fix: invalid code span detection in reference extractor

Signed-off-by: squidfunk <martin.donath@squidfunk.com>

python/tests/unit/collectors/test_references.py

@@ -1819,6 +1819,27 @@ class TestInlineCode:
         assert text(md, links[0].text) == b"text"
         assert text(md, links[0].href) == b"href"
 
+    @pytest.mark.parametrize(
+        "md",
+        [
+            pytest.param(
+                b"a`\n\n`[b]`",
+                id="blank-line",
+            ),
+            pytest.param(
+                b"a`\n \n`[b]`",
+                id="whitespace-only-line",
+            ),
+            pytest.param(
+                b"a`\r\n\r\n`[b]`",
+                id="blank-line-crlf",
+            ),
+        ],
+    )
+    def test_inline_code_does_not_cross_blank_line(self, md: bytes) -> None:
+        refs = collect(md)
+        assert len(refs) == 0
+
     # --- negative cases ---
 
     @pytest.mark.parametrize(

python/zensical/collectors/references/cursor.py

@@ -1083,6 +1083,19 @@ def _scan_inline_code(cursor: Cursor) -> int | None:
 
     # Search for matching closing backticks
     while pos < cursor.end:
+        # Inline code spans may contain line endings, but cannot cross a
+        # blank line because that starts a new block in Python-Markdown.
+        if cursor.data[pos] in (_CR, _NL):
+            line = pos + 1
+            if (
+                cursor.data[pos] == _CR
+                and line < cursor.end
+                and cursor.data[line] == _NL
+            ):
+                line += 1
+            if _is_blank_line(cursor, line):
+                return None
+
         if cursor.data[pos] == _BACKTICK:
             end = pos