.isWithin(ALLOWED_ERROR)
          .of(expectedLargeDatasetPercentile(index));
    }

    // Assert that the dataset contains the same elements after the in-place computation (although
    // they may be reordered). We only do this for one index rather than for all indexes, as it is
    // quite expensive (quadratic in the size of PSEUDORANDOM_DATASET).

          .isWithin(ALLOWED_ERROR)
          .of(expectedLargeDatasetPercentile(index));
    }

    // Assert that the dataset contains the same elements after the in-place computation (although
    // they may be reordered). We only do this for one index rather than for all indexes, as it is
    // quite expensive (quadratic in the size of PSEUDORANDOM_DATASET).

 *  Upgrade: [Unicode® IDNA 15.1.0][idna_15_1_0]


## Version 5.0.0-alpha.12

_2023-12-17_

We took too long to cut this release and there's a lot of changes in it. We've been busy.

Although this release is labeled _alpha_, the only unstable thing in it is our new APIs. This
release has many critical bug fixes and is safe to run in production. We're eager to stabilize our
new APIs so we can get out of alpha.

 *   <li>Access to all collection-based utilities via {@link #asList} (though at the cost of
 *       allocating garbage).
 * </ul>
 *
 * <p>Disadvantages compared to {@code int[]}:
 *
 * <ul>
 *   <li>Memory footprint has a fixed overhead (about 24 bytes per instance).
 *   <li><i>Some</i> construction use cases force the data to be copied (though several construction
 *       APIs are offered that don't).

 *   <li>Access to all collection-based utilities via {@link #asList} (though at the cost of
 *       allocating garbage).
 * </ul>
 *
 * <p>Disadvantages compared to {@code long[]}:
 *
 * <ul>
 *   <li>Memory footprint has a fixed overhead (about 24 bytes per instance).
 *   <li><i>Some</i> construction use cases force the data to be copied (though several construction
 *       APIs are offered that don't).

   * abruptly. In other cases, a test may put a thread into an uninterruptible operation
   * intentionally, so there is no other way to clean up these threads. (The better solution,
   * though, would be to run the tests that use TestThread in separate VMs so that their threads
   * don't hang around during other tests.)
   */
  @Override
  public void tearDown() throws Exception {
    try {

  private static final class AvlNode<E extends @Nullable Object> {
    /*
     * For "normal" nodes, the type of this field is `E`, not `@Nullable E` (though note that E is a
     * type that can include null, as in a TreeMultiset<@Nullable String>).
     *
     * For the header node, though, this field contains `null`, regardless of the type of the
     * multiset.
     *

    tests themselves and not the code being tested that needs that access, though there's no
    obvious way to ensure that.

    We could consider arranging things so that only the tests we know need this would get
    the add-opens. Right now that doesn't seem worth the effort, though.
    -->
    <test.add.opens>
      --add-opens java.base/java.lang=ALL-UNNAMED

    tests themselves and not the code being tested that needs that access, though there's no
    obvious way to ensure that.

    We could consider arranging things so that only the tests we know need this would get
    the add-opens. Right now that doesn't seem worth the effort, though.
    -->
    <test.add.opens>
      --add-opens java.base/java.lang=ALL-UNNAMED

   * abruptly. In other cases, a test may put a thread into an uninterruptible operation
   * intentionally, so there is no other way to clean up these threads. (The better solution,
   * though, would be to run the tests that use TestThread in separate VMs so that their threads
   * don't hang around during other tests.)
   */
  @Override
  public void tearDown() throws Exception {
    try {