nixos/modules/flake/source-info: new module

[herberteuler]

Description of changes

This PR introduces a new module that improves configurationRevision handling as discussed in #178837.

This option can be set as in the following example:

{
  inputs.nixpkgs.url = ...;
  outputs = { self, nixpkgs }: {
    nixosConfigurations = {
      host1 = nixpkgs.lib.nixosSystem {
        system = ...;
        modules = [
          { system.nixos.configuration.sourceInfo = self.sourceInfo; }
        ] ++ ...;
      };
    };
  };
}

See more details in the module description.

Things done

• Built on platform(s)
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• For non-Linux: Is sandbox = true set in nix.conf? (See Nix manual)
• Tested, as applicable:
  • NixOS test(s) (look inside nixos/tests)
  • and/or package tests
  • or, for functions and "core" functionality, tests in lib/tests or pkgs/test
  • made sure NixOS tests are linked to the relevant packages
• Tested compilation of all packages that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review rev HEAD". Note: all changes have to be committed, also see nixpkgs-review usage
• Tested basic functionality of all binary files (usually in ./result/bin/)
• 22.11 Release Notes (or backporting 22.05 Release notes)
  • (Package updates) Added a release notes entry if the change is major or breaking
  • (Module updates) Added a release notes entry if the change is significant
  • (Module addition) Added a release notes entry if adding a new NixOS module
  • (Release notes changes) Ran nixos/doc/manual/md-to-db.sh to update generated release notes
• Fits CONTRIBUTING.md.

[roberth]

Can we set it to "dirty" when we see that sourceInfo is from git but does not have rev?
We don't have a type-like attribute in there (yet?), but the presence of submodules should give it away.

[roberth]

NixOS/nix#6747

[herberteuler]

The intention of setting it to dirty, to my understanding, is to help identify the state of the configuration better. To answer this question well, I checked the relevant code, and have two proposals:

1. We can do it for all supported VCSes, not just git. For instance, mercurial is also supported (albeit immature). This is partly due to only git getting special handling while examining local paths, and we may improve that too to enable support for other VCSes. This also implies a standardized, fine-grained specification of source information, as will be discussed in sourceInfo does not tell us whether it's git or not nix#6747.

2. To further tell differences among dirty configurations, we can fetch the latest update time of a repository and set something like dirty-<timestamp>. The latest update time can be defined as the most recent modification time of all files in the repository.

[herberteuler]

@roberth Do we solve this later as a part of NixOS/rfcs#125?

[roberth]

We can scope this discussion out regardless.

[roberth]

host1 = nixpkgs.lib.nixosSystem {
system = ...;

Plug: system shouldn't be necessary when specified in a module or in hardware-configuration.nix (merged).

[herberteuler]

host1 = nixpkgs.lib.nixosSystem {
system = ...;

Plug: system shouldn't be necessary when specified in a module or in hardware-configuration.nix (merged).

I am using nixos-22.05, perhaps I should switch to unstable...

[roberth]

We could avoid storing the rev in the derivation output and make the non-derivation parts of NixOS responsible for propagating this information. NixOS/rfcs#125 (comment)

[roberth]

You could do this and then write markdown if you like

Suggested change

	description = ''
	description = lib.mdDoc ''

[herberteuler]

Updated. The example also gets improved.

[roberth]

Doesn't it also occur in the grub and systemd-boot boot menus?

[herberteuler]

It does not. Currently, the menus contain only the shortRev of nixpkgs. Should we add the configuration revision to versionSuffix too? If both revisions are to be included, they should be made distinctive.

Note that the nixpkgs revision can be dirty too; we may develop a uniform way to obtain revisions, and apply it to both occasions. What do you think about my earlier proposals on setting the dirty tag? I can start to work on a prototype after we discuss and reach an agreement.

[roberth]

It does not.

Ok. That can be done in another PR if desirable.

we may develop a uniform way to obtain revisions

Sounds good, but these are all quite independent problems. It's best to resolve them in separate PRs, so that maintainers don't need to keep (in the limit) all of NixOS in their head in order to approve them.

[herberteuler]

Sure, there will be at least three PRs: this one, one for obtaining revisions, and another one for applying the feature. I will try to write a detailed design later.

[roberth]

Suggested change

	re-deployments if the flake repository changes but the
	boot loader entries or re-deployments if the flake repository changes but the

[roberth]

Implementation lgtm. I think we'll need to create a section in the NixOS manual as well. There currently isn't one, but that's going to be a problem when flakes are released. If we're going to write those docs last minute, they'll be incomplete, and probably skip over this feature.

[herberteuler]

Implementation lgtm. I think we'll need to create a section in the NixOS manual as well. There currently isn't one, but that's going to be a problem when flakes are released. If we're going to write those docs last minute, they'll be incomplete, and probably skip over this feature.

I just checked that the missing docs were not only flakes for NixOS but also flakes for Nix. But, unfortunately, writing a whole bunch of such docs is out of my reach right now. Hence I added a section for this feature but did not integrate it into the current NixOS manual. This way, someone will notice that it is there in the future.