IRC channel logs

2019-11-26.log

back to list of logs

<bandali>anyone? :/
<g_bor[m]>bandali: I can't really help you out here, as I don't have a separate home right now.
<bandali>g_bor[m], thanks for the reply, no worries :)
<janneke>bandali: like g_bor[m], i am not using a separate /home; only on one system that also boots debian
<janneke>and i find 30GB a bit too tight for /
<bandali>janneke, ah, so you have one / partition for guix and another / for debian?
<pkill9>i give guix 100 gb
<bandali>does that include /home ?
<pkill9>no
<janneke>bandali: on the one system, yes and a third for /home
<bandali>janneke, i'm a bit confused; since you just said you don't use a separate /home ?
<janneke>i do most of my debian tests and stuff in a vm, my kids still use debian some times
<janneke>bandali: sorry -- i don't, except for this one system that also has debian
<bandali>janneke, ah okay gotcha; and no worries at all!
<bandali>pkill9, btw, out of curiosity, how large is the hard drive that your 100gb guix root sits on?
<pkill9>211GB apparently
<bandali>gotcha
<bandali>i think i'll do something like that too: give ~100gb to / for guix, another ~100gb or so for /home, and maybe ~50gb for debian or trisquel
<buenouanq>anyone here use guile-sqlite3 ?
<buenouanq>what does the sqlite-exec function return and how to I use it?
<buenouanq>so far I just get errors regarding #<unspecified>
<g_bor[m]>buenouanq: I believe cuirass uses it. Maybe you can look at the source...
<g_bor[m]>buenouanq: http://git.savannah.gnu.org/cgit/guix/guix-cuirass.git/tree/src/cuirass/database.scm
<g_bor[m]>this should have what you are éppling for
<g_bor[m]>looking
<buenouanq>thank you g_bor[m]
<bandali>btw, some feedback regarding the installer: it'd be nice if the user had the ability to edit the configuration shown to them right before the install, in case they want to make any manual small tweaks before proceeding
<civodul>g_bor[m]: yes, you have to use ssh forwarding
<civodul>(for zabbix)
<civodul>GC is still running on berlin
<janneke>guix pull succeeded for me
<civodul>good
<civodul>so i'm leaving but i'll check tomorrow morning where we are
<civodul>cuirass is stopped right now
*janneke is considering to let bootstrap-tarballs build run overnight and open bug tomorrow :)
<civodul>bug, you say bug?! ;-)
*civodul -> zZz
<janneke>eh, patch proposal!
*janneke -> zZzzz
<civodul>ah ha!
<g_bor[m]>ok, l believe I do not have access. We could discuss later what is needed.
*g_bor[m] ->zZz
<bandali>hmm, the substitute keeps dying on me :(
<bandali>when trying to install guix system
<bandali>guix substitute: error: TLS error in procedure 'read_from_session_record_port': The TLS connection was non-properly terminated.
<bandali>guix system: error: substituter `substitute' died unexpectedly
<bandali>any thoughts?
***daviid`` is now known as daviid
***jonsger1 is now known as jonsger
***apteryx_ is now known as apteryx
<alextee[m]>how is mono support in guix?
<pkill9>alextee[m]: the version of mono is pretty old
<alextee[m]>hmm
<ScaredySquirrel>ok guys I wanna find out why my keyboard and mouse just don't work
<ScaredySquirrel>under Xorg as a normal user, that is
<atw>are you using tlp-service? I had to add (tlp-configuration (usb-autosuspend? #f))
<ScaredySquirrel>what is that?
<ScaredySquirrel>I'm not using tlp-service
<atw>in that case nevermind :)
<ScaredySquirrel>the local xorg log under ~/.local/share/xorg/Xorg.0.log says No input driver specified, ignoring this device
<ScaredySquirrel>so no input drivers
<ScaredySquirrel>even though one is installd
<alextee[m]>does guix not have a dotnet package yet? (i know no one likes dotnet, including me, but im trying to build a game written in C#...)
<alextee[m]>afaict it's free software https://wiki.archlinux.org/index.php/.NET_Core
<alextee[m]>grepping for "dotnet" in guix sources doesn't show anything
<alextee[m]>arch has something: https://git.archlinux.org/svntogit/community.git/tree/trunk/PKGBUILD?h=packages/dotnet-core
<bandali>hey folks, anyone know if it's possible to have guix/gruix prompt for luks password only once?
<bandali>i *think* trying the installer i'd noticed it happening, but i'm not sure
<lfam>bandali: It's an old user experience bug with a Guix-specific reason. I can't remember the exact reason but if you search the archives of guix-devel, bug-guix, and help-guix you will definitely find discussion
<bandali>lfam, ha, i found something from a couple years ago, but wasn't sure if it's still relevant
<bandali>especially since i *think* i'd seen the installer wizard take care of it
<bandali>maybe it was a pipe dream
<lfam> https://lists.gnu.org/archive/html/help-guix/2016-12/msg00096.html
<lfam>Quoting: "first GRUB needs to access the partition, and then the kernel needs to access it."
