Unix process API

The two most important function calls to use when programming with several processes are fork and exec:

 fork() creates a copy of current process. It gives a different return value to each process and works based on Copy On Write;

 exec() replaces a process with an executable.

(The Windows CreateProcess(...), taking ten arguments, is analogous.)

Notice that fork() implies that each process descends from another process. In fact, in Unix everything descends from a single process called init: basically, init forks a process and then "replaces its code" with, say, the code of bash, using exec().

Example of how to use fork:
#include
#include
#include
int parentid = getpid();
char program_name[1024];
gets(program_name); // reads the name of program we want to start
int cid = fork();
if (cid==0) { // i'm the child
execlp(program_name, program_name, 0); // loads the program and runs it
printf("if the above worked, this line will never be reached\n");
}
else { // i'm the parent
sleep (1); // give my child time to start
waitpid(cid, 0, 0); // waits for my child to terminate
print("program %s finished\n", program_name);
}
Is the sleep(1) call necessary to allow the child process to start? The answer is no, it is not at all necessary. In general, if you think you need to sleep in a program, you are probably doing something wrong, and just slowing down your program. The call to waitpid() is a blocking wait, and will ?rst wait to let the child process start (if it hasn't already), then will wait until it ends.