fix(comments): emit required <a:bodyPr> in <p188:txBody> (maintainer UAT)

Maintainer opened uat_issue25_comments.pptx in PowerPoint: file opened
clean (no repair) but ZERO comments displayed. Root cause: the add
path builds the body via get_or_add_txBody() (a bare <p188:txBody/>)
then sets text via _set_txBody_text, which only PRESERVED a pre-existing
<a:bodyPr> and never created one. <p188:txBody> is a DrawingML
a:CT_TextBody whose schema REQUIRES <a:bodyPr> as first child;
PowerPoint silently drops comments with a bodyPr-less txBody — no
repair dialog, the comment is just absent. The entire green trinity +
17/17 UAT could not see this; only maintainer visual review did.

Fix: _set_txBody_text now GUARANTEES a leading <a:bodyPr/> (inserts it
when absent) for both comments and replies. Added
DescribeThreadedCommentBodyPrRegression (2 tests) asserting every
txBody carries <a:bodyPr> before <a:p>, so the silent-drop class is
now test-caught. uat_issue25.py OUT path moved under ./uat/ per
CLAUDE.md §6a update.

Trinity: pytest 3753, ruff clean, behave 1062, 0 failed.
Refs #25

Author: MHoroszowski

src/pptx/oxml/comments.py

@@ -54,13 +54,19 @@ def _txBody_text(txBody) -> str:
 def _set_txBody_text(txBody, value: str) -> None:
     """Replace `txBody`'s body with a single paragraph/run holding `value`.
 
-    Existing ``a:p`` children are removed and one fresh
-    ``<a:p><a:r><a:t>value</a:t></a:r></a:p>`` is appended after the
-    (preserved) ``<a:bodyPr>``. This is the minimal round-trip-safe shape
-    PowerPoint accepts for a threaded-comment body.
+    ``<p188:txBody>`` is a DrawingML ``a:CT_TextBody`` whose schema REQUIRES
+    ``<a:bodyPr>`` as its first child. A txBody without it is silently
+    dropped by PowerPoint's threaded-comment reader — the comment never
+    appears in the Comments pane (no repair dialog, just absent). The
+    add-path builds the txBody via ``get_or_add_txBody()`` (a bare
+    ``<p188:txBody/>``), so this function must GUARANTEE the leading
+    ``<a:bodyPr/>``, not merely preserve a pre-existing one. A trailing
+    ``<a:p><a:r><a:t>value</a:t></a:r></a:p>`` follows it.
     """
     for p in list(txBody.findall(qn("a:p"))):
         txBody.remove(p)
+    if txBody.find(qn("a:bodyPr")) is None:
+        txBody.insert(0, parse_xml('<a:bodyPr xmlns:a="%s"/>' % _A_NS))
     p = parse_xml(
         '<a:p xmlns:a="%s"><a:r><a:t>%s</a:t></a:r></a:p>' % (_A_NS, _escape_xml_text(value))
     )

tests/test_issue25_comments.py

@@ -626,5 +626,53 @@ def it_treats_a_legacy_comment_replies_as_empty_read_only(self):
         assert list(legacy.replies) == []
 
 
+class DescribeThreadedCommentBodyPrRegression:
+    """Regression: <p188:txBody> MUST carry <a:bodyPr> (issue #25 silent-drop).
+
+    The add-path builds txBody via get_or_add_txBody() (a bare element);
+    a CT_TextBody without <a:bodyPr> is schema-malformed and PowerPoint
+    SILENTLY drops the comment (no repair dialog, comment just absent —
+    caught only by maintainer visual review). Trinity was green while
+    every comment was invisible. This locks the fix.
+    """
+
+    def _modern_xml(self):
+        import io
+        import zipfile
+
+        from pptx import Presentation
+
+        prs = Presentation()
+        s = prs.slides.add_slide(prs.slide_layouts[1])
+        c = s.comments.add("Body check", author="Rev")
+        c.replies.add("reply body check", author="Rev2")
+        buf = io.BytesIO()
+        prs.save(buf)
+        buf.seek(0)
+        z = zipfile.ZipFile(buf)
+        name = next(n for n in z.namelist() if "modernComment" in n)
+        return z.read(name).decode()
+
+    def it_emits_bodyPr_in_every_comment_txBody(self):
+        import re
+
+        xml = self._modern_xml()
+        bodies = re.findall(r"<p188:txBody>(.*?)</p188:txBody>", xml, re.S)
+        assert bodies, "expected at least one <p188:txBody>"
+        assert all("bodyPr" in b for b in bodies), (
+            "every threaded-comment/reply txBody must contain <a:bodyPr> "
+            "or PowerPoint silently drops the comment"
+        )
+
+    def it_places_bodyPr_before_the_paragraph(self):
+        import re
+
+        xml = self._modern_xml()
+        body = re.search(r"<p188:txBody>(.*?)</p188:txBody>", xml, re.S).group(1)
+        assert body.index("bodyPr") < body.index("<a:p"), (
+            "<a:bodyPr> must precede <a:p> (CT_TextBody child order)"
+        )
+
+
 if __name__ == "__main__":
     raise SystemExit(pytest.main([__file__, "-q"]))