Anyway, so recently I ran into a problem where I needed to solve a set of 686 linear equations. (To calculate coeffcients of a 686 degree polynomial). So I quickly ran into the problem of not having enough bits in the built in c/c++ numerical types to store intermediate results. A number raised to the power of 686 needs a lot of bits to represent itself!!!!

So I implemented integer math code that works on arbitrary size arrays of unsigned int, thus giving me as many bits as I need. This was actually not too trivial, especially the multiplication and division!

Anyway I thought this could be a good opportunity to give c++ operator overriding another try, as I always stayed away from that on the grounds of it being very obfuscated to debug later on. And very quickly I indeed confirmed my initial views! Here is the case:

I have the following code:

// constructor
integer::integer(const unsigned int bits, const unsigned int *pValue)
{
assert((bits & 0×1F) == 0);
mpValue = new unsigned int[bits / 0×20];
assert(mpValue);
unsigned int i;
for(i = 0; i > 32);
}
}

// operator -
integer integer::operator-() const
{
integer neg(mBits, mpValue);
neg.twos_compliment( );
return neg;
}

So lets say I do something like:

unsigned int arr[2] = { 1, 0 };
integer b(64, arr);
b =-b;

Looks good, right? But it crashes in a destructor of “integer” trying to delete mpValue.

Why?

Well what seems to happen is that when the - operator is called c++ creates a ‘hidden’ instance of “integer” when executing

return neg;

Since there is no ‘=’ anywhere in that statement the default assignment operator gets called so the mpValue in this ‘hidden’ instance is now pointing to the same place as mpValue of ‘neg’.

Then when the - operator goes out of scope destructor of ‘neg’ gets called and deletes mpValue, but the ‘hidden’ instance is still alive because it is needed for the b =-b statement, however once b=-b is executed, destructor for the ‘hidden’ instance gets called and it crashes trying to delete its mpValue, because it already has been deleted….

So the fix was to implement another constructor:

// constructor
integer::integer(integer &src)
{
mBits = src.bits( );
mpValue = new unsigned int[mBits / 0×20];
unsigned int i;
for(i = 0; i