Very old files (Python <= 2.2) are stored in latin-1 with a correct Content-Type.

JulienPalard · JulienPalard · commit ae2a8fc29c1f · 2022-05-25T00:10:20.000+02:00
diff --git a/build_docs.py b/build_docs.py
@@ -980,15 +980,15 @@ def proofread_canonicals(www_root: Path) -> None:
         """<link rel="canonical" href="https://docs.python.org/([^"]*)" />"""
     )
     for file in www_root.glob("**/*.html"):
-        html = file.read_text(encoding="UTF-8")
+        html = file.read_text(encoding="UTF-8", errors="surrogateescape")
         canonical = canonical_re.search(html)
         if not canonical:
             continue
         target = canonical.group(1)
         if not (www_root / target).exists():
             logging.info("Removing broken canonical from %s to %s", file, target)
             html = html.replace(canonical.group(0), "")
-            file.write_text(html, encoding="UTF-8")
+            file.write_text(html, encoding="UTF-8", errors="surrogateescape")
 
 
 def main():
