[Debian-pan-maintainers] Bug#1018811: Three new test issues preventing upload of patch (Was: pyfai: autopkgtest regression on armel and i386)

[Jerome Kieffer]

Hi Andreas,

Thanks for taking care of the packaging of pyFAI. I had a look at what
the logs and apparently you are packaging the version 0.21.3 ... while
I released the verison 2023.1 last week:
https://pypi.org/project/pyfai/

I just checked on the i386-debian sid chroot I have and none of the 3 test you found were failing (they are skipped)... but there is another one failing:

======================================================================
FAIL: test (pyFAI.test.test_error_model.TestErrorModel.test)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/tmp/pyFAI/build/lib/python3/dist-packages/pyFAI/test/test_error_model.py", line 121, in test
    self.assertGreaterEqual(cormap(ref.__getattribute__(array), res.__getattribute__(array)), epsilon, f"array {array} matches for {k} vs numpy")
AssertionError: 0 not greater than or equal to 0.002 : array
sum_normalization matches for ('poisson', 'opencl', 'integrate') vs numpy

I double checked and this failure is relative to the PoCL
implemantation used on my i386-chroot. If OpenCL is disabled for the
tests in debian, this test should pass OK.

Cheers,

Jerome

On Fri, 20 Jan 2023 22:20:57 +0100
Andreas Tille <andreas at an3as.eu> wrote:

> Hi Jerome,
> 
> I've applied the suggested patch to relax the tests on 32 bit architectures.
> Unfortunately there are new test suite errors as you can see in Salsa CI[1]:
> 
> 
> ======================================================================
> FAIL: test_count_csr (pyFAI.test.test_histogram.TestHistogram2d.test_count_csr)
> Test that the pixel count and the total intensity is conserved
> ----------------------------------------------------------------------
> Traceback (most recent call last):
>   File "/builds/science-team/pyfai/debian/output/source_dir/.pybuild/cpython3_3.11_pyfai/build/pyFAI/test/test_histogram.py", line 339, in test_count_csr
>     self.assertTrue(delta == 0, msg="check all pixels were counted")
> AssertionError: False is not true : check all pixels were counted
> ======================================================================
> FAIL: test_numpy_vs_cython_vs_csr_2d (pyFAI.test.test_histogram.TestHistogram2d.test_numpy_vs_cython_vs_csr_2d)
> Compare numpy histogram with cython simple implementation
> ----------------------------------------------------------------------
> Traceback (most recent call last):
>   File "/builds/science-team/pyfai/debian/output/source_dir/.pybuild/cpython3_3.11_pyfai/build/pyFAI/test/test_histogram.py", line 373, in test_numpy_vs_cython_vs_csr_2d
>     self.assertTrue(delta_max <= self.err_max_cnt, "pixel count difference numpy/csr : max delta=%s" % delta_max)
> AssertionError: False is not true : pixel count difference numpy/csr : max delta=8.0
> ======================================================================
> FAIL: test_2d_nosplit (pyFAI.test.test_csr.TestCSR.test_2d_nosplit)
> ----------------------------------------------------------------------
> Traceback (most recent call last):
>   File "/builds/science-team/pyfai/debian/output/source_dir/.pybuild/cpython3_3.11_pyfai/build/pyFAI/test/test_csr.py", line 195, in test_2d_nosplit
>     self.assertLess(error.mean(), 1e-3, "img are almost the same")
> AssertionError: 244.15215998872887 not less than 0.001 : img are almost the same
> ======================================================================
> 
> Any idea how to fix these?
> 
> Kind regards
>    Andreas.
> 
> 
> [1] https://salsa.debian.org/science-team/pyfai/-/jobs/3826387
> 
> -- 
> http://fam-tille.de