BUG: TestOnenormest.test_onenormest_table_6_t_1 is failing on my dev branch

[Kai-Striega]

Describe your issue.

The TestOnenormest.test_onenormest_table_6_t_1 test is failing on my dev branch. I believe this has not been checked in the CI as it is marked as XSLOW

Reproducing Code Example

`SCIPY_XSLOW=1 python dev.py test -t scipy/sparse/linalg/tests/test_onenormest.py::TestOnenormest::test_onenormest_table_6_t_1`

Error message

Installing, see meson-install.log...
Installation OK
SciPy from development installed path at: /home/kai/Projects/scipy/kai/scipy/build-install/lib/python3.10/site-packages
Running tests for scipy version:1.10.0.dev0+0.699529f, installed at:/home/kai/Projects/scipy/kai/scipy/build-install/lib/python3.10/site-packages/scipy
================================================================================= test session starts ==================================================================================
platform linux -- Python 3.10.6, pytest-7.2.0, pluggy-1.0.0
rootdir: /home/kai/Projects/scipy/kai/scipy, configfile: pytest.ini
plugins: timeout-2.1.0, cov-4.0.0, xdist-3.0.2
collected 1 item                                                                                                                                                                       

scipy/sparse/linalg/tests/test_onenormest.py F                                                                                                                                   [100%]

======================================================================================= FAILURES =======================================================================================
______________________________________________________________________ TestOnenormest.test_onenormest_table_6_t_1 ______________________________________________________________________
scipy/sparse/linalg/tests/test_onenormest.py:183: in test_onenormest_table_6_t_1
    assert_(0.7 < proportion_exact < 0.8)
E   AssertionError
        A          = array([[ 1.0543604 +0.03847635j, -0.33208373-0.22861055j,
         0.40431944+0.0792976j , ..., -0.05681823-0.97139639...     -0.72327521+0.42464523j, ...,  0.86198444+1.35242015j,
        -1.4396941 -0.77624368j, -0.03960976-0.01487513j]])
        A_inv      = array([[0.29119074+0.65104849j, 0.58305885+0.67128938j,
        0.06722355+0.93823077j, ..., 0.6889593 +0.84052724j,
 ...
        0.45835081+0.39487713j, ..., 0.39236538+0.22490301j,
        0.37966209+0.48231849j, 0.52764845+0.90868658j]])
        est        = 286.6717944249084
        expected   = array([ 71.97806709,  70.56821698,  42.24911326, ...,  42.74919002,
        71.59204191, 287.08535285])
        i          = 4999
        itmax      = 5
        max_nresamples = 0
        n          = 100
        nexact     = 2781
        nmult_list = [5, 7, 4, 6, 7, 4, ...]
        nmults     = 4
        nresample_list = [0, 0, 0, 0, 0, 0, ...]
        nresamples = 0
        nsamples   = 5000
        observed   = array([ 71.97806709,  70.56821698,  35.59762277, ...,  42.74919002,
        63.85697738, 286.67179442])
        proportion_exact = 0.5562
        relative_errors = array([0.        , 0.        , 0.15743503, ..., 0.        , 0.10804364,
       0.00144054])
        self       = <scipy.sparse.linalg.tests.test_onenormest.TestOnenormest object at 0x7fb79cad6770>
        t          = 1
        underestimation_ratio = array([1.        , 1.        , 0.84256497, ..., 1.        , 0.89195636,
       0.99855946])
        underestimation_ratio_mean = 0.9414421444954721
        v          = array([0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
       0., 0., 0., 0., 0., 0., 0., 0., 0., ...0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
       0., 0., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.])
        w          = array([ 1.93101609+4.88613457e-02j,  0.95209743+2.99680743e-01j,
        3.85029705-1.27986728e+00j,  1.89775519+4.326...578313 -9.43259458e-01j,  1.87828784-6.22941868e+00j,
        0.80356713+1.08910034e+00j, -2.96027481+1.61222120e+00j])
=============================================================================== short test summary info ================================================================================
FAILED scipy/sparse/linalg/tests/test_onenormest.py::TestOnenormest::test_onenormest_table_6_t_1 - AssertionError

### SciPy/NumPy/Python version information

1.10.0.dev0+0.699529f, '1.23.4', sys.version_info(major=3, minor=10, micro=6, releaselevel='final', serial=0)

[WarrenWeckesser]

This was noted as one of the tests that fails when SCIPY_XSLOW=1 in gh-15384. I currently get 11 failures--including this one--with the latest development branch.

