From 626169fa20b16d579fb5fc4973ac04c5bde67a4a Mon Sep 17 00:00:00 2001
From: dryheatwindbag <office@jpwlegal.com>
Date: Sat, 4 Jul 2026 16:07:50 -0700
Subject: [PATCH] fix: preserve unknown empty elements inside runs
 (w:commentReference)

CT_R::from_xml's Event::Empty branch handled tab/br/footnoteReference/
endnoteReference and silently dropped every other empty element inside
<w:r>. This destroys w:commentReference anchors on round-trip (the
comments part survives but every comment is orphaned), and loses any
other empty run-level markup.

Add the same unknown-element fallback the function already has for
Event::Start (and CT_P has for both): capture the element into
extra_xml via capture_empty_element.

Includes a regression test: a run containing <w:commentReference/>
round-trips verbatim.
---
 crates/rdocx-oxml/src/text.rs | 26 ++++++++++++++++++++++++++
 1 file changed, 26 insertions(+)

diff --git a/crates/rdocx-oxml/src/text.rs b/crates/rdocx-oxml/src/text.rs
index c2f7b66..4156db8 100644
--- a/crates/rdocx-oxml/src/text.rs
+++ b/crates/rdocx-oxml/src/text.rs
@@ -174,6 +174,11 @@ impl CT_R {
                             .and_then(|a| std::str::from_utf8(&a.value).ok()?.parse::<i32>().ok())
                             .unwrap_or(0);
                         content.push(RunContent::EndnoteRef { id });
+                    } else {
+                        // Capture unknown empty child elements (e.g.
+                        // w:commentReference) as raw XML, mirroring the
+                        // Event::Start fallback above.
+                        extra_xml.push(capture_empty_element(e)?);
                     }
                 }
                 Ok(Event::End(ref e)) if matches_local_name(e.name().as_ref(), b"r") => {
@@ -652,6 +657,27 @@ mod tests {
         assert_eq!(parsed.hyperlinks[0].run_end, 2);
     }
 
+    #[test]
+    fn unknown_empty_run_children_roundtrip() {
+        // Unknown empty elements inside a run (e.g. w:commentReference)
+        // must be captured and re-emitted, not silently dropped.
+        let p = parse_paragraph(
+            r#"<w:r><w:t>flagged</w:t></w:r><w:r><w:commentReference w:id="1"/></w:r>"#,
+        );
+        assert_eq!(p.runs.len(), 2);
+        assert_eq!(p.runs[1].extra_xml.len(), 1);
+
+        let mut output = Vec::new();
+        let mut writer = Writer::new(&mut output);
+        p.to_xml(&mut writer).unwrap();
+        let xml = String::from_utf8(output).unwrap();
+        assert!(
+            xml.contains(r#"<w:commentReference w:id="1"/>"#),
+            "comment reference must survive round-trip: {xml}"
+        );
+        assert!(xml.contains("flagged"));
+    }
+
     #[test]
     fn parse_fld_simple_page() {
         let p = parse_paragraph(