IRC channel logs

2020-10-05.log

back to list of logs

<nckx>Good night, sweet #guix.
<helaoban>I'm running into a build error trying to package hledger. running 'guix import hackage --recursive hledger@1.19.1 > hledger.scm' gives me package definition for hledger along with all of its dependencies that aren't yet included in guix.
<helaoban>The build fails while trying to build one these dependencies, specifically ghc-cassava.
<helaoban>here is the error log: https://paste.debian.net/plain/1165907
<helaoban>looks like a dependency conflict with the bytestring package.
<helaoban>can anybody here with haskell experience point me in the right direction?
<cypherclerk[m]>hello I'm sponsoring development of some guix packages https://ethlance.com/#/job/437
<cypherclerk[m]>please contact me if you are interested
<joshuaBPMan>Sweet! My gnucode email address can now receive email...I haven't figured out how to send email yet, but at some point I will!
<ryanprior>What's gnucode?
<ryanprior>cypherclerk: I can package that, I'll email you.
<joshuaBPMan>ryanprior gnucode.me
<joshuaBPMan>That's the domain that I own.
<ryanprior>Oh cool, nice site!
<joshuaBPMan>ryanprior thanks. It's running guix system too.
<joshuaBPMan>config is here: https://notabug.org/jbranso/linode-guix-system-configuration/src/master/linode-locke-lamora-current-config.scm
<ryanprior>That's awesome, thanks for sharing.
<joshuaBPMan>ryanprior: thanks. I've only really got the website working. Everything else still doesn't work as well as I'd like.
<ryanprior>I haven't had any experience with prosody or opensmtpd on any platform much less Guix!
<ryanprior>But I hope you can get them figured out.
<ryanprior>I would like to get experience with running services using Guix and Shepherd. I'm not particularly interested in using Guix System right now, but if I could define services that run in Docker that would be great.
<joshuaBPMan>ryanprior Me too. I've been putting up vidoes of me trying to set up email recently
<joshuaBPMan> https://video.hardlimit.com/accounts/joshua_branson/video-channels
<joshuaBPMan>ryanprior may I ask why you prefer docker? With guix you could define update-able services. Docker is frozen pizza development
<ryanprior>Oh yeah I looked at one of your videos the other day. My patience is very short for video and podcast content though, so I'll have to ask your forgiveness that I didn't watch very long ':D
<ryanprior>Docker containers are a snapshot, like a Guix profile.
<ryanprior>Complaining about not being able to update a Docker container is like complaining about not being able to update a profile - they're cheap and disposable, move on and create a new one.
<joshuaBPMan>ryanprior No need to ask for forgiveness. Those videos are going for quantity not quality. I don't even watch them :)
<joshuaBPMan>ryanprior I guess I thought, the way docker does things makes it hard to update one. With guix you can specify all of your dependencies. And update them all too.
<ryanprior>Docker container images are immutable, you can't update them whatsoever. They are stored using a persistent immutable data structure to save space, so that you can create a new image based on an old one and only save the bits that changed.
<ryanprior>Docker is agnostic to whether you use a declarative, pure process like Guix to build your image or whether you use imperative scripts and blobs. It treats all that as just being data.
<joshuaBPMan>fair enough
<ryanprior>Docker's concept of "tags" is like Guix's concept of "profile generations," it's fast and space-efficient to create new tags and saving the old ones lets you roll back updates or time-travel to previous images.
<ryanprior>There's lots of parallels like that. So I see Guix as being a nice complementary technology to Docker.
<ryanprior>Docker images are content-addressed by their hashes, just like Guix profiles, and they can be automatically garbage-collected too.
<ryanprior>Now I'm thinking I should do a talk about Guix for Docker users, might bring some more people into the fold.
<gnutec>Two news. IceCatMobile is great and GNU Taler is avable in f-droid.
<ryanprior>Ooh cool, I've been looking forward to trying Taler. I know they've been undergoing an audit, I haven't heard how that's going.
<gnutec>I just think people should know. ;) See ya!
<bdju>got 504 error trying to download the "latest" guix iso
<bdju>anyone know if intel I211 gigabit nic works with linux-libre? or how to check
<bdju>a friend had non-working ethernet on guix in the past and is hesitant to put it on a new machine now because of that
<joshuaBPMan>bdju you could check h-node.org
<bdju>ah, good idea. I had only used it for WLAN stuff before. thanks
<bdju>I hear sway is not an option in the guix installer. any plans to add it?
<nalaginrut>how can I install gcc 32bit?
<nalaginrut>or I may ask in another way, how can I install a package in system i686-linux
<baconicsynergy[m>I just wanted to stop by this channel and say I miss you guys so much <3
<nalaginrut>I want to compile with -m32, usually I need to install glibc-i686
<baconicsynergy[m>I'm going to be building more Guix machines in the future hopefully! I'm excited to see how far things have come the last time I was here
<nalaginrut>nice!
<nalaginrut>oh, I see, guix install gcc -s i686-linux
<r-gt>is all guix code in scheme?
<nalaginrut>r-gt: I don't think so, the core part is nix, which is written in C
<r-gt>hmm, I thought nix was done in c++ ?
<nalaginrut>which is written in C
<r-gt>oh ok
<nalaginrut>wow, is my answer cut by the IRC client?!
<r-gt>I see "which is written ..." twice
<nalaginrut>now it looks ok
<nalaginrut>r-gt: sorry, it's in C++
<mroh>good morning, #guix!
<janneke>good morning mroh!
<kmicu>( ^_^)/
<civodul>Hello Guix!
<civodul>Hello Guix!
<janneke>hello civodul!
<mroh>hello civodul!
<civodul>damn i'm repeating myself
*civodul not completely awake
<civodul>the latest icecat crashes for me, is that a known issue?
<Brendan[m]2>mine doesn't, but i get javascript errors
<Brendan[m]2>did you run icecat from a terminal to see if there are any errors
***apteryx is now known as Guest73071
***apteryx_ is now known as apteryx
***stikonas_ is now known as stikonas
<civodul>Brendan[m]2: yes, it just segfaults right away, before it displays anything
<Brendan[m]2>is it not possible to join #guile via matrix like #guix? I could not find it via the same public servers search
<efraim>hello everyone!
<efraim>`wc -l gnu/packages/*.scm | sort -n -r | head` it might be time to split a few more modules. I'm poking crates-io ATM
<civodul>hey efraim! why not, indeed
<efraim>I bumped librsvg-next to 2.50.1 and so far moved some gtk crates to their own modules. I think the next one will be graphics related
<civodul>M-x guix-locations says that the biggest files (in number of packages) are: 1. crates-io, 2. lisp-xyz, 3. python-xyz, 4. cran, 5 emacs-xyz
<civodul>graphics-related prolly makes sense, if there are enough of them
<mfg>at least the xyz files could just be split into for example a-f and g-z?
<mfg>so just sort and split
<efraim>does it take longer to compile based on number of lines or number of packages?
<efraim>I think themed makes the most sense, even if it's only a dozen at a time
<civodul>efraim: -O2 is roughly linear in the number of lines, but i think the new -O1 isn't
<civodul>or at least with smaller factors
<civodul>i agree with themed module names
<Brendan[m]2>i wonder how difficult it would be do develop a refactoring tool that could automatically reorganise all the modules based on some heuristics
*mothacehe added support for --image-type=pine64-raw
<janneke>mothacehe: ooh, nice!
<mothacehe>volatile image support is broken due to some overlayfs shenanigans though
<mothacehe>next step --image-type=pinebook-pro or --image-type=mnt-reform :)
<janneke>ah, right
*janneke is working on a patch for gettext-minimal for hurd
<jeko>hey ! I am playing with guix system vm
<janneke>mothacehe: are you planning on changing the hurd-hello job/name?
<jeko>and in my freshly created vm when I guix pull, I got a message : guix pull: error: changing ownership of path '/gnu/store': Read-only file system
<mothacehe>sure I'm on it
<mothacehe>janneke: I guess you saw: https://ci.guix.gnu.org/build/3259477/details
<mothacehe>is 'hurd-master' fine ?
<janneke>yes, that's fine with me, civodul?
<janneke>no, i didn't look into the failure -- just saw it failed
<janneke>does it say which dependency failed?
<mothacehe>it seems that all the dependencies listed in red failed for different reasons
<civodul>janneke, mothacehe: "hurd-master" seems to be working as expected
<civodul>i saw the list of build failures, yes
<civodul>now we know what to work on!
<mothacehe>jonsger: you can zcat /tmp/cuirass-tests/var/log/cuirass/evaluations/3.gz to have more info
<jonsger>oke
<janneke>oh, they are all failing depedencies!
<janneke>weird
<mothacehe>janneke: rename done
<janneke>mothacehe: thank you!
<jeko>Hm ok I think I just read in the doc why I encounter this behavior... haha
<civodul>thanks, mothacehe!
<civodul>SQL rulez :-)
<mothacehe>haha starting to master sql against my will
<civodul>heheh
<jlicht>hey guix!
<efraim>got crates-io.scm down to 28563 lines, still the largest
<Brendan[m]2>what did you do? delete old unused crates? split it?
<efraim>split off some crates into a new module
<Brendan[m]2>did you merge wip-swc first?
<Brendan[m]2>i hope it doesn't become a pain to merge that now if everything is moved
<Brendan[m]2>also, i notice many use the same snippet for finding libclang. perhaps that could be abstracted out.
