <svinoba>If I see Xorg log, it states vesa/fbdev modules not found and cannot open virtual console 2 permission denied. <anadon>Where is the "MIT" license? It isn't under licenses.scm. <svinoba>I first installed xorg-server, xinit, xfce as root. Cannot use GUI. I then installed them as myself (normal user). Still cannot launch GUI. If I launch X as root, it launches xfce but mouse click and keyboard does not work at initial panel selection of xfce. <rekado_>civodul: I wonder if that might even be faster than properly bootstrapping Bazel. <rekado_>I’m opening an issue on Github about improving the bootstrapping story. <anadon>What package has `bzless`? `guix search bzless` produces no results. <anadon>Huh. I wouldn't have guessed that. <mbakke>svinoba: I guess you'll need drivers such as xf86-input-libinput to make peripherals work. <svinoba>mbakke: I have installed them. xf86-input-libinput, xf86-video-vmware, xf86-video-vesa. <svinoba>I am trying it on Debian in a virtualbox so I'm not able to get Xorg log to share. <mbakke>svinoba: display managers usually need integration with the host system (think PAM, or setuid), so it's probably best to use one from your distro and point it to Xorg from Guix. <svinoba>I actually had installed lightdm from distro. It would launch but then when I entered my credentials, it would try to login to DE but comeback to login screen. <mbakke>svinoba: I haven't used "plain" Xorg for many years, so I don't know why the drivers aren't picked up. Hopefully others around here can chime in. <mbakke>Any particular reason you want to use Guix's Xorg, and not Debians? <mbakke>I have it on good authority that it works on Debian with a tiny tweak to the xinitrc script <svinoba>mbakke: any help is appreciated. Thank you. As for using Guix xorg, I'm just trying to learn. Use Guix for everything except minimal install of Debian. If you say, it is better to use it only as package manager I will. <ngz>I noticed an annoyance in manifests : packages are installed in reverse order, i.e., they appear in reverse order when you list them from guix package -I. <svinoba>mbakke: thanks for the link. I will read up and try. :-) <mbakke>svinoba: Cool :-) let me know if you have any questions! <rekado_>so, you *can* build Bazel without Bazel, but you still need a large number of pre-built third-party jars <rekado_>five of them are part of Bazel, so building them might be tricky. <tsmish>hi #guix. I installed guile-next and on start it warns me about incompatible bytecode. From what I can understand it is related to GUILE_LOAD_COMPILED_PATH. So, is there a way to fix this without hacking %load-path in guilerc. <rekado_>I think this might actually work… It’s just 120 jars… <mbakke>rekado_: but each of those probably has a few dozen dependencies, right? :) <rekado_>but at least it’s a packaging problem, not an immediate bootstrapping problem. <mbakke>sounds like quite the challenge! <rekado_>the 120 dependencies might turn out to lead to 120 bootstrapping problems, of cours <rekado_>I think the 120 jars are due to laziness, though <rekado_>I can’t believe that *all* of them are needed. <rekado_>I guess we could cut out a few features <rekado_>5 are JDK jars. I assume that we can use our JDK. <rekado_>one is junit, another is jsr305, another is hamcrest-core. We already got them. <rekado_>jcommander and asm also already exist. <mroh>tsmish: try removing ~/.cache/guile if the bytecode is from there. <rekado_>I’ve had enough Java for this decade. <civodul>rekado_: i was half serious regarding Bazel on Guix, because there's probably a one-to-one mapping of the primitives <rekado_>civodul: for a subset of Bazel that might be true. <rekado_>but last time I asked they weren’t even willing to specify what features are optional and what are core features. <mroh>and next week, they implement something fancy and they are 180 jars as deps... <NieDzejkob>tsmish: I fixed it by removing all guile2.2 packages :PP <tsmish>NieDzejkob: doesn't guix itself depend on guile2 <NieDzejkob>tsmish: it's using guile3 for me when I do `guix repl' <NieDzejkob>besides, packages in your profile aren't^W shouldn't be used as dependencies by anything <NieDzejkob>so if you removed the packages, it wouldn't break <tsmish>actually, shouldn't envvars be set in shell shim script as guix already does for some packages. <infertux>Guix noob here, how can I find out about the service provided by a package? E.g. I just installed go-ipfs but 'herd status' does NOT show any IPFS service :/ <str1ngs>infertux: packages do not provide services. you need to add them with config.scm assuming you are using guix system? <infertux>str1ngs: so I need to add a new ipfs service in config.scm and configure it there as well similarly to openssh-service? <str1ngs>infertux: that's correct. reconfigure should be enough. then