xcode LLMV 3.1 compiler array issue [duplicate]

Question

I came across with the following issue On Apple LLVM compiler 3.1:

int numIndex = 0;
int *indices = (int *)malloc(3 * sizeof(int));
indices[numIndex] = numIndex++;
indices[numIndex] = numIndex++;
indices[numIndex] = numIndex++;
for (int i = 0; i < 3; i++) {
    NSLog(@"%d", indices[i]);
}

Output: 1 0 1

And

int numIndex = 0;
int indices[3];
indices[numIndex] = numIndex++;
indices[numIndex] = numIndex++;
indices[numIndex] = numIndex++;
for (int i = 0; i < 3; i++) {
    NSLog(@"%d", indices[i]);
}

Output: 0 0 1

I'm expecting 0 1 2 as output. The same code using LLVM GCC 4.2 produces the right output. It's there any optimization flags that I'm missing or something I'm misunderstanding?

Answer 1

So it seems behavior is as follows

int numIndex = 0;
int indices[3];
indices[numIndex] = numIndex++;

here the right hand side is evaluated first, returns 0, and increments numIndex by one, then the right side is evaluated, so indices[1] gets 0

indices[numIndex] = numIndex++;

here the right hand side is evaluated first , returns 1, and increments numIndex by one,then the right side is evaluated, so indices[2] gets 1

indices[numIndex] = numIndex++;

here the right hand side is evaluated first , returns 2, and increments numIndex by one,then the right side is evaluated, so indices[3] gets 2 (and you are actually out of bounds)

And note you are never really assigned indices[0], so it could be anything (in my test it was the max int number)

EDIT- Seems from the comment given that this is behavior is actually undefined, so even though i observed this, its not a definte answer