* placed adjacent to each other, and so will most often share
   * cache lines (with a huge negative performance impact) without
   * this precaution.
   *
   * In part because Cells are relatively large, we avoid creating
   * them until they are needed.  When there is no contention, all
   * updates are made to the base field.  Upon first contention (a
   * failed CAS on base update), the table is initialized to size 2.

    final int len = string.length();
    final long longSize = (long) len * (long) count;
    final int size = (int) longSize;
    if (size != longSize) {
      throw new ArrayIndexOutOfBoundsException("Required array size too large: " + longSize);
    }

    final char[] array = new char[size];
    string.getChars(0, len, array, 0);
    int n;
    for (n = len; n < size - n; n <<= 1) {
      System.arraycopy(array, 0, array, n, n);

     */
    abstract SetBuilderImpl<E> copy();

    /**
     * Call this before build(). Does a final check on the internal data structures, e.g. shrinking
     * unnecessarily large structures or detecting previously unnoticed hash flooding.
     */
    SetBuilderImpl<E> review() {
      return this;
    }

    abstract ImmutableSet<E> build();
  }

   * ranges {@code [3..3)} and {@code [4..4)}.
   *
   * <p><b>Warning:</b> Be extremely careful what you do with the {@code asSet} view of a large
   * range set (such as {@code ImmutableRangeSet.of(Range.greaterThan(0))}). Certain operations on
   * such a set can be performed efficiently, but others (such as {@link Set#hashCode} or {@link

   * ranges {@code [3..3)} and {@code [4..4)}.
   *
   * <p><b>Warning:</b> Be extremely careful what you do with the {@code asSet} view of a large
   * range set (such as {@code ImmutableRangeSet.of(Range.greaterThan(0))}). Certain operations on
   * such a set can be performed efficiently, but others (such as {@link Set#hashCode} or {@link

    // It's possible to race and suspend the thread just before the park call actually takes effect,
    // causing the thread to be suspended for 3.5 seconds, and then park itself for 2 seconds after
    // being resumed. To avoid a flake in this scenario, calculate how long that thread actually
    // waited and assert based on that time. Empirically, the race where the thread ends up waiting
    // for 5.5 seconds happens about 2% of the time.

  public static <E extends @Nullable Object> ArrayList<E> newArrayList(E... elements) {
    checkNotNull(elements); // for GWT
    // Avoid integer overflow when a large array is passed in
    int capacity = computeArrayListCapacity(elements.length);
    ArrayList<E> list = new ArrayList<>(capacity);
    Collections.addAll(list, elements);
    return list;
  }

  /**

   *       userspace buffer (byte[] or ByteBuffer), then copies them from that buffer into the
   *       destination channel.
   * </ol>
   *
   * This value is intended to be large enough to make the overhead of system calls negligible,
   * without being so large that it causes problems for systems with atypical memory management if
   * approaches 2 or 3 are used.
   */
  private static final int ZERO_COPY_CHUNK_SIZE = 512 * 1024;

  }

  public void testFindLast_refStream() {
    assertThat(findLast(Stream.of())).isEmpty();
    assertThat(findLast(Stream.of("a", "b", "c", "d"))).hasValue("d");

    // test with a large, not-subsized Spliterator
    List<Integer> list =
        IntStream.rangeClosed(0, 10000).boxed().collect(Collectors.toCollection(LinkedList::new));
    assertThat(findLast(list.stream())).hasValue(10000);

  private enum Timeout {
    MIN(Long.MIN_VALUE, "-oo"),
    MINUS_SMALL(-SMALL_TIMEOUT_MILLIS, "-" + SMALL_TIMEOUT_MILLIS + "ms"),
    ZERO(0L, "0ms"),
    SMALL(SMALL_TIMEOUT_MILLIS, SMALL_TIMEOUT_MILLIS + "ms"),
    LARGE(UNEXPECTED_HANG_DELAY_MILLIS * 2, (2 * UNEXPECTED_HANG_DELAY_MILLIS) + "ms"),
    MAX(Long.MAX_VALUE, "+oo");

    final long millis;
    final String label;

    Timeout(long millis, String label) {