fix another FreeBSD $HOME symlink issue

[ivanov]

without these changes, I get this error in IPython.core on FreeBSD 9.0

FAIL: test_cwd_x (IPython.core.tests.test_prompts.PromptTests)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/home/pi/code/ipython/IPython/core/tests/test_prompts.py", line 107, in test_cwd_x
    self.assertEquals(p, '~')
AssertionError: u'/usr~' != '~'
    "u'/usr~' != '~'" = '%s != %s' % (safe_repr(u'/usr~'), safe_repr('~'))
    "u'/usr~' != '~'" = self._formatMessage("u'/usr~' != '~'", "u'/usr~' != '~'")
>>  raise self.failureException("u'/usr~' != '~'")

I've verified that this change does not break any tests under Ubuntu when HOME happens to be symlinked.

I don't know how this will affect windows

[takluyver]

Isn't it simpler just to do HOME = os.path.realpath(HOME) unconditionally? If it wouldn't trigger this if, then doing it unconditionally would make no difference.

[minrk]

Why not use existing utils.path.get_home_dir, where you fixed this the first time?

[ivanov]

Thomas Kluyver, on 2012-06-16 05:32, wrote:

Isn't it simpler just to do HOME = os.path.realpath(HOME)
unconditionally? If it wouldn't trigger this if, then doing
it unconditionally would make no difference.

good point, just applied this change

Min RK, on 2012-06-16 21:23, wrote:

Why not use existing utils.path.get_home_dir, where you fixed
this the first time?

That function ends up returning a directory under windows, where
I guess $HOME isn't typically defined. It looks like in this
code, windows ends up getting a nonsensical HOME so that no
directory ends up getting replaced with ~.

Does that seem reasonable enough?

[minrk]

ok, makes sense. It seems like this can go in, then.

[takluyver]

Test results for commit d39ba53 merged into master
Platform: linux2

• python2.7: Failed, log at https://gist.github.com/2944217 (libraries not available: oct2py rpy2 tornado)
• python3.1: OK (libraries not available: cython matplotlib numpy oct2py pymongo qt rpy2 wx wx.aui zmq)
• python3.2: OK (libraries not available: cython oct2py pymongo rpy2 wx wx.aui)

Not available for testing: python2.6

[takluyver]

Test results for commit d39ba53 merged into master
Platform: linux2

• python2.7: OK (libraries not available: oct2py rpy2 tornado)
• python3.1: OK (libraries not available: cython matplotlib numpy oct2py pymongo qt rpy2 wx wx.aui zmq)
• python3.2: OK (libraries not available: cython oct2py pymongo rpy2 wx wx.aui)

Not available for testing: python2.6

[takluyver]

The failure the first time was a sporadic failure, not related to this. Merging now.

I think I see that failure once each time I turn my computer on. I must get round to debugging it.