model-writtenrun 7340 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Amul Sul proposed testing recovery to an LSN whose WAL resides in a segment bearing a higher timeline ID, while acknowledging that his test did not directly verify EndOfLogTLI != replayTLI in #1.
- Andres Freund reported failures across platforms in #2, prompting Amul to remove recovery_target_inclusive = off in #3; after Julien Rouhaud reported continuing Windows failures in #4, Amul added a wait for WAL archival before standby shutdown in #5.
- Julien reported further Windows and macOS failures in #7, and Amul subsequently identified incorrect Windows archive-path handling, adding configurable archive and restore paths to the TAP helpers and revising the test in #9; he left open whether node information should reflect the custom archive directory.
- Michael Paquier identified Heikki Linnakangas's independently submitted test for the same case and recommended adopting its clearer description and structure in #10; Kyotaro Horiguchi supported explicit WAL switching and simpler restore configuration in #11, and Amul agreed to the replacement in #12.
- Heikki committed the alternative from #497589 in thread #44400 as 50e5bc582a, with light edits removing an initial blank line and replacing the fixed test count with done_testing(); the standalone test also verifies recovery through the resulting end-of-recovery timeline-switch record, and he confirmed its landing here in #13.
- Amul acknowledged completion in #14; the requested coverage landed through Heikki's alternative, superseding the attached test series and its proposed configurable archive-path helpers.