Re: [announce] python-kmod 0.9

[Andy Grover <agrover@redhat.com>]

On 03/26/2014 01:49 PM, Lucas De Marchi wrote:
> On Wed, Mar 26, 2014 at 5:32 PM, Andy Grover <agrover@redhat.com> wrote:
>> On 03/26/2014 10:45 AM, Lucas De Marchi wrote:
>>>
>>> On Mon, Mar 24, 2014 at 10:12:26AM -0700, Andy Grover wrote:
>>>>
>>>> On 03/24/2014 05:02 AM, Lucas De Marchi wrote:
>>>>>
>>>>> I've just pushed a branch named "python" which contains the python
>>>>> bindings. I did it a bit different than you:
>>>>>
>>>>>    - All the code from your repository was imported maintaining the
>>>>> history. I would like to keep it, so I did a merge of the final import
>>>>> (fast forward, but forced to contain a commit).
>>>>>    - Python bindings are built with autotools. This allows to easily
>>>>> express the dependency with libkmod... but I'm not sure this is ideal
>>>>> as opposed to having a target to explicitely calling setuptools.  Any
>>>>> opinion?
>>>>>
>>>>> Then I noticed the example given in the README file doesn't work.
>>>>> Neither by installing they original python-kmod package :-/
>>>>>
>>>>>       >>> import kmod
>>>>>       >>> km = kmod.Kmod()
>>>>>       Traceback (most recent call last):
>>>>>         File "<stdin>", line 1, in <module>
>>>>>       AttributeError: 'module' object has no attribute 'Kmod'
>>>>>       >>> dir(kmod)
>>>>>       ['__builtins__', '__doc__', '__file__', '__loader__', '__name__',
>>>>> '__package__', '__path__', '__version__', 'list', 'version']
>>>>>
>>>>
>>>> The old 'python-kmod' pkg works for me:
>>>>
>>>> [agrover@work ~/.../python/kmod ((fac4d09...))]$ ipython
>>>>
>>>> In [1]: import kmod
>>>>
>>>> In [2]: km = kmod.Kmod()
>>>>
>>>> In [3]: sc = list(km.lookup("soundcore"))[0]
>>>>
>>>> In [4]: sc.path
>>>> Out[4]: u'/lib/modules/3.13.6-200.fc20.x86_64/kernel/sound/soundcore.ko'
>>>>
>>>> but when I try to use the new stuff:
>>>>
>>>> [agrover@work ~/.../python/kmod ((fac4d09...))]$
>>>> PYTHONPATH="/home/agrover/git/kmod/libkmod/python/kmod/.libs" ipython
>>>>
>>>> In [1]: import kmod
>>>>
>>>> ---------------------------------------------------------------------------
>>>> ImportError                               Traceback (most recent call
>>>> last)
>>>> <ipython-input-1-ff824b795612> in <module>()
>>>> ----> 1 import kmod
>>>>
>>>> ImportError: dynamic module does not define init function (initkmod)
>>>>
>>>> Is this what you're seeing as well?
>>>
>>>
>>>
>>> humn... indeed. The error was obfuscated for me because in the
>>> __init__.py you catch the exception and make it pass silently.
>>>
>>> In the build system I was inheriting the CFLAGS from the library and
>>> unfortunately PyMODINIT_FUNC doesn't include the visibility attribute.
>>> Since we use -fvisibility=hidden by default, this was breaking the
>>> module.
>>>
>>> It's working now, though I don't like to let it as -fvisibility=default.
>>>
>>> I just pushed the python branch again. Could you ack on it?
>>
>>
>> I'm still getting this error.
>>
>> [agrover@work ~/git/kmod (python)]$
>> PYTHONPATH="/home/agrover/git/kmod/libkmod/python/kmod/.libs" python
>> Python 2.7.5 (default, Feb 19 2014, 13:47:28)
>> [GCC 4.8.2 20131212 (Red Hat 4.8.2-7)] on linux2
>> Type "help", "copyright", "credits" or "license" for more information.
>>>>> import kmod
>>
>> Traceback (most recent call last):
>>    File "<stdin>", line 1, in <module>
>>    File "list.pxd", line 21, in init kmod.kmod
>> (libkmod/python/kmod/kmod.c:3831)
>> ImportError: No module named list
>>
>> I'm sorry I'm not being more help.. maybe we could get W. Trevor King or
>> another Cython contributor to help us out here? I played around with it a
>> bit and changing "kmod.list" to "list" on the line reporting an error in
>> kmod.c seemed to help, and then repeating for additional errors, but kmod.c
>> is a generated file so that's not the actual solution.
>
> Humn... yes. If I set PYTHONPATH as opposed to installing I get this
> error. I'm not sure what's happening.
>

Ah ok! PYTHONPATH needs to be to "...site-packages", not 
"...site-packages/kmod". so I'm all good :-)

Acked-by: Andy Grover <agrover@redhat.com>

Thanks! -- Andy