IRC channel logs
2024-11-21.log
back to list of logs
<elevenkb>Hello there, how does one use a package like texlive-stix2-otf which packages an otf font from texlive and so doesn't seem to add it to the fontconfig cache? <anemofilia>Hm, I was going to run `guix size plasma` and somehow got 0.0MiB <anemofilia>running `ls -a (guix build plasma)` shows me the derivation output is empty <anemofilia>Ah, got it, `plasma` itself is just a metapackage <dajole>Okay, giving this another go. I'm trying to do a straight-forward install of Guix System on an old laptop. With the latest Guix image, when I follow the graphical installer, selecting the full system encryption in the guided partitioning step, and making no manual changes to the generated config, upon reboot I end up in Grub rescue with an `error: no such cryptodisk found`. I've tried multiple times with the same result. <dajole>I'm a little confounded at what could be going wrong. Is this a bug in the image / installer? Given that I had previously installed Guix on the same computer, and given the hard drive is wiped as part of the installation, it doesn't seem like the system itself could be the issue. Any ideas? <dajole>Yesterday someone mentioned the generated config and grub.cfg would be helpful for debugging. Is there a more straight-forward way of getting these from the non-booting system other than booting from usb, manually copying the files to the usb stick and then transferring them to another comupter? <ieure>dajole, Weird, I've never run into that. <wakyct>dajole, is the whole drive wiped? There's not some other older partition or boot partition? <dajole>In the installer step it told me all old partitions would be removed (if I recall correctly) <dajole>I just double-checked, and it looks like using the guided partitioning selecting a single partition should overwrite all existing ones. <dajole>Hm...perhaps I'll try doing the partitioning and encryption manually. <COMPL_EXE>Does the image only support booting with a microSD card right now? <apteryx>I think my migration from WeeChat to ERC is about complete. <apteryx>I'm very happy about the much upgraded text editing capabilities and search ;-) <apteryx>haven't figured out why my channel buffers don't get log to files yet though. <janneke>ACTION wouldn't know about logs, uses znc for logging <janneke>efraim: headsup, i'm building gcc-final right now <janneke>when i'm stuck, talking to someone really helps, so thanks for yesterday <janneke>i'll push another squash! commit to hurd-team when it builds <civodul>wait, gcc-final natively built on x86_84-gnu? <janneke>the build takes about 50 to succeed (less time to fail :) <janneke>using gcc-cross-boot0's timings, gcc-final didn't succeed just yet <efraim>gcc-bootstrap-0 built on riscv64 but failed on aarch64 <efraim>looks like guile-2 doesn't like the wrap-compiler definition inside the #:builder <janneke>efraim: oh, would it be happier inside a (let () ... )? <efraim>definition in expression context, where definitions are not allowed, in form (define (wrap-compiler compiler) ...) <efraim>at least its easy to test out on x86_64-linux <efraim>./pre-inst-env guix build --no-grafts -e '(@@ (gnu packages bootstrap) %bootstrap-gcc)' <efraim>literally wrapping it in a (let () made it pass <janneke>ah, the procedures must come first in guile-2 i guesss <janneke>right below (bindir (string-append out "/bin"))) <efraim>I bet we could also skip copying the tarball to cwd before running xz <efraim>as long as we're touching %bootstrap-gcc <janneke>efraim: nah, seems to be there for a reason <efraim>yeah, I couldn't figure out how to unxz to a different location without outputting to stdout <janneke>xz: Input file has more than one hard link, skipping <efraim>/gnu/store/qc9b01x31ayxb36r0zw5cw28awisdq98-xz: /gnu/store/ipzn49z81cvsgkg9pghcxvb8fsxk06i1-gcc-4.8.2.tar: Cannot set the file group: Operation not permitted <efraim>that got around the hardlinks, but it still tries to do it in place, not in cwd <janneke>but be my guest to get a fix in place <efraim>tar (child): xz: Cannot exec: Not a directory <efraim>can't decompress directly from the store into the store. That would've been really nice <Kabouik>Could anyone explain this error (https://paste.debian.net/1336275/) to me? I already uninstalled ffmpeg and mpv prior to running this install command, and they are not in my config.scm either. I'm not sure why there are remnants