* reduced the available size of the buffer each time a call to read didn't fill up the available
   * space in the buffer completely. In general this is a performance problem since the buffer size
   * is permanently reduced, but with certain Reader implementations it could also cause the buffer
   * size to reach 0, causing an infinite loop.
   */

   * reduced the available size of the buffer each time a call to read didn't fill up the available
   * space in the buffer completely. In general this is a performance problem since the buffer size
   * is permanently reduced, but with certain Reader implementations it could also cause the buffer
   * size to reach 0, causing an infinite loop.
   */

 * @since 2.0
 */
@GwtCompatible
public abstract class ImmutableMultimap<K, V> extends BaseImmutableMultimap<K, V>
    implements Serializable {

  /**
   * Returns an empty multimap.
   *
   * <p><b>Performance note:</b> the instance returned is a singleton.
   */
  public static <K, V> ImmutableMultimap<K, V> of() {
    return ImmutableListMultimap.of();
  }

  /** Returns an immutable multimap containing a single entry. */

 * the columns. The reverse is also available, associating a column with a row key / value map. Note
 * that, in some implementations, data access by column key may have fewer supported operations or
 * worse performance than data access by row key.
 *
 * <p>The methods returning collections or maps always return views of the underlying table.
 * Updating the table can change the contents of those collections, and updating the collections

          Function<? super T, ? extends V> valueFunction) {
    return CollectCollectors.toImmutableRangeMap(keyFunction, valueFunction);
  }

  /**
   * Returns an empty immutable range map.
   *
   * <p><b>Performance note:</b> the instance returned is a singleton.
   */
  @SuppressWarnings("unchecked")
  public static <K extends Comparable<?>, V> ImmutableRangeMap<K, V> of() {
    return (ImmutableRangeMap<K, V>) EMPTY;
  }

 * as an {@link Iterable} or {@link Iterator}, prefer {@link Ordering#leastOf(Iterable, int)}.
 *
 * <p>This uses the same efficient implementation as {@link Ordering#leastOf(Iterable, int)},
 * offering expected O(n + k log k) performance (worst case O(n log k)) for n calls to {@link
 * #offer} and a call to {@link #topK}, with O(k) memory. In comparison, quickselect has the same

  2. If you were previously overriding the kube-proxy configuration for performance reasons, this may no longer be necessary. See https://kubernetes.io/docs/reference/networking/virtual-ips/#optimizing-iptables-mode-performance. ([#115138](https://github.com/kubernetes/kubernetes/pull/115138), [@danwinship](https://github.com/danwinship))

## Version 2.7.4

_2016-02-07_

 *  Fix: Don't crash when finding the trust manager if the Play Services (GMS)
    security provider is installed.
 *  Fix: The previous release introduced a performance regression on Android,
    caused by looking up CA certificates. This is now fixed.


## Version 2.7.3

_2016-02-06_

 *  Fix: Permit the trusted CA root to be pinned by `CertificatePinner`.

 * <p>These features are all optional; {@code new MapMaker().makeMap()} returns a valid concurrent
 * map that behaves similarly to a {@link ConcurrentHashMap}.
 *
 * <p>The returned map is implemented as a hash table with similar performance characteristics to
 * {@link ConcurrentHashMap}. It supports all optional operations of the {@code ConcurrentMap}
 * interface. It does not permit null keys or values.
 *

      ),
    )
    assertThat(foxPart.body.readUtf8()).isEqualTo("Fox")

    assertThat(reader.nextPart()).isNull()
  }

  /**
   * Read 100 MiB of 'a' chars. This was really slow due to a performance bug in [MultipartReader],
   * and will be really slow if we regress the fix for that.
   */
  @Test
  fun `reading a large part with small byteCount`() {
    val multipartBody =
      MultipartBody