Details

Type: Story

Status: Done

Resolution: Done

Fix Version/s: None

Component/s: None

Labels:None

Story Points:1

Epic Link:

Team:Data Release Production
Description
Using a bleed version of the conda environment, meas_modelfit test TruncatedGaussianTestCase is failing
Error Message
AssertionError: True is not false : 1/4 elements differ with rtol=2.220446049250313e16, atol=2.220446049250313e16 2.500000000000001 != 2.5 (diff=8.881784197001252e16/2.500000000000001=3.5527136788004996e16)
Stacktrace
self = <test_truncatedGaussian.TruncatedGaussianTestCase testMethod=test2d> def test2d(self): if scipy is None: return for i in range(5): x = numpy.linspace(1, 1, 5) model = numpy.zeros((x.size, 2), dtype=float) model[:, 0] = x model[:, 1] = x**2 + x data = numpy.random.randn(x.size) + model[:, 0]*0.9 + model[:, 1]*1.1 q0 = 0.5*float(numpy.dot(data, data)) gradient = numpy.dot(model.transpose(), data) hessian = numpy.dot(model.transpose(), model) sigma = numpy.linalg.inv(hessian) > self.assertFloatsAlmostEqual(numpy.linalg.inv(sigma), hessian) tests/test_truncatedGaussian.py:161: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ../../stack/Linux64/utils/16.018g0a50484+1/python/lsst/utils/tests.py:735: in assertFloatsAlmostEqual testCase.assertFalse(failed, msg="\n".join(errMsg)) E AssertionError: True is not false : 1/4 elements differ with rtol=2.220446049250313e16, atol=2.220446049250313e16 E 2.500000000000001 != 2.5 (diff=8.881784197001252e16/2.500000000000001=3.5527136788004996e16)
{{ }}
{{ }}
The problem is just when building in Linux.
Numpy new version is 1.5.4. The old version is 1.4.5.
Scipy new version is 1.2.0. The old version is 1.1.0.
In attachment the sdiff between environemnts.