<sameerynho>running the same file with guile throws an expception <str1ngs>sameerynho: running the same file with guild might not give the same results. it might not find some modules <str1ngs>sameerynho: I might have cause some confusion with my guile -c suggestion <leoprikler>how do some packages build gtk-doc based documentation without gtk-doc? <sameerynho>str1ngs: nah, to be honest i read the docs and some blog posts but i still have to clear picture of the correct development process. the tracebacks looks weird to me in compare to other lisps i know and i have to setup my repl to start using the right env <str1ngs>sameerynho I'm kinda use the guile's trace backs now. sbcl though ... esh <str1ngs>always feel like sbcl is screaming at me <sameerynho>str1ngs: :))) it's similar bug i can't tell which way it is <NieDzejkob>lispmacs[work]: why would emacs be a bootstrap binary? <lispmacs[work]>NieDzejkob: It's pretty much impossible to get anything done without emacs, right? I just assumed it would be a build dependency of everything <leoprikler>I'm pretty sure you can boot Emacs without Emacs. *str1ngs hits C-g and dodges <str1ngs>inti=/run/current-system/profile/bin/emacs <valignatev>Be careful. What appears to be a meme today might as well be reality tomorrow <jonsger>does someone have a config with a working nginx server? no matter what it's exactly, I would be interested <roptat>on that last system, I use the default nginx service, and I extend it with simple-services (to add server blocks to the configuration) <roptat>cat-avatar-generator-service also extends nginx <jonsger>roptat: but the cert thing is by hand? <roptat>you'll find the certbot-service procedure used in ene and hermes <roptat>however there's a trick when you first use certbot <roptat>it needs a running nginx, but nginx won't start if the certs don't exist <roptat>so at first, you have to configure nginx to listen only on http <roptat>then have certbot create your certificate and reconfigure with a new nginx config <atw>that's right, I forgot about that <valignatev>I've finally packaged alacritty and it seems to be working. Now the only thing left is to convert this 9000 LOC dff into a series of patches and submit it. I officially name Rust ecosystem to be npm2.0. It is a horrible mess <atw>Yikes, and congratulations! <leoprikler>Are you all excited for the C++ package manager? <valignatev>I can't wait for 9 millions of C++ packages where each of them is just a 100 LOC class with empty virtual functions <valignatev>Just so, you know, you can depend on it and IMPLEMENT <leoprikler>Well, so far none is mandated by the standard, but there already exist several. <jonsger>I know only need to get docker working, run Nextcloud in a container and bringing all together :P <roptat>no, you need to make a nextcloud package and service and run it with guix :p <jonsger>I know. But I need it know, because my ubuntu server with Nextcloud died yesterday <leoprikler>if you want something more complex you'd have to look at configuration <wdkrnls>dmenu_path started working again for me after a reboot :) ***flor_ is now known as flor
<roptat>wdkrnls, have you looked at pavucontrol? <roptat>I'm thinking maybe something is muted, or maybe pulseaudio crashed (although a process that uses it should restart it) <wdkrnls>Thanks, installing pavucontrol and using correlated with me being able to get the speakers to make sound. <wdkrnls>My scripts all used amixer to monitor and control the sound system. That no longer works, though. <drakonis1>roptat: has the patch to icecat been pushed yet? <drakonis1>firefox has a directory containing default preferences that doesn't require patching to modify defaults <roptat>i'm waiting for Mark's input, but if you can create a better patch, feel free to do it! <bandali>roptat, does the patch whitelist all of /gnu/store? i don’t think that’s the best idea <roptat>well, it's mostly equivalent to the default whitelist of /lib and /usr/lib <roptat>(actually, I don't know what the default whitelist is) <bandali>could the store contain any sensitive information? <bandali>(i do realize that, like nix, it’s world-readable as of now) <drakonis1>a reminder that nix has problems with secret management <roptat>it's world-readable, so it shouldn't <drakonis1>they're dealing with that but i'm not sure on how guix is dealing with it <roptat>we can't deal with secrets either <bandali>it might be a better idea to “do it right” for this patch, and ideally restrict it only to the necessary subpaths in the store <roptat>right, but I'm really not sure what the necessary parts are <bandali>i imagine the media-related libraries? could we get the exact paths for those dependencies? <bandali>nckx and i were talking about this the other day <bandali>drakonis1, ah, yeah it’d be nice if you confirm how they do this <drakonis1>nix leans really hard on wrappers for things <drakonis1>nix the package manager is available on guix, so it should be a service away to verify that <roptat>not only media, the whitelist is also required for webgl for instance <drakonis1>speaking of which, i should upgrade it and push a patch to the repositories <bandali>what does webgl need in terms of libs? <bandali>speaking of nix and wrapping, incidentally, our icecat packages seems to have a wrap-program phase too <roptat>wrap-program simply adds environment variables before calling the real thing <drakonis1>that's what they use for a lot of applications there <roptat>time for me to go to bed :) see you later! <bandali>so i guess we now wait to hear from mark <bandali>i haven’t heard from him in a couple weeks <Moondhum>Hello, while installing a package does guix compile it locally like gentoo or does it come compiled? <PurpleSym>Moondhum: By default is uses “substitutes”, i.e. binaries that come from a build farm if available. <Moondhum>I ran "guix system reconfigure .." and "build phase" has been running for a while. Good to know I wont have to go through this every time. <str1ngs>Moondhum did you recently do guix pull? <str1ngs>also by your use of the term 'install' I'm wondering if you might want install packages to your user profile instead. <slyfox>I see guix/progress.scm had a few fixed in guix-master around it. Should new release be cut? <efraim>disabling tests on libnslog so netsurf built didn't make netsurf work, it looks like there's an actual bug :/ <civodul>g_bor[m]: nope, but close to it: it was in the http-multiple-get changes ***emyles` is now known as myles
***myles is now known as emyles
<sneek>Welcome back peanutbutterandc, you have 2 messages. <sneek>peanutbutterandc, nckx says: Indeed I was mistaken: it was a Yeelong machine but was stolen in Argentina ??? He now uses a Librebooted TP like a pleb. <sneek>peanutbutterandc, str1ngs says: . as far as I can see tuxguitar is missing some plugins. probably minimum tuxguitar-alsa. these need to be added to the build. will try to figure out more when I have time. <peanutbutterandc>nckx, TPs seem to be the most popular thing among those wanting all-free system. Perhaps this will change once RISC-V laptops hit the market? Risc-v pinebooks, perhaps? <peanutbutterandc>str1ngs, I see. I look forward to being able to use only guix-installed tuxguitar on my system <str1ngs>peanutbutterandc I will try to look at when I have more time. <peanutbutterandc>So, I have encountered this strange issue: `guix upgrade` has been stuck (even after another `guix pull`) on "building /gnu/store/.....-mercurial....drv" at the 'check' phase for the longest time. Strange. <str1ngs>peanutbutterandc: I did get Tuxguitar-alas plugin to build. but then there was a jni issue <str1ngs>on the bright side, I got jack and fluidsynth working finally <str1ngs>so my midi keyboard can play a gande piano <str1ngs>peanutbutterandc: pianobooster is next on my list to create a package for. I've been pretty busy with qtwebengine and qutebrowser. though I think that's finally done. *kmicu now needs to sub to all guix mls cuz news.gmane.org is down and news.gmane.io is not acceptable. Never trust a centralized projects, yey ヽ(*＾▽＾)／ <str1ngs>peanutbutterandc: actually guitar is my main instrument <FriendFX>Hey experts, I just tried to start my PC from a USB flash drive with GuixSD and was expecting an option to start a live system to try before installing, but couldn't find anything obvious... is that expected? <str1ngs>FriendFX: that is expected there is not GUI installer at this time. so a liveCD does not make sense <str1ngs>unless the installer has changed, if so someone correct me :P <peanutbutterandc>str1ngs, I see. I'm so glad that someone else besides me likes the project. Perhaps you should write to the original developer, too. The project is on my list of almost-dead-projects-that-I-will-someday-try-to-revive-if-I-become-competent-enough. I see. Then I don't need to worry about finding digital guitar amps in the gnu distribution. :) <str1ngs>peanutbutterandc: which project pianobooster? <peanutbutterandc>str1ngs, You do not happen to have a YT channel or something where you play guitar, do you? Yes, sir. Pianobooster. I did find two forks in GitHub but none from the original dev. <str1ngs>peanutbutterandc: I don't have a YT channel, can't say I'm that good haha <peanutbutterandc>I think pianobooster is a great thing. Synthesia/Linthesia are like the junk food. But pianobooster is proper music educational game, I think. One actually can practice sight-reading. <str1ngs>peanutbutterandc: actually pianobooster was one of the reason I got a midi keyboard to begin with. since it helps with musically theory. with guitar it's pretty easy to bypass musical theory. <str1ngs>peanutbutterandc it's late don't mind all the typo's <FriendFX>kmicu: Thanks for the link, but wouldn't that defeat the purpose of testing it on my real hardware? I'm mainly interested about support for WiFi cards etc at this point. <peanutbutterandc>str1ngs, Absolutely not. Typos are fine. I am glad that you like pianobooster. This repo: https://github.com/captnfab/PianoBooster seems to be the closest to getting pianobooster up again. Perhaps you might consider this instead of the sourceforge one. But I am not sure. Since I don't have a MIDI keyboard now, I haven't tried it yet. <str1ngs>peanutbutterandc: I will look at that when I have more time. <peanutbutterandc>str1ngs, Of course. I wish you all the best. Thank you for your hard work. I, for one, will surely benefit from it. <kmicu>FriendFX: if that’s the goal then indeed, vm is not useful. Regarding WiFi upstream Guix System (and generally libre firmware) only supports ar9271 and ~2012 broadcom b43. <kmicu>(and generally if you have a modern wifi hardware then that’s not supported for sure) <FriendFX>kmicu: Thanks for the info. My Ubuntu system reports I have a BCM4322, currently using a proprietary driver. I guess the best bet is to install GuixSD on another flash drive on the same machine to test... <peanutbutterandc>Since I don't use mercurial, really, and was just trying to `gource` a repo that I liked; I'm going to retry by uninstalling mercurial. *kmicu in old laptop has BCM4311 802.11b/g WLAN (rev 01) which is supported. Lucky me. <peanutbutterandc>kmicu, If one can use wifi without any additional drivers on a machine using debian (official distro), does that mean wifi would work for guix too? <kmicu>I don’t know what Debian installs by default (e.g. on Live ISO) so cannot answer for sure. <NieDzejkob>FriendFX: you can test the wifi when you boot the installer <peanutbutterandc>I see. Debian don't user non-free software in their default official distribution. So I was curious if wifi-cards compatible with debian would work with guixsd <kmicu>That could be the case but someone needs to verify Debian’s LiveCD and its defaults to be sure. <peanutbutterandc>I just removed subversion and tried again and it is still trying to build /gnu/store/....mercurial... I think I will have to go for `guix package --upgrade --do-not-upgrade mercurial`...... and it's still trying to build /gnu/store/...mercurial...drv. Curioser and curioser. <roptat>peanutbutterandc: hg is probably a dependency, not a package in your profile <roptat>do-not-upgrade only prevents upgrade of a package in your profile <roptat>It has no effect if the package is not in your profile <jonsger>I can make it for the FOSDEM and the Guix days :) <peanutbutterandc>roptat, I see. That does make sense. Do you have any idea how I could figure out the culprit and remove it? I know guix has guix graph but I am not that well versed in it <roptat>It coull be a package dependency or a source dependency <roptat>Maybe guix graph -t bag your-package | grep hg <roptat>If there's an output, the package depends on mercurial ***modula is now known as defaultxr
