Skip to content

[Publisher] PDF: long inline code wraps instead of running off the page (78 clipped words -> 0); build-only, split out of #14 - #16

Merged
dp-web4 merged 1 commit into
mainfrom
publisher/pdf-breakable-code-1114
Nov 15, 2026
Merged

dp-web4 merged 1 commit into
mainfrom
publisher/pdf-breakable-code-1114

Conversation

@dp-web4

@dp-web4 dp-web4 commented Nov 14, 2026 •

Copy link
Copy Markdown
Owner

Build-only. No section content changes. The deployed PDF has printed 78 words past the physical page edge, on 40 pages. Most are provenance paths (Research/proposals/<long_name>.md, synchronism-site/explorer/findings/<long-name>.md) that a reader cannot finish reading. The fix has been on #14 since 2026-10-31 (62f7610a). #14 is held on editorial and scientific grounds, so the formatting fix has waited with it for 14 days. This PR carries only the build part, so it can merge on its own.

What changes: whitepaper/pdf-breakable-code.lua (pandoc filter, PDF only). For inline code of 20 or more characters, it emits the \texttt itself with \allowbreak after each / _ - .. Nothing is inserted into the text, and the HTML builds are untouched. make-pdf.sh gains --lua-filter=pdf-breakable-code.lua, and PUBLISHER_CONTEXT.md §5 gains the geometry check. All three hunks are byte-identical to 62f7610a. git merge-tree of #14's head (08227d0b) with this branch is clean.

Measured on origin/main 97f638a9 (CI=true make-md / make-pdf rc 0):

deployed (main) this branch
words past page width − 36 pt 96 1 (p6 plain-text table cell: the 10-31 baseline)
words past 612 pt (clipped at the edge) 78 on 40 pages 0
pages 292 292
Missing character / U+FFFD 0 / 0 0 / 0

Text layer check. After removing whitespace, hyphens and page-number-only lines, every difference falls into one of two groups:

  • 99 insertions in the new build (1,780 chars). These are path tails restored, e.g. …the-baseline-is-o → …the-baseline-is-off-by-3-dex.md.].
  • 16 layout moves: page numbers caught mid-line by pdftotext, one TOC entry 184 → 183, and table reading order where a cell with long code can now wrap.

The one character-level case (site+4b9c681 on main vs site 4b9c681 here) matches the source (dark_matter.md L177 has no +). Page 15 was rendered before and after: the path that ran off the edge now wraps at / and -.

Artifacts are not committed. make-pdf.sh is under the workflow's trigger paths, so CI rebuilds and deploys on merge.

🤖 Generated with Claude Code

— cbp-claude

…age (78 clipped words on 40 pages -> 0), split out of #14

The fix landed on #14 on 2026-10-31 (62f7610). #14 has been held on
editorial grounds since, so the deployed PDF has kept the clipped paths for
14 days. This carries only the build part, with no section content: the Lua
filter, one make-pdf.sh line, and one PUBLISHER_CONTEXT check line. The
three hunks are identical to #14's, so #14 merges over it cleanly.

Measured on origin/main 97f638a (sections unchanged), CI=true make-md/make-pdf rc 0:
- words past page width - 36 pt: 96 -> 1 (p6 plain-text table cell, the 10-31 baseline)
- words past the page edge (612 pt, clipped): 78 on 40 pages -> 0
- pages 292 -> 292; Missing character 0; U+FFFD 0
- text layer: after removing whitespace, hyphens and page-number lines, the
  differences are 99 insertions (1,780 chars, restored path tails) plus 16 layout
  moves (page numbers caught mid-line, one TOC entry 184 -> 183, table reading
  order). The one character-level case ("site+4b9c681" -> "site 4b9c681") matches
  the source, which has no "+".
- p15 rendered before and after: the path that ran off the edge now wraps at / and -.

Build artifacts are not committed; CI rebuilds and deploys on merge
(make-pdf.sh is under the trigger path).

Seat: cbp-claude (interactive maintenance pass)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@dp-web4
dp-web4 force-pushed the publisher/pdf-breakable-code-1114 branch from 1d32a06 to bdea342 Compare November 14, 2026 12:44
dp-web4 added a commit that referenced this pull request Nov 14, 2026
…ted datum); clipping fix split out to #16; count HELD at 6; web4 no change

Window 97f638a relays site 6f25275 + 2e4a627. The density law at the archive
knee (+8,404), solar boost 3.5e4 -> 2.7-10.5, L3 read by keying, Hyades, DR4
default, the Navier-Stokes banner and the door-3 pointer: none reach the paper.
N-S is already caveated at 4 sites, and the paper's L3 is its own division-law
label.

The deployed PDF clips 78 words past the page edge on 40 pages. The 10-31 fix was
stacked here under the hold, so its build-only part is now #16 (96 -> 1
past margin, 78 -> 0 clipped). Frame point: the site exclusion table's
strictest rule keeps none of the six.

Seat: cbp-claude (interactive maintenance pass)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

@dp-web4 dp-web4 left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

NOVA exact-head GO for independent build-only merge. Three-file PDF linebreak repair for long inline code, no scientific content or headline claims changed. The Lua filter applies to >=20-character Code nodes in LaTeX only, escapes TeX specials and inserts allowbreak after separators, with documented bbox geometry check. Green CI. It can land separately from held scientific/editorial #14. Nonblocking: retain glyph and PDF geometry inspection after merge and verify readable breaks in final PDF.

@dp-web4
dp-web4 merged commit 8e23c11 into main Nov 15, 2026
2 checks passed
dp-web4 added a commit that referenced this pull request Nov 15, 2026
…but never inscribed, now CHECKED (P(k) pins the fluid local horn; striction at the archive knee); count HELD at 6; web4 no change

L366: the 08-18 locality fork ("no order-eps0 channel", "~10x underpowered")
holds only without the pressure term. Read at local density as a fluid
(fluid Cardassian), the Jeans term makes existing P(k) exclude the horn
unless gamma = 1/2 +/- 5e-6; gamma = 0.487 survives only on the
mean-density reading. Correction dated 09-15, backfill-audit item 6;
explorer script + controls re-run byte-identical.

L117: "polarization force <= 2e-5 of g ... observationally free" was
computed at the 0.029 V^2 units slip. At the archive 0.028 V^0.5 knee
(site explorer 11-14 6d97ef9): radial striction median 0.5-1.6% of
gravity, vertical 9-27% of vertical gravity, compressive. Three scripts
re-run, output identical apart from timing lines. EFE statements left to
open item 4.

Merges origin/main (#16 build fix, f3b3c5b back-annotation).

Seat: cbp-claude (interactive maintenance pass)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant