TF_CAPI_EXPORT extern TF_Output TF_OperationInput(TF_Input oper_in);

// Get list of all inputs of a specific operation.  `inputs` must point to
// an array of length at least `max_inputs` (ideally set to
// TF_OperationNumInputs(oper)).  Beware that a concurrent
// modification of the graph can increase the number of inputs of
// an operation.
TF_CAPI_EXPORT extern void TF_OperationAllInputs(TF_Operation* oper,

   * without contention. Because assignment of entries to these partitions is not necessarily
   * uniform, the actual concurrency observed may vary. Ideally, you should choose a value to
   * accommodate as many threads as will ever concurrently modify the table. Using a significantly
   * higher value than you need can waste space and time, and a significantly lower value can lead

   * without contention. Because assignment of entries to these partitions is not necessarily
   * uniform, the actual concurrency observed may vary. Ideally, you should choose a value to
   * accommodate as many threads as will ever concurrently modify the table. Using a significantly
   * higher value than you need can waste space and time, and a significantly lower value can lead

	}

	if srcVolume != minioMetaMultipartBucket {
		// srcFilePath is some-times minioMetaTmpBucket, an attempt to
		// remove the temporary folder is enough since at this point
		// ideally all transaction should be complete.
		Remove(pathutil.Dir(srcFilePath))
	} else {
		s.deleteFile(srcVolumeDir, pathutil.Dir(srcFilePath), true, false)
	}
	return res, nil
}

In case, `access` format was used, then `minio_events` would be a list, and the MinIO server would have performed an `RPUSH` to append to the list. A consumer of this list would ideally use `BLPOP` to remove list items from the left-end of the list.

## Publish MinIO events via NATS

Install NATS from [here](http://nats.io/).

### Step 1: Add NATS endpoint to MinIO

    private static final String ANSI_RESET = "\u001B\u005Bm";

    private void configure(CliRequest cliRequest) throws Exception {
        //
        // This is not ideal but there are events specifically for configuration from the CLI which I don't
        // believe are really valid but there are ITs which assert the right events are published so this
        // needs to be supported so the EventSpyDispatcher needs to be put in the CliRequest so that

         * the new one. The result is a ConcurrentModificationException or other bad behavior.
         *
         * (If we decide that we really, really hate allocating two Iterators per cycle instead of
         * one, we can optimistically store the new Iterator and then be willing to throw it out if
         * the user calls remove().)
         */

         * the new one. The result is a ConcurrentModificationException or other bad behavior.
         *
         * (If we decide that we really, really hate allocating two Iterators per cycle instead of
         * one, we can optimistically store the new Iterator and then be willing to throw it out if
         * the user calls remove().)
         */

      // and hence expect the input to be forwarded to the return tensor.
      EXPECT_EQ(orig_ptr, TF_TensorData(t));
    } else {
      // In sync mode, forwarding can't really happen since the client code will
      // have a reference count on the input tensor while the kernel is being
      // executed and thus it cannot be re-used for the return tensor.
      EXPECT_NE(orig_ptr, TF_TensorData(t));

      throw new NullPointerException(Platform.lenientFormat(errorMessageTemplate, p1, p2, p3, p4));
    }
    return reference;
  }

  /*
   * All recent hotspots (as of 2009) *really* like to have the natural code
   *
   * if (guardExpression) {
   *    throw new BadException(messageExpression);
   * }
   *