[Numpy-discussion] Segmentation fault during tests with Python 2.7.2 on Debian 6?

[Bryan Van de Ven]

On 4/12/12 8:41 PM, Charles R Harris wrote:
> It seems that python2.7 is far, far, too recent to be part of Debian 
> 6. I mean, finding python 2.7 in recent Debian stable would be like 
> finding an atomic cannon in a 1'st dynasty Egyptian tomb. So it is in 
> testing, but for replication I like to know where you got it.
>
> Chuck

Just to add a data point, Maggie built python 2.7.2 on Debian 32 and 64 
bit VMs yesterday and saw the same error. Interestingly, it segfaults in 
different places:

32 bit
====
Program received signal SIGSEGV, Segmentation fault.
PyObject_Malloc (nbytes=48) at Objects/obmalloc.c:788
788                if ((pool->freeblock = *(block **)bp) != NULL) {

#0  PyObject_Malloc (nbytes=48) at Objects/obmalloc.c:788
#1  0x08095ad5 in _PyObject_New (tp=0xb7b80180) at Objects/object.c:244
#2  0xb7af433f in PyArray_DescrNew (type_num=18)
     at numpy/core/src/multiarray/descriptor.c:1457
#3  PyArray_DescrNewFromType (type_num=18)
     at numpy/core/src/multiarray/descriptor.c:1106
#4  0xb7b1631c in PyArray_DTypeFromObjectHelper (obj=<value optimized out>,
     maxdims=<value optimized out>, out_contains_na=0xbfffbae0,
     out_dtype=0xbfffbae8, string_type=18)
     at numpy/core/src/multiarray/common.c:259


64 bit
====
Program received signal SIGSEGV, Segmentation fault.
PyType_IsSubtype (a=0x2d00330030002d, b=0x7ffff6624b20) at 
Objects/typeobject.c:1132
1132        if (!(a->tp_flags & Py_TPFLAGS_HAVE_CLASS))
(gdb) print a
$1 = (PyTypeObject *) 0x2d00330030002d

#0  PyType_IsSubtype (a=0x2d00330030002d, b=0x7ffff6624b20) at 
Objects/typeobject.c:1132
#1  0x00007ffff639b8ad in STRING_setitem (op=0x134c930, ov=0x154f630 "", 
ap=0x144f1d0)
     at numpy/core/src/multiarray/arraytypes.c.src:476
#2  0x00007ffff639d466 in UNICODE_to_STRING (ip=0x1111f00 "2", 
op=0x154f630 "", n=<value optimized out>,
     aip=0x14fb5c0, aop=0x144f1d0) at 
numpy/core/src/multiarray/arraytypes.c.src:1378
#3  0x00007ffff62fa27f in _strided_to_strided_contig_align_wrap (dst=0x1 
<Address 0x1 out of bounds>,
     dst_stride=<value optimized out>, src=0x7fffffff9c40 "\003", 
src_stride=<value optimized out>,
     N=140737488329280, src_itemsize=40, data=0x154f5c0) at 
numpy/core/src/multiarray/dtype_transfer.c:354

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/numpy-discussion/attachments/20120413/bd4ae98f/attachment.html>