Fix MissingInputError when scan use hidden variable to calculate condition (#6597)

theano/scan_module/scan.py

@@ -820,6 +820,8 @@ def scan(fn,
 
     # extract still missing inputs (there still might be so) and add them
     # as non sequences at the end of our args
+    if condition is not None:
+        outputs.append(condition)
     fake_nonseqs = [x.type() for x in non_seqs]
     fake_outputs = scan_utils.clone(outputs,
                                     replace=OrderedDict(izip(non_seqs,
@@ -836,8 +838,6 @@ def scan(fn,
     dummy_args += extra_inputs
 
     dummy_outs = outputs
-    if condition is not None:
-        dummy_outs.append(condition)
     # Perform a try-except to provide a meaningful error message to the
     # user if inputs of the inner function are missing.
     try:

theano/scan_module/tests/test_scan.py

@@ -5718,3 +5718,23 @@ class TestGradUntil(unittest.TestCase):
         numpy_grad = np.array([0, 0, 0, 5, 6, 10, 4, 5, 0, 0, 0, 0, 0, 0, 0])
         numpy_grad = numpy_grad.astype(theano.config.floatX)
         utt.assert_allclose(theano_gradient, numpy_grad)
+
+def test_condition_hidden_inp():
+    max_value = theano.tensor.scalar("max_value")
+    n_steps = theano.tensor.iscalar("n_steps")
+    def accum(prev_value, step):
+        new_value = prev_value + step
+        new_step = step + 1
+        condition = theano.scan_module.until(new_value > max_value)
+        return [new_value, new_step], condition
+
+    rs, updates = theano.scan(
+        fn=accum,
+        outputs_info=[0, 0],
+        n_steps=n_steps)
+
+    f = theano.function(
+        inputs=[max_value, n_steps],
+        outputs=rs)
+
+    _sum, total_steps = f(100, 100)