Connect to a socket

leamassiot · 03-09-2007, 04:58 AM

I want to connect to a server socket identified by :
- its IP address : X.Y.Z.W (decimal dotted notation)
- and its TCP port : n

I cannot use the usual:
gethostbyname();
inet_ntoa();
sequence of instructions
because I do not assume I know hosts
according to their host names
but according to their IP addresses.

I need to provide the connect() system call with the proper arguments.
int connect(int socket, const struct sockaddr * address, socklen_t address_len);

Let us say I declare :
struct sockaddr_in in;
I do not know what to put in :
in.in_addr.s_addr (which is an unsigned int representing the IP address in network byte order).

Please help me.

wjevans_7d1@yahoo.co · 03-09-2007, 08:36 AM

First, the normal sequence (where a hostname such as "www.google.com" is given) is not

Code:

gethostbyname()
inet_ntoa()

but simply

Code:

gethostbyname()

because what is returned from gethostbyname() is the address (or more than one address) already in the four-byte network-ordered format.

Second, inet_ntoa() converts not from decimal-dotted format to four-byte network-ordered format, but the other way around. "n" stands for "network", and "a" stands for "ASCII". inet_ntoa() goes from network to ASCII. What you want instead is inet_aton().

Third, instead of using inet_aton() and inet_ntoa(), I strongly recommend you use inet_pton() and inet_ntop(). (The "p" stands for "presentation" format.) There are two reasons to do this.

If you ever go back from 4-byte network addresses to printable dotted-decimal form, inet_ntop() is re-entrant. It stores the data in memory that you allocate, not in a static buffer. This will make your program easier to maintain if it gets much larger than you originally envisioned, which is something that happens often.
You'll be that much closer to the day when you want to convert your program to be IPV6-aware.

The following code demonstrates it all:

Code:

#include <stdio.h>
#include <stdlib.h>
#include <arpa/inet.h>
#include <netinet/in.h>

int main(int    argc,
         char **argv
        )
{
  char   ntop_buffer[INET_ADDRSTRLEN];

  struct sockaddr_in inxyz;

  if(argc!=2)
  {
    fprintf(stderr,"we need one argument: a dotted decimal IP address\n");

    exit(1);
  }

  if(inet_aton(argv[1],&inxyz.sin_addr)==0)
  {
    fprintf(stderr,"invalid address\n");

    exit(1);
  }

  printf("Network-ordered address from inet_aton(), printed as a CPU-ordered integer:\n"
         "0x%08X\n",
         inxyz.sin_addr.s_addr
        );

  printf("Dotted decimal representation from inet_ntoa():\n%s\n",
         inet_ntoa(inxyz.sin_addr)
        );

  if(inet_pton(AF_INET,argv[1],&inxyz.sin_addr)==0)
  {
    fprintf(stderr,"invalid address\n");

    exit(1);
  }

  printf("Network-ordered address from inet_pton(), printed as a CPU-ordered integer:\n"
         "0x%08X\n",
         inxyz.sin_addr.s_addr
        );

  printf("Dotted decimal representation from inet_ntop():\n%s\n",
         inet_ntop(AF_INET,&inxyz.sin_addr,ntop_buffer,INET_ADDRSTRLEN)
        );

  return 0;

} /* main() */

Another point: You can actually call gethostbyname() instead if you want. It returns four-byte network-ordered addresses, and you can give it either "www.google.com" or "66.102.7.147". It will work either way.

Finally, you'll notice in the source code that I used variable inxyz, not in as you had in your example. When choosing the name of a variable, it's always better to choose something that's not a substring of some other identifier. Why? Because if you use your text editor to search for all instances of variable in, for example, you'll also find many other uninteresting occurrences of "in".

leamassiot · 03-09-2007, 08:40 AM

Thanks a lot to you

,
it is exactly what I was searching for and it works

Thanks for the code and the comments.

Good day in Mariposa.