Hi I.P., Thank you for the report and the reproducer; the README made it easy to follow.
I confirmed it: your reproducer fails on 2.18.0, and a new test fails the same way on master. The remote node's events reach the listener as BinaryObjectImpl, while the local node's events arrive as the transformer's class. Ticket: https://issues.apache.org/jira/browse/IGNITE-29226 If a committer agrees with the issue, I'm happy to work on a fix. Best regards, Reza Taroosheh On Wed, Oct 28, 2026 at 1:31 PM <[email protected]> wrote: > hi devs > > I'm back with two more complicated cases, which require your review and > confirmation. I'm going to split them into two emails. > > So the first one is this one and is about a potential problem > with ContinuousQueryWithTransformer that provides a remote node's > transformed events as BinaryObjects on keep-binary caches. And only other > nodes' events arrive as BinaryObjects. Events for the node's own primary > keys arrive correctly as the transformer's class, so a single listener gets > a mix of both. > Expected behavior would be that the listener receives exactly what the > transformer returned, for every event, whichever node it came from. > > As always - the reproducer is attached with a detailed README.md > description. > > Please have a look and if you confirm a defect, please provide a JIRA > ticket number. > Thank you in advance. > > Best regards, > I.P. >
