BF: safeguard closing of log handlers + close in reverse order

otherwise there might be "stuck" handler in the queue. and closing
exceptions can occur -- even stock logging guards in recent versions

server/server.py

@@ -367,11 +367,20 @@ class Server:
 					logSys.error("Unable to log to " + target)
 					logSys.info("Logging to previous target " + self.__logTarget)
 					return False
-			# Removes previous handlers
-			for handler in logging.getLogger("fail2ban").handlers:
-				# Closes the handler.
+			# Removes previous handlers -- in reverse order since removeHandler
+			# alter the list in-place and that can confuses the iterable
+			for handler in logging.getLogger("fail2ban").handlers[::-1]:
+				# Remove the handler.
 				logging.getLogger("fail2ban").removeHandler(handler)
-				handler.close()
+				# And try to close -- it might be closed already
+				try:
+					handler.flush()
+					handler.close()
+				except ValueError:
+					if sys.version_info >= (2,6):
+						raise
+					# is known to be thrown after logging was shutdown once
+					# with older Pythons -- seems to be safe to ignore there
 			# tell the handler to use this format
 			hdlr.setFormatter(formatter)
 			logging.getLogger("fail2ban").addHandler(hdlr)