<kyamashita>Quick question: if I were packaging the pygame library, would that go under game-development.scm or python.scm? I'm leaning more toward game-development.scm right now. <bavier>kyamashita: game-development seems appropriate <kyamashita>bavier: I assume a similar opinion for libraries built on top of Pygame, like onpon4's SGE Game Engine? <davexunit>ACTION is a little frustrated with people saying that it is hard to contribute to guix <davexunit>it is, without a doubt, *significantly* easier to contribute to guix than any mainstream distro ***fkz is now known as Guest403
<Sleep_Walker>we spent a lot of effort on making easy contributions to openSUSE and it doesn't require Scheme (which is not exactly the most popular language around) <Sleep_Walker>we have almost the same way as github/gitlab - fork, fix, submit, review, accept/reject <lfam>We have basically the same workflow: clone, fix, submit, review, accept/reject <lfam>Sleep_Walker: Any advice on how we could make it easier to contribute packages to Guix? <Sleep_Walker>and sometimes Guile error messages are not exactly helpful <lfam>I was mystified the first time I read a Guile backtrace <Sleep_Walker>is there a way how to override the list of required kernel modules without reimplementing initrd generation code? I have my own kernel configuration and have all code compiled in (to disable module insertion at all) <lfam>I'm not sure. There has been recent discussion on help-guix about using custom kernel configuration. Would that help? <Sleep_Walker>it seems that they're still struggling with kernel build and not yet initrd troubles :) <Sleep_Walker>the thing is that I'd need to express that that I have no modules to be added into initrd from system configuration, unfortunatelly it is now hardcoded and not set as some default value <Sleep_Walker>maybe optional ignoring errors on module non-existence can be easier <ng0>one thing i like about guix, people no not read between the lines and assume you have magical knowledge of discussions which were not visible for you. If I were not interested in solving it and had 2 people with positive reactions, I'd just unsubscribe from the bug at gentoo and solve it in private. <wingo>ng0: you can always read the archived chat logs to know the context of a conversation :) <ng0>this was not related, sorry <ng0>just a general feeling of rant towards guile bug at gentoo <ng0>i ask clear question and point out clear points of failure, make suggestions and all I get is minimal answers with assumed knowledge of internal discussions <jlicht>my freshly pulled guix is complaining about the source hash for wxwidgets@2.8.12 <jlicht>before I assume I am being singled out by some nefarious government agency, is anyone else also noticing this? <civodul>with --no-substitutes i get: `/gnu/store/9q5n4985lh5bp0xavwddc13lr7dc8yfl-wxWidgets-2.8.12.tar.bz2' should have sha256 hash `1gjs9vfga60mk4j4ngiwsk9h6c7j22pw26m3asxr1jwvqbr8kkqk', instead has `01zp0h2rp031xn6nd8c4sr175fa4nzhwh08mhi8khs0ps39c22iv' <civodul>so i think it's been modified in place <civodul>could you propose an update on the list? <jlicht>civodul: as long as this update entails looking at the changed files and subsequently changing the hash, sure :-) <jlicht>civodul: found some more sourceforge-hosted source archives that seem to have been slightly modified <Sleep_Walker>it would be almost year anniversary since last incident ;b <jlicht>is there some way in which to update all of my packages to their latest versions (via package -u), but exclude texlive? <jlicht>nvm, just found the --do-not-upgrade flag <ng0>once a year they change? maybe we need a sourceforge-fix-changes thing <frerbst>Hi! Currently I'm installing GuixSD on an old machine to play around with it a bit and test it. I have the problem that "guix system init" keeps failing due to "networking issues" while downloading python2-setuptools-18.3.1. It always downloads exactly 345KiB and fails then. I tried to build it locally, but this adds it to the store of the live system (I#m not very experienced with Guix, but this is what I guess). Any ideas on how to continue? <ng0>i have a solution… which differs and will need time. a byproduct of the gnunet-fs thing. <ng0>frerbst: you need to restart ncsd or what the name was <ng0>herd ncsd restart (I'm not good with remembering services) <frerbst>ng0: Shouldn't it be "herd restart ncsd" (which fails with service ncsd not found)? <ng0>it probably isn't named ncsd then <ng0>we should have this in the manual.. maby there's even an open patch about it <frerbst>ng0: So what is this service doing that makes the download failing in the middle of the process? <frerbst>Or not doing, if it has to be restarted. <ng0>it's the dhcp daemon i mean <efraim>its related to name-server recognition iirc <ng0>herd restart dhclient? check if you can ping gnu.org, else it has to be restarted <efraim>civodul: I rechecked the hash of gnuplot that I got from updating it, the hash is 08vpmhl85l48xcccx8jrkamalih2d6z9ppqpsppwii9y2l1p3297. I'm sure its the one I got from one of the sourceforge mirrors <civodul>efraim: could you compare the old and new tarballs? <efraim>i'm putting a copy at flashner.co.il/~efraim , old is the one I downloaded, new is the one I just downloaded <efraim>i'll look at them too, in case anyone else wants to take a look at them also <efraim>looks like they might've changed the tarball themselves <efraim>diff -r? I should've read the man pages <efraim>it doesn't look like its malicious <Sleep_Walker>it does look like two different snapshots where the older one is used instead <civodul>Sleep_Walker or efraim: could you post the two hashes and the diff to the list, along with a patch to update the thing? <efraim>you did new to old, old to new it looks like they added a check for something <koosha>I've just installed the openssh package . I wanted to start the ssh service but I don't know it's name . What's it's name ? Thank you . <civodul>currently there's on sshd (OpenSSH) service in GuixSD; there's lsh-service instead <koosha>Hi , so you mean I can't use openssh ? <davexunit>koosha: also note that installing a package doesn't install system services to the global system config <davexunit>packages and services are completely separated, unlike in Debian and co. <davexunit>you can start the openssh daemon manually using the software provided in the openssh package, however. <koosha>davexunit: Oh , I tought it's like Debian . thank you . <davexunit>we don't do imperative global config changes like that <koosha>So how can I start the openssh daemon ? <davexunit>you can start the daemon program manually if you'd like. <davexunit>for it to be integrated into guixsd, you need to write some Scheme code to create a new Shepherd service. <koosha>Yes , I see that . But how could be done manually ? <davexunit>I don't know the ins and outs of every program. <davexunit>all I know is that the openssh package provides the client and the server, so you can start the server somehow. <koosha>Sorry , I'm noob . I installed lsh package but it seems that lshd is not installed . <civodul>davexunit: really a quick hack, but i hope Someone can make a proper guile-libgit2 package eventually <davexunit>civodul: I sent a reply to the list that I had started such a project awhile ago <civodul>i have a slight preference for GPLv3+, but LGPLv3+ is fine too <davexunit>me too, but I don't see a particular reason to use GPLv3+ for this <rekado>I have a weird error trying to build Guix: <rekado>./random:156256:1: error: exponent has no digits <rekado> 000000EEEEEEEEEEEEEEEEEEEEEEEE000000EEEEEE424242848484E6E6E6808080424242EEEEEE <rekado>there was a file "random" and for some reason it affected the build. <rekado>(the file was a result of unpacking a tarbomb) <rekado>it affected the build of nix/libutil/libutil_a-archive.o <lfam>So, how many SourceForge package hashes were changed in place? :( <lfam>I spot-checked while updating the URLs, but I didn't verify all of them <jlicht>at least two of them (gnuplot and wxwidgets) <lfam>I don't like this news :( <efraim>gnuplot was a rerelease, so that one we can pin on them <lfam>Very happy to read about Git bindings for checking signatures <lfam>I subscribe to bug-guix so it was in my Guix mailbox :) <lfam>Currently re-downloading all the source tarballs of our SourceForge packages... <civodul>lfam: thanks a lot for the SourceForge work! <lfam>It's a significant set of packages! <lfam>And it's something I'm able to do, so I'm happy to do it :) <civodul>yes, you're doing everyone a great service <jlicht>what are you using to download all the source tarballs of our SF packages? <jlicht>I might want to download _all_ sources, as I'll be without Internet for about a day <lfam>I use a shell for loop to run `guix build --source --no-substitutes` on each one <lfam>I think downloading *all* the source code might take a while <lfam>This has already been running for over 30 minutes, and that's only the SourceForge stuff <lfam>It would have been smarter to use GNU Parallel <lfam>I'll send my findings to the list <Sleep_Walker>I got HTTP error 410 Gone on texlive-texmf-minimal-2016 from hydra <lfam>Sleep_Walker: Add --fallback to the command <lfam>We had to stop serving that because it's so insanely large <lfam>The method of not serving it is sheduled for improvement ;) <lfam>Sleep_Walker: Hydra doesn't build it anymore <lfam>There's been some discussion of the pros and cons of different methods of addressing this problem on guix-devel. Feel free to chime in <Sleep_Walker>I'm not sure now which package pulled this as dependency <Sleep_Walker>I have worked on patch where I can specify no kernel modules and I didn't get to the testing yet :) <rekado>ACTION just pushed a commit without signing it :O <lfam>Yeah... something always pulls it in. Since they seeem to update infrequently, I've made the texlive-texmf source a gcroot on my systems (symlink it to /var/guix/gcroots) <rekado>still haven't set up a subkey for my workstation in the office. <lfam>I think there's a repo option we can turn on to only allow signed commits <rekado>many of us have been pushing only signed commits for many weeks now. <lfam>Yes, I checked recently, and prodded some people to upload their public keys, so I can verify the entire signed Git history now, IIRC <lfam>I think we are ready to make the change, assuming we have access to that level of configuration on Savannah <lfam>Can anybody do `guix build --source --no-substitutes utfcpp`? <rekado>FYI: I'll be traveling (probably without internet access) for one week, starting tomorrow. <jlicht>lfam: I get stuck in a redirect loop <jlicht>but I might have some gnutls issues as well <jlicht>lfam: no wait, it ends up at tarballs.nixos, where I get a 404 <lfam>I thought I was having gnutls issues with the sourceforge packages, but I think that error was unrelated <lfam>I can provide a "raw" URL, copied from my web browser. But when it is constructed with Guix's sourceforge mirrors and string concatenation, it fails <lfam>I wonder if it has something to do with the whitespace in the URL <lfam>When I started fixing all the other URLs, I found that the http -> https redirection only happens when the http URL is bad. So, I think the gnutls warnings are spurious in this case <efraim>I think some of the lx* packages moved to github, but that might've been the lxqt packages <bavier>lfam: IIRC civodul already enabled the signing option on the repo; it'll reject non-signed commits for some weeks now <lfam>bavier: Apparently not. See commit f21403e2b <efraim>tig doesn't show if its signed, what are you guys using? <lfam>`git log --show-signature` <lfam>There is also `git verify-commit` <civodul>do you know of a push hook that i could ask the savannah admin to install? <civodul>a hook that would reject unsigned commits <lfam>civodul: The pre-push hook I shared a while ago could show the way, although it uses `git verify-commit`, which requires the public keys and GnuPG to be available <lfam>So, perhaps it's not appropriate in this case <lfam>The question is, do we want to only accept signed commits, or do we want to only accept verified commits? <lfam>The latter is ideal, but it's significantly more complicated <efraim>we could scrape the keys from savannah and create a keyring to compare the pushed commits against <ng0>how would I link to a collapsed entry on the packages html list? <lfam>ng0: I don't think there's a way to do it. I've found that the anchor links always unfold the entry. <lfam>Why don't you want the entry expanded? :) <lfam>Add #foo to the URL to link to package foo <ng0>updating curl at the moment, then I will create a version bump patch and right now I'm writting a message to curl with information to list us <lfam>ng0: We have to update curl on core-updates, since so many packages depend on it <lfam>Your link works, too. I never tried that :) <lfam>And I was wrong. The packages are not expanded. <lfam>ng0: On core-updates, curl is 7.47.0 <ng0>okay I'll do the update on core-updates <lfam>Let's hold it for core-updates next unless it's a security update. We are just finishing the current core-updates cycle, finally <ng0>i wanted to read up on the notes, it could be a security update <ng0>i see no cve, but in a previous version above 7.47 there is <lfam>No, we decided not to update in response to that. If curl was easier to update, I would have done it. <lfam>But, if I had known the current core-updates cycle would keep being restarted, I would have updated there <lfam>Next cycle I will be less timid about changing things <pmikkelsen>Hi guys, am I right when I think the only way to update the system is to run `guix system reconfigure`? <adfeno>pmikkelsen: "update" to what extent? <bavier>pmikkelsen: users may update their own software with 'guix package -u' <pmikkelsen>adfeno: update all the system packages, like the kernel, xfce, mesa etc. <pmikkelsen>bavier: Yes I know that, but I dont have a lot of packages installed compared to what the system has <bavier>pmikkelsen: yes, 'guix system reconfigure' is the way to update the system packages <ajgrf>i'm packaging a python program. anyone have an idea of the correct way to get the function call ctypes.util.find_library("c") to work correctly in guix? <lfam>ajgrf: I don't, but there are two results when grepping for ctypes.util in gnu/packages <lfam>Maybe they will show the way? <lfam>It dies in 'fix-path-references' <lfam>In unknown file: ?: 1 [string-append "\\"" #f "\\""] <janneke>looks like a problem with gs (or sh) <lfam>Yes, although I don't see why there would a problem <janneke>ghostscript is listed in inputs, shouldn't it be in native-inputs too? <lfam>If it's in inputs, it will be available while building <lfam>native-inputs is more appropriate when it's only needed for building. I don't know whether or not that's the case with lilypond and ghostscript