<mbakke>agaric: I'm not familiar with custom keymaps, but I would guess you don't need to mess with "kbd" at all, as long as your custom keymap installs to "out/share/keymaps". <agaric>mbakke: so make a new package? ah! hadnt thought of that, thanks <mbakke>agaric: Let us know how it goes, you'll probably find 'trivial-build-system' helpful. <mbakke>Feel free to reach out to help-guix@gnu.org if you get stuck :) <agaric>mbakke: will do. much appreciated. gonna restart my box real quick <taylan>has anyone ever had problems with a guix-installed GRUB being ignored by the PC under various circumstances? I can boot from my SSD with GuixSD only when it's plugged via USB, but not when it's in the HP "upgrade bay". this used to work with an HDD that had Debian on it, so the upgrade bay itself is principally capable of booting... <zybell_>May it be that the BIOS boots only HDD (fixed drive) not SSD detected as SD-Card(removable drive)in "upgrade" bay? <zybell_>Additionally it could enforce a particular partitioning. <taylan>hmm, I could try swapping the internal SSD and the one in the upgrade bay and see how it behaves then. the internal one just has MS Windows on it. yeah I'll just try that, thanks for the idea. <zybell_>So I did hear once of a BIOS that wanted to boot a drive only when its only partition was set up in the fourth slot. <zybell_>I did hear once of a BIOS that wanted to boot a drive only when its only partition was set up in the fourth slot. <taylan>the Windows SSD booted when in the upgrade bay, and the GRUB one did not while within the notebook. I suppose the problem lies in the combination of the SSD's configuration (partition scheme or whatnot) and SATA, as it works when connected through USB. (the upgrade bay uses SATA) <taylan>legacy. was the same with my Debian HDD so it can't be that though. <taylan>I'll see if VMWare can boot it :P <zybell_>When it doesnt boot, is it because it doesnt show in boot menu? <taylan>zybell_: I choose "upgrade bay" from the boot menu, and it just jumps to the other SSD, booting windows, without saying anything. <taylan>(and while I had them swapped, same thing except I chose "laptop hard drive" in the boot menu, yet it booted the windows one which was in the upgrade bay) <zybell_>There was a patch for grub when the BIOS didnt swap the drive ids? <zybell_>Could it be that the BIOS assigns drive ID by serial number? Normally you can manage boot order in legacy BIOS. Are both drives selectable there? <mange>Hey, has anyone had issues with GuixSD not reconfiguring because it is looking for initrd modules in the wrong place? If I reconfigure it says I need sdhci_acpi, but then when I add it to initrd-modules it fails because it looks for sdhci_acpi.ko instead of sdhci-acpi.ko, and so can't find the module. <mange>I can patch (gnu system linux-initrd) to fix it, I assume, but I wanted to check if anyone's already aware of it before I file a bug. <zybell_>mange Are you sure that is the case? modprobe changes - and _ correctly afaik <mange>zybell_: Yeah, but linux-initrd uses find-files. <agaric>mange: not sure if this applies, but i had reconfiguration troubles with the nvme module. turns out i could ignore reconfigure's warnings with --skip-checks and things would build fine. <mange>agaric: I don't think that'll work, because it's not a check. It's actually crashing when trying to build a flat directory of the modules (see `flat-linux-module-directory` in linux-initrd.scm). <zybell_>In guix I would expect looking into the wrong path more likely. Why does find-files work with the other module with -? <mange>Some modules have the _ in the filename, too. <mange>So I need two modules, mmc_block and sdhci_acpi. mmc_block is in a file called mmc_block.ko, but sdhci_acpi is in sdhci-acpi.ko. <mange>I just don't understand why this has suddenly broken. I haven't had a problem reconfiguring from here before, and it doesn't look like the code has changed since my last reconfigure. <zybell_>yeah _ in name - in filename is the rule <mange>But mmc_block.ko seems to violate that rule, then. <zybell_>seems. but find_files should at least find both files or should use a list. What if that list is wrong? <mange>The list that find-files is returning is empty. The error message is "module not found" (it would be different if there were too many). <zybell_>no, using /lib/modules/$ver/modules.* <taylan>problem solved: no bootable flags were set on any partition. when that is the case, you can boot over USB but not SATA apparently! <taylan>(FWIW this is my fault not the guixsd installer's. I did my partitioning manually.) <wxie>Has anybody installed a Chinese input, like fcitx? ***fkz is now known as Guest54768
<rekado>sneek: later tell wxie For Chinese input I use ibus with ibus-libpinyin. <IntoxicatedHippo>Is there a way to override a package that's a dependency of a different package? I want to add a compile flag to xfce4-power-manager which is a dependency of xfce <rekado>IntoxicatedHippo: you can create your own variant of xfce4-power-manager and then use input rewriting to let xfce use it. <rekado>or you could use Guix from a modified git checkout. <rekado>is that compile flag generally useful? Would it make sense to add it to the package definition for all users? <IntoxicatedHippo>The flag is -DXFPM_SYSTRAY which adds an option to show an icon in the system tray (at least I think it does, it's not documented) <IntoxicatedHippo>Also while talking about XFCE, xfce-desktop-service doesn't extend polkit with the actions from xfce4-power-manager if someone wants to fix that <ocmylife>Good morning guys. Does anyone else have the problem that module nvme.ko couldn't be found in linux-libre-4.15.15/lib/modules? <IntoxicatedHippo>Ignore my messages about the tray icon, it's enabled by default but it was added in 1.4.4, guix is currently using 1.4.3 <thorwil>ocmylife: roscap had that issue, mentioned here 3 days ago <sneek>thorwil, you have 1 message. <wigust>rekado: Thank you, I didn't know that there is a difference. <ocmylife>@wigust I hadn't known this as well, as english isn't my native language, but I will switch guys with folks next time :-) <ocmylife>@thorwil So the easiest way to fix this issue during the installation, is to add nscd to (packages (cons* ? I'm a noobie on guixsd ;-) <thorwil>ocmylife: some *folks* started using guys as gender-neutral. even cases of women referring to other women with guys are known. but at it's roots, it's not and thus is best avoided if you want to be inclusive <thorwil>ocmylife: nscd is apparently a solution to a problem when running guix on a foreign distro <thorwil>ocmylife: but it sounds like you are busy with guix sd proper? <ocmylife>I had a problem with the grub-bootloader installation after running the guix system init command, so I thought it would be a clever idea to run guix pull first to update the packages. But then I was running into the missing nvme problem I posted before. So for now I think, that adding nscd to packages is the right decision to get my installation successful. I'm familiar with compiling kernels myself, and I will need to build IT <ocmylife>myself, cause my Dell XPS9360 has a Atheros ath10k (QCA6174) wifi adapter, which requires a binary blob. But at first I want to finish my installation. Step by step :-) <ocmylife>So far I get my internet connection running by usb tethering from my phone. It's not that comfortable, but I'm quite confident, that I could get the wifi adapter running as well. I know, that it isn't officially supported on guixsd <thorwil>so i switched to an emacs provided by guix, then noticed that fonts are not available. wanted to install just noto mono, but found out there's only a google-fonts-noto <thorwil>more than 1 GiB hd space later, i decided to remove the noto package and use adobe source code pro instead <thorwil>despite a gc run inbetween, emacs still lists noto, but not the adobe font <thorwil>though it does print: (emacs-25-3:5799): Pango-WARNING **: shaping failure, expect ugly output. shape-engine='PangoFcShapeEngine', font='Noto Mono Bold Oblique 21.119140625', text='Noto Mono Bold Italic' <thorwil>`xset +fp ~/.guix-profile/share/fonts/truetype` fails, listing possible causes: doesn't exist, wrong permissions, missing fonts.dir, incorrect font server address or syntax <thorwil>well, it does exist, has fonts.dir. ownership is root:root, though <civodul>thorwil: i think you need to "readlink" that <civodul>xset +fp `readlink -f ~/.guix-profile/share/fonts/truetype` <rekado>civodul: thank you for the prlimits64 patch! <rekado>I wasn’t able to allocate time to testing this until now. <grillon>hi there, I gave up the idea to use guixSD as main system but I would like to improve to make it possible. What is the good path to follow? (I had two main problems 1)Use japanese IME / 2) configure multiboot) these problems show me I don't know how programs are linked together. <pkill9>grillon: why did you give up on using GuixSD as main system? <civodul>for multiboot you should share your experience on help-guix@gnu.org <grillon>I gave up because I need a good language support. But when ready I'll give a new try because I like the philosophie of that system <rekado>grillon: I only use a Chinese input method. To make it work I only had to install ibus and ibus-libpinyin, and then set the suggested environment variables (e.g. GUIX_GTK3_PATH). <grillon>hi civodul, pkill9, I'm not alone to use the system so I could not keep it without a good language support. <civodul>installing ibus like rekado suggests may help <civodul>but maybe we should make that work out of the box? <grillon>civodul : a language selector as other distro is welcome but it's not a problem to install ibus and ibus-anthy. the problems is it's not sufficient to make it work :( <grillon>how