y
is getting sign-extended in the first snippet, which would overwrite x
with -1
whenever y < 0
.
In the second snippet, the cast to int
is done before the shift, so x
actually gets the value of y
.
long l = (((long)x) << 32) | (y & 0xffffffffL);
int x = (int)(l >> 32);
int y = (int)l;