When malloc gave a core dump

Just now i fixed a core dump caused by a call to malloc. I was wondering y malloc wud give a core dump. With the help of google and purify, i found that an Array Bounds Write that occured well before the call to malloc, resulted in the Segmentation Fault. ABW must have corrupted Malloc's allocation table and that resulted in chaos malloc.

An instance of memory corruption in a structure

Today i was working on a structure that looked like this

struct a_{
U8bit i;
U8bit data[0];
};
typedef a_ a_t;

As can be seen the 'data' is an expandable array. This was used to point to another structure say b_t;

typedef struct b_{
int integer;
}b_t;

a_t *a;
b_t *b; // another structure of arbitrary type
memcpy(a->data, b, sizeof(b));

Assume that the memory location of structure 'a' is 200 and that of 'b' is '210'. Now since the first element of 'a' is of size 1 byte, the statement memcpy(a->data, b, sizeof(b)); had copied the structure 'b' to the memry location 201. Everything seemed to go well till now. but when i tried to unpack the intger of 'b' structure from 'a' structure I got a Segmentation Fault. (OS: Solaris ). This is because i was trying to read the integer from an memory location that was not a muliple of 4. Purify categorises this as Misaligned Memory Read.

With the knowledge i have today, this can be avoided in two ways.
1. The structure declaration of a_t shud be like this
struct a_
{
U8bit data[0];
U8bit i;
};
The parameter 'i' has been moved down to prevent misalignment of the structure to be packed into 'data' field

2. Using memcpy.

Say we wanna use this statement..

int temp = (b_t *)(a->data)->integer;

This will result in a Segmentation fault. Instead use a temporary struture of type b_t, memcpy a->data into it and then use it.

b_t *temp_struct;
temp_struct = (b_t *)malloc(sizeof(b_t)); // This structure will be properly alligned
memcpy(temp_struct, a->data, sizeof(b_t));

temp = temp_struct->integer.

This wud do the trick .

Sometimes (as was the case with me), the solution 1 wont be plausible. In my case, I dint want the structure definition to change.