Github user hequn8128 commented on the issue:
https://github.com/apache/flink/pull/5094
Thanks for your replies.
I agree with that it is valid to join late input data.What I concern is the
watermark has not been hold back correctly.
Take `testRowTimeJoinWithCommonBounds2` as an example.
1. Output watermark with timestamp 1000 (hold back with 5000 from 6000)
2. Output record1 with timestamp 6000 (valid)
3. Output record2 with timestamp 1000 (invalid)
The record2 is outputted invalidly because it's timestamp is equal to the
previous output watermark (1000). In my pr, I hold the input watermark back
with 5001 to make record2 valid. So the test will output with:
1. Output watermark with timestamp 999 (hold back with 5001 from 6000)
2. Output record1 with timestamp 6000 (valid)
3. Output record2 with timestamp 1000 (valid)
---
