Use nmt and nmd from Nixpkgs

[rycee]

Description

Fixes #4879

Depends on NixOS/nixpkgs#280467 being in NixOS unstable.

Checklist

• Change is backwards compatible.

• Code formatted with ./format.

• Code tested through nix-shell --pure tests -A run.all or nix develop --ignore-environment .#all using Flakes.

• Test cases updated/added. See example.

• Commit messages are formatted like

  {component}: {description}
  
  {long description}
  

[oxalica]

This makes it IFD again, isn't it? Should I open a new issue?

error:
       … while calling the 'head' builtin

         at /nix/store/b0l3v8509hrl3qbv6fj2zl0mjb2rwhdy-source/lib/attrsets.nix:960:11:

          959|         || pred here (elemAt values 1) (head values) then
          960|           head values
             |           ^
          961|         else

       … while evaluating the attribute 'value'

         at /nix/store/b0l3v8509hrl3qbv6fj2zl0mjb2rwhdy-source/lib/modules.nix:809:9:

          808|     in warnDeprecation opt //
          809|       { value = builtins.addErrorContext "while evaluating the option `${showOption loc}':" value;
             |         ^
          810|         inherit (res.defsFinal') highestPrio;

       (stack trace truncated; use '--show-trace' to show the full trace)

       error: cannot build '/nix/store/yachqrw4zahp0b2qg9d8gz5kl59xr809-nix-lib-nmd-0.5.0.drv^out' during evaluation because the option 'allow-import-from-derivation' is disabled

[rycee]

@oxalica Hmm, not sure. It works fine without --impure so I would have assumed there is no problem. In any case PRs are welcome to improve the situation.

[rycee]

Would it work if the packages were fixed-output derivations? It should be possible to fix their output since they don't really have any dependencies.

[rycee]

See NixOS/nixpkgs#281111. Hopefully that would fix this issue.

[iphydf]

https://discourse.nixos.org/t/is-nix-lib-nmd-missing-from-23-11/38268

In my dev container, to reduce disk space, I use the 23.11 tag, not the updated channel. This means that now home-manager has to be pinned at the commit before this one. I'd say it's a little surprising that "release-23.11" is not compatible with nixos tag 23.11. I tend to think we should be backwards compatible at least for half a year (one release cadence).

[rycee]

@iphydf I definitely would suggest not pinning to the 23.11 tag, it will never be updated, even for security reasons. I would instead suggest pinning to a suitable commit hash along the history of the 23.11 channel and occasionally consider updating the pin.

That said, if there is a big demand for it, then we could revert the HM 23.11 branch back to using the Sourcehut source since it is available again.

[lovesegfault]

I'm still seeing IFD errors, even with the latest nixos-unstable which includes NixOS/nixpkgs#281111:

github:lovesegfault/nix-config/eec1fe759743d9ef3508f509c82deddf418bd316
├───checks
│   ├───aarch64-darwin
│   │   └───pre-commit-check omitted (use '--all-systems' to show)
│   ├───aarch64-linux
│   │   └───pre-commit-check omitted (use '--all-systems' to show)
│   ├───x86_64-darwin
│   │   └───pre-commit-check omitted (use '--all-systems' to show)
│   └───x86_64-linux
│       └───pre-commit-check: derivation 'pre-commit-run'
├───darwinConfigurations: unknown
├───deploy: unknown
├───devShells
│   ├───aarch64-darwin
│   │   └───default omitted (use '--all-systems' to show)
│   ├───aarch64-linux
│   │   └───default omitted (use '--all-systems' to show)
│   ├───x86_64-darwin
│   │   └───default omitted (use '--all-systems' to show)
│   └───x86_64-linux
│       └───default: development environment 'nix-config'
├───homeConfigurations: unknown
├───hosts: unknown
├───nixosConfigurations
│   ├───aurelius: NixOS configuration
│   ├───bohr: NixOS configuration
│   ├───fourier: NixOS configuration
│   ├───hegel: NixOS configuration
│   ├───jung: NixOS configuration
│   ├───nozick: NixOS configuration
│   ├───riemann: NixOS configuration
│   └───spinoza: NixOS configuration
├───overlays
│   ├───default: Nixpkgs overlay
│   ├───scripts: Nixpkgs overlay
│   ├───sway-scene-graph: Nixpkgs overlay
│   ├───truecolor-check: Nixpkgs overlay
│   ├───uboot-cm4: Nixpkgs overlay
│   └───vim-plugins: Nixpkgs overlay
├───packages
│   ├───aarch64-darwin
│   │   ├───cachix omitted (use '--all-systems' to show)
│   │   ├───default omitted (use '--all-systems' to show)
│   │   ├───nix-eval-jobs omitted (use '--all-systems' to show)
│   │   ├───nix-fast-build omitted (use '--all-systems' to show)
│   │   └───poincare omitted (use '--all-systems' to show)
│   ├───aarch64-linux
│   │   ├───aurelius omitted (use '--all-systems' to show)
│   │   ├───cachix omitted (use '--all-systems' to show)
│   │   ├───default omitted (use '--all-systems' to show)
│   │   ├───luther omitted (use '--all-systems' to show)
│   │   ├───nix-eval-jobs omitted (use '--all-systems' to show)
│   │   ├───nix-fast-build omitted (use '--all-systems' to show)
│   │   └───riemann omitted (use '--all-systems' to show)
│   ├───x86_64-darwin
│   │   ├───cachix omitted (use '--all-systems' to show)
│   │   ├───nix-eval-jobs omitted (use '--all-systems' to show)
│   │   └───nix-fast-build omitted (use '--all-systems' to show)
│   └───x86_64-linux
error:
       … while calling the 'head' builtin

         at /nix/store/wcidiyklj0nrljlz5m3qlkvhv8f2ddv8-source/lib/attrsets.nix:960:11:

          959|         || pred here (elemAt values 1) (head values) then
          960|           head values
             |           ^
          961|         else

       … while evaluating the attribute 'value'

         at /nix/store/wcidiyklj0nrljlz5m3qlkvhv8f2ddv8-source/lib/modules.nix:809:9:

          808|     in warnDeprecation opt //
          809|       { value = builtins.addErrorContext "while evaluating the option `${showOption loc}':" value;
             |         ^
          810|         inherit (res.defsFinal') highestPrio;

       (stack trace truncated; use '--show-trace' to show the full trace)

       error: cannot build '/nix/store/flv4m459a0wml9377yr5ckyrhc1d37sx-nix-lib-nmd-0.5.0.drv^out' during evaluation because the option 'allow-import-from-derivation' is disabled

Curiously, this only happens when I eval on my x86_64-linux and aarch64-linux machines, on aarch64-darwin it works AOK.

[iphydf]

I'll pin the commit hash. It's not super important to be up to date because it's only used for integration tests and our local dev environment, not for releases, but I think using commit hashes will work for us.

[rycee]

Ok, I've opened the above PR to restore the use of fetchTarball for nmd and nmt. Since Sourcehut is available again, I figure it should be safe to switch back.

[oxalica]

Ok, I've opened the above PR to restore the use of fetchTarball for nmd and nmt. Since Sourcehut is available again, I figure it should be safe to switch back.

I've seen the regression several times. Would it be better to add a CI task to build with --option allow-import-from-derivation false so we can catch it early?