Code Analysis rule C6386 seems to report many false Write Buffer Overrun message

Question

Hi,
why does Code Analysis complain on next peace of code ?
we have many loops where the same issue is reported where we loop on (int i = 0 ; i < size ; i++ ) 
_TCHAR * Name = new _TCHAR[String.GetLength() + 2];
Name[0] = LanguageKey;
for (i=0; i < String.GetLength(); i++)
Name[i + 1] = String.GetAt(i);
Name[i+1] = 0x00;
i get next details :
C6386 Write overrun
Buffer overrun while writing to 'Name':  the writable size is '(String.public: int __cdecl ATL::CSimpleStringT<char,0>::GetLength(void)const ()+2)*1' bytes, but '4' bytes might be written.
WinRoute
gr_configuration.cpp
453
'Name' is an array of 3 elements (3 bytes)
446
Enter this loop, (assume 'i<String.GetLength()')
450
Continue this loop, (assume 'i<String.GetLength()')
450
'i' may equal 2
450
Exit this loop, ('i<String.GetLength()' is false)
450
Invalid write to 'Name[3]', (writable range is 0 to 2)
453
I think that actually if "i" may be 2 then it means the String.GetLenght() is at least 3 ( because i < String.GetLenght() ) in the
loop condition. So then the array has at least lenght of  (3 + 2) * sizeof(_TCHAR) , making it writable to at least Name[3] ( and even Name[4] ) 
Thanks, 

Answer 1

It seems to me Code Analysis is wrong.
You can also try single stepping using VS debugger, and check that index (i+1) at the end of the loop is just fine and in-bounds.
Anyway, I wonder why don't you just use string concatenation operations (i.e. operator+=) instead of writing manual concatenation code?
If you need a raw character pointer from the built string (e.g. for some legacy C API interop), you can always call CString::GetString() to get it (or use implicit CString conversion operator).
Giovanni