IRC channel logs
2024-12-30.log
back to list of logs
<homo>it appears that installing git via home config instead of system config results in broken tls <jmes>When writing Guix Home services, I find myself wanting to provide a #:pid-file to make-forkexec-constructor that references $XDG_RUNTIME_DIR, but when I try to start the service, it complains that "PID file '$XDG_RUNTIME_DIR/example.pid' did not show up". Right now I'm hardcoding /run/user/1000 which works on my single-user system but I wonder how to send XDG_RUNTIME_DIR to make-forkexec-constructor... <jmes>i.e. XDG_RUNTIME_DIR doesn't get expanded in the #:pid-file field, but I can't know it at home reconfigure time - my knowledge ends there. <homo>aren't they part of %base-packages installed by system? <cancername>jmes: is it possible that you can use (ungexp (getenv "XDG_RUNTIME_DIR")) <homo>then I have them installed, considering I also use netsurf and hexchat installed via home config <cancername>what are the values of GIT_SSL_CAINFO and SSL_CERT_DIR <homo>GIT_SSL_CAINFO=/run/current-system/profile/etc/ssl/certs/ca-certificates.crt SSL_CERT_DIR=/run/current-system/profile/etc/ssl/certs <homo>netsurf and hexchat work fine with tsl, only git doesn't <homo>unless I use "guix shell git", then git works with tls <homo>it doesn't matter which repo I try to clone or pull <jmes>cancername: Ah yes, perfect! That worked a charm. I am not well versed in using g-expressions yet <cancername>jmes: no worries, am new as well and had to bang my head against the wall quite a bit, manual helped tho, especially the bit about the store monad <cancername>homo: the only thing that comes to mind is also installing curl <homo>it didn't have that error when installing through system config <homo>is this dependency picked randomly depending whether git is installed in home or system? <homo>when using git from guix shell, it doesn't have curl either, but https works <homo>cancername after running "guix shell git" and inside it "which curl" it says "which: no curl in (...)" <homo>libcurl is not installed and guix shell git doesn't install it either <homo>I checked the lib directory of profile generated by shell <cancername>homo: in my "git shell -CN git", "git-remote-https" is available and links to libcurl, so I'm not sure what the deal is here <homo>like I told already, git-remote-https is available in shell, but not outside of it <cancername>homo: are you sure your GIT_EXEC_PATH is set correctly? <homo>ah, it's set to /run/current-system/profile/libexec/git-core <homo>it appears I have to relogin or reboot <homo>well, thanks, git is so git <homo>btw, just now I fixed fontsrv in plan9port, however, I have no idea why openfont1() from src/libdraw/openfont.c doesn't convert "/lib/font/bit" to "$PLAN9/font" <homo>or it does, but doesn't do that for acme... <homo>"9 acme -f /lib/font/bit/fixed/unicode.8x13.font" works fine, it appears the problem is in parsing extra . in "latin.1.7.font" <jmes>Anyone know how to extend XDG_DATA_DIRS with Guix Home? I read some of the pertinant source code and found that XDG_DATA_DIRS is treated a bit differently than the other Base Directories—almost certainly to ensure the various guix profiles' data dirs remain exposed. But that leaves me wondering how to append another data dir of my own. <jmes>I was hoping to add something like ("XDG_DATA_DIRS" . "$XDG_DATA_DIRS:$HOME/...") to home-environment-variables-service-type, but no dice. <graciousgnu>im currently chatting from GNU erc on GNU emacs on GNU guix, while having having gnu.org open on GNU eww lol <graciousgnu>do people here mainly use guix system or just the package manager and on another gnu/linux distro? <jmes>graciousgnu: No clue, there was a survey recently that would provide some insight but I don't think the results are public yet. Personally I use Guix System. <jmes>Now that the survey is on my mind, does anyone know when those results will be available? <Kolev>Hi all. graciousgnu, I use Guix System. Currently configuring Sway and Guix Home to my liking. <Lumie>graciousgnu: I use Guix System, been on it for couple of years plus <Kolev>Lumie, agh, OK. Looking for advice on configuring it. <Lumie>That's the problem and reason why I never put it <Lumie>Last time I heard, it wasn't that great on Guix <Lumie>But that was literally years ago <Kolev>Lumie, oh no, a lot of people use Sway on Guix as a lightweight desktop. <Lumie>I see no point in switching at this point because it works just fine for me <cancername>I'd really