Thread: Endian swap error

Hello,
I am trying to perform an endian swap using the union method:

Code:

#include <pthread.h>
#include <stdio.h>

typedef union
{
    char c[sizeof(int)];
    int i;
} buff;

int endianSwap(int i)
{
    buff eswap;
    eswap.i = i;

    eswap.c[0] = (8>>eswap.c[0] | eswap.c[0]<<8 );
    eswap.c[1] = (8>>eswap.c[1] | eswap.c[1]<<8 );
    eswap.c[2] = (8>>eswap.c[2] | eswap.c[2]<<8 );
    eswap.c[3] = (8>>eswap.c[3] | eswap.c[3]<<8 );

    return eswap.i;
}

int main(void)
{
    int i = 1;
    printf("%d is i\n",  i);
    i = endianSwap(i);


    printf("%d  is i\n", i);
    return 0;
}

If I use "1" as a parameter it does not makes the ingeger that must have the rightmost bit set to 1 gives a different value? What is wrong here? I was trying to manipulate byte by byte using the union addressing. Maybe a misstake?

Fixed the swap error but still not swaps

Code:

eswap.c[0] = (eswap.c[0] >> 8) | eswap.c[0]<<8 );
...

I though that the union and the char array will be in the same address so I can acces 4 8 bits separately and perform 8 bit swap per array? Is that correct?

You're mixing two different methods. Either use your union to swap the bytes around or use bit-shifting to move the bits around in the original object. However, bit-shifting with signed values (particularly right-shifting, I believe ... laserlight?) is not portable (different compilers could do different things).

Right-shifting a signed value is only implementation-defined when it is negative.

6.5.7 Bitwise shift operators

5.
The result of E1 >> E2 is E1 right-shifted E2 bit positions. If E1 has an unsigned type
or if E1 has a signed type and a nonnegative value, the value of the result is the integral
part of the quotient of E1 / 2E2. If E1 has a signed type and a negative value, the
resulting value is implementation-defined.

Originally Posted by kmdv

It does not change anything - it is still not portable.

I did not say it was portable, I was confirming what algorism said by quoting the relevant section of the standard.

I`ve written a simple type universal macro that swaps int 16 or 32 also floats with no problem at all and works for my problems:

Code:

#define ENDIAN_SWAP(Type, Var)                  \
    {                                           \
    union { char c[sizeof(Type)]; Type i; }     \
    ibuff[2];                                   \
    ibuff[0].i = Var;                           \
    int j = 0;                                  \
    int h = sizeof(Type)-1;                     \
    for(; j < sizeof(Type); j++, h--)           \
    {                                           \
        ibuff[1].c[h] = ibuff[0].c[j];          \
    }                                           \
    Var =  ibuff[1].i;                          \
    }                                           \

I do agree with your solution. Thanks