Scoped repository with non-ASCII path name errors

[richard.liao.i@…]

When I config trac using scoped repository, which path name contains non-ascii characters, will raise UnicodeDecodeError in trac-admin resync and browser:

UnicodeDecodeError: 'ascii' codec can't decode byte 0xe9 in position
8: ordinal not in range(128)

Test in trac 0.10.4, 0.11 and 0.12, all have the same problem.

Here is the patch:

Index: svn_fs.py
===================================================================
@@ -111,7 +111,7 @@
    """Check whether the given `fullpath` is within the given
`scope`"""
    if scope == '/':
        return fullpath is not None
-    fullpath = fullpath and fullpath.lstrip('/') or ''
+    fullpath = to_unicode(fullpath and fullpath.lstrip('/') or '')
    scope = scope.strip('/')
    return (fullpath + '/').startswith(scope + '/')

@@ -404,7 +404,7 @@
        Repository.__init__(self, name, authz, log)

        if self.path != path:
-            self.scope = path[len(self.path):]
+            self.scope = to_unicode(path[len(self.path):])
            if not self.scope[-1] == '/':
                self.scope += '/'
        else:

[Christian Boos]

Well, yes, I see the bug: self.scope is sometimes supposed to be utf-8 str, sometimes unicode. We need to clarify what is supposed to be kept as utf-8 and what not, in a more explicit way.

([OT] and yes, to add some weight to #7018, I just found myself entering "unicode svn" in the Cc: field :-) )

[Christian Boos]

Was incidentally fixed during the r7842 clean-up.

[Christian Boos]

Needed follow-up fix r7884 and modified the tests so that they take care of this condition as well (r7885).