<mikadoZero>The emacs and guile-emacs packages have the same synopsis and description. What is the difference between them? <mikadoZero>guile-emacs has three additional dependencies guile-for-guile-emacs, automake and autoconf. <nckx>mikadoZero: guile-emacs replaces the Emacs Lisp interpreter with Guile. So the descriptions matching isn't a bug, but guile-emacs missing an extra line mentioning this important detail is. <nckx>(It's also still buggy IIRC.) <g_bor>ok, now I have a final word on the neovim segfault issue, but I still don't havev a solution. <g_bor>It is curses that returns an error on a terminfo lookup, and the libtermkey code is not handling that. <g_bor>It looks like the lookup should not return an error in the first place, and it also looks like that correct error handling should be added at the call site. <mikadoZero>nckx: I could submit a patch for the guile-emacs description. <nckx>mikadoZero: Very welcome! <nckx>(To be clear: elisp isn't being replaced, just the interpreter.) *nckx .oO Or should we call it a compiler now? <rvgn>I have a confusion with understanding LISP. 1) Is LISP a programming language by itself or a programming language model? 2) Scheme is a dialect of LISP right? Is dialect a synonym for paradigm? <rvgn>So LISP in general is a language model not a language. Using that model there are many LISP languages which are LISP dialects. Am I on right track here? <mikadoZero>rvgn: Yes there is the lisp family of languages. <bavier>any reason we haven't upgraded our patchelf to version 0.9? <mikadoZero>nckx: On further inspection my find was a symptom of something larger which I have submitted to the guix-devel mailing list. ***catonano_ is now known as catonano
<leungbk>i was thinking about packaging a rust app, but i noticed that there aren't any packages that invoke the cargo-build-system. is there no interest at present for rust packaging bc of the long bootstrapping process? <bavier>leungbk: afaik it's just a difficult process, the cargo-build-system is a wip <lfam>"It would be nice" if (guix build union) was able to create in a union in a directory that already exists <lfam>I'd like to build a union in the root of the TMPDIR <lfam>However, mkdir fails because TMPDIR is already created ***kensington_ is now known as kensington
<g_bor>From a file with guix package -f this almost works, it misses the module (guix gexp) <g_bor>could not yet find out how to make it work with -e ... <efraim>looks like go doesn't like compiling from arm to arm64, got an integer overflow <efraim>I might have to revive my gccgo patches <g_bor>I've found out what is happening in neovim. <g_bor>I just don't know why, and don't know what would be the correct fix. <g_bor>The story is that curses returns a -1 from tigetstr, that is passed to the hook in neovim, but the hook is not prepared for that, and the -1 is passed as a pointer causing the segfault. <g_bor>Only two question remains: why curses returns -1, and should we fix the hook, and if yes, how... <efraim>i assume you checked if any other distros are getting this issue? that's normally my first check, see if anyone else already fixed it <efraim>could it be that we're using an "older" version of ncurses? <g_bor>efraim: did not check it yet. Good idea. I still feel that, even if the curses is faulty that the hook should also be fixed, as a segfault on startup is not a really nice ux. <g_bor>It seems that there are some unfortunate combinations... like this: <g_bor>however we don't use unibilium... <g_bor>efraim: on the upstream tracker I see no more instances related to that, and we are on ncurses 6.1, so that should not be the issue either... ***dddddd_ is now known as dddddd
<nly>I can't even connect to bluetooth device without starting pulseaudio first <g_bor>It seems that we can have the neovim segfault resolved soon. <g_bor>I talked to someone on their IRC and it seems that a one-line patch can solve this. <g_bor>roptat: I promised them a bug report and a pull request later today. <g_bor>How should we handle this on the guix side? <g_bor>Would it be better to have a patch, or should we just add a build phase? <g_bor>I guess I would go for the build phase... <g_bor>ok, I will file the pull request first, so that I have something to refer to on the comment. <rekado>nly: the audio application should cause pulseaudio to be started automatically. <rekado>nly: if you have a headless system and you want to use bluetooth audio I suggest using bluez-alsa. <rekado>I’m still having trouble with mpd playing back audio over bluetooth this way, but for all other players it seems to work fine. <nly>I got as far as running `bluealsa` and the device connects with no pulseaudio running <nly>Trying to play any sound results in `Unknown PCM` or something like `no file` <nly>package-description contains newlines which is a problem <rekado>nly: you’ll need to set up a new device in /etc/asound.conf <rekado>I’ll probably add a service for this in the next few days <nly>that would be very nice <rekado>you’d still need to pair and connect the device manually, though (with bluetoothctl) <nly>huh, it connects automatically after (in bluetoothctl) `pair` and `trust` <nly>and one more thing, `bluealsa` must be running ***nly is now known as nlyy
