IRC channel logs
2023-05-31.log
back to list of logs
<ieugen[m]>hi, noob question, how can I chmod a file (make it exec) after install-plan phase? <ieugen[m]>I have ` (chmod (string-append #$output "/bin/regbot") #o755)` and I get "Unbound variable: ungexp" <lemos1898>can i have multiple (source camps in a package? i need to fetch some C++ libs from github for the package <bjc>ieugen[m]: ‘ungexp’ (or ‘#$’) is only valid inside a gexp; otherwise you get that error <bjc>you'll need to wrap something with ‘#~’, but i couldn't tell you where <lemos1898>also, someone had to deal with meson wrap dependencies? <lemos1898>those are the things i am dealing with, gentoo had a network sandbox, seems like guix have the same <lemos1898>but if there's someway i can let those dependencies be fetched, i will do it, since it would keep the package going on for some time with easy maintenance <ieugen[m]>I think I found that gexp are not working in builders. <bumble[m]>does anyone have a recommended pattern for configuring swaylock these days? <bumble[m]>it seems there was a recent change related to pam and setuid that broke my previously working swaylock configuration <bumble[m]>I've browsed the mailing list and attempted some of the patterns there but those patterns introduce xord dependencies which is not un-desirable for one who would use swaylock <lemos1898>search for "swaylock hyprland" and you will probably find it <efraim>it's for the test data, I'd skip the build <efraim>AwesomeAdam54321: also, have you taken a look at the rust-team branch? It might have some of the crates you want packaged <AwesomeAdam54321>efraim: Thanks, I want to rebase on top of the rust-team branch but I don't know how <efraim>the easiest might be a git worktree, then its like you have a checkout for each branch <efraim>I would then cherry-pick the commits from one worktree to the other <civodul>cbaines: hi! with non-blocking exec-query, one has to make sure that only one fiber calls exec-query on a given connection at a time <civodul>otherwise you end up starting several queries and you get an exception ("another command is already in progress") <civodul>hence, i'm resurrecting with-critical-section in Cuirass <cbaines>indeed, there's still a need to manage having only one fiber/thread access each connection at a time <cbaines>I did try running the data service locally with the squee changes, and I didn't spot any issues <cbaines>have you any further insights in to what went wrong in Cuirass yesterday? <civodul>so far in Cuirass, database operations were carried out in a non-fiberized thread <civodul>so the whole suspend-while-command-is-in-progress thing cannot work <civodul>now i'm removing the worker thread and replacing it by critical sections <cbaines>I don't follow what you think wont work? <civodul>which is funny, because we went in the opposite direction in the past :-) <civodul>cbaines: i think current-read-waiter assumes you're in a fiber <cbaines>fibers probably sets that in a way that assumes you're in a fiber <cbaines>but that probably indicates you're doing something funny when you start the threads <cbaines>does Cuirass somehow start those threads after it's setup fibers? <civodul>the new thread inherits the dynamic environment of its parent, doesn't it? <cbaines>yep, but the data service for example makes sure to start those threads prior to starting fibers <civodul>this mixture of fibers and non-fiber threads is best avoided anyway :-) <civodul>there was a time where we'd (parameterize ((current-fiber #f)) ...) in non-fiber threads <cbaines>e.g. for using SQLite in the build coordinator, you do that in non-fiber threads <civodul>here i'm talking about a pure-squee context <cbaines>yeah, you mentioned critical sections though, I'd consider structuring this as a connection pool <civodul>so you'd grab a connection from the pool, do your thing, and release it? <cbaines>indeed, and a simple pool could just have a fixed number of connections <civodul>but what's the advantage of having several connections to postgres? <civodul>it's not gonna make it any faster, right? <cbaines>since you'll be able to execute queries in parallel <cbaines>more importantly, less expensive queries will be less likely to be blocked by slower ones <cbaines>so the performance will be more consistent (assuming you don't run out of connections) <civodul>i knew i'd rather discuss this with you :-) <cbaines>civodul, have you tried moving the with-database call in (cuirass scripts web) yet? <cbaines>if you move it to wrap run-fibers, that might solve the immediate issue <civodul>i thought (foolishly?) that i'd rather go straight and refactor the thing <roman1>So, I have followed the instructions from the manual on fiddling with substitute-urls, but I don't seem to understand why nothing works :/ <cbaines>roman1, if nothing works, I'd start by fixing your keyboard :) <roman1>I am trying to alter the default substitute urls during the installation of Guix system, and to achieve that I retyped the code from the manual, the one that uses (modify-services ...) function, I tried to modify both %desktop-services and the resulting list comprising my own services (the default ones suggested by installer) and the %desktop-services <roman1>Actually, everything works pretty fine. It just doesn't do what I'm intending to do :( I must be missing something... <cbaines>have you got Guix installed and you're reconfiguring, or are you trying to do something else? <roman1>I am trying to install the system on a virtual machine (qemu). So far my guess is that config.scm doesn't matter during the installation, but I guess it's wrong <cbaines>and are you trying to affect the substitute behaviour while installing, or of the installed system, or both? <roman1>For both. I used to change tty after finishing doing stuff in graphical installer and pass --substitute-urls="bordeaux...etc" to some "guix system init" command (I don't remember the exact command now) <roman1>But I guess that I could redefine substitute urls from config.scm, couldn't I? <cbaines>note that you might need to also change the ACL so that the substitutes are accepted <roman1>I am trying to remove "ci.guix.gnu.org" from the substitutes list actually, so I think there's no need to change it, as far as I understood how it works <roman1>Because it's rather unreacheable for me, so I have to wait a lot while installing the system as guix numerously tries to connect to it <cbaines>ok, I'm not sure if it's possible using the graphical installer, but adding --substitute-urls=... when doing guix system init should avoid trying to use ci.guix.gnu.org <cbaines>I'd address the configuration of the installed system once you've got it installed <roman1>Ah, thank you, I'll go with