Bug 2046907

Created attachment 1856338 [details]
build.log

file build.log too big, will only attach last 32768 bytes

Created attachment 1856339 [details]
root.log

file root.log too big, will only attach last 32768 bytes

Created attachment 1856340 [details]
state.log

This bug appears to have been reported against 'rawhide' during the Fedora 36 development cycle.
Changing version to 36.

This fails on aarch64 and ppc64le. With different failures.


aarch64:


____________________________ test_cwt_batch[0-fft] _____________________________
axis = 0, method = 'fft'
    @pytest.mark.parametrize('axis, method', product([0, 1], ['conv', 'fft']))
    def test_cwt_batch(axis, method):
        dtype = np.float64
        time, sst = pywt.data.nino()
        n_batch = 8
        batch_axis = 1 - axis
        sst1 = np.asarray(sst, dtype=dtype)
        sst = np.stack((sst1, ) * n_batch, axis=batch_axis)
        dt = time[1] - time[0]
        wavelet = 'cmor1.5-1.0'
        scales = np.arange(1, 32)
    
        # non-batch transform as reference
        [cfs1, f] = pywt.cwt(sst1, scales, wavelet, dt, method=method, axis=axis)
    
        shape_in = sst.shape
        [cfs, f] = pywt.cwt(sst, scales, wavelet, dt, method=method, axis=axis)
    
        # shape of input is not modified
        assert_equal(shape_in, sst.shape)
    
        # verify same precision
        assert_equal(cfs.real.dtype, sst.dtype)
    
        # verify expected shape
        assert_equal(cfs.shape[0], len(scales))
        assert_equal(cfs.shape[1 + batch_axis], n_batch)
        assert_equal(cfs.shape[1 + axis], sst.shape[axis])
    
        # batch result on stacked input is the same as stacked 1d result
>       assert_equal(cfs, np.stack((cfs1,) * n_batch, axis=batch_axis + 1))
E       AssertionError: 
E       Arrays are not equal
E       
E       Mismatched elements: 61080 / 65472 (93.3%)
E       Max absolute difference: 2.3052588782060247e-15
E       Max relative difference: 4.507026999251287e-14
E        x: array([[[ 0.004042+0.026095j,  0.004042+0.026095j,  0.004042+0.026095j,
E                ...,  0.004042+0.026095j,  0.004042+0.026095j,  0.004042+0.026095j],
E               [-0.005609+0.012741j, -0.005609+0.012741j, -0.005609+0.012741j,...
E        y: array([[[ 0.004042+0.026095j,  0.004042+0.026095j,  0.004042+0.026095j,
E                ...,  0.004042+0.026095j,  0.004042+0.026095j,  0.004042+0.026095j],
E               [-0.005609+0.012741j, -0.005609+0.012741j, -0.005609+0.012741j,...
pywt/tests/test_cwt_wavelets.py:406: AssertionError
____________________________ test_cwt_batch[1-fft] _____________________________
axis = 1, method = 'fft'
    @pytest.mark.parametrize('axis, method', product([0, 1], ['conv', 'fft']))
    def test_cwt_batch(axis, method):
        dtype = np.float64
        time, sst = pywt.data.nino()
        n_batch = 8
        batch_axis = 1 - axis
        sst1 = np.asarray(sst, dtype=dtype)
        sst = np.stack((sst1, ) * n_batch, axis=batch_axis)
        dt = time[1] - time[0]
        wavelet = 'cmor1.5-1.0'
        scales = np.arange(1, 32)
    
        # non-batch transform as reference
        [cfs1, f] = pywt.cwt(sst1, scales, wavelet, dt, method=method, axis=axis)
    
        shape_in = sst.shape
        [cfs, f] = pywt.cwt(sst, scales, wavelet, dt, method=method, axis=axis)
    
        # shape of input is not modified
        assert_equal(shape_in, sst.shape)
    
        # verify same precision
        assert_equal(cfs.real.dtype, sst.dtype)
    
        # verify expected shape
        assert_equal(cfs.shape[0], len(scales))
        assert_equal(cfs.shape[1 + batch_axis], n_batch)
        assert_equal(cfs.shape[1 + axis], sst.shape[axis])
    
        # batch result on stacked input is the same as stacked 1d result
>       assert_equal(cfs, np.stack((cfs1,) * n_batch, axis=batch_axis + 1))
E       AssertionError: 
E       Arrays are not equal
E       
E       Mismatched elements: 61080 / 65472 (93.3%)
E       Max absolute difference: 2.3052588782060247e-15
E       Max relative difference: 4.507026999251287e-14
E        x: array([[[ 0.004042+0.026095j, -0.005609+0.012741j, -0.004250+0.074014j,
E                ..., -0.012503-0.087315j, -0.010488+0.00735j , -0.008362+0.051594j],
E               [ 0.004042+0.026095j, -0.005609+0.012741j, -0.004250+0.074014j,...
E        y: array([[[ 0.004042+0.026095j, -0.005609+0.012741j, -0.004250+0.074014j,
E                ..., -0.012503-0.087315j, -0.010488+0.00735j , -0.008362+0.051594j],
E               [ 0.004042+0.026095j, -0.005609+0.012741j, -0.004250+0.074014j,...
pywt/tests/test_cwt_wavelets.py:406: AssertionError



ppc64le:


