<Aurora_iz_kosmos>dftxbs3e: You can mess with your $PATH, though that's a hacky way to do so. <dftxbs3e>Aurora_iz_kosmos: yeah that's what I am doing.. hah <dftxbs3e>created a symlink to it in a subdir that's called "guile" and added it in front of the PATH <dftxbs3e>gnutls-guile-3.6.7-1.fc29.ppc64le.rpm 296 kB/s | 64 kB 00:00 <dftxbs3e>----------------------------------------------------------------------------------------------------------------------------------- <samplet>I think my old build notes won’t work anymore, since they use Guile 2.0. <Aurora_iz_kosmos>Besides the icecat issue, xfce and MATE are broken for me with the GuixSD iso. (Tried in VM) <apteryx>arg... now it's bison that fails on core-updates <Aurora_iz_kosmos>se-sm-ca: Yeah, on x86. It installed, but attempting to login after the install simply crashed right back to the display manager. <apteryx>ah, it doesn't fail, it "differs", i.e. it is not reproducible. <Aurora_iz_kosmos>Ah, the XFCE notice right at the top of the page. Okay then, it's known. <dftxbs3e>samplet: I'll fork your project, I have ppc64le gitlab runner setup already :) <samplet>dftxbs3e: Nifty! And what a coincidence! :) <gromni>Hmm, managed to get my network running from another tty, can ping and guix pull no problem but can't get past the network section of the guided installer.... I guess I'll have to install from the shell <nixo_>Hi, how do I install wireguard? I see there is out and kernel-patch.. is adding "wireguard" to system packages enough? <samplet>gromni: That might be your best bet. Sorry! (I don’t know much about the installer yet – it’s new.) <nckx>nixo_: No, you have to build your own custom kernel that applies <wireguard:kernel-patch>/wireguard.patch. It's non-trivial. <nckx>Did I dream up a custom kernel blog post, or is it still work in progress? <nixo_>nckx: thanks, are you using it? <Aurora_iz_kosmos>samplet: Any idea what's up with "guix pack"'d graphical programs not displaying characters properly? <samplet>nckx: There was some discussion about how to expose the kernel package interface and whether putting a bunch of “(@@ (...) ...)” forms in a blog post is good idea. <samplet>dftxbs3e: You can build it from the org-mode file. <samplet>dftxbs3e: I think the YAML file describes how. <nckx>nixo_: I have it built into my kernel, but haven't tested it yet. <dftxbs3e>samplet: does it convert a README.org to a build script o.O? <nckx>samplet: Good, it's not a great generic interface (it didn't have to be; it was private :-p) <esper0s>love "literate programming". I cannot understand how especially big organizations have not adopted more. <samplet>Aurora_iz_kosmos: Sorry, no clue. It could be a font thing, but I don’t know. <nckx>samplet: Thanks for confirming I didn't imagine it. <se-sm-ca>Aurora_iz_kosmos: ah I never actually logged in, ran into the same thing. Thanks apteryx ! *samplet goes AFK for a while <nixo_>nckx: I hope it will get mainlined soon then :) Or I'll try getting the custom kernel. On nix it was quite easy to apply kernel patches <nixo_>I'll go to sleep, see you :) <Formbi>why is ungoogled-chromium almost never available as a substitute? <kabo>I created a bug report for the guix pull thing, #35561 <kabo>I reinstalled with gnome instead of xfce btw, seems much better supported, there's network manager and browser installed from the get-go <se-sm-ca>the xfce workaround did succeed, though I had to boot into tinycore to edit /etc/config.scm. emacs was returning command not found. There's probably a better way to fix it than to boot into another iso but just sharing what I did <Aurora_iz_kosmos>samplet: Now however I find myself wondering if there's a way to 'pack' a program without having to set variables on the destination-machine. <nckx>Aurora_iz_kosmos: What exactly was wrong? <nckx>You *want* applications to respect environment variables, or (for example) every single graphical application would depend on every single font (or only support the one it was compiled with, both are bad). <samplet>Aurora_iz_kosmos: You could try to wrap it with a script that sets up the variables the way you want. <samplet>That is, include the wrapper in the pack. <samplet>dftxbs3e: Funny story: GNOME and GDM work just for me in VirtualBox. <samplet>That’s VirtualBox on Debian 9 with essentially the default settings. <PotentialUser-51>file "...glibc-2.28/include/bits/errno.h" contains "#include <linux/errno.h>", so can't compile <Aurora_iz_kosmos>nckx: samplet: The idea essentally is to package the program as a fully self-contained unit. A bit like a Container but not as a container. <nckx>Aurora_iz_kosmos: Ignoring users' preferences? <Aurora_iz_kosmos>nckx: The alternative is going through the trouble of managing all the nodes via saltstack. <sielicki>guix pack is one of the coolest things I've seen in a long time <nckx>#f it, I'm going to bed. <dftxbs3e>"guix pack: error: could not find bootstrap binary 'tar' for system 'powerpc64le-linux'" <nckx>May you be finished compiling libreoffice by the morrow. <vagrantc_>dftxbs3e: someone was working on it a few months ago, not sure how far they got, obviously not integrated yet <dftxbs3e>well can someone push powerpc64le-linux-gnu bootstrap binaries? it seems they build out of the box <dftxbs3e>samplet: hey, can you create guix packages? <samplet>dftxbs3e: Like, am I personally capable of writing a package definition? Yes! <dftxbs3e>it seems that powerpc64le-linux-gnu bootstrap-tools compile out of the box <dftxbs3e>samplet: run this: guix build --target=powerpc64le-linux-gnu bootstrap-tarballs <samplet>dftxbs3e: I’m a little confused. Did your PowerPC machine build the Guix bootstrap tarballs? <dftxbs3e>"guix pack: error: could not find bootstrap binary 'tar' for system 'powerpc64le-linux'" <dftxbs3e>so I tried to build bootstrap-tools on my x86_64 box <samplet>And you were able to cross compile them? <dftxbs3e>and it says: error binary128 floating point type (GCC >= 6.2) is required on powerpc64le <dftxbs3e>which is solved by using a higher gcc version <samplet>dftxbs3e: That binary128 thing sounds familiar from the mailing list. <samplet>Guix installed correctly in the log you linked. What failed was building the binary distribution. (If I’m reading it right.) <dftxbs3e>I don't know if you can tell guix pack to use system tar and xz <samplet>I think what you would want to now is to tell Guix to use your cross-compiled bootstrap tarballs. <dftxbs3e>samplet: yes of course, but they don't build because of missing binary128 floating point? <dftxbs3e>so we would need to build bootstrap binaries with gcc-6.2+ <dftxbs3e>samplet: how you think we can tell it to do that? <samplet>I think you can do that by modifying the package definitions. If you just add a newer GCC to native-inputs, I’m pretty sure it will get used. (Although the bootstrap stuff often works differently.) <samplet>Not sure. You could try “--with-input”, but it might not work for the GCC used when building. <samplet>I can look at the code. Gimme a sec. <dftxbs3e>samplet: I'm seeing this: "(("tar" ,(search-bootstrap-binary "tar" (%current-system)))" <samplet>You should look at “make-bootstrap.scm”. That’s where the action is. <samplet>No dice with the system GCC. Guix is picky about that. <samplet>I believe the way to do it is to cross compile the bootstrap binaries, and then tell Guix to use them for bootstrapping. <samplet>Right, but I think it would be easier to cross compile for the bootstrap binaries, because then you can use Guix. <samplet>Let me check for a second, I haven’t read the bootstrap code very much before. <kabo>would be cool to run guix on a raptor talos <samplet>dftxbs3e: Have you read about running Guix from a Git checkout? You will need that to be able to modify the package definitions. <samplet>However, there’s no harm in giving it a try. <dftxbs3e>samplet: if I understand it right, guix can compile everything from source, no need for binaries from your side? <dftxbs3e>so if I can get an environment where guix work, I can do everything? <samplet>It should work, yup. You will have to make sure Guix finds your locally built bootstrap binaries, but that’s a bridge that can be crossed later. <samplet>dftxbs3e: It looks like you can cross-build GNU hello, but not GNU sed. There might be some issues yet. I’ve never tried! <dftxbs3e>they say they worked on the wip-gcc7 branch <samplet>Actually, “core-updates”. I think “wip-gcc7” was merged into “core-updates”. <samplet>So, yeah, it looks like there’s hope. <dftxbs3e>samplet: can I drop into a shell with core-updates branch? <dftxbs3e>I know you told me about the git checkout <samplet>You will need to clone our repo, checkout “core-updates”, and follow the instructions in the manual to build Guix. <samplet>Then, you can try things like “./pre-inst-env guix build --target=powerpc-blah-blah bootstrap-tarballs” <samplet>(Sorry – already forget the triple!) <leungbk>i linted and submitted a series of patches a couple of weeks ago, but under the emacs-debbugs interface my patch series is highlighted as red. how can i find the reason for this? i linted again just now and didn't spot anything. i've tried looking on the mailing archive but don't see any relevant linting info about the patch. <leungbk>would the hydra.gnu site have any record of these sorts of linting issues? <apteryx>leungbk: I think it just means that nobody replied yet (the red color in debbugs) <samplet>leungbk: I think the 1.0 release has put everyone a little behind on patch review. <leungbk>@samplet i see, thanks for your guys' hard work. <samplet>apteryx: What do the other colours mean? The manual doesn’t mention it. <samplet>dftxbs3e: It looks like we don’t have substitutes for GCC 7 on the “core-updates” branch. That means Guix will build it locally, which takes quite a bit of time. <dftxbs3e>samplet: would you need ppc64le CI builders? <dftxbs3e>or if everything can be cross compiled,, <apteryx>samplet: IIRC green means that the ticket was replied to. <samplet>dftxbs3e: I don’t know. Once things start rolling, you should contact the mailing list. That’s where you’ll find other PowerPC people and the people in charge of the build farms. <dongcarl>Hey all, I installed Guix thru the binary path, but I'm now finding that I need to build from git to test my changes. I've looked at the manual, but I'm wondering how I might install it after a successful `./bootstrap && ./configure --localstatedir=/var && make` <dongcarl>`make install`? Just don't wanna mess up my system <lisbeths`>When I run the installer script it gives me a gpg error and tells me to run a gpg command <lisbeths`>but when I run that commmand it says gpg keyserver receive failed: no data <samplet>dongcarl: You can use the “pre-inst-env” script to run Guix without installing. <samplet>lisbeths`: Did you use the <pool.sks-keyservers.net> key server? <lisbeths`> gpg --keyserver pool.sks-keyservers.net --recv-keys 3CE464558A84FDC69DB40CFB090B11993D9AEBB5 <dongcarl>Are people getting failures for `make check`? Or should I be getting all PASSes? *samplet should run “make check” from time to time. <samplet>lisbeths`: I am able to download the key. Does the error message say anything else? <lisbeths`>maybe the server was giving me a timeout for too many requests <lisbeths`>I hate compiling and I was very impressed by nixos a few years ago and the nix fs but it was too much in the early days for me to play with it <lisbeths`>fail a previous guix installation was found. Refusing to overwrite it <lisbeths`>aht was my bothced attempt to install it when I didn't see the installer script. I want to force overwrite it <samplet>I think you can remove /gnu and /var/guix and /etc/guix to “uninstall” Guix. <lisbeths`>afraik the idea from nix came out around the time that haskell was really popular and it was originally involved with an immutable filesystem <samplet>The idea is that the store manages files like a memory-managed language manages heap memory. <samplet>Functional languages were a big influence for sure. <lisbeths`>yeah afaik from some cryptic video on youtube which used to be the only video on nix, some of the haskell people who were involved in the immutable package manager for haskell called cabal began working on an immmutable filesystem that worked on the same principles <lisbeths`>only making a filesystem is super hard and making a system for organizing files is alot easier <lisbeths`>before the nix operating system therew was and probably still is this project called bedrock linux which used a very similar idea to guix except he claimed that symlinks were a problem and swore by using procmounts instead of symlinks and I tend to believe him <dftxbs3e>samplet: does guile transpile to go then machine code? <samplet>dftxbs3e: Here, “.go” means Guile object code, and not the Go programming language. <lisbeths`>guile is a version of lisp it should compile to machine code <brendyyn>lisbeths`: I think Guile 3.0 will do that. <samplet>Guile compiles to bytecode, and the bytecode gets interpreted. Soon, Guile 3 will have JIT compilation, and then after that, hopefully, native ahead-of-time compilation. <buenouanq>I will be able to 3D print lisp processors at home. <dftxbs3e>wouldnt it be rather easy to write an LLVM frontend for Guile? <buenouanq>And then the computer revolution will finally begin. <brendyyn>andy justt merged his jit code the other day, so looks like interesting stuff is coming ;) <samplet>dftxbs3e: That’s essentially what’s happening, except instead of LLVM it’s GNU Lightning. <lisbeths`>fpgas are near the performance of printed scilicon but you can make just 1 instead of printing millions <lisbeths`>the main most common alternatives to the jvm are: jvm, llvm, common language runtime (microsoft open source version of jvm), and webassembly <dftxbs3e>parsing lisp with an fpga would be insane <lisbeths`>afaik that is how boston dynamics robots work. I think they have a lisp fpga in each joint of their robots <lisbeths`>I am having trouble installing cool-retro-term <lisbeths`>it still warns me that guile failed to install locale even though I installed that package <lisbeths`>and even after I run a guix package -u it still warns me that I need to run it <samplet>I also seem to always get the “guix pull” warning. <lisbeths`>yeah I tried guix pull && guix package -u and it warns that I need to pull and -u <dftxbs3e>so guix embeds all the package definitions inside itself <brendyyn>Make sure you don't have guix in your profile <dftxbs3e>I'm on Guix system itself and I get the same <samplet>lisbeth`: For the locale warning, I think you need to set “GUIX_LOCPATH” in the daemon’s environment. <lisbeths`>lel I have no idea how to do that. I assume that's not the same as my system path <lisbeths`>no worries I understand for a newly 1.0'd software that there are bound to be some rough edges <dftxbs3e>GUIX_LOCPATH probably doesnt have another solution <samplet>lisbeths`: If you are using systemd, it would involve editing the guix-daemon service file. <samplet>In a few minutes, I’ll be able to check how to do it on a non-Guix System machine. <samplet>Okay, it looks like I installed the locales using Guix as root (so they ended up in the root profile). <samplet>Then, I modified /etc/systemd/system/guix-daemon.service. <samplet>There’s a line in the “Service” section that says “Environment=GUIX_LOCPATH=/var/guix/profiles/per-user/root/guix-profile/lib/locale”. <samplet>Oh. That line’s been there since 2016. I guess just installing the locales into root’s profile should do the trick, then. <apteryx>lisbeths`: make sure you have glibc-locales rather than glibc-utf8-locales (the later only contains a few locales). <apteryx>in the root profile of your foreign distro, plus the correct GUIX_LOCPATH environment var set for your guix-daemon systemd service <dftxbs3e>samplet: how do I run guix-deamon twice? <dftxbs3e>I have built another version of guix under a guix environment <dftxbs3e>and I want to run the daemon to build bootstrap tools but it doesnt work <samplet>dftxbs3e: There’s no need to run the new daemon. It almost never changes, so the new client code will be compatible with whatever daemon happens to be running. <lisbeths`>export PATH="/root/.guix-profile/bin${PATH:+:}$PATH" <lisbeths>It works I just have to figure out how to permanantly export my path <lisbeths>he he he now I don't have to compile cool-retro-term which is nice <lisbeths>once guix becomes a bit more polished on non guix distributions I think it will be a huge competitor to the arch aur *samplet runs “guix environment --ad-hoc cool-retro-term” and is not disappointed. <lisbeths>what's really nice is when you get the burn in and the horizontal sync and a translucent terminal <lisbeths>combine it witht mux or emacs and it becomes very comfy <lisbeths>samplet: if you right click and selectr preferences you should be able to change the colors <brendyyn>perhaps they can upgrade it to cool-coloured-term <lisbeths>One of my favorites is the futuristic template though I always go with green <samplet>Thanks! I couldn’t guess how to configure it. <brendyyn>I can only run the Arch version. For me the guix version is borked <lisbeths>the sweet sauce is the horizontal sync and the burn-in. <lisbeths>horizontal sync makes your screen flicker randomy and burn-in makes it look liek it takes a few seconds for the screen to update <calher>Congratulations on the 1.0 release! :D <brendyyn>I'm just wondering where I should post a bug report for this. I don't know where the problem lays exactly <brendyyn>samplet: with Guix on Arch, both Godot and cool-retro-term look like this. https://brendan.scot/p/bogot-borked.png. But the Arch versions work fine. I've seen this kinda glitch before with KDE on Arch with my AMD GPU, but I'm not sure where the problem is coming from <lisbeths>brendyyn: did ou follow the arch guide for guix? <samplet>brendyyn: That link does not work for me. I get a 404. <samplet>Just a 404 with a link for “creating a new conversation”. Clicking that link takes me to a page that says my browser is no good. <brendyyn>is it going to my jitsi meet instance for some weird reason <samplet>Also, it says the certificate is only valid for “meet.brendan.scot”. <brendyyn>probably my certs are wrong somehow then <brendyyn>samplet: My firefox and Icecat both tell my my certificate is fine. <dftxbs3e>I've resized it using parted but now the file system needs to be grown <samplet>Nope. Neither Icecat nor EWW nor cURL nor wget. It may be a certificate problem on my machine. <brendyyn>It's just as likely my server is misconfigured.. hmmmmmmm <samplet>It fails on a different machine on my network as well. <brendyyn>We don't have ipv6 here because the government is a bunch of crooks that destroyed the plan to improve it and funneled the money into some rich peoples pockets <brendyyn>ssllabs shows that it may only be valid for ipv4 <samplet>lisbeths: Did you try “sudo guix install glibc-locales && sudo systemctl restart guix-daemon”? <samplet>brendyyn: I think that’s it, but how do I force Icecat to use IPv4? <samplet>cURL is probably easier. I’ll try that. <samplet>brendyyn: “curl -4 …” did the trick. <brendyyn>samplet: thanks for making me aware of this issue <samplet>brendyyn: Godot looks terrible alright. You may want to ask on <help-guix@gnu.org> or even <bug-guix@gnu.org>. This looks kinda familiar, but I don’t know how to fix it. <samplet>lisbeths: Do you have /var/guix/profiles/per-user/root/guix-profile/lib/locales? <samplet>That’s right. Okay, then, what does “/etc/systemd/system/guix-daemon.service” for “Environment”? <lisbeths>GUIX_LOCPATH=/var/guix/profiles/per-user/root/guix-profile/lib/locale <vagrantc_>so, i can use (substitute* "path/to/file" (("foo") "bar")) to monkey-patch a file ... how can i add a line to a file? <vagrantc_>is it possible to substitute EOF or something? <samplet>vagrantc_: (substitute* "file" (("foo") "foo\nbar")) <samplet>lisbeths: Are you sure you ran “systemctl restart guix-daemon”? Everything looks good, so I’m a little flummoxed. <lisbeths>well this time when I restarted it it did something different <lisbeths>though it still says failed to install locale. Doing a guix pull right now <samplet>[x] Install glibc-locales into root profile and restart daemon. <samplet>[ ] Install glibc-locales into user profile. <samplet>[ ] Make sure that GUIX_LOCPATH is set to $HOME/.guix-profile/lib/locale when running Guix commands. <samplet>That’s it. Did you do those other two things? <samplet>You can check with “guix package --list-installed”. <samplet>Or checking if “$HOME/.guix-profile/lib/locale” exists. <dftxbs3e>samplet: I'm getting a /gnu/store/*linux-libre* not found, is that normal?\ <samplet>dftxbs3e: When do you get that message? <samplet>dftxbs3e: Is there any more context? Is it during compilation? <dftxbs3e>samplet: it's an error returned by guix itself <dftxbs3e>samplet: brr stuck forever checking perl-net-ssleay-1.85 <samplet>dftxbs3e: No, though it has been discussed. <samplet>dftxbs3e: I get a substitute for it. You shouldn’t have to build it. <dftxbs3e>samplet: I updated to core-updates earlier <dftxbs3e>I did git pull --branch core-updates etc <dftxbs3e>but then I got guix recompiled and figured that was useless <samplet>dftxbs3e: It’s possible core-updates is broken. You are jumping into the abyss here. ;) <dftxbs3e>samplet: honestly the error is probably caused because I used core-update <samplet>You could try going back to e337061b3a7af892d3a14f5191efb2487ec0fad9 (still on core updates, but two weeks old). It’s a long shot, but it might work better. ***janneke_ is now known as janneke
***vagrantc_ is now known as vagrantc
<kmicu>And I thought 512 lines of backlog should be enough for #guix. NOT. *rekado_ build Guix on Debian GNU/Hurd again <brendyyn>Is it possible in a build system to have a dependency that is only pulled in under certain conditions like if the package source contains certain files? <rekado_>hmm, Debian has guile-2.2 but no guile-2.2-dev, only guile-2.0-dev <rekado_>oh, it only has guile-2.2-doc, no guile-2.2 <rekado_>…so I need to build guile 2.2 from source first :-/ <efraim>rekado_: I have a couple of minutes if that'll help <rekado_>after the Guix database got corrupted on my other old Hurd VM I now have to do everything from scratch on another VM to build Guix. <efraim>well actually I'm not sure of an easy way, but it defiantely isn't in stable or testing <efraim>not sure about experimental, but I wouldn't bother searching there myself <rekado_>in /etc/apt/sources.list it shows sid and debian-ports unreleased. <efraim>i haven't actually heard of unreleased, but not a big deal <efraim>it looks like guile-2.2 failed one of the tests so it shouldn't be a big deal to just ignore it locally <rekado_>I’ll install guile 2.0 and then use that guile to speed up the from-source-build of guile 2.2 <efraim>i started with guile-2.2 already though <rekado_>I’ve done this all last year and ended up with Hurd bootstrap binaries and all <rekado_>but then the database got corrupted and I haven’t been able to fix it. *efraim has to run off again <another-user>hello, i need to create package for elasticsearch but i have no idea where to start, by reading manual i understood that i probably have to use trivial-build-system <rekado_>ArneBab: what is the value of PYTHONPATH? <rekado_>another-user: what makes you think you need to use trivial-build-system? It’s a very inconvenient build system. Often it makes more sense to modify the gnu-build-system instead. <dongcarl>Would really appreciate if someone could review my GCC 9.1.0 patch <dongcarl>I've got another one coming that's on top of it <dongcarl>And makes cross-gcc from cross-base.scm work with GCC 9.1.0 <Forza>Hello everyone. Is there a way to fix the missing 'ls' and similar when installing from the iso <rekado_>Forza: yes, you can run “guix install coreutils” <rekado_>Forza: in the coming weeks we’ll make another release with a fix for this bug in the installer. <Forza>I think I tried that but it didn't help. I installed both as user and as root <gagbo>I thought this might be a guix-daemon service issue, so I enabled and started the service manually, but it didn't solve this problem <rekado_>gagbo: this would mean that the daemon is not actuall running. Is SELinux enabled? <Forza>rekado_: the ls and other tools were not in the PATH and so didn't work <gagbo>rekado_: yes, it looks like SELinux is out there making my life hard again <rekado_>have you reconfigured your system as recommended in the manual (see link above)? <rekado_>gagbo: we have an SELinux policy for the daemon, but I think it has a minor problem and won’t be sufficient to make the daemon work with SELinux enabled. <Forza>rekado_: reading now. Yea. I have some more steps to do :) <rekado_>Forza: the good news is that you can get out of this <rekado_>gagbo: you would need to use SELinux auditing to see what actions have been denied. <rekado_>gagbo: and you’d need to install the guix-daemon.cil policy file. <gagbo>rekado_: will do, as disabling SELinux (setenforce 0) did work <rekado_>but setting SELinux to permissive is faster. <rekado_>you need to reboot after disabling SELinux, no? <gagbo># setenforce (0|1) works runtime <rekado_>gagbo: the policy file… seems not to be installed, but it is part of the source tarball and is instantiated upon configuring. <rekado_>if you want to use it without configuring you’d need to replace the placeholders <rekado_>i.e. @guix_sysconfdir@ –> /etc, @guix_localstatedir@ —> /var, @storedir@ –> /gnu/store, @prefix@ –> … “/usr” maybe? <gagbo>I'll try that. I know one day I'll get away from RH linux and SELinux hell, but not today <rekado_>by using this you might step a little more into that hell. <rekado_>I know of one report claiming that the policy is insufficient when using Guix. <rekado_>I’d love to fix this one day, but I can’t do it these days. <gagbo>this is a lisp file, I don't know exactly how I'm supposed to run it yet <rekado_>it’s not a lisp file. It’s written in the CIL, the SELinux intermediate language, which uses S-expressions. <rekado_>the manual says in section 2.4.3 SELinux Support that you need to run semodule -i etc/guix-daemon.cil <another-user>rekado_: well, because there is nothing to build - it's just set of sh/jar files <rekado_>jar files are usually the result of a build. <amz3>hello, what is the equivalent of build-essential package? <amz3>it think it is gcc-toolchain <brendyyn>I wanted to run tinc on debian with a guix pack, but It tells me I don't have permission to access /dev/net/tun <gagbo>rekado_: I don't have much more time to work on this today, but I wrote where I was at in this snippet https://framagit.org/snippets/3535 sorry for the error messages in french, I'll try to fix it later tonight/this week, I really need to go now. <brendyyn>strange. I'm just adddding #:use:module (guix build-system gnu) to my module and it says home/b/code/guix/unrar.scm:6:13: warning: unknown keyword or bad argument <amz3>#:use-module with dash not : <nee`>How do I build with a different gcc version? When I put gcc-8 in native-inputs it gets used, but stdlib.h is not found. I already tried make-stdlibc++ and putting ("gcc" ,gcc-8 "lib") into inputs both don't work. When I put gcc-8 into inputs instead of native-inputs gcc-5 is used to compile. <brendyyn>wow so in other news, i have my laptop and desktop connected via direct ethernet, with a tinc vpn on top. when im pinging between them over tinc, its about 0.8ms. but if i start wiggling my mouse around real fast it goes down to 0.1. its very consistent too. i wonder what kinda gremlins cause that to happen <brendyyn>actually it does that even without the vpn <mbakke>nee`: You might need to also add a phase that does `(for-each unsetenv '("C_INCLUDE_PATH" "CPLUS_INCLUDE_PATH"))`. <dongcarl>when adding a new patch, do you need to add them to `gnu/local.mk`? <dongcarl>It would seem that I've made a mistake in a patch I've submitted, I'd just submit another? Should I note in the previous one that it's invalid? <dongcarl>brendyyn: Anything special to do to indicate that it's v2? <brendyyn>dongcarl: they can be generated with git *dongcarl is sorry he's new to submitting patches via email <brendyyn>But it's nothing special, they just make the email look like [PATCH v2 1/3] ... <brendyyn>So you could put that in the subject line if you like <r3s1stanc3>how is the current maintainer story in guix? i've heard from nix that they have dedicated maintainers per package and that there's a bot in place to notify maintainers if the source upstream gets updated. has guix something similar? <efraim>We share maintainace across all the packages, with some people working harder on a subset they care more strongly about. We also have 'guix refresh' to help keep track of many upstream packages and also of CVEs <ArneBab>rekado_: it’s echo $PYTHONPATH -> /home/arne/.guix-profile/lib/python2.7/site-packages:/home/arne/.guix-profile/lib/python3.7/site-packages, but I don’t find where that is set <ArneBab>another problem I currently have is: $ LANG=C sudo <ArneBab>sudo: /home/arne/.guix-profile/bin/sudo must be owned by uid 0 and have the setuid bit set <efraim>ArneBab: you need the setuid sudo in the OS config <brendyyn>r3s1stanc3: people have their own tools but there is also guix refresh <brendyyn>I've never used it but I don't look after any "important" <ArneBab>efraim: does that mean I need to add sudo as package to /etc/config.scm and reconfigure the system? <efraim>ArneBab: it should already be there in /run/setuid...something *efraim is on a phone ATM <davexunit>I'm going to update guile-next to the latest commit from guile master <davexunit>just gotta compile guix for the first time in awhile... <ArneBab>efraim: yes, it is there, but it doesn’t seem to be taken <ArneBab>efraim: does this happen when I install it as user? <M4R10zM0113R>How'd I go about forcing everything to go without pulseaudio? *dongcarl is glad contributing to Guix forced him to set up `git send-mail` <kmicu>dongcarl: Do you enjoy not using 500MB of RAM to send a patch or is there something else more cool in your case? <dongcarl>kmicu: Haha true... I use Protonmail, so had to set up hydroxide for a bridge SMTP server, but wasn't that hard *dongcarl notices that guix doesn't have a hydroxide package yet <manzerbredes>hi ! I just started using guix yesterday (on ArchLinux) and I am not able to install the hello application (the installation phase takes to much time) <manzerbredes>guix is installing python since more than 4 hours (it is on the check phase) *sirgazil is letting go the bugs he collected until yesterday... <sirgazil>This is the first one: Installer: Graphical installation failed because of truncated UUID *nckx is happy to see that we've not been overrun by spammers. <nckx>sirgazil: A nice weird bug right off the bat… o_O <nckx>Silly Guile, that's not what null-terminated string means. <sirgazil>How do you check in a terminal if a DVD is mounted. When I insert CDs or DVDs in the GNU system (installed with Guix 1.0 ISO installer) GNOME does not say anything and GNOME Files doesn't list the DVD drive. <esper0s>sirgazil if the dvd gets automatically mounted you could use the command "findmnt" <esper0s>which lists all the mounted filesystems, the most recently mounted one is listed last *nckx points out lsblk too. <nckx>esper0s: So it's the same output as ‘mount’ but prettier? ☺ <nckx>(Don't get me wrong, I love prettier.) *nckx runs dfc to prove it. <sirgazil>esper0s: Thanks. "findmnt" does not display the DVD. <nckx>That will display unmounted block devices as well. <sirgazil>nckx: lsblk says "sr0 11:0 1 1024M 0 rom" <esper0s>i could also recommend the chapter 3 "Devices" and especially the subchapter 3.4.2 of the book "How linux works" 2nd ed by "Brian Ward" <esper0s>which can also be found freely in pdf online *nckx runs sed s|Linux|GNU/Linux|g on the Internet again. <nckx>esper0s: Thanks for the book recommendation. <esper0s>You recommened lsblk so you might actually know most of the staff, although its always good to keep a library *nckx agrees; looks at the pile of unread Scheme PDFs on their block device; sighs. *esper0s sighs knowningly <nckx>sirgazil: Can you manually mount /dev/sr0 somewhere? <nckx>Alternatively, ‘udisksctl mount -b /dev/sr0’ will work without root privileges, if it, er, works. <sirgazil>nckx: I'm not trying to do that. I expected the system to do it for me, so I reported a bug about that as well. <sirgazil>nckx: But I can try if the results of mounting manually adds more information to solve the bug. <nckx>sirgazil: That was the intention. I'd read your bug report ☺ <nckx>‘Doesn't work’ isn't much to go on, and I know nothing about GNOME so that's the only approach I can suggest. <esper0s>by the way i saw the bug report and it seems that lsblk does indeed list the CDROM <esper0s>its just that its not mounted i think <sirgazil>nckx: Yes, I "Doesn't work" is all I saw (I'm a high-level user) :) <nckx>M-hm. That was the ‘rom’ posted here too. *nckx didn't even know ‘rom’ was the name used, so hipsterly ODD-less are they. <sirgazil>esper0s: Yes, it is listed, but not mounted. <sirgazil>nckx: So, should I run "udisksctl mount -b /dev/sr0" <nckx>sirgazil: Nothing wrong with that, just a bit hard to debug ;-) <nckx>With a little luck it will helpfully say ‘service foo not running’, but this is GNOME… <nckx>so it might just hibernate your machine instead. <sirgazil>nckx: Object /org/freedesktop/UDisks2/block_devices/sr0 is not a mountable filesystem. <nckx>…hum. Now that I was not expecting. <nckx>sirgazil: Just to make sure: ‘sudo mount /dev/sr0 /mnt && sudo umount /mnt’ does work? *nckx tries udiskctl on a USB drive on their very minimal system and it works (っ ºДº)っ ︵ ⌨ <sirgazil>nckx: Sorry, I have to go. Thanks for the help :) <amz3>M4R10zM0113R: what are you trying to achieve? <M4R10zM0113R>want it to do all from source, have no pulseaudio, or be able to configure each package / package flags globally <GuiXUser-44470>r the partition, i am unable to proceed further or click / select on OK button. please help with the next steps <GuiXUser-44470>ng the values for the partition, i am unable to proceed further or click / select on OK button. please help with the next steps <amz3>GuiXUser-44470: processing please wait... <amz3>M4R10zM0113R: there is nothing like package flags in guix <amz3>GuiXUser-44470: the keyboard is unresponsive? did you try to reboot and retry? <amz3>GuiXUser-44470: is the keyboard plugged in? <M4R10zM0113R>amz3: globally and locally? I see there's guix edit, but I'm unsure about a lot of the syntax <amz3>GuiXUser-44470: reboot and retry <M4R10zM0113R>... and having to do every single file manually, if it's the only way around <amz3>M4R10zM0113R: you want to install your packages from sources? guix edit is useful to edit package definitions. <amz3>GuiXUser-44470: hmm, idk wait a little someone will chime in <M4R10zM0113R>amz3: yes, all from source, so I can edit the packages into not using some stuff, like PA/NM/etc <nckx>M4R10zM0113R: It's a pretty conscious decision not to fill our packages with (if (use FOO) '("--enable-foo'#")) stuff, so we don't have anything like USE flags & the like. <GuiXUser-44470>amz3, the laptop was with a windows 10 OS, had the boot order set to DVD and rebooted it - landed in the GUIX installation wizard and opted for Guided Graphical User Interface,.. set the keyboard and other timezone related selections... and on the partition section.. its looping (unable to go back or forward) please share some shortcut keys to fix <nckx>…and probably never will. <amz3>M4R10zM0113R: to do that you will to rewrite the package definitions. There is an article about the lack of package flags in guix, tl;dr: in the long term it is not useful to have them. <amz3>GuiXUser-44470: can you Ctrl+Alt+1 and tell me what you see <nckx>Does the installer work in a VM? Is there a version that does? (I tried a quick & lazy ’qemu-system-x86_64 -cdrom/-hda guix-system….iso’ but that panics due to an ‘incomplete write’) *nckx has never even seen the installer and would like to help people using it. <GuiXUser-44470>amz3: Ctrl+Alt+1 (when tried on the 'Partition Creation' GUI - no response <nckx>M4R10zM0113R: Have you read the manual? Specifically ‘Defining Packages’? <GuiXUser-44470>amz3: the arrow keys and enter / return keys are working.. but unable to reach the OK button on the bottom of the screen <lambda-smith>Hi guys, so I just defined a new guix package! (That was cool.) And now I'm wondering how can I add my new package as a dependency to one of the existing package in the "core" guix. <amz3>M4R10zM0113R: if you are getting started it is preferable to add new packages first to get a feeling how things work <amz3>M4R10zM0113R: or updat existing packages <lambda-smith>I tried searching the doc, but didn't find any pointer on how to "edit" the package defitinion to add new dependency (I only found how to replace existing ones). <amz3>lambda-smith: don't be sorry. <amz3>lambda-smith: it is for your own use or you want to publish the package? <lambda-smith>Both, I want to test it on my system first, then publish. <amz3>lambda-smith: I am asking because there is several answers to your question. <lambda-smith>Ah, then publish it is. Essentially, I'm trying to add Thai language support to Guix. <lambda-smith>Which require adding new font package and a new library, which is needed to be linked to pango. <amz3>lambda-smith: thx, I was asking for context actually <amz3>lambda-smith: to contribute packages you need to setup guix git <amz3>lambda-smith: basically, guix environment guix && ./configure --prefix=/ && make <nckx>amz3: Did we get rid of the --localstatedir=/var requirement? <nckx>lambda-smith: Otherwise, add that ☝ to ./configure too. <amz3>nckx: no, but --prefix=/ does work on my side <lambda-smith>quick question, is it okay to point to the /var directory in my actual system? <nckx>lambda-smith: Yes, it's even required to properly use Guix from git at all. <amz3>lambda-smith: then you open the guile module where you want to add the package definition, IDK where it should land. To test the package definition do something like ./pre-inst-env guix build package-name <amz3>lambda-smith: yes, the --prefix=/ works using default directories, if you have something else there is several options available look into configure --help. <lambda-smith>nckx & amz3: Ah, I see. Thank you. I was afraid I might break my current guix store <amz3>lambda-smith: nckx: I only tried, --prefix=/ on top a foreign distro <amz3>lambda-smith: you do that as root? <amz3>lambda-smith: It possible to break guix or guix system, if you chown files inside ~/.guix-profile (or /gnu/store) <amz3>maybe they are other ways but I don't know <amz3>lambda-smith: building is fine I doesn't break <amz3>lambda-smith: you are on another distro or do you use guix as an OS? <davexunit>nvm re: updating guile-next. test suite broken. I remember having this issue before. I think only the pre-release tarballs reliably build. moving on. <lambda-smith>amz3: another quick question: --prefix=/ is required, right? <nckx>lambda-smith: (not amz3 but) Yes. <lambda-smith>Incidentally, great job guys. I've only been using Guix for a couple of days, but I'm already in love. <lambda-smith>It's especially great how there are so many up-to-date lisp compilers/interpreters in Guix. :D <nckx>lambda-smith: Thanks! Guix is still under heavy development (even for a 1.0) but is awesome indeed. <amz3>lambda-smith: if you do not provide --prefix=/ what happens is that when you try to build something, 'guix' command will try to reach 'guix-daemon' that is responsible to execute the build, 'guix' will be unable to reach 'guix-daemon <amz3>lambda-smith: eventually, if you trick 'guix' to contact guix-daemon, you will end up building or installing in the wrong directory, but AFAIK it is harmless <lambda-smith>So the official doc should be update then? I didn't see any mention of the flag anywhere. <nckx>lambda-smith: The manual suggests --localstatedir=DIRECTORY (my /var above), which is equivalent to amz3's --prefix=/ in this case. <nckx>--localstatedir=/var does less & is a bit more ‘correct’ for a traditional package (it would install into /usr/local, for example) but that's not very relevant for guix: you'll never ‘make install’ it. <nckx>atw: ‘Pesky ^L glyphs certain developers use…’ yeesh. Someone's opinionated ;-) <g_bor[m]>I have a vm consistently failing to start sshd on boot. <g_bor[m]>How could I get more info than service failed to start? <g_bor[m]>Where does shepherd log the service outputs? <amz3>sheperd rely on syslog IIRC <GuiXUser-44470>@amz3 Ctrl + Alt + 1 is not responsive at the create partition GUI <amz3>GuiXUser-44470: you should memory check your RAM and check the health of your hard disk <amz3>maybe someone will chime but It is unlikely that Ctrl + Alt + 1 doesn't work <amz3>GuiXUser-44470: it is F1 not 1 <GuiXUser-44470>@amz3, thanks for your quick response - have a 6GB RAM and 320G HD - didnt forsee any issue earlier... will try to reboot again and will follow the steps.. looks like the create partitiion step is just looping without moving to the OK / selection <amz3>GuiXUser-44470: sorry, the keys are Ctrl + Alt + F1 <amz3>GuiXUser-44470: yes, but you are the first to have *that* issue <amz3>GuiXUser-44470: so *maybe* it is an hardware issue? again try Ctrl + Alt + F1 <GuiXUser-44470>@amz3 - yes I tried Ctrl + Alt + F1 with no response. [tried on the create partition window] <amz3>GuiXUser-44470: it that console, type 'dmesg' a look for suspicious messages <amz3>GuiXUser-44470: try F2 then, sorry I don't xp with that installation method <amz3>GuiXUser-44470: Ctrl + Alt + F2 <nckx>atw: Ah, thanks; I've stumbled onto that site before. <nckx>(Most of the seasoned users here didn't use the installer to install, just in case anyone's wondering why we all sound so clueless about it. ^_^) *nckx still trying to boot it in QEMU with no success. <GuiXUser-44470>@amz3 Ctrl + Alt + F3 switched to a shell / "Welcome to the Installation with GNU Guix!" for installation "Manually" - with a root@gnu prompt. Should I follow the manual installation steps? let me search online for the steps if GUI installation is not working <nckx>M4R10zM0113R: We'll take that praise. <M4R10zM0113R>I have to dhclient manually because for some reason it never did the connection to the internet, continuing on the network step warning shouldn't be trobule much <M4R10zM0113R>so far only the guided partitioner without encryption works <GuiXUser-44470>M4R10zM0113R, Did you try GUI installation in a windows 10 system? which ended in a guided partition suggestion and then 'STOPPED' unable to select the 'OK' button / selection on the bottom of the "Create Partition" window? <M4R10zM0113R>I booted into the iso, which should be independent of the installed OS in the machine <amz3>GuiXUser-44470: what dmesg says? <amz3>GuiXUser-44470: indeed, it doesn't depend on whether windows or something else is installed in the hard disk. guix and usually all gnu/linux livecd start before windows has chance.. <GuiXUser-44470>yes M4R10zM0113R, amz3 I did boot into the DVD iso , got the Guix installation steps ...chose the GUI guided option and 'unable to proceed' after partition options were selected (the OK button was not getting selected) and now using Ctrl+Alt+F3 - entered into manual installation mode with the root shell <GuiXUser-44470>amz3 M4R10zM0113R any reference material or steps to expedite the installation in the manual / from the root shell will be my next try.. in the meantime.. is there a place where i can raise it as a bug so that someone could fix it for future users? <amz3>GuiXUser-44470: you don't have enough information to fill a bug <amz3>GuiXUser-44470: that's why I asked you to type 'dmesg' in the terminal you got while pressing Ctrl + Alt + F3 ***jonjitsu is now known as johnzorn
<cinterloper>hello, is there some good information available about customizing the guix livecd? <cbaines>cinterloper, what kind of customisation did you have in mind? <cinterloper>some other software, I use a livecd for day-to-day computing, so basically my normal suite of user software <cbaines>I'm not sure, but I think the livecd and the installer are the same thing... <cinterloper>huh ok, ill look through that, thanks for the pointer <cinterloper>also, i noticed in the root of the VM image, when i extracted the rootfs, theres no /bin/bash, everything seems to be invoked in /run/current-system ... is this a container system? is there a good explination on how the init system works to reach this from boot? <cbaines>cinterloper, Guix packages are quite similar to "containers" as some people think of them, as the dependnecies are well managed <cbaines>Not having /bin/bash removes any risk that things will work differently if it changes in behaviour (between different bash versions for example) <cbaines>As for the init system, I'm not sure what you mean? <lambda-smith>amz3: nckx: Hmmm... guys, do I need to perform this step: "$ sudo -E ./pre-inst-env guix-daemon --build-users-group=guixbuild"? <nckx>lambda-smith: I haven't been paying attention, but assume you have a working ‘guix’ command on this system already? <nckx>If so: no. A compatible guix-daemon is already running. <nckx>You can skip that, then. <cinterloper>Im not saying its a bad thing, im just wondering what the path is from kernel entrypoint to running /run/current-system/profile/bin/bash <cbaines>cinterloper, as far as I understand it, Linux starts the Shepherd (the init