<liberalartist>What's the current state of a npm importer? I'm not a Node developer, but I just started looking at a package where it would be very useful. Even the "binary" importer I've heard of would help. <iskarian>liberalartist, as far as I'm aware there's no npm/node importer, but you might try to get in contact with the contributors to the node build system <iskarian>at first glance it looks like it's a single commit, so you could probably just cherry pick it into wherever you want to test it <ixmpp>Is there a nice way to add something to the user PATH declaratively? <ixmpp>Actually, might be an rde flag... <ixmpp>Or not... there's always zshrc i guess <PotentialUser-10>Hello. Will I be able to bring up Xen using Guix as host (dom0)? I installed Xen from the repository, but I do not understand how to enable it, usually the kernel is simply replaced, I did not find any documentation either <PotentialUser-10>I'm worried about how much of an impact the arrival of corporations and SJWs will have in the GNU world? To what extent is RMS radicalism justified? <vagrantc>PotentialUser-10: i don't know that anyone has actively worked on using Xen with guix *vagrantc is happy to see some corners of the GNU world realizing the connections to free software and social justice <dstolfa>i imagine that it's not different to any other boot time option you'd pass to linux and the grub configuration you'd do on guix, though maybe not as polished as one would like it to be <vagrantc>it's been ages since i've used xen, but the grub configuration for dom0 was a bit different than a typical linux distro ... <dstolfa>vagrantc: yeah, but i don't see why you couldn't do it using the guix bootloader configuration record <vagrantc>e.g. you pass the hypervisor as the kernel, and then pass the kernel and initrd as modules or some such <dstolfa>(note i say could, i haven't done it and it's probably not as polished as other more common options) <vagrantc>dstolfa: it's different enough that it may require improving the guix bootloader configuration record <vagrantc>theoretically, yes, but it's not a simple matter of just passing the right arguments <dstolfa>vagrantc: i mean, i'm all to aware that it's probably going to be a pain... i've set up xen from scratch on a server recently :P <dstolfa>but i don't think there's anything really stopping you from going through said pain <dstolfa>except maybe wanting to keep your sanity *dstolfa prefers KVM since it's infinitely easier to set up <vagrantc>PotentialUser-10: so, the short of it is, theoretically possible, but you may have to get your hands dirty and submit patches to get it working reasonably <vagrantc>PotentialUser-10: GNU has also been tightly coupled with a corporation since 1985 ... <jackhill>muradm: I think it'd be okay to send shepherd patches to guix-patches@gnu.org <jackhill>just mark in the subject that it's for shepherd <muradm>jackhill: do you think this is ok: [PATCH] shepherd: add keep-tty flag to skip setsid ***irrealis1 is now known as optative
***optative1 is now known as optative
<jackhill>muradm: that's probably ok, but I think the common pattern I've seen is [PATCH shepherd] … <jackhill>muradm: ah, I see you had already sent it, I'm sure that's fine <muradm>jackhill: sorry waited for 5 and submitted... ) but i will note [PATCH shepherd] <jackhill>muradm: I should probalby have just included that bit in my original reply. I think the reason to prefer taht is that `git am` will do the right thing when creating a commit from that patch, but it's not too much work for the person applying the patch to get the commit message right. <jackhill>unfortunatley, I'm not in a position to review the patch, so you'll have to wait on someone else :) <muradm>jackhill: actually intention of the patch is more to raise an issue, shepherd last commit is like oct-2020 :) so i'm not sure if this will make to git at all :) <jackhill>hopefully :) there have been others thinking about using sheperd as a user recently at least <muradm>i use it like at least 6 months or so with few modifications, where the above patch is most critical show stopper, other than that it is started from .bashrc and works fine with xorg and now with sway/wayland <easbarbosa>hey, I need glibc@2.33, and so far its newest version is 2.31 <easbarbosa>is there a way to --force a build to that version? <easbarbosa>'version `GLIBC_2.33' not found (required by /home/easbarbosa/Builds/expense_tracker/.gems/gem <muradm>easbarbosa: guix package --list-available=glibc shows no 2.33 version <muradm>you could inherit 2.31 and try building 2.33 <lfam>You could try using one of the package transformation tools for this. Exactly which one depends on what you are trying to do <lfam>It's surprising a Ruby program would a strict dependency on a particular version of such a foundational component <lfam>Like, where could this program actually run? <lfam>easbarbosa: Did you build this expense_tracker on a different system? Maybe one that is using glibc 2.33? <lfam>Well, you could 1) try packaging glibc 2.33 2) try the core-updates branch (might not work yet, not sure of its status), or 3) try using an older version of your Ruby program from between Feb 2020 and August 2020 <lfam>(That's the range of dates from glibc 2.31 to 2.32) <easbarbosa>lfam: yep, the book I am trying these examples are from 2017 so it might work, thanks! <lfam>Now you are a distro maintainer :) <lfam>It's like putting together a puzzle <lfam>I feel like it's usually possible to handle this stuff automatically but it gets tricky when you cross the boundaries between different languages, like from Ruby to C <lfam>Their tooling that manages version compatibility is not compatible and someone would have write some code to translate between them <lfam>That is going to be a huge improvement <easbarbosa>lfam: I also found that, as you mentioned early, that gem was using arch glibc version(2.33), what means that direnv was using arch gcc too <easbarbosa>I had to list both glibc and gcc-toolchain to it wont use foreign toolings :) <PotentialUser-10>Can I get Xen up and running for Guix myself? I can learn the necessary languages, I only need a direction where to go <lfam>PotentialUser-10: Do you know how Xen is setup on other operating systems besides Guix? <lfam>If you could describe that, even at a high level, you might get some advice about how to make it happen for Guix <PotentialUser-10>It actually does not need to be configured, after installation, the Xen hypervisor appears in the kernel boot options, after which the host machine becomes dom0, from where Xen takes the drivers to work with everything else <iskarian>is there a Guix policy on when pypi packages don't include tests? should the source be changed to upstream so we can test in all cases, or is there a "relevance" bar to pass first? <schmillin>does anyone run guix with a nonfree kernel for wifi and such? <cehteh>you likely need that when your hardware isnt otherwise supported <jahor>Hi! Is it possible to specify channels in config.scm? <jahor>Cool. Thanks for the information :) ***V__ is now known as V
<tissevert>what's the shortest way to override local domain name resolution ? <tissevert>I want to make sure what data a form is posting <leoprikler>abrenon: (guix licenses) has cecill, cecill-b and cecill-c <leoprikler>iskarian: when there's no tests, there are no tests, you don't need to go out of your way to provide them when upstream doesn't <abrenon>is there an equivalent to guix/git.scm's update-cached-checkout to retrieve and expand a tar.gz file ? <abrenon>url-fetch/tarbomb looks like a good candidate, am I correct ? <abrenon>thanks for this more complete description, I was finding the comments in the implementation a bit unhelpful in term of return type <abrenon>so when it says a "fixed-output derivation", I suppose it's not the same type as the location returned by update-cached-checkout ? <abrenon>hmm plus I have no idea what the hash is supposed to be at that point so I guess this is not the right tool for my needs <civodul>abrenon: right, see the "Derivations" section of the manual for more <civodul>update-cached-checkout updates Git checkouts in ~/.cache/guix "here and now" <abrenon>but it's the one I'm after, since it's the one used now in the opam importer <abrenon>and the repo I'm trying to import from is distributed as a .tar.gz <civodul>then maybe you could use http-fetch from (guix http-client)? <abrenon>currently I'm trying to fool the importer by hosting a git repos containing the archive I extracted locally <jlicht>civodul: amazing blog post on the stat storm, very elucidating! <abrenon>I don't understand how cloning works in update-cached-checkout, it seems to call a function clone* which in turn calls clone but I can't find where it's defined <vivien>Hello, I would like to set up emacs to send emails, and I need a "sendmail" program. <vivien>There is a "sendmail" package in guix, but it does not provide a "sendmail" command. <vivien>I also tried "nullmailer", but still no sendmail command. <vivien>What package should I install to have a "sendmail" command? <yoctocell>vivien: why do you need a "sendemail" program, won't msmtp work fine? <sneek>yoctocell, you have 1 message! <sneek>yoctocell, apteryx_ says: how does the prefix trick work for define-maybe? It seems I'd still be bound to a single (define-maybe something (prefix myprefix-)), otherwise there'd be name clashes at the top level for 'something'. <apapsch>vivien: sendmail is broken, try opensmtpd <vivien>yoctocell, it looks like msmtp would do what I want. <yoctocell>apteryx_: not sure what you mean exactly, but you can define multiple define-maybes for the same `something'. (define-maybe integer (prefix foo-)) (define-maybe integer (prefix bar-)) <vivien>apapsch, thank you for the suggestion, I think this is a server, correct? <abrenon>thanks ! I can't understand how it enters guix/git.scm's scope though <bricewge>Is there a package receipe for vagrant lying around ? <apapsch>have yet to test mail functionality myself though :-) <yoctocell>abrenon: guix/git.scm imports the (git) module (git.scm file in guile-git) which seems to expose a bunch of other modules <dootano[m]>I have the GuixSystem on my computer for several months. And I do not know much about GNU/Linux. <dootano[m]>When I connect to a PPPOE modem via Wi-Fi, the internet is disconnected after about 10-15 minutes (Wi-Fi is connected) <dootano[m]>After the internet is disconnected from the computer, I will not be able to connect to the internet with any other device! And after disconnecting the Wi-Fi connection in GuixSystem, everything goes back to normal and the problem disappears! <abrenon>I cannot for the life of me figure out how the guile implementation manages to be more restrictive than the actual git CLI <dootano[m]>I really have more expectations from the Guix community!! <abrenon>but seriously, it's complaining about "text/plain" not being a proper content-type for the first git, after I fixed this by taking the content-type from a repos that I checked to work <tissevert>dootano[m]: do you mean you have no trouble with other distros ? <dootano[m]>tissevert: I did not have any problems in Debian before <tissevert>I don't know much about guix myself but I don't see how a basic ethernet connection could not work <tissevert>maybe you have enabled services that do weird thing with the network like NetworkManager (for what it's worth, I have it running here and it works) <tissevert>I'd be curious to know if the network works in a live USB system of any other distribution <abrenon>I really have to understand where all these imports come from <tissevert>have you had that issue since the begining with guix ? <apapsch>dootano[m]: the next time it happens, are there errors in "sudo dmesg" output? also check out log files in /var/log <abrenon>understanding guile error messages is really hard when you're not used to it. any tip ? (I'm trying to replicate a call I've found in guix' code, have guile's doc open and the error message I receive isn't even consistent with what I read in the doc) <wingo>civodul: for guile, python, etc -- i think it can make sense at build time to make the equivalent of ld.so.cache, as a symlink tree. wdyt <wingo>then your GUILE_LOAD_PATH isn't a whole load of things, it's just the one symlink tree <abrenon>how do you people manage this without types ? I'm more or less getting a vague understanding of what is going on in the code I read, but not being able to look things up by type is huge pain <apapsch>abrenon: it seems to me guile community really wants you to rack your brain. the cryptic error messages, few breadcrumbs on the web (if any), and silence on irc imploring you to *think again*, it all conveys: a good hacker bears the fruits of hard wrought thought work ;-) <abrenon>I do not believe in such things, and it certainly isn't what I'm trying to convey <abrenon>I just wish I could become more independent <apapsch>abrenon: (ice-9 match) lets you inspect values, including record types <civodul>wingo: hi! yes, profiles are essentially that symlink tree, so you can end up with a single GUILE_LOAD_PATH entry <abrenon>because so far all the meta-questions I've had have been: 1) asked for a concrete instance 2) this concrete instance answered, leaving me with absolutely no knowledge of how to fix the next occurrence of a similar problem <abrenon>http-fetch returns a cryptic "port" value which doesn't download in itself but doesn't seem to be even used by the few occurrences of (http-fetch ) I've found here and there <civodul>wingo: so it's actually not that bad, though one typically has at least two entries in GUILE_LOAD_PATH (~/.guix-profile + /run/current-system/profile) <abrenon>url-fetch expects a hash-method but I can't find any actual application of the function so I'm clueless as to what acceptable value can be passed to it <abrenon>I tried "foo", but that doesn't work : ) <abrenon>yeah, without reading that page it's the impression the proper choice of naming had left me <abrenon>but it doesn't explain why the example I've found magically gets the actual download to happen without seeming to do anything more than opening said port ^^ <tissevert>for instance this page claims that "There are a number of different port types" and I have no idea what that means, but worse, I don't know how to even find out by myself <tissevert>how can I inspect the type of a given port ? <apapsch>tissevert: there are definitely types: (string-append 23 42) will throw "Wrong type (expecting string)" <vivien>In guile, a port is just a bunch of callbacks <vldn>hey :) how to set cores and no-grafts options for the cuirass buildservice? <vldn>my own substituteserver seems to build every package with one core <civodul>vldn: hi! you don't need to set no-grafts for Cuirass, it does that automatically (on the grounds that grafts are "cheap") <civodul>as for the number of cores, i believe it just takes that configured for guix-daemon, which defaults to --max-jobs=1 and --cores=N, where N is the number of cores on the machine <Rampoina>hello everyone, does anyone know if guix has an appstream database and where I can find it? <Noisytoot>Rampoina, I don't think it does. but "guix search" and "guix show" output is in the GNU recutils format, so it should be possible to convert it <Noisytoot>Guix packages are Guile objects, so you could also write a program in Guile to convert it into appstream <Rampoina>would that include the same information? especially application screenshots, logos <Noisytoot>Guix packages don't include screenshots or logos <Rampoina>I don't use guix myself, I was looking for information to use guix in a software manager (athenaeum, a game software manager) <Rampoina>the applications themselves provide the appstream files, there's a converter but it needs a backend for guix <vldn>mhh how to get the auth keys from my cuirass server to use it as a substitute server from my laptop? <Rampoina>if you don't have that already then I guess I'll wait and open an issue somewhere <Rampoina>huh no, there's no hardcoded paths anywhere, that's not how it works <Rampoina>the final .xml is generated from repositories <Rampoina>the upstream xmls can be read from wherever <Rampoina>a backend can be written for that to gather the information wherever it is <Rampoina>basically it needs to collect the upstream xml files and merge them together with the correct name ids and so on <roptat>Rampoina, it sounds interesting, but if the xml is provided by upstream, I guess we need to build the package before we can provide appstream support? That sounds backwards <Rampoina>I'm not sure I understand what you're saying <Rampoina>the final file contains metadata for all of the packages <roptat>yeah, but the question is how do you create that file? <Rampoina>it's meant for software managers to be able to find package metadata <Rampoina>mostly you just copy it from the upstream project (like the .desktop file) <Rampoina>I'm not sure how that converter works though, I just a basic understanding (just using the generated files) <roptat>mh... do you need the appstream file only for installed packages, or for the whole set of packages of the distribution? <roptat>then, how do you collect these files when you haven't built the packages yet? :) <Rampoina>I don't know, clone the repo, copy the file? <Rampoina>> I'm not sure how that converter works though <roptat>mh... is the file always