Python Tests Fail "ImportError: dynamic module does not define module export function (PyInit__base)"

[mamoll]

Original report by Anonymous.

---

Hi,

I couldn't find this issue yet. I ran

#!c++

cmake -DPYTHON_EXEC=/usr/local/bin/python3 -DOMPL_BUILD_PYBINDINGS=ON -DOMPL_BUILD_PYTESTS=ON ../..

make -j 8 update_bindings

make -j 8

and all ran without error. Then I ran the tests

#!c++
make test

All worked except

21:test_base.py

22:test_geometric.py

23:test_geometric_compoundstate.py

24:test_control.py

I've attached the LastTest.log
All the tests failed at

Traceback (most recent call last):
File "/Users/Adam/Documents/UofT/Thesis/code/omplapp/ompl/tests/control/test_control.py", line 48, in
import ompl.base as ob
File "/Users/Adam/Documents/UofT/Thesis/code/omplapp/ompl/py-bindings/ompl/base/init.py", line 2, in
from ompl.base._base import *
ImportError: dynamic module does not define module export function (PyInit__base)

Any thoughts?

I'm installing on MacOSX Mojave (10.14.2).

[mamoll]

Original comment by Mark Moll (Bitbucket: mamoll, GitHub: mamoll).

---

Which version of Boost are you using? It’s working fine for me on OS X 10.14.2 with Boost 1.66.

[mamoll]

Original comment by Adam Hall (Bitbucket: [Adam Hall](https://bitbucket.org/Adam Hall), ).

---

It looks like version 1.70. Should I revert back to 1.66?

‌

brew info boost

boost: stable 1.70.0 (bottled), HEAD

Collection of portable C++ source libraries

https://www.boost.org/

/usr/local/Cellar/boost/1.68.0 (13,712 files, 460.2MB)

Poured from bottle on 2018-12-18 at 14:32:00

/usr/local/Cellar/boost/1.70.0 (13,983 files, 521.8MB) *

Poured from bottle on 2019-06-13 at 16:18:56

From: https://github.com/Homebrew/homebrew-core/blob/master/Formula/boost.rb

==> Dependencies

Required: icu4c ✔

==> Options

--HEAD

Install HEAD version

==> Analytics

install: 53,542 (30 days), 190,823 (90 days), 635,367 (365 days)

install_on_request: 18,137 (30 days), 63,790 (90 days), 197,333 (365 days)

build_error: 0 (30 days)

[mamoll]

Original comment by Mark Moll (Bitbucket: mamoll, GitHub: mamoll).

---

If you have an easy to revert to an older version of Boost, then it’d be nice if you can confirm that the problem is indeed with Boost 1.70. I installed Boost and other dependencies of OMPL via MacPorts, but I think it’s most likely a change in Boost that’s to blame. I don’t know when I will have time to track this down.

[mamoll]

Original comment by Adam Hall (Bitbucket: [Adam Hall](https://bitbucket.org/Adam Hall), ).

---

Okay I’ll try a reinstall using Boost version 1.66. I’ll let you know how it goes. Thanks!

[mamoll]

Original comment by Adam Hall (Bitbucket: [Adam Hall](https://bitbucket.org/Adam Hall), ).

---

Hi Mark,

I tried installing with 1.68 (brew doesn’t let you get 1.66 easily) and had the same problem. Is it possibly a problem with libspot or py++? Upon closer inspection I found this output after running make -j 8 update_bindings.

‌

I’m currently trying to install with boost 1.59. I will let you know how it goes.

‌

file "vectorPlannerSolution.pypp.cpp" - updated( 0.000558 seconds )

Traceback (most recent call last):

File "/Users/Adam/Documents/UofT/Thesis/code/omplapp/ompl/py-bindings/generate_bindings.py", line 930, in

globals()['ompl_'+module+'_generator_t']()

File "/Users/Adam/Documents/UofT/Thesis/code/omplapp/ompl/py-bindings/generate_bindings.py", line 154, in __init__

code_generator_t.__init__(self, 'base', ['bindings/util'], replacement)

File "/Users/Adam/Documents/UofT/Thesis/code/omplapp/ompl/py-bindings/ompl/bindings_generator.py", line 151, in __init__

self.mb.split_module('bindings/' + name, use_files_sum_repository=True)

File "/usr/local/lib/python3.7/site-packages/pyplusplus/module_builder/boost_python_builder.py", line 384, in split_module

, encoding=self.encoding)

File "/usr/local/lib/python3.7/site-packages/pyplusplus/file_writers/__init__.py", line 37, in write_multiple_files

mfs.write()

File "/usr/local/lib/python3.7/site-packages/pyplusplus/file_writers/multiple_files.py", line 396, in write

self.split_classes()

File "/usr/local/lib/python3.7/site-packages/pyplusplus/file_writers/multiple_files.py", line 300, in split_classes

for cls in class_creators: self.split_class(cls)

File "/usr/local/lib/python3.7/site-packages/pyplusplus/file_writers/multiple_files.py", line 288, in split_class

self.split_class_impl( class_creator )

File "/usr/local/lib/python3.7/site-packages/pyplusplus/file_writers/multiple_files.py", line 262, in split_class_impl

, self.create_function_code( function_name ) ) )

File "/usr/local/lib/python3.7/site-packages/pyplusplus/file_writers/multiple_files.py", line 56, in write_file

raise RuntimeError( os.linesep.join(msg) )

RuntimeError: `Py++` is going to write different content to the same file(bindings/base/__self.pypp.hpp).

The following is a short list of possible explanations for this behaviour:

* `Py++` bug, in this case, please report it

* module_builder_t contains two or more classes with the same alias

* module_builder_t contains two or more classes with the same wrapper alias

Please carefully review `Py++` warning messages. It should contain an additional information.

-- Build type: Release

-- Checking for module 'libspot'

-- No package 'libspot' found

-- Could NOT find SPOT (missing: SPOT_LIBRARIES SPOT_INCLUDE_DIRS)

-- Code for module control not found; type "make update_bindings"

-- Code for module geometric not found; type "make update_bindings"

-- Code for module tools not found; type "make update_bindings"

-- Code for module app not found; type "make update_bindings"

-- The following features have been enabled:

[mamoll]

Original comment by Adam Hall (Bitbucket: [Adam Hall](https://bitbucket.org/Adam Hall), ).

---

boost 1.59 had the same issue with the failed tests.

[mamoll]

Original comment by Mark Moll (Bitbucket: mamoll, GitHub: mamoll).

---

I committed a temporary workaround: https://bitbucket.org/ompl/ompl/commits/eece2cddbb3a43292c499bcaf711aa4ac1339b75

[mamoll]

Original comment by Adam Hall (Bitbucket: [Adam Hall](https://bitbucket.org/Adam Hall), ).

---

Work around seemed to work! Thanks! I used boost 1.59 and boost 1.70. Both worked.

‌

Does this workaround cause limited functionality?

[mamoll]

Original comment by Mark Moll (Bitbucket: mamoll, GitHub: mamoll).

---

Yes, in the sense that there won’t be any Python bindings for the ConstraintIntersection class, but that’s a rather specialized class used by few people at this point, I suspect.