C is very confusing in this regard.The older fmax function (C99-C23) has option 1 behavior. I believe it is because Kahan recommended it before it got into IEEE 754.
C also has fmaximum/fminimum (which are *different* than fmax/fmin) in C23, which implement option 2. They are not in cppreference yet. See https://port70.net/~nsz/c/c23/n3220.html#7.12.12.4
-- Peter McGoron
