/*abs_error*/ 0));
  ASSERT_NO_FATAL_FAILURE(CheckTensorValue(outputs[1], {0.0f}, /*dims*/ {},
                                           /*abs_error*/ 0));
  outputs[0]->Unref();
  outputs[1]->Unref();
}

#ifdef PLATFORM_GOOGLE
INSTANTIATE_TEST_SUITE_P(
    UnifiedCAPI, CppGradients,
    ::testing::Combine(::testing::Values("graphdef", "mlir"),

static void RenameFile(const TF_Filesystem* filesystem, const char* src,
                       const char* dst, TF_Status* status) {
  // If target is a directory return TF_FAILED_PRECONDITION.
  // Target might be missing, so don't error in that case.
  struct stat st;
  if (stat(dst, &st) != 0) {
    if (errno != ENOENT) {
      TF_SetStatusFromIOError(status, errno, dst);
      return;
    }
  } else if (S_ISDIR(st.st_mode)) {

  TFE_DeleteOp(op);

  // The device of var_handle1 is remote device, which was replaced during
  // cluster update. Removing resource with invalid device should fail
  // gracefully (i.e., with error status) instead of crashing with segfaults.
  op = TFE_NewOp(ctx, "DestroyResourceOp", status);
  ASSERT_EQ(TF_OK, TF_GetCode(status)) << TF_Message(status);
  TFE_OpAddInput(op, var_handle1, status);

  if (!tensorflow::protobuf::TextFormat::ParseFromString(text_proto, &gdef)) {
    status->status = tensorflow::errors::Internal(
        "Invalid text proto for GraphDef: ", text_proto);
    return {};
  }
  const auto& fdef_lib = gdef.library();
  if (fdef_lib.gradient_size() > 0) {
    status->status = tensorflow::errors::Internal(
        "GradientDef is not supported in reading Dataset related functions: ",
        text_proto);

  }
  registry_.insert({op_name, gradient_function_factory});
  return absl::OkStatus();
}
Status GradientRegistry::Lookup(
    const ForwardOperation& op,
    std::unique_ptr<GradientFunction>* gradient_function) const {
  auto iter = registry_.find(op.op_name);
  if (iter == registry_.end()) {
    const string error_msg = "No gradient defined for op: " + op.op_name + ".";
    return errors::NotFound(error_msg);

  OwnedTFStatus c_status(TF_NewStatus());
  init_fn(&params, c_status.get());
  TF_RETURN_IF_ERROR(tsl::StatusFromTF_Status(c_status.get()));
  TF_RETURN_IF_ERROR(ValidateTPOptimizerRegistrationParams(params));
  TF_RETURN_IF_ERROR(ValidateTPOptimizer(optimizer));
  TF_RETURN_IF_ERROR(ValidateTPOptimizerConfigs(optimizer_configs));

  CGraphOptimizerRegister(

      }
    }
  });
  // Loop until either block content is successfully fetched, or our request
  // encounters an error.
  absl::MutexLock l(&block->mu);
  TF_SetStatus(status, TF_OK, "");
  while (true) {
    switch (block->state) {
      case FetchState::ERROR:
        // TF_FALLTHROUGH_INTENDED
      case FetchState::CREATED:
        block->state = FetchState::FETCHING;

  EXPECT_EQ(out.size(), 1);
  // Now read the first block; this should yield an INTERNAL error because we
  // had already cached a partial block at a later position.
  Status status = ReadCache(&cache, "", 0, block_size, &out);
  EXPECT_EQ(status.code(), error::INTERNAL);
}

TEST(RamFileBlockCacheTest, LRU) {
  const size_t block_size = 16;
  std::list<size_t> calls;

  // Step 1: Load plugin
  Env* env = Env::Default();
  void* dso_handle;
  TF_RETURN_IF_ERROR(env->LoadDynamicLibrary(dso_path.c_str(), &dso_handle));

  // Step 2: Load symbol for `TF_InitPlugin`
  void* dso_symbol;
  TF_RETURN_WITH_CONTEXT_IF_ERROR(
      env->GetSymbolFromLibrary(dso_handle, "TF_InitPlugin", &dso_symbol),
      "Failed to load TF_InitPlugin symbol for DSO: ", dso_path);

}

class FunctionErrorInjectionPass : public tensorflow::FunctionOptimizationPass {
 public:
  FunctionErrorInjectionPass(string error_node, string error_device)
      : error_node_(error_node), error_device_(error_device) {}
  tensorflow::Status Run(const std::string& function_name,
                         const tensorflow::DeviceSet& device_set,
                         const tensorflow::ConfigProto& config_proto,