on inet_ntoa in hp-ux

[Vishal Augustine]

Hi

After an accept system call, when I perform an inet_ntoa, I get the result as "0.0.0.0" ? Why is it so ? The same code works fine in Linux with one exception. For the 1st accept it gives "0.0.0.0" and later it shows proper value !!!!! (Linux RedHat 6.2 Kernel 2.2.14-12 /glibc2.1.3-15)

Below is the portion of the sample program. Program was compiled using g++

struct sockaddr_in acc_addr;
memset(&acc_addr,0,sizeof(acc_addr));
if((sad = accept(s_sd,(struct sockaddr *)&acc_addr, &len)) < 0)
{
exit(1);
}

if(!fork())
{
unsigned char readbuff[SOCK_BUFF_SIZE];
cout << inet_ntoa(acc_addr.sin_addr)) << endl;
close(sad)
close(s_sd)
exit(0);
}


Thanks and Regards
Vishal

[Vishal Augustine]

Forgot to give the hp details

HP-UX 11.0
aCC 3.3 compiled

Thanks and Regards
Vishal

[Vishal Augustine]

Got the answer !!!

Thought will share with u ...

The 'len' value passed to 'accept' was not set. Hence the sockaddr_in structure - acc_addr was not getting populated properly.

An interesting aspect is .. it works fine in Linux !! If len is set to zero and we pass acc_addr, the len value is populated but not the acc_addr. In HP-UX , we need to let 'accept' know the value of len.

So the solution is ..

len = sizeof(acc_addr);
if((sad = accept(s_sd,(struct sockaddr *)&acc_addr, &len)) < 0)
{
exit(1);
}

Thanks and Regards
Vishal

[rick jones]

that Linux is happy to fill-in what you have told it is a zero-length sockaddr implies that Linux may not be doing proper parameter checking on the accept() call...

[Vishal Augustine]

I am not a strict guy ... so I wud not say what linux is doing is bad

Vishal