<roptat>Or another solution could be to run the upgrade, when mercurial is being built, killall gcc or make, then you'll get a backtrace and guix will tell you what failed, ending with the most top-level package that depended on that build <roptat>Or another solution could be to run the upgrade, when mercurial is being built, killall gcc or make, then you'll get a backtrace and guix will tell you what failed, ending with the most top-level package that depended on that build <roptat>Or another solution could be to run the upgrade, when mercurial is being built, killall gcc or make, then you'll get a backtrace and guix will tell you what failed, ending with the most top-level package that depended on that build <roptat>That was my last essage in case you didn't receive it <peanutbutterandc>What is the command that starts with a 'r' that we pipe from guix package -I to? I forgot it <peanutbutterandc>I am trying to do the first thing `guix graph -t bag pkg | grp hg` but with a for loop <peanutbutterandc>how do I install recsel with guix, again? `guix search/show recsel` didnt give me anything <peanutbutterandc>No, sir, I wouldn't think so. I am on a foreign distro and `apt show coreutils` does not list it <peanutbutterandc>roptat, I just tried with for i in `guix package -I | cut -f 1`;...` and turns out everything from emacs to python to tinyscheme has hg as a dependency.... :[ <peanutbutterandc>roptat, You'll never guess. Of the 61 packages in my profile, 60 have hg as a dependency. <NieDzejkob>hg is short enough that it's likely to match some other stuff <alextee[m]>i think the icecat package is missing a font dependency <alextee[m]>NieDzejkob: i still don't get why there's no hook for that after installing it <digash`>peanutbutterandc: I've ran into the same mercurial issue, let me know if find a solution <Moondhum>I have been uncessfully trying to setup config.scm with wayland and sway. Any tips guys? <puoxond>sneek: later tell peanutbutterandc: Maybe some package in your profile (or one of its dependencies) uses hg-fetch to download its source. <pkill9>Moondhum: are you able to start sway from the console? <NieDzejkob>Moondhum: what is the problem? Like, what errors are you getting? ***calher is now known as KE0VVT
***ng0_ is now known as ng0
<peanutbutterandc>digash, I see. That's good news, perhaps. That means it is not something I did to mess up my system. :) I will. <sneek>Welcome back peanutbutterandc, you have 1 message. <sneek>peanutbutterandc, puoxond says: Maybe some package in your profile (or one of its dependencies) uses hg-fetch to download its source. <peanutbutterandc>puoxond, Hmm... so I would probably have to search for hg-fetch as a dependency? Or is it in the build system itself? <puoxond>peanutbutterandc: I used the guix repl to look for packages where the origin-method of the packages-source is equal to hg-fetch. <peanutbutterandc>puoxond, I see. It is possible that one of the packages I have has one of those packages as a dependency, then. Guix repl, hun? Super cool. I still haven't gotten around to that section of the reference manual yet <digash`>puoxond: very cool, what is the guile code that you've had to run to get that? I had to brute force it with grep -ire :) <puoxond>peanutbutterandc: Unfortunately the programming interface is not yet fully documented. I had to look at the source code to figure out how e.g. fold-packages works. <gnutec>when I did "guix environment --conteiner --ad-hoc gcc-toolchain" the system add the mes project. Is it because the mes is the new gcc? <cnmne>hello! I access my guix box on my home network through ssh, and I'm using a program that hosts a webpage on localhost:8080, but I can't access the page unless I'm on the guix machine itself. I'm guessing that I need to allow that port somehow. I've looked at the Web Services section of the Reference Manual and all the options seem like hosting an entire server (which is already provided by my program). The Networking Services section <cnmne>has an iptables-service-type for filtering packets. Is that what I'm looking for ? <digash`>puoxond: very cool, thank you, just map package-name, pp and i've got the list. package-fold is very usefull. <alextee[m]>i don't see operating system options in virt-manager. does it work for anyone else? <alextee[m]>like where you're supposed to specify the OS of the image you're installing <bavier>alextee[m]: it wasn't working properly for me yesterday either <bavier>i'm not familiar with the detection mechanism it uses. <str1ngs>cnmne is your http program binding to the nic interface as well? <drakonis1>isnt there a xfce post on foreign system integration? <str1ngs>cnmne port forwarding if you need it at all is usually done at the gateway/router <str1ngs>alextee[m] I noticed that too. I have not run into any issues using generic. though I only used linux VM's myself <alextee[m]>str1ngs: should i send a bug report or do u know if this is a known issue? <puoxond>peanutbutterandc: On a different note: You could try finding the affected package using this: https://paste.debian.net/1126135/ , where `my-packages' is a list of the