handling exits: main calls a function which creates processes

Code:

void call_me(){
	pid_t child;
	
	if((child=fork()) < 0){
		exit(-1);
	}
	else if(child==0){
		execl("./someexample","./someexample",(char*)0);
		exit(1);		
	}
	else{
		int status;
		pid_t   waitpid = wait(&status);
		exit(0);
	}
}

main(){
	
   call_me();
}

Question I have is where do I check the exit(-1), exit(1), exit(0)? Inside main? or check exit(-1),exit(1) inside parent process of call_me and check exit(0) inside main?

1. If any of the exit() are called, there is no return to main.
2. If execl() is successful, exit(1) never happens.

>1. If any of the exit() are called, there is no return to main.

Salem, I don't understand what you mean by no return to main.
For example, main had other statements to execute after "call_me" and call_me was called. Inside call_me, one of the exit occurs, does it mean it will never came back to main. Main program will crash or call_me processes parent-child crash, or call_me memory stack will be removed from memory and main function continues?

Originally Posted by 911help

>1. If any of the exit() are called, there is no return to main.

When exit() is called, the program terminates. Any statement after exit want be executed at all. If you want to pass values you may use return(). e.g:

Code:

int test(){
 return(1);
}

int main() {
  int i;

  if(test() == 1)
     printf("test returned 1\n");
  else
     printf("test returned not 1\n");

  exit(0);
}

When you fork a process, what memory is passed to the forked process (whole program memory or in the above example, only "call me" function's memory stack is passed to the forked process?)

Each process runs in its own virtual memory space - it inherits nothing from the process that launched it.

Originally Posted by Elysia

Each process runs in its own virtual memory space - it inherits nothing from the process that launched it.

fork() causes a copy of the current process to be made, including the value of all variables.

Since everything is virtual, in terms of addressing, from the application point of view, you're dealing with an exact copy that points to the same variables, which of course you're not. The value of all variables are copied.

Case in point:

Code:

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/types.h>
#include <sys/wait.h>
#include <unistd.h>

int main(void)
{
        size_t i;
        pid_t p = fork();
        char *szMsg = "Testing!";

        if(p == 0)
        {
                /* Child */
                printf("szMsg is at address: &#37;p\n", szMsg);
                for(i=0;i<strlen(szMsg);i++)
                {
                        printf("szMsg[%d] = [%d, %c]\n", i, szMsg[i], szMsg[i]);
                }
        }
        else if(p > 0)
        {
                /* Parent */
                wait(NULL);
                printf("szMsg is at address: %p\n", szMsg);
                for(i=0;i<strlen(szMsg);i++)
                {
                        printf("szMsg[%d] = [%d, %c]\n", i, szMsg[i], szMsg[i]);
                }
        }
        else
        {
                fprintf(stderr, "Error.  Unable to fork.\n");
                return 1;
        }

        return 0;
}

Output:

Code:

>gcc -W -Wall -ansi -pedantic -O2 fork.c
>a.out
szMsg is at address: 108d8
szMsg[0] = [84, T]
szMsg[1] = [101, e]
szMsg[2] = [115, s]
szMsg[3] = [116, t]
szMsg[4] = [105, i]
szMsg[5] = [110, n]
szMsg[6] = [103, g]
szMsg[7] = [33, !]
szMsg is at address: 108d8
szMsg[0] = [84, T]
szMsg[1] = [101, e]
szMsg[2] = [115, s]
szMsg[3] = [116, t]
szMsg[4] = [105, i]
szMsg[5] = [110, n]
szMsg[6] = [103, g]
szMsg[7] = [33, !]
>

Even though the addresses match, you're not dealing with the same exact memory space (because it's virtual).

So I would argue you are actually inheriting everything, or most of everything, from the parent process. In this context, "inheriting" means all variables retain their same values. In some cases, such as file descriptors, you actually are pointing to the same files, but this is somewhat indepth to get into atm.

Ah, so fork does copy mostly everything from the parent process to the child process? I would never have imagined such a thing.

I would recommend you try some *nix programming some time. I think you might enjoy it.

Been thinking about linux, but hardware support and all the different versions (for lack of a better name) and all that has steered my away all this time.

You might be able to try cygwin if you want to stay in the safety of your Windows machine, but practice with the *nix system calls. It provdes wrappers for them, so you're still able to get into a bit of the mindset of how things should run on Unix or Linux. It may not be the best if you really wanted to get into it, but it might be fun to try.

Perhaps this as well.
http://www.colinux.org/

Originally Posted by Salem

Perhaps this as well.
http://www.colinux.org/

Every time I try to install this Cygwin it download the package and all goes right and i can get through the installation process as well. But when I try to launch the application, doom the console comes and disappears. And found that one of those Cygiconv-2.dll files and many are missing and on top it says reinstalling will solve this problem.

Ohh, i have re-installed million times by now lol, no success I gave up. I will try Colinux now. Buy the way have anyone had this problem before.

ssharish

> Every time I try to install this Cygwin it download the package...
No, works every time for me.

Have you got the latest setup.exe from cygwin.com ?
Do you run setup.exe as admin?

So what's the final verdict?

Program memory is copied, not function memory stack, even when fork is called from within a function?