Does not work well with custom options from flakes and overlays that don't evaluate but are unused

[SuperSandro2000]

$ RUST_BACKTRACE=1 ./target/release/nix-olde -n ~/src/nixpkgs
Fetching 'installed' ...
Fetching 'available' ...
Fetching 'repology' ...
ERROR: command ["nix-instantiate", "<nixpkgs/nixos>", "-A", "system", "-I", "nixpkgs=/home/sandro/src/nixpkgs"] failed with status: ExitStatus(unix_wait_status(256))
  stdout: ""
  stderr: "error: The option `nix.deleteChannels' d"
thread '<unnamed>' panicked at 'assertion failed: cmd.status.success()', src/main.rs:27:5
stack backtrace:
   0: rust_begin_unwind
   1: core::panicking::panic_fmt
   2: core::panicking::panic
   3: nix_olde::run_cmd
   4: nix_olde::get_local_installed_packages
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.
ERROR: command ["nix-env", "-qa", "--json", "--arg", "config", "import <nixpkgs/pkgs/top-level/packages-config.nix>", "--option", "build-users-group", "\"\"", "-f", "/home/sandro/src/nixpkgs"] failed with status: ExitStatus(unix_wait_status(256))
  stdout: ""
  stderr: "error: attribute 'nextcloud26' missing\n\n"
thread '<unnamed>' panicked at 'assertion failed: cmd.status.success()', src/main.rs:27:5
stack backtrace:
   0: rust_begin_unwind
   1: core::panicking::panic_fmt
   2: core::panicking::panic
   3: nix_olde::run_cmd
   4: nix_olde::get_local_available_packages
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.
... 'repology' done.
thread 'main' panicked at 'a scoped thread panicked', src/main.rs:262:8
stack backtrace:
   0: rust_begin_unwind
   1: core::panicking::panic_fmt
   2: std::thread::scoped::scope
   3: nix_olde::main
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

nextcloud26 is an entry in my overlays that does not exist yet. It would be great if such entries could be skipped.
nix.deleteChannels is an option that only exists in a flake input and then is added to the module list in the flake.nix. Can nix-instantiate be made aware of this?

[trofi]

Good question. In the end we just need a .drv file that defines the whole system.

I never tried to set the system up with flakes. Looking at pkgs/os-specific/linux/nixos-rebuild/nixos-rebuild.sh there is a chance to use nix eval to get it. I'll try to set up a flake-based system today and get it to do something sensible.

nix-env -qa failure looks more worrying. It should already skip the overlays via config.

[SuperSandro2000]

You could get the drv with nix eval /etc/nixos#nixosConfigurations.$HOSTNAME.config.system.build.toplevel.drvPath.

[SuperSandro2000]

pkgs/top-level/packages-config.nix is importing config but pkgs/top-level/impure.nix is also taking overlays as another input. Adding --arg overlays "[]" to the nix-env -qa fixes it and streams quite some json.

[trofi]

Converted my system to flake configuration with impurities and removed nixos channel to break compatibility with nix-env.

I wonder if 96cade0 already makes it work for you.

Ideally packages from the overlays should be pullable as is as well (they will just fail to be found on repology, it's ok). I'll try to find flake equivalent for nix-env -qa, but meanwhile i slightly unbroke it with 533ec02.

I think the below has highest chance to succeed as it always forces nixpkgs= override whether it is present or not in default channels:

./mkrun.bash -n https://github.com/NixOS/nixpkgs/archive/refs/heads/staging.tar.gz

[trofi]

With 83cfa55 I can how run ./mkrun.bash without extra parameters to get some output on flake-based system. I still use nix-env and not overriding overlays.

Can you give it a try?

[trofi]

Ah, I now realized you have explicitly broken overlay and it has no way to evaluate at all, not just due to different contexts (flake vs configuration vs nixpkgs). Reproduced failure locally by adding:

--- a/overlays/local-packages.nix
+++ b/overlays/local-packages.nix
@@ -32,4 +32,6 @@ final: prev: {

   xmms2 = final.callPackage ../xmms2 {};
   xmms2_0_8 = final.callPackage ../xmms2/0.8 {};
+
+  broken_package = final.callPackage does/not/exist {};
 }

Fails as

$ ./mkrun.bash
Fetching 'installed' ...
Fetching 'repology' ...
Fetching 'available' ...
... 'available' failed.
... 'repology' failed.
... 'installed' done.
Error: MultipleErrors([Canceled("Repology fetch"), CommandFailed { cmd: ["nix-env", "-qa", "--json", "--arg", "config", "import <nixpkgs/pkgs/top-level/packages-config.nix>", "--option", "build-users-group", "\"\"", "-I", "nixpkgs=/nix/store/2z54qg6klpnp6l2gagppzwj9pinadxvp-source", "-f", "/nix/store/2z54qg6klpnp6l2gagppzwj9pinadxvp-source"], output: Output { status: ExitStatus(unix_wait_status(256)), stdout: "", stderr: "error: getting status of '/home/slyfox/.config/nixpkgs/overlays/does/not/exist': No such file or directory\n" } }])

Looking.

We will probably need to get rid of nix-env and have an equivalent code to traverse attrsets to map them to pname.

[SuperSandro2000]

Yes, I didn't describe that as clear.

I removed the overlays form ~/.config/nixpkgs for now and got some results.

[trofi]

I tried to filter problematic attributes with builtins.tryEval and failed: tryEval does not catch No such file or directory and can only handle throw and assert-based well behaved failures.

Let's declare working pkgs as a requirement and declare the current state good enough for flakes based systems.