model-writtenrun 722 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Chao Li demonstrated that failed BASE_BACKUP commands could leave stale progress entries while the replication connection remained open, and submitted a patch moving progress cleanup into the sink callback used on both success and failure #1.
- Fujii Masao endorsed the patch and asked about backpatching #2; Chao Li considered backpatching unnecessary because ordinary pg_basebackup clients disconnect after errors #3.
- Fujii Masao committed the patch with only function-placement and commit-message edits as 55f0a13e96, backpatching through v15 while skipping v14 because it lacked sink infrastructure and the bug did not affect backup correctness #4.
- Andrew Dunstan subsequently reported that removing the exported wrapper had triggered the ABI compliance checker and suggested updating the affected branches' ABI history #6.
- Fujii Masao instead proposed restoring basebackup_progress_done() on stable branches because external backup tools might still call it #7, and Tom Lane endorsed that approach #8.
- The discussion produced e2ad214dd8, restoring the wrapper and declaration without undoing sink-based cleanup; Fujii Masao confirmed restoration in v15–v18, leaving both the original fix and compatibility repair complete #9.