* we've chosen to instead use a signature that benefits existing users -- and is still safe.
   *
   * For convertAll, I haven't looked as closely at *how* much existing users benefit, so we should
   * keep an eye out for problems that new users encounter. Note also that convertAll could support
   * both use cases by using @PolyNull. (By contrast, we can't use @PolyNull for our superinterface

    final Service service;
    // We store the state in a weak reference to ensure that if something went wrong while
    // constructing the ServiceManager we don't pointlessly keep updating the state.
    final WeakReference<ServiceManagerState> state;

    ServiceListener(Service service, WeakReference<ServiceManagerState> state) {
      this.service = service;
      this.state = state;
    }

  // array. The top bits of this value are the remaining bits of the hash value that were not used
  // in the short hash. We saw that a mask of 0x7f would keep the 7-bit value 0x6f from a full
  // hashcode of 0x89abcdef. The imaginary `hash` value would then be the remaining top 25 bits,
  // 0x89abcd80. To this is added (or'd) the `next` value, which is an index within `entries`

 * href="https://www.javadoc.io/doc/com.github.ben-manes.caffeine/caffeine/latest/com.github.benmanes.caffeine/com/github/benmanes/caffeine/cache/AsyncLoadingCache.html">their
 * deep support for asynchronous operations</a>. But if you want to migrate to Caffeine with minimal
 * code changes, you can use <a

     * is the only one in use, but if we encounter nested concatenations, we start a deque of
     * meta-iterators rather than letting the nesting get arbitrarily deep.  This keeps each
     * operation O(1).
     */

    @CheckForNull private Iterator<? extends Iterator<? extends T>> topMetaIterator;

    // Only becomes nonnull if we encounter nested concatenations.

      HashMap<K, V> newHashMapWithExpectedSize(int expectedSize) {
    return new HashMap<>(capacity(expectedSize));
  }

  /**
   * Returns a capacity that is sufficient to keep the map from being resized as long as it grows no
   * larger than expectedSize and the load factor is ≥ its default (0.75).
   */
  static int capacity(int expectedSize) {
    if (expectedSize < 3) {

    return requireNonNull(DEST_TL.get());
  }

  /**
   * A thread-local destination buffer to keep us from creating new buffers. The starting size is
   * 1024 characters. If we grow past this we don't put it back in the threadlocal, we just keep
   * going and grow as needed.
   */
  private static final ThreadLocal<char[]> DEST_TL =
      new ThreadLocal<char[]>() {
        @Override

 *
 * <p>This class can't depend on other Guava code. If we were to load this class in the same class
 * loader as the rest of Guava, this thread would keep an indirect strong reference to the class
 * loader and prevent it from being garbage collected. This poses a problem for environments where

          // reset state in case of an error so that later dispatch calls will actually do something
          synchronized (this) {
            isThreadScheduled = false;
          }
          // Log it and keep going.
          logger
              .get()
              .log(
                  Level.SEVERE,
                  "Exception while running callbacks for " + listener + " on " + executor,
                  e);

<module>
<source path="">
  <!-- Hack to keep collect from hiding collect.testing supersource: -->
  <exclude name="**/testing/**"/>
</source>

<!--
    We used to set this only for packages that had manual supersource. That
    worked everywhere that I know of except for one place: when running the GWT
    util.concurrent tests under Guava.

    The problem is that GWT responds poorly to two .gwt.xml files in the same