scipy: deadlock on osx for python 3.8

We are still seeing something weird with a python 3.8 build on osx over on conda-forge

==== testing package: scipy-1.3.1-py38h82752d6_2 =====
running run_test.py
============================= test session starts ==============================
platform darwin -- Python 3.8.0, pytest-5.2.2, py-1.8.0, pluggy-0.12.0 -- $PREFIX/bin/python
cachedir: .pytest_cache
rootdir: $SRC_DIR
collecting ... collected 17186 items / 1225 deselected / 15961 selected

_lib/tests/test__gcutils.py::test_set_gc_state PASSED                    [  0%]
_lib/tests/test__gcutils.py::test_gc_state PASSED                        [  0%]
_lib/tests/test__gcutils.py::test_assert_deallocated PASSED              [  0%]
_lib/tests/test__gcutils.py::test_assert_deallocated_nodel PASSED        [  0%]
_lib/tests/test__gcutils.py::test_assert_deallocated_circular PASSED     [  0%]
_lib/tests/test__gcutils.py::test_assert_deallocated_circular2 PASSED    [  0%]
_lib/tests/test__testutils.py::test__parse_size PASSED                   [  0%]
_lib/tests/test__testutils.py::test__mem_available PASSED                [  0%]
_lib/tests/test__threadsafety.py::test_parallel_threads PASSED           [  0%]
_lib/tests/test__threadsafety.py::test_reentering PASSED                 [  0%]
_lib/tests/test__util.py::test__aligned_zeros PASSED                     [  0%]
_lib/tests/test__util.py::test_check_random_state PASSED                 [  0%]
_lib/tests/test__util.py::TestMapWrapper::test_serial PASSED             [  0%]

https://dev.azure.com/conda-forge/feedstock-builds/_build/results?buildId=91407&view=logs

This is with a patch from the 1.3.x branch here

https://github.com/conda-forge/scipy-feedstock/blob/02371e5393763b42fef7f50552829e57f2eb79f3/recipe/scipy1.3.2.patch

@rgommers for viz

relates to https://github.com/conda-forge/scipy-feedstock/pull/114

About this issue

  • Original URL
  • State: closed
  • Created 5 years ago
  • Comments: 23 (19 by maintainers)

Most upvoted comments

Got it, so we do still have a real problem. Release is already tagged, and wheels are being built. I think it’s best to continue with the release, and try to debug this problem asap and push out a 1.3.3 fix if needed. People are blocked upgrading to py38 by no SciPy release, so this seems better than delaying since we don’t have a root cause or fix yet. @tylerjereddy FYI

I am using the latest setuptools and unfortunately it is not solved.

So I am pretty sure the commit you mentioned was already applied.

yes, checked the diff and the patch was already applied

For sure but it will be a while since I need to run to a kids birthday party. This is a conda forge build so we are linking to whatever openblas version is in the global pinnings.