06-02-2008, 03:48 PM
Sure you can do that (rare) case portable:
The code won't be as efficient as the integer version, but that is always the case with pointers. We are discussing safety and not efficiency, so that is irrelevant.
It would also go against the advisory rule 11.3, but as long as you know the size of the system address bus (and document it), and use \"uint32\" rather than \"int32\", there won't be any problems.
The argument MISRA wrote was that pointers may go out of bounds, and I still don't see why integer indices would be less prone to do that.
Code:
uint32 lookup(uint32 const *vp)
{
uint32 r;
uint32 ptr_address = (uint32)vp;
uint32 array_address = (uint32)a;
if ( (ptr_address < array_address) ||
(ptr_address >= array_address + N) )
{
/* handle error. */
r = 0;
}
else
{
r = *vp;
}
return r;
}The code won't be as efficient as the integer version, but that is always the case with pointers. We are discussing safety and not efficiency, so that is irrelevant.
It would also go against the advisory rule 11.3, but as long as you know the size of the system address bus (and document it), and use \"uint32\" rather than \"int32\", there won't be any problems.
The argument MISRA wrote was that pointers may go out of bounds, and I still don't see why integer indices would be less prone to do that.