model-writtenrun 7555 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Tom Lane reported that heap_update() passed otid to its SSI conflict check despite a preceding warning to use oldtup because otid can alias newtup->t_self #1.
- Thomas Munro supplied a patch replacing that argument with &oldtup.t_self, the saved original tuple location #2.
- Munro's debugging found no differing tuple locations at the check in regression or isolation tests, and he observed that newtup->t_self appeared to be updated only afterward by RelationPutHeapTuple() #2.
- The patch landed with identical code as b1df6b696b, whose description clarified that the change removes a potential hazard and follows the existing comment rather than fixing a live bug; it was also backpatched to PostgreSQL 13.
- Munro closed the thread by confirming that he had pushed the fix, leaving no outstanding patch revisions #3.