[cable] generated python wrapper signals EXC_BAD_ACCESS on MacOSX
Kaben Nanlohy
k_cable.public.kitware.com at random.poofygoof.com
Mon Feb 9 17:28:08 EST 2004
Hi;
I hope that others have used CableSwig on Apple's Panther, and can share
some advice with me : I'm struggling with the 'Module' example (at
Utilities/CableSwig/Testing/Module beneath the Insight root) wrapped for
python with CableSwig/Python2.3(fink).
(Early warning and apology : I hope that I'm giving sufficient
information below, but I'm probably giving too much.)
I'm not using or building the example within the Insight project -- I've
copied the example as a test case into my own project (which builds
happily using CMake), where the example builds without problems. Here is
what I see when I try to use the imported module:
$ python2.3
Python 2.3.3 (#1, Dec 29 2003, 09:45:44)
[GCC 3.3 20030304 (Apple Computer, Inc. build 1495)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> from TestModule import *
>>> dir()
['__builtins__', '__doc__', '__file__', '__name__', 'bar', 'barPtr', 'foo', 'fooPtr', 'types']
>>> b=bar()
Bus error
$
Running python2.3 from gdb (this may not be a good way to debug -- I am no
python expert), I see:
>>> b=bar()
Program received signal EXC_BAD_ACCESS, Could not access memory.
PyErr_Occurred () at Python/errors.c:81
81 Python/errors.c: No such file or directory.
in Python/errors.c
(gdb) bt
#0 PyErr_Occurred () at Python/errors.c:81
#1 0x013497f4 in PyObject_Call (func=0x1442af0, arg=0x0, kw=0x13497f4) at Objects/abstract.c:1756
#2 0x013b2358 in PyEval_CallObjectWithKeywords (func=0x614cb0, arg=0x706030, kw=0x0) at Python/ceval.c:3346
#3 0x005516b0 in SWIG_NewPointerObj (ptr=0x0, type=0x0, own=0) at /Users/kaben/Projects/software/avida/branch.kaben/build/cableswig_example/SwigRuntime/swigrunPython.cc:589
#4 0x005204c8 in _wrap_new_bar__SWIG_1(_object*, _object*) (self=0x0, args=0x31e030) at /Users/kaben/Projects/software/avida/branch.kaben/build/cableswig_example/Module/barPython.cc:760
#5 0x005205a4 in _wrap_new_bar(_object*, _object*) (self=0x0, args=0x31e030) at /Users/kaben/Projects/software/avida/branch.kaben/build/cableswig_example/Module/barPython.cc:778
#6 0x0002235c in PyObject_Call (func=0x1442af0, arg=0x0, kw=0x13497f4) at Objects/abstract.c:1755
.
.
.
#30 0x000039d4 in start ()
The problem seems to begin in the CableSwig-generated function
SWIG_NewPointerObj() in cableswig_example/SwigRuntime/swigrunPython.cc, where
the relevant lines read:
/* Create a new pointer object */
565: SWIGRUNTIME(PyObject *)
SWIG_NewPointerObj(void *ptr, swig_type_info *type, int own) {
PyObject *robj;
if (!ptr) {
Py_INCREF(Py_None);
570: return Py_None;
}
#ifdef SWIG_COBJECT_TYPES
robj = PyCObject_FromVoidPtrAndDesc((void *) ptr, (char *) type->name, NULL); #else
575: {
char result[1024];
char *r = result;
*(r++) = '_';
r = SWIG_PackData(r,&ptr,sizeof(void *));
580: strcpy(r,type->name);
robj = PyString_FromString(result);
}
#endif
if (!robj || (robj == Py_None)) return robj;
585: if (type->clientdata) {
PyObject *inst;
PyObject *args = Py_BuildValue((char*)"(O)", robj);
Py_DECREF(robj);
inst = PyObject_CallObject((PyObject *) type->clientdata, args);
590: Py_DECREF(args);
if (inst) {
if (own) {
PyObject *n = PyInt_FromLong(1);
PyObject_SetAttrString(inst,(char*)"thisown",n);
595: Py_DECREF(n);
}
robj = inst;
}
}
600: return robj;
}
During a break at the pertinent line 589 I find:
Breakpoint 1, SWIG_NewPointerObj (ptr=0x613de0, type=0x528054, own=1) at /Users/kaben/Projects/software/avida/branch.kaben/build/cableswig_example/SwigRuntime/swigrunPython.cc:589
589 inst = PyObject_CallObject((PyObject *) type->clientdata, args);
(gdb) p type
$1 = (swig_type_info *) 0x528054
(gdb) p type->clientdata
$2 = (void *) 0x614bc0
(gdb) p args
$3 = (PyObject *) 0x706030
(gdb) p *args
$4 = {
ob_refcnt = 1,
ob_type = 0x1432d88
}
(gdb) p *args->ob_type
$5 = {
ob_refcnt = 1,
ob_type = 0x1433140,
ob_size = 0,
tp_name = 0x141e024 "tuple",
tp_basicsize = 12,
tp_itemsize = 4,
tp_dealloc = 0x13848f0 <tupledealloc>,
tp_print = 0x1384a60 <tupleprint>,
tp_getattr = 0,
tp_setattr = 0,
tp_compare = 0,
tp_repr = 0x1384b24 <tuplerepr>,
tp_as_number = 0x0,
tp_as_sequence = 0x1432d34,
tp_as_mapping = 0x1432d7c,
tp_hash = 0x1384ce0 <tuplehash>,
tp_call = 0,
tp_str = 0,
tp_getattro = 0x1377e78 <PyObject_GenericGetAttr>,
tp_setattro = 0,
tp_as_buffer = 0x0,
tp_flags = 17899,
tp_doc = 0x1432c9c "tuple() -> an empty tuple\ntuple(sequence) -> tuple initialized from sequence's items\n\nIf the argument is a tuple, the return value is the same object.",
tp_traverse = 0x13851c8 <tupletraverse>,
tp_clear = 0,
tp_richcompare = 0x1385240 <tuplerichcompare>,
tp_weaklistoffset = 0,
tp_iter = 0x138583c <tuple_iter>,
tp_iternext = 0,
tp_methods = 0x1432d5c,
tp_members = 0x0,
tp_getset = 0x0,
tp_base = 0x0,
tp_dict = 0x0,
tp_descr_get = 0,
tp_descr_set = 0,
tp_dictoffset = 0,
tp_init = 0,
tp_alloc = 0,
tp_new = 0x1385460 <tuple_new>,
tp_free = 0x13f1a28 <PyObject_GC_Del>,
tp_is_gc = 0,
tp_bases = 0x0,
tp_mro = 0x0,
tp_cache = 0x0,
tp_subclasses = 0x0,
tp_weaklist = 0x0,
tp_del = 0
}
Thus on this side of PyObject_CallObject(), the variables 'type' and
'args' seem to have valid data. I suppose that PyObject_CallObject is a
macro to PyEval_CallObjectWithKeywords, but I'm not sure. In any case,
on the far side of PyEval_CallObjectWithKeywords() something has been
clobbered.
Since this is the limit of my knowledge, I thought I'd seek help before
I start researching the innards of Python and CableSwig.
I should note that the same thing happens when I try to instantiate my
own hello-world example wrapping, but that, when I use cswig in
swig-compatible mode to wrap my hello-world, instantiation works fine.
***
Okay. I hope you aren't overloaded by the length of this email. Any
thoughts / help are appreciated.
-- Kaben
More information about the cable
mailing list