could use new flow control capacity before acknowledging it, causing strict HTTP/2 servers to
    fail the call.

 *  Fix: Recover gracefully when a coalesced connection immediately goes unhealthy.

## Version 3.14.2

_2019-05-19_

 *  Fix: Lock in a route when recovering from an HTTP/2 connection error. We had a bug where two
    calls that failed at the same time could cause OkHttp to crash with a `NoSuchElementException`

    key: String,
    index: Int,
  ) = cacheDir / "$key.$index.tmp"

  private fun readFile(file: Path): String {
    return filesystem.read(file) {
      readUtf8()
    }
  }

  private fun readFileOrNull(file: Path): String? {
    return try {
      filesystem.read(file) {
        readUtf8()
      }
    } catch (_: FileNotFoundException) {
      null
    }
  }

* Add subPath to mount a child dir or file of a volumeMount ([#22575](https://github.com/kubernetes/kubernetes/pull/22575), [@MikaelCluseau](https://github.com/MikaelCluseau))
* Handle image digests in node status and image GC ([#25088](https://github.com/kubernetes/kubernetes/pull/25088), [@ncdc](https://github.com/ncdc))

    val out1 = stream1.getSink().buffer()
    out1.write(ByteArray(Settings.DEFAULT_INITIAL_WINDOW_SIZE))
    out1.flush()

    // Check that we've filled the window for both the stream and also the connection.
    assertThat(connection.writeBytesTotal)
      .isEqualTo(Settings.DEFAULT_INITIAL_WINDOW_SIZE.toLong())
    assertThat(connection.writeBytesMaximum)

# the openxla/xla repo, also include debug info for files under xla/.
build:dbg --per_file_copt=+.*,-tensorflow.*,-xla.*@-g0
build:dbg --per_file_copt=+tensorflow/core/kernels.*@-g0
# for now, disable arm_neon. see: https://github.com/tensorflow/tensorflow/issues/33360
build:dbg --cxxopt -DTF_LITE_DISABLE_X86_NEON

event store. The persistent store will backup events when the AMQP broker goes offline and replays it when the broker comes back online. The event store can be configured by setting the directory path in `queue_dir` field and the maximum limit of events in the queue_dir in `queue_limit` field. For eg, the `queue_dir` can be `/home/events` and `queue_limit` can be `1000`. By default, the `queue_limit` is set to 100000.

To update the configuration, use `mc admin config get notify_amqp` command...

    get_compat_files(TENSORFLOW_API_INIT_FILES_V1, 1) + get_nested_compat_files([
        1,
        2,
    ])
)

TENSORFLOW_API_INIT_FILES_V1 = (
    TENSORFLOW_API_INIT_FILES_V1 +
    get_compat_files(TENSORFLOW_API_INIT_FILES, 2) +
    get_compat_files(TENSORFLOW_API_INIT_FILES_V1, 1) + get_nested_compat_files([
        1,
        2,
    ])
)
# copybara:comment_end