model-writtenrun 7457 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Tatsuo Ishii opened the thread with a patch adding a missing 'be' to the memory-ordering explanation in src/backend/storage/lmgr/README.barrier #1.
- David Rowley identified two additional passages that predated atomic-operation support #2, and Ishii expanded the patch to recommend specific atomic functions #3.
- Rowley supplied revised wording because naming atomic functions alone could imply that they accept ordinary variables; his patch instead required atomic variables, referenced port/atomics.h, and preserved locking as an alternative #4.
- Ishii accepted the revision and favored backpatching #5; Michael Paquier considered backpatching optional and recommended waiting until the release tag #9, while Ishii argued that README corrections deserved the same backpatch treatment as code comments #10.
- Rowley committed the revised patch #4 unchanged as 2ded19fa3a and backpatched through PostgreSQL 9.6, reporting that the balance of opinions favored backpatching #11.
- After the commit, Tom Lane explained why even low-risk changes should generally wait until a release window closes, while accepting that this README change caused no harm #12; Paquier reiterated that caution in the final message #13, leaving the documentation changes fully committed with no further revision pending.