Re: [Discuss-gnuradio] make check test failures, Mac OS X, Python/SWIG

[Stephane Fillod]

On Wed, Jan 12, 2005 at 11:28:48AM -0800, Jonathan Jacky wrote:
> 
> I built gnuradio-core from CVS on Mac OS X.
[...]
> PS Here are the messages I got from "make check" after I built with swig
> 1.3.22.  This first one seems to come from C++, so it is probably not
> related to swig:
> 
> Making check in tests
> make  check-TESTS
> .Testing gr_vmcircbuf_sysv_shm_factory...
> gr_vmcircbuf_sysv_shm: shmat (1): Too many open files
> gr_vmcircbuf_sysv_shm: shmget (1): Invalid argument
> gr_vmcircbuf_sysv_shm: shmat (1): Too many open files
> ....... gr_vmcircbuf_sysv_shm_factory: Failed

Can these system limits be pushed further? (I know next to nothing of OSX)
Have you tried the tricks at http://comsec.com/wiki?MacOsX ?

If you have at least one factory working, you can ignore the sysv_shm.

> Then here are the ones that seem to be related to swig:
> 
> Making check in gr
> make  check-TESTS
> Traceback (most recent call last):
>   File "./qa_add_and_friends.py", line 23, in ?
>     from gnuradio import gr, gr_unittest
>   File
>   
> "/Users/jon/gnuradio/gr-build/gnuradio-core/src/python/gnuradio/gr/__init__.py",
>  
> line 27, in ?
>     from gnuradio_swig_python import *
>   File
>   
> "/Users/jon/gnuradio/gr-build/gnuradio-core/src/lib/swig/gnuradio_swig_python.py",
>  
> line 5, in ?
>     import _gnuradio_swig_python
> ImportError: Failure linking new module: : dyld: python Undefined
> symbols:
> _SWIG_Python_ConvertPtr
> _SWIG_Python_InstallConstants
> _SWIG_Python_NewPointerObj
> _SWIG_Python_TypeClientData
> _SWIG_Python_TypeQuery
> _SWIG_Python_TypeRegister
> _SWIG_Python_addvarlink
> _SWIG_Python_newvarlink

Under Cygwin and MinGW, I fixed this problem with the attached patch.
The "-no-undefined" are required for win32 system, harmless under others.
The "address@hidden@" is what may solve your problem. Make sure
the library is in your linker path (maybe autoconf should check
whether "address@hidden@" links fine).

Eric, what are your thoughts about this?


-- 
Stephane

pythonwin.patch
Description: Text document