MATH.RANGE : Floating Point Range Error

Summary

An attempt to call a floating point function on a value such that the result of the corresponding mathematical function would be unrepresentable, when no other MATH.RANGE.* warning class is appropriate.

Properties

WARNING_FILTER += discard class="Floating Point Range Error"

Triggers

CodeSonar ships with library models that allow it to recognize functions such as libm atanh() that compute a mathematical function whose result is unrepresentable for some inputs. If one of these functions is called with an unsuitable value in the input parameter position, a warning will be issued.

If you have created a custom library model for some function f() in terms of one of these existing models, calls to f() will also be capable of triggering Floating Point Range Error warnings.

Example

#include <math.h>

double call_atanh(double ind){
    double a = atanh(0.1);                /* ok */
    double b = atanh(1);   /* 'Floating Point Range Error' warning issued here */
    double c = atanh(10);  /* domain, rather than range, error: 'Argument Too High' (MATH.DOMAIN.TOOHIGH)
                            * warning issued here */
    double d = atanh(ind); /* EVIDENCED ONLY: 'Floating Point Range Error' warning issued here
                            * if CodeSonar has evidence that call_atanh() is called on a value equal to 1 or -1. */
    return a+b+c+d;
}

Relevant Configuration File Parameters

The following configuration file parameters affect checks for this warning class.

• REPORT_IMPLIED_EQUALITY
• REPORT_IMPLIED_INEQUALITY
• REPORT_SIMILAR_WARNINGS
• TIME_LIMIT_INTER_CLASSIFY
• TIME_LIMIT_INTRA_CLASSIFY
• TIME_LIMIT_INTRA_EXPLORE
• TIME_LIMIT_REFINE
• TIME_LIMIT_RESOLVE
• WARNING_FILTER