<apteryx>sneek: later tell civodul and case that motivated this; linux-libre: 24 cores, 2.9 GiB, 36.76 s. On master: 1 core, 95 MiB, 4 m 10 s <nojr>andreas-e: thanks for the response!! <zimoun>Once “docker load $(guix pack -f docker –save-provenance)”, how I can find the ’provenance’ only with the Docker image? (docker tools) <shcv>it seems like the zfs package provides a module, but I don't know how to add it to the system config so that it's available for the initrd <nckx>shcv: Are you trying to boot from ZFS /gnu? I'm afraid that's not supported yet. It would need patches adding ZFS to gnu/build/file-systems.scm, at least. <nckx>I'd be happy to be proven wrong. Using ZFS filesystems from Guix System once it's booted isn't a problem. <pineapples>nckx: Hi! I updated knot-resolver and it works without an issue! :) <nckx>Ah, great. Looks like I got the Lua stuff right then. <pineapples>Replacing ' "-f" "1" ' with option "-n" should do the trick *nckx lies low and hopes that someone™ else will fix it before it becomes fatal. I already don't use knot-resolver the programme; ‘fixing’ the service would be a complete shot in the dark. <nckx>pineapples: Could you file a bug? That should tease out a volunteer 🙂 <nckx>Thanks for providing the solution in any case. <pineapples>nckx: I'm sorry, but I can't :( In any way, it's not a fatal error and knot-resolver continues to work with option "-n" for now. By the way, can we change the hard limit for number of file descriptors on Guix System? knot-resolver nags me about the limit being too low (4096). <pineapples>Systemd already raised it to 524288. Why shouldn't we too? <nckx>Well, I'd say ‘that's a good bug report’ but... 😛 <nckx>Is the ‘sorry, but I can't’ anything we can help with? <pineapples>nckx: It'd be awesome, if an ability to open new issues via issues.guix.gnu.org was implemented :D Not a fan of creating special-purpose email addresses <Gooberpatrol66>How do I prevent a laptop from sleeping when the lid is closed on AC power with gnome? <nckx>Gooberpatrol66: Look at elogind-configuration under ‘Desktop Services’ in the manual. Not familiar with the G{NOME,UI} way. <vagrantc>or use "elogind-inhibit" if you just want it temporarily *nckx wonders if ‘docked’ == AC. <Brendan[m]2>The documentation says "Available parameters and their default values are: " <Brendan[m]2>but from what I can see only default values are there, not all the available parameters <nckx>Brendan[m]2: By parameters they mean the ‘keys’, not the values. Is there a logind manual we could link to instead of maintaining an easily outdated copy? <Brendan[m]2>I think it probably means plugged in to the power point <Brendan[m]2>Gooberpatrol66 maybe this (elogind-service-type config => (elogind-configuration (handle-lid-switch 'ignore))) <c4droid>Have anyone tried to pack termonad to guix? <xelxebar>c4droid: I don't see anything pending in guix-patches. <xelxebar>Cool project. I used to use xmonad. Might try packaging this. <c4droid>xelxebar: termonad have a lot of haskell package depend to resolve, I tried twice to resolve the depend but I failed, so I'm waiting for other packager can pack that. And, I'm a xmonad user too. :) <ryanprior>I'm about 30% of the way through packaging Hugo =D <ryanprior>checks notes 37 dependencies down, 33+ to go. So possibly more than 30%! Depending on how many transitive deps the remaining 33 have! <c4droid>I just post a issue on termonad's github homepage, hope termonad author or other packager can see that. :) <ryanprior>So far it's been 14 direct deps, 23 indirect deps. So at that rate I'd expect my remaining 33 deps to have 87 further indirect deps X.X <ryanprior>Assuming that holds, I'd be 29% done currently. So I think my original guess was pretty good! <ZombieChicken>What the heck are you setting up for? Sounds like something that is as much a mess as calibre <ryanprior>Hugo is a static site generator. When it's finally compiled it's just a single static binary that only depends on libc. <ryanprior>Oh no, just 14 direct deps that I've packed so far. 47 direct deps total. <ryanprior>Hugo has been a major economic success, it's open source but many profitable services have sprung up around it offering value-adds. <Brendan[m]2>Haskell I think distributes software in large numbers of small dependencies. <ryanprior>It's suffered greatly for its success in the form of scope creep. <ryanprior>rekado did a monumental effort reducing the amount of ocean-boiling necessary to compile Pandoc =D <ryanprior>(it's still not ideal afair but a fraction of what it was) <ZombieChicken>For me to install pandoc on my current Gentoo system, I only need to install 124 packages totaling roughly 160MiB downloaded <c4droid>I got termonad author respond, author say waiting the other guix packager pack termonad to guix <Kimapr[m]>alternatively you can copy-paste it from the terminal if your terminal supports that <efraim>glibc-final-with-bootstrap-bash built, just need to fix the glibc patch ***amfl_ is now known as amfl
<efraim>sneek: later tell civodul I think I messed up the patch and included the git diff line twice, so it ignored one of them <bdju>OBS is missing a lot of the icons <bdju>I'm on sway which might matter <bdju>so the show/hide symbol is invisible + and - for sources is invisible <bdju>they still work if I blindly click around the space enough, but it's not ideal <bdju>I also seem to be missing a lot of options in the 'add sources' menu, but that's probably a separate issue. <bdju>oh yeah and the icon left of the sources showing what type they are is also gone <bdju>sure, good idea. sometimes I just tell irc first in case it's a quick fix <iyzsong-w>bdju: do you have adwaita-icon-theme installed? it's the default gtk3 icon theme. <Brendan[m]2>I do and i still get this bug. btw obs is a Qt application <alextee[m]>I'm occasionally getting this error when running gtk apps, not sure what's wrong: unable to create file '/run/user/1000/dconf/user': Permission denied. dconf will not work properly. <efraim>looks like that wasn't the fix for the patch <Brendan[m]2>I notice that system was replaced with systems in the offload code, but use of system is still in the manual <sneek>Welcome back civodul, you have 2 messages! <sneek>civodul, apteryx says: and case that motivated this; linux-libre: 24 cores, 2.9 GiB, 36.76 s. On master: 1 core, 95 MiB, 4 m 10 s <sneek>civodul, efraim says: I think I messed up the patch and included the git diff line twice, so it ignored one of them <civodul>apteryx: ah yes, linux-libre compression is definitely a good motivation! <bdju>iyzsong-w: yep, I do. my icons appear fine in pcmanfm and such. <efraim>I watched the nix presentation on guix, I think it'd actually be interesting to play for guixdays and then discuss <wehlutyk>but current rustup cargo also looks for libgcc, which I don't know how to include in /lib64 <wehlutyk>(maybe a cleaner solution would be to package wasm32-unknown-unknown? I'm up for that if someone can provide a bit a guidance) <xelxebar>efraim: What's this Nix presentation on Guix? Mind sharking a link? <wehlutyk>should I link /lib64 directly to /run/current-system/profile/lib ? (Also, a `guix install gcc;lib` gives "package 'gcc' has been superseded by 'gcc-toolchain'", then "error: package `gcc-toolchain@10.2.0' lacks output `lib'") <Brendan[m]2>guix hash is actually twice as fast as sha256sum for me now on a 2.6GiB file O.o did not expect that. <efraim>civodul: I'd like to try dropping glibc-for-bootstrap down to 2.30 but bumping %guile-static to 3.0 <efraim>I don't have a lot of energy for debugging 2.32 but I don't mind guile-3.0 :) <rain1>maybe a font would be my guess <civodul>efraim: that's for the PPC port, right? <efraim>might be relevant for ppc64(el) too, not sure how they'd feel about 2.30 <efraim>although poking make-bootstrap for 2.30 I'd almost rather poke 2.32 which of course would actually be more useful <civodul>i'm kinda reluctant to get stuck an old version but... let's discuss it :-) <msavoritias[m]1><rain1 "maybe a font would be my guess"> Hmm. That is what i was thinking maybe. But shouldnt it come with icecat? <mfg>msavoritias[m]1: what fonts do you have installed? i'm also using sway and i have numbers in icecat :p <rain1>ideally, but it could be a bug <efraim>also reading glibc's source and commits it seems there might be an issue with certain combinations of configure flags with gcc-7.5.0 and ppc64(le) <mfg>"font-google-noto" "font-adobe-source-code-pro" "font-dejavu" i only have those fonts installed <efraim>and have you run 'fc-cache -frv' recently <mfg>efraim: i just wanted to ask the same :) <bdju>msavoritias[m]1: http://ix.io/2ymW here are the fonts I have installed. I'm on Sway and numbers show up for me, last I checked. (I haven't used IceCat in a week or two due to an unrelated issue) <bdju>you won't need all those, of course, but you could copy them all, see if it works, and if it works, then remove some <mfg>what would be the equivalent of xorg-dev in guix? the whole xorg-server package? <zimoun>Once “docker load $(guix pack -f docker –save-provenance)”, how I can find the ’provenance’ using the Docker image only? (docker tools) <efraim>I assume (char*) foo is casting foo to a char*. is that different than (char *) foo? <zimoun>efraim: space does not matter in C <andreas-e>msavoritias[m]: The one font I have and you do not is gs-fonts. Maybe you could try to add this package. <msavoritias[m]1><mfg ""font-google-noto" "font-adobe-s"> I will check also these. Thanks for your help <msavoritias[m]1><efraim "and have you run 'fc-cache -frv'"> Hmm I have never run that command. <civodul>zimoun: provenance metadata is hidden, muahaha <wehlutyk>re: rustup binaries. Actually linking `/lib64` directly to `~/.guix-config/lib` gives me the same `.rustup/toolchains/stable-x86_64-unknown-linux-gnu/bin/cargo: error while loading shared libraries: libgcc_s.so.1: cannot open shared object file: No such file or directory`, even though `ldd` finds it <efraim>looking at the patch more, I'm not sure why hardcode bootstrap bash instead of calling out to sh, wherever it may be <efraim>i'll make that change and see how it works *efraim throws up his hands and says: you get sh! and you get sh! everyone gets sh! <civodul>in other news, i have WIP where i can do --with-graft=glibc=glibc@2.29 (and likewise for --with-input) <efraim>does it work? I thought you couldn't downgrade glibc <zimoun>If I upload the Docker image to docker hub or whatever, then the person who pulls this image, how can they find this “manifest” file? <efraim>$61 for DHL to ship a pinebook pro to my house :/ ***rEnr3n_ is now known as rEnr3n
<xelxebar>efraim: Not sure exactly what you're working on, but I've had problems with guix using sh but using scripts with bashisms. In that case, I'd have preferred a hard-coded bash reference. <PurpleSym>It it intentional that package-input-rewriting stops recursion after replacing a package? I’m having trouble rewriting b and c in a graph like a→c, a→b→c. The last c stays as is, but will cause a conflict, because it’s propagated (Python). <efraim>it's for the bootstrap binaries for building out the system. It doesn't end up user facing <eayus>Is anyone running GuixSD with a RX 5700? I had some issues relating to the graphics card when I tried installing GuixSD a while ago, and I'm wondering if anyone knew whether it's working now? <numerobis>Is there a way to enable zsh completion of guix commands? Such as with `guix package <tab>` or something like that? <civodul>efraim: ah well, it works only as long as the ABI is compatible, which is indeed not guaranteed in this direction :-) <civodul>in practice it works for simple cases <jonsger>no almost everything, visually et al. <numerobis>nckx: Thank you very much! I'll try sourcing that file from .zshrc. :) <Brendan[m]2>I think it's just expected that you have some fonts installed <msavoritias[m]1>I understand also that when I had gnome installed on guix the font appeared correctly. It was after I removed gnome <msavoritias[m]1>So it seems exclusive to the people without a full desktop environment <andreas-e>msavoritias[m]: I do not have font-google-noto in my store, and icecat works with numbers; did you also try gs-fonts? <jonsger>has someone an idea how to add a patch to icedove? Adding "(patches (search-patches "icedove-78-fix-new-page.patch"))" to thunderbird-sources does not work, it then fails in the prepare-thunderbird-sources phase with "In procedure copy-file: Is a directory" <efraim>there was a commit touching the relevant parts of glibc, i'm going to try reverting that and building new bootstrap binaries like that <andreas-e>Question on docker: I am creating an image with "guix system docker-image"; but the result does not contain an Entrypoint, where it used to have one (composed of a "boot-program" and a "system"). <msavoritias[m]1><andreas-e "msavoritias[m]: I do not have fo"> You were right. I installed the packages suggested and it worked. <andreas-e>The problem with manuals is that they are long... Nice to hear that you got it working! <alextee[m]>> The problem with manuals is that they are long... <alextee[m]>lol yes, you need to pair them with "getting started" guides and tutorials <msavoritias[m]1>Maybe it should be in a getting started guide without desktop environment <mfg>Is it possible to also set environment variables in a manifest file for the created environment? <morgansmith>Should I send my patch (43277) to guix-devel? It's really rather important since it fixes emacs-next which has been broken since Aug 31st. <apteryx>andreas-e: booting a docker image Guix system is currently done via 'docker exec image', IIRC. See the examples in the manual. <apteryx>otherwise the system is not initialized and you'll be missing things. <andreas-e>Concerning docker: my mistake, I issued "docker import" instead of "docker load". <zimoun`>On foreign distro, how can I have “bin/sh” in my PATH? Where ’sh’ is provided by dash or gash. <Brendan[m]2>you could for example include dash/gash in your profile and then append ~/.guix-profile/bin to your PATH <jonsger>fingers crossed that this icedove build will work better :P <apteryx>zimoun`: doesn't installing dash suffice? <Brendan[m]2>that will not however make it so that scripts refering to the absolute path with #!/bin/sh will refer to it <zimoun`>dash does not provide ’sh’ in ’<profile>/bin’ <morgansmith>zimoun`: usually you just put a symlink at /bin/sh to you shell. So doing "guix build dash" will show you the location of dash and you can make a symlink. Unfortunately, that doesn't provide easy upgrade... <morgansmith>dash provides a bin/dash and gash provides a bin/gash <zimoun`>I am trying to use the GitLab CI with a Docker image produced via “guix pack -f docker” and I miss something… Because it always fails somehow with “sh not found” (even when changing the entrypoint to dash). <zimoun`>I thought that “guix package -i dash -p /tmp/foo“ will provide ’sh’ in /tmp/foo/bin but it is not. Is it expected? <apteryx>does your system declaration contains dash? <efraim>zimoun`: correctly? something more like -S /run/current-system/profile/bin=/bin <zimoun`>apteryx: I do not use ‘guix system’ but ‘guix pack’ <zimoun`>efraim: on foreign disto, I do not have /run/current-system/profile <zimoun`>efraim: thanks. Everything works locally with something similar to Pjotr notes. But it fails on the GitLab CI because somehow it expects ’sh’ and I do not know how to pack it with Guix. <efraim>I would toss in bash-minimal, but I haven't gotten gitlab ci to work for me in the past <apteryx>zimoun`: I just tried: guix pack -S /bin=bin busybox -f docker then 'docker load < /gnu/store/5p95235l0dfnkrg8dd7d1502r3l7l1d6-docker-pack.tar.gz' and finally docker run -it busybox sh. <zimoun`>apteryx: thanks! I am trying. And busybox seems the package I was looking for. :-) <morgansmith>So when I first joined #guix, my nick was butterypancake but I've since decided to use my real name. However, in guix-devel there is now someone using the name Buttery Pancake. I don't really have any right to complain but it certainly feel weird. Is that just a common pseudonym? <apteryx>It seems the spammer decided to use your pseudonym, likely to create more confusion about their identity. <morgansmith>are you sure they're a spammer? They started a nice thread about folder locations <zimoun`>apteryx: thanks! GitLab CI is happy now :-) <apteryx>morgansmith: posting "normal" messages before going on with spam seems to be part of the strategy. <morgansmith>apteryx: I can't believe Guix is getting spammers. Like sure passive attacks, but isn't posting "normal" messages first an active attack? Like someone took the time to figure out what a normal message looks like. <dlowe>I'm always surprised when Freenode gets spammers. It seems awfully niche. <pushcx>Programmers are an expensive audience to advertise to. <efraim>when I launch openshot from a terminal I get a message that it can't find libvdpau_nouveau.so. sound familiar to anyone? <wehlutyk>Has anyone done rust/wasm development on Guix recently? Is there a way around making Rustup work to get the wasm32 toolchain? <jsoo>wehlutyk: I tried to package the wasm32-unknown-unknown target and found it possible but a lot of work. <jsoo>The blocker I ran into most recently was that the standard crate dependencies of rustc need the cargo-build-system treatment. I had to move on but that's where I left it <wehlutyk>jsoo: good to know, thanks. I think I'll try that too one day. If you have stuff left around that I can pick up, feel free to send it over. <wehlutyk>ah no, without the target that wouldn't work <zimoun`>Is it possible to pass several commands to ‘guix environment’? For example, ‘guix environment -C –ad-hoc foo bar – (foo; bar)’. <apteryx>zimoun`: you could start a shell like: guix environment --ad-hoc foo bar -- sh -c 'foo; bar;' <jonsger>icedove 78 brings finally a working matrix support :) <efraim>Reverting the commit in glibc and