EXPECT_EQ(dltensor_out->ndim, dltensor_in->ndim);
  EXPECT_EQ(dltensor_out->dtype.code, dltensor_in->dtype.code);
  EXPECT_EQ(dltensor_out->dtype.bits, dltensor_in->dtype.bits);
  EXPECT_EQ(dltensor_out->dtype.lanes, dltensor_in->dtype.lanes);
  for (int i = 0; i < dltensor_in->ndim; ++i) {
    EXPECT_EQ(dltensor_out->shape[i], dltensor_in->shape[i]);
    if (dltensor_out->strides) {

    return;
  } else if (S_ISDIR(st.st_mode)) {
    TF_SetStatus(status, TF_FAILED_PRECONDITION, "source path is a directory");
    return;
  }

  // Do the actual rename. Here both arguments are filenames.
  if (rename(src, dst) != 0)
    TF_SetStatusFromIOError(status, errno, dst);
  else
    TF_SetStatus(status, TF_OK, "");
}

  switch (dtype.code) {
    case DLDataTypeCode::kDLBool:
      if (dtype.bits != 8) {
        return tensorflow::errors::InvalidArgument(
            "Only DLPack bools of bitwidth 8 are supported, got: ", dtype.bits);
      }
      *tf_dtype = TF_DataType::TF_BOOL;
      return absl::OkStatus();

    case DLDataTypeCode::kDLUInt:
      switch (dtype.bits) {
        case 8:
          *tf_dtype = TF_DataType::TF_UINT8;

    TF_RETURN_IF_ERROR(
        RunAndMaybeSum(ctx, forward, theta_inputs, f_outputs, use_function));
    AbstractTensorHandlePtr fMinus(f_outputs[0]);

    // Take Difference of both estimates: (f(theta + eps) - f(theta - eps)).
    TF_RETURN_IF_ERROR(
        ops::Sub(ctx, fPlus.get(), fMinus.get(), f_outputs, "sub_top"));
    AbstractTensorHandlePtr fDiff(f_outputs[0]);

// second worker, and run a distributed function (VariableAddFunction) whose ops
// span the local and remote workers. If the graph optimization pass is executed
// on both the main function side and the component function side, an error will
// be thrown in the registered graph optimization pass.
TEST(CAPI, DistributedFunctionGraphPassOnlyOnce) {

  ~GraphOperation() override {}

 private:
  friend class GraphContext;  // For access to op_.
  TF_Graph* g_;
  std::unique_ptr<TF_OperationDescription> op_;
  // Hold `op_type` and `op_name` till both are available since we need both
  // to build a graph operation.
  string op_type_;
  const char* op_name_ = nullptr;
  // TODO(srbs): Use this.
  string device_name_;
};

  // Define a second function
  TF_Function* func1;
  DefineFunction("FooFunc1", &func1);
  TF_GraphCopyFunction(host_graph_, func1, nullptr, s_);
  ASSERT_EQ(TF_OK, TF_GetCode(s_)) << TF_Message(s_);

  // Get both function from host_graph_
  EXPECT_EQ(TF_GraphNumFunctions(host_graph_), 2);
  EXPECT_EQ(TF_GraphGetFunctions(host_graph_, funcs, 0, s_), 0);
  ASSERT_EQ(TF_OK, TF_GetCode(s_)) << TF_Message(s_);

    EXPECT_EQ(got_offset, want_offset);
    EXPECT_EQ(got_n, want_n);
    calls++;
    memset(buffer, 'x', got_n);
    TF_SetStatus(status, TF_OK, "");
    return got_n;
  };
  // If block_size, max_bytes, or both are zero, or want_n is larger than
  // max_bytes the cache is a pass-through.
  tf_gcs_filesystem::RamFileBlockCache cache1(1, 0, 0, fetcher);
  tf_gcs_filesystem::RamFileBlockCache cache2(0, 1, 0, fetcher);

  }

  TF_DeleteAbstractTensor(arg0);
  TF_DeleteAbstractTensor(arg1);

  // Finalize the function by providing the returned values.
  TF_AbstractFunction* func;
  {
    // We want to return the output of both add operations, create a new list
    // and populate it.
    TF_OutputList* func_outputs = TF_NewOutputList();
    TF_OutputListPushBack(func_outputs, add_output1, s);
    ASSERT_EQ(TF_OK, TF_GetCode(s)) << TF_Message(s);

    EXPECT_TRUE(GetGraphDef(expected_graph_, &expected_gdef));
    EXPECT_TRUE(GetGraphDef(graph_, &gdef));
    TF_EXPECT_GRAPH_EQ(expected_gdef, gdef);

    // Compare that the output of the gradients of both graphs match.
    RunGraphsAndCompareOutputs(grad_outputs, expected_grad_outputs);
  }

  void TestGradientsError(bool grad_inputs_provided) {
    TF_Output inputs[1];
    TF_Output outputs[1];