the request already contains an `Accept` header.

 *  Fix: Don't crash with a `NullPointerException` if a server sends a close while we're sending a
    ping. OkHttp had a race condition bug.


## Version 4.6.0

_2020-04-28_

 *  Fix: Follow HTTP 307 and 308 redirects on methods other than GET and POST. We're reluctant to

    assertThat(hostnameVerifier.calls).isEqualTo(
      Arrays.asList("verify android.com"),
    )
  }

  /** Tolerate bad https proxy response when using HttpResponseCache. Android bug 6754912.  */
  @Test
  fun connectViaHttpProxyToHttpsUsingBadProxyAndHttpResponseCache() {
    initResponseCache()
    server.useHttps(handshakeCertificates.sslSocketFactory())

    // javax.net.ssl.SSLHandshakeException: No new session is allowed and no existing
    // session can be resumed
    //
    // Report https://bugs.java.com/bugdatabase/view_bug.do?bug_id=JDK-8264944
    // Sessions improvement https://bugs.java.com/bugdatabase/view_bug.do?bug_id=JDK-8245576
    if (!platform.isJdk9() && !platform.isOpenJsse() && !platform.isJdk8Alpn()) {
      reuseSession = true
    }

    "EEE dd-MMM-yyyy HH-mm-ss z",
    "EEE dd-MMM-yy HH:mm:ss z",
    "EEE dd MMM yy HH:mm:ss z",
    "EEE,dd-MMM-yy HH:mm:ss z",
    "EEE,dd-MMM-yyyy HH:mm:ss z",
    "EEE, dd-MM-yyyy HH:mm:ss z",
    // RI bug 6641315 claims a cookie of this format was once served by www.yahoo.com:
    "EEE MMM d yyyy HH:mm:ss z",
  )

private val BROWSER_COMPATIBLE_DATE_FORMATS =

    for a graceful shutdown before it performs an abrupt shutdown.

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

 *  Fix in `okhttp-coroutines`: Don't leak response bodies in `executeAsync()`. We had a bug where
    we didn't call `Response.close()` if the coroutine was canceled before its response was
    returned.

 *  Upgrade: [Okio 3.9.0][okio_3_9_0].

 *  Upgrade: [Kotlin 1.9.23][kotlin_1_9_23].

    }.also { expected ->
      assertThat(expected.message).isEqualTo("Too many follow-up requests: 21")
    }
  }

  /**
   * We had a bug where we were passing a null route to the authenticator.
   * https://github.com/square/okhttp/issues/3809
   */
  @Test
  fun authenticateWithNoConnection() {
    server.enqueue(
      MockResponse(

# IDN-ized. This is "Go" in fullwidth UTF-8/UTF-16.
http://\uff27\uff4f.com  s:http p:/ h:go.com

# URL spec forbids the following.
# https://www.w3.org/Bugs/Public/show_bug.cgi?id=24257
http://\uff05\uff14\uff11.com
http://%ef%bc%85%ef%bc%94%ef%bc%91.com

# ...%00 in fullwidth should fail (also as escaped UTF-8 input)
http://\uff05\uff10\uff10.com
http://%ef%bc%85%ef%bc%90%ef%bc%90.com

    Thread.sleep(500)
    server.send("Hello, WebSockets!")
    clientListener.assertTextMessage("Hello, WebSockets!")
    closeWebSockets(webSocket, server)
  }

  /**
   * We had a bug where web socket connections were leaked if the HTTP connection upgrade was not
   * successful. This test confirms that connections are released back to the connection pool!
   * https://github.com/square/okhttp/issues/4258
   */

      val reference = references[i]

      if (reference.get() != null) {
        i++
        continue
      }

      // We've discovered a leaked call. This is an application bug.
      val callReference = reference as CallReference
      val message =
        "A connection to ${connection.route().address.url} was leaked. " +
          "Did you forget to close a response body?"

    assertThat(response3.body.string()).isEqualTo("def")
    assertThat(server.takeRequest().sequenceNumber).isEqualTo(1) // New connection.
  }

  /**
   * We had a bug where we'd perform infinite retries of route that fail with connection shutdown
   * errors. The problem was that the logic that decided whether to reuse a route didn't track