New submission from Roland Lezuo:

class Int(ctypes.Structure):
           _fields_ = [ ("_i", ctypes.c_uint64),
                                   ("undef", ctypes.c_bool)]

class Int {
        public:
                Int();
                Int(uint64_t i);

                uint64_t _i;
                bool undef;
};

extern "C" Int foo(const Int& a , const Int& b , const Int& c)
{
    ret = ...
    return Int(ret);
}

(gdb) p ret
$3 = 16
(gdb) fin
Run till exit from #0  BVSignExtend (a=..., b=..., c=...) at foo.hpp:130
0x00007ffff784eea4 in ffi_call_unix64 () from 
/usr/lib/python2.7/lib-dynload/_ctypes.so
Value returned is $4 = {_i = 18577824, undef = 16}

My guess: The value 18577824 was not expected to be on the stack.

The following actions "solve" the problem:

1) add another int the class Int (after bool) and adopt _fields_ accordingly.
2) in foo C++ function:
   Int reti = Int(ret);
   return reti;

Because of the "above" solutions I strongly suspect a bug in ctypes.

----------
components: ctypes
messages: 171876
nosy: iroli
priority: normal
severity: normal
status: open
title: python2.7.3 struct misaligned when returned
versions: Python 2.7

_______________________________________
Python tracker <rep...@bugs.python.org>
<http://bugs.python.org/issue16117>
_______________________________________
_______________________________________________
Python-bugs-list mailing list
Unsubscribe: 
http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

Reply via email to