the cli-way then <janneke>jpoiret: headsup, (re)building guile-final right now, praying for it to get past phase `check'... <efraim>Am I supposed to use the gcc-2.3.3 testsuite with gcc-2.7.2.3? <janneke>ACTION is sure to use --keep-failed this time round to have a better chance of fixing any remaining problems so that it may succeed the next build <janneke>anyway, gives /me lots of time to sit in the sun today ;) <civodul>hopefully there won't be new Guile test failures <janneke>yes thanks, ...finally approching "hello" <AwesomeAdam54321>Is it possible to have Guix System and another distro to coexist and show up in GRUB for BIOS? <cdo256_>Hey all, does anyone know what stage fish is in for guix home? It's in the source code but not documented in the manual. <cdo256_>I'm getting: "In procedure service-type-default-value: Wrong type argument: #<<service> type: #<service-type cdo-fish-extension ..." <cdo256_>It was okay until I put in the (config ) argument and then I moved it to an extension because I got "type argument in position 1 (expecting empty list): #<gexp ("") ..." <cdo256_>But it wasn't clear why it expected an empty list <cdo256_>I had a simple gexp #~("") to test it out. The same thing happens with bash on the (bash-profile ) argument so it must be something I'm doing wrong. <cdo256_>I've found it difficult to find examples of people's guix homes. <cbaines>ngz, yeah, I think there's some issue with querying issues.guix.gnu.org <cow_2001>doing a `guix pull` on a Debian(-like) would populate ~/.cache/ and not some system-wide cache. is there a way of using a system-wide cache? i have a bunch of users all using guix in their respective home directories. <civodul>cbaines: the connection pool idea works great! but... there are parts of the code (the zmq processes) that are multi-threaded and not fiberized 😢 <attila_lendvai>cow_2001, set up a global dir with the right permissions, and ln -s it there? <cbaines>civodul, cool, maybe I can copy the connection pool for the data service once you're finished :D <futurile>Q: is it true that there isn't a binary substitution for the latest ungoogled-chromium? And, is there some way I can see where it is in the queue? <cbaines>civodul, I'd presume the bits using zmq are already separate from fibers though, so that should be fine? <civodul>cbaines: not really, because i implemented the connection pool as a fiber <futurile>attila_lendvai: yes thanks - I checked `guix weather` and it says there's no substitute which Im finding surprprising. I have the same response as you to the CI and Data system - I find it hard to decipher heh :-) <civodul>this is starting to look like a rabbit hole <attila_lendvai>futurile, there's ongoing discussion about how to keep substitute availability higher than currently. bear with the opensource fairies for a wee bit longer... :) <cow_2001>attila_lendvai: isn't there some guix daemon in charge of speaking with users? <attila_lendvai>cow_2001, i don't know how guix works on a foreign distro, sorry. <cbaines>civodul, you can probably use the fiber stuff from a non fiber thread in some way. Alternatively just open a connection in that thread and ignore the pool. <futurile>cbaines: ah OK thanks for the link - that's helpful - will add it to my list <civodul>cbaines: yeah, i'll ignore the pool in non-fiber context <hmmmhm1>Hi, Guix (: I have a quick question about Guix system installation process. So, there is /mnt directory, it is where all the compiled stuff will be copied. But the stuff gets compiled and stored under /gnu/store directory residing in RAM, am I right? <cbaines>hmmmhm1, the herd start cow-store /mnt bit affects this <hmmmhm1>Ah, right, just saw it in the installation manual, thank you for reminding about this option, that's the thing I needed <mrvdb>hi guix, getting an error on 'guix install btrfs-progs` https://dpaste.org/giKXD has the scheme form and the error. This is on a powerpc64le machine. Anything I can do locally, or file this as an issue? <ngz>mrvdb: you probably want #$, not #$@ <mrvdb>ngz: this is directly from gnu/packages/linux.scm <ngz>Then linux.scm probably wants #$@, not #$ :) <mrvdb>as in, what I see when doing `guix edit btrfs-progs` <ngz>mrvdb: let me fix that. <val[m]>bumble深く傷を: William ieugen pjals FYI, the Matrix bridge is currently broken, that's why this channel is so silent on the Matrix side <val[m]>@ the others: I don't think your messages got through to IRC <cdo256_>They seem to be getting through now at least <vlorentz>cdo256_: IRC->Matrix is completely unavailable, Matrix->IRC only works for people who joined the room before Friday <ngz>mrvdb: I think this is now fixed. <parnikkapore>yeah IRC messages don't seem to be arriving on the Matrix side still <parnikkapore>So, I'm using `guix shell` to (say) set up a Python development environment. I have (say) python-protobuf in the invocation and an incompatible version of python-protobuf in my home directory's site-packages. What's the official way to make sure that the Guix-provided protobuf is used? <parnikkapore>I could `PYTHONPATH="$GUIX_PYTHONPATH" but that sounds like a hack <parnikkapore>Also the same question for when packaging Python-based apps ig <Parnikkapore_m>> <vlorentz> cdo256_: IRC->Matrix is completely unavailable, Matrix->IRC only works for people who joined the room before Friday <ieugen[m]>thanks val : is someone working to fix the bridge? any ETA? <civodul>cbaines: i pushed the Cuirass changes; now to see if i can test them in "real" conditions on a test instance <ngz>mrvdb: Thanks for the report! <vlorentz>ieugen[m]: Element isn't answering at all <vlorentz>(Element is the company hosting the bridge) <val[m]><ieugen[m]> "thanks val : is someone working..." <- Element (the company hosting the bridge) isn't answering at all <cel7t>I think I'll just set up my own IRC bouncer <sozuba>Hi, I am just exploring (reading about Guix). https://packages.guix.gnu.org/ has all the available binary packages right? Where can i find the template/scheme list for available source packages? or is that something i have to setup <Guest28>On system boot, I need to run "sudo herd restart libvirtd" or without I can't connect in virt-manager. Does anyone else have the same issue? <vbimvnc>I just installed guix on a fresh instance of debian bookworm in a vm to test it out. I used `guix home import` to create a home configuration and switched to it. It currently only has emacs and glibc-locales installled. I'm running KDE and when I logged out and back in, none of the applications (including system applications installed by debian) are detected. I did some testing and found out that this only happens when the `XDG_DATA_DIRS` environment <vbimvnc>variables contains both `$HOME/.guix-home/profile/share` and `$HOME/.guix-profile/share`. I can have either of them, but not both. Does anyone know what is going on here? <jpoiret>vbimvnc: you should add XDG_DATA_DIRS=/usr/local/share/:/usr/share/ before the guix profiles are sourced <jpoiret>basically, XDG_DATA_DIRS being empty or not set, the spec defines it as being equivalent to it having the value /usr/local/share/:/usr/share/, but so if you append something to it these two dirs "disappear" <jpoiret>this spec is just silly tbh, either these dirs are *always* included or there shouldn't be such default behavior <vbimvnc>jpoiret: The /usr/local/share and /usr/share are being set. The full XDG_DATA_DIRS I have is "$HOME/.guix-home/profile/share:$HOME/.guix-profile/share:/usr/local/share:/usr/share" <vbimvnc>For some reason when both the .guix-home and .guix-profile are in XDG_DATA_DIRS I'm having the issue <vbimvnc>This might be a KDE bug, but it only seems to happen when both those two directories are included. <vbimvnc>This is really weird. I've moved .guix-profile and .guix-home so that they don't exist and explicitly set XDG_DATA_DIRS to $HOME/.guix-home/profile/share:$HOME/.guix-profile/share:/usr/local/share:/usr/share, I still have the issue. If I put a typo on one of the guix entries, e.g. .goix-home instead of .guix-home, everything is find. <jpoiret>maybe more than 3 entries is too much? :p <jpoiret>do you have anything installed in your guix profile? <vbimvnc>I did, but tried uninstalling everything and since I tried using that XDG_DATA_DIRS when the directories they refered to didn't exist I'm assuming is some bug in KDE. <vbimvnc>My nixos install has a far longer XDG_DATA_DIRS with more entries so it shouldn't be that. Especially since changing one character fixes things. <helaoban>is there a reason substitute downloads from ci.guix.gnu.org are so slow? It feels like I'm getting throttled. Download will spike to 3mib/s at the beginning and then drop to 660k... <rekado>helaoban: there is a reason but it’s not something we have control over <rekado>if you try from AWS EC2 instances in Frankfurt, for example, you’ll get really good speed <helaoban>rekado: This is a problem for me trying to download the texlive package, which is around 4 gigs. <helaoban>I'm after a particular package "pgfplots" that isn't defined as its own package <helaoban>I could quickly define a package, but the tex packages are complicated. There's a template procedure called simple-texlive-package that makes things much easier. <helaoban>any reason no to export it so one could quickly spin up a standalone tex package outside the tex.scm module? <rekado>helaoban: you can access it with (@@ (gnu packages tex) simple-texlive-package) <rekado>it was never meant for public use, though <rekado>I added it because the tex build system wasn’t very good, and we hadn’t quite figured out what to do with tex packages <rekado>the goal was to eventually remove simple-texlive-package and improve the tex build system <ngz>rekado: which is currently done in some branch of the repository <jackhill>Is there a way to get my nginx service to start after my wireguard service? Currently nginx doesn't start on boot becasue it can't bind to the wireguard IP. <helaoban>rekado: ah! I knew there was a way to get a non-exported binding but couldn't remember the syntax. Do you know where in the guile docs this is documented? <bumble[m]>for the last day or two guix home reconfigure fails with this message here <bumble[m]>I guess I need to start using irc directly again <bumble[m]>but guix home reconfigure fails with the error I posted above and... so I can't install an irc desktop client until that is resolved :( <elevenkb>Hello there, I asked a similar question before but I can't find it on the IRC logs (perhaps I should use a bouncer some day). <elevenkb>How do you populate "/etc/guix/machines.scm" on Guix System? <bumble[m]>I'm glad there is still activity around guix I was a little worried people were leaving guix and disappearing <elevenkb>Nvm. I stole [cough...] copied code from the Guix Sysadmins. <elevenkb>A positively *wonderful* way to use free cloud credits. <elevenkb>Enjoying the peacefully fanless build right now. Another cool thing about this, is that I can probably just use a package transformation to build with debug symbols. <lemos1898>is there someway i can let the guix build command access the internet? <lemos1898>if there is some way i can include them as source without a error of duplicate field <elevenkb>lemos1898: n00b q: is there anything wrong declaring those as separate packages and including them as inputs? <lemos1898>i wish i could do that, but those are kinda really unstable libs and stuff <lemos1898>if trhere is someway i can run a script just to git clone them and place where they can be found <elevenkb>i'm pretty sure that it's possible to let guix download the latest version of the source code without checking hashes. <lemos1898>hmmm, how? i tried with source code or uri/url fields, all of them led to duplicate field errors :/ <elevenkb>ah, i guess your issue is that maybe you have a package that depends on both of those libraries and you don't want to rebuild /that/ package from scratch each time its dependencies change. <lemos1898>i just want to download them and place correctlyt <lemos1898>include them like in SRC_URI and to the changes in a script like src_prepare <apteryx>guix time-machine experts; have you even seen: error: %guix-register-program: unbound variable ? <apteryx>I get that with: LANG=C guix time-machine --commit=9b9e147117e6009451d7acc1f8f156e041263e32 -- shell --pure -D openconnect autoconf automake libtool <al1r4d21>new user here :) .. how to remove all `guix-system` packages? <bumblehead>I don't know the answer to that but may I ask; why would one want to remove all guix-system packages? <al1r4d>Hi, why i cant install telegram-desktop 4.8.1? I only got 4.2.2 <bumblehead>did you try `guix pull` before trying to install telegram-desktop? <Guest53>Good morning, I am new to guix and currently trying to create a home configuration. After running `guix home import` and testing the resulting config with `guix home container`, i tried adding a `home-files-service-type`. The altered home config can't be build and i don't really see why. <bumblehead>did you check the file at "View build log at '/var/log/guix/drvs/9v/lp4knbhcq593dldszqwr1xkibzikm9-files.drv.gz'."? <Guest53>Ohh boy. Wrong quote. Sry for the noise. <bumblehead>it would be great if there would be more noise here <bumblehead>there was a lot of activity and discusttion here until some point a few weeks ago <listentolist>Hey! When running `gtags` with the pygments plugin it complains about not finding python although it is installed: /usr/bin/env: ‘python’: No such file or directory. How can I solve this problem? <Guest53>bumblehead: is there a reason for it? Logs seem quite active. <rekado>bumblehead: I would not call this channel quiet <ChocolettePalett>A prospective topic for discussion: I can not install GNU/Guix System for more than a month of trying. Opinions? <rekado>ChocolettePalett: it may be better to write to help-guix@gnu.org with details <bumble[m]>guix is difficult to install. As a point of comparison, arch linux or gentoo might be technically more difficult to install than guix but those systems have a lot of documentation and various walkthrough guides that make those systems overall easier to install (my opinion) <ChocolettePalett>I guess I'll write to the mailing list but later, since I'm pretty sure I mess something up in a very embarrassing way <iyzsong>i'd say it needs some determination, and i always install a minimal system first, and then reconfigure later (so at least that's usually something else fails instead of installation) <iyzsong>ChocolettePalett: seems nothing wrong to me <ChocolettePalett>Actually GNU/Guix isn't very difficult to install and it has all the documentation one might need, it is just very different from all the other distros, e.g. when I was installing Gentoo and nothing worked I almost immediately thought about checking fstab and it was the right thing. When GNU/Guix shows me some errors I never know what to do... <ChocolettePalett>Though GNU/Guix makes me feel like it's the first time I use a computer as a superuser, which is great in some way <ChocolettePalett>Also I guess I'll try to install the minimal system now, which is a very good idea <bumble[m]>maybe the "manual" cli install would work well for you <ChocolettePalett>And I think that cli install won't make any difference since all the errors I get emerge after I do guix system init, therefore there is no difference how I do the preceding steps <rekado>ChocolettePalett: why can’t you use ci.guix.gnu.org? <iyzsong>🤔 guix usually report errors in the context/viewport of guile code, some understand of the code is needed to know what to do.. <rekado>I’ve decided to get back into recording some music again, but I’m intimidated by the changes since I’ve last done this. <rekado>back then I’d just launch jackd on the second sound card <rekado>but with jack-2 (or pipewire?) that’s all a little more complicated <rekado>for example, when I launch ardour it wants to launch jack by itself, over dbus <rekado>but when it has launched jack successfully no other application can see jack <rekado>so I guess I need to make other applications aware of the dbus session bus <rekado>but messing with /proc/*/environ is quite a buzz kill <iyzsong>um, does those application shave same DBUS_SESSION_BUS_ADDRESS, it usually set by DE/WM with dbus-run-session once for all. <rekado>I’m doing this in Gnome, so there really should be a shared dbus session <jpoiret>rekado: are you already using pipewire? <jpoiret>if so you should just be able to start pipewire-jack in a console or similar <jpoiret>nb: i've never tried it though so take it with a grain of salt <rekado>if I need to do something consciously to use pipewire then no: I’m not using pipewire <jpoiret>ah, then it might be quite annoying, esp. because you're on gnome <jpoiret>i do recommend making the switch though, it's much nicer than pulseaudio <rekado>I’m willing to do without gnome, but I like the convenience of not having to mess with xrandr when doing presentations ;) <iyzsong>seems the irc bridge works for a while but now break again.. ChocolettePalett said ci doesn't work due to "Availability issues depending on the region" <iyzsong[m]>look more like a networking issue to me, i'll feel the same pain without substitutes mirrors 🥲 <jpoiret>rekado: there's arandr if you still want to use gnome <jpoiret>not sure about wayland alternative though (but you should use wayland too!) <efraim>rewriting the comments in the openblas package is like bikeshedding with myself <cel7t><ChocolettePalett> "*keeping you informed*..." <- > <@chocolettepalette:matrix.org> *keeping you informed* <cel7t>> Installing an austere terminal-only system worked well! <cel7t>Were you having issues with running a graphical system? <cel7t>Usually when that happens it's because your system's graphics drivers aren't present in the libre kernel <rekado>cbaines: yes. I see that berlin was rebooted a little over a day ago <cel7t>Also wow the bridge is still broken 😞 <rekado>I wasn’t told, so the debbugs sync isn’t running <ChocolettePalett>cel7t: That's a good hint actually, I'll check if my PC is compatible with the LibreKernel <cel7t>Looks like the bridge is back up <rekado>cbaines: started the sync; gotta wait a bit for mumi to index things <cel7t>ChocolettePalette: if a pure black screen is showing up when you're trying to run it graphically I'm 99% sure that's the reason <ChocolettePalett>Oh, it isn't the issue then. I failed to install the system in the first place... Well, I will try to install Xorg & Co. and see if it works that way <cel7t>Hmm I think the ISO comes with XFCE <civodul>rekado: oh my bad; i'm the one who rebooted (as discussed on guix-sysadmin) <civodul>should we set up a service for debbugs sync? <ChocolettePalett>Isn't it supposed to be downloaded / built and installed? But yes, there was an option to install XFCE via the graphical installer (didn't work well though) <ChocolettePalett>Ah, and it might be an important thing to note: I installed the system using a Qemu virtual machine, but I don't think it could cause more problems than installing to physical hardware <cel7t>I personally would recommend using the official QEMU qcow2 image when running it in a Virtual Machine <cel7t>The last time I tried to install Guix inside a virtual machine it caused some strange errors for me too, but the qcow2 image works flawlessly <iska>I'm trying to use cryptsetup reencrypt on my hard drive, but even after umount it still counts as open. <iska>`Failed to open /dev/sdc1 in exclusive mode (already mapped or mounted).` <iska>grepping /proc/*/fd/ shows nothing. <rekado>civodul: no worries! I remember seeing the mail but failed to realize that I’d have to do something about it :) <cel7t>There's also a specific way to create a VM image if you don't wanna use a prebuilt one <rekado>civodul: we can automate this, but we’d probably want to keep the rsync password secret <cel7t>It's generally faster than scp too <civodul>cbaines: hi! qa-frontpage is currently down (on purpose maybe?) <ChocolettePalett>cel7t: > <@cel7t:matrix.org> There's also a specific way to create a VM image if you don't wanna use a prebuilt one <ChocolettePalett>I was just training to see if I can manage to install Guix before letting it format my drive (reinstalling Gentoo in case the installation doesn't go well takes quite a lot of time) <cel7t>I frequently rsync over ssh to sync notes and stuff across my devices <cel7t>rsync -avzP username@remotehost:/some/path /some/path/on/this/machine <cel7t>It'll ask you for the password for your id_rsa file <cbaines>civodul, yeah, bayfront is currently on the MDC naughty list, so I've turned it off for now <civodul>oh, because it generates too much traffic? <cel7t>ChocolettePalette: I see. The install process itself is extremely straightforward, what you should watch out for is if your hardware is supported <cbaines>civodul, I'm not really sure. The qa-frontpage does make quite a few requests regularly, to stay up to date with all the guix-patches issues. <rekado>I’ll open a ticket with the networking team today <rekado>other than bayfront, do we need more IPs on the “nice” list? <mekeor>ACTION realizes it makes sense to have two local workspaces of the guix repository, so that one can be kept 'built', so that it's much faster (in fact, only then usably fast) to load modules from it into geiser/emacs, because when you checkout a new commit, geiser will attempt to recompile all scheme code <mekeor>AwesomeAdam54321: compared to plain scheme-mode? <mekeor>guix-devel-mode is recommended at (info "(guix) Formatting Code") <mekeor>AwesomeAdam54321: are you interested in my emacs config regarding guix/geiser? <mekeor>make sure you have needed packages installed <jpoiret>mekeor: right, i also have a worktree for rebasing branches <Kabouik>What are my options when guix pull stays stuck at "building /gnu/store/...-module-import.drv" for hours? I already disabled extra channels (because I had a guix pull issue on my other machine recently due to the guix-cran channel) but this does not help. I'm not sure what is wrong, even --verbosity=2 doesn't provide much extra information. <rekado>cbaines: I opened a ticket. I’ll get back to you once I get a response. <rekado>oh, that’s a first for me: “collecting ... Fatal Python error: Aborted” <rekado>that’s when trying to run pytest in a python package <rekado>it happens only when python-pyqt is among the inputs <ChocolettePalett>Another quick GNU/Guix-riddle: after doing guix pull I got 10101 new commits but in order to reconfigure the system, I had to use --allow-downgrades. Would downloading the latest GNU/Guix system installation image save me from having to --allow-downgrades? <cbaines>I've also changed a bit of the qa-frontpage to pipeline some requests to issues.guix.gnu.org, so if the firewall was disliking the rate of new connections, that might help avoid issues <rekado>ChocolettePalett: a common point of confusion is the effective version of Guix <rekado>ChocolettePalett: when you run “guix pull” you’ll get a new “guix” executable in ~/.config/guix/current/bin. <rekado>you need to make sure that this directory comes first in the list of directories in PATH <rekado>and then make sure that your shell is aware of the updated executable (with “hash guix”) <rekado>and when you use sudo or switch to the root user to reconfigure your system you need to make sure that the effective “guix” executable is in fact the one “guix pull” installed <ChocolettePalett>This is a truly helpful piece of advice, I wish it were on the (after) installation manual <rekado>Guix itself prints that sort of information after “guix pull”, but it’s easy to lose it in all the other output <gabber>i'm having issues getting Guix to run on some aarch64 platform. first question (though i am not 100% sure this is an actual show-stopper): can i create a vfat boot partition that is shown (with `fdisk -l`) as "W95 FAT32 (LBA)" (Id: c) and not "EFI (FAT-12/16/32)" (Id: ef) through Guix disk image creation? <gabber>mirai: there it says a small subset of Ada83 :) thanks for the info! should i open an issue on the tracker for that? <chomwitt>When we want to mount a usb flash do we use a file system entry in /etc/config.scm ? <gabber>chomwitt: you can add your custom entries to your /etc/fstab through your config.scm file, but this doesn't prohibit you from mounting drives manually <jpoiret>you can also use udisks if you're so inclined <janneke>jpoiret: headsup, i was almost ready to send my commencement/native build patch set but coreutils-final had some new hanging and failing tests <janneke>identified those, patched coreutils,...but that means rebuild world <jpoiret>only on Hurd? we can special case those patches <jpoiret>ie. make it a Hurd-only world rebuild <janneke>sure, that's what i did and what i mean <janneke>touching coreutils, even if its' "only" for the hurd, means you rebuild all of commencement from coreutils-boot0 up <janneke>even though coreutils-boot0 has #:tests? #f, but yeah, it inherits all stages... <jpoiret>oh, you're patching the inherited coreutils itself? right, that's annoying <janneke>ACTION is building gcc-final right now, fingers crossed <jpoiret>are those patches upstreamed? or does Debian have them? <janneke>i'm just disabling these patches right now <jpoiret>Hurd is such a fast moving target when it comes to things like coreutils, glibc and friends <janneke>it could be anything, our bootstrap-binaries, the fact that commencement uses glibc-2.35, dunno <jpoiret>oh! could we not use libc-for-target there? <janneke>i guess that disabling tests at this point is less than great, but OK to get everything to build, and we can worry about upgrading glibc in commencement for hurd re-enabling tests later? <phf>Hello Guix! Assume that ~list_of_channel~ is a list of channels like in ~channels.scm~. How to set ~home-channels-service-type~ directly? Somehting like: ~(service home-channels-service-type ...)~ ? What is ~...~? <phf>Yes, e.g. é(define list_of_channel (list (channel ...) ...))~ <phf>Yes, e.g. ~(define list_of_channel (list (channel ...) ...))