Thank you very much. You explanation was very helpful. so it just looks like I got lucky on a 64bit platform. On 8/10/07, John Love-Jensen <eljay@xxxxxxxxx> wrote: > Hi Raghu, > > > what am I doing wrong. > > Change this: > > Base *temp = new Child [1024] (3, 2); > > ...to this... > > Child *temp = new Child [1024]; > > Add in a default constructor for Child, and probably a default constructor > for Base too -- since you cannot use initialization parameters for a new[]. > > When the delete[] deletes the Base array, the first Child element happens to > coincide enough (on your platform; although I think its not guaranteed to be > portable) in memory with temp[0] that the code actually executes the virtual > destructor and gets to Child::~Child as desired. > > When it moves on to the destructor for temp[1], the stride is off because > temp does not really point to an array of Base, it points to an array of > Child... and then you get a bus error when the code attempts to dereference > who-knows-what garbage memory as if it were a virtual functional table > pointer to get to the virtual destructor. > > You can use Base* to point to a single Child object. > > You cannot use Base* to point to an array of Child objects. (Well, you can > get to the first Child object, but none of the subsequent ones.) > > You could change your code to use a std::vector<Base*>. Or use one of the > Boost managed containers and/or smart pointers. > > HTH, > --Eljay > > -- Regards Raghu
