Fix encoding issue when manually loading templates for forms

We used to manually load the template files for the edit pages using
turbogears.meta.load_kid_template(). Unfortunately this went through
the one code path where encoding was completely ignored. It ended up
defaulting to sys.getdefaultencoding() which is 'ascii'. So even though
most of the templates are loaded as 'utf-8' the few that really mattered
weren't.

The fix is to call kid.load_template() ourselves and set the encoding of
the class we just loaded to either the setting in the app.cfg file or
to the normal default value of 'utf-8'.

454076

1 files changed, 26 insertions, 0 deletions

diff --git a/ipa-server/ipa-gui/ipagui/helpers/ipahelper.py b/ipa-server/ipa-gui/ipagui/helpers/ipahelper.py
index 4eb7644c..9b340483 100644
--- a/ipa-server/ipa-gui/ipagui/helpers/ipahelper.py
+++ b/ipa-server/ipa-gui/ipagui/helpers/ipahelper.py
@@ -17,6 +17,10 @@
 
 import re
 import logging
+import turbogears
+import kid
+from turbokid import kidsupport
+from pkg_resources import resource_filename
 
 def javascript_string_escape(input):
     """Escapes the ' " and \ characters in a string so
@@ -60,3 +64,25 @@ def fix_incoming_fields(fields, fieldname, multifieldname):
         logging.warn("fix_incoming_fields error: " + str(e))
 
     return fields
+
+def load_template(classname, encoding=None):
+    """
+    Loads the given template. This only handles .kid files.
+    Returns a tuple (compiled_tmpl, None) to emulate
+    turbogears.meta.load_kid_template() which ends up not properly handling
+    encoding.
+    """
+    if not encoding:
+        encoding = turbogears.config.get('kid.encoding', kidsupport.KidSupport.assume_encoding)
+    divider = classname.rfind(".")
+    package, basename = classname[:divider], classname[divider+1:]
+    file_path = resource_filename(package, basename + ".kid")
+
+    tclass = kid.load_template(
+        file_path,
+        name = classname,
+        ).Template
+    tclass.serializer = kid.HTMLSerializer(encoding=encoding)
+    tclass.assume_encoding=encoding
+
+    return (tclass, None)