IRC channel logs
2023-03-30.log
back to list of logs
<Guest2919>Hi Guix! Is there a way to read the link of a directory in guile ? <spiderbit>I notice that my cpu fan starts and stops here in guix but I can't remember that happening in nixos at all and if then much more seldom <spiderbit>the one thing I found on nixos is in the hardware-configuration this option <spiderbit>powerManagement.cpuFreqGovernor = "powersave"; <spiderbit>I don't really understand the difference betweer the first and the last tlp setting <spiderbit>currently the fan goes up with nearly each firefox webpage load <f1refly>When I log in with gdm, my session fails to initialize properly. my window manager starts, but some environment variables seem to be missing. my session is sway and I use waybar, which isn't launched. my font config seems to be missing. all of this doesn't happen when I log into a vconsole and launch sway manually. How can I see debug messages from gdm regarding the launched session? <f1refly>I checked /var/log/gdm/greeter.log and /var/log/messages, but both didnt show any messages regarding sway <f1refly>Figured I should probably set debug? #t in the gdm config, doing that now.. <Ckyle>Hi, I got a question about guix home. Can I edit .profile in my scm file somehow? In my .profile I want to source .config/bash/rc <jpoiret>f1refly: which env variables are missing? <f1refly>I'm not quite sure yet, but waybar doesn't launch so I'd guess my PATH is missing something. Also my fonts configuration isn't applied. I found a bug report from january (?) where someone claimed gdm is missing with his PATH, maybe its got something to do with that <f1refly>sorry for typos, I can hardly read what I'm writing because some letters overlap <jpoiret>gdm on guix system should launch your session through the login shell <jpoiret>but that's because we patched it to behave that way <f1refly>My gdm config right now is just wayland? #t and an keyboard configuration <jpoiret>but even then, what we do is really just run the session command through the login shell, nothing else <f1refly>no, I never came around to migrate to guix home <f1refly>maybe my dotfiles are messing the variables up <f1refly>I just had a debugging session to figure out why i couldn't log into tty1 on my other computer <f1refly>maybe i messed something up for this one <jpoiret>because our script does `<loginshell> --login -c 'thecommand'` <f1refly>i still have a symlink to a zprofile file which i deleted because it has become obsolete <unwox>Ckyle: yes, try extending home-shell-profile-service-type <Ckyle>Thanks! My knowledge Guile is not that great... thought I've tried that already, will try this asap <f1refly>PATH is definitly garbled somehow, by dmenu doesn't find binaries, but when I open a shell in a new terminal they're available <rekado>PotentialUser-76: IIUC Web uses gstreamer for media decoding. Have you installed any gstreamer / gst-* packages? <mirai>what should I do regarding #61015 ? <mirai>since core-updates is soon to be a relic of the past, what do I do with it <jlicht>f1refly: how do you start your WM/DE? <f1refly>I'm starting a sway wayland session with gdm <jlicht>yeah, in that case you are responsible for setting up everything that is normally done via a login shell :/ <f1refly>I just moved .zshrc and .profile, but that didn't have any effect on my problem <f1refly>oh, so I have to add the guix paths to PATH myself? <f1refly>someone told me here that gdm has been patched to launch a login shell in guix, which should set up the environment <f1refly>but sure, I can add those paths myself <jlicht>but you can also replace "sway" by a wrapper that simply runs it in a login shell <f1refly>I'm on <10kb/s, I'll check the ticket later :) <jlicht>f1refly: sorry I don't have a better fix, exactly this gap in (IMHO) important functionality lead to me ditching gdm. <f1refly>I've been cleaning up my dotfiles of superflous things, I thought the vars would be set in /etc/profile when logging in with gdm <f1refly>I just want to launch a wayland session, I'm not particularly picky about my dm <f1refly>will it set my environment up for me? <f1refly>greetd looks pretty nice from the documentation <jlicht>the one with terminal-vt "6" is the sway one <jlicht>I can also confirm it works pretty well with guix home things :-) <f1refly>so it's a replacement for my getty service and will launch different things depending on which tty i log into? <jlicht>pretty much; there's also more graphical greeters, but I haven't been able to get those to work well <f1refly>that's actually i very neat idea, i like it! <f1refly>vt 4 is your "rescue" terminal without any sourcing, right? <jlicht>And I have the login-shell snippet to make sure it always uses my login shell, as I had one user to try out zsh with that was kind of annoying to