From 7c137f4b750dfccc945ea06007ddd69e269a7d46 Mon Sep 17 00:00:00 2001
From: PythonWoods-Dev <gianluca.catalano@gmail.com>
Date: Sat, 4 Jul 2026 10:02:17 +0200
Subject: [PATCH] fix(core/validator): harden polyglot extractor comment
 masking and fence evasion

Signed-off-by: PythonWoods-Dev <gianluca.catalano@gmail.com>
---
 src/zenzic/core/validator.py | 17 +++++++++---
 tests/test_fix.py            | 51 ++++++++++++++++++++++++++++++++++++
 uv.lock                      |  2 +-
 3 files changed, 66 insertions(+), 4 deletions(-)

diff --git a/src/zenzic/core/validator.py b/src/zenzic/core/validator.py
index 1a5fe9d..0b6e278 100644
--- a/src/zenzic/core/validator.py
+++ b/src/zenzic/core/validator.py
@@ -223,7 +223,11 @@ class LinkInfo(NamedTuple):
 _POLY_INFO_SCHEMES: frozenset[str] = frozenset({"mailto:", "tel:", "ftp:"})
 
 # Pattern fence per PolyglotExtractor._mask_fences (subset di SuppressionTracker).
-_POLY_FENCE_RE: re.RegexPattern = re.compile(r"^(?P<fence>[`~]{3,})")
+_POLY_FENCE_RE: re.RegexPattern = re.compile(r"^(?P<fence>[`~]{3,})(?P<info>.*)$")
+
+# HTML and MDX Comment Regex Patterns for masking
+_POLY_COMMENT_RE: re.RegexPattern = re.compile(r"<!--.*?-->", re.DOTALL)
+_POLY_MDX_COMMENT_RE: re.RegexPattern = re.compile(r"\{\/\*.*?\*\/\}", re.DOTALL)
 
 # Strip whitespaces and control characters from URLs prima del check Z205.
 _POLY_CLEAN_URL_RE: re.RegexPattern = re.compile(r"[\s\x00-\x1F]+")
@@ -294,7 +298,7 @@ def extract(self, text: str) -> list[HtmlNodeInfo]:
             Lista di :class:`HtmlNodeInfo`, uno per ogni tag ``<a>``/``<img>``
             trovato fuori dai blocchi di codice.
         """
-        masked = self._mask_inline_code(self._mask_fences(text))
+        masked = self._mask_inline_code(self._mask_fences(self._mask_comments(text)))
         nodes: list[HtmlNodeInfo] = []
         for m in _RE_POLY_TAG.finditer(masked):
             tag = m.group(1).lower()
@@ -304,6 +308,12 @@ def extract(self, text: str) -> list[HtmlNodeInfo]:
             nodes.append(self._parse_node(tag, attrs_str, line_no, m.group(0)))
         return nodes
 
+    def _mask_comments(self, text: str) -> str:
+        """Mask HTML and MDX comments with spaces of equal length to preserve offsets."""
+        text = _POLY_COMMENT_RE.sub(lambda m: " " * len(m.group(0)), text)
+        text = _POLY_MDX_COMMENT_RE.sub(lambda m: " " * len(m.group(0)), text)
+        return text
+
     def _mask_inline_code(self, text: str) -> str:
         """Sostituisce blocchi inline code con spazi bianchi preservando gli offset."""
         from zenzic.core.validator import _INLINE_CODE_RE
@@ -338,7 +348,8 @@ def _mask_fences(self, text: str) -> str:
             else:
                 if fm:
                     fence = fm.group("fence")
-                    if fence[0] == open_char and len(fence) >= open_len:
+                    info = fm.group("info").strip()
+                    if fence[0] == open_char and len(fence) >= open_len and not info:
                         inside = False
                 result.append(" " * len(line))
         return "\n".join(result)
diff --git a/tests/test_fix.py b/tests/test_fix.py
index d1bd15f..2b3afac 100644
--- a/tests/test_fix.py
+++ b/tests/test_fix.py
@@ -80,3 +80,54 @@ def test_formatted_empty_link_validation_and_mutation() -> None:
 
         serialized = serialize(new_ast)
         assert serialized == "[MISSING LINK LABEL](url)", f"Got: {serialized}"
+
+
+def test_polyglot_extractor_comment_masking() -> None:
+    """An HTML tag or a Z205 scheme inside an HTML or MDX comment is ignored by PolyglotExtractor."""
+    from zenzic.core.validator import PolyglotExtractor
+
+    extractor = PolyglotExtractor()
+
+    # 1. HTML comment with Z205 scheme
+    text = "<!-- <a href='javascript:alert(1)'>XSS</a> -->"
+    nodes = extractor.extract(text)
+    assert len(nodes) == 0, "Should ignore tags inside HTML comments"
+
+    # 2. MDX comment with Z205 scheme
+    text = "{/* <a href='javascript:alert(1)'>XSS</a> */}"
+    nodes = extractor.extract(text)
+    assert len(nodes) == 0, "Should ignore tags inside MDX comments"
+
+    # 3. HTML tag outside comments but comments present
+    text = "<!-- comment -->\n<a href='safe.md'>link</a>\n{/* comment */}"
+    nodes = extractor.extract(text)
+    assert len(nodes) == 1
+    assert nodes[0].href == "safe.md"
+    assert nodes[0].line_no == 2
+
+
+def test_polyglot_extractor_fence_evasion() -> None:
+    """A closing fence with trailing characters is NOT recognized as a closing fence, keeping masking active."""
+    from zenzic.core.validator import PolyglotExtractor
+
+    extractor = PolyglotExtractor()
+
+    # A fence is opened, then closed with ```extra. It should NOT be closed.
+    # Therefore, the tag <a href="safe.md"> inside/after it should remain masked.
+    text = (
+        "```\n<a href='safe.md'>inside</a>\n```extra\n<a href='safe.md'>after-fake-close</a>\n```\n"
+    )
+    nodes = extractor.extract(text)
+    # The first tag is inside the block. The second tag is also inside the block because ```extra didn't close it.
+    # The last ``` finally closes the block.
+    # So both should be masked, resulting in 0 extracted nodes.
+    assert len(nodes) == 0, (
+        "Should treat both tags as inside the code block because of malformed closing fence"
+    )
+
+    # If it is closed correctly (without extra info), the tag after should be extracted
+    text_closed = "```\n<a href='safe.md'>inside</a>\n```\n<a href='safe.md'>after-real-close</a>\n"
+    nodes_closed = extractor.extract(text_closed)
+    assert len(nodes_closed) == 1
+    assert nodes_closed[0].href == "safe.md"
+    assert nodes_closed[0].line_no == 4
diff --git a/uv.lock b/uv.lock
index fcc5c40..a27d8cd 100644
--- a/uv.lock
+++ b/uv.lock
@@ -2465,7 +2465,7 @@ wheels = [
 
 [[package]]
 name = "zenzic"
-version = "0.19.3"
+version = "0.19.4"
 source = { editable = "." }
 dependencies = [
     { name = "google-re2" },