IRC channel logs
2025-02-22.log
back to list of logs
<sobol>Aurora_v_kosmose : been that way for days now :( <CoopCoding>So I just installed guix in a vm and chose awesomeWM as the window manager, but when it starts there is no terminal so I cant install anything. I got around this buy doing Ctrl+Alt+f2 to get to a tty, but it feels like there should be a terminal installed with the default install. awesomeWM is expecting xterm to be there, so perhaps it could be <meaty>Is there a way to open internet access to a build process? A package I'm making just needs to download some testing data <meaty>also, do substitute servers keep the hash of built packages after deleting them due to lack of downloads? if not, it would be useful 'cos then you could check if a locally-built version of the package like, 5 years later is bit-perfect to the substite that would be served then <meaty>should 'delete-all-but' be moved into (guix build utils), or some other guix module? It's used in a LOT of package defs <rekado>meaty: you cannot open internet access (i.e. add the current state of the internet as an input) to the build process. <rekado>meaty: many build systems give you a way to override the location of test data, so you can download them, record the hash, and add them to the native-inputs. <qzdl>setting up build-vm on an x86 box; but receive the 'unsupported-platform' error with `(targets '("aarch64"))' & some `guix build efivar -s aarch64-linux' <qzdl>inspecting the launch command from shepherd, it is booting qemu into x86 <qzdl>^^ oops, misrepresenting the issue here, actual config snippet is `(systems '("aarch64-linux"))' <futurile>zlqrvx: basically your build probably can't find zlib because it's expecting it in a standard FHS location and it's not there. There's probably a couple of ways you can solve it such as providing a configure flag or specifying an env variable. Checkout the guix source and look in gnu/packages - there's quite a few packages where zlib is manipulated <meaty>Is there a "native" way in scheme/guile to compare two file paths and tell if they're equal, taking into account all the ways a path can be expressed? <meaty>i.e. something that can tell that ./this/file and ././/this/file are the same? <meaty>is it safe to assume that "canonicalize-path" will work to "normalize" the names? <mra>meaty: that seems safe. "canonicalize" generally means to reduce something down to a unique representation <mra>indeed, the documentation for Guile says that canonicalize-path returns an absolute path with no . or .. éléments, no repeated / separators, and no symlinks. <meaty>true, I just wanted to be sure that there would never be a situation where containers/permissions made the root un-seeable or something <meaty>and if so, would canonicalize-path do a good best-effort attempt <apteryx>someone I know reports facebook video conference is not working in icecat, librewolf or even ungoogled-chromium; anyone knows what it requires that isn't provided by our browsers? (I don't have a FB account to check). <apteryx>another question: aren't GCDs supposed to be discussed in a Debbugs issue? That'd make it much easier to recover the whole thread. <futurile>apteryx: I think the final vote gets tallied in the debbug, but the discussion is on guix-devel <apteryx>ah, I missed that. that's unfortunate <apteryx>well, guix-devel is fine, but finding the full thread is made harder <dariqq>is there a way I can add an account from a service-type to a supplementary group? <dariqq>tried restricting access to the guix-daemon socket to a special guix group but now guix-publish cant connect <mra>hm, I guess I should take a crack at solving 41602 <dariqq>succeeded in adding the guix-publish user to the group by adding a second entry , but it does not work because the guix-publish script does not set supplementary groups when dropping privileges <mra>does anyone else have guix pull failing on their machine? <ieure>mra, It is broken for me, "&missing-dependency-error: module: (gnu build hurd-boot)" <ieure>Hm, I don't see anything obvious that would cause this. <csantosb>luca: we have a (release ?) tag in commit history <mra>ieure: mine's also failing, although for a different reason <mra>just a sec, there's a problem with the log file. <mra>I'm using a fresh install of Guix, just with some standard non-free stuff so that my GPU works <mra>not to complain about an issue that I'm not solving, but shouldn't some CI verify that changes don't break `guix pull`? that seems like a rather bad thing to break... <ieure>I think it might be 8451e489bd. <csantosb>This: 013cc1952a * v1.4.0+154701.013cc1952a8 gnu: Add x86-video-armada-novena <ieure>csantosb, You're saying that commit is what broke `guix pull'? <mra>csantosb: I'm on codeberg right now. seems fine <csantosb>ieure: no, no, I just noticed a new tag, I guess, related to new release planning; I just guix pull with no issue <mra>csantosb: is your guix channel pinned to a specific commit? <ieure>I'm trying to reproduce with `guix time-machine', which I could run in a `git bisect', but it's extremely slow. <ngz>When I try to run a Python package (using pyproject-build-system), I get "Could not load a pixbuf from /org/gtk/libgtk/theme/Adwaita/assets/check-symbolic.svg", then no icon is available. What am I missing in the package definition? <ieure>ngz, Probably need to specify that icon theme in a package input and patch wherever that path is located in your package to point inside it. <ngz>I tried to propagate (librsvg-for-system) and wrap GDK_PIXBUF_MODULE_FILE to no avail. <ngz>I also added adwaita-icon-theme to the inputs. <ieure>ngz, Does that path you pasted exist in the Python project you built? <ieure>mra, Okay, yeah, won't be able to isolate this broken commit. But if you find it, I can revert in a bit, of nobody else gets to it first. <ngz>ieure: The /org/…/ is not a path, is it? <ngz>Hmmm nothing is installed under share/ <dariqq>ieure: f302e6bc1a2d42ef3a2088b6c2e82a36ff03cd05 is the first bad commit. I guess because of the import of (gnu) module <ieure>dariqq, Thanks, were you using `guix pull --commit=' to verify that? <mra>oh, did you find the bug? <mra>I'm starting to go a bit insane with Guix. may be time for me to take the rest of the evening off <dariqq>ieure: have not tested the (gnu) theory yet (was struggling to convince guix to use my local checkout) but it is pulling now <dariqq>ieure: removing the (gnu) import made guix pull not error and apcupsd still builds <rekado>dariqq: thanks for testing this! Do you have a patch you'd like me to apply to fix it? <dariqq>rekado: have not sent a patch but the fix is to delete the #:use-modules (gnu) in gnu/packages/power.scm <rekado>would you like attribution in the commit? <rekado>running "make"... pushing the fix in a minute <wakyct>untitledgoose42: best to just ask your question <untitledgoose42>OK, 2 questions. 1: I got a AR2971 USB wifi adaptor and plugged it in but have no internet. 2: I can't get Wireguard to work (need to do 1 first) <untitledgoose42>I run lsusb and it shows (Atheros) but no internet. But on docs says Atheros is supported. <ieure>untitledgoose42, Atheros made different models, some supported, some not. ath9k is the one which is well-supported. <ieure>ath10k isn't supported, because it requires firmware blobs. <untitledgoose42>Thank you, I believe I have ath9k. However Explaining this just made me realise I need to actually find a network and connect to it, if I have any problems with that I will say <ieure>Yep, that's important all right. <rekado>shepherd 1.0.2 fails to build when I want to deploy to my aarch64 machine. <untitledgoose42>I have now connected to my network. My second question is, I try to connect to commercial VPN provider. I have autogenerated Wireguard config file that works on other machines. I put same details into Guix wireguard-service-type but does not connect. When I run nmcli c the name that shows is different to the peer name I say, so I think it is create a new Wireguard interface rather than use the one I say. <fnat>I understand one can use `--expression='(@ (module) symbol)'`, how about something like this instead '(@ (module) expr)'? <fnat>E.g. `--expression='(@ (module) (list symbol))'`. <sneek>Welcome back civodul, you have 1 message! <sneek>civodul, csantosb says: strange: looking at it until I run out of battery, ... no clue about right here about the unbound variable <jA_cOp>When defining a custom shepherd service in my guix home config, how can I see its log? Or rather, my service fails to start and I'm pretty sure I know why, but I want to know how to properly debug that <civodul>jA_cOp: you can check the main log, either with ‘herd status root’ or by looking at ~/.local/state/shepherd/shepherd.log <civodul>fnat: hey! the @ form is special, the second part must be a symbol <civodul>so it’s always (@ (module) variable) <civodul>if you can several variables, you have to write (list (@ (module) var1) (@ (module) var2)) <fnat>civodul: Brilliant, thanks! That will do it in my case. <civodul>rekado: it successfully built on overdrive1; ci.guix should have substitutes now <civodul>/gnu/store/sxj06mkkavdssnwaacxhxv191p96z10g-shepherd-1.0.2 <jA_cOp>What's the usual way to set environment variables for a Shepherd service? I'm using forkexec-constructor. Do I just wrap the command in a shell script? <jA_cOp>Oh, forkexec-constructor has #:environment-variables, nvm!