<rekado>on wip-java-bootstrap-simplify I upgraded icedtea-8, and now I’m no longer sure it was a good idea. <rekado>a couple of tests are now failing in Java packages far removed from icedtea, and it makes me worry that I botched the upgrade in a way that’s hard to track. <samplet>(It’s exactly what you did with Disarchive from before, just for Guile-LZMA.) <civodul>i guess we have to propagate whatever Disarchive needs <samplet>civodul: Thanks! I will push it soon. There’s still an unrelated issue.... <vldn[m]>is it possible to change the path of the /gnu/store ? like in termux on android? <jlicht>vldn[m]: you can/could, but this also means you can no longer make use of substitutes from the official substitute servers <jlicht>which probably means that you will be compiling till the end of the year ;-) <vldn[m]>then maybe some mount proot magic? :/ *lfam pushed the fix for #52051 <rovanion>Guix System seems to set PYTHONPATH to ~/.guix-profile.../site-packages which is messing with how virtualenvs works. I thought running guix remove python would solve this, but to no avail. There is no python in my path, yet PYTHONPATH is set! <civodul>rovanion: on current master, Guix sets GUIX_PYTHONPATH rather than PYTHONPATH to avoid interfering with virtualenv & co. <KE0VVT>telegram-desktop always conflicts <KE0VVT>Agh! Telegram lost my settings after an update too! <KE0VVT>Is the fix for bug 52051 pushed to master? <KarlJoad>So I installed Guix System yesterday, using grub-efi-bootloader as the bootloader. But now no disks show up for booting from my boot selection menu. I used the installer to generate an iniital config file and mount the disks, but then switched to TTY for the `guix system init` portion. Anyone experience anything similar? <asdf->Building 'pbr-guix-1.3.0-15" fails... Command failed with exit code 1... a search doesn't come up with anyone else reporting this issue <asdf->any suggestions on how to diagnose this? <asdf->i booted the guix livecd off a usb... downloaded the guix-install shell installer script... it says the only requirements for the script are tar and xz but the script requires GPG <apteryx>that bit comes before mentionning the script; it is about the manual binary installation <asdf->yes, i just noticed that it was not a continuation.... thanks <apteryx>the script checks for its dependencies itself <xelxebar>Uhhh... `guix package -m my-manifest.scm` ends up error out for me: guix package: error: opening file `/gnu/store/09v5jx2dmrm62l8y7xvdhlx9cr0sj386-ld-wrapper-0.drv': No such file or directory <xelxebar>Last pulled less than 24hrs ago. On commit 9ebc9ca070a21ba0aeb3f61202a8071bb250ccc8 <apteryx>KarlJoad: hmm, someone had a similar struggle recently; was this with the 1.3.0 image? <KarlJoad>Does the kernel-loadable-modules in the operating-system record automatically load those specified modules? The documentation does not make that very clear. <KarlJoad>Is there a way to specify that all users should use the same Guix channel? I know `sudo guix pull` and `guix pull` can cause differences with system reconfigure. <xelxebar>KarlJoad: Yes, that directive should actually load the modules in memory, not just place them in the filesystem. <KarlJoad>Ok, cool. Just need to figure out why the modules already present and loaded on my system are not working in the `(list ...)` I have... <xelxebar>You can check gnu/tests/linux-modules.scm, which has some tests to this effect. Nice catch in the docs, though. Might be a decent patch to clarify the wording. <KarlJoad>Perhaps when I get my system working (able to shut down), I will submit a patch for that... <apteryx>rekado_: hm, tbb failed with: ../../src/test/harness_barrier.h:117, assertion n_seconds >= 0: Time is out while waiting on a barrier; retrying; perhaps nondeterministic (load sensitive?) <KarlJoad>I think the Cookbook documentation might be out of date... I cannot use the %linux-libre-hash value. In fact, I cannot find any hash being defined for use. <KarlJoad>apteryx: I just saw your older message about install versions. I don't know what happened, but I re-init-ed the system using the boot media and a guix pull today. <lfam>KarlJoad: The cookbook should have been adjusted to account for 1ad9c105c208caa9059924cbfbe4759c8101f6c9 <KarlJoad>lfam: Was that commit made recently? I don't know how to check for that off the top of my head. <KarlJoad>The cookbook online must be out of date then, because it still uses %linux-libre-hash, as does my built-info info pages. <lfam>The information is out of date, surely <KarlJoad>Then I am a bit stumped on how to build my own kernel then. <lfam>KarlJoad: It's described a few paragraphs after that outdated code snippet <lfam>But, we should adjust the other parts to stay current <KarlJoad>So I must use a config file that I define/generate myself? That is a little annoying... I would like it to be done with Scheme... <lfam>No, you can do it in Scheme <lfam>Do you see the example using #:extra-options ? <lfam>And the format is shown in the example for %default-extra-linux-options <KarlJoad>lfam: I see the example using #:extra-options with the "macbook41-full-config" defined, but that is for the example that uses %linux-libre-hash <lfam>So, your goal is to use a kernel with some custom config options? And you want to do the config in Scheme? <lfam>We are getting bogged down in the details of this example <KarlJoad>I want to add all the IPMI modules (which are available in linux-libre) because they are not available as loadable modules in Guix, and I need one of them to shut my computer down. <KarlJoad>I already identified the relevant config flags to set, I just need a way to pass them to the linux-libre kernel so that i can use them in my operating-system record. <KarlJoad>I want to keep all other configurations for building, but just change the %default-extra-linux-options. <KarlJoad>Thanks for finding that. Implementing that now. <KarlJoad>That seemed to do it. I am now building the new generation. I'll update you if I actually build the kernel. <lfam>I'll prepare an update for that blog post / cookbook chapter <KarlJoad>Downloading the linux-libre source, so that's a good start. <KarlJoad>Awesome. Glad we found that out-of-date part. <the_tubular>Is there a torrent of guix that I could help seeding ? <KarlJoad>lfam: The build failed because guix failed to build in the check phasebuilder step. I am going to bed now and will try to figure this out tomorrow. <lfam>Okay. Later send more details *nckxmas wondered that too. <nckxmas>You can't lie in a domain name. It's illegal. <apteryx>rekado_: hmm, no it passes (tbb@2020)... but on my branch I had to link with ld.gold... just that 2020 release... wut <robin>don't remember if i mentioned the culprit, but the 'expected integer from server' guix upgrade error was due to texlive. fixed by building locally, haven't managed to reproduce it but i have some strace logs to pore over <robin>which is of course a great opportunity to learn a bit about the relevant protocol ;) <robin>in the manual, might it be useful to have a program index (@pindex) for the various subcommands? i mix up 'm' and 'i' frequently which might be PEBCAK :) <phf-1>Hello Guix! I'm using Guix to build relocatable binaries. The workflows looks like: 1) edit sources 2) edit package.scm from a private local channel 3) guix pull 4) guix pack -RR ... <phf-1>Problem is: it's slow, in the 10min ballpark. <phf-1>anything else is really good, but is there anything to be done to make is faster? <rekado_>phf-1: for something like that I would use GUIX_PACKAGE_PATH <rekado_>this lets you bypass the “guix pull” step, which is the most expensive in your workflow <rekado_>you set GUIX_PACKAGE_PATH to the root directory containing your modules and then use “guix pack” as before. <rekado_>but for quick turnaround I’d always use GUIX_PACKAGE_PATH to temporarily extend the modules Guix sees. <rekado_>apteryx: I don’t know much about tbb@2020 — only that it’s the last version of tbb that looks like tbb. <rekado_>the new versions have completely different headers, so most of our packages don’t work with them. <zamfofex>Another option is the use ‘-L.’, which I think does the same thing as (or a similar thing to) ‘GUIX_PACKAGE_PATH’. <pukkamustard>Hello Guix! I'm a bit confused about where exactly substitutes are downloaded when doing `guix build` for example. There are some functions in `guix/scripts/subtitute.scm` but I don't see how they are used - they seem to be only for testing purposes. Where is the code that the guix daemon uses to fetch substitutes? <jpoiret>pukkamustard: The daemon code is in nix/ <jpoiret>it hasn't (yet) been rewritten in guile <mothacehe>pukkamustard: more specifically have a look to build.cc line 3032 <mothacehe>that's where the "guix substitute" process is spawned <pukkamustard>jpoiret: mothacehe: thanks! that makes sense. So `guix/scripts/substitute.scm` is not only a testing script it is spawned by the daemon to fetch subsitutes. <mothacehe>it used to be spawned for each substitute request <pukkamustard>i see. there is only one instance of `guix substitute` running at a time and the daemon interacts with it by sending it text commands. agent is a C++ concept? ***heisig_ is now known as heisig
<vldn[m]>any example on wrapping a qt program with qt-build-system? <vldn[m]>alright so far so good, but QQmlApplicationEngine is missing <singpolyma>I am trying to get guix foreign going on an armhf board. I'm running guix pull and it did the git operations and the computing guix derivation spinner, then newline and now nothing. top shows no activity and strace on the guix process shows it is waiting for a read. Any guesses what it thinks it's doing? Should I ^C and try again? <ss2>singpolyma: just try it again. <singpolyma>Exactly the same result. Spinner finished now staring at me strace says read(13 <mroh>vldn: not in a guix build container. For QQmlApplicationEngine, try adding qtdeclarative as an input. <vldn[m]>yeah i have the problem that some git submodules are not fetches with git-fetch <singpolyma>Is it