Re: member strings one character too long

The behavior has been changed back to what you expect.

The changes are in the CVS version.

Marcelo

digitalpenis@xxxxxxxxxxxxxx wrote:

> hello,
>
> i have encountered a problem with character array
> members in classes in swig-1.3.22. here is an example:
>
> the interface is for the following test case...
>
> class zap {
> public:
> char str[4];
> zap() {
> strncpy(str, "FOO", 3);
> str[3] = 0;
> }
> };
>
> the wrapper code has
>
> static PyObject *_wrap_zap_str_get(PyObject *, PyObject
> *args) {
> PyObject *resultobj;
> zap *arg1 = (zap *) 0 ;
> char *result;
> PyObject * obj0 = 0 ;
>
> if(!PyArg_ParseTuple(args,(char
> *)"O:zap_str_get",&obj0)) goto fail;
> if ((SWIG_ConvertPtr(obj0,(void
> **)(&arg1),SWIGTYPE_p_zap,
> SWIG_POINTER_EXCEPTION | 0)) == -1) SWIG_fail;
> result = (char *)(char *) ((arg1)->str);
>
> resultobj = SWIG_FromCharArray(result, 4);
> return resultobj;
> fail:
> return NULL;
> }
>
> which appears to make the returned string one character
> too long (ie. it includes the terminating NULL). the
> outcome is that, if you do something like:
>
> import example
>
> z = example.zap()
>
> print z.str
>
> and redirect the output to a file (you don't see the
> NULL in a terminal!), the it looks like:
>
> FOO^@
>
> i trust that i am not doing something stupid here
> (although that is always a possibility!).
>
> i have investigated this a little further. the same
> example code with swig-1.3.17, gives:
>
> static PyObject *_wrap_zap_str_get(PyObject *self, PyObject
> *args) {
>    PyObject *resultobj;
>    zap *arg1 = (zap *) 0 ;
>    char *result;
>    PyObject * obj0  = 0 ;
>
>    if(!PyArg_ParseTuple(args,(char
> *)"O:zap_str_get",&obj0)) goto fail;
>    if ((SWIG_ConvertPtr(obj0,(void **) &arg1,
> SWIGTYPE_p_zap,SWIG_POINTER_EXCEPTION | 0 )) == -1) SWIG_fail;
>    result = (char *)(char *) ((arg1)->str);
>
>    resultobj = PyString_FromString(result);
>    return resultobj;
>    fail:
>    return NULL;
> }
>
> this works perfectly, as i would expect.
>
> does this mean that something unintended has been introduced
> between versions 1.3.17 and 1.3.22?
>
> cheers,
> andrew.
>
>  



_______________________________________________
Swig maillist  -  Swig@xxxxxxxxxxxxxxx
http://mailman.cs.uchicago.edu/mailman/listinfo/swig