[
https://issues.apache.org/jira/browse/QPIDJMS-325?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16173225#comment-16173225
]
Robbie Gemmell commented on QPIDJMS-325:
----------------------------------------
I don't entirely agree with your assessment, though I'm not necessarily against
changing things to be more consistent.
I think ActiveMQ's StreamMessage behaviour is incorrect, as the javadoc says it
is, however I don't think either clients behaviour is invalid for BytesMessage.
The definition there is much less specific and doesn't so strictly state what
should occur in that case whereas it does for StreamMessage, which in part
makes sense because the two are actually fairly different. In StreamMessage,
the precise behaviour around the start and end of a readBytes process is far
more important because there could be multiple such elements in the
StreamMessage back to back, and mixed with other types of element, and the end
of the message or wrong element type being encountered causes exceptions, so
you need to better demarcate which element is being read and that one elements
inspection is complete before the next begins etc. With BytesMessage its just a
single group of bytes, the end is the end, and no exception is thrown at the
end of the message so it makes less difference what is returned, which is
presumably why the javadoc is so much less specific.
> reading from empty buffer of a BytesMessage should return 0, not -1
> -------------------------------------------------------------------
>
> Key: QPIDJMS-325
> URL: https://issues.apache.org/jira/browse/QPIDJMS-325
> Project: Qpid JMS
> Issue Type: Bug
> Components: qpid-jms-client
> Affects Versions: 0.25.0
> Reporter: Jiri Danek
> Priority: Minor
>
> Consider the following test. According to
> http://docs.oracle.com/javaee/7/api/javax/jms/BytesMessage.html#readBytes-byte:A-
> the #readBytes method should return 0 when it is first called, as the number
> of bytes in the buffer is 0 and it read 0 bytes. Only on subsequent calls it
> should return -1. What happens now is that the method returns -1 the first
> time.
> See the commented lines to try the same thing with ActiveMQ JMS Client
> library, and with StreamMessage instead of BytesMessage. The behavior there
> should be the same.
> ActiveMQ passes the test with BytesMessage and fails it with StreamMessage.
> Qpid JMS fails with BytesMessage and passes with StreamMessage.
> {code}
> import org.apache.activemq.ActiveMQConnectionFactory;
> import org.apache.qpid.jms.JmsConnectionFactory;
> import org.junit.Test;
> import javax.jms.*;
> import static com.google.common.truth.Truth.assertThat;
> public class EmptyBufferInputTest {
> @Test
> public void testEmptyBufferInput() throws JMSException {
> // ConnectionFactory connectionFactory = new
> ActiveMQConnectionFactory("tcp://127.0.0.1:61616");
> JmsConnectionFactory connectionFactory = new
> JmsConnectionFactory("amqp://127.0.0.1:5672");
> Connection connection = connectionFactory.createConnection();
> Session session = connection.createSession(false,
> Session.AUTO_ACKNOWLEDGE);
> final byte[] BYTE_LIST = {1, 2, 4};
> // StreamMessage message = session.createStreamMessage();
> BytesMessage message = session.createBytesMessage();
> message.clearBody();
> byte[] readList = new byte[BYTE_LIST.length - 1];
> byte[] emptyList = {};
> message.writeBytes(emptyList);
> message.reset();
> final int IS_EMPTY = 0;
> assertThat(message.readBytes(readList)).isEqualTo(IS_EMPTY);
> }
> }
> {code}
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
