views:

89

answers:

1

Hello. Maybe I'm being stupid but I can't figure out why a pointer is being changed in this function. I have included "printf" and "puts" to show the problem. The left braket pointer is changing and invalidates the pointer arithmetic I want to do for string manipulation.

Another question I have - Is it dangerous to expect char pointers to remain constant for the same string when using C functions and do the pointers always point to addresses in a contiguous manner for strings? I can see problems arising with those concerns.

Edit: "//Changes here" placed to function which appears to change the point under my debugging.

char * py_to_c_expr(char * py_expr,int line,bool do_brakets){
    //Recursively converts python expression for use in C
    /*
     Order of evaluation. Top done last
     lambda - Not implemented yet
     if – else - Not implemented yet
     or
     and
     not
     in, not in, is, is not, <, <=, >, >=, <>, !=, == - in, not in and is not, not implemented yet
     |
     ^
     &  Bitwise AND
     <<, >>
     +, -
     *, /, //, %
     +x, -x, ~x - Positive useless
     **
     x[index], x[index:index], x(arguments...), x.attribute - Not implemented yet
     (expressions...), [expressions...], {key:datum...}, `expressions...` - Not implemented yet
     */
    char * find;
    //Convert "or", "and" and "is" to other characters to prevent conversion output confliction during processing
    char * scan;
    if (do_brakets) {
        py_expr = str_replace(py_expr, "or", "OR");
        py_expr = str_replace(py_expr, "and", "AND");
        py_expr = str_replace(py_expr, "is", "IS");
        //Convert brakets
        int nest = 0; //Amount of brakets gone into for processing
        //Stack to keep track of braket positions
        char ** left_bracket_ptr_stack;
        int left_bracket_ptr_stack_size = 0;
        scan = py_expr;
        while (true) {
            char * prior_scan;
            prior_scan = scan;
            scan = strstr(scan, "(");
            if (scan) {
                scan++;
            }
            if (scan) {
                nest++;
                //Entered braket. Add start position to stack
                left_bracket_ptr_stack = realloc(left_bracket_ptr_stack, sizeof(char *)*left_bracket_ptr_stack_size);
                left_bracket_ptr_stack[left_bracket_ptr_stack_size] = scan;
                printf("During left bracket scan:\n\nLeft braket pointer - %lli\nPython expression pointer - %lli\n",left_bracket_ptr_stack[left_bracket_ptr_stack_size],py_expr);
                left_bracket_ptr_stack_size++;
                printf("Length of stack - %i\n\n\n",left_bracket_ptr_stack_size);
            }else{
                scan = prior_scan;
                scan = strstr(scan, ")");
                if (scan) {
                    scan++;
                }
                if (scan) {
                    if (!nest) {
                        printf("Error: On line %i an expression closes too many brakets",line);
                        exit(1);
                    }
                    nest--;
                    //Exited bracket, process expression for braket
                    puts("During right bracket scan and processing:\n\n");
                    int len = scan - left_bracket_ptr_stack[left_bracket_ptr_stack_size - 1];
                    char * inner_braket = malloc(sizeof(char) * (len - 1));
                    strncpy(inner_braket,left_bracket_ptr_stack[left_bracket_ptr_stack_size - 1],len - 1); //Changes here
                    char * bracket_expression = py_to_c_expr(inner_braket,line,true);
                    char * new_py_expr = malloc(sizeof(char) * (strlen(py_expr) - strlen(inner_braket) + strlen(bracket_expression)));
                    free(inner_braket);
                    printf("Left bracket pointer - %lli\n",left_bracket_ptr_stack[left_bracket_ptr_stack_size -1]);
                    printf("Length of stack - %i\n",left_bracket_ptr_stack_size);
                    printf("Python Expression pointer - %lli\n\n\n",py_expr);
                    strncpy(new_py_expr,py_expr,left_bracket_ptr_stack[left_bracket_ptr_stack_size - 1] - py_expr - 1);
                    new_py_expr[left_bracket_ptr_stack[left_bracket_ptr_stack_size- 1] - py_expr - 1] = '\0';
                    strcat(new_py_expr,bracket_expression);
                    free(bracket_expression);
                    strcat(new_py_expr,scan);
                    free(py_expr);
                    py_expr = new_py_expr;
                }else{
                    scan = prior_scan;
                    if (nest) {
                        printf("Error: On line %i an expression does not close all brakets",line);
                        exit(1);
                    }
                    break;
                }
            }
        }
    }
    //Find operators and replace them with the functions. Recursively process each side.
    char * new_expr;
    scan = py_expr; //Reset scan to begining of the expression now that brakets are done
    scan = strstr(scan, "OR");
    if (scan) {
        char * left_side = malloc(sizeof(char) * (scan - py_expr + 1));
        char * right_side = malloc(sizeof(char) * (strlen(py_expr) - strlen(scan)));
        strncpy(left_side,py_expr,scan - py_expr);
        left_side[scan-py_expr] = '\0';
        strcpy(right_side,scan + 3);
        char * new_left_side = py_to_c_expr(left_side,line,false); //Recursively process left
        char * new_right_side = py_to_c_expr(right_side,line,false); //and right operands
        new_expr = malloc(sizeof(char) * (strlen(new_left_side) + strlen(new_right_side) + 5));
        strcpy(new_expr,"or(");
        strcat(new_expr,new_left_side);
        strcat(new_expr,",");
        strcat(new_expr,new_right_side);
        strcat(new_expr,")");
        return new_expr;
    }
    return py_expr;
}

