V2596. MISRA. The value of a composite expression should not be assigned to an object with wider essential type.

This diagnostic rule is based on the software development guidelines developed by MISRA (Motor Industry Software Reliability Association).

This diagnostic rule is relevant only to C. The C language allows much freedom in casting between arithmetic types. But such implicit conversions can also lead to hidden problems such as loss of sign, value, or precision.

Code example:

void foo()
{
  ....
  uint16_t var_a = 30000;
  uint16_t var_b = 40000;
  uint32_t var_sum;
  var_sum = var_a + var_b;  /* var_sum = 70000 or 4464? */
  ....
}

When you calculate the 'var_sum' variable value an implicit type conversion from the 'uint16_t' type to 'int' occurs. In consequence, the assignment result depends on the 'int' type size.

If 'int' has the 32-bit size, the modulo 2^32 operation is performed, and the expected '70000' value is written to the 'var_sum' variable.

If 'int' has the 16-bit size, the modulo 2^16 operation is performed, and the '70000 % 65536 == 4464' value is be written to the 'var_sum' variable.

The MISRA standard introduces the Essential type model to prevent such errors. A variable might have the following types in this model:

• Boolean, if it operates true/false values: '_Bool';
• signed, if it operates signed integer numbers, or is an unnamed enum: 'signed char', 'signed short', 'signed int', 'signed long', 'signed long long', 'enum { .... };';
• unsigned, if it operates unsigned integer numbers: 'unsigned char', 'unsigned short', 'unsigned int', 'unsigned long', 'unsigned long long';
• floating, if it operates floating-point numbers: 'float', 'double', 'long double';
• character, if it operates only characters: 'char';
• Named enum, if it operates a named set of user-defined values: 'enum name { .... };'.

Use the Essential type model to reduce the number of such subtle problems. Avoid assigning composite expressions that have a narrower essential type to variables of a wider essential type or passing such expressions to a function as an argument of a wider type.

To correct the code above, use an explicit conversion to 'uint32_t':

void foo()
{
  ....
  uint16_t var_a = 30 000;
  uint16_t var_b = 40 000;
  uint32_t var_sum;
  var_sum = (uint32_t)var_a + var_b;  /* var_sum = 70 000 */
  ....
};

Now the modulo 2^32 operation is performed in all cases, no matter what size the 'int' type has, and the error doesn't occur even if 'int' has the 16-bit size.

This diagnostic is classified as: