fix: doppelte Tabellen-Anker und Bild-Links im Linkcheck

- Komplexe Tabellen mit verlinkter id erzeugten den Anker zweimal
  (<a id> und <table id>), also ungültiges HTML.
- Der Linkcheck übersah das äußere Ziel von Bild-Links
  [![alt](img)](ziel) und startete Links an escapten \[-Klammern.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

epub2md/html_converter.py

@@ -616,7 +616,7 @@ class Renderer:
             if self.ch:
                 self.ch.stats.tables_html += 1
             self.diag.warn("table-html", f"Tabelle enthält {reason}; als HTML erhalten.", self._source(el), self.file)
-            return ([anchor] if anchor else []) + [self.clean_html(el)]
+            return [self.clean_html(el)]  # clean_html keeps the id on <table> itself
         out = [anchor] if anchor else []
         if caption_el is not None:
             cap = join_inline(self.inline_children(caption_el), hard_break=" ")

epub2md/linkcheck.py

@@ -10,7 +10,9 @@ from urllib.parse import unquote
 FENCE = re.compile(r"^(`{3,}|~{3,}).*?$.*?^\1\s*$", re.M | re.S)
 # inline code spans never cross a blank line and never start at an escaped backtick
 INLINE_CODE = re.compile(r"(?<![\\`])(`+)(?!`)((?:[^`\n]|\n(?![ \t]*\n)|`(?!\1))+?)(?<!`)\1(?!`)")
-MD_LINK = re.compile(r"!?\[(?:[^\]\\]|\\.)*\]\((<[^>]*>|[^)\s]+)(?:\s+\"[^\"]*\")?\)")
+# link text: never starts at an escaped \[, never spans a blank line, may contain one bracket level
+# (image links like [![alt](img)](target))
+MD_LINK = re.compile(r"(?<!\\)!?\[((?:[^\[\]\\\n]|\\.|\[(?:[^\[\]\\\n]|\\.)*\]|\n(?![ \t]*\n))*)\]\((<[^>]*>|[^)\s]+)(?:\s+\"[^\"]*\")?\)")
 HTML_ATTR = re.compile(r"""\b(?:href|src)="([^"]*)\"""")
 ANCHOR = re.compile(r"""\bid="([^"]+)\"""")
 HEADING = re.compile(r"^#{1,6}\s+(.*?)\s*#*\s*$", re.M)
@@ -38,7 +40,14 @@ def anchors_in(text: str) -> set[str]:
 
 def extract_links(text: str) -> list[str]:
     body = strip_code(text)
-    links = [m.group(1) for m in MD_LINK.finditer(body)]
+    links: list[str] = []
+
+    def collect(text: str) -> None:
+        for m in MD_LINK.finditer(text):
+            links.append(m.group(2))
+            collect(m.group(1))  # links/images inside the link text
+
+    collect(body)
     links += HTML_ATTR.findall(body)
     return [link[1:-1] if link.startswith("<") and link.endswith(">") else link for link in links]
 

tests/test_links.py

@@ -66,3 +66,18 @@ def test_link_checker_detects_problems():
     }
     broken = check_links(files, {"media/images/a.png"})
     assert {b["link"] for b in broken} == {"002-b.md#nein", "003.md"}
+
+
+def test_link_checker_ignores_escaped_brackets():
+    """An escaped \\[ must not start a link that swallows the following paragraphs."""
+    files = {
+        "a.md": '\\[»\\] Hinweis\n\n<a id="x"></a>Text [kaputt](fehlt.md) und \\[1\\]\n\n[ok](#x)\n',
+    }
+    broken = check_links(files, set())
+    assert [b["link"] for b in broken] == ["fehlt.md"]
+
+
+def test_link_checker_handles_image_links():
+    files = {"a.md": "[![Abb.](bild.png)](fehlt.md) [![ok](bild.png)](#x) <a id=\"x\"></a>\n"}
+    broken = check_links(files, {"bild.png"})
+    assert [b["link"] for b in broken] == ["fehlt.md"]

tests/test_tables_lists.py

@@ -63,3 +63,10 @@ def test_unknown_elements_are_reported_and_keep_html(convert):
     kept = convert(Book(docs=[Doc("c.xhtml", body)]), "--keep-html")
     assert '<foo-widget data-x="1">Inhalt bleibt</foo-widget>' in kept.chapter("001")
     assert "evil" not in kept.chapter("001")
+
+
+def test_complex_table_anchor_is_emitted_once(convert):
+    body = ('<h1>T</h1><p><a href="#t1">Tabelle 1</a></p><table id="t1"><tr><td colspan="2">a</td></tr>'
+            "<tr><td>b</td><td>c</td></tr></table>")
+    text = convert(Book(docs=[Doc("c.xhtml", body)])).chapter("001")
+    assert text.count('id="t1"') == 1