model-writtenrun 2759 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
Sep - Dec 2024 ยท 17 entries
- Tatsuo Ishii initially described an apparent underscore in config.sgml and submitted its removal in #1; Yugo Nagata identified the UTF-8 non-breaking space and found another occurrence in ref/drop_extension.sgml in #4.
- Daniel Gustafsson proposed preventing recurrence with a Makefile checker in #7; Yugo found detection problems in #8 and switched to portable Perl matching the complete UTF-8 sequence in #10.
- Tatsuo committed the original correction plus Yugo's additional occurrence as 3b1a377def, announcing both fixes in #12.
- Yugo corrected the checker's exit status and clarified its portability comments across #14 and #15; Tatsuo and Daniel approved it in #17 and #18, and 5b7da5c261 landed the final patch unchanged.
- Bruce Momjian asked about detecting a broader range of unsupported characters in #16, while discussion identified legitimate non-ASCII examples and contributor names; Daniel explained the special rendering risk of non-breaking spaces in #25.
- Yugo supplied a broader checker with an allowlist in #26, but Tatsuo objected to maintaining that list in #27, and Yugo accepted that it might not justify its cost in #28.
- Bruce replaced the allowlist approach with entity conversions and an unconditional non-ASCII check in #29; Yugo supplied Perl portability and SVG-coverage improvements in #31, and Bruce incorporated them while creating README.non-ASCII in #32.
- Peter Eisentraut objected that the restriction also prohibited valid Latin1 text in #35, and Tom Lane explained its burden on contributor names in #36 and #39, leading Bruce to drop the mandatory check in #40.
- Bruce's testing found missing glyphs without useful warnings in #37, whereas Peter reported FOP warnings on his system in #41, exposing differences in toolchain behavior.
- Bruce landed the relaxed cleanup from #40 as 641a5b7a14, additionally renaming a Makefile variable and extending non-breaking-space checks to image files, and confirmed that no new UTF-8 restrictions were imposed in #45.
- Tatsuo challenged the remaining entity conversions in #46 and suggested iconv-based validation in #48; Peter distinguished encoding style from renderable characters and suggested improving FOP diagnostics in #50.
- Yugo began a new two-patch approach in #51, combining FOP missing-glyph detection with iconv input checking, then revised locale handling and added optional iconv discovery across #53 and #55.
- Tatsuo found the CLANG typo and objected to deleting PDFs containing glyph warnings in #56; Bruce dropped the iconv and configure portions in #59, but Tom still rejected treating those warnings as build failures in #60.
- Bruce changed the remaining patch to print a concluding warning without failing the build in #61, then committed it with revised wording and a README reference as e4c8865196, announcing the landing in #62.
- With PDF warnings available, Bruce reopened the entity-conversion question in #63; Peter explicitly requested reversal in #66, and 8b318a168a restored the affected characters and stylesheet text without an attached patch.
- Peter separately spotted that the FOP-to-awk pipeline lost FOP's exit status in #67; Bruce supplied the sentinel-based correction in #70, which landed unchanged as 498f130756.
- At the final message #70, the space cleanup, preventive non-breaking-space checker, documentation guidance, warning-only PDF check, entity-conversion reversal, and exit-status repair had landed; the allowlist, blanket non-ASCII ban, and iconv/configure series had been dropped, with no further patch revision announced.