IRC channel logs


back to list of logs

***daviid is now known as Guest51516
***Guest51516 is now known as daviid
<civodul>mange: you mean password authentication?
<civodul>never tried that
<mange>It should authenticate however SSH does (ie. in my case using my private key).
<mange>It turns out my repo was also available over unauthenticated HTTPS, so I'm just using that instead.
<civodul>well, it'd be worth investigating at some point
<reepca-laptop>Hm, so my desktop just sort of froze up... can't ssh to it, much less ping it. Does a kernel panic always result in a black screen with a stack/register dump? Mine still have what was last displayed on them, and the last quarter second or so of audio repeating on loop. Anyone else had a problem like this, or is my hardware just breaking down?
<reepca-laptop>also, are there any particular diagnostics I should try before restarting it?
<reepca-laptop>... I think I got that mixed around. "Can't ping it, much less ssh to it" is what I meant.
<reepca-laptop>also, this is the second time in the past week or so that this has happened
<reepca-laptop>I don't think I've reconfigured in awhile, so I'm doubtful that has caused it. The freeze up only seems to happen when I'm there (considering how long it sits idle with me not there and has never frozen up, it seems very likely that it's related to my input somehow)
<reepca-laptop>I can't find any abnormalities in /var/log/messages, according to it the system shut down normally (It says: Oct 17 19:41:51 localhost syslogd (GNU ineutils 1.9.4): restart)
<reepca-laptop>makes me think it may be a (strange) userspace issue, which would make sense, since I have actually pulled and updated packages recently.
<brendyyn>back from china. guix pull worked fine after 5 weeks :)
<brendyyn>and then run guix package -u ., updated 219 packages without any error, I'm very impressed.
<laks>Hi, Where I am trying to do the Binary Installation step for GUIX, I'm getting the following error on verifying the .sig file gpg: can't open 'guix-binary-0.15.0.system.tar.xz.sig': No such file or directory
<laks>gpg: verify signatures failed: No such file or directory
<laks>I tried to get the public key, which gives warnings like gpg: WARNING: unsafe ownership on homedir '/home/lakshmi/.gnupg'
<laks>gpg: /home/lakshmi/.gnupg/trustdb.gpg: trustdb created
<laks>gpg: key 090B11993D9AEBB5: public key "Ludovic Courtès <>" imported
<laks>gpg: no ultimately trusted keys found
<laks>Could someone help me here ? Running the .sh script to install GUIX, I crashed my ubuntu system.
<g_bor>hello guix!
<roptat>hi guix!
<jlicht>hey guix!
<roptat>hey jlicht, I've been playing a bit with your node-build-system and node importer, and it's great!
<roptat>I've modified the importer to use inputs instead of propagated-inputs (since inputs are included in the package output)
<roptat>and I've modified the install phase to only link to inputs instead of inputs and native-inputs
<roptat>more precisely, to link to dependencies instead of devDependencies by parsing the package.json file
<roptat>I've also modified the way binaries are installed and I think we can drop global?
<jlicht>roptat: awesome! What did you do regarding how binaries are installed?
<roptat>parsing the "bin" entry in the package.json file
<roptat>it can contain an array of couples (name - file) or just a filename (in which case the binary is called the same as the node package name)
<roptat>when there is no "bin" and global? is #t, I do the same as before
<jlicht>yeah, I think global? is on longer needed in that case. I'm happy the linking situation seems to work for you :-)
<jlicht>and of course the linking allows us to have inputs instead of propagated-inputs, very nice. What do you do with devDependencies now?
<roptat>I just ignore them
<roptat>they are linked for the build, then I remove the node_modules directory and I relink everything after installation
<roptat>but I still have to find a package I can run the testsuite of...
<roptat>why is the test phase after the install phase?
<jlicht>roptat: that is an artifact of the earlier way of doing things, with your linking-based version, it can be moved back
<roptat>ok, I'll do that
<roptat>also I've rebased your patches, like I introduced (guix build json) wihh the patch for the node-build-system because it's required there
<jlicht>roptat: do you have your work available somewhere?
<roptat>not yet
<jlicht>roptat: awesome. Do you think we should/could add the minimisation step as well?
<roptat>does npm do that?
<jlicht>not by default, but we have a `minify-build-system' which already does this
<roptat>ok, we could do that too
<jlicht>regarding testing frameworks; we could use the prebuilt version as hosted on the npm registry to try to run some tests for the packages you already had a look at
<roptat>I could add mocha rather easily
<jlicht>I think janneke created a modified version of the importer that has a "binary" flag to simply get the sources from npm
<roptat>other test frameworks have a lot more dependencies
<roptat>I don't really want to depend on pre-built packages, even if only for tests
<roptat>also the situation is a bit different because we all have the same code
<roptat>for compiled programs it's useful that distributions test the version they built, because developpers have a different compiler, library versions, ...
<roptat>but in javascript, we all get the same thing, so developpers and distributions actually test the same thing
<roptat>so it doesn't feel as important that we can run tests for npm packages
<jlicht>it was suggested to use the "binary" testing packages to bootstrap to the actual testing packages
<jlicht>that would probably still involve several thousands of packages for anything that is not mocha, but still
<roptat>I'm going to eat
<jlicht>enjoy :-)
<kristofer>hello guix!
<kristofer>I'm trying to setup a mail server with moderate success. My issue now is opensmtpd will not authenticate with pam? I see in guix/gnu/packages/mail.scm that opensmtpd has the --with-auth-pam configure flag.
<kristofer>investigating guix/gnu/services/mail.scm I find that dovecot-service-type extends pam-root-service-type with (unix-pam-service "dovecot") -- opensmtpd does not, but should it?
<roptat>not sure... do you know how to try?
<kristofer>it's been awhile. is there a document about using guix from a git clone? the perfect setup or something like that
<roptat>yes exactly
<roptat>let me find it
<kristofer>got it
<roptat>( has the manual for a recent master, while has the manual for the last release)
<roptat>if it works, we'd appreciate a patch or at least a bug report :)
<g_bor>kristofer: I did not try to this yet, but it seems that it should extend pam-root-service-type.
<g_bor>kristofer: see this patch for an example
<snape>kristofer: you can also authenticate with 'table passwd file:/etc/mail/passwd', as an alternative.
<sneek>Welcome back snape, you have 1 message.
<sneek>snape, civodul says: i think Cuirass should register GC roots for its evaluations under /var/guix/gcroots
<snape>I find it easier to deal with when emails are not correlated with the Unix users
<roptat>silly question, how do I select a numbered argument in format?
<roptat>for instance in (format #t "~a ~a" "a" "b"), I print "a b" but I'd like to print "b a" without changing the order of "a" and "b"
<roptat>(the reason I want that is in a translation it makes more sense to reverse the order of two arguments)
<kristofer>I have built guix with some changes to the opensmtpd service. What is the best way to go about testing it with my existing mail server?
<kristofer>I also have a couple questions. My preference would be to run certbot --standalone in this particular case, rather than extend the nginx-service-type configuration with the location /.well-known {} block/redirect to https on / but I'm not sure what the best approach is for that. It's not really a service. wdyt?
<bavier>roptat: you can use the '~*' formatter for argument jumping
<bavier>roptat: see the section on (ice-9 format) in the manual
<AlishbaKabeer>can I install this package manager on slackware?
<kristofer>AlishbaKabeer: use the binary install
<AlishbaKabeer>and then I can install anything right?
<AlishbaKabeer>with this package manager?
<kristofer>you could install anything packaged with guix
<AlishbaKabeer>how much qouta is required for a minimalist guix installation on x86_64 included the iso?
<kristofer>It can get large fast
<kristofer>depends on the packages you want to install, and whether or not there are binary substitutes available for them
<AlishbaKabeer>kristofer: I think I know you. saw you in many channels. do you use this distro?
<kristofer>yep! it's fun if you like to tinker with stuff
<AlishbaKabeer>kristofer: how would you describe this distro?
<AlishbaKabeer>tinkering? well I love that.
<kristofer>it's progressive? I dunno really, there's a lot of really neat ideas that float around here
<AlishbaKabeer>I would install a WM only as I'm low on quota
<kristofer>it's awkward though, unlike any other distribution I've used. It makes hacking on nearly everything pretty easy
<kristofer>maybe unfamiliar is a better characterization than awkward
<AlishbaKabeer>How do you compare it with arch?
<kristofer>I'm not really familiar with arch. I was a debian user. I found guix because of an interest in scheme and guile.
<roptat>bavier: thanks, it works! now, will msgfmt let me do that? I'll try :)
<AlishbaKabeer>kristofer: I just heard from a friend about it on irc
<kmicu>As a former Arch user I’m not using Arch anymore. 😺
<kmicu>(In other words Guix(SD)/Nix(OS) fixed many pain points in Arch+Pacman though Arch+Guix/Nix is also a pleasure to drive.)
