[Numpy-tickets] [NumPy] #890: test failures with mingw when SIZEOF_LONG_DOUBLE == 8

[NumPy]

#890: test failures with mingw when SIZEOF_LONG_DOUBLE == 8
------------------------+---------------------------------------------------
 Reporter:  cdavid      |        Owner:  cdavid 
     Type:  defect      |       Status:  closed 
 Priority:  normal      |    Milestone:  1.3.0  
Component:  numpy.core  |      Version:  none   
 Severity:  normal      |   Resolution:  wontfix
 Keywords:              |  
------------------------+---------------------------------------------------
Changes (by cdavid):

  * status:  new => closed
  * resolution:  => wontfix

Comment:

 Forcing SIZEOF_LONG_DOUBLE to 8 cannot work: mingw defines its own long
 double math functions like tanl and co, which assume long double to be 12
 bytes. The declarations then become incompatible.

 One solution may be to use our own npy_ math functions, with npy_ types
 instead of plain C types. Instead, I used a workaround for long double
 print on mingw (see r6215->r6220).

-- 
Ticket URL: <http://scipy.org/scipy/numpy/ticket/890#comment:4>
NumPy <http://projects.scipy.org/scipy/numpy>
The fundamental package needed for scientific computing with Python.