python 2.7 & sem_open

[Patryk Zawadzki]

On Tue, Sep 28, 2010 at 7:58 PM, Jakub Bogusz <qboosh at pld-linux.org> wrote:
> On Tue, Sep 28, 2010 at 05:11:01PM +0200, Patryk Zawadzki wrote:
>> Multiprocessing seems to be broken in our python packages:
>>
>> $ python -c 'from multiprocessing.queues import SimpleQueue'
>> Traceback (most recent call last):
>>   File "<string>", line 1, in <module>
>>   File "/usr/share/python2.7/multiprocessing/queues.py", line 22, in <module>
>>   File "/usr/share/python2.7/multiprocessing/synchronize.py", line 33,
>> in <module>
>> ImportError: This platform lacks a functioning sem_open
>> implementation, therefore, the required synchronization primitives
>> needed will not function, see issue 3770.
>>
>> Mentioned bug 3770? claims that sem_open support is checked for at
>> build time. The test case is this:
> IMO API check should be enough on builders (when build host is not
> the destination machine).

Python devs say it has to do a syscall check at build time as there is
no way to modify the behavior at runtime (the whole module has to be
disabled).

>> It seems our builders don't have semaphore support in their kernels
>> (or the syscalls are being filtered by some security mechanism).
>>
>> This results in partially broken python (and renders me unable to use celery).
> /dev/shm mounted?

Yup.

-- 
Patryk Zawadzki