Very Computer

Even my lecturer can't figure this 1 out.... help!  Any resolver gurus
know what's happening?

I've got 2 samples. 1st sample was written by my lecturer to display the
contents of hosent structure.  The 2nd sample is my game with his code
(juz that 1 line with gethostbyname()) ported over.  It works perfectly
fine with the 1st sample - any text string gets accepted by
gethostbyname().

gethostbyname() causes segmentation fault in the 2nd sample though.
I've tried other alternative ways like assigning to a new char ptr,
copying to a new buffer space, & even hardcoding the string into the
function call.  all will still gimme segmentation fault.  The error
pattern I noticed is strings above 11 or 12 chars will produce the fault
:

"brady.sgt.g" will be accepted & actually queries the DNS server (I can
see it thru the modem). This is generate the "Server host invalid" msg
of course.

"brady.sgt.gm" will cause segmentation fault immediately, without
sending any DNS query packet.

if I try "131.181.112.1" gethostbyname() accepts it, but returns invalid
host. I thought gethostbyname() is able to return a hostent struct with
the same addr?  giving "131.181.112.2" will produce segmentation fault
again, & "131.181.112.3" will get invalid host, & so on.

Anybody knows how gethostbyname manipulates the string?

#1
int main(int argc, char *argv[])
{
        char *ptr, **pptr, str[INET_ADDRSTRLEN];

        struct hostent  *hptr;

        while (--argc > 0)
        {
                ptr = *++argv;
                printf("%s", ptr);
                fflush(stdout);
                if ((hptr = gethostbyname(ptr)) == NULL)
                {
                        fprintf(stderr, " not valid hostname?\n", ptr);
                        continue;
                }
                printf(" official hostname : %s\n", hptr->h_name);
...........
...........

#2
void start_player(char name[NAME + 1], char *servername, uint16_t port,
int pos[2])
{
   int sockd;
   struct sockaddr_in serversock;
   struct hostent *serverhost;

        /* test param string */
   printf("HMS %s to enter %s:%d at %d long %d lat\n", name, servername,
port, pos[0], pos[1]);

   if ((serverhost = gethostbyname(servername)) == NULL)
   {
      fprintf(stderr, "Server host invalid.\n");
      exit(PARAM_FAULT);
   }

        /* test string */
   printf("Set server socket now...\n");
   bzero(&serversock, sizeof(serversock));
   inet_pton(AF_INET, serverhost->h_addr_list[0], &serversock.sin_addr);
   serversock.sin_family = AF_INET;
   serversock.sin_port = htons(port);
..........
..........

trapped,
Aaron

Oops, extra parameter without matching conversion specifier.

Quote:>                        continue;
>                }
>                printf(" official hostname : %s\n", hptr->h_name);
>...........
>...........

>#2
>void start_player(char name[NAME + 1], char *servername, uint16_t port,

                   ^^^^^^^^^^^^^^^^^^^

You know that this is just a pointer to char, right? The array notation only
has documentary value.

Quote:>int pos[2])

Don't use uint16_t unless you for some reason need a type that is
exactly 16 bits wide. The types type int or short are more than
adquate for representing a port number.

The uint16_t type is a bleeding edge C9X feature, not yet part of
standard C. It is defined in <inttypes.h>, but an implementation doesn't
have to provide it (what if it the hardware have a 16 bit type?)

Quote:>{
>   int sockd;
>   struct sockaddr_in serversock;
>   struct hostent *serverhost;

>        /* test param string */
>   printf("HMS %s to enter %s:%d at %d long %d lat\n", name, servername,
>port, pos[0], pos[1]);

If you want to printf() a uint16_t, you have to use a macro
from <inttypes.h> to supply a conversion specifier.

        #include <inttypes.h>

        uint16_t u;

        /*...*/

        printf("u = " PRIu16, u);

The PRIu16 macro expands to a little string literal containing the
implementation's appropriate conversion specifier that matches a uint16_t
parameter. Assuming the implementation supports a uint16_t of course.

The new <inttypes.h> interface gives the programmer a way to use the
preprocessor to detect whether exact-sized types are available.
If a type is not available, then there is no typedef for it and
its corresponding macros are absent.  So you can do

        #include <inttypes.h>
        #ifdef UINT16_MAX
        /* uint16_t type is available! */
        #else
        /* uint16_t is not available; provide some alternate proram text
           or use #error to indicate that the program can't work. */
        #endif

Quote:>   if ((serverhost = gethostbyname(servername)) == NULL)
>   {
>      fprintf(stderr, "Server host invalid.\n");
>      exit(PARAM_FAULT);
>   }

You say that the gethostbyname() function crashes. It's impossible to know whyt
his happens because you have not revealed enough context information. The call
simply repeats the servername parameter, but you have not shown how the data
for that parameter came to be.  Without delving into my psychic powers, I can
only guess that maybe it is not a properly null terminated string!

You said that the crash is sensitive to the number of characters in
the string. Perhaps your string buffer doesn't have enough space.
It's easy to misuse a function like strncpy() to cause this kind of
error. For example:

        char host_name[16];

        /*...*/
        strncpy(host_name, source_string, sizeof host_name);

In this case, if the source_string is 16 characters or longer, then surely
enough, only 16 bytes get copied into host_name, but host_name then does not
contain a null terminated string!

Or worse!

        strcpy(host_name, source_string);       /* buffer overflow */