possible my kernel is too old? Does guix require anything specific from kernel for normal use? <vldn[m]>aah the (file-name) option make it break.. <vldn[m]>after commenting it out it fetching all submodules <rekado_>singpolyma: no. Computing the Guix derivation is expensive. I don’t know how long it would take on an armhf board. <singpolyma>It's right after that it gets stuck. No CPU usage <nixo>Hi guix! Is pygame currently broken on master? <sneek>nixo, notmaximed says: #{\x2019} is the symbol consisting of the character ’ (a curly quote) <sneek>nixo, notmaximed says: Probably missing "quotes" around a string, or using the wrong type of quotes? <sneek>nixo, notmaximed says: (symbol->string '#{\x2019;}#) ---> ’ <nixo>mothacehe: thanks, I'm getting ModuleNotFoundError: No module named 'pygame.base' <nixo>but I just realized it's using the path of python 3.8, so there's a problem somewhere in my setup <vldn[m]>mhh alright my package succeeded building and compiling, but afterwards the package binary isn't found in the output path :( <vldn[m]>maybe i have to add some function to copy it to output after the install phase? <nckxmas>florhizome[m]: Your understanding is correct. wrap-program produces a static wrapper that is ‘frozen’ at build time; search-paths are composed much later at profile generation time depending on which *other* packages that profile contains. <florhizome[m]>And also, by default (separator is „:“) search paths should just append? <nckxmas>What you need depends on what you're trying to do. They are not equivalent. <nckxmas>I mean they are not different ways of doing similar things. <nckxmas>I'm not certain enough to say off the top of my head what the default is, but you seem to be looking at the code? <nckxmas>sneek later ask phf-1: Are you using any third-party channels? <nckxmas>florhizome[m]: Yep, that's definitely correct. <nckxmas>(So it's not technically a path at all, just a single file name, but that's a niggle.) <nckxmas>IMO. Maybe you have a valid use case. <nckxmas>They just tend to be used for different things is all. <nckxmas>Be sure not to subvert user expectation is all. <nckxmas>florhizome[m]: I'm not sure what that last message means. <nckxmas>Let's discuss specifics if it happens or if you think you need it. <nckxmas>You can choose whether a wrapper appends or overrides. <nckxmas>You realise I have no idea which variables you're setting and what problem you're debugging. 😛 <ss2>hope I'm not to quick again reporting problems. Just trying out another checkout. <nckxmas>Using third-party channels by any chance? <florhizome[m]>I want to install wayfire in the system profile and plugins somewhere else. <florhizome[m]>there is ...PLUGIN_PATH and ...PLUGIN_XML_PATH that denote the paths of shared objects and xml metadata (for settings, f.e.). <florhizome[m]>i put a wrapper around wayfire bc it needs to load all the plugins right when they are started. After all, normally one would do just that, create a wrapper–script. but maybe only search paths would be enough. and I don’t have unforeseen overrides. <florhizome[m]>In any way I should use the suffix/prefix notation for the wrapper I understand. <nckxmas>Right, I don't see a use case for wrap-program/wrap-script so far. <florhizome[m]>It’s maybe different then other plugin systems, wayfire needs to know the paths of the plugins, the plugins don’t have agency by themselves. <nckxmas>You just need to install both in the same profile. <vldn[m]>how to put the binarys after building into the output path? >package-defintion < <nckxmas>vldn[m]: No make install or equivalent target? <nckxmas>florhizome[m]: Having two unrelated profiles ‘find each other’ isn't supported within Guix. If the user wants that, the user can set ...PLUGIN_PATH and ...PLUGIN_XML_PATH themselves, but that still shouldn't require a wrapper? <nckxmas>Some packages hard-code ~/.guix-profile in a wrapper, but that's a pretty nasty hack, and will only work with the default profile. Maybe that's enough for you? <vldn[m]>but no binary is given after successful building <ss2>nckxmas: thx for the hint. <nckxmas>vldn[m]: Why are you replacing 'install with ‘make release’? <AIM[m]>How does use-package-modules work? <vldn[m]>because the standart install was not working, and it was suggested by the git readme of moonlight-qt <nckxmas>Yes, but it doesn't sound like a substitute for installation, it sounds like 'build. <vldn[m]>i changed the order already to configure and build and let qt-build-system install do it thing but i got a error, now it build successfull :D <vldn[m]>but i will try again and look what was the error <nckxmas>Yes. I made a similar change to see what ‘make install’ gives as error but it's still building. <nckxmas>The Makefile's install target is too complex to follow by hand. <vldn[m]>and maybe i detected an error: if file-name is activated (not needed here) git-fetch recursive #t does not work <nckxmas>My money is currently on ‘tries to install to / and needs to be lovingly taught not to’. <vldn[m]>%exception #<&invoke-error program: "make" arguments: ("install") exit-status: 2 term-signal: #f stop-signal: #f> <unmatched-paren>only pasting in tree-sitter itself worked with nvim, but emacs-tree-sitter needs external parsers, it has no :TSInstall afaik <florhizome[m]><nckxmas> "florhizome: Having two unrelated..." <- But I thought search paths append? And I always have at least two profiles running, right? how do they correspond about setting env vars? <nckxmas>vldn[m]: <if file-name> I doubt that. Probably made a typo. <AIM[m]>How do I know which modules to import for installing packages? <AIM[m]>I don't get what you put in use-package-modules <nckxmas>vldn[m]: The %exception line is just ‘something went wrong, bye’, not useful on its own. The error is (what do I win??): mkdir: cannot create directory ‘/usr’: Permission denied <vldn[m]>@AIM something to automate that would be really cool.. <AIM[m]>florhizome[m]: No for installing packages through config.scm <nckxmas>So you'll have to tell this complicated build system that /usr is not a place into which we want to see it installing things. <vldn[m]>@nckxmas, possible :D will try later the days <florhizome[m]>AIM[m]: Well those that your packages are defined in :) but probably some more for syntax you use in the scheme file <nckxmas>florhizome[m]: I'm not sure if search paths append *across profiles*. You'll have to test that assumption. <AIM[m]>florhizome[m]: Yeah Idk where my packages are defined in... <nckxmas>vldn[m]: Killer plan :) Obviously you just need to run guix build with sudo. <florhizome[m]>nckx:that would be kinda catastrophic if they would just override env vars on a different host? <nckxmas>AIM[m]: You can also use (map specification->package (list "stringy" "package" "names")) which will automatically import the right modules for you. <nckxmas>With Guile modules, (foo bar biz) == foo/bar/biz.scm <vldn[m]>even in package definitions @ nckxmas? <nckxmas>use-package-modules implies (gnu packages), so you just need the last part. <vldn[m]>:D already enjoying this scheme magic in my config.scm <vldn[m]>should be default if working good.. :D <florhizome[m]>i could maybe append the ..XML Path to XDG_DATA_DIRS bc it’s under /share but it seems kinda hacky to me <AIM[m]>What about software from other channels? Do use-package-module work only for guix official channel? <nckxmas>It's just a convenience function for (gnu packages). <vldn[m]>nice @nckxmas will try that for development next time <vldn[m]>do you use some emacs magic to automatically include the defs at the top? <AIM[m]>(use modules (channelname packages superhackedstuff)) <nckxmas>vldn[m]: You just need to import (gnu) to use it. <nckxmas>There's a suggestion. Not that I think a package already exists. <vldn[m]>i know i know, but it's not the right way for package submission in guix i think.. <nckxmas>AIM[m]: It depends on the channel! The channel could put all packages in a mycoolchan/package/foo.scm file and you'd import (mycoolchan package foo), but that would be a pretty trolly thing to do. <vldn[m]>or i just don't see a package with this little scheme magic already <vldn[m]>so i thought there is a way to simply this importing and handwriting stuff <nckxmas>AIM[m]: Trolly because everyone will write ‘packages’ out of habit. *nckxmas → nckxmas preparations. o/ <vldn[m]>so maybe it would be a good way to make a flag for autosearching , aka including this scheme magic, and enjoy some fewer typing while development <florhizome[m]>so the weird thing is, I commented the wrapper out, and now one of the env vars appends and the other one doesn’t <Kabouik>Hey #guix. I'm not a programmer and don't know much about lisp except all people I know who understand it love it. I use Nyxt (a c-lisp based browser) though, and am considering StumWM to sink some time during my holidays. <Kabouik>Now that I'm there, I'm thinking why not be bold and try Guix altogether. <roptat>unless you want to create packages, you don't need to know the language to use guix ;) <Kabouik>Is there a way to make a live-USB of Guix without a priori install? I read that `guix system disk-image` can make a Live USB of an existing installation, but can I do the opposite, i.e., install Guix on a thumb drive to test it, play a bit with it without installing on my computer, configure a WM, play with the package manager, and when I'm satisfied, used the modivied Live USB to install on a SSD? <roptat>Kabouik, you can install guix, the package manager, on your current distro, and use that to install the system on the usb <vivien>Kabouik, you should try guix on top of your current distribution. <roptat>(we make a distinction between Guix, that you can install on top of any linux distro, and Guix System which is a distro) <Kabouik>I did already but I wanted to try the distribution this time; I was not a big fan of mixing the guix package manager with my distribution since they really work in different ways file-hierarchy wise <Kabouik>Guix System is what I wanted to try here <vivien>You can also test guix system in a virtual machine <Kabouik>But I'm running short on SSD space here on my laptop, so trying from a Live USB and being able to save my changes, and then possibly install the customized Live USB on a real computer, would be awesome <roptat>I don't remember how you do that, but sounds like you're just one command away <Kabouik>It's `guix system disk-image` that made me think of that, as it kinda does the opposite: clone an existing Guix System installation into a Live USB to replicate it to new computers. <roptat>guix system image -t iso9660 config.scm <roptat>that will create an ISO from the config, and you can burn that to a USB <roptat>(you need to come up with a system config) <nckxmas>…aand we're back to ‘shouting at computers’ (my hobby according to $partner :) because my advanced home media subsystem won't boot because /etc/static/dbus-1 is missing and this is worth an unhandled exception. Is this familiar to anyone? <vldn[m]>or you could make a guix usb stick with your guix install on your foreign distro <Kabouik>That sounds a bit similar to `guix system disk-image` except disk-image clones the full disk roptat, correct? <roptat>guix system disk-image won't clone anything I think. It's the old name for guix system image <Kabouik>The last time I tried to install Guix (package manager) on my distribution, the thing that annoyed me is there was no uninstall script, and Guix did scatter files and dirs a bit everywhere so it made me feel uneasy for something that was just a test at the time <vldn[m]>is there something like ublock origin for nyxt @ Kabouik? <KarlJoad>Is there a way to make all users on the system default to the guix channel of root? That way guix system reconfigure always uses the root user's definition. <Kabouik>Ok roptat, so indeed those two commands should ultimately do similar things. <Kabouik>vldn[m] no, not yet, but (1) there are blocker-mode and no-script-mode builtin Nyxt which kind of cover the same usage as uBlock, except they're not as efficient eyet, and (2) one of the developers is actively working on supporting web extensions, uBlock being one of the main reasons since it's often requested by users <AIM[m]><vivien> "You can also test guix system in..." <- Or do like me... I just leaped directly in Guix.... Without any considerations... <vldn[m]>ublock origin is so damn nice if you watching videos or streams often <vldn[m]>just blocks the whole youtube and twitch ads <AIM[m]>vldn[m]: Use Invidious or Freetube <vldn[m]>without fregmented ads or popups with erros <vivien>I use mpv to watch youtube or twitch <Kabouik>So roptat, creating an ISO image of my Guix config will be helpful. But can I make a *writable* Live USB of Guix to fiddle with that config in the first place without installing Guix on a computer? <AIM[m]>It gets video data directly from yt without any ads <roptat>Kabouik, ah maybe you want to install the guix system on that USB? like with "guix system init"? <Kabouik>Or, in other words, install Guix to a USB thumb, make it bootable and writable, so that later on I can export my config from it and even install from it when I'm satisfied? <Kabouik>vldn[m] to be honest I use ytfzf all the time, kitty + mpv for the win, no moar bloat (and ytfzf can even manage subscriptions without a Google account) <roptat>Kabouik, you should read about the manual installation steps in the manual <vldn[m]>that's what i thought of @roptat, but it is a problem for the bootloader? <vldn[m]>or could someone just create /mnt/boot and mount the standart bootloader on that usb <roptat>I think you can install the bootloader on the USB <roptat>and then select the USB as the mount point from the bios <vldn[m]>yeah me too, but sometime when you are searching it is anoying to get adds, ublock just gets rid of it all :) <vldn[m]>and browser plugin support are a huge plus for crypto browser addons @ kabouik :) <Kabouik>But yeah, I can't wait to get WebExtensions support in Nyxt; don't know when it'll be though, but it's planned and under progress <vldn[m]>let's hope they are getting it to work :) qutebrowser claims to support plugins since 3 years.. :D <Kabouik>Don't hesitate to come and ask questions to the devs directly on #nyxt vldn[m], they're very responsive (although they might be off during holidays now) <Kabouik>I think there's even basic support for WebExtensions already, but very basic and vastly undocumented. aartaka could tell. <aartaka>Kabouik: I haven't expected you pinging me here xD Yes, basics of WebExtensions support are there, but it's of terrible alpha quality :) <AIM[m]>Hey, how do I do guix show for services? <nckxmas>There's no ‘show’ equivalent but there's ‘guix system search’. <AIM[m]>Btw, where does gdm come from? I don't want display manager since I'm using a single window manager *nckxmas AFK .oO ‘…maybe if I shout louder…’ <KarlJoad>I am trying to build a new system generation, but the initial ramdisk is missing the "ahci" module. It throws an error from the default-initrd-modules in find-module-file function. <KarlJoad>I have added additional config options to build extra modules into the kernel, but the initrd is the one failing. Why might that be? <AIM[m]>Anyway to remove gdm and use startx instead? <nckxmas>KarlJoad: initrd-modules defaults to a simple list, it is not intelligent, and the initrd builder will happily choke if one of the modules in that list does not exist. I'm brooding on a more general solution because of bug 52667, but until then (if that even happens :) you can manually delete modules your custom kernel lacks: https://issues.guix.gnu.org/52667#2 <nckxmas>You can use (fold delete …) if it's more than one. <nckxmas>So rolling back didn't work either, so I tried deleting /etc/static/dbus-1 from the initrd REPL, rebooted, Guix happily recreates it brokenly and the bournish ls lacks an -l option. This is great. <KarlJoad>nckxmas: Thing is, I only _added_ to the list of kernel config options. I prepend my new options to the %default-extra-linux-options list, so I should not be changing anything with the ramdisk. <nckxmas>I thought you said you built in extra modules. <nckxmas>Could that not lead AHCI to be built in as well? <KarlJoad>I added extra CONFIG flags to the kernel. ahci is _already_ defined as a default initrd module, so I believe I should not be removing any functionality, only adding. <nckxmas>We're going in circles. What does your new .config look like? ***Xenguy_ is now known as Xenguy
<KarlJoad>nckxmas: (append `(("CONFIG_IPMI_POWEROFF" . #t)) (@@ (gnu packages linux) %default-extra-linux-options)) is the basic format. This is passed to make-linux-libre* as #:extra-options <nckxmas>Right. I meant the final .config. It's saved in the build output as .config, e.g. /gnu/store/my-custom-kernel/.config <nckxmas>You'd expect CONFIG_SATA_AHCI=m for the initrd builder to find the "ahci" module. <nckxmas>Barring a bug in the builder, I expect yours to say CONFIG_SATA_AHCI=y for some as yet undebugged reason. <nckxmas>Like maybe your custom build is reverting to defconfig for some reason, can't say. <KarlJoad>nckxmas: CONFIG_SATA_AHCI=y is present in the .config file. <nckxmas>I'm going to check whether your IPMI option implies that, which I very much doubt, so I suspect that your custom kernel isn't what you think it is. <nckxmas>However, I don't use linux-libre or know how make-linux-libre* works. <Kabouik>roptat just to make sure I understand correctly, the GUI installer of Guix System is just an installer, not a live version of the stribution? <Kabouik>Trying to check if the only way to get Guix on a thumb drive is to use the manual installation (or maybe the GUI installer, but on a second thumb drive, and select the other thumb as install destination?) <nckxmas>‘(configuration-file #f)’ yeah you're starting from upstream defconfig, then applying the relatively puny %default-extra-linux-options to it. This will result in a kernel vastly different than what Guix expects. <nckxmas>I see all make-linux-libre* calls in linux.scm have an explicit #:configuration-file kernel-config. <nckxmas>I'm not sure how this magically DTRT, but obviously does. <nckxmas>You want to start with gnu/packages/aux-files/linux-libre/5.15-x86_64.conf, not defconfig. <AIM[m]>How do I add new entry to gdm for xmonad? <nckxmas>KarlJoad: Try, I guess, (configuration-file (@@ (gnu packages linux) kernel-config)) <KarlJoad>Thanks for pointing that configuration-file out. Didn't even notice that one. Trying that now. <nckxmas>AIM[m]: Sorry you've been rather ignored lately :) Using startx with Guix System is far from trivial, at least it used to be. I don't know if much has changed. Try searching for recent search engine hits for guix + startx etc. <nckxmas>Most people who don't want GDM use SDDM or another DM, not startx. <KarlJoad>nckxmas: Close, but not quite yet. Same issue, but I compiled a lot more of the kernel this time around. <AIM[m]>I just add it as a package and do guix reconfigure the system? <florhizome[m]><AIM[m]> "Anyway to remove gdm and use..." <- You can remove gdm, not sure where that lands you. <florhizome[m]>But there is a YouTube video about s.o. installing guix with sway without dem ,and they posted their config somewhere <KarlJoad>nckxmas: Actually, I removed a line I added to my personally-specified options, and everything compiled correctly. <Gooberpatrol66>Is there a simple way to add auto-start to a service that doesn't have it? <lilyp>Not really, the hard thing is to make it meaningful :) <nckxmas>KarlJoad-: Was it CONFIG_SATA_AHCI=y 😛 <KE0VVT>nckxmas: Has the installer been updated to fix the HDD login issue? <nckxmas>I look like the VLC cone with a jaunty hat. <attila_lendvai>i was told it's better to sudo -E guix system reconfigure (as opposed to do it as root), but i keep sumbling into inwritable files created by root... <tissevert>I'm so happy I finally got the time to put guix on my personal laptop <tissevert>this distro is so great, it's such a delight to use <nckxmas>You should be able to remove (if caches) or chown (otherwise) the incorrectly-owned files and continue. <attila_lendvai>nckxmas, the last one just now was ~/.cache/guix/authentication/channels/guix <nckxmas>Sounds like you pulled as root, is that possible? <nckxmas>So ‘as root’ but with user environment variables messing up the place. <attila_lendvai>nckxmas, i don't remember pulling as root. my routine is pull as my user, update packages, and then sudo -E reconf. ***lukedashjr is now known as luke-jr
<nckxmas>attila_lendvai: It would be easier if you could reproduce this directly, for example rm -rf ~/.cache/guix && guix pull && guix whatever && sudo -E guix system reconfigure … && find ~/.cache/guix -user root <nckxmas>If it's ‘something that happens after a while somehow’ it's hard to debug. <attila_lendvai>nckxmas, i'll keep an eye on why this happens. for now i was just wondering if it is expected or not. <nckxmas>/home/$USER/.cache/guix should always be owned by $USER. <nckxmas>As long as you're following recommended practice, that is, and it sounds like you are. <nckxmas>I think even -E is optional on Guix System (because it's the default) but it's good that you're being explicit here. <nckxmas>Plenty of things to fix if you're feeling itchy. ***lukedashjr is now known as luke-jr
<jgart>Hi Guix! If I build say many (>20) packages in one go, Is there currently a way to provide a summary of the ones that failed and the ones that built successfully at the end as a report? <nckxmas>You could parse (well, grep) the output with --keep-going but I don't think there's a nicer way. Interesting idea, jgart. <tissevert>just enjoying the perfection of it for a while <tissevert>(there are already a couple fix I wanna make, but not tonight) <jgart>nckxmas, thnks for the tip!. I'll try to grep/parse that first. Merry nckxmas! <nckxmas>It will only print successful builds actually performed (if you run ‘guix build a b’ and b fails and a was already in the store, you'll be able to grep for b's failure but a won't be mentioned) but maybe that's enough for your use case. <AIM[m]>How do I specify version for package list in config.scm ***Xenguy__ is now known as Xenguy
<tissevert>it will take the latest version by default (that is the version the package is defined at in guix) <tissevert>but I suppose you could make fixed versions of packages by inheriting them and overriding the version / hash fields ? <KE0VVT>I need to know if it's safe to run Guix System on my X200 again. <jackhill>KE0VVT: I'm not sure what you mean by the installer, but if this is your issue, we think we've fixed it: https://issues.guix.gnu.org/52051 If it turns out you can't login, I guess you could always boot into and older genereation <nckxmas>I don't know if it's in the Guix *package*. <AIM[m]><tissevert> "but I suppose you could make..." <- Okay I did it like softwarename-version, like ghc-8.6 <nckxmas>If you do try the installer, I'd pull before installing. <KE0VVT>nckxmas: Can I use Stable and do 'pull'? <nckxmas>I *think* so but unless latest doesn't work for you I wouldn't. <nckxmas>You'll be downloading more and possibly building more. <KE0VVT>nckxmas: It installed fine. Just couldn't log in. <KE0VVT>Oh, I guess I'll get the snapshot. <KarlJoad>Is there a list of home services that can be used with guix home? <minikN>Hello. Possibly a silly question, but I never know whether I should install a particular package in my system or my home profile. Is there a convention? What's the difference, besides system package being executable by multiple users? <KarlJoad>minikN: I think a big reason is the time to update the system when doing a guix system reconfigure. <minikN>Hm, that's a good point, never thought of that. <KarlJoad>But from my experience in NixOS, and from what I have experienced with Guix so far, there is no real convention as to where to put packages. <KE0VVT>If I give the entire ~/ to Borg, will it back up all the stuff linked in .guix-profile? :o <nckxmas>minikN: The only reason beyond user preference that I can think of is when something is somehow bound to the system, i.e. you run the postgresql service, so you add some postgresql command-line tools to your system configuration packages so it will always be ‘in sync’. <nckxmas>But it's really just user preference even then. <nckxmas>(Like, why would it even do that to begin with?) <minikN>nckxmas: Thank you. Was the 'I already answered that question.' in regards to me? <KE0VVT>nckxmas: I know you answered that question, but you said hard links, and guix-profile is symlinks. <drakonis>KarlJoad: there's no convention indeed, its typically because the goal is to ensure everything works fine, despite the way it is made available to the user <nckxmas>(That was a winky face, for those fontless.) <nckxmas>Back-up software *will* ‘follow’ hard links because there is nothing to follow, they are just regular files. <nckxmas>It takes expensive work to treat them as anything but, which is why rsync won't do it by default and can slow to a crawl when it does, but that's drifting off-topic. <nckxmas>So… my home server won't boot because the Guix boot process deliberately creates /etc/static/dbus-1 (even if I delete it) as… a dead symlink. <nckxmas>To /gnu/store/something-that-doesn't-exist-configuration <nckxmas>Damaged file system is not plausible while I did GC right before rebooting. <ss2>this just doesn't make sense now. Anyone willing to glance over this: ;;; GNU Guix --- Functional package management for GNU <lilyp>ss2: what about it doesn't make sense? <vldn[m]>@Kabouik @ aartaka - do you know of something of "i don't care about cookies" for nyxt? <ss2>and I just don't know why.. <dissent>Hey guix. I'm still quite perplexed about a situation with my audio. I'm running <dissent> GNU Guix on a Lenovo Thinkpad L580. My wm is Xmonad, although I ran <dissent>into the same issue with a fresh install with GNOME. Basically it looks <dissent>like all of my my audio output also gets channeled to the audio input. <dissent>So if I'm talking to someone they'll hear all of my audio notifications, <dissent> etc. while there is no input of from my voice. Since I ran into this <dissent>issue on a fresh install and so I'm having trouble discerning whether <dissent>alsa/pulse need to be configured or if it is a problem with my hardware. Anyways here is some outputs for a couple of different commands: https://termbin.com/1jz3 <ss2>I hadn't made a patch for inclusion in Guix yet. :) <lilyp>ss2: For one, try using gtk+ as a regular input <nckxmas>There is no ‘sh’ is what I surely meant to type 🤦 <lilyp>other than that, what error do you get? <ss2>lilyp: %exception #<&invoke-error program: "sh" arguments: ("autogen.sh") exit-status: 127 term-signal: #f stop-signal: #f> <ss2>looking at the output in /tmp, nothing is extracted to work on. <lilyp>hmm, bash-minimal ought to be an implicit native input, but try adding it to see if it changes the error <vldn[m]>and is there a simple vnc service for guix? <nckxmas>ss2: (file-name (string-append name "-" version ".tar.gz")) <nckxmas>You get a nonsense ‘xxx-download’ store item without it. <lilyp>okay, nckxmas wins the channel once again <nckxmas>I would like to exchange my Guixcoin for monies. <nckxmas>Hmkay, so I could (mkdir (readlink "/etc/static/dbus-1")) from the initrd REPL, and now boot panics on a different file (nssswitch.conf), so this is the way ‘forward’. <nckxmas>But what a tedious and error-prone way it will be. <ss2>nckxmas: where to? not your coin :), the string-append? <lilyp>did you nuke your drive or why are these files missing? <ss2>it worked before[tm], it is working now. <ss2>nckxmas: I do own you at least a gift for a beer at some point. <nckxmas>(link-xkcd-regex "spacebar heating") <nckxmas>ss2: Sweet! I hope you make it to Brussels one year, when such things are once again legal. <nckxmas>I hoped that before the beer anyway, to be clear. <nckxmas>Oh, thanks, I literally just finished typing (with-output-to-file blah "") but that might not work. <ss2>in the future I can only hope. :) <nckxmas>IWBN if the initrd would perform a graceful reboot. <nckxmas>It took me ~3 to figure out I needed to (sync) :-/ <nckxmas>‘No such file or directory: "/etc/static/sudoers"’ It's going to be a long night. <lilyp>for the record, this is also the case if you boot older entries, right? <nckxmas>You do not wish to see in which bodily positions I must position my body in order to type this stuff. ‘It's headless dear’, I lied, ‘it'll be fine’, I lied. <Kolev>What is this (mkdir) business? I'd live to shell script in Guile <lilyp>you don't want to use gash as a serious shell <nckxmas>As is always the case, this is the one time this year I cleaned up old entries before rebooting. <nckxmas>OK, but it's a step up from bournish. <lilyp>if you get it to work before tomorrow, it'll be a christmas miracle :) <Kolev>lilyp: what's wrong with gash? <nckxmas>Apart from a sane ‘mount’ helper I'm quite thankful to have Scheme composability here. Imagine having to write mkdir $(readlink foo) without $() support. <nckxmas>I can also define things. It's quite nice apart from the completely broken part. <nckxmas>Tab-completion is all that's missing, but what are a few hashes between friends. <Kabouik>vldn[m] I don't, but pretty sure it could be configured in lisp without an extension. I have no idea how hard that would be though. <lilyp>For my personal tastes, having a REPL that doesn't even close parens is preposterous. <lilyp>But yeah, readline missing at boot and all :) <nckxmas>Oh yeah, you make one typo, you dead. <lilyp>still beats writing you commands in ed <stikonas>hmm, gash hasn't seen any development for 6 months... <nckxmas>Want know a thing? I have never used ed. Like, not even tried, since I was ~11. ‘?’ ‘Nope.’ ^C. The end. <nckxmas>‘load-thunk-from-file: No such file or directory.’ Damn. Level 2 is much harder. <nckxmas>I might not be able to write a thunk by hand. <nckxmas>Ugh, and this is whilst starting *nginx* (I think)?? Why isn't the base system up yet? <nckxmas>Uhh… the REPL I'm at turns out to be the Shepherd REPL. <lilyp>isn't this the best level 2 you could hope for? 