Import sklearn dumps core

[monikasoni]

I built sklearn and trying to import with Python3.7 getting core dump

Python 3.7.5 (default, Nov  9 2019, 17:49:52) 
[GCC 5.4.0 20160609] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import platform; print(platform.platform())
Linux-4.19.46-4.19.1-amd64-9cf031ffa09555f2-x86_64-with-debian-stretch-sid
>>> import sys; print("Python", sys.version)
Python 3.7.5 (default, Nov  9 2019, 17:49:52) 
[GCC 5.4.0 20160609]
>>> import numpy; print("NumPy", numpy.__version__)
NumPy 1.17.1
>>> import scipy; print("SciPy", scipy.__version__)
SciPy 1.3.1
>>> import sklearn; print("Scikit-Learn", sklearn.__version__)
Segmentation fault (core dumped)

Stack Trace

======
#0 0x00007fffed400d4e in PyTuple_New (size=) at Objects/tupleobject.c:133
#1 0x00007fffed40824e in PyType_Ready (type=type@entry=0x7fffed80fe80 <PyBaseObject_Type>) at Objects/typeobject.c:5196
#2 0x00007fffed4081d6 in PyType_Ready (type=0x7fffed808b80 <PyModuleDef_Type>) at Objects/typeobject.c:5178
#3 0x00007fffed3f2250 in PyModuleDef_Init (def=0x7fffed7ef600 <__pyx_moduledef>) at Objects/moduleobject.c:46
#4 0x00000000004d7345 in _PyImport_LoadDynamicModuleWithSpec (spec=spec@entry=0x7fffed8b28d0, fp=fp@entry=0x0)
at ./Python/importdl.c:159
#5 0x00000000004d5195 in _imp_create_dynamic_impl (module=, file=, spec=0x7fffed8b28d0)
at Python/import.c:2170
#6 _imp_create_dynamic (module=, args=, nargs=) at Python/clinic/import.c.h:289
#7 0x000000000043a544 in _PyMethodDef_RawFastCallDict (method=0x844460 <imp_methods+320>, self=0x7ffff7f6bcb0,
args=, nargs=1, kwargs=) at Objects/call.c:530
#8 0x000000000043a5f5 in _PyCFunction_FastCallDict (func=0x7ffff7f847d0, args=, nargs=,
kwargs=) at Objects/call.c:586
#9 0x0000000000429e26 in do_call_core (kwdict=0x7fffed8b0aa0, callargs=0x7fffed8b22d0, func=0x7ffff7f847d0)
at Python/ceval.c:4641
#10 _PyEval_EvalFrameDefault (f=, throwflag=) at Python/ceval.c:3191
#11 0x00000000004be574 in _PyEval_EvalCodeWithName (_co=0x7ffff7f73150, globals=, locals=locals@entry=0x0,
args=, argcount=2, kwnames=0x0, kwargs=kwargs@entry=0x7ffff650fb00, kwcount=kwcount@entry=0,
kwstep=kwstep@entry=1, defs=0x0, defcount=0, kwdefs=kwdefs@entry=0x0, closure=closure@entry=0x0,
name=name@entry=0x7ffff7f71300, qualname=qualname@entry=0x7ffff7f71300) at Python/ceval.c:3930
#12 0x000000000043987d in _PyFunction_FastCallKeywords (func=, stack=, nargs=,
kwnames=) at Objects/call.c:433
#13 0x0000000000429b3c in call_function (kwnames=0x0, oparg=, pp_stack=) at Python/ceval.c:4616
#14 _PyEval_EvalFrameDefault (f=<optimized
...
...
...

[glemaitre]

Could you describe the way that you try to install scikit-learn: from source, via pip, via conda?
In case you installed it from source, could you also give the version of Cython.

[monikasoni]

I am trying to build from source, Using Cython-0.28.5
Also set env variable SKLEARN_NO_OPENMP to 0.
'export SKLEARN_NO_OPENMP=0'

[glemaitre]

You will have support for OpenMP with the version of gcc that you are using.
When compiling with Cython, do you have any issues? You can post the output of the command make clean && make in from the root folder.

[monikasoni]

The problem is custom way my packages were being built. We use company specific build framework and it introduced linkage with -lpython3.7m from some configuration file. When linking the dynamic library is removed import works fine. Closing to the issue now.

Thanks for your time @glemaitre.

[glemaitre]

Thanks for giving us the reason :)

…

On Fri, 15 Nov 2019 at 18:23, Monika ***@***.***> wrote: Closed #15580 <#15580>. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#15580?email_source=notifications&email_token=ABY32P4ZFN44IH23OWUQKFDQT3LIDA5CNFSM4JLJE65KYY3PNVWWK3TUL52HS4DFWZEXG43VMVCXMZLOORHG65DJMZUWGYLUNFXW5KTDN5WW2ZLOORPWSZGOU4MHYOI#event-2803399737>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABY32PY2RFDT6IVMAEWCCYTQT3LIDANCNFSM4JLJE65A> .

-- Guillaume Lemaitre Scikit-learn @ Inria Foundation https://glemaitre.github.io/