New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
multiple test failures on s390x in fedora 23 #6338
Comments
Scipy 0.16 is outdated and is unlikely to receive any updates. |
the issue still happens in latest scipy-0.18.0rc1 |
Thanks for testing it! |
Note that the platform is s390x, which I think very few work with.
|
the following tests failed on platform s390x (tested with scipi-0.18.0rc1)
|
The first two and the last one are the only failures I don't recognize. The |
FWIW the so most likely the failures are these tests: https://github.com/scipy/scipy/blob/master/scipy/special/tests/test_data.py#L221 (I think this is what @rgommers was mentioning? Sorry if I read that wrong.) |
@person142 indeed. Finding the data files is the first pain, but worse is then to find the 1 or 2 out of 160 points that is failing and figuring out why that's platform-dependent and/or intermittent. |
The failure message prints the exact points that fail, what value was produced, what was expected, and by how much it is off. I'm not sure if there's something that can be improved here. |
It's almost magical that there are only a couple of reasonably mild failures on a platform we hardly support and definitely do not test on. I'm sure we won't object to patches which fix failures specific to s390x and do not break more common architectures. |
Of course, if someone feels like setting up an s390x buildbot or periodically run tests on this arch, that would be appreciated :-). |
Right, sorry I described that quite wrong. Looking at it in more detail again, the trouble I had was after location the failure working back to a standalone example to reproduce it. All the |
@rgommers: yes, what i'm saying is that the error message contains the numbers that are needed to produce the failure standalone. IOW, it is an allclose() check. |
@ev-br: i'm willing to test scipy for s390x and another platform like powerpc. Is there any idea how to fix these issues? |
@ngothan Great that you're volunteering! To be clear: I don't think we'll be holding the release for s390x specific failures when more common platforms (Windows, OS X and Linux on x86) are clean. Now, like Ralf said, some of the failures you report are seen on other platforms. See e.g. open gh issues with 0.18.0 tag and PRs with backport-candidate tag (both open and closed). The second failure looks to be some endianness issue, |
scipy-0.16 does not build on s390x, it just crash in the tests by running PYTHONPATH=$RPM_BUILD_ROOT/usr/lib64/python3.5/site-packages /usr/bin/python3 -c "import scipy; scipy.test('full', verbose=2)"
The text was updated successfully, but these errors were encountered: