fix(NixOS): check nix paths for binaries in udev hiding rules

[ShadowApex]

Since adding the devnode moving logic, inputplumber would fail in finding the correct binary paths for chmod and mv. This change fixes that by checking for the binaries in NixOS paths. Otherwise it will try to use the which built-in shell command to discover the path.

WARN  inputplumber::input::composite_device] Failed to hide device '/dev/input/event4': Os { code: 2, kind: NotFound }

[github-actions]

🎉 This PR is included in version 0.75.1 🎉

The release is available on:

• v0.75.1
• GitHub release

Your semantic-release bot 📦🚀

[Pandapip1]

Wouldn't the better approach be for InputPlumber itself to traverse through the path everywhere instead of using some hardcoded paths and if that fails shelling out? This should be relatively trivial to implement ourselves, or there's this crate if you're okay with adding another dependency.

[Pandapip1]

Also, completely separately: why is the chmod command needed in the udev rule if we're setting MODE="000" anyway? I'm sure there's a good reason, but I'd appreciate a comment explaining that.

Thank you both on your work on InputPlumber and making sure it works on NixOS!

[ShadowApex]

@Pandapip1

Wouldn't the better approach be for InputPlumber itself to traverse through the path everywhere instead of using some hardcoded paths and if that fails shelling out? This should be relatively trivial to implement ourselves, or there's this crate if you're okay with adding another dependency.

You're absolutely right, that is a much better solution 😄 . Since this is relatively trivial to implement (parse PATH and check if binary exists at each path) I would much rather just implement this logic ourselves instead of adding a new dependency. If you would like to add this an open a PR for it, we will accept it. Otherwise I will try and add this myself in the next few days. Thanks!

Also, completely separately: why is the chmod command needed in the udev rule if we're setting MODE="000" anyway? I'm sure there's a good reason, but I'd appreciate a comment explaining that.

Thank you both on your work on InputPlumber and making sure it works on NixOS!

Thanks for your comments! In our testing, we found that just setting MODE="000" was not sufficient for actually setting the permissions on the device for reasons we're still not fully sure about. We suspect that other rules are overriding the mode.