the first time you might need to start the service manually. which should avoid having to reboot. <infertux>how can I find out what are the available config options for a service? <infertux>e.g. "permit-root-login #t" for the openssh service <vagrantc>anyone able to get audio working with ungoogled-chromium? <vagrantc>how do i find the browser audio settings? <vagrantc>i suspect it's just using the wrong device <str1ngs>infertux: I'm not sure if there is a service for ipfs <infertux>str1ngs: I can't see any service with 'guix edit go-ipfs' - I assume it should be defined there if there is one? <str1ngs>guix edit go-ipfs will not show the service. services are separate from the packages. services are normally find in ./gnu/services/* <str1ngs>as far as I can tell there is no ipfs service. <tsmish>also i think there is guix system search for service search <str1ngs>vagrantc: would it not just use pulseaudio? does audio work in general? <infertux>str1ngs: just curious about your init.scm file - is that a standard file? I can't find it in my filesystem <str1ngs>infertux: my init.scm example should help you out. edit ~/.config/sheperd/init.scm and add the ipfs constructor etc. <str1ngs>infertux: init.scm is for shepherd user services. <str1ngs>you might need to tweak this contractor to stop the service cleanly. I personally only worried about starting the service. <str1ngs>then I use [[ ! -S /run/user/$UID/shepherd/socket ]] && shepherd in ~/.bashrc to one shot start somethings :) <str1ngs>non of this is all that sane, but it should get you running :) <infertux>str1ngs: would the proper way to do it be to create a dedicated ipfs user and run IPFS as a system service? <guix-vits>anadon: maybe (method git-fetch)? -- again, i'm not a developer. <str1ngs>infertux: I've used ipfs for sometime. personally I prefer to run as my user. maybe running as a special user might be useful. but really it depends on what file permissions you want. <str1ngs>infertux: maybe a guix service would be helpful as well for system deployments. <infertux>str1ngs: I don't want to use IPFS as a client in this case, I want to install it on a server along with ipfs-cluster to serve a bunch of data <str1ngs>infertux: in that case you might want to create a guix services for ipfs and ipfs-cluster. <str1ngs>you would need a way to configure ipfs-cluster keys as well. something to concider <guix-vits>str1ngs: is there a way to use the stylesheets (for HTML) in nomad? <str1ngs>also maybe away to connect to the swarm. it would be interesting configuring this with guix . and might be a good addition :) <str1ngs>guix-vits: hello guix-vits, do you want to change the default look of pages? <str1ngs>guix-vits: I think this possible. I do know that changing the GTK theme can effect page styles. it's a gtk thing. I'm not a fan of it though. <str1ngs>guix-vits: how much fine control do you want over this? <infertux>str1ngs: yes although I'm a Guix newbie... I like the ideas behind Guix System and would like to use it in production but I must admit the learning curve is pretty steep... <str1ngs>infertux: like any lisp the learning curve is worth the effort :) <guix-vits>str1ngs: color: #ddd; background: black; font-size: 18px; -- all i can remember. <guix-vits>suckless surf have this configured at build time, probably with WebKit2GTK tools. <infertux>str1ngs: haha I hope so, I can see the potential but need to consider the slow productivity during the learning phase ;) <vagrantc>str1ngs: not sure if i have pulseaudio installed <vagrantc>str1ngs: sound works without doing anthing in icecat <str1ngs>infertux: the slow learning curve. is easily made up for by repeatability :) <infertux>str1ngs: yes I would love to ditch tools like Ansible and Chef and deploy everything through Guix Scheme files <str1ngs>guix-vits: let me see how suckless does this. maybe I can add a trivial way to do this with nomad. for nomad I plan to add as much DOM introspection as possible. but it's tricky because it requires using a web extension to do it right. I have the frame work for using the web extension in nomad. but that's about it for now. <str1ngs>guix-vits: C-f does some trivial DOM introspection if you want to test that. <str1ngs>guix-vits: #nomad-browser is a good place to discuss this as well. since nly the other contributor wants to add more DOM introspection support. he might find this discussion interesting. <str1ngs>guix-vits: I'm just checking how surf does this seems you add files to ~/.surf/styles/ ? <NieDzejkob>Can offloading make use of things I've already built before on the slave machine? <guix-vits>str1ngs: i'm not remember so cleanly, but there is some hostname matching patterns in source code, and the .surf/styles/ used to describe what to do with those sites. <rekado_>NieDzejkob: it would transfer things that