<civodul>roptat: going through my backlog i see a lot of notifications from the TP to guix-devel <civodul>should we do a mass update of all the .po files? <roptat>I remember pushing the danish and german translations recently <roptat>I see there's a Swedish file from yesterday too ***nlyy is now known as nly
<roptat>ah, there's a new German po file for the shepherd, but I don't have commit access <roptat>I can take care of the Swedish po if you want <rekado>civodul: looking at guile-json I see that version 3.0 switched from the hash table representation back to alist. <rekado>so… I wonder if maybe the upgrade to 3.x won’t be so difficult after all. <rekado>the only difference may be JSON arrays that are no longer defined as lists but as vectors. <rekado>civodul: I saw your new slides just now. Looking good! <civodul>rekado: re guile-json@3, yes, it switched to alists, so the main differences are that it used to return hash tables and no longer does, and returns vectors instead of lists <civodul>rekado: speaking of the talk, it was interesting because it was a workshop on containers for HPC <civodul>and they invited me initially to talk about container image generation with Guix <civodul>so of course i slightly retargeted the topic ;-) <civodul>the person after me presented a web UI to generate... Singularity def files <civodul>someone from the audience asked: "wait, shouldn't you generate containers directly via Guix?" :-) <civodul>so we need to get guix-pack-as-a-service going <rekado>I can get some old file servers soon; could help with hosting this. <efraim>Sounds like a use case for my travisci config <g_bor>Regarding the video repository, one thing that should be done is to license the files, and set up a licensing policy for contributions. Wdyt? <civodul>it should be copyleft, but we need to think about how to choose between GPL/GFDL/CC-BY-SA <g_bor>civodul: that does not seem trivial... <g_bor>I would say that the code parts are fine by the GPL, but not sure about the rest... Also would we like to use a single license for the whole thing... <g_bor>I would rather say no. There are two types of very different materials mixed. <bgardner>Is postfix not available due to licensing issues or does it just need packaged? <roptat>bgardner, I'm not aware of any licensing issue with postfix <g_bor>civodul: Maybe we could go for GPL for stuff not in the video directories and CC-BY-SA for the content of the video directories. <civodul>g_bor[m]: i haven't looked into details, but that sounds like a good plan <rekado>g_bor[m]: since the documentation plays a role in the videos please make sure that the licenses are compatible. <quiliro>i am not using guixsd now...i had a problem on epiphany with fonts...arial is not displayed...i got squares with numbers on certain characters <quiliro>but i have some difficulties exporting <quiliro>has anyone had success exporting from emacs? <quiliro>but i am very enthusiastic...i found elisp intro is a great way to learn the syntax of guix <quiliro>i know it is a different variation...but that manual is great to introduce basic concepts <quiliro>does anyone know how to make the desktop or even just emacs tell me when there are new messages on emacs irc (erc)? <mikadoZero>quiliro: What do you mean when you say exporting? The function that is bound to C-x C-e is `eval-last-sexp`. <mikadoZero>quiliro: For rcirc: (add-hook 'rcirc-mode-hook (lambda () (rcirc-track-minor-mode 1))) It is in the rcirc manual. As well as omit-mode which I like. <mikadoZero>quiliro: You might have more success with emacs question in an emacs channel. Like #emacs or #emacs-beginners. <quiliro>mikadoZero: rcirc is the same as erc? <bavier`>nly: the definition is infinitely recursive <nly>there is somehow an origin object in this '(map cadr (package-inputs package)) <bavier`>nly: package inputs are not necessarily packages <bavier`>so if you're interested in only the packages, you'll have to filter them out <Copenhagen_Bram>How does guixSD decrypt a LUKS partition to something like /dev/mapper/rootfs as specified in the config file? <rvgn>Copenhagen_Bram By the use kernel module dm-crypt. <mbakke>Copenhagen_Bram: Yes, the decryption occurs in an initrd. <mikadoZero>quiliro: I have not had trouble exporting with org-mode. I suggest you read the org manual. orgmode.org or C-h i in emacs. *Copenhagen_Bram is installing gentoo and was wondering how GuixSD did it <nly>C-x C-e, I think you meant C-c C-e for org-export...? <quiliro>mikadoZero: i did what the manual said...it works on debian...not on GSD <g_bor>i had a look around the licensing stuff, and it turned out to be inconclusive. <g_bor>it's all fine. But I am confused. <g_bor>Now we have the documentation videos, and there are at least two incompatible licenses in play. <g_bor>it is not even very clear to me what does it mean for example, that gpld software cant include gdfl licensed text... <quiliro>g_bor: they are licenses which prevent changing license. Including a gfdl text on a gpl work would