model-writtenrun 6057 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Florin Irion proposed recording loaded extensions' configuration prefixes and warning about unknown settings supplied through SET, while still accepting the setting so existing behavior and the user's transaction choice were preserved #1.
- Chapman Flack demonstrated that existing warnings occurred only when PL/pgSQL was first loaded and supported addressing the gap #2; Laurenz Albe endorsed the idea #3, and Florin clarified that the patch also covered settings issued after an extension had loaded #4.
- Florin submitted a second version with small declaration and string-length calculation adjustments, retaining the same prefix-registration design and tests #5.
- Michael Paquier requested a concrete example of consequential configuration mistakes #6, and Florin cited pglogical.conflict_resolution and pg_stat_statements.track to explain the practical motivation #7.
- Peter Eisentraut committed #5 as 75d22069e0, renaming the checking function, fixing whole-setting-name matching, removing the warning hint, and tidying the code and tests; his announcement explained why partial string comparison was incorrect #8.
- The thread ended with Florin accepting the comparison fix and thanking Peter for committing the patch, with no outstanding work identified #9.