bugdown: Hide alphanumeric characters in exception logs, for privacy

(imported from commit 39481494b7910307f56e566035c1b464c83d196e)

zephyr/lib/bugdown/__init__.py

@@ -1,6 +1,7 @@
 import markdown
 import logging
 import traceback
+import re
 
 from zephyr.lib.avatar  import gravatar_hash
 from zephyr.lib.bugdown import codehilite
@@ -42,6 +43,16 @@ MAX_MD_ENGINE_USES = 30
 _md_engine = None
 _use_count = 0
 
+# We want to log Markdown parser failures, but shouldn't log the actual input
+# message for privacy reasons.  The compromise is to replace all alphanumeric
+# characters with 'x'.
+#
+# We also use repr() to improve reproducibility, and to escape terminal control
+# codes, which can do surprisingly nasty things.
+_privacy_re = re.compile(r'\w', flags=re.UNICODE)
+def _sanitize_for_log(md):
+    return repr(_privacy_re.sub('x', md))
+
 def _linkify(match):
     url = match.group('url')
     return ' [%s](%s) ' % (url, url)
@@ -62,14 +73,9 @@ def convert(md):
         html = _md_engine.convert(md)
     except:
         # FIXME: Do something more reasonable here!
-        #
-        # NB: For security, we must not print the bare Markdown input.
-        # It could contain terminal control codes, which can do
-        # surprisingly nasty things.
-
         html = '<p>[Humbug note: Sorry, we could not understand the formatting of your message]</p>'
-        logging.getLogger('').error('Exception in Markdown parser: %sInput was: %s'
-            % (traceback.format_exc(), repr(md)))
+        logging.getLogger('').error('Exception in Markdown parser: %sInput (sanitized) was: %s'
+            % (traceback.format_exc(), _sanitize_for_log(md)))
 
     _use_count += 1
     if _use_count >= MAX_MD_ENGINE_USES: