[Wheel-builders] [manylinux] how to build numpy/scipy wheels for x86

Fri Apr 22 04:17:17 EDT 2016

Thank you very much Matthew, they work on our wheezy-i686 systems.
Interestingly, I got two failures reported from scipy.test(), not sure if I
should maybe report that somewhere. I'll paste it here in case somebody is
interested:

======================================================================
FAIL: test_data.test_boost(<Data for btdtri_comp:
ibeta_inv_data_ipp-ibeta_inv_data>,)
----------------------------------------------------------------------
Traceback (most recent call last):
  File
"/home/testsys/test/tc/local/lib/python2.7/site-packages/nose/case.py",
line 197, in runTest
    self.test(*self.arg)
  File
"/home/testsys/test/tc/local/lib/python2.7/site-packages/scipy/special/tests/test_data.py",
line 481, in _test_factory
    test.check(dtype=dtype)
  File
"/home/testsys/test/tc/local/lib/python2.7/site-packages/scipy/special/_testutils.py",
line 292, in check
    assert_(False, "\n".join(msg))
  File
"/home/testsys/test/tc/local/lib/python2.7/site-packages/numpy/testing/utils.py",
line 71, in assert_
    raise AssertionError(smsg)
AssertionError:
Max |adiff|: 7.90146e-13
Max |rdiff|: 8.34163e-07
Bad results (1 out of 1210) for the following points (in output 0):
        0.00022049853578209877                    107380.34375
  0.135563462972641 =>         6.193264077428975e-293 !=
6.193269243624209e-293  (rdiff          8.341628679530342e-07)

======================================================================
FAIL: test_orthogonal.test_la_roots
----------------------------------------------------------------------
Traceback (most recent call last):
  File
"/home/testsys/test/tc/local/lib/python2.7/site-packages/nose/case.py",
line 197, in runTest
    self.test(*self.arg)
  File
"/home/testsys/test/tc/local/lib/python2.7/site-packages/scipy/special/tests/test_orthogonal.py",
line 722, in test_la_roots
    vgq(rootf(50), evalf(50), weightf(50), 0., np.inf, 100, atol=1e-13)
  File
"/home/testsys/test/tc/local/lib/python2.7/site-packages/scipy/special/tests/test_orthogonal.py",
line 286, in verify_gauss_quad
    assert_allclose(vv, np.eye(N), rtol, atol)
  File
"/home/testsys/test/tc/local/lib/python2.7/site-packages/numpy/testing/utils.py",
line 1391, in assert_allclose
    verbose=verbose, header=header)
  File
"/home/testsys/test/tc/local/lib/python2.7/site-packages/numpy/testing/utils.py",
line 733, in assert_array_compare
    raise AssertionError(msg)
AssertionError:
Not equal to tolerance rtol=1e-15, atol=1e-13

(mismatch 0.08%)
 x: array([[  1.000000e+00,   4.552139e-16,  -4.564824e-16, ...,
          1.224186e-16,   9.222704e-17,  -1.530426e-17],
       [  4.552139e-16,   1.000000e+00,   1.328770e-15, ...,...
 y: array([[ 1.,  0.,  0., ...,  0.,  0.,  0.],
       [ 0.,  1.,  0., ...,  0.,  0.,  0.],
       [ 0.,  0.,  1., ...,  0.,  0.,  0.],...

----------------------------------------------------------------------
Ran 20344 tests in 167.768s

FAILED (KNOWNFAIL=98, SKIP=1700, failures=2)
<nose.result.TextTestResult run=20344 errors=0 failures=2>

~/test$ pip freeze
Error [Errno 2] No such file or directory while executing command git
rev-parse
backports-abc==0.4
backports.ssl-match-hostname==3.5.0.1
bokeh==0.11.1
certifi==2016.2.28
check-manifest==0.31
devpi-client==2.4.1
devpi-common==2.0.8
futures==3.0.5
Jinja2==2.8
MarkupSafe==0.23
ngc-decoder==1.2.1
nose==1.3.7
numpy==1.11.0
pkginfo==1.2.1
pluggy==0.3.1
py==1.4.31
python-dateutil==2.5.3
PyYAML==3.11
requests==2.9.1
scipy==0.17.0
singledispatch==3.4.0.3
six==1.10.0
tornado==4.3
tox==2.3.1
virtualenv==15.0.1

~/test$ uname -a
Linux am2-nur-d732-02 3.2.0-4-686-pae #1 SMP Debian 3.2.57-3 i686 GNU/Linux

~/test$ lsb_release -a
No LSB modules are available.
Distributor ID: Debian
Description: Debian GNU/Linux 7.5 (wheezy)
Release: 7.5
Codename: wheezy

Cheers,
Matthias

Matthew Brett <matthew.brett at gmail.com> schrieb am Fr., 22. Apr. 2016 um
06:22 Uhr:

> Hi,
>
> On Thu, Apr 21, 2016 at 3:04 PM, Matthew Brett <matthew.brett at gmail.com>
> wrote:
> > On Thu, Apr 21, 2016 at 2:13 PM, Matthias Hafner <hafner87 at gmail.com>
> wrote:
> >> Hi Matthew,
> >>
> >> Thanks for pointing me to your scripts, that was maybe what I missed.
> >> However, if you could add i686 to Travis, I guess that means it would be
> >> uploaded to pypi? That would be awesome - otherwise I'd have to automate
> >> that myself.
> >
> > They don't get uploaded to pypi automatically, and they would be a bit
> > more difficult to test, but it would certainly make it easier to
> > provide them...
>
> I built i686 cython, numpy, scipy wheels.  You can test with something
> like:
>
> python -m pip install --upgrade pip  # to get latest pip
> pip install -f https://nipy.bic.berkeley.edu/manylinux numpy scipy
>
> Cheers,
>
> Matthew
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/wheel-builders/attachments/20160422/8877699b/attachment.html>