***rekado_ is now known as rekado
*pineapples calls it a night <gn21[m]><leoprikler "gn21: see also #46314"> I'm sorry, I didn't understand <jsoo>Have fun at guix day everyone! I won't be able to make on account of my time zone but I wish I could! <dftxbs3e>how can we wrap a line to 80 in GNU Guix's source when we can't split it? Like a link? Is it possible to split a link on multiple lines? <lfam>You could use string-append <lfam>Also, although we prefer lines shorter than 80, the linter doesn't complain until 90, and sometimes it's just better to have longer lines than to break them up <lfam>It's a matter of judgement <dftxbs3e>okayy well if it was just me I'd like it as-is <lfam>Does it work in other Guix programs raghavgururajan? It might be helpful to compare <lfam>Gajim won't start at all on my Debian <raghavgururajan>Oh you mean camera? yes, it works in others like linphone, browser etc .. *raghavgururajan was under the assumption that lfam was using Guix System :P <pkill9>how do you 'evaluate' things into store paths in a guile script/at the guile repl? <dftxbs3e>hello, how would you append to a '("list" "of" "strings") inline? '("list" "of" ,@(if #t '("more") '()) "strings") ? ***dongcarl8 is now known as dongcarl
<apteryx>dftxbs3e: so you want that "more" string to be at that exact location in the list? (before "strings") ? <apteryx>dftxbs3e: you can also escape a newline and continue the string on the next line <dftxbs3e>apteryx, I succeeded using ,(if #t "string") <dftxbs3e>had another mistake making me believe everything was wrong.. <apteryx>if you just want to add an entry to a list the simpler is (cons "entry" your-list) <apteryx>otherwise you can also use something like (append the-list '("some" "other" "list")) <apteryx>or (cons* "entry1" "entry2" the-list) <dftxbs3e>apteryx, would this ^ work when the condition is not met? <apteryx>it would insert some bogus #<unspecified> in the list <dftxbs3e>apteryx, hmm so how do I make it not insert anything? <apteryx>you could use some unquote splice trickery to have the else do a noop <dftxbs3e>apteryx, can you show me? I'm too bad at Scheme to attempt trickery :-( <apteryx>or you could do (if your-condition the-list (cons "luajittex the-list)) <dftxbs3e>apteryx, where it's being used? I have another problem: %current[-target]-system isnt available there and adding (guix utils) to #:modules doesnt work either, it says it can't find the module.. when (guix elf) works for example.. I don't understand why <apteryx>you probably want to unquote that information from the host layer into the build layer <apteryx>so (guix utils) would be imported at the top level of the module <dftxbs3e>apteryx, I'll try moving the definition within the unquote itself <apteryx>this is using the "empty list splices in as nothing" trick <apteryx>ah, and my "luajittex" should be in a list of its own, I think, as in '("luajittex") <apteryx>as the unquote-splicing operator expects a list. <apteryx>see info '(guile) Expression Syntax' <dftxbs3e>since the second declaration references the first <dftxbs3e>apteryx, I see well.. I hadnt copied it verbatim and now that I did it seems to work. Thanks a bunch! <dftxbs3e>marusich, hello! :-) Good news! We're very close, I think a test of the "guix" package fails and that's it. <sneek>Welcome back marusich, you have 2 messages! <dftxbs3e>marusich, Build still running, I will know for sure at the end. The old result is way back now. <marusich>OK. On debian the tests/containers.scm (or something like that) fails, it seems like a user namespace issue maybe, but I am not sure <dftxbs3e>It's Fedora 33 here, so user namespaces are enabled by default. <marusich>they were supposedly enabled on my debian distro too, so i'm not sure what the deal is <marusich>i didn't do much testing, i just tried building stuff with the built guix <marusich>we'll see what your guix build failure turns out to be <marusich>It's nice to hear we're getting closer, though! <dftxbs3e>marusich, I'll be here today after all, I think! <marusich>Just double checking since it hasn't connected for me yet <marusich>cool! I'm going to be here off and on for a couple hours. <marusich>I'm in a shared space with other people so I may not always be able to talk, but I intend to listen in and chat. <dftxbs3e>efraim, hello!! Could you try something on your PowerPC 32-bits machine? <efraim>dftxbs3e: sure, I just need to find room on my desk. Haven't powered it on in a month or two since I got my new desktop, which is taking space on my desk for some reason <dftxbs3e>efraim, I suggest you apply the patch then try to compile build cmake <dftxbs3e>cmake-bootstrap/minimal's test suite is good too IIRC <efraim>I'll have to locally revert glibc to 2.30 <dftxbs3e>efraim, what's important is that the patch is applied to the glibc version cmake will use <dftxbs3e>I mean the patch I gave you, current GNU Guix master <bonz060>Shouldn't the big blue button be open now? Or I'm I missing something? <marusich>It is not open for me either, so perhaps it hasn't started yet. <jackhill>Good Morning Guix! It's actually the end of the day for me, but I wanted to wish everyone a happy Guix Days! <dftxbs3e>marusich, yes I can hear you. Your voice scared me aa! Too loud. <dftxbs3e>marusich, no no, my computer sound is loud. <dftxbs3e>So for a bit less than 3 hours there's nothing on the programme, so we are just here with time together I believe! <phant0mas>dftxbs3e: we can start working on something, any ideas anyone? <dftxbs3e>There is that ZFS file system support contribution <dftxbs3e>I am horrible at Scheme so not sure I can review but it's really interesting in GNU Guix <bonz060>dftxbs3e: That's (one of) the points of today :) <marusich>Oh, I thought there was going to be a 'state of the union' or some such discussion. is that scheduled to happen later? <efraim>probably. I assumed that was ludo <marusich>I see now that those sessions are planned for a few hours from now. <phant0mas>Ludo will join around 9AM Paris time (UTC+1) <marusich>So yes, I suppose right now is free time! <dftxbs3e>phant0mas, FYI along with marusich we are working on PowerPC 64-bits Little Endian support in GNU Guix. <dftxbs3e>phant0mas, Well! We'll need reviews, but for now I'm just waiting for builds. Do you have a PowerPC 64-bits Little Endian machine at hand? I could make available a virtual machine to you (and others). <dftxbs3e>There's not that *much* work, mostly waiting for builds, to be very honest. <phant0mas>I don't, is it too hard to run a powerpc vm locally? <dftxbs3e>phant0mas, you mean emulating it? It'll be very slow <wonko_the_sane>hmmm this seems brain damaged on guix invoking 'lastlog' fails and emits '/etc/lastlog: No such file or directory' (yeah i would expect not thanks) , /var/log/lastlog file exists as expected. It's cool that your trying to deprecate the filesystem hierarchy standard but, but this just seems like a practical joke or something :D I'm pretty sure the lastlog program I'm familiar with is hardcoded to read <marusich>cross-arch emulation is surprisingly slow. <wonko_the_sane>/var/log/lastlog or at least it doesn't expose any options to change that... /etc/login.defs has only 2 lines (setting ENV_PATH and ENV_SUBPATH) is it incomplete may be ? <dftxbs3e>There's also something else that's relevant to PowerPC 64-bits Little Endian, it's about Golang in GNU Guix, the bootstrapping needs to be updated to use gccgo instead of go 1.4. <wonko_the_sane>i've been having some weird behaviours with sshd and i was thinking it could be related. So i grep -ri src/guix/ -e lastlog and the only matches are from gnu/services/ssh.scm hmmmm <efraim>dftxbs3e: I've applied the patch on top of my local wip-ppc branch, but it could take a couple of days to build out to cmake-minimal <phant0mas>dftxbs3e: let me update my guix install, start having a look at the patches and get up to speed <efraim>I've enabled the binfmt emulation on my ryzen machine, for aarch64 it's comprable to building on my pine64 <dftxbs3e>efraim, ah right.. your machine's very old and slow. <marusich>Just FYI, I will be here, but I will be slow in replying, due to distractions where I am. <dftxbs3e>efraim, Oh.. I thought you had a Macbook PowerPC. <marusich>dftxbs3e, regarding your latest changes, I will do my best to get them merged in the next couple days to the wip-ppc64le branch on savannah (which is where we are doing the POWER9 work, for those who have not heard of it). Maybe you can nag Ludo and the others for commit access later :) <dftxbs3e>marusich, I need one more "vouch", maybe that will be Ludo, don't know! <efraim>gcc and guile-3 take a couple of hours to build each <phant0mas>dftxbs3e: thank you! I will give the gccgo a try <efraim>phant0mas: I suspect it'll need something similar to the *-toolchain to work correctly <dftxbs3e>phant0mas, if you can make it work on x86_64-linux, it will be useful on PowerPC 64-bits Little Endian as well. <dftxbs3e>Basically the problem is that go 1.4 does not support powerpc64[le] code generation at all, but we can only bootstrap from there. The only alternative is gccgo (not written in go even today). <efraim>that's how we tested the binutils-gold for aarch64, built it on x86_64 and when it was working tested it on other machines <lfam>Re lastlog, from the build log of shadow: <lfam>configure: WARNING: utmp file not found <lfam>checking location of faillog/lastlog/wtmp... /etc <timmydo>has anyone use radvd on guix or know if there is an equivalent? <marusich>I've dropped off of BBB for the moment; if someone is sharing a video etc. I'll rejoin <dftxbs3e>marusich, This is the logs, apparently. There's another file? Do I need to -K? Trying now. <marusich>also, what is the exact reason guix claims the build failed? *dftxbs3e waits again for long build <marusich>Yeah, I don't have access to my ppc64le machine right now, either <marusich>So unfortuantely I'll have to wait a day or so to try out your patches myself <dftxbs3e>marusich, I'll create an OpenSUSE Tumbleweed VM and give access. <phant0mas>@dftxbs3e should I use master or core-updates as the base for any work? <dftxbs3e>phant0mas, I suppose for this problem you should use core-updates <dftxbs3e>Thanks a lot! Nice to see people motivated :-) <dftxbs3e>marusich, Debian Bullseye could do as well, since there's a guix package there now,. <dftxbs3e>This patch could be useful to review as well: https://issues.guix.gnu.org/46305 - it allows to setup a 4to6 tunnel easily to give one IPv6 access when they only got IPv4. It's zero-configuration so miredo just needs to be run as root and it works. So for example if I give access to VMs, I only have an IPv6 range to assign them IPs at home, without miredo some times people wont be able to connect to my VMs. <dftxbs3e>Miredo embeds some public teredo servers by default that it uses (that's why it's zero-configuration). <dftxbs3e>phant0mas, how difficult do you think it would be? <phant0mas>dftxbs3e: not sure yet, I have issue with my setup which I am fixing now :) <dftxbs3e>sneek later tell marusich the "guix" package failure is weird, certainly looks like a bug in the test suite. It may be because powerpc64le-linux isnt a "supported" one (needs --with-courage) <kondor41>dftxbs3e btw, can't share my `config.scm` as i have none -- usually on a gnu linux distro + guix as a package manager <dftxbs3e>For the early of GNU Guix Days, I propose we can do some config.scm sharing :-) <kondor41>but, curious to see what others have in there <kondor41>though i have a recent "reproducible" set of packages i created with a help of an inferior <kondor41>i don't know if this is the right way, so i'll share that, interested in comments <dftxbs3e>kondor41, please do share :-) But I am no Scheme/GNU Guix APIs expert. <phant0mas>dftxbs3e: just to make sure I understood correctly, is go 1.4 part of the guix bootstrap process in any arch? <dftxbs3e>phant0mas, yes, but go 1.4 does not support ppc64[le] code generation so it does not work on PowerPC 64-bits Any Endian. <g_bor[m]>What I did a bit differently was that I pulled the packages to channels. <dftxbs3e>g_bor[m], speaking of the config? What do you mean pulled packages to channels? <g_bor[m]>I can't share right now, as I am moving to a new dev machine, and not everything is here yet. <phant0mas>dftxbs3e: do any of the newer ones support ppc64 or our only option is to use gccgo? <g_bor[m]>There are two package definitions on the top. <dftxbs3e>phant0mas, newer ones need previous ones to compile, as they're made with some golang. So gccgo is our only solution yes. <dftxbs3e>g_bor[m], oh.. with git? well this is my dirty hacks, I hope to merge something some time.. <phant0mas>btw just fyi, the environment issue I had was me using an older guix version :( <dftxbs3e>g_bor[m], getting gnome-keyring to work with XFCE was a challenge. <g_bor[m]>also, on one machine I have guix-home-manager setup. <dftxbs3e>g_bor[m], the trap of forgetting pinentry-gnome3 and gcr, means you wont get any prompts. <kondor41>So, I needed something like specifications->manifest and came up with that cludge starting at line **145** <g_bor[m]>If I will have a bit of time to spin up the old setup later I would like to share that. <dftxbs3e>g_bor[m], I'm still trying to get polkit prompts and all to work properly.. on XFCE. <dftxbs3e>liltechdude, the best is you strip that zlib folder or any other folder and package it separately then make that main package use these separate packages as inputs. <dftxbs3e>Otherwise when there is mixed licenses, license field also takes a list <hulten>For many years I have a relatively constant config.scm that slowly clogged up with experiments (mostly commented out). <hulten>dftxbs3e, I see you have a list of many "user" packages like icecat and kodi in your config.scm. <hulten>Is that usual to do this, instead of defining them in your user profile? <dftxbs3e>hulten, yes, I don't use my user profile. Whenever I need something as a user, I use: $ guix environment --ad-hoc <name> -- name <g_bor[m]>liltechdude: you can have a list of licenses there <dftxbs3e>hulten, or if I need that thing a bit more permanently I add it to the config.scm <hulten>Such that it is system-wide and available for all users? <dftxbs3e>hulten, it allows me to manage updates better, system-wide, always. <g_bor[m]>If the license is only for a small part of code, you can also add a comment to see which part is covered by that. <hulten>That might be a good approach – right now I have three users (for personal use, work, and my wife), and 'guix package -i' works quite well if something is not available, especially if the version is the same in another profile where it is already installed (so roughly same guix checkout). <g_bor[m]>What I do is that I have a manifest for my user profile. This does not make clutter in config.scm, and I can still keep track of what is happening. <hulten>But it is a bit of a hassle; I'm often not sure if a package is available. <dftxbs3e>kondor41, what is the point of the inferior here exactly? <hulten>g_bor[m], that's probably the best approach. <dftxbs3e>I don't like having two places to update stuff. Especially with the unattended-upgrades service (it cannot update user profiles yet). <dftxbs3e>I don't like it that my user and system GNU Guix versions are separate too. <hulten>One can keep their system profile minimal, and put everything in the user profile. <hulten>But then there are maybe "border packages" like nmap, wget &c. that I do have in my system profile. <kondor41>dftxbs3e it looked like a way to decludge channels.scm <kondor41>and use my custom channel with custom , unofficial packages in a straightforward way <dftxbs3e>kondor41, I see, without the inferior it does not work? <kondor41>dftxbs3e i used to have multiple channels <dftxbs3e>kondor41, and you're not adding anything to /etc/guix/channels.scm? <kondor41>(rather my .config/guix/channels as i am not on Guix distro) <dftxbs3e>kondor41, I see! So that's how it's useful I believe then! I feel like it's better to define everything in a single file! <kondor41>dftxbs3e there are disadvantages though, `guix -s package pkg-in-inferior` won't return anything <hulten>dftxbs3e, I see you have 'call-with-input-file "/etc/guix/channels.scm"', but not '%default-channels'. <hulten>Is that defined in the former file and/or are you not using the default channels? <hulten>I don't have any such thing in my config.scm; it's also possible that I don't understand. <kondor41>is any of the bioinformatics people on the channel yet? <dftxbs3e>hulten, basically I use that to pass my channel configuration in /etc/guix/channels.scm to the unattended-upgrades service, otherwise it doesnt read from it. <dftxbs3e>It's for the "channels" field, that file contains %default-channels and some others. <bonz060>kondor41: I work in that space. I guess pjotrp is also around. <kondor41>bonz060 i am wondering when will we need to stop adding bioconductor and cran packages and guix just inhales them all automatically :) <pjotrp>we can make bioinformatics a topic for today <pjotrp>on my second coffee - disclaimer <pjotrp>I think we'll start a plenary when Ludo is here <efraim>i drink my coffee from a beer mug, 1/2 liter at a time <dftxbs3e>aaa bad coffee.. I choose not to drink any. <nixo_>good morning and happy guix day! <bonz060>kondor41: There was talk of using minikanren in guix(with program synthesis at the very end!). Perhaps that's a (wild) use case? <dftxbs3e>Maybe that's why I end up leaving all jobs feeling so exhausted /lol <kondor41>bonz060 i'll need to look into that first <hulten>dftxbs3e, what's your poison then? <kondor41>i find irc more compatible with handling children during a conference <hulten>Apropos, I just looked up and my wife put a cup of green tea right here! It's still warm. <dftxbs3e>hulten, not sure, fine buffala-milk mozarella? aha <hulten>Thanks for the reminder of my tea. <pjotrp>I can tell it is going to be a good day :) <hulten>Well... actually it is Jasmine tea.. and that *is* her name. <hulten>so I'm not sure if I should drink it. <kondor41>my wife set a mug of something which i drank but was apparently hers <efraim>bonz060: oops, forgot to mute myself <pjotrp>I imagine opening the room at ICAB and getting tables ready and a beamer, people dropping in slowly - wet but happy <phant0mas>pjotrp: we should have named the room virtual-ICAB <kondor41>pjotrp i remember your awesome talk at last years FOSDEM <pjotrp>with luck we can do next year :) <kondor41>and use some hardware which is guix based and yet have no idea what they are using <dftxbs3e>kondor41, oh cool! where do you work? how do you use GNU Guix? <efraim>vagrantc! have you figured out how to build debian packages on guix? I have a WIP package for pbuilder but i fear it has a ways to go <kondor41>and i'm the only scheme/guix afficionado for the moment <kondor41>so we have a couple of workstations/servers/VMs that all run guix with our own selection of mass-spec related packages <dftxbs3e>kondor41, I see! And can you share config.scm for those machines? It would be interesting! <kondor41>so, thats the thing, i set them all up as hybrids <dftxbs3e>I see.. so no config. I understand. GNU Guix's not that mature yet. <hulten>dftxbs3e, about the mcron jobs you have in your config.scm, (1) are they running properly, and (2) why not a single 'guix gc -F 100G --optimize' command (as it is running at the same times/frequencies)? <kondor41>and the computers we have are not thinkpads :D <vagrantc>efraim: i haven't hardly had a chance to try <hulten>(Cron jobs are not working for me, so maybe I can follow your example.) <dftxbs3e>kondor41, I mean even without the freeness part GNU Guix needs some work. And for the rest, some third party channels can well cover the need so :-) <dftxbs3e>hulten, in my experience it's either --optimize or -F 100G, AFAIK they work, though not 100% certain. <vagrantc>efraim: i've mostly been helpfully using it to extract debian packages and debian source packages on guix to compare against guix and such <vagrantc>efraim: using ar and tar and manually extracting them was ... more-or-less fun, but error prone :) <efraim>vagrantc: I'm torn between trying to make pbuilder work on guix or making a debian-build-system that uses a minimal debian install in a chroot to build packages <kondor41>i think i saw it on gnu.org as one of the packages needing a maintainer (maybe i remember wrongly) <efraim>the debian-build-system sounds more interesting, but I don't want to start thinking about dependency resolving or updating the base debian image <hulten>If not mcron, what is a good (clean and maintained) cron? <dftxbs3e>hulten, I think mcron is maintained because it's what GNU Guix uses everywhere, so. <dftxbs3e>brrr I keep confusing hulten and kondor41 when tagging :-S <vagrantc>efraim: my crazy idea is to create derivations which are essentially extracted .deb packages ... <vagrantc>efraim: and then compose them into an environment somehow... <kondor41>dftxbs3e i am probably wrong about mcron, then <vagrantc>efraim: but i hardly have an idea how to even start such a crazy project :) <vagrantc>efraim: i wonder if pbuilder or sbuild would be easier... <cbaines>vagrantc, what's the aim with composing Debian packages and Guix things? <dftxbs3e>kondor41, it may need a maintainer, but seeing it used in so many places, I am sure people will show up with patches in case something does not work. <dftxbs3e>cbaines, contribute to Debian while running GNU Guix? <vagrantc>cbaines: but i kind of want to leverage guix's containerized build environment while building valid, proper debian packages <dftxbs3e>It would be nice to be able to generate .rpm, .deb, etc. binary packages (with provenance information) from GNU Guix packages. <efraim>slack packages could be an easier starting point, IIRC they're just the output of 'make install' <cbaines>vagrantc, given how many build tools Debian has, a guix builder could probably be written, similar to pbuilder, whalebuilder, ... <vagrantc>cbaines: it would require at least emulating a chroot, as obviously building it against guix libraries and such would result in non-functional .deb packages <vagrantc>cbaines: right, something along those lines <dftxbs3e>I'll be back really quick buy something to eat outside at the bakery before 9AM! BRB! <phant0mas>@dftxbs3e gccgo 5 fails to build on guix, I am debugging this now <cbaines>Is the BigBlueButton stuff up and running? <dftxbs3e>phant0mas, I am thinking maybe it's possible to use even later gccgo? <kondor41>what's the status of maven support, when we're talking about building? <dftxbs3e>gradle as well.. needed it to build some .apk for Android the other day. Apparently an obscure binary lying down there. <vagrantc>BBB seems to be up and running at least enough that i could join *something* <vagrantc>hrm. wasn't able to update diffoscope to the latest version ... suspect some issue with python 3.8 vs. python 3.9 <vagrantc>most of the tests error, some of the tests fail, many of the tests are skipped, and very few actually pass <janneke>jonsger: since some time now; the day before the v1.2 release iirc ***htgoebel2 is now known as htgoebel
<pjotrp>I added the topics on BBB shared notes, feel free to modify <janneke>yes, in no particular order, right ;) <cbaines>pjotrp, I'm not signed in, I'm not sure I have an account <pjotrp>what else - we can add during the day, mind. The point of an unconference! <kondor41>pjotrp why not rename bioinformatics to scientific computing? <kondor41>it's not like we have thousands of scientific computing members today :) <htgoebel>pjotr: I'm a bit curious about the "shared notes". What is it about? <htgoebel>pjotr: In the wiki we tarted a kind-of-schedule yesterday <pjotrp>yeah, this is the unconference ideas <kondor41>why is not _every_ conference organised on IRC <cbaines>I'd missed the link on the wiki page <pjotrp>cbaines: do you want to start? First session on QA :) <civodul>teythoon: hey thanks! your talk on the Hurd back then inspired me :-) <bonz060>If you aren't logged in, you can only access it through read-only mode :( ***apteryx is now known as Guest20319
