Add test that the storage is really cleared.

theano/scan_module/tests/test_scan.py

@@ -18,6 +18,7 @@ from theano.compile.pfunc import rebuild_collect_shared
 from theano.gof.python25 import any
 from theano.tests import unittest_tools as utt
 import theano.scalar.sharedvar
+from theano.scan_module.scan_op import Scan
 from theano.gof.python25 import OrderedDict
 from theano.compat import PY3
 
@@ -261,6 +262,45 @@ class T_Scan(unittest.TestCase):
         theano_values = my_f(state, steps)
         utt.assert_allclose(numpy_values, theano_values)
 
+    # Test that the inner input_storage and output_storage are
+    # properly cleared
+    def test_inner_storage_leak(self):
+        def f_pow2(x_tm1):
+            return 2 * x_tm1
+
+        state = theano.tensor.scalar('state')
+        n_steps = theano.tensor.iscalar('nsteps')
+        output, updates = theano.scan(f_pow2,
+                                      [],
+                                      state,
+                                      [],
+                                      n_steps=n_steps)
+
+        f = theano.function([state, n_steps],
+                            output,
+                            updates=updates,
+                            allow_input_downcast=True)
+
+        scan_node = [node for node in f.maker.fgraph.toposort()
+                     if isinstance(node.op, Scan)]
+
+        assert len(scan_node) == 1
+        scan_node = scan_node[0]
+
+        # Make sure they start out as None
+        assert all(i.value is None for i in scan_node.op.fn.input_storage)
+        assert all(o.value is None for o in scan_node.op.fn.output_storage)
+
+        rng = numpy.random.RandomState(utt.fetch_seed())
+        state = rng.uniform()
+        steps = 5
+
+        f(state, steps)
+
+        # And that they stay that way
+        assert all(i.value is None for i in scan_node.op.fn.input_storage)
+        assert all(o.value is None for o in scan_node.op.fn.output_storage)
+
     # generator network, only one output , type scalar ; no sequence or
     # non sequence arguments
     def test_generator_one_output_scalar(self):