then applying the glibc-bootstrap-system patch seems to have worked. I'll have to ask upstream what's up with that patch <efraim>sneek: later tell civodul fixing glibc 2.32 it turns out is easier than forcing 2.30 <zacts>what is the status of KDE on GuixOS? <zacts>(and I hope I'm referring to the distro correctly) <jonsger>zacts: there are some packages of KDE available, but no the full desktop (Plasma) yet <mbakke>efraim: do you have a link to the commit? <zacts>and is GuixOS correct for describing the complete gnu/linux-libre distro? <dannym>zimoun`: Docker hardcoded /bin/sh for all RUN statements that don't have list operands. Easiest is to use lists <jonsger>zacts: I think it's called Guix system but... <zacts>wait it looks like you can turn an existing distro into a full Guix system? <zacts>one idea I have is to call the package manager 'guix' and the full distro 'Guix' <roptat>I think it's best to have the two sound different <zacts>(that's how Perl and Ruby name their stuff kind of) <roptat>I think I'm halfway through the update of groovy :) I had to update our ant package for groovy-ant, so the update will have to go to staging (500 dependents) <zacts>roptat: are you running Java on guix? <roptat>yes, groovy is a programming language that compiles to the JVM <zacts>what is the status of Java on Guix? <zacts>I'm kind of interested in jruby just a bit <roptat>it's usable, we have bootstrapped all the JDKs from 6 to 14 <roptat>we have maven and a very new and shiny maven-build-system <roptat>we're missing gradle though, and that's something I want to work on <roptat>in total we have ~500 java packages (but that's counting one package per artifact, not one per source repository) <zacts>so you can use guix directly to package each java dependency? <zacts>kind of like packaging emacs or vim plugins via guix versus the emacs or vim plugin manager <roptat>yes, but it's hard, because there are many cyclic dependencies (esp. when counting test dependencies) and you can't always use the proper build system <zimoun`>dannym: thanks for the explanations. Busybox does the trick. :-) Good to know about RUN. But you mean in the Dockerfile, right? <roptat>we try to bootstrap everything, which means not binary (or bytecode) version of the packages are allowed <zacts>I think I'll install guix this week and try it out <zacts>the **only** thing I think I want is a normal firefox though <zacts>I need javascript enabled for college <zacts>roptat: when you return, can I use the javac command to compile .java files? <zacts>or must I use some sort of guix environment to do this? <roptat>install the :jdk output of any icedtea or openjdk package and you'll get javac <roptat>(icedtea 1 to 3 for java 6 to 8, and openjdk has the same version numbers as java itself) <apteryx>what must be done to get the hicolor-icon-theme to be honored in a --pure environment? <roptat>(well to be precise, icedtea for java 1.6 to 1.8, and openjdk 9 and beyond for java 1.9 and beyond) <mbakke>efraim: could it be that the problems you had on PPC was because that function was silently failing, and it only now becomes visible? <efraim>It's unlikely, I think the code it's replacing was already in place with the aarch64 port and I would've seen it then <efraim>and it wasn't a problem with glibc-2.30 <efraim>I'll stop the current build and start it again with -K. if it was silently failing then there won't be the environmental file at the root of the guix build <mbakke>efraim: right, I thought I had updated it with glibc 2.31, but reading the commit log it was 2.29 <efraim>oh, so it would've been with the hurd then <efraim>the environmental file is there so it looks like a bug somewhere <mbakke>efraim: there are related substitutions in the glibc package definition (search for SHELL_PATH), can you double check that those are still effective? <mbakke>I think the libio/iopopen.c substitution has been obsolete since 2.29 <efraim>the libio part still applies, it's unchanged <efraim>SHELL_PATH is defined at the top as "/bin/sh" and SHELL_NAME as "sh". I tried with SHELL_PATH substituted as SHELL_NAME and just as "sh" <efraim>my guess is status should've been initialized as 0 or ret never introduced <efraim>either that or it's supposed to fail if libc isn't reentrant (i'm not sure what that means) and there's a bug in guile-2.0.14 with how we're using it in %static-guile <efraim>new idea: do we even have a "sh" in the bootstrap? <efraim>there's a sh->bash symlink in static-binaries <efraim>I'm going to build the bootstrap binaries again but this time with the previous patch (adapted for the changes) and status initialized to 0. ***ae is now known as Guest42286
