crash fix to the new tests that will be in the next commit.

theano/tensor/basic.py

@@ -593,6 +593,8 @@ def get_scalar_constant_value(orig_v, elemwise=True):
             # mess with the stabilization optimization and be too slow.
             # We put all the scalar Ops used by get_canonical_form_slice()
             # to allow it to determine the broadcast pattern correctly.
+            elif isinstance(v.owner.op, ScalarFromTensor):
+                return get_scalar_constant_value(v.owner.inputs[0])
             elif isinstance(v.owner.op, scal.ScalarOp):
                 if isinstance(v.owner.op, scal.Second):
                     # We don't need both input to be constant for second

theano/tensor/subtensor.py

@@ -16,7 +16,7 @@ from theano.gof.python25 import maxsize
 from theano.printing import pprint
 from theano import scalar as scal
 from theano.tensor.basic import (addbroadcast, clip, get_scalar_constant_value,
-                                 ARange, TensorType)
+                                 ARange, TensorType, NotScalarConstantError)
 from theano.tensor.elemwise import DimShuffle
 from theano.tensor.type_other import NoneConst, SliceType, make_slice
 from theano import config
@@ -470,12 +470,19 @@ class Subtensor(Op):
         broadcastable = []
         for i, (p, bc) in enumerate(izip(padded, x.type.broadcastable)):
             if isinstance(p, slice):
-                if bc and p.start in [None, 0]:
+                if bc:
+                    start = p.start
+                    try:
+                        start = get_scalar_constant_value(start)
+                    except NotScalarConstantError:
+                        pass
+                    if start in [None, 0]:
                         start = p.start
                         if start is None:
                             start = 0
                         if (p.stop is None or
-                        (isinstance(p.stop, (int, numpy.integer)) and
+                            (isinstance(p.stop, (int, numpy.integer,
+                                                 numpy.ndarray)) and
                              p.stop > start)):
                             broadcastable.append(True)
                             continue