***apteryx_ is now known as apteryx
<bonz060>I take that back ^^; I can do some edits <leoprikler>raghavgururajan: I think it's either some missing KDE source (given the leading k) or some obscure GStreamer plugin. <dftxbs3e>cbaines, could your patchwork thing post results automatically on the debbugs mailing list? <dftxbs3e>cbaines, another thing, pre-merge automated GNU Guix builds for all dependents affected by the change, a bot on the mailing list for example, make it soft/hard requirement for merge. <db48x>perl6-terminal-ansicolor; am I doing it wrong or can anyone else replicate the problem? <kondor43>community organisation as a separate topic <phant0mas>@dftxbs3e the gcc-go needs a runpath to be set. Should I set it in the main gcc package or the gccgo one? <efraim>db48x: quite possibly it doesn't build, perl6/raku hasn't gotten much love since it was added <dftxbs3e>cbaines, problem: relying on humans to be rigorous on testing and rebuilding all dependents *before* merge is hard to require. <bonz060>I don't know if this would be practical; but would a discourse channel(that's more discoverable than IRC/ ML) be practical? <efraim>phant0mas: I'd try the gccgo one first <dftxbs3e>phant0mas, I really don't know, so follow other's advice :-D <dftxbs3e>cbaines, I feel anxious/stressed doing anything, I like to rely on some tools to guarantee some things for me, it helps me feel more comfortable mentally working on something. <dftxbs3e>cbaines, e.g. Rust programming language being solution to my security bugs anxiety/stress writing C, being extra careful and doubting every single line of code spending too much time on them. <kondor43>phant0mas hahaha, rust as a stress reducing instrument, love that <kondor43>on the topic of guix@foerign distros, if we could get more of the guix system together with the packager (ie, substitute systemd with shepherd), it would be easier to do QA <dftxbs3e>cbaines, are you looking for problems that are about QA basically? <cbaines>dftxbs3e, sort of, what we've got is a really good list of areas at least <dftxbs3e>cbaines, not sure what the question was, but maybe define specific configurations that should always be tested (instead of observing all packages, observe specific high-profile configurations, such as XFCE or GNOME with default apps and config) <civodul>htgoebel: do you know how to eventually export notes from the pad? <dftxbs3e>require that these configurations never break, builds and system tests, also with UI tests it would be great, not sure UI tests are possible but ensure specific features are working and well configured somehow <dftxbs3e>cbaines, alternatively, somehow make a database of user configurations (user-contributed) moderated lightly but with minimum gate-keeping and present results in some QA page so we can easily find out when things break for users. <htgoebel>civodul: In the right top corner here is a menu (three dots) and there is "export" <kondor43>cbaines one thing i forgot, multiple profiles have become a thing for many Guix users; it could be useful to test specific packages how well they work when there are parallel profiles on the system <dftxbs3e>cbaines, as an example, for the rustc compiler, they have something called crater that runs builds on all crates in the eco-system to find bugs in the compiler, similar thing for GNU Guix would be nice but for user configurations. It essentially gives us what packages are used, to prioritize. <dftxbs3e>cbaines, civodul: I mean that GNU Guix packages are managed by GNU Guix itself, "crates" are user-data, people's configurations would also be user data. <dftxbs3e>It can be overwhelming to see all packages fail, I suggested ensure some high-profile configurations are tested. <civodul>dftxbs3e: oh i see; pre-release, there's "make assert-binaries-available" that checks substitutes are available for high-profile packages <civodul>see the etc/release-manifest.scm file in the source <dftxbs3e>civodul, I mean high-profile whole-system configurations more like <dftxbs3e>like fully-featured GNOME (not like currently out of the installer), fully featured XFCE, .. <civodul>oh i see, then that's more like system tests, no? <dftxbs3e>civodul, but system tests need to be approved into GNU Guix, any way system tests could be submitted with less gate-keeping? <civodul>dftxbs3e: code that goes into Guix passes review, i think that's reasonable :-) <civodul>it's possible for a user to have their own system tests, though <dftxbs3e>civodul, I have in mind that user could submit their configuration as-is (config.scm) and have it tested automatically with feedback on every change to developers making changes in case something breaks <dftxbs3e>it could also mix-up with a platform for users to easily share their configuration with the community <civodul>but that's not really testing much though, because you'd need to interactively figure out whether it works "as expected" <civodul>(whereas system tests are automated) <efraim>like a list of "please don't break these packages" list <dftxbs3e>to obtain that list of don't break package list.. we need configurations, but yes system tests are better civodul, but require more effort to create. <dftxbs3e>civodul, cbaines: I think it is also important to bake every automated QA mechanism with notifications on the mailing list in the concerned bugs (for example for new patches, creating whole new issues automatically would be overkill) <dftxbs3e>So that it solves the discover-ability problem of QA tools <cbaines>I clicked close, that that maybe worked? <janneke>i hear you cbaines and pjotr, but my web interface is gone <dftxbs3e>cbaines, about speed, please provide a ready to use configuration to run a mirror of GNU Guix substitutes that also builds things on its own and all. <dftxbs3e>I want to run one, but can't because hard to configure <PotentialUser-97>Why is there a program that is not completely open source (like Telegram) in Guix packages? <dftxbs3e>cbaines, and embed a default list of these alternate substitute servers and it will only download identical substitutes to official ci.guix.gnu.org available on these other mirrors rather than require more signing keys to be approved <dftxbs3e>PotentialUser-97, the desktop client is Free Software, that's enough for inclusion in GNU Guix, antifeatures like supporting nonfree services is another thing relevant but not prohibiting inclusion in GNU Guix. <wonko_the_sane>The crypto of telegram isn't taken seriously by any cryptographer is it ? <dftxbs3e>wonko_the_sane, you could be using Telegram for it's awesome UI/UX and responsiveness also being keen to privacy even though very imperfect because favors usability over privacy/security. (E2EE not on by default, custom obscure crypto protocols) <PotentialUser-97>dftxbs3e: The server to which the user connects is completely proprietary; Yet Guix has no problem with that? <dftxbs3e>PotentialUser-97, GNU Guix packages could have packages for any Free Software, whether they have the antifeature of connecting to proprietary services or not. <dftxbs3e>Telegram may fall into the relaying information for you rather than doing computing for you, thus would be less of a problem. It's not the worst. <civodul>cbaines: thanks for hosting the QA session! <civodul>you hard the hard task of being the first one :-) <PotentialUser-97>However, it 's best for the user to be aware of this antifeature in the Guix packages section, as well as during installation. <dftxbs3e>PotentialUser-97, yes, that would be a good feature to add! <dftxbs3e>PotentialUser-97, Note: In general GNU Guix seems more focused on building tools for practical software freedom than enforce policies strictly though it still has to with GNU FSDG. <rekado_>that the server you are talking to is not released as free software? <rekado_>should “wget” also get a warning then? <dftxbs3e>rekado_, it's different, Telegram has only one server <dftxbs3e>It's configured by default, you have no choice with Telegram <dftxbs3e>rekado_, FDroid already applies such labeling for Antifeatures, it makes sense to me to have some hierarchy here between proprietary services with Free Software clients and Free Software clients to Free Software servers you can reinstall if you want to. <hulten>Since the Telegram client is free software, there is a protocol at least implicitely documented there in the code to talk to servers; hence, one could in principle install a (free software) Telegram server. <hulten>(I mean: first write, compile, then install.) <dftxbs3e>hulten, that Free Software Telegram server does not exist however, and the