The decimal_to_floating() functions convert the decimal record at *pd into a floating-point value at *px, observing the modes
specified in *pm and setting exceptions in *ps. If there are no IEEE exceptions, *ps will
pd->sign and pd->fpclass are always taken into account. pd->exponent, pd->ds and pd->ndigits are used when pd->fpclass is fp_normal or fp_subnormal. In these cases pd->ds
must contain one or more ascii digits followed by a NULL and pd->ndigits is assumed to be the length of the string pd->ds.
Notice that for efficiency reasons, the assumption that pd->ndigits == strlen(pd->ds) is NEVER verified.
On output, *px is set to a correctly rounded approximation to
Thus if pd->exponent == -2 and pd->ds == "1234", *px will get 12.34 rounded to storage precision. pd->ds cannot have more than DECIMAL_STRING_LENGTH-1 significant digits because one character is used to terminate the string with a NULL. If pd->more != 0 on input then additional nonzero digits follow those in pd->ds; fp_inexact is set accordingly on output in *ps.
*px is correctly rounded according to the IEEE rounding modes in pm->rd. *ps is set
to contain fp_inexact, fp_underflow, or fp_overflow if any of these arise.
pm->df and pm->ndigits are not used.
strtod(3C), scanf(3C), fscanf(3C), and sscanf(3C) all use decimal_to_double().