IRC channel logs

2024-09-10.log

back to list of logs

<fnat>futurile: Lovely! Looking fwd to catching up in person, yay!
<wolfdog>does the recently found Nix vulnerability wrt NAR unpacking affect Guix in any way? https://puckipedia.com/7hkj-98sq/qixt
<ieure>wolfdog, Unclear, cwebber was asking about that earlier.
<ieure>Probably a good idea to post it to guix-devel.
<wolfdog>ah, alright
<aswjrisp>Rutherther: I tried your code snippet for the grafts. It worked for sway and qutebrowser. I rebooted and tried some other graphical programs that needed mesa. I tried them without applying grafts and they also worked. This was not the behavior I was expecting. I though I would need to apply the graft to each program that needed mesa based an guix graph.
<podiki>anyone familiar with qt know why tests might be failing on mesa-updates? https://ci.guix.gnu.org/build/5726766/details
<podiki>maybe i should try the kde update patches on top in case it is because of newer wayland, etc.
<rynn>Anyone know if I'll need to do a reconfigure to get herd to operate my swap properly? After some partition juggling, herd can't start the swap service anymore, but after I reformatted the swap partition I was able to swapon just fine.
<rynn>Doing a herd start just throws a failed to start error.
<aswjrisp>podiki: How frequently does mesa-updates make it's way into master?
<aswjrisp>rynn: Have you tried rebooting?
<podiki>aswjrisp: no set schedule. hopefully this can merge within a few weeks (needs substitutes built for other archs andfix some build failures)
<rynn>Not yet, was getting ready to
<rynn>Hurm. Reboot didn't fix it. However, now that I'm looking at it, it does look like the UUID of my swap partition changed. Is there a way to correct that doesn't require a reconfigure?
<aswjrisp>rynn: Maybe check the Swap Space section of the manual or the manual install section of the manual as it also covers swap space.
<rkazak>What do people recommend in terms of hardware to run Guix?
<rynn>aswjrisp: I didn't find antyhing in the manual for my particular situation, but updated the UUID in my config and then doing a reconfigure+reboot did fix it.
<meaty>is there a way to upload compiled binaries to the build server? i don't want compiling the latest firefox or kernel version to be for nothing
<meaty>ik there is 'guix publish' or smth but isn't that for LAN
<meaty>or is there a directory of substiture servers somewhere that people can put their details on
<elpogo>meaty: from https://guix.gnu.org/manual/en/html_node/On-Trusting-Binaries.html "In the future, we want Guix to have support to publish and retrieve binaries to/from other users, in a peer-to-peer fashion. If you would like to discuss this project, join us on guix-devel@gnu.org. "
<meaty>word
<rynn>Huh. There was some talk on the mailing list recently about who's responsible for reviewing and signing updates too.
<Guest26>Hello everyone. I am looking for an URL to download properly guix iso image. The URL on the guix website is too low. Do you know a good URL? Thank you in advanced.
<rynn>The ftp mirror should automatically redirect you to a working mirror. Are they all too slow, or is it just one?
<rynn>For example, this is the one it just gave me: https://mirrors.ocf.berkeley.edu/gnu/guix/
<Guest26>rynn: It is too slow.
<Guest26>I was trying one hour some times and the mirror speed is slow.
<podiki>Rutherther: i think qtwayland can't build with newer wayland which blocks ~270 packages. maybe it is fixed with a qt update but i haven't tried
<Rutherther>podiki huh, I didnt expect that. I will take a look today, but gotta work first
<podiki>i tried just the older wayland for qtwayland, but that doesn't build (i think due to differing version mesa is built with)
<podiki>i locally have reverted it and just added wayland-next for use with wlroots@0.18; but not sure it that will be usable for other packages if wayland is different versions in mesa etc.
<Rutherther>You mean that mesa depends on wayland, and now its lower version?
<cow_2001>when using a manifest file with (packages->manifest some-package-list), how do you select a specific output for a package that has multiple outputs?
<Rutherther>cow_2001 try `(,package "output)
<Rutherther>Just don't forget to close the " like me
<cow_2001>ooh
<cow_2001>;p
<cow_2001>thank you
<podiki>Rutherther: yeah, with qtwayland using older wayland but mesa using newer wayland, there is a linking error if i remember (some symbol not found or something like that). i didn't try very hard though. so i imagine that could happen through wlroots, though wlroots builds fine with newer wayland and mesa older wayland
<podiki>Rutherther: got to go now, but if you happen to see some other way, let me know
<podiki>ACTION away
<saba25>hallo
<saba25>does anyone here have any idea how I can install gnu guix on a laptop because unfortunately I have problems when I do it on an intel based laptop because I always get a message that I cannot use the wifi driver because what the message says is proprietary
<saba25>does anyone here have any idea how I can install gnu guix on a laptop because unfortunately I have problems when I do it on an intel based laptop because I always get a message that I cannot use the wifi driver because what the message says is proprietary
<saba25>gibt es hier aktive leute . is anyone here active
<efraim>o/
<efraim>it's still a bit early apparently in Europe
<efraim>I think the standard suggestion for the wifi is to use a dongle that is supported by free software
<meaty>is there a way to modify a package's profile outside of installing it manually? I just want to test some plugins to a tool that haven't been packaged yet
<meaty>specifically I want to add some python packages to its environment
<Rutherther>Not sure if I understand correctly. Guix shell also uses profiles, so just put both the package that provides search path and packages that provide path of the search path into a shell, and you will get environment with plugins
<apteryx>any good team/branch to bundled an update to libvpx? it has 1009 dependents
<apteryx>or maybe a single branch to build it could do, but it sounds like a lot of complications to go through for just that (setting up a one time branch in CI, for example)
<z572>apteryx: i have a patch update libvpx to 1.14.1 on qt-team, see https://git.savannah.gnu.org/cgit/guix.git/commit/?h=qt-team&id=1375078336f4b6769ff4a2556dddec364a3e539f
<apteryx>ah! that's why I got a substitute for it
<apteryx>sorry, I haven't yet looked into your Qt updates, but soon will
<apteryx>I'm trying to update jami (and hopefully fix its current build issue)
<mrvdb->good morning guix. I'm getting a hash mismatch from a package (see https://dpaste.org/1yZcb ) Bit surprised, this can't be a local issue, can it?
<civodul>👋
<Lumine>👋🏻
<civodul>mrvdb-: it’s available as a substitute: https://paste.debian.net/1328950/
<civodul>that said, could you report it to bug-guix@gnu.org?
<civodul>ideally with a diff compared to the one that we have
<civodul>(most likely the tarball was modified in place upstream)
<futurile>Morning
<efraim>o/
<efraim>futurile: I have mesa on mesa-updates building for aarch64. do you know if it acrtually needs lua or libarchive?
<futurile>efraim: I'm not working on that, were you think of podiki and possibly there was someone else working on this in the backscroll
<efraim>oh, right, probably
<nikolar>my rock 5a arrived :)
<efraim>nice!
<efraim>which architecture is it?
<nikolar>aarch64
<nikolar>maybe i could donate some of its time to serve as a substitute build server for guix
<nikolar>:)
<efraim>is there an environment variable for the top of the build tree?
<efraim>oh, i'm now still curious, but I no longer need it
<mrvdb->civodul will file a bug indeed. I think because I am on ppc64le the substitute does not apply for this machine?
<mrvdb->s/will/I will/
<civodul>since it’s a fixed-output derivation, it should apply as well
<civodul>just tried `guix build libjpeg -S -s powerpc64le-linux` and it works for me (gets substitutes)
<mrvdb->I'm getting a warning when I do that: "guix build: warning: could not determine current substitute URLs; using defaults"
<civodul>it means the guix-daemon being used is quite old
<civodul>i’m surprised you get that warning in this case though
<PuercoPop>Is package-develop-inputs the right place to put inputs that are needed for the check phase (running tests)?
<aswjrisp>I am using sway and have mouse functionality but there are no cursor icons so I can not see it. I have looked up how to change the cursor theme, but you need the path for the cursor theme. Where does guix put cursor themes?
