--- loncom/loncnew	2004/10/04 10:30:50	1.62
+++ loncom/loncnew	2004/10/05 10:10:31	1.64
@@ -2,7 +2,7 @@
 # The LearningOnline Network with CAPA
 # lonc maintains the connections to remote computers
 #
-# $Id: loncnew,v 1.62 2004/10/04 10:30:50 foxr Exp $
+# $Id: loncnew,v 1.64 2004/10/05 10:10:31 foxr Exp $
 #
 # Copyright Michigan State University Board of Trustees
 #
@@ -302,6 +302,44 @@ sub SocketTimeout {
     }
 
 }
+#
+#   This function should be called by the child in all cases where it must
+#   exit.  If the child process is running with the DieWhenIdle turned on
+#   it must create a lock file for the AF_UNIX socket in order to prevent
+#   connection requests from lonnet in the time between process exit
+#   and the parent picking up the listen again.
+# Parameters:
+#     exit_code           - Exit status value, however see the next parameter.
+#     message             - If this optional parameter is supplied, the exit
+#                           is via a die with this message.
+#
+sub child_exit {
+    my ($exit_code, $message) = @_;
+
+    # Regardless of how we exit, we may need to do the lock thing:
+
+    if($DieWhenIdle) {
+	#
+	#  Create a lock file since there will be a time window
+	#  between our exit and the parent's picking up the listen
+	#  during which no listens will be done on the
+	#  lonnet client socket.
+	#
+	my $lock_file = GetLoncSocketPath().".lock";
+	open(LOCK,">$lock_file");
+	print LOCK "Contents not important";
+	close(LOCK);
+	
+	exit(0);
+    }
+    #  Now figure out how we exit:
+
+    if($message) {
+	die $message;
+    } else {
+	exit($exit_code);
+    }
+}
 #----------------------------- Timer management ------------------------
 
 =pod
@@ -332,18 +370,8 @@ sub Tick {
 	    $IdleSeconds = 0;	# Otherwise all connections get trimmed to fast.
 	    UpdateStatus();
 	    if(($ConnectionCount == 0) && $DieWhenIdle) {
-		#
-		#  Create a lock file since there will be a time window
-		#  between our exit and the parent's picking up the listen
-		#  during which no listens will be done on the
-		#  lonnet client socket.
-		#
-		my $lock_file = GetLoncSocketPath().".lock";
-		open(LOCK,">$lock_file");
-		print LOCK "Contents not important";
-		close(LOCK);
-		
-		exit(0);
+		&child_exit(0);
+
 	    }
 	}
     } else {
@@ -1177,7 +1205,7 @@ sub MakeLondConnection {
 	#
 	my $Socket = $Connection->GetSocket();
 	if($Socket eq undef) {
-	    die "did not get a socket from the connection";
+	    &child_exit(-1, "did not get a socket from the connection");
 	} else {
 	    &Debug(9,"MakeLondConnection got socket: ".$Socket);
 	}
@@ -1491,7 +1519,11 @@ sub SetupLoncListener {
     unless ($socket =IO::Socket::UNIX->new(Local  => $SocketName,
 					    Listen => 250, 
 					    Type   => SOCK_STREAM)) {
-	die "Failed to create a lonc listner socket";
+	if($I_am_child) {
+	    &child_exit(-1, "Failed to create a lonc listener socket");
+	} else {
+	    die "Failed to create a lonc listner socket";
+	}
     }
     return $socket;
 }
@@ -1621,8 +1653,8 @@ sub ChildProcess {
 	undef         $parent_dispatchers{$listener};
 
     }
-    $I_am_child    = 1;		# Seems like in spite of it all I'm still getting
-                                # parent event dispatches. 
+    $I_am_child    = 1;		# Seems like in spite of it all I may still getting
+                                # parent event dispatches.. flag I'm a child.
 
 
     #
@@ -1679,7 +1711,7 @@ sub ChildProcess {
     my $ret = Event::loop();		#  Start the main event loop.
     
     
-    die "Main event loop exited!!!";
+    &child_exit (-1,"Main event loop exited!!!");
 }
 
 #  Create a new child for host passed in:
@@ -1716,7 +1748,6 @@ sub CreateChild {
 #    a connection request arrives.  We must:
 #     Start a child process to accept the connection request.
 #     Kill our listen on the socket.
-#     Setup an event to handle the child process exit. (SIGCHLD).
 # Parameter:
 #    event       - The event object that was created to monitor this socket.
 #                  event->w->fd is the socket.
@@ -1821,6 +1852,39 @@ sub listen_on_all_unix_sockets {
     }
 }
 
+#   server_died is called whenever a child process exits.
+#   Since this is dispatched via a signal, we must process all
+#   dead children until there are no more left.  The action
+#   is to:
+#      - Remove the child from the bookeeping hashes
+#      - Re-establish a listen on the unix domain socket associated
+#        with that host.
+# Parameters:
+#    The event, but we don't actually care about it.
+sub server_died {
+    &Debug(9, "server_died called...");
+    
+    while(1) {			# Loop until waitpid nowait fails.
+	my $pid = waitpid(-1, WNOHANG);
+	if($pid <= 0) {
+	    return;		# Nothing left to wait for.
+	}
+	# need the host to restart:
+
+	my $host = $ChildHash{$pid};
+	if($host) {		# It's for real...
+	    &Debug(9, "Caught sigchild for $host");
+	    delete($ChildHash{$pid});
+	    delete($HostToPid{$host});
+	    &parent_listen($host);
+
+	} else {
+	    &Debug(5, "Caught sigchild for pid not in hosts hash: $pid");
+	}
+    }
+
+}
+
 #
 #  Parent process logic pass 1:
 #   For each entry in the hosts table, we will
@@ -1893,6 +1957,14 @@ ShowStatus("Parent keeping the flock");
 
 
 if ($DieWhenIdle) {
+    # We need to setup a SIGChild event to handle the exit (natural or otherwise)
+    # of the children.
+
+    Event->signal(cb       => \&server_died,
+		   desc     => "Child exit handler",
+		   signal   => "CHLD");
+
+
     $Event::DebugLevel = $DebugLevel;
     Debug(9, "Parent entering event loop");
     my $ret = Event::loop();