<bandali>right
<lfam>It's worth a solution
<bandali>it sure is
<bandali>i see https://lists.gnu.org/archive/html/help-guix/2017-06/msg00018.html
***user__ is now known as some_dude
<some_dude>hey guys I would like some help installing guix on a chromebook. I'm getting an error on boot time about a linux module not found and then the keyboard does not respond. please advise
<Blackbeard[m]>Hi
<Blackbeard[m]>Is any of you using Wayland on Guix?
<elais[m]>some_dude hey I installed guix on a Chromebook pixel. If you're on matrix PM me
<alextee[m]>anyone know how to get maven working? i am getting "No compiler is provided in this environment. Perhaps you are running on a JRE rather than a JDK?"
<alextee[m]>oh the jdk is a separate output, i had to do guix install icedtea:jdk
<alextee[m]>i can't build either of these https://github.com/ppy/osu https://github.com/itdelatrisu/opsu
*alextee[m] gives up
<thomassgn>Hi alextee[m], by build - do you mean to write a packagedefinition or build as a dev build in the source tree somewhere in your FS? Or something else entirely
<alextee[m]>thomassgn: i tried both
<alextee[m]>the dotnet one is out of the question but i managed to build the java one locally, but i get an error when i run the jar that it's missing libxcursor.so.1
<alextee[m]>i didnt get too far with the package either
<thomassgn>I need to leave in a wee bit. If you want to package it, look at some of the packages in /gnu/packages/java.scm. if... Ah
<alextee[m]>i did look there, and i searched a lot. i can't find any maven packages otehr than maven itself
<thomassgn>then what package provides xcursor lib? is it something java specific or is it libx11?
<alextee[m]>i already have libxcursor installed, but i think the build scripts are doing something weird
<thomassgn>ah, I would use guix environment, forgot to mention that.
<alextee[m]>anyway, i'll start developing a similar game, java/dotnet are just too disgusting :-)
<thomassgn>run: guix environment <list the packages whose dependencies you need> --ad-hoc <list of packages you need>
<thomassgn>Agree to that last one.
<alextee[m]>shouldnt it work without guix environment if i already have the packages installed and im building locally?
<alextee[m]>(ie, not a package)
<thomassgn>managed to get jitsi running yesterday, but I don't think I'll make a patch.
<alextee[m]>the problem with the package was some maven stuff, maven missing its local repository or cant download packages, etc.
<thomassgn>Oh, yes. But you'll fill your env with stuff you don't need.
<thomassgn>Ah, probably because the store (/gnu/store/) is immutable (can't be changed - sort of). So maybe you can tell maven to use a different directory than normal for storing things?
<thomassgn>and maven is (in the end) run from the store.
<thomassgn>gotogo. Good luck alextee[m] :)
<alextee[m]>i tried telling maven to use a local dir and i started getting unresolvable host errors
<alextee[m]>but thanks anyway!
<alextee[m]>i will give up on it for now, until someone packages a project using maven so i can see how it's done
<efraim>got the crate importer to do versioning, still needs some clean-up, now to make it work with the recursive importer
<Franciman>great efraim !
<efraim>recursive importer doesn't see the packaged crates, probably because they're hidden
<Franciman>efraim, I was thinking about what you said regarding Cargo.lock
<Franciman>if a package has a Cargo.lock, isn't it enough to provide reproducibility?
<efraim>we have to provide the crates for the build, and if there are any scripts we patch them. With that we necessarily change the hash and have to recompute it. Plus there's everything else in the build environment, not just the crates
<Franciman>understood
<civodul>Hello Guix!
<g_bor[m]>Hello guix!
<g_bor[m]>What is needed for me to get ssh to berlin? Is there some procedure in place?
<civodul>hello!
<civodul>g_bor[m]: you could add your public key to maintenance.git and add you as a sysadmin in berlin.scm
<civodul>and drop a mail to guix-sysadmin@gnu.org ;-)
<civodul>GC is still running BTW, which is a bit worrisome
<g_bor[m]>Ok
<efraim>berlin has spinning rust? the last time I ran 'guix gc -C 1' on bayfront it took about 90 minutes, and it has a much smaller pool
<anon987321>hi guix
<anon987321>does anybody know how to fix that "guile: failed to install locale" message?
<anon987321>it actually shows up for other guix-installed languages such as perl too
<anon987321>i did follow what it tells me to do, but it doesn't fix anything
<jonsger>anon987321: it's only a warning
<anon987321>is there no actual issue?
<civodul>anon987321: see https://guix.gnu.org/manual/en/html_node/Application-Setup.html#Locales
<civodul>in essence you need to provide locale data for Guix' libc
<efraim>looking more at the php-fpm socket-group mismatch I noticed earlier, apache has the socket group owned by httpd so it looks like the mistake is in the manual, not in the service
<anon987321>i already did it civodul
<civodul>ok, and did you set LC_ALL and related environment variables accordingly?
<civodul>what's the value of $LC_ALL and $LANGUAGE?
<Franciman>efraim, is there any place where I can follow the news on your updates for the crate importer?
