%cst = "tf.Const"() {value = dense<3> : tensor<i32>} : () -> tensor<i32>
  %cst_1 = "tf.Const"() {value = dense<[2, 0, 1]> : tensor<3xi32>} : () -> tensor<3xi32>
  %cst_2 = "tf.Const"() {value = dense<[512, 64]> : tensor<2xi32>} : () -> tensor<2xi32>
  %0 = "tf.ExpandDims"(%arg0, %cst) {device = ""} : (tensor<8x64xf32>, tensor<i32>) -> tensor<8x64x1xf32>
  %1 = "tf.AddV2"(%0, %arg1) {device = ""} : (tensor<8x64x1xf32>, tensor<8x64x64xf32>) -> tensor<8x64x64xf32>

  tfg.func @test_func_name0(%test_func_name0: tensor<*x!tf_type.resource> {tf._output_shapes = #tf_type.shape<*>, tf._resource_arg_unique_id = 0 : i64, tfg.name = "test_func_name0"},...

  %cst = arith.constant dense<[-1, 7]> : tensor<2xi32>
  %0 = "tf.TensorListReserve"(%cst, %arg0) : (tensor<2xi32>, tensor<i32>) -> tensor<!tf_type.variant<tensor<?x7xf32>>>
  %1 = "tf.TensorListGetItem"(%0, %arg1, %cst) : (tensor<!tf_type.variant<tensor<?x7xf32>>>, tensor<i32>, tensor<2xi32>) -> tensor<?x7xf32>
  func.return %1 : tensor<?x7xf32>

  func.return %0 : tensor<!tf_type.variant<tensor<2xi32>>>

      %cst_1 = arith.constant dense<2.0> : tensor<3xf32>
      %0 = "tf.XlaCallModule"(%arg0, %cst_0, %cst_1) <{Sout = [#tf_type.shape<1x4x4x3>], module = "", version = 9 : i64}> {_entry_function = @composite_fn_1, _original_entry_function = "composite_fn_1", _tfl_quant_trait = "fully_quantizable"} : (tensor<1x4x4x3xf32>, tensor<3x1x1x3xf32>, tensor<3xf32>) -> tensor<1x4x4x3xf32>

module attributes {tf.versions = {bad_consumers = [], min_consumer = 0 : i32, producer = 268 : i32}} {
  func.func @main() -> () {
     %s0 = "tf.Const"() {value = dense<[4, 1]> : tensor<2xi32>} : () -> tensor<2xi32>
    func.return
  }

      } : (tensor<?x!tf_type.qint8>, tensor<f32>, tensor<i32>) -> tensor<?xf32>
      func.return %1 : tensor<?xf32>
    }
}

// -----

// expected-error@+1 {{Input shapes provided but no `main` function found.}}
module attributes {tf.versions = {bad_consumers = [], min_consumer = 0 : i32, producer = 268 : i32}} {
    func.func @non_main(%arg0 : tensor<?xf32>) -> tensor<?xf32> {

      %outputs_4, %control_5 = tf_executor.island wraps "tf.BiasAdd"(%outputs_2, %outputs) {data_format = "NHWC", device = ""} : (tensor<*xf32>, tensor<32xf32>) -> tensor<*xf32>
      %outputs_6, %control_7 = tf_executor.island wraps "tf.Identity"(%outputs_4) {device = ""} : (tensor<*xf32>) -> tensor<*xf32>
      tf_executor.fetch %outputs_6 : tensor<*xf32>
    }
    func.return %0 : tensor<*xf32>
  }

func.func @side_effect_free_user() -> tensor<2xi32> {
  // CHECK: [[handle:%.*]] = "tf.VarHandleOp"
  // CHECK-SAME: shared_name = "x"
  // CHECK: [[value0:%.*]] = "tf.ReadVariableOp"([[handle]])
  // CHECK: [[value1:%.*]] = "tf.ReadVariableOp"([[handle]])
  // CHECK: "tf.ConcatV2"([[value1]], [[value0]]
  // CHECK-SAME: (tensor<i32>, tensor<i32>, tensor<i32>) -> tensor<2xi32>

  %16 = "tf.Mean"(%15, %1) : (tensor<?x256x56x56xf32>, tensor<2xi32>) -> tensor<?x256xf32>

  // Mean should compute reduction over NHWC spatial dimensions.
  // CHECK: %[[MEAN_DIMS:.*]] = "tf.Const"() <{value = dense<[1, 2]> : tensor<2xi32>}>
  // CHECK: %[[MEAN:[0-9]*]] = "tf.Mean"(%[[RELU]], %[[MEAN_DIMS]])
  // CHECK-SAME: (tensor<?x56x56x256xf32>, tensor<2xi32>) -> tensor<?x256xf32>
  // CHECK: return %[[MEAN]] : tensor<?x256xf32>