[Cython] NumPy output works on 32bit not on 64bit

[Dag Sverre Seljebotn]

Marcin Cieślik wrote:
> Hello,
> 
> Does anyone know why this code fails on a 64bit linux machine (it is
> working on 32 just fine):
> The compiler complains:
> 
> bit.c:792: warning: passing argument 3 of '*(PyArray_API + 744u)' from
> incompatible pointer type
> 
> and I get a segmentation fault
> 
> If it is not an issue with Cython could you please direct me
> somewhere, if have almost no C/C++ skill.

See comments below.

> #####
> cimport numpy as np
> ctypedef np.double_t DTYPE_t

This isn't related to the problem, but when you are using the NPY_DOUBLE 
etc. constants directly you should instead do

ctypedef np.npy_double DTYPE_t

This is because there isn't a 1:1 correspondance between the "shorthand" 
types and the NPY_TYPE constants (see bottom of 
Cython/Includes/numpy.pxd for details. BTW, patches for numpy.pxd to add 
  the Py_Array_SimpleNew... calls are welcome if submitted).

> from numpy cimport NPY_DOUBLE, NPY_UINT
> from stdlib cimport malloc, realloc, free
> 
> cdef extern from "arrayobject.h":
>     cdef object PyArray_SimpleNewFromData(int nd, int *dims,\
>                                             int typenum, void *data)

dims should be declared with the type npy_intp instead (and this is also 
what your compiler complained about).

>     cdef void import_array()
> import_array()
> 
> print sizeof(DTYPE_t)
> print sizeof(void*)
> print sizeof(DTYPE_t*)
> 
> cdef int i, k = 100

Here is the problem. If you make k an npy_intp instead then it will be a 
64-bit int. So:

cdef np.npy_intp k = 100

This wouldn't matter in many circumstances but as you are taking the 
address to k the call to SimpleNewFromData it is crucial. (So NumPy 
thought your array had a random billion elements coming from whatever 
bits were after the 32-bit "k" variable on the stack...)

-- 
Dag Sverre