[NumPy-Tickets] [NumPy] #2159: add.accumulate w/ out array same as input causes memcpy overlap

[NumPy Trac]

#2159: add.accumulate w/ out array same as input causes memcpy overlap
------------------------+---------------------------------------------------
 Reporter:  shauncutts  |       Owner:  somebody   
     Type:  defect      |      Status:  new        
 Priority:  normal      |   Milestone:  Unscheduled
Component:  numpy.core  |     Version:  1.6.1      
 Keywords:              |  
------------------------+---------------------------------------------------
 source looks like:

  N.add.accumulate( count, axis = 1, out = count )

 where count is 2-dim array.

 valgrind reports:

 ==6607== Source and destination overlap in memcpy(0x23cdfdc0, 0x23cdfdc0,
 8)
 ==6607==    at 0x4C23ADE: memcpy (mc_replace_strmem.c:838)
 ==6607==    by 0xAA24AAA: PyUFunc_ReductionOp (ufunc_object.c:3263)
 ==6607==    by 0xAA25CD9: PyUFunc_GenericReduction (ufunc_object.c:3482)
 ==6607==    by 0x41BAFC: PyObject_Call (abstract.c:2522)
 ==6607==    by 0x17684B64:
 __pyx_pw_9factfiber_4stat_6pmodel_1c_9partition_9find_emm_partition
 (partition.c:8867

 This was a call from cython, but I don't think that is important.

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/2159>
NumPy <http://projects.scipy.org/numpy>
My example project