// The scalar const oper also has a consumer.
  ASSERT_EQ(1, TF_OperationOutputNumConsumers(TF_Output{three, 0}));
  TF_Input three_port;
  EXPECT_EQ(1,
            TF_OperationOutputConsumers(TF_Output{three, 0}, &three_port, 1));
  EXPECT_EQ(add, three_port.oper);
  EXPECT_EQ(1, three_port.index);

  // Serialize to GraphDef.
  GraphDef graph_def;
  ASSERT_TRUE(GetGraphDef(graph, &graph_def));

  // Remove the blocks from "a".
  cache.RemoveFile("a");
  // Both blocks from "b" should still be there.
  TF_EXPECT_OK(ReadCache(&cache, "b", 0, n, &out));
  EXPECT_EQ(out, b);
  TF_EXPECT_OK(ReadCache(&cache, "b", 8, n, &out));
  EXPECT_EQ(out, B);
  EXPECT_EQ(calls, 4);
  // The blocks from "a" should not be there.
  TF_EXPECT_OK(ReadCache(&cache, "a", 0, n, &out));
  EXPECT_EQ(out, a);
  EXPECT_EQ(calls, 5);

  p.text += "\n      Most TF jobs run tests three times to root out flakes."
  if seen[key] == 3:
    p.text += (
        "\n      Since there were three failures, this is not flaky, and it"
    )
    p.text += "\n      probably caused the Kokoro invocation to fail."
  else:
    p.text += (
        "\n      Since there were not three failures, this is probably a flake."
    )
    p.text += (

  TensorHandlePtr three_vector =
      VectorFloatTensorHandle({5., 6., 7.}, status.get());
  ASSERT_TRUE(TF_GetCode(status.get()) == TF_OK) << TF_Message(status.get());

  std::vector<TensorHandlePtr> vector_handles;
  vector_handles.reserve(2);
  vector_handles.push_back(std::move(two_vector));
  vector_handles.push_back(std::move(three_vector));

  std::vector<float> size_two_value{1., 2.};
  std::vector<float> size_three_value{1., 2., 3.};
  TensorHandlePtr size_two(
      VectorFloatTensorHandle(size_two_value, status.get()));
  ASSERT_EQ(TF_GetCode(status.get()), TF_OK) << TF_Message(status.get());
  TensorHandlePtr size_three(
      VectorFloatTensorHandle(size_three_value, status.get()));
  ASSERT_EQ(TF_GetCode(status.get()), TF_OK) << TF_Message(status.get());

        "operations on them have been supplied.");

  return OkStatus();
}

// Copies a function table from plugin memory space to core memory space.
//
// This has three benefits:
//   * allows having newer plugins than the current core TensorFlow: the
//     additional entries in the plugin's table are just discarded;
//   * allows having older plugins than the current core TensorFlow (though

# TFCI_BAZEL_TARGET_SELECTING_CONFIG_PREFIX. "Available" means there is
# an entry in tensorflow/.bazelrc for build:rbe_TFCI_BAZEL_TARGET...
#                                           ^^^^
# This env is only valid when RBE is available on the selected platform, since
# it enables a derived --config setting.  If RBE is not available (i.e. there
# is no --config setting), bazel would fail and quit. This script does a quick

void RamFileBlockCache::Trim() {
  while (!lru_list_.empty() && cache_size_ > max_bytes_) {
    RemoveBlock(block_map_.find(lru_list_.back()));
  }
}

/// Move the block to the front of the LRU list if it isn't already there.
void RamFileBlockCache::UpdateLRU(const Key& key,
                                  const std::shared_ptr<Block>& block,
                                  TF_Status* status) {
  absl::MutexLock lock(&mu_);

build/installation issues on GitHub.

-   type: markdown
    attributes:
      value: |

These tests use hermetic Python. They also require a built TensorFlow wheel file
and a requirements_lock file. The requirements_lock file is generated by the
[requirements_updater](https://github.com/tensorflow/tensorflow/tree/master/ci/official/requirements_updater)
tool using the path to this wheel file.

### Hermetic Python

For details about hermetic Python and setting its toolchain version, see