have been built and are not available on the offload machine. <rekado_>sirgazil: if this is a git checkout you may need to make the files writable first <str1ngs>guix-vits: I think I can easily add this with webkit_user_content_manager_add_style_sheet <rekado_>str1ngs: would be great to support transform procedures that take the DOM as SXML and return new SXML. <str1ngs>guix-vits: do you want per URI styles or just a mster style for now? <str1ngs>rekado_: right I planed to use sxml for this <sirgazil>rekado_: Thanks, I'll look for how to do that. <rekado_>sirgazil: search for make-file-writable in gnu/packages <NieDzejkob>rekado_: no, the other way around. I've built something on the offload machine, and now I need it as a dependency for something I'm offloading. It seems to be building it again, but I don't know whether I changed the package definition <str1ngs>guix-vits: I think for now I'll create a styles alist with uri's for keys and then an associated style if the uri matches it will set the style. and also a default style for all none matching uri's WDYT? that would be helpful. then in ~/.nomad you can just create your alist something like. (set! %styles (("https://gnu.org) . "markup here"))) <str1ngs>rekado_: probalby I'll use a string, then people can use strings or sxml as the preffer. <guix-vits>str1ngs: sorry, was flooded on ask-fedora. idk, but support for individual styling of some sites can be good, as some sites get broken badly with simple stylesheets; and favorite sites can be customized further (to `background:` not "disable" images that set as background) <mfg>How can i see all open issues on issues.guix.gnu.org? with the query is:open it only shows until ticket ~25000? is there an option to show more? <rekado_>mfg: for any query we limit the result set to 100. <rekado_>this is probably a hold-over from when we queried Debbugs directly. <rekado_>mfg: now we filter the set of ids for which we request information from Debbugs. <rekado_>mfg: we could probably get rid of this, but I stopped working on mumi after fixing the most urgent problems <rekado_>patches are very welcome, though! I can also answer questions here. <str1ngs>guix-vits: I think can manage per uri style sheets. It might be a little rudimentary but it will work. <guix-vits>str1ngs: important think is, if this stylesheets will be enforced, though. Seen this in Epiphany, but not in old surf versions. qutebrowser not enforces stylesheets at all: so if site says "white background", then "white background". <str1ngs>guix-vits: nomad will support more then just webkitgtk. webkitgtk just happens to be the reference implimentation. <str1ngs>ideally though users don't need to worry about GTK or webkit. I've tried to make things as generic as possible <apteryx>civodul: your updated Guix package doesn't build :-) (from 10 days ago) <guix-vits>str1ngs: good; also, about qutebrowser: C-f hints as a letters are better (aren't they?). <apteryx>./pre-inst-env guix build -e '((@@ (gnu packages package-management) current-guix))' --> /gnu/store/9kzrrccpzl6i1sfwb0drb00gi2gwk0x0-coreutils-8.31/bin/mkdir: cannot create directory ‘/etc/init.d’: Permission denied (in the install phase) <str1ngs>guix-vits: yes, I'm using a hint.js. that I need to rewrite it's not ideal it's kinda a place holder. It would help to have a JS expert to help with this. also I'd like to use a more scheme way of writing JS if possible <str1ngs>guix-vits: maybe a scheme to JS transpiler. or pure DOM introspection using scheme if possible <guix-vits>str1ngs: idk what DOM is, though will look at Wikipedia soon. <apteryx>civodul: I've created an issue with #40153 <str1ngs>guix-vits: document object model. it's what javascript uses to modify the document <str1ngs>guix-vits: mainly though hints.js needs to be rewritten which could use a JS expert. if not I'll have to rewrite it when I have more time to focus on it. <str1ngs>the large goal is to do DOM instrospection with scheme. then that solves more problems at once :) <guix-vits>str1ngs: also it will probably allow to block JS more reliable (as JS isn't used to make the hints work)? <str1ngs>guix-vits: if you have any other feed back regards to nomad let me know. <str1ngs>guix-vits: javascript can be disabled etc I have just not looked at a mechanism to do that generically quite yet. we have cookies and proxy setting no javascript settings yet. <str1ngs>guix-vits: #nomad-browser is a good place to discuss things yep <apteryx>is there an implicit dependency ordering enforced by the order in which my file system objects appear? that'd be convenient <apteryx>the list of shepherd services is returned in the order the file systems are given... now does that mean the assumed implicit dependency is respected? hmm <apteryx>I guess it could break the day Shepherd will start trying to start