of ffmpeg-6.1.1 preventing the installation. <Rutherther>Kabouik: there are no remnants. The problem here is that mpv propagates older version ffmpeg, so you can't currently install both a standalone ffmpeg and mpv <Rutherther>Kabouik: also config.scm has no relation here, profiles are "standalone", the conflicts matter only in one single profile, guix install installs to ~/.guix-profile <janneke>you may have guessed by now, gcc-final failed again (this time the chilhurd hung itself before i could get any info out) <Kabouik>Thanks Rutherther. I checked the mpv package definition but I don't see it hardcoding a ffmpeg version, is that done in the sources? <Rutherther>Kabouik: what do you mean by its definition here? <Kabouik>`guix edit mpv`, and it just sets ffmpeg as propagated input, not ffmpeg@6.1.1 <Rutherther>yes, but ffmpeg symbol it refers to is ffmpeg 6.1.1 <civodul>janneke: i remember ext2fs crash and/or Mach crashed upon memory pressure for 32-bit childhurds <civodul>maybe just throw more memory at it? :-) <Rutherther>but since you are doing guix install ffmpeg, it looks for all packages named ffmpeg and chooses the latest one, being ffmpeg 7.0.2, that's bound to symbol ffmpeg-7 in gnu packages videos <Kabouik>It's not clear to me what is the ffmpeg symbol and how the mpv package refers to it <Rutherther>it refers to it by having ffmpeg in propagated inputs, as you saw it in there <janneke>civodul: interesting, i'm using 4GiB currently <janneke>youpi had another (ext2fs) patch for me yesterday, but i'm not using that yet <civodul>“here’s a patch for you, hope you like it” :-) <janneke>ACTION likes it, but not so much the (half) world rebuild that comes with it atm <Kabouik>Hah, I see, ffmpeg 7.0.2 is actually ffmpeg-7 in gnu/video.scm, and mpv propagates just ffmpeg, and therefore uses ffmpeg@6.1.1. Would upgrading the mpv package so that it uses ffmpeg-7 be a solution (provided it doesn't bring new issues to compile mpv, but I assume it would, else ffmpeg-7 would be called ffmpeg and be the default)> <Rutherther>Kabouik: yeah, it has a reason it's not at the newest version of ffmpeg. I think the best solution here would be to change mpv to not propagate ffmpeg, but it was probably hard to do that, hence it propagates it <Rutherther>Kabouik: as for having both mpv and ffmpeg, you can just install mpv and it will give you ffmpeg 6.1.1 as well <Kabouik>Yeah, I didn't notice it was a propagated input initially, and thought it would just build with it but not keep ffmpeg in my store. <Kabouik>(Yet I did submit a mpv update patch recently, I just forgot) <Kabouik>Thanks for the explanations Rutherther! <ekaitz>efraim: what's the status of the riscv bootstrap? <Arni>i had a question and a problem regarding guix system, as im interested in it and am trying to install it on my system : <Arni>- i hear that guix lets you create isolated environments. something I'd really like to do is to be able to make an environment of my entire system (with such things as the DE and etc), so as to have a "sketch" system where i can do whatever without actually affecting my system. imagine something like making a quick virtual machine of your system *aside* of your main system, so you dont affect this main system despite changes i <Arni>the VM. would guix let you do that? <Arni>- how would one dualboot guix? i tried to do exactly that (with Void Linux being on the system), but it turns out osprober doesn't find guix, despite there being a config file for guix in one of the folders inside /boot, and alongside void's config. am i missing anything? <Arni>odd formatting, im not used to irc chat.. sorry if its hardly readable! <janneke> total used free shared buff/cache available <janneke>Mem: 5.7Gi 3.7Gi 2.0Gi 0B 2.8Gi 2.0Gi <noe>Arni: check “guix system vm” <janneke>Mem: 5.7Gi 4.8Gi 917Mi 0B 2.8Gi 917Mi <Arni>hella, ill check it out. thank you so much! <noe>Alse check “guix system container” :) <Rutherther>Arni: you can't use other os's grub / other bootloader for guix. Guix needs special parameters to launch the kernel. These parameters are crafted by guix. So either use guix's grub for booting into guix as well as other system or use other means of dualbooting, ie. if you have uefi system, just have separate bootloaders for both and choose which one to boot with uefi boot menu <Rutherther>Arni: the parameters will be different for every generation (and that means after applying any