[NumPy-Tickets] [NumPy] #1757: Segfault in pandas.

NumPy Trac numpy-tickets@scipy....
Sun Mar 6 11:07:07 CST 2011


#1757: Segfault in pandas.
------------------------------+---------------------------------------------
 Reporter:  charris           |       Owner:  somebody
     Type:  defect            |      Status:  new     
 Priority:  high              |   Milestone:  1.6.0   
Component:  Other             |     Version:  devel   
 Keywords:  segfault, pandas  |  
------------------------------+---------------------------------------------

Comment(by charris):

 The immediate problem is

 {{{

 #0  discover_dimensions (s=0x0, maxndim=0x7fffffffd14c, d=0x7fffffffd440,
     check_it=1, stop_at_string=1, stop_at_tuple=0, out_is_object=
     0x7fffffffd3cc) at numpy/core/src/multiarray/ctors.c:666
         e = <value optimized out>
         n = <value optimized out>
         i = <value optimized out>
         buffer_view = {buf = 0xb, obj =
     <unknown at remote 0x-53425ff84629df63>, len = 269111108832,
           itemsize = 11584208, readonly = -1468220032, ndim = 62, format =
     0x1 <Address 0x1 out of bounds>, shape = 0x0, strides = 0x3ea8449025,
           suboffsets = 0x7ffff0b9f710, smalltable = {269111386692,
 11584208},
           internal = 0x7ffff0b9f510}

 }}}

 That is, s is a null pointer to a PyObject. This comes from line 825 in
 ctors.c


 {{{
         e = PySequence_GetItem(s, 0);
         r = discover_dimensions(e, &maxndim_m1, d + 1, check_it,
                                         stop_at_string, stop_at_tuple,
                                         out_is_object);

 }}}

 So numpy should to check the value of e and raise an error if it is the
 Null pointer and also check the value of r for an error return.

 So why is it the Null pointer? None may be involved.

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/1757#comment:1>
NumPy <http://projects.scipy.org/numpy>
My example project


More information about the NumPy-Tickets mailing list