Division by result of sizeof() when using _countof macro

[Sander Bouwhuis]

I have this:

int32 i32NewLen;
wchar_t pwcText[256];
// ...
memmove(pwcText, &pwcText[i32NewLen], (_countof(pwcText) - i32NewLen) * sizeof(*pwcText));

I get this warning:
[E:/Development/Dms v5.0.1/Classes/Miscellaneous.cpp:86] (warning) Division by result of sizeof(). memmove() expects a size in bytes, did you intend to multiply instead? [sizeofDivisionMemfunc]

By multiplying the amount of characters (countof(pwcText) - i32NewLen) with the size of a single character sizeof(pwcText) I get a byte count.

Doesn't CppCheck understand the countof macro? In Win32 API countof is defined like this:

#ifdef __cplusplus
    extern "C++"
    {
        template <typename _CountofType, size_t _SizeOfArray>
        char (*__countof_helper(_UNALIGNED _CountofType (&_Array)[_SizeOfArray]))[_SizeOfArray];

        #define __crt_countof(_Array) (sizeof(*__countof_helper(_Array)) + 0)
    }
#else
    #define __crt_countof(_Array) (sizeof(_Array) / sizeof(_Array[0]))
#endif

[versat]

When specifying a windows platform and/or using the windows library Cppcheck is aware of the _countof macro. This is from windows.cfg:

<define name="_countof(_Array)" value="(sizeof(_Array) / sizeof(_Array[0]))"/>

Without this library configuration the warning message disappears.
I guess that Cppcheck issues the warning as soon as it sees the division with the result of the sizeof(). I assume that it is not verified that the number of elements is NOT multiplied with the result of the sizeof() again at a later time.
Maybe, to avoid false positives, Cppcheck could avoid this warning when more than the division with a sizeof() result is seen. At least it could be marked as inconclusive then.

[versat]

Thank you for the report.
I created a ticket for the issue: https://trac.cppcheck.net/ticket/9648

[Sander Bouwhuis]

This is always how I do it. It's basically, tot_bytes = tot_items * item_size

I use the same thing for all the places I have to supply a buffer in bytes instead of objects.
How else would you do it? Is there another way it's done in Linux for instance?

[versat]

I do not see any problem with it, you are correct.
It is a false positive from Cppcheck, the check should get enhanced I think.

[Daniel Marjamäki]

yeah it seems to be a bad warning. Your code makes perfect sense.