change to your system's configuration) <Rutherther>also note for the first question you have, you can still use any other linux distro with guix. You do not need to be booted to guix system for that <gabber>Rutherther: interesting. so the obvious fix would be to add python to native-inputs? <Rutherther>gabber: it's already in inputs by being propagated from python toolchain. The problem is that it's shadowed by python-toolchain that is put in with name "python" <Rutherther>gabber: additionally, you shouldn't wrap with python that's in native-inputs as with cross compilation the python it would try to run is the one from build platform, not the one from target platform <gabber>ACTION got once again confused by the native- in native-inputs <janneke>configure: error: Specified CC_FOR_BUILD doesn't seem to work <janneke>make[2]: *** [Makefile:5849: configure-stage2-gmp] Error 1 <gabber>Rutherther: so, we could add a named input (e.g. "real-python") and refer to that specifically? <janneke>aargh: conftest.c:4:3: error: implicit declaration of function 'exit' [-Wimplicit-function-declaration] <janneke>efraim: do you know how to add "-Wno-implicit-function-declaration" to the stage-2-gmp configure? <Rutherther>gabber: yes, though I don't think it's the correct way to go. The python from #:python should be used. It's just that the underlying python inside the toolchain should be referenced, not the python itself. In case same keys are not thrown out of the alist at some point, it should still be there, so even looking for the last/second instance in the alist would work instead of going with the first one <gabber>Rutherther: i see! thanks for the pointers, i'll see if i can manage (: <janneke>oh yeah, nice! ... the (manually restarted) build continues <janneke>reading the makefile, i guess we should probably just override CFLAGS,CFLAGS_FOR_BUILD, CFLAGS_FOR_TARGET instead of the CC*/CXX* compiler variants <efraim>janneke: another option is to pre-build gmp&co <efraim>also, wrapping g++ breaks building libstdc++ for aarch64, armhf and riscv64, so it should probably be x86_64-gnu only <Arni>thank you for your answers rutherther! (irc client was weird so i had to rejoin, lmao) <Arni>im going back on the graphical installer then, and am doing the manual install. it seems to automatically choose the void linux /boot/efi and set its ESP flag to it <Arni>if i uncheck the ESP flag of this void /boot/efi, from the guix installer, and then make a separate fat32 /boot/efi for guix, would that affect void in any bad way? or am i doing exactly a second bootloader which is the end goal <efraim>I just pushed a fix, and then a fix for that fix <efraim>of course if I had spent more than a few seconds on it I probably would've gotten it right the first time, or make it so %bootstrap-gcc didn't have to be rebuilt <janneke>efraim: yeah, i can relate; did something similar late last night <janneke>i guess i'll try and move the comma one level higher, just before the if <janneke>efraim: the "scary" thing / puzzle here is that i made some changes in commencement that look at gcc version 14, instead of target-hurd64? <janneke>i guess there's no safe choice here, maybe i/we should go for target-hurd64? initially instead? <efraim>for commencement target-hurd64 is probably best <janneke>that's what i did for the bash-minimal and git-minimal cross builds, until i found that the same changes were needed for x86_64 too (and others for i686/i586) <efraim>I guess unless it would make sense that another architecture could come through with gcc-14 <janneke>efraim: ah indeed; until that happens target-hurd64 and gcc > 14 are identical <janneke>and then target-hurd64? is the better choice, thanks! <efraim>yeah, I think so. per-architecture is normally the most targeted change, so the easiest to undo if it's no longer true <Rutherther>sneek: later tell Arni: if you have just one disk, do not make separate esp, there is a good chance your system wont support it. Most systems support one esp per disk. <efraim>janneke: I pushed another fixup commit. We'll get %bootstrap-gcc correct eventually! <janneke>efraim: i'm working to get gcc-final built (manually, after an interrupded build) <janneke>once that works, i'll give you a headsup and reset hurd-team with everything squashed and cleaned-up <apteryx>when using wayland with GDM (the default), is the xorg-configuration field of gdm obsolete? <apteryx>I'm setting my