IRC channel logs

2024-11-29.log

back to list of logs

<apteryx>why do I get this kind of error on the python-team branch? ModuleNotFoundError: No module named 'setuptools' Setuptools should be available since it's bundled with Python itself.
<apteryx>adding it as a native-inputs resolves the build error, but then 'guix lint' complains
<apteryx>bng python-flake8-bugbear
<msglm>can you post direct output?
<apteryx>msglm: https://paste.debian.net/1337380/
<apteryx>some projects like libcst expects to fetch rust deps from online; nixpks has this rustPlatform.fetchCargoTarball procedure it uses to prepare a vendored deps tarball it extracts in the source to workaround it
<apteryx>looks like we'll need to come up with something similar
<apteryx>efraim: do you know if someone has tried implementing an equivalent to nix's fetchCargoTarball in Guix yet?
<apteryx>it takes the project name / version, and then runs 'cargo vendor' with it, I believe
<apteryx>or would listing the cargo inputs as #:development-inputs achieve the same?
<apteryx>nah, it's a pyproject
<apteryx>for the record, this is nix's sledgehammer of a solution: https://raw.githubusercontent.com/NixOS/nixpkgs/refs/heads/master/pkgs/build-support/rust/fetch-cargo-tarball/default.nix
<apteryx>it's a bit opaque, it seems it'd not very clear/easy to ensure the result contains just free software
<apteryx>unless crates.io follows the GNU FSDG, which I'm sure it does not
<apteryx>msglm: must be some upgrade I made one my branch, every pyproject package seems affected
<msglm>tried compiling on my end on that branch with the latest commit only to get complaints about recursive dependencies
<apteryx>err, nope, it happens on origin/python-team
<apteryx>try e.g.: ./pre-inst-env guix build python-markdown2
<msglm>tried that and got:
<msglm>Throw to key `match-error' with args `("match" "no matching pattern" (unbound-variable "resolve-interface" "no binding `~A' in module ~A" (shared-mime-info
<msglm>(gnu packages freedesktop)) #f))'.
<apteryx>seems there's some error in your source; did you 'get reset --hard origin/python-team' ?
<msglm>yup
<msglm>i'll just do it again for good measure
<apteryx>weird. the next try would be to rebuild the source, 'make clean-go && make -j($nproc)'
<apteryx>but I just found out that this appears to be expected -- see (default-python) in (guix build-system pyproject), it defaults to a pip (and setuptools) less variant of python
<apteryx>The main reasons appears to be that 'Using python-toolchain here might cause dependency cycles.'
<apteryx>which means most pyproject packages could be broken... hm.
<apteryx>it's a 1.5 year old change that is only found in the python-team branch
<apteryx>(commit f3299c4a9f6)
<apteryx>ACTION goes on to add setuptools when they face an error about it
<PotentialUser-61>Hello, is there a way to get Guix on a proot on a non-FHS distro? I've seen some guides but they all share /etc and other things. I've gotten close to it but it says permission denied when it tries to link something in the store.
<adanska>does anyone get a crash when they run `guix refresh --list-updaters`?
<adanska>i get this backtrace: https://paste.debian.net/1337393/
<civodul>Hello Guix!
<futurile>Morning all
<jakef>hi Steve, hope you're well
<futurile>all good here jakef - just trying to send some final emails to Guix contributors to take the survey!
<futurile>Just found that Sharlatan has sent over 1400+ patches!
<futurile>good grief - I'm never getting there heh heh heh
<civodul>futurile: last day, right? :-)
<civodul>ACTION sent v2 of https://issues.guix.gnu.org/74542
<gabber>is there some writeup on *how* to upgrade postgres (for cuirass)? i guess it's a dumpall, upgrade service, restore all of sorts, but i am somewhat insecure of how to achieve it all in an as guix-y as possible manner
<gabber>e.g. after dumping, do i have to set up a postgres service to restore all data before setting the cuirass service to use the new postgres version?
<civodul>gabber: hi! i followed the postgres instructions whose link was in the news entry
<civodul> https://www.postgresql.org/docs/current/upgrading.html
<civodul>this has yet to be done for ci.guix though, and it’s a bit more frightening
<futurile>civodul: yeah, I'm shutting it down on Monday morning (traveling Sunday). I'm down to emailing the first 200 people who started contributing to Guix.
<futurile>one nice side-effect is I'll have a count of "active" contributors to the project (active meaning have sent patches in the last 2 years which was my arbitrary decision); and how many contributors we added in 2023 and 2024 (meaning how many new people started sending patches)
<futurile>just out of curiousity I find both quite interesting
<futurile>(also inspecting 1200 contributors is quite painful so it was a way to justify it to myself heh)
<rekado>random mumi idea: extract number of patch-set versions and display links to the first email belonging to each revision. Bonus: collapse all messages prior to the latest patch revision.
<futurile>collapsing would be great, I've sometimes missed that there's a revision lower down that I should read
<elevenkb>Hello there,
<elevenkb>I want to make sure that the hid_apple module has the option fnmode set to 2 by default, and I can't do it by putting a file in /etc/modprobe.d/hid_apple.conf with the contents "options hid_apple fnmode=2" there as I should.
<Deltafire>heh, i did this exact thing 2 days ago
<elevenkb>lol.
<Deltafire>used 'simple-service' to extend 'ext-service' and add the file into etc
<Deltafire>one sec, i'll get my laptop and copy/paste
<Deltafire> https://pastebin.com/LV6YUF3d
<Deltafire>add that to your list of services
<elevenkb>I did, but for some reason my kernel doesn't seem to load the setting.
<Deltafire>hmm.. worked for me, using a bluetooth keyboard
<elevenkb>wild. thanks for helping though.
<Deltafire>if you're using something hardwired, maybe it needs to be loaded earlier on in the initrd
<elevenkb>ahh!
<Deltafire>have a look at Kernel Module Loader Service on https://guix.gnu.org/manual/en/html_node/Linux-Services.html
<elevenkb>thanks I'll try that.
<taeaad>I've installed Calibre through Guix on a foreign distro, and have tried to set CALIBRE_USE_SYSTEM_THEME=1 on my environment, but Calibre still seems to use its own default theme and font. Any ideas?
<gabber>civodul: hi! thanks - i saw that and am well aware of upgrading DBs in general. what i'm not sure is the part in that manual that refers to upgrading the service (as it is done in old-school distros). i guess i'll have to figure it out. if there is interest i can write up my findings to a hands-on guide..?
<elevenkb>re the hid_apple issue, I can't even set it via the kernel command line apparantely.
<elevenkb>so, I'm probably going to ask in the #linux channel maybe.
<jaadu>How would I write a package that creates a link to icecat named firefox?
<Rutherther>jaadu: you can just use trivial-build-system and use "symlink" function you will give file-append to with icecat package
<civodul>gabber: oh yes, that’d be much appreciated
<civodul>looks like qa.guix never built https://issues.guix.gnu.org/70393 :-/
<civodul>i think i’m going to apply it despite the 414 dependents because it’s blocking other things
<orahcio>Hi, someone have problem to run qutebrowser on wayland? After `guix install qutebrowser` I try to run and I get the error ""10:07:32 WARNING: Could not find the Qt platform plugin "wayland" in ""
<orahcio>10:07:32 CRITICAL: This application failed to start because no Qt platform plugin could be initialized. Reinstalling the application may fix this problem.
<z572>orahcio: i think you need qtwayland too
<orahcio>z572: just do `guix install qtwayland`?
<z572>yes, qtwayland set QT_PLUGIN_PATH env, That should solve the problem.
<orahcio>z572: thank you very much, I could see on foruns to install on arch, but not yet on guix, thanks again
<orahcio>z572: problem solved! :)
<ekaitz>civodul: you just reviewed the doc patch... i don't know why `make info` is failing with it... and if I open the erroring file (the .de translation) has no empty lines where the original texi does
<ekaitz>it works only in the english version
<ekaitz>hmmm
<ekaitz>maybe an old artifact... rebuilding
<divya>Has anyone installed lean4 on guix in some other way? Currently Guix only has the old lean 3.
<divya>There's this patch that can does the upgrade, but it hasn't been touched since August. https://lists.gnu.org/archive/html/guix-patches/2024-08/msg00956.html
<lynn_sh>hey geeks. trying to setup so i can send-email with git, and after some painful steps i finally have it almost working. the problem is im getting an SSL error: STARTTLS failed! SSL connect attempt failed error:0A000086:SSL routines::certificate verify failed at
<lynn_sh>/gnu/store/g1q9wmxvfr0xmx869ynlk4rcrq9dp990-git-2.46.0-send-email/libexec/git-core/.git-send-email-real line 1638. any ideas?
<lynn_sh>this is just me sending the email after setting up my bridge (protonmail)
<Rutherther>lynn_sh: so you are using the bridge from proton, right? Did you configure it for starttls then?
<lynn_sh>Rutherther [sendemail]
<lynn_sh>    smtpServer = 127.0.0.1
<lynn_sh>    smtpServerPort = 1025
<lynn_sh>    smtpEncryption = tls
<lynn_sh>    smtpUser = lynn@lynn.sh
<Rutherther>lynn_sh: that's not what I mean. I am talking about the bridge, not git settings
<lynn_sh>the bridge is default starttls
<Rutherther>lynn_sh: so I suppose you will have to export the certificate from Bridge and configure git to trust it - sendemail.smtpSSLCertPath
<z572>for me 'smtpEncryption' is ssl ,not sure what the difference is
<lynn_sh>Rutherther yup that did it. i am not sure why it does not just provide it by default when asked.. sorry, i have never had to set up any of this stuff before.
<lynn_sh>z572 ssl is for TLS i think StartTLS is tls. that is what i researched at least.
<Rutherther>lynn_sh: what doesn't provide what by default?
<lynn_sh>Rutherther the git docs say the default, unset path should be the best option in most cases
<lynn_sh>for the SSL path
<lynn_sh>this solution is very unfortunate because it either means i need to put the cert in my declarative setup or always export it.. since it must be referenced directly in the gitconfig. i can't say im a huge fan of the git patch workflow :)
<Rutherther>lynn_sh: but this is a different case. How the internet works is that you have trusted CA authorities that give out certificates. But since you are using your own local service instance, you cannot have a certificate from trusted CA authority as Bridge assumes you will be using domain out of your control and hence you cannot get a trusted CA authority to give you signed certificate. That would mean for the CA authority to become untrusted, if it...
<Rutherther>... started giving out certificates like that
<lynn_sh>yeah i get all of that it just is a big pain, you know? im not saying it is guix's fault or anything. but needing to setup git send-email with modern tools just feels like im fitting a lot of things into old systems and it causes friction.
<lynn_sh>i appreciate the help though i can now send patches without them being wrong anymore.
<unmush>I don't suppose any of you happen to know of a mail provider one can sign up with anonymously? I'm not keen on linking this alias, but I do want the patch series I linked earlier to get seen...
<Rutherther>unmush: I am not sure what you consider anymous registration specifically, maybe protonmail?
<unmush>registration over tor would be ideal, which protonmail blocks, and last I heard they only offered webmail? Not sure how to use 'git send-email' with that
<unmush>I could maybe see if I can build off of some non-tor proxy
<Rutherther>unmush: I don't think protonmail blocks it, unless something changed in last few months. I've already registered there via tor
<unmush>well, they don't block it outright, but they demand a phone number in that case (I'll try again in case this has changed since I last tried)
<Altadil>unmush: protonmail does not block tor. For interacting with desktop clients (including git send-email) there is a "bridge" app.
<Rutherther>unmush: I use hydroxide with protonmail. They claim you can use bridges only with paid accounts, I paid only for proton pass, not for e-mail, and I can still use it for sending e-mails... so I am not sure if someone who has completely free account might use it for sending e-mails or not
<Rutherther>(hydroxide bridges smtp/imap to protonmail)
<jaadu>Rutherther: Could you help me write such a package?
<old>Rutherther: hydroxidy is working now on Guix???
<old>I've been waiting almost 2 years for it for my proton mail!
<jaadu>Sorry for being slow
<old>jaadu: You packaed hydroxide?
<old>thanks for that!
<old>I really don't like packaging go-package, which is probably why I did not tried to package it myself. Kudos!
<Rutherther>jaadu: https://paste.debian.net/1337446/
<jaadu>Rutherther: Big thanks!
<unmush>I don't know whether it's because I used a non-tor proxy this time, but it only demanded an email address this time
<unmush>so I was able to send the series as attachments
<orahcio>Hi everyone, I start to contribute with Protuguese (Brazil) translation on the guix-manual, some tips to do this better?
<futurile>orahcio: oh nice
<gabber>orahcio: cool! i do not have much experience with translating work. what field/workflow do you wish pinters for?
<orahcio>gabber: I think it could be nice if I can translate the same topic on manual that I read to do my studies on guix
<gabber>orahcio: sorry, i have almost(?) no experience with translation workflows (: let's wait here a bit (maybe someone else can help), otherwise send an email with your question to help-guix@gnu.org. the mailing list is a more sure way to get help, though it tends to be a bit slower
<orahcio>gabber: Thanks for the attention help, someone told me about to contribute with translation is a good start point. I will to note the email, but this channell is also very nice to get help too. Everytime I ask for help somebody can help, great community!
<wolfdog>is there any way to build a QCOW2 image using the images API but using an MBR layout/BIOS bootloader instead?
<bjoli>Ahoyhoy! Does anyone have a way to install guix on aeon desktop? Is it enough to create a subvolume of /gnu and marking it "ro false"?
<gabber>civodul: i just successfully upgraded postgres (and got cuirass running again). where is the right place to share my insights? help-guix@gnu.org?
<bjoli>I will answer my own question. Yes. Yes it seems like that is enough.
<gabber>bjoli: nice!
<bjoli>I don't really know about what will start happening when the system is updated. I disabled CoW which means snapshots shouldn't work, which is what I want.
<bjoli>aeon is great, but handling everything that can't be handled by flatpaks using distrobox is painful. If this works out I am going to be the most happy fish.
<civodul>gabber: yes, or guix-devel even
<civodul>i’m interested anyway!
<gabber>it's been posted to help-guix, since it is an intention to help ;)
<civodul>heh perfect, thanks :-)
<gabber>you're welcome
<gabber>developing on some code for an application i run into record-abi-mismatch (i use guix as packaging, build and deployment system). do i need to delete some guile bytecode files? where would i find those?
<noe>IIRC The manual says to “make clean-go” and build everything again
<noe>so delete every .go file
<gabber>i am not working "on guix". i am working on unrelated code (which uses guix to package and deploy)
<noe>My bad, if the .go files are not in the repository maybe you can find them in $HOME/.cache/guile/ccache
<gabber>noe: thanks!
<noe>you’re welcome :)
<bjoli>So... I had to rollback my aeon system, and re-install guix. My user guix install is looking for something in /gnu that is obviously not there. Is there a way to purge everything related to a user? I have removed everything in my home directory, but that does not seem to cut it.
<bjoli>it tries to look for a thing in /gnu/store/hashhashhashhash
<gabber>bjoli: this is somewhat expected. pretty much everything in guix works from stuff from the store, which is at /gnu/store/
<bjoli>oh yes, but it doesn't exist anymore
<gabber>so guix won't work anymore
<bjoli>guix is re-installed and works for root
<Rutherther>bjoli: so what are you executing exactly that is looking for that path exactly, and what is that path?
<bjoli>but my main user had run some commands
<bjoli>guix pull
<gabber>ah, i see. what is the (full) error you see?
<Rutherther>bjoli: what is output of "type -p guix pull"?
<Rutherther>sorry, "type -p guix"
<bjoli>gabber: Does guix pull: error: reading file `/gnu/store/i11f4vrgrdp0g666zcvk6k8pqj9ywjr8-le-certs-1.drv': No such file or director
<bjoli>Rutherther /usr/bin/guix
<bjoli>so, my profile points to something is /var
<Rutherther>bjoli: how did you "reinstall" guix?
<bjoli>with the opensuse package
<Rutherther>so why would stuff from /gnu/store go missing?
<bjoli>i had done a rollback, so the /gnu directory did not exist
<Rutherther>I see
<Rutherther>then remove /var/guix/profiles/per-user/$USER if that does exist
<bjoli>great
<bjoli>thanks
<bjoli>because /vas is not affected by snapshots :)
<bjoli>var
<wolfdog>having a weird problem, i'm using static-networking on my system definition, and when booting networking works, but the networking service is not started and gives an exception "netlink-response-error errno: 17" when trying to start
<wolfdog>but all interfaces are up and working
<weary-traveler>sneek: botsnack
<sneek>:)
<Rutherther>bjoli: also rethinking it. I would recommend deleting everything in /var/guix. Then you will have to start over even with the root stuff. But it's safer, because if your database stayed, but /gnu/store didn't, you will be in a lot of trouble
<bjoli>will do
<Rutherther>bjoli: if you already have something in gnu store now, remove both /var/guix and /gnu/store again. Otherwise you won't be able to get the store and db in sync
<bjoli>Rutherther only .links
<Rutherther>bjoli: no, the db has the state of the store. If state of the whole store is different than what the db has, there is trouble
<Rutherther>and for future, the db has to always follow the store. I don't know how your rollbacks work, but it's important to never roll back /gnu/store without rolling back the db, ideally the whole /var/guix
<Rutherther>if you will be doing any rollbacks in the future that roll back only the store, your guix stuff will be breaking
<bjoli>Rutherther: thanks! The /var and /gnu dir are excempt from btrfs snapshots, so this should not be a problem. I just messed up my first subvolume for /gnu and that is where all the trouble started. thank you for the help.
<wolfdog>ok, i did "ip addr flush eth0" and started the networking service again, and now it's erroring with "In procedure u8-list->bytevector: Wrong type argument in position 1: #f" :S
<bjoli>Rutherther: I managed to create the first subvolume as a part of the current snapshot, which worked fine until i did a sys upgrade. Everything is supposed to work now.
<orahcio>Somebody knows any reference to manage fish shell configuration with `guix home`?
<Rutherther>orahcio: there is fish-service-type in guix. It's undocumented for some rea
<Rutherther>s/rea/reason
<orahcio>Rutherther: I could see on manual home-fish, but just a output example, we have bash and zsh topics, I would like some blog entry or github guix configuration with this service. I use on my main machine the fish shell, I want to do the same but now with guix
<sneek>weary-traveler: Greetings :D
<ieure>Hmmm, I'm trying to build a package from my channel with `guix build -L. librewolf', but the module the package is in won't load.
<ieure>> In procedure abi-check: #<record-type <origin>>: record ABI mismatch; recompilation needed
<ieure>There are no .go files in the channel, so I don't know what ABI mismatch it's complaining about.
<ieure>What might be wrong and how can I fix it?
<Rutherther>ieure: you can try removing ~/.cache/guile. I am not completely sure it's going to fix it though
<ieure>Rutherther, That fixed it, thank you!
<wolfdog>hm, guix deploy is failing to connect to my server via SSH, saying "Connection refused", but the port is open and I'm logged in on another terminal
<wolfdog>ACTION thinks
<wolfdog>hm, using the IP address worked
<anemofilia>orahcio: I use fish as my main shell (including login shell), and have everything configured declaratively (including plugins), if you want to take a look: https://codeberg.org/anemofilia/zero
<anemofilia>(Ah, I'm from Brazil btw :D)
<orahcio>anemofilia: Thanks (good to see a brazillian)
<anemofilia>:)
<anemofilia>A little detail is that I'm currently using a fork I did for the home-fish-service-type, that adds better support for abbreviations and support for plugins
<anemofilia>I plan to upstream it, but I would like to implement suport for declaring completions prior to the patch submission
<anemofilia>orahcio: ping
<PuercoPop>anyone has setup a test runner for their channel that doesn't involve autoconf?