Search code examples
cllvmxcode4.5lldb

Xcode C debugging changes values of pointers while stepping

I am debugging a C code where I have a pointer inside a pointer to a struct named board. There is a function where I am printing the board:

static void board_print(board *b){
    int i,j;
    char data;
    for (i = 0; i < size; i++) {
        for (j = 0; j < size; j++) {
            data = b->data[i * size + j];
            if(data){
                printf("X ");
            }else{
                printf("O ");
            }
        }
        printf("\n");
    }
}

Here is the weird part. When I hit my breakpoint initially at the beginning of first for loop, everything is OK, my data is correct, all the pointers work etc. as seen below:

first

Then, I step into the loop for the first time, with i and j equal to zero, and b->data[0] should be perfectly valid, as it was two steps ago. Suddenly, as soon as I step on the line data = b->data[i * size + j];, the data pointer changes into a null pointer. When I execute the line, I (obviously) get a bad access error, as seen below:

second

What could be the reason? I've used C before, and I've pretty got a grasp of it, but I've never seen a pointer value changing suddenly to null before while stepping in a single-threaded simple C program. I am using Apple LLVM Compiler 4.1 to compile and lldb to debug, which are the defaults with XCode 4.5.

Update: The same behavior observed with compiling with gcc and debugging with gdb. Almost hundred percent an arror at my side, but I have no idea what is wrong with the code..

Update #2: I've noticed something even stranger on gcc/gdb now. The just before executing the line data = b->data[i * size + j];, I can access everything from the debugger with no issues. Right after executing that line, I can't access b->data entirely, including the values that I've accessed right before stepping:

third

After the $4 = ... line which executed successfully in debugger, I've stepped over the line. Then I've got various addressing errors as see above. I really have no idea what's going on...

Update #3: I've noticed something very weird. Here, first look at the fix that I've implemented. This one started to work with no problem when I got rid of the variable named data completely:

fourth

Now, look closely at the screenshot that I've uploaded with the Update #2: Right after I assign a value to the local variable named data, also b->data gets its address changed. It looks like a side effect of the assignment. But I have no idea what is the reason behind it.

Solution

  • Obviously something / somebody changes the property "data" of the board structure.

    Why ? I can only see 3 reasons :

    • Your application is multi-threaded and an other thread updates the data pointer to NULL (which is not your case, you are using only one thread, sorry didn't notice).

    • The board structure is allocated from the stack but the content is not any more valid... For example : The pointer of a local variable is returned, but the variable (the struct) is destroyed (because this is a local variable) when returning the pointer...

    • The board structure is allocated from the heap then freed, and finally this freed pointer is still used : the heap memory is corrupted by something else...

    My guess : The second point which is a frequent mistake

    So a simple question : The board struct is stored in HEAP or STACK memory ?