model-writtenrun 9170 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
May - Sep 2019 ยท 24 entries
- Bruce Momjian opened the thread with an already-committed draft of the PostgreSQL 12 release notes in #1, inviting corrections to the feature descriptions and organization.
- Oleg Bartunov argued that JSON Path deserved recognition as SQL-standard functionality in #2 and #10; after consulting Alexander Korotkov, Bruce explained why the supporting datatype and functions remained grouped under functions in #7.
- Jonathan S. Katz supplied the duplicated-heading correction in #4, which Bruce reported updating in #6, while David Rowley's independent contributor-name normalization in #5 landed unchanged as f86b0c3c46.
- Fabien COELHO explained that Peter Eisentraut authored the pgbench error-reporting work in #9, leading to the attribution correction in 5d971565a7.
- Nickb reported a redundant word in the partition-bound description in #13, and Bruce removed it in 34d40becfa.
- Andres Freund's extensive review in #15 challenged feature grouping, technical explanations, and attribution; Bruce adopted selected generated-column, system-table-option, pgbench, pg_dump, and table-access-method changes in #29 and 32fe2e3194.
- Andrew Gierth clarified the old and new floating-point output rules in #16 and proposed replacement prose in #19; Andres shortened it in #21, and Bruce's condensed version in #31 landed as b84a801d6a.
- Tom Lane supplied a precise explanation of RECORD and RECORD[] in table-function column definitions in #17, which became 3468a04a3e while retaining the existing documentation link.
- Tom's proposed DETACH PARTITION explanation and combined planner-support description in #23 became 8e719d33fd, although Bruce's version still omitted custom index conditions.
- Shawn Debnath proposed an additional internal-change entry in #27, but Bruce declined it in #28 because it primarily affected internal developers.
- Tom identified the missing index-condition capability and reiterated the Source Code placement in #36; Bruce removed the in-core examples, expanded the extension capabilities, and moved the entry in #42 and 728840fe13.
- Peter Geoghegan explained in #24 and #37 that independent B-tree locking work had been mixed with duplicate-handling and space-efficiency improvements, prompting Bruce to study the implementation before rewriting those entries.
- Ian Lawrence Barwick questioned the placement of TRUNCATE and VACUUM under Authentication in #38, and Michael Paquier explained the unauthorized-locking issue and proposed fuller wording in #40.
- Andres requested combining the three partition-introspection functions in #43, and Bruce merged their descriptions into a single entry in 6a631a4546, confirming completion in #49.
- Adrien Nayrat supplied the list of newly supported stemming languages in #45; Bruce capitalized the names and moved the list into the explanatory paragraph in 9680728371.
- Bruce returned to the B-tree explanation in #46, and Peter corrected his interpretation of locking, page splits, and VACUUM locality in #51, leading to the first attached B-tree rewrite in #52.
- Peter's detailed patch review in #53 prioritized duplicate-heavy indexes, emphasized VACUUM's ability to recycle pages, and recommended dropping the split-after-new-tuple explanation; Bruce incorporated those changes in #54, which landed unchanged as 4bfb79ff6b.
- Bruce declined to merge the recovery-setting behavior changes into one entry in #47, reasoning that users needed explicit notice of changed behavior even after moving settings out of recovery.conf.
- Bruce resisted spelling out an unbackpatched denial-of-service mechanism in #48 and #58, choosing the concise unauthorized-locking explanation in #59 for 4f41a72275; Michael accepted it afterward in #60.
- John Naylor caught the Tami spelling inherited from the language-list patch and supplied #61, which landed unchanged as 178ca6b26c.
- Laurenz Albe proposed documenting the Windows file-opening changes in #62; after Michael questioned their suitability as low-level release-note material in #65, Bruce focused the entry on corrected pg_test_fsync timings and moved it to Server Applications, committing 551394ff3a and posting the exact result in #66.
- Peter requested an entry for amcheck's rootdescend verification option in #64, and Bruce posted the independent addition as already applied in #68.
- Jonathan resumed work on the release presentation in #70, proposing major-feature highlights and an ICU ordering clarification; Andrey Borodin identified the incorrect SP-GiST covering-index claim in #71, and Jonathan corrected it to GiST in #72.
- The thread ends with Jonathan's additional PG_COLORS example patch in #73; the supplied record does not show application of that patch, the revised highlights, or the separate ICU clarification.