IRC channel logs


back to list of logs

<pkill9>I want bonzibuddy
<TopExpert>found one with a girl dancing by the clock iirc, it was the most VR thing ever
***stikonas_ is now known as stikonas
***taiju is now known as Guest540
***taiju` is now known as taiju
<polyex>so rust is going in linux 6.1, libre-linux/guix gonna be ready to receive the glorious benefits?
<apteryx>sneek: later tell civodul looks like it'll be faster to plug an extra storage device (in a pinch, a USB key will do) to balance overdrive1.
<sneek>Will do.
<Aurora_v_kosmose>If I were to package the i2p daemon, would it be preferable to manually move & modify files or just build the installer and use it with the appropriate output location? Although the feasibility of that last one depends on whether Guix/Guile has anything like Expect.
<rekado>there was a campus-wide power outage at the MDC campus
<rekado>I don’t know if our servers were affected
<rekado>some power supplies were shut off
***LispyLights is now known as Aurora_v_kosmose
<jpoiret>i got substitutes from bordeaux this morning so was not really impacted
<rekado> is online, but I can’t ssh into it through the usual bastion server on campus.
<rekado>many switches are still offline
<PotentialUser-70>Good morning! I have just downloaded `guix-system-install-1.3.0.x86_64-linux.iso` and ``. But when doing ` wget -qO - | gpg --import -`, I get the error message "zsh: no matches found:
<PotentialUser-70>, gpg: no valid OpenPGP data found." What can I do. Thanks a lot!
***rgherdt_ is now known as rgherdt
<rekado>I cannot reproduce this.
<PotentialUser-70>Thanks for your prompt answer, rekado! Maybe I'm doing something wrong? 1) Change into the directory where the `.iso` and the `.iso.sig`files are. 2) `gpg --verify guix-system-install-1.3.0.x86_64-linux.iso.sig` returns: "gpg: Can't check signature: No public key". 3) `wget -qO - | gpg --import
<PotentialUser-70>-` returns the error message mentioned in my first post. Is there some mistake in the commands I have entered?
<cbaines>PotentialUser-70, the documentation is probably written assuming bash, but you're getting a zsh error
<cbaines>I'd perhaps try quoting the URL given to wget, it might be that zsh is trying to do something with it#
<cbaines>once you've got that bit of the command working and can download the PGP key, try passing it to gpg
<pkal>Is there a reason that "guix import" doesn't generate the necesary use-module statements?
***Dynom_ is now known as Guest8698
<PotentialUser-70>Thanks a lot for your tips, cbaines! I have switched to bash and quoted the URL, and now it works! The output says, among othe things: "key 1260E46482E63562: 2 signatures not checked due to missing keys ... key 1260E46482E63562: public key "Maxim Cournoyer <>" imported ... no ultimately trusted keys found". Is this ok? I'm
<PotentialUser-70>asking because the wording of the last message differs from the one mentioned in the instructions: “This key is not certified with a trusted signature!”. Thanks again!
<cbaines>PotentialUser-70, the main thing to check is that GPG thinks the binary tarball is fine
<PotentialUser-70>Okay, so I can proceed with the installation. Thanks again for you help! I have read about Guix System in the October issue of the German "Linux User" magazine and am looking forward to trying it out. I have been using Arch Linux for many years, but it's always interesting to try out potential alternatvives :-)
<pkal>There was an article about Guix? Can I read it online somewhere?
***wielaard is now known as mjw
<zimoun>jgart[m]: hi, your mailbox seems full.
<bost>Hi. Which package contains the `ss` binary for socket statistics (like netstat)?
<fnstudio>ok, here's a weird one, do you think i can install sway via guix as a non-root user on my machine (a foreign distro installation)
<fnstudio>oh well, maybe i can... but do you think i'll be able to launch it from a TTY and will it work?
<jpoiret>bost: iproute2
<jpoiret>i often look at the manpages on the internet to find out which project it is part of
<davidl>I have a luks-encrypted disk with a btrfs partition Im trying to install on, but it won't boot after and I have no idea why.
<davidl>Can someone see what's wrong with this config:
<florhizome[m]>pkal should be this Magazine but the octobre issue doesn't seem to be available here yet
<bost>jpoiret: thanx. (I had a look at the man page, too, but haven't seen it, sorry.)
<bdju>could someone please update the fzf package? I could paypal you $20 or something
<bdju>my history search in zsh is broken and updating to the latest version should fix it
<bdju>I might be able to work around something with rollbacks but I'm not actually sure which package broke it (clearly not fzf as our package is over 1.5 years old) or when it broke, as any old shells I have work and it's just sourcing the new config or opening a new shell that gets me the broken ones
<bdju>something in the recipe would need to be tweaked as I already tried and failed to use the --with-commit= thing
<bdju>I have like 3 or 4 fzf versions on my system but I don't really know how to make it use the newest one for this functionality
<sughosha>Hi everyone, is it possible to add a folder to runpath while packaging?
<pkill9>sughosha: what do you mean runpath? liike where it searches for libraries?
<sughosha>like the same way how `patchelf --set-rpath`, but append new folder.
<pkill9>ah yeah you can use pathcelf
<pkill9>idk if that overrides the existing runpathsx
<sughosha>Yes, it overrides, I tried it.
<pkill9>but you could also wrap the binary with LD_LIBRARY_PATH which will tell it to look for libraries in specified paths in addition to hardcoded search paths
<sughosha>But the problem is there is also a .so lib file, which I probablu cannot wrap like that.
<lechner>Hi, I know about search-input-directory and this-package-input. Is there something similar for outputs?
<polyex94>does guix let me pick which specific version i wanna install of a pkg? or just whatever version is latest?
<lechner>polyex94: i think you are generally stuck with whatever is presently packaged
<polyex94>that's kinda fucked up. any fix coming?
<polyex94>look at package managers like cargo. they let you get any crate version that isn't yanked due to security
<lechner>polyex94: which package?
<polyex94>just in general
<lechner>polyex94: for many user-facing packages you can run whichever version you like by adapting the packaging code. my comment applied primarily to prerequisites for other packages
<polyex94>so if i upgraded my system and let's say a browser update made it stop working, i could modify the system config to make it install the prior browser version instead?
<rekado>polyex94: you can use “guix time-machine” to install any package from the past.
<rekado>in the Guix world we don’t consider version numbers to be all that important or descriptive. What matters is *everything* that went into producing a package’s output.
<rekado>and all that is not captured by the version string but by the complete state of Guix at the time
<rekado>so naturally our solution to accessing older package versions is to access them through the lense of Guixes past.
<lechner>the reliability gains of that approach are legendary
<rekado>of course this can lead to problems of its own: since these old packages were built with older toolchains and older libraries you may not be able to use them directly in a modern environment.
<polyex94>ya, sure
<rekado>for cases like this we would transplant old packages into the modern world; but that’s done in a separate channel: Guix-Past
<pkill9>i like seeing all these different uses of guix channels
<remyd1>Hi,I tried to add some channels and after doing `guix refresh`, I get the following error:
<remyd1>So, I commented github channels, but I still have this issue
<davidl> seems down
<remyd1>@davidl Indeed... Do you think it is related to my issue ?
<davidl>remyd1: probably is also not serving substitutes currently
<remyd1>hum, ok
<remyd1>So, now `guix pull` is stuck... (I launched it before you said it was down :/ )
<lechner>Hi, i know ci.guix.gnu.o is down, but why do i get substitution of /gnu/store/7cl0bgi0i84317r6nr04ny0nfjzb5wak-dosfstools-4.2 failed even with --fallback, please?
<cbaines>lechner, fallback means try building if fetching substitutes fails
<cbaines>--no-substitutes means don't try substituting
<attila_lendvai_>is there any ETA on when the substitute servers come back?
***attila_lendvai_ is now known as attila_lendvai
<cbaines>attila_lendvai, it looks like MDC hosted services are down, everything else should still be up, including
<rekado>there was a campus-wide outage, and some of the switches are out of order
<attila_lendvai>should it fall back automatically? because i'm getting a 3 out of 3 failures at the same package (cdrkit-libre-1.1.11)
<lechner>cbaines: but just before that, it said guix home: error: some substitutes for the outputs of derivation `/gnu/store/y8c9qgfrx69r0b8kikf1g556p5bz00aq-dosfstools-4.2.drv' failed (usually happens due to networking issues); try `--fallback' to build derivation from source
<rekado>apteryx: are you around?
<cbaines>lechner, right, so maybe it should say "try --fallback to enable building the derivation from source", but that still seems like the sensible option to suggest
<rekado>nodes 128 and 129 are online, and they have been online for days, so my guess is that the power supply of our rack hasn’t been affected
<rekado>maybe one of the switches is offline
<attila_lendvai>the fallback mechanism -- if there's any -- seems to be failing, because things start working if i add --substitute-urls=
<apteryx>rekado: hi! yes
<rekado>apteryx: are you able to connect to the serial console on
<apteryx>yes -- through hydra-guix-129
<rekado>I can connect to 129, but not to
<rekado>.41 is fine, .40 is not
<cbaines>attila_lendvai, if you can share the full output (via and maybe email as well, that would be useful
<rekado>uptime on 129 also suggests that the rack stayed powered through the outage.
<apteryx>rekado: I'm at the racadm>> screen
<apteryx>ah, I remembered what to type now: 'console com2'
<rekado>but no response from the console?
<rekado>(it was blank when I tried it)
<apteryx>ah, right. I see the same.
<rekado>can we reboot it safely?
<lechner>cbaines: hi, sorry to be so obtuse. why doesn't it try to build the derivation from source even though --fallback was given, please?
<apteryx>nckx: you around? we may need your bootloader surgeon hands
<cbaines>lechner, I'd expect it to, if it doesn't could you share the full command+output via
<attila_lendvai>cbaines, i've sent it to bug-guix, too
<apteryx>civodul: re overdrive1, I suggest to add an extra storage device, 'btrfs device add $device /', then run balance, then 'btrfs device remove $device /'
<lechner>cbaines: here you go
<apteryx>the device can be a simple flash thumb drive, although if you have something faster that'd speed up the balance
<civodul>apteryx: i'm missing something; it's 96% empty, can't we do better?
<viaken>pkill9: kakoune has clippy integrated.
<civodul>are there hidden snapshots or something?
<civodul>reminds me of defrag on MS-DOS (yeah, i'm getting old)
<Unthahorsten>We have problem reaching with "warning: connection failed: No route to host" and "error: corrupt input while restoring archive". Do you have the same problem ? What could be done about it ?
<rekado>Unthahorsten: we’re aware of it.
<rekado>apteryx: FYI I’ll be mostly away for the next 7 hours or so
<rekado>in the meantime maybe you could mount the SAN on node 129 and swap the DNS record for to .41
<rekado>so that at least substitutes can be served
<Unthahorsten>rekado: Ok good luck to solve this problem
<lechner>cbaines: or this one, for system
<tricon>civodul: (source code:
<apteryx>civodul: oh, it has 47.61 GiB unallocated now, that's new! Device unallocated: 47.61GiB
<apteryx>per 'btrfs fi usage /'
<Unthahorsten>rekado: Thanks
<civodul>apteryx: that's 5%, whereas "df" says 96%
<civodul>apteryx: but maybe i'll let you focus on ci.guix for now; i'm chatting on #btrfs
<civodul>(happy to give a hand on ci.guix as well if there's anything i can do)
<apteryx>rekado: so is the infrastructure back in order, but berlin still unreachable?
<apteryx>otherwise I'd suggest we just way for the infra to come back up
<rekado>apteryx: there was an outage early in the morning, but things have come back up as far as I know
<rekado>it was only much later that stopped responding
<rekado>this suggest that it’s on us
<rekado>and none of the other nodes in the same rack are affected
<Luk6655>anyone needing the guix manual, here is a cached copy (as is down)
<apteryx>civodul: I'm attempting this: 'sudo btrfs balance start -dusage=5 /' on overdrive
<pkill9>ah yes kakoune
<civodul>Luk6655: alternatively, one can run "info guix" locally
<civodul>apteryx: alright!
<pkill9>I want a desktop wide 3D bonzibuddy
<pkill9>maybe suitable as a wyafire plugin
<Luk6655>civodul: thanks
<civodul>apteryx: do i get it right that you can access berlin via the serial console?
<apteryx>civodul: seems to be doing stuff, can be followed via /var/log/messages or 'btrfs balance status /' -> 37 out of about 730 chunks balanced (53 considered), 95% left
<apteryx>and unallocated space is growing
<cbaines>lechner, right, maybe there's no fallback to other substitute servers if an error is encountered
<apteryx>w.r.t. berlin, I normally can, but the serial returns nothing currently, which suggests the machine is completely hung up
<civodul>it's not like we've had a week to reconfigure + reboot
<civodul>can we hard-reboot it?
<rekado>yes, we can always reset it via idrac
<rekado>but I have to leave now
<civodul>rekado: yup, tx & tty later!
<apteryx>we can, but then we'll need to hand-tune bootloader, and only nckx has done so so far
<lechner>cbaines: thanks!
<civodul>apteryx: i guess we'll just wait for nckx to come back and then cross finger, then?
<lechner>hi, are the space requirements for our substitute servers documented somewhere, please?
<apteryx>lechner: what do you mean?
<lechner>apteryx: how much space would it take to set up a mirror, please?
<cbaines>lechner, it's not really documented anywhere, it's a number that keeps changing, but for there's 6.5TiB's of nars
<lechner>is bordeaux a copy of ci?
<apteryx>civodul: I'm in a meeting now but I can try something after
<mitchell>seems went down :(
<civodul>mitchell: yes, it's a known problem, stay tuned
<civodul>in the meantime, you can get substitutes from (which is enabled by default)
<mitchell>hm, my guix doesn't seem to even try. It looks at and dies when it doesn't connect or goes to build from source when --fallback is used
<civodul>lemme see
<civodul>if that's the case, you can force it with --substitute-urls=
<mitchell>good to know
<abrenon>hi guix
<civodul>hey abrenon :-)
<abrenon>: )
*abrenon is regretting that she's making such a silly face on the group picture
<civodul>oh? your face didn't strike me as silly; i'd have described it as a happy face :-)
<abrenon>definitely happy
<abrenon>but a bit silly too
<abrenon>but I'm fine with it ending up online : )
***ChanServ changes topic to '!! is down (website, mumi, berlin substitutes) !! GNU Guix | | videos: | bugs & patches: | paste: | Guix in high-performance computing: | This channel's logged:'
*apteryx puts diving suite on
<ggoes>ruh roh
<abrenon>diving suite ?
<abrenon>oooohh I see
*nckx is summonèd.
<apteryx>nckx: o/
<nckx>DNS records wha?
<abrenon>\o nckx
<abrenon>(my other hand is used to hold the umbrella)
<fnstudio>hi, anyone knows if there's a recommended way to add a one-liner to my bash configuration, as part of home-bash-configuration? i'm thinking of adding something along the lines of 'shopt -s histappend'
<mitchell>fnstudio: I am always going back and forth between how much of my rc/profiles should remain files from the dot file repository. On one hand I don't want to break my configuration on machines that don't have guix, but on the other machines that don't have guix break my configuration anyway.
<apteryx>so we're rehearsing the fun 'manually fix the bootloader' procedure again, eh
<apteryx>nckx: I'm rebooting berlin
<nckx>I don't see movement on the console.
<apteryx>nckx: is there anything to do under /boot/efi ? I guess not. Only copying stuff to /boot and editing /boot/grub/grub.cfg, right?
<fnstudio>mitchell: hm, right, that's a good point re maintaining somewhat fully-functional bash scripts (as files); have you thought of keeping as much as possible under guix home and then export them somehow to files?
<apteryx>nckx: yeah, the console was stuck, probably the machine had frozen
<nckx>apteryx: Yes, that sounds right.
<lechner>Hi, in 'guix home' packages are referenced by their names rather than their global variables. How may I refer to this custom package declaration, please?
<nckx>It's not (that) complex in the end, it was just a pain to figure out what the right configuration was because the boot ‘randomly’ failed to boot it.
<fnstudio>mitchell: was that to say that i could simply keep "shopt -s histappend" in a minimalist bashrc and load that via '(bashrc `((,local-file ".bashrc" "bashrc")))'?
<mitchell>fnstudio: I reference my dot files git repo in guix home
<apteryx>nckx: idrac doesn't seem to cause an actual reboot of berlin, at least the console is still stuck
<apteryx>I'll reset iDRAC
<mitchell>that way i can pretend its immutable but still not have to port every little thing into scheme
<nckx>Connection to closed. :)
<fnstudio>mitchell: i see, it makes sense, i'm adopting a similar approach with a few config files where i gently bring them in, initially as files and then we'll see... i can gracefully/incrementally transition them to scheme when i feel like
<mitchell>fnstudio: I have a define like (define dot-files (git-origin ...)) and use (computed-file "bashrc" #~(#$(file-append dot-files "/bashrc")))
<nckx>Imma return in about an hour, apteryx, but you should be able to fix this (I believe in you, and I believe in hardly anything). The firmware stuff is all correct now, so it should just boot from whatever you put on /boot.
<abrenon>lechner: well you're using specifications->packages which does some work to find packages from a string
<abrenon>but actually since you have the variable in the context, the easiest thing to do would be to simply add it to the list constructed by the call to specifications->package
<apteryx>nckx: thanks!
<abrenon>because (package …) expects just that, a list, so you could simply add a (cons guile-wm-lechner (specification->packages …)) like this and it should do the trick
<dabobeshai>is down?
<tsyesika>seems down for me
<dabobeshai>okay, I see. I think its planned maintainance. Thank you
<zimoun>hum, “guix build --fallback“ also fails with «guix substitute: error: connect*: No route to host». And worse, my laptop uses my desktop as offload machine, the substitute is in the desktop’s store, but it raises an error.
<mitchell>seems like we need to pin a message or something
<mitchell>use --substitute-urls=
<zimoun>mitchell: thanks. I know :-)
<zimoun>Berlin being down appears to me a good test, so I am testing. ;-)
<mitchell>oh i see
<civodul>zimoun: same here! looking for bugs with the URL fallback mechanism in 'guix substitute'
<apteryx>civodul: not super useful right now, but unallocated went up to 772.61, re-running balance with 'btrfs balance start -dusage=50 /' to free some more
<apteryx>(on overdrive1)
<apteryx>nckx: do you remember what was the serial option to pass to the ubuntu pxe boot so that we can see the output?
***wielaard is now known as mjw
<apteryx>console=ttyS0 ?
*apteryx tries
<apteryx>oh, it seems to boot after all
<apteryx>(berlin was still bootable)
<apteryx>it's on 'Generation 9 Sep 07 2022 08:10:56' currently
<nckx>apteryx: Yes, console=ttyS0,115200 but I don't actually know if the latter's faster. Maybe it already defaults to the fastest. Never tried.
<nckx>Also, back.
<nckx>I see it's up.
<nckx>I assume a ‘but’.
<nckx>apteryx: Whatcha currently doin'?
<lechner>abrenon: thanks! may i drop specification->package altogether, if the appropriate modules are listed in (use-modules) ?
<apteryx>nckx: authoring doc/ ^^
<nckx>lechner: Probably.
<apteryx>for the guix-maintenance repo. the amount of knowlegde to gather from places was becoming unwieldy
<nckx>I always felt lost, and felt like I sent a bothersome number ‘uhm how do we x’ e-mails. Thank you.
<abrenon>yes, this is what I do, for some reason I didn't get the habit of using specifications->packages and always import everything I use (just so that I know where it comes from)
<apteryx>civodul: after 'btrfs balance start -dusage=50 /' df -h -> 866G free, btrfs fi usage / -> 864.61GiB unallocated
<apteryx>seems we're done here, although you'll want to setup the mcron job to balance with -dusage=5 (I'll sent a doc in a minute with the info in)
<lechner>abrenon nckx: thanks!
<apteryx>also while you're at it you can add the compression options to basically double capacity
<nckx>apteryx: Is this still about berlin?
<apteryx>sorry, no that was overdrive1
<apteryx>before I revert the topic back, I'll try the script copying things to /boot and a reboot
<apteryx>to validate that this is enough
<nckx>No, don't let my barging (back) in without context slow you down.
<apteryx>I meant IRC topic :-)
<nckx>That makes a lot more sense.
<nckx>I'm a bit lost in general. Haven't been here for, what, a month? Unless I can still be of service, I'll be off again now.
<nckx>(Not for a month :)
<apteryx>I'm sending some doc about idrac and stuff, if you'd like to review the info it contains
<lechner>Hi, one of my first patches has not been merged, presumably due to a deficiency. With a freeze of core-updates in the works, I retitled the bug to include "core-updates". Do I have to set any debbugs tags to get mbakke's attention?
<nckx>CC'ing is how it's normally done.
<nckx>apteryx: Sure!
<apteryx>sent to guix-sysadmin
<apteryx>I'll try the berlin reboot after lunch
<abrenon>I'm off too
<civodul>apteryx: yay! i'll add that mcron job and reconfigure
<civodul>thank you!
<civodul>actually there *was* such a job already, running every 3 days
<civodul>it's in (sysadmin overdrive)
<apteryx>perhaps it was not running for some reason; because after balance with -dusage=5, it had regained most of the space
<apteryx>or perhaps the 'guix gc' was not running in tandem?
<civodul>"guix gc -F50G" was running every day
<civodul>/var/log/mcron.log has "No space left" messages from "btrfs balance" though
<apteryx>Perhaps we could raise that 50G to 100G (that's 1/10th of the drive capacity)
<civodul>yeah, but OTOH it runs every day
<civodul>i was going to reconfigure but without substitutes it's going to take a while :-)
<apteryx>berlin is back up
<apteryx>it rebooted fine
<civodul>yay, thank you!!
<apteryx>it's probably not on the last generation though (september 9th)
<civodul>now we need to get things straight, reconfigure, and reboot
<civodul>maybe better do that during office hours tomorrow
<civodul>but in the meantime, we can get the config in place re copying things around
<apteryx>I'm looking at that %copy-kernel-and-initrd thing.
<lechner>Hi, i cannot find the package variable guile anywhere. Is guile always installed and in PATH?
<apteryx>nckx: wait, /boot is on the SAN, right?
<lil>Hello! What i'm doing wring with install target:
<lil>install: test
<lil> install -d $(PREFIX)/share/doc
<lil> $(INSTALL_DATA) meme.txt $(PREFIX)/share/doc
<apteryx>lil: hi! please use for multi-line pastes :-)
<lil>meme.txt does not appear in $(builddir)/share/doc
<lil>apteryx: ok
<apteryx>PREFIX != builddir, I suppose
<apteryx>prefix is the installation prefix
<apteryx>nckx: the reason, I ask, is that if GRUB doesn't see the SAN, how would copying files to the SAN help?
<lil>here problem and code for Makefile and guix.scm
<lil>hope this problem is so easy that everyone can understand what i'm doing wrong
<apteryx>nckx: also, the last generation (9th from September 9th) booted fine; had you hand-fixed it, or does it mean that the system now boots without any hack?
<nckx>Hm, lots to swap in.
<nckx>apteryx: OK… say that's true: why is it booted to a much older one now?
<apteryx>it's not Generation 9 Sep 07 2022 08:10:56 (but I misremembered the date as september 9th)
<nckx>lechner: No, it's not magical like that. There just isn't a ‘guile’ variable. There are several ‘guile-X.Y’ variables.
<nckx>apteryx: The currently booted generation is from 2022-07-30 though?
<apteryx>I'm running 'guix system describe' and it says that generation 9 is current. Am I missing something?
<nckx>Then I wouldn't trust ‘guix system describe’ with a very long stick.
<rekado>FYI: the problem in the data centre was that power for the cooling pumps was lost, so systems started overheating. The pumps also didn’t come back online when power came back up.
<nckx>apteryx: Look at the currently booted kernel: it's /@root/gnu/store/86ymw35fiw475vmsq6fpa8hi6vqybx6y-linux-libre-5.18.14/bzImage
<nckx>here's the fun part
<nckx>doesn't even exist on / anymore
<nckx>rekado: …wow.
<nckx>apteryx: Whatever is the *real* boot isn't on the SAN, no.
<civodul>/run/current-system = /run/booted-system, so i guess "guix system describe" is right
<apteryx>but uname reports an older kernel
<nckx>civodul: ‘guix system describe’ literally claims ‘GNU with Linux-Libre 5.19.7’.
<rekado>cooling failure is pretty bad; they have annual failure tests to make sure this kind of thing doesn’t happen.
<rekado>I’m looking forward to a post-mortem report.
<nckx>It's… interesting that it even came up, presumably without modules (well, it had the modules from the initrd, at least).
<podiki[m]>(guix shell goodness on hackernews front page)
<rekado>we still have the SSDs plugged into that server, no?
<rekado>could the kernel + initrd have been loaded from the SSDs?
<nckx>rekado: By definition. We can't boot from the SAN.
<rekado>and then when the SAN came up during boot we pivoted to the current system there?
<nckx>Hm, I don't think it's that complex, but who knows.
<dirtcastle>a month back I talked to you and nckx you guys were teaching me how to use send patches through email. and I took screenshot of everything you said. and I read the important parts of progit book (will read the remaining too) just like you guys told me to. and rn my question is that day unmatched-paren mentioned that someone tried to
<nckx>It's definitely interesting. I didn't know ‘guix system describe’ could be so thoroughly misled.
<nckx>Yo dirtcastle.
<dirtcastle>there have been attempts to add services to guix. but it's a pretty large project and noone cares about it.
<nckx>*enough to put in as much work as it requires :) I've heard the occasional ‘oh yeah in Guix would be nice’.
<nckx>But if it's nontrivial work that doesn't go far.
<apteryx>nckx: re berlin, so hmm, if GRUB can't see the SAN at all, wether / or /boot, how were you able to fix the bootloader previously?
<dirtcastle>hi nckx !
<dirtcastle>I care abt it. the next thing I want to do is write the service. can u show me who were the ones that tried to write it? I wanna see their code. (I'm yet to read large portions of guix manual that's required to write package definitions and services. that will be next after reading progit or I'll read alongside it)
<nckx>apteryx, rekado: So I'm *relatively* confident that we're currently booting from /dev/sda2 (FAT UUID 4563-9DEF).
<nckx>Currently mounted at /mnt for your convenience.
<nckx>The ‘fix’ was simply to copy the required @root files there, see.
<nckx>Because I seem to have lo^Wmisplaced my notes I'm less sure if that is also the grub.cfg that's used, but it would make some sense if it were?
<nckx>dirtcastle: Hm, I didn't mean that others tried. Maybe someone did, I don't know, I just meant that ‘no working package’ != ‘nobody cares’.
<civodul>nckx: well, if you fiddle with the GRUB entries, you can certainly boot a kernel other than the one in that system generation
<civodul>no doubt
<civodul>but you have to want it
<nckx>I don't think that much fiddling took place.
<civodul>well, my understanding is that it doesn't boot out of the box, does it? :-)
<nckx>It's not something I would do knowingly, because I'd never expect it to work (I'm still very surprised/sceptical that it seems to have done).
<nckx>Maybe something got desynced the past few months, of course.
<nckx>It was rebooted in that time IIUC.
<civodul>my point is: claiming that "guix system describe" is misleading in general is misleading
<apteryx>nckx: OK, so if I understand, we need two things to industrialize the config: 1. have /boot mount from a non-SAN drive partition, such as /dev/sda2. 2. Copy kernel image and initrds to it, straight under /boot
<nckx>I made no such claim.
<civodul>apologies then, that's how i understood it
<nckx>It is misleading *here*.
<nckx>civodul: Sorry if that's how it came across. The ‘Then I wouldn't trust’ above is to be read as ‘In this case, seeing what we are clearly seeing, it can clearly not be trusted’ :)
<nckx>Or at least not used as a source of truth.
<nckx>It *is* surprising, no disagreement there.
*civodul nods
<nckx>apteryx: Sounds right apart from ‘straight under /boot’, it expects /@root, not /.
<apteryx>ah, my script was adjusting grub.cfg for that
<nckx>Oh, OK.
<apteryx>so if I want to leave the grub.cfg alone, I keep the prefix
<nckx>Mild preference for leaving GRUB with the illusion that the fake store matches the real one, but both should work.
<apteryx>actually I'm not sure we can leave grub.cfg alone, we probably want to remove the "search --fs-uuid --set d5d1a040-7f2a-4c38-9a89-82f08866f6ec" at the top that would set root to the store?
<lil>next error is "suspicious ownership or permission on (builddir)"
<nckx>apteryx: I thought it couldn't see the store?
<nckx>I don't see how that line can have an effect.
<apteryx>right, that's why I suggest it's probably best to remove that grub.cfg directive
<apteryx>ah, you mean it'd simply fail without any impact?
<apteryx>I see! Clever.
<clever>my ears are burning... :D
<nckx>‘You can't tab-complete stupid’ (in this channel anyway).
<nckx>apteryx: Again, this is just my preference for leaving grub.cfg intact if we can, I get why it's ugly and you'd want to remove it. But I fear finding an artisanal grub.cfg with subtle modifications in a few months with no clear recollection of what it does :)
<nckx>I know you're working on automation, but there are too many grub.cfgs already. I deleted one just now that I'm… pretty sure wasn't used.
<apteryx>yep, agreed, until we have the grub.cfg cleanup automated in the grub.cfg generating Guile module
<nckx>And I'm almost sure that UUID E1B3-BF92 isn't used…
<apteryx>so my plan is to adjust the /boot mountpoint with the UUID of /dev/sda2, then adjust the copy-kernel+initrd-to-/boot snippet
<nckx>Sounds good.
<apteryx>which is an ad-hoc solution to a more general problem
<apteryx>(I'm not ready to tackle that one just now)
<nckx>Do you mean the problem that GRUB can't see the SAN or something else?
<apteryx>I meant that having a standalone /boot is a genuinely useful use case in some circumstances such as this one
<apteryx>(other circumstances: old BIOS machines with >2TB drives)
<nckx>Or bcachefs weirdoes.
<apteryx>such that it'd make sense to allow setting that up from operating-system without resorting to scotch tape.
<nckx>This must be one of the lowest bug/wishlist #s to be unfulfilled :)
<nckx>(And I'm fully guilty of writing a ‘good enough’ bash script and calling it a day, to be clear.)
<nckx>I was going to ask earlier, but don't think I did: are we still going to move the SSDs (but one)?
<nckx>Keeping sda around as a very expensive overprovisioned /boot is OK for now.
<rekado>we talked about it on the sysadmin list, but it’s not clear what the next steps are on the receiving machine.
<rekado>we still have the original spinning disks; those are perfectly fine for a 2TB /boot
*nckx sees ‘QVO’ in ‘lsblk -S’, deletes ‘expensive’.
<lil>what access mode (644, 744 or so) should be on directory before call 'install' target?
<apteryx>ideally /boot should be on RAID, otherwise the day the deice fails we're screwed
<nckx>lil: 7xx at least, if you're not root (and even then it's… odd not to have +x on a directory).
<civodul>apteryx: note that /boot/efi is a separate mount point, so we must make sure we don't shadow it or something like that
<apteryx>I think the mount order is smart enough to mount parents first?
<apteryx>in any case I'll list /boot first in the file-systems list
<apteryx>we just need to choose a device
<nckx>I wonder if we're in the minority now, still mounting it at /boot/efi.
<civodul>i'm afraid there's nothing smart going on when it comes to mount order :-)
<nckx>I'm surprised.
<nckx>I thought I'd observed smartness once.
<civodul>however, you can explicitly add a 'dependencies' field to ensure the right ordering
<apteryx>uh, that's why it has to support "dependencies" ?
<civodul>yes, and for mapped-devices
<civodul>like when a file system depends on a mapped device
<apteryx>right, we need it for mapped devices
<apteryx>rekado: is this MD34xx blackbox storage device a redundant array of disks?
<nckx>It's unlikely to be a 1U chassis with a single lonely 40T drive in it, but maybe the MDC is weird.
<lil>nckx: 755 is ok
<nckx>2U, apparently, sorry.
<lil>thank for help
<lil>now i have reproducable resume
<nckx>755 is the classic. Don't deviate from it without valid reason.
<rekado>apteryx: that’s the old storage array, isn’t it? I think that was RAID 6 or something, don’t remember the specs of the individual disks, though
<apteryx>rekado: yes, it's the old storage array. I'll go ahead and make a /boot partition on that
<rekado>how’s that going to help thougH?
<rekado>we can’t boot off it
<apteryx>we can't? I thought it was bootable.
<apteryx>nckx: I'm still confused about E1B3-BF92; on what drive is it hosted?
<apteryx>seems to be on a 1022M partition on the... SAN?
<nckx>apteryx: I think it's the SAN?
<apteryx>but I thought the whole point was choosing a non-SAN device because GRUB can't see it? :-)
<nckx>Just going to paste this for safety, bad time for typoes:
<nckx>21 Sep 19:29:24<nckx> And I'm almost sure that UUID E1B3-BF92 isn't used…
<nckx>21 Sep 19:14:20<nckx> So I'm *relatively* confident that we're currently booting from /dev/sda2 (FAT UUID 4563-9DEF).
<apteryx>ah, thanks! I had misread that as "I'm almost sure it's the one used"
<nckx>Oof :)
<nckx>Yeah, hence my caution.
<nckx>Heh, I see there is still evidence of the highly advanced debugging technique I used back then (…/this-is-xxx).
<nckx>(You can delete those, unless you want to double-check again from GRUB.)
<avp>Hello, Guixers!
<apteryx>nckx: I will, for our sanity
<nckx>Hi avp.
<apteryx>so /boot/efi is currently from the SAN sa well... is this alright?
<rekado>apteryx: we were not able to boot from the external storage array, just like we are now unable to boot from the SAN.
<apteryx>rekado: I see, thanks for reminding me
<rekado>back then we also just copied the kernel and initrd to /store on the local root disks
<rekado>(hence that old activation service extension)
<nckx>apteryx: Are you using /mnt?
<nckx>I want to tidy up before I go AFK otherwise.
<apteryx>I ain't
<nckx>Oki. umounted.
<apteryx>I think I'll create boot and efi partitions o /dev/sda and /dev/sdb, and list them in the bootloader targets
<apteryx>this will make for some cheapo redundancy
<apteryx>are the rotative drives still plugged in berlin? I don't see them in 'lsblk -S'
<rekado>apteryx: I don’t remember well, but I think I didn’t remove them. I only added SSDs until we ran out of slots.
<apteryx>OK! weird, they are not exposed to the OS if still there.
<apteryx>you might want to remove them physically next time you visit
<apteryx>or verify that they are connect alright
<atomsk298[m]>Quick question about building new packages
<wwalker>I wanted to give guix a whirl. I ran the installer , then "guix shell coreutils" -> "guix: shell: command not found" "guix --help" doesn't show "shell" as a command. What am I missing?
<atomsk298[m]>I run ./pre-inst-env guix build <insert package name here> --keep-failed, but get an unknown package error
<viivien>wwalker, that’s because you have guix 1.2.0, which is quite old
<viivien>You should run guix pull first to get the latest commit
<viivien>Or even 1.3.0 does not have it I think
<wwalker>viivien: Thank you!
***mark_ is now known as mjw
<trevdev>Any gnome-deskop-service-type users on? I'm noticing that programs like "gnome-photos" and "gnome-music" aren't propegating from my ~/Pictures and ~/Music folders respectively. I assume I'm missing some feature of gnome that's not in the convenience service?
<lilyp>wdym propagating?
<trevdev>lilyp: I open these applications. It says "we'll find the files from ~/Folder" and it never does.
<two[m]>i have a problem
<two[m]>> ERROR: C shared or static library 'm' not found
<two[m]>what to install?
<two[m]><trevdev> "Any gnome-deskop-service-type..." <- tracker?
<apteryx>I want to extend etc-file-service with a bunch of generated files. I think I know how to do that. But I also need to have some "compilation" pass that would consider these files, "dconf compile" or "dconf update". Any suggestions as to how to approach this problem?
<civodul>apteryx: you could first have a 'computed-file' that runs "dconf compile" on its inputs
<civodul>and then you'd pass that as an extension to etc-service-type?
<apteryx>I fear that 'dconf' must expect the /etc/dconf hierarchy to exist before hand
<apteryx>so I'd need to file-union my computed-files, and use this as an input to another computed-file that'd do the compilation?
