model-writtenrun 8576 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Tom Lane reported that pi was misaligned in PDF output and proposed a pi-specific processing instruction to select the symbol font explicitly, while asking for a more elegant mechanism #1.
- Alexander Lakhin replaced the processing instruction with generic phrase markup so the same font selection could support other symbols #2; Tom adopted that implementation with naming and comment adjustments in 5ac2475548 and confirmed the decision in #3.
- After Tom described an unsuccessful attempt to share the font-switching logic #3, Alexander supplied a named-template follow-up #4, but both considered the two invocation styles less clear than keeping the existing duplication #4 #5.
- Peter Eisentraut proposed replacing the underlying PDF body font with DejaVu Serif, adding FOP font discovery and Cyrillic examples, and explained that this could make the explicit pi workaround unnecessary #6.
- Tom initially objected to requiring readers to install additional fonts #7, but accepted the possibility subject to PDF size after Peter explained that the font would be embedded #8 #9.
- Alvaro Herrera tested the font configuration, found a stray XML character and Java dependency problems, and obtained a successful build after local fixes, but reported that footnotes still used a different font #10.
- Alexander closed the thread by offering Postgres Pro's FOP configuration while noting that its pi rendering still looked unusual and favoring the symbol font #11; the targeted pi fix had landed, the shared-template cleanup was declined, and the broader font proposal remained unresolved.