IRC channel logs


back to list of logs

<elevenkb>is anyone in the mood to help me with `emacs-substitute-variables`?
<gabber>how can i find the reason why package XYZ is being built on `guix home reconfigure`? is it some `guix graph`?
<lechner>gabber / most likely, the needed substitute was not available from elsewhere
<gabber>i've just set my own up :) but i'm not sure why my (slowish) machine still wants to build the packages.....
<lechner>i also share substitutes local, but only via 'guix publish'
<gabber>IIUC that's the same thing as substitutions? at least i see my host show up when i invoke reconfigure (at the very beginning, with the other substitute servers)
<lechner>gabber / i think so, but your substitute server probably does a lot more, such as pull Guix and rebuild needed packages
<gabber>that's possible, yes :)
<lechner>i use guix deploy for my systems, so only users really benefit from sharing their items
<gabber>huh.. so you build centrally and only _deploy_ and refrain from local `guix install`, `guix shell`, `guix foo reconfigure` calls?
<lechner>i use guix shell extensively, or rather it's Emacs equivalent buffer-env
<lechner>like you, i also use guix home reconfigure
<gabber>ACTION just learned about buffer-env
<gabber>i was sure this *had* to exist, but i just didn't know what it was called :) thanks lechner !
<elevenkb>nvm. y'all's i managed to fix my problem.
<elevenkb>i have a style question: i'm trying to package emacs-eat and it comes with a shell script that one is recommended
<elevenkb>to add to .bashrc, etc. is there a standard convention about where (in the build result) to put such things?
<elevenkb>eh, maybe i should just hand in the patch and then get feedback that way.
<apteryx>which is faster? a Cortex A9 Quad (native) vs a x86_64 Core 2 Duo Quad (QEMU) at compiling for armhf-linux? :-)
<apteryx>ACTION is benchmarking wih u-boot
<elevenkb>let me guess... cortex a9 quad native.
<apteryx>we'll know in at most 4 minutes (the time it took on the ARM board)
<apteryx>QEMU does sleem very sluggish
<elevenkb>is it 4 minutes yet?
<elevenkb>apteryx ?
<apteryx>yes! and it's still running
<apteryx>phew. 13m45.077s on Q6700 Core 2 Duo vs 4m18.704s on the Cortex A9 Quad
<apteryx>you're flair was right!
<elevenkb>always be enjoying moore's law?
<elevenkb>flair? do you mean username?
<elevenkb>anyway... gotta go to bed. see y'alls tomorrow
<apteryx>perhaps not used much in English "skill or instinctive ability to appreciate or make good use of something"
<apteryx>elevenkb[m]: o/
<apteryx>cross-built for arm-linux-gnueabihf: 1m6.613s
<argv>Hey, i'm trying to run guix build from guix repository root with pre-inst-env script but for some reason its not able to find my updated package. Instead its picking up the older definition.
<argv>I followed the Building from Git ( and then following the Packaging Guidelines ( but not sure what's wrong
<argv>ah looks like the GUILE environment variables were a bit off so it had the guix profile ones before the local repository one
<unmatched-paren>morning guix
<mechalain>Hi there, I was having some issues with a system reconfigure without any changes to my config after a pull. Im getting an odd shepherd error. shepherd: Evaluating user expression (and (defined? (quote transient?)) (map (# ?) ?)).
<unmatched-paren>mechalain: that's not an error
<unmatched-paren>it's just a status message
<unmatched-paren>a status message that probably should be hidden somehow
<mechalain>Ok odd, I was having boot issues but it may be unrelated, thanks I'll check back in
<raghavgururajan>guix system: warning: exception caught while executing 'eval' on service 'root':
<raghavgururajan>error: build-directory: unbound variable
<raghavgururajan>Oh! Already been posted.
<raghavgururajan>mechalain: Same here, system won't boot.
<sepi>I'm trying to add a service type to (services web) and then use it in my system config but get a unbound variable error for that service type. I added my definition in guix/gnu/services/web.scm, put the service's symbol in the #:export part of the package definition and (use-service-modules web). Did I miss anything?
<sepi>I'm using ./pre-inst-env guix system reconfigure /etc/config.scm. It seems that my changes in my guix git repo do not get picked up
<sepi>Argh, nevermind. I was still in a guix environment :)
<haugh>sepi, why do you need to edit Guix' source? Can't you edit closer to the system configuration? Not trying to talk you out of it, just surprised
<apo11o>How can I use hugo in guix?
<sepi>haugh: I don't exactly understand what you mean with closer to the system configuration. I was told that the prime way of developing your own services was to just clone the guix repo and add it there. Like this it's also easy to contribute it back
<sepi>apo11o: you mean the static site generator?
<sepi>I could not find a package for it
<apo11o>sepi: me neither thats why I asked, if there's a workaround available
<sepi>apo11o: the best work-around is creating a package and submitting it ;)
<apo11o>sepi: I don't know how to create packages, I'm new here, but I hope I'll be able soon.
<sepi>apo11o: yeah, I'm new too and I've been trying to figure out how to install wordpress :D
<sepi>apo11o: if you don't want to contribute or at least package your own stuff, you will not get extremely far with guix I'm afraid. Unless all your use-cases are covered already by chance
<apo11o>sepi: haven't tried it and not sure if it's recommended but maybe you can using nix. My usercase is mostly covered, I only use flatpak for anki. I just wanted to try out hugo, but I'll probably just settle with org + customize org-html-themes
<rekado>apo11o: we have a package for anki
<apo11o>rekado: yeah, I can run it only using QTWEBENGINE_CHROMIUM_FLAGS="--disable-seccomp-filter-sandbox" anki But it's a really old version and none of my addons support it, even my custom ones that I have made.
<sepi>apo11o: nah, I'm here to automate a bunch of VMs, so I'm really interested in having all the packages I used integrated into the distro
<cbaines>raghavgururajan, there was a problem with the nix service
<cbaines>I've just pushed a fix. Hopefully you can boot in to a previous generation
<raghavgururajan>cbaines: I see. Thanks for letting me know.
<raghavgururajan>Yeah, previous gen was good. Hail Guix!
<Maya[m]1>Hi! I’d like to buy some guix stickers for my laptop, is there any official way to do so?
<cbaines>Maya[m]1, there are often stickers at events involving Guix
<cbaines>the next one I'm aware of is FOSDEM, including the Guix specific event in the two days prior
<Maya[m]1>cbaines: i dont believe i can afford the travel sadly :(
<Maya[m]1>(and the hotel stay)
<efraim>certainly understandable
<apo11o>It'd be cool if we could buy stuff online to support guix development, +it has a really cool logo. I might come to Brussels just to take one of those stickers !
<ekaitz>hi guix! I'm working on a build-system where should I leave the output? I'm leaving everything in `out` but the package appears empty!
<ekaitz>oh maybe i just figured it out
<ekaitz>oh yes i did
<sepi>Is it normal that I need to "make" guix everytime I want to reconfigure a system based on my custom guix clone? Does guix system reconfigure not pick up changes in the scm sources automatically?
<haugh>Who told you that editing the source code was the "prime" way to manage your own services?
<AwesomeAdam54321>Can someone commit this patch to add the dataparksearch package?
<sepi>haugh: don't remember really
<sepi>haugh: so how should I do it?
<haugh>I'm really, really hesitating to give you advice because I'm new here. Personally I've been putting my own service definitions in the same repository as my operating-system definitions, which import the service defs through the Guile module system
<haugh>I just can't imagine trying to manage everything through an entire fork and wondering what the advantage is
<haugh>sepi, ping
<sepi>haugh: pong
<sepi>haugh: so you just put it next to the system config.scm and use it
<sepi>I must admit, this is something that seems more straight forward
<sepi>Unfortunately the manual does not afaik tell you how to do it
<florhizome[m]>Sepi you can add custom modules via a Channel or load path for pretty much anything, Not only packages
<sepi>florhizome[m]: Is it possible to have a local channel that's not a poblished git repo?
<florhizome[m]>If it's a channel it needs to be a git repo
<sepi>florhizome[m]: I need to think of my system more like any other application I develop where I could just define my own stuff and use/include it in my main script file
<florhizome[m]>But yeah
<florhizome[m]>You can also have package repos locally
<sepi>florhizome[m]: Ok, I'll start just putting my stuff in a regular file next to my system config and see if I can just use it normally, maybe by modifying the GUILE_PATH
<florhizome[m]>You can start by using -L
<florhizome[m]>to prepend a load path
<unwox>sepi: there is also -L flag for specifying load path
<florhizome[m]>I have also Seen people adding a load path on top of a scheme module
<haugh>with include-from-path, or what?
<sepi>florhizome[m]: unwox: oh, sure. I'll try that
<florhizome[m]>Not Sure, i thought it was add-to-load-path or sth Like that
<florhizome[m]>specifically i saw that in pkill9 s gitlab
<florhizome[m]>a while ago though
<sepi>Btw, what's your strategy of checking generated configs when developing services? I had the case where the php-fpm config generated was referencing a group (nginx) which didn't extist and therefore didn't start. I had to use globs to figure out the right derivation in the store that was genereated to understand this problem
<nckx>Mornin' Guix.
<ekaitz>did we get the mentoring thing going finally? i need a mentor for some patches :D
<haugh>sepi, I make HEAVY use of the REPL, to the point where I can't imagine working in Guile without it. Basically everything that makes it into a file has already been run iteratively through every step of the development process with tangible results. This dramatically increases your confidence in the outcome.
<haugh>You still raise a valid question and I'm definitely interested in getting unit/integration tests into the system layer. If you've got the stomach for some richer Scheme concepts, check out SRFI-64, which is implemented in Guile.
<ardumont>zimoun: hello, file search thingy, i've iterated over adding back another indexation method (i had to rewrite all the cli stuff)
<ardumont>my question is for replying to the thread, how should i do regarding the patch.
<ardumont>should i just send the new commit(s) as patch?
<ardumont>(just trying to figure out stuff prior to be too noisy by email ;)
<ekaitz>I sent an email to guix-devel, if anyone wants to play with me to make a zig-build-system, your help will be more than welcome! I have something working but it needs polishing
<permcu>Hi, I am working on a variant of digital-ocean-environment-type for guix deploy. Does deploy-digital-ocean create a new digital-ocean droplet, "invect" it and then run deploy-managed-host each time guix deploy is run? I hope someone here can tell me if my understanding is right. This is not the behavior I would have expected. Imho "infecting" the droplet _once_ and then falling back to normal deploy-managed-host would make much more
<permcu>sense. But I do not think digital-ocean.scm does this.
<sepi>haugh: I'm mostly looking for a convenient way to figure out all derivations produced by "instantiating" a service
<haugh>sepi, you want to generate a list of store hashes so that you can go and manually inspect them on the filesystem?
<haugh>feels kind of XY but I think I see where you're coming from
<sepi>haugh: it's probably not really the way to go but it's a very down to the earth way of seeing what's wrong and why a service might have troubles starting
<efraim>I've never looked closely at the digitial-ocean-machine specification but I think it was meant more for deploy and destroy than to deploy and reconfigure. Once it's deployed it can just be a regular system in terms of 'guix deploy'
<efraim>ekaitz: at the risk of being more overwhelmed, I'd be interested in taking a look. The furthest I've gone in that direction was ncdu2
<ekaitz>efraim: please do and tell me what you think
<ekaitz>efraim: with the build system I added in the patch it shouldn't be difficult to be able to add ncdu2 to guix, but I'd like to fix the build system first :D
<permcu>efraim: Your explanation or intuition fits the code and makes sense to me too.
<efraim>ekaitz: returned with comments :)
<ekaitz>thanks for the email efraim , I'll review it!
<jlicht>lembrun[m]: I was just reading the logs, but did you happen to run into "... authentication error: pam_open_session ..."?
<lilyp>can we make it so that mumi supports qa tags in search?
<rekado>I reconfigured my system, rebooted, and now input methods don’t work as they are supposed to…
<ekaitz>efraim: just a simple question is any build system wrote using gexps? or they have to be built with the keyword args in order to get the outputs and inputs?
<rekado>ibus-libpinyin works only in the gnome-shell search input field
<rekado>nowhere else
<rekado>does anyone know how user-installed input methods are supposed to be found by Gnome?
<rekado>are they supposed to be installed to the system profile now?
<gnucode>I am trying to get pinetry to work on a wayland only session. I can't seem to figure out how to configure pinetry-bemenu. Anyone using pinetry-bemenu? Or using pinetry in a wayland only session?
<apteryx>would someone know how to completely reset the nfs service without a reboot?
<rekado>apteryx: what does “reset” mean?
<rekado>you can kill all its programs
<apteryx>restart every components it uses, the same as if I rebooted
<apteryx>after booting my board with nfsroot v3, u-boot which uses v2 started being unhappy: Loading: *** ERROR: File lookup fail
<apteryx>a reboot previously fixed it
<efraim>ekaitz: um, I'm not sure exactly. I'd have to check the other build systems
<apteryx>rekado: ah! user error
<ekaitz>efraim: don't worry, I'll make some research
<lechner>apteryx / there is an NFS kernel module you may have to unload
<lechner>several, actually
<rekado>lechner: the service doesn’t load or unload kernel modules FWIW
<rekado>it just starts all the components
<rekado>hmm, just made my rockpro64 boot from USB, using the /boot/extlinux/extlinux.conf from the SSD
<rekado>it loaded the new kernel and initrd, but then the screen froze with a weird yellow pattern
<lilyp>rekado: re ibus, I think you need to clear your ibus cache and then restart the daemon
<rekado>lilyp: I deleted ~/.cache/ibus and ~/.config/ibus and rebooted
<rekado>it’s a little sad to see how difficult this has become compared to the early days when all you needed was a search path :-/
<rekado>(I feel there’s a little too much friction in using Gnome; might finally give up on it.)
<apteryx>rekado: cool; is this using U-Boot's distro_bootcmd facility?
<rekado>apteryx: no. Frankly, I don’t know where all these variables come from that “env print” shows me.
<rekado>there’s a boot_targets variable
<rekado>and set it to “usb0” (so it won’t try to boot from scsi)
<rekado>and then it takes care of the rest, reading /boot/extlinux/extlinux.conf etc
<rekado>ACTION —> repairs bike
<lechner>ACTION ditched Gnome too
<apteryx>rekado: OK, so probably that board is configured with DISTRO_DEFAULTS=y
<apteryx>these are shared scripts made available by default when that option is set
<apteryx>or in older u-boot such as the one I have on my board (v2015), it had to be enabled via an #include
<apteryx>rekado: you hadn't luck with the grub-efi chain option?
<apteryx>it seems that should be easy, with a U-Boot that supports UEFI
<apteryx>(2016 and up, I think)
<apteryx>s/easy/possible/, perhaps :-)
<corydalis470>Is anyone using greetd with Sway? I'd be curious to see some configs and how autologin and can be configured.
<acrow>Anyone know the trick to get vlc working? It appears to be trying to use a invalid video acceleration module on my guix system. mpv, however, works fine.
<nckx>Works here:
<nckx>I do have libva set up for mpv.
<nckx>Where ‘set up’ just means installed, I don't remember doing anything more.
<lechner>acrow / i have vlc working (but I do not recall installing libva separately). which type of graphics card do you have and what kind of error do you see, please?
<jonsger>my vlc works just fine on Guix System :)
<nckx>btw I use Sway.
<lechner>acrow / are you om wayland?
<sam-d[m]>is their a way in guix to query which package provides a certain library or file?
<acrow>lechner: I didn't do anything special regarding libva it just came along for the ride. For my trivial purposes it would probably be best if I was able to just turn off video acceleration. I was trying to find a --intf <module> that would work but no joy so far. I have no graphic card mojo. Just running X11 though I occasionally see mention of a deblobbed attempt to load a radeon module that I ignore.
<lechner>sam-d[m] / i am not aware of a way, but i am working on a proposed solution, which is one among many
<acrow>I've noticed a few tickets out on mesa. Could this problem be associated?
<nckx>So, it's coming.
<nckx>acrow: Does ‘LIBVA_DRIVERS_PATH=/var/empty vlc’ work?
<sam-d[m]>nckx: cool, I will be waiting 😀
<acrow>nckx: Yes!, you hammered it again nckx. Thank you.
<nckx>I am the godhead of superficial work-arounds 💪
<acrow>nckx: issues/tickets.
<nckx>Mesa has just (and barely) shy of 3k open issues :-/
<nckx>The only r600-related SIGFPE does not look relevant.
<torb>Hi and thanks for making GUIX=) I have a question: How should I go about getting GTK-programs to show text? When I installed GUIX I chose EXWM as desktop environment. I guess it is kind of minimalistic in its dependencies and maybe I should install some package to get gtk-fonts? I have installed cantarell
<torb>but see not text in gtk programs
<torb>(Sorry for the line break)
<lechner>i don't see the GTK icons with EXWM
<nckx>torb: No text at all? Anything?
<acrow>The recent one I was looking appears to be associated with the linux kernel but I'm running linux-libre 6 and with mpv working I guess this is strictly an libav associated issue.
<nckx>torb: Not ever [a-zA-Z]?
<nckx>acrow: But weird that mpv doesn't trigger it, as it uses libav too.
<torb>No text at all in gtk widgets (event emacs menu-bar). Iceweasel displays text in web pages, but the open file-dialog has no text in any elements.
<acrow>nckx: which brings us back to vlc...
<nckx>I did not mean to imply otherwise.
<torb>lechner: I do get icons in inkscape for example
<acrow>nckx: Would creating an issue be appropriate/helpful?
<nckx>torb: I'd install at least font-dejavu, but if you also install font-google-noto (warning: huge) and run fc-cache -rv (and perhaps log out & back in, just in case) and still don't have text, I'm flummoxed.
<torb>nckx: I will try=)
<nckx>acrow: It's certainly appropriate, but (I guess) only those with similar hardware will be able to investigate at all.
<acrow>nckx: See:, looking at dmesg, and that issue on nouveau, perhaps this IS that upstream kernel issue. Maybe your workaround should get into that issue while we wait for maxim to get a response from upstream.
<acrow>issue 59361 seems to cover it.
<acrow>But I don't want to downgrade my kernel. The workaround and mpv will do it. I do wonder if there are any guix approved ways to use the installed VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Mars XTX [Radeon HD 8790M]
<nckx>I don't see a connection to 59361.
<nckx>I mean, with r600
<nckx>Does nouveau handle AMD now?(??)
<nckx>ACTION confused.
<acrow>nckx: nouveau driver seems to be implicated in
<nckx>Yeah, but how do you get from that to ‘issue 59361 seems to cover it’.
<acrow>nckx: Yeah, the radeon/AMD acceleration confuses me too.
<nckx>How does <> lead to a SIGFPE (IIRC?) on AMD hardware.
<nckx>(I don't know either way, I'm just confused.)
<nckx>Have you tried downgrading your kernel?
<acrow>nckx: I thought maxim identified a linux-libre kernel problem. No, I have not tried a downgrade. I could just do a system rollback but I don't see how I could contribute to what has already been documented. I have not looked at your commit yet...
<acrow>Argh, icecat blocks it. Is that the commit that Maxim identified?
<torb>Hi again! Installing the suggested fonts and running fc-cache -rv did
<torb> not work, so I removed all config files from my home directory, logged
<torb> in and out and voila - text in gtk-programs=)
<nckx>I think you're confusing two things, but I don't blame you (and I don't have a direct explanation for your dmesg). The problem with ‘Maybe your workaround should get into that issue’ is that the former is very specific to libva, and the issue is about OpenGL.
<nckx>acrow: Yes.
<nckx>It is.
<torb>However, after reconfiguring from my guix-home-configuration, they are gone again=/ Something fishy is going on!
<apteryx>acrow: it's not linux-libre, just linux, and it regresses with just one card so far (mine) a 8800 GTS 640 Mo. a mesa developer tried it with 8800 GTS 320 Mo and didn't have the issue, oddly.
<nckx> and friends are ‘video’ as in moviefilms; not ‘video’ as in OpenGL.
<nckx>torb: Hm. I don't use guix home. Did you save the broken configuration files, or can you otherwise try to bisect the issue to narrow down which file/setting it is?
<torb>nckx: I think I just kind of did, killing xsettingsd fixes it - now I will just have to test a bit more=)
<nckx>acrow: Super nitpick, but I have to defend the innocent: IceCat doesn't block anything.
<torb>To se what settings causes the problem
<nckx>It's Cloudflare.
<nckx>torb: …old school. What made you suspect xsettingsd so fast, if anything?
<torb>nckx: Just luck, I had no idea what could cause the problem. I mostly use xsettingsd to set the x-cursor before starting exwm. Would there be a better way of settings that?
<nckx>ACTION wouldn't know, has never customised their cursor.
<acrow>nckx: fair enough regarding icecat/adblock.. I am thankful for your workaround. :)
<lembrun[m]><jlicht> "lembrun: I was just reading..." <- hey there i'm writing my reply right now
<lembrun[m]> So I know why I do get the auth error for pam for the greetd-wlgreet-sway-session
<lembrun[m]>It was because the fucking keyboard layout was in qwerty
<torb>nckx: Turns out there is an environment variable for that: XCURSOR_THEME. This is useful on hidpi-displays, where the default cursor is so small it's difficult to find.
<lembrun[m]>Linux usability is so abysmal in those cases
<lembrun[m]>* those cases, it's not too much asked to have a message with the current keyboard layout, I had the same shit happening with grub, running around in circles before finding out it was the keyboard layout long time ago
<acrow>apteryx: acking your comment and interpreting the call chain to be application(vlx) -> opengl -> mesa -> libva -> kernel. Any layer could contribute to the failure.
<lechner>lembrun[m] / Passwords in Winblows might be equally difficult to enter
<nckx>acrow: I'm no expert but I'm not sure that's quite accurate. As in, it's not one pipeline like that. VA-API is not an OpenGL function. You can decode a video with libva & then render it to an OpenGL surface, but it's not the same pipeline. IIUC!
<nckx>Now, how to to use that info to narrow down the culprit…
<lembrun[m]>Now we need to focus on propagating the keyboard-layout to wlgreet...
<lembrun[m]> * Linux usability is so abysmal in those cases, it's not too much asked to have a message with the current keyboard layout, I had the same shit happening with grub, running around in circles before finding out it was the keyboard layout long time ago
<lembrun[m]>It still does not works for seat for me though. (stuck with blinking cursor and the get_children something error)
<lembrun[m]> * Linux usability is so abysmal in those cases, it's not too much asked to have a message with the current keyboard layout, I had the same shit happening with grub, running around in circles before finding out it was the keyboard layout long time ago
<lembrun[m]>It still does not works with seatd for me though. (stuck with blinking cursor and the get_children something error)
<lechner>i am not sure your keyboard is working right
<apteryx>can't seem to build inkscape on armhf-linux; I think the board overheated on the last attempt
<lembrun[m]>lechner: it does just works for you ?
<lechner>linux, since 1993
<lechner>but guix is a bit harder
<nckx>acrow: <But I don't want to downgrade my kernel.> That's your right, but there's not much point in pursuing this 6.x regression angle further then. That's how one tests.
<nckx>Unless you want to revert just that commit (assuming it reverts cleanly), but that's a lot more work.
<apteryx>acrow: the only thing that changed between works and fails is the kernel version
<nckx>Guix: when Linux gets boring.
<apteryx>acrow: so the rest of the layer is sane. I also tried with latest mesa and could reproduce it.
<djeis1>It looks like something about how emacs' build process discovers libxrender isn't directly compatible with guix, although so far it seems outside of my C build tooling skill to understand why. At the very least, explicitly adding libxrender to the inputs wasn't sufficient for emacs to build against it. The configure tests for xrender seem to fail even with the dep present. Does anyone know if this is a known issue, or know of any workarounds?
<lechner>lembrun[m] / you are right, however, that the support for alternative keyboard layout is not great in any Linux (especially when using XKB). we have some ideas and promise to take a look at all patches
<acrow>nckx: I'm willing to do the rollback to a linux-libre 5 kernel if that is helpful
<mechalain>Hi there, Guix is freezing on boot when populating etc after the most recent pull. I can
<mechalain>t seem to find any relevent boot messages but would love some assistance
<lechner>djeis1 / do you know for a fact that libxrender is incompatible, or is it just not being found?
<lembrun[m]>Yeah ok my bad
<lembrun[m]>for misreading the sway-greetd.conf
<nckx>acrow: I don't see how it wouldn't be? Just add (kernel linux-libre-5.15), reboot, try, then roll back. I don't see the harm. But OTOH what's useful is up to you, since you're the only one who currently sees the issue.
<nckx>I wish I could be more helpful.
<lembrun[m]>you just need to set this... (full message at <>)
<djeis1>lechner: I suspect it's just not being found, all I can really say is that the test program emacs' config script does against it fails (either to compile or run, not sure). I've stepped through the script a little in a dev shell.
<lechner>djeis1 / can you post what you see?
<lechner>pastebin, please
<djeis1>lechner: There isn't really anything too see, apart from after the build completes there's no XRENDER in emacs' system-configuration-features.
<lechner>djeis1 / i'll take the configure script, if there is one
<djeis1>Oh, sure, I can extract the relevant bit of the configure script. One sec.
<lechner>lembrun[m] / i do not currently use wayland but would like to pool ideas and experiences together. my situation is even worse, as i hope to use my own custom keyboard but cannot currently
<lechner>djeis1 / did it create a config.h?
<acrow>nckx: you are quite helpful. I'll give your suggestion a try and see if I get the same outcome apteryx found. Be back in a little bit.
<lechner>djeis1 / or in other words, did it set HAVE_X11?
<nckx>acrow: 👍!
<lechner>djeis1 / i haven't seen such a pedestrian configure script in a while. do we know what the $LINENO test is all about?
<djeis1>It did set HAVE_X11 to 1 in the config.status.
<djeis1>No idea about the LINENO stuff, no.
<djeis1>This is generated from autoconf, I could start digging through that but I have only the barest experience there.
<djeis1> looks like the corresponding bit of the
<lechner>djeis1 / i would go into something like guix shell --development -f guix.scm and make sure the headers are available. then the question is what XRenderQueryExtension probes. Why is the script looking at a runtime variable instead of merely checking the presence of the headers? Do you have to have XRENDER loaded in X for the test to succeed?
<lechner>i guess it could be a function. i though the parentheses were required, but maybe not in modern C
<djeis1>Those headers are on C_INCLUDE_PATH.
<lechner>and the symbols is not found, i.e. "checking for XRenderQueryExtension in -lXrender... NO" ?
<djeis1>I don't get that output, actually. It's the first program that fails, if I've followed what's happening right.
<djeis1>That is, I don't get any normal output about xrender at all.
<lechner>if it's not looking for the symbol, it probably did not find the library
<djeis1>Yeah but, what's weird about that (to me) is that the first program also refers to XRenderQueryExtension.
<argv>Hey I'm confused about setting up guix repository for building packages, need some help. Do I add guix as an ad-hoc dependency in my pure environment? Also what are my GUILE environment variables supposed to be.
<lechner>djeis1 / first off, i think $LINENO refers to the test script
<djeis1>If I've read the script right, I'd only get the "checking ..." output if the first program succeeds, and the boolean is to do with the second program. But, both programs talk about that symbol.
<djeis1>I agree that LINENO is probably to do with the test script.
<lechner>djeis1 / is there anything like CHECK_LIB or SEARCH_LIB earlier in the script?
<djeis1>Nope, and this is the first reference to xrender outside of a long list of variable names near the top of the config script.
<lechner>actually, i think this code does exactly that manually
<lechner>do you see this in config.h? ac_cv_lib_Xrender_XRenderQueryExtension
<lechner>i don't know where that is cached, actually
<lechner>config.status, maybe?
<djeis1>I don't have that in config.status, and I don't have a config.h
<lechner>does line 52 in your paste produce any output?
<djeis1>I'm going to extract out the full text of the first program it generated and see what that gets me...
<djeis1>Oh! I am missing a header: X11/Intrinsic.h
<lechner>there you go
<djeis1>Was so focused on xrender specifically that that one slipped my fingers.
<lechner>djeis1 / no worries. happy building!
<lechner>and thank you for your contributions
<djeis1>Thanks, and likewise!
<apteryx>does the atheros ar9382 wifi chip have free drivers?
<apteryx>I guess so:
<nckx>I swear I'm not just saying this, but I expected nothing else.
<nckx>But certainty is something.
<acrow>nckx: Your workaround again gets the job done.
<nckx>Yeah but it sucks.
<nckx>🤷‍♀️ oh well.
<apteryx>hm, still trying to build inkscape@1.1.1 for armhf-linux
<apteryx>--cores doesn't work with offloading (#44995)
<apteryx>and the usual workaround GUIX_DAEMON_SOCKET=ssh://my-host guix build ... fails with "guix build: error: SSH connection to 'tm' failed: kex error : no match for method compression algo client->server: server [none], client [,zlib]" (the remote runs dropbear, not openssh)
<apteryx>I can probably 'guix build /path/to/package.drv' on the target
<acrow>nckx: how can I dump a summary of my graphics configuration? I'm trying to identify the vlc module that works so that, hypothetically, I can just tell vlc what module to use and avoid the buggy introspection.
<nckx>acrow: I'm not sure what you mean, and probably don't know the answer. I never use VLC and don't know how it organises its modules…
<nckx>There's libva-utils' vainfo (not, by far, a summary of your graphics; just VA). I wonder if it would throw the same FPE…
<nckx>If you want to dive into Mesa, there's this:
<nckx>ACTION not at a PC right now.
<kaelyn>Hi #guix! Does anyone use udisks-service in their system config? I added (udisks-service) to the list of services in a working config and got an error with dbus-system-services:
<kaelyn>wait, I think I found my problem. It looks to already be included--I must've missed it in the source code.
<lechner>kaelyn / i have it here
<kaelyn>lechner: thank you for sharing that! :) My error turned out to be from the service getting added twice.
<apteryx>how can I list the mDNS host names on my local network?
<apteryx>ah, avahi-browse looks like something useful
<lechner>apteryx / i usually use avahi-browse --resolve --all
<apteryx>strangely it hangs after outputting many entries, and misses the one I was looking for
<lechner>try without --resolve
<apteryx>hangs the same
<apteryx>thanks anyway :-)
<apteryx>is there something else than lm-sensors for embedded boards? lm-sensors doesn't show any sensors for the TS-7970 board I'm looking at
<nckx>avahi-browse never ‘finishes’.
<lechner>yeah, that's what i thought
<lechner>apteryx / a patch seems to exist
<apteryx>ah, so that's normal behavior. interesting. it never seems to discover a particular device
<lechner>does that device advertise?
<apteryx>I think so, I can ssh to its .local host name
<nckx>Cool, a Devuice Tree.
