diff options
Diffstat (limited to 'pkgs/test/nixpkgs-check-by-name/src/eval.rs')
| -rw-r--r-- | pkgs/test/nixpkgs-check-by-name/src/eval.rs | 17 |
1 files changed, 13 insertions, 4 deletions
diff --git a/pkgs/test/nixpkgs-check-by-name/src/eval.rs b/pkgs/test/nixpkgs-check-by-name/src/eval.rs index b411a2a3c601..c4dad8257cf9 100644 --- a/pkgs/test/nixpkgs-check-by-name/src/eval.rs +++ b/pkgs/test/nixpkgs-check-by-name/src/eval.rs @@ -168,14 +168,23 @@ pub fn check_values( // so the UsesByName ratchet is already as tight as it can be NonAttributeSet => Success(Tight), NonCallPackage => Success(Tight), - // This is an odd case when _internalCallByNamePackageFile is used to define a package. + // This is the case when the `pkgs/by-name`-internal _internalCallByNamePackageFile + // is used for a package outside `pkgs/by-name` CallPackage(CallPackageInfo { call_package_variant: Auto, .. - }) => NixpkgsProblem::InternalCallPackageUsed { - attr_name: attribute_name.clone(), + }) => { + // With the current detection mechanism, this also triggers for aliases + // to pkgs/by-name packages, and there's no good method of + // distinguishing alias vs non-alias. + // Using `config.allowAliases = false` at least currently doesn't work + // because there's nothing preventing people from defining aliases that + // are present even with that disabled. + // In the future we could kind of abuse this behavior to have better + // enforcement of conditional aliases, but for now we just need to not + // give an error. + Success(Tight) } - .into(), // Only derivations can be in pkgs/by-name, // so this attribute doesn't qualify CallPackage(CallPackageInfo { |