configure otherwise <f1refly>ACTION will proceed to steal this snippet <jlicht>f1refly: small correction, the login-shell snippet does all the sourcing it needs to do already (at least for both bash and zsh), so I explicitly disable sourcing here so I don't double-source some files <jlicht>so my vt 4 still sources things, just whatever my login shell would have sourced with the old-school getty/login services <ardon>Hi, I've tried sending out a message over to help-guix twice (yesterday and an hour ago), but I can't see the message on the archive, do I need to ask for any permission? <jlicht>ardon: the Powers That Be have to manually let your first message(s) through, so waiting and/or politely pinging here from time to time is going to be your best bet <ardon>jlicht: Ah ok, thanks for letting me know <apteryx>perhaps there's a -Dflatpak=no or similar <apteryx>then I'd report an issue and try to patch the check in the code <apteryx>(an issue to the valent upstream bug tracker) <apteryx>the GCCs aren't hidden package anymore... should we fix that, or stop hiding them? <mirai>I've noticed that the mails with X via Guix-patches via <guix-patches@...> have one thing in common <mirai>X-ACL-Warn: , <!SENDER NAME!> via Guix-patches <guix-patches@gnu.org> <apteryx>does soemone know which packages are failing on staging which aren't on master? <apteryx>perhaps it's just the aarch64 and powerpc64-le builds that haven't yet caught up <rekado>apteryx: one of the aarch64 honeycomb nodes has I/O errors, so builds fail <apteryx>ah, there are some ruby failures on staging apparently <rekado>I’m currently trying to get rust 1.67 from the rust-team branch built on aarch64, so I guess there’s not much free capacity for this architecture at the point <apteryx>rekado: OK! could we disconnect that failing machine until it can be serviced? <apteryx>except the few ruby failures, staging looks pretty much like master <cel7t>Hi is there a way to check if a program got built with a given configure flag? <cel7t>Or is there any program that has an easy-to-check configure flag <etienne_>You could look at the manifest for that specific program you are interested in; I'd imagine the substitute you would download if no compilation is needed follow from that manifest anyway. <jpoiret>you can `guix build --check <your-package>` and look at the output of the configure phase <cel7t>I'm basically testing a patch that works with #:configure-flags <jpoiret>or you could look at the package source <cel7t>jpoiret: I'll give that a try <jpoiret>you can also make the package build fail right after the configure phase (by adding a failing dummy phase after it) and then with `-K` look at the configure.log in the build directory <f1refly>jlicht: zsh still can't find sway, even when using your login-shell construct with greetd :( <dthompson>thank you 'guix pack -f docker'. makes gitlab ci easy to setup. :) <gnucode>hey guix, I am trying out OpenBSD for this coming week just to play with it a bit. So far I like it quite a bit, and I might be staying there for a while....However, I do wish OpenBSD had a config.scm. Can someone help me brainstorm how much of guix I could port to OpenBSD? I'm not even talking about packaging at the moment. <gnucode>I'm talking about etc-service-type as a start. <mirai>probably the kind of work to add an additional kernel? <jpoiret>no guix service will work without a daemon to build anything <gnucode>I'm hoping to create a simple declarative config.scm for OpenBSD that creates various config files in /etc and ~/ . <mirai>there's debian/kfreebsd that you might be able to steal some ideas from <gnucode>jpoiret: why do I need to port the C++ build daemon? I'm not talking about building packages. I'm talking about creating config files as a first step. <jpoiret>how do you think the config files are created? using gexp which are then lowered by the daemon to get some directory in the store, that then gets copied on boot to /etc via the activation-service-type <jpoiret>but that's the main issue: guix (even guix system) is tied to the guix daemon, and uses gexps, lowering and friends everywhere <jpoiret>even local files are handled through (local-file ...) which is lowered by the daemon <gnucode>How difficult do you think porting the C++ build daemon to OpenBSD will be? <gnucode>And can you link me to the C++ source code? <gnucode>jpoiret does guix use guix-daemon.cc and not nix-daemon.cc? <jpoiret>for example nix/libstore/build.cc is one of the biggest (if not the biggest0 file <gnucode>jpoiret: would it be easier to just convert the C++ daemon to guile ? <rekado>rather than porting the old daemon to a different system it’s probably more fruitful to continue the port to Guile. <jpoiret>but you'll face the same difficulties when porting the daemon to openbsd in C++ or in Guile <rekado>I like fruit, so I chose that word over “noble” ;) <jpoiret>just now you have to also rewrite it in guile <gnucode>jpoiret I feel like porting guile code will be easier than porting C++ code. That's just naive me. <rekado>note also that all that process isolation business looks very different in the BSDs <jpoiret>and guile has some rough edge-cases with multi-threading, fibers and spawning programs <rekado>user namespaces is a Linux feature <jpoiret>yep, that's the main obstacle to port to openbsd, the rest should work just fine <gnucode>rekado: I know. Do you need process isolation to reproducibly build config files? That's the main thing that I want to build at the moment. <gnucode>this is probably crazy talk. I don't normally have the follow-through to pull something like this off. But man it would be awesome to use guix to declare all of my config files on OpenBSD. <jpoiret>i don't know if you can disable chroot at build time, maybe that's possible <jpoiret>then you could take a look if the daemon builds <jpoiret>if it does, then you'd probably have to port some guix guile code as well since we rely on linux syscalls to build some system stuff <gnucode>jpoiret: I wonder how likely Guix maintainers would be to commit any work that I did. To be fair this is all conjecture. I highly doubt I will actually do any of this stuff. <jpoiret>if it works they'd probably be happy to add it <gnucode>can someone point me to the source code location for the work that has been done to create a guile daemon that replaces the C++ daemon? <jpoiret>if you run it from a checkout, i guess that should be enough, but to support guix pull, you'd also need a full toolchain for openbsd <gnucode>I wonder if the OpenBSD guys would be willing to add an option to their bootloader to roll back to the latest generation. <hwpplayer1>Do we have a brief information for Thinkpad T490 laptop compatibility with GNU Guix ? <unwox>hwpplayer1: you'll probably need nonfree firmware for wifi card and bluetooth. there is another channel for details <unwox>other than that everything should work fine. i use guix with t480 <jlicht>f1refly: That's pretty weird actually. Can you use the "recovery vt4" (so with login-shell, but without the "-c sway") and run `which sway'? <uniavix>I'm willing to expand as many acronyms and abbreviations as I can but I'm not able to do it on my own. Anyone interested on this idea? To have Guix and Gnu without any acronym or abbreviation at all? It would make understanding the code and the system much easier. <apteryx>hwpplayer1: I'd actually advise to purchase a cheap and compact USB dongle for using wifi, it should make your life easier in the end <apteryx>(and is more respectful of your software freedoms; e.g. you can study or fix a problem in the firmware used) <hwpplayer1>apteryx: If I get one of that USB can I run %100 free software ? <apteryx>you'll have to be careful which one you buy of course <jpoiret>uniavix: wdym by expanding acronyms/abbrevs? <hwpplayer1>apteryx: Understood Do you have some model in your mind ? <unwox>apteryx: is there an adapter you could recommend? <apteryx>I've purchased some from thinkpenguin in the past; they still work well <hwpplayer1>Okay I need to end my session for now. See you then friends <uniavix>That would be non_free, wifi would be wireless_fidelity, and bluetooth would be blue_tooth. USB, Universal_Serial_Bus... Long options would have underscore and not dash, example given, --dired --directory_editor and all variables, constants, and function names would be written in full. At first I can understand wireless_fidelity but not wifi and so on... So... For the sake of understanding... writing it all in full is better than <uniavix>A better long option to demonstrate it: --numeric-uig-gid --numeric_user_identification_and_group_identification <unwox>apteryx: i believe my desktop is beyond saving unfortunately <apteryx>err, no, it's 300 Mbps, so megabits :-) <unwox>and my options for free hardware are quite limited as 80% of the world won't ship to my country anyway <apteryx>if you're on a budget or other constraints, the important part is finding a dongle that uses the same chipset <apteryx>so in theory anything you find that uses Atheros AR9280 should work (since the Linux kernel driver will work the same) <unwox>apteryx: got you, thank you for the advice <f1refly>jlicht: it returns /run/current-system/profile/bin/sway <f1refly>it also complains about not finding `uname` when initializing the shell <apteryx>AprilGrimoire: if the bug is still open probably it hasn't <jlicht>f1refly: it shouldn't make a difference, but I have sway installed in my user profile <jlicht>separate issue: Could someone using guix home try to log in and out (via TTY) several times in a row? If I don't have any other active sessions for that user, I