<efraim>Brendan[m]2: oh, I haven't merged that. hopefully it doesn't cause too many problems :(
<Brendan[m]2>i would be happy if you could. my next contributions builds on that branch because it has added 2 or 3 packages im using, so it will be convenient for me. i can move back to master and send patches for upstream master
<efraim>i'll start taking a look at the branch
<Brendan[m]2>i think some patches did some incidental indenting of surrounding code, but nothing terrible
<raghavgururajan>Hello Gu...Gu...Gu...Guix!!!
<zzappie>Helloo!1
<efraim>Brendan[m]2: overall looks good. rust-diesel-derives is 1.4, not 1? I don't understand rust-rustc-hash-1.0, I think the zstd crates bundle zstd (hence the version number), syn-test-suite should be 0.0
<Brendan[m]2>is a mistake perhaps? fix it up?
<pinoaffe>I want to do things such as `guix package -m blarg -p qwer` from a guile script, is there a better way to do this than running the corresponding shell command?
***stikonas__ is now known as stikonas
<nckx>Mernin gerks.
<nckx>Brendan[m]2: ‘/join #freenode_#guile:matrix.org’ works fine for me. Not familiar with how search works (or doesn't).
<rekado>pinoaffe: yes, you can use the commands from (guix scripts) directly
<rekado>e.g. (use-modules (guix scripts package)) (guix-package my-args)
<rekado>that’s almost exactly the same as calling the shell command, except that you don’t have to involve the shell.
<rekado>the invocation is very similar to that of the shell command
<helaoban>hey guix, is there a way to scope an instance of postgres to a development environment à la docker, eg if I wanted a postgres db for my app that gets setup on 'guix environment' with the required pg users ddeclared upfront, which then gets destroyed when I exit the environment shell?
<helaoban>it looks like containers are the right mechanism but after reading a bunch of the docs I'm not sure what the optimal setup for this might be.
<davexunit>helaoban: not currently. supporting this use-case is something I had in mind when I wrote the initial version of 'guix environment' but it was never implemented.
<davexunit>I was hoping to use containers + shepherd to make it work.
<greg>Is it possible to view derivations on Cuirass? I've pulled to the same commit as a successful ci build (http://ci.guix.gnu.org/build/3256575/details) of a fio dependency which is requiring a build on my local system, but am seeing different hashes for the derivations.
<greg>I don't understand why Cuirass is building with Python 3.8 and my local build is attempting with Python 2.7.
<g_bor[m]1>hello guix!
<g_bor[m]1>Anyone arrived inquiring about outreachy?
<g_bor[m]1>I believe that the contribution period is started today.
<apteryx>mothacehe: is offload critical to our CI performance? (I think so, but would like this confirmed :-))
<apteryx>also, LibreJS blocks the jschart javascript at https://ci.guix.gnu.org/metrics.
<helaoban>davexunit: yeah that was an obvious extension to the container idea that jumped out at me. still on the roadmap for the future?
<helaoban>davexunit: what happens to the postgres data directory when I remove the postgresql-service from my system config and run 'guix system reconfigure', it doesn't exist in the new derivation, but is it garbage collected at some point from the old derivation?
<mothacehe>apteryx: yes it should, but offloading is so ineffective that many things are actually built on berlin itself
<mothacehe>metrics about that will come
<mothacehe>why are you asking?
<davexunit>helaoban: no the data stays around. guix doesn't manage state like that.
<helaoban>davexunit: got it.
<apteryx>mothacehe: do you have an idea why it is ineffective? If the offload machines are often busy waiting because of high load, this would help: http://issues.guix.gnu.org/43773
<apteryx>we can configure the load threshold to keep them hot (and the default threshold makes more sense on multi core machines)
<roptat>cbaines, so I found out what the issue was: my gitolite.rc file was owned by root, with permission 440, so gitolite falled back to its default configuration
<roptat>not sure why it's not readable by others
<ryanprior>davexunit, helaoban: I am super interested in that kind of rich environment specification with service-level dependencies. Can I read more about past work? Is that something Nix has that I can study how they do it?
<helaoban>ryanprior: re, here are some links where this has been discussed (from a cursory search):
<helaoban> https://discourse.nixos.org/t/how-to-configure-postgresql-declaratively-nixos-and-non-nixos/4063
<helaoban> https://mgdm.net/weblog/postgresql-in-a-nix-shell/
<helaoban>it looks like most of the existing solutions leverage shell hooks to run a series of commands on shell entry to setup the database.
<helaoban>so not really declarative
<helaoban>in guix the postgresql-service currently allows specifying the data directory with 'data-directory', so I guess that can be mapped to a location in your cwd.
<helaoban>you could then have some sort of hook on shell entry that spawns that postgresql service as a shepherd user service (ie as part of a container-scoped os config va guix system reconfigure) and then that system config could just be rolled back on shell exit?
