misra.py: Fix Rule 4.1

[jubnzv]

This commit will fix issues with MISRA rule 4.1 caused by improper escape sequences handling.

References:

1. https://trac.cppcheck.net/ticket/9370
2. https://sourceforge.net/p/cppcheck/discussion/development/thread/7274ed3842/?limit=25#799a

[jubnzv]

@matzeschmid @whoopsmith could you please review these changes before merging? This solves the problems on your codebase?

[matzeschmid]

If you add a max. sequence length check in isOctalEscapeSequence() then the following non-compliant examples will be catched.

int c41_12        = '\12323';  // 4.1
int c41_13        = '\1232\3'; // 4.1

def isOctalEscapeSequence(symbols):
    """Checks that given symbols are valid octal escape sequence.
    Reference: n1570 6.4.4.4"""
    if len(symbols) < 2 or len(symbols) > 4 or symbols[0] != '\\':
        return False
    return all([s in string.octdigits for s in symbols[1:]])

[matzeschmid]

I removed the restriction to just check assignments to see how the fix works for non-assignment test cases. After that it generates false positives because the escape sequence split adds a '\' to each sub-token. Thus the line const char *s41_3 = "\x41" "g"; generates false positive because split result for token g is \g .
There is also a false positive for (void)printf("%i", '\141'); because %i gets \%i .

Sequence split line in script.
for sequence in ['\\' + t for t in token.str[1:-1].split('\\')]:

[danmar]

you could compress this condition I think:

if len(symbols) < 3 or symbols[:2] != '\\x':

[whoopsmith]

Doesn't work for me at all. It throws 4.1 errors all over the place for things that are not octal sequences. Like below:

static const char * mfg_pass = "pass";

[jubnzv]

Thanks for the thorough review.
It seems there are more pitfalls, I'll fix it.

[matzeschmid]

Thanks for the reasonable review.
It seems there are more pitfalls, I'll fix it.

I think the wrong sequence split described above generates most of the false positives.

[whoopsmith]

I ran the latest revision and all the false positives are gone. I haven't tested if a valid error gets caught yet.