fixup 51602f5

lightningd/lightningd.c

@@ -782,12 +782,22 @@ static struct io_plan *sigchld_rfd_in(struct io_conn *conn,
 	static u8 ignorebuf;
 	static size_t len;
 	pid_t pid;
-
-	/* Reap any terminated child we can; for the rest we ignore these */
-	while ((pid = waitpid(-1, 0, WNOHANG)) != 0)
-		/* No child processes: this causes a hang sometimes, so break that */
-		if (pid == -1 && errno == ECHILD)
+	int status;
+
+	/* Reap any terminated child we can; for the rest we ignore these.
+	 * This elaborate while loop appears to solve a rare hang issue in the
+	 * case all children are gone (somehow ?). */
+	while (true) {
+		pid = waitpid(-1, &status, WNOHANG);
+		if (pid == 0)
+			break;
+		if (pid == -1){
+			log_debug(ld->log, "waitpid returned -1: errno %s",
+				   strerror(errno));
 			break;
+		} else
+			log_unusual(ld->log, "reaped child pid %i", pid);
+	}
 
 	return io_read_partial(conn, &ignorebuf, 1, &len, sigchld_rfd_in, ld);
 }