From d08fddeb2713286b208115b8ca601906aa1dd092 Mon Sep 17 00:00:00 2001
From: Aliasghar Jawadwala <sharksurfauto@gmail.com>
Date: Mon, 26 Jan 2026 16:57:35 +0530
Subject: [PATCH] Add YAML frontmatter syntax validation for license data

---
 src/licensedcode/frontmatter.py           |  2 +-
 src/licensedcode/models.py                |  4 +--
 tests/licensedcode/test_license_models.py | 32 +++++++++++++++++++++++
 3 files changed, 35 insertions(+), 3 deletions(-)

diff --git a/src/licensedcode/frontmatter.py b/src/licensedcode/frontmatter.py
index 2e3f6ee9971..05c60d12ed3 100644
--- a/src/licensedcode/frontmatter.py
+++ b/src/licensedcode/frontmatter.py
@@ -135,7 +135,7 @@ def load_frontmatter(fd, encoding="utf-8", **defaults):
         text = fd.read()
 
     else:
-        with codecs.open(fd, "r", encoding) as f:
+        with open(fd, "r", encoding=encoding) as f:
             text = f.read()
 
     text = return_unicode(text, encoding)
diff --git a/src/licensedcode/models.py b/src/licensedcode/models.py
index 354d93f52d3..5b84230cd0c 100644
--- a/src/licensedcode/models.py
+++ b/src/licensedcode/models.py
@@ -498,7 +498,7 @@ def dump(self, licenses_data_dir):
         content = get_yaml_safe_text(content)
         output = dumps_frontmatter(content=content, metadata=metadata)
         license_file = self.license_file(licenses_data_dir=licenses_data_dir)
-        with open(license_file, 'w') as of:
+        with open(license_file, 'w', encoding='utf-8') as of:
             of.write(output)
 
     def load(self, license_file, check_consistency=True):
@@ -2418,7 +2418,7 @@ def dump(self, rules_data_dir, **kwargs):
             metadata.update(kwargs)
         content = self.text
         output = dumps_frontmatter(content=content, metadata=metadata)
-        with open(rule_file, 'w') as of:
+        with open(rule_file, 'w', encoding='utf-8') as of:
             of.write(output)
 
     def load(self, rule_file, with_checks=True):
diff --git a/tests/licensedcode/test_license_models.py b/tests/licensedcode/test_license_models.py
index 6c47d92a594..b407a194532 100644
--- a/tests/licensedcode/test_license_models.py
+++ b/tests/licensedcode/test_license_models.py
@@ -660,3 +660,35 @@ def test_get_key_phrases_ignores_nested_key_phrase_markup(self):
             raise Exception('Exception should be raised')
         except InvalidRuleRequiredPhrase:
             pass
+
+
+class TestLicenseYamlFrontmatterSyntax(FileBasedTesting):
+    """
+    Validate that all license data files have valid YAML syntax.
+    See: https://github.com/aboutcode-org/scancode-toolkit/issues/3947
+    """
+    test_data_dir = TEST_DATA_DIR
+
+    def test_license_yaml_frontmatter_integrity(self):
+        """
+        Ensure all .LICENSE files in licenses_data_dir have valid YAML syntax
+        in their frontmatter section.
+        """
+        from pathlib import Path
+        from licensedcode.frontmatter import load_frontmatter
+        from licensedcode.models import licenses_data_dir
+
+        licenses_path = Path(licenses_data_dir)
+        errors = []
+
+        for license_file in sorted(licenses_path.glob('*.LICENSE')):
+            try:
+                load_frontmatter(str(license_file))
+            except Exception as e:
+                errors.append(f'{license_file.name}: {e}')
+
+        if errors:
+            error_msg = '\n'.join(errors[:20])  # Show first 20 errors
+            if len(errors) > 20:
+                error_msg += f'\n... and {len(errors) - 20} more errors'
+            assert False, f'Invalid YAML in {len(errors)} license files:\n{error_msg}'