[SciPy-user] Bus error on Intel mac pro w/Leopard, gcc 4.2, numpy from svn head

Robert Kern robert.kern@gmail....
Mon Jun 16 11:49:12 CDT 2008


On Mon, Jun 16, 2008 at 10:58, Anand Patil
<anand.prabhakar.patil@gmail.com> wrote:
> Hi all,
>
> I'm getting a bus error using the numpy head on an Intel Mac Pro
> running Leopard with gcc 4.2. It happens intermittently 1-2 hours into
> a long computation, so I haven't been able to boil the situation
> down... I'm hoping the gdb output and crash report below are enough to
> go on.
>
> I built Python from source using gcc 4.2, so I had to get rid of the
> 'no-cpp-precomp' and 'Wno-long-double' options.
>
> The really long call stack in the crash report below makes me wonder
> whether the problem is in numpy itself, or other code is misusing
> numpy somehow. I'm using several f2py'ed Fortran modules from PyMC and
> also PyTables here.

It could be either. The long chain of array_dealloc() calls means that
you have a view of a view of a view ... of and array. When the tail
end of the view chain gets decrefed to 0 and deallocated, it will
decref its reference to the array it is a view of (myarray.base, if
you want to check this out at the Python level). If that was the only
reference, it gets deallocated and decrefs its .base array.

It is possible, though unlikely, that a cycle got formed somehow.

Does your program crash using the www.python.org binary built with gcc 4.0.1?

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


More information about the SciPy-user mailing list