Github user tzulitai commented on the issue:
https://github.com/apache/flink/pull/5568
@aljoscha I've made a few extra changes, and made sure that the end-to-end
test fails properly without the fix:
- use the correct data formats in the Kafka e2e test
- allows the e2e test to print error logs (in this case, the logs would
have a `ClassNotFoundException: CustomWatermarkAssigner` without this PR's fix)
in case of a timeout interruption. This should improve visibility in a failure
scenario.
Could you have another last look?
---
