_(): Encode any Unicode values in the interpolation dictionary back

into encoded 8-bit strings.  Returning Unicode from here caused too
much breakage, so we go in the other direction.  Eventually, Mailman
should use Unicode for everything (maybe).

This should fix (one of) Peer's problems, and maybe the one recently
reported by Dan Buchmann.

1 files changed, 15 insertions, 4 deletions

diff --git a/Mailman/i18n.py b/Mailman/i18n.py
index d38eba85b..c58534384 100644
--- a/Mailman/i18n.py
+++ b/Mailman/i18n.py
@@ -1,4 +1,4 @@
-# Copyright (C) 2000,2001,2002 by the Free Software Foundation, Inc.
+# Copyright (C) 2000-2003 by the Free Software Foundation, Inc.
 #
 # This program is free software; you can redistribute it and/or
 # modify it under the terms of the GNU General Public License
@@ -17,7 +17,7 @@
 import sys
 import time
 import gettext
-from types import StringType
+from types import StringType, UnicodeType
 
 from Mailman import mm_cfg
 from Mailman.SafeDict import SafeDict
@@ -74,8 +74,19 @@ def _(s):
     # missing key in the dictionary.
     dict = SafeDict(frame.f_globals.copy())
     dict.update(frame.f_locals)
-    # Translate the string, then interpolate into it.
-    return _translation.gettext(s) % dict
+    # Translating the string returns an encoded 8-bit string.  Rather than
+    # turn that into a Unicode, we turn any Unicodes in the dictionary values
+    # into encoded 8-bit strings.  BAW: Returning a Unicode here broke too
+    # much other stuff and _() has many tentacles.  Eventually I think we want
+    # to use Unicode everywhere.
+    tns = _translation.gettext(s)
+    charset = _translation.charset()
+    if not charset:
+        charset = 'us-ascii'
+    for k, v in dict.items():
+        if isinstance(v, UnicodeType):
+            dict[k] = v.encode(charset, 'replace')
+    return tns % dict