model-writtenrun 9226 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Daniel Gustafsson opened the thread with #1 after his separate catalog multi-insert work triggered a logical-decoding assertion, proposing to adjust the assertion in DecodeMultiInsert().
- Heikki Linnakangas observed in #2 that the proposed assertion was stricter, prompting Daniel to post a revised expression in #3 that he reported testing both with his catalog changes and on HEAD.
- Michael Paquier identified incorrect tuple-data traversal as the underlying problem in #4 and supplied a replacement that scans every tuple entry regardless of XLH_INSERT_CONTAINS_NEW_TUPLE, preserves the final length assertion, and asserts that tupledata is non-NULL.
- Daniel tested Michael's replacement and reported success in #5, supporting its adoption independently of the separate catalog multi-insert work.
- Michael committed #4 as 75c1921cd6 with an added explanatory comment; his closing message #6 confirmed the fix had landed and deferred handling potentially absent tuple data for catalog multi-inserts to Daniel's separate patch.