keep leaking shepherd instances <jlicht>f1refly: and you can just run "sway" from your tty? <jlicht>AprilGrimoire: as in, how to do it? Or rather, how to see if it's been done? <jlicht>If the first: send an email to XYZ-done@debbugs.gnu.org (or XYZ-close@likewise) for an issue <jlicht>the issues frontend is updated shortly afterwards <apteryx>-close is obsolete per debbugs control doc, IIRC <PotentialUser-12>I have read some Guix documentation and I'm attempting to setup some Guix stuff for my particular needs (pin versions, offline guix (no access to "official" substitutes servers, etc.), and I'm currently running through tons of issues.... <PotentialUser-12>I have a "guix publish" instance running locally on my own machine but I can't have it serving neither generating the cache (NAR ?) files <PotentialUser-12>is there some documentation describing the cache severs files+folder hierarchy ? <PotentialUser-12>other questions: how Guix knows if a package was "installed" by building it from source or by downloading some substitutes ? <PotentialUser-12>ex: when I try to challenge a "hello" package installed by substitutes, I get this warning message "guix challenge: warning: could not challenge '/gnu/store/s5pd3rnzymliafb4la5sca63j86xs0y0-hello-2.12.1': no local build" <PotentialUser-12>when I try a "guix weather" with my substite server, I get tousands of such messages "acquiring write lock on `/var/guix/temproots/32081' ; downgrading to read lock on `/var/guix/temproots/32081'", I'm not sure of what it is <apteryx>PotentialUser-12: the package needs to be built, or made available first <apteryx>guix build hello && guix challenge hello <apteryx>you'll probably want to build it without substitutes for a useful cross-comparison <PotentialUser-12>guix challenge: warning: could not challenge '/gnu/store/s5pd3rnzymliafb4la5sca63j86xs0y0-hello-2.12.1': no local build <unmatched-paren>PotentialBear-86: you were quieted by the bot for pasting too much :) please use paste.debian.net or similar <PotentialUser-12>for the first challenge, were no substitute-url was provided, what has been used for challenging and determining that the packages were identical ? <PotentialBear-86>hi i was curious if there is a way to manager extra profiles how we manager guix home (with reconfigure) rather than with the manifests? manifests seem a lot fussier and im just looking for a no-fuss way to have profiles <PotentialUser-12>and for the second challenge where I specify the substituteURL, guix raises the warning that there is no local build (no local build where ? on my machine or on the substiute server ?) and result is unconclusive <minima>hi, i was exploring the idea of a lightweight system image that only "exposes" one single (GUI) application - think of, say, a browser that you then run as a VM <minima>i was thinking (but i haven't verified this and i might be wrong) that i might get away without window manager, just a login manager (or a simple scripts) that "autologins" me in <minima>and a xinitrc that launches the GUI app (perhaps, plus setting the required screen resolution) <minima>so my plan is either naive or might require some fiddling for it to work, i imagine? <minima>hey unmatched-paren, thanks! that looks interesting indeed, i was briefly reading about it this morning <minima>and wayland is ok, i guess, so that wouldn't be a problem <minima>i should still build a guix service for it though, if i understand the flow correctly, unless there's one already <minima>ok, neither of them seems to be packaged, at least in GNU Guix, but i might be able to handle that part <minima>understanding how to create a service might be a bit more complicated, but i'll see when i get there <minima>oh that's so sweet of you, thank you so much! <unmatched-paren>i wrote home-dbus-service-type and home-batsignal-service-type, which are in guix <minima>lovely, thanks, i'll use those as a reference <unmatched-paren>the way it's all structured is the same, it's just that system services often have to be more sophisticated <minima>this might be a bit out of scope guix-wise but, other than cage/matchbox, do you think one could do with just execing the app from a xinitrc or startx? <Guest19>Where are logs stored for shepherd? I switched my HDD and now sometimes I can't shutdown the system anymore since it says "service root is not started" and I can see an syslog that shepherd exited forcefully <Guest19>With switching I mean it happens sometimes after I disc my HDD with Guix and connect the new one, after some time I disc the new HDD and connect my old HDD with Guix again (need to switch OS that is way I am doing it) <podiki[m]>i get network timeout but the address does work when I try e.g. in a browser, maybe the importer timeout is too short? <apteryx>I reproduce: In procedure connect*: Connexion terminée par expiration du délai d'attente <apteryx>or 'connect*: Connection timed out' in terse English <podiki[m]>without -r works for the package, just won't ever go more than one package deep with -r <podiki[m]>but the files.pythonhosted.org package it tries to download is a live link for me in my browser, hrm <apteryx>weird; so perhaps the a timeout needs to be relaxed <apteryx>at least it's reproduce, that's very lucky for a network-related bug :-) <podiki[m]>seems pretty quick in browser too, so < 1 second <podiki[m]>figured i'd check before i ran around my network (virtually) <apteryx>two things to do: 1. handle the exception and 2. bump the timeout <apteryx>rekado: do you know if overdrive1 online? failed to connect to 'overdrive1.guix.gnu.org': No route to host <Guest19>apteryx: Thanks. I assumed shepherd may have its own log. I also see "Mar 30 20:15:41 localhost -- MARK --" what does it mean? <Guest19>apteryx Also it doesn't really show anything about shepherd. It only says that it is not responding and and it sends a sigkill <podiki[m]>apteryx: what are the guile/guix procedures I should be looking for? just skimming the pypi importer right now <jpoiret>podiki[m]: someone recently reported the same but I couldn't reproduce on my side <podiki[m]>mind pasting that somewhere then? save me the trouble of one by one imports :) <podiki[m]>either way I would think a timeout should be in the 1 second or so range <podiki[m]>I think everything uses url-fetch with a time out of "10"...what units? second? <podiki[m]>weird, in my guix checkout it works (before making any changes), but not in my current guix <rekado>apteryx: no idea about the overdrives <apteryx>ACTION is about to revert "gnu: network-manager: Update to 1.43.4." <jpoiret>hopefully this will fix my issue as well <jpoiret>did you check that it would revert the buggy behaviour? <jpoiret>but i guess if it's not an actualy release then it's good either way <apteryx>but it didn't exist prior to the update, according to a few victims <jpoiret>why do projects have this even/odd release policy, it's super confusing <apteryx>it should have been abandoned in GNOME, but it's not enforced <jpoiret>heh, it's just one more thing I have to manually do when resuming from suspend though, so I'm not too bothered <jpoiret>I have to restart my fan control program because it just dies on suspend <jpoiret>and I have to manually send an ACPI call for my touch screen and pen to wake up <podiki[m]>apteryx: thanks! do we want to go to the current (even) release then? or just hold on to the previous one for now? <apteryx>i'm not sure. given I haven't been able to touch base with the #nm people I'll leave it at that for now <apteryx>but if you are interested in testing things, that's more than welcome <podiki[m]>selfishly happy with that, unlike jpoiret it is my only manual action after suspend <podiki[m]>I can try out the current (stable/even) nm release and see <podiki[m]>the other thing to test is some linked issue, forget the number that Ludo mentioned <spiderbit>Hi, stupid question, I want to use the ssh homeservice... I have a public and private key in my nixos home directory... the homeconfig you seem to only have the public key... so do the private key live outside of the homeconfig and I have to copy that directly into my .ssh/ manually? <spiderbit>well actually it's not really a stupid question, why do I use this introduction :D <jpoiret>you shouldn't put the private key in the config, since it would be put into the store which is world readable <apteryx>subversion fails to build on i686. odd <jpoiret>gtk+ depends on sambda and gpgme, great! <spiderbit>herd: exception caught while executing 'load' on service 'root': <spiderbit>In procedure fport_write: Input/output error <spiderbit>not sure it's the ssh I also added redshift-service-type <spiderbit> (authorized-keys (list (local-file "foo.pub"))))) <spiderbit>maybe it has to do with the emacs daemon still <spiderbit>I would expect that this pub key get's copied to .ssh/ but there is none and there is no redshift process running <eriedaberrie>Looking to move from NixOS, I was wondering if there's an equivalent to agenix? So I can store encrypted files alongside the rest of my config that get decrypted automatically when I reconfigure the system. <spiderbit>I have put this 2 service types under (home-environment (services (list ...))) <mvnx>eriedaberrie: Maybe git-crypt? No experience yet but I'll also be looking for a solution soon. SOPS also comes to mind but there are is no Guix package for it yet. <eriedaberrie>mvnx: Would there be an easy way to get files from git-crypt into my filesystem without also being in the world-readable store? Agenix would decrypt them into somewhere in /run. <apteryx>jpoiret: and gtk transitively depends on qt ^^'. There's