[ic] Server.pm

[Rich Bodo]

Trying to get acquainted with the code.  Got a few developer
questions.

First, is there a developers mailing list?

Second, does anyone meet on IRC anymore?  I hang out on irc.cursor.biz
every day until the camper bot kicks me off, waiting for someone to
show.

Third, an actual question about the code: Why are two children forked
in Server.pm, run_server()?  I'm looking at the following code from
the 4.8 version of Server.pm:

        fcntl($pidh, F_SETFD, 0)
            or die ::errmsg(
			"Can't fcntl close-on-exec flag for '%s': %s\n",
			$Global::PIDfile, $!,
			);
        my ($pid1, $pid2);
        if ($pid1 = fork) {
            # parent
            wait;
			sleep 2;
            exit 0;
        }
        elsif (not defined $pid1) {
            # fork error
            print "Can't fork: $!\n";
            exit 1;
        }
        else {
            # child 1
            if ($pid2 = fork) {
                # still child 1
                exit 0;
            }
            elsif (not defined $pid2) {
                print "child 1 can't fork: $!\n";
                exit 1;
            }
            else {
                # child 2
                sleep 1 until getppid == 1;

                my $running = grab_pid($pidh);
                if ($running) {
                    print errmsg(
		"The Interchange server is already running (process id %s)\n",
						$running,
						);
                    exit 1;
                }
                print server_start_message(
		"Interchange server started in %s mode(s) (process id %s)\n",
						1,
					 ) unless $Vend::Quiet;

                setsid();

                fcntl($pidh, F_SETFD, 1)
                    or die "Can't fcntl close-on-exec flag for '$Global::PIDfile': $!\n";

				$next = server_both();

				unlockfile($pidh);
				opendir(RUNDIR, $Global::RunDir)
					or die "Couldn't open directory $Global::RunDir: $!\n";
				unlink $Global::PIDfile;
                exit 0;
            }
        }
    }

It appears to me that the above code basically does the following:
	   1) set close on exec of the current server (parent) to one
	   2) fork off child process 1 (child1)
	   3) parent waits for child1 to terminate then exits
	   4) child1 forks child2 and child1 exits immediately
	   5) child2 waits for it's parent to die then calls
	   server_both()

So the parent, will hang around until a grandkid to exist, then both
the parent and child will exit.  Is my analysis incorrect, or is there
a reason for the family tree here?

BTW, I have taken it upon myself to understand the workings of
Interchange.  If I can get access to just one interchange developer
for infrequent coding questions like this I will pay them back
with documentation and code in spades.  Thanks.

-Rich

Rich Bodo | rsb@ostel.com | 650-964-4678