new files. **Do not add RELNOTE=yes comments in CLs.** Instead, add a file to
the CL (or ask the author to do so).

At the end of the development cycle, the files will be merged by being
concatenated in sorted order by pathname. Files in the directory matching the
glob "*stdlib/*minor" are treated specially. They should be in subdirectories
corresponding to standard library package paths, and headings for those package

 * </ul>
 *
 * To force selection of our fallback strategies we load {@link AggregateFutureState} (and all of
 * {@code com.google.common.util.concurrent}) in degenerate class loaders which make certain
 * platform classes unavailable. Then we construct a test suite so we can run the normal FuturesTest
 * test methods in these degenerate classloaders.
 */

@NullUnmarked

      return emptySet();
    }
  }

  /**
   * Creates a suite whose map has some elements filtered out of view.
   *
   * <p>Because the map may be ascending or descending, this test must derive the relative order of
   * these extreme values rather than relying on their regular sort ordering.
   */
  final TestSuite createSubmapSuite(
      FeatureSpecificTestSuiteBuilder<

# above is tagged with "py3-none-any". We cannot change the tags by simply
# renaming the wheels as uploading to PyPI would fail with "File already exists"
# error. In order to upload to PyPI, we unpack the wheel and change the tag
# inside a metadata file to the one we want (e.g cp38-cp38-win_amd) and then
# re-pack it to generate it as a platform specific wheel with this new wheel
#tag.

 * }
 *
 * <p>Note that this try-catch-finally block is not equivalent to a try-catch-finally block using
 * try-with-resources. To get the equivalent of that, you must wrap the above code in <i>another</i>
 * try block in order to catch any exception that may be thrown (including from the call to {@code
 * close()}).
 *
 * <p>This pattern ensures the following:
 *
 * <ul>

  }

  /*
   * The behavior of toString() and iteration is tested by LinkedHashMultiset,
   * which shares a lot of code with HashMultiset and has deterministic
   * iteration order.
   */

    /**
     * Select the best available RDMA provider
     *
     * @return best RDMA provider, or null if none available
     */
    public static RdmaProvider selectBestProvider() {
        // Try providers in order of preference
        List<RdmaProvider> providers = Arrays.asList(new DisniRdmaProvider(), // InfiniBand/RoCE - highest performance
                new TcpRdmaProvider() // TCP fallback - always available
        );

    // useful testable guarantees about the behavior of that dispatcher in a multithreaded
    // environment. Here we simply test that all the expected dispatches happened in some order.
    assertThat(dispatchedSubscribers).containsExactly(i1, i2, i3, s1, s1, s1, s1, s2, s2, s2, s2);
  }

  public void testImmediateDispatcher() {
    dispatcher = Dispatcher.immediate();

                    return new StandardImplForwardingSortedMultiset<>(
                        TreeMultiset.create(asList(elements)));
                  }

                  @Override
                  public List<String> order(List<String> insertionOrder) {
                    return Ordering.natural().sortedCopy(insertionOrder);
                  }
                })
            .named("ForwardingSortedMultiset with standard impls")

    @Override
    public Double[] createArray(int length) {
      return new Double[length];
    }

    /** Returns the original element list, unchanged. */
    @Override
    public List<Double> order(List<Double> insertionOrder) {
      return insertionOrder;
    }
  }

  public static class SampleDoubles extends SampleElements<Double> {
    public SampleDoubles() {
      super(0.0, 1.0, 2.0, 3.0, 4.0);