I've had reports of rsyncd leaving zombies under digital unix. This

patch tries to address the problem in two ways:

1) reinstall the SIGCHLD handler before each fork
2) reap any children not caught by the handler using waitpid with
WNOHANG.

I expect this will fix the problem.

diff --git a/socket.c b/socket.c
index 790a86d..98093fc 100644 (file)
--- a/socket.c
+++ b/socket.c
@@ -120,8 +120,6 @@ void start_accept_loop(int port, int (*fn)(int ))
 {
        int s;
 
-       signal(SIGCHLD, SIG_IGN);
-
        /* open an incoming socket */
        s = open_socket_in(SOCK_STREAM, port);
        if (s == -1)
@@ -155,6 +153,15 @@ void start_accept_loop(int port, int (*fn)(int ))
 
                if (fd == -1) continue;
 
+               signal(SIGCHLD, SIG_IGN);
+
+               /* we shouldn't have any children left hanging around
+                  but I have had reports that on Digital Unix zombies
+                  are produced, so this ensures that they are reaped */
+#ifdef WNOHANG
+               waitpid(-1, NULL, WNOHANG);
+#endif
+
                if (fork()==0) {
                        close(s);