model-writtenrun 3554 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Kyotaro Horiguchi reported that initdb -c produced redundant configuration entries for differently cased names and proposed case-insensitive replacement preserving the existing spelling, while asking whether lowercasing input names would be preferable #1.
- After several months, Daniel Gustafsson revived the patch, endorsed its behavior, rebased it, and added a regression test covering repeated mixed-case work_mem settings #2.
- Alvaro Herrera suggested rejecting repeated settings #3, but Tom Lane favored the configuration files' existing last-setting-wins behavior #4; Daniel Gustafsson and Robert Haas supported preserving file casing, and Tom supported backpatching to PostgreSQL 16 for consistency #5 #6 #7.
- Tom Lane criticized the scattered replacement construction and initially suggested that changing the comparison alone would suffice #7 #8; Daniel explained why preserving file casing required more work and submitted a smaller revision #9, while Kyotaro clarified that the original patch still stopped after the first match #10.
- Tom Lane acknowledged that comparison alone was insufficient, identified premature replacement construction when a requested name merely prefixes another name, and supplied an implementation that validates the assignment before copying the canonical spelling into the prepared replacement #12.
- Daniel Gustafsson approved Tom Lane's implementation without further restructuring #13; after Daniel registered the work for the next commitfest #14, cfbot exposed the need to use pg_strncasecmp, which Tom corrected in the final patch #15.
- Daniel Gustafsson approved the final patch before commitment #16, and Tom Lane landed #15 with only test line wrapping as fce2ce797c, with the same fix backpatched to PostgreSQL 16 as b78f4d22b2.
- The thread ended with Kyotaro Horiguchi thanking Tom Lane for the fix, with the change landed and no outstanding work identified #17.