diff --git a/tensorflow/core/kernels/tensor_array_ops.cc b/tensorflow/core/kernels/tensor_array_ops.cc
index 075bacb43..c795ecfb6 100644
--- a/tensorflow/core/kernels/tensor_array_ops.cc
+++ b/tensorflow/core/kernels/tensor_array_ops.cc
@@ -597,15 +597,24 @@ class TensorArrayPackOrGatherOp : public OpKernel {
     // If there are no elements to return, return a zero-element Tensor with
     // shape [0] + element_shape_
     if (num_indices == 0) {
-      OP_REQUIRES(ctx, element_shape_.IsFullyDefined(),
+	    PartialTensorShape partial_empty_shape;
+	    OP_REQUIRES(ctx, element_shape_.MergeWith(tensor_array->ElemShape(),
+				    &partial_empty_shape).ok(),
+			    errors::InvalidArgument(
+				    "Attr element shape ", element_shape_.DebugString(),
+				    " is not compatible with TensorArray element shape ",
+				    tensor_array->ElemShape().DebugString()));
+
+	    OP_REQUIRES(ctx, partial_empty_shape.OsFullyDefined(),
                   errors::Unimplemented(
                       "TensorArray has size zero, but element shape ",
-                      element_shape_.DebugString(),
+                      partial_empty_shape.DebugString(),
                       " is not fully defined. "
                       "Currently only static shapes are supported when packing "
                       "zero-size TensorArrays."));
+
       TensorShape empty_shape;
-      element_shape_.AsTensorShape(&empty_shape);
+      partial_empty_shape.AsTensorShape(&empty_shape);
       empty_shape.InsertDim(0, 0);
       Tensor* empty_unused;
       OP_REQUIRES_OK(ctx, ctx->allocate_output(0, empty_shape, &empty_unused));