nixos/tests/misc: support old and new path-info output structure

[mweinelt]

Description of changes

#315262 (comment)

Things done

• Built on platform(s)
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• For non-Linux: Is sandboxing enabled in nix.conf? (See Nix manual)
  • sandbox = relaxed
  • sandbox = true
• 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/)
• 24.11 Release Notes (or backporting 23.11 and 24.05 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
• Fits CONTRIBUTING.md.

---

Add a 👍 reaction to pull requests you find important.

[fricklerhandwerk]

@nixos/nix-team Why in the world is Nixpkgs even testing against Nix experimental commands? That's exactly how you make them unchangeable or annoy everyone with breakages.

Looking at commit history, this has been added before experimental features were a thing. I suggest removing that test and only adding it back when the command is stabilized.

[K900]

Going to untag as channel-blocker as the Nix bump was reverted.

[Mic92]

@NixOS/nix-team Why in the world is Nixpkgs even testing against Nix experimental commands? That's exactly how you make them unchangeable or annoy everyone with breakages.

Looking at commit history, this has been added before experimental features were a thing. I suggest removing that test and only adding it back when the command is stabilized.

I think we are beyond a point where we can no longer reasonable treat these commands as experimental. Even if it doesn't break in nixpkgs, it will still break other users, since these commands have been cooking for too long in an "experimental" state.

[Mic92]

@mweinelt thoughts about: #315878 ?

[mweinelt]

If it broke, then yes, passthru.tests seems appropriate. Naming the test misc 🤷

[Mic92]

Yeah. Terrible name. I will come up with something else.

[lilyinstarlight]

I think we are beyond a point where we can no longer reasonable treat these commands as experimental. Even if it doesn't break in nixpkgs, it will still break other users, since these commands have been cooking for too long in an "experimental" state.

This command has been present since Nix 2.0 and was added to the NixOS misc tests during the Nix 2.0 bump PR (https://github.com/NixOS/nixpkgs/pull/34636/files#diff-5eb2258233433ba8710976a266f30f16a70d92c633d36706072a095f91dc50a9R38)

I'm just providing context, though, so I won't make any comment either way about whether it's appropriate the command was later made "experimental" when that ability was added (like the nix repl command) or whether it should be removed from the test until the command is "stabilized"

[Mic92]

Tested that this, fixes the nix upgrade. I currently don't see a good replacement for this command in the old cli interface. If we make sure that we run this NixOS test on every nix version bump, than we can also fix it.

[Ericson2314]

I will say, I mean for things to be experimental because I intend to make breaking changes. It's quite simple. I caused this breaking change, and I have further changes to the format in NixOS/nix#9995, and probably more after that. Per the issue through, the JSON guidelines need to be further fleshed out, and this (and other interfaces) need to be brought into compliance.

The unstable interface of Nix grew much faster than our ability to maintain it, or make it adequate quality. It's a hard problem to solve without a more developer hours.

[Ericson2314]

I currently don't see a good replacement for this command in the old cli interface.

There is a mishmash of nix-store commands to query individual fields of the path info JSON. Agreed that is much less nice to use!

If we make sure that we run this NixOS test on every nix version bump, than we can also fix it.

That's true. If we the tests make clear they are not locking down features users should rely on, but just sanity checking Nixpkgs' own internal usage of experimental stuff, it is less bad.