model-writtenrun 9760 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Andrew Gierth proposed distinguishing plain SELECT from SELECT FOR UPDATE because locking can change performance, attaching a patch that fingerprints row-mark targets, strengths and wait policies while excluding pushed-down clauses #1.
- Vik Fearing offered regression tests and asked about an extension version bump #2; Tom Lane explained that hash-calculation changes did not require a version bump and should not be backpatched #4.
- Tom Lane advocated processing RowMarkClause through the generic expression jumbling machinery #3 #6, but Andrew defended a separate walker because RowMarkClause is not an expression and the existing organization parallels PostgreSQL's tree walkers #5 #8; the dedicated walker was retained.
- Vik posted Andrew's implementation with regression tests for locking targets, strengths and wait policies #7; Sergei Kornilov subsequently reported successful testing and requested a clearer comment explaining why one query shares the plain FOR UPDATE statistics entry #9 #10.
- Thomas Munro marked the patch ready for a committer #11, and Andrew committed the tested version from #7 with the requested comment clarification as 6e74c64bcf.
- Andrew confirmed that he changed the comment during commit #12, and Sergei acknowledged completion #13; the submitted work had landed with no remaining patch revisions requested.