point of Telegram is to discuss with people so that alternate server certainly wont federate with the main network, the main network wouldnt allow it a.k.a the company behind Telegram. <kondor4>completely unrelated (as i got in here back from another meeting), can anyone recommend a more GNU friendly variant of Amazon cloud where I could run an instance of Guix (as a KVM image) for a few days (and use it for a student course)? <dftxbs3e>kondor4, Digital Ocean there is a deploy script for it. See guix deploy - but it's just because it exists, Amazon Cloud would be as suitable if it did have the same deploy script. <rekado_>the license of the server software really doesn’t factor in it. <hulten>dftxbs3e, yes I understand, and my proposal does not make much sense; but it's correct nonetheless. A bit advocate-of-the-devil. <rekado_>what would make DigitalOcean “more GNU friendly”? <dftxbs3e>rekado_, GNU Guix-friendly I guess. Misread. <hulten>I like the anti-feature warnings in F-Droid. <kondor4>rekado_ i've seen some openstack solutions <dftxbs3e>rekado_, in FDroid, the license of the server software does matter for the Antifeatures labelings. <hulten>And the question is whether we want to do that as well. <dftxbs3e>I think it is quite informative for users and is relatively easy to do. So for me it's a yes. <hulten>But who is going to decide if certain properties of software are "anti-features"? <hulten>I agree with the Telegram case, but it may not be always as clear. <dftxbs3e>rekado_, wget can fetch anything, the Telegram client can only fetch telegram.com <hulten>(but of course you're arguing: not enough :-)) <rekado_>shall we also label shooter games with warnings? <dftxbs3e>rekado_, Xonotic server list service? I think it's Free Software. <hulten>And what things would be too much privacy-invading. <bdju>looks like hedgewars still fails to build <rekado_>I prefer to stay clear of labeling packages with warnings <rekado_>the FSDG are controversial enough; we shouldn’t go even farther. <dftxbs3e>rekado_, it is relevant to the whole political software freedom battle to put hierarchy between solutions that care about software freedom fully and those who consider it only second priority (like Telegram vs GNU Jami) <dftxbs3e>Besides developers who can inspect code, non-developer users are not aware about the difference. <vagrantc>rebuilding the world for armhf ... that will be fun! <dftxbs3e>janneke, how hard is it to port this to PowerPC 64-bits? <hulten>What about informing and educating users through different ways than 'guix package', e.g. an article on the Guix blog explaining that packages are FSDG-compliant but they may always have some other issues? <dftxbs3e>janneke, I didnt know it was all Scheme! <dftxbs3e>civodul, OTOH RISC-V hardware is less expensive and thus more accessible (more like ARM) <rekado_>hulten: good idea, thought I’d like it to be a little more positive <leoprikler>I personally think an 'anti-features property would be wonderful <jonsger>I don't think at least not for leaf packages <vagrantc>"do we even need guile" quote of the day from ludo! <janneke>pjotrp: mescc-tools builds fine for both riscv and powerpc <janneke>dftxbs3e: yeah, all except for the scheme interpreter itself: mes.c and mes lib c, is scheme -- guile-compatible scheme even <avalenn>I would have liked to participate to today's unconference but did not manage to pause my life. <avalenn>I hope participants have a good time. <efraim>i'm off to lunch, also, looks like we want to adjust the ppc64le patches so they can be applied on master without rebuilding the world on the other architectures <civodul>looks like a case where we're almost there <civodul>janneke, vagrantc: great to see Debian build mescc on all the architectures! <jonsger>civodul: out-sourcing the CI work to other distros seems to be a thing now :P <vagrantc>civodul: but it's important to be able to get there :) <dftxbs3e>civodul, I also forgot another next step: Linux-libre config and installer images.. <bqv>cheers: did you ever get anywhere? With your "flakes-esque" tool? <bqv>I keep forgetting jami is a gnu thing. It's not what I expect gnu software to look like... <dannym>I think the breakout room timed out <janneke>dannym: saw your arm messages on bootstrap; nice <janneke>and yes, wip-arm-bootstrap is up to date <dannym>janneke: I had a deja yu about those messages though :) <janneke>yes, we had a similar problem before <dannym>My suspection is that the numeral to number conversion is broken <dannym>That's why 5+5 worse (one digit each) but 10 doesn't <dannym>That's why 5+5 works (one digit each) but 10 (two digits each) doesn't <dannym>in /gnu/store/wdi63jz8ryxxcphikd9fc8yhzig0d4x0-gcc-core-mesboot0-2.95.3/lib/gcc-lib/arm-unknown-linux-gnu/2.95.3/cpp0 <dannym>( ~/src/guix-wip-arm-bootstrap/guix/pre-inst-env guix build -S -e '(@@ (gnu packages commencement) gcc-core-mesboot0)' ) <civodul>dannym: we missed you this morning in the boostrapping session :-) <civodul>mothacehe: cool; should you take the lead soon? *bqv building guile-emacs 👀 <civodul>mothacehe: hope you're fine with the round of intros <civodul>mothacehe, lfam: we should be able to start the "CI" session in 15mn i suppose <lfam>Okay, that sounds good for me civodul and mothacehe <dftxbs3e>phant0mas, hey! are you in BBB? what's your username there? <phant0mas>I sent a patch related to gccgo, have a look in the ml <dftxbs3e>phant0mas, this quick?! Ahh thanks a lot! <phant0mas>well nothing to fancy, just adding an updated gccgo <phant0mas>and I still need help with validating the runpath <phant0mas>but it does work locally if you disable the runpath <phant0mas>then we can start working with a new build system *dftxbs3e doesnt see the email in the ML yet <dftxbs3e>phant0mas, thanks, looks like I don't receive emails from debbugs yet. <phant0mas>check the link and respond to 46380@debbugs.gnu.org <phant0mas>have you registered to the guix-patches mailing list? <dftxbs3e>phant0mas, yes, I receive other patches, I'm just confused now. Also very tried so not the best time to respond to anything heh I'm sorry.. <phant0mas>dftxbs3e: don't worry too much about it, they will come <zimoun>mothacehe: the link in RSS are relative and so broken in the RSS reader. For instance «Link: ../../build/279909/details» instead of https://ci.guix.blabla, I guess. *janneke is afk for a bit <efraim>i thought it looked interesting, haven't tried building it though <spk121>ArneBab: as the guy that wrote a binding for guile for ncurses, this is pretty exciting <jeko>when packages have outputs like i can select with i.e. `rust:cargo` if i install rust, is cargo install by default ? or do I have to specify it ? <dftxbs3e>jeko, you have to select, default is name:out <bonz060>Wrt the Rust stuff, is it Room 1 or Room 2 <jeko>dftxbs3e: ok thank you ! <civodul>efraim, htgoebel: room 1 is now yours! <lfam>How do we switch between rooms? <civodul>there's a "Separate rooms" menu to the left <spk121>lfam: for me, if I reload the screen, it starts me at the join breakout room menu <efraim>htgoebel: I keep timing out when connecting the audio <bonz060>efraim: Ha! Audio problems! I've had to use my phone :( <dannym>janneke: Maybe gcc/cexp.y though, according to gdb <bqv>starting to understand why guile-emacs isn't cached <bqv>build has been running for like 3 hours <rekado_>bqv: guile-emacs is effectively broken, even if it were to build <rekado_>bqv: I had a version that did build but would crash… <euandreh>how you feel about adding a c99 alias to cc in the gnu-build-system, so that packages that use the c99 compiler don't need to give the "CC=,(cc-for-target)" flag? <euandreh>the downside would be the cost of rebuilding everything, the upside is having a more POSIX friendly base environment <rekado_>I have never seen any package that invoked GCC as “c99”. <lfam>New problem: gnome-calculator takes very long to start <pkill9>has anyone run guix on a pinephone? <civodul>Outreachy/'guix git log' session in ~15mn hosted by lemes, zimoun, and g_bor[m] <efraim>it copies, it doesn't actually work <euandreh>rekado_: after reading "How to Write Portable C Without Complicating Your Build", I started using c99 on