<vagrantc>hrm. just updated a machine that hasn't been booted for a few months (last guix pull was in mid-june) ... but i get an error trying to update the system:
<vagrantc>sudo -E guix system reconfigure --keep-going --fallback /home/vagrant/agxsd.scm
<vagrantc>guix: system: command not found
<vagrantc>this is an guixsd aarch64 system, fwiw
<vagrantc>ah, export PATH="/home/vagrant/.config/guix/current/bin${PATH:+:}$PATH"
<vagrantc>missed that message
<vagrantc>figured i had already done that, since "guix pull -l" listed three prior generations
<kristofer>is it reasonable for me to run ./pre-inst-env guix system disk-image
<efraim>vagrantc: the other option is to prefix with 'guix environment guix --'
<vagrantc>is there a way to get all the currently unset but recommended environment variables?
<vagrantc>after various guix commands, it occasionally recommends setting some variable ... but is that state-dependent on the running action and hard to guess later?
<vagrantc>it's easy to miss one of those messages
<kristofer>vagrantc: I think if you were to logout and login the new session would be updated
<vagrantc>kristofer: if it's supposed to, it's not
<kristofer>vagrantc: do you set PATH in .bashrc or some other place possibly interfering with your profile?
<vagrantc>i don't think i've successfully updated the system since the switch to the newish method with guix pull generations
<vagrantc>so /etc/profile isn't including anything to do that
<kristofer>maybe you could specify a commit in between and step your way up to current?
<kristofer>guix pull --commit=
<kristofer>I don't know what the experts recommend
<vagrantc>guix pull completed successfully
<vagrantc>hopefuly the system successfully reconfigures and then it works after reboot :)
<mbakke>The Swig Guile tests are failing for i686 (on core-updates).
<AlishbaKabeer>I don't have ethernet cable. I have usb dongle for internet connection. does this iso have got the drivers to detect it?
<moaz>depends on the chipset
<AlishbaKabeer>I was able to install arch using same usb dongle.
<AlishbaKabeer>but arch iso also supports wifi
<moaz>were you able to install debian normally?
<AlishbaKabeer>slackware doesn't. no wifi, no usb dongle.
<moaz>or did you need to download the firmware separately
<AlishbaKabeer>yeah with netinstall iso, I was able to install debian
<moaz>and it didn't ask you for separate firmware or anything?
<AlishbaKabeer>for wireless I have to download firmware.
<AlishbaKabeer>for installation, I used usb ethernet
<moaz>um, yeah, if you needed the firmware then guix won't work with your wireless adapter
<AlishbaKabeer>I'm asking for installation only. as I think installation media pulls packages
<moaz>nonfree wireless firmware isn't included in guix repos
<AlishbaKabeer>Once I'm done with installation I can download and install firmware
<moaz>there is a bug in the linux-libre kernel (used by guix) that causes nonfree firmware to not load
<AlishbaKabeer>so I shouldn
<AlishbaKabeer>I shouldn't try this ?
<moaz>yeah, I guess not
<moaz>do you have an old phone lying around?
<moaz>you can use that as a wifi adapter using USB tethering
<AlishbaKabeer>yeah but cable isn't working.
<moaz>I see
<AlishbaKabeer>my usb dongle is came from huawei
<AlishbaKabeer>can I configure mirrors while installation? This iso download has too slow speed. 40kbps.
<rekado>sneek: later tell civodul I don’t know how to process metapost files. For metafont we run “mf” on the files in build phases. Maybe you need to run “mpost” in a similar fashion.
<AlishbaKabeer>here is my partition table. is it okay to just mkfs / and mount it /mnt for guix install?
<AlishbaKabeer>I'd also mkfs on esp.
<AlishbaKabeer>moaz: kristofer ^^
<moaz>it seems okay to me
<jic>Hey guys do you know what bournish@guile-user is? Because when I run GuixSD, it always auto-start, and I need to type ,quit to continue booting
<mbakke>jic: Bournish is the "rescue shell" in GuixSD. If you drop to it, it means something failed to start.
<jic>ops, mmm
<jic>ok thanks mbakke i'll check it
<mbakke>jic: Perhaps `herd status` in the booted system can provide some clues.
<jic>herd status shows started: + + + + and stopped: - -
<jic>stopped just: term-auto and user-homes
<jic>but I needed to run it with sudo
<mbakke>jic: That is normal. Maybe /var/log/messages has some information?
<rekado>I found it puzzling that “guix pull” compiles Python.
<mbakke>rekado: Grafts work in mysterious ways.
<efraim>speaking of which, pango@1.24.4 is 'define-public'
<mbakke>efraim: Is that a problem?
<efraim>probably not, we normally make the replacement packages hidden
<mbakke>efraim: python-fonttools has ~2700 dependent packages.
<mbakke>I am considering reverting Swig too, which ticks in at ~1700 dependents. But, the fix looks important.
<efraim>according to hydra python-fonttools is failing on 3 arches on core-updates
<mbakke>efraim: Nothing is using that variant, though.
<rekado>huh, and now it’s building python 2.7.
<mbakke>Perhaps it would be better to define a different variable, and "hide" the buggy version.
<efraim>i see
<efraim>i'll go ahead and revert it
<mbakke>efraim: Thanks!
<rekado>computing the Guix derivation should not require Python. Am I missing something?
<mbakke>rekado: I think the Python dependency comes from libgit2.
<mbakke>Which is used by Guile-Git, which in turn is used by Guix.
<moaz>AlishbaKabeer says she is unable to boot
<moaz>She selects the option in her BIOS's boot manager, and it flashes, then goes back to said manager
<moaz>any ideas?
<moaz>nvm. she figured it out - she had misflashed the image
<rekado>mbakke: this makes sense. A bit sad to have such a big dependency.
<thomassgn>I'm getting 'Unbound variable: ungexp' from the following in a package definition: (string-append "-DCMAKE_AR=" #$binutils "/bin/gcc-ar"); the context is cmake configure flags. I have ' #:use-module (guix gexp)' at the top of the file and 'gnu packages base' is also "used"/imported...
<thomassgn>any tips for continued troubleshooting?
<rekado>thomassgn: use unquoting instead of gexps here.
<thomassgn>ah, so (unquote binutils)? or (unquote (string-append ...))?
<rekado>I’d write ,(string-append …)
<thomassgn>ah, the comma is sugar for (unquote ...)? like ' (or is it `) is for quote?
<rekado>“'” is “quote”, “`” is “quasiquote”, and “,” is “unquote”.
<rekado>you can only unquote in a quasiquoted expression.
<rekado>I like to think of ` and , as two states of a switch: up is data mode and down is code mode.
<thomassgn>hah, nice. Thanks :)
<thomassgn>Now it tells me it is getting a package and not a string from binutils... in ,(string-append "-DCMAKE_RANLIB=" binutils "/bin/gcc-ranlib"))))
<thomassgn>ie. type error. gexps don't seem to make a big difference.
<thomassgn>I guess I need to get the path somewhere else? or maybe add binutils as an input...
<thomassgn>heh. Have to admit I'm fumbling at this point :P
<rekado>oh, sorry, you need to get those from the %build-inputs.
<rekado>Use (assoc-ref %build-inputs "binutils") instead of just binutils
<rekado>and then don’t do any unquoting.
<rekado>you can search for %build-inputs in gnu/packages/bioinformatics.scm for examples.
<thomassgn>ah, cool. :) will do.
<mbakke>rekado: libgit2 does not retain a Python reference, so I suspect it comes from somewhere else, actually.
<EternalZenith>Do unquotes and unquote-splices work outside of quasi-quotes?
<thomassgn>rekado: awesome, now it builds again. :)
<thomassgn>If it works now, we might have openclonk on guix soon. :)
<rekado>ah, nice!
<rekado>I wanted to try it some years ago but never got around to it
<AlishbaKabeer>I had slackware installed before now I want to install jio u i x
<AlishbaKabeer>I had an encrypted in partition table
<AlishbaKabeer>I have formatted the first partition with fat
<AlishbaKabeer>Now I have open the second partition and with cryptsetup
<AlishbaKabeer>So root partition now please guide me how should I format the root partition
<vagrantc>are there any aarch64 builders right now, or just running behind?
<EternalZenith>I'm having trouble getting i3 or i3-gaps to build
<EternalZenith>I keep getting this error: "/gnu/store/f95ghy8mx00fc22nrvswvnpqlfdkf2nk-module-import/guix/build/utils.scm:616:6: In procedure invoke:
<EternalZenith>Throw to key `srfi-34' with args `(#<condition &invoke-error [program: "autoreconf" arguments: ("-vif") exit-status: 127 term-signal: #f stop-signal: #f] 9e0f40>)'."
<EternalZenith>Here's the package definition:
<bavier>EternalZenith: autoconf is failing, does it indicate why?
<pkill9>i'm curious why it's running autoconf, it doesn't do that by default
<bavier>pkill9: gnu-build-system does that now
<bavier>but iirc only if it needs to
<pkill9>oh looking at the source, it does need to be run
<EternalZenith>Is it alright to paste longer errors in here, or should I use a pastebin-ish service?