<lilyp>trevdev: and are the respective missing dbus services
<two[m]>these are graphics
<two[m]>when opening music i see
<two[m]>destination=org.freedesktop.Tracker3.Miner.Files serial=42 path=/org/freedesktop/Tracker3/Endpoint; interface=org.freedesktop.Tracker3.Endpoint; member=Query string "CONSTRAINT SERVICE
<two[m]>CONSTRAINT GRAPH tracker:Audio SELECT ?added ?song ?albumDiscNumber ?trackNumber ?duration ?album ?artist ?url ?title...
<two[m]>must be tracker
<two[m]>there is "FILTER(STRSTARTS(nie:isStoredAs(?song), 'file:///home/...'))" at the end
<two[m]>if it's running but there are no results try running tracker3 reset -s
<lilyp>I don't get any of that with gnome, where do you get that from?
<trevdev>Hokay I'm grokking the gnome-desktop-service-type, which is what comes with the choice of the "gnome desktop" option you may choose from your initial install from Guix.
<trevdev>It's just gnome, gnome-settings-daemon, gnome-control-center, gnome-system-monitor, gvfs.
<trevdev>In Gnome's inputs I don't see tracker. I do see tracker-miner, however
<trevdev>I am admittedly mostly just in acceptance of Gnome's usual magic so I don't really get things when they break
<two[m]>is the miner running?
<lilyp>the miner is positively running – look at d-feet
<trevdev>2 PIDs for .tracker-miner-
<trevdev>Probably not related but the thumbnails don't show up in nautilus either. Maybe I should avoid the use of the service module and just put a GNOME desktop together myself.
<two[m]>do you have anything in the audio index?
<two[m]>tracker3 sparql --dbus-service org.freedesktop.Tracker3.Miner.Files --query "constraint graph tracker:Audio select * {?s ?p ?o}"
<trevdev>tracker3 is not on path for my user nor root
<apteryx>is file-union able to merge multiple same-named directories with differing content?
<trevdev>apteryx: Yep
<trevdev>No, directory-union
<trevdev>But file-union too.
<apteryx>OK. etc-service-type uses file-union
<apteryx>hm, file-union doesn't support it. it expects regular files, not directories
<trevdev>Sorry I failed to convey that
<apteryx>no worries
<apteryx>I then tried having files on the right side of the alist file-union wants, and same-named directory on the left
<trevdev>I have used (directory-union) for that use-case in the past though. Not sure if that helps
<apteryx>but it doesn't accept it either, it fails with a symlink error (I guess it tries to symlink over the existing directory)
<trevdev>I used it to hack together paths to my common-lisp source repositories before I found a more sane way to get my env right
<trevdev>Lots of nested, similar directories
<tricon>echo $? #=> -1
<two[m]>i have am getting this when building gnome-control-center
<two[m]>Compiler stderr:
<two[m]>ld: cannot open output file /home/user/??????/gnome-control-center/build/meson-private/tmpki66j7yy/output.exe: No such file or directory
<two[m]>collect2: error: ld returned 1 exit status
<two[m]> ERROR: C shared or static library 'm' not found
<two[m]> * >Compiler stderr:... (full message at <>)
<two[m]> * >Compiler stderr:... (full message at <>)
<two[m]> * >Compiler stderr:... (full message at <>)
<rekado>is there a reason why you don’t build this as a Guix package?
<two[m]>rekado: i want to code it
<apteryx>are you using gcc-toolchain?
<apteryx>how did you setup the environment? 'guix shell -D gnome-control-center' ?
<two[m]>apteryx: yes
<two[m]>i have tried to add gcc-toolchain, no success
<rekado>do you have ‘gcc’ installed? (you probably shouldn’t)
<rekado>can you reproduce this in a pure environment?
<two[m]>rekado: this is a --container
<apteryx>what's the exact command you used
<apteryx>(to create the containerized environment with guix)
<lechner>nckx: thanks!
<rekado>two[m]: is this a clean checkout or had it previously been configured outside of the container?
<apteryx>civodul: Merry thoughts to end your day: what you started and that we've all been building on has grown amazingly in the last 10 years. I can't wait to see the kind of growth and adoption that GNU Guix will enjoy in the next decade. Thank you for starting this project!
<apteryx>(and for sticking around!)
<lechner>Guix may well replace Debian. You folks are geniuses in software design, and the product is maturing quickly
<two[m]><apteryx> "what's the exact command you..." <- guix shell -C -D gnome-control-center
<two[m]><rekado> "two: is this a clean checkout or..." <- it's clean
<two[m]>lechner: i'm afraid it won't on most hardware
<rekado>lechner: I hope it won’t replace Debian but augment it. Debian’s contribution is much bigger than the software in the distribution.
<rekado>two[m]: does the directory it references exist at all in the container? Is it just the file that’s missing?
<two[m]>i don't have meson installed outside
<two[m]>rekado: no, i don't have a "??????" directory
<rekado>oh, literally question markes
<lechner>rekado: i suppose there is always a chance Debian will adopt Guix as the package manager...
*vagrantc chuckles
<rekado>if there’s a chance that a tree falls but nobody is around to hear it, what sound does a single clapping hand make as it catches it?
<singpolyma>rekado: 💬
<lechner>what if the tree is called dpkg?
<rekado>is ‘dpkg’ onomapoetic?
<lechner>yes, it sounds like a crashing tree, doesn't it!