my personal projects, and when I package the first I found that gnu-build-system doesn't have c99 <efraim>debian keeps crate sources in /usr/share/cargo/registry/<name>-<full-version>/ <euandreh>however it is true that very few places invoke the compiler via c99 <dftxbs3e>efraim, htgoebel: what do you think about "possible solutions" in the write up ? <rekado_>I just joined the guixbbb thing; is anything scheduled to happen right now? I only hear the occasional sniffing :) <pkill9>does anyone know of a cli email client htat displays mesages in the style of reddit/ <mroh>euandreh: On the first look, I found only one package (spiped) that tries to invoke c99 <rekado_>which of the many reddit styles do you mean? :) <pkill9>showing messages indented to the message they are replying to <rekado_>is it an option for you to use emacs+gnus only as a mail client? <euandreh>mroh: it probably isn't worth the cost, given that so few packages do that <dannym>tcc used is /gnu/store/v37svarzv82x5c4zalflmiqsd8mld1i0-tcc-boot-0.9.27/bin/tcc <lfam>pkill9: I think mutt does what you ask <dannym>Not yet--I think it's just because i used printf with %d <dannym>%lld doesn't exist on tinycc, though <janneke>ah...right mes lib c may well have problems with %lld <janneke>note that tcc uses mes lib c (eg for printf) <stikonas>doesn't mes libc ignore those l quantifiers? <janneke>strange, but possibly that feature is not used on x86... <janneke>either that, or mes lib c printf %lld works on x86 <janneke>stikonas: yes i think so...the result could be "it works" on x86, and "it breaks" on arm? <dannym>gcc/cexp.y: max_over_base = (unsigned HOST_WIDEST_INT) -1 / base; // 0 <dannym>│ 0x8238 <parse_number+236> bl 0x10c0c <__udivdi3> <dannym>that _udivdi3 implementation in there tries to load the arguments from the stack... but they are put into registers, not on the stack <janneke>oh my, we have armeabi confusion somewhere? <dannym>tcc -D __GLIBC_MINOR__=6 -DIN_GCC -g -DHAVE_CONFIG_H -D __GLIBC_MINOR__=6 -I. -I. -I./config -I./../include -c ./cexp.c <dannym>when compiling a "/" to assembly <pinoaffe>on my laptop, inkscape keeps segfaulting whenever I attempt to open / import a file <pinoaffe>this is on latest guix - anyone know what the cause could be? or should I go to the inkscape folks? <lfam>Is there a way to do `make check --keep-going`, so it doesn't stop if one of the tests fails? <mroh>pinoaffe: Is this on a foreign distro (I can't reproduce it on guix system)? I think I remeber some issues with the gnome "open file" dialog on foreign (debian) distros in the past. <dannym>The values that are divided in gcc cpp0 parse_number are -1ULL and 10 <janneke>ah, the cast to unsigned long is fubar? <janneke>these functions were supposed to be stubs <janneke>at least when compiled with mes and the early tinycc's that don't have long long <janneke>now they are also being used "for real" by the final tcc <civodul>guile-netlink demo by roptat right now :-) <wingo>guix time-machine: error: Git error: the index is locked; this might be due to a concurrent or crashed process <wingo>i guess due to invoking multiple "guix time-machine" inside "make -j40" <vagrantc>thanks all for sharing your thoughts for guix days <dannym>The original location does this: <dannym> max_over_base = (unsigned HOST_WIDEST_INT) -1 / base; <dannym>Does that really *need* to be "long long" ? <dannym>Or we could just implement long division for unsigned long long, too, I guess. <janneke>dannym: what happens if we wrap the casting to long inside #if !HAVE_LONG_LONG? <janneke>otoh, avoiding long long sounds nice too <janneke>ah, but we would need to implement unsigned long long division .. right <dannym>Also, there's __mesabi_udiv ... but that only improves precision by one bit :) <civodul>starting an open-ended "community" session now :-) <dannym>I'm not sure it's necessary since it's only a warning anyway--that warning comes when the overflow checker found an overflow--it doesn't break at that place yet <dannym>I'll run ./pre-inst-env guix build -e '(@@ (gnu packages commencement) glibc-mesboot0)' again to find an actual error to look at, now that I know <dannym>Possible problems on all the *other* places where it uses _udivdi3, though <dannym>Aha, I eventually get: program cpp0 got fatal signal 11 <dannym>janneke: Hmm, the segfault is also because of division by zero <dannym>#0 0x00030e68 in __mesabi_div0 () <dannym>So I think we should handle this after all <janneke>arg --- too bad...and also have a hard look at other functions like _udivdi3 <dannym>Reproducer in glibc-mesboot0 so far: <dannym>dannym@novena /tmp/guix-build-glibc-mesboot0-2.2.5.drv-0/glibc-2.2.5/iconv$ /gnu/store/wdi63jz8ryxxcphikd9fc8yhzig0d4x0-gcc-core-mesboot0-2.95.3/lib/gcc-lib/arm-unknown-linux-gnu/2.95.3/cpp0 -lang-c -nostdinc -I../include -I. -I.. -I../libio -I../sysdeps/arm/elf -I../sysdeps/unix/sysv/linux/arm -I../sysdeps/unix/sysv/linux -I../sysdeps/gnu -I../sysdeps/unix/common -I../sysdeps/unix/mman -I../sysdeps/unix/inet -I../sysdeps/unix/sysv <dannym>-I../sysdeps/unix/arm -I../sysdeps/unix -I../sysdeps/posix -I../sysdeps/arm/fpu -I../sysdeps/arm -I../sysdeps/wordsize-32 -I../sysdeps/ieee754/flt-32 -I../sysdeps/ieee754/dbl-64 -I../sysdeps/ieee754 -I../sysdeps/generic/elf -I../sysdeps/generic -D__GNUC__=2 -D__GNUC_MINOR__=95 -Dunix -D__arm__ -Dlinux -D__ELF__ -D__unix__ -D__arm__ -D__linux__ -D__ELF__ -D__unix -D__linux -A'system(unix)' -A'system(posix)' -A'cpu(arm)' -A'machine(arm)' <dannym>-D__CHAR_UNSIGNED__ -D__OPTIMIZE__ -g -Wall -Winline -Wstrict-prototypes -Wwrite-strings -D__ARM_ARCH_3__ -D__APCS_32__ -D MES_BOOTSTRAP=1 -D BOOTSTRAP_GLIBC=1 -isystem /gnu/store/wdi63jz8ryxxcphikd9fc8yhzig0d4x0-gcc-core-mesboot0-2.95.3/lib/gcc-lib/arm-unknown-linux-gnu/2.95.3/include -isystem /gnu/store/clzk6rv0smi6hxb3agfy0vb35zakh4hj-mesboot-headers-0.22-123-g9cd754498/include -include ../include/libc-symbols.h gconv_simple.c /tmp/guix-build-gli <dannym>bc-mesboot0-2.2.5.drv-0/ccmwcf1b.i <sirmacik>how can I run guix pull from an installation image? <sirmacik>under tty3 it tell's me that there is no guix-profile for root <iis>How can I set gnu guix to automatically mount usb pendrives when attached? Actually, I open terminal and after plug in the usb memory stick, I run sudo fdisk -l then mkdir $HOME/pendrive ; sudo mount /dev/sdc1 and for unmount the following: sudo umount /dev/sc1 . This process lets me see the usb pendrive contents, but in thunar I cannot unmount because it requires root privilegies <efraim>I think I have a working prototype for re-integrating rust with the graph generation <efraim>and when we have reusable rust build artifacts it'll drop right in <dannym>janneke: Reason for crash is "#if MAX_NEEDED_FROM % MIN_NEEDED_FROM != 0" in skeleton.c <dannym>So I guess generally modulo and division for long long is broken :P <dannym>"#if (2 % 2)" also crashes with segv (division by zero) <dannym>The crash is in __moddi3 (double -> int conversion in there O_o) <dannym>What is up with those stubs... those are so weird <dannym>__moddi3 should NOT have doubles as parameters (at least in gcc) <janneke>initially i just wanted the symbols to resolve <janneke>then, on x86, it turned out that a stub implementation (return 0) worked <janneke>i guess the problem has to do with division on arm <janneke>tcc actually /using/ these stubs, then we (i?) started to add some half-baken implementations that "work" when compiling tcc <lfam>sirmacik: I downloaded the installer from our website, switched to tty3, and ran `guix pull`. It worked fine for me <lfam>Is that the same as what you tried? <sirmacik>now i got database disk image is malformed :f <sirmacik>is there a way to fix it without reflashing image on usb? <lfam>What do you mean you got database disk image is malformed? <lfam>My guess is that the process of writing the ISO to the flash drive was not completed <lfam>What tool did you use to do it? <sirmacik>but started giving me this error on the second one <lfam>When using dd for this, I always include "conv=fdatasync" <lfam>It's hard to say if you could fix it without having more detail <lfam>The fastest thing would be to reflash it <lfam>Reading the manual, it seems like "conv=fsync" is even better <dannym>janneke: Which branch should I do mes fixes on now so that they are easy to test in guix wip-arm-bootstrap? (those things are not ARM specific) <lfam>Does anyone have a list of all the Guix system tests? <lfam>I'm trying to run them all with `make check-system` but the test suite quits after the first failure <lfam>Hm, looking at Makefile.am I see there is a Guile script that can run them <lfam>Hm, it fails with "error: channel-source->package: unbound variable" and "hint: Did you forget `(use-modules (gnu ci))'?" <lfam>It needs to be run in a Guix development environment <lfam>Nice, it can be run with `guix build` so it accepts --keep-going <pinoaffe>mroh: it's on guix proper, and it does indeed seem to be the file dialog, since running `inkscape file.ext`works without any issues <pinoaffe>I don't know whether this is relevant, but I installed it to a separate profile, which I enabled by "sourcing" it's etc/profile <janneke>dannym: currently we have: master => wip (which is arm, used in arm-bootstrap) <janneke>i'd like to release 0.23 asap anyway, so i'd say "wip" is the best place <janneke>wip-arm-bootstrap is using mes "wip" ***zimoun` is now known as zimoun
