在C中使用wait（）vs waitpid（）

Question

So I'm trying to traverse a directory(and subdirectories) and create new processes to sort files and traverse subdirectories. However, I'm having a little trouble understanding how useful my code will be. To my understanding, wait() will sleep the parent process until the child process terminates. Currently my recursive function is

void iterate_dir(char* name){

    DIR *dd = opendir(name);
    struct dirent *curr;

    while((curr = readdir(dd))!=NULL){

        if((strcmp(curr->d_name,".")==0 || strcmp(curr->d_name,"..")==0) && curr->d_type==DT_DIR){

            continue;

        }

        int status = -1;
        int pid = fork();

        if(pid==0){

            //child process
            if(curr->d_type==DT_DIR){

                printf("DIRECTORY:\t%s\n", curr->d_name);
                char new_path[strlen(curr->d_name)+strlen(name)+2];
                sprintf(new_path,"%s/%s",name,curr->d_name);
                //recurse, iterate sub directory
                iterate_dir(new_path);
                _exit(getpid());


            }else{

                printf("FILE:\t%s\n", curr->d_name);
                //sort the file
                _exit(getpid());

            }


        }

        wait(&status);

    }

    closedir(dd);

}

given and initial directory, it works, but im concerned with the wait() function. I would like the code to continue traversing the directory while the child processes are executing, and currently wait prevents this from happening. But I still need to prevent zombie children. Would using waitpid() instead allow me to have this functionality(ie. allow the loop to continue through the rest of the directory while the child process executes), can I just use 1 wait in main which will prevent all processes created from becoming zombies, or is there a different approach I should take? This is for a school project and I'm required to use fork (not exec) to create a new process for traversing sub directories and a new process for sorting each file.

Answer 1

No, changing to waitpid() won't solve the problem, since it will still suspend the loop until the child exits.

Take the wait() out of the main loop and do it at the end:

while (wait() > 0) {
}

wait() will wait for any child processes, and return -1 if there are no children to wait for.

Answer 2

For that code to be really useful, there can only be one process running each time, else the output of several simultaneously running processes will mix in the standard output.

So wait() is a good choice, since you want each process to wait for the child to end before launching a new child process or end its execution.

In a different case where you want different children processes to be run at the same time you can use waitpid(-1, &status, WNOHANG) , that will (not) wait for a child process to end, ie it will not hang if no child process has finished but it will prevent any finished child process from becoming a zombie.