<aswjrisp>I have looked in /usr/ .local/share and ~/.icons
<Rutherther>aswjrisp did you switch to mesa-updates branch? As for where icons are stored, they can be either on ~/.icons as you said, or in any of XDG_DATA_DIRS folders, so installing themes to your profile or to your system packages is enough to add an icon theme
<aswjrisp>Rutherther: I will look into how I can pull from a branch that is not master.
<aswjrisp>I did find the Adwaita theme in the /gnu/store using the find command. Then came up with what I think will be a more stable path with /run/current-system/profile/share/icons/Adwaita
<aswjrisp>Rutherther: It sounded like mesa-updates is a development branch that is still being worked on.
<Rutherther>Yes, it is a development branch, has a lot of updates like mesa and its deps
<aswjrisp>Even though it is a development branch it is okay to use, not in a broken state.
<Rutherther>Yes, you seem to have icon theme installed with system packages. That share folder should be in Xdg data dirs, so it can contain icon themes, among other stuff
<Rutherther>aswjrisp there can appear issues, but in worst case you just revert back to master
<aswjrisp>Yes, that is a nice benefit of guix. I will try it once I figure out how to pull a non master branch.
<PuercoPop>I'm trying to check a change in the guix tree, it doesn't seem that guix build is using the updated package definition. Does using a guile expression (-e) to refer to the package an alternative to ensure that the packages come from the checked out tree? For some reason -L is running way more code than I expect and failing in an unrelated area
<aswjrisp>I am going to try to explicitly set the cursor theme in the sway config now that I have a path for Adwaita.
<futurile>PuercoPop: you're using the ./pre-inst-env guix right?
<PuercoPop>futurile: it is one of the options I've tried, yes
<futurile>PuercoPop: here's the section of how to run guix when you have the tree - https://guix.gnu.org/manual/devel/en/guix.html#Running-Guix-Before-It-Is-Installed
<bilke>Hi, is it somehow possible to combine time-machine and shell to run a command, like this
<bilke>guix time-machine -C channels.scm -- shell -C -- python --version
<futurile>PuercoPop: so lets say you want to build your package: guix shell --development guix (to get the guix dev inputs), and then ./pre-inst-env guix build <package-name-I-changed>
<bilke>This does not work, I guess because the -- is interpreted only once. I tried all combinations of quoting the arguments without success...
<PuercoPop>I'm trying to fix https://issues.guix.gnu.org/73155. My understanding is that the fix is to regen the certificates before runnning the tests. So I'm trying this as the fix https://paste.sr.ht/~puercopop/499b2118c0912e883a742bc08651cb0a83161a6c
<PuercoPop>I even tried putting (raise-exception "fubar") in the function I'm inserting and the failure seems to be the same, when running the check phase. It seems the before-check phase modification is not run
<PuercoPop>futurile: I'm using guix shell -D guix -CPW to run things, it is inside that environment that I call ./pre-inst-env guix build ruby-net-smtp
<futurile>PuercoPop: OK, and what happens
<PuercoPop>futurile: I see the error reporting on the issue. Due to expired certs. I would expect to see an error related to the raised exception if its running the on disk version
<futurile>PuercoPop: uh does ruby have a special development-inputs field, AFAIK that's only in the cargo-build-system?
<futurile>PuercoPop: are you doing ./pre-inst-env guix build --no-substitutes <package-name> - if you don't it will download the version from the Guix subsittute server and use it
<PuercoPop>futurile: Not from what I found by 'C-s'-ing their build system, but I did do (development-inputs ruby-net-smtp) in a guix repl (in said container) and I did see a lot of packages returned, when I wouldn't have expected a package to be there
<PuercoPop>no, I'm not using --no-substitutes, let me try
<futurile>PuercoPop: if you have ALREADY downloaded the substitute binary version, then you need to do gc --delete /path/to/thingy - which guix build will show you
<PuercoPop>futurile: Thanks, I was wondering how to do a 'clean' build. guix gc --delete is the way?
<futurile>PuercoPop: yep - that's the easiest way - just rm it from the Store basically.
<PuercoPop>futurile: the --path is to the .drv file or the dir it was unpacked to?
