Github user jbertram commented on a diff in the pull request:
https://github.com/apache/activemq-artemis/pull/2187#discussion_r207218441
--- Diff:
artemis-core-client/src/main/java/org/apache/activemq/artemis/core/protocol/core/impl/ActiveMQSessionContext.java
---
@@ -189,16 +185,24 @@ public void kill() {
this.killed = true;
}
+ private void setHandlers() {
+
sessionChannel.setCommandConfirmationHandler(commandConfirmationHandler);
--- End diff --
I realize that your original intent was that either the confirmation or the
response handler would be set here rather than both, and this is definitely the
reason for the duplicate confirmation. However, I couldn't determine any other
way to maintain compatibility with existing bridge behavior without setting
both. If this method is changed back to your original semantic (i.e. only set
one) then lots of tests in
org.apache.activemq.artemis.tests.integration.cluster.bridge.BridgeTest fail.
org.apache.activemq.artemis.tests.integration.client.SendAckFailTest was broken
at some point as well. I will certainly admit that my solution here isn't the
most elegant, but I haven't found a more clever way of dealing with the failing
tests. I think eventually the whole confirmation implementation should be
refactored to deal with these issues, but I don't see how that's feasible in a
minor release.
---
