| Ondra Chaloupka Yes, IMO it is very weird to have the after failure observers invoked for an issue with placing the synchronisation during committing phase. See also my recent comment on https://issues.jboss.org/browse/CDI-724?focusedCommentId=13643091&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-13643091 . I think we really should place one synchronisation during TX.active , and then process the placed events from a queue there, to work around such odd edge cases |
