model-writtenrun 4319 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Yugo Nagata reported that pg_node_attr annotations were being mistaken for structure field names and proposed ignoring that macro with ctags -I in #1.
- Alvaro Herrera recommended excluding temporary build trees and improving file discovery in #2; Yugo incorporated those suggestions and restricted the ignore option to Exuberant-compatible ctags in #3, omitting shell and SGML files because he saw little benefit for ctags.
- Tatsuo Ishii found the same problem in make_etags and introduced an Emacs mode in make_ctags to share the implementation in #4; discussion in #5, #6, #7, #8, and #9 settled on an executable wrapper instead of a repository symlink, including consideration of Windows compatibility.
- Alvaro requested optional suppression of per-directory tag symlinks in #10, while Bruce Momjian confirmed their existing convenience for Emacs users in #11; Tatsuo retained the default behavior, added -n, and converted make_etags into a wrapper in #13.
- Yugo fixed simultaneous use of -e and -n in #14, Tatsuo aligned formatting and naming with existing scripts in #15, and Yugo accepted the result in #16.
- Tatsuo committed #15 unchanged as d1e2a380cb and announced the landing in #17, completing the macro fix and shared tag-generation enhancement.
- Yugo separately proposed ignoring tags and TAGS in the repository's .gitignore in #18, but Tatsuo cited prior consensus favoring personal ignore settings in #19, and Yugo accepted that decision in #20.
- Yugo also reported Vim encountering Emacs TAGS files and suggested a cleanup option in #20; Tatsuo questioned the need in #21, and Yugo agreed manual cleanup was sufficient in #22, leaving that option unimplemented.
- Months later, Fujii Masao reported that the shared wrapper failed on macOS because its non-Exuberant ctags lacked -e in #23; Tatsuo clarified the previous dependence on a separately installed etags program in #25.
- Yugo proposed restoring an etags fallback inside make_etags in #26, but Tatsuo identified lost -n support and moved fallback selection into make_ctags in #27.
- Yugo challenged the ctags capability probe and recommended preferring etags in #28, leading to #30; Fujii then reported invalid options passed to etags, premature deletion of existing tags, and misleading usage output in #31, prompting the corrections in #32.
- Fujii found that empty quoted arguments still caused failures and supplied #33, which simplified program selection, removed the unsupported assumption that another ctags implementation needed -e probing, and added make_etags-specific usage validation.
- Tatsuo approved Fujii's final patch in #34 and committed it unchanged as 87f21d2c68, announcing the fix in #35.
- Masahiko Sawada subsequently found that sorting corrupted the section-based Emacs TAGS format on his Mac and supplied #36 to skip sorting in Emacs mode and update the explanation.
- Tatsuo reproduced the corruption, verified Masahiko's fix on macOS, and checked that Linux with Exuberant ctags remained unaffected in #37; he committed the patch unchanged as ae66716bf3 with a PostgreSQL 15 backport and announced completion in #38.
- The thread ended with Masahiko's acknowledgment in #39; the enhancement and both portability fixes had landed, while the repository ignore entries and cleanup option had been declined or set aside.