Output:

During left braket scan:

Left braket pointer - 4296016033
Python expression pointer - 4296016032
Length of stack - 1


During left braket scan:

Left braket pointer - 4296016034
Python expression pointer - 4296016032
Length of stack - 2


During left braket scan:

Left braket pointer - 4296016035
Python expression pointer - 4296016032
Length of stack - 3


During right braket scan and processing:


Left braket pointer - 3473436750496411705
Length of stack - 3
Python Expression pointer - 4296016032


Program received signal:  “EXC_BAD_ACCESS”.

Thank you very much for any answer.

+2  A: 

You're allocating one less pointer than you need here:

left_bracket_ptr_stack = realloc(left_bracket_ptr_stack, sizeof(char *)*left_bracket_ptr_stack_size);

(it should be (left_brack_ptr_stack_size + 1) that you multiply by).

In addition:

  • %p is the correct printf format string for printing pointers, not %lli;
  • strncpy() is not really a "safer strcpy()" - it is designed for filling fixed-length text fields, which means that it does not always nul-terminate the destination.
caf
Thank you for the answer. I'm about to look into it but I will say that "%p" gives unreadable hexadecimal (I, like many others, like to read decimal). "%lli" works. I have used casts with my latest debugging just in case it broke something...
Matthew Mitchell
You are very correct about the memory allocation. It seems to have fixed the pointer change issue. Now I have some other memory access error when the function recursively converts the bracketed expressions. I'll need to debug that now.
Matthew Mitchell
Seems I forgot to terminate that "strncpy" string. Oh dear, this is very bad on my part.
Matthew Mitchell
Funnily enough I've removed the stack after finding an alternative method to convert the expressions.
Matthew Mitchell
I like to read decimal for things that are well-represented in decimal, and hex for things that are well-represented in hex. Pointers are terribly represented in decimal. 4296016034 gives no information and is useless. Its hex equivalent, 0x1001000A2, is quite informative: among other things, it shows that you must be running on a 64-bit system.
Reinderien
@Reinderien: Agreed. It's also easy to see the power-of-2 alignment of a pointer in hex (for example, any pointer aligned on a 4k page boundary must end in `000` in hex) and it's easier to compare them to debugging info which shows addresses in hex, like `/proc/<pid>/maps`.
caf