IRC channel logs
2025-04-28.log
back to list of logs
<apteryx>there's indeed no include/format.h in the clang-toolchain@18 package <apteryx>ah, it's not format.h, it's just format. for libcxx from llvm/clang it's under include/c++/v1/format <apteryx>snappy-with-clang6 does something for that (in its adjust-CPLUS_INCLUDE_PATH phase) <dhoffman>is there a mirror for the ci results? i've been trying to access the ci.guix.gnu.org website for an hour and consistently get 504 errors <dhoffman>does Guix not have the funds to run the website? is the stack behind the CI particularly weird in a way that doesn't scale up? <dhoffman>are there a billion AI crawlers blasting it? <ieure>dhoffman, There is no mirror for CI. Guix infrastructure is all donated, some from GNU, some from corporations, some from contributors. <ieure>dhoffman, Guix is a small project, CI is pretty much a one-man show. <dhoffman>would money solve the issue or is the bottleneck technical expertise? <ieure>dhoffman, For what it's worth, the non-HTTPS CI site seems to be working for me. <dhoffman>maybe SSL negotiations are the reason why it's hitting a gateway timeout? <ieure>dhoffman, It's... the same site. You don't need a link. <dhoffman>the phrasing non-HTTPS CI site made it seem like a different URL <ieure>The protocol is part of the URL. <ieure>Anyway, both HTTP and HTTPS seem to be working for me at the moment. <dhoffman>that's diving into semantics at this point <ieure>I don't think there's a CDN involved. <ieure>Hmm yeah, it's intermittent. <apteryx>lilyp: in a fun twist of 'I will ensure you miss meeting' from GNOME calendar, now we can add event reminders, but they don't seem to trigger any notification. <cow_2001>just wanted to say that i love you. guix is awesome <lilyp>looks like there's some defense in depth going on 😁 <apteryx>is there a reason we are keeping gimp-2 around? <apteryx>lilyp: I don't really understand how extending GI_TYPELIB_PATH can crash an application as discussed in bug#77921, but perhaps in this specific case it could be OK to wrap exactly (=) and not allow extending GI_TYPELIB_PATH, since it seems sensitive? <lilyp>apteryx: I mean, you could, but you'd be breaking gfeeds <apteryx>I meant in the specific case of gimp, not in general <apteryx>how does gfeeds make use of extending GI_TYPELIB_PATH? <lilyp>ahh, thought of the wrong bug, sorry <lilyp>pinning typelibs for applications should be fine <lilyp>it could break extensions, but it's likely not going to <adanska>I dont know whats caused this, but the repeat paths are causing issues for some of my programs. <apteryx>would someone know why attempting to add a bind mount to %base-file-systems cause: statfs-raw: No such file or directory ? <apteryx>hako: I was try6g ot just add %runtime-variable-data/bind-mount to the %base-file-systems, I guess you walked that path too <apteryx>I don't see the modification to mkdir-p/perms in your linked commit <apteryx>it's also not happy with a bind mount? <tclaesson>Hi! I have a question for team-r: I'm slowly working towards <tclaesson> building r-cardx from guix master. Multiple dependencies of this <tclaesson> package fail tests, which is typically fixed upstreams and easily <tclaesson> remedied on the guix side by a simple point release update. Do you <tclaesson> prefer separate issues/patches for each of these <tclaesson> dependency-updates, or would you prefer one patch with all <cbaines>tclaesson, providing the separate packages are changed in separate commits, either approach is fine <nutcase>Hi guix! if a (new) package builds, but fails in phase 'validate-runpath' (no libs are found), what is a good starting point to solve this kind of issue? <apteryx>hako: even with the bind mount, the jami-provisioning test still fail. i have no clue why. <apteryx>hako: oh, some clue: warning: failed to delete /run: Device or resource busy <apteryx>are the system tests using grafted packages or not? <cbaines>I think they can, it probably depends on how they're computed <ArneBab>cow_2001: I’d like to second that! Thank you all for your work on Guix! <cow_2001>i just copy and paste stuff then change some fields <dcunit3d>is it typical to include tests that require network access? <futurile>dcunit3d: tests in packages/services right? These don't have access to the network at build-time so they need to be disabled <dcunit3d>well i was building libtpms and swtpm, but swtpm fails so many tests that idk. and it's tough to disentangle the networking (basically the entire test suite requires network) <hako>apteryx: jami-provisioning tests hang on my system it seems, not sure why. <apteryx>on mine too, after any attempt to make /run a tmpfs <apteryx>I tried symlink and also the bind mount approach, same result. <apteryx>even had a grafted dbus that used /run instead of /var/run without change. I guess I need to start paying attention to *what* the actual problem is <dcunit3d>idk if i can even use it, but i wanted to put ssh host keys in tpm so a sops file can be decrypted with them ... but that's alot of process lol. also, idk if a deployment with sops-guix could really give me a chance to enter a pin <jlicht>hiya guix, good praxis question: I've only joined a team after some GCDs were shared: can(/cq should) I still share my position on any? <apteryx>hako: are you sure the bind mount approach is better? does it behave well in containerized (e.g. least-authority environments). <jlicht>dcunit3d: GCDs 2 and 3, specifically. <apteryx>jlicht: I guess you can express your positions on the GCDs that will be submitted after you joined a team but not before <jlicht>makes sense to me, thanks for the rubber ducking :) <dcunit3d>yeh i saw those, but idk. i think the email workflow is more robust to interference/censorshop. the discussion on GCD 003 was helpful because i was never really sure what the rationale was for master/main <dcunit3d>idk that my opinion really needs to be taken into account though. obviously these issues affect the core devs more. it seems like a lot of work to migrate though, but migrating off of savannah seems to be happening on a few other GNU projects <apteryx>hako: at least fedora seems to be using a symlink for /var/run -> ../run <tom2342342311>I just installed guix and getting this on my first `guix pull`: <tom2342342311>guix pull: error: Git error: unexpected http status code: 502 <hako>apteryx: seem to be a requirement for systemd <tom2342342311>Seems like the package website is also having troubles.. searching or trying to load a specific package just times out: <tom2342342311>Or is there another way to find what versions of a package are available? <hako>but there's no reason to not use a bind mount <identity>tom2342342311: guix show python | grep version <ruther>identity: that shows just the latest version <tom2342342311>Currently just shows a very old python 3.9.9, but I'm waiting for "guix pull" to finish.. quite slow at indexing. Hopefully there is something more recent than that.. <ruther>tom2342342311: it shows the ones that are available from the guix version you're using. <ruther>tom2342342311: latest guix has python 3.11. <tom2342342311>Ah I just downloaded the image last week, I guess it needs an update. 3.11 is not bad <ruther>tom2342342311: what image did you download exactly? <ruther>it is the latest. And latest release is two years old. So it's two years old <tom2342342311>That's okay. So far installation was much smoother than NixOS. Guix docs are sooo much nicer. <cobra_>getting "libc.so.6: CPU ISA level is lower than required" when dbus-uuidgen is run on the latest installation image build <cdegroot>quick question - I'd like to look at contributing tailscale client + headscale server, both 3 clause BSD. That latter license is Guix-compatible, not? <ruther>cdegroot: there are plenty of bsd licensed packages in guix channel <ieure>PotentialUser-96, It has been down often for a couple weeks. <identity>cdegroot: all free licenses are fair game, definitely those defined in guix/licenses.scm (so bsd 0 through 4 clause are fine) <cdegroot>identity: thought so, but it'll be Some Work so rather have someone else confirm :). Thanks for the pointer to the approved licenses, I forgot about that file (I've been on a Nix stint, mostly to help get things introduced at work, now reverting my hardware back to Guix and noticing a couple of missing packages) <ruther>PotentialUser-96: guix pull to earlier commit is completely fine, but I would rather avoid guix system/guix home downgrades if you don't have a reason to do that. Should be safe most of the time though <old>how can I tell guix-edit(1) to prefer definitions in guix channel and not my channel? <cdegroot>ruther: I know that there's multiple tailscale Guix packages (but rather have something that important from Official Upstream), the big thing that seems to be missing is headscale, the open source control server. It allows you to create a tailscale overlay network without relying on the (closed source/SaaS) Tailscale service. <PotentialUser-42>Could you point me to some resources that explain *how to manage Emacs packages from within Guix System*? <ieure>PotentialUser-42, I don't know of an existing resource for this (!= none exists). I plan to submit a talk proposal for the EmacsConf on the subject. <erik``>When in Emacs, following the definition of an Elisp function defined in C is unable to locate the Emacs source directory. Is there some way to change my home configuration so that it can find it? <hako>erik``: Set find-function-C-source-directory in Emacs. <erik``>hako: Are the sources contained in the guix emacs package? I am unsure what I should set the variable to. <hako>No, Emacs package in Guix uses tarball for source, you need to unpack it somewhere. <tom2342342311>Is there a list anywhere of the wifi chipsets with firmware supported by the linux-libre kernel? <ieure>tom2342342311, That's the newest one. <identity>tom2342342311: (info "(guix) Hardware Considerations") <tae>If I am at the REPL, how can I get the source code of a procedure. For example, "> build-expression->derivation" prints $2 = #<procedure ... at guix/derivations.scm:1268:0, but not the source code. <identity>tae: i think `,help all` should list some interesting commands <tae>Yes, but I want to jump to the place where the procedure is defined. <tae>Something similar to M-x describe-variable in Emacs. <identity>tae: Geiser might provide something like that, i'm not sure <luca>Hi, does oci-container-service-type support running podman instead of docker? <orahcio>Hi, How about the guix-home-service-type? I could not search this service at gnu/home.scm, but the guix manual shows it module. How to use it to deploy my home configuration with the guix system instead guix home reconfigure? <sham1>(service guix-home-service-type `(("user" . ,home-environment))) <sham1>From (gnu services guix) IIRC <tadhgmister>I am currently in a guix installer on a machine that thinks it is Jan 2022, how can i get it to update the time? <ieure>tadhgmister, `guix shell ntp -- ntpdate -u pool.ntp.org' <untrusem>tadhgmister: you can access shell and run the command ieure sent <tadhgmister>ok but i cant `guix build` or `guix pull` as it fails to download anything as the ssl certs are not happy about the time <ruther>tadhgmister: just set the date with `date` command <ieure>tadhgmister, Oh, that's a bummer. ... yeah, `date' will let you set it manually.