diff --git a/pubs/pubs_cmd.py b/pubs/pubs_cmd.py
index e849f69..e9c8b46 100644
--- a/pubs/pubs_cmd.py
+++ b/pubs/pubs_cmd.py
@@ -83,13 +83,17 @@ def execute(raw_args=sys.argv):
 
         # Eventually autocomplete
         autocomplete(parser)
+
         # Parse and run appropriate command
+        # if no command, print help and exit peacefully (as '--help' does)
         args = parser.parse_args(remaining_args)
+
         if not args.command:
             parser.print_help()
-        else:
-            args.prog = "pubs"  # FIXME?
-            args.func(conf, args)
+            sys.exit(0)
+
+        args.prog = "pubs"  # FIXME?
+        args.func(conf, args)
 
     except Exception as e:
         if not uis.get_ui().handle_exception(e):
diff --git a/tests/test_usecase.py b/tests/test_usecase.py
index ce870cf..7ce85fc 100644
--- a/tests/test_usecase.py
+++ b/tests/test_usecase.py
@@ -29,7 +29,7 @@ PRINT_OUTPUT   = False
 CAPTURE_OUTPUT = True
 
 
-class FakeSystemExit(Exception):
+class FakeSystemExit(SystemExit):
     """\
     SystemExit exceptions are replaced by FakeSystemExit in the execute_cmds()
     function, so they can be catched by ExpectedFailure tests in Python 2.x.
@@ -131,9 +131,9 @@ class CommandTestCase(fake_env.TestFakeFs):
             exc_class, exc, tb = sys.exc_info()
             if sys.version_info.major == 2:
                 # using six to avoid a SyntaxError in Python 3.x
-                six.reraise(FakeSystemExit, exc, tb)
+                six.reraise(FakeSystemExit, *exc.args, exc_traceback=tb)
             else:
-                raise FakeSystemExit(exc).with_traceback(tb)
+                raise FakeSystemExit(*exc.args).with_traceback(tb)
 
     def tearDown(self):
         pass
@@ -187,6 +187,26 @@ class URLContentTestCase(DataCommandTestCase):
 
 # Actual tests
 
+class TestAlone(CommandTestCase):
+
+    def test_alone_prints_help(self):
+        # capturing the output of `pubs --help` is difficult because argparse
+        # raises as SystemExit(0) after calling `print_help`, and this gets
+        # caught so no output is captured.  so comparing outputs of `pubs` and
+        # `pubs --help` isn't too easy unless substantially reorganization of
+        # the parser and testing context is made.  instead, the exit codes of
+        # the two usecases are compared.
+        self.execute_cmds(['pubs init'])
+
+        with self.assertRaises(FakeSystemExit) as cm1:
+            self.execute_cmds(['pubs'])
+
+        with self.assertRaises(FakeSystemExit) as cm2:
+            self.execute_cmds(['pubs --help'])
+
+        self.assertEqual(cm1.exception.code, cm2.exception.code, 0)
+
+
 class TestInit(CommandTestCase):
 
     def test_init(self):