Add Word bisection fixture batch for full-book open failure

Local audit of the full-book output is now exhaustive and clean (package
parts, content types, rels, styles, numbering, bookmark ids, fields,
sectPr/settings ordering, control chars, header/footer machine). LibreOffice
renders the file fine but Word hard-refuses it, so the trigger must be
isolated empirically: each candidate removes exactly one feature family.

- tests/make_bisect_files.py: regenerates six candidates from the final
  full-book output (no running heads; no oddPage/titlePg; no field runs;
  no style refs; first 8 sections only; no images).
- tests/word_bisect/bisect_0[1-6]_*.docx: the six fixtures (all verified to
  load via python-docx and convert via LibreOffice).

.gitignore

@@ -13,6 +13,7 @@ umbraformat/formatted/
 # Documents / user scratch files (not part of the app source)
 *.docx
 !tests/*.docx
+!tests/word_bisect/*.docx
 summary.txt
 Sections_Tags
 Word_to_PDF_print_formatting

tests/make_bisect_files.py

@@ -0,0 +1,144 @@
+#!/usr/bin/env python
+"""Generate Word-diagnostic bisection files from the real full-book output.
+
+Word refuses the full-book output with a hard "Word experienced an error
+trying to open the file" dialog while the mini-book fixture opens fine, and
+every structural OOXML audit on the full file passes. LibreOffice renders all
+of these files, so Word itself is the only oracle that can tell them apart.
+
+Each candidate starts from the final full-book DOCX and removes exactly one
+feature family:
+
+  01 no_running   - every sectPr headerReference/footerReference removed
+  02 no_oddpage   - sectPr w:type -> nextPage, titlePg dropped, evenAndOddHeaders off
+  03 no_fields    - all fldChar/instrText runs removed (TOC PAGEREF machinery)
+  04 no_styles    - all w:pStyle / w:rStyle removed (style-system interaction)
+  05 first8       - body truncated after the 8th section boundary
+  06 no_images    - drawing paragraphs removed
+
+Open each in Word; whichever show the error dialog pinpoint the feature
+family at fault.
+"""
+import os
+import sys
+
+from docx import Document
+from docx.oxml.ns import qn
+from lxml import etree  # noqa: F401  (kept for parity with the other makers)
+
+W = '{http://schemas.openxmlformats.org/wordprocessingml/2006/main}'
+
+HERE = os.path.dirname(os.path.abspath(__file__))
+DEFAULT_BASE = os.path.normpath(os.path.join(
+    HERE, '..', 'umbraformat', 'formatted',
+    '3a7d086e-c5ab-47bc-a3c5-0ad6352bd9ae_formatted.docx'))
+OUTDIR = os.path.join(HERE, 'word_bisect')
+
+NAMES = [
+    'bisect_01_no_running.docx',
+    'bisect_02_no_oddpage.docx',
+    'bisect_03_no_fields.docx',
+    'bisect_04_no_styles.docx',
+    'bisect_05_first8.docx',
+    'bisect_06_no_images.docx',
+]
+
+
+def _save(doc, name):
+    os.makedirs(OUTDIR, exist_ok=True)
+    path = os.path.join(OUTDIR, name)
+    doc.save(path)
+    return path
+
+
+def make_no_running(doc):
+    """Remove header/footer references from every section (incl. first-page)."""
+    for sectPr in doc.element.body.iter(qn('w:sectPr')):
+        for tag in ('w:headerReference', 'w:footerReference'):
+            for el in sectPr.findall(qn(tag)):
+                sectPr.remove(el)
+
+
+def make_no_oddpage(doc):
+    """Neutralise the oddPage section-start / odd-even running machinery."""
+    for sectPr in doc.element.body.iter(qn('w:sectPr')):
+        t = sectPr.find(qn('w:type'))
+        if t is not None:
+            t.set(qn('w:val'), 'nextPage')
+        tp = sectPr.find(qn('w:titlePg'))
+        if tp is not None:
+            sectPr.remove(tp)
+    settings = doc.settings.element
+    for el in settings.findall(qn('w:evenAndOddHeaders')):
+        settings.remove(el)
+
+
+def make_no_fields(doc):
+    """Drop every run carrying field code/result machinery."""
+    for p in doc.element.body.iter(qn('w:p')):
+        for r in list(p.findall(qn('w:r'))):
+            if r.find(qn('w:fldChar')) is not None or r.find(qn('w:instrText')) is not None:
+                p.remove(r)
+
+
+def make_no_styles(doc):
+    """Remove style references from all paragraphs/runs (renders as Normal)."""
+    for el in list(doc.element.body.iter()):
+        if el.tag in (qn('w:pStyle'), qn('w:rStyle')):
+            el.getparent().remove(el)
+
+
+def make_first8(doc):
+    """Truncate the body after the 8th section boundary (move its sectPr to
+    body level and drop every later sibling, incl. all later sections)."""
+    body = doc.element.body
+    markers = []
+    for s in body.iter(qn('w:sectPr')):
+        parent = s.getparent()
+        if parent is not None and parent.tag == qn('w:pPr'):
+            markers.append((s, parent.getparent()))
+    sect, para = markers[7]  # 8th boundary (0-based)
+    sect.getparent().remove(sect)
+    nxt = para.getnext()
+    while nxt is not None:
+        nxt2 = nxt.getnext()
+        body.remove(nxt)
+        nxt = nxt2
+    for fin in body.findall(qn('w:sectPr')):
+        body.remove(fin)
+    body.append(sect)
+
+
+def make_no_images(doc):
+    """Remove paragraphs containing inline drawings."""
+    for p in list(doc.element.body.iter(qn('w:p'))):
+        if p.find(qn('w:drawing')) is not None:
+            p.getparent().remove(p)
+
+
+MAKERS = [
+    make_no_running,
+    make_no_oddpage,
+    make_no_fields,
+    make_no_styles,
+    make_first8,
+    make_no_images,
+]
+
+
+def main():
+    base = sys.argv[1] if len(sys.argv) > 1 else DEFAULT_BASE
+    if not os.path.exists(base):
+        sys.exit(f"base output not found: {base}\n"
+                 "Regenerate it via the API first, or pass its path as argv[1].")
+    for maker, name in zip(MAKERS, NAMES):
+        doc = Document(base)
+        maker(doc)
+        path = _save(doc, name)
+        # sanity: it must still open via python-docx
+        Document(path)
+        print(f"wrote {path}")
+
+
+if __name__ == '__main__':
+    main()