[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bus error on armhf (Was: Bug#877419: Bug#877700: RM: pandas [arm64 armel armhf mips mips64el mipsel s390x] ...)



On Tue, Oct 24, 2017 at 02:15:15PM +0200, Andreas Tille wrote:
> Hi,

> I made some progress with excluding some tests on some architectures.  The
> last target I was trying to tackle was s390x which builds now.  Not all
> architectures have finished [1] currently but armhf has now failed with

> ...
> ../debian/tmp/usr/lib/python2.7/dist-packages/pandas/tests/io/test_pytables.py::TestHDFStore::test_append_all_nans PASSED
> ../debian/tmp/usr/lib/python2.7/dist-packages/pandas/tests/io/test_pytables.py::TestHDFStore::test_append_frame_column_oriented Bus error
> XIO:  fatal IO error 11 (Resource temporarily unavailable) on X server ":99"

> debian/rules:112: recipe for target 'python-test2.7' failed
>       after 15 requests (13 known processed) with 0 events remaining.

> make[1]: *** [python-test2.7] Error 135

> (see build log[2]).  My suspicion is that this is not really an error in
> the test suite but some problem with the autobuilder?  Could somebody
> please have a look?  If not the only chance I see is to deactivate armhf
> for pandas.

test_api_default_format (pandas.io.tests.test_pytables.TestHDFStore) ... ok
test_append (pandas.io.tests.test_pytables.TestHDFStore) ... ok
test_append_all_nans (pandas.io.tests.test_pytables.TestHDFStore) ... Bus error (core dumped)
XIO:  fatal IO error 11 (Resource temporarily unavailable) on X server ":99"
      after 15 requests (13 known processed) with 0 events remaining.
debian/rules:114: recipe for target 'python-test2.7' failed
make[1]: *** [python-test2.7] Error 135

(https://launchpad.net/ubuntu/+source/pandas/0.19.2-5.1ubuntu4/+build/13337694)

This is not a problem with the Debian autobuilder.  Something is attempting
an unaligned access, which is not portable on armhf.  Historically, the
Linux armhf kernel would fix up these unaligned accesses by default; but
Android kernel builds, for example, would not, and Linux arm64 kernels also
do not, and while we could enforce that Debian kernels fix up unaligned
access on armhf, the reason they don't all do this is that the fixups are
expensive and it's better to fix the code.

(I am somewhat surprised to see that this particular package build is an
armhf build on an "armel" host; but I'm not sure that's material here.)

-- 
Steve Langasek                   Give me a lever long enough and a Free OS
Debian Developer                   to set it on, and I can move the world.
Ubuntu Developer                                    http://www.debian.org/
slangasek@ubuntu.com                                     vorlon@debian.org

Attachment: signature.asc
Description: PGP signature


Reply to: