<donotshake[m]>Newb question: Where do I find what exactly the %base-packages refers to? <lfam>I'm struggling to figure out what's wrong <lfam>If I could figure out *where* it's failing, I might be able to continue <lfam>This pattern of (string->uri (git-reference-url (origin-uri origin))) is used elsewhere, so I think that's okay. It makes sense to me anyways <lfam>And then (uri-scheme uri) should also work, assuming it got a Guile URI <lfam>But, IIUC, the backtrace doesn't say which part failed <xelxebar>$ guix package --with-commit=nyxt=2-pre-release-4 -n -i nyxt <xelxebar>The following package would be upgraded: <xelxebar>Something is chopping off that version string O.o <lfam>Looks like a bug xelxebar! <lfam>Wow, there are really a lot of emails coming from the patch tracker without a "from" address <lfam>It says "From: guix-patches--- via <guix-patches@gnu.org>" <apteryx>bavier[m]: yes, we should fix it! (the incomplete texinfo support) <dissoc>is there a way to manually run the certbot service command? <Bumblehorse>I can't seem to get port-columns and lines to wrok porperly. For example I open a file for input and output that only contains the string "cats". I then evaluate (use-modules (ice-9 textual-ports)) and use (set-port-column! *PORT* 2) but when I use (get-char) on the port it returns the first character #\c. <Bumblehorse>port-column returns the same value that I set with (set-port-column!) <lfam>dissoc: Do you mean, to use certbot manually? <dissoc>lfam: i just wanted to test the certbot manually instead of waiting for mcron to run it. i ended up just finding the job file via ps and manually running it but it failed for some reason <sundbry>Hello fellas, how can I export guix user profiles from one machine to the other? It this currently supported? <sundbry>I have a new computer and want to copy all my packages over <ryanprior>sundbry: yeah I was going to say you're in luck, that feature should land soon! <sundbry>the .guix-profile/manifest format looks like its not importable yah <ryanprior>Right now there's no slick automated way to do it yet, but the basic flow is: on your existing machine you run `guix package --list-installed` and then on your new machine you run `guix package --install ...` <ryanprior>sundbry: No, that manifest file is not the same as manifests that you can install. It's an overloaded term in Guix unfortunately. <sundbry>ahh ok looks like I will have to run that and clean up the output from guix package --list-installed to script it <sundbry>The document says that grafted packages must have the same length name and version. What if I have a package that has only a --configure-flags change? no change in version. I can't add a "-1" to the version string? <sundbry>(I want to graft jemalloc for some flags to support java-rocksdb and it causes a rebuild of the world practically) <ryanprior>You can do that, but it makes it ungraftable. The reason is that grafts use binary-level patching to avoid rebuilds, which means that pre-calculated offsets can't change. <sundbry>ok and if I keep the same version I get a warning about ambiguous package versions <ryanprior>Add "-1" to the end of the version and now we've got to recalculate the offsets to account for 2 more bytes :) <sundbry>its kind of a catch-22 in this scenario I think <sundbry>The configure flags in question, without them will crash any process that calls dlopen(libjemalloc.so) <ryanprior>Possibly you kept around the previous package definition and added a new one, which would lead to an ambiguous version warning. <ryanprior>You could resolve that by removing the previous package definition you aren't using anymore. <sundbry>But I need to keep it around so I can instruct guix to graft it? <sundbry>I am trying to avoid triggering a rebuild of the world <sundbry>Rigtht now it's named 'jemalloc-5.2.1' I could call it 'jemalloc-5.21a' 'jemalloc-5.2-1' noone of these are ideal of course? What would be advisable for something that could land upstream? Since I want to contribute my java-rocksdb package <sundbry>I suppose I will just do 5.2-1 that seems to make the closest compromise <ryanprior>I think that would work. I can't say whether it would be accepted upstream, I'm not a committer. <ryanprior>java-rocksdb would be a great package to have though <sundbry>or a separate package entirely perhaps. jemalloc-dynamic-safe-5.2.1 then when we actually upgrade jemalloc we can get rid of the duplicate package <sundbry>when we are ready to rebuild everything <sundbry>yes it took a long time to get this sucker working... lol <apteryx>is there only 2 maximum remote build slots when offloading? <ryanprior>That sounds doable ^^ we have a core-upgrades branch that we use to stage patches that require rebuilding the world, and we batch them to do it all at once. <apteryx>I think if you provide an absolute profile to it it'll copy the closure of the profile over <apteryx>guix copy $(realpath ~/.guix-profile) --to=your-remote <ryanprior>Ooh cool, I did not know guix copy worked with profiles. <sundbry>the man page doesn't say anything about actually installing the packages tho <apteryx>right, it'll just copy them to the remote store <apteryx>then you need the actual list of packages *and* run the install command from the very same Guix revision <ryanprior>Ah like "okay all copied, if you wanna actually use them you're on your own" <apteryx>*and* without grafts? it's tricky to get it right <ryanprior>But then when you go to install the packages it should have less work to do than if you had to download & build everything yourself <apteryx>it's probably much easier to setup your machine A as a substitute server <apteryx>configure your machine B to use it as such <apteryx>then from the same Guix install a manifest of your packages or a list of them <apteryx>and it'll fetch them from machine A automatically <raingloom>heyyyy, i have an openimageio patch that i needed to build Blender but it's uuuuh. well, my machine can't handle building it. Blender that is. <raingloom>could someone build and test it? (and then maybe send me the package coz i wanna do some art :D ) <leoprikler>Is there a way to get in contact with someone who moderates/spam filters the MLs? <jmarciano>how do I solve: /gnu/store/6rn4l3h0p9x0m615pp1ynlv9v0743kl3-guix-1.2.0/libexec/guix/guile: symbol lookup error: /gnu/store/gpv8zbls7vqbimaaq4dvl69an5ayy38b-libgit2-1.0.1/lib/libgit2.so: undefined symbol: SSL_CTX_set_options, version OPENSSL_1_1_0 -- after installation of guix binary? <jmarciano>I think it requires & on the end of the line: ~root/.config/guix/current/bin/guix-daemon --build-users-group=guixbuild <jmarciano>Another issue is that ./guix-install.sh supports (so I guess) only systemd and openrc, but I think I have openrc, yet it failed to find the files and shows errors that it cannot start from /etc/init.d/guix-daemon if I remember well <jmarciano>but script could manually start daemon in that case <jmarciano>well, I have tried many times with installation of guix binary on Hyperbole, it is basically not useful due to: undefined symbol: SSL_CTX_set_options, version OPENSSL_1_1_0 -- <cbaines>jmarciano, did you download a substitute for libgit2, or did you build it? <cbaines>PotentialUser-48, what's the problem? I don't know about /gnu/store/mjmn2yhsy6g20csciyl2w3blrssq35yj-openimageio-2.0.13.drv so I can't try building it myself. How was this derivation generated? <jmarciano>Yes, I expect that type of output, so that libgit is self-contained <cbaines>jmarciano, OK, what options are you running the guix-daemon with? <jonsger>how do people split PDFs per page on guix? <jmarciano>cbaines: ~root/.config/guix/current/bin/guix-daemon --build-users-group=guixbuild & <cbaines>jmarciano, OK, and what user is it running as? <cbaines>jmarciano, what do you get if you run: md5sum /gnu/store/gpv8zbls7vqbimaaq4dvl69an5ayy38b-libgit2-1.0.1/lib/libgit2.so.1.0.1 <cbaines>right, that's the same as me... I wonder if there's something odd with your environment then <PurpleSym>Is there any trick to getting teeworlds running? I fixed the audio loading issues, but it’s just showing a black screen now. Maybe an OpenGL/SDL2 issue? <cbaines>jmarciano, do you have LD_LIBRARY_PATH set? <jmarciano>actually now when I do ldd, I do not see those problems, all is pointing to /gnu <jmarciano>OK now when unsetting that variable I cannot move forward <cbaines>jmarciano, hmm, I'd perhaps try running guix pull as root, and then restarting the guix-daemon <jmarciano>if I install package as root, is it available to users automatically? Or I need to install it as user? <cbaines>PurpleSym, I can get to the menu at