model-writtenrun 9030 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Justin Pryzby reported that adapting scripts to PostgreSQL 12's required pg_restore -f - syntax silently created a file named '-' on older versions, breaking pipelines; his /dev/stdout workaround depended on operating-system support #1.
- Tom Lane proposed backpatching optional -f - support and considered relaxing PostgreSQL 12's requirement, while Andrew Gierth defended explicit destinations because inexperienced users frequently mistook SQL printed to their terminals for a completed restore #2 #3 #4.
- Euler Taveira initially opposed changing older branches because existing scripts could depend on the old behavior, while Stephen Frost and Yoshikazu Imai considered intentional use of a file named '-' unlikely enough to justify the compatibility fix #7 #8 #9.
- Justin explained why deployments spanning several major versions needed shared maintenance scripts and advocated retaining PostgreSQL 12's enforcement while documenting the compatibility issue; Stephen supported that approach, while Tom emphasized the missing transition period #11 #15 #16 #17.
- Ahead of the November minor releases, Alvaro Herrera volunteered to implement the backpatch, Tom accepted the approach, and Alvaro posted a partial cherry-pick that recognized '-' as standard output without requiring explicit destinations on older branches #18 #19 #22.
- Tom's review caught that the documentation must still describe standard output as the default on older branches, and Alvaro supplied corrected wording; Euler also flagged translation breakage from the help-text update, but Alvaro retained it to make the new behavior discoverable #23 #24 #25 #26.
- Alvaro committed the attached patch with only that documentation correction as 3574c0ac05, providing optional -f - support across older supported branches, and announced the push in #27.
- After the commit, Peter Eisentraut argued that large installations needed PostgreSQL 12's requirement relaxed, and Tom supported deferring enforcement until PostgreSQL 13; Stephen opposed changing the released behavior, while Alvaro pointed to the backpatch and existing workaround #28 #31 #33 #36.
- Robert Haas cautioned that changing released behavior again could cause more damage, and Tom accepted the emerging consensus to leave things as they stood: the compatibility backpatch had landed, PostgreSQL 12's explicit-destination requirement remained, and no further patch was pending in the thread #37 #38.