三元运算符作为C宏中的三元运算符的操作数

Question

I'm trying to implement Sieve of Erathostenes to get prime bits of an unsigned long array, so I wrote a macro to check the value of certain bit.. (function would be easier but it's school asignment so it has to be macro) I needed to test whether the index is in range of bits, so there is function call FatalError , which is void function with exit(1) call, so the comma operator makes the whole macro possible to be inside of condition

#define GetBit(array_name, index) \
  (((index) < (array_name)[0]) && ((index) >= 0)) ? \
  (((array_name)[((index) / BYTE) + 1] & ( (unsigned long)1 << \
    ((index) % BYTE))) ? 1 : 0) : \
  (FatalError("Index %ld out of range 0..%ld\n", (long)(index), \
    (long)(array_name)[0]), 0)

More readable form:

#define GetBit(array_name, index) \
  (range check) ? \
  ((bit shift, and) ? 1 : 0) : \
  (function call, 0)

On array_name[0] , there is the size of the array in bits.

So my problem is, that not even the first index can get through range check, the Sieve starts with index 2 and the program immediately ends with

~ $ gcc primes.c fatalerror.c -pedantic -Wall -g -std=c99 -lm; ./a.out
FATAL ERROR: Index 2 out of range 0..1000
~ $ 

Answer 1

Well, it seems, that the problem was in Erathostenes function in

if( ! GetBit(pole, m) )

with the ! macro of course expanded as,

!(range check) ? op1 : op2

so the condition was negated

I removed the exclamation mark and switched the return operands of op1

((shift, and) ? 0 : 1)

on first look the values seem a little illogical, but work just fine

thanks to everyone for your time

Answer 2

Clearly there is a bug, so ignore the 'correct' form of answer, and make it easy for yourself to find the bug. It'll likely be quicker to find the problem if you go an easier route. As long as you try to stick with a single macro, it is likely to be harder to debug.

Concretely, I'd turn the macro into a function, to make it easier to debug, then break the expression down into component parts, and add a load of print statements to get a better understanding of what is happening:

printf("(((index) < (array_name)[0]) && ((index) >= 0)) = %d\n", 
        (((index) < (array_name)[0]) && ((index) >= 0)));
printf("(array_name)[((index) / BYTE) + 1] = %d\n", 
        (array_name)[((index) / BYTE) + 1]);
printf("( (unsigned long)1 << ((index) % BYTE)) = %d\n", 
        ( (unsigned long)1 << ((index) % BYTE)));
printf("BYTE=%d\n", BYTE);

// etc.

and redefine getbit to use

printf("Index %ld out of range 0..%ld\n", (long)(index), (long)(array_name)[0]);