<futurile>PuercoPop: to the dir it unpacked to
<futurile>bilke: should be possible
<futurile>bilke: I have an example in a blog post I did where I set the timemachine like this: guix time-machine --channels=channels.scm -- shell --container weechat@1.8 -- weechat
<futurile>bilke: https://www.futurile.net/2023/11/07/guix-time-travel-tricks/
<sepeth>Hi Guix, I am trying to create a Rust package in Guix, but I see the deps under target_os=windows in Cargo.toml are also requested. Is this expected?
<futurile>sepeth: I think the Rust build-system just sees dependencies, as we have MacOs and Android packages (I think). Some of the packages alter the Cargo.toml file, you could try that.
<sepeth>futurile: Oh ok, thanks ^-^ I will see if I can find an example.
<PuercoPop>futurile: any tips on how to list the store paths for a derivation? because the messages seem to point to a top level ruby in the store, not a gem specific one . ie. /gnu/store/xhx0nzp7m1vg856knsy0jygbm8ffabgr-ruby-3.1.4/
<PuercoPop>And the last error mentions the builder derivation, ie. builder for `/gnu/store/zxxyx4hgn02gq97ylvla6ndh8y9nqf98-ruby-net-smtp-0.3.3.drv' failed with exit code 1
<PuercoPop>which is checkout out on /tmp/
<PuercoPop>ie. /tmp/guix-build-ruby-net-smtp-0.3.3.drv-0/source
<z572>podiki: I push a commit to mesa-updates to fix qtwayland-6, see https://git.savannah.gnu.org/cgit/guix.git/commit/?h=mesa-updates&id=b9419722e0801334607530fd1afb487e1d74f9c9
<futurile>PuercoPop: so if you do ./pre-inst-env guix build --no-substitutes ruby-net-pop - it should give you a path in the Store no?
<futurile>PuercoPop: or is it failing to build it, so it errors when you try and build?
<bilke>futurile: thanks! it works if I specify a manifest file or a package, so both of the following works:
<bilke> guix time-machine -C channels.scm -- shell -C -m manifest.scm -- python --version
<bilke> guix time-machine -C channels.scm -- shell -C python-wrapper -- python --version
<futurile>bilke: ah cool. I found the errors a bit hard to interpret when I was playing with it - glad it works!
<bilke>Yea, initially i used ` ... -- shell -C` only because I relied in the implicit manifest loading of the shell command
<bilke>futurile: btw. thanks a lot for your awesome blog series! Looking forward for the next post ;-)
<futurile>bilke: ah cool, glad they're useful. Currently trying to get my heard around modifying-phases in package definitions
<PuercoPop>futurile: the build phase pases. https://paste.sr.ht/~puercopop/5b266ba7cf59e9c63e8264bb149796de94f354d0 I did find the path while selecting what to put on the paste bin, it does go after the build phase again
<PuercoPop>Is it possible that there is no before check phase?
<futurile>PuercoPop: no, it's actually running the check phase in your output - grep - starting phase `check'
<futurile>PuercoPop: lets go back one step. You have a checkout of the guix source right. And you added the patch you showed to the package ruby-net-smtp which is in <top-level>/gnu/packages/ruby.scm . And then you're in the top directory of your checkout doing guix shell --development guix ; and then guix build --no-substitutes ruby-net-smtp@0.3.3 .
<PuercoPop>yes, although I'm also using the -CPW flags in the guix shell
<podiki>efraim: thanks for checking in on mesa-updates, let me know if there's anything you need for other archs. as you saw, had to clean up some options that were changed before and hopefully consolidated the different inputs now that mesa needs them on other archs
<futurile>PuercoPop: OK, I can get it to fail. So change your `development-inputs` line to (inputs (list openssl))
<podiki>z572: thanks for the qtwayland update! did not expect that but love when i wake up to problems being solved for me :)
<futurile>PuercoPop: then when I try to build it - it will try and run the phase you created - starting phase `regenerate-certificate' - where it fails on taht make command
<PuercoPop>its beciae it should be (invoke "make" "rengen_certs"), each arg on its own string
<PuercoPop>*because
<PuercoPop>But I can't get to that part failure
<PuercoPop>hmm, maybe its because I wrote something in a way its a load error and guix uses the cached version in those scenarios? I was writing (development-inputs openssl) instead of (list openssl) after all
<futurile>PuercoPop: no - there is no such field as development-inputs - it only exists in the Cargo build system. Every other build system it's `inputs`, 'native-inputs' and 'propagated-inputs' - https://guix.gnu.org/manual/devel/en/guix.html#Defining-Packages
<PuercoPop>So where do dependencies only needed to run tests go? inputs?
<futurile>PuercoPop: inputs or native-inputs
<yelninei>Finally was able to build gcc-final and beyond on my little childhurd. Had to restart the bootstrap process several times because of random build errors in glibc-intermediate and some early gcc just segfaulting immediately
<yelninei>Is there a way i can exclude that gcc from garbage-collection? Because I don't really want to go through that process again.
<Rutherther>yelninei make a gc root for it somewhere, ie the -r flag to stuff like build or shell. Or Alternatively install it to one of your profiles
<civodul>yelninei: hi! you can register a “GC root” for the thing you don’t want to GC, with “guix build whatever -r gc-root”
<yelninei>gc-root looks exactly what I want. Is it enough to gc-root gcc-final only or should i add others as well just to be sure?
<Rutherther>Dependencies will be kept as well, if thats what you mean
<Rutherther>If there are other things you want to keep that are related to it, but not dependencies, they wont. Also the source and stuff used to build it can be gc'd unless you use extra flags to your daemon that prevent that
<Rutherther>So only runtime dependencies are kept by default, not all dependencies, sorry for the confusion
<aswjrisp> /run/current-system/profile/share/icons/Adwaita
<aswjrisp> /run/current-system/profile/share/icons/Adwaita/cursors
<aswjrisp>Not sure what else to try.
<Rutherther>Programs usually dont accept full paths afaik. Try the mesa updates first, since I think its possible its related to a package missing mesa, so it doesnt work
<aswjrisp>Rutherther: Okay I will try a guix pull from the mesa-updates branch.
<Rutherther>You can try setting env var WLR_NO_HARDWARE_CURSORS=1 Prior to starting sway, but if it fixes it my bets would be on the cause by mesa even higher. Or something else broken in the drivers
<yelninei>I've added another one for (@@(gnu packages commencement) guile-final) as that also took forever
<Rutherther>Are you bootstrapping something without substitutes?
<yelninei>Rutherther: hurd packages failed to build on ci :(, so i am doing it myself
<Rutherther>I see. Is hurd usable on real hw or only v so far?
<Rutherther>s/v/vm
<yelninei>just using the childhurd-service and trying to have some fun playing around with it
<civodul>yelninei: you can add GC roots for glibc, etc.
<civodul>(you can have the roots on the host rather than in the childhurd, to be safe)
<civodul>yelninei: BTW, did you file a bug?
<yelninei>no apologies, I don't have an email with not my real name. Should probably create one
<civodul>oh, np; maybe you can chance the ‘From’ field in your mail client when sending
<yelninei>The system test passes when i change to C.utf8 locale in the shepherd service. Not sure why that was not working when i ran that manually in the ssh session
<civodul>hmm, do you know why?
<civodul>i experimented briefly yesterday with std::stoi under different locales and everything seemed fine
<yelninei>no idea. Tried to setlocale to different locales and that was all working for available locales
<galois``>How do I make my controller work on guix? Do I need driver software for it? I want to make an experiment with raylib.
<ieure>galois``, You'll need to be more specific. What kind of controller? What make/model?
<dthompson>most controllers ought to "just work" over usb, but sometimes bluetooth requires extra work
<galois``>ieure: It's a wireless pro game controller for the switch. It has usb connection though and the controller blinks when plugged in
<dthompson>I have gotten my switch pro controller working in guix with usb and bluetooth
<ieure>galois``, There's a driver for that in the Linux kernel, my friend Emily wrote it.
<dthompson>I didn't have to do anything in particular for usb to work
<galois``>I wanted to test it on some gamepad test website, but it coudln't be detected.
<galois``>dthompson: Thanks for the insight
<ieure>I agree that it should Just Work, but I haven't tried it mysefl.
<ieure>*myself
<galois``>ieure: Hah, cool.
<ieure>She wrote drivers for a bunch of the Switch wireless controllers.
<galois``>dthompson: Does a random website also "just work" for controller testing?
<dthompson>I have never used a controller with my web browser before so I have no idea
<galois``>Ok. Where did you try it out for testing?
<dthompson>idk I can't remember which game I was playing at the time
<ieure>galois``, I'd start by using evtest to make sure your controller input events are making it to the kernel.
<ieure>You might also need a udev rule to give your user permission on the evdev device node.
<galois``>Thanks, I'll try that
<galois``>Yeah that worked great.
<dthompson>woooo
<galois``>Also this raylib example worked.. I should have just tried that first before asking. I was confused by why it doesn't work on the browser. Not on icecat nor on ungoogled-chromium
<aswjrisp>I am installing guix on a 64 bit computer. I selected the manual install option. I can open the manual on tty2.
<aswjrisp>The manual included with the installer has the version number error 0.0-git. This breaks links in the install section of the manual.
<aswjrisp>If the version number issue is fixed it would be good to update the installer image so it is also fixed.
<podiki>sneek later ask z572 do you know the fix for qtwayland@5 as well?
<sneek>Got it.
<efraim>podiki: mesa passed the `configure phase on riscv64 so I hope it'll build without any issues. I won't have time to test 32-bit ppc any time soon though, so you can continue to ignore it like normal :)
<podiki>thanks, and haha will do! at least i know to bug you about non-x86, that's a step in the right direction right? :)
<podiki>still seeing some qt/kde related failures, some fixed on master (kmime) others i'm checking locally. but at least on x86_64 looks pretty good, i686 hopefully getting there
<efraim>yeah. I'll pull out my pinebook pro probably tomorrow and see if I had anything else in my git stash for aarch64 for mesa
<efraim>I heard something from debian about mesa 24.2 and gtk4 tests so I'm building out to there locally
<efraim>ACTION has to run off again
<podiki>i updated gtk4 for that reason if i remember, at least on x86_64
<podiki>f2c8b4c07e7e73b8d0bfede7fc886b8b372bc781 (on master, was pre-mesa update failure)
<peanuts>"guix.git - GNU Guix and GNU Guix System" https://git.savannah.gnu.org/cgit/guix.git/commit/?id=f2c8b4c07e7e73b8d0bfede7fc886b8b372bc781
<Franciman>hi, what's happening to icedove?
<Franciman>there haven't been substitutes available for 2 days so far
<olndrxyz>Hi, I'm getting the "recovering journal" message when booting/restarting after a guix system reconfigure. Is this normal? System boots just fine, fsck returns no errors, the smartctl test on my device passes.
<nikolar>fsck returns no errors because the journal did its job
<ieure>olndrxyz, It means a filesystem wasn't unmounted before the machine rebooted. Did you run `sudo reboot' to restart it?
<olndrxyz>ieure: I run loginctl poweroff/reboot but it's the same with sudo reboot/poweroff.
<olndrxyz>nikolar: I run the fsck from a guix live usb
<olndrxyz>ieure: sudo reboot/shutdown sorry
<ieure>olndrxyz, Hm, that definitely shouldn't be happening.
<olndrxyz>ieure: it does not happen if I don't reconfigure or guix pull before restarting.
<meaty>are the bluetooth service defaults supposed to 'Just Work'?
<ieure>meaty, They work fine for me, although for some reason I don't understand, Bluetooth is always disabled when I wake my laptop from sleep and I have to manually reenable it.
<ieure>I don't know why that is.
<meaty>ieure: oh? how do you re-enable/check that it's enabled?
<ieure>meaty, That box runs GNOME, there's a widget in the upper-right menu.
<ieure>Works good other than that.
<PuercoPop>I have seen chdir used to change the current directory during a phase, is there something like with-chdir utility?
<PuercoPop>with-directory-excursion seems to be what I'm looking for