This looks to have been a typo in
ebd6191b0eec6e23d96c92aaa06212de339207e3 that
switched this code to operate on double points instead of int points, in that it
did not also update the local temporaries used in this calculation. As a result,
this function produced inaccurate answers due to rounding when points were too
close together and downright wrong answers when points did not fit in an int
data type. Squashes 4 -Wfloat-conversion warnings.
static int record_path(node_t * n, port * prt, int side, boxf rv[],
int *kptr)
{
- int i, ls, rs;
+ int i;
+ double ls, rs;
pointf p;
field_t *info;