CERT C: Rec. FLP07-C

R2026b

Cast the return value of a function that returns a floating-point type

Since R2026b

Description

Cast the return value of a function that returns a floating-point type.¹

Polyspace Implementation

Polyspace^® checks for the issue Imprecise returned float.

Examples

Imprecise returned float

Issue

The issue occurs when the return type of a function is float but:

• The return expression is a hard-coded literal without an explicit cast to the return type.

• The return expression is an arithmetic expression involving literals without an explicit cast of the entire expression to the return type.

The checker does not report violations on return expressions that do not contain any literals.

Risk

The returned value can have different precision than the declared return type on different platforms, leading to inconsistent program execution across architectures. Comparisons or subsequent arithmetic with the returned value can produce unexpected results because of differences in precision.

Fix

Cast the return expression to the declared return type of the function.

Example — Floating-Point Return Value Without Explicit Cast

In this example, the function calc_percentage returns a floating-point expression without an explicit cast. Polyspace reports a violation on the return statement.

float calc_percentage(float value) {
    return value * 0.1f;    // Noncompliant
}

void use_percentage(float x) {
    double y = calc_percentage(x);
}

The expression value * 0.1f involves a hard-coded literal and can retain higher precision than float on some platforms. Without an explicit cast in the return statement, the caller receives a value with unpredictable precision.

Correction — Cast Return Expression to Declared Type

The corrected code casts the return expression to float, removing any excess precision.

float calc_percentage(float value) {
    return (float)(value * 0.1f); // Compliant
}

void use_percentage(float x) {
    double y = calc_percentage(x);
}

Check Information

Group: Rec. 05. Floating Point (FLP)

PQL Name: std.cert.FLP07_C

Version History

Introduced in R2026b