<EternalZenith>It's only a dozen lines or so
<bavier>EternalZenith: a pastebin would be best
<pkill9>put it in EternalZenith
<EternalZenith>I was just about to ash which one, thanks
<pkill9>i think it's unable to find autoreconf
<EternalZenith>I'll try to add the module
<pkill9>yeah, try adding autoconf package to native-inputs
<EternalZenith>Shouldn't it be part of the gnu-build-system?
<pkill9>i would have thought so, but it's worth a try
<EternalZenith>I assume it's part of the autoconf package?
<tune>can someone package "terasology"?
<EternalZenith>Aha, that did something
<EternalZenith>Although it looks to be giving the same error
<EternalZenith>Now it's a problem with aclocal that's messing up autoreconf
<pkill9>what's the error?
<EternalZenith>One sec, I think I fixed it
<EternalZenith>I had to add automake to native-inputs too
<EternalZenith>It worked!
<EternalZenith>Thing is, I now have i3 and i3-gaps installed simultaneously
<EternalZenith>I assume the user's profile overrides the system one?
<pkill9>nice, i think there is one benefit to having to add those inputs actually, it makes it simple to see if a package needs to generate a configure file, which could help wiht debugging
<bavier>tune: we can add it the wishlist (
<pkill9>yeah it does i think EternalZenith
<bavier>tune: in the meantime, guix has a "minetest" package
<tune>Yeah, I'm familiar with minetest, but not much with terasology. I was hoping to compare them a bit.
<tune>Didn't know about the wishlist, that's cool.
<EternalZenith>I'm building up several package definitions I feel I should submit to the official repository
<pkill9>i've built up a bunch of package definitions i should submit to the official repository also
<EternalZenith>pkill9: Do you want to submit your polybar definition? I have the one I made that I'll try to submit later if you don't want to
<pkill9>nah it's fine, submit yours, i never get round to submitting mine
<EternalZenith>Alright, I want to get a few more things packaged and then I'll submit a bunch of packages at once
<EternalZenith>A couple of these seem to be more tricky
<thomassgn>huh, after digging a bit I find the linker doesn't find pthreads... Isn't pthreads part of GCC?
<bavier>thomassgn: it needs the '-pthread' linker flag
<moaz>Is it okay to replace gnome with i3 in configuration.scm?
<EternalZenith>How do you download and get the hash of a git repository at a certain tag?
<EternalZenith>Is it 'guix hash -r'?
<thomassgn>bavier: do you mean 'c++ ... -lpthread' on the CLI or -pthread flag as part of the package def?
<thomassgn>the ld/c++ command fails because 'ld: cannot find -lpthreads'
<pkill9>EternalZenith: `guix hash -xr`
<pkill9>the -x makes it not include the .git files i think
<EternalZenith>Thanks, that seems to have fixed an error I was attempting to find
<rekado>sneek: later tell civodul The refcard is beautiful! I’d love to customize it. Will you push the sources to the maintenance repo?
<sneek>Will do.
<EternalZenith>Whenever I try to run "guix gc" It fails after several seconds with "guix gc: error: build failed: executing SQLite statement: FOREIGN KEY constraint failed"
<user_oreloznog>moaz: No. Insert instead (services (cons* (gnome-desktop-service)
<EternalZenith>Does guix use libinput as the default input driver?
<EternalZenith>Why doesn't libevdev include the C header files?
<EternalZenith>I can't get interception-tools to compile since it cannot find the necessary header files from libevdev, which apparently does not provide them in its only output
<rekado>EternalZenith: it does provide an “include” directory.
<rekado>EternalZenith: the headers are in a sub-directory, so you may have to tweak the include path.
<EternalZenith>Would that require patching the package?
<rekado>this depends on the build system used by that package
<EternalZenith>It's just the regular cmake-build-system
<EternalZenith>It's looking for them in "libevdev/libevdev.h" and "libevdev/libevdev-uinput.h"
<EternalZenith>Should the source be patched to have those changed to "libevdev/include/libevdev.h" and "libevdev/include/libevdev-uinput.h"?
<EternalZenith>When I'm changing the path to the header files, what should the path be relative to?
<EternalZenith>rekado: I fixed part of the issue, but now the files included from libevdev are giving me the same error: "In file included from /tmp/guix-build-interception-tools-0.1.1.drv-0/source/udevmon.cpp:24:0:
<EternalZenith>/gnu/store/a66ld37hcwj5gv69g1pgwi84wg3ydgdw-libevdev-1.5.9/include/libevdev-1.0/libevdev/libevdev-uinput.h:30:31: fatal error: libevdev/libevdev.h: No such file or directory"