Bug 1148173

Summary: python-numpy has 6 failing tests for ppc64le since version 1.16.3
Product: [openSUSE] openSUSE Tumbleweed Reporter: Michel Normand <normand>
Component: DevelopmentAssignee: Dirk Mueller <dmueller>
Status: RESOLVED UPSTREAM QA Contact: E-mail List <qa-bugs>
Severity: Normal    
Priority: P5 - None CC: alynx.zhou, code, evgueni.tchubykalo, ioannis.bonatakis, mcepl, toddrme2178
Version: Current   
Target Milestone: ---   
Hardware: Other   
OS: Other   
See Also: http://bugzilla.opensuse.org/show_bug.cgi?id=1172555
Whiteboard:
Found By: --- Services Priority:
Business Priority: Blocker: ---
Marketing QA Status: --- IT Deployment: ---
Attachments: python_numpy_standard_ppc64le_201908231005.log

Description Michel Normand 2019-08-26 13:44:38 UTC
Created attachment 815730 [details]
python_numpy_standard_ppc64le_201908231005.log

python-numpy has 6 failing tests for ppc64le since version 1.16.3

as per attached python_numpy_standard_ppc64le_201908231005.log there are 6 failing tests (1)
And the osc jobhist (2) lets assume failure since 20190522, so with package version 1.16.3

(1) extract of python_numpy_standard_ppc64le_201908231005.log
===
[  251s] ============================= test session starts ==============================
[  281s] ../core/tests/test_einsum.py::TestEinsum::test_einsum_sums_cfloat64 FAILED [ 12%]
[  780s] ../linalg/tests/test_linalg.py::TestInv::test_sq_cases FAILED            [ 60%]
[  781s] ../linalg/tests/test_linalg.py::TestInv::test_generalized_sq_cases FAILED [ 60%]
[  796s] ../linalg/tests/test_linalg.py::TestCond::test_sq_cases FAILED           [ 61%]
[  800s] ../linalg/tests/test_linalg.py::TestCond::test_generalized_sq_cases FAILED [ 61%]
[  827s] ../linalg/tests/test_linalg.py::TestCholesky::test_basic_property FAILED [ 63%]
[ 1125s] _____________________ TestEinsum.test_einsum_sums_cfloat64 _____________________
[ 1125s] >           assert_equal(np.einsum("...i, ...i", a, b, optimize=do_opt), np.inner(a, b))
[ 1125s] E           AssertionError:
[ 1125s] E           Arrays are not equal
[ 1125s] ____________________________ TestInv.test_sq_cases _____________________________
[ 1125s]         assert_almost_equal(dot_generalized(a, a_inv),
[ 1125s] >                           identity_like_generalized(a))
[ 1125s] ______________________ TestInv.test_generalized_sq_cases _______________________
[ 1125s] E                x: array([[[ 1.000000000000e+00,  1.297115638874e-18, -4.724840362098e-17,
[ 1125s] E                        -7.773841053956e-16, -1.146365979120e-16,  4.796253199340e-17,
[ 1125s] E                        -9.425056501197e-18,  2.183139631582e-16],...
[ 1125s] E                y: array([[[1., 0., 0., 0., 0., 0., 0., 0.],
[ 1125s] E                       [0., 1., 0., 0., 0., 0., 0., 0.],
[ 1125s] E                       [0., 0., 1., 0., 0., 0., 0., 0.],...
[ 1125s] ____________________________ TestCond.test_sq_cases ____________________________
[ 1125s] E               Arrays are not almost equal to 11 decimals
[ 1125s] E                ACTUAL: 136.713644070272
[ 1125s] E                DESIRED: 158.85751255811152
[ 1125s] ______________________ TestCond.test_generalized_sq_cases ______________________
[ 1125s]         Arrays are not almost equal to 5 decimals
[ 1125s]         Mismatch: 33.3%
[ 1125s]         Max absolute difference: 6.e-05
[ 1125s]         Max relative difference: 2.57136612e-05

[ 1125s]          x: array([1.     , 2.33333,     nan])
[ 1125s]          y: array([1.     , 2.33339,     nan])
[ 1125s] _______________________ TestCholesky.test_basic_property _______________________
[ 1125s] >       raise LinAlgError("Matrix is not positive definite")
[ 1125s] E       numpy.linalg.LinAlgError: Matrix is not positive definite
===

(2) osc jobhist output:
===
$osc jobhist  openSUSE:Factory:PowerPC/python-numpy standard ppc64le
time                 package      reason           code      build time      worker
...
2019-03-04 09:36:50  python-numpy source change    succeeded 22m 59s   obs-power8-04:18
2019-03-23 06:00:24  python-numpy source change    succeeded 23m 28s   obs-power8-04:2
2019-03-27 16:17:06  python-numpy source change    succeeded 22m 54s   obs-power8-04:18
2019-04-04 11:14:38  python-numpy source change    succeeded 23m 17s   obs-power8-04:19
2019-05-22 09:33:05  python-numpy source change    failed    13m 38s   obs-power8-04:9
2019-06-02 13:55:29  python-numpy source change    failed    14m  3s   obs-power8-04:11
2019-06-22 01:08:13  python-numpy new build        failed    14m  6s   obs-power8-02:5
2019-07-30 04:31:24  python-numpy new build        failed    13m 52s   obs-power8-02:7
2019-08-19 19:31:37  python-numpy source change    failed    24m  0s   obs-power8-02:3
===
Comment 1 Michel Normand 2019-09-06 10:05:05 UTC
create  an upstream issue https://github.com/numpy/numpy/issues/14438
Comment 2 Benjamin Greiner 2020-06-30 13:16:18 UTC
https://build.opensuse.org/request/show/817874

Not a solution, but skipping the two remaining tests to have an updated package.
Comment 5 Matej Cepl 2022-05-18 09:37:43 UTC
*** Bug 1194863 has been marked as a duplicate of this bug. ***
Comment 6 Dirk Mueller 2022-09-19 09:25:25 UTC
I don't have the resources to fix this. please report upstream.