IRC channel logs


back to list of logs

<cbaines>interestingly it seems like the downtime effects a lot of FSF hosted stuff, e.g. and too
***ChanServ sets mode: +o nckx
***nckx changes topic to 'GNU Guix | guix pull servers are currently down | 1.0.1 is out! get it at | videos: | bugs and patches: | paste: | Guix in high-performance computing: | This channel is logged:'
***ChanServ sets mode: -o nckx
<nckx>"FSF services are temporarily down for maintenance. Thanks for your patience!"
*vagrantc wonders about high availability with guix
<reepca>I wonder how much space all sources for every package in guix would take, cumulatively...
<reepca>just in case I one day want to go full hermit mode
<vagrantc>"guix build --source ." to find out :)
<jlicht>hey guix!
<reepca>vagrantc: guix build: error: .: unknown package
<vagrantc>doh. had assumed it'd take regexps
<nckx>reepca: They should fit in 100 GiB, although that info stale af.
<nckx>‘Remarkably doable.’
<idnull>it seems that this problem persists In procedure module-lookup: Unbound variable: openscenegraph (I copied suckless.scm to my repo) channels.scm
<jlicht>idnull: you are missing %default-channels in your declaration, AFAICS
<jlicht>I have something like (cons* (channel ....) %default-channels) in my channels.scm, FYI
<mbakke>idnull: I believe the openscenegraph problem should be fixed with commit 84a0c6ff63afc296c1bc5873f783693dccc5ded2
***ChanServ sets mode: +o nckx
***nckx changes topic to 'GNU Guix | 1.0.1 is out! get it at | videos: | bugs and patches: | paste: | Guix in high-performance computing: | This channel is logged:'
<reepca>Welp, reconfigured with gdm and sure enough I can't start a graphical session now. 'New session c506 of user gdm', yikes. Meanwhile whenever I booted with sddm recently, I had to ctrl+alt+f2 my way to a tty to restart it, since it wouldn't start on its own (though herd reported it started). So I guess I've come full circle now, back to slim!
***ChanServ sets mode: -o nckx
***Digitteknohippie is now known as Digit
<retropikzel>I have distorded audio in icecat, any ideas what could help?
<shrdlu68>Hello world, I believe I've found my next distro.
<rk4>(suggestion (topic #guix-converations) (body should be conducted in s-exps))
<shrdlu68>How do I install the system without a USB, DVD, or PXE? Is there something like Gentoo's Stage3 tarball?
<rk4>if you can access the boot disk and installation disk, while running in a linux distro, i believe you can do an install
<shrdlu68>Scenario: I want to install Guix System in a cloud VM, where I can't boot from the Guix iso. However I can boot the VM in "rescue mode", which runs some distro like Ubuntu, where I can partition the file system and install Guix System
***ng0_ is now known as ng0
<shrdlu68>What do I I've gotten as far as `herd start cow-store /mnt` in, however it seems herd is part of Guix system, which I don't have.
<rk4>i haven't used the new installer, but the standard old method was basically 'boot off a livecd, then run a bunch of manual commands and then tell the guix command to dump a bunch of files corresponding to a bootable system in /mnt/foo' - which is still documented here -
<rk4>when i did that i used a guix live image, but i don't recall anything remarkable about it, the main advantage was that the guix package manager was already installed
<shrdlu68>The manual says that the the command above "This makes /gnu/store copy-on-write, such that packages added to it during the installation phase are written to the target disk on /mnt rather than kept in memory. This is necessary because the first phase of the guix system init command (see below) entails downloads or builds to /gnu/store which, initially, is an in-memory file system."
<shrdlu68>I do have the guix package manager, after running the installer script referenced in the Manual. But I don't have herd.
<rk4>see i thought you could just mkfs, mount, guix system init blah.scm /mnt, but i don't know enough to tell if that step is essential. :( wait a while, i'm europe is waking up :)
<shrdlu68>Alright, waiting for Europe.
<ramajd>Hi, does anyone knows that `nslookup` is located in which package in guix ? thanks
<pinoaffe>retropikzel: I also have distorted (seemingly lowpassed) audio in some media players in icecat, particularly the twitter one, but I have no clue what could fix it
<pinoaffe>I've just been playing twitter videos using mpv as a workaround
<retropikzel>pinoaffe: I usually play all my videos with mplayer or such, but countrys public broadcasting companys live radio is distorted. Not all audio, so I would guess it has something to do with codecs maybe?
<pinoaffe>it might, or it might be one of pulseaudios "special" features
<pinoaffe>shrdlu68: you can install guix (the distro) using guix (the package manager) from nearly any linux system, in a similar manner to how one would bootstrap, say, a debian system
<shrdlu68>pinoaffe: So `herd start cow-store /mnt` is not essential?
<roptat>nckx, no idea what went wrong, I didn't touch it
<pinoaffe>shrdlu68: it is
<roptat>maybe a filesystem issue, or a gc that didn't work as expected?
<shrdlu68>pinoaffe: So how do I bootstrap Guix System from an Ubuntu Live disk?
<pinoaffe>shrdlu68: it is not documented nor supported, but it should be possible if you get both herd and guix set up on the ubuntu system
<shrdlu68>pinoaffe: A tarball that I can untar onto / and chroot from there, Gentoo-style, is the only other option I see. I've tried `guix system init /mnt/etc/config.scm /mnt`, but I end up running out of space, since the filesystem is on the livedisk.
<pinoaffe>huh, so why aren't you mounting a proper blockdevice onto /mnt?
<pinoaffe>I guess it might be possible to prepare a tarball on another machine, but that might give ya some issues with BLKIDs and the like
<roptat>pinoaffe, the issue is that guix system init fills /gnu/store, and then copies it to the target
<efraim>are ilmbase and openexr even supposed to be installed separately?
<pinoaffe>roptat: ah, yeah that's an issue
<roptat>pinoaffe, I think that's the goal of cow-store: instead of installing on /gnu/store, it installs on /mnt/gnu/store directly, but can still access it as /gnu/store
<roptat>but you can probably replicate the behavior of cow-store manually on the command line
<pinoaffe>or maybe do some mounting tricks or even a pivot_root to ensure /gnu/store is backed by a blockdevice from the get-go
<g_bor>hello guix!
<civodul>Hello Guix!
<civodul>what's up?
<idnull>hello once again, I've manage to build my packages, but I have one more question. I need to put config.h in the build dir (for dwm/st)? How can I do this?
<pinoaffe>idnull: by making a derivative package with a phase added to the build process, see
<idnull>Am I doing something wrongly here?
<idnull>lambda*: invalid argument list in subform (#:allow-other-keys) of (#:allow-other-keys)
<pinoaffe>idnull: I don't see what you'd be doing wrong in the lamba* definition, but you're not supposed to add a string describing the location of your desired config.h, you're supposed to add guile that modifies the tarball, adding your desired configuration
<lprndn>idnull: I might be wrong but if you don't need any keys you could go with (lambda* _ )
<pinoaffe>what you did now is add a phase that does nothing but return a string
<pinoaffe>and you could indeed probably just do (lambda _ )
<idnull>pinoaffe: how can I make a substitution of a config.h then?
<oldosfan>does anyone here know when Guix is going to bundle a newer version of GNOME?
<oldosfan>3.28 is getting really old now
<pkill9>that's all cow-store does?
<civodul>hi oldosfan!
<civodul>oldosfan: i think there was work towards an upgrade, but it takes time
<civodul>would you like to give a hand?
<oldosfan>perhaps, if I can get building GNOME from source to actually work on Guix lmao
<civodul>well the package definitions are already there
<oldosfan>yeah ik that
<civodul>so it's a matter of updating them and testing things
<oldosfan>for some reason 3.34 doesn't build
<oldosfan>I might work on that though
<shrdlu68>I seem to have successfully done `guix system init` by chrooting. Haven't rebooted yet, though.
<roptat>there's a new version in core-updates, no?
<roptat>oldosfan, ^
<roptat>are we going to merge core-updates soon? :)
<oldosfan>I'll take a look
<roptat>also, I think there's wip-gnome-updates or gnome-updates branch
<idnull>ambiguous package specification `dwm choosing dwm@6.2 from gnu/packages/suckless.scm:79:2
<idnull> how to choose from my repo not guix one?
<lprndn>idnull: I would suggest you to change the name of your custom package ;)
<idnull>And while I'm here, I want to build st from master, can one skip sha256 somehow?
<roptat>if st has a source from git, --with-branch=st=master I think
<g_bor>hello guix!
<roptat>that should skip hash validation
<roptat>hi g_bor :)
<g_bor>:) jsut bumped into the next home-manager thing :)
<g_bor>trying to use gnupg
<roptat>not an issue for me: I use a USB key, so I just set up a link from ~/.gnupg to a directory where the key is always mounted
<idnull>ty:all: for help I'll try later rebuilding from master without sha256 once again later
<g_bor>right now it fails with can't create directory ~/.gnupg
<roptat>as usual, you can create a symlink to your data directory
<jonsger>oldosfan: Kei Kebreau worked on gnome3.32. He said that it lies here: But I can't see any gnome3.32 patches
<g_bor>:) yes, but I would like to this without it if possible
<roptat>.gnupg is supposed to be at least partially mutable
<roptat>(key updates, ...)
<jonsger>oldosfan: gnome3.30 is in the core-updates branch and should arrive in the next days in "normal" guix :)
<roptat>hm... haven't watched yet, but maybe of interest to us:
<g_bor>yes, I am aware of that... I would like to go as far as possible with this.
<roptat>if you can come up with a home definition, it'd be great :)
<g_bor>roptat: I'm on it. Finally I would like to do something like %guix-development-environment, and just having to put the gpg and shh key in place,and have everything else working out of the box.
<idnull>At last!
<idnull>it only took me two days (i'm a little slow yeah)
<g_bor>roptat: what would be an easy way to create an empty directory in the resulting profile, ala mkdir-p?
<g_bor>It looks like to me that is not exported directly. I guess it would help while developing.
<g_bor>Maybe I am missing something...
<roptat>g_bor, the root-home-type doesn't support that directly
<jonsger>idnull: nice :)
<roptat>it only creates symlinks
<roptat>you can symlink to an empty directory though
<g_bor>roptat: an make that read only... neat trick
<roptat>you can use a file-like object too
<idnull>jonsger: this was the easiest part: now I need to copy new moon, slstatus, wireguard and find a way to make a theme package (and set this theme as a transaction)
<roptat>like (computed-file "empty" #~(mkdir-p #$output))
<g_bor>roptat: I was aware of that, thanks.
<roptat>I think that will work :)
<g_bor>I will go with it then :)
<civodul>roptat: speaking of which, did you see ?
<civodul>i haven't watched it yet but it might be of interest
<civodul>you should blog about Guix Home, BTW!
<roptat>didn't watch it yet, but that's how I found the other video :)
<civodul>heheh :-)
<roptat>yeah... now is not a good time for me
<civodul>i reached it via
<civodul>which also looks interesting
<roptat>a few other videos from that conference look interesting
<roptat>there's one about nix too
<shrdlu68>Can't seem to find my hard disk after rebooting. Am I missing kernel modules?
<shrdlu68>I'm trying the "bare bones" configuration.
<roptat>shrdlu68, probably, but guix should have warned you during guix system init
<civodul>roptat: i saw that and i wasn't convinced that "recursive attribute sets in Nix" were a good fit for an intro to FPM :-)
<civodul>it's as if we were discussing syntax-case in an introductory talk on Guix ;-)
<g_bor>roptat: I guess I made some mistake somewhere... I now get guix home command not found.
<g_bor>I remember it was working earlier though.
<roptat>g_bor, check your GUILE_LOAD_PATH
<roptat>did you change anything in guix/scripts/home.scm?
<g_bor>not, and I've just pulled.
<g_bor>The laod path only has the /run/current-system...
<shrdlu68>roptat: I didn't get a warning, althoug I'm doing all this in a chroot.
<roptat>g_bor, there's a bug in guix where it doesn't load the channel before checking for subcommands, so you have to set GUILE_LOAD_PATH
<roptat>shrdlu68, maybe that's the reason
<roptat>shrdlu68, you should boot on another distro and check what is needed for your disk
<g_bor>roptat: how do you manage that? Do you st it up from a shell startup fiel?
<g_bor>Ok, let's try it again: do you set it up from a shell startup file?
<roptat>currently I do it manually, but I could add it to my .bash_profile
<roptat>(which is a file-like object, by the way :))
<g_bor>:) I was also thinking about that.
<g_bor>I remember talking about this before, but I just can't find it right now. What is the correct value?
<roptat>g_bor, not current, but:
<shrdlu68>Does someone have the .config of the default Linux-Libre kernel?
<jlicht>hey guix
<roptat>just add (simple-file-home bashrc ".bashrc") to the list
<g_bor>roptat: thanks, that helps.
<g_bor>roptat: what is the correct value of GUILE_LOAD_PATH?
<roptat>it should contain an item that points to your guix home git checkout, or to ~/.config/guix/current/share/guile/site/2.2
<g_bor>ok, thanks. just what I was looking for.
<g_bor>roptat: one more question, why did you use computed-file for bashrc? It looks like a plain-file?
<roptat>don't know :)
<pkill9>wow the silver searcher is a lot faster than grep
<pkill9>is any work being done on making the create importer able to import recursively?
<jonsger>shrdlu68: gnu/packages/aux-files/linux-libre/ there you should find them
<jlicht>pkill9: I think there is a patch floating somewhere that does exactly that
<jlicht>I recall it being functional with some rough edges
<jonsger>pkill9: I think efraim has something for that
<pkill9>also, does anyone know why this rust package doesn't install the binaries?
<pkill9>it just installs the license files
<efraim>the 'import crate foo@version' and import crate foo --recursive' got approved together and foo@version got merged first, the recursive one wasn't rebased
<efraim>pkill9: are you sure there's a binary?
<pkill9>there should be, I'm trying to check with the -K option, but I can't cancel the build in time
<efraim>looking at the source it looks like it's a library
<pkill9>this is what I want, maybe it's not the right crate:
<pkill9>oh yea, the importer has the URL
<efraim>also i'm not sure the recursive importer makes sure it imports the correct version
<atw>Hello, I am trying to move everything that's not an implementation of lua into lua-xyz.scm (this is because I plan to add some lua libraries and this seems like best practice). But, when attempting a clean build, I get a confusing failure: I'm stumped!
<roptat>atw, did you forget to import (gnu packages lua) in (gnu packages lua-xyz)?
<atw>I promise I didn't!
<roptat>I think lua-5.1 is used somewhere where it's not imported
<atw>hmm, good thought, I'll just grep everywhere for it
<roptat>the error with zip is just a side-effect
<roptat>I get it every time I make a syntax error or there is an unbound variable
<atw>every occurrence of "lua-5.1" I could find is in a file with the appropriate "#:use-module (gnu packages lua)"
<roptat>did you transfer its definition to lua-xyz by mistake? :p
<atw>just checked again and I did not
<atw>roptat: I've got to run to work, would you leave a message with sneek if anything occurs to you? TIA!
<civodul>comrades, look at the code at
<civodul>and move your mouse over it
<civodul>and enjoy! :-)
<idnull>wow, this feels great!
<mbakke>civodul: very nice!
<mbakke>efraim: do you remember why freeimage was reverted to use the bundled libjxr? (commit f347c24acc14e080dc2801561edca0d525a90257)
<roptat>civodul, well done!
<roptat>looks like what I used for my own website :)
<jonsger>civodul: and what happens if we have 11 parens :P
<jonsger>but serious, this is nice and even without JS :)
<apteryx>civodul: pretty cool1
<nckx>civodul: Fun!
<efraim>mbakke: wasn't that a while ago? Likely it either didn't compile correctly or it wouldn't actually link to it, probably the later
<efraim>I'd be tempted to merge ilmbase and openexr into one package since the double inclusion of headers drives freeimage crazy and openexr already propagates ilmbase
<pkill9>ooo very nice :)
<pkill9>did the Guix manual have a margin before?
<civodul>jonsger: ten parens should be enough for everyone :-)
<civodul>i stole the trick from the Chicken web site:
<civodul>pkill9: the manual had a margin but i've made it wider recently
<civodul>rather, i've reduced the max-width of the body
<pkill9>ah, it feels nicer I think, it's easier to read
<mbakke>efraim: wdym by driving freeimage crazy?
<civodul>davexunit: perhaps we should provide this syntax-highlighting mode directly as part of your package?
<mbakke>efraim: it's from april last year... AFAICT none of the freeimage dependents work since the unbundling
*mbakke applies and tries to get guile-sly working again
<efraim>Hmm, the unbundling was so it would build on arm and aarch64. If none of the dependents work that's a problem
<efraim>I'm away from my computer ATM, I thought the freeimage issue now was the double declarations between the headers
<jonsger>hm, something is wrong with the po/texi generation:
<davexunit>civodul: I think I'm missing context
<jonsger>civodul: I think introduced a bug: en@boldquot.po:2849: format specifications in 'msgstr[0]' are not a subset of those in 'msgid_plural' and en@boldquot.po:2856: format specifications in 'msgstr[0]' are not a subset of those in 'msgid_plural'
<jonsger>nckx: thanks
***nckx is now known as bug_bot
***bug_bot is now known as nckx
<idnull>is it possible to bundle service with a package? I want to package iNET Wireless Daemon (it can substitute wpa_supplicant and works with connman and NM)
<idnull>package with a service*
<roptat>you typically define a service that uses a package, not the other way around
<idnull>ok, so firstly I create a package, then service, then I put package as a service dependency?
<roptat>as a simple example, you can have a look at the opensmtpd service for instance
<roptat>or my initial version of
<jetomit>idnull: I just started packaging iwd yesterday :) not sure when I'll have the time to finish, so maybe you can use as a starting point
<jetomit>ell should be OK, I haven't figured out the dbus bits for iwd yet
<idnull>New Moon is working, but with an issue tar (child): palemoon-: Cannot open: No such file or directory
<idnull> so it can
<idnull>'t do something with archive (maybe create one)
<nckx>idnull: You can try strace'ing New Moon (or its children).
<rain>heyyy, so, is there a nice way to make the regex in substitute* into a literal? like, all the special characters that have special meanings in regexes, i wanna opt out of them.
<nckx>rain: So like a REGEXP-ESCAPE procedure? I dunno.
<rain>nckx, hm. then i guess i'll just have to figure out scheme regexes. :/
<nckx>Welp, it's literally called regexp-quote 🙂
<nckx>rain: ‘Scheme regexes’? What's different about them?
<rain>nckx, oh nice, it exits. well, i'm not sure if they are compatible with any regex i know. (eg. Lua has a very different (and simpler) regex implementation compared to Perl)
<nckx>I've always considered non-Perl regexes to be ‘normal’, and Perl regexes to be, well, not.
<nckx>Doesn't Lua call them patterns, not regexes, to avoid exactly this confusion?
*nckx agrees there are too many ‘regexes’ though.
<nckx>Guile's regexes are bog-standard to me but it's the double-escaping that makes them fun™!
<rain>nckx, that looks like a group of people raising their hands (like the \o/ smiley)
<nckx>I thought of a marathon runner crossing the finish 🙂
<nckx>‘Yay, finally wrote a regexp that works.’
<civodul>we need srfi-117!
<civodul>ah no, wrong number
<civodul>well anyway
*civodul tries again
<civodul>we need srfi-115!
<nckx>Or a (match-regexp (perl-regexp "…")), (scheme-regex "…"), … syntax. All the regexes in one codebase! Imagine the fun.
<nckx>civodul, rain: Thanks though, I had no idea there was really such a thing as a ‘Scheme regexp’ (sigh).
<rain>welp, i'm trying to escape but it seems to affect an (assoc-ref) call somewhere for... some unknown reason...
<rain>any idea why these regexp-quote calls introduce an error?
<rain>nckx, i think regexp-quote isn't quoting something properly: (make-regexp "LIBCLANG_LIBS += $${CLANG_LIB}" 1) Invalid preceding regular expression
<nckx>(regexp-quote "LIBCLANG_LIBS += $${CLANG_LIB}") → "LIBCLANG_LIBS [+]= \\$\\$[{]CLANG_LIB[}]"
<nckx>Which make-regexp accepts without complaint.
<nckx>At the REPL.
<nckx>rain: Not saying you're wrong, but I don't see what led you to that diagnosis. OTOH, I'm working, not in the backslash-counting zone.
<nckx>Does it work outside of the (substitute* …) syntax? In a (let …)?
<nckx>`(,( sometimes fucks with macros in ways that aren't entirely clear to me.
*nckx → wurk.
<rain>nckx, it doesn't seem to :/
<rain>but i gotta go, class is over
<rain>thanks for the help! ^u^
<nckx>rain: 😃
<nckx>You're welcome. Good luck.
<nckx>sneek: later tell rain: It's always helpful to include the actual error message in addition to your code: after downloading & building your package, Guix said: Wrong number of arguments to #<procedure assoc-ref (_ _)>
<sneek>Will do.
<nckx>sneek: later tell rain: which 1) is suspiciously relevant for a Scheme error message, what's going on and 2) immediately makes the obviously wrong ‘(assoc-ref inputs "clang" "/lib"))’ obvious.
<sneek>Will do.
*nckx is unsure whether rain2 is rain.
<rain2>different person
<sneek>Welcome back rain2, you have 2 messages.
<sneek>rain2, nckx says: It's always helpful to include the actual error message in addition to your code: after downloading & building your package, Guix said: Wrong number of arguments to #<procedure assoc-ref (_ _)>
<sneek>rain2, nckx says: which 1) is suspiciously relevant for a Scheme error message, what's going on and 2) immediately makes the obviously wrong ‘(assoc-ref inputs "clang" "/lib"))’ obvious.
<nckx>Well, you fooled sneek 😃
<civodul>nckx: srfi-115 is not really about "scheme regexps", it's about providing a non-stringy DSL for regexps
<civodul>thus without ambiguities, special characters, special characters that aren't that special, and all that
<civodul>IOW, it rocks
<nckx>civodul: OK, now that sounds sane.
<nckx>As in not reinventing/forking regexpen yet again. Whether '(+ (or space punct)), '(* digit), or '(: "*" ($ word) "*") are more or less readable, well…
<nckx>I obviously still need a bit more time in the Scheme re-education mines.
<civodul>one can do regexp->sre... :-)
<nckx>I got to the ‘SRE Syntax’ syntax header right after typing that.
<nckx>s/syntax header/header/
<nckx>civodul: Right, but I assume your ‘without ambiguities, special characters, special characters that aren't that special, and all that’ advantages go back out of the window then, no?
<nckx>It's interesting but I don't grok the ‘rock’ yet. I'm gonna read this straight through when I have the time to do so 🙂
*civodul regained SIEVE access, mail filtering should improve \o/
<refpga>Hello, I need to edit the source code of a package before installing it (using the same compilation process). Any ideas on how to proceed?
***Digitteknohippie is now known as Digit