Skip to content

Test failures on nightly ARM64 wheel builds. #20426

Closed
AnacondaRecipes/numpy-feedstock
#43

Description

@charris

See https://app.travis-ci.com/github/MacPython/numpy-wheels/jobs/549013361.

FAILED core/tests/test_mem_policy.py::test_new_policy - AssertionError: asser...
FAILED core/tests/test_umath.py::TestComplexFunctions::test_loss_of_precision[complex256]

Activity

  1. mattip commented on Nov 21, 2021

    @mattip
    Member

    The test_new_policy failure is due to running the whole test suite with a new memory allocation policy and failing on the same TestComplexFunctions.test_loss_of_precision[complex256] test. So there is a single failure of arcsinh with np.long128 np.longcomplex. Just before the check is

            x_series = np.logspace(-20, -3.001, 200)
            x_basic = np.logspace(-2.999, 0, 10, endpoint=False)
    
            if glibc_older_than("2.19") and dtype is np.longcomplex:
                if (platform.machine() == 'aarch64' and bad_arcsinh()):
                    pytest.skip("Trig functions of np.longcomplex values known "
                                "to be inaccurate on aarch64 for some compilation "
                                "configurations.")
    
                # It's not guaranteed that the system-provided arc functions
                # are accurate down to a few epsilons. (Eg. on Linux 64-bit
                # So, give more leeway for long complex tests here:
                check(x_series, 50.0*eps)
            else:
    >           check(x_series, 2.1*eps)
    

    Did something change with arcsinh recently and we need to blocklist this glibc function?

  2. changed the title [-]Test failures on ARM64 wheel builds.[/-] [+]Test failures on nightly ARM64 wheel builds.[/+] on Nov 21, 2021
  3. charris commented on Nov 21, 2021

    @charris
    MemberAuthor

    That check was updated in #20274.

  4. charris commented on Nov 21, 2021

    @charris
    MemberAuthor

    Manylinux2014 has glibc 2.17, so should have used the lower precision check, but should probably just xfail/skip the test instead.

  5. added a commit that references this issue on Nov 21, 2021
    a5d002e
  6. charris commented on Nov 21, 2021

    @charris
    MemberAuthor

    Fixed. Travis-ci Python 3.9 arm64 has gone missing, but the working versions built wheels and passed the tests.

  7. added a commit that references this issue on Nov 21, 2021
    5a4ecc0
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions