model-writtenrun 2931 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
Nov 2023 - Jan 2024 · 12 entries
- Peter Smith opened the thread by cataloguing inconsistent quoting of configuration names and suggesting a substitution macro in #1 and #2; Daniel Gustafsson advocated reducing duplicate translations, while Tom Lane explained why language-specific quotation marks must remain translatable in #3, #4 and #5.
- Peter Eisentraut initially favored leaving recognizable names unquoted, Laurenz Albe accepted an exception for names containing underscores, and Nathan Bossart preferred uniform quoting in #6, #7 and #9.
- Alvaro Herrera proposed documenting conditional quoting in #10; Tom Lane approved it in #11, while Nathan Bossart warned that recognizing natural English words was subjective in #12.
- Peter Smith separately proposed common wording for hints about increasing configuration limits in #15; Kyotaro Horiguchi approved the patch but questioned the replication-state PANIC advice in #16, and Peter temporarily set that work aside to keep the quoting discussion focused in #23.
- Peter Smith implemented the conditional policy in #18, including mixed-case names after Alvaro's guidance in #17; Michael Paquier identified missed messages and suggested automatic quoting for dynamically substituted names in #19.
- The dynamic quoting implementation progressed from a static-buffer helper in #22 to a predicate and three-substitution macro in #31, following Michael's proposal in #27 and Tom's warning against locale-dependent character classification in #25.
- Michael committed most static-message changes from #31 as 8d9978a717, omitting DateStyle and fixing missing commit-timestamp expected output, and committed Alvaro's documentation with wording edits as a243569bf6, as reported in #33.
- Michael supplied a revised dynamic-quoting remainder in #33, but Kyotaro identified its inability to translate quotation marks in #34; Alvaro suggested a translatable helper in #37, and Peter Eisentraut rejected content-dependent quoting as excessive complexity in #38.
- Peter Smith abandoned dynamic quoting and returned to the deferred DateStyle correction plus a new IntervalStyle canonical-name fix in #39, preserving the documented spellings that Laurenz and Tom had defended for compatibility in #28, #29 and #32.
- Alvaro endorsed the case fixes, added an IntervalStyle test and supplied a broader common-message cleanup in #40; Peter Eisentraut accepted the case corrections but objected to removing their quotes in #41.
- Peter Eisentraut argued that internal refactoring must not change visible quoting and that quoting should be uniform for a category in #45 and #48, prompting Peter Smith to rewrite the documentation and quote GUC names throughout the tree in #49.
- Peter Eisentraut committed the documentation and broad quoting portions of #49 with small fixes as 17974ec259, superseding the earlier conditional policy while leaving the case fixes and common-message work pending, as confirmed in #56.
May - Jun 2024 · 2 entries
- After several months, translation work brought Peter Eisentraut back to the unfinished policy change in #50; Daniel Gustafsson supported the uniform quoting patches, and Daniel and Tom discussed the awkward spelling of combined names and values in #52 and #54.
- Peter Smith resumed the remaining work in #58, carrying forward the case and translation patches and adding a fresh quoting sweep split by GUC type; #59 fixed a missing comma, and #60 rebased the series after further development.
Aug - Oct 2024 · 8 entries
- By #61, the boolean quoting changes were already present in HEAD, so Peter removed that patch and refreshed the remainder; Michael then requested merging the remaining quoting patches in #62, which Peter did in #63.
- Michael committed the quoting component of #63 as b4db64270e, adding overlooked search_path quotes in oid2name and vacuumlo and making a small message-spacing adjustment, as described in #64.
- Michael committed selected common-message substitutions from #63 as a68159ff2b, dropping substitutions that provided no translation savings or introduced platform inconsistency and adding translator comments, while leaving both case fixes pending in #64.
- Alexander Lakhin separately questioned the wording of pg_test_fsync's wal_sync_method explanation in #65; the supplied thread contains no resolution of that wording question.
- Peter resent the two remaining case fixes in #66, and Michael committed its unchanged DateStyle component as 2e7c4abe5a, reporting the landing in #67.
- Michael expanded the remaining IntervalStyle fix in #67 to use canonical table names across more guc.c paths and remove a redundant function argument, while explicitly leaving the earlier ACL check in AlterSystemSetConfigFile() outside that cleanup.
- Peter proposed splitting set_config_with_handle() to restrict the caller-supplied name's scope in #68, but Michael judged the extra wrapper confusing and committed the preceding patch #67 unchanged as f3f06b1330, explaining the choice in #69.
- Peter declared the year-long thread complete in #70, leaving additional opportunities to consolidate common messages for another thread.