least, although I don't hear any audio. Do you not see anything at all? <cbaines>jmarciano, typically, each user, including root has separate profiles, so they only see packages installed in their own profiles <cbaines>jmarciano, of course, everything is in the same store, so this is just about what environment variables are set <PurpleSym>cbaines: Nope, the window disappears after the audio starts (I fixed the audio). I have a AMD GPU, maybe there’s drivers missing? <cbaines>PurpleSym, I'm testing with Intel integrated graphics. If you run it from a terminal, do you see any relevant output in the terminal? <PurpleSym>But if it works on Intel this must be a GPU issue. <PurpleSym>extremetuxracer works, so it could be an SDL2 issue. <leoprikler>I don't think it's much related to configuration: `guix build teeworlds`/bin/teeworlds works for me (NVidia card) <leoprikler>Also Guix system, but that should hardly make a change when running directly from the build directory <PurpleSym>Well, supertuxkart is working too (in windowed mode). Strange :/ <PurpleSym>Ah yeah, teeworlds now works in windowed mode. <nly>roptat: how to do .onion resolution with unbound? ***janneke_ is now known as janneke
<cbaines>PotentialUser-48, did you see the questions I asked? <PotentialUser-48>cbaines, I do not understand what you mean (I do not know much about Guix) <nckx>jonsger: qpdf. Or pdfarranger if you crave the GUI. <cbaines>PotentialUser-48, I don't understand what you're asking? <cbaines>putting that aside, it does look like openimageio fails to build, since an update to pybind11 <aecepoglu[m]>A plugin (written in Rust) for a text editor. Should it go into 'text-editors.scm' or elsewhere (like how emacs has a file of its own) <cbaines>aecepoglu[m], probably in text-editors.scm, assuming that's where the text editor the plugin is for is defined <rekado>I successfully reconfigured my i686 on staging (haven’t rebooted yet, though) <rekado>tried upgrading my packages but znc fails to build <rekado>I also need to build gst-plugins-good from source <jmarciano>cbaines: alright, after guix pull as root, it started working for user <jmarciano>great to have system that works on top of other system independently <cbaines>PotentialUser-48, a change was introduced 4 days ago which broke openimageio <iyzsong>PotentialUser-48: not you, but the latest guix have that problem (unable to build blender because openimageio failed to build). <rndd>hi everyone! how i can edit /etc/hosts ? <mdevos>rndd: search for "/etc/hosts" in the manual, section "‘operating-system’ Reference" <mdevos>rndss: for creating a file-like object, "plain-file" will probably suit your purposes <mdevos>There's an example in the manual, see %facebook-host-aliases <adfeno>Hi there, how to get lua package's Equivalent of “-dev” files mostly seem in distros such as Trisquel ? I want to get the files needed to build a local project that is using pkg-config. <mroh>nice pkg name `guix show stuff` ;) <apteryx>why is the --max-jobs implicitly set to 2 for remotes when offloading? <adfeno>under Lua's derivation path + /lib/pkgconfig <raghavgururajan>apteryx: I think Makefile and/or make sometimes pass 'jobs' arguments. <rndd> how many users have "stuff" in their packages's name 0_o? <nixo_>hi! guix pull seems to be broken (at least, it is for me) <nixo_>I get: (exception unbound-variable (value #f) (value "Unbound variable: ~S") (value (rust-dirs-2.0)) (value #f)) <nixo_>(-builder for `/gnu/store/zlij12n2xlfqvvf22c0rq0cpn6z5fmc0-guix-package-cache.drv' failed to produce output path `/gnu/store/c5jhqqpvs8bfn3a9wnl5nq52pac71zn8-guix-package-cache') <raghavgururajan>leoprikler: Is this correct for rlottie-for-telegram-desktop? (git-version "2.4.4" revision commit) <ngz>cbaines: Hi. I'm currently building the latest release of openimageio. Then I'll try to build blender and blender-2.79 in order to check if this is sufficient to fix them. <ngz>It might take a while, tho. <leoprikler>rlottie upstream version seems to be 0.1, but there is a 0.2 as well <leoprikler>can you try checking if telegram builds against 0.2? <cbaines>ngz, sounds good, I had a little look at building a newer openimageio, but just bumping the version to the latest one failed to build, I think due to new dependencies that were missing <leoprikler>well, duh, but it might