everything it thinks can be parallelized in parallel :-) <anadon>I think I got lost in the steam of text here. <anadon>It has to do with importing guile modules under python <Formbi>anadon: what is the error message? <bavier`>anadon: you should be importing the guix module instead <bavier`>anadon: e.g. the "python-six" package is found in (gnu packages python-xyz) <anadon>bavier`: how is ` #:use-module (gnu packages python-six)` not that? Or is it literally python-xyz? <Formbi>you can get info in which file a package is by doing «guix edit some-package» <Formbi>it will show you the file with the module <anadon>`guix show: error: python-xyz: package not found` <bavier`>anadon: "python-six" is the package name <anadon>Then why is that not the name I use to import it? <bavier`>so 'guix show python-six' will give you a location of gnu/packages/python-xyz.scm, so the guile import line would be (gnu packages python-xyz) <bavier`>imports work on guile modules, not packages <guix-vits>anadon: as in manual: `guix search 'PACKAGE_NAME'| recsel -p location` -- recsel is in recutils. <guix-vits>actually, if using search `-p name,location` is better. <anadon>Everything is working now except for their builtin tests which require some undocumented envirnmental configuration stuff. I've submitted a ticket on this, but how do I disable python-build running setuputils' tests? <guix-vits>anadon: try search in python-xyz.scm, search for `(modify-phases` -- (delete , (replase and a such. Hope that setuputils' is a standard thing. <jsoo_>anadon: I think you can also give #:tests? #f to arguments <jsoo_>you will need to edit package descriptions and pull for the effect to take place <guix-vits>jsoo_: python-iocapture is a good primer in case that some tests steel needed? <guix-vits>'check is deleted, but after 'install `(invoke "py.test" "-v" "tests")` added <anadon>I'm trying to understand the information you're giving me. What? <anadon>`(build-system python-build-system)` -> ? <guix-vits>anadon: probably (build-system python-build-system) (arguments `(#:tests? #f)) <guix-vits>at least that it's looks like in offlate package definition in python-xyz.scm <anadon>guix-vits: That was it. I have to say, being dropped into emacs with no knowledge of emacs for the edit command makes `guix edit` really, very hard to use. <anadon>Arch got democracy? That's unexpected. I don't know how that will go for them. <str1ngs>anadon guix edit probably uses EDITOR <str1ngs>export EDITOR=vim or your editor of choice <PotentialUser-16>hello all, what's the canonical approach for managing dotfiles with guix? <bavier`>PotentialUser-16: I use stow and it works for me <PotentialUser-16>@bavier thank, what about rollbacks? Maybe a stupid question but if something in a config file broke guix on startup, would it be possible to roll-back to a previous generation, incl. a rollback of the config? <lfam>If you only use the defaults, then you don't need to manage dotfiles <guix-vits>anadon: you're packaging an client to "Send, sign, and approve documents..."; why do you need a Software Development Kit (SDK) now? <lfam>Hey rekado, how can I test changes to goggles.scm (the IRC log page generator)? <anadon>It is a python module they intend external developers to use. <anadon>And I expect to be using it soon, so I'd like to get it packaged and tracked. <str1ngs>guix-vits quick feedback question. are you primarily wanting to change the background color and forground color? <guix-vits>str1ngs: rather fonts size, colors are the second. <guix-vits>anadon: but that SDK isn't a dependency of python-docusign? <anadon>guix-vits: It is python-docusign. <guix-vits>str1ngs: The main reason i'm prefer the emacs-w3m now, aside from JS needs, is a font-size. <lfam>Nvm rekado, I figured out to use goggles.scm <str1ngs>guix-vits: okay I have a per "host" implementation now for style sheets. but I think a regex implementation would be better or per domain <str1ngs>guix-vits: style sheets are probably not ideal. for example "body { color: black; background-color: #cccccc; }}" works on histo.io but not with www.gnu.org <str1ngs>guix-vits: webkit has away to change background color specifically so I might experiment with that. <jonsger>how can i remove a channel from guix pull? <guix-vits>str1ngs: about hosts, regexes and domains: i'd read that adblock-functionality of qutebrowser uses rather hosts approach due to speed concerns, if i got this right. <str1ngs>guix-vits: hosts would provide easier implementation. since I can just use an alist. regex add more complexity to the implementation. really it's how much granularity needs to be provide. I think per host is a good basis to start with. <Krafter>Hi I had an issue with guix adding things to my $INFOPATH I can't figure out where this happens. It happens regardless whether or not I load