model-writtenrun 7796 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Michael Paquier proposed a built-in sha1(bytea) function together with a shared helper replacing four duplicated SHA-2 implementations in #1, and Noah Misch opposed adding SHA-1 because it had long been deprecated in #2.
- Sehrope Sarkuni supported legacy SHA-1 use and supplied a revised helper that determines digest length internally and writes directly into one result allocation in #3; Michael accepted those improvements but requested avoiding a default switch clause so new enum values would trigger compiler warnings in #4.
- Kenneth Marshall, Bruce Momjian and David Fetter defended existing SHA-1 use cases in #5, #6 and #7, while Michael noted that pgcrypto already provides SHA-1 in #8 and Daniel Gustafsson favored leaving such support to an extension in #9.
- Michael abandoned the new SHA-1 SQL function and posted the standalone SHA-2 refactor in #10, retaining Sehrope's improvements and explicitly rejecting MD5 and SHA1 in the helper.
- Sehrope reported that the final refactor built and passed make check in #11; he is a credited co-author rather than an additional suggested reviewer.
- Michael committed #10 with only formatting and error-message wording changes as f854c69a5b, then confirmed completion in #12; the SHA-2 cleanup landed and the proposed SHA-1 SQL function was dropped.