Hi Bob,
We have investigated this issue and the problem is related to dlopen flags set in Ice for load plug-ins, this settings caused the segmentation when numpy load one of its extensions or modules.
We have tested that in modern linux (python-2.5, gcc-4.x) this ldflags are no longer needed, we will investigate further if this is needed for any of the supported platforms.
To work around this issue you can comment the following lines at the top of Ice.py
Code:
#try:
# import dl
#
# This is necessary for proper operation of Ice plug-ins.
# Without it, RTTI problems can occur.
#
# sys.setdlopenflags(dl.RTLD_NOW|dl.RTLD_GLOBAL)
#except ImportError:
# If the dl module is not available and we're running on a linux
# platform, use the hard coded value of RTLD_NOW|RTLD_GLOBAL.
# if sys.platform.startswith("linux"):
# sys.setdlopenflags(258)
# pass
We'll include a fix for the issue in next Ice release.
Thanks for point this bug out, and let us know if you need further help with this
Regards,
José