guix provided profile scripts. <Krafter>guix-vits: Hi again. I never got any further than last time. :-( <str1ngs>guix-vits: webkit_user_script_new has whitelist and blacklisting but, not sure how generic I can make that. ie qtwebeninge or a ncurses engine might not provide that kinda feature. atleasy my current implimentation is generic enought. <guix-vits>str1ngs: Is ncurses wersion will be ran in a terminal emulator? <guix-vits>anadon: ah, Github: "This open-source SDK...", i see now. <guix-vits>str1ngs: good; but how do you see a ncurses stylesheets (console is already years ahead of GUI in this regard, because same font-size, good contrast and all of that)? <str1ngs>guix-vits: in some feature might not apply well. I guess for me it's still important to consider the abstraction <str1ngs>guix-vits: can you play with the style sheets . "body { color: black; background-color: #cccccc; }" does not seem to work on gnu. maybe my css fu is not good enough,. <guix-vits>str1ngs: yes, kitty t. emulator already uses imagemagick to output pictures. Future :) <str1ngs>aww okay, I think an ncurse guile web browser is nice. but it will be one of the last implementations. but theoretically later you can do nomad --platform gtk or qt or ncurses. that's the plan anyways. <Veera>running make authenticate is failing in guix.git with unknown commit key <str1ngs>guix-vits: also C-u will open things in new buffer. and M-x: ibuffer helps with switching buffers. <str1ngs>guix-vits: there are many emacs features that are available that are not readily apparent <Veera>this key is not available in keyservers <Veera>commits are d68de95 to 9ae3e79 <str1ngs>guix-vits: also you can edit *scratch* and M-x eval-buffer. also M-: is handy as well. and you can start a repl that gieser can use. <str1ngs>guix-vits: then you can use geiser-connect-local from emacs <guix-vits>str1ngs: i'll try. actually `background-color: pink` not work for me on histo.io too (changed default). ***amfl_ is now known as amfl
<guix-vits>str1ngs: worked with * instead of body, but its wrong approach, just more rules needed. <guix-vits>str1ngs: also gnu.org works, just change pink to red. <str1ngs>guix-vits: gnu.org works for background? <str1ngs>guix-vits: I guess the main thing. does this do what you need ? <lfam>Veera: Savannah is the "canonical" source of keys for Guix <guix-vits>i fear that this even worse, as !important is definetely a bad thing, like the global-variables afaiu. <guix-vits>str1ngs: added `font-size: 48px !important;`, fonts are huge-huge. <Veera>lfam: well civodul key was available in keyservers. And I download and imported the key from savannah. It still says key missing. <mehlon>in git? I think I had that problem too <Veera>lfam: that key is someone else's <str1ngs>guix-vits: do you have a hiDPI monitor? <civodul>Veera: "make authenticate" uses its own keyring <civodul>Veera: ↑ it's in ~/.config/guix/keyrings/channels <Veera>civodul: please can you tell how? <civodul>Veera: gpg --import --no-default-keyring --keyrin ~/.config/guix/keyrings/channels/guix.kbx --recv-keys XYZ <civodul>well, either --import or --recv-keys <Veera>civodul: the key is not available in that keyring <guix-vits>str1ngs: no, just tested it -- funny: if i use M-x edit-uri and switch from www.gnu.org to any another site, then gnu's stylesheet applied anyway. Even for M-x load-uri <civodul>Veera: right, so you need to import it in that keyring with a command along the lines of the one i wrote above <Veera>civodul: from which keyserver <Veera>civodul: it is not available in pgp.mit.edu pgp.keyserver.io <civodul>Veera: the most reliable way is the URL lfam gave you above <str1ngs>guix-vits: this works off the 'committed event. meaning the final uri that is loaded. uri's can change though you might load http://gnu.org. the final redirected uri is probably https://www.gnu.org so we use the finaly uri heree <anadon>I'm trying to bootstrap a just pulled guix repo with a patch to add my "python-docusign" in python-xyz.scm, but it seems to be failing due to something missing about the PKG_PROG_PKG_CONFIG macro or variable, and says that the PKG_CHECK_MODULES thing is undefined. What might be causing the guix building bootstrap to fail here? <str1ngs>guix-vits: I miss read what you meant there. yes this seems like a bug. <str1ngs>guix-vits: webkit-user-content-manager-add-style-sheet does not work like I thought. probably I need to remove all styles before adding them <civodul>anadon: you need to make sure pkg.m4 (provided by the 'pkg-config' package) is in ACLOCAL_PATH <civodul>normally, if you do "guix environment guix", you get a working dev env <janneke>hww, wip-pipe fails with the dreaded In procedure scm_lreadr: #<unknown port>:15:133: Unknown # object: #\< <janneke>probably something silly *unspecified* or something...but where?