throw e;
    }
  }

  /**
   * Returns a string containing the result of {@link Throwable#toString() toString()}, followed by
   * the full, recursive stack trace of {@code throwable}. Note that you probably should not be
   * parsing the resulting string; if you need programmatic access to the stack frames, you can call
   * {@link Throwable#getStackTrace()}.
   */
  @GwtIncompatible // java.io.PrintWriter, java.io.StringWriter

       *
       * 1. On Android, querying a WeakReference blocks if the GC is doing an otherwise-concurrent
       * pass.
       *
       * 2. We would probably choose to compare exceptions using == instead of equals() (for
       * consistency with how weak references are cleared). That's a behavior change -- arguably the
       * removal of a feature.
       *

This is especially the case for user models, because:

* The **input model** needs to be able to have a password.
* The **output model** should not have a password.
* The **database model** would probably need to have a hashed password.

/// danger

Never store user's plaintext passwords. Always store a "secure hash" that you can then verify.

      try {
        File testFile = new File(testFileUrl.toURI());
        testDir = testFile.getParentFile(); // the testdata directory
      } catch (Exception ignore) {
        // probably URISyntaxException or IllegalArgumentException
        // fall back to copying URLs to files in the testDir == null block below
      }
    }

    if (testDir == null) {

    //
    // All this makes me really wonder if there's any value in queueing here at all. A dispatcher
    // that simply loops through the subscribers and dispatches the event to each would actually
    // probably provide a stronger order guarantee, though that order would obviously be different
    // in some cases.

    /** Global event queue. */

  }

  @GwtIncompatible // different overflow behavior; could probably be made to work by using ~~
  public void testConcat_overflow_negative() {
    int dim1 = 1 << 16;
    int dim2 = 1 << 15;
    assertThat(dim1 * dim2).isLessThan(0);
    testConcatOverflow(dim1, dim2);
  }

  @GwtIncompatible // different overflow behavior; could probably be made to work by using ~~
  public void testConcat_overflow_nonNegative() {

    try {
      return Thread.class.getConstructor(
          ThreadGroup.class, Runnable.class, String.class, long.class, boolean.class);
    } catch (Throwable t) {
      // Probably pre Java 9. We'll fall back to Thread.inheritableThreadLocals.
      return null;
    }
  }

          Comparator<? super E> comparator = spliterator.getComparator();
          if (comparator == null) {
            // A sorted spliterator with no comparator is already using natural order.
            // (We could probably find a way to avoid rawtypes here if we wanted.)
            @SuppressWarnings({"unchecked", "rawtypes"})
            Comparator<? super E> naturalOrder =

          Comparator<? super E> comparator = spliterator.getComparator();
          if (comparator == null) {
            // A sorted spliterator with no comparator is already using natural order.
            // (We could probably find a way to avoid rawtypes here if we wanted.)
            @SuppressWarnings({"unchecked", "rawtypes"})
            Comparator<? super E> naturalOrder =

   * This is useful if the function instance already exists, or so that you can supply a lambda
   * expressions. If those circumstances don't apply, you probably don't need to use this; subclass
   * {@code TreeTraverser} and implement its {@link #children} method directly.
   *
   * @since 20.0