Manual calculation of a struct/class size is not just an indication of poor design and a source of maintenance problems; it may also lead to bugs that are very hard to detect:
struct Person{char firstName[5];char lastName[8]}; //the actual size of Person is most likely larger than 13 bytes Person p= {{"john"}, {"lippman"}}; memset(&p, 0, 5+8 /* surprise */ ); //not all members of p are completely erased!Simple addition of the sizes of each data member in a struct/class will probably yield a wrong size. Wrong sizes occur because the compiler is allowed to add additional padding bytes between members whose size does not fit exactly into a machine word. Padding enables programs to execute significantly faster (especially on RISC machines). On a 32 bit (four bytes) processor, three additional bytes will be inserted between the two members of Person, thus making the first member occupy eight bytes–a size that fits neatly into two processor’s words. In our case, it also means that Person’s size is 16 bytes and not 13. (You should consult your compiler’s manual in order to find out what its default alignment is). Consequently, on some implementations, the above memset function will leave the last three bytes of p.lastName uninitialized. Therefore, the size of a struct/class should always be calculated by the sizeof operator:
memset(&p, 0, sizeof(Person)); //correct