keyboard layout via it, and it doesn't do anything it seems, which suggests so <apteryx>perhaps it could warn if that's the case <janneke>ACTION adds a co-authored-by to that patch <reedm>Hi all. I'm trying to set up cgit and gitolite. For cgit to access the gitolite repositories folder, the fcgiwrap user needs to be part of the git group. Is there a nice way to do this? <reedm>I tried to change the gitolite group to "fcgiwrap", which works, but I get a warning about fcgiwrap group being declared twice when I do guix system reconfigure <efraim>with the qt-team branch merged I should rebase the rust-team branch again <efraim>civodul: I got a bug report about shepherd-run, apparently I was still using the deprecated goops interface and it was removed with the 1.0.0rc1, so there are people out there testing shepherd and other users of shepherd-run! <civodul>i wanted to try and add shepherd-run-like functionality directly into shepherd <efraim>we could always add the gawk version ;P <janneke>hmm, we have --without-tests, but not --with-tests :) <civodul>there’s much less demand than for --without-tests, go figure :-) <dariqq>civodul: hi, did you see #74284 already? I think the issue is that a oneshot service is treated as having started before it actually has (via %one-shot-services-started) <dariqq>i tried to swap the order but got errors because subsequent oneshot services return the running value which is #f <civodul>dariqq: i saw it but haven’t investigated yet :-) <lispmacs[work]>does somebody web paste a /etc/profile file from a recent guix installation? I'm wondering if that file has changed since my install four or five years ago <civodul>lispmacs[work]: it’s stateless so if you reconfigure today, you’ll have the same /etc/profile as everyone else <lispmacs[work]>civodul: the file is rewritten every reconfigure? It doesn't appear to be a shortcut to anything in thestore <lispmacs[work]>something annoying me is that my /etc/profile inserts a bunch of ~/.guix-profile stuff in my environment variables, but since I switched to guix home, I don't keep the old guix profile up to date <ieure>But the timestamp on my /etc/profile corresponds to the timestamp of the current system generation, so this must be getting done on activation of the new system generation. <Rutherther>lispmacs[work]: why don't you just uninstall everything you installed to ~/.guix-profile? <lispmacs[work]>Rutherther: that sounds like a good idea, though it leaves me with some doubt still whether .guix-profile would still be having any effect on my environment, to things like GUILE_LOAD_PATH. I would rather that .guix-profile was not referenced in my environment in any way <Rutherther>lispmacs[work]: it shouldn't affect anything after you uninstall it all. There will be no more search paths to add since it won't have any paths <Rutherther>maybe even the whole folder gets deleted, I am not sure now, though <lispmacs[work]>so, in the past I was using manifest files. Can I just pass it an empty manifest? <lispmacs[work]>that didn't seem to be working, but I did a manifest with just one package and them removed that package from the CLI <lispmacs[work]>there is still a .guix-profile, but the only thing it has in it is /etc/profile <x8dcc>Hello, I just installed the guix system on a laptop, and I am having some trouble setting up the network stuff. I normally use "iwd" in other distros, so I tried "guix install iwd", but I can't use "iwctl" or start any iwd service with "herd start iwd" <sneek>Welcome back x8dcc, you have 1 message! <sneek>x8dcc, Rutherther says: I can't really say whether font packages _should_ add XDG_DATA_DIRS, but it's not something they would be doing now. You have to do it yourself, even on Guix System this is handled by adding lines directly to /etc/profile of the whole system. (or for guix home from profile files)There is currently no "Guix" native solution to this that wouldn't be more of a workaround (like the dummy package I proposed) <spiderbit>Hi, I try to get stumpwm modules installed in a way that they are in the right path to load them <vagrantc>x8dcc: might need to install it as a service in your system profile, not as a user-installed package. not sure, never used it. <spiderbit>what tool do you use for uploading some code lines <x8dcc>vagrantc: And how would I install it as a service? <x8dcc>I searched the documentation, but I found little information about iwd (some