Berkeley Socket : recv system call - c++

Question

//server side
void* s2(void *arg){
info *s = (info*)arg;
char buffer[MAXS];
int k;
sockaddr_in addr;
socklen_t aSize = sizeof(sockaddr_in);
int sfd = accept(s->fd,(sockaddr*)&addr,(socklen_t*)&aSize);
if(sfd<0){
s->current--;
pthread_exit(0);
}
while(1){
k = recv(sfd,buffer,MAXS,0);
cout<<buffer<<"\n";
//1. k ==0 socket has been closed by client
//2. k==-1 error in recv
//3. recv quit
if((k==-1)||(!strncmp(buffer,"quit",4))||(k==0))break;
sprintf(buffer,"%d\n",(int)strlen(buffer)); //Convert length to string using sprintf()
send(sfd,buffer,strlen(buffer),0); //send buffer to client
}
close(sfd);
if(s->limit==s->current)
FD_SET(s->fd,&sfds);
s->current--; //decreament the client number
pthread_exit(0);
}
//client side
1. send(sockfd,"sadhdag",8,0);
2. send(sockfd,"ss",3,0);
Server recv sadhdag in first call to recv .
In second call to recv server recv
ss
dag
server side:
function s2 is run by a thread and argument is passed with socket information connection is accepted there and send and recv is called onn newly accepted client.
Why this is happening ? or how to avoid this ?

Answer 1

You're ignoring the count returned by recv(). Contrary to suggestions in comments, it isn't necessary to memset() the buffer to zero before recv(), but it is necessary to only use that buffer afterwards up to that count. For example:
printf("%s", buffer);
is wrong, and
printf("%.*s", count, buffer);
is right.
NB
if((k==-1)||(!strncmp(buffer,"quit",4))||(k==0))break;
isn't correct either. It should be
if((k==-1)||(k==0)||(k >= 4 && !strncmp(buffer,"quit",4))) break;
It isn't valid to look into the buffer at all unless k is positive, and it isn't valid to compare four characters in it unless there are four characters in it.

Answer 2

You forgot to actually implement any kind of protocol or application-level logic of any kind. This sums it up:
k = recv(sfd,buffer,MAXS,0);
cout<<buffer<<"\n";
This code suggests you're terminating your messages with a newline:
sprintf(buffer,"%d\n",(int)strlen(buffer)); //Convert length to string using sprintf()
send(sfd,buffer,strlen(buffer),0); //send buffer to client
If so, where's the code to parse that on the other end? You throw the returned length away, so you wouldn't even know which characters to check for a newline.
TCP is a byte-stream protocol that does not preserve application-level message boundaries. If you want to use it to send strings or other application-level messages, you have to specify and implement an application-level protocol to do it.
You call recv to get the raw TCP data, then you output it as if it was a string. If you're going to define and implement a protocol to send and receive strings over TCP, you have to actually do it. You can't just imagine it happens by magic.
If you want to receive newline-terminated messages into C-style strings, you can certainly do it. But you have to actually write the code to do it.