did you configure environment variables rekado? <grillon>because I could launch the manager et select language but it's not taken into account. I guess the problem comes from my environment variables <pkill9>grillon: if you put them into .profile then you need to logout and log back in <grillon>XFCE but I tried with gnome and it's not working too <rekado>my GUIX_GTK3_PATH is /run/current-system/profile/lib/gtk-3.0 and I (now) use GNOME. <rekado>I have to regularly delete ~/.cache/ibus, though, because the cache may refer to stale store paths. <grillon>thank you rekado I don't have that env var it may be the cause <roptat>grillon: did you configure ibus? <roptat>did you run ibus-setup to select the anthy input system? <roptat>in what profile did you install ibus? user? root? system? <grillon>I have not tried to add GUIX_GTK3_PATH rekado gave me. I could not try until this evening :( <roptat>you may need to set XMODIFIERS=@im=ibus and QT_IM_MODULE=ibus too <grillon>is it different to install as root or as user? <roptat>yes, it's different but you shouldn't install ibus as root, because it wouldn't be accessible to your user <grillon>ok I see. you mean in the ideclaration of the env var in .profile <rekado>I should say that after switching to GNOME the *only* variable I have set that relates to this is GUIX_GTK3_PATH. <roptat>or maybe the file is simply not loaded by your graphical environment <rekado>I didn’t need to set any of the other variables. <roptat>I could only make it work with GUIX_GTK{3,2}_IM_MODULE_FILE, and don't have GUIX_GTK3_PATH (or maybe it's set automatically by guix) <roptat>also I never had to remove ~/.cache/ibus <rekado>I had to remove it often, usually after upgrades. <Sleep_Walker>can be guix container used for running other linux distribution in container? can guix run docker images? is anyone working on packaging docker? <roptat>guix uses container technology to isolate processes while giving them access to some subset of the store. You could run another linux distribution in one of them but it's not easy <roptat>you can generate docker images containing a guix system with guix, but not another linux distribution <roptat>and I don't know the state of docker on guix <Sleep_Walker>unfortunatelly I am in need to run docker stuff to work with others in team <civodul>my new Guix-HPC intern at work is starting today \\o/ <civodul>there are maybe 10 of them we can get rid of, but the rest is going to be tricky <rekado>I should remove bazel from the best practices section on bootstrappable.org <civodul>maybe you need a Phoronix headline first <civodul>"bootstrappble.org To Remove Bazel from Best Practices Section" <rekado>also worrying: “You are building Bazel from source (bootstrapping). This is a resource-hungry operation: you'll need at least 4 GB RAM, maybe even more (6 or 8 GB).” <rekado>tensorflow itself also bundles lots of things and expects you to build them with bazel. <civodul>damn it, it's worse than 'guix pull' <zybell_>Did you really read the best practices page on bootstrappable? There is only explained how bazel, that normally has to be installed before it can be build, can be build anyway without installing it first. Only for this problem that page applies. There is *no* mention of 'minimal system','few dependencies','no resources' and other connotations one thinks of when the word 'bootstrap' is mentioned. <rekado>zybell_: I published the page on bootstrappable, so yes: I did read it. <rekado>the point is not that Bazel needs more than a few dependencies. <rekado>bundling 60+ binaries to build the thing is not what I consider to be in line with “best practices”. <zybell_>If you need bazel, you must bootstrap bazel. If you need 60+ binaries for that, you must bootstrap *them*, not bundle. <zybell_>Maybe that should be explained more fully on that page. <rekado>but that’s not what we want from developers of other build systems or compilers. <rekado>bootstrapping these 60+ binaries requires bootstrapping of two other recursively implemented build systems and bootstrapping of the JDK. <rekado>if we were okay with that we should also be okay with GHC needing GHC, because it just takes a very long bootstrap chain <zybell_>GHC has a shorter chain available through another Haskell provider. If bazel has a shorter chain available, then I would mean: use it. But until that is and the longer chain is not obvius per template 'look up deps;build them;use deps to build ...', then that longer chain should be described. Yes that is what I would expect from that site. If the page 'Best Practices' is *named* appropriately could be a matter of debate. <Ulrar>Hey, I was wondering if there is anything like nixos's "nixpkgs/nixos/lib/make-iso9660-image.nix" for guixsd ? <Ulrar>I use it to build an iso from a config file, to build a stateless webserver running nixos without a hard drive <rekado>zybell_: I don’t understand