system, PID 1), then it starts all the services and runs through the system activation proecess <rekado_>ArneBab: your PYTHONPATH looks dangerous, and that’s not your fault. <rekado_>ArneBab: you have both Python 2 and Python 3 directories on there, and Python isn’t smart enough to not load incompatible binaries. <rekado_>ArneBab: this is not your fault because Guix does this for you when you have both Python 2 packages and Python 3 packages in your profile. <rekado_>Once every two years I post to the mailing list in the hopes of finding a way to do without PYTHONPATH, but it hasn’t yet revealed itself. <rekado_>ArneBab: you can move your Python 2 things to another profile so that Guix won’t be tempted to add them to PYTHONPATH. <rekado_>or you can manually remove the Python 2 things from PYTHONPATH. <Aurora_iz_kosmos>I have another HDD with more free space on one of my machines where I'd rather place guix-related files. <Aurora_iz_kosmos>amz3: Yeah, I'm aware of how to do it manually. I just wanted to know if there was a way to do it with the autosetup script. Seems like not really. <rekado_>Aurora_iz_kosmos: your best bet is bind mounting. Changing the store directory name means that you won’t be able to use the pre-built binaries at all. <rekado_>because store file names end up embedded in binaries, so you can’t just move everything to /home/aurora/guix/store, for example. <rekado_>I’m just saying that it *must* be /gnu/store if you want to use our binaries. <pkill9>that's what i did when i ran guix on a foreign distro, it works well <ArneBab>rekado_: and yes, my pythonpath is dangerous … <ArneBab>efraim: after guix remove sudo sudo works again — thank you! <ruffni>my computer shut down unexpectedly during 'guix system reconfigure'; now the same command complains: "guix system: error: error parsing derivation '/gnu/store/hash-xf86-input-libinput-0.28.2.drv': expected string 'Derive(['"... how can i fix it? <pkill9>ruffni: maybe try removing it by running `guix gc --delete /gnu/store/hash-xf86-input-libinput-0.28.2.drv` and try reconfiguring - maybe the file was corrupted by the shutdown? <rekado_>ruffni: you can run “guix gc --verify=repair,contents” as root <sirgazil>Have you experienced errors like "[Status: Error, Address: <bug-guix@gnu.org>, ResponseCode 421, Host not reachable.]" when reporting bugs or sending messages to the mailing lists. Some of my messages are not delivered on time because of this... <Aurora_iz_kosmos>There's mention of nscd and recommending it for foreign distros, but not all that much about how to ensure the right nscd is running. <ruffni>guix gc --verify=repair,contents finds stuff, repairs, but "error: cannot repair path '/gnu/store/bla'" ... should i just manually delete them all? why is guix gc unable to fix them? <g_bor[m]>I still have no idea why openssh fails to start on this vm. I only have the single line from shepherd in /var/log/debug, that says the service ssh-daemon could not be started. <g_bor[m]>I can start it without problem after logging in. <Aurora_iz_kosmos>Hm. Seems like I misunderstood. nscd is used as a proxy from the foreign system to guix, and should be installed via the foreign system's tools. <jfred>Does anyone here use flatpak applications on a guix system? I've downloaded flatpak from the guix repos, but when I try to add flathub as a remote, the flatpak client segfaults <jfred>...I realize it's a bit weird to mix flatpak and guix, but it'd be helpful for some gnome apps that aren't packaged in guix yet/that are officially packaged as flatpaks <Aurora_iz_kosmos>jfred: They may be conflicting. (Means you might have to write yourself a wrapper) <jfred>mmm, I have not - I've only just started using guix <rekado_>jfred: a segfault is generally not okay. Could you please report a bug by sending a description of the problem via email to bug-guix@gnu.org <rekado_>g_bor[m]: this sounds familiar because it came up on the mailing list a couple of times. There seems to be some sort of timing problem. <rekado_>Aurora_iz_kosmos: yes, you need the one from the host system *rekado_ builds all of the updated bioconductor packages… <rekado_>NSCD allows us to connect via socket instead of having to load any binary plugins directly <rekado_>so software from Guix simply talks the host system’s NSCD via socket, which in turn takes care of talking to the host system’s plugins. <ruffni>i'm unable to fix my system; guix system reconfigure fails and i'm unable to resolve the issues using "guix gc --verify=repair,contents".. any ideas? removing the failing packages by hand fails (read-only file system)....? <rekado_>ruffni: you shouldn’t ever delete anything from /gnu/store manually <rekado_>to delete things you need to use “guix gc -d /the/thing” <amz3>I always the guix hash wrong. Should I use 'guix hash -rx .' to generate to guix hash? <gagbo>rekado_: the .cil file seems to work once I added the typeattribute file_type to basically everything (maybe unnecessary), and when I start the daemon manually (sudoing the command in ExecStart part of the service), everything works as expected (I can guix pull and guix install glibc-locales elsewhere) <gagbo>For some reason, it doesn't work when I use systemctl to start the service, but at this point I'll just kill selinux I think <amz3>by the way regarding my guix hash -rx issue, I forgot (several times!) to clean the repo from files that are not tracked by git. Something like 'git clean -fxd' is required before calling guix hash -rx . <kmicu>nckx: [jokin’] how dare you bash my beloved page delimiters! Get Off My Lawn! <nckx>That was quoted from atw_'s link. *kmicu thinks it’s time to increase backlog from 512 to 1024. You folks are on fire today. <kmicu>Quoted w/o quotes‽ Fake news. <kmicu>Oh, now I see quotes. I shouldn’t IRC before sleep. <nckx>Everything else is technically before sleep. <kmicu>Time before-bedtime, wine and #guix do not mix well. <nckx>I'm guessing wine is not an emulator in that sentence… <nckx>Here, have a tasty Belgian beer instead. <kmicu>Were our amazing IceCat users affected by recent Mozilla kerfuffle? <pkill9>i didn't notice anything, i only have on eextension installed, umatrix <thomassgn>can't find MIT under licenses, does guix use a different name for it? <g_bor[m]>rekado_: thanks, I also noticed this on several bare-bones like system, but did not happen to me on desktop-like systems. Does this give any clues on where the timing problem might lie? <nckx>thomassgn: *Probably* expat. <nckx>(‘MIT’ is ambiguous and can refer to two licences, the most common is expat.) <rekado_>gagbo: wow, that’s good news. I don’t know if “typeattribute file_type” is necessary (nor where exactly it would be needed). Could you please send a diff to guix-devel@gnu.org so that I can play with it on my Fedora machine with SELinux? <kmicu>GNUzilla’s IceCat 1:0 Mozilla’ Firefox <nckx>kmicu: My IceCat is unaffected, I noticed that a certain about:config option (which I forget) that someone recommended disabling was already disabled for me. <gagbo>rekado_: will do, I didn't know if there was an issue tracker or something else. It's going to be very crude though <nckx>Something validate signatures something. Maybe we got lucky. <nckx>thomassgn: Don't guess, choose the right one based on the exact wording 😛 <rekado_>gagbo: our issue tracker can be reached by sending email to bug-guix@gnu.org, but I think in this case guix-devel@gnu.org is fine, unless you are certain that the file we provide is actually wrong. <gagbo>I definitely am not certain :D <rekado_>heh, okay, then let’s discuss this on guix-devel. Thank you so much for testing this! <rekado_>you’re one of the few who bothered actually installing that policy <rekado_>(I wrote it for myself for my Fedora workstation a long time ago, but then stopped using SELinux for other reasons) <pkill9>is ungoogled-chromium failing? or just not been built yet? <kmicu>Thank you Miguel for fixing all those typos! <kmicu>nckx: do you see QEMU 4.0.0 in Guix? <nckx>It's not in Guix. Should it be? <nckx>kmicu: protip: guix search ^qemu$ <kmicu>nckx: thank you for the confirmation. I wasn’t sure whether ‘define qemu’ gives me all definitions. <nckx>Regexes are never trivial, but better than yet another (non-)standard ☺ <nckx>kmicu: I assume you meant ‘define.*qemu’ but hacks like those are never guaranteed to work. <nckx>That said, of course I use them all of the time 😊 *nckx was just battling regexps in Guile. *kmicu is using counsel-rg for an interactive fuzzy search. <nckx>Fuzzy's great until it isn't, but I'll check that out, thanks. <kmicu>It’s an instant search in Nixpkgs/Guix repos on my old second-hand laptop. <g_bor[m]>one more thing I just noticed, the same vm that consistently fails to start openssh on poweron can start it on reboot without problems, *nckx had an ‘nckx/find’ bash script that did the same thing until quite recently. <g_bor[m]>There is also a strange looking error message: error in finalization thread: Success ... <nckx>kmicu: Your pretty modeline distracts me. <rekado_>hey, is magit’s commit feature broken for any of you? <g_bor[m]>I always see this, and I guess it is not related in any way... <rekado_>I wanted to commit a few things before working through the patch backlog, but whenever I hit “c” I get an error (which seems to be related to emacs-dash). <nckx>g_bor[m]: It isn't related. *kmicu inspired by nckx’s comment rechecks whether rms.sexy is still online. <nckx>kmicu: Now that's something I'd not visit before bed time. <nckx>rekado_: I'm guessing -ag is for the Silver Searcher and -rg is for… grep? <rekado_>so! I just built Guix (and all dependencies, manually) on the Hurd. I also still have all the bootstrap binaries from my earlier attempts last year. <rekado_>now I just need to patch gnu/packages/bootstrap.scm, I think <nckx>kmicu: Which (probably bog-standard) package do you use for it? <kmicu>Yes. counsel-* is the important part; -grep/ag/ack/rg is not really relevant. <kmicu>nckx: emacs-ivy, which points to abo-abo/swiper, which has counsel. Don’t ask why ;) <rekado_>I really don’t like ivy’s completion in M-x shell. <nckx>kmicu: I meant the modeline, sorry for being confusing while you're tired. I've just installed emacs-ivy ;-) <rekado_>I often hit TAB to complete to a common prefix and then append * (e.g. to delete all matching files) <rekado_>but ivy insists on selecting *one* matching file <nckx>Eh. Yes. Why would-oh. No. <rekado_>ivy/counsel is otherwise nice, but I really can’t stand this <rekado_>(and I haven’t found any nice documentation to tweak this behaviour) <nckx>That's literally how Windows's CMD shell does ‘completion’ and I've always used it as an example of horribleness. Why would you *add* that…? <rekado_>what’s worse is that aborting the completion *removes* the text that was added by any partial match up to this point. <nckx>I've gone on an emotional roller coaster from WANT to DO NOT WANT in jiffies. <rekado_>so you do "/tmp/“ TAB, type ‘gu’, hit TAB to complete to ‘/tmp/guix-’, and then abort — now you’re back to just /tmp/. *nckx grabs a beer themselves. *rekado_ sips on a cup of cold coffee <kmicu>nckx: emacs-smart-modle-line if I read my ancient config properly. <rekado_>there’s *got* to be a way to change this; how else do other people use this? <rekado_>I also use smart mode line, but it’s not nearly as pretty as the thing in that screenie. <gagbo>rekado_ C-j does the action with the currently typed text iirc <pkill9>could it be possible to add some kind of hook to `guix pull` so that it replaces a package with a graft for all packages? (without needing to maintain a separate repository) <rekado_>gagbo: you’re right. With C-M-j I can keep my text. Quite a handful for something so simple. <rekado_>pkill9: I don’t understand. Why do you want this in ‘guix pull’? <nckx>Aurora_iz_kosmos: There isn't an XDG variable for fonts. Could you be more specific? <nckx>Maybe XDG_CONFIG_DIRS, but mine don't have anything to do with fonts in them 🤔 <nckx>Are you sure this is XDG-related? Have you tried ‘fc-cache -fr’ (and possibly restarting IceCat and/or everything)? <nckx>Ah, sorry, XDG_DATA_DIRS. <nckx>That does have (links to) fonts in it on one machine here. <kmicu>Aurora_iz_kosmos: are you on Guix System or Guix-on-foreign-distro? <nckx>They're just included through ~/.guix-profile, though, so I'm stil sceppie. <nckx>kmicu, Aurora_iz_kosmos: …or still using ‘guix pack’? *rekado_ uses git commit on the command line (because magit is broken) and feels like a cave dweller <kmicu>I had the same issue on a fresh Guix System 0.16 and fixed it by installing new fonts and restarting IceCat (and probably doing fc-cache -frv but I don’t remember for sure). <nckx>Aurora_iz_kosmos: In your XDG_DATA_DIRS? <nckx>rekado_: Hey, command-line git is my cave, get your own. <kmicu>Aurora_iz_kosmos: iirc I’ve seen similar question on the mailing list and someone suggested installing fontconfig from Guix too to fix the issue. *nckx has never used Guix on a foreign distro, sorry. <Aurora_iz_kosmos>nckx: It works nicer when you aren't trying to do something that isn't quite supported. <kmicu>(Plus fc-cache -fr is always recommended after any change related to fonts ;) <kmicu>(-r means --realy-force cuz -f --force is not really force 🤷) <nckx>Bike-shed straw poll: I'm adding basic package-output searching support (so ‘guix search kernel-patch’ now matches wireguard:kernel-patch, for example). On a scale of 4 (as relevant as the name) to 1 (not very relevant, show it near the end), how weighty should that field be in sorting the results? *nckx went for 1 because nckx is an opinionated bastard. <pkill9>rekado_: in my mind, it would change the guix source that's pulled so that package A is replaced with package B using a graft, then that source would be used to build guix <pkill9>so doing that in `guix pull` makes the most sense to me, i don't know enough guile to be able to do such a thing though <rekado_>pkill9: changing the guix source is better done with a git checkout; you can tell Guix to pull from your modified checkout with ~/.config/guix/channels.scm (overwrite the 'guix channel). <nckx>It can be everything from ‘cargo’ (the command you're actually looking for) to ‘doc’ (that tells you nothing interesting) so I was uncertain. <pkill9>am i misunderstanding how grafts work? if I run `guix build --with-graft=mesa=hello xorg-server`, then it starts compiling xorg-server <pkill9>I thought it would just patch the xorg-server output or something to point to "hello" instead of mesa <pkill9>and then put that in a new output <nckx>(Aside: graft store file names need to be of equal length so unless that happens to be the case (-mesa-1234 and -hello-123, for example) that won't work, even as a proof of concept.) <nckx>pkill9: Turns out that that's the reason: guix build --with-graft=glibc=glibc@2.23 xorg-server → no rebuild. <Aurora_iz_kosmos>... So, installing icecat doesn't tell me to set 'export XDG_DATA_DIRS="$HOME/.guix-profile/share${XDG_DATA_DIRS:+:}$XDG_DATA_DIRS"' despite it being necessary for it to work properly. <nckx>I'm pleasantly surprised that --with-grafts transparently falls back to building from source, but I understand that it's surprising. 