names of the packages in your profile. <peanutbutterandc>puoxond, Whoa. Super cool. Is there some way I could give it the list of packages in my profile with an s-expression? `guix packages -I` like thingy? (I am not a scheme/guile wizard yet) <str1ngs>drakonis1: ahh depends which hooks you mean build hooks? I know profile are in guix/profile.scm <drakonis1>there's a few things missing hooks right now <drakonis1>gnome doesnt refresh its application menu when the profile is modified <puoxond>peanutbutterandc: It's supposed to be called from guix repl (interactively). <str1ngs>alextee[m]: I'm not sure if it's a bug. <str1ngs>drakonis1: you need to start your X11 sessions on fortunately <drakonis1>that is, welp, i need to keep it in a single line <str1ngs>you can't really compare the two. fedora uses FHS which make things easier. because guix is atomic etc. there are technically hurdles to over come. <peanutbutterandc>puoxond, You break my heart. :D I was not in the mood to type that much. :[ <str1ngs>though for application menu it's possible it should update. so maybe someone else can confirm this expected behaviour <str1ngs>if anyone uses qutebrowser if you could test the switch to qtwebengine it would be appreciated thank you. I don't use qutebrowser myself. <bavier>str1ngs: you committed the switch already? <str1ngs>bavier: Pierre handled the commit with review. I just update the package and the depends. <bavier>str1ngs: thanks to both of you. one less thing for me to do ;) <peanutbutterandc>puoxond, *Ahem* but really, is there not a function call to return all the packages in my profile? I have 61 packages here. <str1ngs>bavier: I'm using qutebrowers as a reference for any bugs in qtwebengine. It's the package I can think of. <str1ngs>bavier: no problem. next and nomad's qtwebengine support is still a ways off. qutebrowser seems quite stable in this regards. <puoxond>peanutbutterandc: I believe it's possible but I'm still figuring it out. <str1ngs>bavier: one other thing in regards to qtwebengine. is I'm assuming qt will be bumped to 5.14 so I should probably track that somehow. <NieDzejkob>peanutbutterandc: (map (lambda (entry) (string-append (manifest-entry-name entry) "@" (manifest-entry-version entry))) (manifest-entries (profile-manifest %current-profile))) <peanutbutterandc>NieDzejkob, *a lisp wizard appears with a magic yin-yang of eval-apply* Thank you! :) <NieDzejkob>though note you will need to import some modules: (use-modules (gnu) (guix profiles)) <leoprikler>as if either of them were not already turing complete <peanutbutterandc>Oh that was a reference to the image at the beginning of MIT lisp lectures on the Internet. <cnmne>str1ngs: I'm sorry but I'm not sure how to tell if the program is binding to the nic interface. I'm using the guile program haunt and `haunt serve' to statically generate my site and serve locally. my router settings are definitely out of the way. I can access the site on the machine that's hosting, but not from any other machine on the network. Is there a better place to ask this? I'm thinking now it might not be guix relevant... Also <cnmne>to butt into your other thread, I use qutebrowser a lot; I'll start using it on guix now! I have to head to work soon, but I'll be on later! <str1ngs>peanutbutterandc: (load "/path/to/script.scm") <str1ngs>cnmne: my guess you need to tell haunt serve to bind to your network nic as well. not just the local loop back 127.0.0.1 <cnmne>str1ngs: that makes sense! thanks, the documentation and features seem kind of sparse. so maybe I'll have to play around with it <str1ngs>peanutbutterandc: does adding (guix git-download) help? <str1ngs>though if this hg.. that's an odd error ***drakonis1 is now known as drakonis
<str1ngs>peanutbutterandc from what I gleam you are trying to find what packages in your manifest use hg for fetch <peanutbutterandc>str1ngs, Yes, sir. And it's all puoxond's work. I am not that well-versed in the ways of guile, yet. :) <str1ngs>peanutbutterandc a better approach is to add #:tests? #f to mercurial build arguments. then use ./pre-inst-env guix package -u <str1ngs>peanutbutterandc: better yet, we need to figure out why mercurial hands on check phase. <peanutbutterandc>str1ngs, Umm.... I am trying to upgrade guix using `guix upgrade` after a `guix pull`. So it's the gnu distribution. I am in no way qualified to poke around with the gnu distribution yet, I think. The only thing I do poke around with are my own personal package definitions in my channel... <lispmacs[work]>which command is used to roll back to a previous Guix generation (not package generation) <str1ngs>peanutbutterandc: strangly I can replicate this with guix build tuxguitar is that in your manifest? <str1ngs>I bet if you remove it that resolves this. <str1ngs>well I was going to look at tuxguitar I guess I'll look at mecurial instead lol <str1ngs>peanutbutterandc: it might not fix it. but it some to coincidental we both discovered this issue. <peanutbutterandc>lispmacs[work], You're most welcome! (Can't believe I could just help out someone whose name starts with lisp. Yeah!) :D :) <peanutbutterandc>str1ngs, There was another person who had this same issue. He has asked me to tell him if a solution is found <str1ngs>I would have said use a hammer, it fixes everything once. haha <str1ngs>ummm how do I list the dependancy graph for a package? <str1ngs>drakonis: quite literally, thank you <peanutbutterandc>I wonder if `guix graph --backend=graphviz | graphviz` is what I used once... but I don't know <str1ngs>awww why can't I just have a tree view. this is to fancy for me! <peanutbutterandc>lispmacs[work], I have seen it too. But it isn't documented in the reference manual, I don't think. What does it do again? <lispmacs[work]>peanutbutterandc: It is document in the info manual. Allows you to run any guix command from and commit of guix, without changing your profile guix version <peanutbutterandc>str1ngs, From the manual: `guix graph package | dot -Tpdf > dag.pdf` for an actual graph <lispmacs[work]>peanutbutterandc: e.g., build older stuff and new stuff than your current profile guix <str1ngs>peanutbutterandc: I figured that out thanks <peanutbutterandc>lispmacs[work], I see. It isn't in the reference manual pdf. I should probably start getting comfortable with info pages as well as pdfs... <lispmacs[work]>peanutbutterandc: in emacs that is "C-h i" to bring up the Info page directory, then "C-m Guix" for the guix info node <peanutbutterandc>lispmacs[work], I see. Thank you very much. I am also installing info-reader, currently. <peanutbutterandc>Okay... `info guix` says "no item guix in node (dir) Top" whereas I can get the info of a guix installed package. Any ideas, please? <alextee>can we put screen recording back into gnome :( <peanutbutterandc>lispmacs[work], Um... where in the info manual is the bit about time-machine, please? <lispmacs[work]>peanutbutterandc: fastest trick is to type "C-s time-machine" then enter <lispmacs[work]>peanutbutterandc: you can search through entire guix manual for anything with C-s command <peanutbutterandc>lispmacs[work], "C-s time-machine" didn't quite work. I have emacs-guix manual, it seems, rather than guix manual.... *confused* <lispmacs[work]>peanutbutterandc: emacs-guix manual is different from the Guix manual <lispmacs[work]>emacs-guix manual explains how to use the emacs extension for controlling guix <peanutbutterandc>Yes, sir. How would one get to guix-manual? Also, why is `info guix` from command line not giving me the guix manual? <peanutbutterandc>whereas `info gtypist` (gtypist is in my guix profile) gives me the info for gtypist <lispmacs[work]>peanutbutterandc: info-guix command worked for me. Are you running a Guix System or on top of a distro <asterope>honestly, is disabling mingetty on tty1 safe? It's bugging me every time my machine boots up <lispmacs[work]>peanutbutterandc: hmm, yes, there might be an issue with info page paths then, I'll see if there is a manual section on that <leoprikler>asterope: what about it bothers you? the gratuitous login prompt? <lispmacs[work]>peanutbutterandc: in guix Binary installation section, it mentions some instructions for making the guix info manual available to the regular info command. With the PDF, can you look at section Installation > Binary Installation > item 6 <peanutbutterandc>lispmacs[work], Thank you very much. I will look at it right away. (I was under the impression the installer-script did it for me automatically. But I must have mistaken the official installer for my-custom installer) <sameerynho>hey folks, where is the `modules` function definition ? I can see that some packages used it but when i use it I get `unbound variable: modules` <str1ngs>peanutbutterandc: seems mecurial was update on the 7th which is causing this problem. if you were to do guix pull rollbacl or guix pull --commit=f22b35711323a1e101ed666ec5ce776bf1908a74 should fix this. but obviously that does not upgrade your packages <nckx>asterope: Works fine here. I moved slim to tty1 and disabled all getties. <alextee>i think something is broken with gnome's screen recording <alextee>it's supposed to work out of the box <str1ngs>peanutbutterandc: some changes here made to the tests. so either the tests need fixing or go back to disabling tests. not sure yet <alextee>maybe it's writing something in some log, but i don' tknow how to check <nckx>alextee: System logs are in /var/log but that doesn't mean that everything is logged there. <peanutbutterandc>lispmacs[work], I see it now. Yes. It does make sense. I didn't have info installed during the time of installation so it didn't install the info pages (the official installer script). <peanutbutterandc>However, now that I have guix but do not have the info page, is there a guix package for the guix-manual somewhere? <peanutbutterandc>str1ngs, I see... Perhaps I'll just have to wait till a change is made to the mercurial package definition? <str1ngs>though you might need to install the guix package to get the manual. <str1ngs>which might not to great of a solution. <peanutbutterandc>I just learned that the info pages are more up-to-date than the pdf reference manual <str1ngs>peanutbutterandc: maybe it would be better to add $HOME/.config/guix/current/share/info/ to INFOPATH <alextee>im pretty sure something is wrong with the package because i literally just reinstalled a fresh guix system <str1ngs>peanutbutterandc: adding guix to your profile can have side effects <nckx>alextee: I don't know where (if anywhere) GNOME logs things on Guix. <reepca>so I vaguely recall a time when a failed build would be reported as a cascade of failed builds (can't build X because can't build Y because can't build Z...), and mercurial is failing to build when I run 'guix package -u "" --do-not-upgrade="mercurial"', and I'm not sure why it's trying to build it. <str1ngs>reepca: it was upgraded on the 7th and is hanging on tests <reepca>even when I run it with --verbose it doesn't tell me which derivation has it as an input <str1ngs>so there are probably no substitutes for it. since it probably hangs on the build farm as well <str1ngs>I guess the tests need to be disabled or fixed. <peanutbutterandc>str1ngs, I see. That makes sense: both the env path to include the guix-profile from ~/.config and not installing guix in one's profile. I just tried it and it works, too! <str1ngs>reepca: to fix this I would roll back or pull commit f22b35711323a1e101ed666ec5ce776bf1908a74 <nckx>g_bor[m]: I see Guix² was updated! Does that mean you managed to find the bug? <str1ngs>peanutbutterandc: you could install guix. but if your PATH is wrong it could use the wrong guix. which is probablematic <str1ngs>peanutbutterandc: I personally have guix installed for emacs-guix to work. though maybe there is a better way. not sure. <reepca>str1ngs: I'd prefer to fix it by figuring out what needs it and just adding that to the --do-not-upgrade argument. But I'm not sure how to easily figure out what needs it :-/ <leoprikler>alextee: I think this might be related to gstreamer in some way <peanutbutterandc>str1ngs, I see. I just realized that this means the patch I sent to guix needs to be updated to include the special case for INFOPATH on foreign distros. <str1ngs>reepca: as far as I have gleamed something in tuxguitar's dep graph needs it. but there are probably many more and growing as mecurial can not be built as it stands. <str1ngs>peanutbutterandc: oh for guix-install.sh ? <peanutbutterandc>str1ngs, Yes, sir! `[ -e "$INFOPATH" ] && export INFOPATH=$INFOPATH:$_GUIX_PROFILE/share/info` <nckx>str1ngs: Mercurial broken?? <str1ngs>nckx: I'm surprised this has not been discovered yet <nckx>It's known for having a long suite. <reepca>so I just tested it and when I run "guix build X" and Y is built as an input of X, but fails, I do get a "can't build X: 1 dependencies couldn't be built" message. But when I run "guix package -u ..." I get no such message. <alextee>so is there any other similar screencasting software? lol. im trying a couple but it's a whole science to make them record a simple video <str1ngs>nckx: but maybe a further commit has effected mercurial tests to hang. possibly a change in an input <leoprikler>If you feel comfortable with a little experiment, you could try reverting gstreamer to 1.16.1 and check whether screencast works <nckx>~/guix/pre-inst-env guix build --no-grafts mercurial → /gnu/store/lbpgss5vgwlayh7lk8af1rr0bxm6gsk5-mercurial-5.2.1 <alextee>leoprikler, will that require rebuilding other things? <str1ngs>reepca: I have given you a fix. the method you are trying will just get worst as you do subsequent pulls <alextee>peanutbutterandc, thanks, trying now <nckx>It was already built. This is Guix a650ee0baa6dc257de428524239d456402a59ae0. So we have a non-deterministic test failure that I can't reproduce. <str1ngs>nckx: I'm assuming you are close to tip? <alextee>leoprikler, yeah.....i'll just wait until someone fixes it <nckx>str1ngs: I'm 15 commits ahead of master, nothing relevant to mercurial in any of them. <str1ngs>I will git bisect. maybe that will indicate the efffecting hash <nckx>peanutbutterandc, str1ngs, anyone: Which mercurial derivation (.drv) fails for you? <str1ngs>nckx: could be a a non deterministic test issue then <nckx>Exactly. I'm building again with --check. <str1ngs>nckx: f22b35711323a1e101ed666ec5ce776bf1908a74 does not fail let me check what my tip is <nckx>But this is on spinning drives. Might take a while. <str1ngs>nckx: my tip is 44ce3ebb46a4e63a892ef1aac2c7b5fa3dec28ce <nckx>str1ngs: The commit I pasted above is my ‘tip’ if I understand you correctly. <nckx>I have commits on top of it but they don't change mercurial. <reepca>str1ngs: I don't see how that's the case. It would just be upgrading only the packages that aren't affected by the build failure while I wait for it to be resolved. <str1ngs>nckx: it could be a input of mercurial though <nckx>str1ngs: That's why I want the .drv. <leoprikler>0ed75a5f068541339bc1db6271796a3d382bb681 and 6d15516dc6ffca3622ff49179c5c7c360c11b54b <str1ngs>nckx: please describe how to get the drv <nckx>str1ngs: guix build --dry-run mercurial <str1ngs>I'm using git proper not guix to bisect <reepca>nckx: /gnu/store/m947zzg7k1f0wn66p629vwsh8p08iizb-mercurial-5.2.1.drv is what I have here <str1ngs>nckx: /gnu/store/m947zzg7k1f0wn66p629vwsh8p08iizb-mercurial-5.2.1.drv <alextee>leoprikler, thanks, but i have other work to do, don't want to spend too much time. simplescreenrecorded worked for now..! <nckx>peanutbutterandc: OK, so you can see that the first element is lbp…-mercurial…, which is what I also pasted above. So we're trying to build the *exact* same Mercurial, and it worked for me and fails for you. That's what I wanted to know. <nckx>So we do have a non-deterministic test failure on our hands. <str1ngs>I think it's only logical to consider 10576acbbf496a051d488c2832f1e474ef6074f3 to be the culprit since it effects tests. which were disable before <peanutbutterandc>nckx, I am trying to do `guix build mercurial` right now to see if I have any luck. It fails for me when I try to `guix upgrade`. It doesn't even fail, really. Just times out (as seen in my previous paste) <nckx>str1ngs: Yes, of course. But that doesn't tell us why. <str1ngs>nckx: what is are your --cores --max-jobs. maybe I can replicate a successful build. then we can determine its a race condition. an small note it hangs so not exactly a failure. more a race condition <str1ngs>worst case if it is a race condition. some magic start alignment as caused your build to not hang <str1ngs>in which case I propose just disable the tests :P <nckx>str1ngs: 4 cores, 3 jobs, spinning RAID5 (so slow compared to fast CPU). <nckx>str1ngs: Thanks for the summary, there's a lot of backlog to read. <nckx>I don't see a link to a build log anywhere, does it print anything useful before timing out? <str1ngs>I guess I should set a timeout what's a good timeout for this? <nckx>peanutbutterandc: Thanks. <peanutbutterandc>You're most welcome. Also, my `guix build mercurial` from 8 minutes ago is still stuck at 'check' phase "running 879 tests using 4 parallel processes". So perhaps it is failing for me, again, after all <nckx>str1ngs: As painful as it is with a test suite this slow, I think it's more useful to test with -c1. So I've restarted my build. <str1ngs>nckx: the thing its not CPU churning either <nckx>peanutbutterandc: ‘Stuck’ means something specific that 8 minutes are too short to determine. Is your machine actually idle (CPU, I/O)? Are no new processes being spawned and can you strace the ‘hanging’ test to see if it's looping or hanging on (say) a select()? <str1ngs>nckx: and I don't think these test honor --cores <nckx>str1ngs: Heh, I was just pasting: <nckx>running 879 tests using 4 parallel processes <str1ngs>could be this (invoke "make" "check") should take core arguements? <str1ngs>I don't think guix relies on MAKEFLAGS? <nckx>str1ngs: -j1 applies to make itself, it's not applicable here. <str1ngs>neither (setenv "MAKEFLAGS" "-j1") or (invoke "make" "check" "-j1") effect the parallel tests <peanutbutterandc>nckx, Sorry. No, it isn't. `top` shows that the top 4 are usually guixbuilder users running 'hg'. I just meant that it's doing just 'check' and nothing else besides that. It was never idle. Even this morning, when I'd forgotten about the running `guix upgrade`, a look at the system-monitor to see how my own script was doing made me realize that guix was still running <str1ngs>unless the pythons tests run there own parallelism? <efraim>can I run guix graph to graph the path between two packages? *nckx is running 879 tests using 1 parallel processes, let's wait a week and see which (if any) hangs. <str1ngs>nckx: you were not clear on that point at all <nckx>‘it's not applicable here’ is quite clear. <sneek>Welcome back guixy, you have 1 message. ***janneke_ is now known as janneke
<str1ngs>sameerynho: add #:modules ((guix build utils)) to arguments instead <bavier>sameerynho: and 'modules' -> 'use-modules', line 32 <str1ngs>sameerynho: the phases run in a special context/environment. I don't understand the details precisely so you need that argument <str1ngs>is suspect it's because phases are run by the daemon <str1ngs>sameerynho: I can't say I understand it completely quite yet myself <sameerynho>str1ngs: guile is weird ( at least i feel this way for now ) <str1ngs>sameerynho: source is a record getter for a package recored <nckx>sameerynho: ‘source’ is a field of the ‘package’ record type. The value of source is itself a record type (origin) which has its own fields, one of which is ‘modules’. <str1ngs>sameerynho which is a origin record. see guix/packages.scm <nckx>‘arguments’ is another field of the ‘package’ record type, but its value is a list of (#:keyword value [#:keyword value…]) pairs instead of a record. That's where the difference in notation comes from. <nckx>sameerynho: Lots of this is implemented using ‘macros’ but that's not the reason for this particular difference. <str1ngs>you can see define-record-type* for reference <str1ngs>technically speaking source is a constructor <sameerynho>nckx: i though source is macro with modules defined in its context <str1ngs>sameerynho: source constructs an <origin> records which also has a modules constructor see <orgiin> in packages.scm <str1ngs>sameerynho: bunch of typos sorry, I hope you got the general ideal though. <str1ngs>hello nckx please consider this patch http://paste.debian.net/1126152 . this builds for me with --cores 12. the comment message is WIP. This is based on you input re ./run-tests.py flags. I'm rerunning with /pre-inst-env 'guix build --no-offload --no-grafts --rounds=2 --check mercurial . encase my build was just lucky. <str1ngs>I think my rounds flags should be okay for this <nckx>str1ngs: Thanks. The point of this patch is only to respect p-j-c, correct? <str1ngs>nckx: yes, but also -v helps with determining if the tests have hung or not <str1ngs>which I suspect people are assuming is the case. and then cancel <str1ngs>the tests are more intensive then people are assuming I think. <str1ngs>but. I could have got lucky . maybe if you can confirm my rounds flags. I'll let that run it's course. <str1ngs>also IIRC the build server times out on lack of output? in which case -v could help with that as well. maybe I'm off base on that ? <str1ngs>also we could use TESTFLAGS to test flags. but I kinda like passing the flags manual its more explicit <nckx>str1ngs: Yes, -v would help if there's and timeout and it's due to silence (we have a knob specifically for that, I forget what it's called, something like max-silent-time). <nckx>I'll see what -v produces and add it if it seems useful (probably). <str1ngs>yes, in regards to timeout I heard something to that effect. but did not know the details <nckx>str1ngs: So we've gone from ‘the test suite hangs’ to ‘well it timed out and we just assumed it was hung’? :-/ <nckx>(By ‘we’ I don't mean you.) <str1ngs>nckx: that hung is large, did you remove tests? <nckx>str1ngs: I'm waiting to push that patch until my non-parallel test suite finishes. I'm still not sure if it will include #:tests? #f. <str1ngs>nckx: right the is my assessment, and yes I did assume it had hung :P <nckx>str1ngs: No, that's indentation due to w-d-e at the top. <str1ngs>kk I just want to confirm the hunk, for me all test pass. but lots tests hung "test daemon" some network maybe <nckx>I'm too neurotic to look at (string-append "tests/" …) … (chdir "tests") and not clean it up 😛 <str1ngs>your patch looks good to me, minus the verbose flag. which I think is useful here. <str1ngs>unless you want to bypass the whole issues for now. I'm guess more packages get pushed that use hg fetch this becomes quite the blocker. <efraim>after trying presentty on debian and on guix I can safely say our copy is better <nckx>str1ngs: No, my plan was (now that you've made it clear that it was unlikely that the tests were really hung, thank you!) to push with tests enabled but -v added, and to see whether that's enough. <nckx>str1ngs: And just drop #:tests?. Keep the (when …), it's moar korrect. <nckx>str1ngs: Does this currently break many packages, or a VIP? If it's really bad I won't wait for the 1-job test to finish. <str1ngs>nckx: I think this is a compounded breaked based on how many packages uses hg fetch. ummm forget the term for hg's downloader <nckx>str1ngs: Also, do you want a Co-authored-by: for the "-v" bit? Both yes & no feel a bit weird to me so you decide 🙂 <str1ngs>nckx: oh no co-author is not important it's a trivial change <nckx>str1ngs: Exactly, and guix refresh -l misses those & I don't know how to check that (& no time to write the Scheme code that could probably tell). <nckx>As you can probably tell by my grammar I'm busy. <efraim>python-build-system doesn't