Bug#1014884: statsmodels: FTBFS on mipsel (pythran shifted tolerances)

[Drew Parsons]

Source: statsmodels
Version: 0.13.2+dfsg-2
Severity: serious
Justification: FTBFS
Control: forwarded -1 https://github.com/statsmodels/statsmodels/issues/8341

mipsel is failing to build statsmodels 0.13.2+dfsg-2.

Problem seems to be another shift in tolerance requirements triggered
by scipy using pythran, similar to Bug#1014278.

In this case the discrepancy is mild, gets error 1.12443388e-12,
but expects 1e-12. Could be worked around by relaxing the tolerance to 2e-12

The error message is

____________________ TestGMMStOneiterOLS_Linear.test_basic _____________________

self = <statsmodels.sandbox.regression.tests.test_gmm.TestGMMStOneiterOLS_Linear object at 0x6c0ce0a0>

    def test_basic(self):
        res1, res2 = self.res1, self.res2
        # test both absolute and relative difference
        rtol,  atol = self.params_tol
        assert_allclose(res1.params, res2.params, rtol=rtol, atol=0)
>       assert_allclose(res1.params, res2.params, rtol=0, atol=atol)
E       AssertionError: 
E       Not equal to tolerance rtol=0, atol=1e-12
E       
E       Mismatched elements: 1 / 13 (7.69%)
E       Max absolute difference: 1.12443388e-12
E       Max relative difference: 1.53607078e-12
E        x: array([ 6.195478e-02,  2.712120e-03,  3.083947e-02,  4.216306e-02,
E              -9.629347e-02,  1.328993e-01, -5.420948e-02,  8.058085e-02,
E               2.075915e-01,  2.282237e-01,  2.226915e-01,  3.228747e-01,
E               4.235357e+00])
E        y: array([ 6.195478e-02,  2.712120e-03,  3.083947e-02,  4.216306e-02,
E              -9.629347e-02,  1.328993e-01, -5.420948e-02,  8.058085e-02,
E               2.075915e-01,  2.282237e-01,  2.226915e-01,  3.228747e-01,
E               4.235357e+00])

../.pybuild/cpython3_3.10_statsmodels/build/statsmodels/sandbox/regression/tests/test_gmm.py:207: AssertionError