<helaoban>davexunit: would the implementation that you were talking about earlier work along those lines?
<cbaines>roptat, I'm glad you found the issue. I wonder if the Gitolite service should do something about the RC file permissions
<roptat>cbaines, yeah, I just sent a patch for that
<cbaines>roptat, cool, the patch looks good to me :)
<pinoaffe>rekado: ooh, that's nice!
<mfg>does anyone have aclue what could have happened since 5ad0a4522116007420aeb288f76da241f43cb6bf to python-shapely or why it now doesn't build anymore? error is this: https://paste.debian.net/1166007/
<mfg>i don't know how/where to search :/
<mothacehe>apteryx: I should really help. However there are other sources of contention such as https://issues.guix.gnu.org/34033 and https://issues.guix.gnu.org/24937.
<ko_>Hello! I'm having trouble when trying to use the guile-ncurses library
<ko_>Guile doesn't seem to be able to find it
<ko_>Do I have to do something in particular for it to work in Guix?
<iyra>Hi all, I have a question about radeon on guix. I'd like to use my HD7850 card (AMD Southern Islands), with 2D or 3D accel (whichever is available), and at any rate, something better than the VESA driver (or whatever it is). I don't have integrated graphics, either.
<iyra>I understand that radeonsi (superceded by amdgpu?) might need proprietary blobs. If it comes to it, I wouldn't be averse to using that either.
<cbaines>ko_, is the GUILE_LOAD_PATH environment variable being set correctly?
<mfg>iyra: if your card really needs firmware blobs, they won't be available.
<ko_>I think there's no way to use that card without blobs. IIRC, how to use software with blobs and other nonfree software is not allowed here, but look it up because it can be done
<ko_>cbaines, what should it be?
<ko_>Now it currently is /run/current-system/profile/share/guile/site/3.0
<cbaines>ko_, Ok, and in which profile do you have guile-ncurses installed?
<ko_>As an user package, so I guess that's the problem, right?
<iyra>ko_: similar cards seem to work, at least according to h-node: https://h-node.org/videocards/view/fr/1944/Trinity--Radeon-HD-7520G-/1/1/undef/undef/undef/undef/video-card-works/amd but true, they're not SI.
<cbaines>ko_, well, your load path only references your system profile
<cbaines>ko_, if you check ~/.guix-profile/etc/profile does that contain GUILE_LOAD_PATH ?
<ko_>I still find profiles confusing. Let me check
<ko_>No, there's no GUILE_LOAD_PATH
<ko_>How should I set it correctly?
<cbaines>ko_, OK, that'll be because guile isn't installed in your user profile, so I'd suggest installing it.
<ko_>I can still use guile from my user profile. That is because it is in the system profile, right?
<cbaines>I'm guessing right now, you're using guile from the system profile
<iyra>mfg: I suppose I was meaning to ask if it's known by anyone if I do require the firmware blobs. I know raingloom was in here a couple of months ago asking about SI support to be added to linux-libre.
<ko_>Okay, I'm installing guile in my user profile. Thanks! I'll see if it works now
<mfg>iyra: ah sry, i see :)
<ko_>cbaines, now it works perfectly :) I see I have to read about profiles in the manual. Thanks a lot!
<cbaines>Great :)
<apteryx>mothacehe: I see, thanks!
<commanderkeen>what's the deal with openjfx/javafx on guix? is it not included in the jdk packages?
<apteryx>interesting; I've used (@@ (gnu packages linux) make-linux-libre) to define a kernel with a few %default-extra-linux-options, and when 'guix system build'ing the operating-system using it, it fails at the linux-module-database-builder derivation, apparently because none of the default-initrd-modules are found in my built linux output.
<apteryx>here's the config, if anyone is interested to try: https://paste.debian.net/1166024/
<apteryx>commenting out the kernel field line fixes it.
<apteryx>perhaps the #:configuration-file #f is the problem.
<apteryx>For linux-libre-5.8 (the current version), it is set to kernel-config, which for the x86_64 system resolves to the file gnu/packages/aux-files/linux-libre/5.8-x86_64.conf, which does add CONFIG_SATA_AHCI=m :-)
<apteryx>the following config worked, if anyone else is interested in augmenting our stock kernel with just a few extra config options: https://paste.debian.net/1166026/
<apteryx>I should update the cookbook section regarding customizing the kernel linux, which has already fallen a bit behind.
<rekado>apteryx: would be lovely!
<efraim>agreed!
<efraim>I'm also thinking of putting together a VM kernel but I'd love if someone beat me to it :)
***iyzsong- is now known as iyzsong
***bkv is now known as bqv
***janneke_ is now known as janneke
<apteryx>that'd be useful!
***somethinglittle is now known as nothingmuch
<zimoun>roptat: Hi! I have seen that the French TP is moving. :-) Who is currently translating? I remember a message by Hubert…