Thread: increment question

As is everybody. That code exhibits undefined behavior, so there is no way to know. The following explanations of similar code should clarify this:
Question 3.1
Question 3.9
Question 3.10a

The links you have provided discuss cases where i is used both as array index and in increment expression while in this question that is not the case. It gives same output on GCC and Borland C. Any other explanation ?

Looks like you guys are right. This program gave an output of 10 on a different compiler. So behavior is unexpected...Thanks again

Originally Posted by ranjit89

The links you have provided discuss cases where i is used both as array index and in increment expression while in this question that is not the case. It gives same output on GCC and Borland C. Any other explanation ?

Nope, no other explanation. I guess I should have linked you to Question 3.8 as well. That has the better discussion of it, and covers exactly what in the standard guarantees that you are stepping into undefined behavior.

Oh, and I should mention a few other things:

void main is only supported in unhosted environments (embedded systems and the like). Hosted environments (i.e. those that run on a normal OS) require main to return an int per the C standard. Thus your program should be:

Code:

int main(void)
{
    // whatever code you want, as long as it's not undefined behavior :P
    return 0;
}

Also, multi-dimensional array initializers should be put int the proper amount of curly braces, and nested appropriately. You have a Nx2x3 array, so the compiler will figure out N based on your initializer. But adding the correct braces makes it easier to spot errors:

Code:

static int a[][2][3] = {
    {
        {0, 1, 2},
        {3, 4, 5}
    },
    {
        {6, 7, 8},
        {9, 10, 11}
    },
    12 // who knows what this guy belongs to
};

I got warnings for both of those as well as the undefined behavior by enabling warnings on my compiler. Here's what happened when I compiled your initial code (note the strict gcc options):

Code:

$ gcc -Wall -Wextra -pedantic --std=c99 foo.c
foo.c:2:6: warning: return type of ‘main’ is not ‘int’ [-Wmain]
foo.c: In function ‘main’:
foo.c:4:5: warning: missing braces around initializer [-Wmissing-braces]
foo.c:4:5: warning: (near initialization for ‘a[0]’) [-Wmissing-braces]
foo.c:7:16: warning: operation on ‘i’ may be undefined [-Wsequence-point]
foo.c:7:16: warning: operation on ‘i’ may be undefined [-Wsequence-point]

I would also recommend not using Borland anymore, simply because it's so outdated and no longer maintained. GCC, Pelles C, Code::Blocks/MinGW and even Microsoft Visual C++ Express are all free and currently maintained.

Nice to have the extra info..