<rekado>civodul: we’re still at 100% disk use
<efraim>Franciman: not really. I have a WIP blog post I sent to the guix-devel mailing list and I mention it now and then here on IRC
<anon987321>no, civodul, these are not set
<rekado>civodul: could you stop the gc and only run it on two or three disk-image files?
<rekado>that way we could probably get to a usable system faster
<efraim>I should figure something out for that
<civodul>rekado: stopped!
<Franciman><efraim> Franciman: not really. I have a WIP blog post I sent to the guix-devel mailing list and I mention it now and then here on IRC <- ah cool, I will find it
<rekado>thanks
<civodul>rekado: there's a "guix gc -D" running now, so let's see
<civodul>i'm really surprised it didn't complete overnight
<anon987321>weird, it seems ~/.guix-profile/etc/profile is messing up stuff a bit
<anon987321>it's adding a fixed /gnu/store path to $PATH
<anon987321>and i can't seem to use gnu global because of it
<anon987321>i can't uninstall global
<anon987321>and emacs can't use it
<rekado>civodul: it’s stuck on “finding garbage collector roots...”
<rekado>it’s only supposed to delete *one* item.
<efraim>i'm not sure about the internals of 'guix gc -D', but 'guix gc -C' will also clear any already marked deleted items, so the first one should take a while
<civodul>rekado: right but it has to (1) find roots, (2) delete the thing, (3) delete unused links
<civodul>#1 takes a bit of time on berlin, and #3 takes a lot of time
<civodul>however, looking at process 44763, it's currently deleting stuff from /gnu/store/trash
<civodul>meaning it's in phase #2
<rekado>efraim: berlin has an external 37T storage array with SAS disks.
<anon987321>hey, how can i uninstall guix (just the manager)
<anon987321>gonna nuke it and install it again
<anon987321>i only have a few packages anyway
<anon987321>the global thing was annoying me
<civodul>essentially: rm -rf /etc/guix /var/guix /gnu
<anon987321>thanks
<lurch>hey guix
<lurch>Is it possible to 1) define user profiles from the system definition? (-> like packages and conf) and 2) to modify the touchpad to work with touch? Couldn't find documentation on that
<lurch>my touchpad is suffering from my clicking :P
***avocado is now known as Guest28892
***jonsger1 is now known as jonsger
<alextee[m]>hmm i keep getting this warning with pangocairo include/limits.h:124:3: warning: #include_next is a GCC extension
<roptat>hi guix!
<roptat>lurch, 1 is not possible I think, 2 probably requires that you change some xorg configuration
<roptat>I'm sure there's an example in the manual but I can't find it
<roptat>oh
<roptat>civodul, can you refresh the website? I pushed my blog post on the artwork repo
<roptat>zimoun, oh I found it: https://issues.guix.gnu.org/issue/22138
<civodul>roptat: berlin is gc'ing, which is why nothing's happening
<civodul>soonish hopefully
<roptat>is it automatic?
<civodul>yes, but since the big GC lock is taken, it's just not happening
<civodul>and before that we had ENOSPC
<civodul>and probably mcron breakage as well
<civodul>SNAFU!
<roptat>ok, I'll be patient :)
<g_bor[m]>Hello guix!
<g_bor[m]>I see that babl is still problematic.
<g_bor[m]>Some time earlier I researched it, just because I can't reproduce it on my system I did not propose a patch.
<g_bor[m]>Upstream says the test failures are harmless, so they should be disabled, but I don't know which ones.
<g_bor[m]>It is hardware dependent, amd affects amd ryzen amd similar
<civodul>perhaps there's a patch floating around to make the test more robust?
<ArneBab>Does someone have a package for wrk? https://github.com/wg/wrk/blob/master/INSTALL
<efraim>2 rust import patches sent to the mailing list, still WIP
<lurch>hmm thanks anyway roptat
<lurch>guess I'll dig around some more with that touchpad thing
<lurch>guix really would benefit from a wiki arch/debian/gentoo style
<roptat>lurch, we have the cookbook now, which should contain a few things
<roptat>it's still wip
<roptat>but you're welcome to add anything to it if you want to
<roptat>have you seen "set-xorg-configuration" in the manual?
<roptat>in fact, I can't find the example I was looking for, but I think that's what you need
<roptat>lurch, something like this should work: https://paste.debian.net/1118032/
<nomr>#guix, the build server api is giving me a 502 on every request, preventing my substitutes-only 'guix pull'. Is this a known problem?
<roptat>nomr, yes I think it's related to a recent ENOSPC on the server. The gc is running, but the CI is unavailable while it's running
<nomr>thanks so much
<roptat>hopefully, it's going to be back to normal soon
<civodul>nomr: 502 on what URL?
<nomr>civodul e.g. https://berlin.guixsd.org/api/evaluations?nr=30
<davie`>hi! Anyone knows is there a way to quickly install package using `guix package --install-from-file=` if it has inputs that are not defined in any channel?
