SEZ9 commented on issue #12966:
URL: https://github.com/apache/seatunnel/issues/12966#issuecomment-6406773265

   Thanks @SEPURI-SAI-KRISHNA, both points are clear.
   
   **1. Zero-length assertion.** The test as pasted in `8ad46124f` is what I 
was after: asserting that an empty value remains a decode error and produces no 
row, without pinning the exact exception type from the Kafka converter, is the 
right level. The mutation check (widening the guard to `msg.value() == null || 
msg.value().length == 0` fails only that test, reverting restores 3 of 3 green) 
is exactly the evidence I wanted. Nothing more needed on this one.
   
   **2. IT as a follow-up.** Agreed on keeping it as its own PR rather than 
folding it into #12967, since #12927 has merged as `befdd3655` and the emitter 
skip branch no longer dereferences a null value. The shape you describe is 
good: own topic, a tombstone between two blocks of real records, an assert job 
that fails unless every real record arrives, driving both 
`format_error_handle_way` values, and asserting the offset advanced past the 
tombstone rather than relying on row count alone. The offset assertion is the 
important part, because a replay after restart is the failure this issue is 
about.
   
   Two concrete asks so I can close the loop:
   
   - Please link the follow-up PR here once it is open so this issue references 
both the format fix (#12967) and the IT.
   - Confirm in that PR that the new `compatible_kafka_connect_json` config 
exercises the same guard path as the unit tests (tombstone returns before 
`convertToSinkRecord`), so the IT and the unit tests are covering the same code 
rather than two separate paths.
   
   With #12967 carrying the guard, the tombstone and well-formed-record tests, 
and the zero-length assertion, I have no further asks on that PR from my side.
   
   <!-- streview-comment:2036 -->


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to