MISRA.STDLIB.FENV.MACRO.2012

Floating-point exception feature from fenv.h is used.

MISRA C 2012 Rule 21.12: The exception handing features of <fenv.h> should not be used

C99 [Unspecified 27, 28; Undefined 109–111; Implementation J.3.6(8)]

Category: Advisory

Analysis: Decidable, Single Translation Unit

Applies to: C99

Amplification

The identifiers feclearexcept, fegetexceptflag, feraiseexcept, fesetexceptflag and fetestexcept shall not be used and no macro with one of these names shall be expanded.

The macros FE_INEXACT, FE_DIVBYZERO, FE_UNDERFLOW, FE_OVERFLOW, FE_INVALID and FE_ALL_EXCEPT, along with any implementation-defined floating-point exception macros, shall not be used.

Rationale

In some circumstances, the values of the floating-point status flags are unspecified and attempts to access them may lead to undefined behaviour.

The order in which exceptions are raised by the feraiseexcept function is unspecified and could therefore result in a program that has been designed for a certain order not operating correctly.

Example

#include <fenv.h>

void f ( float32_t x, float32_t y ) 
{  
  float32_t z;
  
  feclearexcept ( FE_DIVBYZERO );           /* Non-compliant */
  
  z = x / y;

  if ( fetestexcept ( FE_DIVBYZERO ) )      /* Non-compliant */ 
  { 
  }

  else 
  { 
#pragma STDC FENV_ACCESS ON
    
    z = x * y; 
  }

  if ( z > x ) 
  { 
#pragma STDC FENV_ACCESS OFF

    if ( fetestexcept ( FE_OVERFLOW ) )     /* Non-compliant */ 
    { 
    } 
  } 
}