<davie`>I could do a local channel with `define-global` package definitions but `guix pull` takes a bit too long for debugging
<roptat>davie`, you could define these additional inputs in the file you're installing from
<roptat>or use -L to load more definitions from a directory
<TheZeus121>davie`: You can make a local channel with all that but instead of commiting and then `guix pull` use something like `guix environment -L <path to channel> --ad-hoc <package name>`
<Parra>hey guys I'm trying to run a cmake build script from Guix, but my cmake script does things like npm install during the build step, so this is problematic with guix, because npm install needs root permissions it seems... I've tried to use node-build-stystem but I don't have any example, what can I do? It's being frustrating
<TheZeus121>Parra: I'm not familiar with npm, does `npm install` installs dependencies or some part of your package during the build?
<TheZeus121>If it installs dependencies you should be able to remove `npm install` from CMake and list the dependencies in package inputs
<Parra>yes, my cmake file just does: execute_process("npm install ..")
<Parra>that was my original idea but I can't find an example of guix using a npm dependency, do you have an example?
<roptat>Parra, the actual issue with your cmake file is that npm install requires network access, which we don't have in the build environment
<Parra>not only that, also the root access to do mkdir in some temp.folders
<roptat>now, the npm package is probably going to be extremely hard to package (because usually npm packages have a huge number of dependencies, and we still have none of them)
<Parra>someone told me that there is a workaround around that too, because I want to install .NET Core runtime from curl-ing a sh script
<Parra>rotpat: it isn't because it's node-addon-api
<davie`>TheZeus121: Wow thanks I didn't know that
<Parra>if you check their repo they don't have dependencies
<Parra>it's only one dependency without third party dependencies
<roptat>ok, nice then :)
<roptat>you should be able to package that with the node-build-system
<nomr>npm install just fills the node_modules folder. Just provide the contents of that folder
<roptat>there's nothing very hard with it, you declare a package just like you would do with the gnu-build-system
<Parra>but then I'm mixing two build systems
<roptat>but you use the node-build-system. report back if you have any trouble with it :)
<roptat>what do you mean?
<roptat>why to build systems?
<Parra>I should mix node+cmake build system in one package
<roptat>two*
<Parra>no?
<roptat>can you build the two parts separately?
<roptat>it's probably easier if you can
<bandali>has anyone noticed the installer setting up full-disk encryption so that it only asks for the luks password once? or am i misremembering?
<Parra>I'm a noob with guix XD
<Parra>I will try
<roptat>I guess the npm package uses the rest of the package, so try building your package without the npm thing first
<roptat>then define a new package for the npm thing, and use the first as an input
<roptat>also, you said there was an npm dependency too, so maybe try to package that dependency first
<bandali>also, anyway one could access wget or curl during the install phase?
<bandali>can i just..install it?
<Parra>I would need that too in a near future bandali
<roptat>you can have curl or wget in the environment, but you won't get network access
<Parra>I need to package netcore runtime, and the only way to install it is via curl+sh
<Parra>there isn't a workaround?
<roptat>unless you're creating a fixed-output derivation (like sources are)
<bandali>roptat, can you elaborate? why not?
<bandali>also if not, would y'all consider making one or the other available in the installation iso?
<bandali>it'd be convenient for retrieving my config.scm, at least
<roptat>when building a standard derivation, guix wants to maximize the chances that it's reproducible, so it builds in a container with exactly the defined inputs and no network access
<roptat>network access is only granted if you build a fixed-output derivation, in which case you have to know what will be downloaded in advance
<davie`>roptat: Thanks! Ill check out the -L option. You also mentioned that I could define additional inputs in the file. I've tried that but failed.. It seems like i can't just substitute input field with a package like (native-inputs `("some-package" ,(package (name "some-package") ...))
<roptat>bandali, oh wait I thought you were talking about the install phase of the building process, not the installer!
<roptat>nevermind
<roptat>of course you can download things from the installer!
<bandali>roptat, haha no i meant the installer iso
<bandali>alright that's cool :)
<roptat>you can switch tty with alt+fn, and then you can do whatever you want
<bandali>nice
<bandali>roptat, and do you happen to know anything about my other question re: installer setting it up to enter luks password only once?
<roptat>no, I think the bug is still there, you have to enter the password twice
<Parra>that's the idea, for example I need netcore as a dependency of.the build, can't I download it before?
<roptat>Parra, you can't download it as part of the build process, but you can create a package for it if it doesn't exist, and use it as an input
<Parra>but that's moving the problem, netcore needs to be installed through a sh script that is commonly executed by means of a pipe with wget or curl
<Parra>the only way netcore can be installed is like that as far as I know,.it is boostrapped by itself
<roptat>so you can use the sh script as the source for a package (it would be downloaded) and execute it as the build procedure ?
<Parra>is that possible?
<roptat>sure
<Parra>perfect
<roptat>although it probably won't be accepted in guix proper
<roptat>because it's not bootstrappable
<Parra>yep but I don't care
<roptat>I understand :)
<Parra>I'm doing all this thing in order to create a self contained package of my software
<Parra>which is a Polyglot
<Parra>I couldn't package it properly yet, after one year
<roptat>mh... actually if the script for netcore wants to download things, it won't be able to
<bandali>ah okay
<Parra>and I'm trying guix now, it seems promising
<Parra>oh shit
<Parra>then? there isn't a workaround?