________________________ test_accuracy_precomputed_cwt _________________________
    @uses_precomputed  # skip this case if pymatbridge + Matlab are being used
    @pytest.mark.slow
    def test_accuracy_precomputed_cwt():
        # Keep this specific random seed to match the precomputed Matlab result.
        rstate = np.random.RandomState(1234)
        # has to be improved
        epsilon = 2e-15
        epsilon32 = 1e-5
        epsilon_psi = 1e-15
        for wavelet in wavelets:
            with warnings.catch_warnings():
                warnings.simplefilter('ignore', FutureWarning)
                w = pywt.ContinuousWavelet(wavelet)
                w32 = pywt.ContinuousWavelet(wavelet,dtype=np.float32)
            psi = _load_matlab_result_psi(wavelet)
            _check_accuracy_psi(w, psi, wavelet, epsilon_psi)
    
            for N in _get_data_sizes(w):
                data = rstate.randn(N)
                data32 = data.astype(np.float32)
                scales_count = 0
                for scales in _get_scales(w):
                    scales_count += 1
                    coefs = _load_matlab_result(data, wavelet, scales_count)
>                   _check_accuracy(data, w, scales, coefs, wavelet, epsilon)
pywt/tests/test_matlab_compatibility_cwt.py:100: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
data = array([ -1.39349312e+00,   8.43433358e-01,  -7.48389740e-01,
         6.96962461e-01,  -2.31877711e-02,   3.14787623e-....61324483e+00,  -2.62751782e-03,
         2.24986089e+00,  -1.20909849e+00,   5.33308212e-02,
         2.85267990e+00])
w = pywt._extensions._pywt.ContinuousWavelet(name='cgau1'), scales = 1
coefs = array([[  5.88904952e-01 -4.22675773e-01j,
         -1.02734763e+00 +2.40005860e-02j,
          5.28246856e-01 +1.4842....74440177e+00 +1.48754656e-01j,
         -7.16701910e-01 -1.17576867e+00j,
         -1.58277318e+00 +7.72076610e-01j]])
wavelet = 'cgau1', epsilon = 2e-15
    def _check_accuracy(data, w, scales, coefs, wavelet, epsilon):
        # PyWavelets result
        coefs_pywt, freq = pywt.cwt(data, scales, w)
    
        # coefs from Matlab are from R2012a which is missing the complex conjugate
        # as shown in Eq. 2 of Torrence and Compo. We take the complex conjugate of
        # the precomputed Matlab result to account for this.
        coefs = np.conj(coefs)
    
        # calculate error measures
        err = coefs_pywt - coefs
        rms = np.real(np.sqrt(np.mean(np.conj(err) * err)))
    
        msg = ('[RMS > EPSILON] for Scale: %s, Wavelet: %s, '
               'Length: %d, rms=%.3g' % (scales, wavelet, len(data), rms))
>       assert_(rms < epsilon, msg=msg)
E       AssertionError: [RMS > EPSILON] for Scale: 1, Wavelet: cgau1, Length: 1000, rms=0.398
pywt/tests/test_matlab_compatibility_cwt.py:161: AssertionError

*** Bug 2099086 has been marked as a duplicate of this bug. ***

This blocks many other packages.

Following the FTBFS policy, asking for a needinfo.

for the record, with pywt updated to 1.3.0 is the error on ppc64le the same, I wonder if this is uncovering an numpy issue rather than one being contained in pywt ...

This also blocks some live spin currently.

> aarch64:
> […]
> E       Mismatched elements: 61080 / 65472 (93.3%)
> E       Max absolute difference: 2.3052588782060247e-15
> E       Max relative difference: 4.507026999251287e-14

Those are tiny differences. It looks like the upstream epsilon may just be a little too tight.

> ppc64le:
> […]
> >       assert_(rms < epsilon, msg=msg)
> E       AssertionError: [RMS > EPSILON] for Scale: 1, Wavelet: cgau1, Length: 1000, rms=0.398

That, on the other hand, is a large difference. Note that epsilon is 2e-15 here.

There is an issue open upstream about failures on ppc64le, although the exact problems reported there appear to be different: https://github.com/PyWavelets/pywt/issues/508

I've filed https://src.fedoraproject.org/rpms/python-pywt/pull-request/5.  I think the i686 failure might be a sporadic failure with xvfb-run.

Closing since the package now builds from source, even if skipped tests perhaps still represent some real issues on ppc64le.