model-writtenrun 7548 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Guillaume Lelarge reported that --schema omitted extensions needed to restore selected objects, and proposed either including extensions automatically or adding an explicit option in #1.
- Laurenz Albe and Daniel Gustafsson explained that extensions belong to the database and can span schemas, while Daniel also identified restoration problems with automatically including extensions from other schemas in #2, #3, and #5.
- Tom Lane argued that selective dumps intentionally leave dependency selection to users and favored an extension-pattern option; Guillaume accepted that direction in #6 and #7.
- After an eight-month gap, Guillaume posted the first implementation in #8; Julien Rouhaud requested tests in #9, Asif Rehman tested the feature in #12, and Guillaume added a TAP test in #14.
- Michael Paquier found that the plpgsql test could not demonstrate dumping a user extension and requested test_pg_dump coverage, updated comments, and documentation corrections in #15; David Fetter also identified the documentation typo in #20.
- Tom clarified that --extension should select extensions independently of --schema and --table while preserving existing behavior when absent, and Guillaume confirmed that intent in #16 and #18.
- With Guillaume short of time and willing to hand off the remaining work in #22 and #26, Michael supplied #27 with positive and negative extension tests and the requested cleanup; it landed unchanged apart from rebasing as 6568cef26e, announced in #28.
- Noah Misch then demonstrated that configuration-table data from an unselected extension could still be dumped depending on its schema and --schema arguments, supplying a reproducer and noting duplicated --no-sync switches in #30.
- Michael separately removed the duplicated test switches in 885a876419, implementing the observation in #30 without a posted removal patch.
- Michael considered schema-selection changes in #33, but Noah explained that the fix should associate configuration tables with their extensions and omit those belonging to unselected extensions without suppressing unrelated schema objects in #34.
- Michael implemented that filtering in processExtensionTables() and incorporated Noah's reproducer plus another schema case in #35; Noah reviewed the fix and requested a clearer test name in #36.
- The configuration-table fix landed as 344487e2db with only test naming and comment edits, confirmed in #37; Guillaume acknowledged the completed fix in the final message #38, leaving no outstanding patch in the thread.