<roptat>depends on how the script works
<roptat>if it's creating a cache somewhere, you could use a few origin records as input, put their content in the cache and then use the script
<Parra>:/
<roptat>hopefully, the script will find that all it needs is in the cache, so it won't try to download stuff
<roptat>but you'll have to understand how the script is working to trick it into thinking it was already run before and it doesn't need to download stuff
<Parra>it's ultra tricky then
<Parra>XD
<roptat>right... maybe you could go low-level and use a derivation directly. Then you could configure it to allow network access, I'm not sure
<Parra>maybe someone can package it before me
<roptat>or you could make it a fixed-output derivation, which is not so low-level
<Parra>ruby is problematic too
<Parra>I tried to build with libpython and it worked
<Parra>but not with ruby, it doesn't find the Lib
<Parra>I need to build against libruby too
<roptat>isn't it just an issue with env vars?
<Parra>I don't know
<Parra>cmake uses common paths to find ruby
<roptat>we have a ruby build system, so I think we have everything needed to build this kind of package
<roptat>mh... so you'll need to fix the cmake file, maybe
<Parra>it's not allowed by me
<Parra>FindRuby is standard of CMake
<roptat>oh, mh...
<Parra>I could reimplement it but I think it should be easier if we solve it from guix side
<roptat>let's search for "FindRuby cmake environment variable" ^^
<Parra>haha
<Parra>wait
<Parra> https://cmake.org/cmake/help/git-stage/module/FindPython.html
<Parra>this one is python
<Parra> https://cmake.org/cmake/help/v3.0/module/FindRuby.html
<Parra>this one is ruby
<Parra> https://github.com/Kitware/CMake/blob/8741d9e262a66de86fac6008bbf466dc25f72ec6/Modules/FindRuby.cmake#L37
<roptat>I'm not a cmake expert, but maybe you could set RUBY_LIBRARY?
<Parra>check the implementation
<Parra>I will try it
<Parra>another thing is the following
<Parra>after generating the whole package, I want to do some tests
<Parra>I want to build a program against my library
<Parra>what I would need?
<Parra>would I *
<Parra>because rpath and ldpath may point incorrectly if using my own gcc instead of guix gcc
<Parra>can I export a .pc or .cmake file for finding properly the libs?
<roptat>you mean you want to build something outside of guix?
<Parra>setting all compiler flags
<Parra>yes, after doing guix pack -RR metacall
<roptat>I think it's possible, but I never tried
<Parra>I need to do that
<roptat>you could certainly add a .pc or .cmake file to the output of your package
<zimoun>roptat: thank you for the pointer to the bug report and the associated patch. I will give it a try... maybe I will be convinced ;-)
<roptat>zimoun, I answered to the bug with some of my thoughts too
<Parra>but is that automated? or have I to make it by hand?
<roptat>by hand I think
<Parra>I don't know, so many questions, I'm trying to do a too complex thing for a beginner like me
<roptat>although usually, packages can create their .pc or .cmake file correctly if upstream configure them to do so
<Parra>my library generates the cmake automatically
<roptat>then you could install that
<roptat>I think it will work even outside of guix
<Parra>but that doesn't include its own ld etc
<Parra>I should try it
<roptat>it should be fine, I think
<Parra>also, I want to do dlopen and see what happens
<roptat>dlopen is way more tricky
<Parra>basically
<Parra>this is what happens:
<Parra>I have a nodejs script where I do require ("metacall");
<Parra>then, there is a metacall_port.node which is a .so file for NodeJS, and NodeJS loads it dynamically
<Parra>this is linked dynamically against libmetacall.so which is the core lib
<Parra>then this lib implements a plugin system which loads other libs, for example, libpy_loader.so
<Parra>libpy_loader.so is compiled against libpython
<Parra>and then I can load python scripts from node, and call the functions
<Parra>that's how the polyglot works
<Parra>my idea was to package this inside a zip, and downloading it when doing npm install
<vagrantc>any guix folks in madrid?
<Parra>or this or compile everything statically but it would be a nightmare
<Parra>I'm from Valencia vagrantc
<roptat>ok, I'm not sure how dlopen works... would it follow the runpath?
<Parra>in theory
<Parra>guix says that packaging with -RR implements $ORIGIN as rpath and ldpath
<Parra>this means that it search from ./ instead of /
<roptat>so I think your dlopen will work
<Parra>well, relative vs absolute
<Parra>me too
<roptat>you'll have to test though :p
<vagrantc>Parra: practically around the corner, but still quite far. :)
<Parra>I think before trying to package other things, I should do a PoC
<Parra>Nodejs <> python
<Parra>python backend already works, I should try to build the nodejs frontend now
<Parra>(aka the node port which offers metacall to node world)
<Parra>vagrantc: it's not a big deal, 3.5h with car, 2h with ave
<pkill9>does anyone get message slike this when trying to play music with "lollypop": `(lollypop:4341): GStreamer-WARNING **: 14:25:54.570: Failed to load plugin '/run/current-system/profile/lib/gstreamer-1.0/libgstreplaygain.so': /gnu/store/h90vnqw0nwd0hhm1l5dgxsdrigddfmq4-glibc-2.28/lib/libm.so.6: version `GLIBC_2.29' not found (required by /run/current-system/profile/lib/gstreamer-1.0/libgstreplaygain.so)`
<Parra>what do you think roptat?
