Re: help to fix ftbfs issue #1052815 for python-sparse
On Sun, Oct 29, 2023 at 1:25 AM Bo YU <tsu.yubo@gmail.com> wrote:
>
> Hi,
>
> I am looking at the issue and trying to fix it.
>
> If upgrading to 0.14.0[0] directly does not fix the issue and it is in
> complicated situation. So below I will explain what I do.
>
> First to fix #1052815[1], wo need to adjust pybuild with custom way:
> ```
> PYBUILD_SYSTEM=custom \
> PYBUILD_TEST_ARGS="{interpreter} -m pytest -v" \
> ```
> Okay, it works. But unfortunately, the test cases failed.
>
> ```
> -- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
> =========================== short test summary info ============================
> FAILED sparse/tests/test_coo_numba.py::TestBasic::test_roundtrip - numba.core...
> FAILED sparse/tests/test_coo_numba.py::TestBasic::test_roundtrip_constant - n...
> FAILED sparse/tests/test_coo_numba.py::TestBasic::test_unpack_attrs - numba.c...
> FAILED sparse/tests/test_coo_numba.py::TestBasic::test_repack_attrs - numba.c...
> FAILED sparse/tests/test_dot.py::test_tensordot[coo-gcxs-a_shape1-b_shape1-axes1]
> FAILED sparse/tests/test_dot.py::test_tensordot[coo-gcxs-a_shape3-b_shape3-axes3]
> FAILED sparse/tests/test_dot.py::test_tensordot[coo-gcxs-a_shape7-b_shape7-axes7]
> FAILED sparse/tests/test_dot.py::test_tensordot[coo-gcxs-a_shape9-b_shape9-0]
> FAILED sparse/tests/test_dot.py::test_tensordot[gcxs-coo-a_shape1-b_shape1-axes1]
> ==== 9 failed, 5402 passed, 36 xfailed, 49 xpassed, 3729 warnings in 44.59s ====
>
> ```
>
> The upstream has reported the bug[2] then I applied the workaround in
> issue 8993[3], but it still got some cases failed:
> ```
> =========================== short test summary info ============================
> FAILED sparse/tests/test_coo_numba.py::TestBasic::test_roundtrip - numba.core...
> FAILED sparse/tests/test_coo_numba.py::TestBasic::test_roundtrip_constant - n...
> FAILED sparse/tests/test_coo_numba.py::TestBasic::test_unpack_attrs - numba.c...
> FAILED sparse/tests/test_coo_numba.py::TestBasic::test_repack_attrs - numba.c...
> ```
>
> It seems the bug was raised by numba and fixed in 0.57 but not sure
> why it fails on Debian here.
>
> But maybe I am wrong to get the conclusion above because from upstream
> issue 393[4], the problem maybe
> due to entry points of sparse. But I have no idea how to fix the issue
> properly. For example, If I install
> python3-hyperspy[5] to run tests and it will pass. I suspect
> python-sparse was installed properly when installing
> python3-hyperspy which depend on python-sparse and installing
> python3-hyperspy should not to as workaround at all.
Now the issue should be fixed by changing the order in which tests are executed,
which was put after installing. The only trouble is still one test
case failed on i386
even applied upstream workaround[0]. I have reported this to upstream also[1]
[0]: https://github.com/pydata/sparse/pull/486
[1]: https://github.com/pydata/sparse/issues/490
BR,
Bo
Reply to: