Question

I have this sample code for converting 32 bit integers to ip addresses.

#include <stdio.h>
int main()
{
 unsigned int c ;
 unsigned char* cptr  = (unsigned char*)&c ;
 while(1)
 {
  scanf("%d",&c) ;
  printf("Integer value: %u\n",c);
  printf("%u.%u.%u.%u \n",*cptr, *(cptr+1), *(cptr+2), *(cptr+3) );
 }
}

This code gives incorrect output for input 2249459722 .
But when i replace

scanf("%d",&c) ;

by

scanf("%u",&c) ;

The output comes out to be correct.

P.S : I know about inet_ntop and inet_pton.
I expect answers other than suggesting those.

Answer 1

You are coding ‘sinfully‘ (making a number of mistakes which will hurt you sooner or later – mostly sooner). First off, you are assuming that the integer is of the correct endian-ness. On some machines, you will be wrong – either on Intel machines or on PowerPC or SPARC machines.

In general, you should show the actual results you get rather than just saying that you get the wrong result; you should also show the expected result. That helps people debug your expectations.

---

Here’s my modified version of your code – instead of requesting input, it simply assumes the value you specified.

#include <stdio.h>
int main(void)
{
    unsigned int c = 2249459722;
    unsigned char* cptr  = (unsigned char*)&c;
    printf("Integer value:  %10u\n", c);
    printf("Integer value:  0x%08X\n", c);
    printf("Dotted decimal: %u.%u.%u.%u \n", *cptr, *(cptr+1), *(cptr+2), *(cptr+3));
    return(0);
}

When compiled on my Mac (Intel, little-endian), the output is:

Integer value:  2249459722
Integer value:  0x8614080A
Dotted decimal: 10.8.20.134 

When compiled on my Sun (SPARC, big-endian), the output is:

Integer value:  2249459722
Integer value:  0x8614080A
Dotted decimal: 134.20.8.10 

(Using GCC 4.4.2 on the SPARC, I get a warning:

xx.c:4: warning: this decimal constant is unsigned only in ISO C90

Using GCC 4.2.1 on Mac – with lots of warnings enabled (gcc -std=c99 -pedantic -Wall -Wshadow -Wpointer-arith -Wstrict-prototypes -Wmissing-prototypes -Werror) – I don’t get that warning, which is interesting.) I can remove that by adding a U suffix to the integer constant.

---

Another way of looking at the problems is illustrated with the following code and the extremely fussy compiler settings shown above:

#include <stdio.h>

static void print_value(unsigned int c)
{
    unsigned char* cptr  = (unsigned char*)&c;
    printf("Integer value:  %10u\n", c);
    printf("Integer value:  0x%08X\n", c);
    printf("Dotted decimal: %u.%u.%u.%u \n", *cptr, *(cptr+1), *(cptr+2), *(cptr+3));
}

int main(void)
{
    const char str[] = "2249459722";
    unsigned int c = 2249459722;

    printf("Direct operations:\n");
    print_value(c);

    printf("Indirect operations:\n");
    if (sscanf("2249559722", "%d", &c) != 0)
        printf("Conversion failed for %s\n", str);
    else
        print_value(c);
    return(0);
}

This fails to compile (because of the -Werror setting) with the message:

cc1: warnings being treated as errors
xx.c: In function ‘main’:
xx.c:20: warning: format ‘%d’ expects type ‘int *’, but argument 3 has type ‘unsigned int *’

Remove the -Werror setting and it compiles, but then shows the next problem that you have – the one of not checking for error indications from functions that can fail:

Direct operations:
Integer value:  2249459722
Integer value:  0x8614080A
Dotted decimal: 10.8.20.134 
Indirect operations:
Conversion failed for 2249459722

Basically, the sscanf() function reports that it failed to convert the string to a signed integer (because the value is too large to fit – see the warning from GCC 4.4.2), but your code was not checking for the error return from sscanf(), so you were using whatever value happened to be left in c at the time.

So, there are multiple problems with your code:

• It assumes a particular architecture (little-endian rather than recognizing that big-endian also exists).
• It doesn’t compile cleanly when using a compiler with lots of warnings enabled – for good reason.
• It doesn’t check that functions that can fail actually succeeded.

---

Alok’s Comment

Yes, the test on sscanf() is wrong. That’s why you have code reviews, and also why it helps to post the code you are testing.

I’m now a bit puzzled – getting consistent behaviour that I can’t immediately explain. With the obvious revision (testing on MacOS X 10.6.2, GCC 4.2.1, 32-bit and 64-bit compilations), I get one not very sane answer. When I rewrite more modularly, I get a sane answer.

+ cat yy.c
#include <stdio.h>

static void print_value(unsigned int c)
{
    unsigned char* cptr  = (unsigned char*)&c;
    printf("Integer value:  %10u\n", c);
    printf("Integer value:  0x%08X\n", c);
    printf("Dotted decimal: %u.%u.%u.%u \n", *cptr, *(cptr+1), *(cptr+2), *(cptr+3));
}

int main(void)
{
    const char str[] = "2249459722";
    unsigned int c = 2249459722;

    printf("Direct operations:\n");
    print_value(c);

    printf("Indirect operations:\n");
    if (sscanf("2249559722", "%d", &c) != 1)
        printf("Conversion failed for %s\n", str);
    else
        print_value(c);
    return(0);
}


+ gcc -o yy.32 -m32 -std=c99 -pedantic -Wall -Wshadow -Wpointer-arith -Wstrict-prototypes -Wmissing-prototypes yy.c
yy.c: In function ‘main’:
yy.c:20: warning: format ‘%d’ expects type ‘int *’, but argument 3 has type ‘unsigned int *’


+ ./yy.32
Direct operations:
Integer value:  2249459722
Integer value:  0x8614080A
Dotted decimal: 10.8.20.134 
Indirect operations:
Integer value:  2249559722
Integer value:  0x86158EAA
Dotted decimal: 170.142.21.134 

I do not have a good explanation for the value 170.142.21.134; but it is consistent on my machine, at the moment.

+ gcc -o yy.64 -m64 -std=c99 -pedantic -Wall -Wshadow -Wpointer-arith -Wstrict-prototypes -Wmissing-prototypes yy.c
yy.c: In function ‘main’:
yy.c:20: warning: format ‘%d’ expects type ‘int *’, but argument 3 has type ‘unsigned int *’


+ ./yy.64
Direct operations:
Integer value:  2249459722
Integer value:  0x8614080A
Dotted decimal: 10.8.20.134 
Indirect operations:
Integer value:  2249559722
Integer value:  0x86158EAA
Dotted decimal: 170.142.21.134 

Same value – even in 64-bit instead of 32-bit. Maybe the problem is that I’m trying to explain undefined behaviour, which is more or less by definition unexplainable (inexplicable).

+ cat xx.c
#include <stdio.h>

static void print_value(unsigned int c)
{
    unsigned char* cptr  = (unsigned char*)&c;
    printf("Integer value:  %10u\n", c);
    printf("Integer value:  0x%08X\n", c);
    printf("Dotted decimal: %u.%u.%u.%u \n", *cptr, *(cptr+1), *(cptr+2), *(cptr+3));
}

static void scan_value(const char *str, const char *fmt, const char *tag)
{
    unsigned int c;
    printf("Indirect operations (%s):\n", tag);
    fmt = "%d";
    if (sscanf(str, fmt, &c) != 1)
        printf("Conversion failed for %s (format %s \"%s\")\n", str, tag, fmt);
    else
        print_value(c);
}

int main(void)
{
    const char str[] = "2249459722";
    unsigned int c = 2249459722U;

    printf("Direct operations:\n");
    print_value(c);
    scan_value(str, "%d", "signed");
    scan_value(str, "%u", "unsigned");

    return(0);
}

Using the function argument like this means GCC cannot spot the bogus format any more.

+ gcc -o xx.32 -m32 -std=c99 -pedantic -Wall -Wshadow -Wpointer-arith -Wstrict-prototypes -Wmissing-prototypes xx.c


+ ./xx.32
Direct operations:
Integer value:  2249459722
Integer value:  0x8614080A
Dotted decimal: 10.8.20.134 
Indirect operations (signed):
Integer value:  2249459722
Integer value:  0x8614080A
Dotted decimal: 10.8.20.134 
Indirect operations (unsigned):
Integer value:  2249459722
Integer value:  0x8614080A
Dotted decimal: 10.8.20.134 

The results are consistent here.

+ gcc -o xx.64 -m64 -std=c99 -pedantic -Wall -Wshadow -Wpointer-arith -Wstrict-prototypes -Wmissing-prototypes xx.c


+ ./xx.64
Direct operations:
Integer value:  2249459722
Integer value:  0x8614080A
Dotted decimal: 10.8.20.134 
Indirect operations (signed):
Integer value:  2249459722
Integer value:  0x8614080A
Dotted decimal: 10.8.20.134 
Indirect operations (unsigned):
Integer value:  2249459722
Integer value:  0x8614080A
Dotted decimal: 10.8.20.134

And these are the same as the 32-bit case. I’m officially bemused. The main observations remain accurate – be careful, heed compiler warnings (and elicit compiler warnings), and don’t assume that “all the world runs on Intel chips” (it used to be “don’t assume that all the world is a VAX”, once upon a long time ago!).