… if LDAPObject._trace_level is non-zero and Python is running in debug mode

…RL into a list to be displayed with self.fail()

…s() for now

Testing of attributetypes and objectclasses parsing from file

Fixes: #79

…n module unittest

Sync with upstream 2.4.35

Those were introduced in 3c46166; however, we strive to be absolutely
explicit on the return type of our functions:

- with `bytes_mode=True` (legacy behavior), we use bytes (Py2 'str')
  on every Distinguished Name like attributes
- with `bytes_mode=False` or under Py3, we always use text (Py2
  'unicode' or Py3 'str') for those attributes

=> `ldap.search_subschemasubentry_s` should return values according to
   those rules.

This is a backport from python-ldap 2.4.37

Fixes: #90

This is a backport from python-ldap 2.4.39+

Fixes: #90

…troduced in 2.4.35

#94

_unbytesify_value is a no-op in Python 3, so add a new function that actually
decodes bytes to text when bytes_mode is set to False, reagrdless of Python
version.

To match the other bytesify functions, the type assertion is only run under
Python 2.

References #39, #54

Rework how pyldap handles converting Unicode to UTF-8 encoded byte-strings to
avoid inconsistencies between Python 2 and Python 3 in how Python's C API
handles the "s" format unit. The inconsistencies result in Python 2 raising
UnicodeEncodeErrors when Unicode strings with non-ASCII characters are passed
to the C bindings.

- In Python 3, the "s" format expects the argument to be a unicode string, and
  it encodes it using UTF-8 (unconditionally).
- In Python 2, the "s" format tries to encode unicode strings using the
  default encoding (sys.getdefaultencoding(), which defaults to 'ascii').

Avoid these inconsistencies (while preserving the warnings/error handling) by
explicitly encoding values to UTF-8 bytes, instead of passing passing unicode
objects to C. Doing this addresses the issue, while still relying on Python
for GC/memory management. The alternatives (e.g. using the "es" format unit)
would require more invasive changes in the C bindings, including explicit
memory management.

Also update `attrs_from_List` to align with these changes:

- In Python 2, assume a sequence of bytes objects (encoded to UTF-8)
- In Python 3, assume a sequence of string (Unicode) objects and rely on
  Python to encode to UTF-8 and handle the memory allocations.

This simplification of attrs_from_List also fixes a leak of the bytes objects;
previously, the new object returned by PyUnicode_AsUTF8String was leaked.

Fixes #4
Fixes #39
Fixes #54

The original comparison made python3 sad with
"BytesWarning: Comparison between bytes and string"
warning.

PEP8-styling accidentally snuck into parts of _unbytesify_result.  Also
comment a commented-out bind test.

Remove usage of dict.keys().pop().

Closes #84.

Improve unicode handling in Python 2

- Improve docstrings
- Rename helpers to clarify their role
- Merge _bytesify_value and _unbytesify_encoded_text

A call to 'result4' returns properly typed entries for Py3, we have to
convert it back to bytes only under Py2 bytes_mode.

Converting from bytes to text should only occur when decoding attribute
*values*.

#100

#81