Fix possible UnicodeDecodeError in warning when msgid already exists.

This happens when comparing 'Read more…' with 'Read more…'.
That should not give a warning, because they are really the same, but that would need a fix somewhere else,
in the code that gets the value of an i18n:attribute in a different way than an i18n:translate.

Added test for the proper printing of the warning.

3 files changed, 32 insertions, 3 deletions

diff --git a/CHANGES.rst b/CHANGES.rst
index 6d53682..17af10e 100644
--- a/CHANGES.rst
+++ b/CHANGES.rst
@@ -5,7 +5,7 @@ CHANGES
 4.0.0a2 (unreleased)
 --------------------
 
-- Nothing changed yet.
+- Fix possible UnicodeDecodeError in warning when msgid already exists.
 
 
 4.0.0a1 (2013-02-15)
diff --git a/src/zope/tal/talgettext.py b/src/zope/tal/talgettext.py
index 80859e8..233751e 100644
--- a/src/zope/tal/talgettext.py
+++ b/src/zope/tal/talgettext.py
@@ -136,8 +136,8 @@ class POEngine(DummyEngine):
                 references = '\n'.join([location[0]+':'+str(location[1]) for location in domain[msgid]])
                 print(("Warning: msgid '%s' in %s already exists "
                        "with a different default (bad: %s, should be: %s)\n"
-                       "The references for the existent value are:\n%s\n" %
-                         (msgid, self.file+':'+str(position), msgid.default.encode('utf-8'), existing_msgid.default.encode('utf-8'), references)).encode('utf-8'), file=sys.stderr)
+                       "The references for the existent value are:\n%s\n".encode('utf-8') %
+                         (msgid.encode('utf-8'), self.file+':'+str(position), msgid.default.encode('utf-8'), existing_msgid.default.encode('utf-8'), references)), file=sys.stderr)
         domain[msgid].append((self.file, position))
         return 'x'
 
diff --git a/src/zope/tal/tests/test_talgettext.py b/src/zope/tal/tests/test_talgettext.py
index d4761ff..3799297 100644
--- a/src/zope/tal/tests/test_talgettext.py
+++ b/src/zope/tal/tests/test_talgettext.py
@@ -46,6 +46,35 @@ class test_POEngine(unittest.TestCase):
                         "POEngine catalog does not properly store message ids"
                         )
 
+    def test_translate_existing(self):
+        engine = POEngine()
+        # This tries to reproduce a big surfacing in a template of
+        # PloneSoftwareCenter when using the i18ndude package to
+        # extract translatable strings, which uses zope.tal.  The
+        # relevant html snippet is this:
+        #
+        # <a href="#" title="Read more&hellip;"
+        #    i18n:attributes="title label_read_more"
+        #    tal:attributes="href release/absolute_url">
+        #    <span i18n:translate="label_read_more">Read more&hellip;</span>
+        # </a>
+        #
+        # Due to the different ways that i18n:attributes and
+        # i18n:translate are handled, the attribute gets passed to the
+        # translate method with the html entity interpreted as a
+        # unicode, and the i18n:translate gets passed as a simple
+        # string with the html entity intact.  That may need a fix
+        # elsewhere, but at the moment it gives a warning.  The very
+        # least we can do is make sure that this does not give a
+        # UnicodeDecodeError, which is what we test here.
+        engine.file = 'psc_release_listing.pt'
+        # position is position in file.
+        engine.translate('foo', 'domain', default=u'Read more\u2026', position=7)
+        # Adding the same key with the same default is fine.
+        engine.translate('foo', 'domain', default=u'Read more\u2026', position=13)
+        # Adding the same key with a different default is bad and triggers a warning.
+        engine.translate('foo', 'domain', default='Read still more&hellip;', position=42)
+
     def test_dynamic_msgids(self):
         sample_source = """
             <p i18n:translate="">