model-writtenrun 6930 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- The thread began with Payal Singh's review of an externally posted pgindent patch, reporting successful tests of multiline formatting, protected comment styles, single-line preservation, and performance in #1, repeated through the commitfest application in #2.
- After a gap, solaimurugan vellaipandiyan reported successful testing and repeat-run stability of v7, but questioned formatting churn in banner comments such as contrib/seg/seg.c in #3.
- Tom Lane argued that standardizing those comments was desirable and posted v8 in #4, replacing restrictive eligibility checks with explicit single-line protection and normalizing leading whitespace and asterisks; he proposed manual preparation of GEQO ASCII-art signatures instead of complicating pgindent.
- John Naylor questioned extra blank space in one reformatted comment in #5, which Tom Lane considered too uncommon to justify special handling in #6; solaimurugan vellaipandiyan then confirmed v8's consistency, repeat-run stability, and preservation of single-line comments in #7.
- Alvaro Herrera flagged a comment needing manual treatment in #9, and Tom Lane agreed in #10 that the obscure ranking formula in tsrank.c should remain aligned with its surrounding code.
- Aleksander Alekseev supplied a separate explanation of the ranking calculation in #11, which Tom Lane committed as 163f20ca12 after relocating and polishing the comments and reordering the final expression to match them.
- In the thread's final message, #12, Tom Lane confirmed the tsrank.c commit and reposted the unchanged pgindent implementation alongside manual GEQO and partbounds.c corrections and regenerated formatting results, describing the remaining work as ready to go.
- The pgindent implementation subsequently landed unchanged from #12 as 60f9467c38; the two supplied commits establish that the formatter change and tsrank.c rewrite landed, but do not establish the landing of the accompanying manual and bulk-formatting patches.
- Although thread #51766 attributes the formatter commit to its earlier patchset #1003066, this thread's #12 is the later exact match and therefore supplies the final committed version.