Fixed issue with (multiple) BOM's in response.

tzuryby · Aug 4, 2010 · a546df7 · a546df7
1 parent ab9818a
commit a546df7
Show file tree

Hide file tree

Showing 3 changed files with 10 additions and 2 deletions.
diff --git a/README b/README
@@ -9,7 +9,7 @@ The code currently handles a couple of simple WSGI application. It also handles
 
 Known issues:
 - this wsgi handler currently doesn't support chunked responses. I haven't figured out yet how to do this.
-- Using the Django demo app I get some strange HTML parsing errors in Chrome. It says "Extra <html> encountered...". Which leads to a couple a follow up parsing errors. It apparently has something to do with unicode and a BOM being present in the response data. Strange.
 
 Fixed issues:
 - [FIXED 4 aug 2010] the handler does not handle query strings correctly right now. Still have to fix and test this.
+- [UGLY FIX 4 aug 2010 - Removing BOM(s) before returning the response] Using the Django demo app I get some strange HTML parsing errors in Chrome. It says "Extra <html> encountered...". Which leads to a couple a follow up parsing errors. It apparently has something to do with unicode and a BOM being present in the response data. Strange. Hmmm, the Django development server (./manage runserver) has the same problem.
diff --git a/test_wsgi_app.py b/test_wsgi_app.py
@@ -33,6 +33,7 @@
     __file__ = '?'
 
 html_template = u"""\
+<!DOCTYPE html>
 <html>
 <head>
  <title>WSGI Test Script</title>

diff --git a/wsgi-handler.py b/wsgi-handler.py
@@ -43,7 +43,7 @@ def wsgi_server(application):
     A WSGI application should return a iterable op StringTypes. 
     Any encoding must be handled by the WSGI application itself.
     '''
-
+    
     while True:
         if DEBUG: print "WAITING FOR REQUEST"
 
@@ -93,6 +93,13 @@ def wsgi_server(application):
         code = response[0][9:12]
         status = response[0][13:]
 
+        # strip BOM's from response data
+        # Especially the WSGI handler from Django seems to generate them (2 actually, huh?)
+        # a BOM isn't really necessary and cause HTML parsing errors in Chrome and Safari
+        # See also: http://www.xs4all.nl/~mechiel/projects/bomstrip/
+        # Although I still find this a ugly hack.
+        data = data.replace('\xef\xbb\xbf', '')
+
         # Get the generated errors
         errors = errIO.getvalue()