// Why 1000?  WHY NOT!
  private static final int MAX_BUSY_WAIT_SPINS = 1000;

  @Override
  public final void run() {
    /*
     * Set runner thread before checking isDone(). If we were to check isDone() first, the task
     * might be cancelled before we set the runner thread. That would make it impossible to
     * interrupt, yet it will still run, since interruptTask will leave the runner value null,
     * allowing the CAS below to succeed.

            //
            // We promise this because, once we cancel that Future, we would no longer be able to
            // tell when any underlying work it is doing is done. Thus, we might start a new task
            // while that underlying work is still running.
            //
            // So that is why we cancel only in the case of CAS success.
            taskFuture.cancel(false);

        /*
         * We don't *need* to use reflection to access Optional: It's available on all JDKs we
         * support, and Android code won't get this far, anyway, because ProcessHandle is
         * unavailable. But given how much other reflection we're using, we might as well use it
         * here, too, so that we don't need to also suppress an AndroidApiChecker error.
         */

}

// freezeServices will freeze all incoming S3 API calls.
// For each call, unfreezeServices must be called once.
func freezeServices() {
	// Use atomics for globalServiceFreeze, so we can read without locking.
	// We need a lock since we are need the 2 atomic values to remain in sync.
	globalServiceFreezeMu.Lock()
	// If multiple calls, first one creates channel.
	globalServiceFreezeCnt++
	if globalServiceFreezeCnt == 1 {

		// Note: we should not be defer'ing the following closeBitrotReaders() call as
		// we are inside a for loop i.e if we use defer, we would accumulate a lot of open files by the time
		// we return from this function.
		closeBitrotReaders(readers)
		if err != nil {
			// If we have successfully written all the content that was asked
			// by the client, but we still see an error - this would mean

        <property name="ignoreFormat" value="^(threadPool)$"/>
    </module>
    -->

    <!-- We don't use Java's builtin serialization and we suppress all warning
      about it. The flip side of that coin is that we shouldn't _try_ to use
      it. We can't outright ban it with ForbiddenApis because it complain about
      every we reference a class that implements Serializable like String or
      Exception.
      -->

    return new SingletonImmutableSet<>(e1);
  }

  /*
   * TODO: b/315526394 - Skip the Builder entirely for the of(...) methods, since we don't need to
   * worry that we might trigger the fallback to the JDK-backed implementation? (The varargs one
   * _could_, so we could keep it as it is. Or we could convince ourselves that hash flooding is
   * unlikely in practice there, too.)
   */

  /**

## Consequences

* X
* Y
* Z
```

## Consequences

* We start to use Architectural Decision Records
* We use the proposed template from this ADR
* We locate `.md` files in the folder `/architecture/standards`
* We highly encourage usage of ADR to communicate decisions

   * methods to accept null. We could avoid having unchecked casts in Converter.java itself if we
   * could perform a cast to LegacyConverter, but we can't because it's an internal-only class.
   *
   * TODO(cpovirk): So make it part of the open-source build, albeit package-private there?
   *
   * So we use uncheckedCastNullableTToT here. This is a weird usage of that method: The method is

 * </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
@GwtIncompatible