Subtract with a complex on the left and a float on the right acts as if the operands were swapped:
2J8 - 3.0
1J¯8
3.0 - 2J8
1J¯8
However, complex minus int or int minus complex is OK:
2J8 - 3
¯1J8
3 - 2J8
1J¯8
FloatCell.cc::bif_subtract is executing "return A->bif_subtract(Z, this)" when
A is not a real, no doubt inspired by the pattern used in FloatCell::bif_add.
It should rather take its inspiration from IntCell::subtract and do
"bif_negative(Z); return A->bif_add(Z,Z);" instead.