<efraim>if ret fails it should spit out an error, and status won't be unconditionally -1. hopefully this works <nojr>is anyone using Emacs and its packages in a foreign distro? I'm having a weird issue with projectile. I't wont get activated, even if it's toggled in my init.el, unless I evaluated manually. <nojr>am I'm not sure how to debug this issue <rekado>nojr: are you mixing Emacs packages from Guix and from M-x package-install? <rekado>nojr: a good thing to try is to move away (or remove) whatever Elisp code might be in sub-directories of ~/.emacs.d/ <rekado>andreas-e: I can only get expired PGP keys for you. Is there some place where I can get a fresh key for you? <nojr>rekado: that didn't work :/ <mbakke>efraim: what commands are you running to reproduce the issue? *mbakke goes afk for a while <mfg>is it possible to give the bootstrap phase of gnu-build-system arguments? <roptat>mfg, only #:bootstrap-scripts, a list of files that contain the code to run for the bootstrap phase <mfg>i see, the riscv-openocd package contains a bootstrap script which takes arguments... i need to pass one because the default is wrong :D *raghavgururajan tried /msg but didn't work <joshuaBPMan>hey #guix! I'm trying to reconfigure my remote guix system...I guess I've got a syntax error somewhere in my services... <roptat>mfg, you could replace the bootstrap phase <joshuaBPMan>but that error message is message is information overload! haha <roptat>(replace 'bootstrap (lambda _ (invoke "./bootstrap" "arg") #t)) <joshuaBPMan>I'm going to try commenting out more of my config and see if the error is still there. wish me luck! <roptat>raghavgururajan, you might have to say something to defuse that bomb ;) <sneek>raghavgururajan, you have 1 message! <roptat>so, asking sneek to forget doesn't actually work <roptat>or it forgot something different *jonsger finally has sent the patches for icedove 78 :P <efraim>mbakke: I'm building bootstrap-tarballs on one machine, uploading them to the internets and then using them to bootstrap guix on my G4. not exactly the most reproducable ***bkv is now known as bqv
<leoprikler>I seem to have found some breakage in guile-sdl2 <apteryx>interesting: this fails silently (exit status 1): guix environment --expose=/etc/ -E '^DISPLAY$' -C --ad-hoc gnome-boxes hicolor-icon-theme strace grep coreutils <leoprikler>trying to use (mixer-init) results in Failed loading udev_device_get_action: /gnu/store/long-hash-sdl2-2.0.12/lib/libSDL2-2.0.so.0: undefined symbol: _udev_device_get_action <apteryx>what do I need to expose it a container to share my X DISPLAY? I'm getting: cannot open display: :0.0 <leoprikler>also do -E 'XDG_*' unless you have a reason not to *leoprikler will forever be known as "that person, who discovered the XAUTHORITY hack" <leoprikler>anyway, me → afk, someone sneek me if they know more about my sdl problem <apteryx>still fails. The problem seems to be: connect(6, {sa_family=AF_UNIX, sun_path=@"/home/maxim/.cache/broadway1.socket"}, 38) = -1 ECONNREFUSED (Connection refused) <apteryx>I'm trying to run gnome-boxes in a container <apteryx>ah, it needs the network (--network) <jsoo>wehlutyk: you're welcome, good luck! <apteryx>leoprikler: actually what helped was --expose=$HOME/.Xauthority along with -E '^DISPLAY$' <apteryx>ah! I could understand what's going on with hicolor-icon-theme. It needs to be added to XDG_DATA_DIRS (used by glib), but nothing propagates it. glib-or-gtk build system wraps with XDG_DATA_DIRS, but unless it appears as an input, there's no default icon fallback. <apteryx>I could be fixed perhaps by resolving #22138, but I think simply adding the hicolor-icon-theme as an implicit input to the glib-or-gtk build system is a lower hanging fruit, and a more localized fixe <leoprikler>it seems we're building it without modplug and (mixer-init) defaults to setting that <nckx>Zoinks, it's already a quarter to midnight. Now I feel like one of those annoying people who call later ‘tomorrow’. <nckx>Whichever one implements read-foo-superblock. I think that's (gnu build file-systems), yes. <nckx>(gnu packages ...) modules are just that: discrete packages. You want the rest, the glue code that boots the OS. <nckx>Guix's initrd scripts are pure Guile, they don't simply use libblkid like other distroes, so we do our own basic file system sniffing. 