"issue at https://github.com/tensorflow/tensorflow/issues.");
        return;
      }
      // We have a convoluted scheme here: Using the C++ graph construction API
      // to add potentially many nodes to the graph without running the checks
      // (such as uniqueness of the names of nodes) we run with other functions
      // that add a node to the graph (like TF_FinishOperation).

#include "tensorflow/core/framework/tensor_reference.h"
#include "tensorflow/core/platform/logging.h"

namespace tensorflow {

namespace {

// Managing context for the DLManagedTensor, will manage the lifetime of
// DLManagedTensor. When calling DLManagedTensor::deleter, it will notify the
// original framework of destruction, and this context will be deleted also.
struct TfDlManagedTensorCtx {
  TensorReference reference;

               const GradientRegistry& registry) {
  TF_RETURN_IF_ERROR(op_->Execute(retvals, num_retvals));
  for (int i = 0; i < *num_retvals; i++) {
    // TODO(srbs): Manage refcount of ForwardOperation's inputs/outputs.
    forward_op_->outputs.push_back(retvals[i]);
  }
  // TODO(b/166669239): This is needed to support AttrBuilder::Get for string

  absl::optional<std::vector<std::unique_ptr<ParallelTensor>>>
      maybe_parallel_results(
          parallel_device.Execute(context, parallel_inputs, operation_name,
                                  attributes, expected_max_outputs, status));
  if (!maybe_parallel_results.has_value()) return result;
  std::vector<std::unique_ptr<ParallelTensor>> parallel_results(
      std::move(maybe_parallel_results.value()));

    // 'read0_maybe_with_graph_error' op having a requested device `dev2_name`.
    // During execution:
    //   * task:0 processes main function `VariableAddFunctionWithGraphError`
    //     and places the 'read0_maybe_with_graph_error' op on task:2
    //   * task:0 partitions the main function with a subgraph containing
    //     'read0_maybe_with_graph_error' sent to task:2

    TF_RETURN_IF_ERROR(dyn_cast<tracing::TracingContext>(ctx)->AddParameter(
        input->DataType(), shape, &handle));
    params->emplace_back(handle);
  }
  return absl::OkStatus();
}

// Runs `model` maybe wrapped in a function.
Status RunModel(Model model, AbstractContext* ctx,
                absl::Span<AbstractTensorHandle* const> inputs,
                absl::Span<AbstractTensorHandle*> outputs, bool use_function) {

Status RegisterFilesystemPluginImpl(const TF_FilesystemPluginInfo* info) {
  TF_RETURN_IF_ERROR(ValidatePluginMemoryRoutines(info));

  // Validate and register all filesystems
  // Try to register as many filesystems as possible.
  // Free memory once we no longer need it
  Status status;
  for (int i = 0; i < info->num_schemes; i++) {
    status.Update(ValidateAndRegisterFilesystems(info, i));

namespace tf_writable_file {
typedef struct GCSFile {
  const std::string bucket;
  const std::string object;
  gcs::Client* gcs_client;  // not owned
  TempFile outfile;
  bool sync_need;
  // `offset` tells us how many bytes of this file are already uploaded to
  // server. If `offset == -1`, we always upload the entire temporary file.
  int64_t offset;
} GCSFile;

                              int* num_retvals) {
  TF_RETURN_IF_ERROR(parent_op_->Execute(retvals, num_retvals));
  for (int i = 0; i < *num_retvals; i++) {
    // TODO(srbs): Manage refcount of ForwardOperation's inputs/outputs.
    forward_op_.outputs.push_back(retvals[i]);
  }
  // TODO(b/166669239): This is needed to support AttrBuilder::Get for string