your point. Also: GHC does *not* have an alternative bootstrap path because there are no Haskell systems capabale of building GHC other than GHC. <Ulrar>works pretty well, just wondering if that would be doable with guix <rekado>Ulrar: we have “guix system” which can build a disk image, among other things. <Ulrar>"When using disk-image, a raw disk image is produced; it can be copied as is to a USB stick, for instance." <Ulrar>But would that boot in read-only ? <Ulrar>I was thinking iso but I guess anything would work as long as it's RO and doesn't have to be duplicated once per VM <Ulrar>I mean, the goal is to spawn VMs in the end, just want to avoid any kind of shared storage <zybell_>I would have to look it up, but on bootstrappable is mentioned that you can build an older ver of GHC with another Haskell system (dont recall interp or compiler). That GHC is capable of building even modern GHCs. <rekado>if you read till the end you’ll see that this is a failed bootstrap attempt. <rekado>I have since tried two more times to bootstrap GHC, both unsuccessful. <pkill9>can you use `guix system disk-image` on foreign distro (non-GuixSD)? <pkill9>wait, does that command create a livecd-type image? <wigust>pkill9: Yes, it's mentioned in the documentation. Also I think you could use anything except ‘guix system reconfigure’ on foreign distributions. <Ulrar>what do you call "livecd" type image ? as I understand the doc, it just makes a raw disk image <Ulrar>a livecd is supposed to live in ram, not on a hard drive <civodul>Ulrar: the installation ISO is a "live CD", but its main application is installing GuixSD <Ulrar>Yes, any change you do to the system on that iso won't be kept after a reboot (besides the installation, of course) <Ulrar>Is that the same for a raw disk you created using guix system disk-image ? <Ulrar>Reading the doc, I'd assume guix system disk-image just created a guixsd installation, not a "live cd" type image <civodul>"guix system disk-image" creates a disk image of the config you pass as an argument <pkill9>just an OS that you can run off removeable storage <Ulrar>Yes, so it's a regular install, persistent if you will <pkill9>ah, as opposed to living only in RAM and never changing on disk? <Ulrar>With nixos, using nixpkgs/nixos/lib/make-iso9660-image.nix, I create an iso with apache, php and some scripts and configurations <Ulrar>A "livecd" with everything on it, ready to run <Ulrar>But as far as I understand it, with guix system disk-image, the result wouldn't be the same (without considering the format iso vs raw) <civodul>Ulrar: i think we answered your question: "guix system disk-image" in itself is "neutral", it's the config file you pass as an argument that determines what goes into the image, live CD or not <Ulrar>Oh, so the "no disk" part would go in the config file <Ulrar>I never used guixsd, I just assumed the process would be similar to nixos but maybe not <pkill9>so you can test your setup on memory stick before installing it to hard drive <Ulrar>The idea for us is to have a git repo with the config, and X virtual machines running that config on X different servers, with an easy way to spawn more as needed <Ulrar>and on git push, having a ci process run to make the iso from that config, boot it, test it and if everything runs fine, reboot the virtual machines on the new version <thomassgn>hmm, I've got a atheros wifi usb card running, but it keeps losing connection to the wifi. Not sure it's the card. Is there a log for dhcp or something? (dmesg gives no useful output) <civodul>Ulrar: you can do that using "guix system vm" or through the (gnu system vm) API if you need more control <civodul>thomassgn: does /var/log/messages have hints? <bavier`>thomassgn: that's been happening for me lately too <thomassgn>huh, it seems there is some info in dmesg after all. I must be blind. <thomassgn>I've always wondered what the difference is between dmesg and messages... (: <thomassgn>didn't see it yesterday on a different network though... <bavier`>another recent oddity is icecat tabs crashing <thomassgn>Right, so not immediately connected to that then. <thomassgn>I'm getting deauth for some reason... Apr 3 17:29:41 localhost vmunix: [22916.234075] wlan0: deauthenticated from 24:a4:3c:04:68:17 (Reason: 6=CLASS2_FRAME_FROM_NONAUTH_STA) weird, but only seen it once so far, let the troubleshooting commence! <thomassgn>haha, and qutebrowser segfaults! this is my day :) <thomassgn>looks like qutebrowser segfaults cause the network goes down... weird, or - strace says lots about connman and my network devices right before the segfault... <mbakke>thomassgn , bavier`: Can you try the 4.14 kernel in case it's a regression in 4.15? <thomassgn>yes, currently building some newer kernel here (as part of a system reconfigure/update) (currently on 4.15.3 <thomassgn>mbakke: how do I specify kernel version? (kernel linux-libre@4.14.32) does not work(tm)... :3 <alezost>thomassgn: if it is a variable name, I think it should be linux-libre-4.14 <thomassgn>this works: (kernel linux-libre-4.14), ofcourse it's a variable name, not @ form. <lo_mlatu>Hi all, I'm trying to load extra modules at startup using the `initrd` field of operating-system, following the example in the "Initial RAM Disk" section in the manual. I was able to load the kernel `intel_lpss_pci` using `modprobe`, but when I try to reconfigure the system, it said the module was not found <sohomb83>when i am using "mutt -H file.patch" to send a patch. How do i add a message as the email body without modifying the patch file ? <buenouanq>ERROR: In procedure dynamic-link: file: "libpq", message: "file not found" <buenouanq>does this look like a guix thing or something in guile-squee I need to change? <zybell_>buenouanq If it is a guix-thing the name libpq is probably wrong. If its guile-sqee its probably a missing pkg. <buenouanq>I installed libpqxx and that didn't help/work. <zybell_>pls note the diff btwn libpq and libpq.so and libpq20.so and libpq.so.2.0 <buenouanq>I haven't really started poking at this, a trial just spat out that error so I came here. <zybell_>guix is a very privacy respecting dist. So there is no tracking of the sort how users set up their machines. That would we need to setup a table: user made this error=>error message the user sees=>solution to correct error. You seem to expect such a table. Then you should ask yourself "Is guix right for me?". <guiks>hello, can someone show me an example of configuration for encrpyted installation of GuixSD? <sneek>Welcome back guiks, you have 1 message. <sneek>guiks, rekado says: For Chinese input I use ibus with ibus-libpinyin. <guiks>I just noticed that root users path changed and now I can't use guix as root <guiks>or some other programs like ls <grillon>I was not at home so not able to try you advices, I it all and it does not work but I have one more information <mbakke>grillon: That looks like your ibus package is using a different C library than the one provided by GuixSD. That can happen if you `guix pull` as your user, but the system configuration is using an older libc. <mbakke>What do you have in "/run/current-system/locale/"? <grillon>I did guix pull as user by mystake then guix system reconfigure it does not work so I reboot on old version and did all as root. then I had to do some gc do you think my system is broken? <grillon>inside it I got many locales in utf8 <mbakke>The glibc in Guix since ~February is 2.26. So `guix pull && guix system reconfigure` as root should fix the problem. <grillon>I'm doing it but it's compiling (55%) <grillon>I did it already 3 times :( when I do a reconfigure something goes wrong and I have to go back <mbakke>grillon: Currently, guix pull requires compiling all the things every time, but there is work underway to make it a little smarter. <mbakke>What is the error you get during reconfigure? <grillon>do not remember but reconfigure should be fast and pull is at 83% <mbakke>Does anyone know why GCC installs C++ headers in the default output, but all other headers in "lib"? <civodul>mbakke: good question, i think libstdc++ just uses $(includedir), whereas the others use $(libdir)/whatever <mbakke>I see. I'll try moving it at some point for science. I found that C++ support in clang was broken, currently trying to add GCC:out on the default include path. <civodul>there's a --with-gxx-include-dir=DIR configure flag in libstdc++ <civodul>clang is typically used alongside libc++ (not libstdc++) <civodul>so perhaps that's what we need to fix <mbakke>Actually it seems most GNU/Linux distros use libstdc++. <civodul>actually you can also just add 'libstdc++' as an input <civodul>i did that for work: you pass clang and libstdc++, and it works <mbakke>Ah, that might be better indeed. <mbakke>civodul: Adding libstdc++ as an input was indeed sufficient, thanks! <mbakke>Side note: I wonder if we should add "libstdc++" variables for each supported GCC so consumers don't need to use 'make-libstdc++'. <grillon>mbakke: don't know it's still working :) <mbakke>Currently building Chromium with Clang 6, since upstream no longer supports GCC... Let's see how this goes. <grillon>I'm going to sleep sorry, thank you for your help and see you tomorrow. computer is still working for reconfigure. It was not working so long last time so I hope it's a good thing. <mbakke>Apparently it's possible to build Chromium 65 with GCC 6/7 (NixOS and Debian does), but Arch and Gentoo just switched to Clang with this release.. GCC is no longer officially supported (and thus not tested in their CI). <mbakke>Unfortunately GCC 6 and later has other problems in Guix. <mbakke>grillon: Okay, let us know how it goes! Feel free to reach out to help-guix@gnu.org if you don't get any response here :-) <grillon>ok thank you, I let you know :) see you 