<moopi>Hello everyone! I am a new user to Guix System and have a question. I type programmer dvorak though Xorg does not recognize it. Do I simply add the required lines in my config? I refer to subsection 10.6 Keyboard Layout. <htgoebel>moopi: There is an example: You need to add it three times: for the console, grub and Xorg. <lfam>Is that in the cookbook nckx? <lfam>Actually the manual section is quite clear <nckx>I was just reading it & thinking the same thing. I think it's been improved ☺ <nckx>moopi: What do you mean by ‘Xorg doesn't recognise it’ though? ‘us dvp’ is quite standard. <lfam>Well, not as clear as your example I suppose :) <nckx>I learn much better from examples than specifications, so... *nckx hurls at man pages with Backus-Naur. <moopi>Hi @nckx - thank you for your help! The manual is great! No so, I do not have the lines " ... (set-xorg-configuration (xorg-configuration ... " <nckx>I haven't used xorg for a year or two, I don't remember where it goes on non-Wayland systems. <moopi>No worries - the manual has everything. I will just issue "guix system reconfigure" afterwards yea? <nckx>If something breaks & you can't type your password you can select a previous generation at the GRUB prompt. Maybe keep the original GRUB keyboard layout until you've tried. <zimoun>civodul: if the TOC is accurate, it is 11 pages about Guix, wow! :-) <nckx>‘Je me suis désabonné depuis qu’ils ont arrêté de distribuer les PDF’ -- guess I'll stop looking for copies of questionable legality then. <kondor>civodul rekado_ what was the name of that tool which was mentioned during the community session? I should probably add it to the pad <nckx>Par Gaspar Émilien (gapz) <nckx>Which statement, leoprikler? <nckx>leoprikler: I was looking for a free-as-in-pirated PDF of the MISC issue that civodul mentioned. <rekado_>kondor: sorry, I wasn’t around during 99% of the discussion. Do you have more context? <nckx>Uh, I pasted the wrong link above and lost the right one to an IceCat crash. <lfam>I wonder about that site that crashed your browser! <nckx>Guix's icecat is so crashy ☹ <roptat>but it's limited to just the few first sentences <kondor>rekado_ it was mentioned in the context of, i believe, patch submission workflow (or, something related. More general context making it easier to contribute to guix) and civodul mentioned you as the author of the thing <nckx>roptat: Yess! Thank you! <kondor>I'm afraid i can offer no less vague context than that <nckx>I just wanted to note that it lists an author, as kondor said. <nckx>pinoaffe: ‘Guix, you'll have hardly any crashes’ shall be our new tagline. <nckx>When swayidle invokes swaylock all IceCat tabs crash, but now it took down the browser (and it was an incognito). <pinoaffe>I think the only times I got icecat to crash was when I ran out of memory/disk/processing power or when I killed random icecat workers to prevent any of those from happening <iis>recently I installed Gnu Guix OS. During the installation i selected to install i3 and xfce as Desktop Environment. But at the login there is no way to select what DE use, simply goes automatically to i3wm. What I should do or install to get the option to select i3 or xfce DE at starting of session? <nckx>Which display (log-in) manager do you use? I think there's one minimal one that doesn't list the session types unless you hit F1/F2 or whatever. <iis>nckx: I guess I have that you say. It is what I got after installation of GNU GUIX OS <iis>Should I install a login manager?, someone told me about SDDM <roptat>if you have multiple DE, you should be able to select one on the left of the field where you type your password, after you select a user <iis>roptat: how I can check that? <roptat>I think it should be a small cog wheel that you have to click <iis>roptat: That wheel is not present. Sadly I am writing from the same machine, so I'll have to re-logging to show you a pic of the login session. <civodul>zimoun: 11 pages looks fishy, there must be ads in the middle <euandreh>is there a function version of (content-hash "..") that takes a variable and computes the hash at runtime? <civodul>euandreh: what do you mean by "computes the hash"? <civodul>it also accepts a non-literal i believe <euandreh>the implementation even mentions that it should behave like a constexpr in C++ <iis>here again, I am know on xfce but it looks that something is not good. The mouse pointer is an X . the hexchat window is not possible to maximize or minimize or adjust the size, it lacks of the close bar I guess <roptat>the X is the default mouse cursor on X11 <iis>Even this text I cannot see after is written, because i can't expand hexchat window <iis>finally I could expand a little! <civodul>euandreh: the last clauses of that syntax-cases are for non-literal bytevectors, no? <euandreh>civodul: hmm, it looks like it, let me try to pin down futher my issue <euandreh>civodul: you were right :) I had an issue previouly where I was providing the hash dynamically, and I somehow confused it with giving the runtime hash value to (content-hash ..) <iis>thanks nckx and roptat. I am now using xfce. Is far different than the xfce I use on LinuxMint, but i'll read and try things to configure it. <roptat>iis, what you describe sound more like a bug though... <iis>roptat: perhaps. Thunar also doesn't have a way to expand or resize the window <nckx>Sounds like your window manager is missing/crashed. <iis>roptat, is far from it. The background is black, the windows of the programs cannot be resized <nckx>What with the X cursor & no decorations. <iis>where i can send a screenshot of my screen? ***bqv_ is now known as bqv
<iis>that image was taken using scrot from terminal, the xfce4-screenshooter doesn't exist as you see in the error message on the image. <nckx>You might need to install it separately. <nckx>iis: All we can say from that image is that xfwm wasn't launched, or that it crashed. <nckx>I don't know where XFCE or GDM log. <iis>nckx, so I should remove xfce first? how I do it? something like guix remove xfce? <roptat>no, xfce was installed as part of the system, not as part of the user profile <roptat>so to remove it you'd remove the declaration from /etc/config.scm and reconfigure <cbaines>civodul, you probably want to follow the link in the "View comparision" check <roptat>but before that, it'd be best if we could figure out what is wrong, and how to fix that *nckx AFK a while; good luck iis. <roptat>iis, can you find anything useful in /var/log/gdm? <civodul>cbaines: oh right; and from there can i directly see how this affects dependents? <cbaines>civodul, yeah, through "Compare package derivations" <iis>roptat, here are some files <iis>greeter.log greeter.log.1 greeter.log.2 greeter.log.3 greeter.log.4 <iis>iis@iish /var/log/gdm$ <roptat>I'd go with greeter.log, it must be the most recent one <cbaines>civodul, this change seems to rebuild the world, judging by the number of affected packages <roptat>do you have a file that looks like ~/.xfce4-session ? <cbaines>civodul, there's no "number of affected builds yet", there probably should be. I just ticked "All results" and watched my browser struggle to load the page <iis>roptat, I am looking for it ~/.xfce4-session. <pkill9>i don't understand this, if i run `guix build -n quassel` it returns the store path, yet when I run `guix build -n qtbase` it says it will download qtbase, but qtbase is an input to quassel so it should already be in the store as quassel is already downlaoded <civodul>cbaines: are you looking at "package derivation changes"? <roptat>iis, or maybe a .log file in .cache/xfce4 or .cache/sessions ? <iis>roptat, no theres no ~/.xfce4-session. I am looking in .cache <cbaines>civodul, ah, so the derivations have changed, but the outputs haven't. guile-json looks to be used in lots of fixed output derivations <cbaines>I didn't anticipate that when I wrote the page... I guess it's both correct, but a little confusing <civodul>perhaps you should map it back to packages via outputs <iis>roptat, there is ~/.cache/sessions/xfce4-session-iish <pkill9>infact this produces the same store path for qtbase as shown in `guix build -n qtbase`: guix gc --references $(guix build quassel) | grep qtbase <iis>roptat, iis@iish ~/.cache/sessions$ ls <iis>thumbs-iish:1/ xfce4-session-iish:1 xfce4-session-iish:1.bak <iis>iis@iish ~/.cache/sessions$ <civodul>cbaines: anyway, i wanted to check the build status of dependents; how can i see that? <zimoun>with Geiser, how do I go to the symbol definition? M-. (geiser-edit-symbol-at-point) returns an error about modules. I guess, I need to do something before. I have tried various things. <cbaines>civodul, the information will be on the "Package derivation changes" page, but that isn't helpful since every package derivation has changed... <cbaines>I need to look again at how this page works, and what changes it should represent, and what changes it should probably ignore <roptat>iis, sorry, that doesn't look very useful :/ <civodul>i'd say yes, but it's almost staging-y <pkill9>oh it doesn't actually download it <iis>roptat, I don't find the .xfce4-session.verbose-log file which is present on xfce linuxMint. Something is wrong here with xfce4. But I don't know how to uninstall xfce4 <iis>I'll gonna exit, reboot and re logging again to see if something changes. <raghavgururajan>leoprikler: I just found that those error is not related to GNU/Linux. One is for Windows and one for MacOS. <pkill9>can anyone else reproduce this? running this command returns nothing, indicating that xdg-utils isn't a referrer for the package guix gc --referrers (guix build qtbase | grep -v debug) | grep xdg-utils <pkill9>but, xdg-utils is an input in the qtbase package <leoprikler>packages using xdg utils have the nasty habit of invoking them via shell IIRC <lfam>pkill9: Mixed up the meanings of "referrers" and "references"? <pkill9>oh yea, it still doesn't show up in --references <lfam>In that case, references to qtbase won't protect xdg-utils from garbage collection <lfam>If it's a run-time dependency it needs to be referenced or propagated <pkill9>leoprikler: quassel uses qtbase, which is calling xdg-utils via the store path, idk if that's good or bad but it's not working because for some reason xdg-utils isn't considered a reference <pkill9>lfam: but it's in the (inputs) field in the qtbase package <pkill9>anything that's in the (inputs) field should be protected from garbage collection <lfam>"References" are strings that look like store items, found in the built output of a package <lfam>After building, the daemon scans the built output and looks for these references. When it finds them, it records them in the database <lfam>Everyone has this TIL :) <pkill9>qtbase seems to have the path to xdg-utils hardcoded, so still idk why it's not considered a reference <lfam>Why do you think it's hard-coded? Do you see the filename path in the qtbase store item? <pkill9>i see the filename store path in the terminal output in quassel when it doesn't exist <lfam>It's likely coming from somewhere else. It could be found at run-time somehow <lfam>It's not a reference in Guix terms, though <pkill9>i hadn't updated xdg-utils though <lfam>If qtbase *does* refer to it, but the reference is obscured somehow (like compression), that's a bug <lfam>You'll have to poke around <jlicht>Not saying someone should do so, but you could force a reference by creating a file (e.g. ".guix-reference") which lists all the references you want/need to keep, no? <lfam>Sometimes there are spurious references, basically created like that <lfam>I missed the beginning of this discussion... why does it matter if qtbase keeps a reference to xdg-utils? <pkill9>because I think it hardcodes the store path to it so it can open things <lfam>efraim: You use quassel, right? ^ <pkill9>unless it has the store path to xdg-utils that it's qtbase input was built with <nckx>jlicht: That's what I thought propagated-inputs were until I got a clue. <lfam>We updated Qt recently, so maybe it stopped working <lfam>If it used to work, I would check if that 'patch-xdg-open' phase still works as expected <cbaines>I'll make it an option at some point <cbaines>unfortunately, not all the builds have been done yet, as the priority would have been quite low, due to the number of "changed" derivations <cbaines>turns out the query already had two ids representing the respective sets of outputs, so the change to the query was minimal <cbaines>I overcomplicated it at first though <civodul>cbaines: something we didn't mention this morning: flaky packages or test suites <civodul>looks like the Coordinator is already equipped for that <cbaines>yeah, making it easy to retry automatically is really useful whether you want high substitute availability, or to work out what packages really fail to build, or are just flaky <cbaines>I think that's one of the key factors why guix.cbaines.net still has higher substitute availability (at least for x86_64-linux) than ci.guix.gnu.org, even though it has far less compute resouces <choozy>Is there a plan to also release images of GNU Guix with the Hurd kernel as ISO? <lfam>We have tools built-in to create Hurd images. I know that's not answering your question <lfam>One could install Guix on another distro and then create the Hurd image <pkill9>i'm gonna try building qtbase with the phase modified to just use xdg-open instead of the store path ***p is now known as Guest16037
<Guest16037>which is the better way to remove guix and purge all the filesfor reinstall <Noisytoot>Guest16037: rm -rf ~/.config/guix ~/.guix-profile && sudo rm -rf /gnu /var/guix <Guest16037>the locale config is crazy and icecat isnt able to show characters in the right way ***rekado_ is now known as rekado
<lfam>You might need to install a font Guest16037 <leoprikler>As for locales, I think we're lacking a bit of info on that, but it's something along the lines of "missing locale" when running guix commands, right? <leoprikler>substitute <something> with the font you want, e.g. dejavu or google-noto <leoprikler>Just take into account the Unicode coverage of those fonts. <leoprikler>yep, hurd is still very much a wip operating system <leoprikler>Plus we're planning to rewrite it completely in WebAssembly on April 1st. <lfam>I'm noticing failures of xorriso in the system tests. I wonder, could the recent update have introduce problems? <raghavgururajan>Guest16037: Yeah, in that case, you can purge as Noisytoot mentioned, and re-run guix install script. <leoprikler>Bear in mind, that you shouldn't need to purge it. <Guest16037>using comand Noisytoot mentioned as root, the output is that is not posible to remove /gnu/* <Guest16037>anyway i remove the binary so now i cant try to fix the locales <leoprikler>you might need to install glibc-locales if you're running on a foreign distro <Guest16037>i am running the shell script installation to recover the guix binary <efraim>lfam: I use quassel, I can click on links and it opens them in icecat <leoprikler>I sadly can't ever recall needing to do so on Guix System. <lfam>pkill9: See efraim's message about quassel ^ <efraim>Single click, double clicking on the links doesn't do anything 