Exiting Endless Loop

Ok , so I have a question. I wrote some programs for some assignments where I have a server and a client/clients and they interact. One was a bulletin board system (very basic and primitive of course) the other a mail server-client and the other a chat...

I completed them but there is one thing bugging me; In all these the server goes into an endless loop where he accepts connections and creates new threads or processes that complete tasks. So my question is this: After the endless for loop I have some statements that detach shared memory from processes and delete it or destroys thread attributes and mutexes. But how do I get these commands to execute. I mean the only way to terminate the server is by ctrl-c. But when I press that doesnt the whole program closes at that point? I thought about doing it like normal programs - using a variable - and when catching the signal of ctrl-c after asking the user whether to exit the server or resume setting the variable so I can break from the for-loop i.e

Code:

---

for (;;)    {
...
    if (c) break;
...
}

---

but I want to end it instantly. With this way the program will most definitely accept another connection and THEN terminate because after a connection is accepted it will loop back and block at the accept statement waiting for an incoming connection.. So any ideas? [This is strictly curiosity from my part I dont need it right now for anything]

My code in the above always looks something like this

Code:

---

int main()    {
...
...
...
for (;;)    {
    sd = accept(...);
    ...
    if (!fork())    {
        child process here
    }
    parent here
}//I am talking about the following statements
sem_unlink(...);
shmctl(...);
}

---

or

Code:

---

int main()    {
...
...
...
for (;;)    {
    sd = accept(...);
    ...
    pthread_create(...);

}//I am talking about the following statements
pthread_attr_destroy(...);
pthread_mutex_destroy(...);       

}

---

So m this is it.

Right, so I presume the real problem here is that you don't know what state the server is in, or where in the endless loop your server happens to be?

In that case, it's quite difficult to solve the problem.

Using a signal handler to let your code leave through a function that cleans up is one possible approach.

Another approach would be to have a your code accept a particular type of packet from a special list of hosts that can send a "kill message" to the socket itself. You obviously need to restrict the hosts that can do this to the ones the you trust and control, otherwise someone from the outside can down your server, which is a bad idea.

--
Mats

Hmm, that's a good idea. Thanks matsp. Expanding on that, how about sending a "kill message" to itself. That could work... I think

Quote:

---

Originally Posted by Phoenix_Rebirth

Hmm, that's a good idea. Thanks matsp. Expanding on that, how about sending a "kill message" to itself. That could work... I think

---

Well, sending a kill message to itself only works if you have a way to tell it to do so - which you may. That would depend a lot on the actual server and it's purpose.

--
Mats

(if possible) it is usually better to not use infinate loops, instead defining the terminating condition. This also saves using an if and logically should manage memory better (even though the program will probably free the memory anyway)