model-writtenrun 4067 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Sami Imseih opened the thread in #1, explaining that “running in parallel” in the work_mem documentation could be confused with parallel query and proposing “simultaneously”, with backpatching to supported releases.
- Peter Eisentraut preferred “concurrently” in #2, but Tom Lane argued for a fuller rewrite and a separate clarification that the hash-memory limit applies per table in #3.
- Gurjeet Singh argued that the text must explicitly describe overlapping operations within one query and should say “sort and hash” in #4; Sami confirmed that interpretation in #5 and incorporated it into the first attached patch in #6.
- David Rowley questioned the simultaneous-operation wording in #7, while Tristen Raab reported successful documentation review and testing in #8; Sami clarified in #9 that the sentence concerned multiple memory consumers within one query execution.
- Bruce Momjian supplied a shorter rewrite and the singular per-table memory-limit wording in #10; Sami approved it in #12, while David requested “and” or “and/or” in #11 and explained the distinction in #13.
- Bruce adopted “sort and hash” in #14, and David approved that final patch in #15.
- The final patch landed unchanged on the development branch as 5f567b3c35; the separately supplied candidate a4a8c0de2b is its PostgreSQL 11 backport, containing only the applicable concurrency wording change.
- Bruce confirmed application back through PostgreSQL 11 in #16, closing the thread with the documentation corrections landed and no outstanding revisions.