Merge pull request #2041 from edugonza/main

fixes #2040 Wrapped urlsplit call in a try-except block

tests/test_url.py

@@ -0,0 +1,32 @@
+"""Test URLs."""
+
+import re
+
+import pytest
+
+from .testing_utils import FakeHTML, capture_logs, resource_path
+
+
+@pytest.mark.parametrize('url, base_url', (
+    ('https://weasyprint.org]', resource_path('<inline HTML>')),
+    ('https://weasyprint.org]', 'https://weasyprint.org]'),
+    ('https://weasyprint.org/', 'https://weasyprint.org]'),
+))
+def test_malformed_url_link(url, base_url):
+    """Test malformed URLs."""
+    with capture_logs() as logs:
+        pdf = FakeHTML(
+            string=f'<p><a href="{url}">My Link</a></p>',
+            base_url=base_url).write_pdf()
+
+    assert len(logs) == 1
+    assert "Malformed" in logs[0]
+    assert "]" in logs[0]
+
+    uris = re.findall(b'/URI \\((.*)\\)', pdf)
+    types = re.findall(b'/S (/\\w*)', pdf)
+    subtypes = re.findall(b'/Subtype (/\\w*)', pdf)
+
+    assert uris.pop(0) == url.encode()
+    assert subtypes.pop(0) == b'/Link'
+    assert types.pop(0) == b'/URI'

weasyprint/urls.py

@@ -151,10 +151,19 @@ def get_link_attribute(element, attr_name, base_url):
     uri = get_url_attribute(element, attr_name, base_url, allow_relative=True)
     if uri:
         if base_url:
-            parsed = urlsplit(uri)
-            # Compare with fragments removed
-            if parsed.fragment and parsed[:-1] == urlsplit(base_url)[:-1]:
-                return ('url', ('internal', unquote(parsed.fragment)))
+            try:
+                parsed = urlsplit(uri)
+            except ValueError:
+                LOGGER.warning('Malformed URL: %s', uri)
+            else:
+                try:
+                    parsed_base = urlsplit(base_url)
+                except ValueError:
+                    LOGGER.warning('Malformed base URL: %s', base_url)
+                else:
+                    # Compare with fragments removed
+                    if parsed.fragment and parsed[:-1] == parsed_base[:-1]:
+                        return ('url', ('internal', unquote(parsed.fragment)))
         return ('url', ('external', uri))