docs: record what a metafile brings and why a line still breaks early

The line-break finding is sharpened: LibreOffice compresses inter-word
spaces to fit a line (91 of 349 full-width justified lines on the
thesis, up to 0.83px per space), which CSS justification cannot. A
matching negative word-spacing was measured and reverted.

CHANGELOG.md

@@ -56,7 +56,9 @@
 **Insert**
 - Tables: insert via size picker, Word-style row/column drag-resize, add / delete rows & columns, delete table, cell borders, merge cells and split cells (N×M, Word/LibreOffice-style), cell background shading, header row / first column toggles, named table styles (see Styles); a table splits cleanly across page boundaries
 - Table border control (Word/LibreOffice-style): per-side cell borders with presets (all / outside / inside / single edges / none), line width and color; buttons show active states matching the current pen and toggle borders off; round-trips to ODF `fo:border-*` and DOCX `w:tcBorders`
-- Images: inline or floating with text wrap (left / right / top-bottom), resize handles, rotation, live size badge; insert via toolbar, drag-and-drop, or paste. A floating frame is placed by the file's own offsets
+- Images: inline or floating with text wrap (left / right / top-bottom), resize handles, rotation, live size badge; insert via toolbar, drag-and-drop, or paste. A floating frame is placed by the file's own offsets, and two top-and-bottom frames set against opposite ends of the text share one band, side by side
+- EMF metafiles are drawn (`import/emf.ts`): the picture is rebuilt as SVG from the metafile's own records — paths, filled shapes, text with its font and colour, embedded bitmaps — so a plot pasted out of MATLAB or Excel arrives as a picture instead of a placeholder
+- Anything wider than the sheet — an oversized formula, a frame reaching past the margin — is cut at the page edge, as it is in LibreOffice
 - Charts are drawn from the file — DrawingML `chartN.xml` and ODF `chart:chart`, bar / line / area / scatter / pie, with their titles, axis titles, gridlines, axis bounds and series colours (read-only; see the limitations below)
 - Special characters picker
 - Date and time fields: picker with 7 date and 4 time formats (live samples) and an "update automatically" toggle — fixed fields keep the inserted moment, auto fields refresh on open. Round-trips to ODF `text:date`/`text:time` (minted `number:date/time-style`) and DOCX `DATE`/`TIME` fields; the field carries the surrounding font
@@ -109,7 +111,7 @@ The gap against Word/LibreOffice, most valuable first. Reviewed 2026-08-08.
 - Comments / annotations: dropped on import (`office:annotation`, `w:comment`)
 - Track changes / revisions: no recording, no accept/reject, no author colors (`text:tracked-changes`, `w:ins`/`w:del`); imported revisions are flattened to their current state
 - Bookmarks and cross-references: no anchor, no reference field ("see chapter 3 on page 7"), none survive an import — this also blocks internal hyperlinks
-- Charts are **drawn** from the file (`import/chart.ts`: DrawingML `chartN.xml` and ODF `chart:chart`), but as a picture, not a chart object — a re-export carries the drawing and the numbers behind it are no longer editable. OLE objects and WMF/EMF/SVM metafiles keep their box and a placeholder label, and export writes that back out (`import/imageFormats.ts`): no JS decoder exists for the metafile formats, and none can be embedded without a backend
+- Charts are **drawn** from the file (`import/chart.ts`: DrawingML `chartN.xml` and ODF `chart:chart`), but as a picture, not a chart object — a re-export carries the drawing and the numbers behind it are no longer editable. The same holds for an **EMF** metafile (`import/emf.ts`): it is drawn, but as the SVG picture it was rebuilt into, and only from the record set a plot consists of — a hatched brush, a clipping region or a rotated bitmap is skipped. **WMF/SVM** metafiles and OLE objects still keep their box and a placeholder label, and export writes that back out: WMF is a different (16-bit) record format, SVM is StarOffice-proprietary, and an OLE object cannot be rendered without its application
 - Shapes beyond rect / round-rect / ellipse: lines, arrows, connectors, polygons, freeform, and rotated shape text (dropped on import with a warning)
 - Text boxes in the DOCX export: lists inside a box are flattened to literal-marker paragraphs (`•` / `1.`) and images inside a box are dropped — the box XML is injected by a post-pack string pass (`export/docx.ts` `applyTextBoxesDocx`) that mints no numbering.xml or media/rels entries. The ODT export has neither limitation
 
@@ -167,11 +169,15 @@ merely unimplemented belongs in the list above, not here.
   reach is a frame a word processor puts *inside* running text: Chromium moves
   every line after a full-width float below it, so such a frame lands after the
   paragraph's text instead. Noted 2026-08-08, revised 2026-08-09.
-- A line takes one word more or fewer than LibreOffice where its natural width
-  lands within ~0.2 mm of the right margin. LibreOffice quantizes every glyph
-  advance, Chromium keeps it fractional, and the two drift apart along the line
-  (measured on the thesis fixture: identical word widths, 0.45 mm apart by the
-  150th mm). Nothing in CSS exposes the quantization. Noted 2026-08-09.
+- A line takes one word more or fewer than LibreOffice, for two reasons neither
+  of which CSS exposes. LibreOffice **compresses** inter-word spaces to fit a
+  line (measured on the thesis: 91 of 349 full-width justified lines, up to
+  0.83 px per space at 12 pt) where CSS justification only expands; and it
+  quantizes every glyph advance where Chromium keeps it fractional, so the two
+  drift apart along the line (identical word widths, 0.45 mm apart by the
+  150th mm). A matching negative `word-spacing` was tried and reverted — it
+  fixes the break but shortens every last line; see
+  `tests/render-parity/README.md`. Noted 2026-08-09.
 - A text box anchored inside a paragraph loses the vertical offset it was
   anchored by: it is a block node here, so the importer lifts it out and it
   simply follows that paragraph (measured 4.7 mm on the thesis' figure page).

tests/render-parity/README.md

@@ -133,13 +133,19 @@ with 51 line-level differences — most of them the two engine rules below, not
   the value odf-kit writes for level 2, where it draws its own flat 0.635cm hanging
   (1.909cm) instead. So reading the attribute back naively moves the markers of our own
   exports. Whatever rule that is has to be understood before the value is trusted.
-- Lines whose natural width lands within ~0.2mm of the right margin may take one word
-  more or fewer than LibreOffice. Sub-0.1mm engine rounding, not a layout rule — but it
-  **accumulates**: measured on the thesis' page 2, matching word widths (11.86mm for
-  `haben,` on both sides) had drifted 0.45mm apart by the 150th mm of the line, which is
-  what decides whether the next word fits. LibreOffice quantizes each glyph advance where
-  Chromium keeps it fractional, and nothing in CSS exposes that. Most of the thesis'
-  remaining `lineBreak` reports are this.
+- Lines take one word more or fewer than LibreOffice, which is most of the thesis'
+  remaining `lineBreak` reports. Two causes, both measured on that fixture:
+  - **LibreOffice compresses inter-word spaces to fit a line**, CSS justification only
+    expands. Of 349 full-width justified lines, **91** are narrower in LibreOffice than
+    their natural width here — up to **0.83px per space** at 12pt (p99 0.68px).
+    Attempted and **reverted**: `word-spacing: -0.05em` on justified blocks. It does fix
+    the break decision (the thesis' page 6 paragraph goes 7 lines → LibreOffice's 6), but
+    the corpus got *worse* — `lineBreak` 45 → 43 while `lineEnd` went 2 → 16, because the
+    tighter spacing also shortens every paragraph's unjustified last line by 1–3mm.
+  - Sub-0.1mm engine rounding, which **accumulates**: matching word widths (11.86mm for
+    `haben,` on both sides) had drifted 0.45mm apart by the 150th mm of the line.
+    LibreOffice quantizes each glyph advance where Chromium keeps it fractional, and
+    nothing in CSS exposes that.
 
 ## Findings that did not survive measurement