edge-case regression introduced with the events cleanup in 3.14.0.

 *  Fix: Provide actionable advice when the exchange is non-null. Prior to 3.14, OkHttp would
    silently leak connections when an interceptor retries without closing the response body. With
    3.14 we detect this problem but the exception was not helpful.

## Version 3.14.0

_2019-03-14_

    only does when assertions are enabled.)

 *  Fix: Honor `RequestBody.isOneShot()` in `HttpLoggingInterceptor`.


## Version 4.2.2

_2019-10-06_

 *  Fix: When closing a canceled HTTP/2 stream, don't send the `END_STREAM` flag. This could cause
    the server to incorrectly interpret the stream as having completed normally. This is most useful
    when a request body needs to cancel its own call.

    when explicitly forbidden.
 *  Fix: Throw a `SSLPeerUnverifiedException` when host verification fails.
 *  Fix: MockWebServer explicitly closes sockets. (On some Android releases,
    closing the input stream and output stream of a socket is not sufficient.
 *  Fix: Buffer outgoing HTTP/2 frames to limit how many outgoing frames are
    created.
 *  Fix: Avoid crashing when cache writing fails due to a full disk.

    }

    @Override
    public void close() {
        if (httpClient == null) {
            return;
        }
        if (logger.isDebugEnabled()) {
            logger.debug("Closing HcHttpClient...");
        }
        if (connectionMonitorTask != null) {
            connectionMonitorTask.cancel();
        }
        if (httpClient != null) {
            try {

                if ( ssn.getExpiration() != null && ssn.getExpiration() < now && !ssn.isInUse() ) {
                    if ( log.isDebugEnabled() ) {
                        log.debug("Closing session after timeout " + ssn);
                    }
                    ssn.logoff(false, false);
                }
            }
        }

is sequenced before the send on <code>c</code>, which is synchronized before
the corresponding receive on <code>c</code> completes, which is sequenced before
the <code>print</code>.
</p>

<p class="rule">
The closing of a channel is synchronized before a receive that returns a zero value
because the channel is closed.
</p>

<p>
In the previous example, replacing
<code>c &lt;- 0</code> with <code>close(c)</code>

  }

  @SuppressWarnings("CatchingUnchecked") // sneaky checked exception
  private static void closeAll(BaseStream<?, ?>[] toClose) {
    // If one of the streams throws an exception, continue closing the others, then throw the
    // exception later. If more than one stream throws an exception, the later ones are added to the
    // first as suppressed exceptions. We don't catch Error on the grounds that it should be allowed

type GetObjectReader struct {
	io.Reader
	ObjInfo    ObjectInfo
	cleanUpFns []func()
	once       sync.Once
}

// WithCleanupFuncs sets additional cleanup functions to be called when closing
// the GetObjectReader.
func (g *GetObjectReader) WithCleanupFuncs(fns ...func()) *GetObjectReader {
	g.cleanUpFns = append(g.cleanUpFns, fns...)
	return g
}

	// errors on a previous write(2) operation are reported only on the final close() that releases
	// the open file descriptor.
	//
	// Failing to check the return value when closing a file may lead to silent loss of data.
	// This can especially be observed with NFS and with disk quota.
	return w.Close()
}

## Changelog since v1.13.7

### Other notable changes

* Fix possible fd leak and closing of dirs in doSafeMakeDir  ([#79534](https://github.com/kubernetes/kubernetes/pull/79534), [@odinuge](https://github.com/odinuge))