All the weird stuff goes here #

This is basically a monorepo where I dump everything that can be made public.

Structure #

• lib: autopilot evaluation entry point, Nix related helper functions
• modules: all modules that will be picked up by autopilot, then loaded with flake-parts
• pkgs: packages from nixpkgs that are globally overridden, my own stuff, it's a mess...

Module portability #

For all modules exposed under ./modules, importApplyWithArgs is used to optionally apply a two level argument. For example, if the file importing is in ./modules/nixos, we can have:

{ lib, inputs, ... } @ arg0:

{ config, lib, modulesPath, options, pkgs, ... } @ arg1:

{ ... } # actual content

Where arg0 is from ./modules/flake/modules/default.nix, includes the lib from builtins // nixpkgs.lib, and with all locally defined helper functions under ./lib and extensions declared in autopilot, and is not compatible with the lib in arg1.

All variables from arg1 are standard NixOS module arguments, but if lib is needed, it must be pulled in from arg0 (precedence), or pkgs.lib:

{ config, lib, ... }: # <-

lib.deepMergeAttrsList ... # pull something from `config`

Will resulted in an error (config not found), and

{ lib, ... }: # <-

{ config, ... }:

lib.deepMergeAttrsList ... # pull something from `config`

Must be used. But if the following is allowed since the application of arg0 is optional based on the names inside:

{ config, ... }:

{ ... } # do something dependent on `config`

Note that this behavior only apply one level deep, i.e. only modules imported using modulesFor will get automatic argument injection. If arg0 is needed in nested modules or files imported from top-level module:

arg0:

{ config, ... }:

{
  imports = [
    (import ./some-module.nix arg0) # `arg0` is not optionally here, the second level arg is the regular module arg
  ];
}