[Numpy-tickets] [NumPy] #641: numpy.sqrt(numpy.array([-1.0], dtype=numpy.complex256)) segfaults

[NumPy]

#641: numpy.sqrt(numpy.array([-1.0], dtype=numpy.complex256)) segfaults
------------------------+---------------------------------------------------
 Reporter:  nils        |        Owner:  somebody
     Type:  defect      |       Status:  new     
 Priority:  high        |    Milestone:  1.0.5   
Component:  numpy.core  |      Version:  devel   
 Severity:  critical    |   Resolution:          
 Keywords:              |  
------------------------+---------------------------------------------------
Comment (by nils):

 I can also reproduce the non-reproducibility of the results on 32-bit
 systems using complex192.
 {{{
 >>> numpy.sqrt(numpy.array([-1.0],dtype = numpy.complex192))
 array([0.0+4.1241196e+1561j], dtype=complex192)
 >>> numpy.sqrt(numpy.array([-1.0],dtype = numpy.complex192))
 array([0.0+1.7102583e+2238j], dtype=complex192)
 >>> numpy.sqrt(numpy.array([-1.0],dtype = numpy.complex192))
 array([0.0+8.4925975e+1643j], dtype=complex192)
 >>> numpy.sqrt(numpy.array([-1.0],dtype = numpy.complex192))
 array([0.0+3.9101809e-4942j], dtype=complex192)
 >>> numpy.sqrt(numpy.array([-1.0],dtype = numpy.complex192))
 array([0.0+1.2078972e+4422j], dtype=complex192)
 }}}

-- 
Ticket URL: <http://scipy.org/scipy/numpy/ticket/641#comment:2>
NumPy <http://projects.scipy.org/scipy/numpy>
The fundamental package needed for scientific computing with Python.