From 177d2071e58740255273bafb6d2f07df3dd93ac9 Mon Sep 17 00:00:00 2001
From: safishamsi <safishamsi98@gmail.com>
Date: Sat, 10 Oct 2026 21:20:29 +0100
Subject: [PATCH] fix(html): stop a node label from breaking out of the
 embedded script block (#4124)

Node data is embedded as JSON inside a <script> element, but the serializers
escaped only `</`. An unclosed `<!--` in one label followed by a later `<script`
drove the HTML tokenizer into script-data-double-escaped state, where the real
</script> no longer closed the tag and graph.html rendered blank. Escape every
`<` as < in both exporters/html.py and tree_html.py; labels still render
because `<` round-trips through JSON.parse.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---
 graphify/exporters/html.py         |  8 +++-
 graphify/tree_html.py              |  7 +--
 tests/test_html_script_breakout.py | 71 ++++++++++++++++++++++++++++++
 3 files changed, 81 insertions(+), 5 deletions(-)
 create mode 100644 tests/test_html_script_breakout.py

diff --git a/graphify/exporters/html.py b/graphify/exporters/html.py
index 14f781249..1fbe8eb10 100644
--- a/graphify/exporters/html.py
+++ b/graphify/exporters/html.py
@@ -613,9 +613,13 @@ def to_html(
         n = member_counts.get(cid, len(communities.get(cid, []))) if member_counts else len(communities.get(cid, []))
         legend_data.append({"cid": cid, "color": color, "label": lbl, "count": n})
 
-    # Escape </script> sequences so embedded JSON cannot break out of the script tag
+    # Escape every `<` as < so embedded JSON cannot break out of the <script>
+    # element. Escaping only `</` is not enough: an unclosed `<!--` then a later
+    # `<script` in a label drive the HTML tokenizer into script-data-double-escaped
+    # state, where the real `</script>` no longer closes the tag and the page goes
+    # blank (#4124). `<` round-trips through JSON.parse, so labels render unchanged.
     def _js_safe(obj) -> str:
-        return json.dumps(obj).replace("</", "<\\/")
+        return json.dumps(obj).replace("<", "\\u003c")
 
     nodes_json = _js_safe(vis_nodes)
     edges_json = _js_safe(vis_edges)
diff --git a/graphify/tree_html.py b/graphify/tree_html.py
index dfb41343a..77630868d 100644
--- a/graphify/tree_html.py
+++ b/graphify/tree_html.py
@@ -568,9 +568,10 @@ def emit_html(
     svg_width: int = 6000,
     svg_height: int = 8000,
 ) -> str:
-    # Escape </script> sequences so embedded JSON cannot break out of the
-    # <script> tag, and HTML-escape values that land in <title>/<h1>.
-    data_json = json.dumps(tree, ensure_ascii=True, separators=(",", ":")).replace("</", "<\\/")
+    # Escape every `<` as < so embedded JSON cannot break out of the <script>
+    # element (escaping only `</` leaves the `<!--` + `<script` breakout open, #4124),
+    # and HTML-escape values that land in <title>/<h1>.
+    data_json = json.dumps(tree, ensure_ascii=True, separators=(",", ":")).replace("<", "\\u003c")
     return _HTML_TEMPLATE.format(
         title=_html.escape(title),
         header=_html.escape(header),
diff --git a/tests/test_html_script_breakout.py b/tests/test_html_script_breakout.py
new file mode 100644
index 000000000..5bee271eb
--- /dev/null
+++ b/tests/test_html_script_breakout.py
@@ -0,0 +1,71 @@
+"""A node label can never break out of the embedded <script> block (#4124).
+
+Node data is serialized to JSON and embedded inside a <script> element. Escaping
+only `</` is not enough: an unclosed `<!--` in one label followed by a later
+`<script` drives the HTML tokenizer into script-data-double-escaped state, where
+the real `</script>` no longer closes the tag — the rest of the page is swallowed
+and graph.html renders blank. Escaping every `<` as \\u003c closes all three
+breakout sequences (`<!--`, `<script`, `</script`) while labels still render,
+since `<` round-trips through JSON.parse.
+"""
+from __future__ import annotations
+
+import json
+
+import networkx as nx
+
+from graphify.exporters.html import to_html
+from graphify.tree_html import emit_html
+
+DANGEROUS = [
+    "foo <!-- bar components the panel",   # unclosed comment opener
+    "code:vue (<script setup>)",           # script opener
+    "end </script> tail",                  # explicit closer
+]
+
+
+def _no_breakout(html: str) -> None:
+    # No label-originated raw breakout sequence survives in the document.
+    assert "<!--" not in html
+    assert "<script setup" not in html
+    assert "</script> tail" not in html
+    # The escaped form is what carries the label through instead.
+    assert "\\u003c" in html
+
+
+def test_to_html_labels_cannot_break_out_of_script(tmp_path) -> None:
+    G = nx.Graph()
+    for i, lbl in enumerate(DANGEROUS):
+        G.add_node(f"n{i}", label=lbl, file_type="code", source_file=f"f{i}.py")
+    G.add_edge("n0", "n1", relation="imports")
+    out = tmp_path / "graph.html"
+    assert to_html(G, {0: list(G.nodes)}, str(out)) is True
+    html = out.read_text(encoding="utf-8")
+    _no_breakout(html)
+    # Exactly the page's own <script> elements survive — no label minted an extra one.
+    assert html.count("<script") == html.count("</script")
+
+
+def test_tree_html_labels_cannot_break_out_of_script() -> None:
+    tree = {
+        "name": "root",
+        "children": [{"name": lbl, "children": []} for lbl in DANGEROUS],
+    }
+    html = emit_html(tree, title="t", header="h")
+    _no_breakout(html)
+
+
+def test_clean_labels_round_trip_unchanged(tmp_path) -> None:
+    """A label with no `<` is untouched, and a `<`-bearing label still decodes
+    back to the original text."""
+    G = nx.Graph()
+    G.add_node("a", label="plain_label", file_type="code", source_file="a.py")
+    G.add_node("b", label="Vector<T>", file_type="code", source_file="b.ts")
+    G.add_edge("a", "b", relation="imports")
+    out = tmp_path / "graph.html"
+    to_html(G, {0: ["a", "b"]}, str(out))
+    html = out.read_text(encoding="utf-8")
+    assert "plain_label" in html
+    # the escaped generic type decodes back to the original
+    assert json.loads('"Vector\\u003cT>"') == "Vector<T>"
+    assert "Vector\\u003cT>" in html