I won't call this a duplicate. gh-15384 is something of a meta-issue, and I think it is reasonable for the individual failures to have their own issues, where the discussion (if any arises) can focus on the specific failure. (See also #16655, which is another failure that occurs when SCIPY_XSLOW=1.)

[WarrenWeckesser]

This is still occurring in main (dbc2d3a). Is anyone who is somewhat familiar with the code in sparse.linalg interested in taking a look at this?

[WarrenWeckesser]

Still failing with version '1.14.0.dev0+1021.bd60b4e':

$ SCIPY_XSLOW=1 python3 dev.py test -m full -t scipy.sparse.linalg.tests.test_onenormest -j 6
💻  ninja -C /home/warren/repos/git/forks/scipy/build -j12
ninja: Entering directory `/home/warren/repos/git/forks/scipy/build'
[2/2] Generating scipy/generate-version with a custom command
Build OK
💻  meson install -C build --only-changed
Installing, see meson-install.log...
Installation OK
SciPy from development installed path at: /home/warren/repos/git/forks/scipy/build-install/lib/python3.12/site-packages
Running tests for scipy version:1.14.0.dev0+1021.bd60b4e, installed at:/home/warren/repos/git/forks/scipy/build-install/lib/python3.12/site-packages/scipy
=================================== test session starts ===================================
platform linux -- Python 3.12.3, pytest-8.1.1, pluggy-1.5.0
rootdir: /home/warren/repos/git/forks/scipy
configfile: pytest.ini
plugins: timeout-2.3.1, xdist-3.5.0, hypothesis-6.100.1, cov-5.0.0
6 workers [7 items]     
......F                                                                             [100%]
======================================== FAILURES =========================================
_______________________ TestOnenormest.test_onenormest_table_6_t_1 ________________________
[gw3] linux -- Python 3.12.3 /home/warren/py3.12.3/bin/python3
scipy/sparse/linalg/tests/test_onenormest.py:183: in test_onenormest_table_6_t_1
    assert_(0.7 < proportion_exact < 0.8)
E   AssertionError
        A          = array([[ 1.0543604 +0.03847635j, -0.33208373-0.22861055j,
         0.40431944+0.0792976j , ..., -0.05681823-0.97139639...     -0.72327521+0.42464523j, ...,  0.86198444+1.35242015j,
        -1.4396941 -0.77624368j, -0.03960976-0.01487513j]])
        A_inv      = array([[0.29119074+0.65104849j, 0.58305885+0.67128938j,
        0.06722355+0.93823077j, ..., 0.6889593 +0.84052724j,
 ...
        0.45835081+0.39487713j, ..., 0.39236538+0.22490301j,
        0.37966209+0.48231849j, 0.52764845+0.90868658j]])
        est        = 286.67179442492255
        expected   = array([ 71.97806709,  70.56821698,  42.24911326, ...,  42.74919002,
        71.59204191, 287.08535285])
        i          = 4999
        itmax      = 5
        max_nresamples = 0
        n          = 100
        nexact     = 2781
        nmult_list = [5, 7, 4, 6, 7, 4, ...]
        nmults     = 4
        nresample_list = [0, 0, 0, 0, 0, 0, ...]
        nresamples = 0
        nsamples   = 5000
        observed   = array([ 71.97806709,  70.56821698,  35.59762277, ...,  42.74919002,
        63.85697738, 286.67179442])
        proportion_exact = 0.5562
        relative_errors = array([0.00000000e+00, 0.00000000e+00, 1.57435032e-01, ...,
       3.32423953e-16, 1.08043636e-01, 1.44054171e-03])
        self       = <scipy.sparse.linalg.tests.test_onenormest.TestOnenormest object at 0x73283e1c5c70>
        t          = 1
        underestimation_ratio = array([1.        , 1.        , 0.84256497, ..., 1.        , 0.89195636,
       0.99855946])
        underestimation_ratio_mean = 0.9414421444954721
        v          = array([0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
       0., 0., 0., 0., 0., 0., 0., 0., 0., ...0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
       0., 0., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.])
        w          = array([ 1.93101609+4.88613457e-02j,  0.95209743+2.99680743e-01j,
        3.85029705-1.27986728e+00j,  1.89775519+4.326...578313 -9.43259458e-01j,  1.87828784-6.22941868e+00j,
        0.80356713+1.08910034e+00j, -2.96027481+1.61222120e+00j])
================================= short test summary info =================================
FAILED scipy/sparse/linalg/tests/test_onenormest.py::TestOnenormest::test_onenormest_table_6_t_1 - AssertionError
=============================== 1 failed, 6 passed in 4.78s ===============================