From 467b070326b8e8f4c2b1419860cf8d3b1428e221 Mon Sep 17 00:00:00 2001
From: dschlueter <dschlueter@kitux.de>
Date: Mon, 5 Oct 2026 23:59:02 +0200
Subject: [PATCH] =?UTF-8?q?fix:=20Notennummer=20und=20R=C3=BCcklinks=20nic?=
 =?UTF-8?q?ht=20an=20Codeblock-Zeilen=20anh=C3=A4ngen=20(QA2-02)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Bei Anmerkungen im Anker-Stil (--format commonmark, --footnotes
end-of-book) wurde der Rücklink [↩] an die schließende Fence-Zeile
angehängt. Der Codeblock blieb dadurch offen und verschluckte den Rest
der Datei; epub2md_merge.py schrieb die Links darin nicht mehr um.
Beginnt oder endet eine Anmerkung mit einem Block (Code, Tabelle,
Liste, Formel), stehen Nummer bzw. Rücklinks jetzt in eigenem Absatz.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---
 epub2md/html_converter.py    | 22 ++++++++++++++++++---
 tests/test_qa_regressions.py | 38 ++++++++++++++++++++++++++++++++++++
 2 files changed, 57 insertions(+), 3 deletions(-)

diff --git a/epub2md/html_converter.py b/epub2md/html_converter.py
index ae76927..f77f94c 100644
--- a/epub2md/html_converter.py
+++ b/epub2md/html_converter.py
@@ -152,6 +152,23 @@ def indent_block(text: str, first: str, rest: str) -> str:
     return "\n".join(out)
 
 
+_BLOCK_SYNTAX = ("```", "~~~", "|", "<table", "- ", "> ", "#", "$$", "\\[", "* * *")
+
+
+def _is_block_syntax(block: str) -> bool:
+    return block.startswith(_BLOCK_SYNTAX) or re.match(r"\d+\. ", block) is not None
+
+
+def _labeled_note(label: str, blocks: list[str], back: str) -> str:
+    """Anchor-style note: label before and backlinks after the body, without gluing them onto a code fence,
+    table or list line (which would break that block)."""
+    body = "\n\n".join(blocks)
+    text = f"{label}\n\n{body}" if _is_block_syntax(blocks[0]) else f"{label} {body}"
+    if back:
+        text += f"\n\n{back}" if _is_block_syntax(blocks[-1]) else f" {back}"
+    return text
+
+
 def _ancestors(el: Any, depth: int = 5) -> list[Any]:
     out: list[Any] = []
     node = el.getparent()
@@ -925,7 +942,7 @@ class Renderer:
                 back = " ".join(
                     f"[↩](#{ref})" for _, ref in self.note_refs.get(f"{self.file}:{note.label}", [])
                 )
-                anchor_style.append(indent_block(body, f'<a id="{note.label}"></a>**{num}:** ', "") + (f" {back}" if back else ""))
+                anchor_style.append(_labeled_note(f'<a id="{note.label}"></a>**{num}:**', blocks, back))
         if anchor_style:
             title = "Anmerkungen" if self.opt.language.startswith("de") else "Notes"
             out = [f"**{title}**"] + anchor_style + out
@@ -946,8 +963,7 @@ class Renderer:
             num = int(note.label.split("-")[1])
             blocks = self.render_note(note) or ["(leer)"]
             back = " ".join(f"[↩]({url_path(rel_link(self.file, f))}#{ref})" for f, ref in self.note_refs.get(note.label, []))
-            body = "\n\n".join(blocks)
-            out.append(f'<a id="{note.label}"></a>**{num}.** ' + body + (f" {back}" if back else ""))
+            out.append(_labeled_note(f'<a id="{note.label}"></a>**{num}.**', blocks, back))
         return "\n\n".join(out).strip() + "\n"
 
     # ------------------------------------------------------------------ images
diff --git a/tests/test_qa_regressions.py b/tests/test_qa_regressions.py
index 11793e0..fc52114 100644
--- a/tests/test_qa_regressions.py
+++ b/tests/test_qa_regressions.py
@@ -21,3 +21,41 @@ def test_pipe_in_inline_code_does_not_split_gfm_table(convert):
         cells = re.split(r"(?<!\\)\|", row.strip()[1:-1])
         assert len(cells) == 2, f"Zeile zerfällt in GFM in {len(cells)} Zellen: {row}"
     assert "||" in text and "a | b" in text and "logisches ODER" in text
+
+
+NOTE_WITH_CODE = ('<h1>K</h1><p>Text<a epub:type="noteref" href="#n1" id="r1">1</a>.</p>'
+                  '<aside epub:type="footnote" id="n1"><p>Note mit Code:</p><pre>code()\n    weiter()</pre></aside>'
+                  '<p>Zweite<a epub:type="noteref" href="#n2" id="r2">2</a>.</p>'
+                  '<aside epub:type="footnote" id="n2"><pre>nur_code()</pre></aside>'
+                  "<p>Danach kommt normaler Text.</p>")
+
+
+def _fences_closed(text: str) -> bool:
+    """Every opening ``` fence has a closing line that consists only of backticks."""
+    open_fence = None
+    for line in text.splitlines():
+        s = line.strip()
+        if open_fence is None:
+            m = re.match(r"(`{3,})", s)
+            if m:
+                open_fence = m.group(1)
+        elif re.fullmatch(r"`{3,}", s) and len(s) >= len(open_fence):
+            open_fence = None
+    return open_fence is None
+
+
+def test_backlink_after_code_block_does_not_break_fence_commonmark(convert):
+    """QA2-02: the ↩ backlink was appended to the closing ``` line, leaving the fence open."""
+    run = convert(Book(docs=[Doc("c.xhtml", NOTE_WITH_CODE)]), "--format", "commonmark")
+    text = run.chapter("001")
+    assert _fences_closed(text), text
+    assert not re.search(r"\*\*\d+[:.]\*\* ```", text), "Fence hinter Notennummer ist kein Codeblock"
+    assert "[↩](#fnref-001)" in text
+
+
+def test_backlink_after_code_block_does_not_break_fence_end_of_book(convert):
+    run = convert(Book(docs=[Doc("c.xhtml", NOTE_WITH_CODE)]), "--footnotes", "end-of-book")
+    notes = run.chapter("999")
+    assert _fences_closed(notes), notes
+    assert not re.search(r"\*\*\d+[:.]\*\* ```", notes), "Fence hinter Notennummer ist kein Codeblock"
+    assert "[↩](001-k.md#fnref-001)" in notes