[yt-users] Fw: mpi4py on Kraken (XT5)

[Matthew Turk]

Hi Stephen,

Okay, you're right -- my suspicion about distutils was a red herring.
Darn.  :)  The symbols should resolve long enough for linking via the
usage of libpython2.6.a.  They weren't deprecated, as in my (shared
lib) build of python2.6 on an x86_64 machine I have them.

In your Python2.6 source directory, what is the output of:

$ grep ^ac config.log | grep thread

In my CNL build of 2.5, I have:

ac_cv_cxx_thread=no
ac_cv_func_pthread_init=no
ac_cv_func_pthread_sigmask=yes
ac_cv_header_pthread_h=yes
ac_cv_header_thread_h=no
ac_cv_kpthread=no
ac_cv_kthread=no
ac_cv_pthread=no
ac_cv_pthread_is_default=yes
ac_cv_pthread_system_supported=yes
ac_cv_sizeof_pthread_t=8

I wonder if there is something going on in the include files; could
you additionally send back your ./configure line, taken from the top
of config.log in your python2.6 directory?

Regardless, it kind of seems likely to me that if these functions are
not part of your libpython2.6.a, that static library might in fact be
defective.  If you cd to your Python2.6 directory and do:

$ nm Python/pystate.o | grep PyGIL

Do you get those symbols as being defined?  If so, the building of
python2.6.a has not worked properly...

-Matt

> I don't see these in libpython2.6.a.
>
>> Can you tell me which version of python you are compiling with, and
>> also send the *actual* C compiler line that failed?
>
> Python 2.6.
>
> http://paste.enzotools.org/show/42/
>
>> In my umath.a I
>> have those two symbols undefined:
>>
>>                  U PyGILState_Ensure
>>                  U PyGILState_Release
>
> I find these in my umath.a.
>
>> Did you modify the distutils code as per the GPAW instructions?  I
>> have noticed their indentation is a bit unclear; the function they
>> provide as "link_shared_object" should in fact be a *member* of
>> UnixCCompiler, so it should be indented an additional level beyond
>> what is obvious in the code sample.  In fact, this seems likely to be
>> the culprit here, now that I think about it...
>
> No, I think the indentation is correct. In particular, I am successful in getting "Append to Setup: " lines when I re-install numpy.
>
> _______________________________________________________
> sskory at physics.ucsd.edu           o__  Stephen Skory
> http://physics.ucsd.edu/~sskory/ _.>/ _Graduate Student
> ________________________________(_)_\(_)_______________
> _______________________________________________
> yt-users mailing list
> yt-users at lists.spacepope.org
> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>