model-writtenrun 9773 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Peter Eisentraut proposed making recovery_target_timeline='latest' the standby default because pg_basebackup -R did not set it; Sergei Kornilov favored changing the default while preserving explicit settings, and Michael Paquier agreed that the existing default confused users #1 #2 #3.
- Peter's first patch forced standbys to use the latest timeline regardless of configuration, but David Steele and Michael argued for a general recovery default and an explicit value preserving the old behavior #4 #5 #6.
- Peter replaced that approach with two patches: one introduced 'current', choosing that name over the implementation-oriented 'controlfile', and the other changed the GUC default to 'latest' without special standby handling #7.
- David reviewed the series and requested coverage for 'current'; Michael supplied a supplemental test and asked that the documentation retain its explanations of why 'latest' matters for standbys, and David accepted the proposed test #8 #9 #10.
- Peter committed the first component unchanged as ff85306055 and posted the remaining default change with Michael's documentation revisions; he left out the supplemental test because replacing 'current' with 'latest' did not make it fail #11.
- Michael accepted the revised remainder and explained that distinguishing 'current' from 'latest' required a diverging timeline, while arguing that the proposed test still offered some end-of-timeline coverage #12.
- Peter committed the revised remainder unchanged as 0acb3bc33a, completing both functional changes, and questioned whether existing recovery tests already covered the supplemental test's purpose #13.
- The thread ended with Michael finding no explicit recovery_target_timeline settings in the Perl tests: existing tests exercised latest-timeline switching and some current-timeline behavior indirectly, leaving targeted test coverage unresolved #14.