model-writtenrun 8968 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Fabien COELHO proposed extensive psql TAP tests in #1, reporting line coverage rising from 40.0% to 78.4% and function coverage from 41.9% to 98.1%; the patch added Expect-based interactive tests and extended TestLib to supply command input.
- Michael Paquier reviewed the rebased submission #2 in #3, requesting better helper documentation, questioning compatibility with the minimum Perl version, and identifying the need to skip readline-dependent tests; he separately committed its unchanged describe.c comment correction as 8a0deae8d9.
- Fabien addressed the readline and documentation issues in #4 and explained that placing backslash commands on separate lines increases coverage because some commands consume the rest of their line.
- Michael extracted the conditional-command regression tests from #4 into aafe2762b1, correcting indentation and moving the new elif case beside related tests, as explained in #5.
- Fabien resent the remaining TAP work in #6, then removed commented-out experiments and debugging code across #8 and #10 in response to vignesh C's reviews #7 and #9.
- Alvaro Herrera argued in #11 that TestLib changes should be separated and preserve existing callers through named parameters and compatibility handling; in #13 he also questioned having separate interactive and noninteractive helpers, while Fabien defended their different execution models in #14.
- After the patch stopped applying and Michael requested an update in #16, Fabien supplied #17, rebasing the remaining work and removing three additional comments identified by vignesh C in #15, without implementing the proposed API redesign.
- The patch again stopped applying by March 2020, as David Steele reported in #19, and Daniel Gustafsson eventually marked it Returned with Feedback in #21 after no replacement arrived.
- Fabien attributed the impasse to objections to Expect and declined to reimplement its functionality in #22, while Daniel clarified in #23 that the return was caused by the stale patch and that a working revision remained welcome.
- Tom Lane opposed a mandatory Expect dependency but accepted conditionally skipped tests in #24, asking whether Expect offered enough benefit over IO::Pty; Andrew Dunstan supported an optional dependency in #25, and Michael asked about using IO::Pty directly in #26.
- Fabien maintained his preference for reusing Expect in #27, and Andrew clarified in #28 that Expect and its PTY dependencies do not work on Windows and require conditional loading and test skipping; the thread ended with only the two extracted changes landed and the larger TAP proposal still returned with feedback.