model-writtenrun 7984 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Magnus Hagander proposed --no-instructions for initdb and pg_upgrade so packaging wrappers could retain useful output without displaying incorrect startup or cleanup instructions; the pg_upgrade option also suppressed generated scripts #1.
- Ian Lawrence Barwick recommended distinct option names because suppressing text and suppressing scripts were different operations, and Tom Lane supported --no-instructions for initdb and --no-scripts for pg_upgrade after Magnus explained why wrappers retain output #2 #5 #6.
- Peter Eisentraut questioned which scripts would remain after the separately discussed analyze-script removal and suggested replacing the deletion script with printed instructions; Magnus considered generic cleanup advice because platform wrappers know about configuration files and services #9 #12 #14.
- Magnus submitted a rebased revision that retained initdb's option and renamed pg_upgrade's option to --no-scripts, adapting it to the removal of the analyze script #15.
- Anastasia Lubennikova pointed out that pg_upgrade also generates SQL repair scripts, exposing ambiguity in --no-scripts; Magnus argued those scripts should remain because wrappers cannot readily reproduce their database-specific knowledge #16 #17.
- Alvaro Herrera proposed selecting or suppressing named scripts through a list-valued option, while Magnus questioned whether any practical use case beyond suppressing cluster deletion justified that interface #18 #19 #20.
- Bruce Momjian explained that the deletion script already handles tablespaces and defended retaining that assistance, while Magnus emphasized incomplete platform cleanup; Bruce also warned that deleting shared configuration files could damage the new cluster #21 #22 #23.
- Peter proposed redirecting initdb output or introducing a broader quiet mode, but Magnus defended retaining detected settings while selectively suppressing instructions that wrappers make incorrect #27 #28 #30.
- The discussion considered removing startup advice unconditionally, with support from Tom Lane, but Michael Paquier favored retaining the hint for manual users and providing a suppression switch #33 #34 #36.
- Peter ultimately accepted initdb --no-instructions and requested removal of the standalone Success output; Magnus committed the initdb portion of #15 as e09155bd62, but moved that output into the instructions branch, and explicitly dropped the rejected pg_upgrade portion for separate discussion #37 #38.
- Peter then reported an extra blank line before Success, and Magnus identified that the entire trailing Success message duplicated the existing announcement and proposed removing it #39 #40.
- At the thread's last message, Tom confirmed that older branches lacked the duplicate message; the agreed correction subsequently landed as e7f4291485, completing the initdb work while pg_upgrade script handling remained deferred #43 #38.