nix-diff seems to prepend 'daemon' before paths to the nix store

[dwt]

Hi there,

while experimenting in the last few days nix-diff started behaving strangely, and started outputting errors like this:

❯ nix-diff /run/current-system /nix/store/l2h9ciy4ynxp1vklgvqaizzmfp52pff7-darwin-system-25.05.349a74c --color always 
nix-diff: daemon/nix/store/bqh2p1bc7kbswp0ip3agibm3irll984y-darwin-system-25.05.349a74c.drv: withBinaryFile: does not exist (No such file or directory)
❯ ll /nix/store/bqh2p1bc7kbswp0ip3agibm3irll984y-darwin-system-25.05.349a74c.drv
Permissions Size User Date Modified Name
.r--r--r--   38k root  1 Jan  1970  /nix/store/bqh2p1bc7kbswp0ip3agibm3irll984y-darwin-system-25.05.349a74c.drv

Is this something I have configured wrong and need to change or is this somehow a bug in nix-diff that it prepends daemon to that path before trying to read it?

[dwt]

I did install nix-diff with nix-darwin, some details:

❯ nix flake metadata
warning: Git tree '/Users/dwt/Code/Projekte/nix/nix-darwin-config' is dirty
Resolved URL:  git+file:///Users/dwt/Code/Projekte/nix/nix-darwin-config
Locked URL:    git+file:///Users/dwt/Code/Projekte/nix/nix-darwin-config
Description:   mkk nix-darwin system flake
Path:          /nix/store/86chrs2za0z91hwnyrsgp01j7fqpdjwn-source
Revision:      14bfb38868eaac4e22d028a89db220eeac00bdab-dirty
Last modified: 2025-01-29 20:17:58
Inputs:
├───lix-module: https://git.lix.systems/api/v1/repos/lix-project/nixos-module/archive/b90bf629bbd835e61f1317b99e12f8c831017006.tar.gz?narHash=sha256-YMLrcBpf0TR5r/eaqm8lxzFPap2TxCor0ZGcK3a7%2Bb8%3D&rev=b90bf629bbd835e61f1317b99e12f8c831017006
│   Last modified: 2025-01-18 22:58:14
│   ├───flake-utils: github:numtide/flake-utils/11707dc2f618dd54ca8739b309ec4fc024de578b
│   │   Last modified: 2024-11-13 22:27:16
│   │   └───systems: github:nix-systems/default/da67096a3b9bf56a91d16901293e51ba5b49a27e
│   │       Last modified: 2023-04-09 10:27:08
│   ├───flakey-profile: github:lf-/flakey-profile/243c903fd8eadc0f63d205665a92d4df91d42d9d
│   │   Last modified: 2024-04-12 07:09:50
│   ├───lix: https://git.lix.systems/api/v1/repos/lix-project/lix/archive/2837da71ec1588c1187d2e554719b15904a46c8b.tar.gz?narHash=sha256-CCKIAE84dzkrnlxJCKFyffAxP3yfsOAbdvydUGqq24g%3D&rev=2837da71ec1588c1187d2e554719b15904a46c8b
│   │   Last modified: 2025-01-18 22:04:46
│   └───nixpkgs follows input 'nixpkgs'
├───nix-darwin: github:LnL7/nix-darwin/349a74c66c596ef97ee97b4d80a3ca61227b6120
│   Last modified: 2025-01-28 03:58:58
│   └───nixpkgs follows input 'nixpkgs'
├───nix-rosetta-builder: github:cpick/nix-rosetta-builder/b32bd2c1915481d79e3c9d1e222cbd3a9642c016
│   Last modified: 2025-01-03 06:37:31
│   ├───nixos-generators: github:nix-community/nixos-generators/f5a0197ccfef7388885fc9455e74d6dd39e0c5e8
│   │   Last modified: 2024-12-09 02:09:07
│   │   ├───nixlib: github:nix-community/nixpkgs.lib/f4dc9a6c02e5e14d91d158522f69f6ab4194eb5b
│   │   │   Last modified: 2024-12-08 02:08:11
│   │   └───nixpkgs follows input 'nix-rosetta-builder/nixpkgs'
│   └───nixpkgs follows input 'nixpkgs'
└───nixpkgs: github:NixOS/nixpkgs/f898cbfddfab52593da301a397a17d0af801bbc3
    Last modified: 2025-01-27 21:31:03

I am invoking it via this script

#!/bin/sh

set -e
nix run nix-darwin -- build --flake . "$@"
computerName=$(networksetup -getcomputername)
outPath=$(nix eval --raw .#darwinConfigurations.$computerName.config.system.build.toplevel.outPath)

# currently fails https://github.com/Gabriella439/nix-diff/issues/98
nix run nixpkgs#nix-diff -- /run/current-system $outPath --color always

Is that somehow the wrong way to do this?

[dwt]

I have since switched back from unstable to stable

    nixpkgs.url = "github:NixOS/nixpkgs/nixpkgs-24.11-darwin";
    nix-darwin.url = "github:LnL7/nix-darwin/nix-darwin-24.11";

Which seems to have fixed this. Therefore there is likely a bug in the version that is part of unstable currently.

Sadly my Haskell skills are non existent, so I have no clue how to further debug this.

However, if there is any more information or experiments I can provide, I would be glad to do so, as this tool is super helpful.

[dwt]

I do not know how to do this, but it seems likely that a git bisect should be able to find the faulty commit.

If you could give me some assistance how to do that, I would gladly do that experiment.

[Sporif]

Probably caused by f31e0c1, as unsetting NIX_REMOTE fixes the issue for me.

[dwt]

Does that mean you know how to fix it? Or can I provide any additional information?