<roptat>sounds good, but the node-build-system might not be mature enough
<Parra>well
<Parra>I only need to install one package with no dependencies
<pkill9>it's odd that glibc version 2.28 is being used, i have similar messages elsewhere but i don't know where glibc-2.28 is being used
<Parra>it should be easy, no?
<Parra>in fact, I can even just clone the git repo, I only need some c/c++ headers and that's it
<roptat>we can try
<Parra>or I can remove that depend from my system and paste the headers directly, although it's a dirty solution it will work
<Parra>XD
<roptat>I mean, the node-build-system should work
<roptat>it just hasn't been tested on a lot of packages yet
<Parra>node sucks dude, it has great design decisions but at same time very poor design decisions, I love it and hate it in the same way xD
<Parra>okay
<Parra>I will try
<roptat>if you have any trouble with it, report here, we'll try to help
<roptat>pkill9, your current system still uses glib 2.28?
<roptat>glibc*
<roptat>or is it lollypop?
<Parra>thanks so much roptat
<pkill9>roptat: it doesn't look like either of them do when I run `guix graph -t references lollypop | grep glibc-2.28`or `guix graph -t references $(guix system build config.scm)`
<pkill9>*`guix graph -t references $(guix system build config.scm) | grep glibc-2.28`
<roptat>and what is your current lollypop?
<roptat>is it the same than the one produced by guix build lollypop?
<pkill9>1.2.7
<roptat>I mean, do you run a lollypop that's from an older guix revision?
<pkill9>ohh i might be actually
<pkill9>because i have it in a separate profile
<pkill9>ahhhh yea
<pkill9>thanks
<Parra>I think the loader won't work
<Parra>because dlopen takes the loader which is in the current process and nodejs will use the loader of the system instead of the one I'm providing in the guix package
<Parra>but linux already handles this in theory
<Parra>since an old version of the kernel
<Parra>so if rpath works is enough
<Parra>it's*
***ng0_ is now known as ng0
<Parra>I will check it
<ScaredySquirrel>ok um...why when xf86-input-libinput is installed the xorg server still doesnt't detect any input devices as a normal user?
<pkill9>i'd like to make a TUI for managing guix profiles
<ScaredySquirrel>it detects my PixArt USB optical mouse with the config/udev module then it says "no input driver specified"
<pkill9>so things like, selecting which packages to upgrade, and showing which packages will have substittutes available
<pkill9>and showing the progress of the whole package upgrade operation
<pkill9>s/package/profile
<ScaredySquirrel>guys how do I put an alternative sudoers file in /etc/config.scm
<roptat>ScaredySquirrel, hi :)
<pkill9>(sudoers-file (plain-file "blah blah blah")) i think
<roptat>there's a sudoers-file field in the operating-system record type
<pkill9>we need a sudoers-file-service so we can add configuration
<ScaredySquirrel>ah
<pkill9>and remove risk of badly written sudoers file
<pkill9>and then services can add their own sudoers file config entries
<roptat>ScaredySquirrel, for your xorg issue, I think libinput is already one of xorg drivers available by default, installing it in your profile won't do anything
<roptat>it's probably a configuration issue, maybe you need to add a few lines to the xorg.conf, using set-xorg-configuration?
<ScaredySquirrel>but why would I need to add anything with set-xorg-configuration? I added just my keyboard layout in there
<roptat>I don't know, xorg should just work by default, but if it doesn't, I suspect your hardware needs more config
<bricewge>Hello Guix!
<ScaredySquirrel>it's just a usb mouse
<ScaredySquirrel>the usb keyboard doesn't work either
<ScaredySquirrel>both work on the linux console
<bricewge>dhcp-client-service fail if one of the interface need non free firmware, even though the other interface support free firmware. Is it expected? Should I open a bug report?
<pkill9>i would have thought the interface wouldn't shop up if it's firmware isn't loaded
<pkill9>s/shop/show
<bricewge>I get an error (in dmesg) that my wifi card doesn't support free firmware but it shows up in `ip a s`.
<bricewge>Which stop the networking service. Removing the wifi card allows the networking service to start successfully through ethernet.
<roptat>ScaredySquirrel: can you share your config?
<ArneBab>Can I force guix environment to use an old glibc? guix environment --ad-hoc --pure glibc@2.28 util-linux xcb-util-image libxcb xcb-util-keysyms sqlite nss libxcomposite libxtst libxrandr alsa-lib libxml2 libxslt libx11 libxkbcommon glib dbus dbus-glib libxt xkeyboard-config libxcursor util-linux coreutils mesa gcc-toolchain:debug findutils fontconfig libdrm libpng mesa-utils
<ArneBab>Usecase: I have a compiled binary that currently segfaults, I’m trying to get it to work again.
<ArneBab>It started segfaulting at some point in the past two months but worked before that.
<ArneBab>Is it possible to use guix environment for that?
<pkill9>ArneBab: what do you want to do with the older glibc?
<pkill9>that command will work yes ArneBab
<roptat>No it won't
<pkill9>but all the other programs will still have been compied with glibc 2.29, but i'm not sure if that was part of your question
<pkill9>comiled*
<ArneBab>I get segfaults, I need to solve the segfaults
<roptat>It will add an old glibc, but the rest will still use the new one
<ArneBab>pkill9: I need to get the libraries compiled with the older glibc
<ArneBab>I *think* that it was the glibc update, but I’m not sure
<ArneBab>trying to run through gdb gets gdb to also segfault
<roptat>You have two solutions: input rewritting or using an old guix
<pkill9>in that case you want to add the flag --with-input=glibc=glibc@2.28 after `guix environment`
<ArneBab>pkill9: nice — it starts to download sources!
<ArneBab>thank you!
<pkill9>you're welcome
<ArneBab>build fails …
<pkill9>which build fails?
<ArneBab>binutils fails
<ArneBab>but I don’t see the errors in the log: /var/log/guix/drvs/xm/058zglhm6wnkv9a3y2d4ppd2c9kf8k-binutils-2.32.drv.bz2
<ArneBab>ends with phase `compress-documentation' succeeded after 0.7 seconds
<pkill9>hmm, it doesn't do anything if i run guix build --with-input=glibc=glibc@2.28 binutils
<ArneBab>it’s a dependency of gcc-toolchain
<ArneBab>cannot build derivation `/gnu/store/vndf10x4h0g9dnlcr6n9nyb9fkv9c58s-gcc-toolchain-9.2.0.drv': 1 dependencies couldn't be built
<ArneBab>guix environment: Fehler: build of `/gnu/store/vndf10x4h0g9dnlcr6n9nyb9fkv9c58s-gcc-toolchain-9.2.0.drv' failed
<roptat>Weird
<roptat>Maybe try again?
<ArneBab>I did
<ArneBab>seeing it for the third time
<roptat>Always the same failure?
<ArneBab>what’s weird is that I don’t see why it should have failed when the logs sounds like a success
<roptat>Right…
<ArneBab>at least two of those failures were the same, I think the third, too
<roptat>Are you sure about the log file?
<roptat>Is it the one given by the guix command?
<roptat>If so, it may be worth a bug report
<roptat>You could work around that by using an older guix that had glibc 2.28
<ArneBab>maybe this?
<ArneBab>\ „compress-documentation“-Phaseoutput (`/gnu/store/cs0gi9yhwcldrcc17gppg2vszzx87h3j-binutils-2.32') is not allowed to refer to path `/gnu/store/wzfdcfmp6ax2agl43bw2pnpjsnxn1v2p-glibc-2.28'
<roptat>Oh sounds like an error
<roptat>Oh, why not?
<ArneBab>I don’t know
<roptat>I'd email help-guix
<lispmacs>so, erm, any update on the subtitute failure problem?
<lispmacs>I'm trying now a --no-substitutes build just to see what happens
<ArneBab>roptat: did that now, thank you!
<pkill9>if you have anewer guix with the new time-machine command, you could run `guix time-machine --commit=<commit-before-glibc-2.28-upgrade> -- guix environment --ad-hoc ...`
<Parra>roptat: I'm working on the node package, it seems to fail due to shebang
<Parra>is the check command important?
<Parra>it runs only test, right?
<Parra>I can avoid it instead of trying to sed the sebangs
<roptat>Parra: what error do you get? Check can be skipped with #:tests? #f
<roptat>Don't sed, use substitutes* :)
<roptat>Or report the issue, because shebangs should be fixed automatically
<Parra>....node-10.15.3/lib/node_modules/npm/node_modules/npm-lifecycle/node-gyp-bin/node-gyp: ...node../...npm/.../node-gyp/bin/node-gyp.js: /usr/bin/env: bad interpeter
<Parra>something like that
<Parra>node-gyp is the extension builder for nodejs
<Parra>I think I can avoid it, I don't care about tests
<Parra>if it doesn't work it will crash later on
<Parra>question, isn't mit license supported? I didn't saw it in licenses.scm
<Parra>and another question, can I do guix build for two packages inside one file? I did this:
<Parra>metacall.scm:
<Parra>(define-public node-addon-api
<Parra>....
<Parra>)
<Parra>(define-public metacall
<Parra>...
<Parra>)
<Parra>then:
<Parra>guix build node-addon-api metacall -L /path/to/the/scheme/file
<Parra>uh, I wanted to send only one message, sorry
<zig>there is several mit-like license, one of them is expat
<Parra>the repo says exclusively mit, can I change that?
<Parra>expat == mit?
<zig>the other is x11
<zig>mit might be x11 or mit depending on the text, look into guix/guix/licenses.scm to find the link to the text.
<zig>mit might be x11 or *expat* depending on the text, look into guix/guix/licenses.scm to find the link to the text.
<zig> https://directory.fsf.org/wiki/License:Expat
<Parra> oh yeah man it worked
<Parra>roptat: I could build node_port.node
<Parra>thanks zig
<Parra>I'm going to check that now
<roptat>Parra: cool :)
<alextee[m]>hi, there is no librejs settings in icecat 68.2.0-guix0-preview3
*alextee[m] uploaded an image: Screenshot from 2019-11-26 18-49-55.png (31KB) < https://matrix.org/_matrix/media/r0/download/matrix.org/GdYkERSARkpyVGoTAZifyBFC >
<alextee[m]>^ anyone else get this?
<Parra>roptat: now I'm facing a new error: node_port: depends on libmetacall.so, which cannot be found in RUNPATH
<anon987321>had some nasty issues with environment variables
<anon987321>just sent a bug report to bug-guix
<roptat>Parra: what is trying to find libmetacall.so in its runpath?
<pkill9>not me alextee[m]
<alextee[m]> https://invidio.us/watch?v=pRagiT78NsY something like this
<alextee[m]>wait did i just send a link? ignore pls, matrix being dumb
<Parra>zig: it was expat, thanks
<Parra>roptat: no idea let me check, libmetacall is my own lib that I already compiled
<Parra>it's inside "validate-runpath" step
<Parra>it says: validating RUNPATH of 8 binaries in ...metacall../lib
<Parra>and then it says that
<Parra>what I do in debian (docker) is to run ldpath on libmetacall in order to workaround this
<roptat>You probably need to pass some flags to the linker/gcc to add the output directory to the runpath
<Parra>so node_port can find libmetacall, but in guix I have no idea how to do that
<finfin>does anyone know how to make Emacs-Guix read packages in the location defined by $GUIX_PACKAGE_PATH?
<Parra>do you have an example?
<roptat>Grep for -Wl in gnu/packages
<roptat>Sorry, I'm not om a computer, it's not easy to access the sources for me rightqnow
<Parra>okay, don't worry
<Parra>I'm going to investigate
<Parra>what is RUNPATH anyway?
<roptat>It's a mechanism for elf binaries to include a list of non standard locations where they can find libraries
<roptat>It's useful in guix because there's no standard location, and it allows us to record precisely a dependency
<roptat>The validate runpath phase checks that the runpath only has store paths, and that they are sufficient to load the binary's shared libraries
<roptat>Same for libraries
<pkill9>it's a sweet phase
<Parra>excellent
<Parra>I didn't know about it
<Parra>I knew about ld and rpath
<Parra>google suggest about patchelf, but I think it's not necessary if I have the code and control over build step
<roptat>Indeed
<Parra>I want to make a meme about this
<Parra>it's like, how many levels of build systems do you use? look at this
<Parra>XD
<roptat>:)
<PotentialUser-13>Hi, I'm following the video tutorials on the website to add new packages to guix. I have followed the instructions for Part One, but in part two, the video uses guix without ./pre-inst-env to guix import cran aspi, I cannot use these after following Part One, nor can I use git to switch over to a new branch in the repo
<PotentialUser-13>My guess is that I'm in an environment that only has the packages I specified with guix environment, but I still don't get why the video does not explain well in Part Two how to follow the steps over from Part One if we're on a guix environment
<nomr>Qt5 isn't working for me on a foreign distro, due to being unable to find libQt5Core.so.5, which is in my profile lib path which is $LIBRARY_PATH
<nomr>I tried setting LD_LIBRARY_PATH to the same value, but then everything segfaults on launch 0_0
<nomr>Other programs seem to find their libs fine
<civodul>nomr: re Qt5, i'm sure that's because the distro runs an old kernel, right?
<civodul>see the "bonus section" at https://guix.gnu.org/blog/2019/running-a-guix-xfce-desktop-on-centos-7/
<roptat>civodul, I've been pointed to https://www.divio.com/blog/documentation/ today
<roptat>they make some good points, and I think it may help us improve the cookbook
<PotentialUser-13>I think the video tutorial for packaging are confusing
<roptat>basically, they divide documentation into four parts: tutorials, how-tos, explanations and reference
<roptat>that's two dimensions: either theoretical or practical, and either work-oriented or study-oriented
<PotentialUser-13>They don't transition well between parts -- part one talks about setting guix environment but part two doesn't mention how this connects to part one, it only starts using guix import right away
<PotentialUser-13>then switchest to ./pre-inst-env
<nomr>(ok ... it is because libQt5Core.so.5 from qtbase-5.11.3 uses ABI 3.16 and I am running linux 3.3.8. other stuff is ABI 2.6.32. hard problem to diagnose, got lucky)
<roptat>I'd like to suggest a change to the cookbook to put everything in one of the first three categories, instead of having everything ordered by topic
<roptat>so the cookbook would start with tutorials (tutorial on guix installation, tutorial on the guix system, tutorial on scheme, on packaging), then with how-tos (guides on advanced uses of the package manager, on how to configure certain things on the guix system, on how to package) and finally with more theoretical discussions and explanations about guix (what is a derivation, how do generations work, etc)
<roptat>does it make sense?
<rockandska>zimoun: thanks for your investigation about "guix pull" who change the symlink. feel free to ask me here if needed
*janneke adds another url to wip-bootstrap, and -> zZzzz