test_magic.test_obj_del fails on win32

[ipython]

Original Launchpad bug 366334: https://bugs.launchpad.net/ipython/+bug/366334
Reported by: ellisonbg (Brian Granger).

test_obj_del fails on win32 because getoutput is called like this:

IP.getoutput('ipython obj_del.py')

But on Windows, the ipython executable is not found. This causes stdout to be empty
and stderr to have an error saying that ipython can't be found.

FAIL: Test that object's del methods are called on exit.

Traceback (most recent call last):
File "c:\python24\lib\site-packages\nose-0.10.4-py2.4.egg\nose\case.py", line 182, in runTest
self.test(*self.arg)
File "c:\python24\external\ipython\IPython\tests\test_magic.py", line 86, in test_obj_del
nt.assert_equals(out,'obj_del.py: object A deleted')
AssertionError: '' != 'obj_del.py: object A deleted'

raise self.failureException,
(None or '%r != %r' % ('', 'obj_del.py: object A deleted'))

---

[ipython]

[ LP comment 1 by: Brian Granger, on 2009-04-24 22:56:21.616906+00:00 ]

We are now finding ipython using platutils.find_cmd, but we still see this error.

[ipython]

[ LP comment 2 by: Brian Granger, on 2009-04-25 22:24:37.298541+00:00 ]

Here is an IPython session showing this bug:

In [4]: from IPython.platutils import find_cmd

In [5]: from IPython.platutils import get_long_path_name

In [6]: cmd = find_cmd('ipython')

In [7]: cmd
Out[7]: 'C:\Python25\Scripts\ipython.exe'

In [8]: _ip.IP.getout
_ip.IP.getoutput _ip.IP.getoutputerror

In [8]: _ip.IP.getoutput(cmd+' obj_del.py')
Out[8]: ''

In [9]: _ip.IP.getoutput('ipython obj_del.py')
Out[9]: ''

I spent some time looking around and getoutput uses genutils.getoutput, which uses os.popen. The odd thing is when I use os.popen3, to get the stderr, it hangs and when I use CTRL-C to abort, IPython crashes hard. :-(

[ipython]

[ LP comment 3 by: Jörgen Stenarson, on 2009-04-26 08:10:57+00:00 ]

Brian Granger skrev:

Here is an IPython session showing this bug:

In [4]: from IPython.platutils import find_cmd

In [5]: from IPython.platutils import get_long_path_name

In [6]: cmd = find_cmd('ipython')

In [7]: cmd
Out[7]: 'C:\Python25\Scripts\ipython.exe'
This exe file is some easy_install magic which has caused me some
problems before. Why not try to do a call to python ipython.py instead?

In [8]: _ip.IP.getout
_ip.IP.getoutput _ip.IP.getoutputerror

In [8]: _ip.IP.getoutput(cmd+' obj_del.py')
Out[8]: ''

In [9]: _ip.IP.getoutput('ipython obj_del.py')
Out[9]: ''

I spent some time looking around and getoutput uses genutils.getoutput,
which uses os.popen. The odd thing is when I use os.popen3, to get the
stderr, it hangs and when I use CTRL-C to abort, IPython crashes hard.
:-(
This crash may be related to the problem I have seen with ipython.exe
before. There have been some threads on the lists about this problem
before. Basically the ipython.exe launches a new process that starts a
python process that runs ipython. But the original process still sees
any ctrl-c action you do and shuts down. My recomendation is to not use
ipython.exe at all.

/Jörgen

[ipython]

[ LP comment 4 by: Brian Granger, on 2009-04-27 20:57:20+00:00 ]

Ahh, that may be it. I will give this a shot.

Brian

On Sun, Apr 26, 2009 at 1:10 AM, Jörgen Stenarson
jorgen.stenarson@bostream.nu wrote:

Brian Granger skrev:

Here is an IPython session showing this bug:

In [4]: from IPython.platutils import find_cmd

In [5]: from IPython.platutils import get_long_path_name

In [6]: cmd = find_cmd('ipython')

In [7]: cmd
Out[7]: 'C:\Python25\Scripts\ipython.exe'
This exe file is some easy_install magic which has caused me some
problems before. Why not try to do a call to python ipython.py instead?

In [8]: _ip.IP.getout
_ip.IP.getoutput      _ip.IP.getoutputerror

In [8]: _ip.IP.getoutput(cmd+' obj_del.py')
Out[8]: ''

In [9]: _ip.IP.getoutput('ipython obj_del.py')
Out[9]: ''

I spent some time looking around and getoutput uses genutils.getoutput,
which uses os.popen.  The odd thing is when I use os.popen3, to get the
stderr, it hangs and when I use CTRL-C to abort, IPython crashes hard.
:-(
This crash may be related to the problem I have seen with ipython.exe
before. There have been some threads on the lists about this problem
before. Basically the ipython.exe launches a new process that starts a
python process that runs ipython. But the original process still sees
any ctrl-c action you do and shuts down. My recomendation is to not use
ipython.exe at all.

/Jörgen

test_magic.test_obj_del fails on win32
https://bugs.launchpad.net/bugs/366334
You received this bug notification because you are a member of IPython
Developers, which is subscribed to IPython.

Status in IPython - Enhanced Interactive Python: Confirmed

Bug description:
test_obj_del fails on win32 because getoutput is called like this:

IP.getoutput('ipython obj_del.py')

But on Windows, the ipython executable is not found.  This causes stdout to be empty
and stderr to have an error saying that ipython can't be found.

FAIL: Test that object's del methods are called on exit.

Traceback (most recent call last):
 File "c:\python24\lib\site-packages\nose-0.10.4-py2.4.egg\nose\case.py", line 182, in runTest
   self.test(*self.arg)
 File "c:\python24\external\ipython\IPython\tests\test_magic.py", line 86, in test_obj_del
   nt.assert_equals(out,'obj_del.py: object A deleted')
AssertionError: '' != 'obj_del.py: object A deleted'

raise self.failureException,
         (None or '%r != %r' % ('', 'obj_del.py: object A deleted'))

---

Brian E. Granger, Ph.D.
Assistant Professor of Physics
Cal Poly State University, San Luis Obispo
bgranger@calpoly.edu
ellisonbg@gmail.com

[fperez]

@jstenar, can you see if this problem still exists?

[minrk]

This test is currently skipped on Windows, but letting it run, it does pass for me. The neighboring test_tclass that is also skipped in Windows still fails.

It does depend on win32api, so it still needs that conditional on it.

[takluyver]

If this is passing on Windows now, can we remove the skip_win32 decorator and close the issue? If test_tclass still fails, should we leave this issue open for it, or start a new one?

[fperez]

@takluyver, I'd remove the deco, close this one, and open a clean one for tclass if that fails, simply with a link to this one in there in case it's useful to read some of this discussion when thinking about the tclass one.