With python 2 it was possible to do exception[0][...] to recover
details about an exception.
It's no longer authorized with python 3.
Now, we must do something like exception.args or exception.urls.
fortunately this syntax also works with python 2.
So we use it for both.
python-ldap talks in bytes,
as the rest of ldapcherry talks in unicode utf-8:
* everything passed to python-ldap must be converted to bytes
* everything coming from python-ldap must be converted to unicode
The previous statement was true for python-ldap < version 3.X.
With versions > 3.0.0 and python 3, it gets tricky,
some parts of python-ldap takes string, specially the filters/escaper.
so we have now:
*_byte_p2 (unicode -> bytes conversion for python 2)
*_byte_p3 (unicode -> bytes conversion for python 3)
*_byte_p23 (unicode -> bytes conversion for python AND 3)
* python-ldap 3 is slightly different than 2 on how it handles modify
the modified attributes used to be transmitted as a dict, now it should
be transmitted as a list of dict)
* changes in urllib imports since quote_plus in urllib with python 2 and
in urllib.parse in python 3
* changes in imports for Sets since set is a native type in python 3 and
doesn't requires an import
* fix in __import__, '-1' level for module path discovery is not supported
anymore, switching to 0 (absolute import only).
Before, no particular treatment was done on the user attributes.
This caused some issues because python-ldap systematically returns
the attribute value as a list (even if it's mono-valuated).
Now we recover the attributes used in the group attr templates,
and we "normalize" the user attributes before using it in add_to_groups
and del_from_groups.
By normalize, we mean, transforming the list to it's unique value.
In case the attribute doesn't exist or is multi-valuated, it raises an
error.
In some forms, the 'cn' attribute might not be accessible.
The _set_password method relied on 'cn' to build the user dn.
Now it accepts the cn or the dn (by_cn switch).
* add display_name parameter for backend
* handle display_name in modify template (lonely groups) and error
message
* fix bug in modify when user is in only one backend
* remove redondant exception in ad/ldap backend
* fix unit test accordingly
* fix License in AD backend (GPLv3 -> MIT, previously tought of using
samba4 python library which is GPL, but this module finaly uses
python-ldap which is MIT compatible)
re-raising the exception without specifying the exception object gives
exactly the previous exception (including the correct stacktrace)
which is nicer to have for debugging.