~ <phf>You mean: « execute (service home-channels-service-type (guix-configuration list_of_channel)) » ? <gabber>i've created a Guix system image for aarch64 -- where are the kernel and initrd files? they don't appear in neither the ext4's /boot nor in the VFat partitions? <phf>HiltonChain[m]: OK, thanks, it works! <gabber>aaaaahhh.. wait. the files are in the store and EFI/foo.efi are binaries pointing to the right places, right? <HiltonChain[m]><phf> "You mean: « execute (service..." <- To be clear, I mean `(service home-channels-service-type list_of_channel)`. If that's what works for you? <phf><HiltonChain[m]> Yes it works. So, to infer how to configure a service-type as home-channels-service-type (reproduced here: https://bpa.st/GDNQS) one needs to examine the ~default-value~ value? In this case, ~default-value~ is a liste of channels so one should set the value of ~~home-channels-service-type~ as a list of channels? <phf><HiltonChain[m]> Ok. Thank you! <hwpplayer1>how to start learning Guix and (it's programming language ?) ? <AwesomeAdam54321>There's also the Guix Cookbook if you're looking for examples you can follow along <civodul>there's one issue left: i'd like it to prescribe a convention that people can blindly follow, but i realized i used a convention that's not the one i would recommend :-) <civodul>namely, i think we should suggest putting files under .guix/ <civodul>like .guix/modules/guile-package.scm <AwesomeAdam54321>civodul: would it be an issue that .guix is hidden as a dotfile when doing ls? <bjc>civodul: i can't parse “so that `guix pull` doesn’t end loading unrelated `.scm` files when someone pulls the channel” does that mean that the pull will fail for some reason? <bjc>also, shouldn't the ‘guix shell’ invocations be ‘guix shell -D’ in order to get access to the delevopment environment inputs? or does ‘shell’ work differently with ‘guix.scm’ files? <AwesomeAdam54321>bjc: It means that `guix pull` won't load .scm files that aren't package definitions, because a channel directory with only package definitions is specified <civodul>bjc: "guix shell" (without arguments) is equivalent to "guix shell -D -f guix.scm" <bjc>so, instead of “doesn't end loading”, maybe just say “doesn't load”? the word “end” is confusing to me in that spot <bjc>i wouldn't regard the current wording as particularly grammatical, anyway <bjc>(“end while loading” would be better) <bjc>“doesn't end up loading” is fine, too. it'd mean the same thing as “doesn't load” <bjc>adding a channel directly to the source repo is clever. i'm going to have to start doing that <civodul>"doesn't load" is probably better, indeed <civodul>(oh there was a typo, t'was supposed to be "doesn't end up loading") <bjc>“It's as simple as this” → “It's as simple as that” on 461 <attila_lendvai>civodul, i suggest using `guix shell --long-names` in such articles instead of `guix shell -CP`. or at least mention both. <attila_lendvai>i'm playing with guix for a year now, but i still had to look up -P <bjc>should the manifest sample code have a “FIXME”? <shtwzrd[m]>howdy guix :) I am attempting to use guix inside a vm (guix system vm) which i understand shares the store with the host. Very cool, this is much faster than building an image and then running qemu with it manually. but I have the problem that I basically can't use guix from within the vm, because all the commands i want to do fail with error: changing ownership of path '/gnu/store': Read-only filesystem <shtwzrd[m]>is that by design or is there something i could do to fix this? My use case boils down to wanting to launch a vm for testing out a system config, and then iterate over that config continuously within the vm with guix system reconfigure. i am doing this with the host being a foreign distro, if that matters <civodul>attila_lendvai: re long options, will do! <civodul>bjc: the FIXME is because like i wrote above i'm proposing a convention that i haven't followed so far :-) <bjc>shtwzrd[m]: you shouldn't be manipulating the guix store from more than one daemon at a time <shtwzrd[m]>/gnu/store is owned by root both on the host and within the vm so i am unsure why guix is attempting to chown it in the first place <bjc>the store is mounted read-only to prevent shenanigans <shtwzrd[m]>bjc: ah, that does make sense. but is there some way to have my vm use the hosts daemon then? <bjc>there's a socket you may be able to forward along, but i'm not sure what the advantage is <bjc>it shouldn't be much, if any, faster to rebuild the system inside the vm compared to outside <shtwzrd[m]>that's a fair point, i assumed (or i guess more like hoped) that it'd be faster and shorten my feedback loop but i don't know that it would be. i suppose arguably it might even be a bit slower because i don't give the vm as many cores, and the natural overhead of 9p filesystem and using a hypervisor <shtwzrd[m]>most of the time is eaten just computing the derivation anyway so maybe killing the vm and continuously rebuilding with guix system vm is really as fast as it'll get. <bjc>if you're just adding and removing packages, subsequent rebuilds should be fairly fast (as long as there's no ‘guix pull’ or ‘guix gc’ in between) <attila_lendvai>civodul, not sure it's worth it, but maybe: also mention git-checkout as an option instead of local-file, which can be used to only test recorded changes. i find that useful sometimes (e.g. when running my code in a VM using a patched shepherd. i don't want shepherd be rebuilt for every changed comment). <shtwzrd[m]>bjc: ahh, you are right. it turns out i had a badly written make target that was causing a pull inbetween my reconfigures. i've fixed it and it's more like 30s to reconfigure now, pretty nice. :) thank you for the help. <tricon>podiki[m]: i'm attempting to get stumpwm to launch via lightdm. following your dotfiles. i keep receiving the lightdm.log error: "Seat seat0: Can't find session 'default'"; ~/.dmrc contains "Session=default"; what provides/resolves this? <civodul>attila_lendvai: re git-checkout, good point; i wonder if should leave it out, to keep things simple, and also because it's not in the manual (which we should fix) <acrow>I guess the only fix is to chroot into the instance and then try another reconfigure. <_graywolf>I am trying to produce docker images using guix pack, and it works nicely, but the produced images are somewhat large (~400MB for image with bash-static and git). Is it possible to decrease the size? Or should I just try to trim the final image once it is build? Any tips? <acrow>Should it be the case that system reconfigure is blocked from succeeding because a required package has and empty <>.info.gz file in it? Of course, why official packages would have such faults in the first place is a sad observation. <attila_lendvai>i `herd stop my-service` and then it starts up. i keep stopping, but it keeps restarting. is this broken expectations, or a bug in shepherd or my service code? do i need to `herd disable myservice` to prevent starting? <attila_lendvai>seems like a `guix system reconfigure` starts up services that i have explicitly disabled, so that they are not started <jackhill>Is there a way to get my nginx service to start after my wireguard service? Currently nginx doesn't start on boot becasue it can't bind to the wireguard IP. <attila_lendvai>jackhill, shepherd service objects have a field called requirement. it's a list of symbols, the name of other services that need to start before this one. <mirai>jackhill: there's a requirement field <mirai>not sure of the exact name of it, see the manual <jackhill>mirai attila_lendvai: thanks! I guess I need to take a look at the service proc. <civodul>jackhill: perhaps you should get nginx to bind to 0.0.0.0, if that's an option for you <lfam>Howdy! Is there a way to specify URLs and commits for channels on the command line? Or is editing channels.scm the only way? <lfam>More generally, what is the workflow for testing changes to 3rd party channels? <jpoiret>lfam: you can't specify channels on the clI other than guix iirc, so i usually copy channels.scm and feed it to `-C` <lfam>Okay, thanks. Just wanted to make sure I wasn't missing something. I guess I could have a file like 'channels-local.scm' and pass that to -C <lfam>It seems tricky on the command line! How to reference each channel and specify them separately? <jpoiret>esp. since we have channel names as symbols, this would fit right into a cli <lfam>Well, since we have the names... :) <rekado>lfam: when I work on guix-science, guix-past, guix-bimsb, etc I have local checkouts and add the locations with -L <batalie>sorry if this is covered somewhere, but i'm having trouble getting substitutes from bordeaux on a foreign distro. it is definitely in my /etc/guix/acl, but guix package operations still only pull from ci.guix.gnu.org, even after many daemon and system restarts, etc. it does work by specifying bordeaux through --substitute-urls though. is that normal? <batalie>jpoiret: yes, both root's and the local user's are a few days old <jpoiret>how do the substitute urls end up in /etc/guix/acl? are they put there by your guix service's configuration? <batalie>i think the "guix archive" suggested in the blog post announcing bordeaux puts them there <jpoiret>can you open a guix repl, type `,use (guix config)` then `%sysconfdir`? <lfam>Thanks rekado! I'll think about it <batalie>jpoiret: interestingly, root's is /etc <batalie>(this is on debian, with the debian package) <jpoiret>right, so i guess you don't have bordeaux in your /etc/guix/acl then (can you check that file?) <jpoiret>(it should be the one with the key starting with 7D <batalie>could it be looking for an acl in /usr/local/etc, for some reason? since that's the local user's %sysconfdir? <jpoiret>ah, no, you mentioned foreign distro, right? do you have the additional --substitute-urls= ... to your guix daemon invocation? <jpoiret>check out "(guix) Getting Substitutes from Other Servers" <batalie>jpoiret: i'm guessing that would be in the systemd service, in a foreign distro's case? no, there's no --substitute-urls flag being used there <jpoiret>yep! everything should be explained in the info manual node i mentioned above <batalie>thanks, i must have missed that changing the daemon invocation was necessary <rekado>civodul: we have a problem with the maintenance repo <rekado>or maybe it’s just me…? Maybe I need to update the keyring branch <rekado>I’m happy it was just me and we didn’t mess up the auth chain :) <vagrantc>pretty sure the defaults for guix-daemon are to use both ci.guix.gnu.org and bordeaux.guix.gnu.org with the debian package ... at least with the most recent version ... it was missing the key for a while... <vagrantc>you would have to manually add the key and update the guix-daemon stuff to use the substitute-urls :) <vagrantc>in ~10 days it will be oldstable ... ugh. i forgot to fix a bug ... <batalie>oh, i didn't realize it was coming out so soon <civodul>vagrantc: good that Debian and Guix are on a similar pace, release-wise :-) <vagrantc>ACTION was beginning to wonder if debian would get in two consecutive releases before guix <vagrantc>actually, it was never in danger ... debian/stable has guix 1.2 :) <civodul>actually, a release team +/- formed at the Guix Days... in February <janneke>jpoiret: thanks for the hint, looks we can better/more use libc-for-target in glibc-final-with-bootstrap-bash and/glibc-intermediate and glibc-final, esp. for the source field <janneke>i'm going to try building that first, and then i'll take another look into the new test failures we had for guile and coreutils-final <janneke>so the patch set for native build will have to wait/mature a bit <mekeor>hello. i'd like to attempt to contribute a "how to set up emacs and geiser". should it go to the cookbook or the manual? <flaminwalrus[m]>Hey y'all, I've just gotten a fresh reinstall of Guix running and as I'm trying to upgrade from the ISO guix I'm running into a cryptic error. I'm on a i686 Thinkpad, and was running into deadlocks (?) when trying to `guix pull` normally, so I followed the directions in the manual, section 7.5 "Channels with Substitutes," so as to avoid building the derivations causing me issues. However, I suspect that the derivation outputs I got were <TylerWolf[m]>I cannot install musl-cross without getting a long error about gimple-match. Does this happen for anyone else? <iyzsong[m]>flaminwalrus: maybe try: env -i $(which guix) ... <iyzsong[m]>perhaps some GUILE_ env variables may lead guix load incompatible files? <flaminwalrus[m]><iyzsong[m]> "flaminwalrus: maybe try: env -i..." <- This seems to be taking a while, which may be confirmation it's working...no output though. <flaminwalrus[m]>The only thing I've messed with in the envvars is $GUIX_PROFILE, sourcing it in order to get the new guix <iyzsong>flaminwalrus[m]: are you using the guix system or guix on a foreign distro? <flaminwalrus[m]>Guix system; this cropped up when upgrading the system immediately on first boot into disk after the installation image <flaminwalrus[m]>Interestingly, when I log out and in again as root, the commit in `guix describe` changes <iyzsong>um, 'which guix' should always return ~/.config/guix/current/bin/guix after 'guix pull' <flaminwalrus[m]>I'm pulling again, but my user account's guix which is having an identical problem returns that. <flaminwalrus[m]>Quick, basic question: do I need to worry about root's guix to do a `system reconfigure`, or will my user's updated guix suffice? If not, is that an artefact of sudo, rather than su? <flaminwalrus[m]>`/run/current-system/profile/bin/guix`...I suppose it's my path-setting at issue, then? <iyzsong>the current one should be prefered. when do a system reconfigure, make sure run the current (pulled) one by 'sudo -E' (will search guix in current PATH) or by absolute path <flaminwalrus[m]>Ok, so I can install and use GNU hello via `env -i $(which guix) install hello`, and `describe` in an environment indicates the guix used is updated. <flaminwalrus[m]>`guix install hello` yielding, of course, the same ABI error I've been up against <flaminwalrus[m]>It did tell me to set my GUIX_PROFILE to `/var/guix/profiles/per-user/user/guix-profile` in the usual environment variables hint, though. <ahriman>Hello, I am trying to change the wine-staging package definition from version 8.0 