> and did you know that IEEE-754 floating point can represent a "negative zero" that you can use for an error code in functions that return float or double?

I am begging, please never ever do this. NaN literally exists for this reason. NaN even allows you to encode additional error context and details into the value.