False positive on misra-c2012-5.9

[José Martins]

The exception to the MISRA rule 5.9 described in the guidelines, regarding static inlined functions in header files, is not taken into account by the cppcheck misra add-on. For example, when using the following files:

// file a.h
static inline int f(void) {
    return;
}

// file b.c
#include "a.h"

// file c.c
#include "a.h"

Running cppcheck with the misra add-on triggers a rule 5.9 violation when I believe it shoulnd't:

[23:29] misratests | cppcheck --addon=misra a.h b.c c.c
Checking a.h ...
1/3 files checked 57% done
Checking b.c ...
2/3 files checked 78% done
Checking c.c ...
3/3 files checked 100% done
a.h:1:19: style: misra violation (use --rule-texts=<file> to get proper output) [misra-c2012-5.9]
static inline int f(void){

Best regards,
José

[Daniel Marjamäki]

Thanks! I have created https://trac.cppcheck.net/ticket/10820

cppcheck --addon=misra a.h b.c c.c

I do not recommend that you run cppcheck on the headers directly.

The command cppcheck --addon=misra b.c c.c should be used instead. since b.c and c.c include a.h , the header will be checked in those translation units more properly.

If you check a header directly you can for instance expect false positives for unusedStructMember.

[José Martins]

Thank you for bringing it to my attention Daniel! I've been passing headers to cppcheck, both with and without misra add-on, since I started using it.

Nevertheless, the issue it's still there if only passing the c files. I took a look at the code yesterday, and I believe it is a simple matter to fix. I'll probably submit a PR later today.

[Daniel Marjamäki]

I assume you fixed https://trac.cppcheck.net/ticket/10820 ? I can't reproduce the FP right now. I have closed it.

[José Martins]

While trying to fix this issue, I came across another issue where sometimes the inline keyword is ignored when we use a multiple word return type (e.g. unsigned long). For example, using misra rule 8.10 for illustration purposes:

//file r8_10.c
inline long f(void) { return; }

Running cppcheck --addon=misra r8_10.c correctly yields:

Checking r8_10.c ...
r8_10.c:2:13: style: misra violation (use --rule-texts=<file> to get proper output) [misra-c2012-8.10]
inline long f(void);

However, a slight change to the file to use a multi-word return type:

//file r8_10.c
inline unsigned long f(void) { return; }

No longer triggers any violation.

I believe I also have a suggestion to patch this issue, although I'm not completely confident it is the best approach. PR later.