rsync - Feb 2002 - Debian bug #128632 && fork

Why the sleep() call?

On 16 Feb 2002, Colin Walters <walters@debian.org>
wrote:
> Index: socket.c
> ==================================================================> RCS
file: /cvsroot/rsync/socket.c,v
> retrieving revision 1.74
> diff -u -u -r1.74 socket.c
> +++ socket.c	16 Feb 2002 23:08:47 -0000
> @@ -387,6 +387,7 @@
>  	   for each incoming connection */
>  	while (1) {
>  		fd_set fds;
> +		pid_t pid;
>  		int fd;
>  		struct sockaddr_storage addr;
>  		socklen_t addrlen = sizeof addr;
> @@ -418,12 +419,17 @@
>                  while (waitpid(-1, NULL, WNOHANG) > 0);
>  #endif
>  
> -		if (fork()==0) {
> +		if ((pid = fork()) == 0) {
>  			close(s);
>  			/* open log file in child before possibly giving
>  			   up privileges  */
>  			log_open();
>  			_exit(fn(fd));
> +		} else if (pid < 0) {
> +			rprintf(FERROR, "could not create child process: %s\n",
> +				strerror(errno));
> +			close(fd);
> +			sleep(2);
>  		}
>  
>  		close(fd);
-- 
Martin

On Mon, 18 Feb 2002, Martin Pool wrote:
> Why the sleep() call?
Also, why close(fd) twice?
> > +		} else if (pid < 0) {
> > +			rprintf(FERROR, "could not create child process: %s\n",
> > +				strerror(errno));
> > +			close(fd);
> > +			sleep(2);
> >  		}
> >
> >  		close(fd);
..wayne..

On 18 Feb 2002, Wayne Davison <wayned@users.sourceforge.net>
wrote:
> On Mon, 18 Feb 2002, Martin Pool wrote:
> > Why the sleep() call?
I guess the point is that the fork() probably failed because the
server is overloaded, and therefore there is no point trying to accept
another connection again immediately.  I just wanted to check this was
the reason.
> Also, why close(fd) twice?
-- 
Martin