Issue 5408: test_osx_env failing - Python tracker

Mac OS 10.4 PPC debug build

test_osx_env
[31492 refs]
Could not find platform independent libraries <prefix>
Could not find platform dependent libraries <exec_prefix>
Consider setting $PYTHONHOME to <prefix>[:<exec_prefix>]
Fatal Python error: Py_Initialize: can't initialize sys standard streams
ImportError: No module named encodings.utf_8
test test_osx_env failed -- Traceback (most recent call last):
  File "/temp/python/py3k/Lib/test/test_osx_env.py", line 27, in
test_pythonexecutable_sets_sys_executable
    self._check_sys('PYTHONEXECUTABLE', '==', 'sys.executable')
  File "/temp/python/py3k/Lib/test/test_osx_env.py", line 24, in _check_sys
    self.assertEqual(rc, 2, "expected %s %s %s" % (ev, cond, sv))
AssertionError: expected PYTHONEXECUTABLE == sys.executable

Is this a standard unix build or a framework build?

It appears the test doesn't work correctly for non-framework builds, 
something I didn't test.  A patch is forthcoming.

ANALYSIS    In a non-framework build, when PYTHONEXECUTABLE is set to a
            invalid value, getpath.c is unable to determine appropriate
            values for sys.prefix and sys.path so the interpreter fails
            in initialization.

SOLUTION    Simplify the test and supply appropriate values so the test
            works for both framework and non-framework builds.

APPLIES     py3k, 3.0

Ned: IMHO your patch is not correct. test_osx_env tests behaviour that's 
only valid for a framework build, and should therefore only run when 
testing a framework build.

The easiest way to accomplish that is to change the test in 
test_osx_env.main to:

   if sys.platform == 'darwin' and 
sysconfig.get_config_var('WITH_NEXT_FRAMEWORK'):

(with an "from distutils import sysconfig" statement just above the 
test).

I will commit this fix when I've actually tested it.

Committed my fix as r70179 (3.x) and r70180 (30-maint).

Benjamin: can you confirm this actually fixes the issue with a non-
framework build, I don't know if I'll be able to test that before the 
first 3.1a release.

I don't have strong feelings about it one way or the other.  As far as I 
can tell from the code and testing and the documented behavior, the effect 
on sys.executable is independent of whether it is a framework build or 
not.  But I'm fine with narrowing the scope of the test.

The effect on sys.executable is always present, the test is only valid for 
a framework build because that's the only build that can find sys.prefix 
without looking at sys.executable.  Setting PYTHONEXECUTABLE with a non-
framework build is only valid if the value of the variable point's to a 
python executable in the right location within the installation tree.

Thumbs up here! Thanks.