IRC channel logs


back to list of logs

<opalvaults[m]>float: if you look at the repo, you should see .org files that have tangle directives either at the very top of the file or in individual blocks. look like podiki beat me to it
<float>opalvaults[m]: Yeah I see now. I'll try to implement it and let you know if I run into any problems.
<opalvaults[m]>float: good luck!
<apteryx>rekado: cheaper Samsung EVO 4 TB (10 of them) drives in an a RAID10 array as we currently have could probably also work
<rekado>we can’t swap them live
<nckx>I'm in the RAID6 camp with opalvaults[m] but if this comes down to a vote and not maths we're screwed.
<rekado>the conversion can take a very long time
<rekado>the device supports a low priority conversion, but this can take days or weeks dependent on how busy the storage is
<rekado>I don’t think we’re in a hurry here
<rekado>we do have 9T available right now
<rekado>only 28T are in use
<rekado>yes, with a few more Ts we could GC less often
<rekado>but I don’t think it makes *that* much of a difference
<rekado>we want to get rid of the big store
<rekado>playing with storage probably isn’t going to help much.
<rekado>we have the SAN, which is currently at 10TB max, and we’ve already moved a lot of stuff there.
<rekado>we could probably bind mount the finished directories from the SAN onto /var/cache and clear the source dirs there.
<rekado>this should give us another 10T right now.
<rekado>well, a little less than that
<rekado>and if that’s *still* not enough to allow us to run GC less often we could still drop to RAID 6 and wait a week or two.
<rekado>but none of this seems to be top priority — if we want redundancy we need to get the substitute cache copied *elsewhere*
***rgherdt_ is now known as rgherdt
<nckx>rekado: I was surprised to hear that the scheduled nightly GC still duefully runs and Mathieu (or someone else) duefully kills it each day. Does it not make sense to stop the mcron job until it actually makes sense again?
<nckx>I thought it sometimes completed, on good days, but that was not what I heard today.
<rekado>I didn’t know it was killed all the time.
<rekado>that’s … uhm, not good, I think
<KE0VVT>nckx: Trying to esc. backslashes. error: ("PS1" . "[\\u@\\h \\W]> "): source expression failed to match any pattern
<nckx>It's just stealing IOPS for no reason. And slowing down builds as a side effect, but if we want that, we can just pause that. I'll comment out the service & reconfigure?
<nckx>KE0VVT: Doesn't sound like anything to do with escapement.
<rekado>nckx: but how soon would we eat up the 9TB that are currently available?
*rekado —> zzZZ
<nckx>I don't have an answer to that, but the answer to *that* would be pause Cuirass, not keep a GC lock for most of the day as a side-effect of a seizure.
<apteryx>rekado: mathieu was saying 14 days
<apteryx>good night!
<RuZzz>How to switch the layout in the console if the config says: (keyboard-layout (keyboard-layout "ru,us" "altgr-intl")) ?
<nckx>Depends on your grp:foo_toggle value. I don't know what the default is. I use grp:win_space_toggle.
<RuZzz>It's the default config from GuixSD ISO.
<the_tubular>Dumb question of the day, Can you take a container image made with guix and use it with firecracker ?
<singpolyma>the_tubular: I don't see why not? I intend to try stuff with that this year
<the_tubular>Firecracker needs a whole OS to boot from no ?
<singpolyma>Ah, you mean a guix pack image vs a guix system image? Only way to find out is to try I guess
***califax- is now known as califax
<KE0VVT>Going to try living in Emacs.
<the_tubular>That's also in my plans KE0VVT
<the_tubular>I'm not sure which command singpolyma, but I'll give it a shot I guess
<singpolyma>Well if you use guix pack you will get only what is in the given packages. guix system will build a full OS image if you need that
<float>Hmmm almost everything works with my emacs setup. It properly handles updating my emacs config when I do guix home reconfigure but for some reason I still get a prompt asking me if I want to overwrite the store path that my init.el gets symlinked to even if I have ":mkdirp yes" set.
<float>It seems like right before it gets added to the store the init file is appended with custom-set-variables for some reason.
<float>There is also these set of messages in the *Message* buffer:
<float>Note: file is write protected
<float>Saving file /gnu/store/phykbdj3xl5kad2zhfxvmxkychjgv3ad-init.el...
<float>face-attribute: Attempt to save to a file that you aren’t allowed to write
<the_tubular>Sorry singpolyma, didn't see your mesage. Can you use guix system for creating a container image ?
<the_tubular>float, wow I didn't even know guix home could mange emacs
<float>the_tubular: yep, opalvaults[m] shared their dotfiles of it working with guix home.
<float>it works for me for the most part it's just that whenever I modify my config I have to answer that prompt which is annoying.
<the_tubular>Are you sure the link is correct ?
<float>well that's their home config file, there is a section on line 225 where they create a simple service to write their local emacs config into the proper config path and then symlink it to the store
<float>I managed to make it work, I had to use (setq custom-file (concat user-emacs-directory "/custom.el")) to make the custom variables be written in a separate file so they could be loaded later with "load-file" instead of being appended.
<sash-kan>hi all! why `guix graph` can't find path between bootstrap-tarballs and gcc?
<sash-kan>am i specifying the parameters incorrectly?
<podiki[m]>florhizome: I think he said to *not* have mkdirp set
<podiki[m]>or they rather (let me not assume)
<apteryx>sash-kan: bootstrap tarballs are static libraries that purposefully cut the dependency graph (being built statically and severing and reference to the storethey may have)
<podiki[m]>whoops, meant to @float
<apteryx>severing any*
<sash-kan>apteryx: but why then `gcc@7.5.0` contained in the output of `guix graph -M 3 bootstrap-tarballs`?
<alMalsamo>Hi all, I heard there is a special IRC channel on libera for bootstrapping from source discussions?
<janneke>alMalsamo: yes, we have the #bootstrappble channel
<alMalsamo>janneke: Nice! I am so excited about GNU Mes btw! I first installed GNU Guix in 2015 and I compiled everything from source but there were SO many bootstrap binaries!
<alMalsamo>It's been years since I used GNU Guix however, but I am massively looking forward to reinstalling it once I install an ath9k-supported Wifi card!
<opalvaults[m]>How do I escape guile procedures in configuration files?
<opalvaults[m]>for example, the package bind
<rekado_>opalvaults[m]: you can be explicit about the module, e.g. (@ (gnu packages foo) bind)
<rekado_>or you can explicitly hide or select certain values when importing
<rekado_>or you can import modules with a prefix
<janneke>alMalsamo: There have never been many bootstrap binaries
<janneke>Guix version 1.0 already had a relatively small closure of bootstrap binary seed: binutils, gcc, glibc, guile, and coreutils&co think: bash, bzip2, coreutils, gawk, grep, gzip, patch, sed, tar, xz
<rekado_>on aarch64 the test fails, because ‘service 'xorg-server' provided more than once’
<rekado_>that’s probably because we use sddm-service-type on all non-x86_64 now
<rekado_>and sddm-shepherd-service already provides xorg-server
<rekado_>and I guess we shouldn’t add gnome at all when the target system is non-x86_64
<alMalsamo>janneke: Hmm well I still consider that a lot... it's better than Gentoo's stage3 tarballs but it's still a lot of trust in opaque binaries signed by Ludo
<sash-kan>alMalsamo: > but it's still a lot of trust in opaque binaries signed by Ludo
<sash-kan>but what about mes? // isn't this part of the guix bootstrap procedure?
<alMalsamo>sash-kan: I know about GNU Mes now but the last time I installed GNU Guix on a Librebooted Thinkpad X200 was in 2015 and GNU Mes didn't even exist then!
<alMalsamo>I am excited to reinstall GNU Guix again soon, but it won't be with Libreboot this time unfortunately
<tissevert>hi guix
<efraim>rekado_: there's two or 3 bugs about it already. basically the sddm service isn't properly wrapped with the xorg-configuration
<efraim>but I agree about gnome on non-x86_64
<jason1234>I land here from recommendation from #gnu.
<jason1234>I am looking for a linux distro, that is 1.) Libre 2.) looks Unix 3.) without systemd 4.) without wayland (still with that old Xorg). Is Guix the right one?
<AwesomeAdam54321>jason1234: Yep
<rekado_>efraim: I have an ugly-looking patch for it
<rekado_>it’s a stop-gap really; we just add gnome conditionally in the example
<rekado_>I’m building this on an aarch64 node in a moment to test it
<AwesomeAdam54321>jason1234: About 2), the installed packages on the Guix System are put in the store, rather than in /bin,/sbin,etc. Apart from that, it's Unix-like
<eyJhb>What is the default umask on Guix?
<abrenon>hello, guix
<rekado_>jason1234: you can use wayland or xorg; not sure if this choice will remain with all combinations of desktop environments, but for now that’s what it is.
*rekado_ uses gnome with wayland
<federico>good morning everyone \o/
***federico is now known as fproverbio
<fproverbio>can someone share some insight on recursive dependencies in cmake based builds?
<fproverbio>the package i'm working on needs pybind11, which is available for guix, but of course refuses to use that and instead tries to git pull into /homeless-shelter/
<fproverbio>is there a known way to bypass that?
<AwesomeAdam54321>fproverbio: Does the package you're working on check that pybind11 is installed, prior to attempting to git pull?
<fproverbio>i don't think so, it tries to clone it without checking
<florhizome[m]><jason1234> "I am looking for a linux distro,..." <- You can definitely avoid wayland on guix , just not sure why you would ;)
<fproverbio>before that it checks for available compilers
<mothacehe>hey guix!
<fproverbio>mothacehe: hi!
<AwesomeAdam54321>fproverbio: I think the only way to fix this problem is to modify the build instructions to check if pybind11 is installed, and if so to skip the git pull step. These patches can be sent to the maintainers
<AwesomeAdam54321>Ideally, a package shouldn't need to download and install its dependencies itself, that's the job of a package manager
<fproverbio>AwesomeAdam54321: i see, i hoped there was a way to patch that via the package definition somehow
<AwesomeAdam54321>fproverbio: The patches can be in either the package definition or in the package itself, I just thought the latter would be more helpful to other people build from source without Guix
<AwesomeAdam54321>s/people build/people who build packages
<fiesh>jason1234: I use spectrwm with X -- doesn't get more UNIX than that ;-)
<sash-kan>hi all! is there documentation on ci/cd organization (with stages, for example, devel -> testing -> ready -> release) using guix?
<sash-kan>yes, i read, but i didn't understand if it was possible to make stages (for example, devel -> testing -> ready -> release) using cuirass.
<fproverbio>AwesomeAdam54321: it could be, but then again it might be harder to get the patches accepted upstream than implementing them on guix
<fproverbio>the concpetually easiest solution for me would be faking the git clone command that cmake attempts by providing the files where it is expecting them. This, however is unfeasable for me right know, as i know next to zero about cmake
***hollowman6 is now known as hollowman
<florhizome[m]><sash-kan> "hi all! is there documentation..." <- I know that dhruvin on sourcehut has made a guix integration into their builds ci service. Don’t know if that helps you though.
<rekado_>efraim: this would fix the test on non-x86_64:
<rekado_>too ugly?
<nckx>eyJhb: 0022
<jason1234>Does PKGSRC work on GUIX Linux?
<GNUHacker>guix GNU
<jason1234>fiesh: cool mine is too pretty much Unix: <-- nice to see taht unix alike on linux is still there. Will Wayland replace XORG soon?
<AwesomeAdam54321>jason1234: As far as I know, Wayland won't replace Xorg anytime soon.
<attila_lendvai>gnome-shell-extension-clipboard-indicator is nasty: by default it saves the clipboard history to the disk in cleartext, and the setting for turning it off is called "Cache only favorites". don't ask me, i tried:
<attila_lendvai>shall i report this as a security issue? where? to the mailing list?
<vivien>attila_lendvai, is it the same problem as when linux swaps memory pages to disk?
<PurpleSym>Do we have a policy for creating branches in the guix.git repo? I’d like to finally start merging the Python PEP 517 build system.
<vldn[m]>would be nice :)
<attila_lendvai>vivien, no really, because non-root users cannot read the swap file, and it is either encrypted, or overwritten at shutdown.
<attila_lendvai>i have added a native-search-paths entry to a package, and a `guix build mypkg` doesn't seem to rebuild it, just return the store path. am i missing something?
<attila_lendvai>this is not very helpful in the manual: "A list of search-path-specification objects describing search-path environment variables honored by the package."
<attila_lendvai>my ultimate goal is to make the gnome-shell-extension part of gpaste work, it's failing with "Typelib file for namespace 'GPaste', version '1.0' not found"
<ngz>You may need to wrap GI_TYPELIB_PATH around your executable
<yewscion>Good Morning, Guix!
<abrenon>hi yewscion
***aya is now known as gyara
<attila_lendvai>ngz, yeah, the native-search-paths thing is probably something else. there's no exe here, so i'm afraid i'll need to use some substitute* as i see in the udisks package
<phodina[m]>Hello there :-)... (full message at
<phodina[m]> * Hello there :-)... (full message at
<efraim>rekado_: I like it better than some of the other patches I think. also on the 'if' line you can use (target-x86-64?)
<apteryx>rekado_: about btrfs zstd, perhaps my 30% is a bit conservative; on one of my machines the compression ratio achieved is 53% for /gnu/store (TOTAL 53% 66G 123G 312G)
<PotentialUser-64>hi guys I tried to install 0ad
<PotentialUser-64>guix install 0ad
<PotentialUser-64>The following package will be installed:
<PotentialUser-64>   0ad 0.0.25b-alpha
<PotentialUser-64>guix install: error: integer expected from stream
<PotentialUser-64>what that mean?
<apteryx>I think that's a transient network related issue
<apteryx>try again, it'll probably work
<PotentialUser-64>No, it isn't. It work with other package
<ngz>You need to upgrade Guix (the daemon part)
<rekado_>PurpleSym: this sounds like a reasonable candidate for a branch.
<rekado_>we don’t have a policy, but I think it makes sense to push this to a branch and have build it all
<rekado_>it’s the easiest way to see if it breaks things
<PotentialUser-64>ngz do You mean a guix pull
<rekado_>PotentialUser-64: no, just ‘guix install 0ad’ again.
<KarlJoad-1>I am searching for a mailing list thread online. Can someone help me? I am trying to find the "Erase your Darlings" thread about a completely declarative system.
<sneek>KarlJoad-1, you have 1 message!
<sneek>KarlJoad-1, lilyp says: my emacs service is meant for user shepherd, not system shepherd
<PotentialUser-64>rekado_ Same error
<rekado_>apteryx: I’m confused about what these numbers mean :)
<rekado_>PotentialUser-64: perhaps something is wrong with the substitute itself; or maybe something’s wrong with your substitute cache.
<ngz>PotentialUser-64: I think so.
<rekado_>(I don’t know how to debug this)
<phodina[m]><PotentialUser-64> "guix install: error: integer..." <- There was some info lately in guix pull --news about fixing this issue
<rekado_>efraim: thanks for the hint about ‘target-x86-64?’; that certainly looks nicer
<PotentialUser-64>phodina[m] not in my guix pull --news'=(
<rekado_>efraim: do you happen to have the bug numbers for this failing test?
<gnoo>how do i view the logs of a service?
<gnoo>herd start wpa-supplicant gives: herd: failed to start service wpa-supplicant
<htsr[m]>mothacehe: How have you fixed gnome bluetooth?
<gnoo>but there's probably something in it's stdout/stderr
<PurpleSym>rekado_: Pushed to wip-python-pep517. That’s alot of emails to guix-commits… Will post an update to
<apteryx>rekado_: hmm, actually I'm confused too; I think the issue where hard links are not properly counted by compsize still exists, but here is the full table:
***wielaard is now known as mjw
<mothacehe>htsr[m]: yes but the fix is still in review
<apteryx>but basically btrfs + zstd leads to about 50% more capacity for /gnu/store
<snor>What's up guys! I got a backtrace when running `guix pull --debug=2` but it worked when running `guix pull`. Want to make a bug report somehow, but I am not sure how the bug tracking/mailing list works. Is there somewhere I can find some pointers, like is there some naming convention I should follow or so?
<htsr[m]>Ok I will wait, thank!
<apteryx>rekado_: another data point: last time my partition was full and I ran 'guix gc' on NVMe storage, it cleared 996 GiB (nearly 1 TiB) in 20 minutes; and that data was stored on a nearly full 466 GiB partition.
<apteryx>so supposing similar performance from a 20 TB drive, it should be able to fully clear it in less than 7 hours.
<jlicht>rekado_: thanks for making quick work of the texlive situation, currently building it locally to see if everything works :)!
***aya is now known as gyara
<rekado_>PurpleSym: I’ve added
<rekado_>PurpleSym: I don’t know if this will automatically start building or only on the next push.
<rekado_>I configured it to build everything for x86_64 only so as not to distract build nodes for other architectures.
<rekado_>apteryx: deleting /gnu/store/trash would likely be faster, but I suspect that there are some problems that only happen when you’ve got a huge ext4 file system with almost all activity happening in the same directory
<rekado_>so my guess is that ‘guix gc’ would still take an annoyingly long time to complete, longer than we would expect just by looking at disk performance.
<rekado_>but I do think we should act now if we want to change something about the situation :)
<rekado_>I’ve replied to the email discussion earlier today.
***aya is now known as gyara
<apteryx>rekado_: to be clear, I'm suggesting to move from ext4 to btrfs atop new SSD storage
<apteryx>(with zstd compression enabled)
*** sets mode: +o ChanServ
***jonsger is now known as 020AAL0BS
<sneek>Welcome back nckx :D
***020AAL0BS is now known as jonsger
<rekado_>apteryx: yes, I understand. In our quick tests btrfs performed not as well as ext4 wrt read and write speeds, though I guess btrfs might perform better with that huge /gnu/store directory.
<rekado_>(do we keep that huge directory or will be remove it?)
<gnoo>is profile-hooks not documented in the manual? how should i change it's value? in /etc/config.scm seems like the wrong place...
<apteryx>rekado_: I think it will still be needed for the time being
<apteryx>but cbaines seemed to have good ideas at improving things (as was done with the nar-herder) for the longer term
<anguixuser>Hi All :)
<notmaximed>The v5 from (the git-fetch updater patch series) seems ready, according to zimount and I
<notmaximed>It's apparently not yet on but you can grab a copy from (branch: git-fetch)
<notmaximed>(commit: 31d1d2089f07084f571bf062e23e2973490a0cca)
<dissent>hey guix, what is the difference between propagated, native, and regular inputs found in package definitions?
<notmaximed>dissent: See ‘package Reference’ in the manual for some explanation
<notmaximed>Feel free to ask for more details if things remain unclear
<notmaximed>* ‘package’ Reference
<notmaximed>You can read it with "info '(guix)package Reference'" in a shell
<dissent>notmaximed: thanks, will do.
<KE0VVT>Is there a guide on how to use the Emacs interface?
<daviwil>Hey folks, is there an official development build of the Guix Live VM image? I'm having some issues with updating to the latest Guix on the official 1.3.0 image that's available on
<daviwil>Basically, if I run `sudo guix system reconfigure /run/current-system/configuration.scm` it renders the VM unbootable for next restart
<daviwil>Trying to prepare a video that shows off `guix home` and `guix shell` so I wanted to base it off the Live VM so that other people can easily try it
<KE0VVT>daviwil: IIRC, Andrew Tobin is making a live image of his RDE setup.
<daviwil>That might be useful, but I'd prefer if it were an image that came from an official source so that people trying it out felt like they were actually trying Guix
<podiki[m]>what about an image from something like the desktop template configuration? although there is no prebuilt version of that as far as I know
<batalie>darn, i saw that anki had an upgrade but the fonts aren't quite fixed yet
<dissoc>i used tor network service with control socket but it no longer creates the domain socket file. anyone else have this problem?
<KE0VVT>M-x guix-all-packages seems to just lock up my Emacs. 😞
<batalie>oh wait, this does fix it
<batalie>i'll comment to that effect
<lfam>attila_lendvai: Did you still need help with understanding search path specifications?
<attila_lendvai>lfam, since then i added --with-typelibdir= to the #:configure-flags but i'll only get back to this tomorrow. starting a movie now, and winding down... thanks for asking, i'll find you later if it still doesn't work!
<lfam>Okay. I'll just say that search-path-specifications are something used when building profiles, not when building packages. That's why it didn't have any effect when building the package
<lfam>See ya later!
<attila_lendvai>lfam, re "if there were any preliminary patches that should be
<attila_lendvai>applied before" -- i don't know what you mean with this. only the fix is left, and it doesn't depend on anything unpushed anymore. the rest is unrelated stuff.
<attila_lendvai>lfam, makes sense, thanks! gn!
<nckx>absolutely no one:
<nckx>sneek: Welcome back nckx :D
<nckx>I must have done some good thing in this life. Morning, Guix.
<Kolev>emacs-guix does nothing but freeze my Emacs.
<rekado_>our package still takes the code from the old repo
<rekado_>didn’t we take over development and put it on savannah?
<Kolev>I thought emacs-guix was supposed to work and that people used it all the time.
<rekado_>this one:
<nckx>Not sure about freezing but g-a-p takes a long time to run here (in fact it's still running).
<rekado_>rather than
<nckx>Interestingly, all 100% CPU is emacs.
<rekado_>looks like the one on gitlab is again up-to-date
<rekado_>oh well
<Kolev>rekado_: Guess I'll stick to the command line. I'm not checking out experimental stuff.
<rekado_>I stopped using it when it got separated from the rest of Guix
<nckx>strace: it is reading all the things.
<nckx>I don't use it either.
<Kolev>Separated? ;asdkjf;ajf
<rekado_>experimental? aoeuhts
<Kolev>It's no longer a part of Guix. :-(
<Kolev>rekado_: I want to stay on Dvorak, but it lacks special characters that are found on UK QWERTY.
***Kabouik_ is now known as Kabouik
<Kolev>I guess I can use Compose < - instead.
<civodul>daviwil: you can build a live image with something like "guix system image config.scm"
<rekado_>I’ll update the guix package to benefit from the fixed tests.
<opalvaults[m]>hey all, I'm running into an issue where nm-applet is crashing despite having setuid set in system configuration. before it crashes it gives me:
<opalvaults[m]>(nm-connection-editor:4343): GLib-GIO-ERROR **: 12:45:32.063: Settings schema 'org.gnome.nm-applet.eap' is not installed"
<opalvaults[m]>Is there a package that provides that schema?
<opalvaults[m]>i'm not using gnome, only GDM
<opalvaults[m]>unsure if that's creating an issue somewhere
<the_tubular>Anyone has a wireguard service config that creates a VPN ?
<the_tubular>I know that the service exist, just unsure on how to use it
<daviwil>civodul: yeah, I was hoping I wouldn't have to, but not a big deal I guess. Thanks!
<robin>the_tubular, i'm curious about that too, i think i got the wg-quick script working on guix at some point but that's obviously not as good as a service
<jackhill>the_tubular: is there something specific you're stuck with for the service? I'm using it on one host like so,
<jackhill>On all my other hosts, I'm configuring wireguard with network manager
<the_tubular>Got it, i'll try that. Might not be today, as I got to go soon, but that definitely looks like what I want.
<the_tubular>Thanks :)
<jackhill>hmm, speaking of VPNs, looks like it might be a neat package/service to have 🤔
<robin>Kolev, if you use x11 it's fairly easy to make a customized dvorak layout with xkb (idk how you'd integrate it on a system level, but guix keyboard layouts *are* all converted from xkb layouts afaict...)
<robin>i made a (US-dvorak-based) "dvorak-lisp" layout where () and <> are switched, for instance, which is only a 10-line config file
<robin>i'll paste it in case it's useful to anyone, as i found it a bit tricky to figure out initially compared to xmodmap
<nckx>opalvaults[m]: The subtletly here is that merely adding a binary to setuid-programs does not ‘install’ it, it does not do all the other work of setting up search paths & propagation &c. which are needed here. I suspect that adding network-manager-applet to your system packages list *in addition* to the setuid entry will make it work.
<nckx>If that doesn't quite fix it something similar definitely will.
<opalvaults[m]>nckx: thanks for the suggestion, unfortunately I've added all of NetworkManager, nm-connection-editor, and network-manager-applet to packages and set them up for setuid and it's not working.
<opalvaults[m]>In fact I fear that installing one of those packages has broken Gnome. none of the gnome icons, gtk, etc. exist when I log into Gnome.
<nckx>It's possible. I just tested it with network-manager-applet and stalonetray here; I don't have the full GNOME stack.
<opalvaults[m]>This is nearly a fresh install with Gnome set up by the installer.
<robin> <- example xkb symbols file for swapping around some keys + a script to load it under x11 (which is run on login, and also gnome resets the layout when you (un)plug input devices)
<nckx>nm-applet should probably be part of the default ‘GNOME experience’ anyway, people expect that.
<opalvaults[m]>nckx: do you get a segmentation fault when you use nm-applet from the cli?
<opalvaults[m]>yeah default i3/Gnome installation the nm-applet segfaults despite having setuid on it
<nckx>Context menu works.
<nckx>It is not setuid however.
<nckx>I'm pretty sure this is complicating matters, although I could reproduce and ‘fix’ your error above without making it setuid.
<nckx>Just by installing/uninstalling it & invoking it directly from /gnu/store.
<opalvaults[m]>interestingly, I figured being a part of the group netdev would even allow me to use any of the networking stuff but I need superuser access for all networking still
<opalvaults[m]>heres my setuid
<nckx>The /gnu/store/vm7q5kz3w15w3xj8398bcjy00bcjxi8q-network-manager-applet-1.22.0/bin/nm-applet invoked from a terminal, running in a crappy stalonetray, just successfully dis- and and reconnected my Wi-Fi, no setuid.
<nckx>I am in the "audio" "disk" "kmem" "kvm" "lp" "lpadmin" "netdev" "root" "video" "wheel" supplementary-groups.
<nckx>I'm pretty sure I can do anything I'd want from this pop-up menu without setuid, although I won't test removing & re-adding my home wi-fi network now.
<opalvaults[m]>running it from gnu/store still aborted
<opalvaults[m]>it gave huge amounts of errors too
<opalvaults[m]>If I run it as root (nm-applet:2678): GLib-GIO-ERROR **: 13:32:37.278: Settings schema 'org.gnome.nm-applet' is not installed
<nckx>Right, that's the error I got before ‘properly’ guix installing network-manager-applet into a profile, so it can get its propagation on (see the comment in the package definition; it needs to).
<opalvaults[m]>I have it listed in packages and I only install packages through guix home reconfigure
<opalvaults[m]>or via guix system .scm
<nckx>Unfortunately I cannot support guix home.
<nckx>~/.guix-profile/share/icons/Adwaita/16x16/status/image-missing.png exists here and is a perfectly valid PNG.
<opalvaults[m]>well I don't think this is Guix Home's fault ;P
<notmaximed>The v5 from (the git-fetch updater patch series) seems ready, according to zimount and I
<opalvaults[m]>My entire Gnome DE is broken (I'm using i3 rn), and all gtk programs segfault.
<nckx>No, the fault is hiding somewhere in the vast space between our hugely different environments, unfortunately :)
<notmaximed>It's apparently not yet on but you can grab a copy from (branch: git-fetch)
<notmaximed>(commit: 31d1d2089f07084f571bf062e23e2973490a0cca)
<nckx>I use Sway with very little GNOME support stuff (the bare minimum to make things work).
<notmaximed>sneek: later tell civodul: git-fetch patches appear ready according to zimoun and me:
<notmaximed>(or anyone with commit privileges)
<notmaximed>sneek: botsnack
<nckx>opalvaults[m]: Is your ~/.guix-profile/share/icons/Adwaita/16x16/status/image-missing.png sane?
<opalvaults[m]>one second let me check
<notmaximed>sneek: botsnack botsnack botsnack?
<notmaximed>sneek apparently takes only one snack at a time.
*nckx goes to poke issues.guix with the sysadmin stick.
<opalvaults[m]>nckx: lrwxrwxrwx 1 root root 110 Dec 31 1969 image-missing.png -> /gnu/store/zwrx3n379whnxfdm9n9l1gr4lia48lxz-gtk-icon-themes/share/icons/Adwaita/16x16/status/image-missing.png
<nckx>But is a pretty /!\
<nckx>*is it
<opalvaults[m]>the image loads properly as well
<nckx>This is beyond my GNOMErstanding.
<opalvaults[m]>dang..I might have to go to another distribution at this point for the time being. Basic functionality keeps breaking such as changes to/from display managers, installing random packages stripping gnome of all of its dependencies, etc.
<opalvaults[m]>all good, thanks again for your help nckx
<nckx>All I helped was illustrate there's a way to make it work, the binary isn't fundamentally broken or anything. Oh well. I tried.
<opalvaults[m]>how do I make it work?
<nckx>That is the several-dollar question?
<opalvaults[m]>No I mean you said you illustrated a way to make it work and I must have misunderstood something
<nckx>No, sorry, I just meant it works fine for me, so it's not an issue with Guix's nm-applet binary or so, but clearly some error in the exactl composition your environment that mine lacks.
<nckx>* -l
<opalvaults[m]>the packages are a part of my profile, setuid is correctly for apps, my configuration file works fine.
<nckx>So nm-applet is setuid, right? Is it designed to be? Does it need to be?
<opalvaults[m]>no, but it didn't work whilst not setuid either :\
<lfam>I use nm-applet on debian and it is not setuid
<nckx>I'd still remove it, it's almost certainly wrong.
<lfam>I wonder if the nmtui setuid example in the manual is similarly misleading?
<opalvaults[m]>I can't even post a screenshot of what I'm seeing because no GTK stuff works :(
<opalvaults[m]>so icecat crashes
<nckx>If it was ever nm-related to start with it's certainly outgrown it now.
<nckx>If this were my system I'd remove *all* later additions first, and ensure the systems works as such.
<podiki[m]>nckx: were you just noticing also that seems a bit behind/stuck?
<opalvaults[m]>I'm trying trying to rack my brain, this is my 3rd re-install due to this issue of everything breaking
<nckx>podiki[m]: notmaximed pointed it out.
<podiki[m]>ok wasn't just me (thought it was a little slow but still hasn't updated from a few hours ago)
<nckx>No, definitely real. Thanks for reporting it too.
<podiki[m]>"reporting", no problem!
<lfam>A few hours ago I tried to view <> but it did not load
<podiki[m]>I just thought about it, if only it was that easy
<nckx>I'm doing as good a job of ‘fixing’ it so we make a great team.
<lfam>It seems to work now, but I thought it was weird
<nckx>Not related but for the record, any 50x now is just me playing with the service.
<opalvaults[m]>Is there any information I can provide to help someone get a picture of whats going on and help right the system?
<opalvaults[m]>Or would it be best if I re-installed?
<rekado_>lfam: loads for me, but it’s unusually slow
<lfam>It's a really big patch series rekado_
<lfam>opalvaults[m]: Guix System can be recreated more or less exactly based on your config.scm and the `guix describe` used to instantiate the config.scm. However, there is always the possibility of complications caused by using shells besides Bash, things in shell startup files, other stateful (mis?)configuration in user dotfiles, Guix Home, etc
<lfam>So, it really depends on how much effort you want to put in to debugging
<opalvaults[m]>lfam: I'm willing to put in any amount of effort. I just don't know where to start at this point.
<opalvaults[m]>Here is my home file:
<lfam>You'll probably need to find a debugging partner who understands Guix more fully
<lfam>Many of us haven't used Guix Home yet, so it's not something we can meaningfully assist with
<lfam>Add on the trickiness of 3rd party channels, and it really will take an expert
<opalvaults[m]>lfam: This is unfortunately not quite the case as two laptops with the same guix describe, and 1 failed to go from SDDM->GDM yesterday to match the laptop with GDM already installed.
<lfam>Looking at guix-home.scm, I have to wonder how much effort has been spent so far ironing out issues with Zsh
<lfam>I use zsh when convenient, which means I don't use it on Guix System
<nckx>Mumi logs a lot of GC Warning: Repeated allocation of very large blocks.
<opalvaults[m]>lfam: zsh has an officially supported service configuration in the manual correct?
<lfam>opalvaults[m]: What is not the case?
<opalvaults[m]>That two systems with the same guix describe will be reproducible
<nckx>Mumi is back up to dating.
<opalvaults[m]>If channel revisions are the indication
<nckx>rekado_: I think what did it was killing the hung-looking rsync process.
<lfam>opalvaults[m]: It's very much the case. However, like I pointed out, stateful configuration (dotfiles) at the user level complicate the situation
<nckx>If so, perhaps it should be run with timeout(1) or (if it's invoked from Guile somewhere) the Guile equivalent.
<nckx>Or just rsync-native timeout options. I don't know if they would have helped here.
<lfam>Zsh may be in the manual. But I'm telling you: Bash has first class support on Guix.
<nckx>I just kill stuff. Don't ask me how anything works.
<opalvaults[m]>lfam: I'm curious what Guix's idea of reproducibility is then.
<rekado_>nckx: it’s running in a while loop
<opalvaults[m]>I just tested this yesterday with the same exact system and one failed to go from SDDM->GDM
<nckx>Ah, the nckx style of critical service management, nice.
<rekado_>yes, the very same
<lfam>opalvaults[m]: Let me put it this way. You could copy a Debian installation bit-for-bit to another computer of the same model, and make only one change: A line in ~/.bashrc that does `rm -r $HOME`
<lfam>Would you expect them to operate in the same way?
<nckx>I feel much less bad about not finding the service anywhere now.
<lfam>My point is that I doubt the two systems are actually the same opalvaults[m]
<opalvaults[m]>lfam: No, but then again I'm not sure I follow. There is nothing in any of my zsh files except for variable bindings and imports for aliases
<lfam>It's just an example of how there are things not accounted for by Guix, and that can't really be accounted for
<rekado_>mumi will improve (there are a few patches by Arun for a graphql API that I’ll apply soon), and I hope whatever surrounds it will too
<lfam>And it's like, some people seem to have no trouble at all with Guix System, and for some people it's just a terrible buggy experience
<lfam>My intuition is that there are some "happy paths" when using Guix System, and the latter group are not on the happy path, but rather are exploring parts of Guix that are not well-trodden
<lfam>Guix Home is probably still somewhat immature
<nckx>rekado_: Nothing of what I said was meant as serious criticism. I am amused, nothing more :)
<lfam>It was added to Guix before the code review process was complete. What happened there, I do not know
<opalvaults[m]>lfam: Well I'm not sure what I'd do then?
<rekado_>nckx: no, I get it :) It is amusing to me too.
<opalvaults[m]>lfam: that's a pretty dismal outlook for me, one who has had a terribly buggy experience.
<lfam>Yeah I know
<rekado_>with the updated Guix package: FAIL: tests/home-import
<rekado_>guess I won’t be pushing that updated version then
<lfam>What works for me with Guix System is to be somewhat conservative about what I handle via the declarative configuration. That means I'm not using Guix Home yet, I don't use 3rd party channels (the kernel channel seems to work reliably), I just keep it simple
<opalvaults[m]>If I have two systems, and I describe to Puppet how I'd like these systems to be, the two systems will be the exact same. Actually I'd say Guix has a lot in common with Puppet except that reproducibility is inconsistent which is a huge deal for someone like me trying to create such an environment between all of his systems. Wouldn't you agree?
<lfam>Guix is definitely in the stage of development were you have to hold it correctly
<rekado_>oh, I’ve never thought I’d see a comparison between puppet and guix where puppet is associated with the better outcome.
<rekado_>I (and my colleagues too) never got puppet to build any system reproducibly
<lfam>I have no experience with Puppet so I can't say much about the comparison
<rekado_>there’s state everywhere
<rekado_>order matters
<rekado_>adding and removing an instruction is not the same as not having had the instruction in the first place, etc
<opalvaults[m]>rekado_: we run a debian farm and puppet works just fine? I'd be curious how you'd actually get inconsistent application between 2 servers
<rekado_>I guess the puppet needs to be held a certain way :)
<lfam>I think this conversation is instructive. For both tools, there are people who can use one effectively, but not the other
<opalvaults[m]>rekado_: how so?
<opalvaults[m]>But this isn't about me using the tool effectively is it?
<rekado_>we’re using it for all the HPC cluster, group servers, web servers, workstations, etc; different OSes too.
<lfam>Wasn't your goal to use Guix?
<rekado_>by far the biggest problem for us with puppet was stateful deployment
<rekado_>by far the biggest annoyance with Guix for me is application caches
<lfam>By the way, I'm sorry if that was hurtful. I'm basically in the same boat. I use Debian primarily because I still find it hard to achieve my goals with Guix System
<rekado_>(e.g. the ibus cache that gets stale whenever I upgrade the system)
<opalvaults[m]>lfam: not at all. I think I'm going to have to migrate away as well and just use it as a package manager.
<lfam>Yeah, there is a lot of state at the application and user level in Guix and it can definitely cause problems
<opalvaults[m]>I think my professional experience using configuration management is marring my experience of Guix when it breaks out of my control.
<opalvaults[m]>(or in my control, who knows?)
<rekado_>I haven’t been following the whole conversation; is there a bug report about the problem you’re having? Are there enough details for us to analyze it?
<opalvaults[m]>rekado_: Good question. I've made a bug report about this yesterday. you missed it, but above I asked if there was any information I could provide to help someone to help me figure this one out
<opalvaults[m]>rekado_: I'm also stuck in a weird spot because I can't open any GTK applications as it crashes Icecat so I can't upload screenshots atm :P
<opalvaults[m]>I'm happy to help provide information :)
<lfam>I would unset XDG_DATA_DIRS and GIO_EXTRA_MODULES and try again
<rekado_>opalvaults[m]: oof, that’s terrible.
<lfam>When ... something in the environment ... related to GTK isn't set up right, I have to unset those variables until the environment can be fixed
<rekado_>these are good suggestions; another common culprit is LD_LIBRARY_PATH.
<lfam>I experience this when mixing GTK from Guix and Debian
<lfam>My hunch is that a similar state mismatch has occurred on your Guix System
<lfam>Either from Guix Home or because of some major version incompatibilities between the system and user level
<lfam>I'd hope there was no LD_LIBRARY_PATH set
<opalvaults[m]>lfam: I will unset them in all shells, reboot and go from there
<lfam>Wait, you don't have to do all that
<lfam>Unsetting these variables will only persist for as long as the shell is open
<lfam>Just unset them and try to run your program (icecat?) again
<lfam>Just unset them and then launch icecat from the same shell
<lfam>It's just a guess anyways
<opalvaults[m]>my misunderstanding, apologies
<lfam>Concretely, you could try running this: `unset XDG_DATA_DIRS; unset GIO_EXTRA_MODULES, icecat`
<lfam>That comma must be replaced by a semicolon
<opalvaults[m]> -- Icecat segfault
<opalvaults[m]>guix package --list-installed -- packages installed (just in case it was a case of gtk/mime-shared-info not being installed correctly)
<opalvaults[m]>lfam: I will try that suggestion as well one moment
<opalvaults[m]>oops, forgot link
<opalvaults[m]>unset XDG_DATA_DIRS; unset GIO_EXTRA_MODULES, icecat
<opalvaults[m]>dangit :(
<opalvaults[m]>same segfault with your unset...;icecat command as well. hmm
<opalvaults[m]>last time this happened, it also took with it the binary gsettings
<opalvaults[m]>so if there is anything trying to access that api its going to fault likely
<rekado_>this error … it looks familiar
<rekado_>bug #38544
<opalvaults[m]>rekado_: I will try with that sudo -E /gnu/store...
<lfam>I just installed the icecat package using Guix on Debian and it runs fine for me
<lfam>I'm not sure why it cause any trouble, but why is GTK 4 installed in your profile?
<rekado_>opalvaults[m]: running with sudo is probably not going to help
<rekado_>but the problem is the same: the program doesn’t find an image loader, but it shoudl
<opalvaults[m]>lfam: apollyon:~:% guix package --list-installed | grep gtk
<opalvaults[m]>gtk 4.2.1 out /gnu/store/1k1ard2r1cda76br7acz92wsb1if25ri-gtk-4.2.1
<lfam>And running it from /gnu/store will definitely not help
<lfam>For any complex application, Guix sets up a lot of necessary things when you install the package
<rekado_>in the case of bug #38544 the solution was to wrap, which is what the glib-or-gtk-build-system does
<lfam>Just executing the binary from the cache is not something useful
<lfam>opalvaults[m]: Yes, I was asking why GTK 4 is installed
<lfam>That's unusual
<rekado_>I think I also saw a very similar problem with rstudio, I think.
<lfam>You could probably remove that unless you have a reason to keep it installed
<rekado_>the fix for rstudio was to wrap the executable in GSETTINGS_SCHEMA_DIR, pointing at "/share/glib-2.0/schemas" of the gtk+ package
<opalvaults[m]>oh lfam , sorry for misunderstanding you. I installed gtk because I was seeing if it would fix the gtk errors.
<rekado_>IIRC it was a search path ordering problem
<lfam>Huh rekado_
<rekado_>rstudio doesn’t even use GTK
<lfam>opalvaults[m]: Okay. Either way, Icecat uses GTK 2 and 3, but not version 4
<rekado_>granted, this was on a foreign distro, but the problem was that the application picked up some glib schemas that are incompatible and then crashed somewhere down the line
<opalvaults[m]>to your point , I'm now going to uninstall packages installed with `guix install`, and remove some of these added $PATHS just in case.
<lfam>Does icecat currently work for *anyone* on Guix System?
<lfam>It works for on Debian
<rekado_>works for me
<lfam>For me
<podiki[m]>icecat works here (just tried)
<lfam>Guix should manage $PATH properly
<podiki[m]>I'm on bare WM for what that is worth (so I do more work on my side for setting things up with gnome etc)
<rekado_>PATH is not an issue here. But search paths in general can mess things up.
*rekado_ has to leave
<opalvaults[m]>take care rekado_ thanks for the help!
<podiki[m]>lfam: thanks for the quick push for syncthing-gtk; been on a patch frenzy lately trying to sort out all the little things and submit my packages I've had locally
<lfam>podiki[m]: Keep 'em coming
<lfam>And feel free to ping me with similar fixlets
<podiki[m]>I think the rest I have in the queue are all new packages (openrgb, mangohud, xinitrc-session) and an update to xmonad (that Lars chimed in on, I might need to revise that one)
<lfam>Okay. Feel free to ping Lars; he will be able to commit it for you
<lfam>I think we are all overwhelmed by the volume of email, so patch submitters have to be persistent in order to land their patches.
<podiki[m]>yeah, was non stop with the core updates too
<potatoalienof13>Does guix allow compiling some packages against glibc and others against musl?
<podiki[m]>lfam: what do you think the medium (or long) term solution is? more reviewers or committers? or some better automated tools (branches building pending patches as a first check that sort of thing)?
<lfam>Hm, it's really big topic
<lfam>I'm not sure there is a "solution" in the sense that I don't think it's a problem that can be solved
<lfam>The situation can definitely be improved in various way, or it could even get worse
<podiki[m]>right, improved, or more manageable
<lfam>There is some early discussion about trying to formalize the structure of the community. I think we are at the point where that is necessary
<lfam>But it's not an easy subject
<lfam>For sure, more reviewers would help. But it's hardly a solution, because how do you get more reviewers?
<singpolyma>More committers is probably part of it, but it takes time for people to land enough patches to build trust and such
<lfam>We have a lot of committers but a relatively small number of committers do most of the code review
<lfam>For example: `git shortlog --numbered --summary --committer v1.3.0..HEAD`
<singpolyma>lfam: ah, well, that's no ideal for sure
<lfam>We ask people to apply for commit access, but in reality most people have been invited. It seems that people don't feel confident to apply, or don't think it would improve their experience with Guix
<singpolyma>I would love to help with review someday, once I can land commits without making obvious errors myself, heh.