model-writtenrun 8208 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Antonin Houska questioned why xl_heap_header's comment claimed that storing t_hoff was free due to alignment, given the WAL code's use of memcpy #1; Andres Freund explained structure padding #2, while Tom Lane observed that adding t_hoff to the otherwise four-byte structure was not free and recommended correcting the comment without removing the field #3.
- Antonin confirmed that his concern was the explanation rather than the field's existence #4, and Bruce Momjian later posted a patch replacing the alignment claim with a statement that t_hoff could potentially be recomputed #5.
- David G. Johnston recommended removing “potentially” or deleting the sentence #6, prompting Bruce to revise the comment to say that t_hoff could be recomputed each time it was needed #7.
- Tom advocated deleting the sentence altogether because removing the field was unlikely to be a worthwhile WAL optimization #8; Antonin proposed explaining why the field remained #9, then withdrew that request in favor of the simpler changes #10.
- Bruce committed the revised patch with its replacement sentence omitted, leaving only deletion of the misleading comment, as 1e0512ff23; the commit credits Antonin as reporter and was backpatched through PostgreSQL 9.5, and Bruce's application notice closed the thread #11.