BUG: Fortify against longdouble cast loss

Base the "magic" value of the double, and use nextafter even for the
equal condition to avoid rounding during the cast to longdouble.
This might further be necessary because the code uses double constants
and not longdouble ones (not sure about this).

Using nextafter on the double value should put us firmly on the right
side and nobody should care about the few ulps in between those values.

0 files changed, 0 insertions, 0 deletions