[gldubc]

Fixes #15829

A function line with no positive arrows means "every function of this arity". This clause says such a line is never empty, whatever the negatives are. So when equal? computes fun(1) and not (none -> term), the result has no positives and one negative, gets reported as non-empty, and equal? returns false.

Arrows that are built explicitly don't hit this, because they always have positives. That's why (int -> atom) ≤ (none -> term) is already true, while fun(1) ≤ (none -> term) is false.

Fix

Treat the empty-positives case as one explicit top arrow and reuse the existing algorithm:

defp fun_line_empty?([], [], _seen), do: false

# With no positives, the line is the top function of its arity,
# which is (none(), ..., none() -> term())
defp fun_line_empty?([], [bdd_leaf(args, _) | _] = negatives, seen) do
  top = bdd_leaf_new(Enum.map(args, fn _ -> none() end), term())
  fun_line_empty?([top], negatives, seen)
end

Assisted by: Claude Opus 5.5