model-writtenrun 998 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- David G. Johnston proposed configurable boolean display strings modeled on \pset null, choosing display_true/display_false while retaining t/f defaults and hoping to improve documentation examples #1.
- David added documentation and regression tests in #2; Tom Lane warned that configurable boolean output could break scripts, while David defended presentation controls as part of psql’s purpose #3 #4.
- Daniel Verite pointed out that substitution would omit booleans inside arrays, composite values, and COPY output; David considered arrays possible future work, composites low priority, and COPY outside \pset’s scope #5 #6.
- Vik Fearing supported the feature and argued that scripts should already use --no-psqlrc to avoid user configuration affecting their output #7.
- After several months, Álvaro Herrera supplied a cleaned-up version with corrected documentation and consistent ordering; David accepted it on a quick read but reopened the question of shortening the option names to true and false #8 #9 #10.
- Chao Li demonstrated misleading output from swapping the boolean strings and proposed basic validation; David briefly supported blocking reversed t/f values, but Tom Lane argued that whitespace variants would defeat such safeguards, while Pavel Stehule suggested predefined short, long, or localized representations #11 #12 #13 #14 #15 #16.
- Álvaro committed #8 as 645cb44c54, adding help and completion entries and simplifying documentation while retaining unrestricted display strings and the display_true/display_false names; he explained that the shorter names were too generic and reiterated --no-psqlrc for scripts #17.
- Months later, a.kozhemyakin reported an AddressSanitizer double-free when a temporary \g option changed display_false, identifying 645cb44c54 as the introducing commit #19.
- David reproduced the failure and supplied a fix making savePsetInfo() and restorePsetInfo() explicitly duplicate and free both boolean display strings #20 #21.
- Bruce Momjian applied #21 unchanged as 34be85f657 and confirmed completion in the final message, leaving both the feature and its reported memory-management fix landed with no attached patch outstanding #22.