model-writtenrun 2738 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Peter Eisentraut began by proposing removal of three unnecessary multibyte ctype wrappers, arguing that their callers only needed ASCII or single-byte handling, and removed pg_trgm's undocumented non-default KEEPONLYALNUM path as part of the cleanup #1.
- Peter expanded the series with replacement of lowerstr*() by the locale-provider-aware str_tolower() using the database default collation, while leaving column-specific pg_trgm collations and provider-aware t_isalpha()/t_isalnum() for later work #2.
- Jeff Davis questioned whether standard libc classification functions sufficiently expressed ASCII-only intent, endorsed the series with a request for an upgrade warning, and accepted Peter's explanation that supported encodings made the replacements appropriate #3 #4 #5 #6.
- Peter committed the first three patches together without code changes as d3aad4ac57 and the lowerstr() replacement separately with only comment reflow as fb1a18810f, both from #2.
- Andreas Karlsson and Peter supported future explicit ASCII helpers, while Peter closed the current commitfest entry, recorded the release-note obligation, and deferred t_isalpha()/t_isalnum() to a new series once provider-aware classification became available #7 #8 #9.
- Peter returned in August with an inline PostgreSQL 18 migration warning recommending reindexing affected full-text-search and pg_trgm indexes after pg_upgrade when non-libc locale behavior differs from LC_CTYPE #10.
- Daniel Verite demonstrated that parsing still used LC_CTYPE and recommended narrowing the warning to lowercasing; Heikki Linnakangas then proposed revised wording and pressed for timely upgrade documentation #11 #12.
- Peter landed the migration note as d1073c3b4c, using 'case conversions' and explicitly including ICU and builtin providers; his final reply confirms completion of the release note while leaving true case folding as possible future work #13.