Fix check for broadcastable pattern.

theano/tensor/opt.py

@@ -618,8 +618,8 @@ def local_upcast_elemwise_constant_inputs(node):
                     new_inputs.append(i)
                 else:
                     try:
-                        cval_i = get_constant_value(i)    # works only for scalars I think
-                        if all((not b for b in i.broadcastable)):
+                        cval_i = get_constant_value(i)    # works only for scalars
+                        if all(i.broadcastable):
                             new_inputs.append(T.cast(cval_i, output_dtype))
                         else:
                             if shape_i is None:
@@ -638,9 +638,11 @@ def local_upcast_elemwise_constant_inputs(node):
                 if rval[0].type != node.outputs[0].type:
                     print >> sys.stderr, "NODE:", node
                     print >> sys.stderr, "NODE INPUT TYPES:", [i.type for i in node.inputs]
+                    print >> sys.stderr, "NODE OUTPUT TYPES:", [o.type for o in node.outputs]
                     print >> sys.stderr, "RVAL:", rval
                     print >> sys.stderr, "NEW INPUT TYPES:", [i.type for i in new_inputs]
                     print >> sys.stderr, "RVAL INPUT TYPES:", [i.type for i in rval[0].owner.inputs]
+                    print >> sys.stderr, "RVAL TYPES:", [o.type for o in rval]
                 assert rval[0].type == node.outputs[0].type, (node, rval[0])
                 return rval