model-writtenrun 5742 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Michael Paquier reported that preloading pg_stat_statements made installcheck fail because query IDs appeared in EXPLAIN output, and proposed disabling computation in regression databases while explicitly enabling it in pg_stat_statements tests #1.
- Julien Rouhaud opposed disabling computation because regression tests provide useful coverage for query-tracking extensions, prompting Michael to propose a regress mode that hides EXPLAIN query IDs while preserving computation; Julien noted that internal query_id_enabled tracking made this approach cleaner, though he questioned its necessity #2 #3 #4.
- Michael supplied the redesigned patch implementing regress mode with auto semantics and selecting it in regression databases; Julien found no problem with the patch, and Michael noted that extending the enum preserved ABI compatibility for backpatching #5 #6 #7.
- Michael committed the revised implementation as ebf6c5249b and backpatched it through PostgreSQL 14, retaining the backend changes from #5 with small documentation and comment edits but omitting automatic selection by pg_regress because PGOPTIONS or postgresql.conf covered the known use cases #8.
- The thread ended with regress mode landed and automatic configuration by pg_regress left as a possible future change; Justin Pryzby noted that the EXPLAIN suppression overlapped work in his separate machine-output proposal #8 #9.