DOCX: large title paragraph under docGrid renders short, gap to body too small

[developer0hye]

Summary

The large title paragraph in 02_contract_ko.docx renders too short vertically and sits too high, leaving the title-to-first-body gap ~10pt short of Word. The per-body-paragraph grid rhythm is correct (fixed in #394), but the title region contributes a constant ~13.6pt upward offset carried through the rest of the page.

Measured (pdftotext at 72 DPI):

	GT	office2pdf
Title baseline (소프트웨어)	65.9	62.3
First body baseline (주식회사)	105.9	92.3
Title→body gap	40.0pt	30.0pt
Title glyph ink height	22.7pt	17.0pt

The title run is w:sz w:val="34" (17pt). Glyph widths match GT exactly (85.0pt), so the horizontal font size is correct, but the rendered glyph height differs (22.7 vs 17.0) — pointing at a font-substitution or line-box difference specific to the bold title line rather than the point size. Under the section's w:docGrid the title line snaps to a single 18pt grid line in office2pdf, whereas Word appears to give it more vertical room, widening the title-to-body gap.

Reproduction

Convert a Korean DOCX with a large bold title paragraph under a w:docGrid section and compare the title height and title-to-body gap against a Word PDF export at 150 DPI.

Expected

Title line height and the title-to-first-body gap match Word (gap ~40pt in the fixture).

Actual

Title renders ~5.7pt shorter and the gap is ~10pt short, shifting all subsequent content up by a constant ~13.6pt.

Related

Related: #394, #354

[developer0hye]

Investigation update (this session): re-measured after #394 (grid boundary leading) and #398 (fixed metric line box) merged — the title→body gap is still 30.0pt vs GT 39.9pt (title at 62.3 vs GT 65.9, body at 92.3 vs GT 105.9), so those grid fixes did not incidentally resolve it. The gap is the large bold title paragraph's line height under the substitute Korean font plus Word's title spacing model; same font-substitution constraint as #404 (Malgun unavailable on this host), so it can't be verified against GT here.

[developer0hye]

Correction: not a font-substitution problem — Malgun Gothic IS available and embedded (verified via pdffonts; Korean glyph widths match GT). The residual is the large bold title's line height / title→body gap (30pt vs GT 39.9pt) under docGrid — Word's title-paragraph spacing model, not the font.

[developer0hye]

Resolving as a documented layout-model limitation (per the standing directive; honest state + path forward).

Corrected root cause: NOT font substitution — Malgun Gothic is available and embedded (verified via pdffonts; Korean glyph widths match GT). Confirmed the merged grid fixes (#394 boundary leading, #398 fixed metric box) did not incidentally resolve it: the title→body gap is still 30.0pt vs GT 39.9pt.

Real residual: the large bold title paragraph's line height / spacing under docGrid. office2pdf gives the 17pt title one grid line; Word gives it more vertical room. This is Word's title-paragraph row-height/spacing model, part of the same table/title row-height modeling as #404.

Work needed to truly close: implement Word's title/large-font paragraph grid row-height behavior, calibrated against Word-exported samples. Tracked here.

[developer0hye]

Fresh business-corpus audit on current main (dc905e9a) confirms this defect remains visible and materially contributes to one of the highest-difference DOCX pages, so I am reopening it under the current GT closing rule.

For 02_contract_ko.docx, pdftotext -bbox gives:

Element	Word GT yMin	office2pdf yMin
17 pt title	65.940 pt	62.350 pt
First body line	105.871 pt	92.350 pt
Title-to-body gap	39.930 pt	30.000 pt

The current 150-DPI comparison has 204,220 differing pixels at 5% fuzz (0.093896). Text, emphasis, borders, fills, page count, and content order are preserved; the unresolved defect is the large-title line box / following-paragraph gap.

Source: 02_contract_ko.docx
Native Word PDF: 02_contract_ko.pdf