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

Re: Bug#790925: pandas: FTBFS on armhf and sparc: Bus error in test_append_frame_column_oriented



On Fri, Feb 19, 2016 at 03:10:08PM -0500, Yaroslav Halchenko wrote:
> 
> On Fri, 19 Feb 2016, Lennart Sorensen wrote:
> 
> > On Fri, Feb 19, 2016 at 12:11:35PM -0500, Yaroslav Halchenko wrote:
> > > Thanks for looking into it. Fwiw, for such cases better to use python-dbg so you could then py-bt and do more introspection within attached gdb. You would need -dbg packages for numpy etc
> 
> > I hadn't ever needed to debug python before, and since the error seemed
> > to be in C code, not python code, I wasn't sure looking for a python
> > way of debugging would help.
> 
> it would show python stack which might give better idea of the scope
> etc.  Not that it is a panacea but I found it handy from time to time
> 
> > I can't seem to guess how to use python-dbg though.  Certainly just
> > replacing python with python-dbg doesn't work in this case.
> 
> you just need to rebuild entire pandas using python-dbg, smth like
> 
> python-dbg setup.py build_ext --inplace
> 
> and then use it straight from there
> 
> P.S. for some packages I did provide -lib-dbg package with extensions
> built using python-dbg, sorry that I took a shortcut with pandas
> 
> Thanks again for looking into it

So I dist-upgraded, which brought in a new cython version, so I rebuilt.
This seems to have changed things.  We no longer get an alignment problem,
but we do segfault (of course maybe we segfault before hitting the
alignment problem now, given other tests are failing now that passed
before).

[New Thread 0xb4e02460 (LWP 11518)]
[New Thread 0xb4602460 (LWP 11519)]
[New Thread 0xb3e02460 (LWP 11520)]
test_append_frame_column_oriented (pandas.io.tests.test_pytables.TestHDFStore) ... 
Thread 1 "python2.7" received signal SIGSEGV, Segmentation fault.
0xb2c25ffc in ?? () from /usr/lib/python2.7/dist-packages/tables/hdf5extension.arm-linux-gnueabihf.so
(gdb) where
#0  0xb2c25ffc in ?? () from /usr/lib/python2.7/dist-packages/tables/hdf5extension.arm-linux-gnueabihf.so
#1  0x000419d4 in PyDict_GetItem (op=<optimized out>, key=0xb5cc1320) at ../Objects/dictobject.c:743
#2  0x734c0800 in ?? ()
Backtrace stopped: previous frame identical to this frame (corrupt stack?)

So either I didn't rebuild it cleanly or something, or the new version
has changed something.

In fact it fails initially here now:
test_complex_across_dimensions (pandas.io.tests.test_pytables.TestHDFComplexValues) ... Segmentation fault (core dumped)

I am going to do a clean extract and build again just to see if it does
it again.

It is starting to look like a bug in cython though as far as I can tell.

-- 
Len Sorensen


Reply to: