model-writtenrun 8120 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Peter Eisentraut reported that pgindent failed on macOS dtrace-generated probes.h; Daniel Gustafsson and Tom Lane favored excluding externally generated files, while Tom and Peter agreed that generators maintained in PostgreSQL should produce pgindent-compatible output #1 #2 #3 #4.
- After a gap of nearly four months, Daniel submitted generator formatting fixes for sql_help.h and plperl_opmask.h together with a probes.h exclusion, and identified unexplained formatting changes in fmgrprotos.h #5.
- Tom traced the fmgrprotos.h discrepancy to functions sharing names with typedefs, and Alvaro Herrera identified the numeric typedef in the ECPG headers; Tom initially discouraged blacklisting those typedef names because it could damage formatting elsewhere #6 #7.
- Tom committed Daniel's patch with additional sql_help.c whitespace fixes and expanded documentation as add105840b, then suggested renaming the conflicting C functions without changing their SQL names #8 #9.
- Daniel supplied inline exclusions for the generated SPI.c and Util.c files, while Tom found more generated files needing attention and preferred excluding keyword-list headers because pgindent made their formatting worse #10 #11.
- Tom broadened exclusions for a fully built development tree, consolidated their explanations into the exclusion file and added explicit comment handling in 74d4608f50, leaving four generated outputs for subsequent cleanup #12.
- Tom fixed the remaining sort and procedural-language generator output formatting in f859c2ffa0 and initially considered the cleanup complete #13.
- Remembering fmgrprotos.h, Tom posted a separate patch renaming the numeric and chr C functions; he explained that renaming the public ECPG typedef would be more disruptive, and Alvaro reported that blacklisting the names badly distorted regex formatting #15 #16 #17 #18.
- Daniel proposed documenting the typedef-name conflict, and Tom clarified that it affects function definitions as well as declarations #19 #20.
- Tom abandoned the function renames because future typedef conflicts could recur, instead excluding fmgrprotos.h and adding documentation plus a .git exclusion in c4133ec169; his final message reported the replacement committed, completing the cleanup without landing the renaming patch #21.