happen, that there's some actual compatibility <ngz>I'm currently at 80% into build <ngz>My CPU is going to burn when building blender =/ *raghavgururajan just realised that he has to change commit and hash for telegram desktop's git-fetch inputs, all over again <leoprikler>tgvoip seems to be 2.4.4 + patches according to configure.ac <ngz>cbaines: openimageio built. Now heading to blender@2.79 <raghavgururajan>leoprikler: So (revision "88") and (git-version "2.4.4" revision commit)? <leoprikler>if there's been 88 commits since and that's what you're trying to indicate, go for it <htgoebel1>Any tips how to debug/fix this kind of issues? <mothacehe>nixo_: hey, it should be fixed, you can run "guix pull" again. <nixo_>mothacehe: wonderful it did work, thanks! <ryanprior>raghavgururajan: I think you can build it by defining a trivial non-hidden package that depends on it. Probably possible to do it via the guix repl as well but I'm not so handy with that. <htgoebel1>raghavgururajan: Something like guix build -e '(@ (gnu packages maths) the-hidden-package)' could do the trick. see <ngz>cbaines: blender should now be fixed. Testing blender-2.79b <civodul>we have an "upgrade-services" system test, but it didn't catch this because it works at the later below <civodul>did you use (guix status) to get build completion info? <civodul>it's going to make a big difference to have this info available <mothacehe> civodul: thanks for the fix, I'll reconfigure berlin :) <mothacehe>glad you like it :) No, it's built upon the remote building mechanism. A message is sent by every worker just before starting the build and on build completion. The database is updated accordingly, and this page is just a nice representation of the database status. <civodul>looking at /workers, i see that -127 was marked as building julia-adapt, but it was in fact (attempting to) build/substitute julia <mothacehe>there's no need for (guix status) as each build triggers a new build-derivation RPC <civodul>like -123 currently has a progress bar showing it's almost done with 'guix' <mothacehe>no it's based on the duration of the last successful build with the same job name <mothacehe>currently a package build can trigger the build of some of it's input, making this duration a bit inconsistent <mothacehe>but in the future I'll implement "BuildSteps" as Hydra does <mothacehe>spamming F5 to check progress bar completion is kinda addictive <civodul>will "guix publish" still have build logs to provide? <civodul>when you run "guix build --log-file icecat --no-grafts", say ;-) <mothacehe>oh I guess it wont, I have to find a way to fix it. <mothacehe>on the overdrive front, I have played with wireguard, it seems like a nice workaround. However, it needs an open UDP port on berlin, I sent a mail to sysadmins about it. <mothacehe>jonsger: thanks, it's largely inspired by the opensuse monitor you sent me <mothacehe>even though they have nicer colors than I do for now <jonsger>yeah, the similarity is not overlook-able :) <jonsger>is it a feature that the evaluation numbers aren't counted +1, they look now like the number of commits or so <mothacehe>the commits that did not trigger new evaluations are incrementing the evaluation primary key <mothacehe>that's a difference between sqlite and postgresql <mothacehe>they have different behaviours on rollback more specifically <civodul>mothacehe: re VPN, we'll see; i suspect it's going to take time before the port is opened <civodul>is there an SSH tunnel kind of hack we could come up with in the meantime? <mothacehe>civodul: the nice point is that once this one is opened we won't need any more port opening <lfam>I'm going to merging staging into master this weekend <lfam>If it goes well on my computer I'll push <mothacehe>yes if rekado can give me an estimation of how long it could take, I'll think about an SSH based work-around <civodul>lfam: i've reconfigured my system on staging, too (but haven't rebooted yet) <lfam>You should try it and let me know how it goes :) <civodul>mothacehe: i guess it depends on the availability/workload of the admin team <civodul>lfam: yup, i'll reboot soonish and we'll see :-) <civodul>i have a childhurd in my config, which makes it quite demanding <apteryx>lfam: I'll try reconfiguring my machine with staging right now <apteryx>that's new: guix system: warning: mapped-device '/dev/sda2' may not be mounted by the bootloader. <apteryx>it comes from 'system: Allow separated /boot and encrypted root.' <apteryx>didn't even know this was possible; cool! <apteryx>OK, apparently the warning above is because "Only devices specified by uuid are supported." <emsyr>Hi everyone. For sometime now I cannot reconfigure system cause kwayland.drv fails during check phase. Is there a way to skip the check phase? For "guix install" there is the --without-tests=[] option but there is not such option for "guix system reconfigure". Thanks. <civodul>emsyr: hi! indeed, this option is not available for "guix system" <civodul>did you report the kwayland build failure? <emsyr>civodul not yet. I'm trying to see if there is a workaround. For "guix package -u" it worked but not for "sudo guix system reconfigure". <emsyr>civodul not yet. I'm trying to see if there is a workaround. For "guix package -u" it worked but not for "sudo guix system reconfigure". <civodul>emsyr: kwayland also fails to build locally? <emsyr>civodul yes, the derivation is being built locally and it fails during the check phase. <emsyr>civodul it can be installed only with the --without-tests=[] option so as check phase is skipped. <shcv>how do I install the zfs modules? If I install the zfs package it just gives me the cli tools <shcv>there are some patches on an email thread that make a service definition, but they don't seem to be merged yet <luis-felipe>I'm reporting some bugs to Jami upstream, and they ask me to run "journalctl --since "24h ago" | grep dring". <mdevos1>zhcv: idk the specifics of zfs, but the usual way of using features implemented by unmerged patches is applying them to your local git clone, then using that patched guix (with ./pre-inst-env guix system reconfigure etcetera) <mdevos1>luis-felipe: probably some file under /var/log <luis-felipe>Also, how do I check the version of the ring daemon. Their instructions say: "/usr/lib/ring/dring -h" <civodul>luis-felipe: there's no Jami service in Guix System, so no equivalent to the journalctl command <mdevos1>luis-felipe: maybe start Jami, run gnome-system-monitor and look where the dring location is located? <mdevos1>luis-felipe: I wonder what Jami upstream hopes to find. The daemon is run as the same user as the Jami application, right? <mdevos1>So why would any information end up in a system log file? Perhaps ... <mdevos1>Perhaps some thread encountered an ‘illegal instruction’, in which case some information about dring would end up in the output of "sudo dmesg" <mdevos1>(or sudo dmesg -k, don't remember the exact command) <mdevos1>(Pure speculation, though, and I'm assuming the Linux kernel here.) <luis-felipe>civodul, mdevos1: thanks, I'll just omit that information from the bug report. <luis-felipe>mdevos1: I could get the exact version of the Jami Daemon with your instructions, thanks. <luis-felipe>Being a GNU package, it's sad that Jami works better in Android than in GNU. ***MidAutumnHotaru8 is now known as MidAutumnHotaru
<luis-felipe>leoprikler: That's what they say. "Jami, a GNU package" <shcv>how do you include files in the system disk-image? E.g. .emacs.d and other dotfiles <fnstudio>anyone that can point me to a guix-packaged guile project, that i can use as an example for packaging? it's for a simple project, so i suppose something that makes use of guile-build-system? <luis-felipe>fnstudio: The first package is just a guile library. The other package is a CLI program. <luis-felipe>fnstudio: The "source" fields of these packages point to private repositories, so you should change the values according to your needs. <fnstudio>luis-felipe: this is great! thanks so much! <fnstudio>luis-felipe: actually the pastebin link seems broken <fnstudio>uh, maybe it's because i'm not logged in? <fnstudio>cool, works now - thanks luis-felipe, very helpful <lfam>I'm going to try merging staging later tonight <efraim>Did I see civodul tested the childhurd process? <lfam>I think (?). He said that "it does work" but it wasn't clear to me if he meant that it worked on staging or not <efraim>I just ran the build instructions from (gnu system examples bare-hurd) to make sure that worked, and fixed a couple of packages <lfam>I didn't notice but somebody pushed some commits to staging yesterday