PRE02-A. Macro replacement lists should be parenthesized

Macro replacement lists should be parenthesized to protect any lower-precedence operators from the surrounding expression. See also PRE00-C. Prefer inline or static functions to function-like macros - CERT Secure Coding Standards and PRE01-A. Use parentheses within macros around parameter names.

Non-Compliant Code Example

This CUBE() macro definition is noncompliant because it fails to parenthesize the replacement list.

#define CUBE(X) (X) * (X) * (X)
int i = 3;
int a = 81 / CUBE(i);

As a result, the invocation

int a = 81 / CUBE(i);

expands to

int a = 81 / i * i * i;

which evaluates as

int a = ((81 / i) * i) * i);  /* evaluates to 243 */

which is not the desired behavior.

Compliant Solution

With its replacement list parenthesized, the CUBE() macro expands correctly for this type of invocation.

#define CUBE(X) ((X) * (X) * (X))
int i = 3;
int a = 81 / CUBE(i);

This compliant solution violates PRE00-C. Prefer inline or static functions to function-like macros - CERT Secure Coding Standards. Consequently, this solution would be better implemented as an inline function.

Non-Compliant Code Example

In this noncompliant code example, EOF is defined as -1. The macro replacement list consists of a unary negation operator '-' followed by an integer literal '1'.

#define EOF -1
/* ... */
if (getchar() EOF) {
   /* ... */
}

In this example, the programmer has mistakenly omitted the comparison operator (see MSC02-C. Avoid errors of omission) from the conditional statement, which should be getchar() != EOF. After macro expansion, the conditional expression is incorrectly evaluated as a binary operation: getchar()-1. This is syntactically correct, even though it is certainly not what the programmer intended. Note that this example also violates DCL00-C. Const-qualify immutable objects.

Parenthesizing the -1 in the declaration of EOF ensures that the macro expansion is evaluated correctly.

#define EOF (-1)

Once this modification is made, the noncompliant code example no longer compiles because the macro expansion results in the conditional expression getchar() (-1), which is no longer syntactically valid. Note that there must be a space after EOF because otherwise it becomes a function-like macro (and one that is incorrectly formed, because -1 cannot be a formal parameter).

Compliant Solution

In this compliant solution, the macro definition is replaced with an enumeration constant in compliance with DCL00-C. Const-qualify immutable objects.

enum { EOF = -1 };
/* ... */
if (getchar() != EOF) {
   /* ... */
}

Exceptions

PRE02-EX1. A macro that expands to a single identifier or function call is not affected by the precedence of any operators in the surrounding expression, so its replacement list need not be parenthesized.

#define MY_PID getpid()

Risk Assessment

Failing to parenthesize macro replacement lists can cause unexpected results.

Automated Detection

The LDRA tool suite V 7.6.0 can detect violations of this recommendation.

Related Vulnerabilities

Search for vulnerabilities resulting from the violation of this rule on the CERT website.

References

[[ISO/IEC 9899:1999]] Section 6.10, "Preprocessing directives," and Section 5.1.1, "Translation environment"
[[ISO/IEC PDTR 24772]] "JCW Operator precedence/Order of Evaluation", "NMP Pre-processor Directions"
[[Plum 85]] Rule 1-1
[[Summit 05]] Question 10.1

PRE01-A. Use parentheses within macros around parameter names      01. Preprocessor (PRE)       PRE03-C. Prefer typedefs to defines for encoding types