lxml-3.5.0 on NetBSD 7.99.22 with python-3.4.3 in the default C locale fails two tests:
Doctest: xpathxslt.txt
======================================================================
ERROR: test_etree_parse_io_error (lxml.tests.test_io.ETreeIOTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
File "/usr/pkg/lib/python3.4/unittest/case.py", line 58, in testPartExecutor
yield
File "/usr/pkg/lib/python3.4/unittest/case.py", line 577, in run
testMethod()
File "/disk/3/archive/obj/textproc/py-lxml/work/lxml-3.5.0/src/lxml/tests/test_io.py", line 276, in test_etree_parse_io_error
dn = tempfile.mkdtemp(prefix=dirnameRU)
File "/usr/pkg/lib/python3.4/tempfile.py", line 295, in mkdtemp
_os.mkdir(file, 0o700)
UnicodeEncodeError: 'ascii' codec can't encode characters in position 5-18: ordinal not in range(128)
======================================================================
ERROR: test_etree_parse_io_error (lxml.tests.test_io.ElementTreeIOTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
File "/usr/pkg/lib/python3.4/unittest/case.py", line 58, in testPartExecutor
yield
File "/usr/pkg/lib/python3.4/unittest/case.py", line 577, in run
testMethod()
File "/disk/3/archive/obj/textproc/py-lxml/work/lxml-3.5.0/src/lxml/tests/test_io.py", line 276, in test_etree_parse_io_error
dn = tempfile.mkdtemp(prefix=dirnameRU)
File "/usr/pkg/lib/python3.4/tempfile.py", line 295, in mkdtemp
_os.mkdir(file, 0o700)
UnicodeEncodeError: 'ascii' codec can't encode characters in position 5-18: ordinal not in range(128)
----------------------------------------------------------------------
Ran 1735 tests in 24.823s
````
If I understand the test correctly, it tries to convert a Russian string to an ASCII string (since the default C locale on NetBSD is ASCII) and fails, and that makes the test fail. In my understanding the test works on Linux because the default C locale is UTF-8 there and the conversion works.
I don't know what the test really wants to test. Please change this test to be more portable.
Having debugged a similar problem for cookiecutter, I'm pretty sure that this problem will also appear on Linux when you unset LANG* and LC_*, since the Linux C locale is ASCII too.