Fix nix why-depends --derivation

[thufschmitt]

Reverts #7337 and re-implements it in a different way in order to not break nix why-depends --derivation.

Fix #7496

Cc @raphaelr

[Radvendii]

• Using Realise::Derivation here tells it to only realise the derivation, and not go further (i.e. it won't try to resolve the output if it hasn't already been resolved).
• operateOn tells it to return the store path of the derivation if invoked with --derivation.

Is that right?

[raphaelr]

Can confirm that this fixes the issue:

$ nix run 'github:NixOS/nix/pull/7539/head' -- why-depends --derivation -f . libsForQt5.qtbase nasm
/nix/store/6mr2fs8z9arz8wfc357szvcnz3vvmxv4-qtbase-5.15.7.drv
└───/nix/store/9kzlffpxzj7zf3fmz637y5glkqnawj99-libjpeg-turbo-2.1.4.drv
    └───/nix/store/wpd81rvxfd3lc4qf0l5dls5fq3p0vdny-nasm-2.15.05.drv

[thufschmitt]

• Using Realise::Derivation here tells it to only realise the derivation, and not go further (i.e. it won't try to resolve the output if it hasn't already been resolved).

  • operateOn tells it to return the store path of the derivation if invoked with --derivation.

Is that right?

Not exactly: Realise::Derivation is some kind of dry-run mode. Unless you also pass OperateOn::Derivation, it will still try to return the output path (despite not building it). So it fails if the derivation is an unbuilt CA derivation (hence the try-catch block).

[nixos-discourse]

This pull request has been mentioned on NixOS Discourse. There might be relevant details there:

https://discourse.nixos.org/t/tweag-nix-dev-update-43/25185/1