On Tue, 30 Jan 2018, Yaroslav Halchenko wrote:
> On Tue, 30 Jan 2018, Adrian Bunk wrote: > > Control: reopen -1 > > On Tue, Jan 30, 2018 at 03:39:04AM +0000, Debian Bug Tracking System wrote: > > >... > > > pymvpa2 (2.6.4-1) unstable; urgency=medium > > >... > > > * debian/rules > > > - run tests under fixed MVPA_SEED=1 to avoid surprises, so should > > > resolve spurious FTBFS (Closes: #880793) > > >... > > Unfortunately this didn't fix the problem: > > https://buildd.debian.org/status/package.php?p=pymvpa2&suite=sid > Thanks for spotting it! > yeah -- I will look "inside" > Note that for mips64el, previous version included, failures happen on > aql builder but not on manda: > https://buildd.debian.org/status/logs.php?pkg=pymvpa2&arch=mips64el > so it is something relating to available resources on the box. might be > that we need to skip the test if resources are tight FTR, I think it is an issue to be fixed in numpy itself -- that error is caused by specifying step in np.float dtype, which leads to ValueError instead of ZeroDivisionError as in the case of regular float: (sid_mips64el-dchroot)yoh@eller:~/pymvpa2-2.6.4$ python -c 'import numpy as np; print(np.arange(-1, -1+0, 0.))' Traceback (most recent call last): File "<string>", line 1, in <module> ZeroDivisionError: float division by zero (sid_mips64el-dchroot)yoh@eller:~/pymvpa2-2.6.4$ python -c 'import numpy as np; print(np.arange(-1, -1+0, np.float16(0.)))' -c:1: RuntimeWarning: invalid value encountered in true_divide Traceback (most recent call last): File "<string>", line 1, in <module> ValueError: array is too big; `arr.size * arr.dtype.itemsize` is larger than the maximum possible size. on my x64 laptop it is yet another kind of fun with the same version of numpy 1:1.13.3-2 : $> python -c 'import numpy as np; print(np.arange(-1, -1+0, np.float(0.)))' Traceback (most recent call last): File "<string>", line 1, in <module> ZeroDivisionError: float division by zero $> python -c 'import numpy as np; print(np.arange(-1, -1+0, np.float64(0.)))' -c:1: RuntimeWarning: invalid value encountered in double_scalars [] $> python -c 'import numpy as np; print(np.arange(-1, -1+0, np.float32(0.)))' -c:1: RuntimeWarning: invalid value encountered in true_divide in current numpy git, unfortunately they release from releaase branches so hard to say 'version': pre-removal-numpybook-5689-gbb7b12672 it is a bit more consistent but still varying: $> PYTHONPATH=. python -c 'import numpy as np; print(np.arange(-1, -1+0, 0.))' Traceback (most recent call last): File "<string>", line 1, in <module> ZeroDivisionError: float division by zero $> PYTHONPATH=. python -c 'import numpy as np; print(np.arange(-1, -1+0, np.float(0.)))' Traceback (most recent call last): File "<string>", line 1, in <module> ZeroDivisionError: float division by zero $> PYTHONPATH=. python -c 'import numpy as np; print(np.arange(-1, -1+0, np.float64(0.)))' -c:1: RuntimeWarning: invalid value encountered in double_scalars Traceback (most recent call last): File "<string>", line 1, in <module> ValueError: arange: cannot compute length $> PYTHONPATH=. python -c 'import numpy as np; print(np.arange(-1, -1+0, np.float32(0.)))' -c:1: RuntimeWarning: invalid value encountered in true_divide Traceback (most recent call last): File "<string>", line 1, in <module> ValueError: arange: cannot compute length asked upstream: https://github.com/numpy/numpy/issues/10496 will fix up pymvpa with my patches tomorrow or so -- Yaroslav O. Halchenko Center for Open Neuroscience http://centerforopenneuroscience.org Dartmouth College, 419 Moore Hall, Hinman Box 6207, Hanover, NH 03755 Phone: +1 (603) 646-9834 Fax: +1 (603) 646-1419 WWW: http://www.linkedin.com/in/yarik
