15 Oct 2024 |
Whovian9369 | In reply to @whovian9369:matrix.org I'm not sure - Doing nix build -L --no-link --inputs-from . lix and the same for lix-module seems to rebuild lix ? Can't immediately confirm, but that's just my anecdotal experience with it. I'll double check later. | 21:39:46 |
Whovian9369 | $ nix eval --inputs-from . --raw lix
warning: Git tree '/tmp/wat/My_Config_Flake-main.bundle_out' is dirty
/nix/store/h3lndwfn54dpfb2ncvb9jbzpvqflw9q7-lix-2.92.0-devpre20241014_326cbec
$ nix eval --inputs-from . --raw lix-module
warning: Git tree '/tmp/wat/My_Config_Flake-main.bundle_out' is dirty
/nix/store/5c585xp03rh11fb542q3xv64xyxx52r2-lix-2.92.0-dev-pre20240930-a16ceb9
Did this as a quick test, and they seem to throw different paths
| 21:42:13 |
@jade_:matrix.org | In reply to @kfears:matrix.org
Misc bug (?) on NixOS:
$ whereis nix
nix: /etc/nix /nix/store/ry4i0aqv98yabkdlq8rd1nqa77xyz19m-system-path/bin/nix
I don't think /etc/nix belong here, like, at all
i dont think this is a bug | 21:42:35 |
@jade_:matrix.org | it is searching in /etc for things called nix | 21:42:44 |
@jade_:matrix.org | if you are trying to ask the actual question "where is my nix binary", you want command -v nix , which nix , type -p nix | 21:43:27 |
KFears (tragedy arc) | So /etc/nix is included because it's called "nix" and it's added to PATH but it's also a directory so it's skipped when you run nix command? | 21:43:38 |
Whovian9369 | I've personally been fond of type -s nix for getting the symlink in PATH and the actual executable. | 21:44:20 |
@jade_:matrix.org | In reply to @kfears:matrix.org So /etc/nix is included because it's called "nix" and it's added to PATH but it's also a directory so it's skipped when you run nix command? ยป whereis hosts
hosts: /etc/hosts /usr/share/man/man5/hosts.5.gz
| 21:45:26 |
@jade_:matrix.org | it's included because whereis does not return you just executables | 21:45:46 |
Whovian9369 | (Gonna gently poke Jade here quickly!) | 22:11:44 |
@jade_:matrix.org | what, about the flakes? i am currently focused on other things. can you file a bug report if you can diagnose it with a smaller example to return unexpected output? no promises it will get fixed but it would be good to have the defect documented. | 22:23:59 |
@jade_:matrix.org | is there a reasonable procedure to debug issues like this one?
error: cycle detected in build of '/nix/store/aksg7yap79slndh0s0bx5vr422vbzcr1-capnproto-1.0.2.drv' in the references of output 'bin' from output 'dev'
so far as i can tell, the output in question is deleted from the disk, which is highly impolite of the lix daemon. keep-failed will keep the build directory but not the output directory. i did file a bug that the error sucks, but we surely also need a way of keeping the bad outputs somewhere? https://git.lix.systems/lix-project/lix/issues/551
| 22:31:41 |
Whovian9369 | jade - I'll take a shot at it, thanks for the time and sorry to interrupt then! | 22:32:17 |
@jade_:matrix.org | okay, keep-failed will only keep around bad outputs if they are non-reproducible, but not in these other instances. that's so unhelpful | 22:36:08 |
KFears (tragedy arc) | This is a relatively small thing, but thank you for how fast Lix rebuilds! | 22:38:33 |
| @enzime:nixos.dev left the room. | 23:59:18 |
16 Oct 2024 |
SigmaSquadron | In reply to @jade_:matrix.org
is there a reasonable procedure to debug issues like this one?
error: cycle detected in build of '/nix/store/aksg7yap79slndh0s0bx5vr422vbzcr1-capnproto-1.0.2.drv' in the references of output 'bin' from output 'dev'
so far as i can tell, the output in question is deleted from the disk, which is highly impolite of the lix daemon. keep-failed will keep the build directory but not the output directory. i did file a bug that the error sucks, but we surely also need a way of keeping the bad outputs somewhere? https://git.lix.systems/lix-project/lix/issues/551
If it's nixpkgs, add breakpointHook to nativeBuildInputs and put a non-zero exit command in the postFixup . You can then use cntr to inspect the outputs.
| 13:09:33 |
hexa | hm, I noticed that ctrl+c'ing a remote build locally does not reliably stop the build on the remote host | 14:04:24 |
hexa | which results in me waiting for a lack the next time I start the build | 14:04:51 |
hexa | is that a behavior anyone else noticed as well? | 14:05:04 |
K900 | Yeah, something got funny with it at some point | 14:07:06 |
hexa | ok, filing a bug | 14:07:41 |
hexa | https://git.lix.systems/lix-project/lix/issues/552 if you want to subscribe | 14:10:54 |
daagaak | Is there a quick way to tell whether lix thinks it can substitute a given derivation? One of my machines suddenly thinks it wants to rebuild all of my homelab from scratch but my local attic cache seems to have it cached. I can't work out why that one machine wants to rebuild everything. | 14:22:38 |
Linux Hackerman | daagaak: you can try nix path-info --store <cache URL> /nix/store/... | 14:50:50 |
Linux Hackerman | oh wait that doesn't really help | 14:51:17 |
Linux Hackerman | but maybe you're running into the negative caching, try adding --narinfo-cache-negative-ttl 0 to your command | 14:51:33 |
KFears (tragedy arc) | Hmm, for lix-installer, any suggestions how to cure existing Nix installs? | 14:59:16 |
KFears (tragedy arc) | * Hmm, for lix-installer, any suggestions how to cure existing CppNix installs? | 14:59:21 |
KFears (tragedy arc) | I'm looking at fixing the test cases, and because Lix is installed using nix-env (gross hack), there's an issue with getting conflicts in the profile between Lix and CppNix | 15:00:03 |