Re: [sage-devel] Sage 4.4.1.alpha2: unable to produce binary for sage.math

[Willem Jan Palenstijn]

On Thu, Apr 29, 2010 at 07:01:10PM +1000, Minh Nguyen wrote:
> Hi folks,
>
> Sage 4.4.1.alpha2 builds OK on sage.math. I wasn't able to run
> doctests on the Sage library. When trying to wrap up a binary for
> sage.math, the reason became apparent:
>
> [mvngu@sage sage-4.4.1.alpha2]$ ./sage -bdist
> 4.4.1.alpha2-sage.math.washington.edu
> Sage works!
> *** glibc detected *** python: corrupted double-linked list:
> 0x0000000002f6a640 ***

I've ran valgrind on the binary you built on sage.math and get the warnings
below, which seem to point at polybori.

To test further, I upgraded to the new polybori package from #8192 at home in a
sage 4.4 install, but that didn't cause any trouble. I'm building alpha2 at
home now to see if I get the errors on exit there at all. Maybe it would also
be worth checking if the errors on sage.math disappear when downgrading
polybori.


Unfortunately it only actually crashes on exit on sage.math about once in 10-15
times.

-Willem Jan



==29134== Invalid read of size 4
==29134== at 0x30A93070: __tcf_1 (sp_counted_base_gcc_x86.hpp:50)
==29134== by 0x570010F: exit (in /lib/libc-2.7.so)
==29134== by 0x4BAAC2: handle_system_exit (pythonrun.c:1716)
==29134== by 0x4BACC9: PyErr_PrintEx (pythonrun.c:1126)
==29134== by 0x4BB925: PyRun_SimpleFileExFlags (pythonrun.c:1035)
==29134== by 0x4163B9: Py_Main (main.c:142)
==29134== by 0x56E91C3: (below main) (in /lib/libc-2.7.so)
==29134== Address 0x3240f490 is not stack'd, malloc'd or (recently) free'd
==29134==
==29134== Invalid read of size 8
==29134== at 0x30A93087: __tcf_1 (CCuddCore.h:208)
==29134== by 0x570010F: exit (in /lib/libc-2.7.so)
==29134== by 0x4BAAC2: handle_system_exit (pythonrun.c:1716)
==29134== by 0x4BACC9: PyErr_PrintEx (pythonrun.c:1126)
==29134== by 0x4BB925: PyRun_SimpleFileExFlags (pythonrun.c:1035)
==29134== by 0x4163B9: Py_Main (main.c:142)
==29134== by 0x56E91C3: (below main) (in /lib/libc-2.7.so)
==29134== Address 0x3180e918 is 8 bytes inside a block of size 64 free'd
==29134== at 0x4C227BC: operator delete(void*) (vg_replace_malloc.c:342)
==29134== by 0x570010F: exit (in /lib/libc-2.7.so)
==29134== by 0x4BAAC2: handle_system_exit (pythonrun.c:1716)
==29134== by 0x4BACC9: PyErr_PrintEx (pythonrun.c:1126)
==29134== by 0x4BB925: PyRun_SimpleFileExFlags (pythonrun.c:1035)
==29134== by 0x4163B9: Py_Main (main.c:142)
==29134== by 0x56E91C3: (below main) (in /lib/libc-2.7.so)
==29134==
==29134== Invalid write of size 8
==29134== at 0x30A93092: __tcf_1 (CCuddCore.h:208)
==29134== by 0x570010F: exit (in /lib/libc-2.7.so)
==29134== by 0x4BAAC2: handle_system_exit (pythonrun.c:1716)
==29134== by 0x4BACC9: PyErr_PrintEx (pythonrun.c:1126)
==29134== by 0x4BB925: PyRun_SimpleFileExFlags (pythonrun.c:1035)
==29134== by 0x4163B9: Py_Main (main.c:142)
==29134== by 0x56E91C3: (below main) (in /lib/libc-2.7.so)
==29134== Address 0x3180e918 is 8 bytes inside a block of size 64 free'd
==29134== at 0x4C227BC: operator delete(void*) (vg_replace_malloc.c:342)
==29134== by 0x570010F: exit (in /lib/libc-2.7.so)
==29134== by 0x4BAAC2: handle_system_exit (pythonrun.c:1716)
==29134== by 0x4BACC9: PyErr_PrintEx (pythonrun.c:1126)
==29134== by 0x4BB925: PyRun_SimpleFileExFlags (pythonrun.c:1035)
==29134== by 0x4163B9: Py_Main (main.c:142)
==29134== by 0x56E91C3: (below main) (in /lib/libc-2.7.so)

--
To post to this group, send an email to sage-...@googlegroups.com
To unsubscribe from this group, send an email to sage-devel+...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

[daveloeffler]

On Apr 29, 11:20 am, Willem Jan Palenstijn <w...@usecode.org> wrote:
> Unfortunately it only actually crashes on exit on sage.math about once in 10-15
> times.
>
> -Willem Jan

I get exactly the same behaviour on our 64-bit Ubuntu box: it fails
randomly every now and again, with tracebacks looking very similar to
what Minh gave above. I can't help noticing that the spkg/standard
directory contains both polybori-0.6.4.spkg and the previous version
polybori-0.6.3-20091028.spkg. Maybe it's trying to install both at
once or something?

David

[Willem Jan Palenstijn]

On Thu, Apr 29, 2010 at 10:20:25AM +0000, Willem Jan Palenstijn wrote:
> On Thu, Apr 29, 2010 at 07:01:10PM +1000, Minh Nguyen wrote:
> > Hi folks,
> >
> > Sage 4.4.1.alpha2 builds OK on sage.math. I wasn't able to run
> > doctests on the Sage library. When trying to wrap up a binary for
> > sage.math, the reason became apparent:
> >
> > [mvngu@sage sage-4.4.1.alpha2]$ ./sage -bdist
> > 4.4.1.alpha2-sage.math.washington.edu
> > Sage works!
> > *** glibc detected *** python: corrupted double-linked list:
> > 0x0000000002f6a640 ***
>
> I've ran valgrind on the binary you built on sage.math and get the warnings
> below, which seem to point at polybori.
>
> To test further, I upgraded to the new polybori package from #8192 at home in a
> sage 4.4 install, but that didn't cause any trouble. I'm building alpha2 at
> home now to see if I get the errors on exit there at all. Maybe it would also
> be worth checking if the errors on sage.math disappear when downgrading
> polybori.

Update:

After install the package from #8192 on top of a working 4.4 install and
running sage -ba, I get the crashes on exit (and valgrind warnings).

After downgrading polybori to 0.6.3.something in a crashing 4.4.alpha2 install
(and running sage -ba), the crashes (and valgrind warnings) disappear.


-Willem Jan

[Burcin Erocal]

This is probably my fault for disabling the static linking flags in the
new package. Unfortunately, I don't have the time at the moment to
fix this. It would be great if someone who already has the testing
environment set up could try enabling the static building again.

I also commented on the ticket here:

http://trac.sagemath.org/sage_trac/ticket/8192#comment:24

Many thanks for looking in to this. Apologies for the inconvenience.

Cheers,
Burcin