some dependency cleanup to do. <jpoiret>it's been a recurring feature request <jpoiret>spiderbit: once you use the ssh service, it doesn't use ~/.ssh afaik <apteryx>eriedaberrie: the usual workaround is to put sensitive files (secrets) under /root/secrets, for example. <jpoiret>apteryx: here I was just patching gpgme thinking it would be easy to test <apteryx>is there something better than (guix build utils)'s strip-store-file-name + package-name->name+version to retrieve the version from a store file name? <bumble[m]>Hello, I'm hoping someone will give me advice. I want to use cfm "cactus file manager" to manage and open files, but it requires the env var $OPENER to be defined. I want guix home to define $OPENER with an executable bash script copied from the internet. How should this be done? <devcarbon>Anyone know what it might take to use flutter in guix? <chipb>eriedaberrie: when I looked at it there's not really much complication in agenix. one might be able to convert into a guix analogue in an afternoon or so. <chipb>I'm sure people'd prefer it be guile-implemented rather than the shell script, but you, uh, bash with the tools you have. <eriedaberrie>I figured it wouldn't be way too complex, but I was hoping someone would've made one already. <eriedaberrie>Maybe I'll give it a shot after I figure everything else out. <Guest19>devcarbon you mean officially or just make it work? Officially will never happen since it is to much work to cleanup since it requires the whole android studio thing and we need to work against Google. If you just want it to make it work maybe it is enough to patch the binaries but it will have problems with not having FHS. I would say if you <Guest19>actually want to use Flutter, don't waste your time making it work in Guix <Guest19>except you want to gain experience but it will be hard counter productive <spiderbit>jpoiret ok ssh login works on the hosts I put my pub keys or how that works... <mvnx>eriedaberrie: No experience but it doesn't look like it. Maybe you could use just `age` directly? <spiderbit>yet this error message makes me nervous, was that about the redshift service or maybe still from yesterday with the emacs service, but I did not change anything and why does "root" then throw a input/output error <devcarbon>Guest19: Just work, really. Official would be nice, of course, but I'm trying to figure out if I can run guix as my whole system on my work machine or if I need to just use it as a package manager on Arch. I am building a project using ClojureDart, as long as I can do that I'm happy. <devcarbon>I may be mistaken, but I don't think android-studio is a hard-dependency, is it? probably android-tools, though, so probably still the same problem. <Guest19>In that case just use Arch with Guix as package manager. I doubt you will have a nice experience otherwise. Guix is complete different and without some touching software will not work out of the box since they assume things in a way that Guix does not do. What I mean is you will probably just have problems with Flutter on Guix. <spiderbit>devcarbon you know there is a nix package manager in guix, maybe that could help you for some of your problems <spiderbit>I do just the same coming from nixos, but nobody pays me so that is just for my private machines <devcarbon>spiderbit: I think I got really close with nix, but I got stuck. Seemed like it was trying to use guix's clang that got it all confused or something. Not that I really know my way around guix or nix, so probably operator error here. <spiderbit>well writing packages is also a big reason for me if I have to do that, I feel more secure than with that horrible nix configuration language :D <devcarbon>Yeah I kinda feel like I'm in a funny situation, tropin's RDE seems like a near perfect solution for what I'm looking for, but if I can't use flutter I suddenly can't get any actual paid work done. :P <devcarbon>Guest19: When you say not officially, do you mean for Libre reasons? (IE. Is it something that nonguix might support?) <Guest19>devcarbon yes. Well if someone would make a pull request they would accept it probably but that would require a person that invests time to get it to work <mirai>jpoiret: are you planning on upgrading shared-mime-info? <jpoiret>i'm focusing on core-updates these days, and rn. the priority is fixing build failures, not upgrading packages <mirai>you're one of the few names I see working on core-updates <jpoiret>If it does not fix a major issue we won't do it <jpoiret>Also, anyone can work on core-updates :) <mirai>I've no idea what's the criteria used for classifying them <jpoiret>Would core-updates be completely unusable for a user without such an upgrade? Do we need it to fix build failures? <mirai>I know of one package that *can fix* some build failures of other packages