DanielLeens commented on PR #11222:
URL: https://github.com/apache/seatunnel/pull/11222#issuecomment-6473942002

   Sorry for not reconciling the vector consumer contracts in my earlier 
clearance. Thanks @SEZ9 for the concrete counterevidence. I rechecked 
CopyFieldTransform.cloneByteBuffer at lines 196-205, 
VectorUtils.toFloatArray/toShortArray at lines 57-83, and MilvusSinkConverter 
at lines 86-92. Copy allocates only remaining bytes, while those consumers use 
capacity or array(). A nonzero-position input therefore changes the copied 
dimension, and a direct binary-vector buffer cannot satisfy the array-backed 
Milvus path. Whether such an input is a supported source representation must be 
demonstrated, rather than assumed: some producer/converter paths already 
require a canonical full heap buffer. I am withdrawing the unconditional 
merge-ready wording while that contract is unresolved. +1 to resolving your 
buffer-contract concern with a test through the actual supported source -> Copy 
-> vector-consumer path, covering position/limit, direct/heap and byte order. 
Please either enforce and test th
 e canonical input contract at its ownership boundary or provide a compatible 
copy/consumer contract. Green CI alone does not close this discussion; this is 
a correction to the prior conclusion, not a second partial full review or a new 
claim of runtime validation.
   
   Gate snapshot (2026-10-28T15:59:52+00:00), head 
`ac877ada9dc2081f5f7df17cd677ece4209611c7`: 
[Build](https://github.com/apache/seatunnel/runs/122617822624) is `SUCCESS`.
   
   The green check does not clear the independent source/documentation or scope 
requests in my [previous 
conclusion](https://github.com/apache/seatunnel/pull/11222#pullrequestreview-5676327685).
 Those requests remain unchanged; this is not an approval.
   
   Comparison recorded in this scan: base `dev`, `diverged`, ahead 7 / behind 
15, no merge conflict. Branch drift is metadata here and is not an additional 
sync request.
   


-- 
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