mentions, but I think for other services) <spiderbit>in this long list the closest I saw was closure am I blind :D <spiderbit>somebody I think must told me something ilke that here <spiderbit>It's kind of installed I think when I install stumpwm? I would think I do then install stumpwm-with-mylibs but that does not exist <spiderbit>(specification->package "stumpwm-with-libs") <spiderbit>guix system: error: failed to load '/etc/config.scm': <spiderbit>guix/utils.scm:699:2: In procedure package-name->name+version: <spiderbit>(specification->package stumpwm-with-mylibs) <spiderbit>is there something with my syntax wrong or do I understand the ideah behind define or something wrong? <spiderbit>does it add it to the normal stumpwm package? <spiderbit>I thought it defines a new package that is modified <Rutherther>spiderbit: yes, there is a problem. A specification should be a string, but you seem to be giving in some kind of a symbol, possibly bound to the package <spiderbit>maybe without the specification thing just as list item in packages? <spiderbit>'create-asdf-configuration' phasebuilder for `/gnu/store/0m5kabvfnmmw6ad58j3dlsk885da5i7z-stumpwm-23.11.drv' failed with exit code 1 <spiderbit>build of /gnu/store/0m5kabvfnmmw6ad58j3dlsk885da5i7z-stumpwm-23.11.drv failed <Rutherther>spiderbit: if stumpwm-with-mylibs already is a package record, then yeah, just don't call any function at all, it's already a package. specification->package means take a specification, and make a package record from it. A specification is a string, that references to a package, by name and possibly version <spiderbit>this modules do not get linked in some systemwide /module directory right... <spiderbit>interesting if I only have cl-str added it works <janneke>ACTION bisects the findutils-gnulib-tests problem to glibc's sigreturn.o module <janneke>so, another patch we're missing -- it works on debian :) <x8dcc>How can I see the value of a variable being used in a configuration file? e.g. %desktop-services <spiderbit>ok I think that was ok for this str lib don't remember for what I needed it but that seems to not work for the modules <spiderbit>maybe having a module in the software list should have a description how you install it :D if it's not just install <package> <vagrantc>x8dcc: there is no simple answer; i have never done it. might also work as a service using guix home, but again, i have never used that. <x8dcc>vagrantc: Okay, don't worry. If anyone else uses iwd, please let me know. <x8dcc>How about my other question for showing the content variables? That can't be an uncommon thing to do <spiderbit>why does the package search tool find cl-wayland when I search mahogavy a wayland port of stumpwm <spiderbit> "This package provides Common Lisp FFI bindings for libwayland, <spiderbit>primarily for the mahogany window manager.") <spiderbit>ahh that's why... but there is no mahogany as package :) <spiderbit>I guess the easiest way is to use git clone and set the path and the second do it somehow over homeconfig to install it therefor not as system service and include it there <spiderbit>Well, thanks anyway, often help is to try to explain the problem and then you see parts of your errors... <bigbookofbug>spierbit: mahogany has a guix.scm that can be used to build it on a guix system <x8dcc>I don't understand how this is so hard to set up. If I can't install iwd, I guess guix is not for me <ieure>x8dcc, I believe Guix uses NetworkManager for that stuff; all that works out of the box. Never heard of iwd. <ieure>But, if easily using iwd is a hard requirement for you, I agree, Guix is probably not a great fit. Nothing wrong with that at all. <x8dcc>I rather not use NetworkManager; it has given me issues in the past, and I think it's too much for what I need <janneke>istm that iwd hasn't gotten enough love yet in guix <x8dcc>janneke: Thanks for that link. I will see if I get something working <janneke>x8dcc: (y) -- i've only ever used wpa-supplicant, in the days before we had (ugh) network-manager <civodul>do you remember the happy days of wicd? <gabber>did issues.g.g.o just swallow my patch-series cover letter? <janneke>sure, otherwise it wouldn't be in the list archives? <gabber>yeah, i added an unrelated commit when creating the patch series (just posted an email where i explain that). <gabber>should i send in revision 2 right away (with just the numbering fixed) or should i wait for other comments?