* AtomicInteger's atomic operations, with some help from ConcurrentMap's atomic operations on
   * creation and removal (including automatic removal of zeroes). If the modification of an
   * AtomicInteger results in zero, we compareAndSet the value to zero; if that succeeds, we remove
   * the entry from the Map. If another operation sees a zero in the map, it knows that the entry is

   * AtomicInteger's atomic operations, with some help from ConcurrentMap's atomic operations on
   * creation and removal (including automatic removal of zeroes). If the modification of an
   * AtomicInteger results in zero, we compareAndSet the value to zero; if that succeeds, we remove
   * the entry from the Map. If another operation sees a zero in the map, it knows that the entry is

   * #add(Range) add} any range not {@linkplain Range#encloses(Range) enclosed} by {@code view}.
   */
  RangeSet<C> subRangeSet(Range<C> view);

  // Modification

  /**
   * Adds the specified range to this {@code RangeSet} (optional operation). That is, for equal
   * range sets a and b, the result of {@code a.add(range)} is that {@code a} will be the minimal

   * concurrent modification, the iterator fails quickly and cleanly, rather than risking arbitrary,
   * non-deterministic behavior at an undetermined time in the future.
   *
   * <p>Note that the fail-fast behavior of an iterator cannot be guaranteed as it is, generally
   * speaking, impossible to make any hard guarantees in the presence of unsynchronized concurrent

   * lists are sublist views of the original list, produced on demand using {@link List#subList(int,
   * int)}, and are subject to all the usual caveats about modification as explained in that API.
   *
   * @param list the list to return consecutive sublists of
   * @param size the desired size of each sublist (the last may be smaller)
   * @return a list of consecutive sublists

   * allocated, the value of {@code metadata} combines the number of bits in the "short hash", in
   * its bottom {@value CompactHashing#HASH_TABLE_BITS_MAX_BITS} bits, with a modification count in
   * the remaining bits that is used to detect concurrent modification during iteration.
   */
  private transient int metadata;

  /** The number of elements contained in the set. */
  private transient int size;

   * #add(Range) add} any range not {@linkplain Range#encloses(Range) enclosed} by {@code view}.
   */
  RangeSet<C> subRangeSet(Range<C> view);

  // Modification

  /**
   * Adds the specified range to this {@code RangeSet} (optional operation). That is, for equal
   * range sets a and b, the result of {@code a.add(range)} is that {@code a} will be the minimal

     * They may be subsequently nulled out by TreeMultiset.clear(). I think that the only place that
     * we can reference a node whose fields have been cleared is inside the iterator (and presumably
     * only under concurrent modification).
     *
     * To access these fields when you know that they are not null, call the pred() and succ()
     * methods, which perform null checks before returning the fields.
     */

   * allocated, the value of {@code metadata} combines the number of bits in the "short hash", in
   * its bottom {@value CompactHashing#HASH_TABLE_BITS_MAX_BITS} bits, with a modification count in
   * the remaining bits that is used to detect concurrent modification during iteration.
   */
  private transient int metadata;

  /** The number of elements contained in the set. */
  private transient int size;

  public static void touch(File file) throws IOException {
    checkNotNull(file);
    if (!file.createNewFile() && !file.setLastModified(System.currentTimeMillis())) {
      throw new IOException("Unable to update modification time of " + file);
    }
  }

  /**
   * Creates any necessary but nonexistent parent directories of the specified file. Note that if