to 8.9. It seems that after 8.0 the `patchinstall.sh` script that was used to apply the wine staging patches was replaced by `patchinstall.sh` and moved from the `patching` to `staging` directory. I updated the package definition to reflect this, but when `'apply-wine-staging-patches` is running, execvp fails with "No such file or dorectory". Does anyone <ahriman>know why cant the `invoke` find the file while `search-input-file` inside it can? <ahriman>(correction, `patchinstall.sh` was replaced by *`patchinstall.py`*) <ahriman>does it being a python script make `invoke` not work? and if so, what should i be using instead of `invoke`? <ryan77627>Hey, anyone have any experience changing `.profile` by extending home-shell-profile or getting dbus to launch with certain env variables that are present in my bash_profile? They do not seem to get reflected, causing issues with xdg-desktop-portal <acrow>Yes, and it was unnecessary. The service entries like (service xfce-desktop-service-type) will do all that dbus wrapping stuff for you without you having to peer behind the curtain. :) <ryan77627>I do not believe that is a thing for sway unfortunately (i could be wrong, i am fairly new to guix still). I need the env variables since I have both xdg-desktop-portal-{gtk,wlr} installed, using gtk for file picker stuff since wlr doesn't implement that and wlr for the screencasting. Unfortunately, only gtk is getting auto launched via dbus activation and I presume it may be because of that env <ryan77627>variable, though i was still troubleshooting <ryan77627>XDG_{CURRENT,SESSION}_DESKTOP is set properly in my env otherwise <acrow>I, perhaps heretically, prefer X with networking capabilities over that newer stuff; so, no I have not done that; but if I was doing it I think info:"(guix)Desktop Services" (which talks about wayland) might provide some helpful information. Good luck. I know others have made that work here and if you are stuck I'm sure they can offer greater assistance. <ryan77627>acrow: I have in fact figured out a solution. I do not mess with (nor understand entirely) good ol' d-bus, and as such forgot the command `dbus-update-activation-environment` was a command. I just slapped that into the top of my sway config with the relevant env vars I need and it now launches both portals properly. Some days I do miss the "just works" of X... in certain regards ;) <flaminwalrus[m]>Ok, so it looks like my problem is solved (at least for now) by unsetting GUILE_LOAD_COMPILE_PATH. However, this makes any guix command take an eternity to run, presumably because guile has to load all the libraries from scratch. <flaminwalrus[m]>I'm not intimate with the tooling for Guile; how might I recompile the modules? <devmsv_web>hi get the following: guix system: error: /home/msv/config.scm:152:12: no value specified for service of type 'pam-krb5' <devmsv_web>well strangly enough it fixed specifying the configuration for the service even thoug i used the default values... <MatoHota-work>Hi, using guix import pypi I have built a configuration for pdf2image <MatoHota-work>I have an account on savannah but I have no experience with this git repository <evilsetg[m]>Hello guix, would anyone mind looking at this package patch: https://issues.guix.gnu.org/62293. I hope it is okay for me to try to bump this every once in a while until something happens. It is my first contribution to guix so I would be happy if it would not get lost. <evilsetg[m]>I just noticed that there are no irc logs for yesterday and today . Does anyone know if anything happened? I use a matrix bridge so I use the logs to sometimes check if it is still working. <next4th>evilsetg[m]: pushed it now, change the name to pcsc-cyberjack :) <trankitron>Thanks to your help, I'm making progress with the pharo-vm package specification. <trankitron>However, still I feel like I'm far from understanding guix enough, so I need more help. <next4th>AwesomeAdam54321: the linenoise patches seems a little heavy, dunno why the upstream didn't merge it.. <trankitron>Is this the proper channel to write these kind of noob questions or it's some other newsletter or channels for that? <next4th>trankitron: yes, it's the right place! <trankitron>I'm trying to replace some paths in a list of files using the substitute* utility, but when I try with a file in a subdir like "bin/pharo" it fails with system-error "mksteemp" "~A" ("No such file or directory"). I don't know what it's wrong with that. <next4th>that mean the given path is wrong (not existed), can you paste your definition to a paste service (eg: https://pastes.sh) <paulvictor47>hello channel. I have been using nix for quite sometime now and am trying out guix. nix has this concept of overlays where you can override a package definition which changes the derivation in the rest of the places referred in the system. Is there something similar in guix ? The reason I am asking is that the version of nix in the guix channel is <paulvictor47>too old(more than 2 years) and i need to use a newer version. <jpoiret>paulvictor47: you can use package transformations <jpoiret>there's no general "rewrite this package definition please" knob though, because guix is functional: if nix is used in a system service for example, you have to find a way to pass the modified nix package to the service <jpoiret>you can also try updating the nix we have :) <trankitron>it works correctly when I only substitute the file "pharo", but it fails when I include "bin/pharo" <paulvictor47>thanks jpoiret . yeah about that, thats what I am looking for, as to how to pass this new nix package to the herd service. today's my first day on guix, while I would be happy to contribute back, i need to get it(nix) working and slowly move things to guix. <PotentialUser-22>is there any reason why guix passes "modprobe.blacklist=usbmouse,usbkbd" to the kernel by default? <pjals>Why does libx265 use build steps instead of outputs to specify 10-bit? It seems unlogical to me that an unstandardized libx265_main10.so would be used to specify 10-bit libx265 when we have isolated software and libraries! <pjals>This is especially bad when I want to use ffmpeg, which has support for x265 10-bit, but I can't use it because it's so complicated to specify that I want main10! I would need to completely refactor the ffmpeg package definition! <acrow>flaminwalrus: Recompiling the modules might be as simple as starting guile and accessing them. <acrow>eg... say the problem is gnutls then in guile run ,m (gnutls). I believe that autocompile will then take care of it. Then you can just exit guile and you are done. <Guest32>Hi, I notice the IRC archive logs for guix an Hurd are not generating <rekado>they are, but no new file has been opened for them