write/flush after response handler has returned.
```

```
λ git cherry-pick 4f3317effea38c203c358af9cb5ce3c0e4173976
```

    }

    @Override
    public Iterator<E> descendingIterator() {
      assertTrue(Thread.holdsLock(mutex));
      return delegate.descendingIterator();
    }

    private static final long serialVersionUID = 0;
  }

  @SuppressWarnings("CheckReturnValue")
  public void testHoldsLockOnAllOperations() {
    create().element();
    create().offer("foo");
    create().peek();
    create().poll();

    Serialization.populateMap(this, stream);
  }

  @GwtIncompatible // not needed in emulated source.
  private static final long serialVersionUID = 0;

      ImmutableSet<C> columnSpace) {
    // use a dense table if more than half of the cells have values
    // TODO(gak): tune this condition based on empirical evidence
    return (cellList.size() > (((long) rowSpace.size() * columnSpace.size()) / 2))
        ? new DenseImmutableTable<R, C, V>(cellList, rowSpace, columnSpace)
        : new SparseImmutableTable<R, C, V>(cellList, rowSpace, columnSpace);
  }

  public void testHash_hashesCorrectlyForSkipping() throws Exception {
    HashCode expectedHash = Hashing.md5().hashBytes(new byte[] {'m', 's'});
    HashingInputStream in = new HashingInputStream(Hashing.md5(), buffer);

    long numOfByteSkipped = in.skip(2);
    assertEquals(2, numOfByteSkipped);

    byte[] buf = new byte[4];
    int numOfByteRead = in.read(buf, 0, buf.length);
    assertEquals(2, numOfByteRead);

   * currently skipping MediumTests on Android, and we skip them by not making them present at
   * runtime at all. I could just make _this_ test a MediumTest, but then it wouldn't run on
   * Android.... The right long-term fix is probably to get MediumTests running under Android by
   * default and then suppress them strategically as needed.
   */
  public static final class ConcretePackageSanityTests extends AbstractPackageSanityTests {}

  public void testHash_hashesCorrectlyForSkipping() throws Exception {
    HashCode expectedHash = Hashing.md5().hashBytes(new byte[] {'m', 's'});
    HashingInputStream in = new HashingInputStream(Hashing.md5(), buffer);

    long numOfByteSkipped = in.skip(2);
    assertEquals(2, numOfByteSkipped);

    byte[] buf = new byte[4];
    int numOfByteRead = in.read(buf, 0, buf.length);
    assertEquals(2, numOfByteRead);

      checkNotNull(fromKey);
      return new StandardRowSortedTable<R, C, V>(sortedBackingMap().tailMap(fromKey), factory)
          .rowMap();
    }
  }

  private static final long serialVersionUID = 0;

      aboutToAcquire(this);
      try {
        return super.tryLock();
      } finally {
        lockStateChanged(this);
      }
    }

    @Override
    public boolean tryLock(long timeout, TimeUnit unit) throws InterruptedException {
      aboutToAcquire(this);
      try {
        return super.tryLock(timeout, unit);
      } finally {
        lockStateChanged(this);
      }
    }

    assertThrows(IndexOutOfBoundsException.class, () -> sink.putBytes(new byte[8], 0, 16));
    assertThrows(IndexOutOfBoundsException.class, () -> sink.putBytes(new byte[8], 0, -1));
  }

  /**
   * This test creates a long random sequence of inputs, then a lot of differently configured sinks
   * process it; all should produce the same answer, the only difference should be the number of
   * process()/processRemaining() invocations, due to alignment.