chang the work's license <quiliro>one can't be included inside the other <quiliro>one work must be separate from the other work with incompatible licenses <quiliro>as far as i can understand...i am not an expert...does what i am writting make any sense? <g_bor>quiliro: that is exactly my understanding, but what does include inside each other, and keep them separate mean? <g_bor>they should not be in the same file? Or in the same repository? <bavier`>ianal, but consider the case of including a source listing of a gpl program in an appendix of a gfdl'd manual <g_bor>Just take an example: guix is gpl. The manual is gfdl. Is this not a problem? <quiliro>i think bavier's example is tolerable because the amount of code is trivial and g_bor's example is solvable by specifying which file is gpl and which is gfdl <bavier`>I think "separate" here means that neither depends on the other <bavier`>the guix program does not require the manual in order to run, and the manual can be read and understood without the program <quiliro>still...the license should be modified to be compatible <ng0>hm.. what about a license "gpdl OR gpl3" choice/dual licensed? i don't remember if it was checked by lawyers then, but I think it was okay. Or is this bda too? <vagrantc>where are the architectures defined in guix? ... i was wanting to try and get a riscv64 cross-compiler; i think it's supported upstream in the various toolchain bits with recent versions of gcc/binutils/glibc/etc <lfam>Do we have a handy procedure for creating the *name* of a temporary file or directory? Like a way to generate a short random string that is safe for the filesystem? <lfam>I don't want to actually create the path; it will be passed to union-build <kmicu>g_bor2: what do you mean by ‘software including text’? <lfam>I am getting lots of truncated build logs from Berlin <civodul>lfam: yeah it seems there's a bug that has to do with offloading <civodul>rekado: i just saw the commit that adds guile-studio, yay! \o\ <mikadoZero>`which recsel` say recsel not found and `guix package -s recsel` returns nothing. I would like to use it to refine package search results as demonstrated in the manual. Where is it located? ***renich_ is now known as renich
<janneke>hmm, i get some errors that i don't understand <janneke>building /gnu/store/nxcq7hw103q2ryncnpa54cdfs7mninhq-gcc-toolchain-8.2.0.drv... <janneke>while setting up the build environment: executing `/gnu/store/d1yz95qbxk4lgbdsi9r4i9kbvzfvc49p-guile-2.2.4/bin/guile': No such file or directory <janneke>things seem to be missing from the store, what could be going on? <lfam>janneke: Try `guix gc --verify` <g_bor>I was about to ask if you ran gc recently... <lfam>janneke: There are further options to that command but they are expensive and potentially dangerous <lfam>If something is missing then something is wrong <janneke>g_bor yes, i ran gc with some specific store paths (guile, gcc) <janneke>after i succesfully setup an guix environment for arm, gcc was missing (dangling symlink in the environment) <janneke>so, i guix gc'd that gcc and then things apparently got worse *janneke now runs sudo guix gc --verify=contents,repair <lfam>janneke: If you are working from a Git checkout, remember that after gc you probably need to do ./configure again *g_bor rebuilding guix, hoping nasty backtrace go away:) <janneke>lfam: yes, thanks -- i have some of those around; running off of guix pull mostly nowadays for non-bootstrap work though :-) <janneke>but dannym did some nice arm things again for mes, and i want to run it! :-) *vagrantc got to learn the excitment of "guix gc --verify=contents,repair ... *sighs* <civodul>janneke: the "while setting up the build environment" error, is it because the file actually doesn't exist, or because its ELF interpreter doesn't exist, for example? <civodul>lfam: great if we can get that fixed! <g_bor>lfam: I'm so happy about that! <janneke>civodul: oh crap, that's it -- guile exists but it's an ARM executable <lfam>I need to do some final tests, update the annotations, and write the commit message, but it seems to work <civodul>janneke: pheew :-) it's still weird that you get this error instead of the one that says you're on the wrong system <civodul>but yeah, maybe you're doing something weird ;-) <civodul>i saw all the ARM commits by Danny and that's exciting! <janneke>lfam, civodul: the problem is solved -- thanks! <rekado>cbaines: I think it may be good to split (gnu packages ruby), so that we can keep the module closure small. <janneke>i need to bother dannym about this, he suggested a nice trick to have qemu for arm show assembly code in single-step upon a segfault, or something <rekado>cbaines: let’s do this after all of your ruby patches have been merged. <g_bor>there seems to be something wrong with guix refresh here... <g_bor>Maybe I'm doing something wrong... <g_bor>It is a wrong argument in position thing, but only when I call it with -u ... <mikadoZero>Does `guix package -s ""` return all guix packages? <vagrantc_>got as far as a kernel panic on the asus-c201pa ... but somehow the wrong uuid got specified in my config.scm ... trying again...