[NumPy-Tickets] [NumPy] #1671: scalars don't raise overflow/underflow/divbyzero/invalid

[NumPy Trac]

#1671: scalars don't raise overflow/underflow/divbyzero/invalid
------------------------+---------------------------------------------------
 Reporter:  m-paradox   |       Owner:  somebody
     Type:  defect      |      Status:  new     
 Priority:  normal      |   Milestone:  2.0.0   
Component:  numpy.core  |     Version:  devel   
 Keywords:              |  
------------------------+---------------------------------------------------

Comment(by m-paradox):

 I dug in a little bit, and the problem appears to be the compiler
 optimizing too liberally.  If I add a printf between the operation and
 PyUFunc_getfperr() in scalarmathmodule.c.src, it works.

 {{{
     @name@_ctype_@oper@(arg1, arg2, &out);
     printf("Result is %g\n", out);
 #endif

 #if @fperr@
     /* Check status flag.  If it is set, then look up what to do */
     retstatus = PyUFunc_getfperr();
 }}}

 I'm guessing the compiler thinks it can reorder the float operation to
 after PyUFunc_getfperr, since PyUFunc_getfperr is independent of it except
 for the float exception state.

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/1671#comment:1>
NumPy <http://projects.scipy.org/numpy>
My example project