Implicit type conversions in switch statements

[dcrocker]

I've a question concerning the following example:

Code:

int8 x;
...
switch(x)
{
    case 0:      /* valid */
    case 255:    /* valid? */
    case 1L:     /* invalid, implicit cast from long to int */
        ...
}

I would expect any good static checker to report that x can never equal 255 here. But is 'case 255:' actually a MISRA 2004 rule violation? The C standard says that the switch expression x gets promoted to 'int' in this case, so the type of the case label 255 does match the type of the promoted switch expression. I can't find any MISRA rule that says that we should consider the underlying type (instead of the promoted type) of the switch variable when deciding whether a case label is allowed or not. If "case 255:" was the only case label on some code, then rule 14.1 (no unreachable code) would apply; but in this example, there is no unreachable code.

[misra-c]

The underlying type of the controlling expression of the switch statement is signed 8-bit.

Conversion of 1L from underlying long int to signed 8-bit violates Rule 10.1.

Conversion of 255 from underlying signed 16-bit to signed 8-bit also violates Rule 10.1.

This conversion is mentioned in Section 6.10.2 under the subsection Assigning Conversions. See also the Exemplar Suite file mc2_1001.c.