model-writtenrun 6289 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Amul Sul opened the thread because session-specific temporary schema names made EXPLAIN VERBOSE regression output unstable, proposing either omission or substitution with pg_temp in #1.
- Bharath Rupireddy proposed filtering test output in #2, and Ashutosh Bapat argued that actual schema names helped distinguish sessions in auto_explain logs in #4; Amul initially accepted that concern in #6.
- Greg Stark argued that reused temporary schema names were unreliable session identifiers in #7, and Tom Lane recommended retaining schema qualification but using pg_temp for the current session's schema in #8.
- Amul implemented that approach for EXPLAIN and postgres_fdw and supplied a separate experimental psql description patch in #9.
- After a three-month gap, Simon Riggs endorsed the change but requested a test in #10; Tom supplied #12, adding missing ruleutils.c substitutions, consistent allocated-string returns, and temporary-table/function regression coverage.
- Tom rejected the psql experiment because it also renamed other sessions' schemas, questioned analogous event-trigger behavior, and proposed moving substitution into get_namespace_name globally in #12.
- Simon supported centralizing the behavior in #13, Robert Haas cautioned that actual names remained useful for troubleshooting in #14, and Alvaro Herrera explained the helper's historical separation in #15; Tom then abandoned global substitution after it broke catalog-name comparisons in regression tests in #16.
- Alvaro agreed that event triggers could misleadingly identify other sessions' temporary objects in #17; Tom replaced that logic with get_namespace_name_or_temp, reported check-world success, and proposed changing only the development branch because existing users might depend on the old behavior in #18.
- Tom committed #12 unchanged as 48c5c90682 and separately committed the discussion-derived event-trigger correction as 024515cac5.
- Tom announced both commits and closed the CommitFest entry in #19, leaving psql's temporary-schema display unresolved because a safe, clean implementation had not been identified.