model-writtenrun 1325 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Tom Lane traced missing diagnostics from tanager's tab-completion failure to IPC::Run::timeout throwing exceptions before explicit expiration checks could run, and proposed switching BackgroundPsql to a timer while warning that callers needed auditing #1.
- Daniel Gustafsson confirmed that the earlier interactive-test implementation had used a timer, and Michael Paquier tested Tom's patch on tanager, obtaining diagnostic output that exposed unwanted backslash escaping by its old libedit #2 #5.
- Erik Wienhold narrowed the libedit fix to the 2019 releases, and Dagfinn Ilmari Mannsåker explained that contemporary Debian versions avoided the affected library; Tom declined to accommodate the obsolete behavior, while Michael kept tanager configured without readline support #12 #14 #16 #17.
- Andrew Dunstan audited expiration checks and found Cluster.pm's exception handling sound but other uses suspicious, leaving a broader examination of timer construction and propagation necessary #15 #20.
- Daniel expanded the proposal into three patches: automatic per-query timer restarting, reporting timeout failures and returning undef instead of aborting, and cleaning up a background session at test completion; he reported that the series passed check-world #22.
- A related discussion examined psql's word-break characters: Robert Haas and Alexander Korotkov defended the SQL rationale for removing the opening brace, while Tom identified limitations for future expression completion without demanding a change for that hypothetical case #21 #23 #24.
- After a six-month gap, Heikki Linnakangas approved the series, requested clearer quit documentation, questioned redundant assertions, and reported that the session-cleanup fix had already been implemented; Daniel resent the remaining two patches with corresponding edits #25 #26.
- Heikki supplied clearer documentation, and Andrew later caught incorrect isnt calls and suggested combining assertions with bare early returns; Daniel incorporated this feedback, rebased the series, reported successful check-world testing, and applied pgperltidy #27 #28 #29 #30 #31.
- Nearly a year later, Alvaro Herrera revived the forgotten submission; Tom confirmed that the diagnostic problem remained relevant, and Daniel resumed looking at the patches #32 #33 #34.
- Tom approved the timer-restart cleanup and extended it to the copied logic in 030_pager.pl, but objected that returning undef would be poorly handled by callers and leave stale output; he recommended improving query_until diagnostics while retaining fatal timeout behavior #35.
- Daniel approved Tom's final attachment and agreed with the diagnostic direction in the thread's last message #36; he subsequently committed the attachment unchanged as 53c6bd0aa3, leaving the broader timeout-handling proposal unresolved in the supplied record.