The glob_include?(site.include, entry) → glob_include?(site.include, relative_to_source(entry)) change in included? looks correct, but there's a subtle interaction: relative_to_source(entry) is also called inside excluded? — if entry is already a relative path, calling relative_to_source again on it could produce an incorrect path (e.g., foo/../bar vs bar). Has this been tested with entries that already have partial source-relative paths?

The glob_include?(site.include, entry) → glob_include?(site.include, relative_to_source(entry)) change in included? looks correct, but there's a subtle interaction: relative_to_source(entry) is also called inside excluded? — if entry is already a relative path, calling relative_to_source again on it could produce an incorrect path (e.g., foo/../bar vs bar). Has this been tested with entries that already have partial source-relative paths?

I've added a specific test for that