enumerateUsers fails with unicode arguments

[Willi Langenberger]

The enumerateUsers method of ZODBUserManager plugin does not
handle unicode arguments right. In the following example (taken
and modified from test_ZODBUserManager.py) we add four users
('foo', 'bar', 'baz', 'bam' ) and then search for u'abc'. I would have
expected that nothing is found. Instead, we get three matches:

  ID_LIST = ( 'foo', 'bar', 'baz', 'bam' )
  for id in ID_LIST:
      zum.addUser( id, '%<email address hidden>' % id, 'password' )
  info_list = zum.enumerateUsers( id = u'abc',
                  exact_match=False )

=> 'bar', 'baz', 'bam'

Reason:

class ZODBUserManager(...):

  def enumerateUsers(self, id, login, ...):
        ....
        if isinstance( id, str ):
            id = [ id ]
        if isinstance( login, str ):
            login = [ login ]

Because u'abc' is no str object, it is *not* transformed to a list.
Thus, later on we search for: "a", "b" and "c".

I dont know what is the best fix for this. Either we can
assume a standard encoding of the user names, then we can
encode to that string. Or we log an error, eg "string argument
required". But i think it is wrong to enumerate 'bar', 'baz' and
'bam' if we look for u'abc'.

[Willi Langenberger (wlang)]

attached is a patch which includes the new unicode test. we expect zero matches but get three ones.

[Tres Seaver (tseaver)]

The attached patch makes your test pass.

[Tres Seaver (tseaver)]

Thanks for the report, and the test. I have checked in the test and a fix
on the 1.4 branch, the 1.5 branch, and the trunk. Both are included in
the 1.5.3 release.