I'm not sure if it makes sense to run a separate eval for each feature like this. I can't really think of any issues with just running a single strict eval to check for issues.

@LnL7 better?

Yeah, but maybe this should be a config option in nixpkgs. That way it's a bit less obscure to reproduce.

@LnL7 I don't want people to avoid extend when they actually need it, so, to be honest, I think this check is not for users, and therefore I don't think it should be an option. If someone really wants this check in their own tools, it's not too complicated to reproduce.

If someone is motivated to make it an option, I can review their PR.

cc @Ericson2314

Having the error message inline in the command just feels a bit weird to me, but yeah it's not that complicated.

@grahamc What do you think?

I'm sorry to be annoying and not provide more actionable feedback sooner.

This test should be applied to the evaluation which happens before and after a PR is merged. Please move the configuration to https://github.com/NixOS/ofborg/blob/ae982a2a683dfb5062ef8fb1b03bb7c3804b7e31/ofborg/src/outpaths.nix and delete the changes to the massrebuilder.rs.

Aliases are tested this way because aliases are desirable for external users, but can easily rot unless they are checked regularly. This is why the check is performed once with, and once without aliases.

I don't believe extends will ever rot like aliases can, so no need to perform the different checks.

I'm cleaning up my open PR list. After almost four years, this has proven itself to be unimportant in many ways. If anyone disagrees, feel free to modify and use the code in any way you like.