throw new UnsupportedOperationException();
  }

  /**
   * Returns an {@code ImmutableList} containing the same elements, in the same order, as this
   * collection.
   *
   * <p><b>Performance note:</b> in most cases this method can return quickly without actually
   * copying anything. The exact circumstances under which the copy is performed are undefined and
   * subject to change.
   *
   * @since 2.0
   */

```

Let's take for example a set of 12 compute nodes with an aggregate memory of _1.2TiB_, we need to do following settings for optimal results. Add the following optimal entries for _core-site.xml_ to configure _s3a_ with **MinIO**. Most important options here are

```
cat ${HADOOP_CONF_DIR}/core-site.xml | kv-pairify | grep "mapred"

mapred.maxthreads.generate.mapoutput=2 # Num threads to write map outputs

  # ROCm / CUDA are mutually exclusive.
  # At most 1 GPU platform can be configured.
  gpu_platform_count = 0
  if environ_cp.get('TF_NEED_ROCM') == '1':
    gpu_platform_count += 1
  if environ_cp.get('TF_NEED_CUDA') == '1':
    gpu_platform_count += 1
  if gpu_platform_count >= 2:
    raise UserInputError('CUDA / ROCm are mututally exclusive. '
                         'At most 1 GPU platform can be configured.')

   */
  @IgnoreJRERequirement
  public void forEachLine(Consumer<? super String> action) throws IOException {
    try (Stream<String> lines = lines()) {
      // The lines should be ordered regardless in most cases, but use forEachOrdered to be sure
      lines.forEachOrdered(action);
    } catch (UncheckedIOException e) {
      throw e.getCause();
    }
  }

  /**

  }

  /**
   * Tests the {@link Object#equals} and {@link Object#hashCode} of {@code cls}. In details:
   *
   * <ul>
   *   <li>The non-private constructor or non-private static factory method with the most parameters
   *       is used to construct the sample instances. In case of tie, the candidate constructors or
   *       factories are tried one after another until one can be used to construct sample
   *       instances.

   */
  public void forEachLine(Consumer<? super String> action) throws IOException {
    try (Stream<String> lines = lines()) {
      // The lines should be ordered regardless in most cases, but use forEachOrdered to be sure
      lines.forEachOrdered(action);
    } catch (UncheckedIOException e) {
      throw e.getCause();
    }
  }

  /**

            init(null as KeyStore?)
          }.trustManagers
          .first() as X509TrustManager

      val sslContext =
        Platform.get().newSSLContext().apply {
          // TODO remove most of this code after https://github.com/bcgit/bc-java/issues/686
          init(null, arrayOf(trustManager), SecureRandom())
        }

      client =
        client
          .newBuilder()

            // Just returning an Integer with the index here is faster, but requires a lot of if/then/else to check for
            // -1
            //  or QUALIFIERS.size and then resort to lexical ordering. Most comparisons are decided by the first
            // character,
            // so this is still fast. If more characters are needed then it requires a lexical sort anyway.

#### Rationale

- SSE-C requires application to remember the keys for all GET/PUT operations, any unfortunate loss of keys would automatically mean the objects cannot be accessed anymore.
- SSE-C is hardly adopted by most widely used applications, applications prefer server to manage the keys via SSE-KMS or SSE-S3.
- MinIO recommends applications to use SSE-KMS, SSE-S3 for simpler, safer and robust encryption mechanism for replicated buckets.

your changes by running the command:

```bash
tensorflow/tools/ci_build/ci_build.sh CPU tensorflow/tools/ci_build/ci_sanity.sh
```

This will catch most license, Python coding style and BUILD file issues that
may exist in your changes.

#### Running unit tests

There are two ways to run TensorFlow unit tests.