model-writtenrun 5483 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
Apr 2022 · 10 entries
- Tom Lane proposed a wildcard-capable psql command for inspecting configuration parameters and their privileges in #1, reviving an idea from earlier privilege work that Mark Dilger identified as his original \dcp implementation in #9.
- Pavel Stehule suggested \show in #2, but Tom argued that it would be confused with psql variables in #6; Jonathan S. Katz favored the describe-command family in #8, and Tom posted a \dconf prototype without documentation or tests in #13.
- Joe Conway and Jonathan questioned the output columns in #16 and #18; Tom kept context together with privileges under the verbose form in #19, while Justin Pryzby's suggestion to use current_setting() in #22 eliminated the separate units column.
- The revised implementation in #25 added documentation and tests, moved type into verbose output, fixed qualified-name handling, and deferred the pre-existing case-sensitive completion problem; Jonathan tested the equivalent SQL and approved the revision in #27.
- Mark preferred \dconfig for terminology consistency in #28, and Pavel identified the dconf naming collision in #34; Tom accepted the longer name in #35 and landed 3e707fbb40 from #25 with the rename, a comment, and revised regression tests.
- After trying the command, Christoph Berg proposed showing only non-default settings without a pattern in #36; Tom retained the \dconfig name in #37, and Julien Rouhaud supported filtering while objecting to blanket suppression of override settings in #38.
- Jonathan initially objected that hiding defaults could obscure useful information in #39, then accepted the explicit wildcard escape and distinguishing table heading in #44; Tom implemented that discussion directly as 5e70d8b5d1 and announced it in #45.
- Jonathan's testing found initdb-written values still appearing despite matching their defaults in #46; Tom proposed combining source and boot-value checks in #47 and rejected making initdb recognize every compiled default as fragile in #49.
- Tom supplied the NULL-safe boot_val comparison in #51, Jonathan tested and approved it in #52, and 139d46ee26 landed that patch unchanged, as announced in #55.
- The mixed-case concern raised in #18 and investigated in #19 also produced b5607b0746, which lowercased completion patterns and flattened the GUC-name queries without an attached patch.
Jun 2022 · 6 entries
- The remaining computed-default problem resurfaced in June when Justin reported shared-memory settings in ordinary output in #56; Tom considered excluding internal settings in #57, while Robert Haas explained the competing diagnostic uses of the command in #58.
- Jonathan noted that in_hot_standby could provide useful context in #62, and Christoph explicitly argued to retain it in #63, making blanket suppression of internal parameters unsuitable.
- Tom changed the approach in #64: instead of adding another psql filter, he corrected backend source labels to identify computed values as dynamic defaults, including shared-memory sizes, automatically selected wal_buffers, and max_stack_depth.
- The revision in #65 preserved visibility of constrained stack limits and true in_hot_standby by using appropriate source labels and SetConfigOption(), while removing an assertion that prohibited the necessary override exception.
- Jonathan tested generated-default suppression and standby behavior in #67, then endorsed retaining the stack-limit distinction in #69; 7ab5b4eb48 landed #65 unchanged, with Tom announcing the push in #70.
- The thread closed with Jonathan's acknowledgment in #71, with all five changes committed and no attached work outstanding; the backend correction remained development-branch-only, so older servers could still expose the computed settings discussed in #64 and #66.