[
https://issues.apache.org/jira/browse/ARTEMIS-2712?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17084089#comment-17084089
]
Robbie Gemmell commented on ARTEMIS-2712:
-----------------------------------------
To be clear, this isnt about connection drop, but explicitly aborted deliveries:
[https://github.com/clebertsuconic/activemq-artemis/blob/d27d61f223fe88fd01f8d98415ddadb75605d374/artemis-protocols/artemis-amqp-protocol/src/main/java/org/apache/activemq/artemis/protocol/amqp/proton/ProtonServerReceiverContext.java#L290]
There is a new retained reference to the 'current large message' which is not
cleared up during an aborted transfer, so if one were aborted it seems like the
next arriving message would appear to reuse the old message bits.
Testing will be awkward as proton-j cant generate aborted transfers.
> updated large message handling not accounted for in aborted message cleanup
> ---------------------------------------------------------------------------
>
> Key: ARTEMIS-2712
> URL: https://issues.apache.org/jira/browse/ARTEMIS-2712
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Components: AMQP
> Affects Versions: 2.12.0
> Reporter: Robbie Gemmell
> Priority: Blocker
>
> ARTEMIS-1975 updated the message handling for 'larger' AMQP messages, but in
> doing so introduces additional behaviours while handling such messages. These
> dont appear to have been similarly catered for during aborted message cleanup
> (irony: those being more likely for this to be needed). As such it currently
> looks like message corruption could occur if a subsequent message arrived
> after an abort.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