support cross building <str1ngs>nckx: #:tests #f is a safer push. since it was off this commit anyways. as long as there is follow up and it's not forgotten. <nckx>I'm not going to protest someone advising me to do less work. #f it is. <nckx>No, wait, I want the build farm to test this. #t it is. <str1ngs>I think so, it resolves the issue. and I think with the compound natures this could become a really blocer. <nckx>Sorry for subjecting you to my inner monologue. <str1ngs>nice this is just coming to head a week after this commit? <str1ngs>it kinda makes sense. nice to catch it now atleast. <str1ngs>I'll run these rounds so far no race conditions <nckx>str1ngs: I've pushed the enhancements. I've been unable to catch any failures either. Now CI can try. <nckx>Thanks for thinking along. <puoxond>It doesn't matter that mercurial is broken if I've already got the source (or there's a substitute), does it? <puoxond>I'm asking because when I run `guix build emacs-openwith --check --no-grafts' it works, but `guix build emacs-openwith --check' starts building mercurial. <brettgilio>civodul: thanks for being so helpful on the formal methods proposal. <nckx>puoxond: Correct on both counts. *nckx puzzled, which client has that. <nckx>brettgilio: Oh, do you have a home? <nckx>formal-method-peeps.guix.community <puoxond>nckx: Why is mercurial needed for grafting in this case? <nckx>puoxond: TBH I can't say for sure off the top of my head. <nckx>‘Grafting == evil dark magic’ is the general answer. *nckx could guess, but then so can you. <brettgilio>nckx: bandali and I are proposing fm.guix.gnu.org *bandali is working on that <brettgilio>I've been trying to get a new job honestly. Which is taking up a lot of time. <brettgilio>And I've been trying to work on the OCaml compiler upstream too. <str1ngs>nckx: I've run aboud 3 rounds with 12 cores. and have not had it fail yet. <nckx>str1ngs: What the hell kind of fast box you got bud. <nckx>brettgilio: fm just means ‘radio station’ to me but sounds good 🙂 If we ever start a podcast guix.fm is still available. <str1ngs>nckx: AMD Ryzen 5 3600 6-Core Processor. though I ear marked some $ for a 3950x lol <nckx>Well, wildcardjacked but available. <nckx>str1ngs: Ah. That explains it. Nice. *nckx is still on round one, albeit with other processes running. <brettgilio>nckx: civodul proposed to bandali the expanded name form. But we found it inconsistent with things like hpc. <bandali>yeah.. i just think formal-methods.guix.gnu.org is a bit of mouthful to say and type <nckx>I would have proposed formal.guix.gnu.org (which applies it to Guix, not ‘just’ the field) but you boys just go paint your pretty shed whichever way you like. <nckx>I mean, the goal is to apply these principles to Guix itself, right? Not just ‘use Guix as a tool in FM’ (which is also valuable). <brettgilio>Approaching the topic from both directions, rather. <nckx>As a non-(computer/exact-)scientist I'd never heard ‘FM’ before. <brettgilio>Formal Methods is a superset of things including type theory. <nckx>We exchange sulty glances across the room but ‘familiar’ might be a bit strong. <nckx>(I realise that's not what you're actually suggesting. :-) ) <civodul>brettgilio, nckx: i like "formal" as well <nckx>NieDzejkob: Why specifically the FSDG? I was expecting some obvious issue but don't see one. <nckx>Setting aside the GUI for now. <NieDzejkob>nckx: it interacts with a server that I couldn't find the source for <nckx>That's actually FSDG-fine (civodul?). <brettgilio>NieDzejkob: technically servers can't be proprietary. The issue of servers is a separate issue. <jonsger>meh, I ran again in the rate limits of lets encrypt :( <brettgilio>So a server being a separate issue should not conflict with any FSDG requirements <nckx>Well, NieDzejkob obviously means ’server software’ but you're both right. <nckx>Obvious example: youtube-dl. <nckx>NieDzejkob: As much as you or I (despite being on Keybase) or rms might dislike such ‘disservices’, they are a separate fight from GNU's. If the client is completely free, it's OK to package. *nckx just registered to avoid squatters TBH. <drakonis>is it like when the fire nation attacked? <nckx>drakonis: More of a reference to ‘At night, the ice weasels come.’ but you've got the gist. <nckx>drakonis: …I believed you 😃 <civodul>tomorrow is a day to celebrate Guile 3! <ng0>ah, so it's definitely out <ng0>was confused to see it on the ftp but no announcement *ng0 continue creating new package <drakonis>guix isn't defaulting to guile 3 instead of 2 <bandali>ng0, i think the announcement is coming tomorrow afternoon <drakonis>i had a really nice talk with the nix folks that got the elephant out of the room <bandali>so it’s not “officially” released just yet <nckx>brettgilio: And civodul was this weird dude popping in once in a while to (politely) ask to be taken off the newest auto-notification-bot, so I decided to check out his weird niche ‘Scheme(??)’ toy thing. <ng0>I can still package it for pkgsrc though :) <drakonis>it was worthwhile because all said the things we wish nix had <nckx>drakonis: Eelco attended & spoke at the Guix Days last year. All the elephants I know of are happy & floofy & tusk-in-cheek. <drakonis>i know both communities are in good terms <ng0>as long as it doesn't differ from 2.x too much, i think I'm fine. but thanks <bandali>yea, i don’t think there’s any animosity between the two communities <ng0>ah, this shows what builds with 3.0 already? that's useful <ng0>i naddition to guix packages <drakonis>its more of a "how can we learn from guix to improve nix" situation <jonsger>ng0: yeah, some packages do require patches <jonsger>I have a tracker bug for openSUSE, but I forgot to submit. I can send you the link tomorrow ng0 <lispmacs[work]>hello, I'm trying to add yubico-pam to my systems pam services. I tried just inserting (pam-services (cons yubico-pam (base-pam-services))) as an extra field in operating-system but that gives an error