<apteryx>lechner: I'm already using that .dtb :-) it's in mainline thanks to the above patch
<apteryx>but still no sensors
<lechner>for some i2c, it helps to load more modules
<lechner>ACTION can see all local ports 22 accepting connections, and never had port 9 (discard) open before
<lechner>apteryx / do you have (service avahi-service-type)
<apteryx>the OS currently running on the thing is not Guix System (it's a raspberry pi with stock os), for the moment
<lechner>apteryx / could the .local resolution come from your router.
<apteryx>ah, perhaps!
<apteryx>it uses librecmc
<apteryx>kaelyn: yes, udisks-service is in the %desktop-services, see in (gnu services desktop)
<lechner>apteryx / do you think i can get librecmc running on this device?
<apteryx>The officially supported devices can be found at
<lechner>yeah, it's not on it
<lechner>this is exotic even for OpenWRT
<apteryx>I got a second hand WNDR3800; installation was as easy as uploading the librecmc image (same as if you'd update from the provider)
<lechner>unfortunately, not having 5 GHz wifi is a blocker, since my 2.4 GHz goes out everytime someone uses the microwave
<lechner>apteryx / maybe have a look at ubuntu's patches, if any
<jeko>yo Guix !
<jeko>I am trying to tidy my guix home configuration a bit but I m stuck !
<jeko>I get a `no code for module (guixrc home emacs)` in my `(guixrc users slim)` module
<jeko>when I do `guix home container $GUIXRC_PATH/users/slim.scm`
<rekado>does the load path include the directory containing the file guixrc/home/emacs.scm?
<jeko>$GUILE_LOAD_PATH ? I appended /home/jeko/Workspace/guixrc to it
<rekado>where is guixrc/home/emacs.scm?
<rekado>the module (guixrc home emacs) must correspond to a file $SOMEWHERE/guixrc/home/emacs.scm and $SOMEWHERE must be on the load path.
<rekado>not $SOMEWHERE/guixrc
<rekado>not $SOMEWHERE/guixrc/home
<rekado>if $SOMEWHERE is too big then move the files to a clean sub-directory
<mekeor[m]>hello. is it possible to use tarballs from guix-pack on mac-os?
<rekado>if “guixrc” *is* that directory put $SOMEWHERE/guixrc on the load path and rename all the modules to drop the “guixrc” part
<rekado>mekeor[m]: no
<rekado>the stuff that “guix pack” generates are sections of /gnu/store wrapped up in an archive.
<rekado>whatever is in /gnu/store is based on the shared bootstrap with glibc
<rekado>our glibc does not work with the macos kernel
<jeko>rekado: oooooh right ! guixrc is in $HOME/Workspace. adding this to the load path did it ! thaaaank you !!!
<rekado>I don’t think there are even any patches out there to make glibc work on darwin.
<rekado>the approved way to build software for macos is to use their proprietary toolchain Xcode.
<rekado>this includes a proprietary C library.
<rekado>IIUC this is what nix people do: cut the dependency graph at an arbitrary point and graft it onto the huge binary blob that is Xcode and the macos SDK.
<rekado>best we can do is to build a virtual machine image and use that on macos.
<KarlJoad>How simple is it to define my own meta-package? Meaning a package composed of other packages that I may want. Would it just be a matter of adding the required packages as propagated-inputs?
<nckx>KarlJoad: That is how, e.g., gnome works.
<rekado>KarlJoad: yes, but depending on the situation a *package* may not be the best tool here
<rekado>for environments we use manifests, for example.
<jackhill_>what do Nix and homebrew use on MacOS? At any rate, depending on the use case, I wonder if a VM with linux-libre is a better option anyway. I don't know of any study about cross-kernel reproducability for example.
<rekado>jackhill: xcode
<jackhill>sadface, but I guess not surprising
<KarlJoad>rekado: It would be for declaration in a guix-home config. I already use manifests for projects.
<lechner>Hi, is the link about "Writing Manifests" here working for you
<nckx>Does guix home not support manifests/lists?
<nckx>KarlJoad: ☝
<KarlJoad>nckx: I have no idea. I haven't looked into it. Just a general question I had. I sometimes want a set of packages to have their %outputs available where each can find their dependencies without needing to clutter the outer layer I interact with.
<nckx>Er, OK.
<nckx>That sounds like there's a deeper packaging flaw at play.
<KarlJoad>nckx: I don't think there is. For example, I use straight.el and org-roam. straight.el will compile an sql binary, which requires a C compiler. The default Emacs does not have one to find, so that fails.
<KarlJoad>But I do enough dumb code testing to warrant having GCC installed all the time.
<jeko>bedtimes! see you all!
<KarlJoad>Does anyone here keep their guix-deploy configs in a channel? Mine keeps breaking Cuirass.