"Segmentation fault: 11" while importing ModelFactory

[nbro]

System Information

OS: Mac OS X, version 10.13.6 (High Sierra)
Python: 2.7.10 (System's one, under /usr/bin) and 2.7.15 (Homebrew-installed one, under /usr/local/bin)
GCC: g++ (GCC) 6.3.1 20170510 (for GNAT GPL 2017 20170515)
Apple LLVM version 9.1.0 (clang-902.0.39.2), Target: x86_64-apple-darwin17.7.0, Thread model: posix

Problem Description

I installed NuPIC in a virtual environment by issuing the command pip install nupic. This also installed all NuPIC's dependencies, including nupic.bindings. I can import nupic normally without getting errors, but if I try to import ModelFactory from nupic.frameworks.opf.model_factory I get a "Segmentation fault: 11".

Python 2.7.10 (default, Oct  6 2017, 22:29:07) 
[GCC 4.2.1 Compatible Apple LLVM 9.0.0 (clang-900.0.31)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import nupic
>>> from nupic.frameworks.opf.model_factory import ModelFactory
Segmentation fault: 11

or, using Homebrew installed one,

Python 2.7.15 (default, Jun 17 2018, 12:46:58) 
[GCC 4.2.1 Compatible Apple LLVM 9.1.0 (clang-902.0.39.2)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import nupic
>>> from nupic.frameworks.opf.model_factory import ModelFactory
Segmentation fault: 11

Even if I install NuPIC from the local source code, using pip install . I get a "segmentation fault 11".

I have looked for similar issues. It's not the first time that people, in the NuPIC community, report a "segmentation fault" error.

Comments

In the README file, you state that NuPIC binaries are only available for OSX 10.9 and 10.10, so I suppose that no one has tried to port NuPIC to (or test on) OS X 10.13. Anyway, NuPIC (not nupic.core) is a Python package, so I don't really get what you mean by "binary" here, in this context.

If this is an issue that all Mac OS X 10.13 users have, then this is a big limitation, I think, and the issue should be solved.

I am a bit puzzled that I can import nupic, but not that particular class. nupic.frameworks.opf.model_factory probably contains the execution of code which causes this segmentation fault.

Question

Can some Mac OS X user reproduce this error?

[nbro]

I've just noticed that I'm getting this "segmentation fault 11" in several other cases (e.g. when simply importing from nupic.encoders.scalar import *), but I've still not understood what's causing it. The strangest thing is that, when I try to do from nupic.encoders.scalar import *, the "segmentation fault 11" seems to be caused by the import statement from nupic.encoders.base import Encoder, EncoderResult, but I can directly import both Encoder and EncoderResult (using from nupic.encoders.base import Encoder, EncoderResult), without any errors.

[nbro]

I installed NuPIC, on an Ubuntu 16.04 virtual machine, using Python 2.7.12. So far, I have not run into this issue. Later, I will try to install it on the Python versions 2.7.10 and 2.7.15. This will be my workaround until I (or someone else) solve this "segmentation fault 11" problem.

[rhyolight]

I think you'll have to compile locally on 10.13.6. We don't continue to support the latest OS builds.

[rhyolight]

I was just talking to @lscheinkman and he is running the same OS as you @nbro. He thinks this has something to do with your virtualenv python environment. The Seg11 does not tell us much. Brew is not going to work either.

I suggest you try to get virtualenv setup as bare as possible, make sure there is no numpy there before you install nupic. In the past, this problem has been a symptom of some numpy dependency mismatch.

If nothing else works, try outside of virtualenv. At least that will tell us that it is a problem with the python environment.

[nbro]

I installed NuPIC on an empty virtual environment (apart from the default Python packages that are added automatically to the environment when you create it, i.e. pip, setuptools and wheel).

I will try to install it outside of the virtual environment, but I don't think this is the problem (I think I already tried it).

[rhyolight]

There is something off in your python environment. Sometimes bad things in your global python env can leak into virtualenv. That has happened to me before.