I have been working on my friends post

here is my test case:
————————————————————————————————————————–

ert@SEMERKAND:~$ cat dnm.cpp

#include <iostream>

union manken_u {
        char    ch;
        int     integer;
        double  db;
};

#pragma pack(2)
struct manken_s_p {
        char    ch;
        int     integer;
        double  db;
};

#pragma pack()
struct manken_s {
        char    ch;
        int     integer;
        double  db;
};

#define sub(x,y) (int)(x)-(int)(y)

int main()

{
        std::cout
                << ”size of union: ” << sizeof(union manken_u) << std::endl
                << ”size of struct: ” << sizeof(struct manken_s) << std::endl
                << ”size of packed struct: ” << sizeof(struct manken_s_p) << std::endl
                << ”size of char+int+double: ” << sizeof(char)+sizeof(int)+sizeof(double) << std::endl;

        manken_s my_struct;
        char* base = &my_struct.ch;
        std::cout
                << ”offset of char ” << sub(base,&my_struct.ch) << std::endl
                << ”offset of int ” << sub(base,&my_struct.integer) << std::endl
                << ”offset of db ” << sub(base,&my_struct.db) << std::endl;

        std::cout << ” pack is 2 ” << std::endl;

        manken_s_p my_struct_p;
        base = &my_struct_p.ch;
        std::cout
                << ”offset of char ” << sub(base,&my_struct_p.ch) << std::endl
                << ”offset of int ” << sub(base,&my_struct_p.integer) << std::endl
                << ”offset of db ” << sub(base,&my_struct_p.db) << std::endl;

        return 0;
}
ert@SEMERKAND:~$ g++ dnm.cpp
ert@SEMERKAND:~$ ./a.out
size of union: 8
size of struct: 16
size of packed struct: 14
size of char+int+double: 13
offset of char 0
offset of int -4
offset of db -8
 pack is 2
offset of char 0
offset of int -2
offset of db -6

————————————————————————————————————————–

So, what is the moral of the story:

#pragma pack is a directive for compiler to determine the size of atomic unit for storing nonprimitive data types (struct for example) on memory.
If you set it to 2, chars in a struct will have a size of 2 bytes not 1 byte, because compiler will align it.
In my example first offset values are for pack(4) (default), so char has a padding value of 3 bytes to expand it to 4 bytes.
Following offset values shows pack(2) results. Char has a total lenght of 2 byte. BUT IT DOES NOT USE 2 BYTE.

And…. Why the compiler does this alignment stuff. Because different arc. have certain types of sizes of types like this

references:
http://publib.boulder.ibm.com/infocenter/macxhelp/v6v81/index.jsp?topic=/com.ibm.vacpp6m.doc/compiler/ref/rnpgpack.htm
http://docs.hp.com/en/8/pragmas.htm
and http://emreknlk.blogspot.com/