q generosity problems #1
Description
There are still some problems with q<=32 in the implementation and tests.
The tests have some epsilon-checks which assume there are enough fractional bits to get close to corresponding float64 values.
{{ message }}