at the same location in a repo? <Rampoina>yes it's a standard thing like the .desktop files <roptat>mh, it doesn't seem so, like gnome boxes has only a .in file, so you need to generate the actual appstream <roptat>I would expect the files to be *installed* in a standard location, but you need to build the package to get them <jahor>Good evening/morning/day! Is it possible to specify local file as a source of the package (like /home/user/src/foo)? And also is it possible to run shutdown and reboot without sudo? Thanks <roptat>jahor, you can use (source (local-file "/home/user/src/foo" #:recursive? #t)) <roptat>and you can probably specify shutdown and reboot as setuid binaries in your config, but sounds scary <Rampoina>roptat: you would need to ask someone that knows more for the specifics <cehteh>can you add sudo configs to the config.scm? <roptat>it sounds interesting, but if we really need to build the packages, then we at least will have some delay, and missing packages <Rampoina>roptat: "from distribution packages." I guess it maybe uses built packages <roptat>for debian it's easy because they distribute binaries directly, but guix is a source distribution <roptat>(with substitutes, that makes it look like a binary distribution, granted) <roptat>for installed packages only, that's very easy to do, because we already have the built packages <roptat>"meson -Ddownload-js=true .." scary :D <Rampoina>roptat: well, I'll leave the specifics to you people :P I guess in the worst case you would "only" need to build all of the packages :P (and maybe you don't need to do the whole build) <roptat>we might implement all that in guile :p ***Kimapr1 is now known as Kimapr
<JorgeTern[m]>guix pull: error: You found a bug: the program '/gnu/store/djwm8n5kyw2ynns4mc866wbn44xwhkky-compute-guix-derivation' <JorgeTern[m]>failed to compute the derivation for Guix (version: "e1ddcf75bf9e68a04698733063dd35a3d5b997ee"; system: "x86_64-linux"; <lupo>Hello Everyone, I'm trying to use Guix in antix linux (debian buster without systemd). Everything installed ok, I can install new packages but can't update. "guix pull" fails at 7% with error "guix pull: error: Git error: SSL error: received early EOF". 10 times I tried, always fails at 7%. Any idea what it may be? ***JorgeTern[m] is now known as JorgeTern[m]1
<podiki[m]>hmm that sounds like a network problem (connection timing out or cutting off?) <lupo>I've tried 10 different times within 3 days with connection problems and with a perfect connection and always fails at 7% <lupo>it's quite a mistery for me, I don't know where to find any log or reference. The verbose switch doesn't do a thing. <podiki[m]>hrm. maybe an SSL certificate thing? I have guix on foreign distro also but didn't have to do anything (installed using the script). anyone else seen this before? <roptat>I think the pull is on the user side, so hopefully, you can set this environment variable before guix pull <lupo>I've tried that, "git config --global http.postBuffer 100000000", same result. <lupo>guix package: warning: Your Guix installation is 18843 days old. <lupo>guix package: warning: Consider running 'guix pull' followed by <lupo>'guix package -u' to get up-to-date packages and security updates. <lupo>51 years? maybe there's something wrong with how it was installed. I did it with the script. <roptat>that would be 1970, unexpected, but believable ^^ <iskarian>leoprikler, thanks; just to confirm, I mean when the pypi dist doesn't have test, but the actual source on github has tests <leoprikler>yeah I don't think it's a hard requirement for you to go the extra mile <iskarian>it's frustrating that so many pypi dists strip tests, and some strip only part of the tests so they're broken <lupo>is there a script or a guide to uninstall? I want to try with a fresh install. <bdju>anyone else run into their system config not applying anymore after an update? I swear I didn't change it in between pulls and suddenly I got an error yesterday <bdju>one sec, I'll supply more info ***emacsoma1 is now known as emacsomancer