like to use River again, but the Zig bootstrapping efforts are still in the wip-zig-bootstrap branch, so I'm using GNOME for now <cancername>Lumie: I've been liking Wayland because I can sandbox my applications that way <Lumie>I'm more on the side of usability <cancername>unfortunately, it still exposes all of /gnu, unlike guix shell -C :/ <Lumie>I remember that some Wayland apps looked horrible <Lumie>cancername: that was it, it had something to do with the rendering, don't recall <meaty>how long should it take for a patch to be seen? is there something wrong with what I sent if there's no response except for the server acknowedgement <homo>if it doesn't work, then wait for moderators to approve <homo>my first patches took many hours to appear <meaty>I mean they're there, I see them in the archive and your link, they've just been there for almost a week now <meaty>which is fine, just want to be sure I'm not missing some registration step or smth before it gets looked at <homo>ah, it takes a while for committers to review, besides they are busy doing other work than reviewing patches <ekaitz>let's see if i can help you with it <homo>meanwhile I'm waiting 3 weeks for 6-line patch to be merged, it simply enables simple direct rendering manager driver, as without it I cannot use neither gdm nor wayland https://issues.guix.gnu.org/74778#2 <homo>when compiling kernel manually, it is as painful as compiling llvm, webkit and qt6, because laptop overheats and it occupies all ram, so I have to enable swap <ekaitz>homo: i'd love to just merge it, but I know nothing about the kernel so IDK if it might break anything else <hako>homo: I suggest you to CC kernel team <homo>I can't find e-mail address of kernel team <Rutherther>homo: it's in /etc/teams.scm under guix, ie. "./etc/teams.scm list-members kernel" <cluelessguixer>Unfortunately I'm clueless, although I've managed to make a virtual network bridge... not sure how it helps me though. <mgd>Hello all. I'm getting this error when trying to install `python-fastapi` https://pastebin.com/RpwzLmRh. It seems that the build for `python-jose` failed, which is a dependency for `python-fastapi`. I would appreciate any guidance on how to troubleshoot this issue or if there are any common solutions for this type of error. <cluelessguixer>Actually, maybe it did work. "herd start wireguard-wg0" worked afterwards, but there's still something missing. <cluelessguixer>Looks like I don't need a bridge for wireguard-wg0 service to successfully start. Now to figure out how that's even possible when there's no indication that the container has internet access... <mgd>I'm happy to share the log output if it helps. I don't know Python well enough to see what's gone wrong <pisemsky>Hi guix! May I ask for a patch review here? I intended to package usbrelay completely, and this intent stuck on issue 72635. <PotentialUser-76>It used to be the case that I could renew certificates manually with /var/lib/certbot/renew-certificates <Franciman>hi, is there any blocker to updating telegram-desktop to a newer version? It is stuck at 5.5.5 <pisemsky>Franciman: I don't have an answer but curious if you experience some problems with it? Because I do - it segfaults when I try to open a channel. <Franciman>pisemsky: for me it segfaults after login succeeds lol <Franciman>i started using the web version, it works quite fine <efraim>oops, broke maturin on rust-team, I guess it's time to upgrade it to fix it <hako>Which Zig version have you reached on powerpc64le btw? <efraim>hako: I hadn't gotten that far on powerpc64le or riscv64, different errors on each, but I had it working completely on aarch64 <efraim>is that supposed to make it Just Work™ on all architectures? <hako>Yes, but I'm not sure if it's substitutable <efraim>I don't think that really works, I think that would just limit it to x86_64-linux <efraim>and then if you removed the field later you'd have other architectures trying to build x86_64 packages as part of the bootstrap I think <efraim>I rebased mesa-updates on a commit that QA/data-service had already processed <janneke>gnu/packages/gtk.scm:926:13: gtk+ would be upgraded from 2.24.33 to 3.94.0 <janneke>as it so happens, there's no newer 2.x version <janneke>ACTION goes trough the "add -Wno-error=" flags motions <efraim>aalib also fails to build, but I'd rather see about tearing it out :) last release 2001 <janneke>mathematics from 500BC still works, why wouldn't software from 2001? <efraim>I followed the instructions for a cvs checkout, no newer code <janneke>after -Wno-error=incompatible-pointer-types, /me now also adds -Wno-error=implicit-int... <efraim>where's the -Wno-error-yolo flag? <janneke>efraim: i don't believe i have an aalib depenency in my system config (that's my current target) <efraim>alternatively, would using a lower C standard decrease the need for the flags or are the checks on by default regardless? <janneke>it would have been so nice to have a --stdc=xx <janneke>it's kinda silly to me that gcc-14 can just refuse to build programs that were previously deemed ok <efraim>at least debian is still patching aalib <efraim>I'd rather write a package transformation to remove it, but that wouldn't fix it for others <efraim>in the meantime I guess I'll see some of the systems in the examples <Lessness>Hello, I am trying to use guix from emacs with the emacs-guix package. However, I run into some issue with the guix-all-packages command. It says it launches a REPL for guix then it just blocks and after a while timeouts. Has anyone else experiences this issue? Is there some configuration needed for it to work? <Lessness>The guix-installed-packages command on the other hand works perfectly fine <jmes>I'm still confused why adding ("XDG_DATA_DIRS" . "$XDG_DATA_DIRS:example/dir") to home-environment-variables-service-type doesn't work. The line gets added to ~/.guix-home/setup-environment where other variables are exported just fine 🤔 <jmes>"Doesn't work" means that "echo $XDG_DATA_DIRS" doesn't show the new dir appended on. <podiki>efraim: re: mesa-updates; does that mean we'll get an actual QA comparison soon? <Rutherther>jmes: did you relog or source the file you poionted to manually? <jmes>Rutherther: Logging in on another tty or sourcing manually worked, I was just expecting guix home to re-initialize the profile like I think it does for the other environment variables I set. Anyway, thank you! I'll keep that in mind next time 😅 <Rutherther>jmes: it doesn't do that, it's impossible to do that (to arbitrarily change environment variables of running processes). The only thing is that most environment variables already point to existing folders that are replaced with new generation, so you get new programs in path, or elsewhere, like that <jmes>Rutherther: Yeah I don't mean for running processes, I mean for a newly opened term emulator for example. I would normally test by opening a terminal emulator which sources the new profile, but that wasn't working this time. Dumb of me not to test relogging though :/ <janneke>sbin/smbd: error: depends on 'libtevent-util.so.0', which cannot be found in RUNPATH <Rutherther>jmes: opening terminal doesn't source new profile, it sources only rc files. If you want to source profile, you need to run with "--login" flag <janneke>ah, libtevent-util is being built by samba hmm <jmes>Rutherther: makes sense why it doesn't work! Thanks for the info. <janneke>hmm, it seems that ...samba-4.17.0/lib/ is not in RUNPATH? <janneke>oh wait, it's samba/pinned and can probably be updated.. <olafes>I think package transformations should have a note in the manual explaining that they are not fully integrated into guix yet <olafes>Some like with-source with-patches are good but with-c-toolchain or tuned-package are inconsistent at best and useless at worst <luca>oh hey I had a RUNPATH issue yesterday as well. I used -rpath, but honestly no idea if that was the correct choice <Lessness>Is there a particular reason why autocompletes of guix commands are slow ? <Franciman>Lessness: i have no idea, i suppose it's guile's slowness <Lessness>Does someone here use emacs-guix to manage guix? <Rutherther>luca: so I suppose something from the bin folder depends on something from lib in the same package? <luca>xbps- binaries depend on libxbps.so <Kolev>I have ` (delete gdm-service-type)` but GDM still shows up when I boot. <dariqq>Kolev: adjust the login manager that is used for set-xorg-configuration or remove it alltogether if you dont want one at all. If no value is specified the default value (gdm) is used. <olafes>Why do we use gcc@11 by default? <Kolev>dariqq, ah. I'm using Sway and would like something appropriate for it. IIRC Sway doesn't use a display manager, just a TTY. <efraim>I wanted something to setup all the magic login bits for me, so I figured I needed a login manager of some sort, and at the time I thought it had wayland support in Guix <efraim>and after installing quassel and a webbrowser I didn't think I was saving much space by not using it 