<bdju> http://ix.io/3uUs here I got this backtrace. it mentions some stuff about swaylock which is near the bottom of my config. I suspect it's not really the problem. <roptat>the way setuid binaries are handled has changed recently, maybe related? <bdju>ah, in that case the backtrace error probably is accurate then... <bdju>any idea what I have to change? <roptat>the list should look like (setuid-programs (list (setuid-program (program (file-append #$swaylock "/bin/swaylock"))))) <roptat>probably replace that with an append, to add %setuid-programs too :) <bdju>what am I replacing with append? I don't actually know scheme... sorry <bdju>list? I know I had it as cons* before <bdju>here was my first (wrong) guess: (setuid-programs (append (setuid-program (program (file-append #$swaylock "/bin/swaylock"))) %setuid-programs)) <bdju>(plus one last paren at the end to close the operating system definition) <roptat>append takes lists as arguments, so (setuid-programs (append (list (setuid-program (program (file-append #$swaylock "/bin/swaylock")))) %setuid-programs)) <roptat>(I haven't checked the parenthesis, append should have two arguments: the list and %setuid-programs) <roptat>this is equivalent to (setuid-programs (cons* (setuid-program (program (file-append #$swaylock "/bin/swaylock"))) %setuid-programs)) <bdju>I keep getting a similar error. `error: setuid-program: unbound variable` and `hint: Did you forget `(use-modules (gnu system setuid))'?` <bdju>not sure if I really need to add a use-module or if I've done something else wrong <bdju> http://ix.io/3uUz here's my config at the moment. I tried adding a use-module and it didn't work so I removed it again. I don't know what to do here <bdju>it'd be nice if there were a page in the style of the archwiki for sway that just said what was needed for this. surely at least a few other people need this same thing and already had to fix it after this update <roptat>bdjjust like the hint says, add that use-module <roptat>we could have something like this in the cookbook <roptat>when you added the use-modules, did you get a different error? <liberalartist>Does anyone have a recomendation for a USB WiFi dongle that works with Guix / Linux-libre? I only need 2.4GHz 802.11b/g/n. <bdju>roptat: can't be sure I added the use-module correctly, but the error had become `error: ungexp: unbound variable` and `hint: Did you forget a `use-modules' form?` <roptat>you should get one that is recommended by the fsf (the "respect your freedom" certification), you'd be sure it works with linux-libre :) <bdju>I'd added a line like (gnu system setuid) below my (gnu services virtualizaton) line <roptat>bdju, oh, I think it's #$swaylock, should just be swaylock maybe? <bdju>roptat: I think I'm finally getting somewhere! I fixed that and then added the use-module line back. thank you <bdju>no error and it's seeming to update stuff normally so far <roptat>ok, then it's a mistake in the manual <roptat>it had that #$ too, which is ungexp, and that can only happen inside a gexp, not outside ***jackhill is now known as KM4MBG
<bdju>updates finished successfully. yay <podiki[m]>the setuid change should have been in the pull news then? or is that not in the main branch? <drakonis>error: native-search-paths: unbound variable <drakonis>oh right, this has to do with the changes to package definitions introduced a little while back <drakonis>i should do pulls more often or just move into guix again <apteryx>drakonis: sorry, perhaps it was my bogus commit 77e2df870b? <apteryx>try guix pull again, it should be resolved now <Noisytoot>Is it possible to create an environment with (or install) a gcc-toolchain that compiles for a certain target (not one that runs on a certain target)? <podiki[m]>I should followup on that (opengl on foreign distro) issue later to pull together the separate threads. In short, should be straightforward to at least build with libglvnd, and then we can go from there irfus reported that being the main step needed on foreign distros to get things ready <roptat>Noisytoot, not sure about a cross gcc toolchain, but you can at least get a cross-gcc from (gnu packages cross-base) <Noisytoot>roptat, How can I get an environment with it/install it? <roptat>you'd need something like (define xbinutils (cross-binutils target)) (define xlibc (cross-libc target)) (define xgcc (cross-gcc target #:binutils xbinutils #:libc xlibc)) and then have xgcc, xbinutils and xlibc in the environment <roptat>(you'd need to import (gnu packages cross-base) <roptat>(I read standard-cross-packages from (guix build-system gnu) for that info) <Noisytoot>although it seems to work for riscv64-unknown-linux-gnu